&EPA
United States
Environmental Protection
Agency
Stategic Studies Staff
Office of Policy Analysis
Office of Policy,Planning
and Evaluation
April 1984
Potential Climatic
Impacts of Increasing
Atmospheric CO2 with
Emphasis on Water
Availability and
Hydrology in the
United States
-------
POTENTIAL CLIMATIC IMPACTS OF INCREASING ATMOSPHERIC C02
WITH EMPHASIS ON
WATER AVAILABILITY AND HYDROLOGY IN THE UNITED STATES
REPORT PREPARED FOR
THE ENVIRONMENTAL PROTECTION AGENCY
BY
NASA GODDARD SPACE FLIGHT CENTER
INSTITUTE FOR SPACE STUDIES
NEW YORK, N.Y. 10025
Principal contributors: David Rind and Sergej Lebedeff
CAUTION: The state of the art of climate modeling is inadequate to accurately
forecast climate changes on the regional level. The grid estimates contained in
this report are for study purposes only and should not be relied upon or used
for the purposes of planning.
-------
TABLE OF CONTENTS
SECTION
TITLE
PAGE
PREFACE i
INTRODUCTION 1
I CLIMATE MODEL ASSESSMENT OF THE
HYDROLOGICAL CHANGES ASSOCIATED
WITH DOUBLING ATMOSPHERIC C02 3
II DIFFERENCE OF PRECIPITATION
BETWEEN WARM AND COLD PERIODS IN
THIS CENTURY 55
III FUTURE REFINEMENTS TO CLIMATE
MODELS 78
APPENDIX A ESTIMATING ANNUAL CHANGES IN
TEMPERTATURES 82
REFERENCES 94
-------
PREFACE
By
John S. Hoffman* and Stephen Seidel**
Water planners, coastal engineers, and agronomists as part
of their jobs necessarily make assumptions about future water
supplies, temperature, droughts, and storms. In general, they
assume that in the future these conditions will repeat those of
the past — there will be the same amount of water available,
the worst storm during the next hundred years will be similar to
the worst storm during the past 100 years, and droughts will
be of similar frequency and duration. Such assumptions are
used to design public works, to demarcate flood plains, and to
establish safety and growth margins for project and planning
purposes. In agriculture, such assumptions form the basis for
developing more productive strains of crops and tor establishing
planting and irrigation practices.
Unfortunately, the underlying assumption that future cli-
mate will essentially repeat the past no longer appears valid.
Research efforts since the turn of the century have produced
considerable scientific support for the belief that increases in
atmospheric levels of carbon dioxide and other greenhouse gases
will increase global temperature and alter precipitation patterns
by trapping infrared radiation.
*Director, Strategic Studies Staff, Office of Policy Analysis,
U.S. Environmental Protection Agency
**Senior Policy Analyst, Strategic Studies Staff, Office of
Policy Analysis, U.S. Environmental Protection Agency
-------
- 11 -
Although a significant rise in temperature appears likely,
much uncertainty still surrounds the timing and magnitude of a
greenhouse warming. Moreover, less confidence exists about
effects other than temperature, including potentially critical
changes in regional rainfall, evaporation, storms, droughts,
run-off, and soil moisture. Yet the importance of possible
shifts in the hydrologic cycle on future agricultural practices,
public works, and development patterns underscores the need to
improve our ability to project likely changes.
This study focuses specifically on possible shifts in a
range of hydrologic conditions that could accompany a doubling
of atmospheric carbon dioxide. It reports on an experiment done
with the Goddard Institute for Space Study's general circula-
tion model (GCM) — a complex computer simulation of the physical
forces that produce weather patterns. GCMs are the most
sophisticated tool used to project climatic changes resulting
from increases in greenhouse gases. Past reporting of GCM
experiments has dealt primarily with likely changes in tempera-
ture accompanying a doubling of atmospheric CC>2 levels; this
analysis shows how these models can be used to examine potential
hydrologic effects. THIS STUDY REPRESENTS A FIRST ATTEMPT AT
PROVIDING DETAILED REGIONAL HYDROLOGIC EFFECTS. THE VALUES
GIVEN FOR HYDROLOGIC CONDITIONS IN PARTICULAR GRIDS SHOULD NOT
BE USED FOR PLANNING PURPOSES.
Section I reports on a range of hydrologic conditions
(e.g.* precipitation, evaporation, soil moisture, run-off) for a
"control" run that seeks to duplicate existing conditions. It
-------
- 111-
then compares these results with output from a model run where
atmospheric CC>2 levels are doubled. Section II shifts the focus
of analysis from the future to the past. It compares hydrologic
conditions during a particularly cold period (1900-20) with a
particularly warm period (1940-60). The final section discusses
future research efforts needed to advance our understanding and to
improve GCMs and thus enhance our ability to respond to changes
resulting from the greenhouse effect. It also describes an
important project underway at GISS to further this effort. The
remainder of this foreward sets the context for this analysis.
Scientific Evidence of the Greenhouse Effect
Greenhouse gases in the atmosphere allow the sun's energy
to penetrate and warm the earth, but then block the escape of
some of the infrared energy given off by our planet. In effect,
these gases form a thermal blanket around the earth. Carbon
dioxide is an important greenhouse gas. Others exhibiting the
same property include water vapor, nitrous oxide, methane, and
the chlorofluorocarbons.
A small amount of these greenhouse gases in the atmosphere
is partly responsible for the climatic conditions under which
our civilization has prospered. For example, the earth is
30°C (58°F) warmer than it would be without the presence of any
greenhouse gases. In contrast, the atmosphere surrounding Venus
is 97 percent CO2 and it is much warmer than earth, while Mars
with little CC>2 or water vapor in its atmosphere is much colder.
-------
- iv -
Although CC>2 constitutes only .03% of our atmosphere, it
has increased more than 20% since the beginning of the Industrial
Revolution. Much of this increase is directly linked to
expanded use of coal, oil, and natural gas. In addition, past
reductions in forests — trees absorb and store carbon dioxide
during photosynthesis — may also have contributed to histo-
rical increases in atmospheric CC>2 levels.
Despite this sizable increase in CC>2 levels, the complexity
and time lags built into our climatic system prevent scientists
from stating unequivocably that the global warming experienced
since 1850 can be attributed to the rise in CO2» The amount
of warming that should accompany the past rise in CO2 is too
small compared to unexplained natural variation in global
temperature to yield statistically significant proof that
CO2 is responsible. Nonetheless, this warming is consistent
with expectations.
In an effort to illuminate this issue, the National Academy
of Sciences has conducted two extensive reviews (1979, 1983)
of what we know and don't know about the greenhouse effect.
In each case, they concluded that temperatures would ultimately
increase somewhere between 1.5°C (2.7° F) and 4.5°C (8.1° F)
for a doubling of pre-industrial atmospheric CO2 levels (to
600 ppm). Increases in other greenhouse gases could increase
the overall warming by an additional 50-100 percent. Current
estimates of fossil fuel usage suggest that we could reach
-------
- v -
atmospheric levels of 600 ppm by 2075 (NAS, 1983).
This rate of temperature change would be unprecedented.
For example, a temperature increase of 2.0°C (3.8° F) by
2050 would be roughly comparable to the highest temperature
of the last 100,000 years. At the height of the last Ice Age,
approximately 15,000 years ago, the earth was only about 5°C
(9.0° F) colder.
Rising temperatures would produce many other climatic
changes. As the difference in temperature between the equator
and the poles shifts -- the poles will warm at a faster rate
because melting ice changes their albedo -- weather patterns
may be radically altered, resulting in changes in precipitation.
Potential Changes in the Hydrologic Cycle
This study first looks at aggregate measures of changes in
the hydrologic cycle such as annual average precipitation and
evaporation for the entire North American continent. These
changes provide a broad sense of whether we are likely to
experience more or less rainfall.
However, most decisions based on assumptions about hydro-
logic conditions require more specific data. Annual averages
across a large geographic scale are not detailed enough
in two respects. They may mask critically important changes
in the seasonality of precipitation. For example, moderate rain-
fall just before spring planting is far more important to
-------
- vi -
agricultural productivity than the same rainfall in the fall.
Similarly, rainfall concentrated in a single season would
require different design standards for many public works than
the same amount dispersed throughout the year.
To be useful to decisionmakers, projections of hydrologic
changes must also be tied to specific locations. Precipitation
can vary dramatically within a hundred miles depending on the
location of mountains, lakes, and oceans, and on prevailing wind
patterns. Other important factors such as soil moisture and
run-off also are linked to conditions at a specific location.
This study seeks to address these issues by examining
possible changes in specific hydrologic characteristics including
ground moisture, lenghth of growing season, frequency and severity
of droughts, run-off, and ground moisture. Changes in these
factors may be particularly relevant to decisionmakers concerned
with long-term projects (25 years or more) that are sensitive
to assumptions concerning hydrologic conditions.
Caveats to This Analysis
The GISS results suggest substantial changes throughout
the hydrologic cycle. If changes of this magnitude were to occur
in the future, the assumption that future hydrologic conditions
would repeat the past would fall far wide of its mark.
However, the results of this modeling effort should be consi-
dered only the first step in the process of planning for future
changes. Because of current limitations, the model does not
-------
- vii -
resolve many important issues. Model input and resolution
must be enhanced in several important respects.
To minimize computer time, the GCM developed by GISS simu-
lates the weather for rather large grid areas. The whole
earth was divided into grids, each consisting of approximately
17,000 sq. miles. The grid size used in GCMs covers a
large area of land that may include diverse topography. Not
only does this prevent a completely reliable representation
of the physical processes that determine weather, but it also
makes interpretation of the output somewhat difficult. Conditions
at a specific location within a grid may vary considerably
from the average for the grid.
The GISS model produces a warming of 4.1°C, a temperature
increase at the higher end of the current NAS temperature range
(1.5-4.5°C) for a doubling of atmospheric C02. As we learn
more about the processes producing climate, it is possible
that this number will shift, though it probably will remain
within the NAS range. Narrowing this range will require more
extensive data from observed characteristics of the oceans
and clouds as they effect our climate system and improvements
in the representation of the processes in models. Changes
also will be needed in detailing ground hydrology, vegetation
response, and the changes in these systems over time as climatic
conditions shift. More accurate model results will become
available only as quickly as support is given to efforts to
-------
- vin -
observe the climate system adequately to enhance and validate
the GCM models.
As they now stand, GCMs are able to accurately recreate
existing climatic conditions only at a very aggregate level.
Thus, when run with existing concentrations of greenhouse gases,
they will produce current global temperature, rainfall, and
seasonal changes in reasonably accurate fashion. Today's models
cannot, however, recreate existing conditions in each grid —
essential information for decisionmakers. For example, in the
control run for this study, efforts to recreate existing hydrolgic
conditions produced results that were generally consistent for the
entire North American continent, but showed 50-100 percent exces-
sive rainfall in the western part of the continent and half the
observed rainfall around Tennessee. Clearly, improvements to the
model will be required to increase the reliability of its output,
particularly where information from individual grids is being
used.
Finally, the model run examined in this report looks only
at changes that will occur once CC>2 levels have doubled and
temperatures have reached equilibrium. Steady state conditions
are useful to simplify the analysis, but focus attention on
changes that probably will not occur before 2075. In the interim
significant hydrologic changes are likely to accompany increases
-------
- ix -
in CO2 and other greenhouse gases. Nor is there a basis for
simply stating that the transitional period will be a linear
interpolation based on the results from a doubling of CO2.
In fact, a region shown to be drier after a doubling of CO2
has occurred could, in fact, first become wetter before
beginning to dry.
As mentioned above, Section III of this report discusses
efforts currently underway to improve the ability of GCM's to
reliably predict climatic changes associated with annual increases
in levels of greenhouse gases. Appendix A details some of the
assumptions that will be made in a first effort to simulate the
yearly evolution of climate. These results should be available
by the end of 1984 and provide a better basis for assessing the
short and long-term vulnerability of decisions to climatic change.
Potential Response by Decisionmakers
Given the uncertainties surrounding the results of this
study, water planners, engineers and agronomists face a difficult
situation. Based on the output presented here and elsewhere,
it is clear that future hydrologic conditions should diverge
from past conditions and that decisions involving long-term
projects may be vulnerable to such changes. Nonetheless, we
are currently limited in our ability to project future hydrologic
conditions at the required geographic scale. WE CAN SAY WITH
SOME CERTAINTY THAT CHANGES WILL OCCUR, BUT WE CANNOT YET ASSIGN
A VALUE TO THOSE CHANGES.
-------
- x -
In light of these uncertainties, decisionmakers need to begin
assessing the vulnerability of their decisions (e.g., design,
sizing, and location of long-term projects) to a range of
potential changes in hydrologic conditions. In many cases, low
cost changes can be made that will reduce future risks. This
analysis can also show the value of reducing existing uncertain-
ties in projections -- whether the costs to narrow these
uncertainties are justified by the potential value of
reducing the accompanying risks. Finally, decisionmakers must
develop an ongoing dialogue with climate modelers to effectively
communicate their priorities and to make certain that useful
and timely information is made available to them.
-------
INTRODUCTION
The increase of C02 in the atmosphere resulting from fossil fuel combustion
is expected to have a profound climatic effect, due to the ability of carbon
dioxide to absorb radiation emitted from the earth atmosphere system. It is
expected that increasing 0)2 will warm the atmosphere. The magnitude of this
warming, and the consequences for the other aspects of the climate system -e.g.
precipitation patterns, clouds, winds, etc. -are currently being investigated by
a variety of scientific researchers.
Two recent reviews of the evidence by the National Academy of Sciences
(1979, 1981) concluded that a doubling of carbon dioxide would probably lead to
an increase in mean global surface temperature of between 1.5°C and 4.5°C. This
range was determined primarily from reviewing results of several mathematical
models of the earth's climate, the most sophisticated models of which, the
General Circulation Models, incorporate representations of many physical pro-
cesses in a three dimensional framework. These models are also capable of indi-
cating changes in the other climate variables, and, in particular, can assess
expected alterations in components of the hydrological cycle.
This report concentrates on analyzing the hydrologic changes over the North
American continent that were produced by doubling the carbon dioxide in the
Goddard Institute for Space Studies (GISS) general circulation model. The first
section of this report describes the model and provides a comparison between the
model output and the actual current climate. The changes that the model pro-
duces with a doubling of the atmospheric carbon dioxide are then presented. The
second section attempts to put the results in perspective by examining histori-
cal variations in precipitation.
The results presented in this report, while giving plausible estimates of
potential hydrological changes at the regional level, should be used and inter-
-------
preted with great care. Many aspects of the GCM lead to uncertainty in the
reliability of the results. The computation of changes at a relatively coarse
geographic scale necessarily introduces uncertainty as to the magnitude of the
change on a smaller scale. Other uncertainties are introduced because some
relevant processes ~ ocean transports and cloud process, for example-are mod-
eled very crudely. Consequently the certainty that can be attached to various
results is problematical. In some grid boxes and regions there may be large
errors. Thus even though it is possible to conclude that climatic normals,
whether they be thirty year or one hundred year averages, cannot be reliably
used as predictors of future means or variation in climate it is as yet impossi-
ble to project future conditions accurately.
RESEARCHERS USING THE RESULTS OF THIS GCM EXPERIMENT IN THEIR STUDIES MUST
RECOGNIZE THAT THE OUTPUTS OF THE GCM CANNOT BE CONSIDERED AS ACCURATELY
DEFINING FUTURE CONDITIONS. More reliable estimate of hydrological change will
become available only as further research, scientific effort, and data collec-
tion allow the scientific community to improve the GCM's representations of the
actual climate processes. The speed at which this is done will directly depend
on the overall level of effort the whole scientific community is able to devote
to this problem. THIS REPORT, IN EFFECT, PRESENTS A METHODOLOGY FOR ESTIMATING
THE HYDROLOGIC IMPACT OF INCREASED ATMOSPHERIC C02, AND SHOULD BE LOOKED UPON AS
A FIRST APPROACH TO A COMPLEX PROBLEM.
The results presented herein apply only to a doubled 0)3 climate, once the
system comes to equilibrium. C02 amounts are not expected to double until
approx-innately the middle of the next century, and it will take additional deca-
des before the equibrium is attained. Furthermore the doubling simulation does
not consider the increases taking place in other greenhouse gases, -- methane,
chloro-flurocarbons, and nitrous oxide -- which will significantly enhance the
-------
expected wanning. In order to consider the transient response of the climate
system between now and then and to consider the increases in trace gases, GISS
is conducting an experiment in which C0£ and these other greenhouse gases are
increased gradually in the model atmosphere, starting from 1958 values. This
later experiment will not only provide a more realistic estimate of expected
decadel changes, but will also provide insight into the expected response of the
climate system in a world of doubled C02 in which the climate system is in dise-
quilibrium. In the third section we describe details of this experiment.
I. CLIMATIC MODEL ASSESSMENT OF THE HYDROLOGICAL CHANGES ASSOCIATED WITH
DOUBLED ATMOSPHERIC C02
Model Description
The general circulation model used for this work has been developed at
Goddard Institute for Space Studies (GISS) over the last several years. Global
in extent, the model has realistic topography, 8° x 10° resolution in the hori-
zontal, and nine layers in the vertical. Climate is simulated by solving the
fundamental equations for conservation of mass, momentum, energy and water,
using numerical schemes developed by Arakawa (1972). Parameterizations for the
source terms in these equations represent the physical processes of radiation,
turbulent transfers at the ground-atmosphere boundary, cloud formation, and con-
densation of rain. A complete description of the model appears in Monthly
Weather Review (Hansen et al., April, 1983).
The only difference between the version of the model used for the experiment
and the one documented in the literature is in the determination of sea surface
temperatures and sea ice. In the documented version these values are specified
climatologically based on monthly-mean values with linear interpolation once per
day. If a calculation is made in that run of the net heat gain or loss from each
ocean grid box, it is then possible to specify what heat gain or loss from each
-------
ocean grid box, it is then possible to specify what heat convergence or divergence
in the ocean would be necessary to reproduce the observed ocean temperatures.
Assuming that these heat fluxes represent horizontal transports, one can calcu-
late the implied meridional ocean heat transport, and compare these values with
observations. As shown in Hansen et al. (Fig. 15) the close agreement of model
II, used in these experiments, with the observations indicates that the ocean
heat exchange with the atmosphere on an annual and zonal average is realistic in
the model. An expanded discussion of this technique and its results, including
a detailed geographical presentation, can be found in Miller et al . (1983).
In the control run, and the climate change experiments, it was necessary to
allow the sea surface temperatures to change. A simple thermal response which
ignores ocean heat transport would result in the low latitude ocean (and atmo-
sphere) being excessively warm, while high latitudes would be too cold, unless
the atmospheric transport increased substantially. Either alternative would be
unrealistic, in comparison with the current climate. It was thus decided to
include the ocean transports which were calculated from the specified sea sur-
face temperature model as necessary to reproduce the observed temperatures,
given the fluxes in or out of the ocean which prevailed in that model. Thus to
first order, the sea surface temperatures would be allowed to vary as radiative,
sensible or latent heat fluxes varied, but would be kept realistic in the cur-
rent climate simulation by the specified fixed transports. To the extent that
the control run experienced fluxes which differed from the specified SST model,
the sea surface temperatures would vary from the climatological average, but not
to the extreme that would occur if no transport were included. The specified
transport, with varying sea surface temperatures, was thus a standard part of
the control run and experiment. Sea ice was also allowed to vary, forming when
the ocean temperature drops below -1.6°C, and melting when it rises above 0°C.
-------
For the purpose of this report we will compare observations and model simu-
lations of the hydrologic cycle. This comparison must be labeled preliminary;
more thorough comparison is expected in a later report.
Three situations will be described:
(1) Observed climate, which represents averages for forty years or more
(depending on location). For the longest averages (8U years) they
cover a time period during which a climatic change of 0.3°C occurred,
for global mean temperature.
(2) The control run, the model's simulation of the current climate, a
period of ^0_ modeled years with C02 at 315 ppm, with no climate trend
(the only exception to this is shown and explained below, for Figure 1).
(3) The doubled C02 or "experiment" run, in which the world is treated as
if the C02 had doubled and it had time to reach equilibrium.
Comparisons of Model Control Run and Observed Climate
Fig. 1 presents the annually averaged precipitation produced by a five year
run of the model with specified climatological sea surface temperatures along
with observations of actual precipitation (left hand side). (Year to year stan-
dard deviations of 25% occur in the model in certain regions, so while five years
are sufficient to show the general patterns of rainfall, the exact values miyht
vary somewhat for a longer time average). The distributions are similar, with
both showing relatively little precipitation in desert areas (Sahara, Gobi,
Australian deserts) and off the west coasts of continents (South America, North
America and the southern portion of Africa). Regions of large rainfall occur in
the tropics in the Amazon and African rainforests and in the central Pacific.
The model produces too much rain over the Bay of Bengal and New Guinea, and some-
what more than observed over certain portions of North America. A more detailed
comparison with observations over North America will be presented below.
-------
Preclpitotion (mm day1), Annual Mean
Observations
Precipitation Interonnuol \fariobihty (percent)
Observations
0.30
•90
-ISO -I2O -60 0 60
Longitude (degrees)
Precipitation (mm doy'l.Annuol Keen
120
Model H-I(L'9.B*»IO*)
-60 0 60
Longitude (degrees)
„„ Ptecipitotign Interonnuol Variability (percent) Model I-I(L"9 B
T" II I I I I •• I
120 IBC
-60 0 60
Longitude (degrees)
60 O 60
Longitude (degrees)
Fig. ]. . Global distribution of annual-mean precipitation (left) and its
interannual relative variability (right). Observed annual-mean precipitation IB
from Schutr and Gates (1971) and Interannual variability from Berry et al. (1973).
-------
In addition to the annual mean precipitation, Fig. 1 also indicates the
relative variability of precipitation from year to year in the model and in
observations (right hand side). This is an important aspect of the model simu-
lations because it indicates whether the model's natural variability is repre-
sentative of the real variability, and implies something about the sensitivity
of the model to internal climate forcings relative to that of the real world.
As can be seen in the figure, the model's relative variability is similar to
that which is observed both in magnitude and pattern, except for underestimating
the variations that occur in the tropical Pacific and Atlantic. This deviation
is presumably related to the use of specified sea surface temperatures in that
area; in reality there is strong interannual variability in the sea surface tem-
peratures in this region associated with the El Nino phenomenon, which occurs
together with large fluctuations in rainfall. Over land, the model appears to
have the proper degree of natural variability.
The control run was extended for 35 years allowing the ocean temperatures
to change in response to thermal forcing, while incorporating inputs or outputs
of heat into each ocean grid box representative of the influence of ocean cur-
rents. This procedure kept the ocean temperatures at reasonable values, while
allowing them to respond to radiative and other thermal variations. A complete
description of this technique is available in Miller and Russell (1982). The
following comparisons with observations are from the last ten years of this run
unless otherwise stated. Fig. 2 shows the model grid, numbered for reference.
Fig. 3a and 3b show the annual rainfall in the model grids over North
America and the observed rainfall on a much finer scale. This comparison is
meant to emphasize the subgrid-scale variation which exists in the observed
rainfall pattern, which the model cannot directly reproduce. Fig. 3c shows as
estimate of the observation on the model grid scale. Given the inhomogeniety
-------
7
II
\ 1
-) /1
Fig. 2. Grid boxes over North America and the surrounding ocean in the GTSS
8 x 10 global climate model. Grid boxes are numbered for reference.
-------
Fig. 3a. Observed annual precipitation (mm) over North America (Korzoun et al 1977)
-------
I35CJ3I4
Fig. 3b. Model produced annual precipitation (iron) from years 26-35 of the
control run.
-------
Fig. 3c. Estimated observed annual precipitation (mm) on the model grid,
-------
12
evident in Fig. 3a the value for a given grid box is probably only accurate to
about 20%. Comparison between these presentations shows that the model produces
realistic rainfall except for the following deficiencies: rainfall is exceesive
in general, by about 50%-100% in the longitude belt centered around 100°W and
110°W stretching from Canada southward to Mexico. In contrast, the grid box
centered in Tennessee (grid box 18) has only one-half the observed rainfall.
Detailed discussion of the reasons for these deficiencies are beyond the scope
of this report; it is believed they are principally related to the larger than
actual water holding capacities of the ground specified in the model.
This emphasizes the necessity to treat the results given below for indivi-
dual grid boxes with extreme caution. As shown in Fig. 1 the model simulation
of precipitation on global and regional scales is quite realistic. The compari-
son in Fig. 3 indicates that in certain regions, and especially in certain grid
boxes, the model does not produce precipitation values in accord with observa-
tions. This introduces a degree of uncertainty into the validity of the changes
produced for any single grid box in climate change experiments. The results for
a given grid box should not be considered to be the expected change for that
particular physical location. The greater the geographic scale of the expected
climatic change, the more confidence can be attached to the result.
Fig. 4a and 4b show the annual evaporation in the model over land along with
the "observed" evaporation. It is important to realize that the observations
are deduced from a mathematical formula, which differs somewhat from the formula
used to calculate evaporation in the model; thus the comparison is not as meaning-
ful as was the case for precipitation. Nevertheless, the large-scale patterns
agree with those determined for precipitation, with somewhat excessive evapora-
tion west of the Mississippi River (by about 30%) and a deficit around Tennessee.
Fig. 5 shows the runoff calculated in the model, and observed river runoff.
Again the question of the similarity of the two diagnostics arises and will be
-------
Fig. 4a. Observed annual evaporation (nun) over North America (Korzoun et al 1977)
-------
Fig. 4b. Model produced annual evaporation from years 26-35 of the control run,
Values are in mm.
-------
'ig. 5a. Observed annual runoff over North America (Korzoun et al 1977).
-------
Fig. 5b. Model produced annual runoff (mm) from years 26-35 of the control run
-------
17
commented upon further in a later report. As shown here the model runoff is
excessive (by a factor of two or more) in the same grid boxes that precipitation
was excessive, while being deficient in the Tennessee area, so the pattern is
qualitatively consistent.
The comparisons presented indicate the extent to which the model repro-
duces the annual average hydrologic cycle. THESE DEFICIENCIES DIRECTLY AFFECT
THE ACCURACY OF ESTIMATES FOR INDIVIDUAL GRIDS AND MUST BE REMEMBERED WHEN EVA-
LUATING THE RESULTS OF THE DOUBLED C02 EXPERIMENT.
Experiment Results
The amount of atmospheric C02 in the model was doubled and the run was inte-
grated for 35 years, enough time to produce an equilibrium climate. The results
discussed in this section are for a comparison of the last ten years of the ex-
periment with the last ten years of the control run unless otherwise stated.
The doubled C02 world has an annual average surface temperature increase of
4.16°C in this experiment. Over the United States the temperatures rise by 4.2°C
in the eastern part of the country, and 4.9°C in the central and western parts.
There is some seasonal variation to this rise, with the increase in winter being
about 40% larger than the increase in summer. One way to appreciate the magni-
tude of this increase is to consider the maximum temperatures observed on an
average daily basis, and for the month as a whole, for different cities in the
United States during July with the model-predicted doubled C02 climate. Fig. 6a
shows the average daily temperature maximum to be expected in degrees Fahrenheit
obtained by adding the model calculated temperature change due to doubled C02 to
the currently observed value. These values average 5-7°F higher than is current-
ly observed. Fig. 6b shows the monthly maximum temperature which would be
expected; again, the difference is 5-7°F above current expectations.
-------
Fig. 6a. Average daily temperature maximum in July in doubled CC>2 climate.
The current average daily temperature maximum for these cities is 5°- 7°F less,
-------
Fig. 6b. Average monthly temperature maximum to be expected in July in
doubled CO2 climate.
-------
20
In the following discussion we will concentrate on the effects of doubling
the C02 on factors related to the hydrologic cycle over North America. A more
complete discussion of the results of this experiment will be published elsewhere.
I. Annual Results: Basic Diagnostics
Fig. 7 shows the change in annually averaged precipitation for grid boxes
over North America. Positive numbers indicate precipitation increase in the
doubled C02 climate compared to the control run. The general pattern shows that
precipitation increased in the north and northwestern portions of the domain,
with variations of alternating sign elsewhere. The numbers in parenthesis at
the bottom of each grid box indicate the percentage change of the annual preci-
pitation relative to the ten years of the control run. Fifteen to twenty per-
cent increases are common in the north and west. An increase in precipitation
on the annual average is characteristic of the doubled COg world, as warmer tem-
peratures lead to greater evaporation of moisture from the ocean. The global
average precipitation increases by 11% when C02 is doubled.
The change in evaporation, along with the percentage change from the control
are shown in Fig. 8. The change refers to the difference in evaporation over the
land portions of the grid box, while the percentage change relative to the con-
trol run uses evaporation from the grid box as a whole, but it should be repre-
sentative, for those grid boxes which are mostly land. The pattern noted for
precipitation is repeated for evaporation with increases of 15-20% (or more)
occurring over the northern and western portions of the map. On a global basis
evaporation increased by 11% similar, to the rainfall.
Fig. 9 shows the change in runoff over land, along with the percentage change
from the control run. Assuming no change in water storage over the last ten years
this would be equal to the difference between the precipitation and evaporation
changes (Fig. 7 and 8). The values shown in Fig. 9 are close to this difference,
-------
Fig. 7. Change in precipitation between the last ten years of the doubled CG>2
run and the last ten years of the control run, for the annual averaqe. The top
number indicates the actual change (mm), the bottom number in parenthesis gives
the change in % relative to the control run.
-------
Fig. 8. Change in evaporation between the last ten years of the doubled CC>2
run and the last ten years of the control run, for the annual average. The
top number indicates the actual change (mm), the bottom number in parenthesis
gives the change in % relative to the control run.
-------
Fig. 9. Change in runoff between the last ten years of the doubled CO2 run
and the last ten years of the control run, for the annual average. The top
number indicates the actual change (mm), the bottom number in parenthesis
gives the change in % relative to the control run.
-------
24
although not exactly equal, indicative of the ground moisture changing somewhat
on the annual average during the period (or perhaps a numerical resolution effect)
The general pattern is one of increased runoff over the northwestern and extreme
southwestern portions of the continent of 20-60%. Some grid boxes in the central
and eastern regions have.decreases of 15-20%.
Fig. 10 gives the ground wetness for each grid box, the percentage of soil
moisture in the first layer in the ground relative to the total water holding
capacity of the earth. The control run values (the lower number) were generally
less than 25% of what could potentially be held in the southwest, rising to over
50% in the northeast. In the doubled COg experiment, small changes (upper number)
in this quantity occur, with a small reduction over most of the area except for
the extreme northeast.
Fig. 11 shows the total earth water (and ice) for the ground extending down
to a depth of 4 meters. The change is the top number, and the percentage change
relative to the control run is shown as the bottom number is parenthesis. The
values in the control run are greatest in the northwest and northeast, and least
in the southwest. The northern and western portions of the continent generally
increase their total water content, by 20-60%, while the southern and eastern
sections generally experience some drying. The results shown in Fig. 11 repre-
sent a concise summary of the direct hydrologic changes experienced in the
doubled ($2 climate for the annual average.
2. Annual Results: Interpretative Diagnostics
To evaluate the increase or decrease of extended dry episodes, a drought
index was formulated. This index is similar to the Palmer drought index except
that it relates current water availability to climatologically expected water
availability, rather than to water use. If P-E is the difference between preci-
pitation and evaporation for a month, P^E the mean difference for the same ten
-------
-i
-447)
-4
-2
(43)
(75^6)
09)
-2
'(25)
Fig. 10. Change in ground wetness between the last ten years of the doubled
C02 run and the last ten years of the control run, for the annual average. The
top number gives the actual change ( in %), the bottom number in parenthesis
gives the ground wetness (in %) for the control run.
-------
Fig. 11. Change in total earth water between the last ten years of the doubled
COj run and the last ten years of the control run, for the annual average.
The top number indicates the actual change (mm), the bottom number in parenthesis
gives the change in % relative to the control run.
-------
27
calendar months in the control run, and S.D. the standard deviation of this dif-
ference from the control run, then the index is:
^current = .897Iprevious month + (
The relationship between the current index and that for the previous month is
the same as that used for the Palmer drought index. Positive values of the
index indicate greater precipitation relative to evaporation than was experi-
enced in the control run, with a deviation of one standard deviation two months
in a row augmenting the index by 1.897. The index is normalized for each grid
box so that the control run has a distribution for each category similar to that
for the Palmer index. Table 1 shows the categories for the Palmer Index, and
the distribution for each category for the control run with the drought index as
defined above. This drought index differs from the Palmer index in that it uses
actual evaporation rather than expected demand as the water loss process.
The difference in the drought index due to doubling 0)2 is shown in Fig.
12. The results show a tendency for droughts to persist longer, on the annual
average, in the south and east, with sequences of wetter than normal months more
likely in the northwest. Since by definition all grid boxes have distributions
corresponding approximately to that shown in Table 1, a change in the drought in-
dex of -2 would in the mean shift all categories two divisions to the left, and
increase extreme dryness from 2% to 10% in occurrence. Similarly an increase in
the index of +2 would increase the occurrences of extreme wet periods in the same
fashion. However, there is no way of telling from the numerical change alone in
what situations the index changed, and thus it is necessary to look at the change
in distribution for any particular grid box. This will be done for specific grid
boxes in the next section. The changes shown in this figure are consistent with
the pattern of effects that has become visible during the course of our review of
-------
Fig. 12. Change in the drought index between the last ten years of the doubled
CO2 run and the last ten years of the control run, for the annual average. A
negative value implies increasing drought frequency; see the text for an
exact definition.
-------
29
the previous diagnostics.
A diagnostic directly related to the influence of climate on agriculture is
the plant water stress, which is defined by the formula
PUK - V is r TGmax - TGmin , >,
PWS ' L 18 ( TSmax - TSmin -1 )
where TG is the temperature of the top layer of the ground (10 cm), TS is the
surface air temperature, and max and min refer to the maximum and minimum values
of these two temperatures recorded during the day. The plant water stress is
Table 1
DROUGHT INDEX DROUGHT % OCCURRENCE IN
CONTROL RUN
DI < -4.0
-4.0 < DI < -3.0
-3.0 < DI < -2.0
-2.0 < DI < -1.0
-1.0 < DI < 1.0
1.0 < DI < 2.0
2.0 < DI < 3.0
3.0 < DI < 4.0
4.0 < DI
Extremely Dry
Severely Dry
Moderately Dry
Mildly Dry
Near Normal
Mildly Wet
Moderately Wet
Severely Wet
Extremely Wet
2
6
10
15
33
15
10
6
3
thus accumulated daily; if the ground temperature variation is twice that of the
surface air temperature, then 18° accumulate per day. With the ground temperature
exceeding the surface air temperature there can be expected to be a flux of heat
and moisture out of the ground which is assumed to "stress" local vegetation.
In general, plant water stress decreases somewhat from south to north. (It should
be noted that the legitimacy of using this index in a higher C02 world, is ques-
-------
30
tionable. Higher C02 will change stomal behavior in plants, and plant transpira-
tion. Consequently the degree to which plants are stressed will change. The
results given here do not include these effects).
Fig. 13 shows the change of plant water stress and the percentage change in
the experiment relative to the control run. In the experiment the stress in-
creased in both the southern and northern portions of the region. This of course
is the yearly average result; the changes for different seasons will be discussed
in the next section.
Another element of the climate system which is of importance to agriculture
is the length of the growing season. This is defined in the model as the dura-
tion between days that the surface air temperature drops below 0°C during August-
January (in the Northern Hemisphere). The change in the growing season, and the
percentage change relative to the control run are shown in Fig. 14. The changes
are quite dramatic and unambiguous - the growing season increases throughout, by
values approaching 50% in the northern grid boxes. This occurs because of the
general warming experienced due to increasing C02, which is apparently much more
pronounced statistically than the precipitation changes, at least on the yearly
average.
One additional factor which might be expected to influence plant growth is
the cloud cover, or equivalently, the amount of sunlight reaching the surface.
The change in this proved to be insignificant. Cloud cover changed by only a
few percent from the range of 30-70% for different grid boxes in the control run.
To summarize the results from this section: the annually averaged hydrolo-
gic cycle increased in intensity over the northwestern portions of the continent,
with greater precipitation, evaporation and runoff. Specific aspects of the hydro-
logic cycle increased also in the northern regions (precipitation, total earth
water, evaporation) and western portions (precipitation, evaporation, runoff).
-------
Fig. 13. Change in plant water stress between the last ten years of the doubled
C02 run and the last ten years of the control run, for the annual average. The
top number indicates the actual change, the bottom number in parenthesis gives
the change in % relative to the control run. See the text for an exact definition,
-------
4
w,
o
Fig. 14. Change in the qrowinq season between the last ten years of the doubled
C02 run and the last ten years of the control run. The top number indicates the
actual change (days), the bottom number in parenthesis gives the change in %
relative to the control run. See the text for an exact definition.
-------
33
The southern and eastern portions of the continent had either mixed changes,
or a tendency towards drying.
The warming associated with doubled C02 occurred throughout the continent
and produced a longer growing season everywhere and increased plant water stress
in the northern and southern sections.
3. Monthly Variation
While the annually averaged results present a picture of the variations on
the continental scale for the climate in general, for many applications it is
necessary to know variations on a monthly basis. In this section we will look
at how various parameters changed monthly in four specific grid boxes: the grid
boxes labeled 9, 16, 17, and 14 in Fig. 2. Grid box 9, which encompasses all or
parts of such states as Idaho, Washington and Oregon experienced the precipitation
increase noted for the northwestern section of the continent in Fig. 7. Grid box
16 includes the major portion of the Colorado River. Grid box 17, which includes
the states of Kansas and Oklahoma, takes in the wheat growing region of the
southern plains. Finally, grid box 14 covers the heavily populated regions of
the northeast, New York and New England.
It is important to again emphasize the uncertain nature of climate change
results for an individual grid box. The seasonal variation of changes for pre-
cipitation are shown for the entire country in the next section; the changes which
occur on regional or larger scales are of more certainty than those which vary
from grid box to grid box. The individual grid boxes chosen provide an indication
of the range and character of changes that are likely to occur due to doubled C02,
and, once again, should not be thought of as necessarily applying to that parti-
cular location.
-------
34
Changes in a grid box with increased wetness
Grid box 9, which the control run had about 50% wetter than the observed
values, experienced an overall increase in the hydrologic cycle, as indicated in
the previous section. Fig. 15 shows the monthly variation of precipitation, eva-
poration and runoff as the change from the control run values for each month; an
increase of all three parameters occurs in all months except September and April.
The rainfall and runoff changes are small from July through September when both
parameters are at their minimum in the control run. Thus to some extent the
largest increases occur when the hydrologic cycle in the control run is largest,
which indicates that the climate change is amplifying an already existing pattern.
Fig. 16 presents the monthly variation of other parameters, related both to
the hydrologic cycle and to vegetation. The total earth water, which is affec-
ted by the difference between evaporation and precipitation, increases except
for the time period August through October, mirrorring with a one month phase
lag the effects noted in Fig. 15. The drought index, defined in the previous
section, shows that months with precipitation greater than evaporation, relative
to the control run, occurred more consecutively throughout the year.
The change in the percent occurrence of extreme dry and wet periods for
this grid box, as defined in Table 1, can be seen in Table 2. In the control
run severe and extreme droughts occurred 12% of the time, while severe and ex-
treme wet periods occurred 8% of the time. In the doubled C02 climate, extreme
and severe dryness occurred only 6% of the time, while extreme and severe wet
periods now occurred 37% of the time. As can be seen from the figure the in-
crease occurred throughout the year, although somewhat less often in summer.
The plant water stress increased throughout the summer months (as did eva-
poration as in Fig. 15) due to the general warming of the ground. The snow
cover decreased by up to 20% in winter or 80% relative to the control run, indi-
-------
E
E
1.6
1.4
1.2
1.0
.8
.6
.4
.2
0
-.4
-.6
-.8
-1.0
-1.2
-1.4
-1.6
M
M
J J
Month
N
Fj.g. ig. Change of precipitation (P) , evaporation (E) and runoff (R)
between the dot
for grid box 9.
between the doubled CO2 run and the control run as a function of month
-------
60
50
40
30
20
10
0
-10
-20
-30
-40
-50
-60
AEW
\
M
M
J J
Month
N
Fig. 16. Chanqe in total earth water (F.^(mm) drought index (DT) (times ten)
plant water stress (PWS), and snow cover (S.C.) (%) between the doubled
C02 run and the control run as a function of month for grid box 9.
-------
Table 2
Extremely
Dry
Grid Box 9
Control 4
2'd C02 2
Grid Box 16
Control 4
2'd CC-2 1
Grid Box 17
Control 0
2'd C02 0
Grid Box 14
Control 0
2'd C02 22
Severely Moderately Mildly Near Mildly Moderately Severely
Dry Dry Dry Normal Wet Wet Wet
8 13 13 38 8 8 7
4 9 11 16 9 11 10
4 13 13 26 19 8 8
8 12 17 23 19 10 12
0 3 21 37 18 7 11
2 16 18 40 12 5 5
3 13 10 33 18 6 10
13 12 12 20 18 0 2
Extremely
Wet
1
27
5
0
3
2
7
1
-------
38
eating that the warmer temperatures offset the increased wintertime precipita-
tion in their influence on snow cover. These diagnostics emphasize the warmer
and better quality of the climate in this area, with the hydrologic cycle least
affected in late summer and early fall.
Changes in a grid box with autumn dryness
Fig. 17 shows the changes in the hydrologic cycle in grid box 16. In the
control run this grid was much wetter than observed, by about a factor of three.
The seasonal cycle in rainfall change noted for the northwest region continues
here, in amplified form. The small change in April seen in Fig. 15 has expanded
into little change throughout the spring; the slight decrease in September has
expanded into decreases from August through October. This grid box is normally
wettest in the months of November through June, so the decrease in spring rela-
tive to winter and summer change the seasonal cycle somewhat; the dryness in fall
accentuates the prevalent pattern. In fact, the rainfall in September decreases
to only 20% its normal value. Also shown is the evaporation which has a similar,
although not identical, monthly variation. The runoff change therefore tends to
be small, with a slight increase in winter, and a slight decrease in fall.
Fig. 18 gives the monthly variation for the changes in diagnostics related
to the hydrologic cycle or vegetation. The earth water increases during the first
half of the year, and decreases during the second half. The first half increases
amount to 10% of the values in the control run, but the decreases exceed 50% in
some months (Sept. and Oct.) and thus, like the rainfall, indicate a substantial
drying of the climate in the fall. The drought index showed there was a tendency
for consecutive dry months to occur in the fall. Reference to Table 2 shows that
dry periods tended to occur 8% of the time in the control run, with very wet periods
13%; in the doubled C0£ climate there is little change in these percentages.
-------
1.6
1.4
1.2
1.0
.8
.6
.4
E -.2
-.4
-.6
-.8
-1.0
-1.2
-1.4
-1.6
M
M
J J
Month
N
Fig. 17. As in Fig. 15 for grid box 16.
-------
60
50
40
30
20
10
0
-10
-20
-30
-40
-50
-60
_L
M
M
J J
Month
N
Fiq. 18. As in Fig. 16 for grid box 16.
-------
41
The plant water stress increases throughout all the warm months, associated
with the greater temperatures; and the snow cover decreased in winter, with the
deficit reaching 17% of the grid box area in January which is an 85% decrease
relative to the control run. For the winter as a whole this grid box is covered
with snow only 5% of the time, compared to 15% in the control run.
Changes in a grid box with increased autumn wetness
The changes in the hydrologic cycle that occurred in grid box 17 are shown
in Fig. 19. In the control run this grid box had about 50% more rainfall than
observed. The results for this grid are quite different from those discussed
previously, especially for autumn. Large decreases in precipitation occur during
the spring and the first half of summer, with strong increases in late summer and
early fall. The precipitation decreases are on the order of 20% of the control
run values, while the increases are as much as 30%. (Remember that this grid
received about 50% more rainfall on the yearly average in the control run than
is observed). The runoff changes follow a similar pattern, although the percent-
age reduction relative to the control run reaches 50%. The evaporation changes
in a less predictable fashion.
The change of the total earth water is shown in Fig. 20, and emphasizes the
reduction of water except in the fall. The value in late spring has been reduced
by a factor of 33% compared to the control run. The drought index indicates the
tendency for consecutive dry months to occur from March through July. As shown
in Table 2 droughts did not occur at all in the control run in this grid box,
while wet periods occurred 14% of the time. In the doubled C02 climate droughts
now occur 2% of the time, with wet periods 7%. (These changes are likely within
the noise limit of precipitation variability.)
The plant water stress increases around the equinoxes and decreases in June
and July (along with evaporation, as in Fig. 19). The snow cover area decreases
-------
1.6
1.4
1.2
1.0
.8
.6
.4
"» 2
•^* •£•
o
0
I -2
-.4
-.6
-.8
-10
-1.2
-1.4
-1.6
AP
M
M J J A
Month
N
Pig. 19. As in Fig. 15 for grid box 17.
-------
60
50
40
30
20
10
0
-10
-20
-30
-40
-50
-60
M
J J
Month
N
Fiq. 20. As in Fig. 16 for grid box 17.
-------
44
by up to 14% in January; for the winter as a whole, this amounted to a 75% reduc-
tion in snow coverage relative to the control run.
Changes in a grid box with increased dryness
The final grid box to be looked at with a monthly resolution is grid box 14.
In the control run its precipitation was close to the observed amounts. The hydro-
logic cycle changes are shown in Fig. 21. In general precipitation decreases
during the last half of the year, with the decrease being most noticeable in early
summer and fall. The decreases in June, October and November amount to 20-30% of
the normal rainfall in the control run for these months. Runoff follows a similar
change; however the reduction in these months reaches 90%, indicating that runoff
has virtually disappeared due to the large precipitation decreases.
Fig. 22 shows the changes in related diagnostics. The earth water is seen
to decrease throughout the year, the result of the annually averaged drying; the
reduction is on the order of 25% of the control run values. The drought index
shows a consistent tendency towards a dryer climate throughout the year. Table
2 shows the change in distribution of dry and wet periods. In the control run
very dry periods occurred 3% of the time, with very wet periods occurring 17%;
in the doubled C02 climate dry periods now occur 35% of the time, with wet periods
only 3%. This indicates the dramatic change toward increased dryness experienced
in this grid box. As shown in the accompanying figure this change maximized in
the fall. In a different doubled C02 experiment run, that used a different repre-
sentation of sea ice (a key factor in climate models), analysis of this region
hydrologically using the Palmer drought index produced similar results, with the
largest effect in the spring. This difference demonstrates that the specific
results for any particular grid box and month are to some extent dependent upon
the unique parameterization representation of a climatic process in the model.
At best, the monthly results can be taken as indicative of overall tendencies for
change, not as reliable predictors.
-------
1.6
1.4
1.2
1.0
.8
.6
.4
o
•o
E
E -.2
-.4
-.8
-1.0
1.2
-1.4
-1.6
M
M
N
• Fig. 21. Same as Fiq. 15 for grid box 14.
-------
60
50
40
30
20
10
0
-10
-20
-30
-40
-50
-60
M
J J
Month
N
Fig. 22. Same as Fig. 16 for grid box 14.
-------
47
The plant water stress increases during the warm season, although the change
in summer is small relative to that seen at the other grid boxes discussed above.
The area covered by snow cover decreases by a small percentage during the winter,
although the change amounts to up to a 50% decrease relative to the control run.
Summary of Changes Over The Year
To summarize the results of this section: a review of the change in the
hydrologic cycle and related diagnostics for four grid boxes for each month in-
dicates that the variation differs from one location to another. Grid box 9
shows a general increase in the hydrologic cycle, with minimum change during late
summer and early fall. Grid box 16 had increased precipitation only near the
solstices, with a drying in fall. Grid box 14 had a generally drier climate,
with maximum precipitation decrease during fall. In contrast, grid box 17 expe-
rienced the southern considerable drying during spring and early summer, with
increased precipitation during late summer and early fall. Thus no general con-
clusion can be drawn about variations in monthly effects on the hydrologic cycle
of doubled C02; each grid must be reviewed separately.
If one combines the extremely and severely dry percentages for the various
grid boxes one can display the results, as in Fig. 22a, of the number of months
in a 30 year period in which there is potential for drought. The results are
shown for the four grid boxes, both for the control run and the doubled C02 run.
As can be seen, there is potential for substantial increase in a particular grid
box; again, these results should be viewed as indicating the types of changes
possible in the hydrologic cycle, rather than as applying to any particular grid
box. Fig. 22b shows the number of months of severely and extremely wet periods
- there is much variation from grid box to grid box, and the potential for strong
changes from the current climate.
Several other diagnostics did show common changes, due largely to the increase
in temperature experienced throughout the year. The plant water stress generally
-------
Grid Box
14
Fig. 22a. Number of severely or extremely dry months which would be expected
during 30 years in four different grid boxes for the control run (blank boxes)
and in the doubled CC>2 climate (blackened boxes) .
Fig. 22b. Same as 22a except for the number of months extremely or severely wet.
-------
49
increased during the warmer months (although even this has to be qualified for
the southern plains region), and snow cover decreased during the winter, by
large percentages relative to the control run.
4. Seasonal Variation
With so much variation occurring from grid box to grid box the question
arises as to what the seasonal pattern of changes looks like for the region as a
whole. We will review this briefly, with reference to the precipitation changes,
which, from the last section, seem to be indicative of the changes seen in other
areas of the hydrologic cycle, especially the earth water. We will show the re-
sults for the four seasons, defined as December through February (winter), March
through May (spring), June through August (summer), and September through November
(fall). This approach does not provide as much detailed information as looking
at the monthly variation, but it does allow patterns to be understood more easily;
it also indicates which grid box changes have larger regional applicability, and
thus more inherent confidence. The presentation will be insensitive to changes
when there are strong differences between early and late parts of a season, (for
example, the rainfall variation in grid box 16 during early fall and late fall).
Nevertheless, the structuring provided is worthwhile.
The rainfall changes in percent relative to the control run for the four
seasons are shown in Figs. 23-26. Fig. 23 and 24 show that most grid boxes
showed precipitation increases, or no change, during the winter and spring; a
notable exception is the southern plains region in spring, as discussed above.
The precipitation increases were greatest in the northern and western portions
as was true for the annual average. The southern regions showed little overall
change, although Florida was considerably drier during winter.
in summer (Fig. 25) the north and northwest continued to experience greater
rainfall, and the southeastern region also had much more rainfall. Little change
was experienced in the center of the country. In fall (Fig. 26) substantial de-
-------
>25%
11% to 25%
O 4-10% to -10%
0 -11% lo-25%
<-25
Fig. 23. Change in precipitation during winter between the doubled CO2 run
and the control run. The change is given in % relative to the control run,
indicated by the shading in the legend.
-------
>25%
% >o 25%
0+10% 10-10%
0 -11% to-25%
<-25
Fig. 24. Change in precipitation during spring between the doubled CO- run
and the control run.
-------
>25%
11% to 25%
Q-MO% 10-10%
0 -11% !o-25%
<-25
Fig. 25. Change in precipitation during summer between the doubled C02 run
and the control run.
-------
>25%
11% to 25%
Q-MO% Io-
0 -11% to-25%
<-25
Fig. 26. Change in precipitation during fall between the doubled CO- run and
the control run.
-------
54
creases in rainfall occur over much of the country, with the northwest being a
major exception. The more detailed review showed that even the Colorado River
grid received less rainfall through part of the fall. The areas which avoided
this loss of precipitation, parts of Canada, Mexico and the northwestern United
States thus experienced increased rainfall throughout the year. It should be
remembered that in all grid cells evaporation increases; therefore it would be
incorrect to assume that increases in precipitation mean greater water availabi-
lity, although decreases can be assumed to mean reduced water availability.
It is beyond the scope of this report to explore in great detail the reasons
for the changes noted above. In general the increased precipitation to the north
is associated with the warmer temperatures; more water can be held by a warmer
atmosphere, and with increased evaporation from the ocean, vapor is transported
northward and condenses at higher latitudes. This same influence is apparently
felt in the northwestern sections, where increased evaporation off the ocean is
advected onto land by the prevailing west winds. The precipitation change at any
grid box and in any season, however, results from a complex interaction between
temperature changes in that season, changes in moisture availability due to changes
in atmospheric dynamics and ground water changes (which are affected by what has
happened in preceding months), and changes in the atmospheric stability and jet
stream position which are associated with global scale circulation effects (e.g.
Hadley cell reaction, etc.).
Manabe et al. (1981) concluded in an experiment with 4 times the current
C02 level, that a substantial drying would occur in spring and summer in mid to
upper mid latitudes. The results shown in Figs. 24 and 25 do not indicate this
effect. Although the increase in ($2 amount is greater in the Manabe et al.
experiments, the global wanning was similar to that reported here. This again
serves 'to illustrate the uncertain nature of the results for specific areas and
seasons at this stage of global climate modeling.
-------
55
II. DIFFERENCE OF PRECIPITATION BETWEEN WARM AND COLD PERIODS IN THIS CENTURY
purpose
In the previous sections we investigated the changes in the hydrologic
cycle indicated by the model for the doubled C02 climate, and then discussed
some consequences of these changes. The hydrologic cycle was significantly
altered by the global wanning (of 4°C) associated with a doubling of 0)2; the
question addressed here is whether any evidence exists in the historical record
that the hydrological cycle changes as global temperature changes?
This section investigates whether changes in precipitation patterns can be
observed in the United States between periods of prolonged global warming and
cooling. The results of this analysis are useful for two related purposes:
• for testing the hypothesis that global warming will yield significant
changes in the patterns of hydrology
• as a potential benchmark for the pattern of change associated with a small
temperature increase, which will be useful in evaluating the planned tran-
sient experiment discussed in section III.
The results cannot be used to validate or invalidate the doubled C02 results
from the GCM experiment reported in Section I, however. That experiment represented
a climate in equilibrium, after a much larger wanning, not the transitional changes
in patterns that would lead to the new equilibrium. It is possible, for example,
that a region which would eventually get drier for doubled C02 may at first get
wetter as the "typical path" of storms tends to shift eastward as the world warms.
This could lead to the region becoming more in the center of the storm track for
a small C02 change, while later with a larger C02 change it could be on the
periphery.
-------
56
Past Research That Is Relevant
The possible influence of global warming on regional precipitation patterns
has been investigated recently by Wigley, et al. (1980), Williams (1980), Pittock
and Salinger (1982), and Jager and Kellogg (1983). Their approach was to compare
the precipitation in the warmest years with the average precipitation (Williams,
and Jager and Kellogg) or with the precipitation of the coldest years (Wigley,
et al. and Pittock and Salinger). However, as pointed out by Pittock and Salinger,
the use of individual years to represent warm and cold climatological periods is
open to serious question, since a one year transient response may have quite dif-
ferent characteristic than a warm period of several decades. This suggests that
investigating the changes in precipitation patterns between periods of prolonged
global warming and cooling would be worthwhile.
One other study should be noted. An analysis of long-term averaged data of
temperature and precipitation over the United States has recently been done by
Diaz and Quayle (1980). Their main interest was in the possible relationship be-
tween changes of temperature and precipitation and the changes in the corre-
sponding variances.
Summary of Results
We find that there is a significant increase in the precipitation during the
warm epoch (defined below) in the eastern half of the United states, particular-
ly in the Southeast. The changes in precipitation in this region have a signi-
ficant correlation with the changes in the global temperature. There are also
precipitation changes in Western Canada and over the basin of the Colorado river.
However the changes -in the latter region have low statistical significance and
thus the relationship between these changes and the global temperature is less
certain.
-------
57
Methodol ogy
We had to perform the following tasks:
• determine appropriate cold and warm periods for study based on global
temperature data,
• establish the area! extent over which the temperature and precipitation
trends of a given station can be taken as representative
• combine trends at individual stations to obtain regional trends.
This section reports on the procedures used to accomplish these tasks:
Cold and warm periods. Examination of the global temperature (see Fig. 27)
shows that the coldest period in the past century was 1880 to 1920, while 1940
and 1960 was a warm period. Because of data limitations in station coverage we
choose the period 1900 to 1920 as the cold period, thus yielding comparable sta-
tion coverage in the cold and warm periods. The global temperature difference
between the two periods was about 0.3°C. The number of stations within each
grid box of our medium resolution general circulation model over the U.S.A. are
shown on the Figs. 28 and 29 for the periods 1900-1920 and 1940-1960.
Establishing stations reliability. The difference between the number of
stations in each grid box during these two time intervals is generally small and
the distribution of stations is reasonably uniform. The average number of con-
tinental stations is 4.4 per grid box during the cold period 1900-1920 and 5.3
station per grid box during the warm period 1940-1960.
Correlation of trends at different stations. In order to gain an indica-
tion of how well these station distributions permit analysis of regional preci-
pitation trends, it is necessary to establish the area for which a given station
provides a representative precipitation trend. For this purpose we used the
correlation coefficient between the two sets of time series data, calculating,
-------
MMJAL AVERAGE
11 YEARS EUMiNG AVTUACE
-O.B
IB80. IB9D.
1970. 1980.
Fig. 27' . Global temperature (deviation from the mean) as a function
of year from 1880-1980.
-------
.sn
«6.90
39.15
31.33
\
?3.50U-
135-0 12'J 0 llrj.D 1U3.U
fo« tt€ PIRIOO igno - 19111 MAXIMUM NUMHTH i» MI'J'.ING
Fig.28 . Distribution of stations in the grid boxes of our GCM
during the period between 1900 and 1920, the cold period. Only
those stations were counted which had no more than five years
missing in the precipitation record.
-------
/'j.u
65.0
5*1.0
FOR IHE Pf.RIM) 1940 - 19611 MAXIMUM MM»M 0* HlbSIWi rl U1\,
Fig. 29. Same as Fig. 28 but for the period between 1940 and I960,
the warm period.
-------
61
the correlation coefficients between all pairs of stations in a given region.
This was done for the time series of both temperature and precipitation; the
results are shown in Figs. 30 and 31. To simplify the graphical representation,
it is restricted to four grid boxes of our GCM model which include the Colorado
river basin and to stations with separation of 500 km or less.
The correlation coefficients were calculated for the period 1880-1980 for
temperatures and for the period 1880-1979 for precipitation, with only those pairs
of stations having more than 20 years of overlap included. The dashed lines in
the figures connect stations between which the correlation is calculated, with the
symbol in the middle of the line indicating the value of the correlation coefficient
For example, [] means that the correlation coefficient is between 0.6 and 1. Thus,
[] indicates that we can state with a 99% or better confidence that there is a
linear association between the temperature (or precipitation) trends at the two
stations, with the linear association accounting for between 36 percent and 100
percent of the variation. The symbol A indicates that the probability for this
association is better than 0.9. In the case of the symbol o the probability of
this association is only better than 0.6, which means that there is a possible
linear relationship between the measurements at two stations, but such can not be
stated with high confidence and the relationship can not account for most of the
variation.
As can be seen in Fig. 30, correlation between the temperature trends of
nearby stations located in the Great Plains is excellent, and even in the Rocky
Mountains this correlation is good. In contrast, as can be seen in Fig. 31,
correlation between the precipitation time series measured at different station
is fair to poor, and this conclusion is the same over plains and over mountains.
The averaging method, described below, takes the poorer correlation for precipi-
tation trends into account by decreasing the area for which a station is assumed
to provide representative coverage.
-------
A' Q| '• OVtlW COI i>AI"J I-'IV
«7.0i
f-
l"
00
39.1
31.3
Fig. 30. Correlation for temperature trend time series between stations
located in the four grid boxes of the GCM. Colorado is in the center.
Dashed lines are drawn between the stations, while a symbol in the middle
of a line indicates the value of the correlation coefficient. Correlation
coefficients were calculated over the period between 1880 and 1980.
Maximum distance between the stations is 500 km.
-------
IATION OVLK1 COLuK'AUu '.-JVLR LJAMN
00
.,1
105.0
D l-l to -.61 a I-.S to -.HI O (-.1 to -.21 0 I-.J to .21 • 1.2 lo .Ml
1.1 to .61
95.D
1.6 to I
Fig. 31. Same as Fig. 30 but for precipitation trends.
-------
64
Combining station data to obtain regional trends. A method was developed to
use data from stations to extract a mean change of temperature and precipitation
for an arbitrary region. This procedure involves: (1) dividing the region into
a large number of cells, and (2) computing the temperature or precipitation in
each cell as a weighted average of all nearby stations, i.e., stations within a
circle of radius R, defined below. If there are no stations inside this circle
the temperature in the cell is undefined. The weight is taken to be a linear
function of the station distance, i.e.
1-d/R dR
The radius R defines a region inside of which the temperature or precipitation
is likely to have a trend similar to that measured at the station. The weight
function is similar to the structure function of Gandin (See P. Morel (1973)).
This linear dependence of weight on station distance was chosen because of its
simplicity and because large scatter of correlation coefficients shown on Fig. 32
does not warrant a more complicated function.
We use correlation coefficients between the time series measured at dif-
ferent stations to estimate the distance R which we assume to be equal to the
distance at which the correlation between these time series is statistically not
significant. The appropriate value for R can be estimated from Fig. 32, which
shows the correlation coefficient for 1000 pairs of stations. The solid line is
an empirical analytic fit to the correlation,
dV2
r "
18.3 + 0.7
where d is in kilometers. There is significant positive correlation out to
distances of ~500 km for precipitation.
-------
,.,
DI^TANCt 'KM. )
Fig. 32. Correlation coefficients between the precipitation trends
of stations located in the region shown on Fig. 28, as a function
of distance between the stations.
-------
CO
c c -
ax- xt
;xc- sot
11-
2BC-3DC
1 251-301
-t e •* t
-c:
t«
UB
Pig. 33. Histogram for the correlation coefficients for
precipitation trends between the stations as a
function of distance. The distances have been divided
into intervals of 500km as shown on the right of the
figure.
-------
67
This Is illustrated in another way in Fig. 33, which is based on all the
stations located in the United States. This figure was obtained by dividing
station separations into groups at intervals of 500 km. The resulting histograms
serve as an estimate to the probability density of the correlation coefficients.
These distributions are approximately symmetric. The average value for the
correlation coefficient is indicated by the arrow.
For stations separated by 500 km or less the average correlation is 0.42,
the standard deviation is 0.18, and more than 99% of all the correlation coef-
ficients are positive. For separations between 500 and 1000 km the average corre-
lation is 0.23, the standard deviation is 0.16 and 95% of the stations are posi-
tively correlated. The correlation becomes progressively poorer for the stations
separated by a larger distance.
We choose the radius R = 500 km, on the basis of the above results. This
yields an area coverage of 76% of the complete region shown on Fig. 28 for the
cold period (1900-1920) and an area coverage of 80% for the warm period (1940-
1960), i.e., these percentages of the region have at least one station within
the correlation radius R. The coverage of land area is significantly higher, as
illustrated in Figs. 34 and 35 in which circles of the radius R=500 km are drawn
around each station. This choice of the radius is also consistent with the
resolution of our GCM, which has grid box diameters of about 1000 km.
Results
Me used the above averaging procedure to compute precipitation trends in
all the grid boxes shown in Fig. 28. The time series in each box was then aver-
aged over the cold and warm periods. The difference between the average values
for warm and cold periods is plotted in Fig. 36, the units being mm/year.
Two distinct regions of precipitation change are apparent in Fig. 36. There
is a region of very large increase in precipitation located over the eastern part
-------
AUEA COVERAGE RY oHsmvio PMCIPMATION DATA
'
31.33
3.1 50
135.0
115.0
103 i!
85.0
75.0
63.0
95.0
PERIOD FROM 1900 TO 1920 MINIMUM NUMBER OF YEARS 15.
DISTANCE 500 KMb.
Fig. 34. Area covered by the stations during the cold period.
55 0
-------
i -
4—
Ol
31.33 -
23.50
135.0
AREA COVERAGE BY OBSERVED PWC1PITATION DATA
115.0 IOb.0
95.0
05.0 75.0 65.0 55.0
PERIOD FROM 19*0 TO I960 MINIMUM NUMBER OF TEARS 15.
DISTANCE 300 CT15.
Fig. 35. Area covered by the stations during the warm period.
-------
0
a
DIFKLHI.NCE IN PM C IPI TAT ION It fWLl N WAKM AM) CCK.U ftfj
•jo.goprsg.- ~^r
rj C*' Tl-t. liLOMAL Tl MPTNATiJRi;
^ ± J
«.07
I2O.O 110.0 100.0
COLO PERIOD 1900 - 1920 WAM1 PERIOD 1940 - I960
90-0
80.0
70.0
60.0
Fig. 36. Precipitation during warm period minus precipitation during
cold period (mm yr ).
-------
71
of the U.S.A. and over Canada, with two apparent maxima: one over Bermuda and a
second over the southern tip of Hudson Bay, in Canada.
The remaining large region has very little change in precipitation between
the warm and cold periods. Located west of the Mississippi valley, it covers
the region of the Great Plains and Rocky Mountains. Two additional, less signi-
ficant regions of change appear in Fig. 36. One of these, located in the
northeast, shows an increase in precipitation. It could be an extension of the
region of precipitation increase located in the eastern part of the U.S.A. and
connected to it through northern Canada. The second is a region of decreased
precipitation, which is of a particular interest because it includes the
Colorado river basin.
Comparison to results of other researchers
In order to compare our results with those of Diaz and Quayle we computed the
precipitation change between the time periods (1955 - 1977) and (1895 - 1970), which
are their time periods C and A. The results of this computation are shown on Fig. 37.
Their results for precipitation change have a broad similarity with ours. Thus both
results show a precipitation increase in the Southeast and Northwest, and a pre-
cipitation decrease over the remaining part of the western coast. We also show a
precipitation decrease over the Midwest, although our decrease appears to be smaller.
There are however some significant differences between the two results in the
Northeastern part of the United States. Some of these difference are partially
due to the fact that their resolution of 5°x5° is smaller than our resolution of
8°xlO°. For instance, instead of their precipitation decrease over the northern
part of Maine and precipitation increase over the New York state, we have a broad
region where the precipitation increase is small and almost constant. The dif-
ference between our results and theirs is further accentuated by differences in
date bases and averaging procedures used by us and by Diaz and Quayle. While we
-------
i1 ' i i i n I I .,i 11 ii I., iwi ' 11 ' n» 'I •'. li I'M i V- l'.*-i
,,,
•i i.ii
3-1.
IJU.U -|^
-------
73
use the measurements from individual stations which we then average over the grids
shown on Fig. 28, they use average precipitation for each of the 48 United States
which they then interpolate over their 5°x5° grid.
Statistical significance of our results
To study the statistical significance of the precipitation changes as a
function of global temperature we first isolated the region of significant pre-
cipitation change by applying statistical significance tests to the precipita-
tion changes in individual grid boxes. First, in order to determine the appro-
priate test for the precipitation change between the warm and the cold periods,
we used the F test to find out if the variances of the precipitation time series
in each grid box were the same during the cold and the warm periods. This hypoth-
esis had to be rejected only in three grid boxes at the 90% confidence level.
Therefore we assumed that variances were the same during the warm and cold periods
and applied the appropriate one sided t test to the precipitation time series to
isolate regions of statistically significant precipitation change. The criterion
was set at the 90% confidence level. The four regions with significant precipi-
tation change are shown on Fig. 38.
Next we examined the time series of the precipitation for these four regions.
The time series of precipitation over the eastern United States is particularly
interesting. The eleven year running mean (solid line in Fig. 39a) is very sim-
ilar to the global temperature trend as computed by Hansen et al. (1981), i.e.,
it increases during the period 1910 and 1940, flattens out during the period be-
tween 1940 and 1960 and has maxima in 1940 and 1960. The correlation coefficient
between the eleven year running means of temperature and precipitation is 0.85,
and even if we assume that the time series have only eight data points (i.e. one
point per decade), then the correlation between the precipitation in the eastern
United States and the global temperature is significant at a 99% confidence level.
The precipitation time series in the region of little precipitation change (Fig.
-------
7 128 145
Fig. 38. Same as Fiq. 36 with the change for each grid box (mm/year)
Only changes significant at the 90% confidence level are shown.
-------
•(a)
JDD.D
0 1 .. L
!
D -i. 1 L J
10D.D-—
•2SD.
INCREASE
_4.1U^M 14 4?—^ -
4 || ij i« ;/ ! u
H ± n
(b)
25D.D
-feelow
•UNC-HAN6ci> -WEC-j P M AT JON
j t
1
1ST
19IC.
13EC J93G
19SD
196C.
1S7D-
ICC.
DATt
Fig. 39. (a) Averaged precipitation over the region of
significant precipitation increase, in the eastern portion
of the continent.
(b) Averaged precipitation over the region with
no precipitation changes.
-------
76
39b) has negligible correlation with the global temperature trend (correlation
coefficient -0.1). The precipitation time series of the Northwestern region is
very noisy (Fig. 40a). However it does shows an increase during the period be-
tween 1910 and 1960; the correlation coefficient between the 11 year running means
of global temperature and regional precipitation is 0.6, which is marginal since
this represents a correlation at the 88% confidence level. Finally the precipi-
tation curve over the Colorado basin (Fig. 40b) shows a decreasing long range trend.
It is negatively correlated with the global temperature trend, with correlation
coefficient -0.5. The sense of the correlation is consistent with the slight ten-
dency toward drying in this region with the increased temperatures in the doubled
CO? experiment discussed in the previous section of this report, however, the
correlation is too weak to permit firm comparison.
By comparing our results with those by Wigley et al. (1980) we can see that the
change in the Eastern United States is broader and includes the region of the preci-
pitation increase computed by Wigley et al. on the basis of individual warm years.
It also appears that their precipitation increase along the West coast of the United
States is a warm year transitory phenomenon not reflected in the climatological mean.
Finally, we note that the precipitation change found in the eastern United
States is quite large. Corresponding to the change of 0.27°C in the global tem-
perature, the precipitation change is 11 cm/yr which represents 10% of the total
precipitation in the region. If the global temperature trend and rainfall changes
are indeed causally connected, the implication may be substantial; note that the
temperature difference between these two periods is small compared to the pro-
jected warming during the next several decades as a result of C02 and trace gas
warming. Closer examination of these relationships as the current warming trend
progresses is obviously warranted, as well as comparison of 3-D global climate
model simulations with the observed patterns of precipitation change. The pre-
paration for one such modeling experiment is described in the next section.
-------
(L
e
19CD.
J97D.
198:
Fig. 40. (a)Averaged precipitation over the northwestern region.
(b)Averaged precipitation over the Colorado River Basin,
-------
78
III. Future Refinements to Climate Models
In the previous two sections we discussed the difference between the
climate in the doubled ($2 atmosphere and the present climate, and the observed
differences between warm and cold climates in the past century. The results in
some instances seem contradictory. For example, in the slightly warmer histori-
cal climate, the east coast received more rainfall, while in the much warmer
doubled C02 climate, the east coast was drier. As mentioned earlier, however,
these results are not necessarily contradictory ~ the doubled C02 experiment
results do not describe the transitional phases of climate, they focus only on
equilibrium conditions after C02 has doubled. In order to adequately anticipate
transitional climates and decide upon the appropriate responses it will be
necessary to know how the system will react ^s_ the climate warms, with magnitu-
des of wanning intermediate between the historical rise and the projected
doubled C02 equilibrium effect. This requires that GCMs be used for experiments
in which carbon dioxide (and other greenhouse gases) is allowed to increase gra-
dually with time, and the response of the system determined as a function of
time.
The Goddard Institute for Space Studies is in the process of doing its
first "transient" run, in which we start with the climate of 1958 and alter the
atmospheric composition in a continuous manner. It is important to emphasize,
however, that one run with one model will not provide a statistically reliable
solution to this question. Many such runs with different models will be
required. Furthermore, uncertainties associated with global climate modeling,
particularly of physical processes such as ocean atmosphere coupling (whose
importance is illustrated by El Nino events), cloud dynamics, sea ice represen-
tation, and accurate hydrological/biological interactions, all need to be exa-
mined. In addition, as is evident in the presentation so far, model resolution
-------
79
would need to be increased if answers are to be provided on a finer spatial
scale. Nevertheless, this first transient run will allow us to begin examining
the evolution of the warming. Simulations with increasingly realistic models
will be possible in the future, if appropriate model development and related
observational studies are carried out.
Use of GCMs holds much promise as a tool for projecting future climatic
change from increases in greenhouse gases. These models, however, are still in
their development stage and require considerable refinement if they are to
achieve their full potential. This chapter highlights critical physical pro-
cesses that need to be better represented if we are to reduce current uncertain-
ties in estimating the responsiveness of the climate system to increases in
greenhouse gases. It also discusses other possible refinements to the models
aimed at providing annual or "transient" changes in climate as greeenhouse gases
increase over time and at improving estimates for individual grids.
Three kinds of uncertainties affect the reliability of output from general
circulation models: how responsive will the climate system be to increases in
greenhouse gases (eg. how much will the temperature rise for a particular
increase in 0)2); how will regional climates change; and how will changes occur
over time given specified increases in various greenhouse gases.
Sensitivity of Climate System
The range of estimates from GCM's is quite large. For a doubling of
atmospheric 0)2 levels, Manabe et al. estimate a 2°C rise, Washington et al.
estimate a 4.2°C rise, and Hansen et al., a rise of 4.1°C. Differences between
the high and low estimates can be attributed to how the GCMs simulate clouds,
and secondarily to initial sea ice conditions.
Cloud behavior is particularly difficult to model. Existing databases and
theory do not provide an adequate basis for defining their appropriate treat-
-------
80
ment. Thus, additional research will be needed to develop a more comprehensive
database and to improve scientific capability to model clouds more reliably.
Until this analysis is completed, large uncertainties will persist in calcu-
lating thermal responsiveness of the climate system to increases in greenhouse
gases.
Improvements will also be needed in knowledge about sea ice and ocean dyna-
mics that could affect such ice. Since sea ice reflects large amounts of
sunlight, changes in it could be important to the total radiation absorbed by
earth, and thus the climate system's responsiveness to perturbation.
Uncertainties Concerning Regional Climates
Regional climates are sensitive to patterns of general circulation. In
particular, sea surface temperature is a key factor influencing the movement of
weather systems. For example, the 1982-3 Nino involved a very large change in
sea surface temperature that caused weather in many parts of the world to follow
highly unusual patterns. The cold weather (for continental U.S.) of 1976 and
1977 was possibly influenced by unusually warm water off the Bering Staights
that may have influenced the polar jet streams to first go north, then dip south.
Global warming on the scale anticipated from increases in greenhouse gases
is likely to alter ocean circulation and thus sea surface temperatures. Current
GCMs generally have very simple oceans, which only passively respond to changes
in heat. Ocean circulation is not ordinarily modeled as a dynamic process.
Thus, GCMs underestimate variations in sea surface temperature that could alter
weather patterns, failing to adequately predict this vital aspect of the climate
system. Until changes in ocean circulation are modeled so that they respond to
temperature, the regional projections produced by GCMs will be less reliable
than desired. Research must be expanded in this area to collect data that can
be used to incorporate more realistic treatment of the oceans.
-------
81
Of less importance, but still of significance, will be improvements in
modeling ground hydrology and the response of plant life to alterations in cli-
mate and to the photosynthetic effects of rising 003. Both the water holding
capacity of land and the efficiency of water use by plants will be altered by
rising COg and climate change. Since these are important determinants of
regional climate, this feedback loop needs to be modeled in GCMs. In the long
term GCMs will need representations of these processes or they will not be able
to forecast regional climate change accurately.
Timetrend of Temperature Change
Finally from the perspective of users the most important aspect of climate
change will be the transitions through time, not the "ultimate" climate after
C02 has doubled. Appendix A discusses key parameters used in an initial tran-
sient run underway at the Goddard Institute for Space Studies.
-------
82
APPENDIX A
ESTIMATING ANNUAL CHANGES IN TEMPERATURES
The Goddard Institute for Space Studies is now in the process of performing
a transient run with its GCM in which the year by year changes in climate are
simulated for past and expected changes in greenhouse gases and aerosols.
Trace Gas and Aerosol Perturbation
In this section we describe the parameterizations developed for the time
evolution of trace gases for the period 1958 to 2030 which will be used in our
transient global climate model experiment. First we consider the parameteriza-
tion for C02 and other trace gases such as Fluorocarbon 11 (Fll), Fluorocarbon
12 (F12), Methane (CH4) and Nitrous Oxide (NeO). Since increase in these gases
is slow compared with their mixing time over the globe, we can take these gases
as being uniformly mixed.
For the COg concentration between 1958 and 1980 we use the annually averaged
amount of 0)3 at Mauna Loa as measured by Keeling et al. (1982). Projection of
the C02 concentration into the future is extremely uncertain, depending espe-
cially on the postulated growth of energy demand and the availability of dif-
ferent fuels. For these reasons it is worthwhile to consider a range of scenar-
ios, for example, as discussed by Hansen et al. (1981). Table 3 shows the COe
trends obtained with their 'fast growth' (~3 percent/year), slow growth (~ll/2
percent/year) and no growth energy scenarios, as well as the C02 trend which we
will use in our 3-0 transient experiment. The latter was supplied to us by EPA,
being constructed on the basis of a moderate growth of energy use with changes
of the airborne fraction of COg accounted for on the basis of a simple carbon
cycle model (loc. cit.). This EPA scenario is quite similar to the slow growth
scenario of Hansen et al. (1981).
-------
83
Table 3. Different C02 Scenarios
Year
1980
1990
2000
2010
2020
2025
2030
2040
2050
EPS
Scenario
339
352
373
396
420
433.75
448
478
513
Fast
339
356
384
429
504
554
614
753
893
Growth Rate
Slow
339
354
373
395
422
438
453
490
532
NO
339
352
366
378
394
401
408
424
440
Minor gases included in our transient run are Fll, F12, Cfy and
Concentrations of fluorocarbons were obtained using estimates for their release
for the period 1950 to 1980 by the Chemical Manufacturer Association (1981).
The annual release of fluorocarbons during the years 1980-2030 is assumed to be
constant and equal to the average value of their release during the decade
1971-1980, i.e. 283.5 millions of kilograms of Fll per year and 367.5 millions
of kilograms of F12 per year.
The annual release of fluorocarbons for the period prior to 1950 was esti-
mated by assuming this release to be a linear function of time.
Annual release = A|<(years -
Here k=ll,12 and stands for fluorocarbon Fll and F12. The proportionality
constants A|< and the years Yfc » at which the annual release is equal to zero,
-------
84
were determined in such a way that the annual release in 1950 and the total
amount released up to 1950 calculated using the expression above were equal to
the values reported in the CMA report. The residence times of fluorocarbons are
not well known, the main sink being the stratosphere. Here we estimate these
residence times t^ to be 75 years for Fll and 150 years for F12. Thus, for the
year m the concentration C|c(m) as a function of the annual release Rfc(m) is
given by:
m -m-
Ck(m) = fk I e Rk(e) . (A)
e=1940
According to our estimate the annual release for both fluorocarbons during
the year 1940 was zero. The constant fk relates the mixing ratios Ck in ppbv of
fluorocarbons to their annual release Rk (in millions kg/year). These constants
of proportionality were determined by comparing the computed concentrations with
observed globally averaged values of Fll and F12 for the years 1977-1979 as
reported by NOAA (1979) and (1980) in Geophysical Monitoring for Climatic Change
No. 7 and No. 8. Global average concentrations were computed from the results
of measurements of concentrations at five stations. Locations of these stations
and the measured values are summarized in Tables 4 and 5. We assumed the con-
centrations to be zonal ly uniform and fitted expresssions
5 ,
"
e=l
ae sn
to the data in Table 4. Here <|> denotes latitude and x longitude. The global
averages are shown in the last column of Table 4. The constants of propor-
tionality fk were obtained by fitting expressions (A) to the data, yielding
= 4.6395.10-5 and f12 = 5.3279.10-5 ppb/(mill ions kg/year).
-------
85
Table 4. Concentrations of Fluorocarbons (ppbv)
Year
1977
1978
1979
1977
1978
1979
BRW
0.159
0.172
0.182
0.292
0.302
0.301
Station
NWR
MLO
SMO
SPO
Fluorocarbon 11 (CC^F)
0.155
0.168
0.175
0.148
0.162
0.174
0.140
0.153
0.164
0.139
0.154
0.175
' GLOBAL
0.145
0.159
0.171
Fluorocarbon 12 (CC12F2)
0.275
0.296
0.301
0.270
0.291
0.296
0.256
0.273
0.276
0.248
0.271
0.306
0.262
0.282
0.290
Table 5. Locations of the Stations
Name
Point Barrow
Niwot Ridge
Mauna Loa
American Samoa
South Pole
Abrev .
BRW
NWR
MLO
SMO
SPO
Longitude
130.60°W
105.63°W
155.58°W
170.56°W
24.80°W
Lattitude
70.32°N
40.05°N
19.53°N
14.25°S
89.98°S
The corresponding fluorocarbon scenarios are shown in Fig. 41. Scenarios
for Methane and Nitrous Oxide were constructed using estimates of Lacis et al.
(1981). Concentrations for Cfy and N£0, in 1958, are respectively 1.4 and 0.295
ppmv. Between 1958 and 1970 the concentration of N20 remains unchanged while that
for CH4 increases by 0.6%/yr. During the decade between 1970 and 1980 concentrations
for CH4 and NgO change respectively by 0.96% and 0.2%. Finally, after 1980, we
estimated the CH4 increase to be 1.5% and N20 increase to be 0.3%. The time
variations of Cfy and N20 are shown in Figs. 42 and 43.
Atmospheric aerosols produced by volcanic explosions are shortlived, lasting
no more than a few years. Also the spread of volcanic dust in space is highly
-------
FLUOROCARBON5 SCENARIO
1.00
0.90
u.ao
0.70
0.60
0.90
0.40
0.30
0.20
0.10
o.tn
19
•
» 1
;
i
1
/
\
^
~z^ \
„.-'• \ X""
••• -^f
.-•• ^-^ i i
i i
,'
.•'
,/'
x'
7
^*-''
,*'
•;JT*"
_
X
s
rs'
^•~'
s
s'
./
^^
..••'
fjr^'~
BO 1970 1980 1980 2000 2010 2O20 20
DATE
0.90
o
O.BO o
r
o
°-"m
z
0.60 -1
T3
>
0.30 ^
o.« o
z
0.30 ^
-o
0-20 01
0.10
0.00
30
Fig. 41. Fluorocarbon scenarios. The observed values are also
shown by circles for Fll and squares for F12.
-------
METHANE SCENARIO
3.60
3.20
2.80
2.«0
2.00
1.60 -
1.20
I960 1970
—.3.60
3.20
ir.BO
' >
['I
z
I
II
' •
z
II
I)
I
Fig. 42. Scenario for methane.
-------
NITROUS OXIDE bCENARIO
0.330
0.330
0.320
0.310
0.300
0.290
I960 1970 I960 1990
0.3X
0.320
I
'I
I
D
i-
0.310
ii
D
3
0.300 ~
2000
DATE
2010 2020
—' .290
2030
Fig. 43. Scenario for nitrous oxide
-------
89
nonuniform. For instance: dust from volcanoes that erupt in the polar regions
tends to be confined to these regions; also volcano Agung erupted near the
equator, but most of its dust went into the Southern Hemisphere; more recently
the stratospheric aerosols from the volcano El Chichon remained confined for a
long time to a latitudinal belt between the equator and 30°N. Therefore, the
specification of the volcanic aerosol distribution has to be done at least on a
monthly basis and the optical depth has to be a function of position over the
globe. A list of volcanos since 1956 that are thought to have produced substan-
tial aerosol amounts is summarized in Table 6.
Table 6. List of Climatologically Significant Volcanoes
Name Date Latitude Longitude Lambs DVI Scaling
Factors
X108
Bezymjannaja
Volcanos in Chil
Gunung Agung
Awu
Fernandina
Fuego
Soufriere
St. Helens
Alaid
El Chichon
3/30/1956
5/20/1960
3/17/1963
8/12/1966
6/12/1968
10/18/1974
4/17/1979
5/18/1980
4/27/1981
4/2/1982
56.0°N
39-45°S
8.5°S
3.5°N
0.5°S
14.3°N
13.3°N
46.2°N
50.8°N
17.3°N
160. 5°E
72.5°W
115. 5°E
125. 5°E
92.0°W
90.5°W
61.2°W
122. 2°W
155. 5°E
92.8°W
30 0
100 ?
800 846
150 162
75 123
83
0
0
0
?
-------
90
We used our medium resolution 9-layer general circulation model ( Hansen et
al. (1983) ) and our tracer model ( Russell and Lerner (1981) ) to determine the
dispersion of the volcanic dust over the globe for each of these volcanoes.
First, winds for the entire globe and for several years were generated by the
General Circulation Model. Then the same amount of dust from every volcano
shown in Table 6 was injected into the stratosphere of the model at the location
and time indicated in the table, and the tracer model was used to study the dif-
fusion of the volcanic cloud over the globe. Dust entering the troposphere was
assumed to be instantly removed. Therefore the exchange of air parcels between
the stratosphere and troposphere served as a sink for the stratospheric dust.
The main scattering and absorbtion of solar radiation is by sulfuric acid
aerosols, which are produced from gas-to-particle conversion of the gases S02,
H2S, and CSO injected by volcanos. The conversion time in the stratosphere be-
tween these gases and sulfuric acid is estimated to be of the order of six
months. Therefore, in order to take into account this conversion time, we
multiplied the computed density of the volcanic dust by a function
t/T tT
where t is the time elapsed since the eruption and T is assumed to be equal to
six months.
In order to determine the optical depth due to the aerosol injected by each
volcano we used the solar irradiance transmission measured at Mauna Loa for the
period 1958 to 1979 (NOAA 1979). The T so obtained contains scattering, due to
Rayleigh aerosol as well as absorption by different atmospheric gases. In order
to isolate the optical depth due to the volcanoes, we used the fact that no
major eruption affecting the atmospheric transmission at Mauna Loa occurred in
the several years prior to Agung. Thus by subtracting the average value for the
-------
91
optical depth between the years 1958 and 1963 the effect of scattering due to
all other components of the atmosphere but the volcanic aerosol was removed.
This residual optical depth is plotted in Fig. 44. Eruption dates of the volca-
nos listed in the table are indicated by arrows on the figure. Volcanos in
Chile are located too far to the south of the equator to have a significant
impact at Mauna Loa. However, Agung, Awu, and Fuego are clearly visible.
Although any signal from Fernandina is barely discernable in the noise, it was
introduced in order to represent the slow decrease in optical depth after the
explosion of Awu. The scaling factors between the computed density and the
optical depth to be used in the model run were obtained by imposing the con-
dition that the computed and observed optical depths at Mauna Loa should be
equal.
The tracer model was integrated for 36 months after the Agung eruption at
which point the volcano Awu exploded. The computed optical depth due to Agung
was subtracted from the observation and the computed density due to Awu as well
as the residual observed optical depth were integrated for 22 months, when the
volcano Fernandina exploded. The scale factor for Fernandina was obtained in
the same way as the scale factor for Awu. The optical depth due to the volcano
Fuego was not contaminated by previous eruptions, and thus the scale factor was
simply obtained by integrating the computed and observed T over a period of 32
months. The computed optical depth at Mauna Loa is plotted as a solid line in
Fig. 44. The agreement between observed and computed optical depths for Awu and
Fuego is very good, and the signal from Fernandina is buried in the noise.
However, in the case of Agung the observed maximum in the optical depth is
delayed by about a year from the computed maximum. The imprecision in repre-
senting the time dependence of aerosol opacity after Agung may be due to
unrealistic transport by the tracer model; however, the model did result in 20%
-------
0.04
-0.01
1958 I960 1962 1964
1966 1968
Date
1970 1978 1974 1976 1978
Fig. 44. Optical depth as measured at Mauna Loa Observatory (dashed line)
after removal of average value from 1958 to 1963. Also shown is
the computed optical depth (solid line). The arrows refer to the
volcanos listed in Table 6.
-------
93
of the aerosol going into the Northern Hemisphere in agreement with obser-
vations. Perhaps the assumed 6-month decay time is inaccurate for an eruption
as large and which penetrated as high as Agung did.
According to M.P. McCormick et al. (1981) the effect of the explosion of
Soufriere on climate should be negligible, and as estimated by Jager et al.
(1982) the same is true for the two volcanos Mt. St. Helen and Alaid. Thus, the
only volcano to be included since 1974 is El Chichon, for which the data just
now are becoming available.
Increasing Our Confidence in the Result
As noted in the first part of this section, this run will provide only one
estimate of the climate change due to the gradual increase in carbon dioxide and
other trace gases. Much work needs to be done to increase our confidence in the
result. The elements of the climate system that are currently poorly modeled,
or not modeled at all, must be investigated closely, both theoretically and
observationally. The highest priority must be given to understanding the ocean
circulation, and how it may respond to the climate change. This will require a
large increase in observations of the ocean, from both ships and satellites, for
it is impossible to confidently model a system when it is uncertain whether the
model is producing realistic simulations. Programs are already under way to
produce a cloud climatology data set for the same reasons. Understanding of
these and other parts of the geophysical system will depend on increased obser-
vational capability of both the particular aspect and of its interaction with
the atmosphere.
We may expect that firm results about the climate's sensitivity will yield
only to long-term analysis. The results obtained in the above sections, and
those to be produced in the transient experiment, are only a beginning of
attempts to understand and model how climate will change in the years to come.
-------
94
References
Arakawa, A., 1972: Design of the UCLA General Circulation Model. Tech. Rep.
No. 7, Dept. Meteor., University of California, Los Angeles, 116 pp.
Berry, F., E. Boll ay and N. Beers, Eds., 1973: Handbook of Meteorology.
McGraw-Hill, 1068 pp.
Chemical Manufacturers Association (CMA): World Production and Release of
Chlorofluorocarbons 11 and 12 through 1980. CMA Report, July 29, 1981.
Diaz, M. F. and Quayle, R. G., 1980: The Climate of the United States since
1895: Spatial and Temporal Changes, Mon. Wea. Rev., 108. p. 246-266.
Hansen, J., Johnson, D., Lacis, A., Lebedeff, S., Lee, P., Rind, D., Russell,
G., 1981: Science. 213, 957-966.
Hansen, J., Russell, G., Rind, D., Stone, P., Lacis, A., Lebedeff, S., Reudy, R.
and Travis, L., 1983: Efficient Three-Dimensional Global Models for Climate
Studies, Models I and II. Mon. Wea. Rev. Ill, in press.
Jager, H., Reiter, R., Carnuth, W., Funk, W., 1982: The Aerosol Increase with
Stratosphere due to Volcanism During 1980 and 1981, observed by Lidar at
Garmish-Parturkirchen, presented at The Symposium on the Application of
Lidar to Radiation and Climate. IAMAP Third Scientific Assembly, Hamburg,
Federal Republic of Germany, 1981.
Jager, J. and Kellogg, W.W., 1983: Anomalies in Temperature and Rainfall
during Warm Arctic Seasons, Climatic Change 5^, 39-60.
-------
95
Keeling, C.D., Bacastaw, R.B., Whorf, T.P., 1982: Measurements of the
Concentration of Carbon Dioxide at Mauna Loa Observatory, Hawaii, pp
377-385, in Carbon Dioxide Review, 1982, W.C. Clark, Ed., Oxford Univ.,
Press, New York.
Korzoun, V., et al., 1977: Atlas of World Water Balance. The UNESCO Press,
Paris.
Lacis, A., Hansen, J., Lee, P., Mitchell, T., Lebedeff, S., 1981: Greenhouse
Effect of Trace Gases, 1970-1980, Geophysical Research Letters, _8,
1035-1038.
Lamb, H.H., 1970: Volcanic Dust in the Atmosphere, Phil., Trans. Roy Soc.
London, 266B. 425-575.
McCormick, M.P., Kent, G.S., Yue, G.K., Cunnold, D.M., 1981: Sage Measurements
of The Stratospheric Aerosol Dispersion and Loading From the Soufriere
Volcanoes, NASA TP-1977.
Miller, J., Russell, G. and Tsang, L., 1983: Annual Oceanic Heat Transports
Computed from an Atmospheric Model, Dynamics of Atms. and Oceans, _7_, 95-109,
Morel, P., 1973: Space and Meteorological Data Analysis, p. 471 - in Dynamic
Meteorology, P. Morel, Ed., D. Reidel Publishing, Co., Boston, U.S.A.
National Academy of Sciences, 1979, 1981: Carbon Dioxide and Climate, A
Scientific Assessment, Washington, D.C.
NOAA, 1979: Geophysical Monitoring for Climatic Change No. 7., Summary Report
1978, B.G. Mendonca, Ed., Environmental Research Laboratories, Boulder,
Colorado.
-------
96
NOAA, 1980: Geophysical Monitoring for Climatic Change No. 8., Summary Report
1979, G.A. Herbert, Ed., Environmental Research Laboratory, Boulder,
Colorado.
Pittock, A.B., Salinger, M.J., 1982: Towards Regional Scenarios for a
C02-Warmed Earth, Climatic Change 4_, 23-39.
Russell, G.L. and Lerner, J.A., 1981: A New Finite-Differencing Scheme for the
Tracer Transport Equation, J. Appl. Meteor, JO, 1483-1498.
Schutz, C. and Gates, W., 1971: Global Climatic Data for Surface, 800mb,
400mb, Reps. R-915-ARPA, R-915/1 ARPA, R-1029-ARPA, Rand Corp., Santa
Monica.
Wigley, T.M.L., Jones, P.O. and Kelly, P.M., 1980: Scenario for a Warm,
High-C02 World, Nature, 283, 17-21.
Williams, J., 1980: Anomalies in Temperature and Rainfall During Warm Arctic
Seasons as a Guide to the Formulation of Climatic Scenarios, Climatic Change
£, 249-260.
•u.s. aornoKart PBUTUO oma i 1904 0-421-062/518
------- |