Executive Summary Improving the Assessment and Valuation of Climate Change Impacts for Policy and Regulatory Analysis Modeling Climate Change Impacts and Associated Economic Damages and Research on Climate Change Impacts and Associated Economic Damages June 2011 Workshop Report: Improving the Assessment and Valuation of Climate Change Impacts for Policy and Regulatory Analysis Part 1 Draft January 2011


Improving the Assessment and Valuation o
Climate Change Impacts for Policy
and Regulatory Analysis j

Sponsored by

r-nA United Slates
Environmontai
^ W L—l Protection Agency

^0^ 0 *. DEPARTMENT Or

mENERGY





Executive Summary:

Improving the Assessment and Valuation of
Climate Change Impacts for Policy and
Regulatory Analysis

Modeling Climate Change Impacts and Associated Economic Damages

and

Research on Climate Change Impacts and Associated Economic Damages

June 2011

Workshop Sponsored by:
U.S. Environmental Protection Agency
U.S. Department of Energy

Workshop Report Prepared by:
ICF International

-------
I. Introduction

In 2009 and early 2010, the U.S. Environmental Protection Agency (EPA) and U.S. Department of Energy
(DOE) joined other U.S. government agencies in conducting an analysis of the social cost of carbon (SCC).
The interagency working group used the DICE, FUND, and PAGE integrated assessment models (1AM) to
estimate a range of values for the SCC from 2010 to 2050 for use in U.S. government regulatory impact
analyses. The U.S. government analysis concluded in February 2010 and the estimated SCC values were
first used in March 2010 in the analysis of DOE's Energy Conservation Standard for Small Electric Motors.
In preparation for future revisions to the U.S. government SCC analysis, EPA and DOE seek to improve
the understanding of the natural scientific and economic impacts of climate change. This enhanced
understanding is also intended to inform ongoing work of the U.S. government to improve regulatory
assessment and policy analysis related to climate change.

To further these objectives, the EPA National Center for Environmental Economics and Climate Change
Division and the DOE Office of Climate Change Policy and Technology sponsored a pair of invitational
workshops on November 18-19, 2010 and January 27-28, 2011. The November workshop focused on
conceptual and methodological issues related to modeling and valuing climate change impacts. It also
addressed the implications of these estimates for policy analysis. The January workshop reviewed recent
research on physical impacts and associated economic damages for nine impact categories (e.g., human
health, agriculture, sea level), with a particular focus on knowledge that might be used to improve lAMs.

This workshop summary was prepared by ICF International on behalf of EPA and DOE. It does not
represent the official position or views of the U.S. government or its agencies, including EPA and DOE,
nor has it been reviewed by the workshop speakers and other participants. The potential improvements
and key findings outlined below represent the perspectives of one or more participants, as expressed at
the workshops and summarized by the planning committee. However, these summaries do not
necessarily represent consensus views, since none was sought at these workshops. This Executive
Summary is organized into six sections: Physical Impacts Assessment; Valuation of Damages;
Representing Impacts and Damages in Models; Communication of Estimates; Research and
Collaboration; and Specific Impacts Sectors.

II. Physical Impacts Assessment

Participants made comments and suggestions related to impacts assessment, including the following:

• More fully incorporate uncertainty. Natural and social scientists should attempt to more fully
characterize the uncertainty in impacts assessments, including parametric, stochastic, and
structural uncertainty at all stages in the modeling process. Many of the current 1AM inputs and
parameters represent too narrow a range of possibilities. Complex and non-linear processes at
the high ends of the impacts probability distribution (i.e., "fat tails") should be better
characterized.

-------
• Consider both top-down and bottom-up approaches. Estimates from both top-down and
bottom-up approaches can help to estimate and bound the range of climate change impacts.
For bottom-up approaches, the appropriate scale and detail may be different for each sector.

• Incorporate threshold effects of physical and biological impacts. Mechanistic and process
models relying on basic principles (e.g., conservations of energy, plant biophysiology, ocean
biogeochemistry) should be used, when possible, to extrapolate responses to new conditions,
since statistical methods may not capture non-linear threshold effects of unprecedented levels
of change. When climate change impacts are expected to be within or close to the range of past
variations, statistical models are appropriate.

• Capture climate variables beyond global mean temperature. A better characterization of
multiple climate variables (e.g., precipitation, storms, seasonal and diurnal temperature
variations, rate of temperature change) and threshold effects on a geographically disaggregated
scale could improve model calibration and the accuracy of local damage projections.

• Focus research efforts on sectors that could have the largest influence on overall damage
estimates. This will include research on impact categories that could comprise a large share of
total damages but where relatively little information has been collected to date. Researchers
should not simply focus on issues that are easiest to approach. Research priorities should be
guided by the combination of potential consequences and uncertainty, not one or the other
alone.

• Increase focus on high-impact events, multi-century impacts. Existing studies tend to examine
the means of the impacts probability distribution, neglecting the low-probability, high impact
tails of the distribution, which can have a significant influence on 1AM results. Impact studies
should address this gap, recognize the potential for unexpected and unpredictable events, and
attempt to model very long-term impacts (e.g., beyond 2100), despite great associated
uncertainty. To do this, modelers should develop more complete multi-century projections for
socio-economic and climate inputs including estimates of socio-economic uncertainty.

• Rigorously test, compare, and evaluate impact models. Model intercomparison projects have
helped to improve physical climate models and could be used to improve impact models.

III. Valuation of Damages

Comments and suggestions related to damage valuation included the following:

• Consider alternate functional forms for damage functions. Representation of damages could be
improved by: evaluating the additive or multiplicative nature of impacts; better incorporating
discontinuities; better capturing natural capital and its interactions with physical and social
capital; and generally considering a broader set of functional forms. Alternate forms are

-------
particularly important given the challenges in extrapolating damage functions calibrated at 2-
3°C warming to considerably higher global mean temperature increases.

• Clearly incorporate human behavioral responses. Adaptation and technological development
should be more fully incorporated in estimates of climate change impacts, and the underlying
assumptions associated with those factors should be clearly articulated.

• Consider different ways of equity weighting when conducting social welfare analysis of
climate policies. Several workshop participants suggested considering different ways of
incorporating equity weights into the SCC or lAMs more generally. For example, most lAMs use
a utility function with a single parameter that controls preferences regarding intra-generational
equity, inter-generational equity, and risk aversion. Future research should explore alternative
functional forms that allow these effects to be disentangled.

• Fully account for non-market impacts and non-use values. This includes improving estimates of
impacts currently included in some models (e.g., health impacts) and incorporating impacts
currently missing from most models (e.g., ocean acidification, loss of cultural heritage). Revealed
and stated preference estimates and benefit-transfer methods should be improved and
estimated jointly to mitigate problems with each.

• Consider "outer measures" of climate damages. Developing a model for a highly simplified but
inclusive "outer" measure of climate change damages may help provide an upper bound on SCC
estimates. Current bottom-up models are "inner" measures that attempt to capture and sum
the individual components of climate damages. Since it is challenging to capture all of the
components and interactions between them, these models will tend toward underestimation.

IV. Representing Impacts and Damages in Models

Throughout both workshops, but especially during the first, participants made suggestions related to
integrating impacts and damages in models. These comments included the following:

• Improve both aggregated and disaggregated models while utilizing the strengths of each.

There are important roles for models across the spectrum of aggregation, as more or less
aggregation may be appropriate for different applications. Model type and analysis time scale
should be matched to analytical objective. Since aggregation can contribute to a bias in impact
estimates, some models should be less aggregated spatially, temporally, and sectorally to more
realistically represent impact mechanisms. Since disaggregated models can incorporate more
realistic impact mechanisms and use empirical data to estimate model parameters, they can be
used to calibrate components of more comprehensive aggregated models.

• Incorporate more sectors. lAMs should include a broader range of sectors. For example, no
lAMs currently represent ocean acidification.

-------
• Incorporate interactions between sectors. Interactions between sectors (and among climate
and non-climate stressors) may be synergistic or antagonistic, additive, multiplicative, or
subtractive, making cumulative impacts larger or smaller than the sum of the individual impacts.
Double-counting should be avoided.

• Use consistent scenarios. Consistent socio-economic and climate scenarios should be used in
impact and damage assessment to facilitate inter-comparison, integration, and combination of
estimates.

• Increase model flexibility to facilitate improvements. lAMs should be (re)designed to facilitate
updates to models or model components as new research develops. A more flexible or modular
structure would allow components to be individually updated or replaced.

• Conduct new empirical studies and better incorporate existing research. lAMs need new
primary impacts research from which to draw. Research needs include empirical studies on:
physical impacts, monetization of damages, decision making under uncertainty, adaptation-
related technological change, adaptive capacity, tipping points, and impacts beyond 2100. lAMs
could also be improved by drawing more on the existing body of research.

V. Identify metrics for model validation. Metrics and methods of
validation are needed to assess models and model results.
Communication of Estimates

Participants, particularly at the first workshop, made comments and suggestions related to the

communication of impacts and damages estimates. These comments included the following:

• Increase transparency. lAMs should be made more accessible and transparent, including their
key assumptions, structural equations, parameter values, and underlying empirical studies.

• Fully and clearly communicate uncertainty. Communication should help decision makers and
the public fully and clearly understand uncertainty and its implications. The full range of model
outputs should be communicated and used, rather than focusing on one central value from a set
of model runs.

• Consider other metrics. Multiple criteria, in addition to the SCC and cost-benefit analysis, should
be used for climate-related regulatory analysis, including additional cost-effectiveness
measures.

VI. Research and Collaboration

Comments and suggestions related to research and collaboration included the following:

-------
• Increase collaboration and communication between natural scientists, economists, and
modelers. Collaboration and communication should be increased between all parties involved in
impacts assessment, damages valuation, and integrated assessment modeling. Impacts
assessment and valuation efforts should be coordinated with existing efforts such as the
National Climate Assessment and international impacts and valuation efforts. 1AM data sources,
damage functions, and outputs should be reviewed by relevant members of the Impacts,
Adaptation, and Vulnerability (IAV) and economic valuation communities to ensure that lAMs
reflect the current state of the primary literature for each of the impact categories.

• Increase capacity to address challenges. Additional funding and staff are needed to help
address existing impacts and damages assessment challenges.

VII. Specific Impacts Sectors

The second workshop focused on the current state of research in nine impact categories. This section

highlights key research findings and recommendations for future research for each of the categories.

Storms and Other Extreme Weather Events

• Fewer tropical storms are expected in the future, but average wind speeds and precipitation
totals are expected to increase. The intensity of the strongest storms is expected to increase.

• Estimates in the literature for increases in cyclone property damages due to climate change
range from 0.002 to 0.006% of global GDP. Increases in property damages from all extreme
events (including cyclones) due to climate change under an A1B scenario, according to one
study, range from $47-$100 billion (2008 dollars) per year, or 0.008-0.018% of GDP, by 2100.

• Fatalities may increase or decrease due to climate change impacts on extreme events, as deaths
from tropical cyclones may decrease more than deaths from other extreme events (e.g., heat
waves) increase. Tropical cyclones are expected to continue to be the dominant cause of
extreme event-related damages.

Water Resources

• Water demand, supply, and management should be modeled on a river basin scale to effectively
estimate climate change impacts.

• National estimates from the literature of climate change damages to water resources range
from $12-$60 billion (2009 dollars) per year for the United States according to analyses in a
range of studies.

• Coupling approaches that model changes using regional hydrologic models and those using
regional economic models could help bridge some gaps in water resources damage estimation.

-------
Human Health

• The majority of climate change health effects result from diarrhea, malnutrition, and malaria.
The World Health Organization estimates that the costs to treat climate change-related cases of
diarrhea, malnutrition, and malaria in 2030 would be $4 to $13 billion under a scenario in which
C02 is stabilized at 750ppm by 2210. The study predicts a 3%, 10%, and 5% increase in cases of
diarrhea, malnutrition, and malaria, respectively.

• Health impact valuation depends largely on mortality valuation, particularly in developing
countries and particularly among children. Adjusting the value of a statistical life for income is
critical for accurate valuation.

Agriculture

• Estimates in the literature project the global range of yield changes in the 2050s to be
approximately -30 to +20% under a 2.3°C mean global temperature increase (relative to 1961-
1990).

• Average global effects of climate change on agriculture are expected to be positive in the short
term and negative in the long term. The location of the inflection points is unknown.

o C02 fertilization from increasing C02 concentrations will benefit some plants (C3 plants)
more than others (C4 plants). Elevated C02 concentrations especially benefit weeds.

• Agriculture contributes only 2-3% of U.S. GDP, but the highly inelastic nature of agricultural
demand means that even a small reduction in agricultural production from climate change could
result in large price changes and large welfare losses.

• Adaptation and technological change can help to mitigate the impacts of climate change on
agriculture. A key challenge will be producing heat and drought tolerant plants with high yields.

Sea Level Rise

• Climate-induced sea level rise will be compounded by both natural and human-induced
subsidence in many densely-populated coastal areas.

• Emissions abatement may stabilize the rate and ultimate total amount (in 100s of years) of sea
level rise, but not reduce the current significant commitment to sea level rise.

• The valuation of sea level rise damages depends heavily on wetland values and adaptation.

Marine Ecosystems and Resources

• Increasing atmospheric C02 concentrations cause ocean C02 concentrations to increase,
decreasing ocean pH, and decreasing saturation states for calcite and aragonite, which are used
by marine animals to produce calcareous parts (e.g., shells).

• Damages from decreased mollusk harvest revenues due to a 0.1-0.2 ocean pH decrease are
estimated at $1.7 to $10 billion in net present (2007) value losses through 2060. Under the A1FI

-------
scenario pH decreases of 0.1 and 0.2 are expected by approximately 2040 and 2060,
respectively.

• Assessments using bio-climate envelopes, minimum realistic models, and ecosystem and food
web models would be beneficial to estimate marine impacts.

• A wide variety of studies to estimate damages is needed, using both revealed and stated
preferences, to estimate total economic value of marine ecosystems and resources. Analyzing
the results available from multiple existing studies could be used in a benefit transfer study to
estimate economic value by transferring available information into the appropriate context.

Terrestrial Ecosystems and Forestry

• Three major types of terrestrial ecosystem impacts are expected: changes in vegetation
distribution and dynamics, wildfire dynamics, and species extinction risks. For example,
predicted global vertebrate extinctions due to land use and climate change range from over 30%
to nearly 60% for >2 degree warming.

• Understanding changes in pest outbreaks, interior wetlands, and snow pack are important gaps.

• Natural scientists and economists need to work together to identify biophysical impacts
assessment endpoints best suited for use in revealed and stated preference valuation studies.

Energy Production and Consumption

• Energy impacts may be beneficial for small to modest climate change, due primarily to
decreases in heating requirements for buildings, but are expected to be dominated by negative
impacts in the long-run and at higher levels of temperature change.

• More data and research are needed to evaluate the effects from wildfire and sea level rise on
power sector infrastructure, and temperature impacts on electricity production, transmission,
and distribution.

Socio-economic and Geopolitical Impacts

• Climate change-induced natural disasters, migration caused by sea level rise and other climate
factors, and increasing resource scarcity may promote conflict; however, the policy debate
regarding socio-economic and geopolitical impacts from climate change is well ahead of its
academic foundation, and sometimes even contrary to the best evidence.

-------
Improving the Assessment and Valuation
Climate Change Impacts for Policy

Sponsored by

4*> United States
^nv'mnrTinnti,i
kl #mProtod»on Agency

y $ DEPARTMENT 0'

and Regulatory Analysis

% ENERGY

November 18-19, 2010 Omni Shoreham Hotel, Washington, DC

Workshop Report:

Improving the Assessment arid Valuation of

Climate Change Impac

Regulator

Parti

Modeling Climate Change Impacts and Associated Economic Damages

January 2011



Workshop Sponsored by:



U.S. Environmental Protection Agency

U.S. Department of Energy



Workshop Report Prepared by:



ICF International

-------
Table of Contents

Table of Contents	1

I.	Introduction	3

Context	3

Workshop Format	4

II.	Potential Future Improvements Suggested by Workshop Participants	4

Overarching comments	5

Comments related to the modeling of natural systems in lAMs	7

Comments related to the modeling of human systems in lAMs	7

Comments related to the communication of 1AM results	9

III.	Chronological Presentation of Workshop Proceedings	10

Workshop Introduction	10

Opening Remarks	10

Progress Toward a Social Cost of Carbon	12

Session 1: Overview of Existing Integrated Assessment Models	13

Overview of Integrated Assessment Models	13

DICE	13

PAGE	14

FUND	15

GCAM and Development of iESM	16

IGSM	17

Session 1 Discussion	18

Session 2: Near-Term DOE and EPA Efforts	19

Proposed Impacts Knowledge Platform	19

Proposed Generalized Modeling Framework	20

1

-------
Session 2 Discussion	21

Session 3: Critical Modeling Issues in Assessment and Valuation of Climate Change Impacts	21

Session 3, Part 1	21

Sectoral and Regional Disaggregation and Interactions	22

Adaptation and Technological Change	23

Multi-century Scenario Development and Socio-Economic Uncertainty	24

Session 3, Part 1 Discussion	25

Session 3, Part 2	25

Incorporation of Climate System Uncertainty into lAMs	26

Extrapolation of Damage Estimates to High Temperatures: Damage Function Shapes	27

Earth System Tipping Points	28

Potential Economic Catastrophes	29

Nonmarket Impacts	30

Session 3, Part 2 Discussion	31

Session 4: Implications for Climate Policy Analysis and Design	32

Implications for Design and Benefit-Cost Analysis of Emission Reduction Policies	32

Implications for Addressing Equity and Natural Capital Impacts	33

Implications for Choice of Policy Targets for Cost-Effectiveness Analysis	34

Implications for Managing Climate Risks	35

Session 4 Discussion	35

Session 5: Workshop Wrap-up	36

2

-------
I. Introduction

This report summarizes the November 18-19, 2010 workshop, Modeling Climate Change Impacts and
Associated Economic Damages, sponsored by the U.S. Environmental Protection Agency (EPA) and U.S.
Department of Energy (DOE). This was the first in a series of two workshops, titled Improving the
Assessment and Valuation of Climate Change Impacts for Policy and Regulatory Analysis.

This report is organized as follows:

• The first section provides an introduction to the report and the workshop, including context and
workshop format.

• The second section provides a summary of the potential future improvements to climate change
integrated assessment models identified by workshop participants. This section aims to
summarize, categorize, and organize the wide variety of recommendations highlighted by
individual participants over the course of the two-day workshop.

• The third section provides a chronological presentation of the workshop proceedings, including
a summary of each presentation, question and answer session, and discussion section.

• The appendix to the report provides the final workshop agenda with charge questions, the
participant list, and extended abstracts of most speaker presentations.

This report serves as the EPA and DOE planning committee's summary of the workshop. It has not
received official endorsement from the workshop speakers and other participants.

Context

In 2009 and early 2010, EPA and DOE participated in the interagency working group on the social cost of
carbon (SCC). The interagency group used the DICE, FUND, and PAGE integrated assessment models
(1AM) to estimate a range of values for the social cost of carbon from 2010 to 2050 for use in U.S.
government regulatory impact analyses (RIA). The SCC working group reported their findings in
February 2010 and the estimated SCC values were first used in the analysis of DOE's Energy
Conservation Standard for Small Electric Motors.1 In preparation for future iterations of this process,
EPA and DOE seek to improve the natural science and economic understanding of the potential impacts
of climate change on human well-being.

To help motivate and inform this process, EPA's National Center for Environmental Economics
(NCEE) and Office of Air and Radiation's (OAR) Climate Change Division and DOE's Office of
Climate Change Policy and Technology sponsored a pair of invitational workshops in late 2010 and early
2011. The first workshop took place on November 18-19, 2010 and focused on conceptual and
methodological issues related to modeling and valuing climate change impacts. It also addressed
implications of these estimates for policy analysis. The second workshop, to be held January 27-

1 See http://go.usa.gov/3fH.

-------
28, 2011, will review the quantitative research that examines the physical impacts and economic
damages for a variety of impact categories (e.g., agriculture, human health, ocean acidification). These
workshops are intended to inform future refinements of the SCC and ongoing work of the U.S.
government to improve regulatory assessment and policy analysis.

Workshop Format

The workshop took place over two days, November 18-19, 2010, at the Omni Shoreham Hotel in
Washington, DC. The workshop was attended by approximately 110 individuals, including
representatives from several U.S. federal government agencies, non-governmental organizations,
academia, and the private sector. A full list of workshop participants is available in the Appendix.

The workshop opened and concluded with remarks by representatives of EPA and DOE. After an initial
background talk on the interagency SCC process, the workshop consisted of four plenary sessions:

• Session 1: Overview of Existing Integrated Assessment Models

• Session 2: Near-Term DOE and EPA Efforts

• Session 3: Critical Modeling Issues in Assessment and Valuation of Climate Change Impacts

• Session 4: Implications for Climate Policy Analysis and Design

Each session included a panel of speakers who gave presentations, responded to questions specific to
their talk, and participated in an open discussion with the audience at the end of each session. The full
workshop agenda, charge questions, and extended abstracts of most presentations are available in the
Appendix.

II. Potential Future Improvements Suggested by Workshop
Participants

Over the course of the two-day workshop, a number of suggestions for improving the assessment and
valuation of climate change impacts were identified by the workshop participants. These suggestions
are related to ways that both integrated assessment modeling generally and SCC estimation specifically
could be improved in the future. This section aims to summarize and categorize those suggestions.

The section is organized into four categories of comments:

• overarching comments;

• comments related to the modeling of natural systems in lAMs;

• comments related to the modeling of human systems in lAMs; and

• comments related to the communication of 1AM results.

The potential improvements outlined below represent the perspectives of one or more participants but,
importantly, do not represent a consensus since none was sought at this workshop.

-------
Overarching comments

Throughout the course of the workshop, many participants made general comments related to the
discipline of climate policy analysis and specific suggestions for potential future improvements related
to the underlying structure of and inputs to integrated assessment models. These comments spanned a
wide range of topics, include the following:

• Improve both aggregated and disaggregated models while highlighting the strengths of each.

There was considerable debate about the appropriate level of disaggregation and the merits of
using more or less aggregated models for different types of applications. Several participants
suggested that increased attention to disaggregation was important to understanding the true
impacts associated with climate change. However, some were skeptical of current capabilities
to downscale global climate models (GCMs) to produce reliable disaggregated estimates of
impacts, at local or regional scales. In the end many participants suggested that a two-track
approach is necessary and that there are important roles for models across the spectrum of
aggregation.

o Build better disaggregated models. Many conference participants recommended using
more disaggregated models, emphasizing that aggregation can contribute to a bias in
impact estimates. (For example, if damages increase at an increasing rate with higher
local temperatures, then using regionally averaged temperature increases would
underestimate the average local damages.) They recommended that models increase
disaggregation spatially and sectorally to allow for more realistic representations of
impact mechanisms. They also emphasized the need to explicitly model the temporal
and spatial variability of climate impacts.

o Better inform calibration of aggregated models with disaggregated models. Some
participants suggested using more disaggregated models to help inform calibration of
more aggregated models. Several noted it is possible to incorporate more realistic
impact mechanisms in disaggregated models, and to more accurately parameterize such
models using empirical data. Participants suggested that the predictions of more
disaggregated models might be useful to calibrate components of the more general and
comprehensive aggregated models (at least within the range of temperature changes
observed in the data).

• Increase model flexibility to facilitate improvements. Several participants suggested that lAMs
should be (re)designed to be more flexible so that it is easier to update the models or model
components to incorporate new research findings. At least two participants suggested moving
to a more modular structure where different components could easily be updated or replaced
by newer modules as research develops. For example, increased modularity could allow
researchers to replace sector-specific damage functions when new research points to different
parameter values or functional forms. While lAMs, which link climate models to impact and
economic models, are somewhat modular in theory, this has not always been the case in
practice. Modularity could be introduced in model implementation in multiple ways. A simple

-------
effort might be to ensure interoperability between existing models of physical impacts and
economic damages and various climate system modules. A more complex effort might allow
researchers to focus in on one specific aspect of the problem without affecting compatibility
with the system.

• Conduct new empirical studies and better incorporate existing research. Participants noted
repeatedly that lAMs need new primary research on impacts from which to draw. Participants
specifically highlighted a need for empirical studies on: physical impacts; monetization of
damages; decision making under uncertainty; adaptation-related technological change; adaptive
capacity; response-time, recovery, and cost related to disasters; tipping points; and impacts
beyond 2050. Participants also noted that lAMs could be improved by drawing more on the
existing body of research. Some participants suggested that assessments of climate change
impacts under high-end warming scenarios would help the integrated assessment modelers
calibrate their damage functions over ranges of temperatures higher than those typically
examined in climate damage assessment studies based on historical data.

• Develop more robust long-term projections of inputs. Several participants emphasized the
need to develop and employ a more complete set of multi-century projections for socio-
economic and climate inputs, in particular projections of population, GDP, and greenhouse gas
emissions that more fully characterize the uncertainty of such long term forecasts. A
standardized set of probabilistic long-term socio-economic projections could be used as a
substitute for, or complement to, the traditional scenario-based approach as exemplified by the
IPCC Special Report on Emissions Scenarios (SRES).2

• More fully incorporate uncertainty. Several participants emphasized the need to more fully
account for uncertainty at all stages in the modeling process from model inputs and parameters
to outputs, using fat-tailed distributions where appropriate. This includes parametric,
stochastic, and structural uncertainty. Participants argued that many of the current inputs and
damage parameters represent too narrow a range of possibilities. Throughout the conference,
speakers and participants identified the need to more fully account for the complex and non-
linear implications at the high ends of the climate change impacts probability distribution.

• Identify metrics for model validation. Several participants highlighted the need to identify
metrics and methods of validation to provide an assessment of models and model results.

These participants argued that without metrics for validation, there is no indication of how well
a model is performing or to what degree the results are accurate.

• Increase communication between natural scientists and economists. Numerous conference
participants and speakers raised the need to increase the communication between natural
scientists and economists in order to facilitate and build a collaborative community.

2 http://www.ipcc.ch/ipccreports/sres/emission/index.htm

-------
• Increase funding for climate economics and integrated assessment research. Throughout the
workshop, participants repeatedly highlighted the currently insufficient level of funding needed
to robustly estimate economic damages of climate change and the SCC. Participants
underscored the large discrepancy between levels of funding for natural science research and
comparatively low levels of funding for economic valuation and integrated assessment research.
Several participants also noted that relatively few researchers are currently working in the field
of climate change economics and valuation. Therefore, the existing body of research in this field
is relatively thin compared to other areas of climate change science.

Comments related to the modeling of natural systems in IAMs

Participants also suggested potential future improvements related to the modeling of natural systems in

IAMs. These suggestions include the following:

• Capture climate variables beyond global mean temperature. Several participants emphasized
the importance of developing more explicit, comprehensive, and detailed characterizations of
the climate variables and threshold effects. Specifically, numerous participants highlighted the
need for climate variables other than global mean temperature (e.g., precipitation, storms,
seasonal and diurnal temperature variations, the rate of temperature change, etc.) to drive
impacts. Participants noted that a better characterization of these climate variables on a
disaggregated scale would provide opportunities for improved model calibration.

• Incorporate the co-variance between climate sensitivity and transient climate response. A few

presenters emphasized the importance of accounting for the co-variance between climate
sensitivity and transient climate response, especially in probabilistic models that consider a wide
range of possible equilibrium climate sensitivity values (e.g., Baker and Roe 2009). Some
participants also highlighted the importance of explicitly modeling relationships between the
strength of the non-C02 forcing, climate sensitivity, and ocean heat capacity. High equilibrium
climate sensitivity is correlated with a more strongly negative current aerosol forcing (and thus
moderately negative total non-C02 forcing). It is also correlated with a higher ocean heat
capacity and a longer timescale to reach equilibrium. As a result of the relationship between
equilibrium climate sensitivity and ocean heat capacity, the probability distribution for the
transient climate response is narrower and has less of a 'fat tail' than the distribution for
equilibrium climate sensitivity.

Comments related to the modeling of human systems in IAMs

Many participants made suggestions of potential future improvements related to modeling of human

systems in IAMs. These suggestions include the following:

• Consider alternative functional forms for damage functions. Numerous conference
participants highlighted the need to re-evaluate the functional form of the models' damage
representations. The suggested improvements included: evaluating whether impacts should be

-------
additive or multiplicative3; better incorporating discontinuities; making damage functions more
reactive to extreme temperature increases; and generally considering a broader set of
functional forms for damage functions. It is important to consider alternative functional forms
given the challenges in extrapolating damage functions calibrated at 2-3 °C global warming to
considerably higher global average temperature increases.

• Better incorporate welfare and equity. Workshop participants identified numerous potential
improvements related to welfare and equity.

• Many participants argued that the formulation of welfare functions should be
reconsidered and refined. Some participants further argued that consumption alone
was not a good measure of welfare, suggesting that more robust measures be used
instead. For example, participants suggested that multivariate utility functions be used,
in order to better account for a variety of goods valued by consumers. These functions
could combine consumption of market and non-market goods such as manufactured
goods and environmental amenities.

• Although discounting was not on the workshop agenda, numerous participants
emphasized the need to re-evaluate discounting assumptions in SCC estimates. Some
participants suggested that discounting be made endogenous to the models and related
to economic growth. Some participants suggested incorporating distributional
considerations into discounting.

• Several workshop participants suggested that models incorporate distributional equity
in ways other than through discounting. For example, this could be done by equity
weighting the estimated monetized damages in each region before aggregating to the
global scale. Some emphasized that ignoring the curvature of utility functions means
that negative impacts on poor countries are equivalent to those in well developed
countries.

• Several workshop participants suggested that risk aversion was not properly
incorporated into the models. These participants suggested that assumptions about risk
aversion should be re-evaluated and refined.

• Incorporate natural capital. Several workshop participants suggested that natural capital be
better incorporated into lAMs. In particular, participants emphasized the importance of
capturing the imperfect substitution between natural and human-made physical capital.

• Incorporate more sectors. Many participants suggested that current lAMs do not include all
impacted sectors. For example, no lAMs currently represent damages from ocean acidification.
They indicated that improvements could be made by incorporating a broader range of sectors.

3 See Weitzman, M. 2010. What is the "Damages Function" for Global Warming - and What Difference Might it
Make? Climate Change Economics 1(1): 57-69.

-------
• Improve valuation of non-market impacts. Several participants emphasized the need to
improve the valuation of non-market impacts and their representation in lAMs. This includes
both improving the estimates of non-market impacts currently included in some models (e.g.,
health impacts) and incorporating non-market impacts currently missing from most models
(e.g., ocean acidification, loss of cultural heritage, etc.).

• Consider "outer measures" of climate damages. A couple of participants highlighted the need
for a highly simplified but inclusive "outer" measure of climate change damages that could
provide an upper bound on the estimates. These participants suggested that current models are
all "inner" measures that attempt to capture the individual subset components of the SCCto
build up to the total SCC. Since it is very difficult to capture all of the individual components,
these estimates tend to be low-end estimates.

Comments related to the communication of IAM results

Finally, many participants suggested potential future improvements related to the communication of
the SCC and its use in decision making. These suggestions include the following:

• Increase transparency. Throughout the workshop, from Deputy Administrator Perciasepe and
Under Secretary Koonin's opening remarks to Dr. Duke and Dr. McGartland's summary
comments, transparency was a recurring theme. Numerous participants and speakers
emphasized the need to increase the accessibility and transparency of the models, including
their key assumptions, structural equations, calibrated parameter values, and the underlying
empirical studies on which these values are based.

• Communicate uncertainty. The effective communication of uncertainty was another theme
that pervaded the comments of participants. Given the significant uncertainty involved in the
estimation of the SCC, numerous participants emphasized the crucial importance of fully and
clearly communicating the uncertainty behind the estimates, including the relationship between
uncertainty and time scale. Much discussion centered on how best to communicate model and
parameter uncertainty so that decision makers and the public properly understand the
uncertainty surrounding SCC estimates and the implications of this uncertainty. One specific
suggestion along these lines was to emphasize that the precision in the final SCC estimates
correlate with the precision that can be supported by the model inputs. For example, reporting
the SCC with several significant figures gives a highly overconfident impression of the precision
of these estimates.

• Use a range of outputs. Related to the communication of uncertainty, many participants
encouraged increased communication and use of the full range of model outputs rather than
focusing on one central value from a set of model runs. Opinions varied regarding the most
effective way to communicate uncertain results, so more work in this area could be useful.

• Consider other metrics. Many participants questioned the usefulness and effectiveness of the
SCC as a single criterion for regulatory analysis. Several participants discussed the potential
shortcomings of cost-benefit analysis in a climate change context. Some participants indicated

-------
that the SCC may be one relevant measure, but they encouraged the use of multiple criteria for
regulatory analysis, in addition to the SCC. Participants suggested using additional measures to
assess cost-effectiveness, such as using the shadow price of a range of policy targets as a
reference.

• Match model to objective. Many participants underscored the importance of matching model
type to analytical objective. Participants noted that a given question may be better addressed
by one type of model than another. For example, a high-resolution model might be most
appropriate for some analytical questions, such as assessing impacts to individual sectors, while
a reduced-form model might be most appropriate for assessing other questions, such as the
sensitivity of the outcomes to a wide variety of policy choices and model assumptions.
Aggregated damage functions might address certain questions best while disaggregated
representations of damages might best address others. Similarly, the time-scale of the analysis
should appropriately match the analytical aims.

III. Chronological Presentation of Workshop Proceedings

This section presents the proceedings of the workshop in chronological order, including: workshop
introduction; session presentations, question and answer sessions, and discussions; and closing remarks.

Workshop Introduction

The workshop commenced with a welcome and introduction by Elizabeth Kopits of the U.S.
Environmental Protection Agency. She noted that this workshop was the first of two EPA- and DOE-
sponsored workshops aimed at an open, scholarly dialog among top researchers about Integrated
Assessment Models and climate change impacts and damage estimations. She explained that the
impetus for the meeting arose from the recent interagency report on the SCC. She highlighted the need
to update and revise the SCC; to incorporate new scientific findings as they emerge; and to improve
transparency, availability, and understanding. She noted the need to spur efforts to fill research gaps,
explaining that some would be difficult to fill while others would be more easily addressed by
improvements in economics and science. Finally, she highlighted the need for increased collaboration
between natural scientists and economists.

Opening Remarks

Following Dr. Kopits' introduction, Bob Perciasepe, U.S. Environmental Protection Agency Deputy
Administrator, shared his opening remarks. Mr. Perciasepe began by thanking the participants for their
work. He underscored the importance of the SCC in helping EPA to be a better decision maker, noting
the important role that cost-benefit analysis (CBA) has played to drive EPA work throughout its 40-year
history. He suggested that the SCC begins another chapter in EPA's history by creating a unifying
measure and tool to use across different programs in the U.S. Government. Mr. Perciasepe also noted
his healthy concern that CBA fails to capture many different issues. He highlighted the more ubiquitous
and difficult aspects of the climate change question, with its numerous effects around the globe. He
concluded that the SCC is an important common building block, but that it needs to be improved.

-------
Mr. Perciasepe then raised a few key questions and challenges to the workshop participants. First, he
asked if the current valuation methods adequately address all costs and catastrophic risks. He
highlighted the possibility of irreversible impacts from climate change, noting the significant
multigenerational effects from climate change. Mr. Perciasepe highlighted numerous impacts that
remain unquantified in the reduced-form lAMs, including ocean acidification and loss of biodiversity. He
questioned whether the breadth of impacts is captured by models, providing agricultural impacts from
weather volatility as an example.

Next, Mr. Perciasepe asked whether there is a way to present the SCC transparently enough for the
public to understand it. He noted that while the current estimate is an incomplete picture, many people
see it as an all-encompassing portrait. He suggested perhaps listing the range of possible impacts and
clarifying which are and are not reflected in current models. Finally, Mr. Perciasepe asked how best to
account for the time horizons of impacts, given that emissions today may set the pattern for centuries.
Mr. Perciasepe concluded his remarks by once again emphasizing that he values this work greatly, that
progress so far has been remarkable, but that improvement is still needed and his challenges are
intended to spur the iterative process forward.

Next, Dr. Steven E. Koonin, Under Secretary for Science at the U.S. Department of Energy, shared his
thoughts from the perspective of DOE's chief scientist. He underscored the importance of the valuation
endeavor, particularly to inform policy. He noted that the interagency report has already been used for
multiple DOE Energy Conservation Standards, including the first U.S. government use of the report in the
Energy Conservation Standard for small electric motors. He emphasized the importance of speaking the
language of economics, to drive action on climate change. Acknowledging the complicated nature of
the problem, he emphasized the importance of addressing it with rigor and transparency so that it is
justifiable to non-experts. Finally, he noted that DOE has and will continue to sponsor integrated
assessment work and climate modeling.

Second, Dr. Koonin presented his thoughts from the perspective of a scientist who has professionally
done modeling work. He explained that the work so far has been good but a lot of progress still needs
to be made. He noted that credible integrated assessment models differ in their results by an order of
magnitude. Dr. Koonin explained his healthy skepticism about models, suggesting that all of the models
are wrong, but some are useful. He asked for the models to be validated, for their differences and
uncertainties to be outlined, and for improvements to be identified. He called for more data, and asked
for metrics to validate model results. He then suggested that more elaborate lAMs are not necessarily
more useful tools than simpler lAMs in every case.

Dr. Koonin concluded his remarks by describing a back-of-the-envelope approach to calculate the social
cost of carbon. He began by noting that - given the long lifetime of carbon dioxide in the atmosphere -
small, marginal changes in C02 emissions will have only minor impacts on the ultimate magnitude of
climate change. Reducing emissions now can therefore be viewed as delaying the time in the future at
which cumulative emissions targets are reached. He finished by suggesting that the notion of buying
time is an interesting avenue to pursue for climate change valuation. If discounted to the present, the
value of time bought might serve as a summary measure of marginal damages.

-------
Progress Toward a Social Cost of Carbon

Dr. Michael Greenstone, who co-chaired the interagency SCC process when he served as chief
economist for the White House Council of Economic Advisors, then presented an overview of the
interagency process, including an example of how the SCC can be useful in a regulatory context. He
started with the background and motivation for developing the SCC. He presented some of the impacts
of climate change and an overview of U.S. climate change regulation. He noted the lack of climate
change legislation and the early efforts to regulate greenhouse gases through the Clean Air Act. Given
these emerging regulations, Dr. Greenstone presented the desire for a social cost of carbon to monetize
benefits during regulatory impact analyses. He explained that the SCC is the monetized damage
associated with an incremental increase in carbon emissions in a given year. He showed how it could be
used to demonstrate net benefits from the otherwise costly emissions standards for light-duty vehicles.

Dr. Greenstone then summarized the key decisions and results from the interagency working group. He
noted that the interagency process selected three commonly used lAMs to estimate the SCC: DICE,

PAGE, and FUND. For socio-economic inputs and emissions trajectories, the interagency process relied
on scenarios from the Stanford Energy Modeling Forum exercise EMF-22. The working group used four
of the ten models and selected four business-as-usual (BAU) paths and one lower-than-BAU path that
achieves stabilization at 550ppm in 2100. The interagency group parsed the Intergovernmental Panel
on Climate Change (IPCC) Fourth Assessment Report (AR4) to define the constraints of equilibrium
climate sensitivity. They calibrated four distributions to the IPCC constraints and selected the Roe and
Baker distribution. He noted that the interagency group decided to use a global measure of the SCC and
decided against equity weighting. Dr. Greenstone explained that the interagency process uses three
discount rates of 2.5, 3, and 5 percent.

The lAMs were run through 2300 to produce 45 separate distributions of the SCC for a given year. The
distributions from each of the models and scenarios were averaged together for each year to produce
three separate probability distributions for the SCC in a given year, one for each discount rate. The
interagency group selected four SCC estimates for use in regulatory analyses. In 2010, these estimates
are $5, $21, $35 and $65 (in 2007 US$). The first three estimates are the average SCC across models and
emissions scenarios for the three distinct discount rates. The fourth value represents higher-than-
expected impacts. The $21 estimate associated with a 3% discount rate is the central value.

Dr. Greenstone finished with a list of key areas identified for future research and advances in calculation
of the SCC. This list included improvements related to: catastrophic impacts; translating physical
impacts into economic damages; interactions between inter-sector and inter-regional impacts;
adaptation and technological changes; incorporation of risk aversion; and valuing reductions of other
GHGs.

During the question and answer session, one participant criticized the misleading presentation of four
significant figures in the SCC estimates, which gives a highly overconfident impression of precision that is
unfounded when the uncertainty ranges are so large. Another participant criticized the negligible
impacts calculated by the models for 2°C of warming, highlighting the conclusion of the Copenhagen

-------
Accord that this level of warming is dangerous. Dr. Greenstone explained that the process used the best
available evidence on economic damages that were incorporated in lAMs at the time.

Session 1: Overview of Existing Integrated Assessment Models

Session 1 was moderated by Stephanie Waldhoff of the U.S. Environmental Protection Agency and
included presentations by Jae Edmonds, Pacific Northwest National Laboratory; Stephen Newbold, U.S.
Environmental Protection Agency; Christopher Hope, University of Cambridge; David Anthoff, University
of California, Berkeley; Leon Clarke, Pacific Northwest National Laboratory; and John Reilly,
Massachusetts Institute of Technology. The session provided an overview of existing integrated
assessment models, including those used for the development of current U.S. government social cost of
carbon values (DICE, PAGE, FUND), as well as other types integrated assessment models (GCAM, iESM,
IGSM).

Overview of Integrated Assessment Models

Dr. Jae Edmonds presented an overview of integrated assessment models. He noted that lAMs
integrate human and natural Earth system climate science and are useful for three reasons: to provide
insights that would be otherwise unavailable from disciplinary research; to capture interactions between
complex and highly non-linear systems; and to provide natural science researchers with information
about human systems such as GHG emissions, land use, and land cover. He further noted that lAMs
were never designed to model the very fine details, rather to provide strategic insights, for example
about non-linear interactions.

Dr. Edmonds then mentioned the diversity of lAMs that are designed for multiple types of questions and
problems, emphasizing the importance of choosing a model appropriate to the question or problem at
hand. He then distinguished between the highly aggregated lAMs and the higher resolution lAMs.

Highly aggregated models are often used to compare the costs and benefits of policy intervention.

These models are typically composed of three components: emissions, natural Earth systems, and
climate damages. Highly aggregated models often summarize information pulled from other, more
detailed models or from off-line research in order to establish parameter values. The less aggregated,
higher resolution models address a different set of questions associated with the details of the
interactions between human and Earth systems. Higher resolution models are focused on cost-
effectiveness rather than cost-benefit analysis, and are often used to identify the best way to
accomplish a given objective.

DICE

Dr. Stephen Newbold presented a summary of Dr. William Nordhaus' DICE model, beginning with an
overview of its historical development and applications. The DICE model, or Dynamic Integrated
Climate-Economy model, includes an optimal economic growth model, a simplified climate change
model, a damage function that represents the loss of economic output due to increased global surface
temperatures, and the projection of abatement costs over time. The model solves for the optimal path
of savings and abatement to maximize present value of discounted aggregate utility.

-------
Dr. Newbold presented a brief overview of the model's structure, noting its Cobb-Douglas production
function, "three-box" climate model calibrated to MAGGIC, and pure rate of time preference set at
1.5%. He noted that, contrary to how it was used in the interagency process, the social cost of carbon in
DICE is typically calculated along an optimal path, where the SCC equals both the change in consumption
in all future years from one additional unit of emissions in the current year, discounted to present value
using the Ramsey consumption discount rate, as well as the tax on C02 emissions. The damage function
in DICE was developed by choosing a functional form for aggregate climate change damages as a
fraction of global economic output, and then calibrating the damage function parameters using a
summary of empirical studies of climate change damages in all major categories, extrapolating among
regions as necessary.

Dr. Newbold then briefly summarized several updates that have been made in the newest version of the
regional counterpart of the DICE model, RICE2010. RICE2010 includes a few changes in parameters, as
well as a revised set of region-specific damage estimates which are a function of temperature, sea level
rise, and carbon dioxide concentrations. RICE2010 produces a near-term carbon price on an optimal
path of approximately $ll/tonC02 as compared to approximately $7.5 in DICE2007.

During the question and answer session, Dr. Newbold clarified the reasons for differences between
DICE's $7.5 SCC estimate and the estimates developed by the interagency group, noting the different
population scenarios, GDP scenarios, discounting, and especially the probabilistic equilibrium climate
sensitivity distribution used in the interagency process. One participant questioned the value in DICE for
the relative risk aversion parameter, believing it to be many times too small. Dr. Newbold explained
that the values were chosen to match observed market interest and savings rates. Another participant
noted, based on his recent research, that if the relative risk aversion parameter is increased from 1.5 or
2, as in RICE and DICE, to 6, which is implied by some research on the "equity premium puzzle," then
DICE produces very different estimates of the social cost of carbon.

PAGE

Dr. Christopher Hope presented a summary of his PAGE model, including its application to the SCC
calculations. Dr. Hope focused on the PAGE09 model, which represents an update to the PAGE 2002
model used by the interagency working group. The PAGE09 model is written in Excel 2007 with an add-
in module to perform Monte Carlo simulations. It considers methane (CH4), nitrous oxide (N20), and
high GWP gases in addition to carbon dioxide (C02). The model evaluates impacts for eight regions, in
10 particular analysis years through 2200, for different impact sectors and discontinuities. The model
conducts 10,000 runs in Monte Carlo distributions to calculate probability distributions of outputs and is
generally used to compare the benefits and costs of two policy options. Dr. Hope noted that while PAGE
incorporates choices and costs of abatement and adaptation, they are not relevant to the interagency
use of the PAGE model.

Dr. Hope then presented the new features of the PAGE09 version. This version of PAGE includes N20 as
a policy gas, includes sea level rise explicitly, models impacts as an explicit function of per capita GDP,
constrains damages with a saturation line of 100% GDP, allows for the possibility of benefits for small
temperature rise depending on input parameters, and measures impacts and costs as expected utility.

-------
Dr. Hope enumerated several of the uncertainties treated by the PAGE model, including climate
sensitivity response, C02 emission levels (which are only estimated by IPCC through 2100), global mean
temperature rise, and global impacts, all of which influence the long right-tail of the impacts and social
cost of carbon estimates. Dr. Hope demonstrated the major influences and sensitivities of the PAGE
model, showing the model to be most sensitive to the transient climate response (TCR), where a change
in the TCR of one standard deviation could increase the SCC by $60. Dr. Hope finished with a
comparison of outputs from PAGE09 and PAGE2002 given the same set of inputs, showing that PAGE09
produces a mean SCC estimate of $100/tonCO2 where PAGE2002 produced a mean estimate of $28. He
noted that the increased impacts in PAGE09 can be attributed to the following characteristics of the new
model: less effective adaptation, a higher chance of a discontinuity, better incorporating the possibility
of very large impacts, and the use of 2005 dollars instead of 2000 dollars.

During the question and answer session, Dr. Hope explained that the extent of the time horizon and
future assumptions are extremely important to the estimates produced by PAGE. For example, if the
time horizon is extended to 2300, even when keeping emissions constant, the SCC estimate is increased
by 20%. One participant raised the point that all of the lAMs incorporate the hidden assumption that
damages are multiplicative which introduces an important bias. Finally, Dr. Hope clarified that the
saturation line for damages of 100% GDP only becomes relevant in a very small number of model runs,
under extreme parameters. He underscored the importance of looking at the full distribution of outputs
rather than a single run when using the PAGE model.

FUND

Dr. David Anthoff then presented a summary of the FUND model, including a description of its basic
structure. Of the three models used by the interagency working group, FUND is the most disaggregated,
with 16 regions, multiple gases, and damage functions that are specified for numerous sectors. The
model includes: a reduced form carbon cycle model for C02, CH4, sulfur hexafluoride (SF6), and sulfur
dioxide (S02); a model to translate greenhouse gas concentrations into temperatures that incorporates a
temperature lag; an ocean model to estimate sea level rise; a biodiversity model to estimate species
loss; an impacts model with impacts based on temperature, sea level rise, species loss, and greenhouse
gas concentrations; and feedbacks where the economic damages of climate change affect the economy
growth rate. In FUND, exogenous variables include GDP, population, energy and carbon intensity, C02
emissions from land use change and deforestation, CH4 emissions, and N2Oemissions. Endogenous
variables include C02 emissions, C02 emissions from natural feedbacks in the "dynamic biosphere", SFs
emissions, and S02 emissions. All of the gas cycles and radiative forcing for each gas are modeled
explicitly, while climate sensitivity is an uncertain distribution.

Dr. Anthoff then presented the impacts that are modeled in FUND, listing: the components of the health
impacts model; the components of sea-level rise impacts as based on the analytical structure of
Fankhauser (1994)4; and other impact categories, including agriculture, tropical storms, extra-tropical
storms, forestry, heating energy, cooling energy, water resources, and species loss. For each impact

4 Fankhauser, S. (1994). "Protection vs. Retreat - The Economic Costs of Sea Level Rise." Environment and Planning
A 27(2): 299-319.

-------
sector, FUND includes a separate damage function that depends on the temperature predicted for that
region and year. He noted that the sign of each impact could vary with geographic location and impact.
The outputs of these damage functions are summed to aggregate impacts. Dr. Anthoff then presented
the planned model modifications for FUND, which include: additions of impacts for ocean acidification,
tourism, and river floods; an update to the energy consumption impacts; and a thorough evaluation of
catastrophes.

Dr. Anthoff finished his presentation with a discussion of the interagency working group's use of FUND.
He explained that he liked a lot of the working group's choices but pointed out three areas in which the
models offer more than what was captured by the interagency process. He indicated that the working
group estimates could be improved by incorporating: a fuller distribution of scenario uncertainty than
the five EMF socio-economic scenarios; endogenous, non-constant discounting where the discount rate
is related to the economic growth rate; and equity weighting to better capture the uneven distribution
of climate change impacts.

During his presentation, Dr. Anthoff distinguished between two types of transparency in lAMs. He
noted that in simpler models like DICE the simple damage function is itself easier to grasp, however the
damage function's foundation and link to underlying studies is less clear. In contrast, in more
complicated models like FUND, the damage functions themselves are more complicated, but their
foundation and link to underlying studies is clearer.

During the question and answer session, one participant questioned the net benefits modeled by FUND
for the first 3 degree Celsius temperature increase, attributing the benefits to agricultural sector
benefits based on research from the early 1990s and health benefits from reduced cold weather deaths.
Dr. Anthoff explained that FUND does not conduct primary impact studies, instead basing impacts on
the existing literature. He further explained that climate damages produce differentiated impacts across
the globe with poor countries most negatively affected. Without equity weighting, he explained, these
damages do not significantly impact the aggregate. Finally, he noted that the social cost of carbon is
related to marginal damages, not total damages, so it is the slope of the damages curve rather than the
absolute value of damages that is important. Another participant agreed with the first participants'
criticism of near-term net benefits, but noted that PAGE also produces some near-term benefits and
there is the added consideration of weather variability. The same participant proposed that the low
slope of the damage function indicates that FUND'S bottom-up approach, while good, is missing some
key aspects.

GCAM and Development of iESM

Next, Dr. Leon Clarke presented the climate impacts representation in GCAM, which is an example of
one of the higher resolution lAMs described by Dr. Edmonds. Dr. Clarke explained that GCAM is a
dynamic-recursive model that includes a climate model based on MAGICC and the energy-economy
model developed by Dr. Edmonds and Dr. Reilly. While the model's basic inputs are similar to the more
aggregate models, GCAM includes a much higher level of detail for each sector. For example, GCAM
includes detail related to energy system resources, technology assumptions, demand technologies, and
agricultural productivity. Dr. Clarke noted that GCAM is particularly useful for examining impacts that

-------
involve interactions among the various systems represented in lAMs. However, he also noted that
aggregating and monetizing all impacts is not a core objective of GCAM or similar, higher-resolutions
lAMs.

Dr. Clarke included a list of priorities for incorporating impacts into PNNL/JGCRI's integrated assessment
modeling. He outlined ways for pursuing these developments, including one dimensional integration
(either all within GCAM or through linkages with other sector-specific models) and incorporating
feedback with other systems by endogenizing interactions within the model or leaving them "hanging"
off of GCAM. Dr. Clarke then presented three examples of areas where GCAM has been used to model
impacts in a more detailed way, related to land use, energy, and water.

Dr. Clarke then provided two examples of linkages between platforms: the integrated Earth System
Model (iESM) and the regional initiative. iESM is a research collaboration between the Pacific
Northwest National Laboratory (PNNL), Oak Ridge National Laboratory (ORNL), and Lawrence Berkeley
National Laboratory (LBNL). The effort has three primary tasks: to create a first generation integrated
Earth System Model linking the human system components of GCAM to a physical Earth System Model
(ESM), the Community Earth System Model (CESM); to further develop components and linkages within
the iESM and apply the model to improve our understanding of the coupled physical, ecological, and
human system; and to add realistic hydrology. Dr. Clarke noted that running GCAM without linkages to
CESM takes approximately 20-30 minutes, but running GCAM with linkages and feedbacks can take as
long as months. The regional initiative is an effort to integrate more detailed regional models into
GCAM (e.g., the crop model EPIC or the whole building engineering model BEAMS).

During the question and answer session, one participant questioned the short-sighted, or "myopic",
nature of recursive-dynamic models, particularly challenging the lack of oil price modeling. Dr. Clarke
clarified that "recursive-dynamic" means that GCAM establishes market equilibrium at each time step
before moving forward. He also noted that the oligopic nature of oil is not modeled in the lAMs.

IGSM

Dr. John Reilly concluded the presentation portion of Session 1 with an overview of the MIT Integrated
Global System Model (IGSM). Dr. Reilly explained that the IGSM is a general equilibrium economic
model with a full inter-sectoral structure. The model includes: impacts from numerous sectors
including, agriculture, forestry, hydrology, trace gas fluxes, sea level change, land use change, and
human health effects; a robust climate model with atmosphere, urban, ocean, and land components;
and model outputs that include GDP growth, energy use, policy costs, global mean and latitudinal
temperature and precipitation, sea level rise, sea-ice cover, and net primary productivity. The model
includes numerous feedbacks and interactions between the economic model and the dynamic terrestrial
ecosystems model. Dr. Reilly noted that the model includes and values the benefits and costs of
adaptation, as well as both market and non-market (e.g., leisure) damages.

Dr. Reilly then discussed the characterization of uncertainty in the IGSM. Uncertainty in the model
arises from: emissions uncertainties (due to uncertain socio-economic inputs); climate system response
uncertainties; and greenhouse gas cycle uncertainties. Dr. Reilly discussed the impacts of different

-------
stabilization targets, including the likelihood of different levels of temperature increase under each
policy. He showed probability distribution functions for five different policy scenarios. He presented an
uncertainty analysis that showed that the five cases used in the interagency process are conservative
estimates of C02 concentration projections and do not capture the full range of IGSM estimates. Dr.
Reilly also compared the IGSM scenarios to the IPCC SRES scenarios for global mean temperature
change. Again, the IPCC results show a low bias and do not cover the full range of IGSM estimates. Dr.
Reilly concluded that the higher impacts estimated by IGSM as compared to IPCC indicates that looking
at the issues in an integrated way can produce different answers than looking at the issues individually.

Session 1 Discussion

Following Dr. Reilly's presentation, the discussion portion of Session 1 began. One participant noted the
importance of lAMs as an essential tool. Acknowledging the difficulty of developing lAMs, he criticized
the narrowness of the current lAMs, particularly regarding incorporation of damages. He noted the
current lAMs' large emphasis on agriculture damages but highlighted the old and new literature that
goes beyond agricultural damages. He noted that the damage levels currently modeled in the lAMs
equate to the world reaching a given GDP level in 2103 instead of 2100, an insignificant change. The
participant suggested that the lAMs should be broadened to incorporate effects such as changes in
savings, investment, and growth rates, and perhaps even things like political stability.

Dr. Hope responded by noting that first, there is an advantage to not disaggregating sectors in that
damage functions are more easily updated, and second, that integrated assessment modelers cannot
claim to do the primary research, rather they incorporate other primary research and build in
uncertainty. He noted that the only thing from the participant's discussion not included in PAGE is the
political stability component, but he noted that if research quantifying political stability impacts existed,
the model could incorporate it.

Another participant criticized the estimation of damages in terms of GDP, arguing it is not a good
measure of human welfare. Dr. Reilly indicated that aspects of welfare are included and that a proper
welfare analysis is done with consumption of different goods and their substitutability specified. Dr.
Anthoff noted that the FUND damage functions are not quantified as a percent of GDP, but as a welfare
loss equivalent to certain consumption loss.

Another participant asked whether there was any way to verify the models given that they are dealing
with unprecedented conditions. Dr. Hope noted that verification is much more difficult for economic
models than for Earth system models, and that more time, money, and research is needed to explore
the issue. Dr. Reilly suggested focusing on mechanistic approaches. Another participant wondered
whether the models could be verified through historical runs projecting forward to today. Dr. Reilly
explained that there are so many degrees of freedom in the model, it is very easy to force the model to
replicate historical events by adjusting input parameters.

Another participant discussed the vast uncertainty and guesswork involved in the lAMs and SCC
estimates. He questioned how best to proceed given the unprecedented uncertainty around the SCC
estimates. He proposed several options, including: forging ahead and producing a number; admitting

-------
the uncertainty is too great and avoiding the exercise altogether; or some hybrid. He further questioned
the applicability of cost-benefit analysis for climate policy decisions. Dr. Hope argued that despite the
uncertainty, it is still beneficial to estimate the SCC. However, it is crucial to always present a range of
values and an explanation of what is and is not included in the estimate, as well as an explanation of
what information is needed to narrow the range.

Finally, a participant asked first how best to characterize various uncertainties that have not yet been
extensively examined quantitatively in the literature (e.g., damages at higher temperatures, degree of
reversibility of impacts and damages), and second, about the importance of feedbacks to growth and
discount rates, noting that only one model incorporates such feedbacks. Dr. Newbold commented that
feedbacks to growth and discount rates are very important if discounting is tied to consumption growth.
Dr. Hope commented that negative discount rates might even be necessary if climate change welfare
effects are significant enough, noting they are exploring the idea of negative discount rates in the latest
version of PAGE. He also commented on the need for a high quality assessment of what the impacts
would be of a much more extreme temperature increase than the typically analyzed 2 or 3 degree C
increase. Finally, Dr. Anthoff commented that existing impact studies only examine a narrow range of
temperature impacts, but that anything beyond these ranges must be extrapolated. He noted that
eventually, assumptions must be made in order to extrapolate to more extreme temperatures, but that
it would be best for the impact scientists to be involved in this exercise.

Session 2: Near-Term DOE and EPA Efforts

Session 2 was moderated by Ann Wolverton of the U.S. Environmental Protection Agency and included
presentations by Robert Kopp, an American Association for the Advancement of Science (AAAS) Science
& Technology Policy Fellow hosted by U.S. Department of Energy; Nisha Krishnan, Resources for the
Future/ICF International; and Alex Marten, U.S. Environmental Protection Agency. The session provided
an overview of near-term DOE and EPA Efforts, including the DOE proposed impacts knowledge platform
and the EPA generalized modeling framework.

Proposed Impacts Knowledge Platform

Dr. Bob Kopp began the presentations by introducing the possibility of an impacts knowledge platform.
This platform would constitute an effort to help overcome the barrier between natural scientists and
economists, to help economists understand and use the best available natural science. Developers of
the platform are working to identify which data should be included and what is needed to inform local
and regional policy making.

Ms. Nisha Krishnan then presented the Global Adaptation Atlas, an existing adaptation planning and
research initiative that DOE partially funded to help inform the consideration of an impacts knowledge
platform. Ms. Krishnan explained that the Adaptation Atlas, which is intended to inform policy making,
is currently in beta form, online, and available (at http://www.adaptationatlas.org/). The Atlas currently
contains twenty studies from the peer reviewed literature on different human impacts of climate
change. The Atlas is a web-based application that enables user-driven, dynamically-generated maps of
climate impacts and adaptation activities, where the user is able to select a location, timeframe, and
scenario and view a map corresponding to their decision filters.

-------
Ms. Krishnan explained that the Atlas was assembled by soliciting data and study results from
approximately 300 studies, which returned only 20-30 responses. She noted that researchers seemed
hesitant to share data, even from peer-reviewed studies. Solicitations focused on five sectors: food,
water, land, health, and livelihood. The data was then translated into a visual, spatial format; every
layer was tagged with IPCC scenarios, timeframes, and locations; and 'meta' filters were applied to
harmonize across time, theme, and assumptions so that the layers could be combined in a simplistic
overlay. Ms. Krishnan explained that the Atlas also attempted to investigate uncertainty, but received
only one response from their solicitations. The Atlas only incorporates sensitivity analysis, which should
be incorporated into the online tool by the end of 2011.

Proposed Generalized Modeling Framework

Dr. Alex Marten then described a preliminary scoping study by EPA to develop a generalized modeling
framework. Dr. Marten explained that the idea arose from the interagency SCC process, and is intended
to explore ways to provide a more transparent and standardized modeling framework that could more
easily incorporate existing and future research on climate science and economic damages. Ideally, such
an approach would also allow for a better understanding of the sources of differences in SCC estimates
and the drivers of model results. Dr. Marten also emphasized the importance of providing detailed up-
to-date documentation and of designing the model code to be open source and freely available to the
public.

Dr. Marten identified the following key characteristics for a more generalized modeling framework:
general and flexible enough to incorporate new research and to nest other commonly used lAMs; fully
transparent; probabilistic; and modular to allow replacement of components over time. Dr. Marten
then provided a brief overview of a prototype for such a framework, highlighting its similarities to other
commonly used lAMs; its current use of MAGICC, a relatively robust climate model compared to some
reduced form models currently being used in lAMs; its potential to represent natural capital; and its
potential to include climate-population feedbacks and endogenous emissions. Dr. Marten explained
that such a framework may be designed to carefully distinguish between several different types of
climate change damages (e.g., market based with sectoral breakdowns, direct capital destruction,
consumption equivalent health damage, etc.) for transparency and accuracy. Dr. Marten emphasized
the concept of creating a general framework as a way to better facilitate incorporation of new research
on climate change-induced damages, as the research becomes available.

Dr. Marten noted that the framework is in an early prototype stage. The basic architecture of the
framework is being tested by using specific parameter settings intended to closely approximate versions
of DICE, PAGE, and FUND as used by the interagency workgroup. Dr. Marten then identified further
steps that would be required for the framework to become fully functional, including: expanding and
modifying the model structure based on feedback from the workshop participants and other informal
reviewers, incorporating currently available and new studies on climate change damages as they are
published, external peer review, and eventual public release.

-------
Session 2 Discussion

During the discussion section, one participant commended the idea of a generalized modeling
framework noting it should be feasible. He also underscored the importance of openness, and criticized
the lack of EPA and DOE policy requiring the projects they fund to be open source. He suggested that
opening up the process would encourage interest in the topic and reduce barriers to entry into the field.
Dr. Kopp responded that DOE has been supporting some efforts to make the process more open.

Another participant noted that the components of the generalized modeling framework are very similar
to FUND, suggesting EPA draw on the capabilities of FUND in developing this framework and noting that
the challenges are programming questions not scientific questions. Another participant noted the
community integrated assessment model in Europe that is looking at non-linear changes and stochastic
models, suggesting it might also be helpful to build on.

Another participant suggested moving away from matching or incorporating existing models as the
existing models need significant improvement and use old research. He highlighted the almost
unanimous comments from the workshop participants indicating a significantly new approach is needed.
Dr. Marten explained that the standardized models are intended to facilitate comparison of existing
models and incorporation of new science. Dr. Wolverton noted the need to change the structure of the
models as well as the underlying science.

Another participant questioned the use of lAMs generally and wondered if it might be worth talking to
OMB about alternative tools. Dr. Wolverton underscored the involvement of OMB in the 2009-2010 full
interagency process, as well as the inclusion of the workshop discussion in future interagency
discussions of the SCC.

One participant highlighted the simplicity of the lAMs, particularly as compared with climate models.
She contrasted FUND, a model built by two people, with climate models that have large teams and $5
million per year for updates and maintenance. She suggested two options moving forward. One option
would be to continue developing what she called "toy models" to transparently run assumptions.
Another option would be to highlight the importance of the exercise and outline exactly what would be
required to develop the models properly.

Finally, a last participant emphasized the need for more basic impacts studies before working to
improve the models themselves.

Session 3: Critical Modeling Issues in Assessment and Valuation of Climate
Change Impacts

Session 3, Part 1

Session 3 was split into two parts occurring in the afternoon of Day 1 and the morning of Day 2. The first
part of Session 3 was moderated by Ann Wolverton of the U.S. Environmental Protection Agency and
included two presentations by Ian Sue Wing, Boston University, one as a replacement for Karen Fisher-
Vanden, Pennsylvania State University, as well as a presentation by Brian O'Neill, National Center for
Atmospheric Research. The session began to explore critical modeling issues in assessment and

-------
valuation of climate change impacts, including: sectoral and regional disaggregation and interactions,
adaptation and technological change, and multi-century scenario development and socio-economic
uncertainty.

Sectoral and Regional Disaggregation and Interactions

Dr. Ian Sue Wing started the Session 3 presentations with a discussion of the sectoral and regional
representation of economic damages in integrated assessment models. Dr. Sue Wing presented the
basic structure of lAMs as a three model structure including an economic model, climate model, and
impact model. He then presented the set of nine disaggregated region- and sector-specific equations
that would be used to construct an 1AM in the absence of resource limitations. He noted that
researchers are most knowledgeable about the economic model components, with 40 years of
experience; relatively knowledgeable about the climate components, with 20-25 years of experience;
and least knowledgeable about the impact model, which is relatively new and the centerpiece of the
workshop's discussion.

Dr. Sue Wing then walked through the nine equations, noting which components comprised each
equation. He highlighted the increasing uncertainty and unknowns as he progressed from the economic
model to the climate model and then to the impact model. He noted the need to separate damages and
costs, creating two separate response surfaces that are multiplicative.

Dr. Sue Wing noted that in the absence of resource limitations, lAMs would be constructed with sectoral
and regional detail in production, consumption, and climate damages. He explained that impacts would
first be elaborated by category of physical endpoint, sector, region, and future time period, based on
simulated climatic changes at the regional scale. Only then would the models aggregate across
endpoints to generate sector-by-region trajectories of shocks. Instead of aggregate damage functions,
the models would incorporate a transparent causal chain from both ex ante shocks and ex-post
adjustments in regional/sectoral output and consumption to ultimate welfare effects.

Dr. Sue Wing noted that in current models, particularly DICE, the complexity and dimensionality of the
issue has been boiled down and combined, with the models dependent only on temperature. Dr. Sue
Wing then enumerated the many difficulties in attempting to build his idea of an ideal model,
emphasizing the lack of empirical or detailed modeling studies, particularly studies that go beyond 2050.
He noted the inherent difficulty in maintaining detailed estimates given increasing uncertainty as
projections extend further forward in time. Dr. Sue Wing identified computable general equilibrium
(CGE) models as a promising new direction, particularly given their increasing skill at regional scales and
their explicitly multi-regional/multi-sectoral approach. However, he also noted their problematic
recursive-dynamic (and therefore myopic) nature and limited time horizon.

During the question and answer session, one participant challenged the notion that intertemporal
valuation is done well and asked how ecosystem services are represented. Dr. Sue Wing suggested
ecosystem services be valued using a Ramsey framework specified with ecosystem service constraints.
The participant commended the answer on how to incorporate ecosystem services but noted there is
generally little knowledge about the welfare derived from non-monetized services, such as ecosystem

-------
services in a climate change context. Dr. Sue Wing acknowledged the current lack of knowledge but
indicated there are ways to make progress. Another participant asked about climate impacts damages
and the regional and local specificity from the perspective of infrastructure risk. Dr. Sue Wing explained
that climate damages can be set to change capital accumulation by reducing investment rates or directly
destroying capital stocks. However, he noted the difficulty associated with projecting specificity into the
future.

Adaptation and Technological Change

Dr. Ian Sue Wing then presented the effects of adaptation and technical change on the SCC, on behalf of
Dr. Karen Fisher-Vanden, who was unable to attend the workshop due to illness. He noted numerous
challenges to incorporating adaptation: the inherent difficulty in modeling adaptation, requiring
advancements in modeling techniques; the limited coverage of empirical work on adaptation and
additional difficulty of incorporating the studies into lAMs; and the lack of adaptation-related
technological change in current lAMs. He emphasized the critical need for empirical studies, as well as
research focused on bringing the results from state-of-the-art empirical studies into modeling
frameworks.

Dr. Sue Wing then walked through the important model features needed to represent adaptation, given
the unique characteristics of the adaptation process. In order to incorporate adaptation, models need
to include: explicit modeling of climate damages and impacts so that reactive expenditures and
proactive investment can be estimated; inter-temporal decision making under uncertainty; endogenous
adaptation-related technological change, as distinguished from mitigation-related technological change,
(which differs in the nature of inducement and the public versus private nature); regional and sectoral
detail since adaptation occurs on local and regional scales; and a connection with empirical work on
impacts and adaptation.

Dr. Sue Wing then examined existing lAMs, noting the four models that deal with adaptation: AD-
WITCH, AD-DICE/AD-RICE, PAGE, and FUND. He noted that only three of the four models are inter-
temporal and only one (AD-WITCH) has proactive adaptation. Dr. Sue Wing then identified the three
main existing empirical summary studies on adaptation and recommended four areas for future
research: decision making under uncertainty; adaptation-related technological change; empirical work
on adaptive capacity; and dynamics of recovery.

During the question and answer session, one participant encouraged the modelers to consider and
incorporate suffering in addition to mitigation and adaptation. Dr. Sue Wing acknowledged that
suffering was missing from the models in their current state using aggregate output good. He suggested
that suffering be incorporated using the regionally and sectorally disaggregated approach, but noted the
difficulty with monetizing effects on culture. Another participant commented on the difficulty in
separating adaptation from other capacity-building exercises, particularly in developing countries. He
also commented on the purely theoretical progress in incorporating adaptation, again calling for more
empirical studies.

-------
Multi-century Scenario Development and Socio-Economic Uncertainty

Dr. Brian O'Neill delivered the last presentation of the day, on multi-century scenario development and
socio-economic uncertainty. He emphasized the vast uncertainty and the importance of years beyond
2100 in SCC estimates. He then presented the assumptions made by the interagency SCC process, along
with alternate estimates that could have been assumed. He explained that the interagency process
used five EMF-22 scenarios, which they extended to 2300 using simple methods. Dr. O'Neill presented a
series of graphs that independently plotted the interagency projections for global population, GDP, and
carbon dioxide emissions along with alternate projections. These graphs demonstrated the narrow
range of uncertainty captured by the interagency process - which sought to capture a wide range of
emission estimates, combined with reasonable and internally consistent assumptions for the other two
factors - compared to estimates of each factor when analyzed independently.

Dr. O'Neill showed that the global population estimates to 2100 used by the interagency process
captured significantly less uncertainty than the estimates produced by the IPCC Fourth Assessment
Report (AR4), the United Nations (UN), and the International Institute for Applied Systems Analysis
(NASA). Dr. O'Neill then demonstrated that the interagency estimates capture an even smaller portion
of the range of UN and NASA estimates when examining global population to 2300. He noted that the
UN long-run estimate that aligns with the interagency estimates is not the most likely scenario, rather a
mathematical benchmark to produce roughly stable population size.

Dr. O'Neill then presented a similar story regarding global GDP. He showed that as compared to the
IPCC AR4 estimates, the interagency process captured a small portion of the range of possible estimates
for GDP to 2100. Compared to a study projecting GDP to 2300, the interagency process only captured a
tiny fraction of the range of estimates - the uncertainty in the study was orders of magnitude larger
than the uncertainty in the interagency process.

Dr. O'Neill finished by showing the interagency scenarios did a better job of capturing the range of
estimates for carbon dioxide emissions through 2100. The interagency estimates for emissions through
2300 covered a higher and wider range than the Representative Concentration Pathways (RCPs). Dr.
O'Neill concluded that the interagency process captured an overly narrow range of uncertainty in
population and GDP over the entire time horizon, especially in the long term, but was reasonably
consistent with the range of emissions in the literature.

Dr. O'Neill listed many issues with multi-century scenario development, noting the fact that uncertainty
ranges in the literature might themselves be too conservative given the vast unknowns of predicting 300
years into the future. He recommended demonstrating the key sources of uncertainty, using full
uncertainty instead of a range of best estimates, considering a substantially wider range of socio-
economic futures through 2100 and 2300, considering simpler approaches to damages in the very long
term, improving how uncertainty in results is characterized, and considering linking to the evolving work
on RCPs and socio-economic scenarios consistent with them.

-------
Session 3, Part 1 Discussion

Following Dr.O'Neill's presentation, the discussion portion of Session 3, Part 1 began. One participant
noted that adaptation should depend on the rate of temperature change, not just temperature.

Another participant defended the models, noting that FUND impacts do depend on the rate of change in
some sectors and that non-market impacts, such as health impacts, are incorporated in models such as
PAGE and FUND. Dr. Sue Wing clarified the distinction between quantifiable non-market impacts and
non-quantifiable non-market impacts such as cultural loss.

Another participant questioned the seeming lack of constraints in the population predictions presented
by Dr. O'Neill. Dr. O'Neill attributed the vast population increases to technological change, explaining
that it was probably hard to imagine 8 billion people on the planet when there were only 500,000.

In response to another question, Dr. Sue Wing recommended representing the elasticity of substitution
dynamically, to capture adaptive capacity.

Another participant questioned the relationship between population and GDP, particularly the
possibility of a low population, high GDP world. Dr. O'Neill clarified that there is no widely accepted
theory between population growth and GDP. The same participant recommended caution in linking the
SCC exercise to RCPs, as the assumptions may differ. He then underscored the importance of ensuring
that assumptions about economic growth are consistent with or feed into the assumptions about
discounting in a Ramsey framework. A different participant noted the need to examine vulnerable
populations within developed countries. Dr. Sue Wing indicated that in addition to more regional
impacts work, there is a need for quantitative historians to quantify damages from historic impacts.

One participant commented that the criticisms of lAMs are great for the modelers to hear, even if not all
are well-deserved. He noted that the importance of scenarios after 2100 also depends on the lifetime of
gases. And finally, he explained that the modelers' choice to narrow uncertainty in population and GDP
was likely a choice to develop reasonable estimates out of profound uncertainty. Dr. O'Neill responded
that clearly communicating uncertainty was critical. The ensuing discussion concluded that even though
projecting through 2300 is very difficult, it is nonetheless important if conditions after 2100 have a
significant effect on results. One participant suggested the only option was to use theoretical, likely
Bayesian techniques to do so. Dr. O'Neill added that the marginal nature of SCC estimation constrains
the conversation, noting the models can be used for other purposes.

One participant noted that a sense of urgency needs to enter the conversation given the small window
of time left to act to address climate change and the importance of these estimates in potentially
influencing the stringency of U.S. regulations. Instead of continuing with incremental adjustments to
SCC estimates, she argued for the addition of normative economics to value things like culture. A final
participant noted that if we continue to emit significant amounts of carbon dioxide, our climate future is
known. He cautioned that even proactive adaptation may not work.

Session 3, Part 2

Session 3 resumed on Day 2 after brief opening comments from Elizabeth Kopits, U.S. Environmental
Protection Agency. The second part of Session 3 was moderated by Robert Kopp on behalf of the U.S.

-------
Department of Energy and included presentations by Gerard Roe, University of Washington; Martin
Weitzman, Harvard University; Timothy Lenton, Unversity of East Anglia; Michael Toman, World Bank;
and Michael Hanemann, University of California, Berkeley. The session continued to explore critical
modeling issues in assessment and valuation of climate change impacts, including incorporation of
climate system uncertainty, extrapolation of damage estimates to high temperatures, Earth system
tipping points, potential economic catastrophes, and nonmarket impacts.

Incorporation of Climate System Uncertainty into IAMs

Dr. Gerard Roe presented an overview of what we do and do not know about climate projections. He
started by stating that given the complexity of the weather and climate systems, any knowledge and skill
regarding climate change is remarkable. Dr. Roe underscored the fact that uncertainty does not imply
ignorance. Dr. Roe then discussed the concept of climate sensitivity, "the long-term change in annual-
mean, global-mean, near-surface air temperature to a doubling of C02 above preindustrial values",
which is used as the benchmark to compare different estimates. Dr. Roe presented several different
estimates of climate sensitivity, showing the long right tail of estimates.

Dr. Roe then demonstrated that climate sensitivity is uncertain because the magnitude of past forcing,
particularly the forcing of aerosols, is uncertain. Through a series of graphs, he showed that all of the
variables in the global energy budget equation, (global mean temperature change, greenhouse gas
warming, and ocean heat storage) are well-observed and well-constrained, except for the cooling effect
from aerosols. This uncertain cooling effect leads to uncertainty in total climate forcing. Dr. Roe then
showed that dividing the well-constrained temperature change by the poorly-constrained climate
forcing results in the fat-tail of climate sensitivity. Dr. Roe further demonstrated the source of climate
sensitivity uncertainty through use of classic feedback analysis models. Dr. Roe noted that the prospects
for narrowing climate sensitivity uncertainty are limited.

Dr. Roe then presented projections of the climate commitment, if all anthropogenic emissions were to
cease immediately. He explained that uncertainty in the climate response to current concentrations
arise from the uncertainty in climate (aerosol) forcing. If radiative forcing has been high, climate
sensitivity is low, and the temperature response could be lower than expected. However, if radiative
forcing has been low, climate sensitivity is high, and the temperature response could be higher than
expected. Dr. Roe concluded that uncertainty in climate sensitivity and climate forcing are not
independent.

Next, Dr. Roe presented the transient evolution of climate impacts, showing that if climate sensitivity is
high, it will take the climate a long time to adjust. This is due to the diffusive nature of ocean heat
uptake and the slow, extended growth of the fat tail. Dr. Roe then explained that fixed carbon dioxide
stabilization targets are an inefficient way to achieve a climate goal. Instead, policies should be
implemented, observed, and then adjusted appropriately. He suggested that a flexible emissions
strategy that adjusts over time could significantly reduce risk and uncertainty, and may be more cost-
effective than rigid policies. Finally, Dr. Roe showed that global climate averages are not strong
predictors of local climate change.

-------
During the question and answer session, one participant underlined the significant unknowns under a
high sensitivity trajectory and the need to fully flesh out the flexible emissions strategy suggested by Dr.
Roe. The value of policy flexibility depends crucially on the feasibility of learning more about key
uncertain parameters in a reasonable span of time. Another participant raised the issue of bio-geo-
chemical feedbacks and their effect on results. A third participant pointed out that the policies under a
flexible emissions strategy would look the same as current policies at the present time.

A final participant suggested that given the uncertainty caused by aerosols, the best way to gather
information and knowledge about climate would be to simply turn off aerosol emissions. Dr. Roe
agreed, noting that a decade would be needed to see the full effects. Dr. Kopp noted a recent paper in
Nature Geoscience on the learning that could occur by turning off aerosols.

Extrapolation of Damage Estimates to High Temperatures: Damage Function Shapes

Dr. Martin Weitzman then presented the issue of damage function shapes, particularly when examining
extreme temperature increases. Dr. Weitzman started by presenting the complicated and challenging
nature of the valuation exercise. He described a long chain of tenuous inferences and deep,
fundamental uncertainties on which impacts valuation relies. Acknowledging that the current models
are reasonable in their assumptions, he explained that very different results can be produced with a
different set of reasonable assumptions. He noted, in particular, the sensitivity of the estimates to how
the tails are modeled and incorporated.

Dr. Weitzman continued by challenging the basic functional form of the damage functions. He argued
that the greatest need to improve the lAMs is not for empirical studies, rather for a reevaluation of the
fundamental structure of the models and damage functions. He questioned the approach of using
quadratic damage functions, criticizing their low reactivity by highlighting an example where a 12 degree
temperature increase only reduces output by 26 percent. He noted the high degree of substitutability
between consumption and avoided impacts in current models, suggesting that an elasticity of
substitution lower than one would greatly influence model results.

Dr. Weitzman then made a series of suggestions. He suggested that it is important to investigate the
influence of extreme events, noting that model results depend non-robustly on seemingly obscure
assumptions such as tail size, functional forms, parameters, and the pure rate of time preference. Dr.
Weitzman suggested that the uncertainty with using cost-benefit analysis to estimate the SCC be
communicated clearly and openly. He suggested that, despite the large inability to estimate extreme
tail behavior and welfare disasters, it would still be beneficial to invest in research in these areas. He
suggested that the fat tail risks of proposed solutions (e.g., nuclear power, carbon capture and
sequestration) be considered alongside the fat tail risks of climate change. He suggested that the worst-
case scenarios in the fat tails of climate impacts provide reason to develop emergency backstop
geoengineering solutions. Finally, Dr. Weitzman concluded by suggesting we hope for the best and
prepare for the worst.

During the question and answer session, one participant seconded the call for backstop research that
will help to promote the ability to undertake mid-course corrections. Dr. Weitzman supported this,

-------
arguing that climate change has the probability of being the worst fat-tailed issue. Another participant
noted that even if the climate trajectory follows the mid to low IPCC projections, the consequences
could be disastrous. He argued that geoengineering is the biggest fat tail problem, with the possibility of
disaster outcomes. He suggested focusing the discussion more on known problems and less on
speculative issues. A third participant noted the huge potential health effects of geoengineering
solutions.

Earth System Tipping Points

Next, Dr. Tim Lenton discussed the issue of Earth system tipping points, which he explained are not
necessarily high impact, low probability events, but may be high impact, high probability events. Dr.
Lenton began with a definition of tipping elements and tipping points; where a tipping element is a
component of the Earth system, at least sub-continental in scale (~1000km), that can be switched, under
certain circumstances, into a qualitatively different state by a small perturbation; and a tipping point is
the corresponding critical point at which the future state of the system is qualitatively altered. He then
presented historical examples of abrupt climate changes, including bifurcations, noting that the
Holocene has been unusually stable so far. Dr. Lenton then explained that policy-relevant tipping
elements are those where: human decisions this century determine whether the tipping point is
reached; the change will be observed this millennium; and a significant number of people care about the
system.

Dr. Lenton then provided several examples of policy-relevant tipping points, including their estimated
proximity in time, or probability of occurrence with increasing levels of global warming above the
present temperature. Dr. Lenton explained that the probability of tipping points being reached under
three different warming scenarios was established using imprecise probability statements elicited from
experts. Experts were asked what the probability of reaching a given tipping point was under the three
different scenarios. Dr. Lenton then presented several examples of tipping elements with the
corresponding likelihood of occurrence based on expert elicitation. His examples of tipping elements
included the Greenland ice sheet, the West Antarctic ice sheet, the Amazon rainforest, and El
Nino/Southern Oscillation. He noted that it is important to assess rate and reversibility, as well as
proximity, when identifying the most policy relevant tipping points. For example, the expert elicitation
indicates that melting of the Greenland ice sheet, melting of arctic summer sea ice, and Amazon dieback
are some of the more near-term thresholds that we face. However, the consequences of crossing a
tipping point are not generally felt immediately when a tipping point is crossed. For example, although
the Greenland ice sheet might be set on an irreversible path to near-complete destruction, the
completion of the process would likely take several centuries. The length of this timescale, across which
the effects of a tipping point are felt, is a key trait affecting policy relevance.

Dr. Lenton then indicated that according to the expert elicitation, there is a 16 percent probability that
one of five tipping points will be passed under 2-4°C warming and a 56 percent probability that one of
five tipping points will be passed under 4°C warming. He explained that there may also be interactions
between tipping points including both positive and negative feedbacks. For example, a weakening of
the Atlantic thermohaline circulation could end up disrupting the seasonal onset of the West African
Monsoon, which in one model could lead to a greening of the region, a rare positive impact. The

-------
strengthening of the Indian summer monsoon is a possible tipping point that is perhaps more sensitive
to aerosols than to temperature changes. GHG impacts on this tipping element are likely being offset by
the already occurring brown haze in the region. Finally, Dr. Lenton included several prospects for early
warning signals, which could help societies manage the risk posed by tipping points. These include
slowing down of a climate system (e.g., lower frequency of oscillation), increasing variability, and
skewness of response.

During the question and answer session, one participant suggested that the dieback of ocean
phytoplankton might be a candidate as a tipping element. Another participant questioned the
classification of changes as tipping points, distinguishing elements that involve tipping physics from
elements that are simply subject to large changes. Dr. Lenton agreed with the distinction. As an
example, he noted summer ice melt involves fluctuation, not bifurcation; but winter- or year-round- ice
melt is actually a switch to an alternate state. Dr. Lenton further noted that this distinction may not
matter for policy purposes. Another participant suggested abrupt change occurs where strong spatial
gradients exist. Dr. Lenton agreed that effective tipping points exist where the underlying climate driver
is smooth.

A different participant posed the layman's question of how to distinguish between natural phenomenon
and man-made events. Dr. Lenton responded that tipping points are affected by a combination of
natural variability and gradual anthropogenic variables. He noted, however, that tipping points are
matters of concern regardless of their drivers. Another participant initiated a discussion about the
economic basis of the precautionary principle. Dr. Weitzman suggested non-linearity in utility was a
more useful concept, pointing out people's natural risk-averse nature.

A final participant noted that two of the three highly aggregated models do incorporate tipping points.
He suggested the need for primary economic studies to quantify impacts. Dr. Lenton acknowledged the
effort made in the models, suggesting room for improvement. He specifically cited a need for multi-
variate forcing, disaggregation, and better impact quantification. He suggested studies on society's
response to other types of historical shocks.

Potential Economic Catastrophes

Dr. Michael Toman then presented his thoughts on the social cost of carbon and risks of climate change
catastrophes. Dr. Toman started by commending Dr. Lenton's presentation, particularly its emphasis
that tipping points may be closer in time and more serious than originally anticipated. Dr. Toman then
outlined the two types of global climate catastrophes: "unfolding" catastrophes and "cascading"
catastrophes. He explained "unfolding" catastrophes are those Dr. Lenton discussed. "Cascading"
catastrophes are the much less studied global catastrophes that arise from the cumulative effect of a
sequence of more localized climate change-induced harms reinforcing each other. Dr. Toman
highlighted the very limited literature on quantitative global catastrophe valuation.

Dr. Toman then presented the standard rational choice approaches and the challenges with applying
them to value global climate catastrophes. He noted the limited information on possible states of the
world, the fat tails of the distribution, and, particularly, the indication from behavioral economics of

-------
systematic assessment errors by the general public. He argued that decision makers need to exercise
their judgment as agents of the general public in evaluations.

Dr. Toman then presented three possible response options: drastic global greenhouse gas reduction;
massive anticipatory adaptation; and particulate injection into the upper atmosphere. He evaluated
each option on four evaluation criteria: effectiveness in mitigating risk; cost of implementation;
robustness to be effective even with surprises in evolution of climate change threats; and flexibility to
modify response as information about risks changes. He finished with a matrix comparing the three
options.

Dr. Toman finished his presentation by explaining that there still exists a large role for standard cost-
benefit analysis in estimating the social cost of carbon. He noted that CBA does not do a good job of
incorporating the fat tails, but noted that was not reason enough to abandon it entirely. He then
presented three approach options for strengthening response options for catastrophe mitigation: the
safe-corridors approach, soliciting expert judgments on alternatives, and soliciting public feedback on
alternatives.

During the question and answer session, several participants questioned aspects of Dr. Toman's matrix
of possible response options. Dr. Toman clarified that the matrix was intended to provide illustrative
examples, rather than present a normative study on policy options. He agreed with two participants'
emphasis on the importance of portfolio approaches and sequence of policy options. He also clarified
several criticisms of the matrix's cost evaluation of different policy options. Finally, in response to
another question, Dr. Toman explained this approach should not be downscaled to individual policies or
categories of within-country investments.

Nonmarket Impacts

Dr. Michael Hanemann concluded the presentation portion of Session 3 with his presentation on
nonmarket impacts. Dr. Hanemann gave his presentation remotely, by phone. He emphasized four
points in his presentation: spatial and temporal aggregation understates impacts; extreme local events
account for most of non-catastrophic damages; risk aversion should be accounted for; and impacts are
multi-attribute and understated by a univariate utility function that treats consumption as a perfect
substitute for environment. Dr. Hanemann showed that non-market impacts from climate catastrophe,
even when underestimated make up the majority of the damages estimated by DICE.

Dr. Hanemann presented impact studies done in California using spatial downscaling. He argued that
increased transparency results from spatial and temporal disaggregation. He noted that impacts and
adaption are spatially and temporally heterogeneous. Any aggregation or averaging of these impacts
results in underestimation of damages. Dr. Hanemann noted the asymmetrical distribution of positive
and negative damages, with greater negative damages. He highlighted that this distribution is often
represented symmetrically in lAMs. Dr. Hanemann also noted the relative importance of increasing
frequency of extreme events as compared to increases in temperature.

Dr. Hanemann concluded that there is a great need to downscale and disaggregate models. He
suggested a modular approach incorporating a network of models. He argued that damage functions

-------
are too simple in current models. Dr. Hanemann suggested that climate change impacts be reframed in
terms of risk, with greater emphasis on downside risk-adjusted impact. He also noted the need to treat
consumption as an imperfect substitute for the environment.

Session 3, Part 2 Discussion

Following Dr. Hanemann's presentation, the discussion portion of Session 3, Part 2 began. During the
discussion session, several participants questioned the ability to downscale data for the entire globe.
Several participants suggested that the data is not good enough globally to support this level of spatial
and temporal disaggregation. They noted that California and the southwest U.S. have particularly good
data and a particularly strong climate signal. One participant wondered whether a bottom up, national
model could help produce a factor that could be used to adjust estimates from existing global aggregate
models. Dr. Hanemann responded that it is still beneficial to disaggregate in addition to working with
global models. He noted that there is a need for several different types of models that can speak to
each other. He highlighted the value of disaggregated information for transparency and
communication. He argued that the level of downscaling might be different for different parts of the
world. For example, he suggested doing a complicated disaggregated sectoral analysis for 3-5 regions,
extrapolating to the U.S., and then conducting a more simple analysis for the rest of the world.

Several participants argued for the need for aggregated models. One participant highlighted the short
time scales and lack of proper climate signal in most regional modeling. Dr. Roe used the example of
river erosion modeling to suggest the need for aggregate functions to encapsulate the principles of very
complicated phenomenon. Another participant cautioned about the indeterminacy of downscaling.
One participant suggested that given the important role of aggregated, simple, reduced-form models, it
is important to reevaluate and refine the form of current damage functions in lAMs. A final participant
suggested the need to rethink and reframe the current aggregate models (e.g., by adjusting the damage
functions) to better qualitatively describe impacts, rather than attempting to introduce a lot of
additional components and details through disaggregation.

Ultimately, several participants argued for a two-pronged approach to modeling: disaggregated,
detailed local modeling and aggregated modeling. One participant noted that the European Commission
is conducting high resolution studies in Europe, which is complementary to highly aggregated studies.

During the discussion, several participants again highlighted the need for better empirical studies on
physical impacts and monetization. One participant highlighted that regional calibration is already
incorporated into current modeling, but that more studies are needed to improve that calibration.
Another participant suggested the incorporation of contingent valuation, choice elicitation, and other
methods of non-use valuation.

Another topic discussed during this session was the role of the SCC and other valuation methods. One
participant distinguished between the need to outline a research agenda to characterize and monetize
impacts and the need to improve the necessarily crude and narrow exercise to develop an SCC number
for OMB guidance. Another participant emphasized the need to articulate regional impacts and to
engage the public, regardless of whether regional impacts are summed to a single number. A third

-------
participant suggested that the economic impacts work, and specifically the SCC, be updated to reflect
the urgency and seriousness of climate change described by natural scientists. Lastly, a participant
underscored the regulatory importance of the SCC as the communication message to the world. As
such, she suggested two short-term improvements to the SCC: to tie down the high end of damages and
to make the discount rate endogenous to growth. Another participant noted this is not as
straightforward as the commenter makes it sound.

Session 4: Implications for Climate Policy Analysis and Design

Session 4 was moderated by Charles Griffiths of the U.S. Environmental Protection Agency and included
presentations by Raymond Kopp, Resources for the Future; Geoff Heal, Columbia University; Nathaniel
Keohane, Environmental Defense Fund; and Roger Cooke, Resources for the Future. The session
examined the implications of assessing and valuing climate change impacts for climate policy analysis
and design, including the following implications: for design and benefit-cost analysis of emission
reduction policies, for addressing equity and natural capital impacts, for choice of policy targets for cost-
effectiveness analysis, and for managing climate risks.

Implications for Design and Benefit-Cost Analysis of Emission Reduction Policies

Dr. Raymond Kopp focused his presentation on the needs of three classes of policymakers and how
lAMs might meet those needs. Specifically, he looked at legislative policymakers, including the U.S.
Congress; international policymakers, including the U.S. Executive Branch; and regulatory agencies,
including the U.S. EPA.

Dr. Kopp noted that legislative policymakers never ask for the social cost of carbon or the benefit-cost
ratio of a given carbon price. Instead, legislative policymakers are interested in: how climate change will
affect the world, the country, and their constituents; worst case scenarios; how adaptation can help;
how their constituents will benefit from mitigation; the cost of mitigation; the distribution of costs to
their constituents; ways to lower costs; and their constituents' willingness-to-pay to avoid damages.

Dr. Kopp then presented the areas of interest and questions of international policymakers. Past and
current areas of interest include: estimates of damage such as the Stern Review, with particular interest
in well-defined sector- and region-specific impacts; estimates of mitigation costs; and distribution of
costs. New questions include: how to measure individual country levels of effort; how to measure
incremental cost; how to estimate realistic offset supply curves that address cost and timing; how a
global carbon market would affect international trade and investment; and how large-scale "green
growth" policies would affect trade and investment.

Next, Dr. Kopp noted that regulatory requirements of executive orders seem to be the sole reason the
Interagency Working Group developed the SCC estimates and continues to refine them. He explained
that there may be roles for lAMs to play in regulatory design other than in regulatory impact analysis,
but that the role will be specific to the regulation in question.

Dr. Kopp outlined the information likely to be of future value to legislation and foreign policy. This
information includes detail on the distribution and severity of damages; characterization of adaptation
potential to lower damages; and estimates of damage sensitivity to the speed of climate change. Finally,

-------
Dr. Kopp highlighted the missing element in current SCC analysis: the complete lack of non-use values,
bequest values, existence values, and passive use values. He noted that these methods of non-market
valuation are those classically used in intra- and inter-generational valuation.

During the question and answer session, a couple of participants asked about breaking down and
allocating the social cost of carbon to more meaningful units, such as domestic SCC and international
SCC or present generation costs and future generation costs, to better answer the questions posed by
policymakers. Another participant noted that the interagency group made the policy decision to focus
on the global SCC and intentionally did not break it down.

A third participant suggested that while there will certainly be costs to climate action, these costs are
mitigated by phasing in policy rather than doing an overnight overhaul and encouraging market
innovation under constraints. She further noted that past actions have not been particularly costly. Dr.
Kopp re-emphasized that when costs do enter, the distribution of costs is very important politically.
Finally, a participant asked how to meaningfully consider the willingness to pay for species extinction of
10-25 percent of species. Dr. Kopp explained the need to clearly articulate the consequences so that
people can value them.

Implications for Addressing Equity and Natural Capital Impacts

Dr. Geoffrey Heal then presented the issues of intragenerational equity and natural capital.
Intergenerational equity is bound up with the pure rate of time preference. Both inter- and intra-
generational equity are affected by the elasticity of the marginal utility of consumption, designated in
this discussion as r|. Dr. Heal presented two contradictory implications of equity. First, he showed how
higher intergenerational equity means a higher value for n, which produces a higher discount rate, and
therefore less concern for future generations and less inclination to act on climate change. Second, he
showed how a higher emphasis on intragenerational distributional equity leads to a higher value placed
on the losses of poor countries, and therefore more inclination to act on climate change. Dr. Heal
explained that in most aggregated lAMs, only the first implication is modeled, so that a higher
intragenerational concern leads to less inclination to deal with climate change. He noted that a
disaggregated model would incorporate the counter-argument.

Next, Dr. Heal considered natural capital. He noted that poor countries are more dependent on the
services of natural capital than rich countries. He proposed that there is some minimum level of natural
capital needed to maintain positive welfare. Dr. Heal then explained that running DICE with this
objective makes a significant difference to model results.

Dr. Heal concluded first that 1AM formulations need to separate the three distinct roles of r|: affecting
intergenerational choices, intragenerational choices, and risk aversion. Second, he concluded that
models need to distinguish environmental services from manufactured goods and rich groups from poor
groups.

During the question and answer session, one participant suggested moving away from the Ramsey
equation, as it builds in aggregation. Dr. Heal agreed, noting that the Ramsey equation promotes

-------
thinking as a representative individual and therefore neglects equity. He explained that the use of
distributional rates is returning after having fallen out of use.

Another participant encouraged disaggregating climate change drivers from intragenerational equity
drivers, so that it is clear model results are motivated by climate change. He noted that other policy
instruments exist to deal with inequality. Dr. Heal agreed, noting that international agreements have
fallen apart due to attempts to address other, unrelated issues in the same policy. A last participant
commented on the outdated nature of the economic methods used in climate economics. He noted
that the Ramsey paradigm is 70 years old and that climate change economics is 30 years old. He
wondered why there has not been more progress. Dr. Heal explained that, until recently, climate
economics has been a thin field with few people.

Implications for Choice of Policy Targets for Cost-Effectiveness Analysis

Dr. Nathaniel Keohane then gave a presentation on the implications for choice of policy targets for cost-
effectiveness analysis. Dr. Keohane started by pointing out that the SCC is not a cost-effectiveness
measure as it does not incorporate the cost of achieving a goal. Instead, he suggested the SCC could be
used in the "spirit" of cost-effectiveness and in establishing consistency.

Dr. Keohane suggested that choosing the appropriate type of target (e.g., emissions target, risk target) is
critical. He also suggested that what other countries do is important. Dr. Keohane noted that the
United Kingdom uses a cost-based shadow price measure. He then presented some concrete ideas for
what a cost-effectiveness approach would look like. First, he suggested a cost-based approach where
shadow prices are set to achieve a global scenario (e.g., 450 ppm C02e or 2°C warming) or a range of
national targets. Second, he suggested a risk-based approach such as a risk management framework or
a direct valuation of the shift in the distribution. He underlined the common thread in these options of
marginal analysis, noting that these options are not mutually exclusive with each other or with a
damages-based SCC approach. He concluded that some number is better than no number but several
numbers may be better than one, depending on the intended use.

Dr. Keohane then discussed the role of the current damages-based SCC. He suggested that the SCC
should not be used as a measure of policy stringency or as the sole input into RIA. Instead, the SCC
should be used to ensure consistency across regulatory agencies and as one of many inputs into RIA. He
noted that the SCC has been used in other proceedings as a tangible, credible measure of the value of
carbon. He explained that these uses show that numbers will be used, that the SCC establishes the
principle that marginal damages are real and can be quantified, and that whether or not the current
estimate is too low, it is still much higher than $0.

Finally, Dr. Keohane noted the disconnect between economics and natural science. He suggested that
the models be unpacked and searched for inputs that do not match the natural science. He highlighted
the damage functions as a likely candidate for improvement. He finished by asking how the results of
the workshop will be incorporated into a process going forward.

During the question and answer session, Dr. Keohane noted that the SCC would be approximately three
times larger if the goal was stabilization. One participant suggested that cumulative emissions would be

-------
a more appropriate metric than emissions concentration. Another participant commended the topic of
the presentation, underlining the importance of cost-effectiveness questions. He suggested that there
are more effective communication tools, such as illustrating how New York will begin to look like DC,
and DC like Florida under the effects of climate change.

Implications for Managing Climate Risks

Dr. Roger M. Cooke concluded the presentation portion of Session 4 with his presentation on managing
climate risks. Dr. Cooke presented from the perspective of mathematical risk analysis. Instead of
modeling impacts around a risk-averse representative customer, he suggested climate change should be
managed by risk-constrained optimization.

Dr. Cooke discussed testing current models using stress tests. He presented an example by stress
testing the DICE model, showing the model's questionable results when pushed outside of reasonable
parameter ranges. Dr. Cooke then discussed the benefit of exploring canonical variations to see if other
simple model forms have structurally different behavior. Again, he presented an illustrative example
using the LotkaVolterra model.

Finally, Dr. Cooke discussed the concepts of inner and outer measures. He explained that there are two
ways to estimate a complicated, or "ugly", sum. First, an inner measure attempts to quantify different
simpler subsets of the sum, with the hope of capturing enough subsets that they add up to the total. An
outer measure estimates a simple sum greater than the total, knowing the goal sum lies within. It tries
to narrow the estimate until it approximates the goal sum. If a set is measureable, the inner measure
will converge with the outer measure. He followed this explanation with a slide presenting the Yale G-
Econ database as an example. He then showed a series of regressions and a plot demonstrating an
"outer" measure with impacts dependent on factors other than average temperature. Dr. Cooke
concluded by emphasizing the need to address model uncertainty and the need to converge the "inner"
and "outer" damage models.

During the question and answer session, one participant asked how to conduct risk-constrained
optimization given uncertainty regarding the distribution of outcomes. Dr. Cooke explained that the
models should be fit to structured expert judgments. Another participant commended the idea of using
expert input but questioned the econometric validity of Dr. Cooke's regressions without numerous other
variables. Dr. Cooke clarified that the regressions were merely an illustration, to be improved upon, of
how one might construct an outer measure.

Session 4 Discussion

Following the questions on Dr. Cooke's presentation, the discussion portion of Session 4 began. One
participant pointed out that the interagency process did produce a range of estimates and questioned
why the focus has been on the central estimate rather than the full range. The panel concurred,
emphasizing the need to communicate the full range. Dr. Heal suggested the interagency-produced
range provides a lower bound to a much wider and higher range. Another participant emphasized a
focus on targets with SCC estimates developed to produce that target. For example, the participant
cited a study that concluded a $75-$100 shadow price would be needed to reduce emissions by 17

-------
percent by 2020. However, a member of the panel emphasized that this may not be possible, given that
there is no nationally agreed-upon emissions goal in the United States against which policies can be
evaluated. Until that happens, analysts must use the tools available to them to evaluate the impacts of
regulations, one of which is benefit-cost analysis.

A third participant criticized funding agencies for funding only the incremental development of existing
lAMs. He suggested this type of funding decision prevented new modelers from entering the field and
developing new and different models as discussed at the workshop. However, another participant
suggested this type of funding decision may allow agencies to spend limited resources in areas with
greater payoff.

During the session, participants discussed how best to meaningfully use the range of SCC estimates. Dr.
Cooke suggested the range as an indication of where the central value might lie in the future. Dr.
Keohane questioned the value of models, such the Department of Transportation's Volpe model that
require a single input. He suggested developing creative ways to visually communicate the data, results,
and tables presented by the interagency working group. Another participant emphasized the
importance of communicating the appropriate degree of precision when presenting SCC estimates by
rounding appropriately. For example, reporting the SCC with multiple significant figures gives a highly
overconfident impression of the precision of these estimates. A different participant cautioned against
presenting subjective judgments objectively, as a number. He suggested communicating SCC
subjectivity to decision makers and perhaps relying more on the statutory process than the regulatory
process.

Dr. Keohane suggested that modelers are not limited to pursue one valuation method or another.
Instead, he commented, if the SCC is pursued, efforts like this workshop exist to try to unpack the
problems. He highlighted the issues of communication; conveying uncertainty; combining and enriching
the SCC with other processes and measures (e.g., risk management); using qualitative analysis; and using
natural units analysis. Ultimately, if one number is needed, he suggested that every effort be made to
identify what it should be, but that it should also be enriched with other numbers.

Session 5: Workshop Wrap-up

The workshop concluded with summary comments by representatives from the U.S. Department of
Energy and the U.S. Environmental Protection Agency. First, Dr. Rick Duke, the DOE Deputy Assistant
Secretary for Climate Policy, presented his closing remarks. He commented that the discussion had
been passionate, rich, and complex, doing justice to the topic. He noted that the SCC is a useful step to
examine the full range of goal-directed options in an economically sensible way, particularly important
to stimulate regulatory action.

Next, Dr. Duke emphasized that this workshop demonstrates DOE's and Secretary Chu's commitment to
integrity in science, economics, and policy. Keeping in that theme, Dr. Duke acknowledged that the
models used by the interagency process use reduced-form damage functions with simple functional
forms. He said that he looked forward to improving them over time. He also noted that DOE is funding
work with the higher resolution models, such as GCAM and IGSM. He highlighted the radically different

-------
nature of these models, remarking that perhaps we have been "looking for the keys under one
streetlight" and instead, "need to build more streetlights." Dr. Duke echoed Dr. Cooke's proposal to
optimize risk under constraint and Dr. Heal's notion of the deeply imperfect substitutability of natural
capital.

Dr. Duke then suggested that even with the most comprehensive suite of bottom-up policies based on
the SCC, the complexities may prevent the attainment of adequate abatement goals. He closed with a
comment on the workshop participation. He noted the thin and disjointed nature of the field and
expressed his pleasure at seeing such good attendance at the workshop. He explained that the
interagency process has encouraged continued refinement of the SCC and expressed his hope that the
workshop attendees would continue to be involved in the refinement process.

Finally, Dr. Al McGartland, Office Director for EPA's NCEE, presented his closing remarks. Dr.

McGartland started by remarking the conversation had been stimulating and thought provoking. He
noted that he thought the idea of unpacking the models and identifying areas for improvement makes
sense. He then shared some broader thoughts on the importance and difficulty of cost-benefit analysis
over the course of EPA's history. He noted the significant traction gained by CBA during air toxics
analysis, particulate matter analysis, and recycling versus disposal analysis.

He explained that despite the inherent difficulties and uncertainties involved, for most environmental
problems, economists tend to band together and "circle the wagons" in support of doing CBA. He then
polled the participants on how they feel about the SCC exercise. He asked for a show of hands for
whether or not they would pursue the SCC exercise if they were decision makers. A few participants
indicated they would 'pull the plug' on the SCC exercise altogether. No one supported forging ahead full
speed and 'circling the wagons' without better communication of the great uncertainties involved in
such estimates. Most of the participants indicated that they would follow a middle path, to continue to
cautiously, bravely pursue the SCC exercise without 'circling the wagons.'

After Dr. McGartland concluded his comments, one of the workshop participants asked how this
workshop will fit into the two-year plan and how the participants could be involved. Dr. McGartland
responded that the first product of the workshop would be a workshop report with a summary of the
proceedings. He noted that the next steps in the interagency process have not yet been completely
defined at this point, but he hoped the interagency group would reconvene in the timeframe outlined in
the 2010 report. He emphasized that EPA is solidly supportive of engaging the public generally and the
research community specifically. Finally, he noted that the second conference focused on damage
functions will take place in late January.

-------
Improving the Assessment and Valuation
Climate Change Impacts for Policy
and Regulatory Analysis

Sponsored by

A r*r\/V United Slalaa
d'&l KH Environn-isntni

Protection Agency

U.S. D1PARTHIHT Of

#ENERGY

November 18-19, 2010 Omni Shore ham Hotel, Washington, DC

APPENDIX



to the



DRAFT Workshop Report:

Improving the Assessment and Valuation of
Climate Change Impacts for Policy and
Regulatory Analysis - Part 1

Modeling Climate Change Impacts and Associated Economic Damages

January 2011



Workshop Sponsored by:
U.S. Environmental Protection Agency
U.S. Department of Energy

Workshop Report Prepared by:
ICF Internationa!

-------
Appendix Contents

Works!	lia with Charge Questions

ripant List
Extendi tracts

-------
Improving the Assessment and Valuation ol
Climate Change Impacts for Policy
and Regulatory Analysis

Sponsored by

A ¦ | \J\ United 5l*t«5
p*Environmental

Protection Agency

U.S. 0IMRTRIINI OP

WENERGY

November 18-19, 2010 Omni Shoreham Hotel, Washington, DC

Workshop Agenda with Charge Questions

MODELING CLIMATE CHANGE IMPACTS AND ASSOCIATED ECONOMIC DAMAGES

Charge Questions: The following charge questions (appearing in boxes) were given to each of
the workshop speakers. Each speaker was asked to write a short abstract (approximately 3-5
pages) and organize their presentations around these questions, though they also were
encouraged to think more broadly and to consider other ideas as they see fit. The purpose of
the papers and presentations was to briefly summarize the current state of the art in each
area and to set the scene for a productive discussion at the workshop, not necessarily to
provide complete answers to all charge questions.

November 18. 2010

Workshop Introduction

8:30-8:35 Welcome and Introductions

Elizabeth Kopits, U.S. Environmental Protection Agency

8:35 - 9:00 Opening Remarks

Bob Perciasepe, Deputy Administrator, U.S. Environmental Protection
Agency

Steve Koonin, Under Secretary for Science, U.S. Department of Energy

9:00 - 9:25 Progress Toward a Social Cost of Carbon

Michael Greenstone, Massachusetts Institute of Techn ology

Session 1: Overview of Existing Integrated Assessment Models
Moderator: Stephanie Waldhoff, U.S. Environmental Protection Agency

Charge: Describe

(1)	the history of climate-economic integrated assessment modeling,

(2)	the major reduced-form and higher-complexity IAMs currently in use,

(3)	the main strengths and weaknesses of each model,

(4)	current areas of active research, and

(5)	how these areas of active research might inform policy and regulatory analysis.

9:25 - 9:50 Overview of Integrated Assessment Models

Jae Edmonds, Pacific Northwest National Laboratory

Models Used for the Development of Current USG SCC Values

A-2

-------
Charge for all model presenters: Describe the current state of your model and any recent,
planned, or potential modifications. Specifically:

(1)	Describe the basic structure of your model. What are key exogenous and
endogenous variables?

(2)	Discuss the physical impacts included in your model and how the
corresponding market and non-market economic damages are
calculated. What major impacts and damage categories are not included
(e.g., ocean acidification and associated damages)? To what extent does
the model incorporate the physical cycles for non-C02 GHGs?

(3)	What assumptions does your model make about adaptation?

(4)	What assumptions does your model make about climate system "tipping
points," catastrophic impacts and the corresponding economic damages?

(5)	How does your model incorporate uncertainty in physical parameters
such as climate sensitivity and economic parameters such as the
discount rate?

9:50-10:15 DICE

Steve Newbold, U.S. Environmental Protection Agency

10:15-10:40 PAGE

Christopher Hope, University of Cambridge

10:40-10:55 Break

10:55-11:20 FUND

David Anthoff, University of California, Berkeley

Representation of Climate Impacts in other Integrated Assessment Models

Charge for all model presenters: Describe the current state of your model and any recent,
planned, or potential modifications. Specifically:

(1)	Describe the basic structure of your model. What are key exogenous and
endogenous variables?

(2)	Discuss the physical impacts included in your model and how the corresponding
market and non-market economic damages are calculated. What major impacts and
damage categories are not included (e.g., ocean acidification and associated
damages)? To what extent does the model incorporate the physical cycles for non-
C02 GHGs?

(3)	What assumptions does your model make about adaptation?

(4)	What assumptions does your model make about climate system "tipping points,"
catastrophic impacts and the corresponding economic damages?

(5)	How does your model incorporate uncertainty in physical parameters such as
climate sensitivity and economic parameters such as the discount rate?

A-3

-------
11:20-11:45 GCAM (JGCRI - UMD/PNNL) and Development of iESM

(PNNL/LBNL/ORNL)

Leon Clarke, Pacific Northwest National Laboratory

11:45-12:10 IGSM (MIT)

John Reilly, Massachusetts Institute of Technology

12:10-12:40 Discussion

12:40-1:40 Lunch

Session 2: Near-Term DOE and EPA Efforts

Moderator: Ann Wolverton, U.S. Environmental Protection Agency

1:40 - 2:00 Proposed Impacts Knowledge Platform
Bob Kopp, U.S. Department of Energy
Nisha Krishnan, Resources for the Future

2:00-2:20 Proposed Generalized Modeling Framework

Alex Marten, U.S. Environmental Protection Agency

2:20 - 2:40 Discussion

Session 3A: Critical Modeling Issues in Assessment and Valuation of Climate Change
Impacts

Moderator: Ann Wolverton, U.S. Environmental Protection Agency

2:40 - 3:10 Sectoral and Regional Disaggregation and Interactions

Ian Sue Wing Boston University

Charge: Review the sectoral and regional representation of economic damages in
integrated assessment models. Specifically, discuss:

(1)	how damages in one category and one region may affect other categories and
regions,

(2)	the relative magnitude/importance of these interactions,

(3)	how these relationships might be represented in an IAM, and

(4)	gaps in the way existing IAMs represent these relationships and major challenges
in improving these representations.

3:10-3:20 Break

A-4

-------
3:20-3:50 Adaptation and Technological Change

Ian Sue Wing, Boston University on behalf of Karen Fisher-Vanden,
Pennsylvania State University

Charge: Drawing from the recent literature, discuss how adaptation may influence the net
social costs of climate change (adaptation costs plus residual climate damages).
Specifically, discuss:

(1)	relevant studies on the observed or potential effectiveness of adaptive measures,
and on private behaviors and public projects regarding adaptation;

(2)	relevant studies on how to forecast adaptive capacity;

(3)	how adaptation and technical change could be represented in an IAM (for at least
one illustrative sector);

(4)	whether the information required to calibrate such a model is currently available,
and, if not, what new research is needed; and

(5)	how well or poorly existing IAMs incorporate the existing body of evidence on
adaptation.

3:50-4:20 Multi-century Scenario Development and Socio-Economic Uncertainty

Brian O'Neill, National Center for Atmospheric Research

Charge: Discuss the methods and difficulties associated with forecasting a baseline
scenario for greenhouse gas emissions and socio-economic variables (e.g., population and
GDP), including the particular challenges in extending these scenarios for multiple
centuries. Specifically, discuss:

(1)	relevant studies on long-term demographic and economic scenarios and the
assumptions used to develop these scenarios;

(2)	relevant studies on the evolution of energy systems and the assumptions used to
develop these scenarios;

(3)	the range of plausible future scenarios extending to at least 2300, including the
range incorporated into major IAMs; and

(4)	what are the main challenges in representing such multi-century forecasts in an
IAM.

4:20-5:00 Discussion

A-5

-------
November 19. 2010

Day 2 Introduction

8:30-8:40 Welcome; Recap of Day 1; Overview of Day 2

Elizabeth Kopits, U.S. Environmental Protection Agency

Session 3B: Critical Modeling Issues in Assessment and Valuation of Climate Change
Impacts (cont.)

Moderator: Bob Kopp, U.S. Department of Energy

8:40-9:10 Incorporation of Climate System Uncertainty into IAMs

Gerard Roe, University of Washington

Charge: Discuss:

(1)	the major sources of climate system uncertainty that could be represented in
reduced-form integrated assessment models (such as DICE, PAGE, and FUND),

(2)	the difficulties/issues with representing the uncertainty surrounding these
parameters in IAMs, and

(3)	relevant studies that estimate probability density functions for these parameters.

9:10-9:40 Extrapolation of Damage Estimates to High Temperatures: Damage
Function Shapes

Marty Weitzman, Harvard University
Charge: Discuss:

(1)	how damage functions behave at high temperatures in the principal reduced-form
IAMs, including DICE, PAGE, and FUND;

(2)	the reasoning underlying the selection of these functional forms and alternative
formulations that have been proposed in the literature;

(3)	the relative strengths of these various functional forms in terms of extrapolating
damage estimates to high temperatures; and

(4)	the difficulties/issues with incorporating uncertainty regarding such "out of
sample forecasts."

9:40-10:10 Earth System Tipping Points

Tim Lenton, University of East Anglia

Charge: Discuss:

(1)	evidence on potential Earth system tipping points, including the most recent
estimates of these tipping points based on modeling studies, paleoclimatic data,
expert elicitation, and other relevant sources; and

(2)	available estimates of their probabilities under different scenarios.

A-6

-------
10:10-10:30 Break

10:30-11:00 Potential Economic Catastrophes

Michael Toman, World Bank

Charge: Discuss:

(1)	the literature on the potential economic damages associated with catastrophic
climate impacts, potentially related to Earth system tipping points;

(2)	how these damages might be incorporated into reduced-form and/or higher-
complexity IAMs; and

(3)	the key challenges associated with translating information on the likelihood and
physical consequences of particular tipping points into economic damages.

11:00-11:30 NonmarketImpacts

Michael Hanemann, University of California, Berkeley

Charge: Discuss:

(1)	recent studies of potential non-market impacts of climate change;

(2)	how the value of such impacts are currently represented in IAMs;

(3)	how such non-market impacts could be better represented in IAMs, possibly
including but not necessarily limited to alternative damage functional forms and
multivariate utility functions; and

(4)	key challenges of quantifying and incorporating non-market impacts into IAMs.

11:30-12:30 Discussion

12:30-1:30 Lunch

Session 4: Implications for Climate Policy Analysis and Design

Moderator: Charles Griffiths, U.S. Environmental Protection Agency

1:30-2:00 Implications for Design and Benefit-Cost Analysis of Emission Reduction
Policies

Ray Kopp, Resources for the Future

Charge: How can improved IAMs, as discussed in Sessions 1-3, aid in the design and
evaluation of domestic emission reduction policies such as cap-and-trade or carbon
taxes, and inform negotiations of international climate agreements?

2:00-2:30 Implications for Addressing Equity and Natural Capital Impacts

A-7

-------
Geoff Heal, Columbia University

Charge: How can improved IAMs, as discussed in Sessions 1-3, help policy analysts
address intra-generational equity concerns, account for impacts on natural capital and
ecosystem services, and better represent the substitutability between ecosystem
services and market goods?

2:30-3:00 Implications for Choice of Policy Targets for Cost-Effectiveness Analysis

Nat Keohane, Environmental Defense Fund

Charge: How can improved IAMs, as discussed in Sessions 1-3, help inform a cost-
effectiveness analysis of various policy actions that reduce C02 emissions? For example,
how could these models help in choosing a temperature or carbon concentration target
for national policies or international agreements? Are there other environmental
endpoints that should be considered in cost-effectiveness analysis of climate policies
(e.g., targets associated with ocean acidification)?

3:00-3:10 Break

3:10-3:40 Implications for Managing Climate Risks

Roger Cooke, Resources for the Future

Charge: How could improved IAMs, along the lines discussed in Sessions 1-3, help
inform a risk management analysis of various policy actions that reduce C02 emissions?
For example, how could these models aid in the design of adaptation policies to manage
increased climate and weather related risks, such as increased flood frequencies and
storm damages?

3:40-4:15 Discussion

Session 5: Workshop Wrap-up

4:15-4:30 Summary Comments by U.S. Department of Energy

Rick Duke, Deputy Assistant Secretary for Climate Policy

4:30-4:45 Summary Comments by U.S. Environmental Protection Agency

A1 McGartland, Director of the National Center for Environmental Economics

A-8

-------
Improving the Assessment and Valuation of"
Climate Change Impacts for Policy
and Regulatory Analysis

Sponsored by

J\ United 5iaujs
Enviranmiintn
Ul Mik Protection Agency

sZtSiLh. U.S. DEPARTMENT Of

fpENERGY

November 18-19, 2010 Omni Shoreham Hotel, Washington, DC

Final Participants List

Frank Ackerman

Stockholm Environment Institute-US
Frank.Ackerman@sei-us.org

Kate Calvin

PNNL/Joint Global Change Research Institute
katherine.calvin@pnl.gov

Lucas Adin

U.S. Department of Energy
lucas.adin@ee.doe.gov

Kate Cardamone

U.S. Environmental Protection Agency
cardamone.kate@epa.gov

Farhan Akhtar

U.S. EPAORD
akhtar.farhan@epa.gov

Linda Chappell

U.S. Environmental Protection Agency
chappell.linda@epa.gov

David Anthoff

Department of Agricultural and Resource
Economics, University of California, Berkeley
david@anthoff.de

Ruth Greenspan Bell

World Resources Institute
rbell@wri.org

Juan-Carlos Ciscar

JRC-European Commission
juan-carlos.ciscar@ec.europa.eu

Leon Clarke

Joint Global Change Research Institute - Pacific
Northwest National Laboratory
leon.clarke@pnl.gov

Aaron Bergman

U.S. Department of Energy
aaron.bergman@hq.doe.gov

Beth Binns

ICF International
bbinns@icfi.com

Jason Bordoff

Council on Environmental Quality
jbordoff@ceq.eop.gov

Dallas Burtraw

Resources for the Future
burtraw@rff.org

Rachel Cleetus

Union of Concerned Scientists
rcleetus@ucsusa.org

Roger Cooke

Resources for the Future
cooke@rff.org

Maureen Cropper

Resources for the Future
cropper@rff.org

Christian Crowley

U.S. Department of the Interior
crow@gwu.edu

A-9

-------
Rita Curtis

OSTP

Rita_E._Curtis@ostp.eop.gov
Michael Dalton

National Marine Fisheries Service
michael.dalton@noaa.gov

Benjamin DeAngelo

U.S. Environmental Protection Agency
deangelo.benjamin@epa.gov

Stephen DeCanio

University of California, Santa Barbara
decanio@econ.ucsb.edu

Terry Dinan

Congressional Budget Office
terryd@cbo.gov

Gabrielle Dreyfus

NOAA

gabrielle.dreyfus@noaa.gov
Rick Duke

U.S. Department of Energy
rick.duke@hq.doe.gov

James Edmonds

PNNL/Joint Global Change Research Institute
jae@pnl.gov

Barry Elman

U.S. DOE Headquarters
barry.elman@hq.doe.gov

Paul Falkowski

Rutgers University
falko@marine.rutgers.edu

Chris Farley

Rutgers University
cfarley@fs.fed.us

Allen Fawcett

Council on Environmental Quality
afawcett@ceq.eop.gov

Peter Feather

USDA-OCE

pfeather@oce.usda.gov
Ann Ferris

Council on Environmental Quality
aferris@ceq.eop.gov

Steve Fetter

Office of Science and Technology Policy
sfetter@ostp.eop.gov

Karen Fisher-Vanden1

Penn State University
kaf26@psu.edu

Arthur Fraas

Resources for the Future
fraas@rff.org

Elisabeth Gilmore

Climate Science and Impacts Branch, EPA
gilmore.elisabeth@epa.gov

Alexander Golub

Environmental Defense Fund
agolub@edf.org

Anne Grambsch

U.S. EPA, Office of Research and Development
grambsch.anne@epa.gov

1 Unable to attend due to illness. Ian Sue Wing
presented in her place.

A-10

-------
Michael Greenstone

MIT, Dept. of Economics
mgreenst@mit.edu

Chris Hope

University of Cambridge
c.hope@jbs.cam.ac.uk

Charles Griffiths

U.S. Environmental Protection Agency
griffiths.charles@epa.gov

Howard Gruenspecht

U.S. Energy Information Administration
howard.gruenspecht@eia.gov

Jay Gulledge

Pew Center on Global Climate Change
gulledgej@pewclimate.org

Tara Hamilton

ICF International
Thamilton2@icfi.com

Michael Hanemann2

Arizona State University
hanemann@berkeley.edu

Reid Harvey

Climate Change Div., US EPA
harvey.reid@epa.gov

Geoffrey Heal

Columbia Business School
gmhl@columbia.edu

Gloria Helfand

U.S. Environmental Protection Agency
helfand.gloria@epa.gov

Kathy Hibbard

PNNL

kathy.hibbard@pnl.gov

2 Attended via teleconference

Asa Hopkins

U.S. Department of Energy

Richard Howarth

Dartmouth College
RBHowarth@Dartmouth.edu

Holmes Hummel

DOE Office of Policy & Int'l Affairs
holmes.hummel@hq.doe.gov

Wendy Jaglom

ICF International
wjaglom@icfi.com

Robert Johansson

USDA-OCE-CCPO
rjohansson@oce.usda.gov

Laurie Johnson

Natural Resources Defense Council
ljohnson@nrdc.org

Benjamin Jones

Council of Economic Advisers
bjones@cea.eop.gov

Kenneth Judd

Hoover Institution
kennethjudd@mac.com

Nathaniel Keohane

Environmental Defense Fund
nkeohane@edf.org

Jim Ketcham-Colwill

U.S. EPA Office of Air and Radiation
ketcham-colwill.jim@epa.gov

A-ll

-------
Heidi King

Office of Management and Budget
Heather Kleminck

U.S. Environmental Protection Agency
Klemick.heather@epa.gov

Charles Kolstad

University of California
kolstad@econ.ucsb.edu

Elizabeth Kopits

U.S. Environmental Protection Agency
kopits.elizabeth@epa.gov

Bob Kopp

U.S. Department of Energy
robert.kopp@hq.doe.gov

Ray Kopp

Resources for the Future
Kopp@rff.org

Nisha Krishnan

Resources for the Future
krishnan@rff.org

Dina Kruger

U.S. Environmental Protection Agency
Kruger.dina@epa.gov

Alan Krupnick

Resources for the Future
krupnick@rff.org

Christine Kym

Office of Management and Budget
Peter Larsen

Lawrence Berkeley National Laboratory
PHLarsen@lbl.gov

Amanda Lee

Office of Management and Budget
Tim Lenton

University of East Anglia
t.lenton@uea.ac.uk

Arik Levinson

Council of Economic Advisers
alevinson@cea.eop.gov

Ines Lima Azevedo

Carnegie Mellon University
iazevedo@cmu.edu

Dan Loughlin

U.S. EPA Office of Research and Development
loughlin.dan@epa.gov

Molly Macauley

Resources for the Future
macauley@rff.org

Michael MacCracken

Climate Institute
mmaccrac@comcast.net

Margaret MacDonell

Argonne National Laboratory
macdonell@anl.gov

Liz Marshall

Economic Research Service, USDA
emarshall@ers.usda.gov

Alex Marten

U.S. Environmental Protection Agency
marten.alex@epa.gov

Damon Matthews

Concordia University
dmatthew@alcor.concordia.ca

A-12

-------
Al McGartland

U.S. Environmental Protection Agency
mcgartland.al@epa.gov

Jim McMahon

Lawrence Berkeley National Laboratory
jemcmahon@lbl.gov

Bryan Mignone

U.S. Department of Energy
bryan.mignone@hq.doe.gov

Antony Millner

UC Berkeley
a.millner@berkeley.edu

Adele Morris

The Brookings Institution
amorris@brookings.edu

Elisabeth Moyer

University of Chicago
moyer@uchicago.edu

Peter Nagelhout

U.S. Environmental Protection Agency
nagelhout.peter@epa.gov

Steve Newbold

U.S. Environmental Protection Agency
newbold.steve@epa.gov

Richard Newell

U.S. Departmentof Energy Information

Administration

Richard.Newell@eia.gov

Robert O'Connor

National Science Foundation
roconnor@nsf.gov

Brian O'Neill

National Center for Atmospheric Research
(NCAR)

boneill@ucar.edu
Inja Paik

U.S. Department of Energy
inja.paik@hq.doe.gov

Don Pickrell

Volpe Center, U.S. Department of

Transportation

don.pickrell@dot.gov

Shaun Ragnauth

U.S. Environmental Protection Agency
ragnauth.shaun@epa.gov

John Reilly

MIT Joint Program on Climate Change
jreilly@mit.edu

Gerard Roe

University of Washington
gerard@ess.washington.edu

Steven Rose

EPRI

srose@epri.com
Arthur Rypinski

U.S. Department of Transportation
Arthur.Rypinski@dot.gov

Ron Sands

Economic Research Service, USDA
rsands@ers.usda.gov

Keith Sargent

U.S. Environmental Protection Agency
sergeant.keith@epa.gov

A-13

-------
Marcus Sarofim

AAAS/EPA

sarofim.marcus@epa.gov

Ian Sue Wing

Boston University
isw@bu.edu

Jayant Sathaye

Lawrence Berkeley National Lab
jasathaye@lbl.gov

Peter Schultz

ICF International
pschultz@icfi.com

Joseph Seneca

Bloustein School, Rutgers University
seneca@rci.rutgers.edu

Sandy Seymour

ICF International
sseymour@icfi.com

Robert Shackleton

CBO

bobsh@cbo.gov
Kristen Sheeran

Economics for Equity and Environment Network
(E3)

ksheeran@ecotrust.org
Michael Shelby

U.S. Environmental Protection Agency
shelby.michael@epa.gov

Jhih-Shyang Shih

Resources for the Future
shih@rff.org

Joel Smith

Stratus Consulting Inc.

jsmith@stratusconsulting.com

303-381-8218

Richard Tol

richard.tol@esri.ie

Michael Toman

World Bank

mtoman@worldbank.org

Robert Vallario

US Department of Energy
bob.vallario@science.doe.gov

Dominique van der Mensbrugghe

World Bank

dvandermensbrugg@worldbank.org

Stepanie Waldhoff

U.S. Environmental Protection Agency
waldhoff.stephanie@epa.gov

Martin Weitzman

Harvard University
mweitzman@harvard.edu

John Weyant

Stanford University
weya nt @ sta nf ord. ed u

Ann Wolverton

U.S. Environmental Protection Agency
wolverton.ann@epa.gov

Craig Zamuda

U.S. Department of Energy
craig.zamuda@hq.doe.gov

A-14

-------
Improving the Assessment and Valuation
Climate Change Impacts for Policy
and Regulatory Analysis

Extended Abstracts

(click on title to go to abstract)

Estimating the Social Cost of Carbon for the United States Government - Michael Greenstone
Summary of the DICE model - Steve Newbold

The PAGE09 model: Estimating climate impacts and the social cost of C02 - Chris Hope
FUND - Climate Framework for Uncertainty, Negotiation and Distribution - David Anthoff
Climate Damages in the MIT IGSM - John Reilly

Modeling the Impacts of Climate Change: Elements of a Research Agenda - Ian Sue Wing

Adaptation and Technological Change - Karen Fisher-Vanden

Knowability and no ability in climate projections - Gerard Roe

Notes for EPA & DOE discussion meeting - Marty Weitzman

Earth System Tipping Points - Tim Lenton

Catastrophic Climate Change - Michael Toman

Natural Capital and Intra- Generational Equity in Climate Change - Geoff Heal
Managing Climate Risks - Roger Cooke

A-15

-------
Estimating the Social Cost of Carbon for the United States Government

Michael Greenstone

3M Professor of Environmental Economics
Massachusetts Institute of Technology
November 2010

The climate is a key ingredient in the earth's complex system that sustains human life and well being.
According to the United Nation's Intergovernmental Panel on Climate Change (IPCC), the emissions of
greenhouse gases (GHG) due to human activity, large the combustion of fossil fuels like coal, is "very
likely" altering the earth's climate, most notably by increasing temperatures, precipitation levels and
weather variability. Without coordinated policy around the globe, state of the art climate models predict
that the mean temperature in the United States will increase by about 10.7° F by the end of the century
(Deschenes and Greenstone 2010). Further, the distribution of daily temperatures is projected to
increase in ways that pose serious challenges to well being; for example, the number of days per year
where the typical American will experience a mean (average of the minimum and maximum)
temperature that exceeds 90° F is projected to increase from the current 1.3 days to a 32.2 days (ibid).
The especially troubling statistic is that the hottest days pose the greatest threat to human well being.

It appeared that the United States and possibly the major emitters were poised to come together to
confront climate change by adopting a coordinated set of policies that could have included linked cap
and trade systems. However, the failure of the United States Government to institute such a system and
the non-binding commitments from the Copenhagen Accord seem to have placed the all at once
solution to climate change out of reach for at least several years.

Instead, the United States and many other countries are likely to pursue a series of smaller policies all of
which aim to reduce GHG emissions but individually have a marginal impact on atmospheric
concentrations. These policies will appear in a wide variety of domains, ranging from subsidies for the
installation of low carbon energy sources to regulations requiring energy efficiency standards in
buildings, motor vehicles, and even vending machines to rebates for home insulation materials.

Although many of these policies have other goals, their primary motivation is to reduce GHG emissions.
However, these policies reduce GHG emissions at different rates and different costs.

In the presence of this heterogeneity and nearly limitless set of policies that reduce GHG emissions, how
is government to set out a rational climate policy? The key step is to determine the monetized damages
associated with an incremental increase in carbon emissions, which is referred to as the social cost of
carbon (SCC).1 It is intended to include (but is not limited to) changes in net agricultural productivity,

1 Under Executive Order 12866, agencies in the Executive branch of the U.S. Federal government are required, to
the extent permitted by law, "to assess both the costs and the benefits of the intended regulation and, recognizing
that some costs and benefits are difficult to quantify, propose or adopt a regulation only upon a reasoned
determination that the benefits of the intended regulation justify its costs."

A-16

-------
2

human health, property damages from increased flood risk, and the value of ecosystem services.
Monetized estimates of the economic damages associated with carbon dioxide emissions allows the
social benefits of regulatory actions that are expected to reduce these emissions to be incorporated into
cost-benefit analyses. Indeed as the Environmental Protection Agency begins to regulate greenhouse
gases under the Clean Air Act, the SCC can help to identify the regulations where the net benefits are
positive.

The United States Government (USG) recently selected four SCC estimates for use in regulatory analyses
and has been using them regularly since their release. For 2010, the central value is $21 per ton of C02
equivalent emissions.4 The USG also announced that it would conduct sensitivity analyses at $5, $35,
and $65. The $21, $5, and $35 values are associated with discount rates of 3%, 2.5%, and 5%, reflecting
that much of the damages from climate change are in the future. The $65 value aims to represent the
higher-than-expected impacts from temperature change further out in the tails of the SCC distribution.
In particular, it is the SCC value for the 95th percentile at a 3 percent discount rate. These SCC estimates
also grow over time based on rates endogenously determined within each model. For instance, the
central value increases to $24 per ton of C02 in 2015 and $26 per ton of C02 in 2020.

I was involved in the interagency process that selected these values for the SCC and this talk summarizes
these efforts.5 The process was initiated in 2009 and completed in February 2010. It aimed to develop a
defensible, transparent, and economically rigorous way to value reductions in carbon dioxide emissions
that result from actions across the Federal government. Specifically, the goal was to develop a range of
SCC values in a way that used a defensible set of input assumptions, was grounded in the existing
literature, and allowed key uncertainties and model differences to transparently and consistently inform
the range of SCC estimates used in the rulemaking process.

The intent of this lecture is to explain the central role of the social cost of carbon in climate policy, to
summarize the methodology and process used by the interagency working group to develop values, and
to identify key gaps so that researchers can fill these gaps. Indeed, the interagency working group
explicitly aimed the current set of SCC estimates to be updated as scientific and economic
understanding advances.

2 All values of the SCC are presented as the cost per metric ton of C02 emissions.

3 Most regulatory actions are expected to have small, or "marginal," impacts on cumulative global emissions,
making the use of SCC an appropriate measure.

4 All dollar values are expressed in 2007 dollars.

5 This process was convened by the Council of Economic Advisers and the Office of Management and Budget, with
regular input from other offices within the Executive Office of the President, including the Council on
Environmental Quality, National Economic Council, Office of Energy and Climate Change, and Office of Science and
Technology Policy. Agencies that actively participated included the Environmental Protection Agency, and the
Departments of Agriculture, Commerce, Energy, Transportation, and Treasury.

A-17

-------
Summary of the DICE model

Stephen C. Newbold

U.S. EPA, National Center for Environmental Economics1

This report gives a brief summary of the DICE (Dynamic Integrated Climate-Economy) model, developed
by William Nordhaus, which "integrate^] in an end-to-end fashion the economics, carbon cycle, climate
science, and impacts in a highly aggregated model that allow[s] a weighing of the costs and benefits of
taking steps to slow greenhouse warming" (Nordhaus and Boyer 2000 p 5). Section 1 of this report
recounts the major milestones in the development of DICE and its regionally disaggregated companion
model, RICE. This section also serves as a convenient reference for more detailed expositions of the
model and applications in the primary literature. Section 2 describes the basic structure of the most
recently published version of DICE, and Section 3 describes some key aspects of the model calibration.
Section 4 gives additional details on the climate damage function in DICE, and Section 5 gives a brief
description of the most recently published version of the RICE model.

Historical development

The DICE integrated assessment model has been developed in a series of reports, peer reviewed articles,
and books by William Nordhaus and colleagues over the course of more than thirty years. The earliest
precursor to DICE was a linear programming model of energy supply and demand with additional
constraints imposed to represent limits on the peak concentration of carbon dioxide in the atmosphere
(Nordhaus 1977a,b).2The model was dynamic, in that it represented the time paths of the supply of
energy from various fuels and the demand for energy in different sectors of the economy and the
associated emissions and atmospheric concentrations of carbon dioxide. However, it included no
representation of the economic impacts or damages from temperature or other climate changes. Later,
Nordhaus (1991) developed a long-run steady-state model of the global economy that included
estimates of both the costs of abating carbon dioxide emissions and the long term future climate
impacts from climate change. This allowed for a balancing of the benefits and costs of carbon dioxide
emissions to help determine the optimal level of near term controls. The analysis centered on the global
average surface temperature, which was "...chosen because it is a useful index (in the nature of a
sufficient statistic) of climate change that tends to be associated with most other important changes
rather than because it is the most important factor in determining impacts" (Nordhaus 1991 p 930). The

1 Prepared for the EPA/DOE workshop, Improving the Assessment and Valuation of Climate Change Impacts for
Policy and Regulatory Analysis, Washington DC, November 18-19, 2010. Please note that the views expressed in
this paper are those of the author and do not necessarily represent those of the U.S. Environmental Protection
Agency. No Agency endorsement should be inferred. Author's email: newbold.steve@epa.gov.

2 While it has not been the focus of the DICE model, it should be emphasized that this type of cost-effectiveness
framework is still useful. For example, if policy makers decide upon a 2 degree target, then the appropriate social
cost of carbon to use is the shadow price associated with that path (Nordhaus, personal communication).

A-18

-------
categories of climate damages that were represented in the model were associated with market sectors
that accounted for roughly 13% of GDP in the United States.3

The DICE model was first presented in its modern form by Nordhaus (1992a,b), who described the new,
fully dynamic Ramsey-type optimal growth structure of the model and the optimal time path of
emission reductions and associated carbon taxes that emerged from it. The full derivation and extended
description of the DICE model and a wider range of applications were presented in a book by Nordhaus
(1994a). The next major advance involved disaggregating the model into ten different groups of nations
to produce the RICE (Regional DICE) model, which allowed the authors to examine national-level climate
policies and different strategies for international cooperation (Nordhaus and Yang 1996). An update and
extended description of both RICE (now with eight regions) and DICE appeared in the book by Nordhaus
and Boyer (2000). The next major update of DICE, modified to include a backstop technology that can
replace all fossil fuels and whose price was projected to decline slowly over time, appeared in another
book by Nordhaus (2008). Finally, Nordhaus (2010) described the most recent version of the RICE model,
which adds an explicit representation of damages due to sea level rise.

In addition to the studies by Nordhaus and colleagues mentioned above, DICE has been adapted by
other researchers to examine a wide range of issues related to the economics of climate change. A
comprehensive review is well beyond the scope of this summary, so only a few examples are mentioned
here. Pizer (1999) used DICE to compare carbon tax and a cap-and- trade-style policies under
uncertainty. Popp (2005) modified DICE to include endogenous technical change. Baker et al. (2006)
used DICE to examine the effects of technology research and development on global abatement costs.
Hoel and Sterner (2007) modified the utility function in DICE to include a form of non-market
environmental consumption that is an imperfect substitute for market consumption, and Yang (2008)
used RICE in a cooperative game theory framework to examine strategies for international negotiations
of greenhouse gas mitigation policies and targets.

Basic model structure

DICE2007 is a modified Ramsey-style optimal economic growth model, where an additional form of
"unnatural capital"—the atmospheric concentration of CO2—has a negative effect on economic output
through its influence on the global average surface temperature. Global economic output is represented
by a Cobb-Douglas production function using physical capital and labor as inputs. Labor is assumed to be
proportional to the total global population, which grows exogenously over time. Total factor
productivity also increases exogenously over time. The carbon dioxide intensity of economic production
and the cost of reducing carbon dioxide emissions decrease exogenously over time. In each period a
fraction of output is lost according to a Hicks-neutral climate change damage function. The output in
each period is then divided between consumption, investment in the physical capital stock (savings), and
expenditures on emissions reductions (akin to investment in the natural capital stock). DICE solves for
the optimal path of savings and emissions reductions over a multi-century planning horizon, where the

3 It should be emphasized that while this model and all subsequent versions of DICE necessarily make assumptions
about climate and economic conditions in the far future, the important question is the extent to which current
policies are robust to changes in assumptions about future variables (Nordhaus, personal communication).

A-19

-------
objective to be maximized is the discounted sum of all future utilities from consumption. Total utility in
each period is the product of the number of individuals alive and the utility of a representative individual
with average income in that period. The period utility function is of the standard constant relative risk
aversion (CRRA) form, and utilities in future periods are discounted at a fixed pure rate of time
preference.

Calibration

The climate model in DICE2007 tracks the stocks and flows of carbon in three aggregate compartments
of the earth system: the lower atmosphere, the shallow ocean, and the deep ocean. The transfer
coefficients linking the flows among the compartments were "calibrated to fit the estimates from
general circulation models and impulse-response experiments, particularly matching the forcing and
temperature profiles in the MAGICC model" (Nordhaus 2008 p 54). The climate sensitivity parameter—
the equilibrium change in global average surface temperature after a sustained doubling of atmospheric
carbon dioxide concentration— was set to 3 degrees Celsius, which is near the middle of the range cited
by the IPCC. The projected temperature change under the baseline scenario (with no climate controls
for the first 250 years) is an increase in global average surface temperature of 3.2 degrees Celsius
around year 2100 with a peak of around 6.5 degrees Celsius around year 2500.

The key economic growth and preference parameters of DICE2007 are calibrated as follows. The global
population is projected to grow exogenously from around 6.5 billion in 2005 to 8.6 billion around 2200.
Total factor productivity growth and the discount rate parameters were calibrated to match market
returns in the early periods of the model: specifically, "We have chosen a time discount rate of V/z
percent per year along with a consumption elasticity of 2. With this pair of assumptions, the real return
on capital averages around 5Vz percent per year for the first half century of the projections, and this is
our estimate of the rate of return on capital" (Nordhaus 2008 p 61).

The abatement cost function is specified such that the marginal abatement cost, measured as a fraction
of output, increases roughly with the square of the fraction of emissions abated. The backstop price—
the marginal cost of eliminating the last unit of emissions in each period—is $1,170 per metric ton of
carbon in the first period and falls exponentially at a rate of 5% per decade to a long run value of $585
per metric ton of carbon. The climate damage function is specified such that for small temperature
changes the fraction of output lost in each period increases with the square of the increase in
temperature above the preindustrial average temperature.4The coefficient of the damage function is
calibrated so that roughly 1.7% of global economic output is lost when the average global surface
temperature is elevated by 2.5 degrees Celsius above the preindustrial average.

4 The DICE2007 damage function has an "S-shape," so for very large temperature changes the fraction of output
lost increases with temperature at a decreasing rate and asymptotes to one. However, it should be emphasized
that the damage function is calibrated to damages in the range of 2 to 4 degrees Celsius. The extent of non-
linearity beyond this range is unknown, so extrapolations beyond this point should not be considered reliable
(Nordhaus, personal communication).

A-20

-------
Damages

The globally aggregated climate damage function in DICE has been calibrated to match the sum of
climate damages in all regions represented in the RICE model. The potential damages from climate
change are divided into seven categories: agriculture, sea level rise, other market sectors, human health,
nonmarket amenity impacts, human settlements and ecosystems, and catastrophes. A full recounting of
the derivation of the damage functions in all categories is beyond the scope of this short summary, but
to the give the reader a flavor for what is involved this section reviews three categories of damages:
agriculture, heath, and catastrophes. This discussion draws heavily on Chapter 4 of Nordhaus and Boyer
(2000), so the reader is referred there for more information.

Agriculture can serve as an illustrative example of some of the other categories not covered here. The
basic strategy for calibrating the damage functions is to draw on estimates from previous studies of the
potential economic losses in each category at a benchmark level of warming of 2.5 degrees Celsius,
extrapolating across regions as necessary to cover data gaps in the literature. Some extrapolations were
made using income elasticities for each impact category. As the authors explain, "United States
agriculture can serve here as an example. Our estimate is that [the fraction of the value of agricultural
output lost at 2.5 degrees Celsius] is 0.065 percent [based on Darwin et al. 1995]... The income elasticity
of the impact index is estimated to be -0.1, based on the declining share of agriculture in output as per
capita output rises" (Nordhaus and Boyer 2000 p 74-75).

The human health impacts of climate change were based on the effects of pollution and a broad group
of climate-related tropical diseases including malaria and dengue fever. The increased mortality from
warming in the summer and decreased mortality from warming in the winter were assumed to roughly
offset and so were not included. The specification of the human health damage function involved "a
regression of the logarithm of climate related [years of life lost] on mean regional temperature
estimated form the data presented in Murray and Lopez [1996]" with judgmental adjustments "to
approximate the difference among subregions that is climate related," and each year of life lost was
valued at two years of per capita income (Nordhaus and Boyer 2000 p 80-82).

The damages from potential catastrophic impacts were estimated using results from a previous survey
of climate experts by Nordhaus (1994b). The experts were asked for their best professional judgment of
the likelihood of a catastrophe—specified as a 25 percent loss of global income indefinitely—if the
global average surface temperature increased by 3 and by 6 degrees Celsius within 100 years. The
averages of the survey responses were adjusted upward somewhat based on "[developments since the
survey [that] have heightened concerns about the risks associated with major geophysical changes,
particularly those associated with potential changes in thermohaline circulation" (Nordhaus and Boyer
2000 p 87). The probability of a 30 percent loss of global income indefinitely was assumed to be 1.2 and
6.8 percent with 2.5 and 6 degrees Celsius of warming, respectively. The percent of income lost was
assumed to vary by region, and a coefficient of relative risk aversion equal to 4 was used to calculate the
willingness to pay to avoid these risks in each region. The resulting "range of estimates of WTP lies
between 0.45 and 1.9 percent of income for a 2.5oC warming and between 2.5 and 10.8 percent of
income for a 6oC warming. It is assumed that this WTP has an income elasticity of 0.1" (Nordhaus and
Boyer 2000 p 89).

A-21

-------
Damages in the remaining categories were estimated in a similar vein, using a combination of empirical
estimates from previous climate impact studies and professional judgments when needed to close the
sometimes wide gaps in the literature. The table below shows the resulting global estimates of damages
in each category in the 1999 version of RICE.

Damages as a percent of global output at 2.5oC of warming

Output Population
weighted weighted

Agriculture

0.13

0.17

Sea level rise

0.32

0.12

Other market sectors

0.05

0.23

Health

0.10

0.56

Non-market amenities

-0.29

-0.03

Human settlements and ecosystems

0.17

0.10

Catastrophes

1.02

1.05

Total

1.50

1.88

(Nordhaus and Boyer 2000 p 91)

With damages in all categories estimated, the DICE damage function was then calibrated "so that the
optimal carbon tax and emissions control rates in DICE-99 matched the projections of these variables in
the optimal run of RICE-99" (Nordhaus and Boyer 2000 p 104).

Recent developments

Nordhaus (2010) presented results from an updated version of the RICE model. A major extension is a
new sea level rise damage function, now explicitly modeled by region as a function of the global average
sea level rise rather than rolled up in the aggregate damage function. "The RICE-2010 model provides a
revised set of damage estimates based on a recent review of the literature [Toll 2009, IPCC 2007],
Damages are a function of temperature, SLR, and CO2 concentrations and are region-specific. To give an
idea of the estimated damages in the uncontrolled (baseline) case, those damages in 2095 are... 2.8% of
global output, for a global temperature increase of 3.4oC above 1900 levels" (Nordhaus 2010 p 3). Other
parameter updates include climate sensitivity, now set to 3.2 degrees Celsius, the elasticity of the
marginal utility of income, now set to -1.5, and parameters that control economic growth rates, which
are re-calibrated such that world per capita consumption grows by an average rate of 2.2% per year for
the first 50 years.

References

Baker E, Clarke L, Weyant J. 2006. Optimal technology R&D in the face of climate uncertainty. Climatic
Change 78:157-159.

Darwin R, Tsigas M, Lewandrowski J, Raneses A. 1995. World Agriculture and Climate Change: Economic
Adaptations. Natural Resources and Environment Division, Economic Research Service, U.S.
Department of Agriculture. Agricultural Economic Report No. 703.

Hoel M, Sterner T. 2007. Discounting and relative prices. Climatic Change 84:265-280.

A-22

-------
IPCC (Intergovernmental Panel on Climate Change). 2007. Climate Change 2007: Impacts, Adaptation
and Vulnerability, Working Group II Contribution to the Intergovernmental Panel on Climate
Change, Summary for Policymakers. Cambridge, UK: Cambridge University Press.

Nordhaus WD. 1977a. Strategies for the control of carbon dioxide. Cowles Foundation discussion paper
no. 443.

Nordhaus WD. 1977b. Economic growth and climate: the carbon dioxide problem. The American
Economic Review 67(l):341-346.

Nordhaus WD. 1991. To slow or not to slow: the economics of the greenhouse effect. The Economic
Journal 101(407):920-937.

Nordhaus WD. 1992a. The "DICE" model: background and structure of a Dynamic /ntegrated Climate-
Economy model of the economics of global warming. Cowles Foundation discussion paper no.
1009.

Nordhaus WD. 1992b. Optimal greenhouse-gas reductions and tax policy in the "DICE" model. The
American Economic Review 83(2):313-317.

Nordhaus WD. 1994a. Managing the Global Commons: The Economics of Climate Change. Cambridge,
MA: The MIT Press.

Nordhaus WD. 1994b. Expert opinion on climatic change. American Scientist 82:45-51.

Nordhaus WD. 2010. Economic aspects of global warming in a post-Copenhagen environment.
Proceedings of the National Academy of Sciences 107(26):11721-11726.

Nordhaus WD. 2008. A Question of Balance: Weighing the Options on Global Warming Policies. Pre-
publication version, http://nordhaus.econ.yale.edu/Balance_2nd_proofs.pdf.

Nordhaus WD, Boyer J. 2000. Warming the World: Economic Models of Global Warming. Cambridge,
MA: The MIT Press.

Nordhaus WD, Yang Z. 1996. A regional dynamic general-equilibrium model of alternative climate-
change strategies. The American Economic Review 86(4):741-765.

Pizer WA. 1999. The optimal choice of climate change policy in the presence of uncertainty. Resource
and Energy Economics 21:255-287.

Popp D. 2005. ENTICE: endogenous technological change in the DICE model of global warming. Journal
of Environmental Economics and Management 48:742-768.

Tol R. 2009. The economic effects of climate change. Journal of Economic Perspectives 23:29-51.

Yang Z. 2008. Strategic Bargaining and Cooperation in Greenhouse Gas Mitigations: An Integrated
Assessment Modeling Approach. Cambridge, MA: The MIT Press.

A-23

-------
The PAGE09 model: Estimating climate impacts and the social cost of C02

Chris Hope (c.hope@jbs.cam.ac.uk)

October 2010

Introduction

PAGE09 is a new version of the PAGE integrated assessment model that values the impacts of
climate change and the costs of policies to abate and adapt to it. The model helps policy makers
explore the costs and benefits of action and inaction, and can easily be used to calculate the social
cost of C02 (SCC02) both today and in the future.

PAGE09 is an updated version of the PAGE2002 integrated assessment model. PAGE2002 was used
to value the impacts and calculate the social cost of C02 in the Stern review (Stern, 2007), the Asian
Development Bank's review of climate change in Southeast Asia (ADB, 2009), and the EPA's
Regulatory impact Analysis (EPA, 2010), and to value the impacts and costs in the Eliasch review of
deforestation (Eliasch, 2008). The PAGE2002 model is described fully in Hope, 2006, Hope, 2008a
and Hope, 2008b.

The update to PAGE09 been made to take account of the latest scientific and economic information,
primarily in the 4th Assessment Report of the IPCC (IPCC, 2007). This short paper outlines the
updated treatment of the science and impacts in the latest default version of the model, PAGE09
vl.7.

PAGE09 uses simple equations to simulate the results from more complex specialised scientific and
economic models. It does this while accounting for the profound uncertainty that exists around
climate change. Calculations are made for eight world regions, ten time periods to the year 2200, for
four impact sectors (sea level, economic, non-economic and discontinuities) which cover all impacts,
with the exception of socially contingent impacts such as massive forced migration and the threat of
war, for which there are currently no economic estimates.

The treatment of uncertainty is at the heart of the model. In the calculation of the SCC02, 45 inputs
are specified as independent probability distributions; these typically take a triangular form, defined
by a minimum, mode (most likely) and maximum value. The model is usually run 10000 times to
build up full probability distributions of the scientific and economic results, such as the global mean
temperature, the net present value of impacts and the SC C02.

The full set of model equations and default inputs to the model are contained in a technical report
available from the author. Initial results from the model are presented in a companion paper, 'The
Social Cost of C02 from the PAGE09 model'.

The changes made to PAGE2002 to create PAGE09 are outlined below under the following headings:
Science, Impacts and Adaptation.

Science

Inclusion of Nitrous Oxide

The number of gases whose emissions, concentrations and forcing are explicitly modelled is
increased from 3 in PAGE2002 to 4 in PAGE09. The forcing from N20 takes the same form as for

A-24

-------
CH4, based on the square root of the concentration. The excess forcing from gases not explicitly
modelled is now allowed to vary by policy.

Inclusion of transient climate response

In PAGE2002, the climate sensitivity is input directly as an uncertain parameter. The climate
sensitivity in PAGE09 is derived from two inputs, the transient climate response (TCR), defined as the
temperature rise after 70 years, corresponding to the doubling-time of C02 concentration, with C02
concentration rising at 1% per year, and the feedback response time (FRT) of the Earth to a change
in radiative forcing (Andrews and Allen, 2008). Default triangular distributions for TCR and FRT in
PAGE09 give a climate sensitivity distribution with a mean of 3 degC, and a long right tail, consistent
with the latest estimates from IPCC, 2007.

Feedback from temperature to the carbon cycle

The standard PAGE2002 model contains an estimate of the extra natural emissions of C02 that will
occur as the temperature rises (an approximation for a decrease in absorption in the ocean and
possibly a loss of soil carbon (Hope, 2006)). Recent model comparison exercises have shown that the
form of the feedback in PAGE2002 works well for business as usual emissions, but overestimates
concentrations in low emission scenarios (van Vuuren et al, 2009).

In PAGE09, the carbon cycle feedback (CCF) is introduced as a linear feedback from global mean
temperature to a percentage gain in the excess concentration of C02, to simulate the decrease in
C02 absorption on land and in the ocean as temperature rises (Friedlingstein et al, 2006). PAGE09 is
much better than PAGE2002 at simulating the carbon cycle feedback results for low emission
scenarios in Friedlingstein et al, 2006, van Vuuren et al, 2009.

Land temperature patterns by latitude

In PAGE2002, regional temperatures vary from the global mean temperature only because of
regional sulphate forcing. However, geographical patterns of projected warming show greatest
temperature increases over land (IPCC, 2007, chlO, p749), and a variation with latitude, with regions
near the poles warming more than those near the equator (IPCC, 2007, chlO, figure 10.8 and
supplementary material).

In PAGE09 the regional temperature is adjusted by a factor related to the effective latitude of the
region, and one related to the land-based nature of the regions. The adjustment is calculated for
each region using an uncertain parameter of the order of 1 degC representing the temperature
increase difference between equator and pole, and the effective absolute latitude of the region, and
an uncertain constant of the order of 1.4 representing the ratio between mean land and ocean
temperature increases.

Explicit incorporation of sea level rise

In PAGE2002, sea level rise is only included implicitly, assumed to be linearly related to global mean
temperature. This neglects the different time constant of the sea level response, which is longer
than the surface air temperature response (IPPC, 2007, p823).

In PAGE09, sea level is modelled explicitly as a lagged linear function of global mean temperature
(Grinsted et al, 2009). The IPCC has a sea level rise projection in 2100 of 0.4 - 0.7 m from pre-

A-25

-------
industrial times (IPCC, 2007, p409). A characteristic response time of between 500 and 1500 years in
PAGE09 gives sea level rises compatible with these IPCC results.

Impacts

Impacts as a proportion of GDP

In PAGE2002, economic and non-economic impacts before adaptation are a polynomial function of
the difference between the regional temperature and the tolerable temperature level, with regional
weights representing the difference between more and less vulnerable regions. These impacts are
then equity weighted, discounted at the consumption rate of interest and summed over the period
from now until 2200. There are several issues with this representation, including the lack of an
explicit link from GDP per capita to the regional weights, and the possibility that impacts could
exceed 100% of GDP with unfavourable parameter combinations.

In PAGE09, extra flexibility is introduced by allowing the possibility of initial benefits from small
increases in regional temperature (Tol, 2002), by linking impacts explicitly to GDP per capita and by
letting the impacts drop below their polynomial on a logistic path once they exceed a certain
proportion of remaining GDP to reflect a saturation in the vulnerability of economic and non-
economic activities to climate change, and ensure they do not exceed 100% of GDP.

Figure 1

Impact by temperature

proportion of GDP

Figure 1 shows such an impact function, with initial benefits (IBEN) of 1% of GDP per degree, with
impacts (W) of 4% of GDP at a calibration temperature (TCAL) of 2.5 degC, with a polynomial power
(POW) of 3, and an exponent with income (IPOW) of -0.5. The impact function has a saturation(ISAT)
starting at 50% of GDP, which keeps the impacts (blue line) below 100% of GDP even for the high
temperatures shown. The red line shows what the impacts would be if they continued to follow the
polynomial form without saturation.

Discontinuity impacts

As in PAGE2002, the risk of a large-scale discontinuity, such as the Greenland ice sheet melting, is
explicitly modelled. In PAGE09 the losses associated with a discontinuity do not all occur
immediately, but instead develop with a characteristic lifetime after the discontinuity is triggered
(Lenton et al, 2008).

A-26

-------
Equity weighting of impacts

In PAGE2002, impacts are equity weighted in a rather ad-hoc way, with the change in consumption
increased in poor regions and decreased in rich ones.

PAGE09 uses the equity weighting scheme proposed by Anthoff et al (2009) which converts changes
in consumption to utility, and amounts to multiplying the changes in consumption by

EQ(r,t) = (G(fr,0)/G(r,t))A EMUC

where G(r,t) is the GDP per capita in a region and year, G(fr,0) is today's GDP per capita in some
focus region (which could be the world as a whole, but in PAGE09 is normally the EU), and EMUC is
the negative of the elasticity of the marginal utility of consumption. This equity weighted damage is
then discounted at the utility rate of interest, which is the PTP rate.

Adaptation

The speed and amount of adaptation is modelled as a policy decision in PAGE. This allows the costs
and benefits of different adaptation decisions to be investigated. In PAGE2002, adaptation can
increase the natural tolerable level of temperature change, and can also reduce any climate change
impacts that still occur.

In PAGE09, there is assumed to be no natural tolerable temperature change, and adaptation policy is
specified by seven inputs for each impact sector. The tolerable temperature is represented by the
plateau, the start date of the adaptation policy and the number of years it takes to have full effect.
The reduction in impacts is represented by the eventual percentage reduction, the start date, the
number of years it takes to have full effect and the maximum sea level or temperature rise for which
adaptation can be bought; beyond this, impact adaptation is ineffective. Both types of adaptation
policy are assumed to take effect linearly with time. An adaptation policy in PAGE09 is thus defined
by 7 inputs for 3 sectors for 8 regions, giving 168 inputs in all. This is a simplification compared to the
480 inputs in PAGE2002.

Figure 2: Temperature and tolerable temperature by date (illustrative)

DegC

Year

A-27

-------
2020 and taking 20 years to implement fully. If the temperature rise is shown by the red line, there
will be 0.5 degC of impacts in 2000, increasing to 1 deg C by 2020, then reducing to 0 from 2030 to
2060. After 2060 the impacts start again, reaching 1 deg C by 2100.

Acknowledgement

Development of the PAGE09 model received funding from the European Community's Seventh
Framework Programme, as part of the ClimateCost Project (Full Costs of Climate Change, Grant
Agreement 212774) www/climatecost.eu and from the UK Department of Energy and Climate
Change. The development of the model also benefited from work with the UK Met Office funded
under the AVOID programme.

References

ADB, 2009, The Economics of Climate Change in Southeast Asia: A Regional Review, Asian
Development Bank, Philippines.

Andrews DG, and Allen MR, 2008, Diagnosis of climate models in terms of transient climate response
and feedback response time, Atmos. Sci. Let. 9:7-12

Anthoff D, Hepburn C and Tol RSJ, 2009, "Equity weighting and the marginal damage costs of climate
change", Ecological Economics, Volume 68, Issue 3,15 January 2009, 836-849.

Eliasch, Johann 2008 Climate Change: Financing Global Forests. Office of Climate Change, UK.Hope C,
2008a, Optimal carbon emissions and the social cost of carbon over time under uncertainty,
Integrated Assessment, 8,1, 107-122.

Friedlingstein P, Cox P, Betts R, Bopp I, Von bloh W, Brovkin V, Cadule P, Doney S, Eby M, Fung I, Bala
G, John J, Jones C, Joos F, Kato T, Kawamiya M, Knorr W, Lindsay K, Matthews HD, Raddatz T,
Rayner P, Reick C, Roeckner E, Schnitzler KG, Schnur R, Strassmann K, Weaver AJ, Yoshikawa
C, Zeng N, 2006, Climate-carbon cycle feedback analysis: results from the C4MIP model
intercomparison. J Clim 19:3337-3353.

Bloomberg, 2010, A fresh look at the costs of reducing US carbon emissions, Bloomberg New Energy
Finance.

EPA, 2010, appendix 15a, Social cost of carbon for regulatory impact analysis under executive order
12866,

http://wwwl.eere.energy.gov/buildings/appliance_standards/commercial/pdfs/smallmotor
s_tsd/sem_finalrule_appendixl5a.pdf

Aslak Grinsted , J. C. Moore, S. Jevrejeva, 2009, Clim Dyn, doi: 10.1007/s00382-008-0507-2.

Hope C, 2008a, Optimal carbon emissions and the social cost of carbon over time under uncertainty,
Integrated Assessment, 8,1, 107-122.

Hope C, 2008b, "Discount rates, equity weights and the social cost of carbon", Energy Economics, 30,
3, 1011-1019.

Hope C, 2006, "The marginal impact of C02 from PAGE2002: An integrated assessment model
incorporating the IPCC's five reasons for concern", Integrated Assessment, 6,1,19-56.

A-28

-------
IPCC, 2007, Climate Change 2007. The Physical Science Basis. Summary for Policymakers.

Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change. IPCC Secretariat Switzerland.

Lenton, T. M., H. Held, E. Kriegler, J. W. Hall, W. Lucht, S. Rahmstorf and H. J. Schellnhuber, 2008,
"Tipping elements in the Earth's climate system", Proceedings of the National Academy of
Sciences USA 105(6), 1786-1793.

Stern, Nicholas. 2007. The Economics of Climate Change: The Stern Review. Cambridge and New
York: Cambridge University Press.

Tol, R.S.J., 2002, "New estimates of the damage costs of climate change, Part II: dynamic estimates.",
Environ. Resour. Econ., 21, 135-160.

Detlef van Vuuren, Jason Lowe, Elke Stehfest, Laila Gohar, Andries Hof, Chris Hope, Rachel Warren,
Malte Meinshausen, Gian-Kasper Plattner, 2009, "How well do Integrated Assessment
Models simulate climate change?", Climatic Change, electronic publication date December
10, 2009, http://www.springerlink.com/content/l841558141481552/

A-29

-------
FUND - Climate Framework for Uncertainty, Negotiation and Distribution

David Anthoff1

University of California, Berkeley, CA, USA
Richard S.J. Tol

Economic and Social Research Institute, Dublin, Ireland

Institute for Environmental Studies, Vrije Universiteit, Amsterdam, The Netherlands
Department of Spatial Economics, Vrije Universiteit, Amsterdam, The Netherlands
Department of Economics, Trinity College, Dublin, Ireland

4 November 2010

FUND (the Climate Framework for Uncertainty, Negotiation and Distribution) is an integrated
assessment model linking projections of populations, economic activity and emissions to simple
greenhouse gas cycle, climate and sea-level rise models, and to a model predicting and monetizing
welfare impacts. Climate change welfare impacts are monetized in 1995 dollars and are modelled
over 16 regions. Modelled welfare impacts include agriculture, forestry, sea level rise, cardiovascular
and respiratory disorders influenced by cold and heat stress, malaria, dengue fever, schistosomiasis,
diarrhoea, energy consumption from heating and cooling, water resources, unmanaged ecosystems
and tropical and extratropical storms (Link and Tol, 2004). The source code, data, and a technical
description of the model can be found at http://www.fund-model.org.

Essentially, FUND consists of a set of exogenous scenarios and endogenous perturbations. The
model distinguishes 16 major regions of the world, viz. the United States of America, Canada,
Western Europe, Japan and South Korea, Australia and New Zealand, Central and Eastern Europe,
the former Soviet Union, the Middle East, Central America, South America, South Asia, Southeast
Asia, China, North Africa, Sub-Saharan Africa, and Small Island States. Version 3.6, the latest version,
runs to the year 3000 in time steps of one year.

The period of 1950-1990 is used for the calibration of the model, which is based on the IMAGE 100-
year database (Batjes and Goldewijk, 1994). The period 1990-2000 is based on observations
(http://earthtrends.wri.org). The 2000-2010 period is interpolated from the immediate past. The
climate scenarios for the period 2010-2100 are based on the EMF14 Standardized Scenario, which
lies somewhere in between IS92a and IS92f (Leggett et a!., 1992). The period 2100-3000 is
extrapolated.

The scenarios are defined by varied rates of population growth, economic growth, autonomous
energy efficiency improvements, and decarbonization of energy use (autonomous carbon efficiency
improvements), as well as by emissions of carbon dioxide from land use change, methane emissions,
and nitrous oxide emissions. FUND 3.5 introduced a dynamic biosphere feedback component that
perturbates carbon dioxide emissions based on temperature changes.

1 Contact: anthoff@berkeley.edu

A-30

-------
Emission reduction of carbon dioxide, methane and nitrous oxide is specified as in Tol (2006). Simple
cost curves are used for the economic impact of abatement, with limited scope for endogenous
technological progress and interregional spillovers (Tol, 2005).

The scenarios of economic growth are perturbed by the effects of climatic change. Climate-induced
migration between the regions of the world causes the population sizes to change. Immigrants are
assumed to assimilate immediately and completely with the respective host population.

The tangible welfare impacts are dead-weight losses to the economy. Consumption and investment
are reduced without changing the savings rate. As a result, climate change reduces long-term
economic growth, although consumption is particularly affected in the short-term. Economic growth
is also reduced by carbon dioxide abatement measures. The energy intensity of the economy and
the carbon intensity of the energy supply autonomously decrease over time. This process can be
accelerated by abatement policies.

The endogenous parts of FUND consist of the atmospheric concentrations of carbon dioxide,
methane and nitrous oxide, the global mean temperature, the effect of carbon dioxide emission
reductions on the economy and on emissions, and the effect of the damages on the economy caused
by climate change. Methane and nitrous oxide are taken up in the atmosphere, and then
geometrically depleted. The atmospheric concentration of carbon dioxide, measured in parts per
million by volume, is represented by the five-box model of Maier-Reimer and Hasselmann (1987). Its
parameters are taken from Hammitt et al. (1992).

The radiative forcing of carbon dioxide, methane, nitrous oxide and sulphur aerosols is determined
based on Shine et al. (1990). The global mean temperature, T, is governed by a geometric build-up to
its equilibrium (determined by the radiative forcing, RF), with a half-life of 50 years. In the base case,
the global mean temperature rises in equilibrium by 3.0°C for a doubling of carbon dioxide
equivalents. Regional temperature is derived by multiplying the global mean temperature by a fixed
factor, which corresponds to the spatial climate change pattern averaged over 14 GCMs
(Mendelsohn et al., 2000). The global mean sea level is also geometric, with its equilibrium level
determined by the temperature and a half-life of 50 years. Both temperature and sea level are
calibrated to correspond to the best guess temperature and sea level for the IS92a scenario of
Kattenberg et al. (1996).

The climate welfare impact module, based on Tol (2002a; Tol, 2002b) includes the following
categories: agriculture, forestry, sea level rise, cardiovascular and respiratory disorders influenced by
cold and heat stress, malaria, dengue fever, schistosomiasis, diarrhoea, energy consumption from
heating and cooling, water resources, unmanaged ecosystems and tropical and extratropical storms.
Climate change related damages are triggered by either the rate of temperature change
(benchmarked at0.04°C/yr) or the level of temperature change (benchmarked at 1.0°C). Damages
from the rate of temperature change slowly fade, reflecting adaptation (cf. Tol, 2002b).

A-31

-------
are monetized. The value of a statistical life is set to be 200 times the annual per capita income.2 The
resulting value of a statistical life lies in the middle of the observed range of values in the literature
(cf. Cline, 1992). The value of emigration is set to be three times the per capita income (Tol, 1995;
Tol, 1996), the value of immigration is 40 per cent of the per capita income in the host region (Cline,
1992). Losses of dryland and wetlands due to sea level rise are modelled explicitly. The monetary
value of a loss of one square kilometre of dryland was on average $4 million in OECD countries in
1990 (cf. Fankhauser, 1994). Dryland value is assumed to be proportional to GDP per square
kilometre. Wetland losses are according to estimates from Brander et al. (2006). Coastal protection
is based on cost-benefit analysis, including the value of additional wetland lost due to the
construction of dikes and subsequent coastal squeeze.

Other welfare impact categories, such as agriculture, forestry, hurricanes, energy, water, and
ecosystems, are directly expressed in monetary values without an intermediate layer of impacts
measured in their 'natural' units (cf. Tol, 2002a). Modelled effects of climate change on energy
consumption, agriculture, and cardiovascular and respiratory diseases explicitly recognize that there
is a climatic optimum, which is determined by a variety of factors, including plant physiology and the
behaviour of farmers. Impacts are positive or negative depending on whether the actual climate
conditions are moving closer to or away from that optimum climate. Impacts are larger if the initial
climate conditions are further away from the optimum climate. The optimum climate is of
importance with regard to the potential impacts. The actual impacts lag behind the potential
impacts, depending on the speed of adaptation. The impacts of not being fully adapted to new
climate conditions are always negative (cf. Tol, 2002b).

The welfare impacts of climate change on coastal zones, forestry, hurricanes, unmanaged
ecosystems, water resources, diarrhoea, malaria, dengue fever, and schistosomiasis are modelled as
simple power functions. Impacts are either negative or positive, and they do not change sign (cf. Tol,
2002b).

Vulnerability to climate change changes with population growth, economic growth, and
technological progress. Some systems are expected to become more vulnerable, such as water
resources (with population growth) and heat-related disorders (with urbanization), or more
valuable, such as ecosystems and health (with higher per capita incomes). Other systems are
projected to become less vulnerable, such as energy consumption (with technological progress),
agriculture (with economic growth) and vector- and water-borne diseases (with improved health
care) (cf. Tol, 2002b).

In the Monte Carlo analyses, most model parameters (including parameters for the physical
components as well as the economic valuation components) are varied. The probability density
functions are mostly based on expert guesses, but where possible "objective" estimates were used.
Parameters are assumed to vary independently of one another, except when there are calibration or
accounting constraints. "Preference parameters" like the discount rate or the parameter of risk
aversion are not varied in the Monte Carlo analysis. Details of the Monte Carlo analysis can be found
on FUND'S website at http://www.fund-model.org.

2 Note that this implies that the monetary value of health risk is effectively discounted with the pure rate of
time preference rather than with the consumption rate of discount (Horowitz, 2002). It also implies that, after
equity weighing, the value of a statistical life is equal across the world (Fankhauser et al., 1997).

A-32

-------
References

Batjes, J. J. and C. G. M. Goldewijk (1994). The IMAGE 2 Hundred Year (1890-1990) Database of the
Global Environment (HYDE), RIVM, Bilthoven, 410100082.

Brander, L., R. Florax and J. Vermaat (2006). "The Empirics of Wetland Valuation: A Comprehensive
Summary and a Meta-Analysis of the Literature." Environmental and Resource Economics
33(2): 223-250.

Cline, W. R. (1992). The Economics of Global Warming. Washington, DC, Institute for International
Economics.

Fankhauser, S. (1994). "Protection vs. Retreat - The Economic Costs of Sea Level Rise." Environment
and Planning A 27(2): 299-319.

Fankhauser, S., R. S. J. Tol and D. W. Pearce (1997). "The Aggregation of Climate Change Damages: A
Welfare Theoretic Approach." Environmental and Resource Economics 10(3): 249-266.

Hammitt, J. K., R. J. Lempert and M. E. Schlesinger (1992). "A Sequential-Decision Strategy for
Abating Climate Change." Nature 357: 315-318.

Horowitz, J. K. (2002). "Preferences in the Future." Environmental and Resource Economics 21: 241-
259.

Kattenberg, A., F. Giorgi, H. Grassl, G. A. Meehl, J. F. B. Mitchell, R. J. Stouffer, T. Tokioka, A. J.

Weaver and T. M. L. Wigley (1996). Climate Models - Projections of Future Climate. Climate
Change 1995: The Science of Climate Change -- Contribution of Working Group I to the
Second Assessment Report of the Intergovernmental Panel on Climate Change. J. T.
Houghton, L. G. Meiro Filho, B. A. Callanderet al. Cambridge Cambridge University Press:
285-357.

Leggett, J., W. J. Pepper and R. J. Swart (1992). Emissions scenarios for the IPCC: an update. Climate
Change 1992 - The Supplementary Report to the IPCC Scientific Assessment. J. T. Houghton,
B. A. Callander and S. K. Varney. Cambridge, Cambridge University Press: 71-95.

Link, P. M. and R. S. J. Tol (2004). "Possible Economic Impacts of a Shutdown of the Thermohaline
Circulation: an Application of FUND." Portuguese Economic Journal 3(2): 99-114.

Maier-Reimer, E. and K. Hasselmann (1987). "Transport and Storage of Carbon Dioxide in the Ocean:
An Inorganic Ocean Circulation Carbon Cycle Model." Climate Dynamics 2: 63-90.

Mendelsohn, R. O., M. E. Schlesinger and L. J. Williams (2000). "Comparing Impacts across Climate
Models." Integrated Assessment 1: 37-48.

Shine, K. P., R. G. Derwent, D. J. Wuebbles and J. J. Morcrette (1990). Radiative Forcing of Climate.
Climate Change - The IPCC Scientific Assessment. J. T. Houghton, G. J. Jenkins and J. J.
Ephraums. Cambridge Cambridge University Press: 41-68.

Tol, R. S. J. (1995). "The Damage Costs of Climate Change - Towards More Comprehensive
Calculations." Environmental and Resource Economics 5: 353-374.

A-33

-------
Tol, R. S. J. (1996). "The Damage Costs of Climate Change: Towards a Dynamic Representation."
Ecological Economics 19: 67-90.

Tol, R. S. J. (2002a). "Estimates of the damage costs of climate change. Part 1: Benchmark
estimates." Environmental and Resource Economics 21(2): 47-73.

Tol, R. S. J. (2002b). "Estimates of the damage costs of climate change. Part 2: Dynamic estimates."
Environmental and Resource Economics 21(2): 135-160.

Tol, R. S. J. (2005). "An Emission Intensity Protocol for Climate Change: An Application of FUND."
Climate Policy 4: 269-287.

Tol, R. S. J. (2006). "Multi-Gas Emission Reduction for Climate Change Policy: An Application of
FUND." Energy Journal 27: 235-250.

A-34

-------
Climate Damages in the MIT IGSM

John Reilly

MIT Joint Program on the Science and Policy of Global Change

Integrated assessment models (lAMs) have proven useful for analysis of climate change because
they represent the entire inhabited earth system, albeit typically with simplified model components
that are reduced form or more highly aggregated than for example, high resolution coupled
atmosphere-ocean-land general circulation models. The MIT Integrated Global System Model has
been developed to retain the flexibility to assemble earth system models of variable resolution and
complexity, however, even at its simplest it remains considerably more complex than most other
lAMs. In its simplest formulation it retains a full coupled general circulation model of the ocean and
atmosphere. Solved recursively, it solution time for a 100-year integration on a single node of
computer cluster is on the order of 24-36 hours, compared with seconds or minutes for other lAMs.
In that form it is not numerical feasible to solve the whole system as a fully dynamic optimizing
model to find an optimal cost-benefit solution as with the DICE, PAGE, or FUND models. Indeed,
inclusion of climate damages is still a work in progress in the MIT IGSM. The slow progress relative
to other efforts stems from a commitment to represent explicitly the physical impacts of climate ad
environmental change on activities (e.g. crop yields, water availability, coastal, inundation,
ecosystem processes and functioning, health outcomes, etc.) and represent market response to
these outcomes and value that response consistent with projections of resource prices as they are
projected to change in the future with economic growth and under different policies to mitigate
greenhouse gas emissions. This is in contrast to most of the optimizing models where climate
damages are estimated as a reduced form relationship in dollars of economic loss as a function of
mean global temperature change as a sufficient indicator of many dimensions of climate change,
and where the damage function is itself completely independent and separable from the economy
as it affects energy use and greenhouse gas emissions. The MIT IGSM is not designed to run well if
the purpose is to estimate a net present value social cost of carbon. The IGSM is best seen as
complementary to such efforts, and probably the focus on uncertainty in future climate outcomes is
one of the areas where it can make the most contribution t the social cost of carbon discussion.

Computationally efficient versions of the IGSM have been assembled for simulating large ensembles
to study uncertainty (Sokolov et al., 2009; Webster et al., 2009). Less complete but more
highly-resolved model components can be combined where research demands them, such as in the
study of the climate effect of aerosols (Wang, 2009; Wang et al., 2009a,b), changes in atmospheric
composition and human health (Selin et al., 2009a) or agricultural impacts and land use change
(Reilly, et al. 2007; Felzer et al., 2005; Melillo et al., 2009). The IGSM framework encompasses the
following components:

•	global economic activity resolved for large countries and regions that projects changes in
human activities as they effect the earth system including emissions of pollutants and
radiatively active substances and changes in land use and land cover;

•	earth system modules linked to the macroeconomy that address effects of climate and
environmental change on human activity, adaptation, and their consequences for the
macroeconomy (this includes modules that represent water use and land use at

A-35

-------
disaggregated spatial scales, energy and coastal infrastructure again at disaggregate spatial
scales, and demography, urbanization, urban air chemistry, and epidemiological
relationships that relate environmental change to human health);

•	the natural and managed land system including vegetation, hydrology, and biogeochemistry
as affected by human activity, environmental change and feedbacks on climate and
atmospheric composition;

•	the circulation and biogeochemistry of the ocean including its interactions with the
atmosphere, and representations of physical and biological oceanic responses to climate
change; and

•	the circulation and chemistry of the atmosphere including its role in radiative forcing, and
interactions with the land and ocean that determine climate change.

The suite of models that have been employed in this framework and their capabilities are briefly
described below.

Human Drivers and Analysis of Impacts

Human activities as they contribute to environmental change or are affected by it are represented in
multi-region, multi-sector models of the economy that solves for the prices and quantities of
interacting domestic and international markets for energy and non-energy goods as well as for
equilibrium in factor markets. The MIT Emissions Predictions and Policy Analysis (EPPA) model
(Paltsev et al., 2005) covers the world economy. It is built on the GTAP dataset (maintained at
Purdue University) of the world economic activity augmented by data on the emissions of
greenhouse gases, aerosols and other relevant species, and details of selected economic sectors. The
GTAP database allows flexibility to represent the world economy with greater country or sector
detail (the data set has 112 countries/regions and 57 economic sectors) that we aggregate further
for numerical efficiency. The model projects economic variables (GDP, energy use, sectoral output,
consumption, etc.) and emissions of greenhouse gases (C02, CH4, N20, HFCs, PFCs and SF6) and
other air pollutants (CO, VOC, NOx, S02, NH3, black carbon, and organic carbon) from combustion of
carbon-based fuels, industrial processes, waste handling, and agricultural activities.

The model has been augmented with supplemental physical accounts to link it with the earth system
components of the IGSM framework. To explore land use and environmental consequences, the
EPPA model (Gurgel, et al., 2007; Antoine, et al.,2008) is coupled with the Terrestrial Ecosystem
Model (Melillo et al., 2009). The linkage allows us to examine the ability of terrestrial ecosystems to
supply biofuels to meet growing demand for low-emissions energy sources along with the growing
demand for food, and to assess direct and indirect emissions from an expanded cellulosic bioenergy
program. The approach generates worldwide land-use scenarios at a spatial resolution of 0.5s
latitude by 0.5s longitude that varies with climate change. To analyze the economic impacts of air
pollution, the EPPA model is extended to include pollution-generated health costs, which reduce the
resources available to the rest of the economy (Nam et al., 2009; Selin et al., 2009a). The model
captures the amount of labor and leisure lost and additional medical services required due to acute
and chronic exposure to pollutants. The GTAP database allows considerable flexibility to represent
the world economy with greater country or sector detail (the underlying data has 112
countries/regions and 57 economic sectors). To assess distributional and regional impacts of carbon

A-36

-------
policy in the US, we use a model that is based on a state-level database and resolves large U.S. states
and multi-state regions and households of several income classes. The U.S. Regional Energy Policy
(USREP) model (Rausch et al., 2009; 2010) is nearly identical in structure to the EPPA model, except
that it models states and multi-state regions in the US instead of countries and multi-country
regions. The main difference from the EPPA model is the foreign sector that is represented as export
supply and import demand functions rather than a full representation of foreign economies. This
sacrifice of global coverage allows explicit modeling of distributional details of climate legislation and
linking the USEP model to very detailed electricity dispatch models. Efforts, under separate funding,
to integrate the USREP database into the GTAP base to provide a complete representation of trade
are underway. Physical impacts of environmental change have been included in the model as a
feedback by identifying factors (land productivity as it affects crops, livestock and forests) or sectors
affected by climate or by introducing additional household production sectors (household health
services that uses leisure and medical services). Thus, the approach is to work with underlying
input-output and Social Accounting Matrix (SAM) that is the basis for the economic model (Matus, et
al., 2008). This provides a framework for potentially linking other impacts such as coastal (Franck et
al., 2010a,b, 2010; Sugiyama, et al., 2008), agriculture (Reilly et al., 2007), health (Selin, et al., 2009;
Nam et al., 2010), or water (Strzepek et al., 2010) impacts.

Hydrology and Water Management

Research on components representing water management are aimed at linking hydrological changes
projected by the atmospheric component of the IGSM to impacts of those changes on water
availability and use for irrigation, energy, industry and households, and in-stream ecological services.
These demands are driven by macroeconomic changes and changes in water supply and will in turn
affect the economy as represented in the EPPA and the USREP models. Techniques have been
developed to take IGSM 2-D GCM outputs and use results from the IPCC AR-4 3-D GCMs to provide
IGSM-generated 3-D climates to the hydrology component of the IGSM-Land Surface Model (NCAR
Community Land Model, CLM) to project runoff. Tests have been conducted for the US, where
adequate data are available, to determine the spatial resolution needed to provide reliable
estimates of runoff using CLM. A Water Resources System (WRS) model has been adapted from and
further developed in collaboration with the International Food Policy Research Institute (IFPRI) to
represent river reaches and natural and management components that affect stream-flow. The
major natural components are wetlands, unmanaged lakes, groundwater aquifers and flood plains.
The major managed components are reservoirs and managed lakes, and water diversions for
irrigation, cooling in thermal power plants, and industrial and household needs. Constraints on use
to preserve in-stream ecological water requirements can be imposed.

A series of models were adapted and developed to represent water use. These include a crop
growth model (CLICROP) developed to be able to run at 2° latitude-longitude grid resolution while
retaining the accuracy of a 0.5° resolution, thereby improving numerical efficiency of the modeling
system (Strzepek et al., 2010a). A model of Municipal and Industrial water demand driven by per
capita GDP was developed jointly with the University of Edinbough (Hughes et al., 2010; Strzepek et
al., 2010a). To investigate changes in thermal electric cooling water demands, a geospatial
methodology based on energy generation and geo-hydroclimatic variables has been developed
(Strzepek et al., 2010b). An assessment of environmental flow requirements to assure aquatic
ecosystem viability has been undertaken and an approach for using the IGSM was selected (Strzepek

A-37

-------
& Boehlert, 2010; Strzepek et al., 2010a). These developments provide the foundation for
completing linkages of the WRS with other IGSM components.

Atmospheric Dynamics and Physics

Research utilizing the IGSM framework has typically included a 2-D atmospheric (zonally-averaged
statistical dynamical) component based on the Goddard Institute for Space Studies (GISS) GCM. The
IGSM version 2.2 couples this atmosphere with a 2D ocean model (latitude, longitude) with
treatment of heat and carbon flows into the deep ocean (Sokolov et al, 2005). The IGSM version 2.3
(where 2.3 indicates the 2-D atmosphere/full 3-D ocean GCM configuration) (Sokolov et al., 2005;
Dutkiewicz et al., 2005) is a fully-coupled Earth system model that allows simulation of critical
feedbacks among its various components, including the atmosphere, ocean, land, urban processes
and human activities. A limitation of the IGSM2.3 is the above 2 -D (zonally averaged) atmosphere
model that does not permit direct regional climate studies. For investigations requiring 3 -D
atmospheric capabilities, the National Center for Atmospheric Research (NCAR) Community
Atmosphere Model version 3 (CAM3) (Collins et al., 2006) has been used with offline coupling.

The IGSM2.3 provides an efficient tool for generating probabilistic distributions of sea surface
temperature (SST) and sea ice cover (SIC) changes for the 21st century under varying emissions
scenarios, climate sensitivities, aerosol forcing and ocean heat uptake rates. Even though the
atmospheric component of the IGSM2.3 is zonally-averaged, it provides heat and fresh-water fluxes
separately over the open ocean and over sea ice, as well as their derivatives with respect to surface
temperature. This resolution allows the total heat and fresh -water fluxes for the IGSM2.3 oceanic
component to vary by longitude as a function of SST so that, for example, warmer ocean locations
undergo greater evaporation and receive less downward heat flux.

In offline coupling between the IGSM2.3 and CAM3, the 3-D atmosphere is driven by the IGSM2.3
SST anomalies with a climatological annual cycle taken from an observed dataset (Hurrell et al.,
2008), instead of the full IGSM2.3 SSTs, to provide a better SST annual cycle, and more realistic
regional feedbacks between the ocean and atmospheric components. This approach yields a
consistent regional distribution and climate change over the 20th century as compared to
observational datasets, and can then be used for simulations of the 21st century.

Urban and Global Atmospheric Chemistry and Aerosols

The model of atmospheric chemistry includes an analysis of all the major climate-relevant reactive
gases and aerosols at urban scales coupled to a model of the chemistry of species exported from
urban/regional areas (plus the emissions from non-urban areas) at global scale. For calculation of the
atmospheric composition in non-urban areas, the atmospheric dynamics and physics model is linked
to a detailed 2-D zonal-mean model of atmospheric chemistry. The atmospheric chemical reactions
are thus simulated in two separate modules: one for the sub-grid-scale urban chemistry and one for
the 2-D model grid. In addition, offline studies also utilize the 3-D capabilities of the CAM3 as noted
above, as well as the global Model of Atmospheric Transport and Chemistry (MATCH; Rasch et al.,
1997), and the GEOS-Chem global transport model (http://geos-chem.org/).

Global Atmospheric Chemistry: Modeling of atmospheric composition at global scale is by the above
2-D zonal-mean model with the continuity equations for trace constituents solved in mass
conservative or flux form (Wang et al., 1998). The model includes 33 chemical species including black
carbon aerosol, and organic carbon aerosol, and considers convergences due to transport,

A-38

-------
convection, atmospheric chemical reactions, and local production/loss due to surface
emission/deposition. The scavenging of carbonaceous and sulfate aerosol species by precipitation is
included using a method based on a detailed 3-D climate-aerosol-chemistry model (Wang, 2004)
that has been developed in collaboration with NCAR. The interactive aerosol-climate model is used
offline to model distributions of key chemical species, such as those utilized in the development of
the urban air chemistry model.

Urban Air Chemistry: A reduced-form urban chemical model that can be nested within coarser-scale
models has been developed and implemented to better represent the sub-gridscale urban chemical
processes that influence air chemistry and climate (Cohen & Prinn, 2009). This is critical both for
accurate representation of future climate trends and for our increasing focus on impacts, especially
to human health and down-wind ecosystems. The MIT Urban Chemical Metamodel (UrbanM) is an
update of our Mayer et al. (2000) model, and applies a third-order polynomial fit to the CAMx
regional air quality model (ENVIRON, 2008) for 41 trace gases and aerosols for a 100 km x 100 km
urban area. While a component of the IGSM, the urban modular UrbanM is also designed to
facilitate inclusion in a number of other global atmospheric models. It has recently been embedded
in the MIT interactive climate-aerosol simulation based on CAM3 in order to assess its influence on
the concentration and distribution of aerosols in Asia (Cohen et al., 2009). Work is underway to
further test the sensitivity of the probabilistic uncertainty results with the IGSM2.2/2.3 to this
improved representation of urban chemistry. The UrbanM is presently being benchmarked in a case
study of the Northeast U.S., and embedded in a global 3-D chemistry-climate model including a
detailed chemical mechanism (NCAR CAM-Chem).

Chemistry-Climate-Aerosol Component: A 3-D interactive aerosol-climate model has been
developed at MIT in collaboration with NCAR based on the finite volume version of the Community
Climate ystem Model (CCSM3; Collins et al., 2006). Focused on analysis of aerosols, this companion
sub-model is not yet integrated into the IGSM but serves as a step toward overcoming the
limitations for analysis of regional issue using the IGSM 2-D atmosphere configuration. The modeled
aerosols include three types of sulfate, two external mixtures of black carbon (BC), one type of
organic carbon, and one mixed state (comprised primarily of sulfate and other compounds coated
on BC); each aerosol type has a prognostic size distribution (Kim et al., 2008). The model
incorporates such processes as aerosol nucleation, diffusive growth, coagulation, nucleation and
impaction scavenging, dry deposition, and wet removal. It has been used to investigate the global
aerosol solar absorption rates (Wang et al., 2009a) and the impact of absorbing aerosols on the
Indian summer monsoon (Wang et al., 2009b). The UrbanM has recently been introduced into this
model to study the roles of urban processing in global aerosol microphysics nd chemistry and to
compute the abundance and radiative forcing of anthropogenic aerosols (Cohen et al., 2010). This
effort also serves as the first step toward introducing the full UrbanM into the 3 -D
aerosol-chemistry-climate framework.

Ocean Component

The IGSM framework retains the capability to represent ocean physics and biogeochemistry in
several different ways depending onthe question to be addressed. It can utilize either the 2-D
(latitude-longitude) mixed-layer anomaly-diffusing ocean model or the fully 3-D ocean general
circulation model (GCM). The IGSM with the 2-D ocean is more computationally efficient and more
flexible for studies of uncertainty in climate response. In applications that need to account for

A-39

-------
atmosphere-ocean circulation interactions, or for more detailed studies involving ocean
biogeochemistry, the diffusive ocean model is replaced by the fully 3D ocean GCM component.

2-D Ocean Model: The IGSM2.2 has a mixed-layer anomaly-diffusing ocean model with a horizontal
resolution of 4e in latitude and 5e in longitude. Mixed-layer depth is prescribed based on
observations as a function of time and location. Vertical diffusion of anomalies into the deep ocean
utilizes a diffusion coefficient that varies zonally as well as meridionaly. The model includes specified
vertically-integrated horizontal heat transport by the deep oceans, and allows zonal as well as
meridional transport. A thermodynamic ice module has two layers and computes the percentage of
area covered by ice and ice thickness, and a diffusive ocean carbon module is included (Sokolov et al,
2005; Holian et al., 2001; Follows et al. 2006).

3-D Ocean General Circulation Model: The IGSM2.3 ocean component is based on a state-of-the-art
3D MIT ocean GCM (Marshall et al., 1997). Embedded in the ocean model is a thermodynamic
sea-ice module (Dutkiewicz et al., 2005). The 3D ocean component is currently configured in either a
coarse resolution (4° by 4° horizontal, 15 layers in the vertical) or higher resolution (2° by 2.5°, 23
layers; or alternate configuration with higher resolution in the topics) depending on the focus of
study and the computational resources available. The efficiency of ocean heat uptake can be varied
(e.g., Dalan et al. 2005) and the coupling of heat, moisture, and momentum can be modified for
process studies (e.g., Klima 2008). In addition, a biogeochemical component with explicit
representation of the cycling of carbon, phosphorus and alkalinity can be incorporated. Export of
organic and particulate inorganic carbon from surface waters is parameterized and biological
productivity is modelle as a function of available nutrients and light (Dutkiewicz et al., 2005). Air-sea
exchange of C02 allows feedback between the ocean and atmosphere components. An additional
module with explicit representation of the marine ecosystem (Follows et al., 2007) has been
introduced in an "offline" (i.e. without full feedbacks to the full IGSM) configuration (see further
discussion in Section 4.2.3).

Land and Vegetation Processes

The Global Land System (GLS, Schlosser et al., 2007) of the IGSM links biogeophysical, ecological, and
biogeochemical components: (1) the NCAR Community Land Model (CLM), which calculates the
global, terrestrial water and energy balances; (2) the Terrestrial Ecosystems Model (TEM) of the
Marine Biological Laboratory, which simulates carbon (C02) fluxes and the storage of carbon and
nitrogen in vegetation and soils including net primary production and carbon sequestratio or loss;
and (3) the Natural Emissions Model (NEM), which simulates fluxes of CH4 and N20, and is now
embedded within TEM. A recent augmentation to the GLS enables a more explicit treatment of
agricultural processes and a treatment of the managed waer systems (Strzepek et al., 2010a). The
linkage between econometrically based decisions regarding land use (from EPPA) and plant
productivity from TEM has been enhanced (Cai et al., 2010). And the treatment of migration of plant
species to include meteorological constraints (i.e. winds) to seed dispersal has been enhanced (Lee
et al., 2009, 2010a,b). The representation of natural and vegetation processes also includes a
diagnosis of the expansion of lakes and changes of methne emissions from thermokarst lake
expansion/degradation (Gao et al., 2010; Schlosser et al., 2010). In addition, continuing updates to
CLM and TEM are also incorporated into the GLS framework. In all these applications, the GLS is
operating under a range of spatial resolutions (from zonal to gridded as low as 0.5s), and is
configured in its structural detail to accommodate various levels of process -oriented research both

A-40

-------
in a coupled framework within the IGSM as well as in standalone studies (i.e. with prescribed
atmospheric forcing).

References

Antoine, B., A. Gurgel, J. M. Reilly, 2008: Will Recreation Demand for Land Limit Biofuels Production?
Journal of Agricultural & Food Industrial Organization, 6(2), Article 5.
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint08-15.pdf)

Babiker, M., A. Gurgel, S. Paltsev and J. Reilly, 2009: Forward-looking versus recursive-dynamic

modeling in climate policy analysis: A comparison. Economic Modeling, 26(6): 1341-1354;
MIT Joint Program Reprint 2009-8

(http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint09 -8.pdf).

Cai, Y., D. Kicklighter, A. Gurgel, S. Paltsev, T. Cronin, J. Reilly, and J. Melillo, 2010; Green House Gas
Mitigation, Bio-fuel and Land-use Change: A Dynamic Analysis. Poster presentation. 2010
DOE Climate Change Modeling Program Science Team Meeting, Gaithersburg, Maryland,
March 29-April 2.

Cohen, J. B., C. Wang, and R. Prinn, 2009: The Impact of Detailed Urban Scale Processing on the
Simulation of the Concentration and Distribution of Aerosols in Asia. Eos. Trans. AGU,
90(52), Fall Meet. Suppl., Abstract A14A-05.

Cohen, J.B., and R. Prinn, 2009: Development of a fast and detailed model of urban-scale chemical
and physical processing. MIT Joint Program Report 181, October, 74 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC_Rptl81.pdf).

Collins, W.D, C.M. Bitz, M.L. Blackmon, G.B. Bonan, C.S. Bretherton, J.A. Carton, P. Chang, S.C.

Doney, J.J. Hack, T.B. Henderson, J.T. Kiehl, W.G. Large, D.S. McKenna, B.D. Santer and R.D.
Smith. 2006: The Community Climate System Model Version 3 (CCSM3). Journal of Climate,
19(11): 2122-2143.

Dalan, F., P.H. Stone, I.V. Kamenkovich, J.R. Scott, 2005: Sensitivity of the ocean's climate to

diapycnal diffusivity in and EMIC. Part I: equilibrium state. J. of Physical Oceanography, 18,
2460-2481; MIT Joint Program Reprint 2005-6

(http://globalchange.mit.edu/files/document/MITJPSPGC_Reprint05 -6.pdf)

Dutkiewicz. S., A. Sokolov, J. Scott and P. Stone, 2005: A Three-Dimensional Ocean-Seaice-Carbon
Cycle Model and its Coupling to a Two-Dimensional Atmospheric Model: Uses in Climate
Change Studies. MIT Joint Program Report 122.
(http://mit.edu/globalchange/www/MITJPSPGC_Rptl22.pdf).

Dutkiewicz, S., M.J. Follows and J.G. Bragg, 2009: Modeling the coupling of ocean ecology and
biogeochemistry. Global Biogeochemical Cycles, 23, GB4017; MIT Joint Program Reprint
2009-24 (http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint09-24.pdf).

Felzer, B., J. Reilly, J. Melillo, D. Kicklighter, M. Sarofim, C. Wang, R. Prinn and Q. Zhuang, 2005:
Future effects of ozone on carbon sequestration and climate change policy using a global

A-41

-------
biogeochemical model, Climatic Change, 73(3): 345-373; MIT Joint Program Reprint 2005-8
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint05 -8.pdf).

Follows, M., T. Ito and S. Dutkiewicz, 2006: On the solution of the carbonate chemistry system in
ocean biogeochemistry models. Ocean Modeling, 12, 290-301.
(http://ocean.mit.edu/~mick/Papers/Follows -etal-OM2006.pdf)

Follows, M., S. Dutkiewicz, S. Grant and S. Chisholm, 2007: Emergent biogeography of microbial
communities in a model ocean. Science, 315(5820): 1843-1846
(http://globalchange.mit.edu/hold/restricted/Follows_Science315_2007.pdf).

Franck, T., 2009a: Coastal Communities and Climate Change: A Dynamic Model of Risk Perception,
Storms, and Adaptation, Ph.D. Dissertation in Technology, Management, and Policy,
Massachusetts Institute of Technology

(http://globalchange.mit.edu/files/document/Franck_PhD_09.pdf)

Franck, T., 2009b: Coastal adaptation and economic tipping points. Management of Environmental
Quality, 20(4), 434-450; MIT Joint Program Reprint 2009-9
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint09 -9.pdf).

Gao, X., C.A. Schlosser, A. Sokolov, and K. Walter, 2010: Quantifying future changes in high-latitude
methane emissions under regional climate change uncertainty, 2010 State of the Arctic
Conference - At the Forefront of Global Change, March 16-19 2010, Miami, Florida

Gurgel, A., J.M. Reilly, and S. Paltsev, 2007. Potential land use implications of a global biofuels
industry. Journal of Agricultural & Food Industrial Organization, 5(2): Article 9.
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint07-14.pdf)

Holian, G., A.P. Sokolov and R.G. Prinn, 2001: Uncertainty in Atmospheric C02 Predictions from a
Parametric Uncertainty Analysis of a Global Ocean Carbon Cycle Model. MIT Joint Program
Report 80 (http://mit.edu/globalchange/www/MITJPSPGC_Rpt80.pdf).

Hughes, G., P. Chinowsky and K. Strzepek, 2010: The Costs of Adaptation to Climate Change for
Water Infrastructure in OECD Countries. Global Environmental Change, in press.

Hurrell, J.W., J.J. Hack, D. Shea, J.M. Caron, and J. Rosinski, 2008: A New Sea Surface Temperature
and Sea Ice Boundary Dataset for the Community Atmosphere Model. J. Climate, 21,
5145-5153.

Kim, D., C. Wang, A.M.L. Ekman, M.C. Barth and P. Rasch, 2008: Distribution and direct radiative

forcing of carbonaceous and sulfate aerosols in an interactive size-resolving aerosol-climate
model. J. Geophys. Research, 113, D16309; MIT Joint Program Reprint 2008-11
(http://globalchange.mit.edu/files/document/MITJPSPGC_Reprint08-ll.pdf).

Klima, K., 2008: Effects of Variable Wind Stress on Ocean Heat Content. Master of Science Thesis,
Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology,
September (http://mit.edu/globalchange/www/docs/Klima_MS_08.pdf).

A-42

-------
Lee, E., C.A. Schlosser, B.S. Felzer and R.G. Prinn, 2009: Incorporating plant migration constraints into
the NCAR CLM-DGVM model: Projections of future vegetation distribution in high latitudes.
Eos Trans. AGU, 90(52), Fall Meeting Supplement, Abstract B41C-0336.

Lee, E., C. A. Schlosser, X.Gao, A. Sokolov, and R. G. Prinn, 2010a: The CLM-DGVM-SEED:

incorporating a meteorological constraint into plant migration in the CLM-DGVM, NCAR
LMWG/BGCWG meetings, February 8, 2010, Boulder, CO.

Lee, E., C. A. Schlosser, X.Gao, and R. G. Prinn, 2010b: Incorporating a meteorological constraint to
plant migration in a dynamic vegetation model: Projections of future vegetation distribution
in the Pan-Arctic, State of the Arctic conference, March 16-19, 2010, Miami, FL.

Marshall, J., A. Adcroft, C. Hill, L. Perelman, and C. Heisey, 1997: A finite-volume, incompressible
Navier Stokes model for studies of the ocean on parallel computers. J. Geophys. Res., 102
(C3), 5753-5766 (http://paoc.mit.edu/paoc/papers/finite.pdf).

Matus, K., T. Yang, S. Paltsev, J. Reilly and K.-M. Nam, 2008: Toward integrated assessment of

environmental change: Air pollution health effects in the USA. Climatic Change, 88(1): 59-92;
MIT Joint Program Reprint 2007-12

(http://globalchange.mit.edu/hold/restrictedReprints/MITJPSPGC_Reprint07-12.pdf).

Mayer, M., C. Wang, M. Webster, and R. G. Prinn, 2000: Linking local air pollution to global chemistry
and climate, J. Geophysical Research, 105(D18): 22,869-22,896;
(http://globalchange.mit.edu/files/document/MITJPSPGC_ReprintOO -5.pdf).

Melillo, J., J. Reilly, D. Kicklighter, A. Gurgel, T. Cronin, S. Paltsev, B. Felzer, X. Wang, A. Sokolov and
C. A. Schlosser, 2009: Indirect Emissions from Biofuels: How Important?, Science 326:
1397-1399; MIT Joint Program Reprint 2009-20

(http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint09-20.pdf).

Monier, E., J. Scott, A. Sokolov and A. Schlosser, 2010; MIT IGSM: Toward a 3-Dimensional Integrated
Assessment Model. Poster presentation. 2010 DOE Climate Change Modeling Program
Science Team Meeting, Gaithersburg, Maryland, March 29-April 2.

Nam, K. M., N.E. Selin, J. M. Reilly, and S. Paltsev, 2010: Measuring welfare loss caused by air
pollution in Europe: A CGE Analysis. Energy Policy, in press
(http://globalchange.mit.edu/hold/pending/NamEtAI-EnergyPolicy2010.pdf).

Paltsev S., J. Reilly, H. Jacoby, R. Eckaus, J. McFarland, M. Sarofim, M. Asadoorian and M. Babiker,
2005: The MIT Emissions Prediction and Policy Analysis (EPPA) Model: Version 4. MIT Joint
Program Report 125 (http://globalchange.mit.edu/files/document/MITJPSPGC_Rptl25.pdf).

Rasch, P.J., N.M. Mahowald, and B.E. Eaton, 1997: Representations of transport, convection, and the
hydrologic cycle in chemical transport models: Implications for the modeling of short-lived
and soluble species. J. Geophys. Res.-Atmospheres, 102 (D23), 28127-28138.

Rausch, S., G. Metcalf, J.M. Reilly and S. Paltsev, 2009: Distributional Impacts of a U.S. Greenhouse
Gas Policy: A General Equilibrium Analysis of Carbon Pricing. MIT Joint Program Report 182
(http://globalchange.mit.edu/files/document/MITJPSPGC_Rptl82.pdf).

A-43

-------
Rausch, S., G.E. Metcalf, J.M. Reilly and S. Paltsev, 2010: Distributional Implications of Alternative
U.S. Greenhouse Gas Control Measures. The B.E. Journal of Economic Analysis & Policy, in
press; also MIT Joint Program Report 185, June, 42 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC_Rptl85.pdf)

Reilly, J., S. Paltsev, B. Felzer, X. Wang, D. Kicklighter, J. Melillo, R. Prinn, M. Sarofim, A. Sokolov and
C. Wang, 2007: Global economic effects of changes in crops, pasture and forests due to
changing climate, carbon dioxide, and ozone. Energy Policy, 35(11): 5370-5383; MIT Joint
Program Reprint 2007-11

(http://globalchange.mit.edu/files/document/MITJPSPGC_Reprint07-ll.pdf)

Schlosser, C.A., D. Kicklighter, and A. Sokolov, 2007: A Global Land System Framework for Integrated
Climate-Change Assessments, Report 147, May 2007, 60 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC_Rptl47.pdf)

Schlosser, C. A., X. Gao, K. Walter, A. Sokolov, D. Kicklighter, C. Forest, Q. Zhuang, J. Melillo, and R.
Prinn, 2010a: Quantifying climate feedbacks from abrupt changes in high-latitude trace-gas
emissions. Presentation to the DOE Integrated Climate Change Modeling Science Team
Meeting, April 1, 2010, Gaithersburg, MD.

Selin, N.E., S. Wu, K.-M. Nam, J.M. Reilly, S. Paltsev, R.G. Prinn and M.D. Webster, 2009a: Global
Health and Economic Impacts of Future Ozone Pollution. Environmental Research Letters,
4(4): 044014; MIT Joint Program Reprint 2009-17

(http://globalchange.mit.edu/files/document/MITJPSPGC_Reprint_09-17.pdf)

Sokolov, A.P., C.A. Schlosser, S. Dutkiewicz, S. Paltsev, D.W. Kicklighter, H.D. Jacoby, R.G. Prinn, C.E.
Forest, J. Reilly, C. Wang, B. Felzer, M.C. Sarofim, J. Scott, P.H. Stone, J.M. Melillo and J.
Cohen, 2005: The MIT Integrated Global System Model (IGSM) Version 2: Model Description
and Baseline Evaluation. MIT Joint Program Report 124, July, 40 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC_Rptl24.pdf).

Sokolov, A., P. Stone, C. Forest, R. Prinn, M. Sarofim, M. Webster, S. Paltsev, C.A. Schlosser, D.
Kicklighter, S. Dutkiewicz, J. Reilly, C. Wang, B. Felzer, J. Melillo, H. Jacoby, 2009a:
Probabilistic forecast for 21st century climate based on uncertainties in emissions (without
policy) and climate parameters. Journal of Climate, 22(19): 5175-5204; MIT Joint Program
Reprint 2009-12

(http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint09-12.pdf).

Strzepek, K., and B. Boehlert, 2010: Competition for water for the food system. Philosophical
Transactions of the Royal Society, in press.

Strzepek, K., A. Schlosser, W. Farmer, S. Awadalla, J. Baker, M. Rosegrant and X. Gao, 2010a.
Modeling the Global Water Resource System in an Integrated Assessment Modeling
Framework: IGSM-WRS, MIT Joint Program Report No. 189, Cambridge, MA.

Strzepek, K., J. Baker, W. Farmer, C.A. Schlosser, 2010b: The Impact of Renewable Electricity Futures
on Water Demand in the United States. MIT Joint Program Report in preparation.

A-44

-------
Sugiyama, M., R.J. Nicholls and A. Vafeidis, 2008: Estimating the Economic Cost of Sea-Level Rise.
MIT Global Change Joint Program, Report 156, April, 40 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC_Rptl56.pdf).

Wang, C., 2004: A modeling study on the climate impacts of black carbon aerosols. J. Geophysical
Research, 109(D3): D03106; MIT Joint Program Reprint 2004-2
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint04-2.pdf)

Wang, C., 2009: The sensitivity of tropical convective precipitation to the direct radiative forcings of
black carbon aerosols emitted from maor regions. Annales Geophysicae, 27(10): 3705-311;
MIT Joint Program Reprint 2009-11

(http://globalchange.mit.edu/files/document/MITJPSPGC_Reprint09-ll.pdf).

Wang, C., R. G. Prinn, and A. Sokolov, 1998: A global interactive chemistry and climate model:

Formulation and testing. J. Geophysical Research, 103(D3): 3399-3418; MIT Joint Program
Reprint 1998-5 (http://globalchange.mit.edu/files/document/MITJPSPGC_Reprint98-5.pdf)

Wang, C., G. Jeong and N. Mahowald, 2009a: Particulate absorption of solar radiation:

anthropogenic aerosols vs. dust. Atmospheric Chemistry and Physics, 9: 3935-3945; MIT
Joint Program Reprint 2009-10

(http://globalchange.mit.edu/files/document/MITJPSPGC_Reprint09-10.pdf).

Wang, C., D. Kim, A.M.L. Ekman, M.C. Barth and P. Rasch, 2009b: The impact of anthropogenic

aerosols on Indian summer monsoon. Geophysical Research Letters, 36, L21704; MIT Joint
Program Reprint 2009-21

(http://globalchange.mit.edu/hold/restricted/MITJPSPGC_Reprint09-21.pdf).

Webster, M., A. Sokolov, J. Reilly, C. Forest, S. Paltsev, A. Schlosser, C. Wang, D. Kicklighter, M.

Sarofim, J. Melillo, R. Prinn and H. Jacoby, 2009: Analysis of Climate Policy Targets under
Uncertainty. MIT Joint Program Report 180, September, 53 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC_Rptl80.pdf).

A-45

-------
Modeling the Impacts of Climate Change: Elements of a Research Agenda

Ian SueWing
Associate Professor

Department of Geography & Environment, Boston University

Karen Fisher-Vanden
Associate Professor

Department of Agricultural Economics, Pennsylvania State University
Elisa Lanzi

OECD Environment Directorate
Introduction: What is an IAM?

As illustrated in Figure 1, an integrated assessment model (IAM) of climate change is typically
constructed from three interlinked sub-models, an economic model (1), a climate model (2) and an
impacts model (3). It is logical to begin with the economic sub-model, which is responsible for
generating time-paths of global emissions of greenhouse gases (GHGs—principally carbon dioxide,
C02) (a). These serve as inputs to the climate submodel, which uses them to project changes in the
magnitude of meteorological variables such as temperature, precipitation or sea level rise (b).
Finally, the changes in climate parameters are translated into projections of global- or regional-scale
economic losses by an impacts sub-model, whose primary role is to capture the feedback effect of
dangerous near-term anthropogenic interference with the climate on economic activity over the
longterm future (c).

Innovation is a key modulator of the clockwise circulation of the feedback loop in the figure.
Improvements in the productivity of labor induce more rapid growth and increase the demand for
fossil energy resources, which has a first-order amplifying effect on emissions (A). Energy- or
emissions-saving technological progress tends to depress the emission intensity of the economy,
slowing the rate of increase in fossil fuel use; conversely, productivity improvements in energy
resource extraction lower the price of fossil fuels and induce substitution toward them, increasing
emissions (B). Lastly, we can imagine that there may be innovations that boost the effectiveness of
defensive expenditures undertaken in response to the threat of climate damages, or investments in
creating new knowledge that enables humankind to mitigate some climate damages (C). This last
category is the most speculative, as impacts will manifest themselves several decades in the future,
when the state of technology is likely to be quite different from today.

A-46

-------
Figure 1: Integrated Assessment of Climate Change and the Effects of Innovation

Land of Cockaigne: An IAM with Regional, Sectoral and Climate Impact Detail

Imagine that there were relatively few constraints to either our computational resources or our
ability to foresee the impacts of climate change. In such a world, what would an IAM look like? We
could then specify a RICE- or AD-WITCH-type IAM that resolved (a) the detailed sectoral structure of
production in various regions, (b) the effects of climate impacts on the productivity of those sectors,
(c) the manner in which different impact endpoints combined to generate the resultant productivity
effects, and (d) the response of the full range of impacts to changes in climatic variables at regional
scale.

Let us write down such a model, and exploit its structure to assess the implications for the social cost
of carbon. Define the following nomenclature:

Set indexes:

* = {o,...,r}

/ = {0,...,Z}
j = {0,
m = {0
/ = {0,...,F}

Control variables:

cllu
qfj,t
Qu

A-47

Time periods
World regions
Industry sectors
Meteorological characteristics
Climate impact endpoints

Sectoral energy input
Sectoral capital input
Aggregate consumption

-------
Oj t	Aggregate jelly capital investment

ci-jj t	Region-, sector- and impact-specific averting expenditure

vjlt	Region-, sector- and impact-specific adaptation investment

Economic state variables:

W	Welfare (model objective)

q\11	Net sectoral product

Ojt	Aggregate net regional product

Of	Aggregate regional energy use

PtE	Global marginal energy resource extraction cost

Of	Stock of aggregate jelly capital

xjj t	Stock of region-, sector- and impact-specific adaptation capital

Environmental state variables:

Gt	Global stock of atmospheric GHGs

M"\	Region-specific meteorological variables

z-jj t	Region-, sector-, and impact-specific endpoint indexes

A ., t	Region- and sector-specific damage induced productivity losses

Functional relationships:

s

Global intertemporal welfare

U,

Regional intratemporal utility



Regional aggregate production functions



Sectoral production functions

©

Global energy supply function

s

Global atmospheric GHG accumulation

ym

11

Regional climate response functions

rit

Regional and sectoral climate impacts functions



Regional and sectoral damage functions

1. Economic Sub-Model
Objective:

T

max W — y^B'3
1= 0

Aggregate net regional product:

lh

Qv

.,u>

\<&.t

Qt,t —

Y	Y

-------
(lb)

Sectoral net regional product = Climate loss factor x Sectoral gross regional product, produced from
energy and capital:

lh,i '

Intraregional and intratemporal market clearance for energy:

jv

E 'ih< = Qtt

/=1

Intraregional and intratemporal market clearance for jelly capital:

E = Qf,t

& c/K

Aggregate regional absorption constraint:

fit, = Ql, -	-EE ("L+4u)

f=\j=\

Global energy trade and marginal resource extraction cost:

.Sf f

E E

if = 0

£=1 5=0

Regional jelly capital accumulation:

<&+i = Q'u + (i - «k)Qo

Accumulation of impact-, sector- and region-specific adaptation capital:

xj,e,t+1 = + '¦ 1 — )xj,ej

2. Climate Sub-Model

Global atmospheric GHG accumulation:

c,+i = £

Ee?,.'G<

Regional meteorological effects of global atmospheric GHG concentration:

Ml = Yf[Gt]

3. Impacts Sub-Model

Physical climate impacts by type, sector and region:

,/

rf

j,e,t ^j,e

M\,o			

Climate damages:

Aj,U ~ Aj,t

1	JP 1	& 1	&

7	7 • fl	fl ' V4	V

''''	' '''	' jJJ

(lc)

(Id)

(le)

(If)

(lg)

(lh)

(li)

(2a)

(2b)

(3a)

(3b)

A-49

-------
From the point of view of period t*, the condition for optimal extraction of carbon-energy is:

<>Qfv.

II. Current marginal
extraction cost

I, Current marginal benefit

1=1*

dUe* dQ^ ^

10. Resource stock effect of contemporaneous energy use

IV. Present value of future marginal climate damage (N.B. dqY/dA < 0 in general)

= 0

(4)

The "social cost of carbon" in this expression is given by the combination of terms (II) + (III) - (IV).
Our interest is in (IV), the marginal external cost of carbon-energy consumption, which, because it
emanates from a globally well-mixed pollutant, is independent of the location in which the energy is
consumed.

It is now clear to see how fundamental gaps in our understanding the render the "land of cockaigne"
unattainable. The difficulty in computing the social cost of carbon stems from the terms in curly
braces. Carbon-cycle modeling is sufficiently advanced to enable us to predict with a fair degree of
confidence the effect of the marginal ton of carbon on the time-path of future atmospheric GHGs
(ds/dOE). Likewise, the IPCC AR4 notes global climate models' substantially improved ability to
capture the future trajectory of consequent changes in temperature, precipitation, ice/snow cover
and sea levels at regional scales (dY"' / dG). But the weak links in the causal chain between climate

change and economic damages continue to be the cardinality and magnitude of the vectors of
physical impact endpoints as a function of climatic variables in each region out into the future
(dg-jj / dM"'), and—to a lesser extent—the manner in which these endpoints translate into

shocks to the productivity of economic sectors (dA -, / dzjt).

A Critical Review of the State of Modeling Practice

To put the key issues in sharp relief, it is useful to consider how implementing the disaggregated 1AM
might improve upon the current state of integrated assessment practice. RICE-type lAMs represent
the productivity losses incurred by climate change impacts through variants of Nordhaus' aggregate
damage function, which specifies the reduction in gross regional product as a function of global
mean temperature. This approach effectively collapses M"' to a scalar quantity in each time period.

Moreover, as reviewed by NRC (2010), it then benchmarks the magnitude of various impacts and the
associated economic losses for a reference level of global mean temperature change, before making

A-50

-------
assumptions about how these costs are likely to scale with income, and finally expressing damage as
a temperature-dependent fraction of regions' gross output. Therefore, the details of climatic
variables' influence on impact endpoints in (3a), and of the latter's effects on economic sectors in
(3b), only affect the calibration of the damage function. From that point on they are entirely
subsumed within the function's elasticity with respect to global temperature change, and, in RICE-
2010, sea level rise. The damage function therefore collapses (3a) into (3b), dealing only with
changes in aggregate global climatic variables, skipping over impacts as state variables and implicitly
aggregating over sectors to express damages purely on an aggregate regional basis.

A similar situation obtains with adaptation. A case in point is the AD-WITCH model, a variant of
Nordhaus' RICE simulation which modifies the damage function by introducing stock and flow
adaptation expenditures which attenuate aggregate regional productivity losses due to climate
change. Formally, using eQY to denote gross regional product, net regional product is given by

y	1 + ADAPTff	~y

Qe'f ~ 1 + ADAPTe,, + CCD a f''	(5)

where CCD is the regional climate damage function and ADAPT is an index of adaptation's
effectiveness. The variable ADAPT is the output of a nested constant elasticity of substitution (CES)
production function which combines inputs of contemporaneous averting expenditures with
adaptation capital and adaptation knowledge according to Figure 2. The key consequence is that
adaptation is able to directly influence the dynamic path of the economy, instead of being implicit in
the curvature of the damage function, as with the RICE model. However, eq. (5)'s assumption that
the effects of ADAPT and CCD are multiplicative seems very strong in light of the fact that the
damage function already explicitly incorporates the influence of adaptation through the studies on
which it is benchmarked—but only at the calibration point, not over the full range of its curvature. A
prime example is Nordhaus and Boyer's (2000) use of Yohe and Schlesinger's (1998) results on the
impact of sea level rise, which optimally balance the costs of abandonment and coastal defenses.
The implication is that because defensive expenditures are likely to be closely associated with the
magnitudes of climate impacts of various kinds within individual sectors, one should not think of
aggregate adaptation expenditure as independent of future changes in the sectoral composition of
output.

Figure 2: The AD-WITCH Adaptation Production Function (Bosello, Carraro and De Cian, 2010)

A-51

-------
By dispensing with the aggregate damage function, our land of cockaigne 1AM explicitly captures the
dynamic evolution of impact endpoints' response to changes in climatic variables, the magnitude
and intersectoral distribution of the follow-on productivity effects, and the optimal intersectoral
adjustments these induce, all at regional scales. An adaptation response may therefore be modeled
more precisely as averting expenditure that mitigates the sectoral and regional productivity loss
associated with a particular category of climate impact. In other words, stock and flow adaptation
reduces the impact elasticity of sectoral productivity shocks. Of course, the problem that besets this
approach is that, except for a very few combinations of impacts, sectors and regions, the relevant
elasticities are unknown.

But the good news is that this is one area in which research is proceeding apace. There are a growing
number of CGE modeling studies of climate impacts (e.g., ICES) which elucidate the magnitude of
both sectoral and regional damages and producers' and consumers' adjustment responses. The
focus of such studies is typically a single impact category (say, f*), whose initial economic effects are
computed using natural science or engineering modeling or statistical analyses. The results are often
expressed as a vector of shocks to exposed sectors and regions, which are then imposed as
exogenous productivity declines on the CGE models' cost functions. In the context of the 1AM in
section 2, this procedure is equivalent to first specifying an exogenous ex-ante effect of a particular
impact cU.; / dz j*, before using the CGE model to compute the ex-post web of intersectoral

adjustments and the consequences for sectoral output, and regions' aggregate net product and
welfare:

This line of inquiry has the potential to yield two critical insights. The first is quantification of the
elasticity of the economy's response to variations in the magnitude and interregional/ intersectoral
distribution of particular types of impact, which has been the type of investigation pursued thus far.
But second—and arguably more important—is comparative analysis of economic responses across
different impact categories for the purpose of establishing their relative overall economic effect,
conditional on our limited knowledge of their relative likelihood of occurrence, and intensity. The
results could at the very least guide the allocation of effort in investigating the thorny question of

how different impacts are likely to respond to climatic forcings at the regional scale, dg-jj / dM"'.

A-52

-------
Adaptation and Technological Change

Karen Fisher-Vanden
Elisa Lanzi
David Popp
Ian Sue Wing
Mort Webster

The purpose of this talk is to provide a brief summary of the state of the science on the influences of
adaptation on the social cost of climate change. Specifically, the charge was to discuss (not
necessarily in this order):

(1)	relevant studies on the observed or potential effectiveness of adaptive measures, and on
private behaviors and public projects regarding adaptation;

(2)	relevant studies on how to forecast adaptive capacity;

(3)	how adaptation and technical change could be represented in an 1AM (for at least one
illustrative sector);

(4)	whether the information reguired to calibrate such a model is currently available, and, if not,
what new research is needed; and

(5)	how well or poorly existing lAMs incorporate the existing body of evidence on adaptation.

A tall order, but important to get our arms around since estimates of the net impact of climate
change could be significantly higher if adaptation is not taken into account.1

As elaborated below, a number of general insights have resulted from our brief foray into this topic
that have implications for the development of a future research program in this area. First, modeling
adaptation is inherently difficult given the nature of the adaptation process, requiring advancements
in modeling techniques. Second, although there has been good empirical work done on impacts and
adaptation costs, the coverage is limited requiring heroic efforts to translate the results into model
parameters. More work is needed to bridge the gap between models and empirical studies. Lastly,
adaptation-related technological change is generally lacking in current models but could significant
lower adaptation cost estimates. This stems from a general lack of understanding of the process
related to this type of technological change. More empirical work is needed in this area.

What is unique about the adaptation process that justifies the need to add features to existing
integrated assessment models (lAMs)? First, adaptation is in response to current or anticipated
impacts and comes in different forms: (a) reactive (e.g., changes in heating/cooling expenditures;
treatment of disease; shifts in production); and (b) proactive (e.g., infrastructure construction (e.g.,
seawalls); early warning systems; water supply protection investments. In some lAMs adaptation

1 For the U.S., Mendelsohn et al. (1994) estimates that the net impact of climate change on the farming sector
will be 70% less if adaptation is included while Yohe et al. (1996) estimates that the net impact on coasts will
be approximately 90% less (Mendelsohn (2000)).

A-53

-------
would occur endogenously in reaction to changes in prices due to climate impacts—e.g., more
power plants built to deal with increases in demand for air conditioning; shifts in production in
reaction to higher prices of factors negatively impacted by climate change. However, many
adaptation activities that would occur in reality, such as investment in flood protection, would not
occur in a simulated model unless there is explicit representation of climate damages to induce
reactive expenditures and proactive investments.

Second, unlike mitigation investments where investments today result in reductions today, proactive
adaptation investments are made today to provide protection against possible future impacts. Thus,
adaptation investment decisions are inherently intertemporal and therefore 2

models need to include intertemporal decision making for proactive adaptation investments, in
order to trade off future damages and current adaptation investment expenditures. Not only are we
making intertemporal adaptation decisions, we are specifically making proactive adaptation
investments under uncertainty. Whether we invest and how much to invest all depends on our
expectations regarding future impacts and how we value the future. Therefore, we need a model
that allows for intertemporal decision-making under uncertainty.

Climate damages and adaptation strategies are locally- or regionally-based. Therefore, ideally the
model will include regional detail or will apply a method to aggregate up to a more coarse regional
representation. Climate damages and adaptation expenditures are also sector specific—e.g., certain
sectors will be impacted more than others and adaptation expenditures will be directed at specific
sectors (e.g., electric power, construction). Thus, a model with sectoral detail or a way to aggregate
these sector-specific impacts and expenditures is desirable.

The demand for adaptation solutions will induce adaptation-related technological change. Do
inducements for adaptation-related technological change differ markedly from mitigation-related
technological change, requiring a different modeling approach? To the extent that adaptation
activities may be region or sector specific, markets for new adaptation techniques will be smaller
than for new mitigation techniques, making private sector R&D investments less attractive. Given
this, as well as the case that adaptation investments are largely public infrastructure investments,
distinguishing between public R&D and private R&D may be important. Note that this is more than a
question of simply basic versus applied science, but driven by the nature of demand for the final
product, much in the same way that the government finances most R&D for national defense. Thus,
the model needs to be capable of distinguishing between private and public investments and include
mechanisms of public revenue raising to fund these projects.

To summarize, to be able to capture adaptation strategies, an ideal 1AM would include the following
features:

•	Explicit modeling of climate damages/impacts

•	Intertemporal decision making under uncertainty

•	Endogenous technological change

•	Regional and sectoral detail for impacts and adaptation strategies

A-54

-------
• Connection with empirical work on impacts and adaptation

Is it feasible or even desirable to have all of these features represented in a single model, since
transparency is lost as more features are added? It is important to measure the trade-offs:

• How much of this needs to be specifically represented in the model and how could be
represented outside of the model

• To cite Jake Jacoby: —different horses for different courses. || Do we need a suite of models
each designed to capture a subset of these features?

• How important is each of these features to the social cost of climate change? Sensitivity
analysis could be useful here to assess whether we even need to worry about including
certain features.

To answer these questions, it is useful to first survey what features currently exist in lAMs. A number
of modeling approaches have been taken to capture impacts and adaptation. Computable general
equilibrium (CGE) models have the advantage of providing sectoral and regional detail and capturing
the indirect effects of impacts and adaptation. Thus, given its structure, CGE models can more easily
accommodate regional and sectoral-specific damage functions. Most CGE models, however, do not
include the type of intertemporal decision making required to model proactive adaptation
investment decisions, given the computational demands required by a model with detailed regions
and sectors. However, there have been a number of CGE models that have been used to estimate
the cost of climate change impacts; for example,

• DART (Deke et al, 2001)—to study the cost of coastal protection

• FARM (Darwin and Tol, 2001; Darwin et al, 1995)—includes detailed land types to study the
effects of sea level rise and impacts of climate change on agriculture.

• GTAP-E/GTAP-EF (Bosello et al, 2006; Bigano et al, 2008; Rosen, 2003)—has been used to
study induced demand for coastal protection; effects of rising temperatures on energy
demand (Bosello et al, 2007); health effects of climate change (Bosello et al, 2006); effects of
climate change on tourism. Focuses on one impact at a time.

• Hamburg Tourism Model (HTM) (Berittella et al, 2006; Bigano et al, 2008)—used to study
the effect of climate change on tourism.

• ICES (Eboli et al, 2010)—models multiple impacts simultaneously: impacts on agriculture,
energy demand, human health, tourism, and sea level rise.

Another set of models used to study climate change impacts and adaptation fall under the category
of optimal growth models. These models include intertemporal optimization but typically lack
sectoral and regional detail given the computational demands this would require. These include:

• DICE/RICE (Nordhaus, 1994; Nordhaus and Yang, 1996; Nordhaus and Boyer, 2000)—DICE
comprises one region, one aggregate economy, and one damage function aggregating many
impacts. RICE comprises 13 regions, each with its own production function and damage
function.

A-55

-------
• AD-DICE/AD-RICE (de Bruin et al, 2009)—DICE/RICE model with adaptation. Adaptation
investment added as a decision variable which lowers damages and faces an adaptation cost
curve. Residual damages are separated from protection costs in the damage function.

There are also a number of simulation models that have been developed to study the effects of
climate change impacts. The major difference from CGE and optimal growth models is that
simulation models do not optimize an objective function, such as intertemporal utility. Instead,
these models represent a number of interconnected relationships that allow for studying the
propagation of perturbations to the system. Two widely used simulation models are:

• PAGE (Plambeck and Hope, 1997; Hope, 2006)—PAGE comprises eight regions each with its
own damage functions for two impact sectors (economic and non-economic). The authors
use information on impacts from IPCC (2001) to generate model parameter values related to
impacts. In addition, PAGE stochastically models catastrophic events where the probability
of an event increases when temperature exceeds a certain threshold. Simple adaptation is
included in the model which reduces damages. Assumes developed countries can reduce up
to 90% of economic impacts while developing can reduce up to 50%. All regions can reduce
up to 25% of non-economic impacts.

• FUND (Tol et al, 1995; Tol, 1995)—referred to as a —policy optimization|| model. Exogenous
variables include population (from the World Bank), GDP per capita (from EMF 14), and
energy use. Endogenous variables include atmospheric concentrations, radiative forcing,
climate impacts (species loss, agriculture, coastal protection, life loss, tropical cyclones,
immigration, emigration, wetland, dryland), emission reductions (energy or carbon efficiency
improvements, forestry measures, lower economic output), ancillary benefits (e.g.,
improved air quality), and afforestation. The model comprises 9 regions with game
theoretics and eight market and non-market sectors, each with its own calibrated damage
function. Adaptation is modeled explicitly in the agricultural and coastal sectors, and
implicitly in other sectors such as energy and human health where the wealthy are assumed
to be less vulnerable to the impacts of climate change. No optimization in the base case-
just simulation. In the optimization case, the model is choosing the optimal level of
emissions reductions by trading off costs and benefits of reductions.

Another class of models involves hybrid combinations of the above model types. For example,

• Bosello and Zhang (2006) couple an optimal growth model with the GTAP-E model of
Burniaux and Truong (2002) to study the effects of climate change on agriculture

• Bosello et al (2010) couple the ICES CGE model with an optimal growth model (AD-WITCH) to
study adaptation to climate change impacts.

• AD-WITCH (Bosello et al, 2010)—an optimal growth model with detailed bottom-up
representation of the energy sector. Comprises 12 regions where the following seven
control variables exist for each region: investment in physical capital, investment in R&D,
investment in energy technologies, consumption of fossil fuels, investment in proactive
adaptation, investment in adaptation knowledge; and reactive adaptation expenditure.

These alternative uses of regional income compete with each other.

A-56

-------
To parameterize these models, most modeling teams look to empirical studies of impacts and
adaptation and are faced with similar frustrations. First, as elaborated in Agrawala and Fankhauser
(2008), the empirical work in the area of adaptation is severely lacking. The authors find that
although information exists on adaptation costs at the sector level, certain sectors (e.g., coastal
zones and agriculture) are studied more heavily than others. Second, most empirical studies are not
done with modeling applications in mind. Most modelers find themselves forced to devise methods
to scale up from the regional and sectoral results generated by empirical studies.

There have been a few recent studies that have attempted to summarize the empirical work on
adaptation costs; e.g.,

• Agrawala and Fankhauser (2008)—provides a critical analysis of empirical work on
adaptation costs. Tables summarize empirical sectoral studies on adaptation costs. Sectors
include coastal zones, agriculture, water resources, energy demand, infrastructure, tourism
and public health.

• World Bank (2010)—report from The Economics of Adaptation to Climate Change (EACC)
study. Seven sector-specific studies: infrastructure, coastal zones, water supply and flood
protection, agriculture, fisheries, human health, extreme weather events. Provides detailed
estimates of adaptation costs; some generated using dose response functions with
engineering estimates and some generated from sector-specific models.

• UNFCCC (2007)—regional studies (Africa, Asia, Latin America, and small island developing
States) on vulnerability; current adaptation plans/strategies; future adaptation
plans/strategies. Most information from national communications to the UNFCCC, regional
workshops, and expert meetings.

A few modeling teams have made serious attempts to integrate existing empirical work on
adaptation into their model; for example,

• AD-DICE/AD-RICE: starts with damage functions of Nordhaus and Boyer (2000) and uses
empirical studies to separate residual damages from adaptation costs. Various studies on
adaptation measures for certain sectors (i.e., agriculture and health) and estimates of
adaptation costs from existing studies are used. Also, other model results—e.g., results from
FUND—are used to estimate adaptation costs in response to sea level rise. Empirical studies
to separate residual damages from adaptation costs are not available for many of the
sectors—i.e., other vulnerable markets; non-market time use; catastrophic risks;
settlements—so assumptions were made in order to separate the damage costs. However,
these sectoral estimates are ultimately aggregated up to one damage cost number and one
adaptation cost number to fit with the one sector structure of the model.

• AD-WITCH: Uses empirical information from the construction of damage functions in
Nordhaus and Boyer (2000), the studies in Agrawala and Fankhauser (2008); and UNFCCC
(2007) to separate residual damages from adaptation costs. Similar to AD-DICE, using these
empirical studies to separate the damage estimates in Nordhaus and Boyer (2000) into
residual damages and adaptation costs.

A-57

-------
Comparing this brief survey of existing work in this area with the list of required modeling features
needed to model adaptation, a couple of key research voids stand out. First, none of these models
include decision making under uncertainty, and for good reason. It is difficult to do. Optimal growth
models like DICE with intertemporal decision making are deterministic and fully forward-looking.

Past approaches to modify such a model to be stochastic usually entail the following steps:

1) Create multiple States of the World (SOWs), each with different parameter assumptions and
different probabilities of occurrence;

2) Index all variables and equations in the model by SOW;

3) Add constraints to the decision variables so that for all time periods before information is
revealed, decisions must be equal across SOWs.

The problem with this approach is that it rapidly becomes a very large constrained nonlinear
programming problem, and often the model will not converge to a solution for more than a trivial
number of SOWs. The general problem of decision making under uncertainty is a stochastic dynamic
programming problem that requires the exploration of a large number of samples of outcomes in
every time period. The challenge is to fully explore the sample space while keeping the model
computationally tractable. Promising on-going research by Mort Webster and his team at MIT could
offer an alternative approach to modeling decision making under uncertainty. Webster's NSF-funded
project team is currently developing a formulation based a new approach called Approximate
Dynamic Programming, introduced by Powell (2007) and others. This approach implements dynamic
programming models by iteratively sampling the state space using Monte Carlo techniques,
approximating the value function from those samples, and using approximate value functions to
solve for an approximate optimal policy, then repeating. This approach has been used successfully in
other contexts for very large state spaces. Mort Webster's team is currently developing an ADP
version of the ENTICE-BR model to study R&D decision making under uncertainty.

Second, adaptation-related technological change is largely absent in current models. Most models
are calibrated using existing knowledge of adaptation strategies and costs with no allowance for
improvements in these strategies and technologies. AD-WITCH (Bosello et al, 2009) does attempt to
account for this by including investment in adaptation knowledge as a decision variable that
competes with other types of investment. Investments in adaptation knowledge accumulate as a
stock which reduces the negative impact of climate change on gross output. However, the lack of
empirical studies on adaptation-related technological change limits the modelers' ability to calibrate
their model based on empirical knowledge. In the case of AD-WITCH, adaptation knowledge
investments only relate to R&D expenditures in the health care sector where empirical data exist.
This suggests that more empirical research in this area is desperately needed.

A-58

-------
technologies as discussed in the previous paragraph, it also captures expenditures to improve the
region's ability to adapt to climate change. Issues arise, however, when the model is calibrated since
the modelers were only able to identify one source of qualitative information on adaptive capacity
(i.e., the UNFCCC (2007) report discussed above) which only covers four aggregate regions (Africa,
Asia, small island developing States, and Latin America). Assumptions were then made to translate
this information to the regional representation and model parameters in AD-WITCH.

Lastly, another area where empirical work to inform models is lacking is in the dynamics of recovery
from climate change impacts. Most models represent climate damages as a reduction in economic
output which is assumed to recover over time. Empirical work on thresholds and time to recover
including factors that influence these variables could help inform models on the type of dynamics
that should be captured in impact and adaptation analyses. Also, better techniques to translate
results from empirical studies to models are needed since the sectoral and regional detail of
empirical studies does not typically align with the sectoral and regional detail in models. In general,
to address the disconnect between empirical studies and modeling needs, we as a research
community need to devise better ways to facilitate communication between empirical researchers
and modelers.

References

Agrawala, S. and S. Fankhauser (2008), Economic Aspects of Adaptation to Climate Change, OECD,
Paris, France.

Berrittella, M., Bigano, A., Roson, R. and Tol, R.S.J. (2006), A general equilibrium analysis of climate
change impacts on tourism, Tourism Management, 25(5), 913-924.

Bigano, A., F. Bosello, R. Roson and R. Tol (2008). Economy-wide impacts of climate change: a joint
analysis for sea level rise and tourism, Mitigation and Adaptation Strategies for Global
Change, Springer, vol. 13(8), pages 765-791.

Bosello, F., C. Carraro and E. De Cian (2009). An Analysis of Adaptation as a Response to Climate
Change, Copenhagen Consensus Center, Frederiksberg, Denmark.

Bosello, F., C. Carraro and E. De Cian (2010). Climate Policy and the Optimal Balance between
Mitigation, Adaptation and Unavoided Damage, FEEM Working Paper No. 32.2010.

Bosello, F., De Cian, E. and Roson, R. (2007), Climate Change, Energy Demand and Market Power in a
General Equilibrium Model of the World Economy, FEEM working paper n. 71.07.

Bosello, F., Roson, R. and Tol, R.S.J. (2006), —Economy wide estimates of the implications of climate
change: human health||, Ecological Economics, 58, 579-591.

Bosello, F. and Zhang J. (2006), Gli effetti del cambiamento climatico in agricoltura, Questione
Agraria, 1-2006, 97-124.

Burniaux, J-M. and T. Truong (2002). GTAP-E: An Energy-Environmental Version of the GTAP Model,
GTAP Technical Papers 923, Center for Global Trade Analysis, Department of Agricultural
Economics, Purdue University.

A-59

-------
Darwin, R., M. Tsigas, J. Lewabdrowski, and A. Raneses (1995). World Agriculture and Climate

Change. Agricultural Economic Report No. 703, US Department of Agriculture, Economic
Research Service, Washington, DC.

Darwin, R. F. and R. S. J. Tol (2001), Estimates of the Economic Effects of Sea Level Rise,
Environmental and Resource Economics 19,113-129. 8

De Bruin, K.C., R.B. Dellink and R.S.J. Tol (2009). AD-DICE: an Implementation of Adaptation in the
DICE Model, Climatic Change, 95: 63-81.

Deke, O., K. G. Hooss, C. Kasten, G. Klepper and K. Springer (2001), Economic Impact of Climate
Change: Simulations with a Regionalized Climate-Economy Model. Kiel Institute of World
Economics, Kiel, 1065.

Eboli, F., R. Parrado and R. Roson (2010), Climate-change feedback on economic growth:

explorations with a dynamic general equilibrium model, Environment and Development
Economics, 15:515-533.

Hope, C. (2006). The Marginal Impact of C02 from PAGE2002: An Integrated Assessment Model
Incorporating the IPCC's Five Reasons for Concern. Integrated Assessment, 6:19-56.

IPCC (2001), Impacts, adaptation, and vulnerability, Contribution of working

group II to the third assessment report, Cambridge University Press.

Mendelsohn, R, Nordhaus, W, and Shaw, D. (1994). "The Impact of Global Warming on Agriculture: A
Ricardian Analysis", American Economic Review, 84: 753-771.

Mendelsohn, R. (2000). —Efficient Adaptation to Climate Change,|| Climatic Change, 45: 583-600.

Nordhaus, W.D. (1994). Managing the Global Commons: The Economics of the Greenhouse Effect.
MIT Press, Cambridge, MA.

Nordhaus, W.D. and Z. Yang (1996). A Regional Dynamic General-Equilibrium Model of Alternative
Climate-Change Strategies, American Economic Review, 86(4), 741-765.

Nordhaus, W.D., and Boyer, J (2000). Warming the World: Economic Models of Global Warming. MIT
Press, Cambridge, MA.

Plambeck, E.L., C. Hope, and Anderson, J. (2007). —The Page95 Model: Integrating the Science and
Economics of Global Warming, || Energy Economics, 19:77-101.

Powell, W.B, (2007), Approximate Dynamic Programming: Solving the Curses of Dimensionality,
Wiley-lnterscience, Hoboken, New Jersey.

Roson, R., (2003), Modelling the Economic Impact of Climate Change, EEE Programme Working
Papers Series, International Centre for Theoretical Physics —Abdus Salam||, Trieste, Italy.

Tol, R.S.J. (1995). The Damage Costs of Climate Change Toward more Comprehensive Calculations,
Environmental and Resource Economics, 5: 353-374. 9

A-60

-------
Tol, R.S.J., T. Van der Burg, H.M.A. Jansen and H. Verbruggen (1995). The Climate Fund-Some
Notions on the Socio-Economic Impacts of Greenhouse Gas Emissions and Emission
Reduction in an International Context (Institute for Environmental Studies, Vrije Universiteit,
Amsterdam).

UNFCCC (2007), Climate Change: Impacts, Vulnerabilities, and Adaptation in Developing Countries,
UNFCCC, Bonn, Germany.

World Bank (2010), The Costs to Developing Countries of Adapt to Climate Change, The Global

Report of the Economics of Adaptation to Climate Change Study, World Bank, Washington,
DC.

Yohe, G., Neumann, J., Marshall, P., and Ameden, H. (1996). _The Economic Cost of Greenhouse-
Induced Sea-Level Rise for Developed Property in the United States', Climatic Change 32:
387-410.

A-61

-------
Knowability and no ability in climate projections

Gerard Roe

Department of Earth and Space Sciences, University of Washington, Seattle, WA
Introduction

The purpose of this note is to provide a referenced summary of the present scientific understanding
about future climate change, tailored towards the kind of global climate factors that are captured in
Integrated Assessment Models (lAMs). In outline, it is organized as follows:

i) Equilibrium climate sensitivity is the long-term response of global temperature to a doubling

of atmospheric CO2. I review the causes of our current uncertainty, and the prospects
for reducing it.

ii) Two other measures of climate change are arguably more important in this context. First the

climate commitment is a measure of the climate change we already face because of
emissions that have already occurred.

iii) The very long timescales associated with attaining equilibrium, especially at the high end of

possible climate sensitivity, mean that the transient climate response is of greater
relevance for climate projections over the next several centuries.

iv) Due to the inherent uncertainties in the climate system, a flexible emissions strategy is far

more effective in avoiding a given level of global temperature change, than a strategy
aims to stabilize CO2 at a particular level.

v) Many important climate impacts are fundamentally regional in nature. Among climate

models, regional climate projections correlate only partially with global climate
projections.

This was prepared for the EPA Climate Damages Workshop, Washington, D.C., Nov 18-19, 2010.
Climate sensitivity

Climate sensitivity (here given the symbol T2x, and sometimes called the equilibrium climate
sensitivity) is the long-term change of annual-mean, globalmean, near-surface air temperature in
response to a doubling of carbon dioxide above preindustrial values. It has long been a metric by
which to compare different estimates of the climate response to greenhouse gas forcing (e.g.,
Charney, 1979). There is a vast literature that has researched climate sensitivity from every possible
angle, ranging from state-of-the-art satellite observations of Earth's energy budget, to geological
studies covering hundreds of millions of years. A fine review of where things stand can be found in
Knutti and Hegerl (2008).

Figure 1 shows a variety of probability distributions (pdfs) of climate sensitivity. A prominent feature
of such estimates is that they all exhibit considerable skewness. In other words, while the lower
bound is confidently known, the upper bound is much more poorly constrained. There is a small but
nontrivial possibility (about 25 %) that the climate sensitivity could exceed 4.5 oC. One concern that
has been raised is that the current generation of IPCC climate models (from the fourth assessment,
or AR4) does not span the range of climate sensitivity that is allowable by observations (the blue

A-62

-------
histogram in figure 1 clusters too narrowly
around the modes of the other pdfs). The
reason for this appears to be that the IPCC
climate models do not sample the full range
of possible aerosol forcing (Armor and Roe,
2010). This should not be surprising since
they are designed to represent the "best"
estimate of climate (something akin to the
mode of the distribution). However, since
these computer models are the only tools
available for modeling regional climates, it
should perhaps be a concern that they are
under sampling the range of possible
futures. I next outline briefly how estimates
are made from observations and models.
The purpose of doing so is to
straightforwardly demonstrate the
important sources of uncertainty.

05

	¦—

0.45



„ 04



U 0.35

1

— 0.3
| 0.25

>*

HI

I o*

| 0.15

If

cl.

111

0.1



0.05
O

J M

M

	ciimateprediction. net

	FeedbacKs (RB07)

	Modern obs. (AR10)

¦Glacial (H84)

I IPCC A R4 (2007)

~	2	4	G

Climate sensitivity < 0C)

Figure 1. Various estimates of climate sensitivity. In
order of the legend: i) from multi-thousand ensembles
from one climate model (Stainforth et al., 20Q5), ii) from
feedbacks with climate models (Roe and Baker, 2007),

iii)	from modern observations (Armour and Roe, 2010),

iv)	from glacial climates (Hansen et al., 1984), v) A
histogram of Ta from 19 main IPCC AR4 models
(IPCC, 2007).

Estimates of climate sensitivity from observations.
A linear approximation of the Earth's energy budget is:

H + A_1T,

(1)

where R is the radiative forcing (units W m"2), H is the heat going into the world's oceans and being
stored there, and A1!" is the climate response in terms of the global-mean, annual-mean, near-
surface air temperature T, and the climate sensitivity parameter, X. (e.g., Roe, 2009, Armour and
Roe, 2010, and many others). For silly historical reasons the terminology here can be confusing. X is
a more fundamental measure of climate system than T2x, since it does not depend on any particular
forcing. A and T2xare related in the following way. Let R2xbe the radiative forcing due to a doubling of
C02 over pre-industrial values (= 4 W m-2). In the long-term equilibrium, ocean heat uptake goes to
zero, and so the climate sensitivity is just:

T2x = XR

2x

(2)

The point of this algebra is to make it clear that the goal of estimating climate sensitivity from
observations is the goal of estimating X from Equation (1):

A = -

T

R-H

(3)

We have observations of T, R, and H, whose probability distributions are shown in figure 2. Hereafter
we refer to R-H as the climate forcing, since it is the net energy imbalance that the atmosphere must
deal with. H and Tare actually quite well constrained, as is the radiative forcing associated with C02
and other greenhouse gases. As is clear from figure, the major source of uncertainty is R and, in
particular, the component of R that is due to aerosols (small airborne particulates that can be either
liquid or solid).

A-63

-------
The reason that aerosol forcing is hard to
constrain is that 1) the spatial pattern and
lifetime is extremely complicated to observe
(they are primarily in the Northern
Hemisphere and downwind of major
industrial economies); 2) some aerosols
have a cooling effect, some have a warming
effect; 3) aerosols alter the thickness,
lifetime, and height of clouds - a powerful
indirect effect that is hard to measure and
attribute properly. The community is
confident, however, that the net aerosol
effect is almost certainly negative. More
information about aerosol uncertainties can
be found in Menon (2004).

4 p

3.5-

3 -

2.5-

2 -

1.5-

U-

1 -

0.5-

(fL

-3

..QCfflfl.

Heatj
Uptakie

Aerosols
" Direct
Indirect,

C02
atone

Nori CO'j
Green house
GaSSS'l

Total
climate,
forcing

"""? Total"
dimate
¦ - -forcing—

: w/o
...^ejasols...

t
t

*
*

|-r

-2

-1 0 1

Climate forcing (WnT2)

Figure 2: Probability distributions of the terms in the
Earth's energy budget, based or IPCC 2007, and
updated for newer ocean heat uptake observations.
See Armour and Roe, 2010 for details. Total climate
forcing is equal to R+I in Eq. 3. Also shown is the
total forcing excluding aerosols, which is the climate
forcing experienced by the Earth, if all anthropogenic
emissions ceased immediately.

Thus, from Eqs. 2 and 3, the probability
distribution of climate sensitivity comes
from combining a relatively narrow
distribution (the well-known temperature

change) in the numerator with a relatively broad distribution (the much less wellknown climate
forcing (i.e., R-H)) in the denominator of Eq. 3. It is this combination that produces the skewed
distribution seen in figures 1 and 3c. The graphs in figure 3 are the fundamental reason why we can
say with great confidence that it is very likely that observed forcing has not been large enough to
imply a climate sensitivity of less than about 1.5 oC. On the other hand, uncertainties in observed
forcing also mean that we cannot confidently rule out the disconcerting possibility that the modern
warming has occurred with small climate forcing, which would imply very high climate sensitivity.
Note that the curves in figure 1 and 3 are consistent with the probabilities given in the 2007 IPCC
report.

Temperature Change t'C)

CJimats forcing (Wnra)

CSrrata Sensitivity fC)

Figure 3: The calculation of climate sensitivity from observations involves combing a relatively
narrow probability distribution of T (panel a) in the numerator, with a relatively broad
distribution of F= H-R (panel b) in the denominator of Eq. (3). This leads to the skewed
distribution of climate sensitivity (panel c). Note tie pdfs must be combined properly - it is not
just a simple division - but the point is hopefully clear.

Estimates of climate sensitivity from models.

A-64

-------
instead are designed with the most likely combination of model parameters, and parameters are
'tuned' to reproduce observed climate history. Clear evidence of that tuning comes from the
correlation of climate sensitivity and imposed aerosol forcing in the models in such a direction that
twentieth century observations tend to be reproduced (Kiehl, 2007, Knutti, 2008). Such tuning is not
problematic if models are interpreted as reflecting combinations of climate sensitivity and aerosol
forcing that are consistent with observed constraints (Knutti, 2008). However AR4 models do not
fully span the range of aerosol forcing allowed by observations (Kiehl, 2007; IPCC, 2007). This is the
likely reason that the AR4 models under sample of the full range of possible climate sensitivity, as
seen in figure 1.

Roe and Baker (2007) explain this in terms of a classic
feedback analysis, summarized in figure 4. The
relationship between feedbacks and response also
produces a skewed distribution because of the way that
positive feedbacks have a compounding effect on each
other (e.g., Roe, 2009). The range of feedbacks as
diagnosed within the AR4 models produces a pdf of
climate sensitivity that is quite consistent with the pdf
estimated from observations (figure 1). This should be
expected since it is observations that ultimately provide
constraints on the models.

Prospects for improved estimates of climate sensitivity.

Can a narrower range of climate sensitivity be expected soon? One can ask: how might more
accurate observations or better climate models change the estimate of T2x?

Reducing uncertainty in either forcing or feedbacks would produce a narrower range. However it is
the nature of these skewed distributions that the mode of T2x moves to higher values as the range of
forcing or feedbacks is narrowed, leaving the cumulative probability of T2x > 4.5°C stubbornly
persistent (Allen et al., 2007; Roe and Baker, 2007; Baker et al., 2010).

It should also be made clear that there are formidable scientific challenges in reducing uncertainty in
climate model feedbacks, or in observing the aerosol forcing better. Progress will occur, but it is
likely that it will be incremental. Another line of attack is to try to combine multiple estimates of
climate sensitivity in a Bayesian approach that might, in principal, significantly slim the fat tail of T2x
(e.g., Annan and Hargreaves, 2006). However, as with all Bayesian estimates, the value of the
analysis is critically sensitive to 1) the independence of different observations; and 2) structural
uncertainties within and among very complex models (e.g., Henriksson et al., 2010; Knutti et al.,
2010). An objective assessment of these factors has proven elusive, rendering the information
obtained by the exercise hard to interpret, and there is an acute risk that it produces overconfident
estimates.

A-65

£
K *

i AjrW /

L/ /

Figure 4: Model feedbacks arid climate
sensitivity. The black curve shows the
mapping between climate feedbacks (x-
axis, greeri curve), and climate response
(y-axls, red curve). See Roe and Baker,
2007 for details.

-------
Overall it is probably prudent to anticipate that there will not be dramatic reductions in uncertainty
about the upper bound on climate sensitivity (Knutti and Hegerl, 2008). On the timescale of several
decades, Nature herself will slowly reveal more of the answer. We will learn about the transient
climate response (see below) more quickly than the equilibrium climate sensitivity. Those interested
in understanding the above arguments in greater depth would do well to read the work of Prof. Reto
Knutti (at ETH in Switzerland) and his collaborators. His research is of extremely high caliber, and
quite accessible for a non-specialist.

The climate commitment.

What if all human influence on climate ceased overnight? Such a scenario— called the climate
commitment—informs us of the climate change we already face due only to past greenhouse gas
emissions. Framing the question this way has proven to be useful in providing a conceptual lower
bound on future climate warming.

Early definitions of the climate commitment simply fixed C02 concentrations at current levels (e.g.,
Wigley, 2005; Meehl et al., 2005), but maintaining current levels actually requires continued
emissions. Lately the focus has been more appropriately on the consequences of establishing zero
emissions (e.g., Solomon et al., 2009). Two important, though sometimes overlooked points should
be made. Firstly the geological carbon cycle means that, although much of the anthropogenic C02
ultimately gets absorbed by the ocean, some fraction — about 25 to 40% — remains in the
atmosphere for hundreds of thousands of years (e.g., Archer et al., 2009). Secondly aerosols, have a
short lifetime in the atmosphere (days to weeks). Thus when human influence ceases, aerosols are
rapidly washed out of the atmosphere and the effect of this is to unmask additional warming due to

Year Vear

Figure 5: Idealized representation of the climate commitment following a cessation of all
human influence on climate Based on Armour and Roe, 2010. Panel (a) shows a simple
view of how uncertainty in forcing has grown since 1800, as allowed by IPCC 2007 observed
uncertainties. After emission cease (here at yr 2000) the uncertain aerosols quickly vanish,
there is a jump in forcing due to sudden unmasking of the (relatively wel^known) radiative
forcing due to C02 and other greenhouse gases, which then declines slowly over time (black
line). Panel (b) shows the temperature over this period, from a simple climate model. For
each possible trajectory of past climate forcing history, a different value of climate sensitivity
is implied, in order that the accurately known past warming is reproduced (low past forcing
requires high climate sensitivity, and vice versa). The light blue curve shows the 90%
confidence range, as permitted by uncertainties in observations, which ultimately grows to
be 0.3 to 6°C at equilibrium. The dark blue curve is the likely' IPCC range (68%). It is this
range that is spanned by the main IPCC AR4 models because they under sample the
allowed range of past forcing. Note that these calculations here only include uncertainties
due to aerosols. The spread would be larger if uncertainties in GHG and ocean heat uptake
were included. Nonetheless the graph highlights that uncertainty in future temperatures is a
result of uncertainty in past forcing.

A-66

-------
the much more slowly declining C02 (illustrated in figure 2 and 5).

Figure 5 shows an idealized calculation of the climate commitment from Armour and Roe (2010),
which contains more details. The purpose of showing this is to highlight that our uncertainty about
future temperature comes primarily from our uncertainty about past forcing. After ceasing all
emissions, the degree and trajectory of future warming depends on the state of the current climate
forcing. We face the disconcerting possibility that our ultimate climate commitment already exceeds
2 °C, because of our current inability to rule out that past warming occurred with relatively little
climate forcing. In other words, the lower flank of the pdf of the past climate forcing distribution
(figure 5a) controls the upper flank of the pdf of the future temperature response (figure 5b).

Climate forcing and climate sensitivity are not independent

Perhaps the most important point to emphasize for the application to integrated assessment models
(lAMs) is that climate sensitivity and climate forcing are not independent of each other. For any
projections made of the future, a starting point for the current climate forcing must be assumed. We
are currently quite uncertain about what that starting point is. If aerosol forcing is strongly negative,
there is a strong implication that climate sensitivity is high. If aerosol forcing is weak, climate
sensitivity must be low. Uncertainties in climate forcing and climate sensitivity must not be assumed
to be independent.

The transient climate response.

Equilibrium climate sensitivity relates to a hypothetical distant future climate after the system has
equilibrated to a stipulated forcing. The transient climate response over the course of a few
centuries may be a more directly useful property of the climate system. A formal definition of the
transient climate sensitivity has been proposed as the global-average surface air temperature,
averaged over the 20-year period centered on the time of CO2 doubling in a 1% yr-i increase
experiment, which occurs roughly at 2070. While this metric may be more relevant for the future, a
negative trade-off is that its exact value depends on this artificially defined trajectory of emissions.

For reasons discussed below, the transient climate response is much better constrained than climate
sensitivity. In the words of the IPCC, it is very likely (> 9- in-10) to be greater than 1°C and very
unlikely (< l-in-10) to be greater than 3 °C. Thus the community is much more confident about the
evolution of the climate over the coming century than it is about the ultimate warming.

The immensely long timescales of high sensitivity climates.

A key factor in the long-term evolution of the climate is the diffusive nature of the ocean heat
storage (figure 6b). In order to reach equilibrium the ocean abyss must also warm, and because of
the relatively sluggish circulation of the deep ocean, the upper layers must be warmed before the
lower layers, and the more the temperature change must be, the longer diffusion takes to work. A
simple scaling analysis (e.g., Hansen et al., 1985) shows that:

Climate adjustment time a (climate sensitivity)2

A-67

-------
ocean circulation. Figure 6a shows one such calculation from Baker and Roe (2009), though there are
others (in particular see Held et al., 2010).

If lAMs are to be used to project out more than a few decades, it is critical that they represent this
physics correctly. A single adjustment time for climate, or a deep ocean that is represented as a
uniform block, cannot represent this behavior.

The extremely high temperatures found in the fat tail of climate sensitivity cannot be reached for
many centuries for very robust physical reasons. Failure to incorporate this fact will lead to a strong
distortion of the evolution of possible climate states, and of the subsequent 1AM analyses based on
them.

Radiation
Balaruce

z-0

250 5to 1000 10WO <=

Tiw jyfBi

Figure 6: (a) The evolution of possible climate trajectories in response to an instantaneous
doubling of C02 given the existing uncertainty in climate sensitivity. From Baker and Roe,
2009. Note the change to a logarithmic x-axis after 500 years. Low climate sensitivity is
associated with rapid adjustment times (decades to a century). High climate sensitivity has
extremely long adjustment times - thousand of years. This results from the fundamentally
diffusive nature of the ocean heat uptake, illustrated schematically in panel (b). Such behavior
is also reproduced in more complete physical models. See Held et al. (2010), for example.

CO2 stabilization targets are a mistake.

A prominent part of the conversation about action on climate change has centered on what the right
level of C02 should be in the atmosphere (e.g., Solomon et al., 2010). Some advocate for 350 ppmv
(e.g., Hansen et al. 2008), though we are already past 380 ppmv and climbing, others contemplate
the consequences of 450 ppmv (e.g., Hansen, et al., 2007), still others 550 ppmv (Pacala and
Soccolov, 2004; Stern, 2007).

A-68

-------
However decreeing and setting in stone a particular target for CO2 is fundamentally the wrong
approach, and a vastly inefficient way to avoid a particular climate scenario. This point was made
very elegantly and powerfully in a study by Allen and Frame (2007), reproduced in figure 7. Panel a)
shows a scenario of what could happen if we decided today to stabilize CO2 at 450 ppmv by 2100,

Figure 7: reproduced from Allen and Frame (2007). Carbon dioxide-induced warming under
two scenarios simulated by an ensemble of simple climate models. (Left) C02 levels are
stabilized in 2100 at 450 ppm; (right) ttie stabilization target is recomputed in 2050. Shading
denotes the likelihood of a particular simulation based on goodness-of-fit to observations of
recent surface and subsurface-ocean temperature trends. The darker the shading, the
likelier the outcome.

and then waited for the climate to evolve. Our current best guess is that would lead to an
equilibrium temperature change of 2 °C, taking us to the edge of what some have called dangerous
climate change. However because of our current uncertainty in climate sensitivity, the envelope of
possible climate states is quite broad by 2150. In other words, our hypothetical choice that we made
today still leaves us exposed to a quite broad envelope of risk. Note, though, that figure 7a is
consistent with figure 6 - temperatures in the fat tail of high climate sensitivity are still very, very far
from equilibrium at 2150.

Panel b) of figure 7 considers an alternative strategy in which we still act according to our best guess
today, but re-compute a new concentration target at 2050, based on the fact that 40 years have
elapsed and Nature has given us more information about what trajectory we are on. Figure 7b
makes it clear that this adaptive strategy is vastly more effective in achieving a desired climate target
(in this case a global temperature change of 2 °C). Because the link between C02 levels and global
temperature is uncertain, and because it is prudent to anticipate only incremental advances in our
understanding, it is common sense to pursue a strategy that has built-in flexibility rather than
declaring a fixed concentration.

How well do global projections correspond to regional projections?

Many of the most important climate impacts - changes in hydrology, storminess, heat waves,
snowpack, etc. - are fundamentally regional in nature. How reliable is global climate change as a
predictor of regional climate change? Since this is a question about the future, we are forced to use
climate models. Figure 8 analyzes how well global climate sensitivity correlates with local climate
change (in this case annual mean temperature and precipitation change in 2100), comparing among
eighteen different IPCC models (IPCC, 2007).

A-69

-------
It takes a correlation of r~ 0.75 before half of the variance (i.e., r2) of the local climate change is
attributable to the global climate change. Only a very few patches of the planet achieve even this
level of correlation in annual temperature (Figure 8a) and nowhere reaches this measure in annual
precipitation (Figure 8b). This highlights that the connection between regional and global climate
change is not that strong. This result should not be surprising: though models may all agree on the
sign of the climate change in a given region, there is a great deal of scatter and individual model
vagaries in projecting the magnitude of the climate change. Research into the limits of regional
predictability is only just beginning. A useful starting point is Hawkins and Sutton (2009).

-1 -OB -0.6 -0.4 -02 0 02 0.4 0.6 O.S 1

Figure 8: a) correlation among 17 IPCC climate models of their global equilibrium climate
sensitivity and their local annual-mean temperature change in 2100,; b) same as a), but for
annual-mean precipitation. Calculation made by N. Feldl from IPCC archived model output
based on the A1B emissions scenario, and similar plots for other variabfes are at
httpJ/earthwe b.ess. Washington.e du/roe/GerardWeb/Publica tions. html.

Summary.

1) The most important point to drive home is that uncertainty is not ignorance. The planet has
warmed in the recent past, and will continue to warm for the foreseeable future. That this is
a result of our actions is beyond rational dispute. The overwhelming preponderance of the
IPCC 2007 report is extremely reliable, and reflects an objective characterization of the best
current understanding about climate. All of the following points are consistent with (and in
many cases drawn from) that report.

2) A traditional measure of the planet's response, equilibrium climate sensitivity is uncertain,
primarily because of uncertainty in the radiative forcing due to aerosols. This precludes us
from calibrating our models of climate with greater accuracy.

3) However a focus on climate sensitivity may be misplaced because of the tremendously long
timescales associated with reaching equilibrium - thousands of years in the case of the fat
tail of high climate sensitivity.

4) If all human influence were to cease today, the rapid loss of anthropogenic aerosols from
the climate would unmask CO?, warming, and the planet's temperature would increase as a
result. The degree of warming is quite uncertain.

5) For related reasons, a strategy that aims to stabilize concentration of greenhouse gasses at a
particular level is a mistake, because the degree of warming is still unpredictable. A strategy
that aims for a flexible emissions will be much more effective at preventing a particular level
of warming.

A-70

-------
6)	lAMs have to make choices about how to represent climate forcing associated with human
activity. We are quite uncertain about what this level is right now. It is crucial to appreciate
that uncertainty in climate sensitivity and uncertainty in climate forcing cannot be treated as
independent.

7)	Many climate damages both to humans and to the biosphere result from regional climate
factors. Unfortunately, there is relatively little agreement among climate models about how
global climate changes relate to local climate changes, and this is especially true in some of
the most vulnerable subtropical regions. Thus the meaning of analyses that use only global
temperature changes to assign climate damages is unclear.

Acknowledgements: I'm grateful for helpful conversations and comments on this report from Marcia Baker,

Kyle Armour, Nicole, Feldl, Eric Steig, Yoram Bauman, David Battisti, and Steve Newbold. All remaining errors

are mine.

References

Allen, M.R., and Frame, D.J., 2007: Call off the quest, Science, 318, 582-583.

Archer, D., et al., 2009: Atmospheric lifetime of fossil-fuel carbon dioxide. Annu. Rev. of Earth and
Planet. Sci. 37, 117-134.

Armour, K.C., and G.H. Roe, 2010: Climate commitment in an uncertain world. Submitted,available
at http://earthweb.ess.washington.edu/roe/GerardWeb/Home.html

Baker, M.B., and G.H. Roe, 2009: The shape of things to come: Why is climate change so predictable?
J. dim. 22, 4574-4589.

Baker, M.B., G.H. Roe, K.C. Armour, 2010: How sensitive is climate sensitivity. In preparation,
available at http://earthweb.ess. washington.edu/roe/GerardWeb/Home.html

Charney, J., and Coauthors, 1979: Carbon dioxide and climate: A scientific assessment. National
Academy of Sciences, 22 pp.

Hansen, J.E., G. Russell, A. Lacis, I. Fung, D. Rind, P. Stone, 1985: Climate Response Times:
Dependence on Climate Sensitivity and Ocean, Science 229, 857-859

Hansen, J. E., et al., 2007: Dangerous human-made interference with climate: a GISS modelE study,
Atmos. Chem. Phys., 7, 2287-2312.

Hawkins, E., and R. Sutton, 2009: The potential to narrow uncertainty in regional climate predictions'
Bull. Am. Met. Soc., 90, 1095, doi: 10.1175/2009BAMS2607.1

Henriksson, S.V., E. Arja. M. Laine, J. Tamminen, A. Laaksonen , 2010: Comment on Using multiple
observationally-based constraints to estimate climate sensitivity by J. D. Annan and J. C.
Hargreaves, Geophys. Res. Lett., 2006, Climate of the Past, 6, 411414.

IPCC, 2007: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the
Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S.,
D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M.Tignor and H.L. Miller (eds.)].
Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

A-71

-------
Kiehl, J.T., 2007: Twentieth century climate model response and climate sensitivity, Geophys. Res.
Lett., 34, L22710, doi:10.1029/2007GL031383.

Knutti, R., 2008: Why are climate models reproducing the observed global surface warming so well?
Geophys. Res. Lett. 35, L18704.

Knutti, R. and G.C. Hegerl, 2008: The equilibrium sensitivity of the Earths temperature to radiation
changes, Nature Geoscience, 1, 735-743, doi:10.1038/ngeo337

Knutti, R., R. Furrer, C. Tebaldi, J. Cermak and G.A. Meehl, 2010, Challenges in combining projections
from multiple models, Journal of Climate, 23, 27Z9-275&, DOI 10.1175/2009JCLI3361.1

Menon, S. 2004: Current uncertainties in assessing aerosol impacts on climate. Ann Rev. Env. Res.,
29, 1-30.

Meehl, G.A., et al., 2005: How much more global warming and sea level rise? Science, 307,1769-
1772.

Pacala, S., R.H. Socolow, 2004: Stabilization Wedges: Solving the Climate Problem for the Next 50
Years with Current Technologies. Science, 305 (5686): 968-972

Roe, G.H., and M.B. Baker, 2007: Why is climate sensitivity so unpredictable? Science 318, 629-632,
doi: 10.1126/science. 1144735.

Roe, G.H., 2009: Feedbacks, time scales, and seeing red. Ann.Rev. of Earth and Plan. Sci. 37, 93-115.

Solomon, S., et al., 2009: Irreversible climate change due to carbon dioxide emissions. Proc.Natl.
Acad. Sci. USA 106, 1704-1709.

Solomon, S. and fourteen others, 2010: Climate Stabilization Targets: Emissions, Concentrations and
Impacts over Decades to Millennia National Research Council, National Academy of Sciences.

Stainforth, D., and Coauthors, 2005: Uncertainty in predictions of the climate response to rising
levels of greenhouse gases. Nature, 433, 403-406, doi:10.1038/nature03301.

Stern, N., 2007: Stern Review on the Economics of Climate Change: Part III: The Economics
ofStabilisation. HM Treasury, London:

Wigley, T.M.L., 2005: The climate change commitment. Science 307, 1766-1769.

A-72

-------
Notes for EPA & DOE discussion meeting

Martin L. Weitzman
November, 2010

First thoughts on '"thinking about' high-temperature damages from potential catastrophes
in climate change."

'Thinking about' is the right phrase. This is a notoriously intractable area even to conceptualize, much
less to model or to quantify. Don't expect miracles or breakthroughs here — too many "unknown
unknowns" with seemingly non-negligible probabilities to feel comfortable with.

What is the nature of the beast?

The economics of climate change consists of a very long chain of tenuous inferences fraught with big
uncertainties in every link: beginning with unknown base-case GHG emissions; then compounded by big
uncertainties about how available policies and policy levers will transfer into actual GHG emissions;
compounded by big uncertainties about how GHG flow emissions accumulate via the carbon cycle into
GHG stock concentrations; compounded by big uncertainties about how and when GHG stock
concentrations translate into global average temperature changes; compounded by big uncertainties
about how global average temperature changes decompose into regional climate changes; compounded
by big uncertainties about how adaptations to, and mitigations of, regional climate-change damages are
translated into regional utility changes via a regional "damages function"; compounded by big
uncertainties about how future regional utility changes are aggregated into a worldwide utility function
and what should be its overall degree of risk aversion; compounded by big uncertainties about what
discount rate should be used to convert everything into expected-present-discounted values. The result
of this lengthy cascading of big uncertainties is a reduced form of truly enormous uncertainty about an
integrated assessment problem whose structure wants badly be transparently understood and stress
tested for catastrophic outcomes.

Let welfare W stand for expected present discounted utility, whose theoretical upper bound is B. Let
D=B-W be expected present discounted disutility. Here D stands for what might be called the
"diswelfare" of climate change. Unless otherwise noted, my default meaning of the term "fat tail" (or
"thin tail") will concern the upper tail of the PDF of InD, resulting from whatever combination of
probabilistic temperature changes, temperature-sensitive damages, discounting, and so forth, by which
this comes about. Empirically, it is not the fatness of the tail of temperature PDFs alone or the
reactivity of the damages function to high temperatures alone, or any other factor alone, that counts,
but the combination of all such factors. Probability of welfare-loss catastrophe declines in impact size,
but key question here is: how fast a decline relative to size of catastrophe? When we turn to theory, it
seems to highlight that the core "tail fattening" mechanism is an inherent inability to learn about
extreme events from limited data.

A-73

-------
What do rough calculations show about this beast?

I have played with some extremely rough numerical examples. GHG concentration implies a PDF of
temperature responses implies a PDF of damages (given a "damages function"). In order to get tail
fatness to matter for willingness to pay to avoid climate change requires a much more reactive damages
function than the usual quadratic. Usual quadratic damages function loses 26% of output for a 12dC
temperature change. At 2% annual growth rate, 12dC change 200 years from now implies that welfare-
equivalent consumption then will still be 37 times higher than today. If you use the standard quadratic
damages function, you cannot get much damage from extreme temperatures. If make a reactive
damages function, such that, say, 12dC temperature increase causes welfare-equivalent consumption to
shrink to, say, 5% of today's level, then get very high WTP to reduce GHG target levels. Model is
terrified of flirting with high C02-e levels, especially above 700 ppm. Incredible dependence on degree
of risk aversion (2, 3, or 4?), fatness of tail PDFs (climate sensitivity PDF: normal, lognormal, Pareto?),
and so forth. My own tentative summary conclusion: tail of extreme climate change welfare-loss
possibilities is much too fat for comfort when combined with reactive damages at high temperatures. It
looks like this could influence such things as social cost of carbon.

Is there anything constructive to take away from this gloomy beast?

My tentative answer: a qualified maybe. Some possible rough ideas follow.

1. Keep a sense of balance. A small but fat-tailed probability of disastrous damages is not a
realization of a disaster. Highly likely outcome is a future sense that we dodged a bullet (like
Cuban missile crisis?). Yet when all is said and done, catastrophic climate change looks to me
like a very serious issue.

2. Try standard CBA or 1AM exercises in good faith. But, be prepared - when dealing with
extremes - that answers might depend non-robustly upon seemingly-obscure assumptions
about tail fatness, about how the extreme damages are specified (functional forms, parameter
values, etc.), assumptions about rates of pure time preference, degrees of risk aversion,

Bayesian learning, C02 stock inertia, CH4 releases from clathrates, mid-course correction
possibilities, etc. Some crude calculations seem to indicate great welfare sensitivity to
seemingly-obscure factors such as the above, most of which are difficult to know with any
degree of precision. Do CBAs and lAMs, study answers, but maybe don't try to deny the
undeniable if these answers are sensitive to tail assumptions in a highly nonlinear welfare
response to extreme uncertainty.

3. Should we admit to the public that climate change CBA looks more iffy and less robust than, say,
CBA of S02 abatement, or would this be self defeating?

4. Maybe there should be relatively more research emphasis on understanding extreme tail
behavior of climate-change welfare disasters. Alas, this is very easy to say but very difficult to
enact. How do we learn the fatness of PDF tails from limited observations or experience?

A-74

-------
5.	A need to compare how fat are tails of climate-change welfare loss with how fat are tails of any
proposed solutions, such as nuclear power, below-ground carbon sequestration, etc.

6.	Suppose that a lot of expected present discounted disutility is in the bad fat tail of the welfare-
loss PDF. Realistically, how can we limit some of the most horrific losses in worst-case
scenarios? Can we filter-learn fast enough to offset residence time of atmospheric C02 stocks
by altering GHG emission flows in time to work? Is tail fatness an argument for developing an
emergency-standby backstop role for fast geoengineering? Any other backstop options? Take-
home lesson here: hope for the best and prepare for the worst. At least we should be prepared,
beforehand, for dealing with ugly scenarios, even if they are low-probability events. Should the
discussion about emergency preparedness begin now?

A-75

-------
Earth System Tipping Points

Timothy M. Lenton

School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, UK

Definitions

A tipping point is a critical threshold at which the future state of a system can be qualitatively altered by
a small change in forcing1. A tipping element is a part of the Earth system (at least sub-continental in
scale) that has a tipping point1. Policy-relevant tipping elements are those that could be forced past a
tipping point this century by human activities. Abrupt climate change is the subset of tipping point
change which occurs faster than its cause'. Tipping point change also includes transitions that are slower
than their cause (in both cases the rate is determined by the system itself). In either case the change in
state may be reversible or irreversible. Reversible means that when the forcing is returned below the
tipping point the system recovers its original state (either abruptly or gradually). Irreversible means that
it does not (it takes a larger change in forcing to recover). Reversibility in principle does not mean that
changes will be reversible in practice.

Tipping elements in the Earth's climate system

Previous work1 identified a shortlist of nine potential policy-relevant tipping elements in the climate
system that could pass a tipping point this century and undergo a transition this millennium under
projected climate change. These are shown with some other candidates in Figure 1.

Melt of —
Greenland Ice Sheet

4 Indian^;
iMonsbonj
Chaotic f
Multistability

. Sahara 1,
Greening

lnstabilit^ftWesT^ta^ticj

~Bj lce[Sheet^H

Figure 1: Map of potential policy-relevant tipping elements in the Earth's climate system overlain on population
density. Question marks indicate systems whose status as tipping elements is particularly uncertain.

Arctic Sea- ce Loss

Climatic

Change-Induced

Ozone Hole?

Permafrost and
Tundra* lloss?

West African

Monsoon Shift

Dieback
of Amazon
Rainforest

Change in ENSOj
Amplitude ortFrequencyJ

population density [persons per km2]

no data 0 5 10 20 100 200 300 400 1000

A-76

-------
We should be most concerned about those tipping points that are nearest (least avoidable) and those
that have the largest negative impacts. Generally, the more rapid and less reversible a transition is, the
greater its impacts. Additionally, any positive feedback to global climate change may increase concern,
as can interactions whereby tipping one element encourages tipping another. The proximity of some
tipping points has been assessed through expert elicitation1,3. Proximity, rate and reversibility have been
also assessed through literature review1, but there is a need for more detailed consideration of impacts4.
The following are some of the most concerning tipping elements:

The Greenland ice sheet (GIS) may be nearing a tipping point where it is committed to shrink1,3. Striking
amplification of seasonal melt was observed in summer 2007 associated with record Arctic sea-ice loss5.
Once underway the transition to a smaller ice cap will have low reversibility, although it is likely to take
several centuries (and is therefore not abrupt). The impacts via sea level rise will ultimately be large and
global, but will depend on the rate of ice sheet shrinkage. Latest work suggests there may be several
stable states for ice volume, with the first transition involving retreat of the ice sheet onto land and
around 1.5 m of sea level rise6.

The West Antarctic ice sheet (WAIS) is currently assessed to be further from a tipping point than the
GIS, but this is more uncertain1,3. Recent work has shown that multiple stable states can exist for the
grounding line of the WAIS, and that it has collapsed repeatedly in the past. It has the potential for more
rapid change and hence greater impacts than the GIS.

The Amazon rainforest experienced widespread drought in 2005 turning the region from a sink to a
source (0.6-0.8 PgC yr"1) of carbon7. If anthropogenic-forced8 lengthening of the dry season continues,
and droughts increase in frequency or severity9, the rainforest could reach a tipping point resulting in
dieback of up to ~80% of the rainforest10"13, and its replacement by savannah. This could take a few
decades, would have low reversibility, large regional impacts, and knock-on effects far away.

Widespread dieback is expected in a >4 °C warmer world3, and it could be committed to at a lower
global temperature, long before it begins to be observed14.

The Sahel and West African Monsoon (WAM) have experienced rapid but reversible changes in the
past, including devastating drought from the late 1960s through the 1980s. Forecast future weakening
of the Atlantic thermohaline circulation contributing to 'Atlantic Nino' conditions, including strong
warming in the Gulf of Guinea15, could disrupt the seasonal onset of the WAM16 and its later 'jump'
northwards17 into the Sahel. Whilst this might be expected to dry the Sahel, current global models give
conflicting results. In one, if the WAM circulation collapses, this leads to wetting of parts of the Sahel as
moist air is drawn in from the Atlantic to the West15,18, greening the region in what would be a rare
example of a positive tipping point.

A-77

-------
land-oceari temperature gradient which is critical in triggering monsoon onset1'. Conversely, greenhouse
gas forcing is acting to strengthen the monsoon as it warms the northern land masses faster than the
ocean to the south. In some future projections, ABC forcing could double the drought frequency within a
decade15 with large impacts, although it should be highly reversible.

Estimation of likelihood under different scenarios

If we pass climate tipping points due to human activities (which in IPCC language are called "large scale
discontinuities"21), then this would qualify as dangerous anthropogenic interference (DAI) in the climate
system. Relating actual regional tipping points to e.g. global mean temperature change is always
indirect, often difficult and sometimes not meaningful. Recent efforts suggest that 1 °C global warming
(above 1980-1999) could be dangerous as there are "moderately significant"21 risks of large scale
discontinuities, and Arctic sea-ice and possibly the Greenland ice sheet would be threatened1,72. 3 °C is
clearly dangerous as risks of large scale discontinuities are "substantial or severe"21, and several tipping
elements could be threatened1. Under a 2-4 °C committed warming, expert elicitation' gives a >16%
probability of crossing at least 1 of 5 tipping points, which rises to >56% for a >4 °C committed warming.
Considering a longer list of 9 potential tipping elements, Figure 2 summarizes recent information on the
likelihood of tipping them under the IPCC range of projected global warming this century.

Figure 2: Burning embers diagram for the likelihood of tipping different elements under different degrees of
global warming33 - updated, based on expert elicitation results' and recent literature.

I o

° 0.5

[ Certain

More likely than not

As likely as not
Less likely than not
Won't happen

Early warning prospects

An alternative approach to assessing the likelihood of tipping different elements is to try and directly
extract some information on their present stability (or otherwise). Recent progress has been made in
identifying and testing generic potential early warning indicators of an approaching tipping point12 2/.
Slowing down in response to perturbation is a nearly universal property of systems approaching various
types of tipping point2 " This has been successfully detected in past climate records approaching
different transitions24'2 ', and in model experiments^4 'Other early warning indicators that have been
explored for ecological tipping points28, include increasing variance28, skewed responses28 2 and their
spatial equivalents'J. These are beginning to be applied to anticipating climate tipping points. For

A-78

-------
climate sub-systems subject to a high degree of short timescale variability ('noise'), flickering between
states may occur prior to a more permanent transition31. For such cases, we have recently developed a
method of deducing the number of states (or 'modes') being sampled by a system, their relative stability
(or otherwise), and changes in these properties over time32.

Applying these methods to observational and reconstructed climate indices leading up to the present,
we find that the Atlantic Multi-decadal Oscillation (AMO) index, which is believed to reflect fluctuations
in the underlying strength of the thermohaline circulation, is showing signs of slowing down (i.e.
decreasing stability) and of the appearance of a second state (or mode of behavior). On interrogating
the underlying sea surface temperature data (used to construct the index), we find that recent
significant changes are localized in the northernmost North Atlantic, and are investigating the possible
relationship with changes in Arctic sea-ice cover. Meanwhile, some other climate indices, e.g. the Pacific
Decadal Oscillation (PDO) show signs of increasing stability.

References

1	Lenton, T. M. et al., Tipping Elements in the Earth's Climate System. Proceedings of the National

Academy of Science 105 (6), 1786 (2008).

2	Rahmstorf, S., in Encyclopedia of Ocean Sciences, edited by J. Steele, S. Thorpe, and K. Turekian

(Academic Press, London, 2001), pp. 1.

3	Kriegler, E. et al., Imprecise probability assessment of tipping points in the climate system. Proceedings

of the National Academy of Science 106 (13), 5041 (2009).

4	Lenton, T. M., Footitt, A., and Dlugolecki, A., Major Tipping Points in the Earth's Climate System and

Consequences for the Insurance Sector, 2009.

5	Mote, T. L., Greenland surface melt trends 1973-2007: Evidence of a large increase in 2007.

Geophysical Research Letters 34, L22507 (2007).

6	Ridley, Jeff, Gregory, Jonathan, Huybrechts, Philippe, and Lowe, Jason, Thresholds for irreversible

decline of the Greenland ice sheet. Climate Dynamics, 1 (2009).

7	Phillips, Oliver L. et al., Drought Sensitivity of the Amazon Rainforest. Science 323 (5919), 1344 (2009).

8	Vecchi, G. A. et al., Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing.

Nature 441, 73 (2006).

9	Cox, Peter M. et al., Increasing risk of Amazonian drought due to decreasing aerosol pollution. Nature

453, 212 (2008).

10	Cox, P.M. et al., Amazonian forest dieback under climate-carbon cycle projections for the 21st

century. Theoretical and Applied Climatology 78, 137 (2004).

A-79

-------
11	Scholze, Marko, Knorr, W., Arnell, Nigel W., and Prentice, I. C., A climate-change risk analysis for world

ecosystems. Proceedings of the National Academy of Science 103 (35), 13116 (2006).

12	Salazar, Luis F., Nobre, Carlos A., and Oyama, Marcos D., Climate change consequences on the biome

distribution in tropical South America. Geophysical Research Letters 34, L09708 (2007).

13	Cook, Kerry H. and Vizy, Edward K., Effects of Twenty-First-Century Climate Change on the Amazon

Rain Forest. Journal of Climate 21, 542 (2008).

14	Jones, Chris, Lowe, Jason, Liddicoat, Spencer, and Betts, Richard, Commited ecosystem change due to

climate change. Nature Geoscience (submitted).

15	Cook, Kerry H. and Vizy, Edward K., Coupled Model Simulations of the West African Monsoon System:

Twentieth- and Twenty-First-Century Simulations. Journal of Climate 19, 3681 (2006).

16	Chang, Ping et al., Oceanic link between abrupt change in the North Atlantic Ocean and the African

monsoon. Nature Geoscience 1, 444 (2008).

17	Hagos, Samson M. and Cook, Kerry H., Dynamics of the West African Monsoon Jump. Journal of

Climate 20, 5264 (2007).

18	Patricola, C. M. and Cook, Kerry H., Atmosphere/vegetation feedbacks: A mechanism for abrupt

climate change over northern Africa. Journal of Geophysical Research (Atmospheres) 113,
D18102 (2008).

19	Ramanathan, V. et al., Atmospheric brown clouds: Impacts on South Asian climate and hydrological

cycle. Proceedings of the National Academy of Science 102 (15), 5326 (2005).

20	Meehl, G. A., Arblaster, J. M., and Collins, W. D., Effects of Black Carbon Aerosols on the Indian

Monsoon. Journal of Climate 21, 2869 (2008).

21	Smith, Joel B. et al., Assessing dangerous climate change through an update of the Intergovernmental

Panel on Climate Change (IPCC) "reasons for concern". Proceedings of the National Academy of
Sciences 106 (11), 4133 (2009).

22	Hansen, J. et al., Dangerous human-made interference with climate: a GISS modelE study. Atmos.

Chem. Phys. 7 (9), 2287 (2007).

23	Lenton, T. M. and Schellnhuber, H. J., Tipping the scales. Nature Reports Climate Change 1, 97 (2007).

24	Livina, V. and Lenton, T. M., A modified method for detecting incipient bifurcations in a dynamical

system. Geophysical Research Letters 34, L03712 (2007).

25	Dakos, V. et al., Slowing down as an early warning signal for abrupt climate change. Proceedings of the

National Academy of Sciences of the United States of America 105 (38), 14308 (2008).

A-80

-------
26	Lenton, T. M. et al., Using GENIE to study a tipping point in the climate system. Philosophical

Transactions of the Royal Society A 367 (1890), 871 (2009).

27	Scheffer, M. et al., Early warning signals for critical transitions. Nature 461, 53 (2009).

28	Biggs, R., Carpenter, S. R., and Brock, W. A., Turning back from the brink: Detecting an impending

regime shift in time to avert it. Proceedings of the National Academy of Science 106 (3), 826
(2009).

29	Guttal, V. and Jayaprakash, C., Changing skewness: an early warning signal of regime shifts in

ecosystems. Ecology Letters 11, 450 (2008).

30	Guttal, V. and Jayaprakash, C., Spatial variance and spatial skewness: leading indicators of regime

shifts in spatial ecological systems. Theoretical Ecology 2, 3 (2009).

31	Bakke, J. et al., Rapid oceanic and atmospheric changes during the Younger Dryas cold period. Nature

Geoscience 2, 202 (2009).

32	Livina, V. N., Kwasniok, F., and Lenton, T. M., Potential analysis reveals changing number of climate

states during the last 60 kyr. Clim. Past 6 (1), 77 (2010).

A-81

-------
Catastrophic Climate Change

Mike Toman

World Bank Development Research Group

Draft, subject to revision; please do not cite or quote. Responsibility for content is the author's alone.
October 25, 2010 version

Introduction

The question of how to assess prospects of climate change catastrophes has been the focus of a great
deal of recent research and debate. An example of the classic conundrum of low probability - high
consequences events, a climate change catastrophe is a highly unlikely event, but if it did occur it would
severely affect well-being across the world - though it would affect poor countries much more seriously
than richer countries.^he larger geographical scale of climate change catastrophes distinguishes them
from more localized extreme events. The consequences of catastrophes also are in varying degrees very
costly, if not possible, to reverse.

Examples of global catastrophes include very large and relatively rapid increases in sea level from faster
melting and collapse of ice sheets, slower changes in ocean currents that have insidious effects on
weather patterns, and large scale destruction of forests and other ecosystems, fairly rapid loss of global
forest cover. Unlike sudden disasters such as earthquakes, the onset of these events is measured in
multiple decades or centuries; but once they occur it is impossible to reverse the impacts. Other
permanent effects of climate change are anticipated to be increases in the frequency and severities of
droughts, floods, and hurricanes, leading to corresponding destruction of crops, water supplies, and
coastal infrastructure. While each of these individual events is a more localized disaster, the cumulative
effect could be a global catastrophe created by the —cascading consequences|| of more localized
disasters occurring in relatively quick succession, each amplifying the effects of others.2

A key step in evaluating risks of climate change catastrophes is to assess not only the impacts on the
physical climate system, but also the consequences in terms of human impacts. The most immediate
implication is that while a physical —tipping point|| may be reached at some unknown future date T°,
the human impacts will evolve more slowly, reaching an intensity viewed as catastrophic only at some
date T1 > T°. This distinguishes climate change from, for example, the risk of catastrophe posed by a
gigantic volcanic eruption, or nuclear war. While a gradual onset of impacts will not prevent a
catastrophe if reversal is not possible, it can provide a window of time for major action to avert or adapt
to the threat - if signals of the changes are detectable. More fundamentally, the assessment of what
constitutes catastrophic human impacts involves not just climate change and earth system science, but

1	In terms of absolute numbers, losses are likely to be larger in richer nations. As a percentage of GDP, however,
less developed countries are likely to face higher damages since most are more dependent on agriculture and less
likely to have the resources to adopt measures that could reduce damages.

2	This possibility appears to have received little systematic attention in reviews of climate change impacts by the
IPCC and others, though it figures prominently in discourse about national security consequences of climate
change.

A-82

-------
also inherent value judgments about what magnitude and speed of consequences are deemed to be
catastrophic. For example, the now-often-cited —scientific near-consensus|| about the urgent need to
hold warming to less than 2°C relative to pre-industrial times reflects more than a natural science
evaluation of climate change impacts.

Climate change catastrophes pose a familiar challenge for assessing the impacts of low probability - high
impact events: while exact quantification is not possible, the most extreme adverse impacts from
climate change—say the worst 1% of scenarios—may account for a large portion of losses in expected
value terms. This implies that focusing primarily on a trajectory of more likely anticipated climate
change damages may miss an important part of the problem.3 Yet, these consequences of an unlikely
but possible climate change catastrophe need to be weighed against a variety of other risks society
faces.

Further complicating the problem is that climate change catastrophes may be better characterized by
ignorance than uncertainty. That is, not only do we not know the probability of a particular mega-
catastrophe occurring, we do not even know many of the possible outcomes. A catastrophe from
climate change could stem from a cause or have impacts that currently receive little attention.4 Some
authors have suggested that this level of ignorance, coupled with the very low probability of an event
and the possibility of extremely severe impacts, hamstrings the use of rational-choice based methods for
analyzing response options. This in turn requires confronting the possibility that attitudes of the broader
public about such events will not align very well with the results of a more systematic evaluation of the
pros and cons of different response options, raising questions about what sets of preferences and
beliefs should govern policy making.

Climate Change Catastrophes

3 For many classes of disasters and catastrophes, the most extreme small percent of the situations represent a
significant proportion of the losses. We have witnessed this —fat tailll phenomenon recently with terrorist deaths
and losses in a financial crisis. 9/11 and the 2008-09 financial meltdown caused more deaths and dollar losses
respectively than all terrorist incidents and financial catastrophes in the post WWII era. With such phenomena,
losses are better characterized by a power law than by a normal or even lognormal distribution. The debate about
fat tails in relation to climate catastrophes has been a subject of lively recent debate among Weitzman, Pindyck,
Nordhaus, and others.

4 The history of the past 40 years is sobering with respect to the ability to identify catastrophe risks. In 1970,
nuclear war would have been the leading contender for any world catastrophe, and looking forward few would
have predicted the major looming threats of the current era, which would include not just climate change, but also
global pandemics and terrorism.

A-83

-------
(based on 1995 baseline conditions) that are relatively small - on the order of 0.5% of world GDP for a 5
m rise. Dasgupta et al. (2007) report figures for developing countries on the order of 6% of GDP, those
these estimates do not take account of possibilities for ex ante efforts to mitigate risks. On the other
hand, estimates based on historical baselines will tend to under-state the economic impacts of sea level
rise by not taking account of likely future growth in the coming years in the share of GDP concentrated
in coastal areas.5

A second important category of global catastrophe risk involves disruptions of ocean circulation from
climate change, with potentially disastrous effects on regional weather patterns and long-term climate
(Vellinga and Wood 2008). Such impacts are most commonly seen as developing over many hundreds of
years. In contrast, very large-scale ecosystem disruptions could occur significantly sooner. Changes in
ecosystems resulting from changes in temperature and rainfall incidence and increased climate
variability have the potential to cause very significant loss of biodiversity—on the order of 20-30%
extinction within a few decades. There is also the prospect of major changes in vegetation, in particular,
irreversible conversion of forest to grassland, desertification, and acidification of the ocean (Smith,
Schneider, Oppenheimer et al. 2009). Another cause for significant concern is the possibility that
positive feedback effects in the climate change process itself could occur (e.g., liberation of trapped
methane from ice, rapid increases in C02from vegetation dieback, or increased heat absorption as
glaciers retreat), causing the abovementioned changes to occur more rapidly.

There also has been significant scientific research on how climate change can effect more localized
disasters, such as heat waves, flooding, droughts, and changes in hurricane frequency or intensity. Less
understood is how a number of smaller disasters all occurring over a relatively short time period could
mutually reinforce each other in such as way that the resulting "cascade of consequences" becomes a
global catastrophe. Extreme events can have secondary consequences that generate substantial
amounts of additional damages; secondary consequences in turn can trigger tertiary consequences that
further amplify the adverse consequences; and so on. One example would be if increased drought from
climate change in different regions successively caused a series of local food shortages to occur in close
proximity, leading to political instability, a breakdown of civil order, large-scale migration for survival,
and regional conflicts. Another example could be a series of local fires occurring in climate-stressed
forests and grasslands overly widely dispersed areas, adding up to a large-scale destruction of resources,
ecosystem services, and livelihoods over a large area.

The compounding or amplifying effects of individual adverse impacts would be the result of exceeding
the resilience of a number of local socioeconomic systems in rapid succession. More frail components of
socioeconomic systems, such as marginal subsistence agriculture, represent potential places of
vulnerability. Cascading-event catastrophes could occur much more rapidly than the slower-onset global
impacts discussed above, especially as climate change accelerates and greater negative impacts occur at
local scales. It is possible that more comprehensive local monitoring of disaster risks may facilitate the

5 Using 1995 data, it has been estimated that around 400 million people would be impacted by a 5 m rise in sea
level and that a WAIS collapse in 100 years could cause, at the peak, 350,000 forced migrations a year for a decade
(Nicholls, Tol and Vafeidis 2008).

A-84

-------
development of early warning indicators for cascading catastrophes. For example, if several years of
historically unusual drought weakened agricultural systems in many vulnerable parts of the world, there
would be a stronger basis for concern about cascading consequences than if agricultural failures were
not occurring in such rapid succession. However, the time interval for action to avert the potential
catastrophe could be short.

Traditional responses to the risk of extreme events are of limited value in mitigating risks of a mega-
catastrophe. The underlying changes in the climatic system could not be reversed over any time scale
relevant for decision-makers. Traditional insurance mechanisms will not function effectively for this type
of event, because the risks are —systemic|| and cannot effectively be reallocated to diversify. Moreover,
significant international transfers from richer to more vulnerable poorer countries are unlikely when a
catastrophe affects broad swaths of the world.

Evaluating Climate Change Catastrophe Risks

The traditional economic model for decision making under uncertainty is expected utility theory, in
which decision makers maximize the utility they receive from potential outcomes weighted by the
probability the outcomes will occur. In the climate change economics literature, GHG abatement policies
with the expected net benefits over time are identified using dynamic Integrated Assessment Models
(lAMs) that compare the anticipated costs of abatement with avoided damages from climate change
over time. By and large these models are deterministic and are used for scenario-based comparisons of
policies under different assumptions about climate change damages and abatement costs. However, a
literature has developed in which catastrophes are treated as (usually known) large-scale rapid-onset
economic damages with an uncertain date of occurrence, the probability of which increases as
atmospheric GHG concentrations rise.6

A common finding in these studies is that while the risk of such catastrophes increases the expected
economic benefits of more rapid GHG mitigation, the effect is not that significant qualitatively unless the
probability of nearer-term catastrophe is quite high, the size of the catastrophe is truly astronomical, or
the discount rate used to value future catastrophic impacts is quite low. The scientific information on
catastrophes summarized above indicates that catastrophes are extremely unlikely in any time frame
short of several decades at the very least, and that while the ultimate effects may indeed be huge, the
most severe impacts will develop only gradually. Until scientific understanding of climate change
catastrophes leads to stronger findings on their proximity and severity, the choice of discount rate will
be the most important determinant of the cost of future catastrophes in the expected-utility framework.

The discount rate issue in turn continues to be very hotly debated, and only a very brief summary of key
points is offered here. Two strands of positive analysis has argued for applying a lower discount rate to
longer-term climate change costs, including catastrophes, than might be inferred from research on
consumer time preference or rates of return on investment. One is that individuals may discount the
future hyperbolically, so rates of discount decline and ultimately plateau at a fairly low number as one

6 References - Kverndokk et a I, Pizer, Nordhaus. Earlier foreshadowing by Manne.

A-85

-------
goes out into the future. The other is that when one accounts for the higher marginal utility of income
for the poor facing more adverse impacts from climate change, then under reasonable assumptions the
effective time discount rate after adjusting for distributional differences is reduced. In addition, if
climate change has the most severe effects on longer-term economic growth when growth itself is more
likely to be weak, then policies to reduce the threat of catastrophe will have a lower effective discount
rate because of their contribution to reducing intertemporal economic risk.7

Even with these considerations, however, the resulting implied discounting of future over current
returns may not be small enough for catastrophes to carry major weight in evaluating the potential
impacts of climate change. Unless the discount rate is under 1%, and perhaps even close to zero, severe
future consequences that will not arrive for some time and are not world-threatening may still be too
—telescoped.|| Stern and others have addressed the issue of discounting by using normative arguments
to suggest a discount rate at or near zero is in fact appropriate. Two other arguments, not so dependent
on normative precepts, may also add weight to the importance of catastrophe risks in evaluating climate
change impacts.

Hypothesis 1: People are Not Expected Utility - Maximizers

There is a growing literature from behavioral economics and psychology which demonstrates that
individuals do not consistently make decisions according to the expected utility paradigm.8 If individuals
are only boundedly rational, they have neither the time nor the capacity to fully assess the
consequences of decisions. In that case, individuals adopt certain rules of thumb and mental shortcuts
to make decisions. These so-called heuristics can lead to choices that depart from predictions of
expected utility theory.

When thinking about possible disasters, it has been found that people tend to be over-optimistic,
thinking negative outcomes are less likely to happen to them. When a risk is highly emotional, however,
people can disregard probabilities altogether, treating all outcomes as equal (—probability neglect||).
Individuals also seem to place an added value on certainty, preferring to reduce a small risk to zero by
more than they value reducing a larger risk by a greater amount. Errors of commission are viewed as
worse than errors of omission. This can lead to a tilt to the side of inaction.

7 [add references] Strictly speaking, the second and third arguments are not about the actual rate of time
preference, but rather about how factors related to distributional impacts and risk that enter the maximand of the
intertemporal utility calculation affect the implied discounting of future over current returns.

8 This discussion is taken from Kousky et al (2009), which contains references to the relevant behavioral economics
literature.

A-86

-------
strongly affected by the way that information is presented. Thus, individuals may make different choices
for the same decision if it is merely phrased differently (—framing effects||). Choices depend upon the
extent to which a risk evokes feelings of dread. Personal utility also is sensitive to individuals'
perceptions of equity and fairness.

These various behavioral attributes can imply higher or lower values attached to catastrophe risks than
would be implied by expected utility theory. The former would follow from dread or the evaluation of all
catastrophes as roughly equal in likelihood. The latter would follow from optimism bias, or a preference
for reducing small and familiar risks to zero over reducing more substantially an unfamiliar risk - of
which climate change catastrophe certainly is an example. While the direction of bias has to be assessed
empirically, the existence of these various —non-rational || attitudes raises an important but not new
question for evaluating climate change catastrophe risks in setting public policy: if decision makers
believe they have better information than the general public and that they are less subject to emotional
biases, to what extent should their valuation of alternatives supersede those of members of the general
public?

Hypothesis 2: People are Non-Egoistic Expected Utility - Maximizers

A second approach that has been taken in the literature for addressing long-term threats posed by
climate change is to see individuals today, imperfect information and all, as interested in more than
maximizing the discounted present value of their lifetime expected utility streams. One can broadly
define this as altruistic preferences, but this label can cover several different forms of preferences.

A traditional approach to altruistic preferences is to include some measure of next-generation or other
future utility in the preferences of members of the current generation. In this setting, individuals will
weigh the potential costs of a climate change catastrophe in terms of its anticipated impacts on future
welfare, as well as the possibly slight impact on current individuals' egoistic well-being. Consequently,
individuals will derive utility in part from the —bequest they leave to the future in terms of a lowered
(endogenous) risk of a climate change catastrophe. However, there are both theoretical and empirical
reasons to expect individuals to discount the welfare of future generations relative to their own egoistic
welfare. This takes us back to the question previously mentioned in the context of time preference, as to
how powerful an influence this form of altruism might be in the current generation's assessment of risks
of climate change catastrophes.9

A second approach is to depart from a purely utilitarian framework by supposing that individuals see
themselves (or should do) as having a moral obligation to future generations. This mixing of obligations
and conventional utilitarian motivations implies some degree of lexicography in individuals' preferences
- or, critics of utilitarianism might say, an innate failure of the standard economic model to describe
what really motivates people. In this view, if a potential future catastrophe threatens to impose a

9 Current individuals also could believe, as Schelling for example has suggested, that other kinds of bequests to the
future would have higher value; or they could further discount bequests of a less risky climate out of concern that
unless the —chain of obligation || is maintained, something impossible to assure, the sacrifice made today would be
wasted in the future.

A-87

-------
morally inacceptable burden on the future, people will be (or at least can be) motivated to endure
potentially extra-ordinary sacrifices to reduce the threat. The expression of that moral sentiment by
individuals as citizens and stewards, versus utilitarian consumers, would be found through public choice
exercises like voting for tough restrictions on future GHG emissions.

This conception is both stimulating and frustrating, since it does not offer any straightforward way of
assessing how economically significant is the threat of a future climate change catastrophe. Aside from
uncertainty about what the triggering level of threat to the future might be, does one regard current
almost universal reticence to support tough GHG restrictions as due to (correctable) moral failing? Lack
of information? Lack of leadership? The result of rational leadership, because the threat of climate
change is seen as less significant than other threats or because international collective action problems
have not been solved?

A third possible approach that has received less attention is that individuals have preferences that
include some notion of —planetary health || as a global public good. Rather than seek to describe
concern about risks of catastrophe from climate change as deriving only from more fundamental
concerns for intergenerational altruism or fairness, one could posit that individuals derive some direct
benefit from having greater confidence in the ability of planetary systems to remain undisrupted,
without the need to unpack the rationales in terms of future human well-being, satisfaction of moral
sentiment, or a pure existence benefit. This approach allows one to sidestep some of the difficulties
encountered in either the altruistic utilitarian or moral-obligations conceptions. In particular, the
normative approach to setting discount rates can be embedded in a framework of preferences without
having to be an ad hoc add-on.10 However, this does not get around the huge empirical problems in
assessing the value that members of the current generation might place on reducing risks of future
climate change catastrophes.11

Catastrophe Risks and Rational Choice Approaches to Policy

While it is certainly possible to debate the capacity of expected -utility types of analyses to adequately
capture the social opportunity cost of climate change catastrophe threats, it is in cases like this that a
disciplined application of rational-choice based analysis more broadly defined can prove most useful. A
thoughtful, systematic, and transparent weighing of benefits and costs, broadly defined, is at the heart
of such an approach. The presence of —deep|| uncertainty or ignorance about the types and likelihoods
of potential catastrophes means that we must include, in addition to sensitivity analysis on these

10	A fundamental criticism of conventional expected-utility analysis for assessing future climate change risks is that
it combines conventional time-preference considerations in assessing the opportunity cost of reducing threats with
the explicitly ethical question of how much the current generation will feel willing or bound to do in protecting the
future.

11	Ideas like this arise often in literature on environmental stewardship, but I am not aware of many treatments of
the idea in economic terms. One example is the paper by Kopp and Portney [ref to add], who describe a thought
experiment in which individuals value —well being of the future, || and the willingness-to-pay for that value can be
discerned through a stated preference valuation effort. While one can debate the merits of the valuation approach
even in a thought experiment, the concept is very similar to what I am trying to describe here. Unfortunately, the
question of how one would ascertain such valuation remains a barrier to empirical implementation of the concept.

A-88

-------
characteristics, focused analysis of the robustness and flexibility of options in addition to the benefits
and costs. With respect to what seem to be behavioral biases in the assessment of catastrophe risks by
individuals, decision makers must make (and then defend) informed judgments on behalf of those they
serve as to when the seeming biases reflect a high degree of economic risk aversion, or dread, and when
the biases reflect other factors (framing effects, optimism bias, and the like) that can be viewed as
inaccurate comprehension of the tradeoffs involved.

Posner (2005) argues that uncertainty over benefits and costs should not prevent using the basic
structure of cost-benefit analysis for evaluating and comparing options, but that this should be framed
in a —tolerable-windows|| approach. This involves using a range of plausible risk estimates to help
identify levels of spending on reducing risk for which benefits clearly exceed the costs, for which costs
clearly exceed benefits. Policies then can be designed with the goal to remain in this window.12This
approach does not provide or depend on —a number|| for how to evaluate the impacts of potential
future climate change catastrophes. In particular, it does not treat them as largely irrelevant
economically given their low probabilities and long time frames to be realized. Instead it provides
flexibility as to how different considerations about climate change catastrophes are brought into the
assessment, including risk aversion and concerns about future sustainability as well as costs of risk
mitigation, while insisting on transparency and a persuasive argument for how these considerations are
to be addressed.

12 This idea is akin to value-of-information approaches. If one has some confidence in the evaluation of costs of
different policies but great uncertainty about the potential benefits, one could investigate how large the potential
benefits might have to be to make a case for the selection of one set of options over another in a portfolio.
Similarly, if the benefits are reasonably well understood conditional on a catastrophe occurring, but there is
uncertainty about the probability of a catastrophe, then one can ask how large the probability would have to be to
justify a particular portfolio of actions.

A-89

-------
Natural Capital and Intra- Generational Equity in Climate Change

Geoffrey Heal
Columbia University1

Introduction

There are two dimensions of equity that are relevant in an evaluation of the impact of climate change -
inter- and intra-generational. It is the former that has been most discussed in the literature to date - all
of the extensive debate about the choice of a discount rate in climate models is in effect a debate about
intergenerational equity and how to model our concerns about this. And clearly this is very relevant in a
climate context - emissions made today will affect generations not yet born, so that issues of
intergenerational fairness are central to any discussion of climate policy. But intragenerational issues
loom large too: climate change is an external cost imposed largely by rich countries on poor ones, and in
addition there is evidence that in any given country it affects poor people more than rich. This
dimension of climate change has not been extensively discussed.

Climate change affects our stock of natural capital - for example, the IPCC has estimated that by 2100 in
the range of 30-40% of currently extant species may be driven extinct by climate-induced changes in
their ecosystems. This would represent a massive transformation of the biospehere, one unprecedented
in human history. Glaciers and snowfields are also likely to diminish greatly in extent, affecting water
supplies to many regions. Changes like this in our natural capital could have far-reaching consequences,
and these are likely to be felt more by poor than by rich countries, and more by poor than rich groups in
any country (World Bank 2006). So intra-generational equity and natural capital impacts are related: the
latter is likely to reinforce concerns about the former. An important question here is whether some
other form of capital - human, intellectual or physical, can replace natural capital. To the extent that
this is possible, it may be possible to ameliorate some of the intra-generational equity impacts of
climate change. In the notes that follow, I begin to develop some of these points, making suggestions
about how they might be modeled.

Equity and Discounting

As anyone who has spent even a short time on the economics of climate change must be aware, a
central issue is the choice of the pure rate of time preference (PRTP), to be distinguished clearly from

the consumption discount rate (CDR). The PRTP is the 6 in the expression J'u(ct)e Stdt where ct is

aggregate consumption at time t, u is a utility function showing strictly diminishing returns to
consumption and we are summing discounted utility over all remaining time.

1 Prepared for an NSF workshop on The Damages from Climate Change, November 2010. Author's contact details:
Columbia Business School, NY 10027, geoff.heal@gmail.com, www.gsb.columbia.edu/faculty/gheal

A-90

-------
The other discount rate concept, the CDR, is the rate of change of the present value of the marginal

good - and we will turn to the case of multiple goods later - it follows from well-known arguments going
back to Ramsey [1928] (see Heal [2005] for a review) that this is equal to the PRTP plus the rate of
change of consumption times the elasticity of the marginal utility of consumption:

elasticity of the marginal utility of consumption and R(ct) is the rate of change of consumption at time t.

What do these two discount rates mean? The PRTP <5 is the rate at which we discount the welfare of
future people just because they are in the future: it is, if you like, the rate of intergenerational
discrimination. Note that there are at least two reasons why we may wish to value increments of
consumption going to different people differently: one is that they live at different times, which is
captured by <5, and the other is that they have different income levels, which we discuss shortly.2 A PRTP
greater than zero lets us value the utility of future people less than that of present people, just because
they live in the future rather than the present. They are valued differently even if they have the same
incomes. Doing this is making the same kind of judgment as one would make if one valued the utility of
people in Asia differently from that of people in Africa, except that we are using different dimensions of
the space-time continuum as the basis for differentiation.

That an increment of consumption is less important to a rich person than to a poor person has long been
a staple of utilitarian arguments for income redistribution and progressive taxation (see Sen [1973]), and
is almost universally accepted. This is reflected in the diminishing marginal utility of consumption, and
the rate at which marginal utility falls as consumption rises is captured by rj(ct). Equation 1 pulls together
time preference and distributional judgments, or considerations based on inter- and intra-generational
judgments: the rate at which the value of an increment of consumption changes over time, the CDR pt,
equals the PRTP <5 plus the rate at which the marginal utility of consumption is falling. This latter is the
rate at which consumption is increasing over time R(ct) times the elasticity of the marginal utility of
consumption rj(ct).

2 We could also value them differently for all manner of other reasons - differences in nationality, ethnicity, and
proximity either physically or genetically. In general we don't do these things, at least explicitly, which to me makes
it strange that we do explicitly discriminate by proximity in time.

utility of consumption, that is, the rate of change of

e dtdu(ct)
dc,

. For the case of a single consumption

A-91

-------
Equity and Climate Change

As we have just seen, there are two dimensions of equity that are important in the context of climate
change: equity between present and future generations, the aspect that has been most extensively
discussed, and equity between rich and poor countries or groups, both now and in the future - inter-
and intra-generational issues. This second dimension is invisible in aggregative one-good models, which
is one reason why we need a many-good model to talk seriously about climate change. The discussions
below will reinforce the need for some measure of disaggregation in the analysis of the economics of
climate change if we are to grapple with equity issues.

The parameter q the elasticity of the marginal utility of consumption, summarizes our preference for
equality: it determines how fast marginal utility falls as income rises. There are two ways in which this
affects the case for action on climate change.

As q rises, the marginal utility of consumption falls more rapidly. If consumption is growing over time,
then this means that the marginal utility of future generations falls more rapidly with larger values of q
and therefore we are less concerned about benefits or costs to future generations. We are less future-
oriented - the consumption discount rate p is higher - and so place less value on stopping climate
change. So via this mechanism, a stronger preference for equality leads to a less aggressive position on
the need for action on climate change. Preferences for equality and action on climate change are
negatively linked here.

There is another offsetting effect, not visible in an aggregative model. Climate change is an external
effect imposed to a significant degree by rich countries on poor countries. The great majority of the
greenhouse gases currently in the atmosphere were put there by the rich countries, and the biggest
losers will be the poor countries - though the rich will certainly lose as well. Because of this, a stronger
preference for equality will make us more concerned to take action to reduce climate change.

So we have an ambiguous impact of a stronger preference for equity on our attitude towards climate
change. Via the mechanism captured in the formula for the consumption discount rate, equation 1, it
makes us less future oriented - provided consumption is growing. (If consumption were to fall, it would
make us more future oriented, and if consumption of some goods were to rise and that of others to fall,
the effect would be a priori unclear.) And via our concern for the poor countries in the world today it
makes us more future-oriented. Unfortunately, without exception analytical models capture only the
first of these effects. They are aggregative one-sector models or models with no distributive weights and
so their operation does not reflect the second mechanism mentioned above. This explains the really
puzzling and counter-intuitive result that a greater preference for equality in Nordhaus's DICE model
leads to less concern about climate change.

To capture fully the contradictory impacts of preferences for equality on climate change policy, we need
a model that is disaggregated both by consumption goods and by consumers, allowing us to study the
consumption of environmental as well as non-environmental goods and also the differential impacts of
climate change on rich and poor nations.

A-92

-------
Natural Capital and Climate Change

Return to equation (1) for the consumption discount rate. Note that if consumption were falling rather
than rising over time (the latter being the universal assumption in lAMs), then the second term in the
expression for pt would be negative and the CDR could in principle be negative, that is the value of an
increment of consumption could be rising over time rather than falling. We would not be discounting
but doing the opposite, whatever that is. It is not impossible that in a world of dramatic climate change
and environmental degradation, consumption might fall at some point. It is even more likely that some
aspects of consumption, or the consumption of some social groups, would fall while other continue to
rise - recognizing this requires that we treat consumption as a vector of different goods that can be
affected differently by climate change. For an early recognition of this point see Fisher and Krutilla
[1975], who comment that increasing scarcity of wilderness areas may drive up our valuation of them. A
more detailed analysis in the context of a growth model is in Gerlagh and van der Zwaan [2002], who
make the interesting point that with limited substitutability between environmental and manufactured
goods and the growing scarcity of environmental goods, there is likely to be a version of Baumol's
disease - an ever larger portion of income being spent on non-manufactured goods.

Let's follow this line of thought and disaggregate consumption at date t into a vector ct= (clit,c2it,—,Cn,t) of
n different goods. (We will mention briefly later the case in which these are the consumption levels of
different countries or social groups.) Utility is increasing at a diminishing rate in all of these goods and is
a concave function overall. In this case we have to change equation 1 for the consumption discount rate.
Now there is a CDR for each type of consumption and we have n equations like equation 1, with a CDR
for each good i equal to the PRTP plus the sum over all goods j of the elasticity of the marginal utility of
consumption of good i with respect to good j times the growth rate of consumption of good j:

(2)

where pit is the CDR on good i at date t, R(c,-t) is the rate of change of consumption of good i at date t,
and rjij{ct) is the elasticity of the marginal utility of good i with respect to the consumption of good j (see
Heal [2005] for details: thp mnct uprwal frampu/nrk nf thk tvnp ran hp fnund in Malinvaud's classic
paper [1953]). The owr Pij — f ,\t )t the cross elasticities /7,y(ct), /*/',
are zero if the utility fu				 J.Tl		..jve either sign.

As an illustration consider the constant elasticity of substitution utility function

(3)

Here we can think of c as produced consumption and s as natural capital, an environmental stock that
produces a flow of ecosystem services. (See Barbier and Heal for a discussion of this concept [2006] and
the World Bank for a detailed review ~f 4-1—	—•' ~apital in the growth process [2006].) In this

case the cross elasticity of the margiij^c° _|_	Jenn depends on whether c and s are

substitutes or complements. For an «			, „ . ^	_ substitutes and the cross elasticity is

positive, and vice versa.

A-93

-------
Let's test our intuitions on this. Take the case where natural capital and produced consumption are
highly complementary, so that indifference curves are near to right angled and the elasticity a is close to
zero. Then the cross elasticity is negative. This means that if the stock of natural capital is rising then this
reduces the consumption discount rate on the regular good. Conversely if the availability of natural
capital is falling then this raises the consumption discount rate on the consumption good. These results
make sense: because of the assumed complementarity, an increase in the amount of the environmental
good will raise the marginal utility of the consumption good and so tend to lower the consumption
discount rate, and vice versa. Of course, the own elasticity on natural capital is positive so that if the
availability of this good is falling then this will tend to make its own consumption discount rate negative.

inputs to a production process, then there is a lower bound to the isoquants on the energy axis. Similarly
one can argue that certain ecosystem services or products, such as water and food, are essential to
survival and cannot be replaced by produced goods. There are therefore lower bounds to indifference
curves along these axes, implying if the utility function is CES that a < 1.

Consumption goods

Minimum level of sendees
from natural capital

Natural Capital

A-94

-------
The figure illustrates this idea: it shows indifference curves for a two-argument utility function,
consumption of produced goods and of ecosystem services, as in equation 3 above. There is a minimum
level of ecosystem services needed for survival - think of this as water, air, and basic foodstuffs, all of
which are ultimately produced from natural capital. For low welfare levels there is no substitutability
between these and produced goods, so that indifference curves are close to right angled. At higher
welfare levels where there are abundant amounts of both goods there is more scope for substitution.
Taken literally, this implies that the elasticity of substitution is not constant but depends on and
increases with welfare levels. This of course is not reflected in the CES function such as 3. A function
with these properties is

which is simply the CES function we noted before, with the zero of the ecosystem service axis
transformed by s > 0. Utility is not defined for s> s. Relative to the transformed origin (e,0) there is still a
constant elasticity of substitution a but relative to (0,0) the elasticity is not constant. For a > 1, every
indifference curve, every welfare level, can be attained with only s of ecosystem services, whereas with
a < 1 greater welfare levels require greater levels of ecosystem services (and of consumption goods).

These ideas can be applied to modeling equity: it is generally recognized that poor countries, or poor
groups within countries, are more dependent on natural capital and its services than are richer groups
(World Bank [2006]). They have less capacity to substitute alternative goods for the services of natural
capital and so show more complementarity between natural capital and other goods. In terms of the
figure, their indifference curves are lower and closer to being right angled. This means that they have
different consumption discount rates from other groups: if the stock of natural capital is falling then
they will have higher consumption discount rates on the common consumption good. In this sense they
will appear to be more impatient. Of course as noted above their discount rate on natural capital will be
negative, so we will have the paradox of an apparently impatient group - with respect to the
consumption good - being willing to invest for low returns in natural capital.

A Sterner Perspective

It's worth looking in more detail at the Sterner and Persson development of this point [2007], They talk
about the effect of changes in relative prices rather than consumption of produced and environmental
goods, but the point is the same. If we consume both produced goods and the services of the
environment, as in the utility function 3, then we can expect that with climate change environmental
services will become scarce relative to produced goods and therefore their price will rise relative to that
of produced goods (the "environmental Baumol disease" that Gerlagh and van der Zwaan refer to
[2002]). Consequently the present value of an increment of environmental services may be rising over
time, and the consumption discount rate on environmental services may thus be negative, precisely the
point that we were making in equation 2 above. This could be the case even with a high PRTP, which is
the main point of the Sterner and Persson paper. They also present an interesting modification of
Nordhaus's DICE model to incorporate this point. They replace the standard utility function, which is an

(4)

A-95

-------
isoelastic function of aggregate consumption, by a CES function along the lines of equation 3 above, but
modified to reflect a constant relative risk aversion:

They assume that the supply of environmental services is negatively affected by temperature according
to the square of temperature, and that the share of environmental goods in consumption is about 20%,
use these assumptions to calibrate the modified DICE model and and then run the model with the PRTP
used by Nordhaus. Their runs show that even with such a high PRTP the presence of an environmental
stock that is damaged by higher temperatures radically transforms the optimal emissions path of CO and
leads to a vastly more conservative policy towards climate change, with emissions both staying lower
and falling faster. In fact it leads to a more aggressive reduction in greenhouse gases than recommended
by the Stern Review.

Natural Capital and Production

I have emphasized so far that natural capital can affect human welfare directly, and needs to be thought
of as an argument of the welfare function. Natural capital also affects a nation's production possibilities:
I mentioned above changes in hydrology such as melting of glaciers and reduction in winter snowfields,
both of which are already in evidence and are affecting agriculture in some regions. They will affect it
further over the coming decades. This is quite separate from any impact that changes in temperature
and precipitation may have on agriculture. Other changes in natural capital will probably affect
agriculture - changes in species abundance and distribution, for example, can affect whether birds and
insects pollinate crops.

Modeling Different Groups

I commented above that equation 2 can be given a different interpretation: instead of

the subscripts i and j referring to different goods, they can be taken as referring to the amounts of a
single good consumed by different groups - these could be social groups within a country or they could
be different countries. I this case we have different consumption discount rates for each group's
consumption, and the elasticities now indicate how the marginal valuation of consumption by one group
depends on the consumption levels of others. Do we value on increment of consumption to the poor
more if everyone else is very rich than if most others are also poor? Presumably the answer to this is
yes, but these are issues that have not featured at all in the discussions to date.

The elasticity of the marginal utility of consumption plays a central role in much of our discussion.
Unfortunately this variable plays two roles in our models: it expresses our distributional preferences,
which is the way we have been using it here, and it also expresses our aversion to risk. Most empirical
estimates of the value of q come from studies of behavior in the face of risk, but it seems clear that
these two interpretations of q are really quite different, and that our aversion to risk tells us little if

(2)

Choosing rf

A-96

-------
anything about our preferences for income equality. Given this, we need to find a way of expressing
preferences that does not conflate distributional and risk preferences. Recursive formulations such as
that of Kreps and Porteus are relevant here.

References

Barbier, Edward and Geoffrey Heal 2006. " Valuing Ecosystem Services," The Economists' Voice, Berkeley
Press, January 2006.

Berry, Stephen, Geoffrey Heal and Peter Salamon 1978. " On a Relationship between Economic and
Thermodynamic Optima," Resources & Energy, vol. 1, pp. 125-137

Dasgupta, Partha and Geoffrey Heal 1979. Economic Theory and Exhaustible Resources, Cambridge
University Press.

Fisher, Anthony and John Krutilla 1975. " Resource Conservation, Environmental Preservation and the
Rate of Discount," Quarterly Journal of Economics Vol. 89 No. 3 August 1975, 358-370.

Gerlagh, Reyer and Robert van derZwaan, 2002," Long-Term Substitutability between the Environment
and Man-Made Goods," Journal of Environmental Economics and Management, 44: 329-45.

Guesnerie, Roger 2004." Calcul economique et development durable," La Revue Economique.

Heal, Geoffrey 2005." Intertemporal Welfare Economics and the Environment," Handbook of

Environmental Economics, Volume 3.Edited by K-G Maler and J.R. Vincent, Elsevier, Chapter 21,
1105-1145.

Kreps David and Evan Porteus 1978 "Termporal resolution of uncertainty and dynamic choice theory"
Econometrica 46(1) 185-200

Malinvaud, Edmond 1953." Capital accumulation and the efficient allocation of resources"

Econometrica Vol 21 No. 2 April 1953

Nordhaus, William 1993." Rolling the DICE: An Optimal Transition for Controlling the Emission of
Greenhouse Gases," Resource and Energy Economics, 15: 27-50.

Ramsey, Frank 1928. " A mathematical theory of saving," Economic Journal, 38: 543-559.

Sen, Amartya 1973. On Economic Inequality, Clarendon Press, Oxford.

Sterner, Thomas and Martin Persson 2007." An Even Sterner Review: Introducing Relative Prices into
the Discounting Debate,", Discussion Paper, Resources for the Future, July 2007, RFF DP 07-37.
Available at http://www.rff.org/rff/Documents/RFF-DP-07-37.pdf

World Bank 2006. Where is the Wealth of Nations? Measuring Capital in the 21st Century. The World
Bank, Washington D.C.

A-97

-------
Managing Climate Risks

Roger M. Cooke
Resources for the Future

Department of Mathematics, Delft University of Technology

Carolyn Kousky
Resources for the Future

Many Integrated Assessment Models (lAMs) maximize the present value of consumption, equating the
marginal benefits of abatement in terms of reduced climate damages with the marginal costs of
reducing emissions. Every trader, banker, and investor knows that maximizing expected gain entails a
trade-off with risk. According to the theory of rational decision, preferences can always be represented
as expected utility, hence from this viewpoint, any aversion to risk could be folded into the rational
agent's utility function. This theory, recall, applies to rational individuals; groups of rational individuals
do not comply the axioms of rational decision theory. The fact is that 'professional risk taking
organizations' do manage risk, and not by bending the utility function of a representative consumer.
Rather, they employ techniques like value at risk, and optimize expected gain under a risk constraint.
Managing risk is a problem of group decision.

Weitzman (2009) has recently called attention to the risks of climate change, arguing that current
approaches court probabilities on the order of 0.05~0.01 of consequences that would render life as we
know it on the planet impossible. What is the plan to manage this "tail risk"? Risk management shifts
the research question from 'how does the optimal abatement level change for different parameter
values?' to 'how does our policy choice fare under the range of potential future conditions and how can
we buy down the risk of catastrophic outcomes?' As such, it places the quantification of uncertainty in
the foreground. Uncertainty quantification is more than a modeler putting distributions on his/her
model's parameters. The antecedent question reads: 'is it the right model? What is the model
uncertainty?' Failing a definitive answer to that question, stress testing our current models for their
ability to handle tail risks, and exploring canonical model variations are essential steps prior to
quantifying uncertainty on parameters. Gone are the days when quantification of the uncertainties was
left to the modelers themselves; at the state of the art, quantification is done by structured expert
judgment in a rigorous and transparent manner.

Stress Testing

Stress testing is preformed to check that models remain realistic and capture the relevant possibilities
when their parameters are given extreme values. Many lAMs specify economic damages as a function of
temperature change, and model their impact on output and utility. For example, damages at time t
induced by temperature change T(t) from pre-industrial mean temperature are represented in DICE as
factor that reduces economic output: 1/[1 + 0.0028388T(t)2]. The standard Cobb Douglas production
function expresses output as a function of total factor productivity, capital stock and labor. Capital
depreciates at rate 10%, and is augmented by savings (in the DICE "Base" case the savings rate is

A-98

-------
optimized with damages set equal to zero, then damages are reinstated). Temperature induced
damages and abatement efforts reduce output. Setting damage and abatement equal to zero, an
illustrative stress test of the Cobb Douglass model with constant population, constant total factor
productivity and DICE values for other parameters is shown in Figure 1. Four output trajectories with
initial capital ranging from 10 times the DICE value ($1800 Trill) to $100 ($1.6xl0~8 for each inhabitant).
The limiting capital value is independent of the starting values - with a vengeance: the four trajectories
are effectively identical after 60 years. Such obviously unrealistic consequences underscore the need for
circumscribing the empirical domain of application of these simple models. Put the factories and
laborers on the Moon and they will produce nothing; other things are involved. Regardless whether the
model adequately describes small departures from an equilibrium state, its use for long term projections
inevitably entails this sort of behavior and putting uncertainty distributions on the model's parameters
will not change that.

Figure 1. Output gross of abatement cost and climate damage ($trill 2000 USD) Base case, no temperature
damage, no abatement, constant population, constant total factor productivity (0.0307951), initial output from

A second stress test examines the effect of adding temperature induced economic damages, again
without abatement. With $180 Trill initial capital, we assume that temperature increases linearly,
leaving other parameters as in the previous case. Figure 2 shows four economic output trajectories,
corresponding to temperature increases of 0, 5, 10, and 15 degrees Celsius in 200 years.

Figure 2. Output after damages before abatement, initial capital = 180 $trill, constant population, constant
productivity, no abatement, temperature in 200 yr (linear increments)

A-99

-------
No scientist claims that life as we know it could exist with 10°C global warming. With a steady
temperature rise leading to 10°C above pre-industrial levels in 200 years, this model predicts that
output would be reduced to 68.% of its value without temperature rise. Such projections seem a bit
sanguine. The essential feature is that climate induced damages hit only economic output; as a result
capital can never decrease faster than its natural depreciation rate, and this rate of decrement is
reached only for infinite temperature. Again, putting uncertainty on other model parameters may cloud
this picture, but will not change this feature.

Canonical Model Variation

It is often noted that simple models like the above cannot explain large differences across time and
geography between different economies, pointing to the fact that economic output depends on many
factors not present in such simple models. To "save the phenomena" researchers have proposed
enhancing the basic model with inter alia social infrastructure, government spending, human capital,
knowledge accretion, predation and protection, extortion and expropriation (see Romer (2006), chapter
3). Before proliferating this model, however, it is well to reflect on its fundamental assumptions about
damage, capital and output. Could different model types with comparable prime facie plausibility result
in macroscopically different behavior?

We illustrate with one variation based on the following simple idea: Gross World Production
(GWP[trillion USD 2005] ) produces pollution in the form of greenhouse gases; pollution, if unchecked,
will ultimately destroy necessary conditions for production. This simple observation suggests that Lotka
Volterra type models might provide a perspective which an uncertainty analysis ought not rule out. The
quantity of anthropogenic greenhouse gases in the atmosphere at year t, GHG(t) [ppm C02], is the
amount in the previous year, less what has decayed at a rate, say, 0.0083, plus any new emissions in
time period t. Assume that new emissions are a fixed fraction, say, 0.024 of GWP (Kelly and Kohlstadt
2001). Different values can be found in the literature, but these are representative. Real GWP has
grown at an annual rate of 3% over the last 48 years (this includes population growth); assume that this
growth is decreased by a damage function D of temperature T, and ultimately of GHG, this gives the
following system:

(1)	GHG(t+l) = (l-0.0083)GHG(t) + 0.024xGWP(t).

(2)	GWP(t+l) =[1+0.03- D(T(GHG(t)))]GWP(t).

If D were linear in GHG, this would be a simple Lotka Volterra type system. With cs as the climate
sensitivity and 280 ppm the pre-industrial level of greenhouse gases, equilibrium temperature follows
T(GHG(t)) = csx ln(GHG(t)/280)/ln(2). Adopting Weitzman's (2010) notion of a "death temperature" of
18°C we write damages as D(GHG)(t) = (T/18)2. Anthropogenic greenhouse gases increase with
production; if GWP(t) were constant, they would increase to a constant 0.024xGWP/0.0083 However, as
GWP increases, GHGs and temperature keep rising as well, lowering the growth rate of GWP. When D >
0.03, GWP starts decreasing. Eventually 0.024xGWP < 0.0083, and then greenhouse gases start
decreasing, reducing damages to a point where production can start growing again. Figure 2 shows

A-100

-------
GWP and GHG as functions of time out to 500 yrs, with all variables at their nominal values. GWP
collapses. Greenhouse gases also collapse, but not to their initial level; hence the next upswing in GWP
is attenuated. A steady state is eventually reached after some 1,500 years. This is not offered as a
plausible model, its role is to spotlight the fundamental modeling assumptions. Evidently, different ways
of modeling the impact of climate change damages give qualitatively different predictions, and steady
state values may not be relevant for current policy choices. Neither theoretical nor empirical evidence
exclude the Lotka Volterra type of interaction between damages and production presented here. A
credible uncertainty analysis should fold in this and other possibilities, which brings us to the next point
of examining a range of future conditions for a given policy choice.

Figure 3: The impact of climate damages on GWP (left) and greenhouse gases (right)

gwp gtig

Structured Expert Judgment for Quantifying Uncertainties

Uncertainty analysis with climate models must be informed by the broad community of climate experts -
not simply the intuitions or proclivities of modelers - through a process of structured expert judgment.
Experience teaches that independent experts will not necessarily buy into the models whose parameter
uncertainties they are asked to quantify. Hence, experts must be queried about observable phenomena,
results of thought-experiments if you will, and their uncertainty over these phenomena must be 'pulled
back' onto the parameters of the model in question. This process is analogous to the process by which
model parameters would be estimated from data, if there were data. The new wrinkle is that data are
replaced by experts' uncertainty distributions on the results of possible, but not actual, measurements.
The 'pull back' process is called probabilistic inversion, and has been developed and applied extensively
in uncertainty analysis over the last two decades (see Cooke and Kelly 2010 and references therein). In
general, an exact probabilistic inverse does not exist, and the degree to which a model enables a good
approximation to the original distributions on observables forms an important aspect of model
evaluation. Four features of the structured expert judgment approach deserve mention: (i) Experts are
regarded as statistical hypotheses, and their statistical likelihood and informativeness are assessed by
their performance on calibration questions from their field whose true values are known post hoc. (ii)
Experts' ability to give statistically accurate and informative assessments is found to vary considerably,
(iii) Experts' uncertainty assessments are combined using performance based weights, (iv) Dependence,
either assessed directly by experts or induced by the probabilistic inversion operation, is a significant
feature of an uncertainty analysis.

A-101

-------
When uncertainty has been quantified in a traceable and defensible manner, an ensemble of possible
futures for each policy choice may be generated. Figure 4 shows 30 Lotka Volterra temperature
trajectories out to 200 years, with BAU emissions at 2.4% GWP (left) and stringent emissions at 1.5% of
GWP (right); and using representative distributions for uncertain variables. Employing a value at risk
management strategy, we would search for an emissions path optimizing consumption while holding the
probability of exceeding a stipulated temperature threshold below a tolerable threshold.

Figure 4: Possible temperature trajectories under (left) emissions at 2.4%GWP and (right) emissions at 1,
GWP (right)

8%

¦:5<

IE0002 1 2E0002;1 .46000^1 6E0002 1 8E00C2 2EC002



These reflections challenge us to deploy risk management strategies on a global scale. We suggest this
begin with (i) stress testing models, (ii) exploring alternative models, and (iii) quantifying uncertainty in
such models via structured expert judgment. We are condemned to choose a climate policy without
knowing all the relevant parameters, but we are not condemned to ignore the downside risks of our
choices.

References

Cooke, R. M. and G. N. Kelly (2010). Climate Change Uncertainty Quantification: Lessons Learned from

the Joint EU-USNRC Project on Uncertainty Analysis of Probabilistic Accident Consequence Code.
Resources for the Future Discussion Paper 10-29. Washington, D.C., Resources for the Future,

Kelly, D.L. and Kohlstadt C.D. (2001) "Malthus and climate chantge: betting on a stable population" J. of
Environmental Economics and Management 41, 135-161.

Romer, D. (2006) Advanced Macroeconomics McGraw Hill Irwin, Boston.

Weitzman, M. (2010). GHG Targets as Insurance Against Catastrophe Climate Damages. Cambridge, MA,
Harvard University.

Weitzman, M. L. (2009). "On Modeling and Interpreting the Economics of Catastrophic Climate Change."
Review of Economics and Statistics 91(1): 1-19.

A-102

-------
November 18-19, 2010

Workshop Agenda

MODELING CLIMATE CHANGE IMPACTS AND ASSOCIATED ECONOMIC DAMAGES

November 18. 2010

Workshop Introduction

8:30-8:35 Welcome and Introductions

Elizabeth Kopits, U.S. Environmental Protection Agency

8:35-9:00 Opening Remarks

Bob Perciasepe, Deputy Administrator, U.S. Environmental Protection Agency
Steve Koonin, Under Secretary for Science, U.S. Department of Energy

9:00 - 9:25 Progress Toward a Social Cost of Carbon

Michael Greenstone, Massachusetts Institute of Technology

Session 1: Overview of Existing Integrated Assessment Models
Moderator: Stephanie Waldhoff, U.S. Environmental Protection Agency

9:25 - 9:50 Overview of Integrated Assessment Models

Jae Edmonds, Pacific Northwest National Laboratory

Models Used for the Development of Current USG SCC Values

9:50-10:15 DICE

Steve Newbold, U.S. Environmental Protection Agency

10:15-10:40 PAGE

Christopher Hope, University of Cambridge

10:40-10:55 Break

10:55-11:20 FUND

David Anthoff, University of California, Berkeley

Representation of Climate Impacts in other Integrated Assessment Models

11:20-11:45 GCAM (JGCRI - UMD/PNNL] and Development ofiESM (PN N L/LBN L/ORN L)
Leon Clarke, Pacific Northwest National Laboratory

-------
11:45-12:10 IGSM (MIT)

John Reilly, Massachusetts Institute of Technology

12:10-12:40 Discussion

12:40 - 1:40 Lunch (on your own; a list of nearby restaurants is provided in the
registration packets)

Session 2: Near-Term DOE and EPA Efforts

Moderator: Ann Wolverton, U.S. Environmental Protection Agency

1:40 - 2:00 Proposed Impacts Knowledge Platform
Bob Kopp, U.S. Department of Energy
Nisha Krishnan, Resources for the Future

2:00-2:20 Proposed Generalized Modeling Framework

Alex Marten, U.S. Environmental Protection Agency

2:20-2:40 Discussion

Session 3A: Critical Modeling Issues in Assessment and Valuation of Climate Change
Impacts

Moderator: Ann Wolverton, U.S. Environmental Protection Agency

2:40 - 3:10 Sectoral and Regional Disaggregation and Interactions

Ian Sue Wing, Boston University

3:10-3:20 Break

3:20-3:50 Adaptation and Technological Change

Ian Sue Wing, Boston University

3:50-4:20 Multi-century Scenario Development and Socio-Economic Uncertainty

Brian O'Neill, National Center for Atmospheric Research

4:20-5:00 Discussion

2

-------
November 19. 2010

Day 2 Introduction

8:30-8:40 Welcome; Recap of Day 1; Overview of Day 2

Elizabeth Kopits, U.S. Environmental Protection Agency

Session 3B: Critical Modeling Issues in Assessment and Valuation of Climate Change
Impacts (cont.)

Moderator: Bob Kopp, U.S. Department of Energy

8:40-9:10 Incorporation of Climate System Uncertainty into IAMs

Gerard Roe, University of Washington

9:10-9:40 Extrapolation of Damage Estimates to High Temperatures: Damage
Function Shapes

Marty Weitzman, Harvard University

9:40-10:10 Earth System Tipping Points

Tim Lenton, University of East Anglia

10:10-10:30 Break

10:30-11:00 Potential Economic Catastrophes

Michael Toman, World Bank

11:00-11:30 NonmarketImpacts

Michael Hanemann, University of California, Berkeley

11:30-12:30 Discussion

12:30-1:30 Lunch (on your own; a list of nearby restaurants is provided in the
registration packets)

Session 4: Implications for Climate Policy Analysis and Design

Moderator: Charles Griffiths, U.S. Environmental Protection Agency

1:30-2:00 Implications for Design and Benefit-Cost Analysis of Emission Reduction
Policies

Ray Kopp, Resources for the Future

3

-------
2:00-2:30 Implications for Addressing Equity and Natural Capital Impacts

Geoff Heal, Columbia University

2:30-3:00 Implications for Choice of Policy Targets for Cost-Effectiveness Analysis

Nat Keohane, Environmental Defense Fund

3:00-3:10 Break

3:10-3:40 Implications for Managing Climate Risks

Roger Cooke, Resources for the Future

3:40-4:15 Discussion

Session 5: Workshop Wrap-up

4:15-4:30 Summary Comments by U.S. Department of Energy

Rick Duke, Deputy Assistant Secretary for Climate Policy

4:30-4:45 Summary Comments by U.S. Environmental Protection Agency

A1 McGartland, Director of the National Center for Environmental Economics

4

-------
Estimating the Social Cost of Carbon for the United States Government

Michael Greenstone
3M Professor of Environmental Economics
Massachusetts Institute of Technology
November 2010

The climate is a key ingredient in the earth's complex system that sustains human life and well
being. According to the United Nation's Intergovernmental Panel on Climate Change (IPCC),
the emissions of greenhouse gases (GHG) due to human activity, large the combustion of fossil
fuels like coal, is "very likely" altering the earth's climate, most notably by increasing
temperatures, precipitation levels and weather variability. Without coordinated policy around
the globe, state of the art climate models predict that the mean temperature in the United States
will increase by about 10.7° F by the end of the century (Deschenes and Greenstone 2010).
Further, the distribution of daily temperatures is projected to increase in ways that pose serious
challenges to well being; for example, the number of days per year where the typical American
will experience a mean (average of the minimum and maximum) temperature that exceeds 90° F
is projected to increase from the current 1.3 days to a 32.2 days (ibid). The especially troubling
statistic is that the hottest days pose the greatest threat to human well being.

It appeared that the United States and possibly the major emitters were poised to come together
to confront climate change by adopting a coordinated set of policies that could have included
linked cap and trade systems. However, the failure of the United States Government to institute
such a system and the non-binding commitments from the Copenhagen Accord seem to have
placed the all at once solution to climate change out of reach for at least several years.

Instead, the United States and many other countries are likely to pursue a series of smaller
policies all of which aim to reduce GHG emissions but individually have a marginal impact on
atmospheric concentrations. These policies will appear in a wide variety of domains, ranging
from subsidies for the installation of low carbon energy sources to regulations requiring energy
efficiency standards in buildings, motor vehicles, and even vending machines to rebates for

-------
home insulation materials. Although many of these policies have other goals, their primary
motivation is to reduce GHG emissions. However, these policies reduce GHG emissions at
different rates and different costs.

In the presence of this heterogeneity and nearly limitless set of policies that reduce GHG
emissions, how is government to set out a rational climate policy? The key step is to determine
the monetized damages associated with an incremental increase in carbon emissions, which is
referred to as the social cost of carbon (SCC).1 It is intended to include (but is not limited to)
changes in net agricultural productivity, human health, property damages from increased flood
risk, and the value of ecosystem services.2 Monetized estimates of the economic damages
associated with carbon dioxide emissions allows the social benefits of regulatory actions that are
expected to reduce these emissions to be incorporated into cost-benefit analyses.3 Indeed as the
Environmental Protection Agency begins to regulate greenhouse gases under the Clean Air Act,
the SCC can help to identify the regulations where the net benefits are positive.

The United States Government (USG) recently selected four SCC estimates for use in regulatory
analyses and has been using them regularly since their release. For 2010, the central value is $21
per ton of C02 equivalent emissions.4 The USG also announced that it would conduct
sensitivity analyses at $5, $35, and $65. The $21, $5, and $35 values are associated with
discount rates of 3%, 2.5%, and 5%, reflecting that much of the damages from climate change
are in the future. The $65 value aims to represent the higher-than-expected impacts from
temperature change further out in the tails of the SCC distribution. In particular, it is the SCC
value for the 95th percentile at a 3 percent discount rate. These SCC estimates also grow over
time based on rates endogenously determined within each model. For instance, the central value
increases to $24 per ton of CO2 in 2015 and $26 per ton of CO2 in 2020.

1 Under Executive Order 12866, agencies in the Executive branch of the U.S. Federal government are required, to
the extent permitted by law, "to assess both the costs and the benefits of the intended regulation and, recognizing
that some costs and benefits are difficult to quantify, propose or adopt a regulation only upon a reasoned
determination that the benefits of the intended regulation justify its costs."

2 All values of the SCC are presented as the cost per metric ton of C02 emissions.

3 Most regulatory actions are expected to have small, or "marginal," impacts on cumulative global emissions,
making the use of SCC an appropriate measure.

4 All dollar values are expressed in 2007 dollars.

-------
I was involved in the interagency process that selected these values for the SCC and this talk
summarizes these efforts.5 The process was initiated in 2009 and completed in February 2010.
It aimed to develop a defensible, transparent, and economically rigorous way to value reductions
in carbon dioxide emissions that result from actions across the Federal government. Specifically,
the goal was to develop a range of SCC values in a way that used a defensible set of input
assumptions, was grounded in the existing literature, and allowed key uncertainties and model
differences to transparently and consistently inform the range of SCC estimates used in the
rulemaking process.

The intent of this lecture is to explain the central role of the social cost of carbon in climate
policy, to summarize the methodology and process used by the interagency working group to
develop values, and to identify key gaps so that researchers can fill these gaps. Indeed, the
interagency working group explicitly aimed the current set of SCC estimates to be updated as
scientific and economic understanding advances.

5 This process was convened by the Council of Economic Advisers and the Office of Management and Budget, with
regular input from other offices within the Executive Office of the President, including the Council on
Environmental Quality, National Economic Council, Office of Energy and Climate Change, and Office of Science
and Technology Policy. Agencies that actively participated included the Environmental Protection Agency, and the
Departments of Agriculture, Commerce, Energy, Transportation, and Treasury.

-------
PROGRESS TOWARD A SOCIAL COST OF
CARBON

Michael Greenstone

3M Professor of Environmental Economics
Massachusetts Institute of Technology

Improving the Assessment and Valuation of Climate Change
Impacts for Policy and Regulatory Analysis

November 18, 2010

-------
OUTLINE

— - —-—- — —		—

Background & Motivation

Social Cost Of Carbon (SCC)

How is the SCC Calculated?

Lifetime Damages of a Ton of C02 Emissions

Results

Limitations

Conclusions

2

-------
I. BACKGROUND & MOTIVATION

-------
RISING TEMPERATURES	

• Human-induced C02 emissions will likely
cause temperature increases

-------
RISING TEMPERATURES

• Global temperatures projected to increase by
18% between 2000 and 2100

CD
i_rj

CO

aj

ra

oj
o_

E

CVJ

-\

"1

A

Projected Mean Annual Temperature, CCSM Model

World, 2000-2099

„ 				

2000	2020	2040	2060	2080	21 OO

Year

Source: Community Climate System Mociel (CCSM) 3. National Center for Atrr»ospheric Research

-------
CURRENT AND PREDICTED CHANGE IN DISTRIBUTION OF
TEMPERATURE FOR 2070-2099, UNITED STATES

80
60
40
20
0
-20
-40

~ 1968-2002 Average ~ Predicted Change, Hadley 3-A1 Fl, Error-Corrected





















































m













_



o

CM

D?

o

1°

V

y.

40 40-W 5ol 6o!zol 7oU 80-90

i

>90





Distribution of Annual Daily Mean Temperatures (F)

-------
U.S. LEGISLATION LANDSCAPE	

House passed Waxmari-Markey cap-and-trade
bill

Senate declined to pursue legislation
Best case in next several years:

Renewable electricity standards
More subsidies for nuclear power

7

-------
CLEAN AIR ACT

EPA has finalized a "tailoring" rule for
Greenhouse Gases (GHG) under the Clean Air
Act to take effect in January 2011

Set Rules that Govern Behavior of 900 Largest
Sources

Statute Requires Use of "Best Available Control
Technology"

Likely to Be Numerous Court Cases

-------
CLEAN AIR ACT

Likely Impact of Clean Air Act Regulations

Reduce GHG Emissions by 5-12% in 2020, relative
to 2005. President Promised 17% in Copenhagen

-------
CLEAN AIR ACT
Will these Regulations have Net Benefits?

A regulatory impact analysis (RIA) will be required
and informs the public of the relative costs and
benefits of this mandate

+ Analyses will use the "social cost of carbon" to
monetize the benefits stemming from C02 reduction

10

-------
II. SOCIAL COST OF CARBON 	

II O AP iwP Am\0' jo| Aw I / R ^ j >|

11

-------
A. DEFINITION

SCC: monetized damages associated with ari
incremental increase in carbon emissions in a
given year

It includes but is not limited to changes in:

Net agricultural productivity
Human health

Property damages from increased flood risk
The value of ecosystem services

12

-------
B. SCC IN ACTION

Up-front Technology Costs and Social Benefits of EPA/DOT GHG
Emissions Standards for Light-Duty Trucks 2010-2050 (NPV 3%
Discount Rate and 2007 Dollars)



2007 $s



Social Benefits

$277.5



Costs -$345.9

Net Benefits, without

SCC

-$68.4



Social Benefits of C02
Reductions (Central
Value)

Total Net Benefits

-------
B. SCC IN ACTION

Up-front Technology Costs and Social Benefits of EPA/DOT GHG
Emissions Standards for Light-Duty Trucks 2010-2050 (NPV 3%
Discount Rate and 2007 Dollars)

2007 $s

Social Benefits

$277.5

Costs

-$345.9

Net Benefits, without

SCC

-$68.4

Social Benefits of C02
Reductions (Central
Value)

$176.7

Total Net Benefits

$108.3

-------
III. HOW IS THE SOCIAL COST OF CARBON
CALCULATED?

15

-------
ESTIMATING SCC

A USG interagency working group developed a
transparent and economically rigorous way to
estimate SCC

Now will Summarize Some of the Key Decisions
and Results. (USG Plans to Revisit as Science
Advances)

16

-------
III. HOW IS SOCIAL COST OF CARBON
CALCULATED

A. INTEGRATED ASSESSMENT MODELS

17

-------
A. INTEGRATED ASSESSMENT MODELS (IAMS)

lAMs combine Climate Processes, Economic Growth, and Feedbacks
between the Climate and the Global Economy into a single model

Specifically, 1AM translate changes in C02 emissions into economic
damages

1.Emissions

[assumptions about GDP and population growth]

2.	Emissions -> Atmospheric GHG Concentrations
[based on carbon cycle]

3.	GHG Concentrations -> Changes in Temperature
[assumptions about climate model and climate sensitivity]

4.	Temperature -> Economic Damages (market and non-market)
[assumptions about damage functions]

18

-------
A. INTEGRATED ASSESSMENT MODELS (IAMS)

			 _ - - 			 - --—			—-—	— -		—-———		 V-- 	¦	

Benefit of these Models is that they Answer
Everything

-------
A. INTEGRATED ASSESSMENT MODELS (IAMS)

	_____ __ — - ——	—-———— ^ •—		 — — — x 	— ^

Benefit of these Models is that they Answer
Everything

Cost of Models is that they Answer Everything

-------
A. INTEGRATED ASSESSMENT MODELS (IAMS)

	_____ __ — - ——	—-———— ^ •—		 — — — x 	— ^

Benefit of these Models is that they Answer
Everything

Cost of Models is that they Answer Everything
Highly Dependent on Validity of Assumptions

-------
A. INTEGRATED ASSESSMENT MODELS (IAMS)

		 • 		 —	—-——-——			\	— *—

Relied on three commonly used lAM's to
estimate SCC:

FUND (Richard Tol)

DICE (William Nordhaus)

PAGE (Chris Page)

All 3 are frequently cited in the peer-reviewed
literature and used in the IPCC assessment

Each model is given equal we ghtto determine
the SCC values

-------
DAMAGE FUNCTIONS

Figure 1: Annual Consumption Loss as a Fraction of Global GDP in 2100 Due to an Increase in Annual

Global Temperature in the DICE, FUND, and PAGE models

G.

a

-
U

XI
_o

DC

T3

T3
Q

0,30

0.2S -

0.20

0.15

£ 0.10

UT-

Q

o.os

0.00

¦DlCEiOO?
PACE Sth%
'PAGE mein
PAGE 9Sth%
¦FUNEHC5-3)

-O.OS -

Tempesture change [deg C]

-------
III. HOW IS THE SOCIAL
CALCULATED?

B. ASSUMPTIONS

¦ 1	IAII	I	I Mi

COST OF CARBON

<"-*¦ -«*> ip i r\\j | j ir^ f\ I a

24

-------
1. SOCIO-ECONOMIC & EMISSIONS TRAJECTORIES

.	^	^ ^ 0' ^ • r h - i • - —r,					1—r — - — ¦ r i~r,		 -	|	1 ^ 1 I L I I *"V

Socio-economic pathways are closely tied to
climate damages

More and wealthier people tend to emit more GHG
Higher WTP to avoid climate disruptions

For this reason, decisions necessary for
several input parameters from present until
2100:

Global GDP
Global Population
Global C02 emissions

-------
1. SOCIO-ECONOMIC & EMISSIONS TRAJECTORIES

.	^ »**V I -«~s	1	- r h ¦ i - - --rt					1—r — - — ¦ r i~r,		 -	|	1 ^ 1 I L I I *"V

Relied on the Stanford Energy Modeling Forum
exercise, EMF-22

Based on 4 of 10 models
Key advantage:

GDP, population and emission trajectories are internally
consistent

Five trajectories selected:

4 business-as-usual (BAU) paths

Correspond to 2100 concentrations of 612 - 889 ppm,
reflecting differences in assumptions about cost of low carbon
energy sources

1 iower-than-BAU path

Achieves stabilization at 550 ppm in 2100

-------
2. EQUILIBRIUM CLIMATE SENSITIVITY

Equilibrium climate sensitivity (ECS): long-term

increase in the annual global-average surface
temperature due to a doubling of atmospheric
C02 concentration relative to pre-industrial
levels

Equivalent to the atmospheric C02 concentration
stabilizing at about 550 parts per million (ppm)

27

-------
2. EQUILIBRIUM CLIMATE SENSITIVITY

According to the Fourth Assessment Report of
the Intergovernmental Panel on Climate
Change (IPCC):

We conclude that the global mean equilibrium
warming for doubling C02... is likely to He in the

range 2°C to 4.5 °C.	with

about 3 °C. Equilibrium climate sensitivity is very
likely larger than 1.5 °C.... For fundamental
physical reasons as well as data limitations, values
substantially higher than 4.5 °C still cannot be
excluded, but agreement with observations and
oroxv data is generally worse for those high values
than for values in the 2 °C

-------
2. EQUILIBRIUM CLIMATE SENSITIVITY

Selected four candidate probability distributions
and calibrated them to the IPCC statement:

Roe and Baker (2007)

Log-normal

Gamma

Weibull

Calibration done by applying three constraints:

Median equal to 3°C

Two-thirds probability that ECS lies between 2 and
4.5 °C

Zero probability that ECS is less than 0°C or greater
than 10 °C

-------
2. EQUILIBRIUM CLIMATE SENSITIVITY

Table 1: Summary Statistics for Four Calibrated Climate Sensitivity Distributions



Roe &
Baker

Log-normal

Gamma

Weibull

Pr(ECS < 1.5°C)

0.013

0.050

0.070

0.102

Pr(2°C < ECS <
4.5°C)

0.667

0.667

0.667

0.667

5th percentile

1.72

1.49

1.37

1.13

Median

3.00

3.00

3.00

3.00

Mean

3.50

3.28

3.19

3.07

95th percentile

7.14

5.97

5.59

5.17

31

-------
2. EQUILIBRIUM CLIMATE SENSITIVITY

Selected the Roe and Baker distribution:

Only distribution based on a theoretical
understanding of the response of the climate
system to increased GHG concentrations

Most consistent with IPCC judgments regarding
climate sensitivity:

"Values substantially higher than 4.5°C still cannot be
excluded"

ECS "is very likely larger than 1.5 °C"

31

-------
3. GLOBAL OR DOMESTIC DAMAGES

Current OMB guidance says Domestic Perspective is
Mandatory and International Perspective is Optional

Determined that a Global Measure of the Benefits
from Reducing U.S. Emissions is Preferable:

Global Externality. Emissions in U.S. Cause Damages Around
the World

The U.S. cannot mitigate climate change by itself

Decided against equity weighting that would place a greater
weight on losses in poor countries

NB: Best available evidence is that US damages are 5-15%
of global damges.

-------
IV. LIFETIME DAMAGES OF A TON OF GHG

I tk * W** I I 1^" I I I A I	% M III	IF"-	¦ % # ¦	1 £k	I	I I ,^N

EMISSIONS

33

-------
A. LONG RUN DAMAGES

V- 		5 T7	~					~			*	

Half Life of a Ton of C02 Emitted is 100 Years

Ton of Emissions Today will Affect Temperatures
and Damages for a Long Period

Net Present Value of Damage due to Ton of
Emissions Today Equals the Sum of the
Discounted Value of the Damages Each Year
Until It Has Disappeared from Atmosphere

The Choice of Discount Rate is a Key Factor

-------
B. DISCOUNT RATES

Choice of a discount rate, especially over long
periods of time, raises difficult questions

USG traditionally employs constant discount
rates of both 3 percent and 7 percent

35

-------
SELECTED DISCOUNT RATES

In light of the above considerations, USG used three discount
rates:

Low Value: 2.5 percent

Interest rates are highly uncertain over time
If climate investments are negatively correlated with market returns
Incorporates normative objections to rates of 3 percent or higher
Central Value: 3 percent

Consistent with estimates in the literature and OMB's guidelines for the
consumption rate of interest

Roughly corresponds to the after-tax riskless rate

High Value: 5 percent

If climate investments are positively correlated with market returns

May be justified by the high interest rates many consumers use to smooth
consumption

Approach is largely descriptive and uses constant discount
rates, but incorporates some key prescriptive concerns

36

-------
C. PUTTING IT ALL TOGETHER	

Running the models produces 45 separate
distributions of the SCC for a given year

(3 models) x (5 socioeconomic scenarios) x (1
climate sensitivity distribution) x (3 discount rates)

The distributions from each of the models and
scenarios are averaged together for each year

Produces three separate probability distributions
for SCC in a given year, one for each discount rate

37

-------
C. PUTTING IT TOGETHER	

For each 1AM, here are steps for calculating the SCC:

1.	Input the path of emissions, GDP, and population and calculate the
temperature effects and (consumption-equivalent) damages in each
year resulting from this baseline path of emissions.

2.	Add an additional unit of carbon emissions in year and recalculate
the temperature effects and damages expected in all years beyond t
resulting from this adjusted path of emissions.

3.	Subtract the damages computed in step 1 from those in step 2 in
each year.

4.	Discount the resulting path of marginal damages back to the year of
emissions using the agreed upon fixed discount rates and calculate the
SCC as the net present value of the discounted path of damages.

38

-------
V. RESULTS

i *¦>

-------
x USG selected four SCC estimates for use in
regulatory analyses

In 2010, these estimates are $5, $21, $35 & $65 (in
2007 US$)

First three estimates are the average SCC across 3
models & 5 emissions scenarios for 3 distinct discount
rates

The fourth value represents higher-than-expected
impacts

Use the SCC value for the 95th percentile at a 3 percent
discount rate

The $21 estimate associated with a 3% discount rate is
the central value

40

-------
HETEROGENEITY BY MODEL AND DISCOUNT RATE

Table 3: Disaggregated Social Cost of C02 Values by Model, Socio-Economic
Trajectory, and Discount Rate for 2010 (in 2007 dollars)



Discount rate:

5%

3%

2.5%

3%

Model

Scenario

Avg

Avg

Avg

95th



IMAGE

10.8

35.8

54.2

70.8



MERGE

7.5

22.0

31.6

42.1

u

Q

Message

9.8

29.8

43.5

58.6



MiniCAM

8.6

28.8

44.4

57.9



550 Average

8.2

24.9

37.4

50.8



IMAGE

8.3

39.5

65.5

142.4

LU

MERGE

5.2

22.3

34.6

82.4

(5

g

Message

7.2

30.3

49.2

115.6



MiniCAM

6.4

31.8

54.7

115.4



550 Average

5.5

25.4

42.9

104.7



IMAGE

-1.3

8.2

19.3

39.7

Q

MERGE

-0.3

8.0

14.8

41.3

Z
D

Message

-1.9

3.6

8.8

32.1

UL

MiniCAM

-0.6

10.2

22.2

42.6



550 Average

-2.7

-0.2

3.0

19.4

-------
Higher discount rates result in lower SCC values,
and vice versa

"here are clear differences in the SCC estimated
across the three main models

FUND produces the lowest estimates
PAGE produces the highest estimates

Results match up fairly well with model
estimates in the existing literature

The SCC increases over time

Physical and economic systems will become more
stressed

-------
RESULTS OVER TIME

Figure 3: Social Cost of C02, 2010 - 2050 (in 2007 dollars)

&
«
Z3

r-
o
o

CN

CN

o
o

CO

o
o

s

o
o
if)

160

140

120

100

80

60

40

20

~5% Avg
3% Avg
2.5% Avg
3% 95th

2010

2015

2020

2025

2030

2035

2040

2045

2050

43

-------
VI. CONCLUSIONS & DIRECTIONS FOR
UPDATING THE SCC

44

-------
The SCC offers a way to measure the economic
value of emissions reductions

The use of the SCC to guide GHG regulations
under the Clean Air Act offers the possibility of
achieving regulations where the benefits exceed
the costs

-------
DIRECTIONS FOR IMPROVEMENTS

Key areas for future research and advances in
calculation of SCC include:

Improvements in how lAM's capture catastrophic
impacts

More attention to how predicted physical impacts
translate into economic damages

Interactions between inter-sector and inter-regional
impacts (e.g., conflict)

More complete treatment of adaptation and
technological changes

Potential Incorporation of Risk Aversion

A methodology for valuing reductions in other GHG's

-------
Overview of Integrated
Assessment Models

Jae Edmonds and Kate Calvin

Improving the Assessment and Valuation of Climate
Change Impacts for Policy and Regulatory Analysis

November 18, 2010
Washington, DC

-------
What is an Integrated

lAMs integrate human and
natural Earth system climate
science.

¦	lAMs provide insights that would be
otherwise unavailable from
disciplinary research.

¦	lAMs capture interactions between
complex and highly nonlinear
systems.

¦	lAMs provide natural science
researchers with information about
human systems such as GHG
emissions, land use and land cover.

lAMs provide important,
science-based decision support
tools.

¦	lAMs support national, international,
regional, and private-sector
decisions.

Human Systems

Economy

Security

Food

Managed
Ecosystems

Population

ENERGY

Transport

Settlements

Science

Technology

Health

Natural Earth Systems

Atmospheric
Chemistry

	



	

Coastal

¦Hnes

Carbon
Cycle

Earth







System

Oceans

Hydrology

Ecosystems

Models

-------
lAMs Are Strategic in Nature

~ lAMs were designed to provide strategic insights.

~	lAMs were never designed to model the very fine details,
e.g.

¦	Electrical grid operation

¦	Daily oil market price paths.

~	lAMs are analogous to climate models in that sense.

¦	Climate models don't forecast weather

¦	They were designed to describe the determinants of 30-year
moving averages of weather.

~	lAMs also span a wide range of models with highly varied
levels of spatial and temporal resolution.

Pacific Northwest

NATIONAL LABORATORY

-------
Example of an IA insight: Sulfur & Land use

~	Carbon tax cases can have higher
radiative forcing than non-control
scenarios.

¦	Sulfur

¦	Land-use change emissions

~	I don't have the original figure because it
predates the age of PowerPoint.

Pacific Northwest

NATIONAL LABORATORY

-------
Example of an IA insight: Sulfur & Land use

~ Consider a reference scenario, e.g. reference to GCAM RCP 4.5, and
a scenario in which fossil fuel and industrial carbon is taxed.

Fossil Fuel & Industrial Carbon Emissions Total Anthropogenic Radiative Forcing

~ Radiative Forcing goes up prior to 2050 because of the
sulfur aerosol and indirect land-use effects.

-------
"HORSES FOR COURSES"

-JAKE JACOBY

Pacific Northwest

NATIONAL LABORATORY

-------
lAMs are a diverse set of tools

~	The diversity of lAMs is a reflection of the diversity of
problems for which the models were designed to address.

¦	What is the optimal climate policy?

¦	Implications of policy regimes for technology choice?

¦	How do policy, energy, the economy, land use and terrestrial
carbon cycle interact?

~	lAMs are evolving to address new questions

¦	How will emissions mitigation and climate impacts interact?

~	The bigger the question, the more aggregated the model.

Pacific Northwest

NATIONAL LABORATORY

-------
THE HIGHLY AGGREGATED IA
MODELS

Pacific Northwest

NATIONAL LABORATORY

-------
Three BIG question models: DICE, FUND &
PAGE

~	As far as I can remember, this line of investigation begins
with a series of discussion papers written by Bill
Nordhaus in 1989 and 1990 leading to the DICE model.

~	These models are characterized by high levels of
aggregation and comprehensiveness.

¦	Typically come in 3 parts.

Emissions

Natural Earth systems (atmospheric composition & climate
change)

Climate Damages

¦	RICE (the regional version of DICE) is ~17 equations

¦	For comparison, GCAM is ~110,000 lines of code

Pacific Northwest

NATIONAL LABORATORY

-------
Sources of Information

~	Highly aggregated lAMs face the problem of establishing
parameter values for the three major components—
emissions, natural Earth systems, and climate damages.

~	Most highly aggregated lAMs summarize information
gleaned from other, more detailed models or from off-line
research.

¦	The relationship between the more highly resolved lAMs and the
highly aggregated lAMs is similar in nature to the relationship
between the Earth system models of intermediate complexity
(EMICs) and the high resolution Earth system models (ESMs).

¦	But the highly aggregated lAMs also derive information from other
research domains, most notably the Impacts, adaptation and
vulnerability (IAV) community.

¦	(Climate research can be divided into IAV, 1AM, and atmosphere-

climate modeling domains.)

Pacific Northwest

NATIONAL LABORATORY

-------
The highly aggregated lAMs are often used for
the purpose of comparing the costs and
benefits of policy intervention. This introduces
several additional issues.

1.	How to compare non-market damages?

¦	Value of a human life—-just ask David Pearce.

¦	Value of unmanaged ecosystems.

¦	While these problems are amenable to economic analysis, actual
values are vigorously debated.

2.	How to include interaction effects?

¦	Across sectors—agriculture, energy and water

¦	Mitigation and adaptation—who gets the land?

¦	Land-use change from mitigation and adaptation affect climate?

3.	How to compare across time—and not just one week or
year to the next, but across multiple generations.

Pacific Northwest

NATIONAL LABORATORY

-------
For the US, how to compare across space—should
damages in distant lands be weighted as heavily as
damages at home?

The tails of the distribution

¦ Climate change potentially pushes the Earth system into regimes
that have not been observed for millions of years.

And, even then big things are different, e.g. the placement of
the continents.

¦ Extreme and catastrophic events are possible.

Both events that might be imagined—e.g. rapid destabilization
of clathrate zones, and.

Events that have not yet been imagined—the rapid emergence
of the ozone hole was the consequence of heterogeneous
chemistry that was not in the models until after the hole
needed to be explained.

• What is the proper weight to give to such events?

Pacific Northwest

NATIONAL LABORATORY

-------
THE HIGHER RESOLUTION IA
MODELS

Pacific Northwest

NATIONAL LABORATORY

-------
The Higher Resolution IA Models Address
Different Problems

~	Higher resolution lAMs address questions associated with
the details of the interactions between human and natural
Earth systems.

¦	The high resolution 1AM economies are more disaggregated;

¦	The high resolution 1AM energy system technologies are highly
varied;

¦	Land use and land cover strongly interact with the economy,
energy systems, and natural terrestrial processes.

~	The higher resolution lAMs tend to focus on outputs in
their natural units.

¦	How many new nuclear builds?

¦	How many Pg of C02 in geologic repositories?

¦	What impact will climate change have on the price of wheat?

Pacific Northwest

NATIONAL LABORATORY

-------
Cost effectiveness

~	The higher resolution IA models have focused on cost-
effectiveness

¦	What is the best way to stabilize C02 concentrations?

¦	What is the best way to limit global mean surface temperature
(GMST) not to exceed 2 degrees?

~	Rich Richels' classic slide

This is a cost-effectiveness study,
NOT a cost-benefit study!!!

Pacific Northwest

NATIONAL LABORATORY

-------
Cost effectiveness—SAP 2.1a

FigureTS. 10 Global Emissions of COj from Fossil and other Industrial Sources Across Scenarios (GtC/yr).

The tighter the constraint on radiative forcing,the faster carbon emissions must decline from those in the reference scenarios.This is
because the stabilization level defines a long-term carbon budget;
that is, the remaining amount of carbon that can be emitted in
the future.The gradual deflection of the emissions from the
reference reflects the assumption of when flexibility, with carbon
prices rising gradually. Under the most stringent radiative forcing
stabilization levels, CO^ emissions begin to decline immediately
or within a matter of decades. Under less stringent radiative
forcing stabilization levels, CO: emissions do not peak until late
in the century or beyond, and they are I Vi to over 2'A times
todays levels in 2100.



ESUadl

A. MERCE_Le«ll



HMK3V1

Lcml



\cai L«d2
K9UmO





MM CAM
MWCAH

_LchS



X MERGE Lra£3

<*



KSM..LbwM

—MERGE L*wM



MWCAM





C5WR.EF

MERGE..R£F

	



_Rff







MiniCAM

25

20

15

*.

U

O JO



i
i
i
i









I
i
i
i

i
i



i
i









i
i

i
i



i
i

1



C

S 1
i

i J

i

i

i
i



^r" iii iii i

f		 i i i i i i

' 	 i 1 1 1 1

i -A i i i i ¦

i

fl

i

pj



I I 1 I 1 I

i i i -»—	A— 1

iii ii ¥ -A 	3

iii i i i i
iii i i i i
iii i i i i

i

r	4

¦

2DOO

2020

2OH0

2060

2080

2100

thwest

LABORATORY

-------
Cost effectiveness

~	Because the higher resolution IA models have focused on
cost-effectiveness, they haven't had to worry that much
about the problems of impacts, and impact valuation. For
example,

¦	They haven't worried about the tails of the distribution—they
simply take the goal of limiting GMST to 2 degrees.

¦	Policy-technology interactions have loomed large.

¦	Discounting has been a lesser issue.

¦	Enumerating a complete set of atmosphere-climate impacts has
not been critical.

~	That situation is changing as the higher resolution IA
models focus more on impacts.

~	The higher resolution of these models mean that
interactions between sectors, regions, mitigation,
adaptation, and climate can begin to be studied,, ., M .

~	^	Pacific Northwest

NATIONAL LABORATORY

-------
Higher Resolution Integrated Assessment Models are
developed by interdisciplinary teams.

18

Model

Home Institution



AIM

Asia Integrated Model

National Institutes for Environmental
Studies, Tsukuba Japan



Japan:

rhe Asia Integrated Mc

,del r>



¦

%





GCAM

Global Change Assessment
Model

Joint Global Change Research
Institute, PNNL, College Park, MD

St*:

GCAM

IGSM

Integrated Global System
Model

Joint Program, MIT, Cambridge, MA





m

IMAGE

The Integrated Model to Assess
the Global Environment

PBL Netherlands Environmental

Assessment Agency, Bildhoven, The
Netherlands





T I .

1 1 ! •"













mmmm.

MERGE

Model for Evaluating the
Regional and Global Effects of
GHG Reduction Policies

Electric Power Research Institute,
Palo Alto, CA

1

B

Merge

MESSAGE

Model for Energy Supply
Strategy Alternatives and their
General Environmental Impact

International Institute for Applied
Systems Analysis; Laxenburg, Austria





% %



jrthwest

\L LABORATORY

-------
Higher resolution lAMs have provided atmosphere &
climate models with both emissions and LULC trajectories

History RCPs

1900 2000 2100

GHG Emissions and Concentrations from IAMs

-	Greenhouse gases: C02, CH4, N20, CFCs, HFC's, PFCs,

sf6

-	Emissions of chemically active gases: CO, NOx, NH3,
VOCs

-	Derived GHG's\ tropospheric 03

-	Emissions of aerosols: S02, Black Carbon (BC), Organic
Carbon (OC)

-	Land use and land cover

Globally averaged surface air temperature

4.0

P 3.0 H

CT>

o
o

2.0 -

o

CD
CD

_L

_L

_L

_L

CCSM4

J	I	1	I	I	I	L_

-RCP 8.5
-RCP 4.5
-RCP 2.6
-20C3M

Note: Preliminary Results
Subject to Change

a> o.o

-2.0

1880

Natural Anthropogenic

T	1	1	1	1	1	r

1920 1960

1	1	1	1	1	1	1	1	1	1	1	1~

2000 2040 2080

Emissions

Chemical reactions

Removal processes:
dry and wet deposition

Radiation

Clouds

Thanks to Warren Washington for CCSM4 preliminary results.

-------
The iESM

• Models that integrate state of the art human Earth system
models	(taken from lAMs) with

(ESMs)	are being actively	devel

• The iESMs will provide feedbacks from atmosphere, oceans,
and climate	on terrestrial systemE.g. climate and

atmospheric composition feedbacks on crop yields, energy
demands, bioenergy prices and climate mitigation.

r



Emissions

Atmosphere

Climate

-------
Growing Overlap in Research

Domains

Impacts,
Adaptation &
Vulnerability

Integrated
Assessment
Models

1980's

imate Models

Pacific Northwest

NATIONAL LABORATORY

-------
Where lAMs Are Headed

Integrated Assessment

NATIONAL LABORATORY

-------
1AM Research Challenges

Impacts, Adaptation &
Vulnerability

Linking to and
Collaborating with Oth
Climate Sciences

egional Scales & Shorter
Time Steps

nergy, Technology,
Water, Land & Science

Evaluating Risk & Scientific Uncertainty, and
Exploration of New Methodologies

-------
DISCUSSION

Pacific Northwest

NATIONAL LABORATORY

-------
Summary of the DICE model

Stephen C. Newbold
U.S. EPA, National Center for Environmental Economics1

This report gives a brief summary of the DICE (Dynamic Integrated Climate-Economy)
model, developed by William Nordhaus, which "integrate [s] in an end-to-end fashion the
economics, carbon cycle, climate science, and impacts in a highly aggregated model that
allow[s] a weighing of the costs and benefits of taking steps to slow greenhouse warming"
(Nordhaus and Boyer 2000 p 5). Section 1 of this report recounts the major milestones in the
development of DICE and its regionally disaggregated companion model, RICE. This section
also serves as a convenient reference for more detailed expositions of the model and
applications in the primary literature. Section 2 describes the basic structure of the most
recently published version of DICE, and Section 3 describes some key aspects of the model
calibration. Section 4 gives additional details on the climate damage function in DICE, and
Section 5 gives a brief description of the most recently published version of the RICE model.

1 Historical development

The DICE integrated assessment model has been developed in a series of reports, peer
reviewed articles, and books by William Nordhaus and colleagues over the course of more than
thirty years. The earliest precursor to DICE was a linear programming model of energy supply
and demand with additional constraints imposed to represent limits on the peak concentration
of carbon dioxide in the atmosphere (Nordhaus 1977a,b).2 The model was dynamic, in that it
represented the time paths of the supply of energy from various fuels and the demand for
energy in different sectors of the economy and the associated emissions and atmospheric
concentrations of carbon dioxide. However, it included no representation of the economic
impacts or damages from temperature or other climate changes. Later, Nordhaus (1991)
developed a long-run steady-state model of the global economy that included estimates of both
the costs of abating carbon dioxide emissions and the long term future climate impacts from
climate change. This allowed for a balancing of the benefits and costs of carbon dioxide
emissions to help determine the optimal level of near term controls. The analysis centered on

1 Prepared for the EPA/DOE workshop, Improving the Assessment and Valuation of Climate Change Impacts

for Policy and Regulatory Analysis, Washington DC, November 18-19,2010. Please note that the views expressed in
this paper are those of the author and do not necessarily represent those of the U.S. Environmental Protection
Agency. No Agency endorsement should be inferred. Author's email: newbold.steve@epa.gov.

-------
the global average surface temperature, which was "...chosen because it is a useful index (in the
nature of a sufficient statistic) of climate change that tends to be associated with most other
important changes rather than because it is the most important factor in determining impacts"
(Nordhaus 1991 p 930). The categories of climate damages that were represented in the model
were associated with market sectors that accounted for roughly 13% of GDP in the United
States.3

The DICE model was first presented in its modern form by Nordhaus (1992a,b), who
described the new, fully dynamic Ramsey-type optimal growth structure of the model and the
optimal time path of emission reductions and associated carbon taxes that emerged from it.
The full derivation and extended description of the DICE model and a wider range of
applications were presented in a book by Nordhaus (1994a). The next major advance involved
disaggregating the model into ten different groups of nations to produce the RICE (Regional
DICE) model, which allowed the authors to examine national-level climate policies and
different strategies for international cooperation (Nordhaus and Yang 1996). An update and
extended description of both RICE (now with eight regions) and DICE appeared in the book by
Nordhaus and Boyer (2000). The next major update of DICE, modified to include a backstop
technology that can replace all fossil fuels and whose price was projected to decline slowly over
time, appeared in another book by Nordhaus (2008). Finally, Nordhaus (2010) described the
most recent version of the RICE model, which adds an explicit representation of damages due
to sea level rise.

In addition to the studies by Nordhaus and colleagues mentioned above, DICE has been
adapted by other researchers to examine a wide range of issues related to the economics of
climate change. A comprehensive review is well beyond the scope of this summary, so only a
few examples are mentioned here. Pizer (1999) used DICE to compare carbon tax and a cap-
and-trade-style policies under uncertainty. Popp (2005) modified DICE to include endogenous
technical change. Baker et al. (2006) used DICE to examine the effects of technology research
and development on global abatement costs. Hoel and Sterner (2007) modified the utility
function in DICE to include a form of non-market environmental consumption that is an
imperfect substitute for market consumption, and Yang (2008) used RICE in a cooperative
game theory framework to examine strategies for international negotiations of greenhouse gas
mitigation policies and targets.

2 Basic model structure

DICE2007 is a modified Ramsey-style optimal economic growth model, where an
additional form of "unnatural capital"—the atmospheric concentration of CO2—has a negative

3 It should be emphasized that while this model and all subsequent versions of DICE necessarily make
assumptions about climate and economic conditions in the far future, the important question is the extent to which
current policies are robust to changes in assumptions about future variables (Nordhaus, personal communication).

2

-------
effect on economic output through its influence on the global average surface temperature.
Global economic output is represented by a Cobb-Douglas production function using physical
capital and labor as inputs. Labor is assumed to be proportional to the total global population,
which grows exogenously over time. Total factor productivity also increases exogenously over
time. The carbon dioxide intensity of economic production and the cost of reducing carbon
dioxide emissions decrease exogenously over time. In each period a fraction of output is lost
according to a Hicks-neutral climate change damage function. The output in each period is
then divided between consumption, investment in the physical capital stock (savings), and
expenditures on emissions reductions (akin to investment in the natural capital stock). DICE
solves for the optimal path of savings and emissions reductions over a multi-century planning
horizon, where the objective to be maximized is the discounted sum of all future utilities from
consumption. Total utility in each period is the product of the number of individuals alive and
the utility of a representative individual with average income in that period. The period utility
function is of the standard constant relative risk aversion (CRRA) form, and utilities in future
periods are discounted at a fixed pure rate of time preference.

3 Calibration

The climate model in DICE2007 tracks the stocks and flows of carbon in three aggregate
compartments of the earth system: the lower atmosphere, the shallow ocean, and the deep
ocean. The transfer coefficients linking the flows among the compartments were "calibrated to
fit the estimates from general circulation models and impulse-response experiments,
particularly matching the forcing and temperature profiles in the MAGICC model" (Nordhaus
2008 p 54). The climate sensitivity parameter—the equilibrium change in global average
surface temperature after a sustained doubling of atmospheric carbon dioxide concentration—
was set to 3 degrees Celsius, which is near the middle of the range cited by the IPCC. The
projected temperature change under the baseline scenario (with no climate controls for the
first 250 years) is an increase in global average surface temperature of 3.2 degrees Celsius
around year 2100 with a peak of around 6.5 degrees Celsius around year 2500.

The key economic growth and preference parameters of DICE2007 are calibrated as
follows. The global population is projected to grow exogenously from around 6.5 billion in
2005 to 8.6 billion around 2200. Total factor productivity growth and the discount rate
parameters were calibrated to match market returns in the early periods of the model:
specifically, "We have chosen a time discount rate of IV2 percent per year along with a
consumption elasticity of 2. With this pair of assumptions, the real return on capital averages
around SV2 percent per year for the first half century of the projections, and this is our estimate
of the rate of return on capital" (Nordhaus 2008 p 61).

The abatement cost function is specified such that the marginal abatement cost,
measured as a fraction of output, increases roughly with the square of the fraction of emissions

3

-------
abated. The backstop price—the marginal cost of eliminating the last unit of emissions in each
period—is $1,170 per metric ton of carbon in the first period and falls exponentially at a rate of
5% per decade to a long run value of $585 per metric ton of carbon.

The climate damage function is specified such that for small temperature changes the
fraction of output lost in each period increases with the square of the increase in temperature
above the preindustrial average temperature.4 The coefficient of the damage function is
calibrated so that roughly 1.7% of global economic output is lost when the average global
surface temperature is elevated by 2.5 degrees Celsius above the preindustrial average.

4 Damages

The globally aggregated climate damage function in DICE has been calibrated to match
the sum of climate damages in all regions represented in the RICE model. The potential
damages from climate change are divided into seven categories: agriculture, sea level rise,
other market sectors, human health, nonmarket amenity impacts, human settlements and
ecosystems, and catastrophes. A full recounting of the derivation of the damage functions in all
categories is beyond the scope of this short summary, but to the give the reader a flavor for
what is involved this section reviews three categories of damages: agriculture, heath, and
catastrophes. This discussion draws heavily on Chapter 4 of Nordhaus and Boyer (2000), so
the reader is referred there for more information.

Agriculture can serve as an illustrative example of some of the other categories not
covered here. The basic strategy for calibrating the damage functions is to draw on estimates
from previous studies of the potential economic losses in each category at a benchmark level of
warming of 2.5 degrees Celsius, extrapolating across regions as necessary to cover data gaps in
the literature. Some extrapolations were made using income elasticities for each impact
category. As the authors explain, "United States agriculture can serve here as an example. Our
estimate is that [the fraction of the value of agricultural output lost at 2.5 degrees Celsius] is
0.065 percent [based on Darwin et al. 1995]... The income elasticity of the impact index is
estimated to be -0.1, based on the declining share of agriculture in output as per capita output
rises" (Nordhaus and Boyer 2000 p 74-75).

The human health impacts of climate change were based on the effects of pollution and
a broad group of climate-related tropical diseases including malaria and dengue fever. The
increased mortality from warming in the summer and decreased mortality from warming in
the winter were assumed to roughly offset and so were not included. The specification of the
human health damage function involved "a regression of the logarithm of climate related [years

4 The DICE2007 damage function has an "S-shape," so for very large temperature changes the fraction of output
lost increases with temperature at a decreasing rate and asymptotes to one. However, it should be emphasized that
the damage function is calibrated to damages in the range of 2 to 4 degrees Celsius. The extent of non-linearity
beyond this range is unknown, so extrapolations beyond this point should not be considered reliable (Nordhaus,
personal communication).

4

-------
of life lost] on mean regional temperature estimated form the data presented in Murray and
Lopez [1996]" with judgmental adjustments "to approximate the difference among subregions
that is climate related," and each year of life lost was valued at two years of per capita income
(Nordhaus and Boyer 2000 p 80-82).

The damages from potential catastrophic impacts were estimated using results from a
previous survey of climate experts by Nordhaus (1994b). The experts were asked for their
best professional judgment of the likelihood of a catastrophe—specified as a 25 percent loss of
global income indefinitely—if the global average surface temperature increased by 3 and by 6
degrees Celsius within 100 years. The averages of the survey responses were adjusted upward
somewhat based on "[developments since the survey [that] have heightened concerns about
the risks associated with major geophysical changes, particularly those associated with
potential changes in thermohaline circulation" (Nordhaus and Boyer 2000 p 87). The
probability of a 30 percent loss of global income indefinitely was assumed to be 1.2 and 6.8
percent with 2.5 and 6 degrees Celsius of warming, respectively. The percent of income lost
was assumed to vary by region, and a coefficient of relative risk aversion equal to 4 was used to
calculate the willingness to pay to avoid these risks in each region. The resulting "range of
estimates of WTP lies between 0.45 and 1.9 percent of income for a 2.5°C warming and
between 2.5 and 10.8 percent of income for a 6°C warming. It is assumed that this WTP has an
income elasticity of 0.1" (Nordhaus and Boyer 2000 p 89).

Damages in the remaining categories were estimated in a similar vein, using a
combination of empirical estimates from previous climate impact studies and professional
judgments when needed to close the sometimes wide gaps in the literature. The table below
shows the resulting global estimates of damages in each category in the 1999 version of RICE.

Damages as a percent of global output at 2.5°C of warming



Output
weighted

Population
weighted

Agriculture

0.13

0.17

Sea level rise

0.32

0.12

Other market sectors

0.05

0.23

Health

0.10

0.56

Non-market amenities

-0.29

-0.03

Human settlements and ecosystems

0.17

0.10

Catastrophes

1.02

1.05

Total

1.50

1.88

(Nordhaus and Boyer 2000 p 91)

5

-------
With damages in all categories estimated, the DICE damage function was then calibrated
"so that the optimal carbon tax and emissions control rates in DICE-99 matched the projections
of these variables in the optimal run of RICE-99" (Nordhaus and Boyer 2000 p 104).

5 Recent developments

Nordhaus (2010) presented results from an updated version of the RICE model. A
major extension is a new sea level rise damage function, now explicitly modeled by region as a
function of the global average sea level rise rather than rolled up in the aggregate damage
function. "The RICE-2010 model provides a revised set of damage estimates based on a recent
review of the literature [Toll 2009, IPCC 2007], Damages are a function of temperature, SLR,
and CO2 concentrations and are region-specific. To give an idea of the estimated damages in
the uncontrolled (baseline) case, those damages in 2095 are... 2.8% of global output, for a
global temperature increase of 3.4°C above 1900 levels" (Nordhaus 2010 p 3). Other
parameter updates include climate sensitivity, now set to 3.2 degrees Celsius, the elasticity of
the marginal utility of income, now set to -1.5, and parameters that control economic growth
rates, which are re-calibrated such that world per capita consumption grows by an average rate
of 2.2% per year for the first 50 years.

6

-------
References

Baker E, Clarke L, Weyant J. 2006. Optimal technology R&D in the face of climate uncertainty.

Climatic Change 78:157-159.

Darwin R, Tsigas M, Lewandrowski J, Raneses A. 1995. World Agriculture and Climate Change:
Economic Adaptations. Natural Resources and Environment Division, Economic Research
Service, U.S. Department of Agriculture. Agricultural Economic Report No. 703.

Hoel M, Sterner T. 2007. Discounting and relative prices. Climatic Change 84:265-280.

IPCC (Intergovernmental Panel on Climate Change). 2007. Climate Change 2007: Impacts,

Adaptation and Vulnerability, Working Group II Contribution to the Intergovernmental Panel
on Climate Change, Summary for Policymakers. Cambridge, UK: Cambridge University Press.
Murray CJL, Lopez AD, eds. 1996. The Global Burden of Disease. Cambridge, MA: Harvard
University Press.

Nordhaus WD. 1977a. Strategies for the control of carbon dioxide. Cowles Foundation

discussion paper no. 443.

Nordhaus WD. 1977b. Economic growth and climate: the carbon dioxide problem. The

American Economic Review 67(l):341-346.

Nordhaus WD. 1991. To slow or not to slow: the economics of the greenhouse effect. The

Economic Journal 101(407):920-937.

Nordhaus WD. 1992a. The "DICE" model: background and structure of a Dynamic /ntegrated
Climate-Economy model of the economics of global warming. Cowles Foundation discussion
paper no. 1009.

Nordhaus WD. 1992b. Optimal greenhouse-gas reductions and tax policy in the "DICE" model.

The American Economic Review 83(2):313-317.

Nordhaus WD. 1994a. Managing the Global Commons: The Economics of Climate Change.

Cambridge, MA: The MIT Press.

Nordhaus WD. 1994b. Expert opinion on climatic change. American Scientist 82:45-51.
Nordhaus WD. 2010. Economic aspects of global warming in a post-Copenhagen environment.

Proceedings of the National Academy of Sciences 107(26):11721-11726.

Nordhaus WD. 2008. A Question of Balance: Weighing the Options on Global Warming Policies.

Pre-publication version, http://nordhaus.econ.yale.edu/Balance_2nd_proofs.pdf.

Nordhaus WD, Boyer J. 2000. Warming the World: Economic Models of Global Warming.

Cambridge, MA: The MIT Press.

Nordhaus WD, Yang Z. 1996. A regional dynamic general-equilibrium model of alternative

climate-change strategies. The American Economic Review 86(4):741-765.

Pizer WA. 1999. The optimal choice of climate change policy in the presence of uncertainty.
Resource and Energy Economics 21:255-287.

-------
Popp D. 2005. ENTICE: endogenous technological change in the DICE model of global warming.

Journal of Environmental Economics and Management 48:742-768.

Tol R. 2009. The economic effects of climate chang e. Journal of Economic Perspectives 23:29-51.
Yang Z. 2008. Strategic Bargaining and Cooperation in Greenhouse Gas Mitigations: An
Integrated Assessment Modeling Approach. Cambridge, MA: The MIT Press.

8

-------
Summary of the DICE model

Steve Newbold
U.S. EPA, National Center for Environmental Economics

The views expressed in this presentation are those of the author and do not
necessarily represent those of the U.S. Environmental Protection Agency.
No Agency endorsement should be inferred.

-------
Outline

1.	Historical development

2.	Applications

3.	One-slide summary

4.	Model structure

5.	The SCC in DICE

6.	Calibration of global damage function

7.	Quick update: RICE2010

2

-------
Historical development

Nordhaus WD. 1977. Strategies for the control of carbon dioxide. Cowles
Foundation Discussion Paper.

Nordhaus WD. 1977. Economic growth and climate: the carbon dioxide
problem. AER 67(l):341-346.

Nordhaus WD. 1991. To slow or not to slow: the economics of the greenhouse
effect. The Economics Journal 101(407):920-937.

Nordhaus WD. 1992. Optimal greenhouse gas reductions and tax policy in the
"DICE" model. AER 83(2):313-317.

Nordhaus WD. 1994. Managing the Global Commons. Cambridge, MA: MIT Press.

Nordhaus WD, Yang Z. 1996. A regional dynamic general-equilibrium model of
alternative climate-change strategies. AER 86(4):741-765.

Nordhaus WD, Boyer J. 2000. Warming the World. Cambridge, MA: MIT Press.

Nordhaus WD. 2008.^4 Question of Balance. Cambridge, MA: MIT Press.

Nordhaus WD. 2010. Economic aspects of global warming in a post Copenhagen
environment. PNAS.

3

-------
Applications

DICE is designed to:

•	"...estimate the optimal path of capital accumulation and GHG -
emissions reductions" (Nordhaus 1992,1994).

•	Compare taxes versus quantity controls under uncertainty, and
investigate value of early information (Nordhaus 1994 Ch 8).

•	Compare business as usual scenario and optimized policy to
alternatives, e.g., Kyoto Protocol, similar to Stern Review, Gore
emission reductions, temperature constraints (Nordhaus 2008).

•	"Examine alternative outcomes for emissions, climate change,
and damages under different policy scenarios" and calculate the
near term carbon prices along alternative policy paths (Nordhaus
2010).

4

-------
Applications

DICE has been modified by others to examine a wide
range of climate change economics issues, e.g.,

•	Pizer (1999) [P vs Q for climate policy]

•	Popp (2004) [endogenous technical change]

•	Baker et al (2006) [optimal R&D policy]

•	Hoel and Sterner (2007) [relative prices of market vs
non-market consumption]

•	Yang (2008) [strategic bargaining in international
negotiations]

•	de Bruin et al (2009) [optimal adaptation policy]

5

-------
One-slide summary

•	Dynamic Integrated Climate-Economy model

•	Optimal economic growth model + a simplified climate change model
+ a damage function that represents the loss of economic output due
to increased global surface temperatures + projection of abatement
costs over time.

•	Solves for optimal path of savings and abatement to maximize present
value of discounted aggregate utility

•	Some key results from DICE2007 (Nordhaus 2008):

>	SCC20o5 in baseline scenario ~ $7.5/tC02 (~ optimal carbon tax)

>	SCC growth rate ~ 0.02/yr

>	Max temp increase ~ 6°C (no controls for 250 yrs); ~ 3.5°C (optimal)

•	New results from RICE2010 in Nordhaus (2010) PNAS

6

-------
Model structure

Net output = gross output from economic production

-	fraction of output lost due to climate damages

-	fraction of output spent on abatement

Consumption = net output - savings

Capital accumulation = savings - depreciation

Temperature = "three-box" climate model calibrated to MAGGIC

Choose savings and abatement to max present value of future
utilities, where utility depends on per-capita consumption in
each period

Key quantities:

>	Pure rate of time preference = 0.015/yr

>	Elasticity of m.u. of consumption = -2

>	Initial per capita consumption growth rate * 0.016/yr

>	Damages at 3 deg C * 2.5% of world GDP

>	Damages at 6 deg C * 9.3% of world GDP

-------
The SCC in DICE

Social cost of carbon = shadow value of emissions -r

shadow value of capital stock

Along an optimal path this will equal:

1.	the change in consumption in all future years
from one additional unit of emissions in the
current year, discounted to present value using
the Ramsey consumption discount rate, and

2.	the tax on C02 emissions.

8

-------
Calibration of damage function

Basic strategy:

1.	Choose a functional form for aggregate climate
change damages as a fraction of global economic
output (e.g., low order polynomial].

2.	Calibrate damage function parameters using
summary of empirical studies of climate change
damages in all major categories, extrapolating
among regions as necessary:

agriculture, sea-level rise, other market sectors, health,
nonmarket amenity impacts, human settlements and
ecosystems, catastrophes.

(Nordhaus & Boyer 2000)

9

-------
Sector by sector

Example 1 - Agriculture:

•	Similar calibration strategy for some other sectors

•	Draw on estimates from previous studies of the potential
economic losses in each category at a benchmark level of
warming of 2.5°C

•	Extrapolate across regions as necessary to cover data gaps
using income elasticities for each impact category

•	"United States agriculture can serve here as an example. Our
estimate is that [the fraction of the value of agricultural output
lost at 2.5°C] is 0.065 percent [based on Darwin et al 1995]...
The income elasticity of the impact index is estimated to be -
0.1, based on the declining share of agriculture in output as

per capita output rises" (Nordhaus and Boyer 2000 p 74-75).

10

-------
Sector by sector

Example 2 - Health:

•	Based on effects of pollution and a broad group of climate-
related tropical diseases including malaria and dengue fever

•	Changes in mortality from more severe summers and less
severe winters were assumed to roughly offset and so were
not included

•	Using data from Murray and Lopez (1996), regress the log of
climate related YLLs [years of life lost] on mean regional
temperature

•	Plus judgmental adjustments "to approximate the difference
among subregions that is climate related"

•	Each YLL valued at two years of per capita income (Nordhaus
and Boyer 2000 p 80-82).

li

-------
Sector by sector

Example 3 - Catastrophes:

•	Based on results from survey of climate experts (Nordhaus
1994). Experts asked for likelihood of a catastrophe (i.e., 25%
loss of global income indefinitely) if the global average temp
increased by 3°C and by 6°C within 100 years.

•	Average responses adjusted upward based on "heightened
concerns about the risks associated with major geophysical
changes..."

•	Probability of 30% loss of global income assumed to be 1.2%
with 2.5°C and 6.8% with 6°C of warming. CRRA = 4 used to
calculate WTP to avoid catastrophic risks.

•	WTPs between 0.45% and 1.9% of income for 2.5°C and
between 2.5% and 10.8% for 6°C warming. Assumed that this
WTP has income elasticity =0.1

-------
Sector by sector

Category

Damages at2.5°C
[ % of global output ]



Output
weighted

Population
weighted

Agriculture

0.13

0.17

Sea-level rise

0.32

0.12

Other market sectors

0.05

0.23

Health

0.10

0.56

Non-market amenities

-0.29

-0.03

Human settlements & ecosystems

0.17

0.10

Catastrophes

1.02

1.05

Total

1.50

1.88

(Nordhaus & Boyer 2000)

-------
Aggregation of damages

RICE/DICE1999 (Nordhaus & Boyer 2000]:

1.	Calculate regional impacts for 2.5°C and 6°C.

2.	Sum across categories to create overall impacts for each
region.

3.	Solve system of 2 quadratic equations for reach region to
obtain quadratic damage function parameters for each
region.

4.	DICE quadratic damage function calibrated "so that the
optimal carbon tax and emissions control rates in DICE-99
matched the projections of these variables in the optimal run
of RICE-99."

14

-------
Update: RICE2010

Nordhaus [2010):

•	Parameters: pure rate of time
preference = 0.015/yr, elasticity
of m.u. of consumption = -1.5,
initial growth rate of per cap
consumption ~ 0.022/yr.

•	"...provides a revised set of
damage estimates based on a
recent review of the literature
[Tol 2009, IPCC 2007]. Damages
are a function of temperature,
SLR, and C02 concentrations and
are region-specific."

•	Near term carbon price on
optimal path * $ll/ton C02

0.35

¦RICE2010

AT [deg C]

RICE2010 damages plotted against
temperature change relative to pre-industrial
in each year.

15

-------
The PAGE09 model: Estimating climate impacts and the
social cost of C02

Chris Hope (c.hope@jbs.cam.ac.uk)

October 2010

Introduction

The changes made to PAGE2002 to create PAGE09 are outlined below under the following headings:
Science, Impacts and Adaptation.

-------
Science

Inclusion of Nitrous Oxide

The number of gases whose emissions, concentrations and forcing are explicitly modelled is
increased from 3 in PAGE2002 to 4 in PAGE09. The forcing from N20 takes the same form as for
CH4, based on the square root of the concentration. The excess forcing from gases not explicitly
modelled is now allowed to vary by policy.

Inclusion of transient climate response

In PAGE2002, the climate sensitivity is input directly as an uncertain parameter. The climate
sensitivity in PAGE09 is derived from two inputs, the transient climate response (TCR), defined as the
temperature rise after 70 years, corresponding to the doubling-time of C02 concentration, with C02
concentration rising at 1% per year, and the feedback response time (FRT) of the Earth to a change
in radiative forcing (Andrews and Allen, 2008). Default triangular distributions for TCR and FRT in
PAGE09 give a climate sensitivity distribution with a mean of 3 degC, and a long right tail, consistent
with the latest estimates from IPCC, 2007.

Feedback from temperature to the carbon cycle

The standard PAGE2002 model contains an estimate of the extra natural emissions of C02 that will
occur as the temperature rises (an approximation for a decrease in absorption in the ocean and
possibly a loss of soil carbon (Hope, 2006)). Recent model comparison exercises have shown that
the form of the feedback in PAGE2002 works well for business as usual emissions, but overestimates
concentrations in low emission scenarios (van Vuuren et al, 2009).

In PAGE09, the carbon cycle feedback (CCF) is introduced as a linear feedback from global mean
temperature to a percentage gain in the excess concentration of C02, to simulate the decrease in
C02 absorption on land and in the ocean as temperature rises (Friedlingstein et al, 2006). PAGE09 is
much better than PAGE2002 at simulating the carbon cycle feedback results for low emission
scenarios in Friedlingstein et al, 2006, van Vuuren et al, 2009.

Land temperature patterns by latitude

In PAGE2002, regional temperatures vary from the global mean temperature only because of
regional sulphate forcing. However, geographical patterns of projected warming show greatest
temperature increases over land (IPCC, 2007, chlO, p749), and a variation with latitude, with regions
near the poles warming more than those near the equator (IPCC, 2007, chlO, figure 10.8 and
supplementary material).

In PAGE09 the regional temperature is adjusted by a factor related to the effective latitude of the
region, and one related to the land-based nature of the regions. The adjustment is calculated for
each region using an uncertain parameter of the order of 1 degC representing the temperature
increase difference between equator and pole, and the effective absolute latitude of the region, and
an uncertain constant of the order of 1.4 representing the ratio between mean land and ocean
temperature increases.

-------
Explicit incorporation of sea level rise

In PAGE2002, sea level rise is only included implicitly, assumed to be linearly related to global mean
temperature. This neglects the different time constant of the sea level response, which is longer
than the surface air temperature response (IPPC, 2007, p823).

In PAGE09, sea level is modelled explicitly as a lagged linear function of global mean temperature
(Grinsted et al, 2009). The IPCC has a sea level rise projection in 2100 of 0.4 - 0.7 m from pre-
industrial times (IPCC, 2007, p409). A characteristic response time of between 500 and 1500 years
in PAGE09 gives sea level rises compatible with these IPCC results.

Impacts

Impacts as a proportion of GDP

In PAGE2002, economic and non-economic impacts before adaptation are a polynomial function of
the difference between the regional temperature and the tolerable temperature level, with regional
weights representing the difference between more and less vulnerable regions. These impacts are
then equity weighted, discounted at the consumption rate of interest and summed over the period
from now until 2200. There are several issues with this representation, including the lack of an
explicit link from GDP per capita to the regional weights, and the possibility that impacts could
exceed 100% of GDP with unfavourable parameter combinations.

In PAGE09, extra flexibility is introduced by allowing the possibility of initial benefits from small
increases in regional temperature (Tol, 2002), by linking impacts explicitly to GDP per capita and by
letting the impacts drop below their polynomial on a logistic path once they exceed a certain
proportion of remaining GDP to reflect a saturation in the vulnerability of economic and non-
economic activities to climate change, and ensure they do not exceed 100% of GDP.

Figure 1

Impact by temperature

proportion of GDP

Figure 1 shows such an impact function, with initial benefits (IBEN) of 1% of GDP per degree, with
impacts (W) of 4% of GDP at a calibration temperature (TCAL) of 2.5 degC, with a polynomial power
(POW) of 3, and an exponent with income (IPOW) of -0.5. The impact function has a saturation(ISAT)
starting at 50% of GDP, which keeps the impacts (blue line) below 100% of GDP even for the high

-------
temperatures shown. The red line shows what the impacts would be if they continued to follow the
polynomial form without saturation.

Discontinuity impacts

As in PAGE2002, the risk of a large-scale discontinuity, such as the Greenland ice sheet melting, is
explicitly modelled. In PAGE09 the losses associated with a discontinuity do not all occur
immediately, but instead develop with a characteristic lifetime after the discontinuity is triggered
(Lenton et al, 2008).

Equity weighting of impacts

In PAGE2002, impacts are equity weighted in a rather ad-hoc way, with the change in consumption
increased in poor regions and decreased in rich ones.

PAGE09 uses the equity weighting scheme proposed by Anthoff et al (2009) which converts changes
in consumption to utility, and amounts to multiplying the changes in consumption by

EQ(r,t) = (G(fr,0)/G(r,t))A EMUC

Adaptation

The green line in figure 2 shows an illustrative tolerable temperature profile over time in an impact
sector that results from an adaptation policy that gives a tolerable temperature of 2 degC, starting in
2020 and taking 20 years to implement fully. If the temperature rise is shown by the red line, there
will be 0.5 degC of impacts in 2000, increasing to 1 deg C by 2020, then reducing to 0 from 2030 to
2060. After 2060 the impacts start again, reaching 1 deg C by 2100.

-------
Figure 2 Temperature and tolerable temperature by date (illustrative)

DegC

3.5 n

2.5 -

1.5

0.5

0 -I—

2000

2020

2040

2060

2080

2100

Year

Acknowledgement

References

ADB, 2009, The Economics of Climate Change in Southeast Asia: A Regional Review, Asian
Development Bank, Philippines.

Andrews DG, and Allen MR, 2008, Diagnosis of climate models in terms of transient climate response
and feedback response time, Atmos. Sci. Let. 9:7-12

Anthoff D, Hepburn C and Tol RSJ, 2009, "Equity weighting and the marginal damage costs of climate
change", Ecological Economics, Volume 68, Issue 3, 15 January 2009, 836-849.

Friedlingstein P, Cox P, Betts R, Bopp I, Von bloh W, Brovkin V, Cadule P, Doney S, Eby M, Fung I, Bala
G, John J, Jones C, Joos F, Kato T, Kawamiya M, Knorr W, Lindsay K, Matthews HD, Raddatz T, Rayner
P, Reick C, Roeckner E, Schnitzler KG, Schnur R, Strassmann K, Weaver AJ, Yoshikawa C, Zeng N,
2006, Climate-carbon cycle feedback analysis: results from the C4MIP model intercomparison. J Clim
19:3337-3353.

Bloomberg, 2010, A fresh look at the costs of reducing US carbon emissions, Bloomberg New Energy
Finance.

EPA, 2010, appendix 15a, Social cost of carbon for regulatory impact analysis under executive order
12866,

-------
http://wwwl.eere.energy.gov/buildings/appliance standards/commercial/pdfs/smallmotors tsd/se
m finalrule appendixl5a.pdf

Aslak Grinsted , J. C. Moore, S. Jevrejeva, 2009, Clim Dyn, doi: 10.1007/s00382-008-0507-2.

Hope C, 2008a, Optimal carbon emissions and the social cost of carbon over time under uncertainty,
Integrated Assessment, 8, 1, 107-122.

Hope C, 2008b, "Discount rates, equity weights and the social cost of carbon", Energy Economics, 30,
3, 1011-1019.

Hope C, 2006, "The marginal impact of C02 from PAGE2002: An integrated assessment model
incorporating the IPCC's five reasons for concern", Integrated Assessment, 6, 1, 19-56.

IPCC, 2007, Climate Change 2007. The Physical Science Basis. Summary for Policymakers.
Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel
on Climate Change. IPCC Secretariat Switzerland.

Lenton, T. M., H. Held, E. Kriegler, J. W. Hall, W. Lucht, S. Rahmstorf and H. J. Schellnhuber, 2008,
"Tipping elements in the Earth's climate system", Proceedings of the National Academy of Sciences
USA 105(6), 1786-1793.

Stern, Nicholas. 2007. The Economics of Climate Change: The Stern Review. Cambridge and New
York: Cambridge University Press.

Tol, R.S.J., 2002, "New estimates of the damage costs of climate change, Part II: dynamic estimates.",
Environ. Resour. Econ., 21, 135-160.

Detlef van Vuuren, Jason Lowe, Elke Stehfest, Laila Gohar, Andries Hof, Chris Hope, Rachel Warren,
Malte Meinshausen, Gian-Kasper Plattner, 2009, "How well do Integrated Assessment Models
simulate climate change?", Climatic Change, electronic publication date December 10, 2009,
http://www.springerlink.com/content/l841558141481552/

-------
Climate impacts in the PAGE09 model

Prepared for the
Climate Damages Workshop
Washington DC
18-19 November 2010

By

Dr Chris Hope
Judge Business School
University of Cambridge
c.hope@jbs.cam.ac.uk

SEI

ClimateCost
CAMBRIDGE

Judge Business School

-------
Plan of talk

•	The PAGE09 model.

•	Impacts and the social cost of C02.

•	Comparison with results from PAGE2002

ClimateCost

-------
The PAGE09 model

•	A development of the PAGE2002 model

•	Excel 2007 worksheet with @RISK 5.5 add-in

•	4 greenhouse gases

•	8 regions

•	10 analysis years

•	3 impact sectors and discontinuities

•	2 policies and their difference

•	10000 runs to calculate probability distributions of outputs

SEI

ClimateCost

-------
Structure of the PAGE09 model

-------
New features of PAGE09

SEI

ClimateCost

CAMBRIDGE

Judge Business School

-------
Possibility of benefits

proportion of GDP

0.14

0.12
0.1
0.08
0.06
0.04
0.02
0

-0.02

Impact by temperature

(] 0.5 1

1.5

2.5

SEI

ClimateCost

3.5 4 4.5
DegC

11 CAMBRIDGE

Judge Business School

-------
Saturation of impacts

Impact by temperature

proportion of GDP

SEI

ClimateCost

11 CAMBRIDGE

Judge Business School

-------
Impacts as a function of GDP/capita

proportion of GDP

0.18 i
0.16 -
0.14 -
0.12 -
0.1 -
0.08 -
0.06 -
0.04 -
0.02 -
0 —

Impact by GDP/capita

\

n	1	1	1	1	1	1	r

0 10000 20000

30000 40000 50000 60000 700CG 80000 90000

$

SEI

ClimateCost

CAMBRIDGE

Judge Business School

-------
Impacts and SCC02

•	Business as usual scenario: A1B.

•	Low emissions scenario: 2016 r5 low.

•	Moderate adaptation.

•	Currency unit: $2005, PPP exchange rates, EU base year GDP/cap.

•	Pure time preference rate: <0.1,1,2> % per year.

•	EMUC: <0.5,1,2>.

SEI

ClimateCost

-------
1.82

Climate sensiti...

4.64

^ SENS

Minimum

1.2967

Maximum

6.7688

Mean

3.0065

Std Dev

0.8668

Values

10000

		.

fM

CO

m

VD

Source: 10000 PAGE09 runs

SEI

ClimateCost

£!Ki

CAMBRIDGE

Judge Business School

-------
ppm

C02 concentration
by date, A1B scenario

1000
900
800
700
600
500
400
300
200
100

5%

2000

2020

2040

2060

2080

2100

2120

2140

2160

2180

2200

Source: 10000 PAGE09 runs

SEI

ClimateCost



CAMBRIDGE

Judge Business School

-------
ppm

1000
900
800
700
600
500
400
300
200
100
0

2000

C02 concentration
by date, low emissions scenario

95%

mean

5%

2020

2040

2060

2080

2100

2120

2140

2160

2180

2200

Source: 10000 PAGE09 runs

SEI

ClimateCost	if Cambridge

Judge Business School

-------
Global mean temperature rise
by date, A1B scenario

DegC

5 n

4 -

3 -

2 -

1 -

95% x'

mean

MAGICC

2000 2010 2020 2030 2040 2050 2060 2070 2080

2090

2100
Year

Source: 10000 PAGE09 runs

SEI

ClimateCost	ERS Cambridge

Judge Business School

-------
Global mean temperature rise
by date, low emissions scenario

DegC

5 n

4 -

3 -

2 -

1 -

95%

mean

MAGICC

5%

2000 2010 2020 2030 2040 2050 2060 2070 2080 2090

2100
Year

Source: 10000 PAGE09 runs

SEI

ClimateCost

m CAMBRIDGE

Judge Business School

-------
$trillion

50

Global impacts
by date, A1B scenario

95%

40

30

20

10

mean

2000 2020

2040 2060 2080 2100 2120 2140 2160 2180

2200

-10

Source: 10000 PAGE09 runs

SEI

ClimateCost

1§ CAMBRIDGE

ISp* Judge Business School

-------
50

Global impacts by date,
low emissions scenario

$trillion

40

30

20

10

95%

_ _ _			mean

o				—			 ~

2000 2020 2040 2060 2080 2100 2120 2140 2160 2180 5°^200

-10

Source: 10000 PAGE09 runs

SEI	CLimateCost	II Cambridge

¦1	w-w Judge Business School

-------
2.5 i

2.0-

d 1.5 ¦

in
-------
7
6
5

o .

i—' 4

2
1
0

NPV of global impacts, low emissions scena...

0.016 0.212

o
o

fNj

Values in Bill!,,,

Source: 10000 PAGE09 runs; 2016 r5 low scenario

5....

I*--

m——

Total impacts / Total

Minimum
Maximum
Mean
Std Dev
Values

-9558655.4006
2.346E+009
83279386.2362
102059827.7632
10000

SEI

ClimateCost

1§ CAMBRIDGE

ISp* Judge Business School

-------
Social cost of C02, A1B scenario

0.007 -
0.006 -
0.005 -
0.004 ¦
0.003 -
0.002 ¦
0.001 -
0.000

SCC02 in 2...

5....

0.01 0.27
•A LJ F

5....

rsi m

Values in Thousa...

SCC02

Minimum 0.6207
Maximum 5317.7561
Mean 102.6004

Std Dev
Values

247.6347
10000

Source: 10000 PAGE09 runs; A1B scenario

SEI

ClimateCost

1§ CAMBRIDGE

ISp* Judge Business School

-------
The social cost of C02 in 2010

2010- 2200 $US (2005) per tonne

5% mean 95%

A1B Scenario 10 100 270

Low emissions 10 45 120

Source: 10000 PAGE09 model runs

SEI ClimateCost

CAMBRIDGE

Judge Business School

-------
Major influences on the SCC02

SCC02 in 2010, A1B scenario

Regression - Mapped Values

TCR

EMUC

PTP

FRT

W 2

CCF

IND

$/tC02

Source: 10000 PAGE09 runs; A1B scenario

SEI

CLimateCost

®| CAMBRIDGE

ISp* Judge Business School

-------
Comparison with results from PAGE2002

SEI ClimateCost II Cambridge

J—* Judge Business School

-------
SCC02 in PAGE09 and PAGE2002

2010 - 2200 $US per tonne C02

5% mean 95%

PAGE09 10 100 270

PAGE2002 3 28 85

Source: 10000 PAGE09 and PAGE2002 model runs; A1B scenario

SEI ClimateCost

CAMBRIDGE

Judge Business School

-------
Why is the SCC02 so much greater in PAGE09?

• Normalised to EU base year GDP/capita.

• Less effective adaptation.

• Higher chance of a discontinuity.

• Proper accounting for very large impacts.

• $2005 not $2000.

SEI

ClimateCost

-------
Supporting documents

The PAGE09 model: A technical description

Describes the changes to the science, impacts, abatement
costs and adaptation. Appendices with all the equations
and default inputs.

The Social Cost of C02 from the PAGE09 model

Default inputs and first impact results from the model

PAGE09 v1.7 user guide

Contains brief instructions on using the model

SFI ClimateCost H Cambridge

' -1- Judge Business School

-------
FUND - Climate Framework for Uncertainty,
Negotiation and Distribution

David Anthoff*

University of California, Berkeley, CA, USA

Richard S.J. Tol

Economic and Social Research Institute, Dublin, Ireland
Institute for Environmental Studies, Vrije Universiteit, Amsterdam, The

Netherlands

Department of Spatial Economics, Vrije Universiteit, Amsterdam, The

Netherlands

Department of Economics, Trinity College, Dublin, Ireland

4 November 2010

*Contact: anthoff@berkeley.edu

-------
FUND (the Climate Framework for Uncertainty, Negotiation and Distribution) is an integrated
assessment model linking projections of populations, economic activity and emissions to simple
greenhouse gas cycle, climate and sea-level rise models, and to a model predicting and monetizing
welfare impacts. Climate change welfare impacts are monetized in 1995 dollars and are modelled
over 16 regions. Modelled welfare impacts include agriculture, forestry, sea level rise, cardiovascular
and respiratory disorders influenced by cold and heat stress, malaria, dengue fever, schistosomiasis,
diarrhoea, energy consumption from heating and cooling, water resources, unmanaged ecosystems
and tropical and extratropical storms (Link and Tol, 2004). The source code, data, and a technical
description of the model can be found at http://www.fund-model.org.

Essentially, FUND consists of a set of exogenous scenarios and endogenous perturbations. The
model distinguishes 16 major regions of the world, viz. the United States of America, Canada,
Western Europe, Japan and South Korea, Australia and New Zealand, Central and Eastern Europe, the
former Soviet Union, the Middle East, Central America, South America, South Asia, Southeast Asia,
China, North Africa, Sub-Saharan Africa, and Small Island States. Version 3.6, the latest version, runs
to the year 3000 in time steps of one year.

Emission reduction of carbon dioxide, methane and nitrous oxide is specified as in Tol (2006). Simple
cost curves are used for the economic impact of abatement, with limited scope for endogenous
technological progress and interregional spillovers (Tol, 2005).

-------
million by volume, is represented by the five-box model of Maier-Reimer and Hasselmann (1987). Its
parameters are taken from Hammitt et al. (1992).

The radiative forcing of carbon dioxide, methane, nitrous oxide and sulphur aerosols is determined
based on Shine et al. (1990). The global mean temperature, T, is governed by a geometric build-up to
its equilibrium (determined by the radiative forcing, RF), with a half-life of 50 years. In the base case,
the global mean temperature rises in equilibrium by 3.0°C for a doubling of carbon dioxide
equivalents. Regional temperature is derived by multiplying the global mean temperature by a fixed
factor, which corresponds to the spatial climate change pattern averaged over 14 GCMs
(Mendelsohn etal., 2000). The global mean sea level is also geometric, with its equilibrium level
determined by the temperature and a half-life of 50 years. Both temperature and sea level are
calibrated to correspond to the best guess temperature and sea level for the IS92a scenario of
Kattenberg et al. (1996).

The climate welfare impact module, based on Tol (2002a; Tol, 2002b) includes the following
categories: agriculture, forestry, sea level rise, cardiovascular and respiratory disorders influenced by
cold and heat stress, malaria, dengue fever, schistosomiasis, diarrhoea, energy consumption from
heating and cooling, water resources, unmanaged ecosystems and tropical and extratropical storms.
Climate change related damages are triggered by either the rate of temperature change
(benchmarked at 0.04°C/yr) or the level of temperature change (benchmarked at 1.0°C). Damages
from the rate of temperature change slowly fade, reflecting adaptation (cf. Tol, 2002b).

In the model individuals can die prematurely due to temperature stress or vector-borne diseases, or
they can migrate because of sea level rise. Like all welfare impacts of climate change, these effects
are monetized. The value of a statistical life is set to be 200 times the annual per capita income.1
The resulting value of a statistical life lies in the middle of the observed range of values in the
literature (cf. Cline, 1992). The value of emigration is set to be three times the per capita income
(Tol, 1995; Tol, 1996), the value of immigration is 40 per cent of the per capita income in the host
region (Cline, 1992). Losses of dryland and wetlands due to sea level rise are modelled explicitly. The
monetary value of a loss of one square kilometre of dryland was on average $4 million in OECD
countries in 1990 (cf. Fankhauser, 1994). Dryland value is assumed to be proportional to GDP per
square kilometre. Wetland losses are according to estimates from Brander et al. (2006). Coastal
protection is based on cost-benefit analysis, including the value of additional wetland lost due to the
construction of dikes and subsequent coastal squeeze.

1 Note that this implies that the monetary value of health risk is effectively discounted with the pure rate of time
preference rather than with the consumption rate of discount (Horowitz, 2002). It also implies that, after equity
weighing, the value of a statistical life is equal across the world (Fankhauser et al., 1997).

-------
impacts, depending on the speed of adaptation. The impacts of not being fully adapted to new
climate conditions are always negative (cf. Tol, 2002b).

Vulnerability to climate change changes with population growth, economic growth, and technological
progress. Some systems are expected to become more vulnerable, such as water resources (with
population growth) and heat-related disorders (with urbanization), or more valuable, such as
ecosystems and health (with higher per capita incomes). Other systems are projected to become less
vulnerable, such as energy consumption (with technological progress), agriculture (with economic
growth) and vector- and water-borne diseases (with improved health care) (cf. Tol, 2002b).

References

Batjes, J. J. and C. G. M. Goldewijk (1994). The IMAGE 2 Hundred Year (1890-1990) Database of the
Global Environment (HYDE), RIVM, Bilthoven, 410100082.

Brander, L., R. Florax and J. Vermaat (2006). "The Empirics of Wetland Valuation: A Comprehensive
Summary and a Meta-Analysis of the Literature." Environmental and Resource Economics
33(2): 223-250.

Cline, W. R. (1992). The Economics of Global Warming. Washington, DC, Institute for International
Economics.

Fankhauser, S. (1994). "Protection vs. Retreat - The Economic Costs of Sea Level Rise." Environment
and Planning A 27(2): 299-319.

Fankhauser, S., R. S. J. Tol and D. W. Pearce (1997). "The Aggregation of Climate Change Damages: A
Welfare Theoretic Approach." Environmental and Resource Economics 10(3): 249-266.

Hammitt, J. K., R. J. Lempert and M. E. Schlesinger (1992). "A Sequential-Decision Strategy for Abating
Climate Change." Nature 357: 315-318.

Horowitz, J. K. (2002). "Preferences in the Future." Environmental and Resource Economics 21: 241-
259.

Kattenberg, A., F. Giorgi, H. Grassl, G. A. Meehl, J. F. B. Mitchell, R. J. Stouffer, T. Tokioka, A. J. Weaver
and T. M. L. Wigley (1996). Climate Models - Projections of Future Climate. Climate Change
1995: The Science of Climate Change -- Contribution of Working Group I to the Second
Assessment Report of the Intergovernmental Panel on Climate Change. J. T. Houghton, L. G.
Meiro Filho, B. A. Callanderet al. Cambridge Cambridge University Press: 285-357.

Leggett, J., W. J. Pepper and R. J. Swart (1992). Emissions scenarios for the IPCC: an update. Climate
Change 1992 - The Supplementary Report to the IPCC Scientific Assessment. J. T. Houghton,
B. A. Callander and S. K. Varney. Cambridge, Cambridge University Press: 71-95.

-------
Link, P. M. and R. S. J. Tol (2004). "Possible Economic Impacts of a Shutdown of the Thermohaline

Circulation: an Application of FUND." Portuguese Economic Journal 3(2): 99-114.
Maier-Reimer, E. and K. Hasselmann (1987). "Transport and Storage of Carbon Dioxide in the Ocean:

An Inorganic Ocean Circulation Carbon Cycle Model." Climate Dynamics 2: 63-90.
Mendelsohn, R. O., M. E. Schlesinger and L. J. Williams (2000). "Comparing Impacts across Climate

Models." Integrated Assessment 1: 37-48.

Shine, K. P., R. G. Derwent, D. J. Wuebbles and J. J. Morcrette (1990). Radiative Forcing of Climate.
Climate Change - The IPCC Scientific Assessment. J. T. Houghton, G. J. Jenkins and J. J.
Ephraums. Cambridge Cambridge University Press: 41-68.

Tol, R. S. J. (1995). "The Damage Costs of Climate Change - Towards More Comprehensive

Calculations." Environmental and Resource Economics 5: 353-374.

Tol, R. S. J. (1996). "The Damage Costs of Climate Change: Towards a Dynamic Representation."

Ecological Economics 19: 67-90.

Tol, R. S. J. (2002a). "Estimates of the damage costs of climate change. Part 1: Benchmark estimates."

Environmental and Resource Economics 21(2): 47-73.

Tol, R. S. J. (2002b). "Estimates of the damage costs of climate change. Part 2: Dynamic estimates."

Environmental and Resource Economics 21(2): 135-160.

Tol, R. S. J. (2005). "An Emission Intensity Protocol for Climate Change: An Application of FUND."

Climate Policy 4: 269-287.

Tol, R. S. J. (2006). "Multi-Gas Emission Reduction for Climate Change Policy: An Application of
FUND." Energy Journal 27: 235-250.

-------
The FUND model

David Anthoff

Department of Agricultural & Resource Economics
University of California, Berkeley

-------
Outline

• FUND model

— Basic structure

— Impacts

— Planed model developments

• Catastrophes

• Social Cost of Carbon-WG

-------
FUND

-------
Scenario

Exogenous

• GDP

• Population

• Energy and carbon intensity

• Land use change and
deforestation C02 emissions

• CH4 emissions

• N02 emissions

Endogenous

• C02 emissions

• C02 emissions from
"dynamic biosphere"

• SF6 emissions

• S02 emissions

-------
Physical Components

• All gas cycles explicitly modeled (C02, CH4,
N20, S02)

• RF for each gas explicitly modeled

• Climate Sensitivity Uncertain

• Adjust transient climate response properly!

-------
Health Impacts

Mortality (#)

Vector born diseases
Dengue fever
Malaria

Schistosomiasis
Diarrhoea
Cardiovascular
Cold
Heat
Respiratory
Extratropical storms
Tropical storms

Morbidity (years)

Vector born diseases
Dengue fever
Malaria

Schistosomiasis
Diarrhoea
Cardiovascular
Cold
Heat
Respiratory

Value of a Statistical Life

WTP

-------
Sea-level Rise

• Based on Fankhauser (1994), updated

• Cost of protection

• Value of lost dryland

• Value of lost wetland

• Cost of Emigration

• Cost of Imigration

-------
More Impacts

• Agriculture

• Tropical Storms

• Extratropical Storms

• Forestry

• Heating Energy

• Cooling Energy

• Water Resources

• Species Loss

-------
Damages

Dfr ^ Ds(T•¦¦)

/ 7

Ds(T,y, ...) = a^—

'7 = 0 additive damage

7 = 1 multiplicative damage

y > 1 "luxury good" type damage

.7 < 0 "Schelling" type damage

Uncertain in Monte Carlo mode

-------
Future Plans on Im

Ocean Acidification

— Corral reefs

— Shell fish

Tourism
River floods

Update energy consumption

-------
Catastrophes

Ceronsky et al. (20052010 hopefully!) „Checking
the price tag on catastrophe: The social cost of
carbon under non-linear climate response/' FNU
Working Paper 87

• Thermohaline circulation collapse

• Marine methane hydrate destabilization

• High Climate Sensitivities

-------
Social Cost of Carbon - WG

-------
Scenario Uncertainty

Scenario YpC in thousand $ Probability

IMAGE 43.4 20%

MERGE 27.8 20%

MESSAGE 32.1 20%

MiniCAM 42.6 20%

5th Scenario 35.7 20%

Calibrated
Roe & Ba ker

Forster/Gregory 06
Frame 05

— Knutti 02

— AndronovaOl

— Forest 06 (02 dashed)

— Gregory 02
Hegerl palaeo 06

— Schneider LGM 06

— - Annan LGM 05

Equilibrium Climate Sensitivity
(°C)

£•
w
c
-------
Discounting

-------
Distribution

Case 1 Case 2

Country

Damage

| Country

Damage

Italy

0.9%

Italy

1.0%

Rwanda

13.2%

Rwanda

1.0%

Country

Damage

World Average

1.0%

Anthoff, D. and R. S. J. Tol (2009). "The Impact of Climate Change on the Balanced Growth Equivalent: An Application of FUND."

Environmental and Resource Economics 43(3): 351-367
Anthoff, D. and R. S. J. Tol (2010). "On international equity weights and national decision making on climate change."

Journal of Environmental Economics and Management 60(1): 14-20

-------
Thank you!

anthoff@berkeIey.edu
http://www.david-anthoff.de

-------
Representation of Climate
Impacts in GCAM

Leon Clarke

Workshop on Modeling Climate Change Impacts and

Associated Economic Damages

Washington DC

Thursday, November 18, 2010

-------
What is GCAM?

Builds on the energy/economy model of Edmonds and Reilly completed three decades
ago.

Combines economics-based energy, agricultural models with an Integrated Climate
Assessment Model (MAGICC).

Dynamic-recursive model.

Technologically detailed ntegrated assessment model.

14 geopolitical regions

Emissions of 16 greenhouse gases and short-lived species: C02, CH4, N20,
halocarbons, carbonacious aerosols, reactive gases, sulfur dioxide.

Runs through 2095 in 15-year time steps (moving to variable time steps).

-------
Geospaciallyjsnergy supply data

. \c£

/ v
tr

vy^ft
( }0 c \

Geospacially explicit land use data

GCAM

NATIONAL LABORATORY

-------
What impacts would we want to consider in
PNNL/JGCRIIA modeling?

~ Goal #1: Pick things that are important.

~ Goal #2: Pick things that involve interactions among the
various systems represented in IA models

~ Goal #3: Pick things that we actually have a chance of doing.

~ Expanding on #3: A primary benefit of IA models is their ability
capture interactions between systems. This leads to a
perspective on impacts in which we distinguish between

¦ Those which are most amenable to an integrated perspective.

¦ Those which can "hang" off of the model and not feed back to other
systems in the model.

~ Although integrated analysis brings impacts together in an
integrated system, aggregating and monetizing all impacts is
not inherently core to considering impacts in GCAM.

Pacific Northwest

NATIONAL LABORATORY

-------
A Plan for Impacts in PNNL/JGCRI's IA
Modeling

Near-term priorities;

¦ Agriculture, Forestry, Land Use and Land
Cover

¦ Energy Use

¦ Ocean Acidification

High priorities with substantial model
development necessary

¦ Water Resources

¦ Sea-Level Rise and Coastal Impacts

¦ Human Health and Demographics

Of substantial interest, but not easily
quantifiable within existing IA models

¦ Extreme Events and Thresholds

¦ Biodiversity

ENERGY

Bringing in Impacts: Strategies for
Integrated Assessment Model
Development

AM Thomson, AC Janetos, RC Izaurralde, LE Clarke, SJ Smith, HM Pitcher,
KV Calvin

July 2008

Prepared for the US Environmental Protection Agency, Climate Change Division under
Interagency Agreement AGRDW89921952-01.

Pacific Northwest

NATIONAL LABORATORY

-------
A Plan for Impacts in PNNL/JGCRI's IA
Modeling

There are many ways to pursue
these impacts:

¦ One dimension

All in GCAM.

• Linkages to other models (iESM
regional initiatives).

¦ Another dimension

Endogenous interactions within
the model - feedbacks with
other systems.

"Hanging" off of GCAM.

ENERGY

Bringing in Impacts: Strategies for
Integrated Assessment Model
Development

AM Thomson, AC Janetos, RC Izaurralde, LE Clarke, SJ Smith, HM Pitcher,
KV Calvin

July 2008

Prepared for the US Environmental Protection Agency, Climate Change Division under
Interagency Agreement AGRDW89921952-01.

Pacific Northwest

NATIONAL LABORATORY

-------
Land Use Impacts

-------
GCAM Moving to an Agro Ecological Zone
Formulation for AgLU

OPTIONS FOR FEEDBACK

Link results from ecosystem
models

(EPIC/BIOME/CENTURY) and
ESMs to GCAM by changing
parameters.

Use sensitivity studies to begin
to develop a concept of the
scale of impacts in the context
of integrated assessment and
adjust GCAM parameters.

Develop a reduced-form
representation of ecosystem
processes and response to
climate change in GCAM.

Leciend

Legend
ctrygaez
Bio2050

1%
4%
I 5%

Bioenergy Land, as a Fraction of Total
Land, in 2050 in a Reference Scenario

Legend
ctrygaez
Crops2050

Crop Land, as a Fraction of Total
Land, in 2050 in a Reference Scenario

Pacific Northwest

NATIONAL LABORATORY

-------
Synthesis of process-level impact studies

(a) Maize, mid- to high-latitude

(b) Maize, low latitude

1 2 3 4 5
Mean local temperature change (°C)

jT_\

1 2 3 4 5
Mean local temperature change (°C)

(d) Wheat, low latitude

1 2 3 4 5
Mean local temperature change (°C)

Pacific Northwest

NATIONAL LABORATORY

-------
Climate impacts interact with mitigation
policy.

No Policy Case with climate change impacts
500: Climate change impacts
No Policy Case
500: UCT policy case

By 2095, ILUC
emissions go below
0 with climate policy
cases

2095

Pacific Northwest

NATIONAL LABORATORY

-------
Impact of Impacts on Costs of Mitigation

0.25

£ 0.20

5 0.15

0.10

0.05

0.00

UCT w/o Impacts
UCT w/ Impacts

450 ppm

500 ppm

550 ppm

-------
Energy Impacts

-------
Effects of Changing Degree Days on
Building Energy Consumption:

The Reference Case of China Buildings

Fixed HDD of 2158
Fixed CDD of 1046

traditional biomass
refined liquids enduse
Idelivered biomass
I district heat

m m m

100

1990 2005 2020 2035 2050 2065 2080 2095

HDD decreasing from 2158 to 1458
CDD increasing from 1046 to 1746

traditional biomass
refined liquids enduse
delivered biomass
I district heat
delivered gas
Idelivered coal
elect_td_bld
•Primary Energy

Other Long-Term Options: (1) Feedbacks on power plant
efficiencies, (2) Feedbacks on water supply for hydroelectric
power.

100

0 ======H 0

1990 2005 2020 2035 2050 2065 2080 2095

Pacific Northwest

NATIONAL LABORATORY

-------
Water

-------
Energy Demand

-------
GCAM: U.S. Energy System Results

120

100

Withdrawals - no policy

1990 2005 2020 2035 2050 2065 2080 2095
gas "coal "nuclear ¦ electricityw/CCS electricity

Consumption - no policy

1990 2005 2020 2035 2050 2065 2080 2095
gas "coal "nuclear ¦ electricity w/CCS electricity

Withdrawals - 515ppm
policy

1990 2005 2020 2035 2050 2065 2080 2095
gas ¦ coal ¦ nuclear ¦ electricity w/CCS electricity

Consumption - 515ppm policy

1990 2005 2020 2035 2050 .,.2065 ,2080 , 2095

Pacific Northwest
gas "coal "nuclear ¦ electricityw/Q&&>i\ electricity ;rv

-------
Examples of Linkages between Platforms:

iESM

-------
A Research Collaboration Between Three
National Laboratories: PNNL, ORNL and
LBNL

ENERGY

J. S Department of Eiicy)
I AC05 76R101830

Improving the Representations of
Human-Earth System Interactions

Principal In\estigators:

James A. Edmonds, Ph.D.. Pacific Northwest National Laboratory (Lead)

John B. Drake, Ph.D.. Oak Ridge National Laboratory

William D. Collins, Ph.D., Lawrence Berkeley National Laboratory

February 2009

Revised May 2009 per client request

Proposal to the U.S. Department of Energy, Office of Science
Cltmate and Environmental Sciences Division

Climate Change Modeling LAB 09-06
Strengthening the Coupling Between Climate and
Earth System Models (ESM) and Integrated Assessment Models (IAM)

Three Primary Tasks

Create a first generation integrated
Earth System Model (iESM) with both
the human components of an 1AM and a
physical ESM;

Further develop components and
linkages within the iESM and apply the
model to improve our understanding of
the coupled physical, ecological, and
human system;

Add realistic hydrology, including

freshwater demand, allocations, and
demands to hold stocks of water as well
as representations of freshwater
availability from surface water, ground
water, and desalinization

Pacific Northwest

NATIONAL LABORATORY

-------
ESM Phase 1 Initial Coupling Strategy

-------
THE CLM: Initial One-way Coupling: Land Use and Land
Cover Change (iESM Control experiment)

-------
iESM multi-phase coupling strategy

-------
Examples of Linkages between Platforms:

Regional Initiative

-------
Beard

Kernel

Leaves

RGCAM: U.S. test Regio

Legend

! ] Field Crops

I ' Herbaceous Vegetation

Woody Vegetation

- 45°N

- *0°N

- 25°N

-i 1— 1 r

12D*W 110°W 100°Yi WW

80°W

974 27$ 77» 3*n 2K aa* ass ses aw kk 298 aoo ace

Regional Climate Model

Major Land Cove/in the 14 States N

0 45 90 180 270 360

Miles

45°N -
40°N -
35°N "
30°N -
25°N "
20°N -

135=W 1SJ°W 1WW WW

7S°W

-------
Buildings Demand Modeling

Whole Building,
Engineering/

BEND Model

~4000 buildings will be simulated in
EnergyPlus to represent the buildings in
the RGCAM U.S. test region:

¦ 4 climate zones

¦ 11 commercial building types

¦ 3 residential building types

¦ 6-9 sizes within each building type

¦ 7-8 vintages of existing buildings and 3
vintages of new buildings

Building characteristic vary for each
combination of attributes

Hourly (8760 hours) electrical output
used to calibrate models and determine
building weights based on actual
weather and actual hourly electric
consumption for test region.

Our challenge is to pass data back and
forth between BEAMS and R-GCAM.

-------
Discussion

Tver green

Pacific Northwest

NATIONAL LABORATORY

-------
Climate Damages in the MIT IGSM
John Reilly

MIT Joint Program on the Science and Policy of Global Change

Integrated assessment models (lAMs) have proven useful for analysis of climate change because they
represent the entire inhabited earth system, albeit typically with simplified model components that are
reduced form or more highly aggregated than for example, high resolution coupled atmosphere-ocean-
land general circulation models. The MIT Integrated Global System Model has been developed to retain
the flexibility to assemble earth system models of variable resolution and complexity, however, even at
its simplest it remains considerably more complex than most other lAMs. In its simplest formulation it
retains a full coupled general circulation model of the ocean and atmosphere. Solved recursively, it
solution time for a 100-year integration on a single node of computer cluster is on the order of 24-36
hours, compared with seconds or minutes for other lAMs. In that form it is not numerical feasible to
solve the whole system as a fully dynamic optimizing model to find an optimal cost-benefit solution as
with the DICE, PAGE, or FUND models. Indeed, inclusion of climate damages is still a work in progress in
the MIT IGSM. The slow progress relative to other efforts stems from a commitment to represent
explicitly the physical impacts of climate and environmental change on activities (e.g. crop yields, water
availability, coastal, inundation, ecosystem processes and functioning, health outcomes, etc.) and
represent market response to these outcomes and value that response consistent with projections of
resource prices as they are projected to change in the future with economic growth and under different
policies to mitigate greenhouse gas emissions. This is in contrast to most of the optimizing models
where climate damages are estimated as a reduced form relationship in dollars of economic loss as a
function of mean global temperature change as a sufficient indicator of many dimensions of climate
change, and where the damage function is itself completely independent and separable from the
economy as it affects energy use and greenhouse gas emissions. In the "horses for courses" metaphor,
the MIT IGSM is not a horse designed (bred) to run well if the course is to estimate a net present value
social cost of carbon. The IGSM is best seen as complementary to such efforts, and probably the focus
on uncertainty in future climate outcomes is one of the areas where it can make the most contribution
to the social cost of carbon discussion.

Computationally efficient versions of the IGSM have been assembled for simulating large ensembles to
study uncertainty (Sokolov et al., 2009; Webster etal., 2009). Less complete but more highly-resolved
model components can be combined where research demands them, such as in the study of the climate
effect of aerosols {Wang, 2009; Wang etal., 2009a,b), changes in atmospheric composition and human
health (Selin etal., 2009a) or agricultural impacts and land use change (Reilly, et al. 2007; Felzer et al.,
2005; Melillo etal., 2009). The IGSM framework encompasses the following components:

• global economic activity resolved for large countries and regions that projects changes in
human activities as they effect the earth system including emissions of pollutants and
radiatively active substances and changes in land use and land cover;

-------
• earth system modules linked to the macroeconomy that address effects of climate and
environmental change on human activity, adaptation, and their consequences for the
macroeconomy (this includes modules that represent water use and land use at
disaggregated spatial scales, energy and coastal infrastructure again at disaggregate spatial
scales, and demography, urbanization, urban air chemistry, and epidemiological
relationships that relate environmental change to human health);

• the natural and managed land system including vegetation, hydrology, and biogeochemistry
as affected by human activity, environmental change and feedbacks on climate and
atmospheric composition;

• the circulation and biogeochemistry of the ocean including its interactions with the
atmosphere, and representations of physical and biological oceanic responses to climate
change; and

• the circulation and chemistry of the atmosphere including its role in radiative forcing, and
interactions with the land and ocean that determine climate change.

The suite of models that have been employed in this framework and their capabilities are briefly
described below.

3.1 Human Drivers and Analysis of Impacts

Human activities as they contribute to environmental change or are affected by it are represented in
multi-region, multi-sector models of the economy that solves for the prices and quantities of interacting
domestic and international markets for energy and non-energy goods as well as for equilibrium in factor
markets. The MIT Emissions Predictions and Policy Analysis (EPPA) model (Paltsev et al., 2005) covers
the world economy. It is built on the GTAP dataset (maintained at Purdue University) of the world
economic activity augmented by data on the emissions of greenhouse gases, aerosols and other relevant
species, and details of selected economic sectors. The GTAP database allows flexibility to represent the
world economy with greater country or sector detail (the data set has 112 countries/regions and 57
economic sectors) that we aggregate further for numerical efficiency. The model projects economic
variables (GDP, energy use, sectoral output, consumption, etc.) and emissions of greenhouse gases (C02,
CH4, N20, HFCs, PFCs and SF6) and other air pollutants (CO, VOC, NOx, S02, NH3, black carbon, and
organic carbon) from combustion of carbon-based fuels, industrial processes, waste handling, and
agricultural activities.

The model has been augmented with supplemental physical accounts to link it with the earth system
components of the IGSM framework. To explore land use and environmental consequences, the EPPA
model (Gurgel, et al., 2007; Antoine, et al.,2008) is coupled with the Terrestrial Ecosystem Model
(Melillo et al., 2009). The linkage allows us to examine the ability of terrestrial ecosystems to supply
biofuels to meet growing demand for low-emissions energy sources along with the growing demand for
food, and to assess direct and indirect emissions from an expanded cellulosic bioenergy program. The
approach generates worldwide land-use scenarios at a spatial resolution of 0.52 latitude by 0.52
longitude that varies with climate change. To analyze the economic impacts of air pollution, the EPPA
model is extended to include pollution-generated health costs, which reduce the resources available to
the rest of the economy (Nam et al., 2009; Selin et al., 2009a). The model captures the amount of labor

-------
and leisure lost and additional medical services required due to acute and chronic exposure to
pollutants. The GTAP database allows considerable flexibility to represent the world economy with
greater country or sector detail (the underlying data has 112 countries/regions and 57 economic
sectors). To assess distributional and regional impacts of carbon policy in the US, we use a model that is
based on a state-level database and resolves large U.S. states and multi-state regions and households of
several income classes. The U.S. Regional Energy Policy (USREP) model (Rausch etal., 2009; 2010) is
nearly identical in structure to the EPPA model, except that it models states and multi-state regions in
the US instead of countries and multi-country regions. The main difference from the EPPA model is the
foreign sector that is represented as export supply and import demand functions rather than a full
representation of foreign economies. This sacrifice of global coverage allows explicit modeling of
distributional details of climate legislation and linking the USREP model to very detailed electricity
dispatch models. Efforts, under separate funding, to integrate the USREP database into the GTAP base
to provide a complete representation of trade are underway. Physical impacts of environmental
change have been included in the model as a feedback by identifying factors (land productivity as it
affects crops, livestock and forests) or sectors affected by climate or by introducing additional
household production sectors (household health services that uses leisure and medical services). Thus,
the approach is to work with underlying input-output and Social Accounting Matrix (SAM) that is the
basis for the economic model (Matus, etal., 2008). This provides a framework for potentially linking
other impacts such as coastal (Francketal., 2010a,b, 2010; Sugiyama, etal., 2008), agriculture (Reilly et
al., 2007), health (Selin, et al., 2009; Nam et al., 2010), or water (Strzepek et al., 2010) impacts.

3.2 Hydrology and Water Management

Research on components representing water management are aimed at linking hydrological changes
projected by the atmospheric component of the IGSM to impacts of those changes on water availability
and use for irrigation, energy, industry and households, and in-stream ecological services. These
demands are driven by macroeconomic changes and changes in water supply and will in turn affect the
economy as represented in the EPPA and the USREP models. Techniques have been developed to take
IGSM 2-D GCM outputs and use results from the IPCC AR-4 3-D GCMs to provide IGSM-generated 3-D
climates to the hydrology component of the IGSM-Land Surface Model (NCAR Community Land Model,
CLM) to project runoff. Tests have been conducted for the US, where adequate data are available, to
determine the spatial resolution needed to provide reliable estimates of runoff using CLM. A Water
Resources System (WRS) model has been adapted from and further developed in collaboration with the
International Food Policy Research Institute (IFPRI) to represent river reaches and natural and
management components that affect stream-flow. The major natural components are wetlands,
unmanaged lakes, groundwater aquifers and flood plains. The major managed components are
reservoirs and managed lakes, and water diversions for irrigation, cooling in thermal power plants, and
industrial and household needs. Constraints on use to preserve in-stream ecological water requirements
can be imposed.

-------
etal., 2010a). A model of Municipal and Industrial water demand driven by per capita GDP was
developed jointly with the University of Edinbourgh (Hughes et al., 2010; Strzepek et al., 2010a). To
investigate changes in thermal electric cooling water demands, a geospatial methodology based on
energy generation and geo-hydroclimatic variables has been developed {Strzepek et al., 2010b). An
assessment of environmental flow requirements to assure aquatic ecosystem viability has been
undertaken and an approach for using the IGSM was selected (Strzepek & Boehlert, 2010; Strzepek et
al., 2010a). These developments provide the foundation for completing linkages of the WRS with other
IGSM components.

3.3 Atmospheric Dynamics and Physics

Research utilizing the IGSM framework has typically included a 2-D atmospheric (zonally-averaged
statistical dynamical) component based on the Goddard Institute for Space Studies (GISS) GCM. The
IGSM version 2.2 couples this atmosphere with a 2D ocean model (latitude, longitude) with treatment of
heat and carbon flows into the deep ocean (Sokolov et al, 2005). The IGSM version 2.3 (where 2.3
indicates the 2-D atmosphere/full 3-D ocean GCM configuration) (Sokolov et al., 2005; Dutkiewicz et al.,
2005) is a fully-coupled Earth system model that allows simulation of critical feedbacks among its
various components, including the atmosphere, ocean, land, urban processes and human activities. A
limitation of the IGSM2.3 is the above 2-D (zonally averaged) atmosphere model that does not permit
direct regional climate studies. For investigations requiring 3-D atmospheric capabilities, the National
Center for Atmospheric Research (NCAR) Community Atmosphere Model version 3 (CAM3) (Collins et
al., 2006) has been used with offline coupling.

The IGSM2.3 provides an efficient tool for generating probabilistic distributions of sea surface
temperature (SST) and sea ice cover (SIC) changes for the 21st century under varying emissions scenarios,
climate sensitivities, aerosol forcing and ocean heat uptake rates. Even though the atmospheric
component of the IGSM2.3 is zonally-averaged, it provides heat and fresh-water fluxes separately over
the open ocean and over sea ice, as well as their derivatives with respect to surface temperature. This
resolution allows the total heat and fresh-water fluxes for the IGSM2.3 oceanic component to vary by
longitude as a function of SST so that, for example, warmer ocean locations undergo greater evaporation
and receive less downward heat flux.

In offline coupling between the IGSM2.3 and CAM3, the 3-D atmosphere is driven by the IGSM2.3 SST
anomalies with a climatological annual cycle taken from an observed dataset (Hurrell et al., 2008),
instead of the full IGSM2.3 SSTs, to provide a better SST annual cycle, and more realistic regional
feedbacks between the ocean and atmospheric components. This approach yields a consistent regional
distribution and climate change over the 20th century as compared to observational datasets, and can
then be used for simulations of the 21st century.

3.4 Urban and Global Atmospheric Chemistry and Aerosols

The model of atmospheric chemistry includes an analysis of all the major climate-relevant reactive
gases and aerosols at urban scales coupled to a model of the chemistry of species exported from
urban/regional areas (plus the emissions from non-urban areas) at global scale. For calculation of the

-------
atmospheric composition in non-urban areas, the atmospheric dynamics and physics model is linked to a
detailed 2-D zonal-mean model of atmospheric chemistry. The atmospheric chemical reactions are thus
simulated in two separate modules: one for the sub-grid-scale urban chemistry and one for the 2-D
model grid. In addition, offline studies also utilize the 3-D capabilities of the CAM3 as noted above, as
well as the global Model of Atmospheric Transport and Chemistry (MATCH; Rasch et al., 1997), and the
GEOS-Chem global transport model (http://geos-chem.org/).

Global Atmospheric Chemistry: Modeling of atmospheric composition at global scale is by the above 2-D
zonal-mean model with the continuity equations for trace constituents solved in mass conservative or
flux form (Wang et al., 1998). The model includes 33 chemical species including black carbon aerosol,
and organic carbon aerosol, and considers convergences due to transport, convection, atmospheric
chemical reactions, and local production/loss due to surface emission/deposition. The scavenging of
carbonaceous and sulfate aerosol species by precipitation is included using a method based on a
detailed 3-D climate-aerosol-chemistry model (Wang, 2004) that has been developed in collaboration
with NCAR. The interactive aerosol-climate model is used offline to model distributions of key chemical
species, such as those utilized in the development of the urban air chemistry model.

Urban Air Chemistry: A reduced-form urban chemical model that can be nested within coarser-scale
models has been developed and implemented to better represent the sub-gridscale urban chemical
processes that influence air chemistry and climate (Cohen & Prinn, 2009). This is critical both for
accurate representation of future climate trends and for our increasing focus on impacts, especially to
human health and down-wind ecosystems. The MIT Urban Chemical Metamodel (UrbanM) is an update
of our Mayer et al. (2000) model, and applies a third-order polynomial fit to the CAMx regional air
quality model (ENVIRON, 2008) for 41 trace gases and aerosols for a 100 km x 100 km urban area. While
a component of the IGSM, the urban modular UrbanM is also designed to facilitate inclusion in a
number of other global atmospheric models. It has recently been embedded in the MIT interactive
climate-aerosol simulation based on CAM3 in order to assess its influence on the concentration and
distribution of aerosols in Asia (Cohen et al., 2009). Work is underway to further test the sensitivity of
the probabilistic uncertainty results with the IGSM2.2/2.3 to this improved representation of urban
chemistry. The UrbanM is presently being benchmarked in a case study of the Northeast U.S., and
embedded in a global 3-D chemistry-climate model including a detailed chemical mechanism (NCAR
CAM-Chem).

-------
impact of absorbing aerosols on the Indian summer monsoon {Wang et al., 2009b). The UrbanM has
recently been introduced into this model to study the roles of urban processing in global aerosol
microphysics and chemistry and to compute the abundance and radiative forcing of anthropogenic
aerosols (Cohen et al., 2010). This effort also serves as the first step toward introducing the full UrbanM
into the 3-D aerosol-chemistry-climate framework.

3.5 Ocean Component

The IGSM framework retains the capability to represent ocean physics and biogeochemistry in several
different ways depending on the question to be addressed. It can utilize either the 2-D (latitude-
longitude) mixed-layer anomaly-diffusing ocean model or the fully 3-D ocean general circulation model
(GCM). The IGSM with the 2-D ocean is more computationally efficient and more flexible for studies of
uncertainty in climate response. In applications that need to account for atmosphere-ocean circulation
interactions, or for more detailed studies involving ocean biogeochemistry, the diffusive ocean model is
replaced by the fully 3D ocean GCM component.

2-D Ocean Model: The IGSM2.2 has a mixed-layer anomaly-diffusing ocean model with a horizontal
resolution of 4^ in latitude and 5^ in longitude. Mixed-layer depth is prescribed based on observations as
a function of time and location. Vertical diffusion of anomalies into the deep ocean utilizes a diffusion
coefficient that varies zonally as well as meridionally. The model includes specified vertically-integrated
horizontal heat transport by the deep oceans, and allows zonal as well as meridional transport. A
thermodynamic ice module has two layers and computes the percentage of area covered by ice and ice
thickness, and a diffusive ocean carbon module is included (Sokolov et al, 2005; Holian etal., 2001;
Follows et al. 2006).

3-D Ocean General Circulation Model: The IGSM2.3 ocean component is based on a state-of-the-art 3D
MIT ocean GCM (Marshall et al., 1997). Embedded in the ocean model is a thermodynamic sea-ice
module (Dutkiewicz etal., 2005). The 3D ocean component is currently configured in either a coarse
resolution (4° by 4° horizontal, 15 layers in the vertical) or higher resolution (2° by 2.5°, 23 layers; or
alternate configuration with higher resolution in the topics) depending on the focus of study and the
computational resources available. The efficiency of ocean heat uptake can be varied (e.g., Dalan et al.
2005) and the coupling of heat, moisture, and momentum can be modified for process studies (e.g.,
Klima 2008). In addition, a biogeochemical component with explicit representation of the cycling of
carbon, phosphorus and alkalinity can be incorporated. Export of organic and particulate inorganic
carbon from surface waters is parameterized and biological productivity is modelled as a function of
available nutrients and light (Dutkiewicz etal., 2005). Air-sea exchange of C02 allows feedback between
the ocean and atmosphere components. An additional module with explicit representation of the
marine ecosystem (Follows et al., 2007) has been introduced in an "offline" (i.e. without full feedbacks
to the full IGSM) configuration (see further discussion in Section 4.2.3).

3.6 Land and Vegetation Processes

The Global Land System (GLS, Schlosser et al., 2007) of the IGSM links biogeophysical, ecological, and
biogeochemical components: (1) the NCAR Community Land Model (CLM), which calculates the global,

-------
terrestrial water and energy balances; (2) the Terrestrial Ecosystems Model (TEM) of the Marine
Biological Laboratory, which simulates carbon (C02) fluxes and the storage of carbon and nitrogen in
vegetation and soils including net primary production and carbon sequestration or loss; and (3) the
Natural Emissions Model (NEM), which simulates fluxes of CH4 and N20, and is now embedded within
TEM. A recent augmentation to the GLS enables a more explicit treatment of agricultural processes and
a treatment of the managed water systems (Strzepek etal., 2010a). The linkage between
econometrically based decisions regarding land use (from EPPA) and plant productivity from TEM has
been enhanced (Cai et al., 2010). And the treatment of migration of plant species to include
meteorological constraints (i.e. winds) to seed dispersal has been enhanced (Lee et al., 2009, 2010a,b).
The representation of natural and vegetation processes also includes a diagnosis of the expansion of
lakes and changes of methane emissions from thermokarst lake expansion/degradation (Gao etal.,
2010; Schlosser et al., 2010). In addition, continuing updates to CLM and TEM are also incorporated into
the GLS framework. In all these applications, the GLS is operating under a range of spatial resolutions
(from zonal to gridded as low as 0.52), and is configured in its structural detail to accommodate various
levels of process-oriented research both in a coupled framework within the IGSM as well as in
standalone studies (i.e. with prescribed atmospheric forcing).

Antoine, B., A. Gurgel, J. M. Reilly, 2008: Will Recreation Demand for Land Limit Biofuels Production?
Journal of Agricultural & Food Industrial Organization, 6(2), Article 5.
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint08-15.pdf)

Babiker, M., A. Gurgel, S. Paltsev and J. Reilly, 2009: Forward-looking versus recursive-dynamic modeling
in climate policy analysis: A comparison. Economic Modeling, 26(6): 1341-1354; MIT Joint Program
Reprint 2009-8 (http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint09-8.pdf).

Cai, Y., D. Kicklighter, A. Gurgel, S. Paltsev, T. Cronin, J. Reilly, and J. Melillo, 2010; Green House Gas
Mitigation, Bio-fuel and Land-use Change: A Dynamic Analysis. Poster presentation. 2010 DOE
Climate Change Modeling Program Science Team Meeting, Gaithersburg, Maryland, March 29-April
2.

Cohen, J. B., C. Wang, and R. Prinn, 2009: The Impact of Detailed Urban Scale Processing on the

Simulation of the Concentration and Distribution of Aerosols in Asia. Eos. Trans. AGU, 90(52), Fall
Meet. Suppl., Abstract A14A-05.

Cohen, J.B., and R. Prinn, 2009: Development of a fast and detailed model of urban-scale chemical and
physical processing. MIT Joint Program Report 181, October, 74 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC Rptl81.pdf).

Collins, W.D, C.M. Bitz, M.L. Blackmon, G.B. Bonan, C.S. Bretherton, J.A. Carton, P. Chang, S.C. Doney,
J.J. Hack, T.B. Henderson, J.T. Kiehl, W.G. Large, D.S. McKenna, B.D. Santer and R.D. Smith. 2006: The
Community Climate System Model Version 3 (CCSM3). Journal of Climate, 19(11): 2122-2143.

Dalan, F., P.H. Stone, I.V. Kamenkovich, J.R. Scott, 2005: Sensitivity of the ocean's climate to diapycnal
diffusivity in and EMIC. Part I: equilibrium state. J. of Physical Oceanography, 18, 2460-2481; MIT
Joint Program Reprint 2005-6 (http://globalchange.mit.edu/files/document/MITJPSPGC Reprint05-
6.pdf)

-------
Dutkiewicz. S., A. Sokolov, J. Scott and P. Stone, 2005: A Three-Dimensional Ocean-Seaice-Carbon Cycle
Model and its Coupling to a Two-Dimensional Atmospheric Model: Uses in Climate Change Studies.
MIT Joint Program Report 122. (http://mit.edu/globalchange/www/MITJPSPGC Rptl22.pdf).

Dutkiewicz, S., M.J. Follows and J.G. Bragg, 2009: Modeling the coupling of ocean ecology and

biogeochemistry. Global Biogeochemical Cycles, 23, GB4017; MIT Joint Program Reprint 2009-24
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint09-24.pdf).

Felzer, B., J. Reilly, J. Melillo, D. Kicklighter, M. Sarofim, C. Wang, R. Prinn and Q. Zhuang, 2005: Future
effects of ozone on carbon sequestration and climate change policy using a global biogeochemical
model, Climatic Change, 73(3): 345-373; MIT Joint Program Reprint 2005-8
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint05-8.pdf).

Follows, M., T. Ito and S. Dutkiewicz, 2006: On the solution of the carbonate chemistry system in ocean
biogeochemistry models. Ocean Modeling, 12, 290-301.
(http://ocean.mit.edu/~mick/Papers/Follows-etal-QM2006.pdf)

Follows, M., S. Dutkiewicz, S. Grant and S. Chisholm, 2007: Emergent biogeography of microbial
communities in a model ocean. Science, 315(5820): 1843-1846
(http://globalchange.mit.edu/hold/restricted/Follows Science315 2007.pdf).

Franck, T., 2009a: Coastal Communities and Climate Change: A Dynamic Model of Risk Perception,

Storms, and Adaptation, Ph.D. Dissertation in Technology, Management, and Policy, Massachusetts
Institute of Technology (http://globalchange.mit.edu/files/document/Franck PhD 09.pdf)

Franck, T., 2009b: Coastal adaptation and economic tipping points. Management of Environmental
Quality, 20(4), 434-450; MIT Joint Program Reprint 2009-9
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint09-9.pdf).

Gao, X., C.A. Schlosser, A. Sokolov, and K. Walter, 2010: Quantifying future changes in high-latitude
methane emissions under regional climate change uncertainty, 2010 State of the Arctic Conference
-At the Forefront of Global Change, March 16-19 2010, Miami, Florida

Gurgel, A., J.M. Reilly, and S. Paltsev, 2007. Potential land use implications of a global biofuels industry.
Journal of Agricultural & Food Industrial Organization, 5(2): Article 9.
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint07-14.pdf)

Holian, G., A.P. Sokolov and R.G. Prinn, 2001: Uncertainty in Atmospheric C02 Predictions from a

Parametric Uncertainty Analysis of a Global Ocean Carbon Cycle Model. MIT Joint Program Report
80 (http://mit.edu/globalchange/www/MITJPSPGC Rpt80.pdf).

Hughes, G., P. Chinowsky and K. Strzepek, 2010: The Costs of Adaptation to Climate Change for Water
Infrastructure in OECD Countries. Global Environmental Change, in press.

Hurrell, J.W., J.J. Hack, D. Shea, J.M. Caron, and J. Rosinski, 2008: A New Sea Surface Temperature and
Sea Ice Boundary Dataset for the Community Atmosphere Model. J. Climate, 21, 5145-5153.

Kim, D., C. Wang, A.M.L. Ekman, M.C. Barth and P. Rasch, 2008: Distribution and direct radiative forcing
of carbonaceous and sulfate aerosols in an interactive size-resolving aerosol-climate model. J.
Geophys. Research, 113, D16309; MIT Joint Program Reprint 2008-11
(http://globalchange.mit.edu/files/document/MITJPSPGC Reprint08-ll.pdf).

Klima, K., 2008: Effects of Variable Wind Stress on Ocean Heat Content. Master of Science Thesis, Earth,
Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, September
(http://mit.edu/globalchange/www/docs/Klima MS 08.pdf).

-------
Lee, E., C.A. Schlosser, B.S. Felzer and R.G. Prinn, 2009: Incorporating plant migration constraints into
the NCAR CLM-DGVM model: Projections of future vegetation distribution in high latitudes. Eos
Trans. AGU, 90(52), Fall Meeting Supplement, Abstract B41C-0336.

Lee, E., C. A. Schlosser, X.Gao, A. Sokolov, and R. G. Prinn, 2010a: The CLM-DGVM-SEED: incorporating a
meteorological constraint into plant migration in the CLM-DGVM, NCAR LMWG/BGCWG meetings,
February 8, 2010, Boulder, CO.

Lee, E., C. A. Schlosser, X.Gao, and R. G. Prinn, 2010b: Incorporating a meteorological constraint to plant
migration in a dynamic vegetation model: Projections of future vegetation distribution in the Pan-
Arctic, State of the Arctic conference, March 16-19, 2010, Miami, FL.

Marshall, J., A. Adcroft, C. Hill, L. Perelman, and C. Heisey, 1997: A finite-volume, incompressible Navier
Stokes model for studies of the ocean on parallel computers.

J. Geophys. Res., 102 (C3), 5753-5766 (http://paoc.mit.edu/paoc/papers/finite.pdf).

Matus, K., T. Yang, S. Paltsev, J. Reilly and K.-M. Nam, 2008: Toward integrated assessment of

environmental change: Air pollution health effects in the USA. Climatic Change, 88(1): 59-92; MIT
Joint Program Reprint 2007-12

(http://globalchange.mit.edu/hold/restrictedReprints/MITJPSPGC Reprint07-12.pdf).

Mayer, M., C. Wang, M. Webster, and R. G. Prinn, 2000: Linking local air pollution to global chemistry
and climate, J. Geophysical Research, 105(D18): 22,869-22,896;
(http://globalchange.mit.edu/files/document/MITJPSPGC Reprint00-5.pdf).

Melillo, J., J. Reilly, D. Kicklighter, A. Gurgel, T. Cronin, S. Paltsev, B. Felzer, X. Wang, A. Sokolov and C. A.
Schlosser, 2009: Indirect Emissions from Biofuels: How Important?, Science 326: 1397-1399; MIT
Joint Program Reprint 2009-20 (http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint09-
20.pdf).

Monier, E., J. Scott, A. Sokolov and A. Schlosser, 2010; MIT IGSM: Toward a 3-Dimensional Integrated
Assessment Model. Poster presentation. 2010 DOE Climate Change Modeling Program Science Team
Meeting, Gaithersburg, Maryland, March 29-April 2.

Nam, K. M., N.E. Selin, J. M. Reilly, and S. Paltsev, 2010: Measuring welfare loss caused by air pollution in
Europe: A CGE Analysis. Energy Policy, in press

(http://globalchange.mit.edu/hold/pending/NamEtAI-EnergyPolicv2010.pdf).

Paltsev S., J. Reilly, H. Jacoby, R. Eckaus, J. McFarland, M. Sarofim, M. Asadoorian and M. Babiker, 2005:
The MIT Emissions Prediction and Policy Analysis (EPPA) Model: Version 4. MIT Joint Program Report
125 (http://globalchange.mit.edu/files/document/MITJPSPGC Rptl25.pdf).

Rasch, P.J., N.M. Mahowald, and B.E. Eaton, 1997: Representations of transport, convection, and the
hydrologic cycle in chemical transport models: Implications for the modeling of short-lived and
soluble species. J. Geophys. Res.-Atmospheres, 102 (D23), 28127-28138.

Rausch, S., G. Metcalf, J.M. Reilly and S. Paltsev, 2009: Distributional Impacts of a U.S. Greenhouse Gas
Policy: A General Equilibrium Analysis of Carbon Pricing. MIT Joint Program Report 182
(http://globalchange.mit.edu/files/document/MITJPSPGC Rptl82.pdf).

Rausch, S., G.E. Metcalf, J.M. Reilly and S. Paltsev, 2010: Distributional Implications of Alternative U.S.
Greenhouse Gas Control Measures. The B.E. Journal of Economic Analysis & Policy, in press; also MIT
Joint Program Report 185, June, 42 p.

(http://globalchange.mit.edu/files/document/MITJPSPGC Rptl85.pdf)

-------
Reilly, J., S. Paltsev, B. Felzer, X. Wang, D. Kicklighter, J. Melillo, R. Prinn, M. Sarofim, A. Sokolov and C.
Wang, 2007: Global economic effects of changes in crops, pasture and forests due to changing
climate, carbon dioxide, and ozone. Energy Policy, 35(11): 5370-5383; MIT Joint Program Reprint
2007-11 (http://globalchange.mit.edu/files/document/MITJPSPGC Reprint07-ll.pdf)

Schlosser, C.A., D. Kicklighter, and A. Sokolov, 2007: A Global Land System Framework for Integrated
Climate-Change Assessments, Report 147, May 2007, 60 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC Rptl47.pdf)

Schlosser, C. A., X. Gao, K. Walter, A. Sokolov, D. Kicklighter, C. Forest, Q. Zhuang, J. Melillo, and R. Prinn,
2010a: Quantifying climate feedbacks from abrupt changes in high-latitude trace-gas emissions.
Presentation to the DOE Integrated Climate Change Modeling Science Team Meeting, April 1, 2010,
Gaithersburg, MD.

Selin, N.E., S. Wu, K.-M. Nam, J.M. Reilly, S. Paltsev, R.G. Prinn and M.D. Webster, 2009a: Global Health
and Economic Impacts of Future Ozone Pollution. Environmental Research Letters, 4(4): 044014; MIT
Joint Program Reprint 2009-17

(http://globalchange.mit.edu/files/document/MITJPSPGC Reprint 09-17.pdf)

Sokolov, A.P., C.A. Schlosser, S. Dutkiewicz, S. Paltsev, D.W. Kicklighter, H.D. Jacoby, R.G. Prinn, C.E.
Forest, J. Reilly, C. Wang, B. Felzer, M.C. Sarofim, J. Scott, P.H. Stone, J.M. Melillo and J. Cohen,
2005: The MIT Integrated Global System Model (IGSM) Version 2: Model Description and Baseline
Evaluation. MIT Joint Program Report 124, July, 40 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC Rptl24.pdf).

Sokolov, A., P. Stone, C. Forest, R. Prinn, M. Sarofim, M. Webster, S. Paltsev, C.A. Schlosser, D. Kicklighter,
S. Dutkiewicz, J. Reilly, C. Wang, B. Felzer, J. Melillo, H. Jacoby, 2009a: Probabilistic forecast for 21st
century climate based on uncertainties in emissions (without policy) and climate parameters. Journal
of Climate, 22(19): 5175-5204; MIT Joint Program Reprint 2009-12
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint09-12.pdf).

Strzepek, K., and B. Boehlert, 2010: Competition for water for the food system. Philosophical
Transactions of the Royal Society, in press.

Strzepek, K., A. Schlosser, W. Farmer, S. Awadalla, J. Baker, M. Rosegrant and X. Gao, 2010a. Modeling
the Global Water Resource System in an Integrated Assessment Modeling Framework: IGSM-WRS,
MIT Joint Program Report No. 189, Cambridge, MA.

Strzepek, K., J. Baker, W. Farmer, C.A. Schlosser, 2010b: The Impact of Renewable Electricity Futures on
Water Demand in the United States. MIT Joint Program Report in preparation.

Sugiyama, M., R.J. Nicholls and A. Vafeidis, 2008: Estimating the Economic Cost of Sea-Level Rise. MIT
Global Change Joint Program, Report 156, April, 40 p.
(http://globalchange.mit.edu/files/document/MITJPSPGC Rptl56.pdf).

Wang, C., 2004: A modeling study on the climate impacts of black carbon aerosols. J. Geophysical
Research, 109(D3): D03106; MIT Joint Program Reprint 2004-2
(http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint04-2.pdf)

Wang, C., 2009: The sensitivity of tropical convective precipitation to the direct radiative forcings of
black carbon aerosols emitted from major regions. Annates Geophysicae, 27(10): 3705-311; MIT
Joint Program Reprint 2009-11 (http://globalchange.mit.edu/files/document/MITJPSPGC Reprint09-
ll.pdf).

Wang, C., R. G. Prinn, and A. Sokolov, 1998: A global interactive chemistry and climate model:

Formulation and testing. J. Geophysical Research, 103(D3): 3399-3418; MIT Joint Program Reprint
1998-5 (http://globalchange.mit.edu/files/document/MITJPSPGC Reprint98-5.pdf)

-------
Wang, C., G. Jeong and N. Mahowald, 2009a: Particulate absorption of solar radiation: anthropogenic
aerosols vs. dust. Atmospheric Chemistry and Physics, 9: 3935-3945; MIT Joint Program Reprint
2009-10 (http://globalchange.mit.edu/files/document/MITJPSPGC Reprint09-10.pdf).

Wang, C., D. Kim, A.M.L. Ekman, M.C. Barth and P. Rasch, 2009b: The impact of anthropogenic aerosols
on Indian summer monsoon. Geophysical Research Letters, 36, L21704; MIT Joint Program Reprint
2009-21 (http://globalchange.mit.edu/hold/restricted/MITJPSPGC Reprint09-21.pdf).

Webster, M., A. Sokolov, J. Re illy, C. Forest, S. Paltsev, A. Schlosser, C. Wang, D. Kicklighter, M. Sarofim,
J. Melillo, R. Prinn and H. Jacoby, 2009: Analysis of Climate Policy Targets under Uncertainty. MIT
Joint Program Report 180, September, 53 p.

(http://globalchange.mit.edu/files/document/MITJPSPGC Rptl80.pdf).

-------
Land Use In the MIT IGSM: The Role of
Biofuels and Forests in Mitigating Climate

Risks

John Reilly

Joint Program on the Science and Policy of Global Change,
Massachusetts Institute of Technology

5 November 2010, Purdue

Economic Models of Land Use and Biofuels

Melillo, et al., 2009, Indirect Emissions from Biofuels: How Important?, Science, 326:1397
99

Gurgel et. al., 2009, Food, Fuel, Forests and the Pricing of Ecosystem Services, ASSA
meeting paper, and to be published in the AJAE.

-------
MIT INTEGRATED GLOBAL SYSTEM MODEL

agriculture,

forestry,
bio-energy,
ecosystem
productivity

human Activity {eppa)

national and/or regional economic
development, emissions, land use

trace gas

fluxes
(C02,CH4.

N,0)
and policy

C02, CH„, CO,
N20, NOx,
SOx, NH3,
CFCs, HFCs,
PFCs, SF6,
VOCs, BC. etc

coupled ocean, atmosphere, and land EARTH SYS7 ci».

solar
forcing

ATMOSPHERE
2-Dimensional Chemical
& Dynamical Processes

URBAN
Air Pollution Process

sunlight water cycles, energy & momentum transfe
air & sea temperatures, COj, CH4, N2O, nutrients,
pollutants, soil properties, surface albedo,
sea-ice coverage, ocean CO2 uptake,
land CO2 uptake, vegetation change...

Ocean

2- or 3-Dimensional
Dynamics, Biological,
Chemical & Ice Processes
(MITgcm)

Land

Water & Energy Budgets

{CLM)

Biogeochemical Processes
ITEM &NEM)

EXAMPLES OF
MODEL OUTPUTS |

GDP growth,
energy use,
policy costs,
agriculture and
health impacts...

ibal mean
anofetitudinal
temp»ture and
precipintion,
sea levelT
sea-ice co®r,
greenhouslgas
concentratl>ns,
air pollutiJl levels...

soil andwgetative
carbon, Mt primary
produjirvity,
tramps emissions
fj^mecosystems,
sermafrost area...

-------
MIT EPPA, 16 Region, multi-sector
CGE model

DYNAMIC
TERRESTRIAL ECOSYSTEMS
MODEL (TEM)

Global Land System Interactions

GHG and Other Pollutants
from energy and agriculture/land use

Coupled Ocean,
Atmosphere

Biogeophysical Land
Processes

Land use shares for crops,
livestock, bioenergy, forestry

GTAP land data/
Spatial disaggregation
algorithm

Temperature, Precipitation,
Solar Radiation

C02, Tropospheric Ozone,
Nitrogen deposition

Spatial data (.5° x .5°) for
land use

Crop, pasture,
bioenergy, forest
productivity

CH4, N20, Net C02
from land use

-------
Table 1. Regions, Sectors, audi Primary Factors, in the EPPA Zvlodel

C'ountn- or Region

Sectors

F actors

developed
United States (USA)

Canada (CATS)

Japan (JP1SJ)

European Union— (EUR)
Australia. N.Zealand (ANZ)
Former Soviet Union (FSU)
Eastern Europe (EET)
Developing
India (END)

China (CHIC)

Indonesia (IDZ)

Higher Inc. East Asia (AST)
Mexico (IvIEXj
Centr. & S. America (LAM)
Middle East (MES)

Africa (AFR)

Rest of World (ROW)

I\~on-Energy
Services (SERV)

Energy-Intensive (E.INT)

Other Industries (OTHR)

Commercial Transp. (TRAN)

Household Transp. (HTRN)

Other HH Consumption - Recreation
Hunting and Fishing (REHF)

W ildlife Viemug in Reserves (REWA R)
Otlier Wildlife Viewing (RIW\ \)

F nets
Coal (COAL)

Crude Oil (OIL)

Refined Oil (ROIL)

Natural Gas (GAS)

Oil from Shale (SYNO)

Synthetic Gas (SY1STG)

Liquids from Biomass (B-OIL)

EIectricity Gen eration
Fossil (ELEC)

Hydro (HYDR)

Nuclear (NUC'L)

Solar and Wind (SOLW)

Biomass (BIOM)

Coal with CCS (IGCAP)

Adv. gas without CCS (NGCC)

Gas with CCS (NGCAP)

"Agriculture
Crops (CROP)

Livestock (LIVE)

Forest products (FORS)

Food Processing (FOOD)

Capital
Labor

Energy Resources
Crude Oil
Natural Gas
Coal
Oil Shale
Nuclear
Hydro
Wind Solar
Land
Cropland
Pascureland
Managed Forest
Nan-Reserved
Natural Forest
Reserved Natural
Forest

Natural Grassland
Other

-------
Expanded SAM—Household
"production" sector, leisure

Domestic

F'roduction

INTERMEDIATE
USE
by Production
Sectors

Biofuels
Bioelec
Crops
Livestock
forestry

Imports

Leisure

Value
added:

-labor

-capital

- natun
resources

INPUT

Environi
al heal
provisi

Household
Services

hh
Prod.

Labor-
Leisure
Choice

aent

recreation

Labor

Land fo
product
environmental

Leisure

Vll

FINAL USE

Private

Gov't

Invest.

Export

OUT-

PUT

land
fu

managed
, recreation
ure value

Added components are in bold italic.

-------
atmospheric carbon dioxide

gross
primary
production

h— ~~~

L - J

...

Carbon

litterfall
production

Vegetation

Labile
Nitrogen

uptake

exchange

Structural
Nitrogen

^uptake

litterfall
production

Carbon

Soil

and
Detritus

Nitrogen

net exchange/
mineralization

Nitrogen
input —

Inorganic
Nitrogen

Nitrogen
^ lost

Terrestrial Ecosystem Model (TEM)

The Ecosystems Center, Marine Biological Laboratory (Woods Hole, Massachusetts)

Figure 3.4: Description of the Terrestrial Ecosystem Model (TEM)

Source: The Ecosystems Center, the U.S. Marine Biological Laboratory (MBL)

Monthly,

0.5° x 0.5°,

Dynamic soils
and vegetation
with multiple
carbon pools,
and multiple
harvest carbon
pools i.e. forest

litter, waste,
paper; lumber

-------
Food, crop, livestock, and forestry price impacts combine impacts of climate
change, ozone, competition for land of biofiiels, and mitigation cost effects on

energy/N20/CH4

o
o

2.2
1.8
1.4
l.O

0.6

—i i 1 1 1 1 1 1 1 i—

2000 2020 2040 2060 2080 2100
Yea r

2.2

1.8

o. 1.4

^ 1.0

0.6 -| 1 1 1 1 1 1 1 1 1 r

2000 2020 2040 2060 2080 2100

Year

— No
Policy

••• Energy-
Only

— Energy
+ Land

4.5

X
QJ

1= 3'5

2.5

-y-

£ 1-5
0.5

4.5

—i 1 1 1 1 1 1 1 1 1—

2000 2020 2040 2060 2080 2100
Year

x
OJ

^ 3.5

-------
FRAMEWORK FOR AIR POLLUTION IMPACTS ANALYSIS

Emissions of
JQx, VOCs, SQ2, BC,

OC, Hg

GHGs

nun in ii

Pollution controls
Technology changes

Atmospheric modeling

Pollutant transport
Atmospheric chemistry
Climate interactions

Concentrations of

Integrated models & tools

ozone, particulates
Hg deposition

Economic activities and
Policy choices

Economic modeling

Population health impacts

Hospital visits

Mortalities
(acute/chronic)
IQ deficits

-------
GLOBAL COSTS OF OZONE POLLUTION IN 2050

a) AMortalities: Climate (Total:-5000) b) AMortalities: Emissions (Total: 817,000)

——

-200 -100 0 100 200 people -200 -100 0 100 200 people

c) AMortalities: Climate+Emissions (Total: 812,000) d) AMortalities: 03 >10 ppb (Total: 2.6x106)

—-—

-200 -100 0 100 200 people

-1000 -500 0 500 1000 people

03 from A1B
scenario [Wu et al.,
2008] to 2050

Calculate change in
welfare due to
health impacts of
ozone changes,
separately for
emissions and
climate drivers

2050 welfare loss from 03 health impacts, climate only scenario:
€790 million (year 2000 €)

2050 welfare loss from climate+emission changes: bil
2050 welfare loss from all 03 above background: €580 bil

[Set in et a I., in prep]

-------
Uncertainty: Due to uncertainty in dose response
relationships and economic modeling of impacts.

o
o
o

en
cn

-9?
Id
c3

J
o

LD
O

C\J

300

U 200
£

5
<3

100

o 0.1500
cz

CD
13

CD
>

0.1000

® 0.0500

0.0000

2010

2020 2030 2040 2050

100 150 200 250

-Q

Q.
Q.

2000

£=
o

1500

1000

"cii

t 0.1000

IX 0.0500

0.0000

0 500 1000 1500
Awelfare ($billion)

2000

-------
Uncertainty Analysis: Methodology

Estimate probability distributions for input parameters controlling the
emissions and climate projections in IGSM sub-models:

(1) Emissions Uncertainties:

Elasticities of Substitution
GDP Growth (based on Labor Productivity Growth)
Autonomous Energy Efficiency Improvement (AEEI)

Fossil Fuel Resource Availability, Population Growth
Urban Pollutant Trends, Future Energy Technologies
Non-COoGreenhouse Gas Trends, Capital Vintaging
(2) Climate System Response Uncertainties (constrained by observations):

Climate Sensitivity
Rate of Heat uptake by Deep Ocean
Radiative Forcing Strength of Aerosols
(3) Greenhouse Gas Cycle Uncertainties:

CO, Fertilization Effect on Ecosystem Sink
Rate of Carbon Uptake by Deep-Ocean
Trends in Rainfall Frequency on natural CH4 & NzO emissions

Five Cases indicated by GHG levels (ppm-equivalent CO^. ppm CO, and
change in Radiative Forcing relative to ~1990 (W/m*) in ~2100:
No Policy (1400 ppm C02-eq; 870 ppm C02; 9.7 W/m2)

Level 4 (900 ppm C02.eq; 710 ppm C02; 7.1 W/m2)

Level 3 (790 ppm C02-eq; 640 ppm C02; 6.3 W/m2)

Level 2 (660 ppm C02-eq; 560 ppm C02; 5.3 W/m2)

Level 1 (550 ppm C02-eq; 480 ppm C02; 4.2 W/m2)

Generate 400 member ensembles (Monte Carlo with Latin Hypercube
Sampling) for each case

-------
95% PROBABILITY BOUNDS
OF GLOBAL AVERAGE GHG

MOLE FRACTIONS AND
RADIATIVE FORCING from
1981-2000 to 2090-2100,
WITHOUT (1400 ppm-eq
C02) & WITH A 550, 660, 790
or 900 ppm-eq C02 GHG
STABILIZATION POLICY?

-------
Cumulative PROBABILITY OF GLOBAL AVERAGE SURFACE AIR WARMING
from 1981-2000 to 2091-2100, WITHOUT (1400 ppm-eq C02) & WITH A 550,
790 or 900 ppm-equivalent C02 GHG STABILIZATION POLICY

(Ref: Sokolov et at, Journal of Climate, 2009)

("Values relative to

No Policy at 1400

100%(*100%)

85%

25%

Stabilize at 900 (L4)

100%(*100%)

25%

0.25%

Stabilize at 790 (L3)

97%(100%)

< 0.25%

|stabilize at 660 (L2)

80%(*97%)

0.25%

< 0.25% |

Stabilize at 550 (L1)

25%(*80%)

< 0.25%

< 0.25^r>£^

-------
Comparison to Range in CCSP 2.1A

0.018

400

600

800

No Policy
Level 4
Level 3
Level 2
Level 1

1000

1200

C02 Concentrations (ppmv) in 2100

-------
Comparison to IPCC

O o

-------
Change in the probability of exceeding illustrative targets for global mean
surface temperature change, as measured by the change between the
average for 1981-2000 and the average for 2091-2100.

100

JD O)

¦ji; CD
CD I—

"S £

— CD

.1= Q_

CD
CD
O
X

LU o

*h— CD

O t

¦S CD

2 ^

Q- 75

REF

Level 4

Level 3

Level 2

Level 1

-------
The Global Adaptation Atlas

Establishing Priorities for Research, Policy and Action on Adaptation

Ray Kopp - Senior Fellow and Director, Climate Policy Program
Nisha Krishnan - Former Atlas Team Member
Improving the Assessment and Valuation of Climate Change Impacts for Policy and

Regulatory Analysis- November 18-19, 2010

RESOURCES

FOR THE FUTURE

-------
Afghanis tart

Tripoli

Mediterranean Set

lorocco

Algeria Algeria

Saudi Arabia

Yemen

Somalia

Zambia

What is the Adaptation Atlas?

Web-based application enables
user driven, dynamically generated
maps of climate impacts and
adaptation activities:

- Database of impacts from peer
reviewed climate studies

- Repository of adaptation projects

- Data available for download and
uploads of new data supported

- User can select different locations,
timeframes, scenarios and overlay
resulting data across sectors

Beta version available at www.adaptationatlas.org

.Sudan

1*?

™r 1

"E~ i

eEJ

jjj

Autumn/winter irrigation as an adaptive mechanism for efficient use of
water resources in Southern Kazakhstan Details

Activity Type: Project
Status: Ongoing
Funding Source: UNDP& GEF
Location: Sadu Shakirov

Duration (years): 2

Start Date: 1/3/2009
Total Funding: $45,286

Description: The project will implement new systems of irrigation - during the
autumn and winter - In pilot sites, to demonstrate the effectiveness of this
technique, and promote its replication by neighbouring ranchers Essentially,
irrigation In Autumn and Winter - the periods of the year with average
temperatures below zero - replicates the same effect of snowfall, which is
declining Water delivered to pastures during these seasons melts and
promotes grass growth during the spring thaw In addition to benefits to the
local community, the project will publish a short booklet aimed at facilitating
replication in areas feeing similar challenges.

Priest Wete'te

Contact Information: Mr Stanislav Kim
67 Tole bi Str, Almaty. 480091
Phone +7-3272 582646/582643
Fax +7-3272 582645
Email stanislav kim@unclp.org

J Save Project to My Atlas |

Caption: Broken down irrigation infrastructure
m Sadu Shakirov Village, Kazakhstan This
CBA project will rehabilitate the local system of
canals, in order to enable cold-season irrigation
as a means of sustainably managing water and
pastures in the face of climate change

RESOURCES

FOR THE FUTURE

-------
Methodological Questions

1. Data Solicitation and Collection

Literature searches
Individual author solicitation

2. Study Harmonization/ Comparability

Each study is its own story
'meta' filters

3. Atlas Outputs

RESOURCES

FOR THE FUTURE

-------
The Uncertainty Issue

- Hales et al. 2002, "Potential Effect
of Population and Climate Changes
on Global Distribution of Dengue
Fever: an Empirical Model"

- 4 sets of sensitivity analyses using
the ECHAM4, CGCMA1, CGCMA2
and CCSR/NIES models.

- Unique layers that fit into the
decision framework of the Atlas.

- Differences between dealing with
sensitivity analyses and uncertainty
analyses.

RESOURCES

FOR THE FUTURE

tfV

o
o
o
o

tj»Mrr*a9iih»nam*. bUWi

•.i-jfitr -rtTi torfto P* ***<', a

Maun j** PtcMiiksn. icahtf f «*«f.

.t «t JCC2

-------
Global Adaptation Atlas

local

various
result

online

Cn tailored

0a s practitioner?, .

.32 ... healthProjecl

collection spatial proCBSS fHltlQatlOn a(^fl

'""'P «ai3: rl 'I

eras', ei- tse Qala

I (Dproblon | (Q support ilTipSCt 1

outreach

tW«W*

priorities

sectors

existing

new

«*Global r™

Dest

. .. .. ..chanqelalreadyfl \

building ^ v-•

;oordination

natural

mapping

patent pending

For more on the Atlas, visit www.adaptationatlas.org or email us at info@adaptationatlas.org

RESOURCES

FOR THE FUTURE

created using wordle.net

-------
A Generalized Modeling Framework
for Climate Change Damage

Assessment

Alex Marten, Steve Newbold,

Charles Griffiths, Elizabeth Kopits,

Chris Moore, Ann Wolverton

National Center for Environmental Economics, NCEE^

U.S. EPA NATIONAL CENTER FOR

ENVIRONMENTAL ECONOMICS

* Please note that the views expressed are those of the authors and do not necessarily represent those of the U.S. EPA.
No Agency endorsement should be inferred.

-------
Lessons Learned

o Need a more transparent representation of the pathways
through which climate change may affect economic
productivity and human well-being

o Need a transparent method of incorporating new research on
climate damages into modeling exercises

o Desire to more transparently map assumptions of economic
behavior (e.g., adaptation, technology diffusion) into
economic damage estimates

o Need for reduced form lAMs that allow for a relatively timely
assessment in a probabilistic fashion

-------
Reasons for a New Framework

o Help facilitate the process of incorporating new climate science and
economic damage research

o To clearly distinguish among damages to market sectors, physical
and natural capital stocks, and human health while also accounting
for defensive expenditures0

o Standardization so that the effects of specific assumptions/pieces
may be better understood

o Increased transparency through complete, accurate, and up-to-date
documentation and open source code

o To make climate-economic integrated modeling more accessible to
government and researchers

H Defensive expenditures is used here to refer to expenditures borne in order to offset the effects of worsening environmental quality.

-------
Key Characteristics of Framework

o General structure that nests commonly used integrated assessment
models, including the three used by the interagency workgroup

o Flexible framework so that new findings and assumptions may be easily
incorporated

o Transparent, in that the code, framework, calibrations, and assumptions
will be well documented and freely accessible to researchers and other
interested parties

o Probabilistic, to allow for formal uncertainty analysis in a Monte Carlo
framework

o Modular design allows for linkage with multiple climate models and future
additions of new impact categories

o For example: Would allow for standardization in climate and economic
assumptions across various calibrations of the damage functions (and vice
versa)

-------
Overview of Structure

o Climate model coupled to a regionalized exogenous growth
model of the economy

o Exogenous technical progress and population growth
(potential for climate-population feedbacks)

o Currently uses exogenous emissions scenarios (retains the
option for endogenous emissions in the future)

o Currently uses MAGICC as the climate model (may use
others; such as those included in DICE, FUND, and PAGE)

o Ability for natural capital to be represented

o Setup to run probabilistically

-------
Representation of Damages

o Distinguishes between different types of climate change damages
to provide for transparency and ensure that they are affecting the
correct end points in the model

o Damages to multiple market sectors

o Damages directly to physical capital

o Defensive expenditures offsetting investment in physical capital
o Defensive expenditures offsetting household consumption
o Consumption equivalent health damages
o Consumption equivalent recreation and nonuse damages
o Use of general functional forms so that the model remains flexible

-------
Current Status

o Prototyping of framework and initial testing
o Development of initial code base

o Including interface for public version of MAGICC, along with
versions of the DICE, FUND, and PAGE climate models

o Ongoing development of clear and accurate documentation
for the framework

o Testing generality by using specific settings to closely
approximate versions of DICE, PAGE, and FUND similar to
those used by the interagency workgroup

-------
Approximation of Other Models

o
o

QO
ro

ro
O

DICE

PAGE

FUND

Approximation

y jb

A ^

>—0--^9==^ O _jO—^

-o -g g—q—e—©-—©—e-__o.__o—o—o—o—

2010 2020 2030 2040 2050 2060 2070 2080 2090 2100

Year

o The central values of parameters are used in this exercise

o Approximation of FUND does not yet include all the damage sectors that are in the full model

-------
Next Steps

o Continual refinement of the model in response to prototyping

o Full approximation of FUND

o Incorporation of feedback from workshops

o Starting from the studies currently used in existing lAMs move
forward with incorporating new studies on climate change
damages

o External peer review

o Eventual public release

-------
Modeling the Impacts of Climate Change:
Elements of a Research Agenda

Ian Sue Wing Karen Fisher-Vanden

Associate Professor Associate Professor

Dept. of Geography & Environment Dept. of Agricultural Economics

Boston University Pennsylvania State University

Elisa Lanzi
OECD Environment Directorate

Abstract

-------
1 Introduction: What is an IAM?

As illustrated in Figure 1, an integrated assessment model (IAM) of climate change is
typically constructed from three interlinked sub-models, an economic model (1), a climate
model (2) and an impacts model (3). It is logical to begin with the economic sub-model,
which is responsible for generating time-paths of global emissions of greenhouse gases
(GHGs—principally carbon dioxide, CO2) (a). These serve as inputs to the climate sub-
model, which uses them to project changes in the magnitude of meteorological variables
such as temperature, precipitation or sea level rise (b). Finally, the changes in climate
parameters are translated into projections of global- or regional-scale economic losses by
an impacts sub-model, whose primary role is to capture the feedback effect of dangerous
near-term anthropogenic interference with the climate on economic activity over the long-
term future (c).

Innovation is a key modulator of the clockwise circulation of the feedback loop in
the figure. Improvements in the productivity of labor induce more rapid growth and in-
crease the demand for fossil energy resources, which has a first-order amplifying effect
on emissions (A). Energy- or emissions-saving technological progress tends to depress
the emission intensity of the economy, slowing the rate of increase in fossil fuel use; con-
versely, productivity improvements in energy resource extraction lower the price of fossil
fuels and induce substitution toward them, increasing emissions (B). Lastly, we can imag-
ine that there may be innovations that boost the effectiveness of defensive expenditures
undertaken in response to the threat of climate damages, or investments in creating new
knowledge that enables humankind to mitigate some climate damages (C). This last cat-
egory is the most speculative, as impacts will manifest themselves several decades in the
future, when the state of technology is likely to be quite different from today

-------
Figure V. Integrated Assessment of Climate Change and the Effects of Innovation

-------
2 Land of Cockaigne: An IAM with Regional, Sectoral and

Climate Impact Detail

Imagine that there were relatively few constraints to either our computational resources
or our ability to foresee the impacts of climate change. In such a world, what would an
IAM look like? We could then specify a RICE- or AD-WITCH-type IAM that resolved (a)
the detailed sectoral structure of production in various regions, (b) the effects of climate
impacts on the productivity of those sectors, (c) the manner in which different impact
endpoints combined to generate the resultant productivity effects, and (d) the response
of the full range of impacts to changes in climatic variables at regional scale.

Let us write down such a model, and exploit its structure to assess the implications
for the social cost of carbon. Define the following nomenclature:

Set indexes:

f = {0, ...,,9'} Time periods

£ = {1,..., World regions

j = {1,..., jY} Industry sectors

m = {1,..., ,J{} Meteorological characteristics
f = {1,Climate impact endpoints
Control variables:

t Sectoral energy input

q^e t Sectoral capital input

Q(n Aggregate consumption

Qj t Aggregate jelly capital investment

Region-, sector- and impact-specific averting expenditure
Region-, sector- and impact-specific adaptation investment

Economic state variables:

Welfare (model objective)

Net sectoral product

Aggregate net regional product

aggregate regional energy use

Global marginal energy resource extraction cost

-------
Qft Stock of aggregate jelly capital

x¦ (t Stock of region-, sector- and impact-specific adaptation capital
Environmental state variables:

Gt Global stock of atmospheric GHGs

Mft Region-specific meteorological variables

r '

z¦ (t Region-, sector-, and impact-specific endpoint indexes

Ajr£rt Region- and sector-specific damage induced productivity losses
Functional relationships:

E Global intertemporal welfare

U/ Regional intratemporal utility

O/ Regional aggregate production functions

ipjri Sectoral production functions

£ Global atmospheric GHG accumulation

Y'" Regional climate response functions

I¦¦ ( Regional and sectoral climate impacts functions
ARegional and sectoral damage functions

1. Economic Sub-Model

Objective:

max ^ W = Y, ft
t=o

r\C „E „K

^e,tnj/,tnj,e,t

iil

Qi,t

Use

Q%,t

(la)

(lb)

Aggregate net regional product:

Qi,t =

Sectoral net regional product = Climate loss factor x Sectoral gross regional product,
produced from energy and capital:

r,e,t

aE aK
<1j,£,t' Hj/,t

Intraregional and intratemporal market clearance for energy:

E ifm

(lc)

(Id)

-------
Intraregional and intratemporal market clearance for jelly capital:

e yfij=Qf,t

7=1

Aggregate regional absorption constraint:

Qlt = Qlt ~ Qit ~ PtQlt -EE Ki,t + vUt

/=17=1

Global energy trade and marginal resource extraction cost:

E E Ql

£=ls=0

Regional jelly capital accumulation:

Pf = ©

Qh+1 = Qej + (1 - «K)Ql,

Accumulation of impact-, sector- and region-specific adaptation capital:

Ut+i = tfl.t + (! -

j,£,t

(le)

(If)

(lg)

(lh)

(li)

2. Climate Sub-Model

Global atmospheric GHG accumulation:

Gf+, =s E
. £

Regional meteorological effects of global atmospheric GHG concentration:

Mlt = Yf [Gt\

(2a)

(2b)

3. Impacts Sub-Model

Physical climate impacts by type, sector and region:

Alt = (L fMlo Mg,

Climate damages:

^j,£,t ^j,£

1 J5" . 1 J5" . 1 J5"

^j,£,t' • • •' ^],£,t' ®j,£,t' • • •' ®j,£,t' %j,£,t' • • • > xj,£,t

(3a)

(3b)

-------
From the point of view of period f*, the condition for optimal extraction of carbon-
energy is:

<)/' j W _y ( d(pe* dipj/* pE

dQp-k i-k / 9Q/* f* \=\ j £* f* j £* f* ^Qf* f* /

' ' ' I V ]' ' ]' ' ' 7, II. Current marginal

I. Current marginal benefit extraction cost

lg 3© £ \ / / Oii OU£*

(fe fci V^^Qo^fv* 7/ V3U<*aQ£,e.

III. Resource stock effect of contemporaneous energy use

X, f f* / / 3E dill*

IV. Present value of future marginal climate damage (N.B. dq /d A < 0 in general)

= 0 (4)

The "social cost of carbon" in this expression is given by the combination of terms (II) +
(III) - (IV). Our interest is in (IV), the marginal external cost of carbon-energy consump-
tion, which, because it emanates from a globally well-mixed pollutant, is independent of
the location in which the energy is consumed.

It is now clear to see how fundamental gaps in our understanding the render the "land
of cockaigne" unattainable. The difficulty in computing the social cost of carbon stems
from the terms in curly braces. Carbon-cycle modeling is sufficiently advanced to enable
us to predict with a fair degree of confidence the effect of the marginal ton of carbon on the
time-path of future atmospheric GHGs (d£/BQE). Likewise, the IPCC AR4 notes global
climate models' substantially improved ability to capture the future trajectory of conse-
quent changes in temperature, precipitation, ice/snow cover and sea levels at regional
scales (BYf/dG). But the weak links in the causal chain between climate change and eco-
nomic damages continue to be the cardinality and magnitude of the vectors of physical
impact endpoints as a function of climatic variables in each region out into the future

-------
f

(3£- (/r)M'"), and—to a lesser extent—the manner in which these endpoints translate into
shocks to the productivity of economic sectors (3A

3 A Critical Review of the State of Modeling Practice

To put the key issues in sharp relief, it is useful to consider how implementing the dis-
aggregated IAM might improve upon the current state of integrated assessment practice.
RICE-type IAMs represent the productivity losses incurred by climate change impacts
through variants of Nordhaus' aggregate damage function, which specifies the reduction
in gross regional product as a function of global mean temperature. This approach ef-
fectively collapses Mf to a scalar quantity in each time period. Moreover, as reviewed
by NRC (2010), it then benchmarks the magnitude of various impacts and the associated
economic losses for a reference level of global mean temperature change, before making
assumptions about how these costs are likely to scale with income, and finally expressing
damage as a temperature-dependent fraction of regions' gross output. Therefore, the de-
tails of climatic variables' influence on impact endpoints in (3a), and of the latter's effects
on economic sectors in (3b), only affect the calibration of the damage function. From that
point on they are entirely subsumed within the function's elasticity with respect to global
temperature change, and, in RICE-2010, sea level rise. The damage function therefore
collapses (3a) into (3b), dealing only with changes in aggregate global climatic variables,
skipping over impacts as state variables and implicitly aggregating over sectors to express
damages purely on an aggregate regional basis.

A similar situation obtains with adaptation. A case in point is the AD-WITCH model,
a variant of Nordhaus' RICE simulation which modifies the damage function by introduc-
ing stock and flow adaptation expenditures which attenuate aggregate regional produc-
tivity losses due to climate change. Formally, using Qy to denote gross regional product,

-------
net regional product is given by

ny = 1 + ADAPTfj ~y
1 + ADAPTfj + CCDft C,t

where CCD is the regional climate damage function and ADAPT is an index of adap-
tation's effectiveness. The variable ADAPT is the output of a nested constant elasticity
of substitution (CES) production function which combines inputs of contemporaneous
averting expenditures with adaptation capital and adaptation knowledge according to
Figure 2. The key consequence is that adaptation is able to directly influence the dynamic
path of the economy, instead of being implicit in the curvature of the damage function,
as with the RICE model. However, eq. (5)'s assumption that the effects of ADAPT and
CCD are multiplicative seems very strong in light of the fact that the damage function al-
ready explicitly incorporates the influence of adaptation through the studies on which it
is benchmarked—but only at the calibration point, not over the full range of its curvature.
A prime example is Nordhaus and Boyer's (2000) use of Yohe and Schlesinger's (1998) re-
sults on the impact of sea level rise, which optimally balance the costs of abandonment
and coastal defenses. The implication is that because defensive expenditures are likely to
be closely associated with the magnitudes of climate impacts of various kinds within indi-
vidual sectors, one should not think of aggregate adaptation expenditure as independent
of future changes in the sectoral composition of output.

By dispensing with the aggregate damage function, our land of cockaigne IAM explic-
itly captures the dynamic evolution of impact endpoints' response to changes in climatic
variables, the magnitude and intersectoral distribution of the follow-on productivity ef-
fects, and the optimal intersectoral adjustments these induce, all at regional scales. An
adaptation response may therefore be modeled more precisely as averting expenditure
that mitigates the sectoral and regional productivity loss associated with a particular cat-
egory of climate impact. In other words, stock and flow adaptation reduces the impact

-------
Figure 2: The AD-WITCH Adaptation Production Function (Bosello, Carraro and De
Cian, 2010)

elasticity of sectoral productivity shocks. Of course, the problem that besets this approach
is that, except for a very few combinations of impacts, sectors and regions, the relevant
elasticities are unknown.

But the good news is that this is one area in which research is proceeding apace. There
are a growing number of CGE modeling studies of climate impacts (e.g., ICES) which
elucidate the magnitude of both sectoral and regional damages and producers' and con-
sumers' adjustment responses. The focus of such studies is typically a single impact cat-
egory (say, /*), whose initial economic effects are computed using natural science or en-
gineering modeling or statistical analyses. The results are often expressed as a vector of
shocks to exposed sectors and regions, which are then imposed as exogenous productiv-
ity declines on the CGE models' cost functions. In the context of the IAM in section 2,
this procedure is equivalent to first specifying an exogenous ex-ante effect of a particular

f-k

impact dXj^/dz-fr before using the CGE model to compute the ex-post web of intersec-
toral adjustments and the consequences for sectoral output, and regions' aggregate net
product and welfare:

| Adaptation |

-------
This line of inquiry has the potential to yield two critical insights. The first is quantifi-
cation of the elasticity of the economy's response to variations in the magnitude and inter-
regional/ intersectoral distribution of particular types of impact, which has been the type
of investigation pursued thus far. But second—and arguably more important—is com-
parative analysis of economic responses across different impact categories for the purpose
of establishing their relative overall economic effect, conditional on our limited knowl-
edge of their relative likelihood of occurrence, and intensity. The results could at the very
least guide the allocation of effort in investigating the thorny question of how different

impacts are likely to respond to climatic forcings at the regional scale, d^g/dMf.

-------
Sectoral and Regional
Disaggregation and Interactions

Ian Sue Wing
Boston University

Improving the Assessment and
Valuation of Climate Change Impacts fo
Policy and Regulatory Analysis

Nov. 18-19, 2010

-------
What Is an Integrated Assessment Model?

-------
Desiderata in Model Development:

If neither empirical estimates nor
computational resources were an issue,
what kind of 1AM would we construct?

-------
A Canonical Intertemporal 1AM

Welfare: max lr Z,={o,...,t) ^ryftt-'UrlCm,C2ir;c,...]

Sectoral Output: Qj rt = A^-J^K,,.,, Ej>rt]
Absorption: Q, rt = Cut + Ij ,t + 27IX.1E, rt
Energy Extraction: \+1 = IjIrEj rs+ Xt
Capital Accum.: Kj-r>t+1 = + f1" 6> Kj,r,t

Carbon Cycle: Gptil = #p[Z,ZrEj,r,t(Gpt]

Regional Climate: M^rt = 3^r[Glt,G2t,...]

Regional Impacts: Z, j-rt = i3iJ_r[M1_rt,M2
Regional Damage: Aw = AJZUWZUW...]

-------
A Canonical Intertemporal 1AM

1. Economy

Welfare: max Ir VP1' U[Cutt + ljrt + HlXJ-Ej ,t
Energy Extraction: Xt+1 = IjIrE: + Xt
Capital Accum.: Kuttl = i|)l,r-I|.li-,r,t + (1 - 6) Kj>r_t

-------
A Canonical Intertemporal 1AM

Carbon Cycle: Gpt+1 =

Regional Climate: Mu rt = 3f^r[Glt,G2

Climate

-------
A Canonical Intertemporal 1AM

Regional Impacts: ZiJ-rt = i3iJ r[Mut,M2_rt,...]
Regional Damage: Aj,,, = AktlZwZ2Uv...]

Impacts

-------
A Canonical Intertemporal 1AM

Welfare: max lr Z,={o,...,t) ^ryftt-'UrlCm,C2ir;c,...]

Sectoral Output: Qj rt = A^-J^Kut,E,,,]
Absorption: Q, rt = Cut + lj rt + 27IX.1E, rt
Energy Extraction: \+1 = IjIrEj rs+ Xt
Capital Accum.: Kj-r>t+1 = tpj,r-Zj'lj-r,t +11" 6> Kj,r,t
Carbon Cycle: Gptil = #p[Z,ZrEj „,Gpt]

Regional Climate: M^rt = 3^r[Glt,G2t,...]

Regional Impacts: Zw = i3iJ_r[M1_rt,M2
Regional Damage: Aw = AJZUWZUW...]

-------
Maximand: Global Intertemporal
Welfare Over a Policy Horizon

Welfare:

max Ir W.,T1 V 3'-U[Ci,r,VC2,rit,...]

Regional .

welfare m

:+l " *-\M

weights 9

Discount

factor

L. A

Regional instantaneous
utility denominated over
consumption of j individual
commodities in r regions

-------
Production is Where We Model
Climate Damages Exerting Their Effects

Sectoral Output: Qut = EUJ

Region-by-sector production
function denominated over
inputs of capital and carbon-
energy

Region-by-sector

instantaneous
economic output

Productivity shock

associated with
contemporaneous
region- and sector-
specific climate damage.
This is the key unknown.

-------
Disposition of Product Determines the
Capacity Constraint of the Economy

-------
Cumulative Carbon-Energy Extraction
Drives Increase in Global Marginal Cost

Cumulative
extraction of
carbon-energy

Energy Extraction: Xt+1 = IjIrEj rt + Xt

Current Past history of

energy use extraction

-------
(Endogenous) Accumulation of Capital
is the Key Engine of Economic Growth

New region-
and sector-
specific capital

Depreciation
factor

Capital Accum.: K: nt+1 = + (1_S) Kj.r.t

Sectoral investment
(sectors enjoy fixed
shares of aggregate
investment)

Extant region-

and sector-
specific capital

-------
Carbon Cycle Model ( Translates GHG
Emissions into Reservoir Concentrations

New GHG concentrations

by reservoir p
(e.g., atmosphere, mixed-
layer ocean, deep ocean)
at global scale

Capital Accurf^W I
Carbon Cycle:

,t+:

Global emissions

from use of
carbon-energy

K:+ (1 — 6) Kj#l.t

Gp,t+1 " -^p[IjIrEj,r,t'Gp,t]

Regional Climate: M

r\ A

H,r,t

Gl,t'G2J

ts I Z;

— S

IV /I

rial Dama

j,r,t^

Extant GHG
concentrations
by reservoir

-------
Climate Model (M) Translates GHG
Concentrations into Meteorology

Meteorological variables

(e.g., temperature,
precipitation, sea levels)
at regional scales

Carbon Cycle:

Extant GHG
concentrations
by reservoir

p,t+l ~

Regional Climate: rt = r[Glt,G2V...]

-------
Impacts Model (12) Translates Regional
Meteorology into Physical Endpoints

Contemporaneous values

of/' physical impact
endpoints by sector, region
and time period

Regional values of
meteorological
variables

Regional Impacts: ZliUt = i3,Ar[Mut,M2/M>...]

-------
Damage Model (A) Translates Physical
Impact Endpoints into Productivity Shocks

Contemporaneous effect
of climate damages on the
productivity of individual
sectors in each region

Distinct physical
effects of climate
change on a given
sector

Regional Damage: A, t = A [Z^ „Z2

-------
Key Points

• lAMs would be constructed so as to have sectoral as well as
regional detail in production, consumption and climate
damages

• Based on simulated climatic changes at the regional scale,
we would first want to elaborate impacts by category of
physical endpoint, sector, region and future time period

• Only then would we aggregate across endpoints to
generate sector-by-region trajectories of shocks

• No aggregate damage function per se, so transparent
causal chain from both ex ante shocks (A) and ex-post
adjustments in regional/sectoral output and consumption
(i.e., reactive adaptation) to ultimate welfare effects

-------
Implications:

The Marginal External Cost of Carbon

Climate impacts of an additional unit of carbon
energy use at t = 0, cumulated over future periods:

Marginal utility of
consumption of output of
affected sector

potential output
by region/sector

Marginal effects of impact
endpoints on productivity of
sectors in each region

IrI,=(0,..,T)
-------
A Critical Review of the State of

Current Practice

-------
The Damage Function Approach

(Nordhaus)

Based on exogenous global-scale climate change
projections, elaborate impacts (some denominated by
category of physical endpoint, some by sector) by region for
a benchmark global mean temperature change (2.5°C)

Monetize, aggregate and express the resulting estimates as
a proportion of future potential GDP

Use assumptions about how proportion will scale with (a)
income and (b) a simplified index of the magnitude of
climate change (global mean temperature change, 1) to
specify aggregate damage function (D)

Some baby steps toward the sector/impact category
disaggregation of the canonical model: sea-level rise in
RICE-2010

-------
The Marginal External Cost of Carbon

as Calculated in RICE

\ /

Marginal utility of
aggregate regional
consumption

Potential
regional GDP

Marginal effect of
temperature on aggregate
output in each region

IrZHo,...,T)r,,P,a,ur/ecr>tx jrt x dvr/d%

Xlf(d%/d G^xdJfp/dEc)

Marginal effects of

reservoir GHG
concentrations on
global mean
temperature change

Marginal effect of
emissions on the
global carbon cycle

-------
Difficult Problems, with Elusive Remedies

• Aggregation is inevitable, but on the modeling side, the
key research need is to explicitly incorporate sectoral
detail (j), impact categories (/) in lAMs

• Major obstacle: lack of empirical or detailed modeling
studies; most of existing ones don't go past 2050 (cf.
World Bank, 2010; Eboli et al., 2009)

• Targeting later decades for quasi-empirical assessment
is critical, as 2050 likely to underestimate the onset of
warming and climate damages late in the century

• But the further one goes out in time the less confidence
one has in detailed estimates, leading to tradeoff
between overall response magnitude and
sectoral/regional specificity

• No easy way to cut this Gordian knot

-------
CGE Models for Climate Impact Analysis

• Promising new direction, particularly given
increasing climate model skill at regional scales

• An explicitly multi-regional/multi-sectoral
approach: compute shocks based on exogenous
information on physical endpoints by sector,
impose consequent shocks on affected sectors
within the various regions

• Key problems are CGE models' recursive-dynamic
character (which precludes anticipation of
impacts), limited time horizon (2050 in ICES)

2.'

-------
Adaptation and Technological Change

Karen Fisher-Vanden, Elisa Lanzi, David Popp, Ian Sue Wing, Mort Webster

The purpose of this talk is to provide a brief summary of the state of the science on the
influences of adaptation on the social cost of climate change. Specifically, the charge was to
discuss (not necessarily in this order):

(1) relevant studies on the observed or potential effectiveness of adaptive measures, and on
private behaviors and public projects regarding adaptation;

(2) relevant studies on how to forecast adaptive capacity;

(3) how adaptation and technical change could be represented in an I AM (for at least one
illustrative sector);

(4) whether the information required to calibrate such a model is currently available, and, if
not, what new research is needed; and

(5) how well or poorly existing IAMs incorporate the existing body of evidence on
adaptation.

A tall order, but important to get our arms around since estimates of the net impact of climate
change could be significantly higher if adaptation is not taken into account.1

As elaborated below, a number of general insights have resulted from our brief foray into
this topic that have implications for the development of a future research program in this area.
First, modeling adaptation is inherently difficult given the nature of the adaptation process,
requiring advancements in modeling techniques. Second, although there has been good
empirical work done on impacts and adaptation costs, the coverage is limited requiring heroic
efforts to translate the results into model parameters. More work is needed to bridge the gap
between models and empirical studies. Lastly, adaptation-related technological change is
generally lacking in current models but could significant lower adaptation cost estimates. This
stems from a general lack of understanding of the process related to this type of technological
change. More empirical work is needed in this area.

What is unique about the adaptation process that justifies the need to add features to
existing integrated assessment models (IAMs)? First, adaptation is in response to current or
anticipated impacts and comes in different forms: (a) reactive (e.g., changes in heating/cooling
expenditures; treatment of disease; shifts in production); and (b) proactive (e.g., infrastructure
construction (e.g., seawalls); early warning systems; water supply protection investments. In
some IAMs adaptation would occur endogenously in reaction to changes in prices due to climate
impacts—e.g., more power plants built to deal with increases in demand for air conditioning;
shifts in production in reaction to higher prices of factors negatively impacted by climate change.
However, many adaptation activities that would occur in reality, such as investment in flood
protection, would not occur in a simulated model unless there is explicit representation of
climate damages to induce reactive expenditures and proactive investments.

Second, unlike mitigation investments where investments today result in reductions
today, proactive adaptation investments are made today to provide protection against possible
future impacts. Thus, adaptation investment decisions are inherently intertemporal and therefore

1 For the U.S., Mendelsohn et al. (1994) estimates that the net impact of climate change on the farming sector will be
70% less if adaptation is included while Yohe et al. (1996) estimates that the net impact on coasts will be
approximately 90% less (Mendelsohn (2000)).

-------
models need to include intertemporal decision making for proactive adaptation investments, in
order to trade off future damages and current adaptation investment expenditures. Not only are
we making intertemporal adaptation decisions, we are specifically making proactive adaptation
investments under uncertainty. Whether we invest and how much to invest all depends on our
expectations regarding future impacts and how we value the future. Therefore, we need a model
that allows for intertemporal decision-making under uncertainty.

Climate damages and adaptation strategies are locally- or regionally-based. Therefore,
ideally the model will include regional detail or will apply a method to aggregate up to a more
coarse regional representation. Climate damages and adaptation expenditures are also sector
specific—e.g., certain sectors will be impacted more than others and adaptation expenditures will
be directed at specific sectors (e.g., electric power, construction). Thus, a model with sectoral
detail or a way to aggregate these sector-specific impacts and expenditures is desirable.

The demand for adaptation solutions will induce adaptation-related technological change.
Do inducements for adaptation-related technological change differ markedly from mitigation-
related technological change, requiring a different modeling approach? To the extent that
adaptation activities may be region or sector specific, markets for new adaptation techniques will
be smaller than for new mitigation techniques, making private sector R&D investments less
attractive. Given this, as well as the case that adaptation investments are largely public
infrastructure investments, distinguishing between public R&D and private R&D may be
important. Note that this is more than a question of simply basic versus applied science, but
driven by the nature of demand for the final product, much in the same way that the government
finances most R&D for national defense. Thus, the model needs to be capable of distinguishing
between private and public investments and include mechanisms of public revenue raising to
fund these projects.

To summarize, to be able to capture adaptation strategies, an ideal IAM would include the
following features:

• Explicit modeling of climate damages/impacts

• Intertemporal decision making under uncertainty

• Endogenous technological change

• Regional and sectoral detail for impacts and adaptation strategies

• Connection with empirical work on impacts and adaptation

Is it feasible or even desirable to have all of these features represented in a single model, since
transparency is lost as more features are added? It is important to measure the trade-offs:

• How much of this needs to be specifically represented in the model and how could be
represented outside of the model

• To cite Jake Jacoby: "different horses for different courses." Do we need a suite of
models each designed to capture a subset of these features?

• How important is each of these features to the social cost of climate change? Sensitivity
analysis could be useful here to assess whether we even need to worry about including
certain features.

-------
To answer these questions, it is useful to first survey what features currently exist in IAMs. A
number of modeling approaches have been taken to capture impacts and adaptation. Computable
general equilibrium (CGE) models have the advantage of providing sectoral and regional detail
and capturing the indirect effects of impacts and adaptation. Thus, given its structure, CGE
models can more easily accommodate regional and sectoral-specific damage functions. Most
CGE models, however, do not include the type of intertemporal decision making required to
model proactive adaptation investment decisions, given the computational demands required by a
model with detailed regions and sectors. However, there have been a number of CGE models
that have been used to estimate the cost of climate change impacts; for example,

• DART (Deke et al, 2001)—to study the cost of coastal protection

• FARM (Darwin and Tol, 2001; Darwin et al, 1995)—includes detailed land types
to study the effects of sea level rise and impacts of climate change on agriculture.

• GTAP-E/GTAP-EF (Bosello et al, 2006; Bigano et al, 2008; Rosen, 2003)—has
been used to study induced demand for coastal protection; effects of rising
temperatures on energy demand (Bosello et al, 2007); health effects of climate
change (Bosello et al, 2006); effects of climate change on tourism. Focuses on
one impact at a time.

• Hamburg Tourism Model (HTM) (Berittella et al, 2006; Bigano et al, 2008)—
used to study the effect of climate change on tourism.

• ICES (Eboli et al, 2010)—models multiple impacts simultaneously: impacts on
agriculture, energy demand, human health, tourism, and sea level rise.

Another set of models used to study climate change impacts and adaptation fall under the
category of optimal growth models. These models include intertemporal optimization but
typically lack sectoral and regional detail given the computational demands this would require.
These include:

• DICE/RICE (Nordhaus, 1994; Nordhaus and Yang, 1996; Nordhaus and Boyer,
2000)—DICE comprises one region, one aggregate economy, and one damage
function aggregating many impacts. RICE comprises 13 regions, each with its
own production function and damage function.

• AD-DICE/AD-RICE (de Bruin et al, 2009)—DICE/RICE model with adaptation.
Adaptation investment added as a decision variable which lowers damages and
faces an adaptation cost curve. Residual damages are separated from protection
costs in the damage function.

-------
increases when temperature exceeds a certain threshold. Simple adaptation is
included in the model which reduces damages. Assumes developed countries can
reduce up to 90% of economic impacts while developing can reduce up to 50%.
All regions can reduce up to 25% of non-economic impacts.

• FUND (Tol et al, 1995; Tol, 1995)—referred to as a "policy optimization" model.
Exogenous variables include population (from the World Bank), GDP per capita
(from EMF 14), and energy use. Endogenous variables include atmospheric
concentrations, radiative forcing, climate impacts (species loss, agriculture,
coastal protection, life loss, tropical cyclones, immigration, emigration, wetland,
dryland), emission reductions (energy or carbon efficiency improvements,
forestry measures, lower economic output), ancillary benefits (e.g., improved air
quality), and afforestation. The model comprises 9 regions with game theoretics
and eight market and non-market sectors, each with its own calibrated damage
function. Adaptation is modeled explicitly in the agricultural and coastal sectors,
and implicitly in other sectors such as energy and human health where the
wealthy are assumed to be less vulnerable to the impacts of climate change. No
optimization in the base case—just simulation. In the optimization case, the
model is choosing the optimal level of emissions reductions by trading off costs
and benefits of reductions.

Another class of models involves hybrid combinations of the above model types. For example,

• Bosello and Zhang (2006) couple an optimal growth model with the GTAP-E
model of Burniaux and Truong (2002) to study the effects of climate change on
agriculture

• Bosello et al (2010) couple the ICES CGE model with an optimal growth model
(AD-WITCH) to study adaptation to climate change impacts.

• AD-WITCH (Bosello et al, 2010)—an optimal growth model with detailed
bottom-up representation of the energy sector. Comprises 12 regions where the
following seven control variables exist for each region: investment in physical
capital, investment in R&D, investment in energy technologies, consumption of
fossil fuels, investment in proactive adaptation, investment in adaptation
knowledge; and reactive adaptation expenditure. These alternative uses of
regional income compete with each other.

To parameterize these models, most modeling teams look to empirical studies of impacts
and adaptation and are faced with similar frustrations. First, as elaborated in Agrawala and
Fankhauser (2008), the empirical work in the area of adaptation is severely lacking. The authors
find that although information exists on adaptation costs at the sector level, certain sectors (e.g.,
coastal zones and agriculture) are studied more heavily than others. Second, most empirical
studies are not done with modeling applications in mind. Most modelers find themselves forced
to devise methods to scale up from the regional and sectoral results generated by empirical
studies.

There have been a few recent studies that have attempted to summarize the empirical
work on adaptation costs; e.g.,

-------
• Agrawala and Fankhauser (2008)—provides a critical analysis of empirical work
on adaptation costs. Tables summarize empirical sectoral studies on adaptation
costs. Sectors include coastal zones, agriculture, water resources, energy demand,
infrastructure, tourism and public health.

• World Bank (2010)—report from The Economics of Adaptation to Climate
Change (EACC) study. Seven sector-specific studies: infrastructure, coastal
zones, water supply and flood protection, agriculture, fisheries, human health,
extreme weather events. Provides detailed estimates of adaptation costs; some
generated using dose response functions with engineering estimates and some
generated from sector-specific models.

• UNFCCC (2007)—regional studies (Africa, Asia, Latin America, and small island
developing States) on vulnerability; current adaptation plans/strategies; future
adaptation plans/strategies. Most information from national communications to
the UNFCCC, regional workshops, and expert meetings.

A few modeling teams have made serious attempts to integrate existing empirical work
on adaptation into their model; for example,

• AD-DICE/AD-RICE: starts with damage functions of Nordhaus and Boyer (2000) and
uses empirical studies to separate residual damages from adaptation costs. Various
studies on adaptation measures for certain sectors (i.e., agriculture and health) and
estimates of adaptation costs from existing studies are used. Also, other model results—
e.g., results from FUND—are used to estimate adaptation costs in response to sea level
rise. Empirical studies to separate residual damages from adaptation costs are not
available for many of the sectors—i.e., other vulnerable markets; non-market time use;
catastrophic risks; settlements—so assumptions were made in order to separate the
damage costs. However, these sectoral estimates are ultimately aggregated up to one
damage cost number and one adaptation cost number to fit with the one sector structure
of the model.

• AD-WITCH: Uses empirical information from the construction of damage functions in
Nordhaus and Boyer (2000), the studies in Agrawala and Fankhauser (2008); and
UNFCCC (2007) to separate residual damages from adaptation costs. Similar to AD-
DICE, using these empirical studies to separate the damage estimates in Nordhaus and
Boyer (2000) into residual damages and adaptation costs.

Comparing this brief survey of existing work in this area with the list of required
modeling features needed to model adaptation, a couple of key research voids stand out. First,
none of these models include decision making under uncertainty, and for good reason. It is
difficult to do. Optimal growth models like DICE with intertemporal decision making are
deterministic and fully forward-looking. Past approaches to modify such a model to be
stochastic usually entail the following steps:

1) Create multiple States of the World (SOWs), each with different parameter assumptions
and different probabilities of occurrence;

2) Index all variables and equations in the model by SOW;

3) Add constraints to the decision variables so that for all time periods before information is
revealed, decisions must be equal across SOWs.

-------
The problem with this approach is that it rapidly becomes a very large constrained
nonlinear programming problem, and often the model will not converge to a solution for more
than a trivial number of SOWs. The general problem of decision making under uncertainty is a
stochastic dynamic programming problem that requires the exploration of a large number of
samples of outcomes in every time period. The challenge is to fully explore the sample space
while keeping the model computationally tractable. Promising on-going research by Mort
Webster and his team at MIT could offer an alternative approach to modeling decision making
under uncertainty. Webster's NSF-funded project team is currently developing a formulation
based a new approach called Approximate Dynamic Programming, introduced by Powell (2007)
and others. This approach implements dynamic programming models by iteratively sampling the
state space using Monte Carlo techniques, approximating the value function from those samples,
and using approximate value functions to solve for an approximate optimal policy, then
repeating. This approach has been used successfully in other contexts for very large state spaces.
Mort Webster's team is currently developing an ADP version of the ENTICE-BR model to study
R&D decision making under uncertainty.

Second, adaptation-related technological change is largely absent in current models.

Most models are calibrated using existing knowledge of adaptation strategies and costs with no
allowance for improvements in these strategies and technologies. AD-WITCH (Bosello et al,
2009) does attempt to account for this by including investment in adaptation knowledge as a
decision variable that competes with other types of investment. Investments in adaptation
knowledge accumulate as a stock which reduces the negative impact of climate change on gross
output. However, the lack of empirical studies on adaptation-related technological change limits
the modelers' ability to calibrate their model based on empirical knowledge. In the case of AD-
WITCH, adaptation knowledge investments only relate to R&D expenditures in the health care
sector where empirical data exist. This suggests that more empirical research in this area is
desperately needed.

Third, differences in adaptive capacity or differences in the ability of regions to adapt to
climate change are also important to capture in model analyses given the implications for
distributional effects but are typically not represented in existing models. The FUND model
implicitly captures adaptive capacity in the energy and health sectors by assuming wealthier
nations are less vulnerable to climate impacts. However, it seems that only one model, AD-
WITCH, attempts to explicitly capture adaptive capacity through the inclusion of investments in
adaptation knowledge as a decision variable. Not only does this variable capture R&D
investments in adaptation-related technologies as discussed in the previous paragraph, it also
captures expenditures to improve the region's ability to adapt to climate change. Issues arise,
however, when the model is calibrated since the modelers were only able to identify one source
of qualitative information on adaptive capacity (i.e., the UNFCCC (2007) report discussed
above) which only covers four aggregate regions (Africa, Asia, small island developing States,
and Latin America). Assumptions were then made to translate this information to the regional
representation and model parameters in AD-WITCH.

-------
better techniques to translate results from empirical studies to models are needed since the
sectoral and regional detail of empirical studies does not typically align with the sectoral and
regional detail in models. In general, to address the disconnect between empirical studies and
modeling needs, we as a research community need to devise better ways to facilitate
communication between empirical researchers and modelers.

References

Agrawala, S. and S. Fankhauser (2008), Economic Aspects of Adaptation to Climate Change,
OECD, Paris, France.

Berrittella, M., Bigano, A., Roson, R. and Tol, R.S.J. (2006), A general equilibrium analysis of
climate change impacts on tourism, Tourism Management, 25(5), 913-924.

Bigano, A., F. Bosello, R. Roson and R. Tol (2008). Economy-wide impacts of climate change: a
joint analysis for sea level rise and tourism, Mitigation and Adaptation Strategies for Global
Change, Springer, vol. 13(8), pages 765-791.

Bosello, F., C. Carraro and E. De Cian (2009). An Analysis of Adaptation as a Response to
Climate Change, Copenhagen Consensus Center, Frederiksberg, Denmark.

Bosello, F., C. Carraro and E. De Cian (2010). Climate Policy and the Optimal Balance between
Mitigation, Adaptation and UnavoidedDamage, FEEM Working Paper No. 32.2010.

Bosello, F., De Cian, E. and Roson, R. (2007), Climate Change, Energy Demand and Market
Power in a General Equilibrium Model of the World Economy, FEEM working paper n. 71.07.

Bosello, F., Roson, R. and Tol, R.S.J. (2006), "Economy wide estimates of the implications of
climate change: human health", Ecological Economics, 58, 579-591.

Bosello, F. and Zhang J. (2006), Gli effetti del cambiamento climatico in agricoltura, Questione
Agraria, 1-2006, 97-124.

Burniaux, J-M. and T. Truong (2002). GTAP-E: An Energy-Environmental Version of the GTAP
Model, GTAP Technical Papers 923, Center for Global Trade Analysis, Department of
Agricultural Economics, Purdue University.

Darwin, R., M. Tsigas, J. Lewabdrowski, and A. Raneses (1995). World Agriculture and Climate
Change. Agricultural Economic Report No. 703, US Department of Agriculture, Economic
Research Service, Washington, DC.

Darwin, R. F. and R. S. J. Tol (2001), Estimates of the Economic Effects of Sea Level Rise,
Environmental and Resource Economics 19, 113-129.

-------
De Bruin, K.C., R.B. Dellink and R.S.J. Tol (2009). AD-DICE: an Implementation of Adaptation
in the DICE Model, Climatic Change, 95: 63-81.

Deke, O., K. G. Hooss, C. Kasten, G. Klepper and K. Springer (2001), Economic Impact of
Climate Change: Simulations with a Regionalized Climate-Economy Model. Kiel Institute of
World Economics, Kiel, 1065.

Eboli, F., R. Parrado and R. Roson (2010), Climate-change feedback on economic growth:
explorations with a dynamic general equilibrium model, Environment and Development
Economics, 15:515-533.

Hope, C. (2006). The Marginal Impact of C02 from PAGE2002: An Integrated Assessment
Model Incorporating the IPCC's Five Reasons for Concern. Integrated Assessment, 6: 19-56.

IPCC (2001), Impacts, adaptation, and vulnerability, Contribution of working
group II to the third assessment report, Cambridge University Press.

Mendelsohn, R, Nordhaus, W, and Shaw, D. (1994). "The Impact of Global Warming on
Agriculture: A Ricardian Analysis", American Economic Review, 84: 753-771.

Mendelsohn, R. (2000). "Efficient Adaptation to Climate Change," Climatic Change, 45: 583-
600.

Nordhaus, W.D. (1994). Managing the Global Commons: The Economics of the Greenhouse
Effect. MIT Press, Cambridge, MA.

Nordhaus, W.D. and Z. Yang (1996). A Regional Dynamic General-Equilibrium Model of
Alternative Climate-Change Strategies, American Economic Review, 86(4), 741-765.

Nordhaus, W.D., and Boyer, J (2000). Warming the World: Economic Models of Global
Warming. MIT Press, Cambridge, MA.

Plambeck, E.L., C. Hope, and Anderson, J. (2007). "The Page95 Model: Integrating the Science
and Economics of Global Warming," Energy Economics, 19:77-101.

Powell, W.B, (2007), Approximate Dynamic Programming: Solving the Curses of
Dimensionality, Wiley-Interscience, Hoboken, New Jersey.

Roson, R., (2003), Modelling the Economic Impact of Climate Change, EEE Programme
Working Papers Series, International Centre for Theoretical Physics "Abdus Salam", Trieste,
Italy.

Tol, R.S.J. (1995). The Damage Costs of Climate Change Toward more Comprehensive
Calculations, Environmental and Resource Economics, 5: 353-374.

-------
Tol, R.S.J., T. Van der Burg, H.M.A. Jansen and H. Verbruggen (1995). The Climate Fund
Some Notions on the Socio-Economic Impacts of Greenhouse Gas Emissions and Emission
Reduction in an International Context (Institute for Environmental Studies, Vrije Universiteit,
Amsterdam).

UNFCCC (2007), Climate Change: Impacts, Vulnerabilities, and Adaptation in Developing
Countries, UNFCCC, Bonn, Germany.

World Bank (2010), The Costs to Developing Countries of Adapt to Climate Change, The Global
Report of the Economics of Adaptation to Climate Change Study, World Bank, Washington, DC.

Yohe, G., Neumann, J., Marshall, P., and Ameden, H. (1996). 'The Economic Cost of
Greenhouse-Induced Sea-Level Rise for Developed Property in the United States', Climatic
Change 32: 387-410.

-------
Adaptation and Technological

Change

Karen Fisher-Vanden, Penn State University

Elisa Lanzi, FEEM

David Popp, Syracuse University

Ian Sue Wing, Boston University

Mort Webster, MIT

-------
Purpose of the talk

Charge: To provide a summary of the state of the science on the
influences of adaptation on the social cost of climate change;
specifically, discuss

(1) relevant studies on the observed or potential effectiveness of adaptive
measures, and on private behaviors and public projects regarding
adaptation;

(2) relevant studies on how to forecast adaptive capacity;

(3) how adaptation and technical change could be represented in an I AM
(for at least one illustrative sector);

(4) whether the information required to calibrate such a model is currently
available, and, if not, what new research is needed; and

(5)

how well or poorly existing lAMs incorporate the existing body of
evidence on adaptation

-------
General conclusions

~ Modeling adaptation is inherently difficult. Requires
advancements in modeling techniques

~ Coverage of empirical work on adaptation limited. Requires
heroic efforts to bring into lAMs. Need to bridge gap
between models and empirical studies.

~ Adaptation-related technological change is lacking in current
lAMs. More empirical work is needed in this area to inform
existing models.

-------
What is unique about the
adaptation process?

i. Adaptation is in response to current or
anticipated impacts. Comes in two forms:

• Reactive—e.g.. changes in heating/cooling expenditures;
treatment of disease; shifts in production

• Proactive—e.g.. infrastructure construction (seawalls);
early warning systems; water supply protection
investments

Need explicit representation of climate damages to induce
reactive expenditures and proactive investment.

-------
What is unique about the
adaptation process?

2. Proactive adaptation investment decisions made
today to provide possible future protection;
decisions are therefore

• Inherently intertemporal

• Made under uncertainty

Need model that can allows for intertemporal decision-
making under uncertainty.

-------
What is unique about the
adaptation process?

3. Is adaptation-related technological change
markedly different from mitigation-related
technological change?

• Public R&D versus private R&D?

• Inducements different?

Need model capable of distinguishing between these two
types of technological change.

-------
What is unique about the
adaptation process?

4. Impacts and adaptation responses are locally-
regionally-based. Adaptation expenditures are
sector-specific.

Therefore, need model that includes

° regional detail
° sectoral detail

° method to aggregate to more coarse representation

-------
Important model features for adaptation

~ Explicit modeling of climate damages/impacts

~ Intertemporal decision making under uncertainty

~ Endogenous adaptation-related technological
change

~ Regional and sectoral detail

~ Connection with empirical work on impacts and
adaptation

-------
Model

Impacts

Regional
detail

Sectoral
detail

Link to empirical
work on adaptation

Intertemporal?

Uncertainty

Adaptation

AD-WITCH

Region-specific
climate damage
functions

1 2 regions

Bottom-up

energy
sector(7)

To separate
adaptation costs
and residual
damages

Optimal
growth-Perfect
foresight

Application
where
uncertain R&D
modeled
implicitly

Investment in
proactive, reactive,
and knowledge
adaptation

AD-DICE/AD-
RICE

Region-specific
climate damage
functions (AD-
RICE)

1 3 regions
(AD-RICE)

One
aggregate
economy
for each
region

Similar to AD-
WITCH

Optimal
growth-Perfect
foresight

Adaptation
investment
included as
decision variable

PAGE

Region-specific

damage
functions for two
sectors
(economic and
noneconomic)

8 regions

One
economic
sector for
each region

IPCCTAR?

Simulation
model

Stochastically

models
catastrophic
events

Simple adaptation

included which
increases tolerable
level

FUND

Damage function
for each of 8
sectors

9 regions

8 market
and non-
market
sectors

Limited

Simulation
model

Application
with monte

carlo
simulation

Explicit in ag and
coastal sectors;
implicit in energy
and human health

GTAP-
E/GTAP-EF

Used for separate
impact studies

8 regions

CGE-8 or
17 sectors

Limited

Static

ICES

Models 5 impacts
simultaneously

8 regions

CGE-1 7
sectors

Dynamic
recursive

FARM

Sea level rise and
impacts on agric

regions—
detailed

i _i j.. _ _

CGE-1 3
sectors

Limited

Static

Coastal protection

-------
Empirical studies on adaptation

~ Agrawala and Fankhauser (2008)—OECD publication which
summarizes empirical work on adaptations costs.

• Sectors include: coastal zones; agriculture; water resources; energy
demand; infrastructure; tourism; and public health.

~ World Bank (2010)—report from the Economics of Adaptation
to Climate Change (EACC) research program at WB

• Seven sector-specific studies on adaptation costs: infrastructure; coastal
zones; water supply and flood protection; agriculture; fisheries; human
health; extreme weather events

~ UNFCCC (2007)—Four regional (Africa, Asia, Latin America,
and small island developing States) studies on vulnerability,
and current and future adaptation plans/strategies.

• Information from UNFCCC National Communications, regional
workshops, and expert meetings.

-------
Recommended future research areas-
Decision making under uncertainty

> Past approaches involve:

1 Create multiple States of the World (SOWs)

2. Index all variables and equations in model by SOW.

3. Solve by constraining decision variable to have single value across SOWs
in all time periods before information is known.

> Problem with this approach: Rapidly becomes intractable for
more than a few SOWs.

> New research by Mort Webster (MIT) applying Approximate
Dynamic Programming introduced by Powell (2007):

Sample state space using Monte Carlo techniques

Approximate value function from these samples

3. Solve for approximate optimal policy using these approximate value
functions

-------
Recommended future research areas-
Adaptation-related technological change

> Adaptation-related technological change largely absent in
current models

> Most models calibrated based on current adaptation cost
estimates. No allowance for technological improvements.

> Exception: AD-WITCH includes investment in adaptation knowledge which lowers
future cost of adaptation. Only applied to health care sector.

> Lack of empirical studies limits modeler's ability to represent
adaptation-related technological change in current models

> More empirical work in this area is desperately needed

-------
Recommended future research areas-
Empirical work on adaptive capacity

> Regional differences in adaptive capacity important to capture
in models. Will affect distributional effects of climate impacts

> Largely absent in existing models

> Exceptions:

FUND model assumes wealthier nations less vulnerable to climate impacts in the
energy and health sectors.

AD-WITCH's investment in adaptation knowledge also captures expenditures to
improve region's ability to adapt

> Although in both cases, modelers were limited by lack of
empirical data. UNFCCC (2007) provides adaptive capacity
measure but only for four aggregate regions.

> Heroic efforts required to translate this little empirical
information to model parameters

-------
Recommended future research areas-
Dynamics of recovery

> Lack of empirical evidence on the dynamics of recovery from
climate change impacts.

E.g., time to recovery, thresholds and factors affecting these variables

> Important for model calibration

> In general, need techniques to better translate results from
empirical studies to models; e.g.,

> Regional and sectoral detail do not typically align

*** Going forward, we need to devise better ways to facilitate
communication between empirical researchers and modelers.

-------
Questions?/Discussion?

-------
Multi-century scenario development and socioeconomic uncertainty

Brian O'Neill

National Center for Atmospheric Research (NCAR)

Abstract presented at the
Workshop on Modeling Climate Change Impacts and Associated Economic Damages
EPA-DOE, Washington, DC, Nov 18-19, 2010

Introduction

The social cost of carbon (SCC) sums the damages resulting from a unit emission of C02 today over the
infinite future. As a result, this quantity depends in principle on socio-economic and climate conditions
over all future time. In practice, SCC calculations are truncated over a finite period, and different factors
can change the relevance of damages that occur in the very long term. On the one hand, future
damages are discounted, which makes damages far in the future contribute less to the net present value
than those that occur in the nearer term. On the other hand, assumed growth in the size of the
economy and damages that increase in proportional terms with the amount of warming will tend to
increase the contribution of damages far in the future relative to those in the nearer term. Thus the
contribution of damages that occur beyond 2100 - and therefore the importance of socio-economic
conditions beyond 2100 - to the net present value of damages from a current emission are ambiguous.
The contribution of long-term damages to the specific calculations carried out in the Interagency
Working Group report on the social cost of carbon (IAWGSCC, 2010; hereafter the "SCC report") are not
specified, so it is unclear how important long-term socio-economic assumptions are to these
calculations. For the present purposes we assume they are relevant, at least in some scenarios. Since
the SCC calculations in the report are carried out to the year 2300, I focus on socio-economic futures
over this time period.

The scenario variables for which long-term assumptions are made in the SCC report include population,
GDP, C02 emissions, and non-C02 forcing. I focus on the first three, and compare the assumptions
made in the report to those available in the literature, for both the 2000-2100 and 2100-2300 time
periods. The quantitative scenarios used in the report are based on a set of scenarios drawn from EMF-
22, a recent model comparison exercise carried out by the Energy Modeling Forum. The report
describes the five scenarios it selected as follows: "EMF BAU scenarios represent the modelers'
judgment of the most likely pathway absent mitigation policies to reduce greenhouse gas emissions,
rather than the wider range of possible outcomes. Nevertheless, these views of the most likely outcome
span a wide range" (IAWGSCC, 2010, p. 16). It is worth noting, however, that typical practice in EMF
exercises is not necessarily to use the most likely socio-economic futures, but rather those that are well
suited to the particular exercise, or most convenient. There is no guarantee that their likelihood has
been judged in any way. In addition, there is no guarantee that they span the range of uncertainty in
the literature, and as we will see in the comparison, they typically do not.

-------
Population

The SCC report's population scenarios, based on the EMF scenarios, span a range of 8.7-10.4 billion
people globally by 2100. In comparison, the most recent long-term projections from the United Nations
(UN, 2004) and the International Institute for Applied Systems Analysis (NASA, Lutz, 2008) span ranges of
5.5 - 14 billion and 4.5 - 12 billion, respectively. Ranges of population assumptions employed in
emissions scenarios contained in the AR4 scenario database are similarly wide (although cover the low
end of this range less well). Thus, the SCC report clearly spans an overly narrow range of population
assumptions in 2100, and can be characterized as essentially clustering around a single medium
population assumption.

The report extends the projections to 2300 by assuming growth rates in 2100 linearly decline to zero,
producing a global population size in 2300 of 8-10.9 billion. Both the UN (UN 2003) and NASA (Lutz and
Scherbov, 2008) have carried out illustrative long-term projections to 2300. In neither case do these
institutions identify a most likely long-term outcome; rather, both emphasize that the projections are
intended to be illustrative of the consequences of different assumptions about fertility and mortality.
The UN produces three projections that differ only in terms of fertility rates, which are assumed to
converge to levels between 1.85 and 2.35 births per woman in the long term. This relative narrow
fertility range produces a range of global population size of 2.3 - 36.4 billion people in 2300. NASA
considers uncertainty in both fertility and mortality, and assumes that fertility converges to levels
between 1.0 and 2.5 births per woman, based on various lines of reasoning regarding determinants of
fertility behavior. These assumptions produce a range of global population size of between 40 million
and 47 billion people in 2300. Thus the SCC report essentially does not consider uncertainty in long
term population size at all, since the range of outcomes it considers vary by a factor of 1.4 between low
and high projections, while those in the demographic literature vary by a factor of more than 1000.

It is also worth noting that other dimensions of population beyond total size are likely important for
impacts and damages, including age structure. In the NASA projections, by 2300 age structures vary
widely as well. The proportion of the population aged 80+ increases from a few percent at present to
between 20% and 65% by 2300, indicating a completely unprecedented demographic structure.

GDP

The global GDP scenarios adopted in the SCC report, based on EMF models, range from a global
economy of $268-$397 trillion (in 2005 US $). In comparison, the scenarios in the AR4 database range
from $136 - $677 trillion, a range spanning a factor of 5 versus the range of a factor of 1.5 assumed in
the SCC report.

Beyond 2100 it is difficult to put the SCC assumptions in perspective given the dearth of long-term GDP
scenarios in the literature. The SCC approach is to assume that growth rates of global GDP decline
linearly to reach zero in 2300, based on the idea that "increasing scarcity of natural resources and the
degradation of environmental sinks available for assimilating pollution from economic production
activities may eventually overtake the rate of technological progress". While this is a plausible

-------
assumption, it is only one of many possible scenarios, and leads to a range of about $750 - $2200 trillion
by 2300.

In contrast, an illustrative exercise by Tonneson (2008) applies a range of different growth rates to
current GDP to project growth over the next 300 years. The growth rates are based on data for GDP per
capita over the past 180 years, defining three scenarios by selecting the slowest and fastest periods of
growth over this time span as well as the overall average growth rate. I combine these per capita
growth rates with the projected population growth from the UN and NASA scenarios, and with current
estimated per capita GDP, to produce illustrative long-range GDP projections. They span a range in 2300
from around $100 trillion to around $1 million trillion - a range of a factor of 1000, far wider than the
range of a factor of 3 covered by the SCC scenarios.

C02 emissions

The range of C02 emissions assumed in the SCC report result in emissions of 13-81 GtC02/yr. The AR4
database includes emissions scenarios that range from -14 to 109 GtC02/yr, which is somewhat larger
but of the same order of magnitude as the SCC range.

Beyond 2100, the report assumes that rates of decline in the carbon intensity of GDP are maintained
through 2300. This is based on the assumption that "technological improvements and innovations in
the areas of energy efficiency and other carbon reducing technologies ... will continue to proceed at
roughly the same pace that is projected to occur towards the end of the forecast period". As in the GDP
case, this is a plausible assumption but only one of many possibilities. It produces a range of emissions
in 2300 of about 10 to 102 GtC02/yr. In the scenario literature, scenarios for emissions beyond 2100
that are based on socio-economic assumptions (rather than simple extrapolations) are scarce. A point
of comparison, however, is provided by the emissions underlying the Representative Concentration
Pathways (RCPs), which are concentration and forcing scenarios that are providing the basis for climate
modeling simulations for the IPCC Fifth Assessment Report (Moss et al., 2010). The RCPs cover a similar
range of emissions as the SCC report through 2100, and then decline to low levels by 2300 (less than 10
GtC02/yr), so the SCC report covers a wider - and higher - range of emissions outcomes than will be
assumed in climate model simulations for AR5.

Discussion and conclusions

In summary, the comparisons carried out here show that the assumptions regarding population and
GDP pathways in the SCC report cover an overly narrow range of uncertainty over the entire time
horizon, but especially in the long term (beyond 2100). In contrast, the range of emissions pathways
through 2100 is reasonably consistent with the range found in the literature. Beyond 2100 the
emissions range is wider and higher than the range found in the RCP extensions, although the RCP
pathways were not designed to reflect uncertainty in very long term emissions. The comparison is
instructive however in that the global average temperature projected from the RCPs reaches 8 degrees
or more by 2300, and therefore the SCC pathways will result in temperature increases even higher than
this.

-------
There are several caveats to these conclusions that must be kept in mind. First, uncertainty ranges in
literature may themselves be too conservative. While the very long term population projections in the
literature have been constructed with an eye toward bounding assumptions that are reasonably well
grounded, the long term GDP projections were constructed in a back of the envelope style that may
underestimate actual uncertainty, and for C02 emissions no similar exercise was found in the literature
at all. Second, we have only examined uncertainty in very aggregate socio-economic variables such as
global population size and global GDP, but future impacts will depend perhaps more strongly on
additional dimensions of these variables, such as the regional and spatial distribution of people and
production, and the sectoral composition of production. It is difficult to interpret what particular levels
of GDP per capita in the long term even mean: what types of economic activities, relying on what types
of technologies, might be taking place 300 years in the future, and how will this affect impacts? Do
current damage functions apply even approximately to the socio-economic conditions that would obtain
in the very long term? Finally, we have ignored the potential for catastrophic impacts and their
implications for socio-economic conditions, despite the fact that some SCC emissions pathways could
lead to more than 8 degrees C in warming over this time period.

Based on these conclusions, and taking into account these caveats, we make the following
recommendations for future versions of the SCC report:

1. Demonstrate the influence of key sources of uncertainty on SCC calculations, including the
contribution to the SCC from different time periods.

2. Drop the use of a range of best estimates as a characterization of uncertainty, which under-
estimates uncertainty, and consider a substantially wider range of socio-economic futures,
through 2100 and 2300.

3. Consider simpler approaches to calculating damages in the very long term, when uncertainty is
highest, such as the use of generic economic sectors and damage types

4. Improve the characterization of uncertainty in SCC results and reconsider the use of probabilistic
outcomes, since the probabilities reflect uncertainty in only some parts of the calculation and
are highly conditional on assumption regarding other components, such as the socio-economic
pathways.

5. Consider linking to evolving work on RCPs and socio-economic scenarios that are consistent with
them.

References

Interagency Working Group on Social Cost of Carbon (IAWGSCC), United States Government, 2010.
Technical Support Document - Social Cost of Carbon for Regulatory Impact Analysis - Under Executive
Order 12866.

Lutz, W., and Scherbov, S. 2008. Exploratory extension of NASA's world population projections: Scenarios
to 2300. NASA Interim Report IR-08-022.

Lutz, W., Sanderson, W. and Scherbov, S. 2008. The coming acceleration of global population ageing.
Nature 451, 716-719.

-------
Moss, R. et al., 2010. The next generation of scenarios for climate change research and assessment.
Nature 463, 747-756.

Tonneson, M. 2008. The statistician's guide to Utopia: The future of growth. Trames 12, 115-126.

United Nations (UN), 2004. World Population to 2300. Department of Economic and Social Affairs,
Population Division. ST/ESA/SER.A/236. United Nations, New York.

-------
Multi-century scenario
development and socioeconomic

uncertainty

Brian O'Neill

National Center for Atmospheric Research (NCAR)

Workshop on Modeling Climate Change Impacts and Associated Economic Damages

EPA-DOE, Washington, DC, Nov 18-19, 2010

NCAR is sponsored by the National Science Foundation

NCAR

-------
Does the long term matter?

Years

-------
Scenario variables and SCC approach

• Population, GDP, C02 emissions, non-C02 forcing

• SCC approach

- "we aimed to select scenarios that span most of the
plausible ranges of outcomes for these variables"

- Select 5 scenarios from EMF-22 exercise, based on 4
models

- "EMF BAU scenarios represent the modelers' judgment of
the most likely pathway absent mitigation policies to
reduce greenhouse gas emissions, rather than the wider
range of possible outcomes. Nevertheless,

the most likely outcome "

- Extend from 2100 to 2300 for SCC calculation

-------
Population and Uncertainty

• 2010-2040

- meaningful projections with well characterized
uncertainty

• 2040-2080

- uncertainty begins to compound, but can still be
usefully characterized

• Beyond 2080

- compounding uncertainty, speculation about new
conditions, limits, and feedbacks

See UN, 2004; Lutz & Scherbov, 2008.

-------
Global Population to 2100

SCC scenarios
AR4 range

— 10

C
O

« 5

2000

2020

2040

2060

2080

2100

-------
Global Population to 2100

— 10
c
o

SCC scenarios

AR4 range
UN H, M, L

« 5

1 1 1

2060 2080 2100

2000

2020

2040

-------
Global Population to 2100

H/L= 1.2 vs. 2.8

SCC scenarios
AR4 range
UN H, M, L
11 ASA H, M, L

2000

Z020

2040

2060

2080

2100

-------
Effect of population on C02 emissions

u
o

-------
SCC extrapolation to 2300

• Growth rates at end of 21st century decline linearly to
zero by 2200

• "reasonably consistent with the United Nations long
run population forecast, which estimates global
population to be fairly stable after 2150 in the medium
scenario"

-------
Global Population to 2300

SCC scenarios

2000

2050

2100

2150

2200

2250

2300

-------
UN Long-Range Projections

. 2000-2300

• Country-specific

• Three fertility variants

- Long-run convergence at 1.85, 2.05, 2.35

• Life expectancy increases throughout the period

- from ~75 in 2050 to ~95 in 2300

• Migration zero after 2050

• Medium to 2300 is not the most likely! Designed to
produce a roughly stable population size

• Value: illustrate the implications of small differences
in future fertility levels

-------
Distribution of national fertility rates, 2005-2010

UN Long-Range

Interval, Total Fertility Rate

Population
replacement
level

Source: UN 2008.

-------
Global Population to 2300

SCC scenarios
UN H, M, L

2000

2050

2100

2150

2200

2250

2300

-------
NASA Long-Range Projections

. 2000-2300

• 13 world regions

• Four long-term fertility levels

- Long-run convergence at 1.0,1.5, 2.0, 2.5

• Life expectancy increases throughout the period

- maximum life expectancy of 120

• Migration zero after 2080

• Extensions to 2300 are not probabilistic

• Value: illustrate the implications of plausible range of
future fertility levels

-------
Distribution of national fertility rates, 2005-2010

Individual Population Above
replacement replacement replacement
level level level

Source: UN 2008.

-------
Global Population to 2300

—

'+->

Q.
O

5
0

SCC scenarios
UN H, M, L
NASA H, L

2000 2050 2100 2150 2200 2250 2300

-------
Global Population to 2300

H/L= 1.4 vs. 1000+

SCC scenarios
UN H, M, L
NASA H, L

11 ASA M + 1.0, 1.5, 2.0, 2.5

2000 2050 2100 2150 2200 2250 2300

-------
Global Population Age Structure

World, Proportion 80+

MaxLE- 120

x—

—

T—

C\J

Year

— - ' TFR= 1 TFR= 1.5 TFR= 1.7 TFR= 2 — -TFR=2.5

Source: Lutz & Scherbov, 2008.

-------
Global GDP to 2100

800

700

SCC scenarios

600

w
lh

o
o

C 500
o

400

300

O 200

100

2000

2020

2040

2060

2080

2100

-------
Global GDP to 2100

H/L= 1.5 vs. 5.0

800

700

600

w
lh

o
o

rsj

c 500
o

400

300

O 200

100

SCC scenarios

AR4 range

2000

2020

2040

2060

2080

2100

-------
SCC extrapolation to 2300

• Growth rates of per capita GDP at end of 21st century
decline linearly to zero by 2300

• Based on idea that "increasing scarcity of natural
resources and the degradation of environmental sinks
available for assimilating pollution from economic
production activities may eventually overtake the rate
of technological progress"

-------
Global GDP to 2300

-------
GDP Projections Based on Historical Experience

8000

7000

¦to-

CT)

6000

•s

5000

4000

CL
A

3000

2000

1000

Medium: 1.23 %/yr

Low: 0.56 %/yr

High: 2.88 %/yr

1800

1850

1900

1950

2000

Source: Date from Maddison, 2010. After Tonneson, 2008.

-------
Global GDP to 2300

H/L = 3 vs. 16

-------
Global GDP to 2300

H/L = 3 vs. 10"

1,000,000

— 100,000

o
o

rsj

c
o

fU
_D

o
o

10,000

£ 1,000

100

SCC scenarios
Medium GDP/cap growth
High GDP/cap growth

Low GDP/cap growth

2000

2050

2100

2150

2200

2250

2300

-------
Global C02 Emissions to 2100

120

tr 100
>«

8 80

SCC scenarios

u 20

2000

-20

D0(

2020

2040

2060

2080

2100

-------
Global C02 Emissions to 2100

120

100

>«
fM

8 80

SCC scenarios

AR4 range

u 20

2000

-20

D0(

2020

2040

2060

2080

- -2100

-------
Global C02 Emissions to 2100

120

-ET 100
>»

rsi

S 80

SCC scenarios

AR4 range
RCPs

2060

2080

- -2100

-------
SCC extrapolation to 2300

• Growth rates (decline rates) of carbon intensity
(C02/GDP) from end of 21st century maintained
through 2300

• "assumes that technological improvements and
innovations in the areas of energy efficiency and
other carbon reducing technologies ... will continue
to proceed at roughly the same pace that is projected
to occur towards the end of the forecast period"

-------
Global Fossil CQ2 Emissions to 2300

-------
Global Fossil C02 Emissions to 2300

140

>• 120

O
u

in
C

'(/)
to

O
u

"to
_Q
O

100

SCC scenarios
RCPs

2150

2200

2300

Source: RCP extensions from Meinshausen et al., submitted.

-------
RCP radiative forcing

Source: Meinshausen et al., submitted.

-------
8

RCP Temperature Projections

c Surface Temperatures

RCP-8.5

RCP-2.6

550 1900 1950 2000 2050 2100 2150 2200 2250 230C

et al., submitted.

-------
Summary: Uncertainty ranges

• Overly narrow range of uncertainty in population and
GDP over the entire time horizon, but especially in the
long term

• Range of emissions through 2100 reasonably
consistent with the range in the literature

• Range of emissions beyond 2100 higher than the
range in the RCP extensions (although not clear that
matters)

-------
Issues

• Current uncertainty ranges in literature may
themselves be too conservative

• Structure of future economy in the long-term: what
does a particular GDP/capita in 2300 mean?

• Regional distribution of people and production

• Catastrophic impacts: high emissions scenarios are
lots of warming! Median of 8+ degrees by 2300

• Do current damage functions apply even
approximately to conditions in the very long term?

- How relevant would a damage function created in 1700 be to
measuring climate damages today?

-------
Recommendations

• Demonstrate key sources of uncertainty, including the
contribution to SCC from different time periods

• Drop the use of a range of best estimates

- "single scenarios" are used for extensions beyond 2100,
when uncertainty is greatest

• Consider a substantially wider range of socio-
economic futures, through 2100 and 2300

• Consider simpler approaches to damages in very long
term: generic sectors and damage types

• Improve how uncertainty in results is characterized

- Use of probabilistic terms for SCC results is problematic
when only sub-components are quantified (i.e., results are
highly conditional)

• Consider linking to evolving work on RCPs and socio-
economic scenarios that are consistent with them

-------
Knowability and no ability in climate projections

Gerard Roe,

Dept. of Earth and Space Sciences,
University of Washington,
Seattle, WA

1. Introduction

The purpose of this note is to provide a referenced summary of the present
scientific understanding about future climate change, tailored towards the kind of
global climate factors that are captured in Integrated Assessment Models (lAMs).
In outline, it is organized as follows:

i) Equilibrium climate sensitivity is the long-term response of global temperature
to a doubling of atmospheric CO2. I review the causes of our current
uncertainty, and the prospects for reducing it.

ii)Two other measures of climate change are arguably more important in this
context. First the climate commitment is a measure of the climate change we
already face because of emissions that have already occurred.

iii)The very long timescales associated with attaining equilibrium, especially at
the high end of possible climate sensitivity, mean that the transient climate
response is of greater relevance for climate projections over the next several
centuries.

iv)Due to the inherent uncertainties in the climate system, a flexible emissions
strategy is far more effective in avoiding a given level of global temperature
change, than a strategy aims to stabilize CC^at a particular level.

v) Many important climate impacts are fundamentally regional in nature. Among
climate models, regional climate projections correlate only partially with global
climate projections.

This was prepared for the EPA Climate Damages Workshop, Washington, D.C.,
Nov 18-19, 2010.

2. Climate sensitivity

Climate sensitivity (here given the symbol T2x, and sometimes called the
equilibrium climate sensitivity) is the long-term change of annual-mean, global-
mean, near-surface air temperature in response to a doubling of carbon dioxide
above preindustrial values. It has long been a metric by which to compare
different estimates of the climate response to greenhouse gas forcing (e.g.,
Charney, 1979). There is a vast literature that has researched climate sensitivity
from every possible angle, ranging from state-of-the-art satellite observations of
Earth's energy budget, to geological studies covering hundreds of millions of

-------
years. A fine review of where things stand can be found in Knutti and Hegerl
(2008).

climateprediction.net

Feedbacks (RB07)

Modern obs. (AR10)

Glacial (H84)

^MlPCCAR4 (2007)

One concern that has been
raised is that the current
generation of IPCC climate
models (from the fourth
assessment, or AR4) does not
span the range of climate
sensitivity that is allowable by
observations (the blue
histogram in figure 1 clusters

too narrowly around the modes of the other pdfs). The reason for this appears to
be that the IPCC climate models do not sample the full range of possible aerosol
forcing (Armor and Roe, 2010). This should not be surprising since they are
designed to represent the "best" estimate of climate (something akin to the mode
of the distribution). However, since these computer models are the only tools
available for modeling regional climates, it should perhaps be a concern that they
are under sampling the range of possible futures. I next outline briefly how
estimates are made from observations and models. The purpose of doing so is to
straightforwardly demonstrate the important sources of uncertainty.

2a. Estimates of climate sensitivity from observations.

A linear approximation of the Earth's energy budget is

R = H + X1T,

2 4 6 8 10

Climate sensitivity (°C)

Figure 1. Various estimates of climate sensitivity. In
order of the legend: i) from multi-thousand ensembles
from one climate model (Stainforth et al., 2005), ii) from
feedbacks with climate models (Roe and Baker, 2007),

iii) from modern observations (Armour and Roe, 2010),

iv) from glacial climates (Hansen et al., 1984), v) A
histogram of T2x from 19 main IPCC AR4 models
(IPCC, 2007).

(1)

where R is the radiative forcing (units W m"2), H is the heat going into the world's
oceans and being stored there, and X1T is the climate response in terms of the
global-mean, annual-mean, near-surface air temperature T, and the climate
sensitivity parameter, A. (e.g., Roe, 2009, Armour and Roe, 2010, and many
others). For silly historical reasons the terminology here can be confusing. A is a
more fundamental measure of climate system than T2x, since it does not depend
on any particular forcing. A and T2x are related in the following way. Let R2x be

-------
the radiative forcing due to a doubling of CO2 over pre-industrial values («4W
m"2). In the long-term equilibrium, ocean heat uptake goes to zero, and so the
climate sensitivity is just:

7~2x =A R

(2)

The point of this algebra is to make it clear that the goal of estimating climate
sensitivity from observations is the goal of estimating A from Equation (1):

X =

R-H

(3)

We have observations of T, R,
and H, whose probability
distributions are shown in figure
2. Hereafter we refer to R-H as
the climate forcing, since it is the
net energy imbalance that the
atmosphere must deal with. H
and T are actually quite well
constrained, as is the radiative
forcing associated with CO2 and
other greenhouse gases. As is
clear from figure, the major
source of uncertainty is R and, in
particular, the component of R
that is due to aerosols (small
airborne particulates that can be
either liquid or solid).

3.5

,2.5

03
.Q
O

1.5

0.5

-3-2-10123
Climate forcing (Wrrr2)

Figure 2: Probability distributions of the terms in the
Earth's energy budget, based on IPCC 2007, and
updated for newer ocean heat uptake observations.
See Armour and Roe, 2010 for details. Total climate
forcing is equal to R-H in Eq. 3. Also shown is the
total forcing excluding aerosols, which is the climate
forcing experienced by the Earth, if all anthropogenic
emissions ceased immediately.

The reason that aerosol forcing
is hard to constrain is that 1) the
spatial pattern and lifetime is
extremely complicated to
observe (they are primarily in the

Northern Hemisphere and downwind of major industrial economies); 2) some
aerosols have a cooling effect, some have a warming effect; 3) aerosols alter the
thickness, lifetime, and height of clouds - a powerful indirect effect that is hard to
measure and attribute properly. The community is confident, however, that the
net aerosol effect is almost certainly negative. More information about aerosol
uncertainties can be found in Menon (2004).

Thus, from Eqs. 2 and 3, the probability distribution of climate sensitivity comes
from combining a relatively narrow distribution (the well-known temperature
change) in the numerator with a relatively broad distribution (the much less well-
known climate forcing (i.e., R-H)) in the denominator of Eq. 3. It is this
combination that produces the skewed distribution seen in figures 1 and 3c. The

-------
graphs in figure 3 are the fundamental reason why we can say with great
confidence that it is very likely that observed forcing has not been large enough
to imply a climate sensitivity of less than about 1.5 °C. On the other hand,
uncertainties in observed forcing also mean that we cannot confidently rule out
the disconcerting possibility that the modern warming has occurred with small
climate forcing, which would imply very high climate sensitivity. Note that the
curves in figure 1 and 3 are consistent with the probabilities given in the 2007
IPCC report.

Figure 3: The calculation of climate sensitivity from observations involves combing a relatively
narrow probability distribution of T (panel a) in the numerator, with a relatively broad
distribution of F= H-R (panel b) in the denominator of Eq. (3). This leads to the skewed
distribution of climate sensitivity (panel c). Note the pdfs must be combined properly - it is not
just a simple division - but the point is hopefully clear.

2b. Estimates of climate sensitivity from models.

Climate sensitivity also can be estimated from climate models. Figure 1 shows
three such efforts. The first is the spread of T2x among the main IPCC AR4
models. One issue is that the mainstream IPCC AR4 climate models are not
designed to explore the edges of the probability distribution, but instead are
designed with the most likely combination of model parameters, and parameters
are 'tuned' to reproduce observed climate history. Clear evidence of that tuning
comes from the correlation of climate sensitivity and imposed aerosol forcing in
the models in such a direction that twentieth century observations tend to be
reproduced (Kiehl, 2007, Knutti, 2008). Such tuning is not problematic if models
are interpreted as reflecting combinations of climate sensitivity and aerosol
forcing that are consistent with observed constraints (Knutti, 2008). However
AR4 models do not fully span the range of aerosol forcing allowed by
observations (Kiehl, 2007; IPCC, 2007). This is the likely reason that the AR4
models under sample of the full range of possible climate sensitivity, as seen in
figure 1.

Climate sensitivity can also be estimated by using thousands of integrations of
the same climate model with the parameters varied by reasonable amounts, a
strategy pursued by the climateprediction.net effort (figure 1, e.g., Stainforth et
al., 2005). This work also found a skewed pdf of T2x. Roe and Baker (2007)
explain this in terms of a classic feedback analysis, summarized in figure 4. The
relationship between feedbacks and response also produces a skewed

-------
distribution because of the way that
positive feedbacks have a
compounding effect on each other
(e.g., Roe, 2009). The range of
feedbacks as diagnosed within the AR4
models produces a pdf of climate
sensitivity that is quite consistent with
the pdf estimated from observations
(figure 1). This should be expected
since it is observations that ultimately
provide constraints on the models.

2d. Prospects for improved
estimates of climate sensitivity.

Can a narrower range of climate
sensitivity be expected soon? One can
ask: how might more accurate
observations or better climate models
change the estimate of T2x?

Reducing uncertainty in either forcing or feedbacks would produce a narrower
range. However it is the nature of these skewed distributions that the mode of T2x
moves to higher values as the range of forcing or feedbacks is narrowed, leaving
the cumulative probability of T2x > 4.5°C stubbornly persistent (Allen et al., 2007;
Roe and Baker, 2007; Baker et al., 2010).

Another line of attack is to try to combine multiple estimates of climate sensitivity
in a Bayesian approach that might, in principal, significantly slim the fat tail of T2x
(e.g., Annan and Hargreaves, 2006). However, as with all Bayesian estimates,
the value of the analysis is critically sensitive to 1) the independence of different
observations; and 2) structural uncertainties within and among very complex
models (e.g., Henriksson et al., 2010; Knutti et al., 2010). An objective
assessment of these factors has proven elusive, rendering the information
obtained by the exercise hard to interpret, and there is an acute risk that it
produces overconfident estimates.

Overall it is probably prudent to anticipate that there will not be dramatic
reductions in uncertainty about the upper bound on climate sensitivity (Knutti and
Hegerl, 2008). On the timescale of several decades, Nature herself will slowly
reveal more of the answer. We will learn about the transient climate response
(see below) more quickly than the equilibrium climate sensitivity.

Those interested in understanding the above arguments in greater depth would
do well to read the work of Prof. Reto Knutti (at ETH in Switzerland) and his

Figure 4: Model feedbacks and climate
sensitivity. The black curve shows the
mapping between climate feedbacks (x-
axis, green curve), and climate response
(y-axis, red curve). See Roe and Baker,
2007 for details.

-------
collaborators. His research is of extremely high caliber, and quite accessible for a
non-specialist.

3. The climate commitment

What if all human influence on climate ceased overnight? Such a scenario—
called the climate commitment—informs us of the climate change we already
face due only to past greenhouse gas emissions. Framing the question this way
has proven to be useful in providing a conceptual lower bound on future climate
warming.

Early definitions of the climate commitment simply fixed CO2 concentrations at
current levels (e.g., Wigley, 2005; Meehl et al., 2005), but maintaining current
levels actually requires continued emissions. Lately the focus has been more
appropriately on the consequences of establishing zero emissions (e.g., Solomon
et al., 2009). Two important, though sometimes overlooked points should be
made. Firstly the geological carbon cycle means that, although much of the
anthropogenic CO2 ultimately gets absorbed by the ocean, some fraction —
about 25 to 40% — remains in the atmosphere for hundreds of thousands of
years (e.g., Archer et al., 2009). Secondly aerosols, have a short lifetime in the
atmosphere (days to weeks). Thus when human influence ceases, aerosols are

1800

2000 2200 2400 2600
Year

2800

1800 2000 2200 2400 2600 2800
Year

Figure 5: Idealized representation of the climate commitment following a cessation of all
human influence on climate. Based on Armour and Roe, 2010. Panel (a) shows a simple
view of how uncertainty in forcing has grown since 1800, as allowed by IPCC 2007 observed
uncertainties. After emission cease (here at yr 2000) the uncertain aerosols quickly vanish,
there is a jump in forcing due to sudden unmasking of the (relatively well-known) radiative
forcing due to C02 and other greenhouse gases, which then declines slowly over time (black
line). Panel (b) shows the temperature over this period, from a simple climate model. For
each possible trajectory of past climate forcing history, a different value of climate sensitivity
is implied, in order that the accurately known past warming is reproduced (low past forcing
requires high climate sensitivity, and vice versa). The light blue curve shows the 90%
confidence range, as permitted by uncertainties in observations, which ultimately grows to
be 0.3 to 6°C at equilibrium. The dark blue curve is the "likely' IPCC range (68%). It is this
range that is spanned by the main IPCC AR4 models because they under sample the
allowed range of past forcing. Note that these calculations here only include uncertainties
due to aerosols. The spread would be larger if uncertainties in GHG and ocean heat uptake
were included. Nonetheless the graph highlights that uncertainty in future temperatures is a
result of uncertainty in past forcing.

-------
rapidly washed out of the atmosphere and the effect of this is to unmask
additional warming due to the much more slowly declining CO2 (illustrated in
figure 2 and 5).

Figure 5 shows an idealized calculation of the climate commitment from Armour
and Roe (2010), which contains more details. The purpose of showing this is to
highlight that our uncertainty about future temperature comes primarily from our
uncertainty about past forcing. After ceasing all emissions, the degree and
trajectory of future warming depends on the state of the current climate forcing.
We face the disconcerting possibility that our ultimate climate commitment
already exceeds 2 °C, because of our current inability to rule out that past
warming occurred with relatively little climate forcing. In other words, the lower
flank of the pdf of the past climate forcing distribution (figure 5a) controls the
upper flank of the pdf of the future temperature response (figure 5b).

3a. Climate forcing and climate sensitivity are not independent.

Perhaps the most important point to emphasize for the application to integrated
assessment models (lAMs) is that climate sensitivity and climate forcing are not
independent of each other. For any projections made of the future, a starting
point for the current climate forcing must be assumed. We are currently quite
uncertain about what that starting point is. If aerosol forcing is strongly negative,
there is a strong implication that climate sensitivity is high. If aerosol forcing is
weak, climate sensitivity must be low. Uncertainties in climate forcing and climate
sensitivity must not be assumed to be independent.

4. The transient climate response.

Equilibrium climate sensitivity relates to a hypothetical distant future climate after
the system has equilibrated to a stipulated forcing. The transient climate
response over the course of a few centuries may be a more directly useful
property of the climate system. A formal definition of the transient climate
sensitivity has been proposed as the global-average surface air temperature,
averaged over the 20-year period centered on the time of CO2 doubling in a 1%
yr_1 increase experiment, which occurs roughly at 2070. While this metric may be
more relevant for the future, a negative trade-off is that its exact value depends
on this artificially defined trajectory of emissions.

For reasons discussed below, the transient climate response is much better
constrained than climate sensitivity. In the words of the IPCC, it is very likely (> 9-
in-10) to be greater than 1°C and very unlikely (< 1 -in-10) to be greater than 3
°C. Thus the community is much more confident about the evolution of the
climate over the coming century than it is about the ultimate warming.

4a. The immensely long timescales of high sensitivity climates.

-------
ocean, the upper layers must be warmed before the lower layers, and the more
the temperature change must be, the longer diffusion takes to work. A simple
scaling analysis (e.g., Hansen et al., 1985) shows that:

Climate adjustment time oc (climate sensitivity)2

Thus if it takes 50 yrs to equilibrate with a climate sensitivity of 1.5 °C, it would
take 100 times longer, or 5,000 yrs to equilibrate if the climate sensitivity is 15 °C.
Although Nature is of course more complicated than this, the basic picture is
reproduced in models with an (albeit simplified) ocean circulation. Figure 6a
shows one such calculation from Baker and Roe (2009), though there are others
(in particular see Held et al., 2010).

If lAMs are to be used to project out more than a few decades, it is critical that
they represent this physics correctly. A single adjustment time for climate, or a
deep ocean that is represented as a uniform block, cannot represent this
behavior.

The extremely high temperatures found in the fat tail of climate sensitivity cannot
be reached for many centuries for very robust physical reasons. Failure to
incorporate this fact will lead to a strong distortion of the evolution of possible
climate states, and of the subsequent IAM analyses based on them.

Time (yrs)

5. C02 stabilization targets are a mistake.

A prominent part of the conversation about action on climate change has
centered on what the right level of CO2 should be in the atmosphere (e.g.,
Solomon et al., 2010). Some advocate for 350 ppmv (e.g., Hansen et al. 2008),

-------
though we are already past 380 ppmv and climbing, others contemplate the
consequences of 450 ppmv (e.g., Hansen, et al., 2007), still others 550 ppmv
(Pacaia and Soccolov, 2004; Stern, 2007).

However decreeing and setting in stone a particular target for CO2 is
fundamentally the wrong approach, and a vastly inefficient way to avoid a
particular climate scenario. This point was made very elegantly and powerfully in
a study by Allen and Frame (2007), reproduced in figure 7. Panel a) shows a
scenario of what could happen if we decided today to stabilize CO2 at 450 ppmv
by 2100, and then waited for the climate to evolve. Our current best guess is that
would lead to an equilibrium temperature change of 2 °C, taking us to the edge of
what some have called dangerous climate change. However because of our
current uncertainty in climate sensitivity, the envelope of possible climate states
is quite broad by 2150. In other words, our hypothetical choice that we made
today still leaves us exposed to a quite broad envelope of risk. Note, though, that
figure 7a is consistent with figure 6 - temperatures in the fat tail of high climate
sensitivity are still very, very far from equilibrium at 2150.

Figure 7: reproduced from Allen and Frame (2007). Carbon dioxide-induced warming under
two scenarios simulated by an ensemble of simple climate models. (Left) C02 levels are
stabilized in 2100 at 450 ppm; (right) the stabilization target is recomputed in 2050. Shading
denotes the likelihood of a particular simulation based on goodness-of-fit to observations of
recent surface and subsurface-ocean temperature trends. The darker the shading, the
likelier the outcome.

Because the link between CO2 levels and global temperature is uncertain, and
because it is prudent to anticipate only incremental advances in our

-------
understanding, it is common sense to pursue a strategy that has built-in flexibility
rather than declaring a fixed concentration.

6. How well do global projections correspond to regional
projections?

Many of the most important climate impacts - changes in hydrology, storminess,
heat waves, snowpack, etc. - are fundamentally regional in nature. How reliable
is global climate change as a predictor of regional climate change? Since this is a
question about the future, we are forced to use climate models. Figure 8
analyzes how well global climate sensitivity correlates with local climate change
(in this case annual mean temperature and precipitation change in 2100),
comparing among eighteen different IPCC models (IPCC, 2007).

Figure 8: a) correlation among 17 IPCC climate models of their global equilibrium climate
sensitivity and their local annual-mean temperature change in 2100,; b) same as a), but for
annual-mean precipitation. Calculation made by N. Feldl from IPCC archived model output
based on the A1B emissions scenario, and similar plots for other variables are at
http://earthweb.ess.washington.edu/roe/GerardWeb/Publications.html.

It takes a correlation of r ~ 0.75 before half of the variance (i.e., r2) of the local
climate change is attributable to the global climate change. Only a very few
patches of the planet achieve even this level of correlation in annual temperature
(Figure 8a) and nowhere reaches this measure in annual precipitation (Figure
8b). This highlights that the connection between regional and global climate
change is not that strong. This result should not be surprising: though models
may all agree on the sign of the climate change in a given region, there is a great
deal of scatter and individual model vagaries in projecting the magnitude of the
climate change. Research into the limits of regional predictability is only just
beginning. A useful starting point is Hawkins and Sutton (2009).

Summary.

-------
and reflects an objective characterization of the best current understanding about
climate. All of the following points are consistent with (and in many cases drawn
from) that report.

2) A traditional measure of the planet's response, equilibrium climate sensitivity is
uncertain, primarily because of uncertainty in the radiative forcing due to
aerosols. This precludes us from calibrating our models of climate with greater
accuracy.

3) However a focus on climate sensitivity may be misplaced because of the
tremendously long timescales associated with reaching equilibrium - thousands
of years in the case of the fat tail of high climate sensitivity.

4) If all human influence were to cease today, the rapid loss of anthropogenic
aerosols from the climate would unmask CO2 warming, and the planet's
temperature would increase as a result. The degree of warming is quite
uncertain.

5) For related reasons, a strategy that aims to stabilize concentration of
greenhouse gasses at a particular level is a mistake, because the degree of
warming is still unpredictable. A strategy that aims for a flexible emissions will be
much more effective at preventing a particular level of warming.

6) lAMs have to make choices about how to represent climate forcing associated
with human activity. We are quite uncertain about what this level is right now. It is
crucial to appreciate that uncertainty in climate sensitivity and uncertainty in
climate forcing cannot be treated as independent.

7) Many climate damages both to humans and to the biosphere result from
regional climate factors. Unfortunately, there is relatively little agreement among
climate models about how global climate changes relate to local climate
changes, and this is especially true in some of the most vulnerable subtropical
regions. Thus the meaning of analyses that use only global temperature changes
to assign climate damages is unclear.

Acknowledgements: I'm grateful for helpful conversations and comments on this report from
Marcia Baker, Kyle Armour, Nicole, Feldl, Eric Steig, Yoram Bauman, David Battisti, and Steve
Newbold. All remaining errors are mine.

References cited:

Allen, M.R., and Frame, D.J., 2007: Call off the quest, Science, 318, 582-583.

Archer, D., et al., 2009: Atmospheric lifetime of fossil-fuel carbon dioxide. Annu. Rev. of Earth and

Planet. Sci. 37, 117-134.

Armour, K.C., and G.H. Roe, 2010: Climate commitment in an uncertain world. Submitted,

available at http://earthweb.ess.washington.edu/roe/GerardWeb/Home.html
Baker, M.B., and G.H. Roe, 2009: The shape of things to come: Why is climate change so

predictable? J. dim. 22, 4574-4589.

Baker, M.B., G.H. Roe, K.C. Armour, 2010: How sensitive is climate sensitivity. In preparation,

available at http://earthweb.ess.washington.edu/roe/GerardWeb/Home.html
Charney, J., and Coauthors, 1979: Carbon dioxide and climate: A scientific assessment. National

-------
Academy of Sciences, 22 pp.

Hansen, J.E., G. Russell, A. Lacis, I. Fung, D. Rind, P. Stone, 1985: Climate Response Times:
Dependence on Climate Sensitivity and Ocean, Science 229, 857-859

Hansen, J. E., et al., 2007: Dangerous human-made interference with climate: a GISS modelE
study, Atmos. Chem. Phys., 7, 2287-2312.

Hawkins, E., and R. Sutton, 2009: The potential to narrow uncertainty in regional climate
predictions' Bull. Am. Met. Soc., 90, 1095, doi: 10.1175/2009BAMS2607.1

Henriksson, S.V., E. Arja. M. Laine, J. Tamminen, A. Laaksonen , 2010: Comment on Using
multiple observationally-based constraints to estimate climate sensitivity by J. D. Annan
and J. C. Hargreaves, Geophys. Res. Lett., 2006, Climate of the Past, 6, 411414.

IPCC, 2007: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I
to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change
[Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M.Tignor and H.L.
Miller (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York,
NY, USA.

Kiehl, J.T., 2007: Twentieth century climate model response and climate sensitivity, Geophys.
Res. Lett., 34, L22710, doi:10.1029/2007GL031383.

Knutti, R., 2008: Why are climate models reproducing the observed global surface warming so
well? Geophys. Res. Lett. 35, L18704.

Knutti, R. and G.C. Hegerl, 2008: The equilibrium sensitivity of the Earths temperature to
radiation changes, Nature Geoscience, 1, 735-743, doi:10.1038/ngeo337

Knutti, R., R. Furrer, C. Tebaldi, J. Cermak and G.A. Meehl, 2010, Challenges in combining
projections from multiple models, Journal of Climate, 23, 2739-2758, DOI
10.1175/2009JCLI3361.1

Menon, S. 2004: Current uncertainties in assessing aerosol impacts on climate. Ann Rev. Env.
Res., 29, 1-30.

Meehl, G.A., et al., 2005: How much more global warming and sea level rise? Science, 307,
1769-1772.

Pacala, S., R.H. Socolow, 2004: Stabilization Wedges: Solving the Climate Problem for the Next
50 Years with Current Technologies. Science, 305 (5686): 968-972

Roe, G.H., and M.B. Baker, 2007: Why is climate sensitivity so unpredictable? Science 318, 629-
632, doi:10.1126/science.1144735.

Roe, G.H., 2009: Feedbacks, time scales, and seeing red. Ann.Rev. of Earth and Plan. Sci. 37,
93-115.

Solomon, S., et al., 2009: Irreversible climate change due to carbon dioxide emissions. Proc.
Natl. Acad. Sci. USA 106, 1704-1709.

Solomon, S. and fourteen others, 2010: Climate Stabilization Targets: Emissions, Concentrations
and Impacts over Decades to Millennia National Research Council, National Academy of
Sciences.

Stainforth, D., and Coauthors, 2005: Uncertainty in predictions of the climate response to rising
levels of greenhouse gases. Nature, 433, 403-406, doi:10.1038/nature03301.

Stern, N., 2007: Stern Review on the Economics of Climate Change: Part III: The Economics of
Stabilisation. HM Treasury, London:

Wigley, T.M.L., 2005: The climate change commitment. Science 307, 1766-1769.

-------
Knowability and no ability
in climate projections

Gerard Roe,

Earth Space Sciences,

Dept. Washington,

Seattle, WA

-------
-------
How sensitive is climate to changes in C02?

A traditional measure

• Climate sensitivity (or equilibrium cl

Definition: the long-term change in
annual-mean, global-mean, near-
surface air temperature to a doubling
of C02 above preindustrial values

(phew!, e.g., Arhenius, 1896, Charney, 1979)

• IPCC 2007 says:

Like/y (2-\n-3) 2.0 < AT < 4.5°C

Very unlikely ( 4.5 °C

(though IPCC says observations are less well fit with these values)

-------
Climate sensitivity

1. Different estimates

_ 0.4-
^ 0.35 -

~ 0.3 -

-g 0.25 -

-f—»

= 0.2 -

-g 0.15-
Q_

0.1 -

2 4 6

Climate sensitivity (°C)

climateprediction. net

-------
Climate sensitivity

1. Different estimates

2 4 6 8 10

Climate sensitivity (°C)

_ 0.4
^ 0.35

~ 0.3

-g 0.25

-f—'

= 0.2

"g 0.15

V—

0.1

-------
Climate sensitivity

1. Different estimates

Modern obs. (Armors Roe '10)

climateprediction.net
Feedbacks (Roe&Baker '07)

_ 0.4
^ 0.35

~ 0.3

-g 0.25

-f—'

= 0.2

"g 0.15

V—

0.1

2 4 6

Climate sensitivity (°C)

-------
Climate sensitivity

1. Different estimates

climateprediction. net
Feedbacks (Roe&Baker '07)
Modern obs. (Armor & Roe '10)
Glacial obs. (Hansen et al, '84)

2 4 6

Climate sensitivity (°C)

So why these values, and why this shape?

-------
Climate sensitivity

1.5 An aside

0.5

0.45

0.4

0.35

-4—'

0.3

'(f)

CD
"O

0.25

—

0.2

J3
O

0.15

0.1

0.05

— climateprediction.net
Feedbacks (Roe&Baker '07)

- - Modern obs. (Armors Roe '10)
Glacial obs. (Hansen et al. '84)
IPCC '07 Climate Models

Climate sensitivity (°C)

The main IPCC climate models under-sample the allowed range.
An issue for regional climate predictions?

-------
Climate sensitivity

2. Estimates from observations

Global energy budget:

AT A 7"

forcing

= storage +
(ocean)

atmospheric
response

In principle, get Rf, F, AT from observations, solve for then:

^ 2xC02 ^

Rf 2xC02 ~4 W m

-2

-------
Climate sensitivity

2. Estimates from observations

How much warming has there been since pre-industrial times?

Temperature change

-1 o 1

Temperature Change (°C)

Global mean temperature change is well observed

-------
Climate sensitivity

2. Estimates from observations

3.5

CM Q

'E d

2.5

in
c

CD r\
"D ^

5 -I c
CO 1 °
_Q
O

0.5

-3

Non CO2
Greenhouse
Gases

-1 0 1

Climate forcing (Wrrf2)

Numbers from
IPCC, 2007

Warming from C02 and other Greenhouse gases (CH4, 03)

(plus a tiny bit from solar)

-------
Climate sensitivity

2. Estimates from observations

3.5

C\J Q

'E ^

2.5

C/)

O r\

"D <-

5 ¦) c

CO 1 -0
X3
O

Ocean
Heat
Storage

Aerosols
Direct &
Indirect

Non CO2
Greenhouse
Gases

-1 0 1

Climate forcing (Wm~2)

Numbers from
IPCC, 2007
and

Lyman et al. (2010)

Cooling from heat storage in ocean, and aerosols

Aerosols: airborne particulates (solid/liquid)

have complicated effects (some warm, some cool, change clouds)

-------
Climate sensitivity

2. Estimates from observations

3.5

C\J Q

•E d

Rf -H = 0.9 ± 0.55Wm

r~2

2.5

CD rs
X3 <-
>.

-D 1 c
CO 1 °

_Q
O

CU*

Total
climate
forcing

-1 0 1

Climate forcing (Wm"2)

• Total climate forcing is quite uncertain and aerosols are
the culprit.

-------
Climate sensitivity

3. Estimates from observations

4 6 8

Climate Sensitivity (°C)

Fat tail is because aerosol forcing be quite negative

-------
Climate sensitivity

3. Estimates from models

Roe &

• Black curve is the relationship between climate feedbacks
and climate sensitivity.

-------
Climate sensitivity

3. Estimates from models

Roe & Baker, 2007

sum of climate feedbacks, f

• Green curve reflects current uncertainty in climate feedbacks.

-------
Climate sensitivity

3. Estimates from models

sum of climate feedbacks, f

• Red curve is resulting uncertainty in climate sensitivity.

-------
Climate sensitivity

3. Estimates from models

sum of climate feedbacks, f

• Red curve is resulting uncertainty in climate sensitivity.

-------
Climate sensitivity

4. Prospects for progress

a. Improved observations/models

Its hard!! Incremental improvements, but probably no
breakthroughs.

b. Combine different estimates?

Very hard to establish the degree of independence of
individual

estimates. (see Knutti and Hegerl, 2008)

c. Use other observations?

(e.g., NH vs. SH; pole-to-eq, \T; seasonality, trap, water
vapor)

Structural errors among models highly uncertain. (see Knutti et al, 2010)
Prudent not to expect big improvements any time soon....

-------
Climate commitment

1. What if all anthropogenic emissions ceased tomorrow?

Lifetimes: C02: centuries to 100,000 yrs+
Aerosols: days to weeks

Climate forcing (Wm~2)

-------
Climate commitment

1. What's already in store for us?

Lifetimes: C02: centuries to 100,000 yrs+
Aerosols: days to weeks

-2-10 1

Climate forcing (Wm"2)

Immediate loss of aerosols uGHG gas warming

-------
Climate commitment

1. What's already in store for us?

Idealized timeline of past and future climate forcing, if we stop everything today

What does
* the
climate do?

1800 2000

2200 2400
Year

2600 2800

90% error bounds,
IPCC numbers,
(Kyle Armour)

-------
Climate commitment

1. What's already in store for us?

Our best guess at what would happen

Year Year

90% error bounds,
IPCC numbers,
(Kyle Armour)

-------
Climate commitment

1. What's already in store for us?

But if past forcing has been high....

Year Year

90% error bounds,
IPCC numbers,
(Kyle Armour)

-------
Climate commitment

1. What's already in store for us?

But if past forcing has been low....

Year Year

90% error bounds,
IPCC numbers,
(Kyle Armour)

-------
Climate commitment

2. Past forcing and climate sensitivity are intrinsically related

If past forcing is strong climate sensitivity is low.
If past forcing is weak climate sensitivity is high.

For Integrated Assessment Models this matters:
- forcing (including aerosol forcing) cannot be
assumed to be independent of climate sensitivity .

-------
Transient evolution of climate

1. Heat uptake of the ocean is diffusive

Radiation
Balance

Upwelling

Mixed Layer

Diffusion Deep Ocean

of heat

Hansen et al. (1985) show this means that

Climate adjustment time
is proportional to
(Climate Sensitivity)2

-------
Transient evolution of climate

2. The fat tail grows very slowly

mean & 95% bounds

• Constraining the details of the far tail of climate sensitivity is not
useful on societally relevant timescales?

climate model response
(mean & 95% bounds)
to an instantaneous
doubling of C02

Time (yrs)

-------
C02 stabilization targets are a mistake

1. Climate response to fixed level of C02 is uncertain

E 3_

Fixed concentration target

EQM

i i

1950 2000 2050 2100 2150 00
Year

(Allen and Frame, 2007)

Stabilization target
of 450 ppm at 2100

High end sensitivities take a long, long time to be realized
There is still considerable uncertainty at 2150.

-------
C02 stabilization targets are a mistake

2. Flexibility is key

^ Adaptive concentration target

^ 4-

0 i i i i i i i

1900 1950 2000 2050 2100 2150 2200

Year

(Allen and Frame, 2007)

Concentration
target adjusted
at 2050.

A flexible emissions strategy is key to reaching a desired goal

-------
Does global climate predict local climate?

1. Is climate sensitivity a good predictor of regional change?

Among models, how well are varns in global climate sensitivity correlated with
varns in regional climate change at 2100?

If |corr. coeff.|
< 0.70 then

Annual mean temperature Annual mean precipitation <50% of local

change is

" y associated with

global mean

/ A \ change.

correlation coeff.

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8

The magnitude of local changes is affected by many factors

19 models from IPCC

. . _ . .. , _ . . _ _ 2007 report,

Global AT is quite a poor predictor of local AT, AP For more calculations

see my web site.

If impacts are local, should global AT be used to calculate damages? N^Feidi)

-------
Summary:

cy-

1. Uncertainty is not ignorance.

The planet is warming and its us that's doing it.

2. Climate sensitivity is uncertain b/c past forcing is

uncertain primarily aerosols .

3. Uncertainty in climate sensitivity and climate forcing

are not independent.

4. If climate sensitivity is high, it takes a very long time to

get there.

5. C02 stabilization targets are not an efficient way to

achieve a climate goal, (flexibility is vital)

6. Global climate is not a strong predictor of local climate

change.

-------
Extra slides....

-------
0

0 100 200 300 400 500 600

Time (yrs)

700 800 900 1000

-------
Time (yrs)

-------
AR4 models undersample climate commitment

¦ Dark blue is the IPCC 'likely7 (68% confidence interval) range of climate
sensitivity (2 to 4.5 C) and implied range of radiative forcing

¦ AR4 climate models span only this 'likely7 range

¦ R and A are correlated within AR4 and older models (Kiehl 2007, Knutti 2008)

-------
Effects of nonlinearity of climate feedbacks

-0.5

0 0.5

Feedback factor

-0

-------
By how much do observations have to change to change climate sensitivity

C\1
l„

c
0
T3

xi
o

5
AR

obs

("Mim Qonc PC*\

r*iim Cone Pr*\

-------
Aspects of feedbacks III.

How does uncertainty in feedbacks translate into uncertainty in
the system response?

Systems of strong positive feedbacks inherently less predictable

-------
Martin L. Weitzman
Notes for EPA & DOE discussion meeting
November, 2010

A. First thoughts on '"thinking about' high-temperature damages from potential
catastrophes in climate change."

'Thinking about' is the right phrase. This is a notoriously intractable area even to
conceptualize, much less to model or to quantify. Don't expect miracles or breakthroughs here
— too many "unknown unknowns" with seemingly non-negligible probabilities to feel
comfortable with.

B. What is the nature of the beast?

The economics of climate change consists of a very long chain of tenuous inferences
fraught with big uncertainties in every link: beginning with unknown base-case GHG emissions;
then compounded by big uncertainties about how available policies and policy levers will
transfer into actual GHG emissions; compounded by big uncertainties about how GHG flow
emissions accumulate via the carbon cycle into GHG stock concentrations; compounded by big
uncertainties about how and when GHG stock concentrations translate into global average
temperature changes; compounded by big uncertainties about how global average temperature
changes decompose into regional climate changes; compounded by big uncertainties about how
adaptations to, and mitigations of, regional climate-change damages are translated into regional
utility changes via a regional "damages function"; compounded by big uncertainties about how
future regional utility changes are aggregated into a worldwide utility function and what should
be its overall degree of risk aversion; compounded by big uncertainties about what discount rate
should be used to convert everything into expected-present-discounted values. The result of this
lengthy cascading of big uncertainties is a reduced form of truly enormous uncertainty about an
integrated assessment problem whose structure wants badly be transparently understood and
stress tested for catastrophic outcomes.

Let welfare W stand for expected present discounted utility, whose theoretical upper
bound is B. Let D=B-W be expected present discounted disutility. Here D stands for what
might be called the "diswelfare" of climate change. Unless otherwise noted, my default meaning
of the term "fat tail" (or "thin tail") will concern the upper tail of the PDF of InD, resulting from
whatever combination of probabilistic temperature changes, temperature-sensitive damages,
discounting, and so forth, by which this comes about. Empirically, it is not the fatness of the tail
of temperature PDFs alone or the reactivity of the damages function to high temperatures alone,
or any other factor alone, that counts, but the combination of all such factors. Probability of
welfare-loss catastrophe declines in impact size, but key question here is: how fast a decline
relative to size of catastrophe? When we turn to theory, it seems to highlight that the core "tail
fattening" mechanism is an inherent inability to learn about extreme events from limited data.

C. What do rough calculations show about this beast?

-------
requires a much more reactive damages function than the usual quadratic. Usual quadratic
damages function loses 26% of output for a 12dC temperature change. At 2% annual growth
rate, 12dC change 200 years from now implies that welfare-equivalent consumption then will
still be 37 times higher than today. If you use the standard quadratic damages function, you
cannot get much damage from extreme temperatures. If make a reactive damages function, such
that, say, 12dC temperature increase causes welfare-equivalent consumption to shrink to, say,
5% of today's level, then get very high WTP to reduce GHG target levels. Model is terrified of
flirting with high C02-e levels, especially above 700 ppm. Incredible dependence on degree of
risk aversion (2, 3, or 4?), fatness of tail PDFs (climate sensitivity PDF: normal, lognormal,
Pareto?), and so forth. My own tentative summary conclusion: tail of extreme climate change
welfare-loss possibilities is much too fat for comfort when combined with reactive damages at
high temperatures. It looks like this could influence such things as social cost of carbon.

D. Is there anything constructive to take away from this gloomy beast?

My tentative answer: a qualified maybe. Some possible rough ideas follow.

1. Keep a sense of balance. A small but fat-tailed probability of disastrous damages is not
a realization of a disaster. Highly likely outcome is a future sense that we dodged a bullet (like
Cuban missile crisis?). Yet when all is said and done, catastrophic climate change looks to me
like a very serious issue.

2. Try standard CBA or IAM exercises in good faith. But, be prepared - when dealing
with extremes - that answers might depend non-robustly upon seemingly-obscure assumptions
about tail fatness, about how the extreme damages are specified (functional forms, parameter
values, etc.), assumptions about rates of pure time preference, degrees of risk aversion, Bayesian
learning, C02 stock inertia, CH4 releases from clathrates, mid-course correction possibilities,
etc. Some crude calculations seem to indicate great welfare sensitivity to seemingly-obscure
factors such as the above, most of which are difficult to know with any degree of precision. Do
CB As and IAMs, study answers, but maybe don't try to deny the undeniable if these answers are
sensitive to tail assumptions in a highly nonlinear welfare response to extreme uncertainty.

3. Should we admit to the public that climate change CBA looks more iffy and less
robust than, say, CBA of S02 abatement, or would this be self defeating?

4. Maybe there should be relatively more research emphasis on understanding extreme
tail behavior of climate-change welfare disasters. Alas, this is very easy to say but very difficult
to enact. How do we learn the fatness of PDF tails from limited observations or experience?

5. A need to compare how fat are tails of climate-change welfare loss with how fat are
tails of any proposed solutions, such as nuclear power, below-ground carbon sequestration, etc.

6. Suppose that a lot of expected present discounted disutility is in the bad fat tail of the
welfare-loss PDF. Realistically, how can we limit some of the most horrific losses in worst-case
scenarios? Can we filter-learn fast enough to offset residence time of atmospheric C02 stocks
by altering GHG emission flows in time to work? Is tail fatness an argument for developing an
emergency-standby backstop role for fast geoengineering? Any other backstop options? Take-
home lesson here: hope for the best and prepare for the worst. At least we should be prepared,
beforehand, for dealing with ugly scenarios, even if they are low-probability events. Should the
discussion about emergency preparedness begin now?

-------
Earth System Tipping Points

Timothy M. Lenton

School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, UK
Definitions

A tipping point is a critical threshold at which the future state of a system can be qualitatively altered
by a small change in forcing1. A tipping element is a part of the Earth system (at least sub-continental
in scale) that has a tipping point1. Policy-relevant tipping elements are those that could be forced
past a tipping point this century by human activities. Abrupt climate change is the subset of tipping
point change which occurs faster than its cause2. Tipping point change also includes transitions that
are slower than their cause (in both cases the rate is determined by the system itself). In either case
the change in state may be reversible or irreversible, Reversible means that when the forcing is
returned below the tipping point the system recovers its original state (either abruptly or gradually).
Irreversible means that it does not (it takes a larger change in forcing to recover). Reversibility in
principle does not mean that changes will be reversible in practice.

Tipping elements in the Earth's climate system

'Arctic Sea-Ice Loss,

MB Melt of
Greenland Ice Sheet

Permafrost and
Tundra'Coss?

P^^jimatic
[Change-Induced
HOzone Hole?

Indian
MonsoonT"
Chaotic!
Multistability

Sahara
[Greening!

West African
Monsoon Shift

Dieback
of Amazon
Rainforest

Change' in ENSdJH
Amplitude or Frequency

rAntarctic Bottorr

Instability^oftWest Antarctic;
•JlaTSheetJH

population density [persons per km2]

' ' i i i i i i i i i i i I

no data 0 5 1 0 20 100 200 300 400 1000

Figure 1: Map of potential policy-relevant tipping elements in the Earth's climate system overlain on
population density. Question marks indicate systems whose status as tipping elements is particularly
uncertain.

-------
We should be most concerned about those tipping points that are nearest (least avoidable) and
those that have the largest negative impacts. Generally, the more rapid and less reversible a
transition is, the greater its impacts. Additionally, any positive feedback to global climate change may
increase concern, as can interactions whereby tipping one element encourages tipping another. The
proximity of some tipping points has been assessed through expert elicitation1,3. Proximity, rate and
reversibility have been also assessed through literature review1, but there is a need for more detailed
consideration of impacts4. The following are some of the most concerning tipping elements:

The Greenland ice sheet (GIS) may be nearing a tipping point where it is committed to shrink1,3.
Striking amplification of seasonal melt was observed in summer 2007 associated with record Arctic
sea-ice loss5. Once underway the transition to a smaller ice cap will have low reversibility, although it
is likely to take several centuries (and is therefore not abrupt). The impacts via sea level rise will
ultimately be large and global, but will depend on the rate of ice sheet shrinkage. Latest work
suggests there may be several stable states for ice volume, with the first transition involving retreat
of the ice sheet onto land and around 1.5 m of sea level rise6.

The West Antarctic ice sheet (WAIS) is currently assessed to be further from a tipping point than the
GIS, but this is more uncertain1,3. Recent work has shown that multiple stable states can exist for the
grounding line of the WAIS, and that it has collapsed repeatedly in the past. It has the potential for
more rapid change and hence greater impacts than the GIS.

The Amazon rainforest experienced widespread drought in 2005 turning the region from a sink to a
source (0.6-0.8 PgC yr"1) of carbon7. If anthropogenic-forced8 lengthening of the dry season
continues, and droughts increase in frequency or severity9, the rainforest could reach a tipping point
resulting in dieback of up to ~80% of the rainforest10"13, and its replacement by savannah. This could
take a few decades, would have low reversibility, large regional impacts, and knock-on effects far
away. Widespread dieback is expected in a >4 °C warmer world3, and it could be committed to at a
lower global temperature, long before it begins to be observed14.

The Sahel and West African Monsoon (WAM) have experienced rapid but reversible changes in the
past, including devastating drought from the late 1960s through the 1980s. Forecast future
weakening of the Atlantic thermohaline circulation contributing to 'Atlantic Nino' conditions,
including strong warming in the Gulf of Guinea15, could disrupt the seasonal onset of the WAM16 and
its later 'jump' northwards17 into the Sahel. Whilst this might be expected to dry the Sahel, current
global models give conflicting results. In one, if the WAM circulation collapses, this leads to wetting
of parts of the Sahel as moist air is drawn in from the Atlantic to the West15,18, greening the region in
what would be a rare example of a positive tipping point.

The Indian Summer Monsoon (ISM) is probably already being disrupted19,20 by an atmospheric
brown cloud (ABC) haze that sits over the sub-continent and, to a lesser degree, the Indian Ocean.
The ABC haze is comprised of a mixture of soot, which absorbs sunlight, and some reflecting sulfate.
It causes heating of the atmosphere rather than the land surface, weakening the seasonal
establishment of a land-ocean temperature gradient which is critical in triggering monsoon onset19.
Conversely, greenhouse gas forcing is acting to strengthen the monsoon as it warms the northern

-------
land masses faster than the ocean to the south. In some future projections, ABC forcing could double
the drought frequency within a decade19 with large impacts, although it should be highly reversible.

Estimation of likelihood under different scenarios

If we pass climate tipping points due to human activities (which in IPCC language are called "large
scale discontinuities"21), then this would qualify as dangerous anthropogenic interference (DAI) in the
climate system, Relating actual regional tipping points to e.g. global mean temperature change is
always indirect, often difficult and sometimes not meaningful. Recent efforts suggest that 1 °C global
warming (above 1980-1999) could be dangerous as there are "moderately significant"21 risks of large
scale discontinuities, and Arctic sea-ice and possibly the Greenland ice sheet would be threatened1'22.
3 °C is clearly dangerous as risks of large scale discontinuities are "substantial or severe"21, and
several tipping elements could be threatened1. Under a 2-4 °C committed warming, expert
elicitation3 gives a >16% probability of crossing at least 1 of 5 tipping points, which rises to >56% for a
>4 °C committed warming. Considering a longer list of 9 potential tipping elements, Figure 2
summarizes recent information on the likelihood of tipping them under the IPCC range of projected
global warming this century.

o
o
o

r\i

0)
>
o

_Q
05
ClO

6 -

5 -

4 -

3 -

2 -

1 -

1 ^ Certain
¦

| More likely than not
¦

0.5 ^ As likely as not
¦

Less likely than not

0 Won't happen

Figure 2: Burning embers diagram for the likelihood of tipping different elements under different
degrees of global warming23 - updated, based on expert elicitation results3 and recent literature.

Early warning prospects

An alternative approach to assessing the likelihood of tipping different elements is to try and directly
extract some information on their present stability (or otherwise). Recent progress has been made in
identifying and testing generic potential early warning indicators of an approaching tipping point1,24
27. Slowing down in response to perturbation is a nearly universal property of systems approaching
various types of tipping point25,27. This has been successfully detected in past climate records
approaching different transitions24,25, and in model experiments24"26. Other early warning indicators
that have been explored for ecological tipping points28, include increasing variance28, skewed
responses'"8'29 and their spatial equivalents30. These are beginning to be applied to anticipating
climate tipping points. For climate sub-systems subject to a high degree of short timescale variability
('noise'), flickering between states may occur prior to a more permanent transition31. For such cases,

-------
we have recently developed a method of deducing the number of states (or 'modes') being sampled
by a system, their relative stability (or otherwise), and changes in these properties over time32.

Applying these methods to observational and reconstructed climate indices leading up to the
present, we find that the Atlantic Multi-decadal Oscillation (AMO) index, which is believed to reflect
fluctuations in the underlying strength of the thermohaline circulation, is showing signs of slowing
down (i.e. decreasing stability) and of the appearance of a second state (or mode of behavior). On
interrogating the underlying sea surface temperature data (used to construct the index), we find that
recent significant changes are localized in the northernmost North Atlantic, and are investigating the
possible relationship with changes in Arctic sea-ice cover. Meanwhile, some other climate indices,
e.g. the Pacific Decadal Oscillation (PDO) show signs of increasing stability.

References

1 Lenton, T. M. et al., Tipping Elements in the Earth's Climate System. Proceedings of the
National Academy of Science 105 (6), 1786 (2008).

2 Rahmstorf, S., in Encyclopedia of Ocean Sciences, edited by J. Steele, S. Thorpe, and K.
Turekian (Academic Press, London, 2001), pp. 1.

3 Kriegler, E. et al., Imprecise probability assessment of tipping points in the climate system.
Proceedings of the National Academy of Science 106 (13), 5041 (2009).

4 Lenton, T. M., Footitt, A., and Dlugolecki, A., Major Tipping Points in the Earth's Climate
System and Consequences for the Insurance Sector, 2009.

5 Mote, T. L., Greenland surface melt trends 1973-2007: Evidence of a large increase in 2007.
Geophysical Research Letters 34, L22507 (2007).

6 Ridley, Jeff, Gregory, Jonathan, Huybrechts, Philippe, and Lowe, Jason, Thresholds for
irreversible decline of the Greenland ice sheet. Climate Dynamics, 1 (2009).

7 Phillips, Oliver L. et al., Drought Sensitivity of the Amazon Rainforest. Science 323 (5919),
1344 (2009).

8 Vecchi, G. A. et al., Weakening of tropical Pacific atmospheric circulation due to
anthropogenic forcing. Nature 441, 73 (2006).

9 Cox, Peter M. et al., Increasing risk of Amazonian drought due to decreasing aerosol
pollution. Nature 453, 212 (2008).

10 Cox, P.M. et al., Amazonian forest dieback under climate-carbon cycle projections for the
21st century. Theoretical and Applied Climatology 78,137 (2004).

11 Scholze, Marko, Knorr, W., Arnell, Nigel W., and Prentice, I. C., A climate-change risk analysis
for world ecosystems. Proceedings of the National Academy of Science 103 (35), 13116

(2006).

12 Salazar, Luis F., Nobre, Carlos A., and Oyama, Marcos D., Climate change consequences on
the biome distribution in tropical South America. Geophysical Research Letters 34, L09708

(2007).

13 Cook, Kerry H. and Vizy, Edward K., Effects of Twenty-First-Century Climate Change on the
Amazon Rain Forest. Journal of Climate 21, 542 (2008).

14 Jones, Chris, Lowe, Jason, Liddicoat, Spencer, and Betts, Richard, Commited ecosystem
change due to climate change. Nature Geoscience (submitted).

15 Cook, Kerry H. and Vizy, Edward K., Coupled Model Simulations of the West African Monsoon
System: Twentieth- and Twenty-First-Century Simulations. Journal of Climate 19, 3681
(2006).

16 Chang, Ping et al., Oceanic link between abrupt change in the North Atlantic Ocean and the
African monsoon. Nature Geoscience 1, 444 (2008).

17 Hagos, Samson M. and Cook, Kerry H., Dynamics of the West African Monsoon Jump. Journal
of Climate 20, 5264 (2007).

-------
Patricola, C. M. and Cook, Kerry H., Atmosphere/vegetation feedbacks: A mechanism for
abrupt climate change over northern Africa. Journal of Geophysical Research (Atmospheres)
113, D18102 (2008).

Ramanathan, V. et al., Atmospheric brown clouds: Impacts on South Asian climate and
hydrological cycle. Proceedings of the National Academy of Science 102 (15), 5326 (2005).
Meehl, G. A., Arblaster, J. M., and Collins, W. D., Effects of Black Carbon Aerosols on the
Indian Monsoon. Journal of Climate 21, 2869 (2008).

Smith, Joel B. et al., Assessing dangerous climate change through an update of the
Intergovernmental Panel on Climate Change (IPCC) "reasons for concern". Proceedings of the
National Academy of Sciences 106 (11), 4133 (2009).

Hansen, J. et al., Dangerous human-made interference with climate: a GISS modelE study.
Atmos. Chem. Phys. 7 (9), 2287 (2007).

Lenton, T. M. and Schellnhuber, H. J., Tipping the scales. Nature Reports Climate Change 1,
97 (2007).

Livina, V. and Lenton, T. M., A modified method for detecting incipient bifurcations in a
dynamical system. Geophysical Research Letters 34, L03712 (2007).

Dakos, V. et al., Slowing down as an early warning signal for abrupt climate change.
Proceedings of the National Academy of Sciences of the United States of America 105 (38),
14308 (2008).

Lenton, T. M. et al., Using GENIE to study a tipping point in the climate system. Philosophical
Transactions of the Royal Society A 367 (1890), 871 (2009).

Scheffer, M. et al., Early warning signals for critical transitions. Nature 461, 53 (2009).

Biggs, R., Carpenter, S. R., and Brock, W. A., Turning back from the brink: Detecting an
impending regime shift in time to avert it. Proceedings of the National Academy of Science
106 (3), 826 (2009).

Guttal, V. and Jayaprakash, C., Changing skewness: an early warning signal of regime shifts in
ecosystems. Ecology Letters 11, 450 (2008).

Guttal, V. and Jayaprakash, C., Spatial variance and spatial skewness: leading indicators of
regime shifts in spatial ecological systems. Theoretical Ecology 2, 3 (2009).

Bakke, J. et al., Rapid oceanic and atmospheric changes during the Younger Dryas cold
period. Nature Geoscience 2, 202 (2009).

Livina, V. N., Kwasniok, F., and Lenton, T. M., Potential analysis reveals changing number of
climate states during the last 60 kyr. Clim. Past 6 (1), 77 (2010).

-------
Earth system tipping points

IEV

University of East Anglia

Melt1

Greenland Ice Sheet

i=ice*

ClimaticRlhange- N

Bor. Permafrost

Induced

Forest Dieback Ozone Hole? Forest Dieback.> _an® >

Tundra

I LOSS?

Sahara
GriSTing

Chang&jrM^^

If Amazon
Rainforest

West African
Monsoon Shift?

~ \
Indian
Monsoon
Chaotic
Multistability

Instability of

West AntaiicJlCfl
¦
m

lte©@CD0@Q

Change s^i n. A n ta rjct i c
Water Formation?

Tim Lenton

School of Environmental Sciences, University of East Anglia, Norwich, UK
Special thanks to John Schellnhuber, Valerie Livina, Elmar Kriegler, Jim Hall

-------
Outline

IEV

University of East Anglia

Evidence on tipping points

Lata lassons from early warnings
pfQt.iuiion.-iry prlndplo 1fi9A-2MO

Probability under different scenarios

Early warning prospects

-------
"Little things can make a big difference" universityofEastAngna

Lenton et al. (2008) PNAS 105(6): 1786-1793

Tipping element

+ A component of the Earth system, at least sub-continental in
scale (~1000km), that can be switched - under certain
circumstances - into a qualitatively different state by a small
perturbation.

Tipping point

+ The corresponding critical point - in forcing and a feature of the
system - at which the future state of the system is qualitatively
altered.

-------
Changing climate states in the past University of East Anglia

Livina, Kwasniok & Lenton (2010) Climate of the Past, 6: 77-82

Number of states: 1, 4

-------
IEV

Policy relevant tipping elements University of EastAnglia

Lenton et al. (2008) PNAS 105(6): 1786-1793

+ Human activities are interfering with the system such
that decisions taken within a "political time horizon" (~100
years) can determine whether the tipping point is reached.

+ The time to observe a qualitative change plus the time to
trigger it lie within an "ethical time horizon" (~1000 years).

+ A significant number of people care about the fate of the
system.

-------
Policy-relevant forcing range

IEV

University of East Anglia

IPCC (2007)

CD
C

-------
Tipping elements in the Earth system University of East Anglia

Revised after Lenton etal. (2008) PNAS 105(6): 1786-1793

i^EGrcenland
I co Shoot]

4VodomaX
Permafrost

Cold Watorf
Coral Roofs?

Boreal
Forest'

Boreal]
Forest

iOccanfc
I Methanol
Hy&ratcSj?!

JwAtlanticy
AThormoKaline
jCirculatlorr

Himalayan
•[Glaciers?,

.SW North
America?

J India n]!
Summw
Monsoon

West African
"Monsoon

PtTropicaip
CoraljReofs?.

Amazon
Rainforest

Hp Niho^lB
SouthemTdscillatiorv

MarinolBioloqlcal
frCar bc'n"1 Fj urn p?;

[Wcst'Antarctic;
¦iotfSiioetM

Melting

Circulation Change
Biome Loss

Population Density [persons per km2]

i i i i i

no data 0 5 10 20

i i t i i i i i

100 200 300 400

1000

-------
Estimates of proximity

IEV

University of East Anglia

Lenton & Schellnhuber (2007) Nature Reports Climate Change

r 4-

£ 2-

o
o

OJ
=3
CO
CD

-o
O

ro
o
o
--j

Results from literature review and workshop

-------
Probabilities under different scenarios University of East Anglia

Kriegler et al. (2009) PNAS 106(13): 5041 -5046

Three different warming scenarios:

Low temperature corridor C1 Medium temperature corridor C2 High temperature corridor C3

2000 2060 2100 2150 2200 2050 2100 2150 2200 2050 2100 2150 2200

Imprecise probability statements elicited from experts for tipping scenarios
Example of collapse of Atlantic meridional overturning circulation:

Self assessment Self assessment Self assessment

2424442322332232 2424442322332232 2424442322332232

1 i i i f i i r i i f > i i r i—i > > »—r—»—i > i r i » > i i i > t r I ' ' M ' ' '—r 1 ' 1 '''''''

C2 3 4 5 7 8 9 1014151618192021 22 C2 3 4 5 7 8 9 1014 15 1618 19 20 21 22 C2 3 4 5 7 8 9 10 14 15 16 18 19 20 21 22

-------
LEi

>
oo

O
2

>
co

30
25
20
15
10

SP450

-! I

-I I—

1300 200022002400 2600 2800 3000

SP750

1 T 1

1 I

——:

s —

_l 1 1

1800 200022002400 2600 2800 3000

GENIE-1 model, data assimilated with Ensembie Kalman Filter, for IPCC AR4, 3°C climate sensitivity

OSP350

1800200022002400 260028003000

SP650

1800 2000 2200 2400 2600 2800 3000

SP550

1800 200022002400260028003000

SP1000

1800 2000 2200 240026002800 3000

-------
Greenland ice sheet

IEV

University of East Anglia

rfa

x ¦ ri

2007 melt days
anomaly relative
to 1988-2006

The Copenhagen Diagnosis (2009)
Net mass balance of Greenland ice sheet

100

03
0
>»

CO
CD
0

cd
E
o

0
CD

1960 1965 1970 1975

1980 1985 1990 1995
Observation period

2000 2005 2010

Low

Medium

High

Expert elicitation
for future warming |
scenarios:

Ml 2 3 4 5 6 8 9 11 12 13 14 15 Ml 2 3 4 5 6 8 9 11 12 13 14 15 Ml 2 3 4 5 6 8 9 11 12 13 14 15

-------
West Antarctic ce sheet

IEV

University of East Anglia

Shepherd & Wingham (2007) Science 315: 1529-1532 The Copenhagen Diagnosis (2009)

Net mass balance of Antarctic ice sheet

Expert elicitation
for future warming
scenarios:

DAIS

-30-20-15 -10 -5 0 5 10 15 20 30

Elevation rate (cm/year)

Medium High

« -150 -

(/)

(TS

E -200 -

I -250
ac

-300

-350

1960

1 1 1 1 1 1 1 1 1

1965 1970 1975 1980 1985 1990 1995 2000 2005 2010
Observation period

0.5 mm yr"1 sea-level rise R

-------
Amazon rainforest

IEV

University of East Anglia

Jones et al. (2009) Nature Geoscience 2: 484-487
Cook and Vizy (2009) Journal of Climate

Cox et al. (2000) Nature 408: 184-187

0.1 0.2 0.4 0.G 0.8 0.9

12 3 4

Global Temperature above Pre-Industrial (K)

Low

Medium

High

Expert elicitation
for future warming
scenarios:

>s 08

Lq 06

O o.*

* 0,
0

1 1 1 2 4 2 1 4 1 2

AMAZ

Al 3 5 6 8 9 10 11 12 14

Al 3 & 6 8 9 10 11 12 14

-------
El Nino I Southern Oscillation University of East Anglia

Guilyardi (2006) Climate Dynamics 26: 329-348, Yeh

Increase in ENSO amplitude in in most
realistic models under 3-6°C warmer
stabilised climate.

No clear change in El Nino frequency

Shift toward Central Pacific Modoki
replacing classic East Pacific El Nino?

Low

Medium

High

3 1 2 2 2 4 3 113

3 1 2 2 2 4 3 1 1 3

Expert elicitation
for future warming
scenarios:

jo 0 6

o 04

NINO

lill I ii

NINO

li'l ' I -i

¦ 1

N1 3 4 5 6 7 8 9 10 14

La Nina Modoki

La Nina

etal. (2009) Nature 461: 511-514

b El Nino Modoki

-------
Combined likelihood of tipping

IEV

University of East Anglia

Kriegler et al. (2009) PNAS
106(13): 5041-5046

Imprecise probability
statements from experts
formally combined

Under 2-4 °C warming:
>16% probability of
passing at least one of
five tipping points

Under >4 °C warming:
>56% probability of
passing at least one of
five tipping points

Low temperature corridor C1

Medium temperature corridor C2

High temperature corridor C3

2000 2050 2100 2160 2200 2060 2100 2150 2200 2060 2100 21&0 2200

Self assessment Self assessment Self a

2424442322332232 2424442322332232 24244

CMOC

l-iii. uILiu

CMOC

2322332232

CMOC

Atlantic

C2 3 4 5 7 8 9 1014 151618 192021 22 C2 3 4 5 7 8 9 1014 151618192021 22 C2 3 4 5 7 8 9 1014 1516 18 192021 22

213221222224 1213221222224

15 06

>• 03

o 0.6
CO

o 04
Cl

MGIS

¦ ¦ll...'ll

MGIS

i'.M ii i

213221222224

i 'i

MGIS

Greenland

VII 2 3 4 5 6 8 9 11 12 13 14 15 Ml 2 3 4 5 6 8 9 11 12 13 14 15 Ml 2 3 4 5 6 8 9 11 12 13 14 15

34443121143234 34443121 143234 3444 3 121143234

DAIS

Lul

.ii

DAIS

DAIS I

Antarctica

D1 3 4 5 6 7 8 9 10 11 12 13 14 15 01 3 4 5 6 7 8 9 10 11 12 13 14 15 D1 3 4 5 6 7 8 9 10 11 12 13 14 15

2 4 2 1 4 1 2

1 1 1 2 4 2 1 4 1 2

2 4 2 1 4 1 2

AMAZ AMAZ

il i ii l> I

IAMAZI

Amazon

Al 3 5 6 8 9 10 11 12 14 A1 3 5 6 8 9 10 11 12 14 Al 3 5 6 8 9 10 11 12 14

3 1 2 2 2 4 3

3 1 2 2 2 4 3 1 1 3

3 1 2 2 2 4 3

15 0.6

NINO

iiiili r ii

NINO

I'm ' i J

,l»1' l' I

El Nino

N1 3 4 5 6 7 8 9 10 14 N1 3 4 5 6 7 8 9 10 14 N1 3 4 5 6 7 8 9 10 14

Expert Expert Expert

-------
Interactions between tipping events

IEV

University of East Anglia

Kriegler et al. (2009) PNAS 106(13): 5041 -5046

Tipping events are connected A—if at least 5
experts judged that triggering A had a direct effect
on the probability of triggering B thereafter

Increase in probability

Decrease in probability

®l

Uncertain direction of change

-------
West African Monsoon

Weakening of the Atlantic
overturning circulation could
trigger collapse of West African
Monsoon (WAM)

Collapse of the WAM could in
turn cause increased inflow of
moist air from West

Requires ~3K warming of Gulf
of Guinea SSTs

Potential for increased food
production in the Sahel region

University of East Anglia

Chang et al. (2008) Nature Geoscience 1: 444-448

1 •• LSw ) yy precipitation

cvcle

January March May i\tf SejUemDer Novemlw "

Model months

Fractional

change in
annual

Sahelian JJAS precipitation change (mm day-1)

MIROC MED Sahel

1950 1960 1970 I960 1990 2000 2010 2020 2030 2040 2060 2060 2070 2OB0 2090 2100

Cook and Vizy (2006) Journal of Climate 19: 3681-3703

-------
Indian summer monsoon

IEV

University of East Anglia

Zickfeld, K. etal. (2005) GRL 32: Li 5707; Ramanathan, V. etal. (2005) PNAS 102(15): 5326

8
6

~ Observation

~ GHGs+S04_2050
¦ ABC 2050

S 4

= 2

¦n in ifl m Hi m IJI

oN oN rN kN rN nN oo

£ & & $ ^ ^ J>

ja' .Q' A
-------
Boreal forest

IEV

University of East Anglia

Lucht et al. (2006) Carbon Balance and Management 1: 6; Kurz et al. (2008) PNAS 105(5): 1551-5

1st Kyoto Period

Canadian forests have recently
switched from carbon sink to source
due to insect outbreaks

More widespread dieback
forecast under ~3°C global warming

(~7°C local warming)

Map shows change in
vegetation carbon content
from 2000 to 2100

LPJ model forced with
SRES A2 climate change
from HadCMS

-------
Yedoma permafrost

IEV

University of East Anglia

Khvorostyanov et al. (2008) Geophysical Research Letters 35, L10703

0 300 600 km

zjfk-—

Extent of permafrost melt
forecast to be proportional to
warming (not a tipping element)

But Yedoma, containing up to
500 PgC, could undergo runaway
meltdown due to biochemical heat
release

Estimated threshold is a 9 °C
regional warming, but note this
region warmed >3 °C in 2007

Provider.

ARCTIC OCiEAN,

. I \>/N

¦nans* ' '[->

Area of 1

oscow A jk S .^orir,k

^Jizhnry Novgorod ^

Kazan ^

'Yekaterinburg

irnd

•v Chelyabinsk

-------
Prospects for early warning

IEV

University of East Anglia

Held & Kleinen (2004) Geophysical Research Letters 31: L23207
Lenton et al. (2008) PNAS 105(6): 1786-1793

Generic early warning signals:
Slowing down
Increasing variability
Skewness of responses

System being
forced past a
bifurcation point

-------
Slowing down at the end of the ice age

IEV

University of East Anglia

Lenton, Livina, Dakos et al. (in prep.) PhiI Trans A

GRIP 5180 :

data 1

-22000

-20000

-18000

-16000

-14000

-12000

Detrended
data

residuals

-22000

j\i J '|i|iiiJ[iLi,| ^ i|||i ^ |L i'i^i i,Ji|ii

-20000

-18000

-16000

-14000

-12000

Early

warning

Indicator

-22000

-20000

-18000

-16000

-14000

-12000

-------
Atlantic Multi-decadal Oscillation index University of East Anglia

Results from Vasilis Dakos arid Valerie Livina

1^^l8^rT88Tl90"0^192 0~194Fl960'19 8 0 2000

time

Number of states: 1, 2, 3, 4

-------
Impacts of tipping

IEV

University of East Anglia

Lenton, Footitt & Dlugolecki (2009)

Allianz / VWVF report:

Increased sea level rise

+ +$25,158 billion
exposed assets in
port megacities

Amazon dieback and
drought

Indian summer
monsoon disruption

Aridification of
southwest North
America

Populations exposed to 1 -in-100-yr flood events

90,000

80,000
70,000
60,000
50,000
40,000
30,000
20,000
10,000

I. fl.rri. , rrl,

Europe Hsm Anerioa Ocearta Sau&i America

100,000
90,000
80,000
70,000
60,000
50,000
40,000
30,000
20,000
10,000

hton OECO i|lrc LOC)

~ Current Exposure ~ No "Tipping ~Tipping

http://knowledqe.allianz.com/climate tipping points/climate en.html

-------
'Straw man' tipping point risk assessment University of EastAnglia

Tipping element

Likelihood of
passing a tipping
point
(by 2100)

Relative
impact** of
change in state
(by 3000)

Risk score
(likelihood x
impact)

Risk ranking

Arctic summer sea-ice

High

Low

Greenland ice sheet

Medium-High*

High

7.5

1 (highest)

West Antarctic ice sheet

Medium*

High

Atlantic THC

Low*

Medium-High

2.5

ENSO

Low*

Medium-High

2.5

West African monsoon

Low

High

Amazon rainforest

Medium*

Medium

Boreal forest

Low

Low-Medium

1.5

8 (lowest)

*Likelihoods informed by expert elicitation

**lnitial judgment of relative impacts is my subjective assessment

-------
Conclusion

IDV

University of East Anglia

+ Tipping elements in the climate system could be triggered this
century by anthropogenic forcing

+ If business-as-usual continues we should expect to pass
tipping points, i.e. high impact high probability events

+ Early warning systems are conceivable and could help
societies manage the risk posed by tipping points

+ More research is needed on the corresponding impacts in
order to do a proper risk assessment

Then put that data in your integrated assessment model!

-------
October 25, 2010 version

CATASTROPHIC CLIMATE CHANGE
Mike Toman, World Bank Development Research Group
Draft, subject to revision; please do not cite or quote. Responsibility for content is the author's alone.

INTRODUCTION

The question of how to assess prospects of climate change catastrophes has been the focus of a great deal
of recent research and debate. An example of the classic conundrum of low probability - high
consequences events, a climate change catastrophe is a highly unlikely event, but if it did occur it would
severely affect well-being across the world - though it would affect poor countries much more seriously
than richer countries.1 The larger geographical scale of climate change catastrophes distinguishes them
from more localized extreme events. The consequences of catastrophes also are in varying degrees very
costly, if not possible, to reverse.

Examples of global catastrophes include very large and relatively rapid increases in sea level from faster
melting and collapse of ice sheets, slower changes in ocean currents that have insidious effects on weather
patterns, and large scale destruction of forests and other ecosystems, fairly rapid loss of global forest
cover. Unlike sudden disasters such as earthquakes, the onset of these events is measured in multiple
decades or centuries; but once they occur it is impossible to reverse the impacts. Other permanent effects
of climate change are anticipated to be increases in the frequency and severities of droughts, floods, and
hurricanes, leading to corresponding destruction of crops, water supplies, and coastal infrastructure.
While each of these individual events is a more localized disaster, the cumulative effect could be a global
catastrophe created by the "cascading consequences" of more localized disasters occurring in relatively
quick succession, each amplifying the effects of others.

A key step in evaluating risks of climate change catastrophes is to assess not only the impacts on the
physical climate system, but also the consequences in terms of human impacts. The most immediate
implication is that while a physical "tipping point" may be reached at some unknown future date T°, the
human impacts will evolve more slowly, reaching an intensity viewed as catastrophic only at some date
T1 > T°. This distinguishes climate change from, for example, the risk of catastrophe posed by a gigantic
volcanic eruption, or nuclear war. While a gradual onset of impacts will not prevent a catastrophe if
reversal is not possible, it can provide a window of time for major action to avert or adapt to the threat - if
signals of the changes are detectable. More fundamentally, the assessment of what constitutes
catastrophic human impacts involves not just climate change and earth system science, but also inherent
value judgments about what magnitude and speed of consequences are deemed to be catastrophic. For
example, the now-often-cited "scientific near-consensus" about the urgent need to hold warming to less
than 2°C relative to pre-industrial times reflects more than a natural science evaluation of climate change
impacts.

1 In terms of absolute numbers, losses are likely to be larger in richer nations. As a percentage of GDP, however, less
developed countries are likely to face higher damages since most are more dependent on agriculture and less likely to have the
resources to adopt measures that could reduce damages.

2 This possibility appears to have received little systematic attention in reviews of climate change impacts by the IPCC and
others, though it figures prominently in discourse about national security consequences of climate change.

-------
October 25, 2010 version

damages may miss an important part of the problem. Yet, these consequences of an unlikely but possible
climate change catastrophe need to be weighed against a variety of other risks society faces.

Further complicating the problem is that climate change catastrophes may be better characterized by
ignorance than uncertainty. That is, not only do we not know the probability of a particular mega-
catastrophe occurring, we do not even know many of the possible outcomes. A catastrophe from climate
change could stem from a cause or have impacts that currently receive little attention.4 Some authors
have suggested that this level of ignorance, coupled with the very low probability of an event and the
possibility of extremely severe impacts, hamstrings the use of rational-choice based methods for
analyzing response options. This in turn requires confronting the possibility that attitudes of the broader
public about such events will not align very well with the results of a more systematic evaluation of the
pros and cons of different response options, raising questions about what sets of preferences and beliefs
should govern policy making.

CLIMATE CHANGE CATASTROPHES

The most widely discussed large-scale impact of climate change is global sea level rise. The collapse of
the West Antarctic Ice Sheet (WAIS) or Greenland ice sheets could lead ultimately to a sea level rise of
several meters, with consequences great enough to be considered a global catastrophe in the absence of
massive and costly relocation because of the number of people living near the coasts. A key uncertainty
is how rapidly this change in sea level might occur. Previously it had been thought that such large
changes might require much longer than a century, but some recent studies suggest that substantial change
could occur in this century. Anthoff et al. (2009) report figures for world losses (based on 1995 baseline
conditions) that are relatively small - on the order of 0.5% of world GDP for a 5 m rise. Dasgupta et al.
(2007) report figures for developing countries on the order of 6% of GDP, those these estimates do not
take account of possibilities for ex ante efforts to mitigate risks. On the other hand, estimates based on
historical baselines will tend to under-state the economic impacts of sea level rise by not taking account of
likely future growth in the coming years in the share of GDP concentrated in coastal areas.5

A second important category of global catastrophe risk involves disruptions of ocean circulation from
climate change, with potentially disastrous effects on regional weather patterns and long-term climate
(Vellinga and Wood 2008). Such impacts are most commonly seen as developing over many hundreds of
years. In contrast, very large-scale ecosystem disruptions could occur significantly sooner. Changes in
ecosystems resulting from changes in temperature and rainfall incidence and increased climate variability
have the potential to cause very significant loss of biodiversity—on the order of 20-30% extinction within
a few decades. There is also the prospect of major changes in vegetation, in particular, irreversible

3 For many classes of disasters and catastrophes, the most extreme small percent of the situations represent a significant
proportion of the losses. We have witnessed this "fat tail" phenomenon recently with terrorist deaths and losses in a financial
crisis. 9/11 and the 2008-09 financial meltdown caused more deaths and dollar losses respectively than all terrorist incidents
and financial catastrophes in the post WWII era. With such phenomena, losses are better characterized by a power law than by
a normal or even lognormal distribution. The debate about fat tails in relation to climate catastrophes has been a subject of
lively recent debate among Weitzman, Pindyck, Nordhaus, and others.

4 The history of the past 40 years is sobering with respect to the ability to identify catastrophe risks. In 1970, nuclear war would
have been the leading contender for any world catastrophe, and looking forward few would have predicted the major looming
threats of the current era, which would include not just climate change, but also global pandemics and terrorism.

5 Using 1995 data, it has been estimated that around 400 million people would be impacted by a 5 m rise in sea level and that a
WAIS collapse in 100 years could cause, at the peak, 350,000 forced migrations a year for a decade (Nicholls, Tol and Vafeidis
2008).

-------
October 25, 2010 version

conversion of forest to grassland, desertification, and acidification of the ocean (Smith, Schneider,
Oppenheimer et al. 2009). Another cause for significant concern is the possibility that positive feedback
effects in the climate change process itself could occur (e.g., liberation of trapped methane from ice, rapid
increases in CO2 from vegetation dieback, or increased heat absorption as glaciers retreat), causing the
abovementioned changes to occur more rapidly.

There also has been significant scientific research on how climate change can effect more localized
disasters, such as heat waves, flooding, droughts, and changes in hurricane frequency or intensity. Less
understood is how a number of smaller disasters all occurring over a relatively short time period could
mutually reinforce each other in such as way that the resulting "cascade of consequences" becomes a
global catastrophe. Extreme events can have secondary consequences that generate substantial amounts
of additional damages; secondary consequences in turn can trigger tertiary consequences that further
amplify the adverse consequences; and so on. One example would be if increased drought from climate
change in different regions successively caused a series of local food shortages to occur in close
proximity, leading to political instability, a breakdown of civil order, large-scale migration for survival,
and regional conflicts. Another example could be a series of local fires occurring in climate-stressed
forests and grasslands overly widely dispersed areas, adding up to a large-scale destruction of resources,
ecosystem services, and livelihoods over a large area.

The compounding or amplifying effects of individual adverse impacts would be the result of exceeding
the resilience of a number of local socioeconomic systems in rapid succession. More frail components of
socioeconomic systems, such as marginal subsistence agriculture, represent potential places of
vulnerability. Cascading-event catastrophes could occur much more rapidly than the slower-onset global
impacts discussed above, especially as climate change accelerates and greater negative impacts occur at
local scales. It is possible that more comprehensive local monitoring of disaster risks may facilitate the
development of early warning indicators for cascading catastrophes. For example, if several years of
historically unusual drought weakened agricultural systems in many vulnerable parts of the world, there
would be a stronger basis for concern about cascading consequences than if agricultural failures were not
occurring in such rapid succession. However, the time interval for action to avert the potential
catastrophe could be short.

Traditional responses to the risk of extreme events are of limited value in mitigating risks of a mega-
catastrophe. The underlying changes in the climatic system could not be reversed over any time scale
relevant for decision-makers. Traditional insurance mechanisms will not function effectively for this type
of event, because the risks are "systemic" and cannot effectively be reallocated to diversify. Moreover,
significant international transfers from richer to more vulnerable poorer countries are unlikely when a
catastrophe affects broad swaths of the world.

EVALUATING CLIMATE CHANGE CATASTROPHE RISKS

The traditional economic model for decision making under uncertainty is expected utility theory, in which
decision makers maximize the utility they receive from potential outcomes weighted by the probability
the outcomes will occur. In the climate change economics literature, GHG abatement policies with the
expected net benefits over time are identified using dynamic Integrated Assessment Models (IAMs) that
compare the anticipated costs of abatement with avoided damages from climate change over time. By and
large these models are deterministic and are used for scenario-based comparisons of policies under
different assumptions about climate change damages and abatement costs. However, a literature has
developed in which catastrophes are treated as (usually known) large-scale rapid-onset economic damages

-------
October 25, 2010 version

with an uncertain date of occurrence, the probability of which increases as atmospheric GHG
concentrations rise.6

A common finding in these studies is that while the risk of such catastrophes increases the expected
economic benefits of more rapid GHG mitigation, the effect is not that significant qualitatively unless the
probability of nearer-term catastrophe is quite high, the size of the catastrophe is truly astronomical, or the
discount rate used to value future catastrophic impacts is quite low. The scientific information on
catastrophes summarized above indicates that catastrophes are extremely unlikely in any time frame short
of several decades at the very least, and that while the ultimate effects may indeed be huge, the most
severe impacts will develop only gradually. Until scientific understanding of climate change catastrophes
leads to stronger findings on their proximity and severity, the choice of discount rate will be the most
important determinant of the cost of future catastrophes in the expected-utility framework.

The discount rate issue in turn continues to be very hotly debated, and only a very brief summary of key
points is offered here. Two strands of positive analysis has argued for applying a lower discount rate to
longer-term climate change costs, including catastrophes, than might be inferred from research on
consumer time preference or rates of return on investment. One is that individuals may discount the
future hyperbolically, so rates of discount decline and ultimately plateau at a fairly low number as one
goes out into the future. The other is that when one accounts for the higher marginal utility of income for
the poor facing more adverse impacts from climate change, then under reasonable assumptions the
effective time discount rate after adjusting for distributional differences is reduced. In addition, if climate
change has the most severe effects on longer-term economic growth when growth itself is more likely to
be weak, then policies to reduce the threat of catastrophe will have a lower effective discount rate because

of their contribution to reducing intertemporal economic risk.

Even with these considerations, however, the resulting implied discounting of future over current returns
may not be small enough for catastrophes to carry major weight in evaluating the potential impacts of
climate change. Unless the discount rate is under 1%, and perhaps even close to zero, severe future
consequences that will not arrive for some time and are not world-threatening may still be too
"telescoped." Stern and others have addressed the issue of discounting by using normative arguments to
suggest a discount rate at or near zero is in fact appropriate. Two other arguments, not so dependent on
normative precepts, may also add weight to the importance of catastrophe risks in evaluating climate
change impacts.

Hypothesis 1: People are Not Expected Utility - Maximizers

There is a growing literature from behavioral economics and psychology which demonstrates that

individuals do not consistently make decisions according to the expected utility paradigm. If individuals
are only boundedly rational, they have neither the time nor the capacity to fully assess the consequences
of decisions. In that case, individuals adopt certain rules of thumb and mental shortcuts to make
decisions. These so-called heuristics can lead to choices that depart from predictions of expected utility
theory.

6 References - Kverndokk et al, Pizer, Nordhaus. Earlier foreshadowing by Manne.

7 [add references] Strictly speaking, the second and third arguments are not about the actual rate of time preference, but rather
about how factors related to distributional impacts and risk that enter the maximand of the intertemporal utility calculation
affect the implied discounting of future over current returns.

8 This discussion is taken from Kousky et al (2009), which contains references to the relevant behavioral economics literature.

-------
October 25, 2010 version

When thinking about possible disasters, it has been found that people tend to be over-optimistic, thinking
negative outcomes are less likely to happen to them. When a risk is highly emotional, however, people
can disregard probabilities altogether, treating all outcomes as equal ("probability neglect"). Individuals
also seem to place an added value on certainty, preferring to reduce a small risk to zero by more than they
value reducing a larger risk by a greater amount. Errors of commission are viewed as worse than errors of
omission. This can lead to a tilt to the side of inaction.

Experimental also has found that context matters, often significantly, a when making decisions. For
instance, when probabilities are unknown and must be estimated, individuals have been found to assess an
event as more likely when examples come to mind more easily (the "availability heuristic"). People can
disproportionately prefer to maintain the status quo in their choices, even if conditions or options change.
Individuals sometimes "anchor" their preferences on an available piece of information, and fail to update
their assessments adequately in the face of new information. Individual choices are also strongly affected
by the way that information is presented. Thus, individuals may make different choices for the same
decision if it is merely phrased differently ("framing effects"). Choices depend upon the extent to which a
risk evokes feelings of dread. Personal utility also is sensitive to individuals' perceptions of equity and
fairness.

These various behavioral attributes can imply higher or lower values attached to catastrophe risks than
would be implied by expected utility theory. The former would follow from dread or the evaluation of all
catastrophes as roughly equal in likelihood. The latter would follow from optimism bias, or a preference
for reducing small and familiar risks to zero over reducing more substantially an unfamiliar risk - of
which climate change catastrophe certainly is an example. While the direction of bias has to be assessed
empirically, the existence of these various "non-rational" attitudes raises an important but not new
question for evaluating climate change catastrophe risks in setting public policy: if decision makers
believe they have better information than the general public and that they are less subject to emotional
biases, to what extent should their valuation of alternatives supersede those of members of the general
public?

Hypothesis 2: People are Non-Egoistic Expected Utility - Maximizers

A second approach that has been taken in the literature for addressing long-term threats posed by climate
change is to see individuals today, imperfect information and all, as interested in more than maximizing
the discounted present value of their lifetime expected utility streams. One can broadly define this as
altruistic preferences, but this label can cover several different forms of preferences.

A traditional approach to altruistic preferences is to include some measure of next-generation or other
future utility in the preferences of members of the current generation. In this setting, individuals will
weigh the potential costs of a climate change catastrophe in terms of its anticipated impacts on future
welfare, as well as the possibly slight impact on current individuals' egoistic well-being. Consequently,
individuals will derive utility in part from the "bequest they leave to the future in terms of a lowered
(endogenous) risk of a climate change catastrophe. However, there are both theoretical and empirical
reasons to expect individuals to discount the welfare of future generations relative to their own egoistic
welfare. This takes us back to the question previously mentioned in the context of time preference, as to

-------
October 25, 2010 version

how powerful an influence this form of altruism might be in the current generation's assessment of risks
of climate change catastrophes.9

A second approach is to depart from a purely utilitarian framework by supposing that individuals see
themselves (or should do) as having a moral obligation to future generations. This mixing of obligations
and conventional utilitarian motivations implies some degree of lexicography in individuals' preferences
- or, critics of utilitarianism might say, an innate failure of the standard economic model to describe what
really motivates people. In this view, if a potential future catastrophe threatens to impose a morally
inacceptable burden on the future, people will be (or at least can be) motivated to endure potentially extra-
ordinary sacrifices to reduce the threat. The expression of that moral sentiment by individuals as citizens
and stewards, versus utilitarian consumers, would be found through public choice exercises like voting for
tough restrictions on future GHG emissions.

This conception is both stimulating and frustrating, since it does not offer any straightforward way of
assessing how economically significant is the threat of a future climate change catastrophe. Aside from
uncertainty about what the triggering level of threat to the future might be, does one regard current almost
universal reticence to support tough GHG restrictions as due to (correctable) moral failing? Lack of
information? Lack of leadership? The result of rational leadership, because the threat of climate change
is seen as less significant than other threats or because international collective action problems have not
been solved?

A third possible approach that has received less attention is that individuals have preferences that include
some notion of "planetary health" as a global public good. Rather than seek to describe concern about
risks of catastrophe from climate change as deriving only from more fundamental concerns for
intergenerational altruism or fairness, one could posit that individuals derive some direct benefit from
having greater confidence in the ability of planetary systems to remain undisrupted, without the need to
unpack the rationales in terms of future human well-being, satisfaction of moral sentiment, or a pure
existence benefit. This approach allows one to sidestep some of the difficulties encountered in either the
altruistic utilitarian or moral-obligations conceptions. In particular, the normative approach to setting
discount rates can be embedded in a framework of preferences without having to be an ad hoc add-on.1
However, this does not get around the huge empirical problems in assessing the value that members of the
current generation might place on reducing risks of future climate change catastrophes.11

CATASTROPHE RISKS AND RATIONAL CHOICE APPROACHES TO POLICY

While it is certainly possible to debate the capacity of expected -utility types of analyses to adequately
capture the social opportunity cost of climate change catastrophe threats, it is in cases like this that a
disciplined application of rational-choice based analysis more broadly defined can prove most useful. A

9 Current individuals also could believe, as Schelling for example has suggested, that other kinds of bequests to the future
would have higher value; or they could further discount bequests of a less risky climate out of concern that unless the "chain of
obligation" is maintained, something impossible to assure, the sacrifice made today would be wasted in the future.

10 A fundamental criticism of conventional expected-utility analysis for assessing future climate change risks is that it combines
conventional time-preference considerations in assessing the opportunity cost of reducing threats with the explicitly ethical
question of how much the current generation will feel willing or bound to do in protecting the future.

11 Ideas like this arise often in literature on environmental stewardship, but I am not aware of many treatments of the idea in
economic terms. One example is the paper by Kopp and Portney [ref to add], who describe a thought experiment in which
individuals value "well being of the future," and the willingness-to-pay for that value can be discerned through a stated
preference valuation effort. While one can debate the merits of the valuation approach even in a thought experiment, the
concept is very similar to what I am trying to describe here. Unfortunately, the question of how one would ascertain such
valuation remains a barrier to empirical implementation of the concept.

-------
October 25, 2010 version

thoughtful, systematic, and transparent weighing of benefits and costs, broadly defined, is at the heart of
such an approach. The presence of "deep" uncertainty or ignorance about the types and likelihoods of
potential catastrophes means that we must include, in addition to sensitivity analysis on these
characteristics, focused analysis of the robustness and flexibility of options in addition to the benefits and
costs. With respect to what seem to be behavioral biases in the assessment of catastrophe risks by
individuals, decision makers must make (and then defend) informed judgments on behalf of those they
serve as to when the seeming biases reflect a high degree of economic risk aversion, or dread, and when
the biases reflect other factors (framing effects, optimism bias, and the like) that can be viewed as
inaccurate comprehension of the tradeoffs involved.

Posner (2005) argues that uncertainty over benefits and costs should not prevent using the basic structure
of cost-benefit analysis for evaluating and comparing options, but that this should be framed in a
"tolerable-windows" approach. This involves using a range of plausible risk estimates to help identify
levels of spending on reducing risk for which benefits clearly exceed the costs, for which costs clearly

exceed benefits. Policies then can be designed with the goal to remain in this window. This approach
does not provide or depend on "a number" for how to evaluate the impacts of potential future climate
change catastrophes. In particular, it does not treat them as largely irrelevant economically given their
low probabilities and long time frames to be realized. Instead it provides flexibility as to how different
considerations about climate change catastrophes are brought into the assessment, including risk aversion
and concerns about future sustainability as well as costs of risk mitigation, while insisting on transparency
and a persuasive argument for how these considerations are to be addressed.

12 This idea is akin to value-of-information approaches. If one has some confidence in the evaluation of costs of different
policies but great uncertainty about the potential benefits, one could investigate how large the potential benefits might have to
be to make a case for the selection of one set of options over another in a portfolio. Similarly, if the benefits are reasonably
well understood conditional on a catastrophe occurring, but there is uncertainty about the probability of a catastrophe, then one
can ask how large the probability would have to be to justify a particular portfolio of actions.

-------
Social Cost of Carbon and Risks
of Climate Change Catastrophes

Mike Toman
World Bank Research Department *
EPA Climate Change Impacts Workshop

November 19. 2010

Views are the author's alone

-------
Interest in the Topic

• Concern that "tipping points" may be closer in time
and more serious than had been anticipated

— calls for rapid and deep cuts in GHG emissions

• Concern for the uncertain fate of international
negotiations

— mitigation may fall short

— adaptation may be under-financed

-------
Challenges in Addressing Topic

• Deep scientific uncertainties about catastrophe risks

• Questions about efficacy of different strategies for
mitigating CC risks

• Perception that standard rational choice methods
are inadequate for assessing risks, identifying policy
approaches

-------
Outline

Potential for Climate Catastrophes
Decision Frameworks
Analysis of Response Options
Implications

-------
Global CC Catastrophes

low probability events with large, global, irreversible impacts
that dramatically reduce long-term human well-being

(probability rises with g

Timely advance warning is uncertain

-------
Types of Catastrophes

• " Unfolding" Catastrophes:

— Sea level rise, ice sheet collapse

— Major increase in natural hazard risks

— Major ecosystem collapses (land, water)

— Shifting ocean currents

• "Cascading" Catastrophes:

— Relatively rapid succession of droughts, crop failures
widespread mitigation, conflicts

— Remain poorly understood

• Methane feedbacks, interactions among types of
catastrophes

-------
"Unfolding" Catastrophes

• Some likely to unfold only over long time periods
(many decades, centuries)

— Even if ice sheets collapse, consequences only develop
and intensify over time

• Ecosystem collapse could occur on much shorter
time scales (decades)

— Depends on unknown magnitude and speed of
temperature responses, other climatic changes

-------
"Unfolding" Catastrophes

• Physical tipping points uncertain and remain
challenging to detect in advance

• Relationship of socio-economic tipping points to
physical tipping points is even more uncertain

— Depends on speed of consequences

— Adaptation capacity

-------
"Cascading" Catastrophes

• Cumulative effect of sequence of more localized
CC-induced harms each reinforcing others

— Series of regional crop disruptions widespread
famine, land degradation, and conflict

— Series of localized extreme weather events larger-
scale economic disruptions, reduced remittances,
refugee problems, and conflict

• Mostly speculation at this point — little has been
done on such risks

-------
Literature on Global Catastrophe
Valuation - Very Limited

• Weitzman simulations; Nordhaus, Pindyck

• Growth theory models with uncertain arrival or
large GDP shock — Nordhaus, Pizer, Gjerde et al

• IAM work-FUND (sea level rise and cities, change
in thermohaline circulation); PAGE

• More has been done on sub-global extreme events:

— Nordhaus, Emanuel, Mendelsohn, FEEM — hurricanes
and other extreme weather events

— Episodically incurred costs are large in absolute terms;
relationship to income less clear

-------
Outline

Potential Climate Catastrophes
Decision Frameworks
Analysis of Response Options
Implications

-------
Standard Rational Choice Approaches

Integrated economy-climate models calculate
"optimal * (dynamic PV-maximizing) emissions paths

Catastrophes represented as large, permanent drop
in welfare with endogenous risk
— Risk rises with atmospheric
GI IG concentration

Approach assumes risks and
impacts can be characterized
quantitatively

-------
Implications of Standard Approaches

• "Optimal" near-term abatement increases with
magnitude of catastrophe risk; but,

• The effect generally is fairly small unless

— catastrophe is VERY large and fairly near-term relative to
discount rate used; Or

— discount rate is low

• Familiar positive and normative arguments for
various discounting approaches inconclusive

-------
Challenges to Standard Approaches

Risk vs. uncertainty vs. ignorance

— Probabilities and even possible states of the world remain
very poorly or largely unknown

"Fat tails" versus expected utility

— Deep uncertainty looms over standard CBA

— Expected utility does not adequately reflect risk
preferences

— Traditional risk management analytical tools have limited
effectiveness in this situation

-------
Issues Raised by
Behavioral Economics

•Risk assessments "anchored"
by particular frames of
reference

•Difficulty in interpreting
small probabilities

•Aversion to extremes or to
ambiguity

Implication is possibility of systematic assessment

errors by general public

-------
Implications for Catastrophe
Risk Assessment

• Assessment "biases" by public could imply more or
less, faster or slower action

— Normal technocratic view is provide more information

• How much can further research on catastrophes do
to reduce such biases?

— Considerable uncertainty on possibility of catastrophe
seems likely to persist for some time

-------
Implications for Catastrophe
Risk Assessment

• Improving knowledge remains useful; but,

• Sound policy decisions cannot simply be based on
what revealed public preferences; however,

• This is not an argument for decision makers to
abandon systematic comparison of gains and losses!

• Decision makers need to exercise their judgment as
agents of the general public in evaluations

— Political economy challenge: myopia, high personal
discount rates, risk aversion

-------
Outline

Potential Climate Catastrophes
Decision Frameworks
Analysis of Response Options
Implications

-------
Evaluation criteria

• Aim is a reasoned comparison of benefits and costs
(broadly defined)

• Given deep uncertainties and several dimensions of
public concerns, multiple criteria can be useful

— Certainly does not preclude economic metrics!

— Practical difficulties to quantify many risk characteristics
in a single common metric

— Use of several metrics can reflect complex risk attitudes

— Given tradeoffs will be made in political give and take,
evaluating multiple criteria adds information

o I 19

-------
Evaluation Criteria: Example

• Effectiveness in mitigating risk

— Several possible ways to quantify

• Cost of implementation

• Robustness — ability to be effective even with
surprises in evolution of climate change threats

• Flexibility — ability to modify response as
information about risks changes

-------
Illustrative Application

1. Drastic and rapid global emission reduction

2. Global-scale anticipatory adaptation to mitigate
prospective consequences of catastrophes

3. Putting particulates into upper atmosphere (form
of geo- engineering to reflect incoming radiation)

-------
Drastic and Rapid GHG Reduction

• Effective for "unfolding" and "cascading" catastrophes

• Costs would be very high unless/until there are
major technology advances for mitigation

• High need for international participation

— More difficult the higher are the costs

• Robust to surprises in nature of risks

— Unless (BIG) surprise is risks are low

• Inflexible — requires sustained commitment to
decarbonization

-------
Global-Scale Anticipatory Adaptation

• Purchase land for mass relocation and begin
preventative relocation

• Drastically limit development in ecosystems and
increase buffer areas to improve resilience

• Massive structural controls against sea-level rise

-------
Global-Scale Anticipatory Adaptation

• Effectiveness would vary with action

— Land acquisition for relocation could sharply limit
natural hazard risks

— Ecosystem protections would have positive impacts, but
magnitude hard to judge

— Structural barriers could be brittle, not performing well
for more severe impacts

— Large-scale adaptation could be particularly effective for
short-circuiting potential cascading catastrophes

-------
Global-Scale Anticipatory Adaptation

• Costs depend on action but could be very high

— Win-win disaster risk reduction policies, ecological
systems protection

• Costlier options have little flexibility

• Portfolio of actions needed to have robustness

— Hazards of sea level rise versus ecosystem collapse

-------
Particulates in Upper Atmosphere

• Successful implementation would be effective and
robust in blunting impacts of GHG accumulation

• Direct costs could well be less than drastic GHG
mitigation, but further R&D costs could be
considerable; but,

• Highly uncertain side effects could create very large
overall costs, non-robust solutions

• Significant RD&D costs needed to establish large
scale feasibility and some confidence in safety

-------
Particulates in Upper Atmosphere

• Could use flexibly, to complement GHG abatement
or responding to warning signs; but,

• This requires adequate capacity to detect risks of
looming catastrophe in time; and,

• Highly inflexible once deployed

• Significant international coordination needed to
deter unilateral use with strong negative spillovers

-------
Summary of Evaluations

Evaluation
Criteria

Drastic Global
GHG Reduction

Massive

Anticipatory

Adaptation

Particulate
Injection to Upper
Atmosphere

Effectiveness

High

Medium

Potentially High

Cost

High w/o major
innovation for
mitigation; Low
post-mitigation

Low (with high co-
benefits) to High
(very disruptive
changes)

Potentially Very
High

Robustness

High

Low (individual
measures) to
Medium (for
portfolios)

Potentially High for
dampening CC; Low
for side effects

Flexibility

Low

Extremely Low
(absent drastic
mitigation later)

-------
Summary of Evaluations

• Certainly potential for effectiveness, robustness

• All options have high cost unless there is massive
advance in low-carbon technology

— All the more if action needed more quickly

• All options have low flexibility once implemented

-------
Outline

Potential Climate Catastrophes
Decision Frameworks
Analysis of Response Options
Implications

-------
Implications for
Social Cost of Carbon

• Cost Benefit Analysis provides much important info
needed to assess expected GHG accumulation cost

— Need also to consider its variance, and its incidence

• Standard CBA provides considerably less help for
evaluating potential impacts of catastrophes and
economic value of mitigation measures

• But the principle of carefully weighing benefits and
costs remains valid; instead we need to consider
different approaches to this assessment

— Problematic nature of vague "precautionary principle"

-------
Implications for
Social Cost of Carbon

• Need to consider SCC vis-a-vis catastrophe risks in
terms of the willingness of public today to bear costs
in an effort to mitigate such risks

— Variety of motivations possible — but for this purpose the
magnitude is the most important to understand

— Willingness to bear costs is not fixed; strongly depends
on individual values, social norms, understanding

-------
Implications for
Social Cost of Carbon

• Willingness to bear costs for reducing prospect of
future catastrophes depends on many unknowns:

— Baseline hazards, public attitudes and values

— Innovation in GHG mitigation that lowers future cost of
rapid, deep emissions cuts

— Ability of large-scale anticipatory adaptation to lower
risks from extreme events

— Possibilities and risks associated with geo-engineering

-------
Thought Experiment for One
Approach to Catastrophe Mitigation

• Define a provisional long-term climate protection
goal (X ppm, or Y° C, or )

• Simulate backwards a set of feasible approach paths

• Evaluate implementation costs and other attributes
of different paths

— Dependence on certain technical advances

— Dependence on certain assumptions

-------
Thought Experiment for One
Approach to Catastrophe Mitigation

• Form expert judgments on alternatives:

— How large would long term risk reduction benefits have
to be to justify mitigation costs?

— How could mitigation costs be reduced by less ambitious
targets or more aggressive adaptation?

— What are the types as well as sizes of residual risks?

• Put the options into the public domain for debate

— Help public understand options and accept choices

— Public feedback helps decision makers refine their
judgments about what protection costs are acceptable

-------
Implications for strengthening
response options

• Uncertainties with all three options imply very high
value of information with larger R&D funding

— New options for drastic decarbonization

— Stronger options for large-scale adaptation

— More research on various types of geo-engineering to
clarify their risks before they are used unilaterally

• Investigation of nature and prospects for "cascading"
catastrophes is needed to evaluate their seriousness

-------
Implications for International
Assistance Measures

• Actions to reducing catastrophe risks need to be
approached at strategic level

— Carbon "shadow price" on a few fossil energy projects
will have minimal impacts

— Same with non-coordinated adaptation

• Priorities for sector — level responses need to be set
(energy, food, water, coastal zones, public safety...)

• Political economy of financing-related "carrots and
sticks" is very complex but needs to be addressed

-------
Implications for

International Cooperation

• Once conditions begin to deteriorate it might be
easier to get international cooperation; but,

• Greater developing country vulnerability may cause
developed countries to turn inward

• Reduction of "adaptation gap" is an urgent priority
with large co-benefits

-------
Thank you!

Comments welcome.

-------
International Cooperation

Experimental economics show people value fairness and cooperation
giving hope that international climate agreements can be successful

Yet consequences are asymmetrically distributed

— Impacts vary by region

— Different populations, among and within countries, will have highly varying
ability to cope with such outcomes.

— Poorer countries or those with closed economies are least capable of
adaptation, and will have to rely on the other countries to bear the risk.

— Migration and international trade may function to diversity risks, especially if
the effects of a catastrophe are geographically concentrated.

— Concerns about equality of outcomes affect social welfare functions

— Even if rich countries decisions agree to bear global costs of CC, it is unclear
how to square that policy decision with policies of foreign aid.

-------
Implications for International
Cooperation

• Prospects for major global actions are limited when
seen as costly, with distant/uncertain payoff

• Without cooperation in risk assessment as well as
implementation, benefits of careful weighing of
options can be negated by others' actions

-------
Nonmarket impacts

Michael Hanemann

UC Berkeley/Arizona State University

-------
Topics

• Spatial and temporal aggregation in
assessment of impacts understates
impacts.

• Extreme local events account for most of
non-catastrophic damages.

• Risk aversion should be accounted for.

• Impacts are multi-attribute. A univariate
utility function, treating consumption as
perfect substitute for environment,
understates damages.

-------
Two of the charge questions

Q: How is the value of non-market impacts
currently represented in lAMs?

A: They are not meaningfully represented in
current lAMs. But neither are many of the
market impacts.

Q: What are the key challenges of quantifying and
incorporating non-market impacts into lAMs?

A: The greatest challenge is not monetization. It is
measurement of the physical impacts. One
needs a disaggregated, bottom-up approach to
the assessment of non-market impacts - and
most market impacts, too.

-------
Damages in DICE 2002

ECONOMIC IMPACT OF 2.5° C WARMING: ANNUAL DAMAGES IN THE US
FROM NORDHAUS & BOYER (2002)

US TOTAL
$ 1990 billions

MARKET IMPACTS

Agriculture 4

Energy 0

Water 0

Sea Level 6

MARKET SUBTOTAL* 11

NONMARKET IMPACTS

Health, water quality, human life 2

Human amenity, recreation, nonmarket time -17

Ecosystem damages, species loss 0

Human settlements 6

Extreme and catastrophic events 25

NONMARKET SUBTOTAL* 17

MARKET + NONMARKET TOTAL* 28

* Totals do not add due to rounding.

-------
• Nordhaus & Boyer (2002) expressed as annual
willingness to pay per US household (2006$)

- Market impacts $126

- Non-climate catastrophe non-market impacts -$103

Subtotal $ 23

- Climate catastrophe non-market impacts $298

- Total $321

-------
What is missing?

-Averaging understates damages

- Neglect of extremes understates damages

- Assumption of symmetry of positive and
negative impacts understates net damages

- Neglect of tail dependence understates
damages

- Failure to allow for risk aversion understates
damages

- Ignoring distributional considerations & loss
aversion understates damages

-------
Climate impact studies

• California has been conducting impact
assessments since ~2000.

• Three rounds of assessment have been
completed (2002, 2006, 2009). Now on
fourth round.

• Key feature of this and other recent work
is spatial downscaling of GCM projections.

• Spatial downscaling has transformed
impact studies in last decade.

-------
GFDL CM2.1 precipitation mm/day

20C3Mrun2 1961-1990 SRESA2 run1 2070-2099 change

235' 240' 245' 250' 235' 240' 245' 250' 235' 240' 245' 250'

1961-1990 2070-2099 Change

Global Climate Models compute
COimate on a coarse grid

So, a "downscaling"
procedure was used
to provide temperature
and precipitation

over a finer mesh that
is more commensurate
with the California
landscape

-1

A hydrologic model is
used to simulate
streamflow, soil moisture
and other hydrologic
properties 8

-------
• Goal: "A more transparent representation
of the pathways through which climate
change may affect productivity and human
well-being."

• While mitigation is global, impacts and
adaptation - both market and non-market
- are local. They are spatially and
temporally heterogeneous.

• Without adequate representation of the
heterogeneity, there is neither a
transparent nor an accurate
characterization of impacts (damages).

-------
Aggregation distorts conception of

temperature change Hayhoe et al PNAS 2004

HOW TO CHARACTERIZE THE CHANGE IN TEMPERATURE, 2070-2099, USING HADCM3

EMISSION SCENARIO**
A1fi B1

Change

global average annual temperature

4.1

Change

statewide a\^rage annual temperature in California*

5.8

3.3

Change

statewide a\^rage winter temperature in California*

2.3

Change

statewide a\^rage summer temperature in California*

8.3

4.6

Change

LA/Sacramento average summer temperature

-10

*Change relative to 1990-1999. Units are °C

-------
• Spatial disaggregation is a major challenge for

economic analysis.

- CGE models are highly spatially aggregated.

• For given AT, yield effect differs by crop and

location:

- Impact on corn different than on wine grapes. Even
for grapes, impact different in Napa County vs Fresno
County.

- Can't represent impact via one "representative farm"

• Two neighboring water districts:

- Different water rights, different sources of supply,
different cost structures, different crops grown, &
different climate impact.

- Water isn't fungible. Can't represent a heterogeneous
area via a "representative farm" with a lumped,
regional supply of water, without distorting the
economic analysis.

-------
• Aggregation: Treat all days with a
temperature above 90°F as the same, as
opposed to, say, 90-94, 95-99, 100-105,

etc [e.g. Deschenes and Moretti (2007)]

• General Consequence:

-With convex damage function (increasing
marginal damage), aggregation understates
damages: E{D(AT)} > D(E{AT}).

-------
Asymmetric negative & positive impacts

• In some cases there can be positive as well as
negative impacts of climate change, depending
on the degree of change.

- Mild warming improves crop yield in cold climates,
extreme warming kills crops.

- Warming in winter reduces mortality, warming in
summer raises mortality.

- Warming in winter lowers energy bills for heating,
while warming in summer raises energy bills for air
conditioning.

• These effects are often represented by a
quadratic, hill-shaped impact function.

• In the DICE model, Nordhaus assumes these
positive and negative effects roughly cancel out.

-------
• However, the empirical evidence suggests
that the effect is generally not symmetric.

• Rather it is highly asymmetric

- e.g. effect of temperature on crop yield

- effect of temperature on energy use

• The empirical evidence suggests that, for
crop yields, energy use and weather-
related mortality in most countries, the
negative impacts of higher temperatures
greatly exceed the positive impacts of
higher temperatures.

-------
Asymmetric Relation of Temperature and Crop Yield
Schlenker & Roberts (PNAS. 2009}

iiCVljTjlV:,

_ ¦¦¦ jT X

¦ /" \

•+' — _ V

! ... "\:

\
\

\
1.

-'cint E^tirrjite

¦icT.Vo Corkle-'Ke Barcl

'ii'j-'c Corkle'xe Bard

¦lnj. T° :¦ :;:cnfic:ence Band

¦ i i i

5 M: IT 2-:: 2T 30 35 40
Terrp entire f Celsius;

-------
Modesto Hourly Load/Temp (Aufhammer)

T3
(0

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

Modesto ID Temperature - Load Spline

50 60 70 80

Temperature (Degree Fahrenheit)

-------
Nonlinear increase in flooding

• In winter storm, waves can be 5-6 ' higher than mean
sea level. Therefore can have flood damage before sea
reaches level of land.

• Scripps analysis based on an extreme wave: occurred 1
hour per year in San Francisco 1960-1980.

• By 2000, it was occurring 15-20 times per year.

• If the mean sea level at San Francisco rises by 20 cm
between 2000 and 2100, expected to occur about 150-
200 times per year.

• If it rises by 40 cm, an extreme hourly event would occur
about 1,500 times per year.

• If it rises by 60 cm, an extreme hourly event would occur
about 7,000 times per year.

• If it rises by 80 cm, an extreme hourly event would occur
about 20,000 times per year.

-------
• Most of the damages to agriculture from
climate change are associated with the
change in frequency of extreme events
rather than the change in average
temperature.

• This is probably true for many other types
of impact as well.

• Weitzman has emphasized the issue of fat
tails in context of updating a prior. There
are also physical reasons - thresholds -
why a fat tail may arise.

-------
Modeling strategy

• The importance of disaggregation and the non-
linearity of impacts has implications for the
modeling strategy.

- Need a US model as well as a global model

- Rather than a single, integrated model, need a
modular approach with a network of models

• GCM

• Spatial downscaling to areas within the US

• Suite of sectoral models/analyses at loc

• al level

• Aggregate to national level for US

• This is more feasible if aim is to calculate SCC,
rather than to determine optimal US emissions. 19

-------
Implication: wrong damage function?

• The special role of extreme events affects the
exponent in the damage function.

• Moreover, damages are represented as a function
of the increase in temperature. But, it is likely that
they are also an increasing (?convex) function of

- The trajectory of increase in temperature (e.g., the
increase measured in degree years).

- The speed of increase in temperature.

• This would significantly change the economically
optimal trajectory of emissions.

-------
Refraining climate change in terms

of risk

• Because the largest part of the damages
from climate change is likely to be
associated with extreme events, one
should think of climate policy in terms of
risk assessment and risk management.

• In assessing potential damages, there
needs to be an allowance for risk aversion.
This is largely absent in most of the
existing economic literature on climate.

-------
• The DICE model allows for risk aversion with respect to
collapse of the thermohaline circulation, but not with
regard to ordinary market and non-market losses.

• These are local impacts (fire, flooding, drought etc), but
the local population which is exposed to them is likely to
have some degree of risk aversion and some WTP to
lower their exposure to these risks.

• There are limits to the extent to which these risks can be
pooled

- Non-financial outcomes (pain and suffering, etc)

- Tail dependence

• Therefore, there should be some allowance for the
public's risk aversion premium to avoid these risks.

• Moreover, the relevant risk concept is likely to be
downside risk aversion.

-------
Downside risk

• This is a modification of the conventional theory
of risk aversion.

• It is based on the notion that there is some
asymmetry in risk attitudes towards outcomes.

• Downside outcomes (defined relative to some
point) are weighed more heavily than upside
outcomes.

• The concept was first applied in the financial
literature in the 1970s - going broke is viewed
differently than making a profit.

• It is likely to apply to many physical outcomes of
climate change - e.g., asymmetry between
having too little water and having too much.

-------
Example of downside risk analysis

(Hanemann et al. 2009)

• Under the downscaled projections from the
GDFL model (a medium-sensitivity GCM), but
not the PCM model (a low-sensitivity GCM),
there is a significant increase in downside risk
with respect to water deliveries for agriculture in
California's Central Valley.

• With downside risk aversion there is a significant
risk premium associated with that change.

-------
Annual deliveries to Central Valley

agriculture, 2085

7,QOO 8,000

9.CDC 10,000 11,000 12,000 13,000 14,000 15,000 16,000

Thousand Acre Feet

-------
Downside risk-adiusted impact

CENTRAL VALLEY AGRICULTURE
ANNUAL NET REVENUE 20S5
($ million)

MEAN

DOWNSIDE
RISK FACTOR

ADJUSTED
VALUE

BASELINE

S415

$132

$2S3

GFDL A2
GFDLB1

$314
$349

$ 17S
$163

$136
$186

PCM A2
PCM B1

$397
$413

$130
$12(5

$267
$287

LOSS COMPARED TO BASELINE

GFDL A2
GFDLB1

$101
S66

$4(5
$31

$147
$97

PCM A2
PCM B1

SIS

-$2
-$<5

$16
-S4

For GFDL, consideration of downside risk increases the estimate
of loss by about 50%.

For PCM, consideration of downside risk reduces the estimate of

-------
Multivariate utility

• Use of an aggregate consumption function
treating consumption as a perfect substitute for,
or a separable from, "the environment" (non-
market impacts) understates damages.

- Weitzman (2009) "Additive Damages"

- Sterner & Persson (2008) "A Sterner View"

- Carbone & Smith (2008) "Evaluating Policy
Interventions with General Equilibrium Externalities"

- Fisher & Krutilla (1975)

-------
Implications for Design and Benefit-Cost
Analysis of Emission Reduction Policies

Ray Kopp

Senior Fellow, Resources for the Future

Improving the Assessment and Valuation of
Climate Change Impacts for Policy and Regulatory Analysis

November 18 -19, 2010, Washington, D.C., Omni Shoreham Hotel

-------
My Task

• Title

¦ Implications for Design and Benefit-Cost
Analysis of Emission Reduction Policies

• Charge

¦ How can improved lAMs, aid in the design and
evaluation of domestic emission reduction
policies such as cap-and-trade or carbon taxes,
and inform negotiations of international
climate agreements?

-------
Frame of Reference

¦ Focus my remarks on three specific classes of
policymakers

• Legislative (domestic policy design)

• Foreign Policy (global policy design)

• Regulatory Agency (policy implementation)

¦ Emphasize policymaker's needs and how those
needs might be met with information from
lAM's

-------
Legislative

• Questions I have never been asked

-------
Legislative

• Questions I have never been asked

¦ What is the SCC, i.e., marginal damage of a ton
of greenhouse gas?

-------
Legislative

• Questions I have never been asked

¦ What is the SCC, i.e., marginal damage of a ton
of greenhouse gas?

¦ What is the benefit-cost ratio of, for example,
a $25/ton carbon price?

-------
Legislative

• Questions posed by Congressional
members and staff amenable to 1AM
analysis

-------
Legislative

• Questions posed by Congressional members and staff
amenable to 1AM analysis

¦ How will the impacts of climate change affect the world, the
country and my constituents (households and employers)?

-------
Legislative

• Questions posed by Congressional members and staff
amenable to 1AM analysis

¦ How will the impacts of climate change affect the world, the
country and my constituents (households and employers)?

¦ What is the worst that can happen?

-------
Legislative

• Questions posed by Congressional members and staff
amenable to 1AM analysis

¦ How will the impacts of climate change affect the world, the
country and my constituents (households and employers)?

¦ What is the worst that can happen?

¦ What can be done to help my constituents adapt to climate
change?

-------
Legislative

• Questions posed by Congressional members and staff
amenable to 1AM analysis

¦ How will the impacts of climate change affect the world, the
country and my constituents (households and employers)?

¦ What is the worst that can happen?

¦ What can be done to help my constituents adapt to climate
change?

¦ How will my constituents benefit from mitigation actions?

-------
Legislative

• Questions posed by Congressional members and staff
amenable to 1AM analysis

¦ How will the impacts of climate change affect the world, the
country and my constituents (households and employers)?

¦ What is the worst that can happen?

¦ What can be done to help my constituents adapt to climate
change?

¦ How will my constituents benefit from mitigation actions?

¦ How much will it cost to mitigate GHG emissions?

-------
Legislative

• Questions posed by Congressional members and staff
amenable to 1AM analysis

¦ How will the impacts of climate change affect the world, the
country and my constituents (households and employers)?

¦ What is the worst that can happen?

¦ What can be done to help my constituents adapt to climate
change?

¦ How will my constituents benefit from mitigation actions?

¦ How much will it cost to mitigate GHG emissions?

¦ How will those costs affect my constituents (distribution of the
costs)?

-------
Legislative

• Questions posed by Congressional members and staff
amenable to 1AM analysis

¦ How will the impacts of climate change affect the world, the
country and my constituents (households and employers)?

¦ What is the worst that can happen?

¦ What can be done to help my constituents adapt to climate
change?

¦ How will my constituents benefit from mitigation actions?

¦ How much will it cost to mitigate GHG emissions?

¦ How will those costs affect my constituents (distribution of the
costs)?

¦ How can we lower the cost?

-------
Legislative

• Questions posed by Congressional members and staff amenable to
1AM analysis

¦ How will the impacts of climate change affect the world, the country and
my constituents (households and employers)?

¦ What is the worst that can happen?

¦ What can be done to help my constituents adapt to climate change?

¦ How will my constituents benefit from mitigation actions?

¦ How much will it cost to mitigate GHG emissions?

¦ How will those costs affect my constituents (distribution of the costs)?

¦ How can we lower the cost?

¦ How much would my constituents be willing to pay to avoid these
damages?

###

-------
International Negotiators

UNFCCC, Major Economies Forum, G20

• Past and Current Areas of Interest

-------
International Negotiators

• Past and Current Areas of Interest

¦ Estimates of damage (e.g., the Stern Review) have a played role and refined
estimates might play a larger role in the future, but greater attention will be paid
if the damages come from well-defined sector/region specific damage functions

-------
International Negotiators

• Past and Current Areas of Interest

¦ Cost of mitigation has attracted more attention than damages and will likely
continue to do so

-------
International Negotiators

• Past and Current Areas of Interest

¦ Cost of mitigation has attracted more attention than damages and will likely
continue to do so

¦ Who bears the burden of cost has been more important than who bears the
damages

-------
International Negotiators

• Past and Current Areas of Interest

¦ Cost of mitigation has attracted more attention than damages and will likely
continue to do so

¦ Who bears the burden of cost has been more important than who bears the
damages

• New Questions

-------
International Negotiators

• Past and Current Areas of Interest

¦ Cost of mitigation has attracted more attention than damages and will likely
continue to do so

¦ Who bears the burden of cost has been more important than who bears the
damages

• New Questions

¦ How can we measure individual country levels of effort?

-------
International Negotiators

• Past and Current Areas of Interest

¦ Cost of mitigation has attracted more attention than damages and will likely
continue to do so

¦ Who bears the burden of cost has been more important than who bears the
damages

• New Questions

¦ How can we measure individual country levels of effort?

¦ How can we measure incremental cost?

-------
International Negotiators

• Past and Current Areas of Interest

¦ Cost of mitigation has attracted more attention than damages and will likely
continue to do so

¦ Who bears the burden of cost has been more important than who bears the
damages

• New Questions

¦ How can we measure individual country levels of effort?

¦ How can we measure incremental cost?

¦ How can we estimate realistic offset supply curves (e.g., REDD and sectoral
offsets) that address cost and timing?

-------
International Negotiators

• Past and Current Areas of Interest

¦ Cost of mitigation has attracted more attention than damages and will likely
continue to do so

¦ Who bears the burden of cost has been more important than who bears the
damages

• New Questions

¦ How can we measure individual country levels of effort?

¦ How can we measure incremental cost?

¦ How can we estimate realistic offset supply curves (e.g., REDD and sectoral
offsets) that address cost and timing?

¦ How would a global carbon market affect international trade and investment?

-------
International Negotiators

• Past and Current Areas of Interest

¦ Cost of mitigation has attracted more attention than damages and will likely
continue to do so

¦ Who bears the burden of cost has been more important than who bears the
damages

• New Questions

¦ How can we measure individual country levels of effort?

¦ How can we measure incremental cost?

¦ How can we estimate realistic offset supply curves (e.g., REDD and sectoral
offsets) that address cost and timing?

¦ How would a global carbon market affect international trade and investment?

¦ How would large-scale "green growth" policies affect trade and investment?

###

-------
Regulatory Agencies

• Requirements of Executive Orders seem to
be the sole reason the Interagency Working
Group developed the SCC estimate and
continues to refine the estimate.

-------
Regulatory Agencies

• Requirements of Executive Orders seem to
be sole reason the Interagency Working
Group developed the SCC estimate and
continues to refine the estimate.

• There may be roles for lAM's to play in
regulatory design other than RIAs, but the
role will be specific to the regulation in
question.

-------
Information Likely to be of Future Value

For Legislation and Foreign Policy

• Detail on the distribution and severity of
damages (by geography, demography and
economic sector)

• Characterization of adaptation potential to
lower damages

• Estimate of damage sensitivity to the speed
of climate change

-------
Improvements in SCC for RIA Purposes

"The economic valuation of an environmental improvement is the dollar value of the private
goods and services that individuals would be willing to trade for the improvement at prevailing
market prices."

EPA,Guidelines for Preparing Economic Analyses, 2008

-------
Improvements in SCC for RIA Purposes

EPA,Guidelines for Preparing Economic Analyses, 2008

• Value of future climate damages is measured
by the preferences of people living today

-------
Improvements in SCC for RIA Purposes

EPA,Guidelines for Preparing Economic Analyses, 2008

• Value of future climate damages is measured
by the preferences of people living today

• People living today will themselves not enjoy
the benefits of mitigation activity

-------
Improvements in SCC for RIA Purposes

EPA,Guidelines for Preparing Economic Analyses, 2008

• Value of future climate damages is measured
by the preferences of people living today

• People living today will themselves not enjoy
the benefits of mitigation activity

• Why then would people living today be willing
to pay anything to avoid climate damages?

###

-------
The Missing Element

"Non-use value is the value that individuals may attach to the mere knowledge of
the existence of a good or resource, as opposed to enjoying its direct use. It can
be motivated for a variety of reasons, including bequest values for future
generations/'

EPA guidelines for Preparing Economic Analyses, 2008

-------
The Missing Element

EPA guidelines for Preparing Economic Analyses, 2008

• WTP to prevent climate damages is the
classic case of intra and intergenerational
bequest value and altruism

-------
The Missing Element

EPA guidelines for Preparing Economic Analyses, 2008

• WTP to prevent climate damages is the
classic case of intra and intergenerational
bequest value and altruism

• Estimates of these values are wholly absent
from the SCC analysis. One wonders why

-------
Natural Capital and Intra- Generational Equity in Climate Change

Geoffrey Heal
Columbia University1

1 Introduction

There are two dimensions of equity that are relevant in an evaluation of the impact of
climate change - inter- and intra-generational. It is the former that has been most discussed in the
literature to date - all of the extensive debate about the choice of a discount rate in climate
models is in effect a debate about intergenerational equity and how to model our concerns about
this. And clearly this is very relevant in a climate context - emissions made today will affect
generations not yet born, so that issues of intergenerational fairness are central to any discussion
of climate policy. But intragenerational issues loom large too: climate change is an external cost
imposed largely by rich countries on poor ones, and in addition there is evidence that in any
given country it affects poor people more than rich. This dimension of climate change has not
been extensively discussed.

Climate change affects our stock of natural capital - for example, the IPCC has estimated
that by 2100 in the range of 30-40% of currently extant species may be driven extinct by climate-
induced changes in their ecosystems. This would represent a massive transformation of the
biospehere, one unprecedented in human history. Glaciers and snowfields are also likely to
diminish greatly in extent, affecting water supplies to many regions. Changes like this in our
natural capital could have far-reaching consequences, and these are likely to be felt more by poor
than by rich countries, and more by poor than rich groups in any country (World Bank 2006). So
intra-generational equity and natural capital impacts are related: the latter is likely to reinforce
concerns about the former. An important question here is whether some other form of capital -
human, intellectual or physical, can replace natural capital. To the extent that this is possible, it
may be possible to ameliorate some of the intra-generational equity impacts of climate change.

In the notes that follow, I begin to develop some of these points, making suggestions
about how they might be modeled.

-------
2 Equity and Discounting

As anyone who has spent even a short time on the economics of climate change must be
aware, a central issue is the choice of the pure rate of time preference (PRTP), to be
distinguished clearly from the consumption discount rate (CDR). The PRTP is the 8 in the

expression | u(ct)e~Stdt where ct is aggregate consumption at time t, u is a utility function

showing strictly diminishing returns to consumption and we are summing discounted utility over
all remaining time.

The other discount rate concept, the CDR, is the rate of change of the present value of the

e Stdu(c.)

marginal utility of consumption, that is, the rate of change of For the case of a single

dct

consumption good - and we will turn to the case of multiple goods later - it follows from well-
known arguments going back to Ramsey [1928] (see Heal [2005] for a review) that this is equal
to the PRTP plus the rate of change of consumption times the elasticity of the marginal utility of
consumption:

pt=S+ri(ct)R(ct) (1)

where pt is the consumption discount rate applied to consumption at time t, 77 ( ct)= — > 0

is the elasticity of the marginal utility of consumption and R(ct) is the rate of change of

,TT , du(c) , „ d ,.
consumption at time t. (Here u = —— and u = —u .)

dc dc

What do these two discount rates mean? The PRTP 8 is the rate at which we discount
the welfare of future people just because they are in the future: it is, if you like, the rate of
intergenerational discrimination. Note that there are at least two reasons why we may wish to
value increments of consumption going to different people differently: one is that they live at
different times, which is captured by 8, and the other is that they have different income levels,
which we discuss shortly.2 A PRTP greater than zero lets us value the utility of future people less

We could also value them differently for all manner of other reasons - differences in nationality, ethnicity, and
proximity either physically or genetically. In general we don't do these things, at least explicitly, which to me makes
it strange that we do explicitly discriminate by proximity in time.

-------
than that of present people, ///.sV because they live in the future rather than the present. They are
valued differently even if they have the same incomes. Doing this is making the same kind of
judgment as one would make if one valued the utility of people in Asia differently from that of
people in Africa, except that we are using different dimensions of the space-time continuum as
the basis for differentiation.

That an increment of consumption is less important to a rich person than to a poor person
has long been a staple of utilitarian arguments for income redistribution and progressive taxation
(see Sen [1973]), and is almost universally accepted. This is reflected in the diminishing
marginal utility of consumption, and the rate at which marginal utility falls as consumption rises
is captured by rj (c,). Equation 1 pulls together time preference and distributional judgments, or

considerations based on inter- and intra-generational judgments: the rate at which the value of
an increment of consumption changes over time, the CDR pt, equals the PRTP 8 plus the rate
at which the marginal utility of consumption is falling. This latter is the rate at which
consumption is increasing over time R(ct) times the elasticity of the marginal utility of

consumption rj (ct).

3 Equity and Climate Change

As we have just seen, there are two dimensions of equity that are important in the context
of climate change: equity between present and future generations, the aspect that has been most
extensively discussed, and equity between rich and poor countries or groups, both now and in the
future - inter- and intra-generational issues. This second dimension is invisible in aggregative
one-good models, which is one reason why we need a many-good model to talk seriously about
climate change. The discussions below will reinforce the need for some measure of
disaggregation in the analysis of the economics of climate change if we are to grapple with
equity issues.

The parameter 77, the elasticity of the marginal utility of consumption, summarizes our
preference for equality: it determines how fast marginal utility falls as income rises. There are
two ways in which this affects the case for action on climate change.

As T] rises, the marginal utility of consumption falls more rapidly. If consumption is
growing over time, then this means that the marginal utility of future generations falls more
rapidly with larger values of T] and therefore we are less concerned about benefits or costs to

-------
future generations. We are less future-oriented - the consumption discount rate p is higher - and
so place less value on stopping climate change. So via this mechanism, a stronger preference for
equality leads to a less aggressive position on the needfor action on climate change. Preferences
for equality and action on climate change are negatively linked here.

There is another offsetting effect, not visible in an aggregative model. Climate change is
an external effect imposed to a significant degree by rich countries on poor countries. The great
majority of the greenhouse gases currently in the atmosphere were put there by the rich
countries, and the biggest losers will be the poor countries - though the rich will certainly lose as
well. Because of this, a stronger preference for equality will make us more concerned to take
action to reduce climate change.

Unfortunately, without exception analytical models capture only the first of these effects.
They are aggregative one-sector models or models with no distributive weights and so their
operation does not reflect the second mechanism mentioned above. This explains the really
puzzling and counter-intuitive result that a greater preference for equality in Nordhaus's DICE
model leads to less concern about climate change.

To capture fully the contradictory impacts of preferences for equality on climate change
policy, we need a model that is disaggregated both by consumption goods and by consumers,
allowing us to study the consumption of environmental as well as non-environmental goods and
also the differential impacts of climate change on rich and poor nations.

3 Natural Capital and Climate Change

Return to equation (1) for the consumption discount rate. Note that if consumption were
falling rather than rising over time (the latter being the universal assumption in IAMs), then the
second term in the expression for pt would be negative and the CDR could in principle be
negative, that is the value of an increment of consumption could be rising over time rather than

-------
falling. We would not be discounting but doing the opposite, whatever that is. It is not
impossible that in a world of dramatic climate change and environmental degradation,
consumption might fall at some point. It is even more likely that some aspects of consumption, or
the consumption of some social groups, would fall while other continue to rise - recognizing this
requires that we treat consumption as a vector of different goods that can be affected differently
by climate change. For an early recognition of this point see Fisher and Krutilla [1975], who
comment that increasing scarcity of wilderness areas may drive up our valuation of them. A
more detailed analysis in the context of a growth model is in Gerlagh and van der Zwaan [2002],
who make the interesting point that with limited substitutability between environmental and
manufactured goods and the growing scarcity of environmental goods, there is likely to be a
version of Baumol's disease - an ever larger portion of income being spent on non-manufactured
goods.

Let's follow this line of thought and disaggregate consumption at date t into a vector
ct = (ci,,c2;,...c„,) of n different goods. (We will mention briefly later the case in which thsee

are the consumption levels of different countries or social groups.) Utility is increasing at a
diminishing rate in all of these goods and is a concave function overall. In this case we have to
change equation 1 for the consumption discount rate. Now there is a CDR for each type of
consumption and we have n equations like equation 1, with a CDR for each good i equal to the
PRTP plus the sum over all goods j of the elasticity of the marginal utility of consumption of
good i with respect to good j times the growth rate of consumption of good j :

where pit is the CDR on good i at date t, R (ci t) is the rate of change of consumption of good
i at date t, and 7]^ (ct) is the elasticity of the marginal utility of good i with respect to the
consumption of good j (see Heal [2005] for details: the most general framework of this type can
be found in Malinvaud's classic paper [1953]). The own elasticities such as Jjtt(ct) are positive
numbers, but the cross elasticities ij^. (c,), j i, are zero if the utility function is additively

separable and can otherwise have either sign.

As an illustration consider the constant elasticity of substitution utility function

(2)

-------
[acCT +(l-a)s,<7] 1 they are
substitutes and the cross elasticity is positive, and vice versa.

Let's test our intuitions on this. Take the case where natural capital and produced
consumption are highly complementary, so that indifference curves are near to right angled and
the elasticity a is close to zero. Then the cross elasticity is negative. This means that if the stock
of natural capital is rising then this reduces the consumption discount rate on the regular good.
Conversely if the availability of natural capital is falling then this raises the consumption
discount rate on the consumption good. These results make sense: because of the assumed
complementarity, an increase in the amount of the environmental good will raise the marginal
utility of the consumption good and so tend to lower the consumption discount rate, and vice
versa. Of course, the own elasticity on natural capital is positive so that if the availability of this
good is falling then this will tend to make its own consumption discount rate negative.

Whether produced goods and environmental services are substitutes or complements in
consumption is not an issue that has been discussed in the literature, as with the few exceptions
mentioned above people have worked with one-good models. There do however seem to be
reasons to suppose that complementarity is the better assumption, with a < 1. Dasgupta and Heal
[1979], following Berry Heal and Salamon [1978], suggest that in production there are
technological limits to the possibility of substituting produced goods for natural resources. In
particular we invoke the second law of thermodynamics (Berry and Salamon are
thermodynamicists) to suggest that if energy is one of the inputs to a production process, then
there is a lower bound to the isoquants on the energy axis. Similarly one can argue that certain
ecosystem services or products, such as water and food, are essential to survival and cannot be
replaced by produced goods. There are therefore lower bounds to indifference curves along these
axes, implying if the utility function is CES that <7 < 1.

-------
Consumption goods

Minimum level of services
from natural capital

Natural Capital

The figure illustrates this idea: it shows indifference curves for a two-argument utility
function, consumption of produced goods and of ecosystem services, as in equation 3 above.
There is a minimum level of ecosystem services needed for survival - think of this as water, air,
and basic foodstuffs, all of which are ultimately produced from natural capital. For low welfare
levels there is no substitutability between these and produced goods, so that indifference curves
are close to right angled. At higher welfare levels where there are abundant amounts of both
goods there is more scope for substitution. Taken literally, this implies that the elasticity of
substitution is not constant but depends on and increases with welfare levels. This of course is
not reflected in the CES function such as 3. A function with these properties is

which is simply the CES function we noted before, with the zero of the ecosystem service axis
transformed by £ > 0. Utility is not defined for s <£. Relative to the transformed origin (e,0)

there is still a constant elasticity of substitution a but relative to (0,0) the elasticity is not
constant. For 1, every indifference curve, every welfare level, can be attained with only e of
ecosystem services, whereas with
-------
ecosystem services (and of consumption goods).

These ideas can be applied to modeling equity: it is generally recognized that poor
countries, or poor groups within countries, are more dependent on natural capital and its services
than are richer groups (World Bank [2006]). They have less capacity to substitute alternative
goods for the services of natural capital and so show more complementarity between natural
capital and other goods. In terms of the figure, their indifference curves are lower and closer to
being right angled. This means that they have different consumption discount rates from other
groups: if the stock of natural capital is falling then they will have higher consumption discount
rates on the common consumption good. In this sense they will appear to be more impatient. Of
course as noted above their discount rate on natural capital will be negative, so we will have the
paradox of an apparently impatient group - with respect to the consumption good - being willing
to invest for low returns in natural capital.

4 A Sterner Perspective

It's worth looking in more detail at the Sterner and Persson development of this point
[2007], They talk about the effect of changes in relative prices rather than consumption of
produced and environmental goods, but the point is the same. If we consume both produced
goods and the services of the environment, as in the utility function 3, then we can expect that
with climate change environmental services will become scarce relative to produced goods and
therefore their price will rise relative to that of produced goods (the " environmental Baumol
disease" that Gerlagh and van der Zwaan refer to [2002]). Consequently the present value of an
increment of environmental services may be rising over time, and the consumption discount rate
on environmental services may thus be negative, precisely the point that we were making in
equation 2 above. This could be the case even with a high PRTP, which is the main point of the
Sterner and Persson paper. They also present an interesting modification of Nordhaus's DICE
model to incorporate this point. They replace the standard utility function, which is an isoelastic
function of aggregate consumption, by a CES function along the lines of equation 3 above, but
modified to reflect a constant relative risk aversion:

[(1 - ry~Va + rsl~Va / (1 _ a)

They assume that the supply of environmental services s is negatively affected by temperature
according to the square of temperature, and that the share of environmental goods in

-------
consumption is about 20%, use these assumptions to calibrate the modified DICE model and and
then run the model with the PRTP used by Nordhaus. Their runs show that even with such a high
PRTP the presence of an environmental stock that is damaged by higher temperatures radically
transforms the optimal emissions path of C02 and leads to a vastly more conservative policy

towards climate change, with emissions both staying lower and falling faster. In fact it leads to a
more aggressive reduction in greenhouse gases than recommended by the Stern Review.

5 Natural Capital and Production

I have emphasized so far that natural capital can affect human welfare directly, and needs
to be thought of as an argument of the welfare function. Natural capital also affects a nation's
production possibilities: I mentioned above changes in hydrology such as melting of glaciers and
reduction in winter snowfields, both of which are already in evidence and are affecting
agriculture in some regions. They will affect it further over the coming decades. This is quite
separate from any impact that changes in temperature and precipitation may have on agriculture.
Other changes in natural capital will probably affect agriculture - changes in species abundance
and distribution, for example, can affect whether birds and insects pollinate crops.

6 Modeling Different Groups

I commented above that equation 2 can be given a different interpretation: instead of

the subscripts i and j referring to different goods, they can be taken as referring to the amounts of
a single good consumed by different groups - these could be social groups within a country or
they could be different countries. I this case we hae different consumption discount rates for each
group's consumption, and the elasticities now indicate how the marginal valuation of
consumption by one group depends on the cosumption levels of others. Do we value on
increment of consumption to the poor more if everyone else is very rich than if most others are
also poor? Presumably the answer to this is yes, but these are issues that have not featured at all
in the discussions to date.

7 Choosing rj

(2)

-------
The elasticity of the marginal utility of consumption plays a central role in much of outr
discussion. Unfortunately this variable plays two roles in our models: it expresses our
distributional preferences, which is the way we have been using it here, and it also expresses our
aversion to risk. Most empirical estimates of the value of Tj come from studies of behavior in the
face of risk, but it seems clear that these two interpretations of Tj are really quite different, and
that our aversion to risk tells us little if anything about our preferences for income equality.
Given this, we need to find a way of expressing preferences that does not conflate distributional
and risk preferences. Recursive formulations such as that of Kreps and Porteus are relevant here.

References

Barbier, Edward and Geoffrey Heal 2006. " Valuing Ecosystem Services," The
Economists' Voice, Berkeley Press, January 2006.

Berry, Stephen, Geoffrey Heal and Peter Salamon 1978. " On a Relationship between
Economic and Thermodynamic Optima," Resources & Energy, vol. 1, pp. 125-137

Dasgupta, Partha and Geoffrey Heal 1979. Economic Theory and Exhaustible
Resources, Cambridge University Press.

Fisher, Anthony and John Krutilla 1975. " Resource Conservation, Environmental
Preservation and the Rate of Discount," Quarterly Journal of Economics Vol. 89 No. 3 August
1975, 358-370.

Gerlagh, Reyer and Robert van derZwaan, 2002, " Long-Term Substitutability between
the Environment and Man-Made Goods," Journal of Environmental Economics and
Management, 44: 329-45.

Guesnerie, Roger 2004. " Calcul economique et development durable," La Revue
Economique.

Heal, Geoffrey 2005. " Intertemporal Welfare Economics and the Environment,"
Handbook of Environmental Economics, Volume 3.Edited by K-GMaler and J R. Vincent,
Elsevier, Chapter 21, 1105-1145.

Kreps David and Evan Porteus 1978 "Termporal resolution of uncertainty and dynamic
choice theory" Econometrica 46(1) 185-200

-------
Malinvaud, Edmond 1953. " Capital accumulation and the efficient allocation of
resources" Econometrica Vol 21 No. 2 April 1953

Nordhaus, William 1993. " Rolling the DICE: An Optimal Transition for Controlling the
Emission of Greenhouse Gases," Resource and Energy Economics, 15: 27-50.

Ramsey, Frank 1928. " A mathematical theory of saving," Economic Journal, 38: 543-

559.

Sen, Amartya 1973. On Economic Inequality, Clarendon Press, Oxford.

Sterner, Thomas and Martin Persson 2007. " An Even Sterner Review: Introducing
Relative Prices into the Discounting Debate,", Discussion Paper, Resources for the Future, July
2007, RFF DP 07-37. Available at http://www.rff.org/rff/Documents/RFF-DP-07-37.pdf

World Bank 2006. Where is the Wealth of Nations? Measuring Capital in the 21st
Century. The World Bank, Washington D.C.

-------
Natural Capital, Equity and
Climate Change

Geoffrey Heal
Columbia Business Scho

-------
Equity

• Two dimensions

— Inter- and Intra-Generational

• Inter-generational equity bound up with pure
rate of time preference delta

• Both affected by elasticity of MU, eta

• We express equity judgments of both types
when we choose delta and eta

-------
Equity

• Famous Ramsey equation for consumption
discount rate ties together both:

P, = S+f](c,)R(c,)

• CDR depends on intergenerational equity
values via delta and intragenerational via eta

-------
Equity

• As eta rises, MU of cons'n falls faster. If cons'n
grows then MU of future generations falls
more rapidly

• Less concerned about benefits to future.

• Consumption discount rate is higher- place
less value on stopping climate change. So a
stronger preference for equality leads to less
action on climate change.

-------
Equity

• Offsetting effect, not visible in aggregative model

• Climate change an external effect imposed by rich
countries on poor.

- greenhouse gases currently in atmosphere were put
there by the rich countries,

- and the biggest losers will be the poor countries

• Because of this, a stronger preference for equality
will make us more concerned to take action on
climate change.

-------
Natural Capital

• Affects well-being in many ways, depending
on stage of development

• Poor countries heavily dependent on services
of natural capital

• Natural capital compromised by climate
change

-------
Natural Capital

Ct ~ (p\,t>C2,f>'"Cn,t^)

• Ramsey equation is now

Pi,t ~ & + a (ct (cj t )+ (ft ifj,t )

• CDR is good-specific and can be + or -

-------
Consumption goods

Minimum level of services
from natural capital

Natural Capital

ac° + (1 -a)

-------
Natural Capital

• For a >1, every indifference curve, every
welfare level, can be attained with only £of
ecosystem services, whereas with 1 greater
welfare levels require greater levels of
ecosystem services (and of consumption
goods).

-------
Sterner and Persson

l~/i \ |—i/
-------
Intra-generational Equity

Pi., =5+ %, (c, )R (c,7 )+ IX (CI )R (C 1,1 )

j*i

Can take subscripts here to be social groups
not goods

-------
Role of Eta

Plays several roles

— Affects intergenerational choices via Ramsey
equn, with larger value making for less concern
CC

— Affects intragenerational choices directly, with
larger values making for more concern for CC

— Affects risk aversion

Really need to find a formulation that
separates these roles

-------
Disaggregation

• Need models that distinguish environmental
services from manufactured goods, and

• Need models that distinguish rich groups from
poor

• Two dimensions of disaggregation

-------
ENVIRONMENTAL DEFENSE FUND

Implications for choice of policy
targets for cost-effectiveness analysis

Nat Keohane
Chief Economist

EPA-DOE Climate Damages Workshop
November 19, 2010

EDF

ENVIRONMENTAL
DEFENSE FUND

Finding the ways that work

-------
ENVIRONMENTAL DEFENSE FUND

Agenda

1. The SCC is not a cost-effectiveness measure

2. What would a c/e approach look like?

3. What should we do with the SCC we have?

- Uses and abuses of the SCC

- Extramural uses of the SCC

4. The economics-science disconnect

5. Where do we go from here?

-------
ENVIRONMENTAL DEFENSE FUND

1. The SCC is not a cost-effectiveness
measure (1/2)

Importance of precision

• "Social cost of carbon" is not a generic term

Specific meaning: present value of the marginal
damage from emitting an additional ton of GHG

• SCC doesn't incorporate the cost of achieving a
goal (-> defn of cost-effectiveness)

-------
ENVIRONMENTAL DEFENSE FUND

1. The SCC is not a cost-effectiveness
measure (2/2)

So what is meant by "cost-effectiveness" here?

1. Contrast with optimal control approach
- SCC computed along BAU trajectory

2. "Letter" vs. "spirit" of cost-effectiveness

- Use in establishing consistency

- Derivation vs. application

Consider derivation first, then application.

-------
ENVIRONMENTAL DEFENSE FUND

2. What would a cost-effectiveness approach
look like? Key issues (1/4)

Considerations for cost-effectiveness analysis:

• What target to use? ("Effectiveness" at what?)

• What other countries do matters.

• Cost estimates aren't perfect either.

-------
ENVIRONMENTAL DEFENSE FUND

2. What would a cost-effectiveness approach
look like? The UK approach (2/4)

UK uses a cost-based shadow price measure

UK experience is instructive:

• National policy target in place

• Participation in the EU ETS cap and trade
program

- Creates a policy need for a c/e approach (trading
and nontrading sectors)

- Observable signal of marginal cost (thus not
entirely model-dependent)

-------
ENVIRONMENTAL DEFENSE FUND

2. What would a cost-effectiveness approach
look like? Some concrete ideas (3/4)

Some concrete ideas:

• Cost-based

- Shadow prices to achieve a "standard set" of
global scenarios (e.g., 450/550/650)

- ... to achieve a range of national targets (17%?)

• Risk-based

- Risk management framework (defer to Roger)

- Directly value the shift in the distribution [*]

-------
ENVIRONMENTAL DEFENSE FUND

2. What would a cost-effectiveness approach
look like? Conclusions (4/4)

Common thread: Marginal analysis

These are not mutually exclusive, either with each
other or with a damages-based SCC approach!

Some number better than no number, but several
numbers may be better than "some number"
(depending on use)

Premise of rest of talk: damages-based SCC has a
role, but what should it be?

-------
ENVIRONMENTAL DEFENSE FUND

3. What should we do with the SCC we have?
Uses and abuses of the SCC (1/3)

Abuses Uses

-------
ENVIRONMENTAL DEFENSE FUND

3. What should we do with the SCC we have?
Extramural uses of the SCC (2/3)

Interagency Working Group SCC has been used in
other unrelated proceedings:

• Colorado PUC proceedings

• DC Court of Appeals cases re: EPA GHG
regulations

• Cape Wind

-------
ENVIRONMENTAL DEFENSE FUND

3. What should we do with the SCC we have?
Extramural uses of the SCC (3/3)

Lessons from the "extramural" uses:

• Numbers have a life of their own

• SCC provides a valuable and concrete benchmark
for uses outside federal rulemaking

• Establishes the principle that marginal damages
are real and can be quantified

• $21/ton » $0/ton

What are the lessons (e.g., conveying uncertainty)?

-------
ENVIRONMENTAL DEFENSE FUND

4. The economics-science disconnect

Ex post approach

Ex ante approach

This value of the SCC
doesn't match the
science"

"This input [parameter
value, assumption]
doesn't match the
science"

Advantages:

• analytic rigor

• strong foundation

Requires something of both economists and scientists. 12

-------
ENVIRONMENTAL DEFENSE FUND

5. Where do we go from here?

• How will the results of this workshop be
incorporated into a process going forward?

-------
ENVIRONMENTAL DEFENSE FUND

An aside: Which damage function? (1/2)

-------
ENVIRONMENTAL DEFENSE FUND

An aside: Which damage function? (2/2)

100,000 T rials F req uency View

20.00 40.00 60

3.200
2, BOO
2.400

~T|

2.000 CD

-9

1,600 as
Z3

1200
800
400
0

Overlay Chart 1

.00 80.00 1 00.00 120.00 1 40.00 1 60.00

— SCC SQ MiniCAM 3% SCC CUBIC MiniCAMi

(Mean, Median, 95th %ile): ($30,28,59) ($56,46,136)

-------
Abstract

Managing Climate Risks

Roger M. Cooke1
Carolyn Kousky2

Many Integrated Assessment Models (IAMs) maximize the present value of
consumption, equating the marginal benefits of abatement in terms of reduced climate damages
with the marginal costs of reducing emissions. Every trader, banker, and investor knows that
maximizing expected gain entails a trade-off with risk. According to the theory of rational
decision, preferences can always be represented as expected utility, hence from this viewpoint,
any aversion to risk could be folded into the rational agent's utility function. This theory, recall,
applies to rational individuals; groups of rational individuals do not comply the axioms of
rational decision theory. The fact is that 'professional risk taking organizations' do manage risk,
and not by bending the utility function of a representative consumer. Rather, they employ
techniques like value at risk, and optimize expected gain under a risk constraint. Managing risk
is a problem of group decision.

Weitzman (2009) has recently called attention to the risks of climate change, arguing that
current approaches court probabilities on the order of 0.05-0.01 of consequences that would
render life as we know it on the planet impossible. What is the plan to manage this "tail risk"?
Risk management shifts the research question from 'how does the optimal abatement level
change for different parameter values?' to 'how does our policy choice fare under the range of
potential future conditions and how can we buy down the risk of catastrophic outcomes?' As
such, it places the quantification of uncertainty in the foreground. Uncertainty quantification is
more than a modeler putting distributions on his/her model's parameters. The antecedent
question reads: 'is it the right model? What is the model uncertainty?' Failing a definitive answer
to that question, stress testing our current models for their ability to handle tail risks, and
exploring canonical model variations are essential steps prior to quantifying uncertainty on
parameters. Gone are the days when quantification of the uncertainties was left to the modelers
themselves; at the state of the art, quantification is done by structured expert judgment in a
rigorous and transparent manner.

Stress Testing

Stress testing is preformed to check that models remain realistic and capture the relevant
possibilities when their parameters are given extreme values. Many IAMs specify economic
damages as a function of temperature change, and model their impact on output and utility. For
example, damages at time t induced by temperature change T(t) from pre-industrial mean
temperature are represented in DICE as factor that reduces economic output: 1/[1 +
0.0028388T(t)2]. The standard Cobb Douglas production function expresses output as a function
of total factor productivity, capital stock and labor. Capital depreciates at rate 10%, and is
augmented by savings (in the DICE "Base" case the savings rate is optimized with damages set
equal to zero, then damages are reinstated). Temperature induced damages and abatement efforts
reduce output. Setting damage and abatement equal to zero, an illustrative stress test of the Cobb
Douglass model with constant population, constant total factor productivity and DICE values for

1 Resources for the Future and Dept of Mathematics, Delft University of Technology

2 Resources for the Future

-------
other parameters is shown in Figure 1. Four output trajectories with initial capital ranging from
10 times the DICE value ($1800 Trill) to $100 ($1.6xl0"8 for each inhabitant). The limiting
capital value is independent of the starting values - with a vengeance: the four trajectories are
effectively identical after 60 years. Such obviously unrealistic consequences underscore the need
for circumscribing the empirical domain of application of these simple models. Put the factories
and laborers on the Moon and they will produce nothing; other things are involved. Regardless
whether the model adequately describes small departures from an equilibrium state, its use for
long term projections inevitably entails this sort of behavior and putting uncertainty distributions
on the model's parameters will not change that.

160.0000

140.0000

0.0003

2005 2325 2045 2055

2105 2125 2145 2165 2185 2205

Figure 1. Output gross of abatement cost and climate damage (Strill 2000 USD) Base case, no temperature
damage, no abatement, constant population, constant total factor productivity (0.0307951), initial output
from production function and DICE defaults for other parameters (DICE 2009 XL version).

A second stress test examines the effect of adding temperature induced economic
damages, again without abatement. With $180 Trill initial capital, we assume that temperature
increases linearly, leaving other parameters as in the previous case. Figure 2 shows four
economic output trajectories, corresponding to temperature increases of 0, 5, 10, and 15 degrees

Figure 20utput after damages before abatement, initial capital = 180 Strill, constant population, constant
productivity, no abatement, temperature in 200 yr (linear increments)

No scientist claims that life as we know it could exist with 10°C global warming. With a
steady temperature rise leading to 10°C above pre-industrial levels in 200 years, this model

-------
predicts that output would be reduced to 68.% of its value without temperature rise. Such
projections seem a bit sanguine. The essential feature is that climate induced damages hit only
economic output; as a result capital can never decrease faster than its natural depreciation rate,
and this rate of decrement is reached only for infinite temperature. Again, putting uncertainty on
other model parameters may cloud this picture, but will not change this feature.

Canonical Model Variation

It is often noted that simple models like the above cannot explain large differences
across time and geography between different economies, pointing to the fact that economic
output depends on many factors not present in such simple models. To "save the phenomena"
researchers have proposed enhancing the basic model with inter alia social infrastructure,
government spending, human capital, knowledge accretion, predation and protection, extortion
and expropriation (see Romer (2006), chapter 3). Before proliferating this model, however, it is
well to reflect on its fundamental assumptions about damage, capital and output. Could different
model types with comparable prime facie plausibility result in macroscopically different
behavior?

We illustrate with one variation based on the following simple idea: Gross World
Production (GWP[trillion USD 2005] ) produces pollution in the form of greenhouse gases;
pollution, if unchecked, will ultimately destroy necessary conditions for production. This simple
observation suggests that Lotka Volterra type models might provide a perspective which an
uncertainty analysis ought not rule out. The quantity of anthropogenic greenhouse gases in the
atmosphere at year t, GHG(t) [ppm CO2], is the amount in the previous year, less what has
decayed at a rate, say, 0.0083, plus any new emissions in time period t. Assume that new
emissions are a fixed fraction, say, 0.024 of GWP (Kelly and Kohlstadt 2001). Different values
can be found in the literature, but these are representative. Real GWP has grown at an annual
rate of 3% over the last 48 years (this includes population growth); assume that this growth is
decreased by a damage function D of temperature T, and ultimately of GHG, this gives the
following system:

If D were linear in GHG, this would be a simple Lotka Volterra type system. With cs as the
climate sensitivity and 280 ppm the pre-industrial level of greenhouse gases, equilibrium
temperature follows T(GHG(t)) = cs/ ln(GHG(t)/280)/ln(2). Adopting Weitzman's (2010)
notion of a "death temperature" of 18°C we write damages as D(GHG)(t) = (T/18)2.
Anthropogenic greenhouse gases increase with production; if GWP(t) were constant, they would
increase to a constant 0.024xGWP/0.0083 However, as GWP increases, GHGs and temperature
keep rising as well, lowering the growth rate of GWP. When D > 0.03, GWP starts decreasing.
Eventually 0.024xGWP < 0.0083, and then greenhouse gases start decreasing, reducing damages
to a point where production can start growing again. Figure 2 shows GWP and GHG as
functions of time out to 500 yrs, with all variables at their nominal values. GWP collapses.
Greenhouse gases also collapse, but not to their initial level; hence the next upswing in GWP is
attenuated. A steady state is eventually reached after some 1,500 years. This is not offered as a
plausible model, its role is to spotlight the fundamental modeling assumptions. Evidently,

(1)

(2)

GHG(t+l) = (l-0.0083)GHG(t) + 0.024xGWP(t).
GWP(t+l) =[1+0.03- D(T(GHG(t)))]GWP(t).

-------
different ways of modeling the impact of climate change damages give qualitatively different
predictions, and steady state values may not be relevant for current policy choices. Neither
theoretical nor empirical evidence exclude the Lotka Volterra type of interaction between
damages and production presented here. A credible uncertainty analysis should fold in this and
other possibilities, which brings us to the next point of examining a range of future conditions for
a given policy choice.

Figure 3: The impact of climate damages on GWP (left) and greenhouse gases (right)

1E0002 1.5E0002 2E0002

gwp

4.7E0002

4.4E0002
4.1E0002
3.7E0002
3.4E0002

3E0002 y—^ ^^^

2.5E00023E0002\ 3.5E0002 4E0002 4.5E0002 5E0002

ghg

6.9E0002

2.7E0002
2.4E0002
2E0002
1.7E0002
1.3E0002

5.7E0002/

^ \ 5.4E0002

1E0002 1.5E0002 2E0002 2.5E00023E0002 3.5E0002 4E0002 4.5E0002 5E0002

5.1E0002
" 4.7E0002
" 4.4E0002
" 4.1E0002
3.8E0002

Structured Expert Judgment for Quantifying Uncertainties

Uncertainty analysis with climate models must be informed by the broad community of
climate experts - not simply the intuitions or proclivities of modelers - through a process of
structured expert judgment. Experience teaches that independent experts will not necessarily
buy into the models whose parameter uncertainties they are asked to quantify. Hence, experts
must be queried about observable phenomena, results of thought-experiments if you will, and
their uncertainty over these phenomena must be 'pulled back' onto the parameters of the model
in question. This process is analogous to the process by which model parameters would be
estimated from data, if there were data. The new wrinkle is that data are replaced by experts'
uncertainty distributions on the results of possible, but not actual, measurements. The 'pull back'
process is called probabilistic inversion, and has been developed and applied extensively in
uncertainty analysis over the last two decades (see Cooke and Kelly 2010 and references
therein). In general, an exact probabilistic inverse does not exist, and the degree to which a
model enables a good approximation to the original distributions on observables forms an
important aspect of model evaluation. Four features of the structured expert judgment approach
deserve mention: (i) Experts are regarded as statistical hypotheses, and their statistical likelihood
and informativeness are assessed by their performance on calibration questions from their field
whose true values are known post hoc. (ii) Experts' ability to give statistically accurate and
informative assessments is found to vary considerably, (iii) Experts' uncertainty assessments are
combined using performance based weights, (iv) Dependence, either assessed directly by experts
or induced by the probabilistic inversion operation, is a significant feature of an uncertainty
analysis.

When uncertainty has been quantified in a traceable and defensible manner, an ensemble
of possible futures for each policy choice may be generated. Figure 4 shows 30 Lotka Volterra
temperature trajectories out to 200 years, with BAU emissions at 2.4% GWP (left) and stringent

-------
emissions at 1.5% of GWP (right); and using representative distributions for uncertain variables.
Employing a value at risk management strategy, we would search for an emissions path
optimizing consumption while holding the probability of exceeding a stipulated temperature
threshold below a tolerable threshold.

Figure 4: Possible temperature trajectories under (left) emissions at 2.4%GWP and (right)
emissions at 1.8% GWP (right)

0.69 0,11

These reflections challenge us to deploy risk management strategies on a global scale.
We suggest this begin with (i) stress testing models, (ii) exploring alternative models, and (iii)
quantifying uncertainty in such models via structured expert judgment. We are condemned to
choose a climate policy without knowing all the relevant parameters, but we are not condemned
to ignore the downside risks of our choices.

References

Cooke, R. M. and G. N. Kelly (2010). Climate Change Uncertainty Quantification: Lessons
Learned from the Joint EU-USNRC Project on Uncertainty Analysis of Probabilistic
Accident Consequence Code. Resources for the Future Discussion Paper 10-29.
Washington, D.C., Resources for the Future,

Kelly, D.L. and Kohlstadt C.D. (2001) "Malthus and climate chantge: betting on a stable

population" J. of Environmental Economics and Management 41, 135-161.

Romer, D. (2006) Advanced Macroeconomics McGraw Hill Irwin, Boston.

Weitzman, M. (2010). GHG Targets as Insurance Against Catastrophe Climate Damages.

Cambridge, MA, Harvard University.

Weitzman, M. L. (2009). "On Modeling and Interpreting the Economics of Catastrophic Climate
Change." Review of Economics and Statistics 91(1): 1-19.

-------
Managing Climate Risks

Roger M. Cooke1'2, Carolyn Kousky1
Resources for the Future
2Dept Math. TU Delft
18-11-10

-------
Key Points

• Why Risk Management?

• Total Uncertainty = Model + Parameter

• Climate Damage: Inner and Outer Measure

-------
Rational Decision Theory

• A rational agent maximizes expected utility

• (subjective) probability and utility unique
individual

• Climate change is a group decision problem

• Professional risk takers don't manage risk by
bending the utility function of a
'representative consumer'

• Probabilistic Design: optimize performance
under risk constraint

-------
Risk Management Approach

~ Risk-averse representative consumer

And / Or

Discounting
Utility function
Utility of civilization

~ Risk-constrained optimization

Capture total uncertainty

Choose probability constraints for set of DAI's

Find efficient ways to satisfy constraints

-------
Pricing Carbon at the Margin

rAssume values of ^
climate variables

Compute path

Compute NPV of
damages from
AltC

Different damage
model

L Different SOW A

GET distribution over
marginal cost of carbon

«•••*

-------
Buying Down Risk

-------
Model Uncertainty

• Stress test

• Canonical variations

-------
Stress Test DICE Growth Model

A = abatement, A = total factor productivity, K = capital stock,
N = labor, 8 = depreciation

[l-A(t)] A(t) K(t)v IVKtpv

Output(t) =

(1 + .0028Temp(t)2)

K(t+1) = (1-8) K(t) + Output(t) - Consump(t)
Bernoulli Equation Consump(t)=r|(t)Output(t):
dK/dt = —5K(t) + B(t)K(t)T;

Put Temp(t) = 0; A(t) = A; N(t) = N; A(t) = 0; r|(t) = r\
K(t) = [(1 - y) B Jx=o t e_(1"Y)6x dx + e~(1~^8t K(0) d-y)]i/(i-Y)

-------
Two capital trajectories with DICE values,
no temperature rise, no abatement
K1(0) = 1$ and K2(0) = 1370 trillion $

1200

-s 1000

£2_
¦73

800

600

400

200

%
*

¦v
*

years

-------
Output[Trill $], outx(t) is output at time t
with linear temperature increase of x [C] in 200 years

with starting capital C = 137 [Trill $]

QUtO
out5
out 10
out 15

-------
Canonical Variations

• Do other simple model forms
have structurally different behavior?

-------
Lotka Volterra instead of Bernoulli Model

GHG(t+l) = (l-0.0083)GHG(t) + 0.024 x GWP(t)

Emissions proportional to
Gross World Output

Kelly & Kohlstadt 2001

GWP(t+l) = [1+ 0.03 - D(T(GHG(t)))] x GWP(t)

Gross World Output
Growth Rate

(World Bank, last 48 yrs)

D(GHG)(t) = (T/18

Weitzman's Death
Temperature

-------
Different Behavior

ghg

-------
Phase Portraits, w / wo Dependence

gwp

-------
Damage: Inner & Outer Measure

-------
Deal with Model Uncertainty?

• Fit your models to 0^?

• Fit your models to probabilistic
data from Structured Expert
Judgment

• Bayes Model Averaging

-------
Yale G-Econ Database: Gross Cell Product

Fig. 1. Economic map of Europe. This figure shows an economic topographical map of Europe with heights proportional to gross domestic product per area
Note how economic activity clusters in the core, whereas the periphery has much lower economic elevations. The observations measure economic activity ii
millions of 1995 U.S. dollars per km2 at a 1° latitude by 1° longitude scale.

Greek isles (0.10)

GCPpp l ime average growth rate:
[Ln(GCPpp) - minflnGCPpp)]

London (33.6) j 1 Pari*(3®')

Irish periphery
(021)

Oenman Rhineland
(8*6.5)

Arctic (0 003)

-------
Regression of Growthrate(400yr) on TEMPAVE

3.0E-2

2.SE-2

2.8E-2

2.4E-2

___

2.2E-2

2.0E-2

l.SE-2

1.6E-2

iij

1.4E-2

j-r

£
n

1.2E-2

1.0E-2

8.0E-3

6.0E-3

4.0E-3

2.0E-3

1.4E-17

X'.'i

-2.0E + 1 -l.OE + l 0.0E+0 1.0E + 1 2.0E + 1

TEMPAVE

~ Samples \7 Linear regression

3.0E + 1

Regression of
-------
Joint Distribution as Cobweb Plot

Conditional on top and bottom 1% growth rates

Sarrinl^s elected: 504.

absLat In(GCPNM) TEMPAVE Growth rate(400yr) TEMPMAX TEMPSD

0 -5.1 -22 -2.2E-06 2.2 0.16

-------
Factor multiplying GCPpp when changing only

TEMPAVE +3C

Indep Vbl

Covariate

Partial reg coeff

Factor

Model 1

In(GCPpp)

TEMPAVE

-0.06167

0.831095973

Model 2

In(GCPpp)

absLat

TEMPAVE

TEMPMAX

0.083428
0.109418
-0.09706

1.388541617

Model 3

In(GCPpp)

absLat
TEMPAVE
TEMPMAX
TEMPSD

0.088856
-0.038029
0.047788
-0.222984

0.892180333

Model 4

In(GCPpp)

TEMPAVE

TEMPMAX

TEMPSD

-0.121162
0.041572
-0.103743

0.695248461

-------
Conclusions

Need to tackle model uncertainty

Need to converge 'inner' and 'outer'
damage models

-------
What Are Predicted Impacts of Warming?

• 5°C

— collapse of Greenland ice sheet

— large-scale eradication of coral reefs

— disintegration of West Antarctic ice sheet

— shut-down of thermohaline circulation

— millions of additional people at risk of hunger, water shortage,
disease, or flooding

(Parry, Arnell, McMichoel et ol. 2001; O'Neill and Oppenheimer 2002; Hansen 2005)

• 11-12X

—regions inducing hyperthermia in humans and other mammals
"would spread to encompass the majority of the human
population as currently distributed"

(Sherwood and Huber 2010)

-------
Value@Risk (Basel II Protocol)

Banks reserve capital to cover "l-in-200 yr" loss event

BAU

Emissions = 2.4% GWP
Prob exceeding 13°C ~0.03

tmp

t
i

¦¦¦¦-. -¦¦¦

11 .::!¦¦¦•

¦ ¦¦

:,,s-

!.!sisi:i;:;i

.2E000a:i .4E0002-1.6E0002 1.8E0002 2E0002

¦¦¦¦¦¦¦¦¦¦¦¦¦¦

-------
Value (g) Risk (Base

Banks reserve capital to cover "l-in-200 yr" loss event

Stringent

Emissions = 1.5% GWP
Prob exceeding 13°C ~0.0001

tmpl

9.4

i-"

.8.4:-

'.7.4

.61

si:

e.a

[5.3"

II!

3:1E0002—

4.2

2.2

1.1

0.11

::=:=E

:¦¦¦
¦¦¦.

1 .2E0002 1.4E0002 1 .6E0002..1.8E0002 2E0002

!¦¦-¦¦¦¦-------•••¦!¦!¦¦¦¦_!.

,——Si—p

:::¦

-------
Risk Constraints

• PROB{ AT > 13°C for 500 yr} < 0.0001

• PROB{Greenland ice sheet melts in 300 yr} < 0.001

• PROB{Oceans become net C emitter in 100 yr} < 0.01

• What is the price?

-------