£EPA
United States
Environmental Protection
Agency
Policy, Planning,
And Evaluation
(PM-221)
EPA-230-05-89-051
June 1989
The Potential Effects
Of Global Climate Change
On The United States
Appendix A
Water Resources
-------
THE POTENTIAL EFFECTS OF GLOBAL CLIMATE CHANGE
* ON THE UNITED STATES:
APPENDIX A - WATER RESOURCES
Editors: Joel B. Smith and Dennis A. Tirpak
OFFICE OF POLICY, PLANNING AND EVALUATION
US. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, DC 20460
MAY 1989
Text Printed on Recycled Paper
-------
TABLE OF CONTENTS
APPENDIX A: WATER RESOURCES
PREFACE iii
INTERPRETATION OF HYDROLOGIC EFFECTS OF CLIMATE CHANGE IN THE
SACRAMENTO-SAN JOAQUIN RIVER BASIN, CALIFORNIA 1-1
Dennis P. Lettenmaier, Thian Yew Gan, and David R. Dawdy
METHODS FOR EVALUATING THE POTENTIAL IMPACTS OF GLOBAL
CLIMATE CHANGE: CASE STUDIES OF THE STATE OF CALIFORNIA AND ATLANTA,
GEORGIA 2-1
Daniel P. Sheer and Dean Randall
THE IMPACTS OF CLIMATE CHANGE ON THE SALINITY OF SAN
FRANCISCO BAY 3-1
Philip B. Williams
EFFECTS OF CLIMATE CHANGES ON THE LAURENTIAN GREAT LAKES
LEVELS 4-1
Thomas E. Croley n and Holly C. Hartmann
IMPACT OF GLOBAL WARMING ON GREAT LAKES ICE CYCLES 5-1
Raymond A. Assel
POTENTIAL CLIMATE CHANGES TO THE LAKE MICHIGAN THERMAL
STRUCTURE 6-1
Michael J. McCormick
THE EFFECTS OF CLIMATE WARMING ON LAKE ERIE WATER
QUALITY 7-1
Alan F. Blumberg and Dominic M. Di Toro
IMPACTS OF GLOBAL WARMING ON RUNOFF IN THE UPPER CHATTAHOOCHEE
RIVER BASIN 8-1
David K. Hains and C. F. Hains
POTENTIAL IMPACTS OF CLIMATE CHANGE ON THE TENNESSEE
VALLEY AUTHORITY RESERVOIR SYSTEM 9-1
Barbara A. Miller and W. Gary Brock
11
-------
PREFACE
The ecological and economic implications of the greenhouse effect have been the subject of discussion within
the scientific community for the past three decades. In recent years, members of Congress have held hearings
on the greenhouse effect and have begun to examine its implications for public policy. This interest was
accentuated during a series of hearings held in June 1986 by the Subcommittee on Pollution of the Senate
Environment and Public Works Committee. Following the hearings, committee members sent a formal request
to the EPA Administrator, asking the Agency to undertake two studies on climate change due to the greenhouse
effect
One of the studies we are requesting should examine the potential health and environmental
effects of climate change. This study should include, but not be limited to, the potential impacts
on agriculture, forests, wetlands, human health, rivers, lakes, and estuaries, as well as other
ecosystems and societal impacts. This study should be designed to include original analyses, to
identify and fill in where important research gaps exist, and to solicit the opinions of
knowledgeable people throughout the country through a process of public hearings and
meetings.
To meet this request, EPA produced the report entitled The Potential Effects of Global Climate Change on the
United States. For that report, EPA commissioned fifty-five studies by academic and government scientists on
the potential effects of global climate change. Each study was reviewed by at least two peer reviewers. The
Effects Report summarizes the results of all of those studies. The complete results of each study are contained
in Appendices A through J.
Appendix Subject
A Water Resources
B Sea Level Rise
C Agriculture
D Forests
E Aquatic Resources
F Air Quality
G Health
H Infrastructure
I Variability
J Policy
GOAL
The goal of the Effects Report was to try to give a sense of the possible direction of changes from a global
warming as well as a sense of the magnitude. Specifically, we examined the following issues:
o sensitivities of systems to changes in donate (since we cannot predict regional dimate change, we
can only identify sensitivities to changes in climate factors)
o the range of effects under different warming scenarios
o regional differences among effects
o interactions among effects on a regional level
• ••
ill
-------
o national effects
o uncertainties
o policy implications
o research needs
The four regions chosen for the studies were California, the Great Lakes, the Southeast, and the Great Plains.
Many studies focused on impacts in a single region, while others examined potential impacts on a national scale.
SCENARIOS USED FOR THE EFFECTS REPORT STUDIES
The Effects Report studies used several scenarios to examine the sensitivities of various systems to changes in
climate. The scenarios used are plausible sets of circumstances although none of them should be considered to
be predictions of regional climate change. The most common scenario used was the doubled CO2 scenario
(2XCO2), which examined the effects of climate under a doubling of atmospheric carbon dioxide concentrations.
This doubling is estimated to raise average global temperatures by 1.5 to 4-5°C by the latter half of the 21st
century. Transient scenarios, which estimate how climate may change over time in response to a steady increase
in greenhouse gases, were also used. In addition, analog scenarios of past warm periods, such as the 1930s, were
used.
The scenarios combined average monthly climate change estimates for regional grid boxes from General
Circulation Models (GCMs) with 1951-80 climate observations from sites in the respective grid boxes. GCMs
are dynamic models that simulate the physical processes of the atmosphere and oceans to estimate global climate
under different conditions, such as increasing concentrations of greenhouse gases (e.g., 2XCO2).
The scenarios and GCMs used in the studies have certain limitations. The scenarios used for the studies assume
that temporal and spatial variability do not change from current conditions. The first of two major limitations
related to the GCMs is their low spatial resolution. GCMs use rather large grid boxes where climate is averaged
for the whole grid box, while in fact climate may be quite variable within a grid box. The second limitation is
the simplified way that GCMs treat physical factors such as clouds, oceans, albedo, and land surface hydrology.
Because of these limitations, GCMs often disagree with each other on estimates of regional climate change (as
well as the magnitude of global changes) and should not be considered to be predictions.
To obtain a range of scenarios, EPA asked the researchers to use output from the following GCMs:
o Goddard Institute for Space Studies (GISS)
o Geophysical Fluid Dynamics Laboratory (GFDL)
o Oregon State University (OSU)
Figure 1 shows the temperature change from current climate to a climate with a doubling of CO, levels, as
modeled by the three GCMs. The figure includes the GCM estimates for the four regions. Precipitation
changes are shown in Figure 2. Note the disagreement in the GCM estimates concerning the direction of
change of regional and seasonal precipitation and the agreement concerning increasing temperatures.
Two transient scenarios from the GISS model were also used, and the average decadal temperature changes
are shown in Figure 3.
iv
-------
FIGURE 1. TEMPERATURE SCENARIOS
QCM Estimated Change in Temperature from 1xC(>2 to 2xCO2
SUMMER
Great Southeast Great California United
Lakes Plain* States'
Great Southeast Graat California United
Lakaa Plains Statas*
Great Southaast Great California United
Lakas Plains States*
GISS
GFDL
[ | osu
* Lower 48 States
-------
FIGURE 2. PRECIPITATION SCENARIOS
GCM Estimated Change in Precipitation from 1xCO2 to 2xCO2
Great Southeast Great California United
Lakes Plains States*
WINTER
Great Southeast Great California United
Lakes Plains States*
Great Southeast Great California United
Lakes Plains States*
GISS
GFDL
OSU
* Lower 48 States
-------
4
3.5
O 3
g 2.5
ui
a.
ui
1.5
1
0.5
0
o
^
3.5
3
2.5
2
£ 1.5
0.5
0
3.72
2.99
2.47
1.72
1.36
0.70
0.88
- 0.30
1980s 1990s 2000s 2010s 2020s 2030s 2040s 2050s
TRANSIENT SCENARIO A
1.28
1.
0.59
0.35
V//A
1980s 1990s 2000s 2010s
TRANSIENT SCENARIO B
2020s
FIGURE 3.
GISS TRANSIENTS "A" AND "B" AVERAGE
TEMPERATURE CHANGE FOR LOWER 48 STATES
GRID POINTS.
vii
-------
EPA specified that researchers were to use three doubled CO, scenarios, two transient scenarios, and an analog
scenario in their studies. Many researchers, however, did not nave sufficient time or resources to use all of the
scenarios. EPA asked the researchers to run the scenarios in the following order, going as far through the list
as time and resources allowed:
1. GISS doubled CO2
2. GFDL doubled CO2
3. GISS transient A
4. OSU doubled C02
S. Analog (1930 to 1939)
6. GISS transient B
ABOUT THESE APPENDICES
The studies contained in these appendices appear in the form that the researchers submitted them to EPA.
These reports do not necessarily reflect the official position of the VS. Environmental Protection Agency.
Mention of trade names does not constitute an endorsement
vifi
-------
INTERPRETATION OF HYDROLOGIC EFFECTS OF CLIMATE CHANGE IN THE
SACRAMENTO-SAN JOAQUTN RIVER BASIN, CALIFORNIA
Dennis P. Lettenmaier
Thian Yew Can
Department of Civil Engineering
University of Washington
Seattle, WA 96195
and
David R. Dawdy
Consultant
3055 23rd Ave.
San Francisco, CA 94132
Contract No. CR814637
-------
CONTENTS
Page
ACKNOWLEDGMENTS "»
EXECUTIVE SUMMARY I'1
CHAPTER 1: INTRODUCTION 1-3
1.1 Background 1-3
1.2 Objectives 1-4
13 Relationship to Other Studies 1-4
CHAPTERS STUDY DESIGN AND METHODOLOGY 1-5
2.1 Overview 1-5
2.2 Sacramento-San Joaquin Basin Description 1-7
23 Study Catchment Description 1-7
23.1 McCloud River 1-9
232 Merced River 1-9
233 North Fork American River 1-9
23.4 Thomes Creek 1-12
2.4 Model Description 1-12
14.1 Snowmelt Model 1-12
2.4.2 Soil Moisture Accounting Model 1-13
2.43 Spatial Disaggregation Model 1-15
2.5 Model Implementation 1-16
2J.1 Precipitation-Temperature Stations and Data Quality 1-16
252 Snowmelt Model Parameter Estimation 1-18
2^3 Soil Moisture Accounting Model Parameter Estimation 1-22
2.6 Model Input Characterization 1-22
2.6.1 Precipitation 1-26
2.6.2 Temperature 1-26
2.63 Potential Evapotranspiration 1-26
CHAPTERS: RESULTS 1-28
3.1 Study Catchment Results for Initial Scenarios 1-28
3.1.1 Snow Water Equivalent 1-32
3.1.2 Runoff 1-32
3.13 Evapotranspiration 1-40
3.1.4 Soil Moisture Storage 1-41
32 Sensitivity Scenarios 1-41
3.2.1 Snow Water Equivalent 1-42
322 Runoff 1-42
323 Evapotranspiration 1-43
3.2.4 Soil Moisture Storage 1-43
33 Spatial Disaggregation: Primary Nodes 1-43
3.4 Spatial Disaggregation: Secondary Nodes 1-45
CHAPTER 4: SUMMARY AND CONCLUSIONS 1-48
REFERENCES 1-51
11
-------
ACKNOWLEDGMENTS
The authors appreciate the assistance of a number of individuals who assisted in various aspects of the
project. Dr. Richard H. Hawkins, then on leave with the EPA Environmental Research Laboratory, Corvallis,
Oregon, and now with the Watershed Management Program, University of Arizona, was instrumental in the
initial planning of the study. Dr. Ronald Neilson of ERL, the Project Officer, participated in the project design
and management, as did Dr. Robert Worrest of ERL, and Mr. Joel Smith of EPA's Office of Policy, Planning,
and Evaluation. Mr. Wendell Tangbora of the HyMet Company graciously provided computer tapes of daily
meteorological data for California. Mr. Roy Jenne of the National Center for Atmospheric Research, assisted
by Mr. Dennis Joseph of NCAR, provided additional meteorological data, as well as computerized summary
output of the Global Climate Model results. Some of the computer simulations and preliminary analysis were
performed by Dr. N. Davies Mtundu, a Postdoctoral Research Associate in the Department of Civil Engineering,
University of Washingtoa The report was reviewed by Dr. Stephen J. Burges, of the Department of Civil
Engineering, University of Washington. Notwithstanding the contributions of these individuals, the content of
the report, and any opinions expressed, are the sole responsibility of the authors.
This paper is Water Resources Series Technical Report No. 110.
ui
-------
Lettenmaier
EXECUTIVE SUMMARY1
Objectives: The objectives of the study were: 1) To develop a method to interpret the possible hydrologic
impacts of global climate change for catchments on the order of tens to hundreds of square miles in mountainous
regions; 2) To assess the interaction of the hydrologic diversity of medium-sized catchments within a large river
basin in California (the Sacramento-San Joaquin) with climate change predictions made using general circulation
models (GCMs); 3) To assess differences hi the hydrologic implications of different GCMs, with specific
attention to the nature of predicted temperature and precipitation variations by month; and 4) To provide
simulated hydrologic data at time and space scales suitable for an assessment of the possible water resource
system impacts of climate change in the Sacramento-San Joaquin basin.
Approach: Four study catchments with area ranging from 203 to 358 square miles were selected in the
headwaters of the Sacramento-San Joaquin basin. The study catchments were selected on the basis of their
geographic and hydrologic diversity, the absence of major upstream diversions or flow regulation, and the
availability of long-term hydrologic and meteorological records. Three of the study catchments, the McCloud
River, the North Fork American River, and Thomes Creek, lie in the Sacramento River basin; one, the Merced
River, is in the San Joaquin basin.
Snow accumulation and ablation, and runoff, were simulated under current (historical) conditions using the
National Weather Service snowmelt and soil moisture accounting models. The snowmelt model operated on a
six hourly time step, while the soil moisture accounting model operated on a daily time step. Once the models
were calibrated to present conditions, simulations were performed under seven alternative climate scenarios.
Four of these were initial cases, based on GCM climate scenarios corresponding to 1) Geophysical Fluid
Dynamics Laboratory (GFDL) model results for CO2 doubling; 2) Goddard Institute for Space Studies (GISS)
model results for CO2 doubling; 3) A transient climate predicted by the GISS model for CO, changing from
current concentrations to double current levels over an eighty year period; and 4) Oregon State University
Meteorology Department (OSU) model results for CO2 doubling. Sensitivity analyses were performed using two
additional climate scenarios: 1) the GISS model temperature predictions, with precipitation assumed to remain
the same as present; and 2) a long-term climate similar to that experienced in the 1930's.
All simulations were performed using 100 years of daily temperature and precipitation data (disaggregated to
a six hourly interval for the snowmelt model) consisting of the years 1951-80 supplemented by 70 additional years
drawn at random from the 1951-80 record. For the alternative climate scenarios, the 100-year temperature and
precipitation records were adjusted as follows: The long-term average GCM temperature and precipitation
means were interpolated to 120°W, 40°N, which is approximately the centroid of the Sacramento-San Joaquin
basin. For precipitation, the ratio of the GCM-predicted long-term monthly mean to the long-term mean for
a base case (nominally, present conditions) for the same GCM was computed. This ratio was applied to all of
the historic precipitation records. For temperature, the same approach was used, except that the difference
between the long-term monthly mean temperature for a given climate alternative and the mean temperature for
the same model's base case was used to adjust the historic precipitation. For the 1930's analog, the precipitation
factors and temperature adjustments were based on an analysis of long-term historic data, rather than GCM
results.
Results: All of the initial scenarios (based on steady-state CO2 doubling, or a transient from present conditions
to CO, doubling) showed that the simulated hydrologic changes were dominated by a shift in the snow
accumulation pattern. Specifically, under the wanner conditions predicted by the GCMs, snow would occur only
rarely at lower elevations, and the snow accumulation would be reduced at the higher elevations. For all but the
highest catchment (the Merced), this resulted in a change from a snow-dominated to a rainfall-dominated
'Although the information in this report has been funded wholly or in part by the U.S. Environmental
Protection Agency under contract no. CR814637, it does not necessarily reflect the Agency's views, and no official
endorsement should be inferred from it.
1-1
-------
Lettenmaier
hydrologic regime. Long-term mean snow accumulations were greatly reduced, and the maximum mean runoff
was shifted from the spring to the winter. Spring and summer runoff were greatly reduced. In addition, soil
moisture increased in the winter months and decreased in the spring and summer. Because of the reduction of
summer soil moisture, and the increased potential evaporation in the spring, there was a shift in maximum
evapo-transpiration earlier in the season. These general shifts were observed for all of the alternative climates.
They were most severe for the GFDL and GISS CO2 doubling scenarios, for which the temperature shifts were
the greatest, and were least severe for the OSU model CO, doubling, for which the predicted temperature
increases were less than one-half those for the GISS and GFDL models for most months.
The sensitivity analyses were designed to assess the extent to which the simulated hydrologic changes resulted
from GCM-predicted temperature, as opposed to precipitation, changes. Simulations based on input from the
GISS model temperature changes, with present precipitation, showed that the simulated hydrologic changes were
dominantly the result of the GCM-predicted warming and not of the predicted precipitation changes. The
simulations for the 1930's analog reflected a vastly different hydrologic regime than that predicted by the GCM
warming scenarios: temperatures were about the same as present, but precipitation was reduced somewhat.
Therefore, for this case, winter snow accumulations were slightly reduced (the reduction was much less than for
CO, doubling scenarios), but the seasonal flow distribution remained about the same, as did the seasonal
distribution of soil moisture and evapotranspiration.
Conclusions: The major conclusion of the study is that, for the snow-dominated hydrology of the
Sacramento-San Joaquin basin, a general warming on the order of that predicted by all the GCMs would cause
a major reduction in winter snow accumulation, and hence increases in winter runoff and reductions in spring
and summer runoff. The simulated changes in annual runoff were minor, and from a practical standpoint,
inconsequential in comparison to the change in the seasonal distribution of runoff. Attendant changes in the
seasonal distribution of soil moisture and evapotranspiration would also occur. From a hydrologic perspective,
GCM-predicted changes in precipitation, for which there is less consensus than temperature, would be less
important than the predicted temperature changes.
This preliminary study suggests a number of aspects of the hydrologic cycle that require further study. These
include the following: 1) The space-time distribution of precipitation under GCM-predicted altered climates;
2) The interaction of long-term shifts in vegetation, particularly as they would affect evapotranspiration and
runoff; 3) Changes in the distribution of extreme floods, given the likely increased incidence of rain-on-snow
events in the mid-winter period of maximum precipitation; and 4) Estimation of potential evaporation under the
GCM-predicted climates.
1-2
-------
Lettenmaier
CHAPTER 1
INTRODUCTION
1.1 Background
Modern civilizations depend critically on water for municipal and industrial supply, irrigated agriculture,
natural resources recovery, and other uses. Throughout history, water shortages have had dramatic cultural
effects. Periods of below-normal flow persisting for months to years are a characteristic of the hydrologic cycle
that must be recognized in water resource system planning. This is normally accomplished by constructing
reservoirs, sometimes augmented by groundwater withdrawals, which act as a buffer to variations in streamflow.
The reliability of a water resource system depends on the withdrawals (demand), as well as the long-term
statistics of the reservoir inflows, including their means, coefficients of variations, skew coefficients, and cross-
and auto-correlations. Most water resource systems are designed and operated based on a requirement that the
system must perform reliably during droughts similar to those experienced in the historic streamflow record.
Statistically this is equivalent to an assumption that the inflows are stationary, that is, the streamflow statistics
are not time-varying.
The possibility of permanent changes in streamflow, such as might result from long-term changes in
climate, complicates the problem of water resources system design and operation considerably. Although
changes in climate over periods of thousands of years are well documented, there has been less agreement
among hydrologists as to whether changes within typical project planning periods (on the order of a hundred
years or less) can be distinguished from the random variations that are to be expected in a stationary time series
(see, for example, Klemes, 1974; Lettenmaier and Surges, 1978). This issue was the topic of extensive research
by hydrologists in the 1%0's and early 1970's, some of which is summarized in the overview to a 1977 National
Research Council report (Wallis, 1977). The general conclusion of the NRC report was that, from a practical
standpoint, evidence for climate change is not detectable in historic streamflow or meteorological records, which
are typically of length less than 100 years.
The advent of general circulation models (GCMs), and some consensus regarding the likely direction
of future global climate change places the problem in a somewhat different light. There is now a
quasi-deterministic basis for assuming the form that future climate change might take, at least with respect to
changes in temperature. A GCM simultaneously solves (numerically) equations representing the conservation
of mass, energy, momentum, and the equations of state on a global grid (Hansen et aL, 1983,1988). The spatial
scale of the GCM results is, however, inadequate for hydrologic interpretation. GCM predictions are provided
as spatial averages over areas on the order of hundreds of thousands of square kilometers. In addition, it is
questionable if GCM predictions on time steps shorter than about one month reflect the observed natural
variability, particularly given that they represent grid-cell averages (see, for example, Rind et aL, 1988).
Interpretation of the effects of changes in meteorological inputs to a hydrologic system (and in turn,
water resource developments) requires that those inputs be specified on a time scale appropriate for modeling
river basin storm response. This is so because the rainfall-runoff process is highly nonlinear, and such
subprocesses as infiltration and evapotranspiration, which play major roles in determining the runoff yield of a
basin, depend strongly on the storage and movement of water within the soil column during storms, and the soil
moisture condition at the onset of storms. For practical purposes, this implies a daily time scale for large basins
(several hundreds of square kilometers and up) and hourly or less for smaller basins. While GCMs can provide
grid cell average results for time steps on the order of a day or less, it is not dear whether the results at these
short time scales properly reflect the short-term dynamics of the atmospheric circulation process.
Therefore, while the GCM models predict long-term changes that could have substantial impacts on
water resource systems, an appropriate interface between the GCM output (most importantly, precipitation,
temperature, and potential evaporation) and hydrologic models has not been developed It is unreasonable to
1-3
-------
Lettenmaier
use GCM results directly as input to a hydrological model to provide streamflow predictions that might be used
in the same sense, for instance, as an estimate of the 100-year flood for flood plain mapping. In the absence of
better.space-time resolution, current hydrologic interpretations are limited to providing descriptive results that
must be interpreted in an alternative scenarios context, or sensitivity analysis context. There is no basis at present
to predict the hydrologic effects of long-term climate change; all results presented in this document must,
therefore, be interpreted in an alternative scenarios context.
12 Objectives
The objectives of this work are as follows:
1. To develop a method to provide a descriptive interpretation of the hydrologic impacts of globalclimate
change for medium sized catchments on the order of tens to hundreds of square miles in mountainous
regions. The approach must account for the changes in snowmelt hydrology that would result from
long-term temperature increases, coupled with possible changes in precipitation and evapotranspiration;
2. To assess the interaction of the hydrologic diversity of medium sized catchments within a large river
basin in California (the Sacramento-San Joaquin) with GCM predictions to interpret the modes through
which climate change could be evidenced in catchment hydrology;
3. To assess differences in the hydrologic implications of different GCMs, with specific attention to the
nature of predicted temperature and precipitation variations by month;
4. To use the methods and results developed in 1) and 2) to provide hydrologic input at time and space
scales that allow descriptive interpretation of the water resource systems impacts of climate change on
the Sacramento-San Joaquin basin;
5. To develop recommendations for research needed to improve the assessment methodology, particularly
in light of the relative uncertainties in the GCM predictions, and spatial and temporal resolution
incompatibilities between the GCM model output and hydrologic data requirements.
13 Relationship to Other Studies
The primary emphasis of this work is the interpretation of the implications of global climate change
as predicted by three GCMs (the models of the Geophysical Fluid Dynamics Laboratory, GFDL; the Goddard
Institute for Space Studies, GISS; and the Oregon State University Department of Meteorology, OSU) on four
medium sized catchments in the Sacramento-San Joaquin River basin of California As part of this work,
predictions of streamflows for larger subbasins of the Sacramento-San Joaquin system were developed and used
in a companion study of the implications of global climate change on the operation of the Sacramento-San
Joaquin water resource system (Sheer and Randall, 1968). Sheer and Randall applied a model of the water
resource system which uses as input monthly streamflow volumes for 13 subbasins. These subbasins have
drainage areas much larger than the study catchments described in this report, for which detailed hydrologic
models were implemented
For several reasons relating to data availability and time and budget constraints, it was not possible to
implement, or develop, a detailed hydrologic model of each of the 13 subbasins. Therefore, to provide the
input required by the water resource systems model used by Sheer and Randall, a statistical model was developed
to relate the subbasin (monthly) streamflows to streamflows in the four study catchments. These subbasin flows
represent the interface between this study and Sheer and Randall's work and (indirectly) with other studies based
on Sheer and Randall's results. In this report, however, we emphasize the interpretation of the hydrologic
implications of climate change on the four study catchments. Because the relationship between the study
catchment and subbasin flows is statistical, there is no basis for a dynamical interpretation of the changed
hydrologies at the subbasin level
1-4
-------
Lettenmaier
CHAPTER 2>
STUDY DESIGN AND METHODOLOGY
2.1 Overview
The overall study design is shown schematically in Figure 2.1. The climatic inputs fall into two categories:
historic data, also referred to as the base case (Case OX and GCM climate alternatives, referred to as Cases 1-7,
identified in Table 2.1. The Case 0 historic data, for the period 1951-1980, were augmented by randomly
resampling (with replacement) from the years 1951-80 to provide a total record of length 100 years. The
hydrologic outputs were either for one of four study catchments (with drainage areas ranging from 203 to 358
square mflesX or for water resource system nodes, which are major subbasins of the Sacramento-San Joaquin
of size up to several thousand square miles. The initial effort in the study was to implement hydrologic models,
one for prediction of snow accumulation and ablation, and another for soil moisture accounting, which could
predict the four study catchment outflows on a daily time scale, for precipitation and temperature maxima and
minima on a daily time scale, and monthly average pan evaporation.
Table 2.1 General circulation model scenarios investigated
Case Description
0 Historic conditions 1951-80
1 Geophysical Fluid Dynamics Laboratory (GFDL) 2xCO, steady-state
2 Goddard Institute for Space Studies (GISS) 2xCO, steady-state
3 GISS 80 year 2xCO, transient
4 Oregon State University (OSU) 2xCO2 steady-state
5 GISS 2xCO2 steady-state with temperature change only (historic precipitation assumed)
6 GISS 2xCO2 steady-state with differential GCM node input for southern and northern
study catchments
7 1930*3 analog precipitation and temperature
The implementation followed a standard procedure of model calibration, or parameter estimation, using
a selected subset of the historic record, and an independent subset of the data for model verification. Section
25 gives details of the calibration and verification procedure. Once the hydrologic models were calibrated, they
were run with the historic data altered by addition of a seasonal shift in the case of temperature, and by
multiplication by a seasonally varying factor in the case of precipitation, to be consistent with predictions from
each of seven long-term GCM predictions. For each climate alternative, study catchment outflows were predicted
on a dairy time step for the 100-year period indicated above.
This report focuses on the upper right quadrant of Figure 2.1, that is, on interpretation of the hydrologic
model output that included soil moisture, snow accumulation, evaporation, and other variables, as well as
streamflow. For the purposes of a companion study (Sheer and Randall, 1988) monthly streamflows at each of
the 13 water resource system model nodes were predicted using a spatial disaggregation model applied to the
study catchment monthly flows; a random noise term was included to assure that selected statistical properties
of the disaggregated flows were preserved The disaggregation model coefficients were estimated using the Case
0 study catchment predictions (summed dafly flows to provide a 100-year monthly flow record) and the
corresponding historic monthly flows at the water resource model nodes, as indicated in the lower left quadrant
of Figure 11. Finally, as indicated in the lower right quadrant of Figure U, the disaggregation model
1-5
-------
Lettenmaier
CL
h-
0
o
o
o
tn
Q
X
Study
Catchment
1 - Merced
2 - North Fork
3 - Thomes
4- McCloud
Water
Resource
Node 1
2
3
4
5
6
7
8
9
10
11
12
13
CLIMATE INPUT
Historic Data (Base) GCMCase! 234567
(Daily precipitation. (Adjusted historic daily
Tmax , Tmin monthly precipitation. Tmax, Tmin
avg pan evaporation) monthly avg pan evaporation)
CALIBRATION
1. 1957-1961
2. 1952-1956
3. 1953-1957
4. 1956-1960
k
*^»
^^
.CO
•c
VERIFICATION
1. 1961-1980
2. 1956-1980
3. 1957-1980
4. 1960-1980
^
g*
"Y"*
MODEL ESTII
100yrhydr
model outpi
study catchi
historic date
Node 1-1 31
MATION
ologic
jtfor
nents
!,for mode
low*? _
OUTPUT:
1 00 year daily
Streamflow Records
a
*Si
-s
MONTHl
RESOUF
PRED
. (100 yea
' i n
K flrtuie a
coefficien
Sir
Is/lx
^ JO
Is
\T
i
.Y WATER
*CE NODE
ICTIONS
rs of monthly
t each node)
Figure 2.1. Schematic overview of study design.
1-6
-------
Lettenmaier
was used to simulate 100-year monthly streamflow sequences at the water resource system model nodes
corresponding to each of the seven climate alternatives. The climatological, hydrologic, and geologic features
of the Sacramento-San Joaquin basin, and the four study catchments, are described, and the study approach is
expanded, in the next section.
2.2 Sacramento-San Joaquin Basin Description
The Sacramento-San Joaquin Basin (Figure 12), or Central Valley, extends nearly 500 miles from north
to south, with an average width of about 45 miles separating the Sierra Nevada from the Coast Range. The
elevation of much of the valley floor is dose to sea level. Beneath its silt and gravel cover is a thick sedimentary
sequence. The Coast Range parallels the Pacific coast from the Oregon border to just north of the Los Angeles
Basin. The mountains in the Coast Range rise abruptly from the narrow coastal plain to peak elevations of 8,000
feet. The Sierra Nevada, with peak elevations over 14,000 feet, form the eastern boundary of the basin. The
Central Valley is terminated in the south by several transverse ranges, which are composed of many overlapping
mountain blocks of nearly east-west trend. The basin terminates to the north near the Oregon-California border,
where the Coast Range, the Sierra Nevada, and the Cascade range converge.
The Sierra Nevada protect the Sacramento-San Joaquin basin from the cold continental air masses that
flow further east in winter. The Coast Range blocks the strong westerly air-flow and accompanying cool summer
temperatures that are experienced by the western slopes of the Coast Range. Most of the precipitation in the
basin is derived from frontal storms originating in the North Pacific between the months of November and
March. Precipitation is strongly orographiq it is heaviest on the west facing slopes of both the Coast Range and
the Sierra Nevada. The valley floor is semi-arid, with average annual precipitation generally less than 15 inches.
Temperatures are influenced by prevailing air masses, elevation, and the drainage of cold dense air from
the higher elevations into the Central Valley. Although the climate on the western side of the Coast Range is
dominated by the Pacific Ocean, with warm winters, cool summers, small daily and seasonal temperature ranges,
and high relative humidities, the Sacramento-San Joaquin Basin experiences a more continental type of climate
with colder winters, warmer summers, greater daily and seasonal temperature ranges, and generally lower
relative humidities than are common for the Coast Range.
23 Study Catchment Description
The four study catchments (Figures A.1 to A.4) were selected to represent the geographic, climatic, and
hydrologic diversity within the Sacramento-San Joaquin River basin, and to represent, via the spatial
disaggregation model described in Section 2.43, the monthly streamflows at the water resource systems nodes.
Initially, 26 candidate study catchments were identified that were upstream of all major reservoirs. Modeling of
streamflow below reservoirs is complicated by the necessity to account for the effect of reservoir storage on
stream flow; in practice, to implement a rainfall-runoff model successfully at a daily or shorter time scale requires
streamflow data relatively unaffected by reservoirs or upstream diversions.
In addition to the requirement that upstream regulation be minimal, it was required that 1) the
candidate study catchments be defined by VS. Geological Survey (USGS) stream gaging stations rated "good"
or better (interpreted by the USGS to mean that the true flow can be expected to be within 10 percent of the
recorded flow 95% of the time); an exception to this criteria was allowed during periods of ice cover; and 2) that
there be at most minor diversions above the stream gage; and 3) that the period of streamflow record include
the years 1951-80. Further screening using these criteria reduced the number of candidate catchments to 19.
The remaining stations were then ranked based on the correlation of their annual flow with the summed
annual flow over all the 13 water resource system node annual flows. The three most highly correlated basins,
the North Fork of the American River at North Fork Dam (USGS 11-4270, drainage area 342 square miles),
Thomes Creek at Paskenta (USGS 11-3820,203 square mifesX and the Mcdoud River near McCloud (USGS
11-3675,358 square mfles) were selected Because of the desirability of representing the San Joaquin subbasin,
1-7
-------
Lettenmaier
N
t
80 miles
38°--
SACRAMENTO
BASIN
SAN JOAQUIN
BASIN
120°*
Figure 2.2. Location of the Sacramento - San Joaquin River Basin.
1-8
-------
Lettenmaier
the southern subcatchment (the Merced River at Happy Isles Bridge, USGS11-2645,181 square miles) havingthe
highest correlation to the total annual basin flow was selected as the fourth study catchment.
Figure 23 shows the locations of the index basins. Precipitation on two of the index basins, the
American and the Merced River, is typical of that occurring on the west facing slopes of the Sierra Nevada;
precipitation in the Thomes Creek Basin is characteristic of the east facing slopes of the Coast Range. The
McCloud River, which drains the extreme northern part of the Sacramento River basin, has precipitation
characteristics intermediate between those of the east facing slopes of the Coast Range and the west facing slopes
of the Sierra Nevada. A brief summary of the particular characteristics of each of the study catchments is
provided in the remainder of this section.
23.1 McCloud River
The annual precipitation, runoff, and temperature extremes (for January and July) are listed in Table
22. The McCloud Basin (358 nu^) has fairly warm summers, with mean annual temperature slightly below 70°F
and cool winters (annual mean of 45°F). Snow and freezing temperatures are not common except at higher
elevations. The basin has annual precipitation in excess of 70 inches but rain is light in summers and more
frequent in late fall and in winter, with over 80% of the annual total falling between November and May.
Thunderstorms occur occasionally in summer; however, they account for only a small percentage of the total
annual precipitation.
The basin is located in an area with large, gentry sloping volcanoes built by outpouring of basaltic and
andesitic lava. Most of these rock formations are deeply weathered, resulting in deep, well-drained soils of
extremely high permeability. Even where unweathered, the lavas are highly permeable and generate little surface
runoff. Consequently, the annual runoff hydrograph has extremely low variability with a peak to average flow
ratio of only about 2.
232 Merced River
The climate in the Merced basin (321 mi2) is elevation-dependent, with hot summers and mild winters
at low elevations and mild summers and cold winters at high elevations. Because of its relatively higher mean
elevation, its hydrology is more controlled by snowfall and snowmelt than other study catchments.
Precipitation is less in the Merced basin than in the McCloud, with an annual mean of about 64 inches.
It increases with elevation, from 36 inches at 4,000 feet elevation to a maximum of about 70 inches between 8,000
and 10,000 feet Most of the precipitation falls in winter, with almost 90% of the annual total falling between
November and April Most precipitation falls as snow at the higher elevations.
The soils in the basin are varied, but are, in general, deep and permeable; some areas are clayey and
have lower permeability. The slopes range from nearly level to very steep. Because of the high permeability of
the soils, the basin has a relatively dampened storm response, as evidenced by the ratio of the mean annual flood
to mean annual flow of about 7.
233 North Fork American River
Figure 23 shows the location of the North Fork American River Basin (342 mi2) and the main stem
of the American River. The North Fork American River has the lowest median elevation (4,000 ft) of any of
the study catchments. Much of the annual precipitation (50-60 inches) falls in late autumn and winter.
Precipitation in summer is tight and generally limited to occasional convective storms.
1-9
-------
Lettemnaier
McCLOUD RIVER BASIN
THOMES
CREEK
BASIN
oi NORTH FORK
AMERICAN RIVER
BASIN
rMERCED RIVER
\ BASIN v
N
t
40 80 miles
\
N
Figure 23. Study catchment locations.
MO
-------
Lettemnaier
Table 22.
Mean annual historical precipitation, runoff and temperature of McCloud River,
Merced River, North Fork American River, and Thomes Creek study catchment
Basin
McCloud
Merced
North Fork
American
Thomes
Creek
Mean
Annual
Runoff
/Inches'^
\ cfs/
35.40
(933.6)
26.30
(621.9)
33.50
(844.0)
19.86
(297.0)
Mean
Annual
Precipitation3
(Inches)
72.90
64.00
60.30
56.50
Mean January Daily
Temperature *Fa
Dally
Average
30.5
31.4
39.8
35.5
Minimum
19.5
20.5
30.4
24.9
Maximum
41.6
42.2
49.0
46.3
Mean July Daily Temperature "Fa
Daily
Average
63.7
66.4
72.9
66.8
Minimum
43.0
47.9
58.5
45.2
Maximum
84.0
84.9
87.2
88.2
Weighted average over elevation bands
Ml
-------
Lettenmaier
The topography varies from nearly level to rolling old valley fill in the vicinity of the Sacramento Valley
to predominantly steep dipping, faulted, and folded metamorphic rocks that have been intruded by several types
of igneous rocks. Most soils in the mountainous uplands formed in place over metamorphic rock, granitic rock,
or andesitic conglomerate, and are not as permeable as those in the Merced Basin. Consequently, the runoff
response of the catchment is flashy, particularly in the upland mountains; the peak to average flow ratio is about
20.
23.4 Thomes Creek
As with the other study catchments, most of the precipitation in the Thomes Creek basin (203 mi2) is
derived from winter frontal storms. Thundershowers, most frequent in May and June, are generally of short
duration but occasionally bring rainfall of high intensity. Much of the western part of the basin, which is above
an altitude of 3,500 feet, has moderately deep to deep steep soils underlain by hard sedimentary and
metamorphic (mainly fractured mica schist) rock. Permeability is low to moderate, resulting in relatively flashy
storm response; the peak to average flow ratio is about 22.
2.4 Model Description
The models selected for this study are a snowmelt model, a soil-moisture accounting (rainfall-runoff)
model, and a spatial disaggregation model Each of the models is described briefly in this section.
2.4.1 Snowmelt Model
The snowmelt model was developed by Eric Anderson of the VS. National Weather Service Hydrologic
Research Laboratory (Anderson, 1973). The model consists of a set of equations that describe the change in
storage of water and heat in the snowpack. The model inputs are ambient air temperature and precipitation at
a six hourly time step. In this study, daily precipitation was interpolated to six hourly increments and six hourly
temperature was estimated from daily temperature maxima and minima using equations given by Anderson
(1973). The model input is limited to air temperature and precipitation because data other than temperature
and precipitation needed for heat budgets are not normally available in mountainous environments.
The model can be summarized as follows. The heat exchange computations are separated into melt and
non-melt periods. Melt periods are further separated into wet (precipitation) and dry periods. If precipitation
occurs, the ground is bare, and the ambient air temperature is greater than 32°F, no computations are performed.
If the ground is not bare, the heat exchange at the air-snow interface is computed. Two conditions are
considered, warm air (air temperature (Ta) >32°F), and cold air (Ta <32°F). For Tfl >32°F, the following
assumptions are made: a) there is no solar radiation; b) incoming longwave radiation is equivalent to blackbody
longwave radiation at T ; c) the snow surface temperature is 32°F; d) the dew point is T ; and e) the rain
temperature is Tfl. On flic basis of these assumptions, the heat balance is computed as meft heat loss - Q +
Qe + Qh + QD» where Q = long wave radiation, Qe * latent heat transfer due to condensation, Qh = sensible
heat transfer (Bowen ratio based on above assumptions), and Q » heat transfer by rainwater (based on
assumed rainwater temperature). If Ta is less than 32°F, it is assumed that the precipitation is falling as snow
and that no melt occurs.
For melt during nonrain periods, the model first checks to determine whether the snowpack is isothermal
at 32°F. If the snowpack is not isothermal, no melt occurs, and the net heat flux is added to the heat content
of the snowpack. If the snowpack is isothermal, and the air temperature is greater than 32°F, melt is assumed
to take place proportionate to a seasonally varying melt factor and the difference between the air temperature
and 32°F (assumed isothermal temperature of the snowpack).
During nonmelt periods (assumed by the model to be any time Ta is less than 32°F), an antecedent
temperature index (ATI) is used as an index of the temperature of the surface layer of the snowpack. The ATI
is similar to the antecedent precipitation index often used for storm hydrograph prediction (see, for example,
Linsley et al., 1975). The net heat exchange at the surface of the snowpack is assumed proportional to the
1-12
-------
Lettenmaier
difference between the ATI and the current air temperature. The proportionality constant is a parameter termed
"the negative melt factor," which varies seasonally in the same manner as does the melt factor used during
nonrain periods.
finally, the model accounts for the area! extent of snow cover. During periods of snow accumulation,
this is assumed to be 100%. During periods of depletion, the model uses an area! depletion curve, which
expresses the percent snow covered area as a function of the ratio of mean area! snow water equivalent to an
index value, where the index value is the smaller of the maximum snow water equivalent since snow began to
accumulate (Le., the beginning of the snow season), or a preset maximum.
The most important parameters in the model are: a) the melt factor for nonrain period melt; b) the
negative melt factor for the heat balance computation during nonmelt periods; c) the recursion parameter in
the antecedent temperature index; and d) the minimum index value in the snow-covered area relationship. This
brief description has omitted some details of the model: there is an expression to account for heat transfer at
the base of the snowpack, and the retention of liquid water in the snowpack is also modeled In addition, the
model allows the rain-snow division to be made at other than 32°F, and also allows nonrain period melt to occur
based on a threshold base air temperature which can be other than 32°F. These expressions and generalizations
involve some additional parameters; however, they usually are not as important as the parameters indicated above
for studies of the type being performed here. For a complete description of the algorithm, the reader is referred
to Anderson (1973).
2.4.2 Soil Moisture Accounting Model
The Soil Moisture Accounting Model was developed by Burnash et al. (1973) and forms the basis of the
UJS. National Weather Service's basic catchment hydrologic response model for operational forecasting. It is
a deterministic, lumped parameter, conceptual model The original model was designed for daily precipitation
input but later versions allow finer time increments (6 hours or less). Input to the model is pseudo precipitation
(snowmelt model output) and potential evaporation (actual, or long-term average).
The structure of the soil moisture-accounting model is shown schematically in Figure 2.4. When rainfall
occurs it is considered to fall on two types of basin covers: (1) a permeable soil mantle, and (2) lakes, channel
networks, and impervious areas. Rain falling on impervious areas always becomes direct runoff, whereas that
which falls on the permeable soil mantle undergoes a complicated sequence which represents the infiltration
process. Below the permeable soil mantle, the soil moisture storage is conceptually made up of upper and lower
zones (Figure 2.4). Each zone stores moisture in two forms, tension moisture and free moisture. Tension
moisture denotes water closely bound to the soil particles, while free moisture is the moisture that fills up the
interstitial soil pores. For medium sized basins such as the four study basins, one set of lumped model
parameters is sufficient to represent the basin hydrology.
The upper zone represents topsoils and the basin interception layer. Upper zone tension water, bound
closely to the soil particles, must be filled before moisture can be stored as free water. Upper zone free water
generates vertical drainage (percolation) to the lower zone and lateral drainage (interflow) to the channel. If
the precipitation rate exceeds the sum of lateral and vertical drainage rates, and the upper zone free water
capacity is completely filled, excess surface runoff will result The actual percolation rate to the lower zone is
governed by the interrelationship between soil drainage characteristics and the relative soil moisture conditions
of the two zones.
The lower zone, which represents a groundwater reservoir, has a tension water storage zone and two
free water storage zones (called primary and secondary). Water goes to the tension water zone first and then
to the two free water zones, which generate primary and secondary baseflow. The reason for using three storage
zones is to allow the nonlinear characteristics of baseflow recession to be represented.
Evapotranspiration (ET) extracts moisture from the upper and lower tension zones and from free water
surfaces. For areas covered by surface water or phreatophyte vegetation, actual basin ET occurs at the daily
potential rate. Over other areas of the soil layers, ET extraction depends on the demand and the volume and
1-13
-------
Letlenmaier
t t
1
[ KIVIOUS AttA I I IM»t«VOg
-59
UPPER
LOWER
ET
Firt WATIR
TENSION WAm f j
lll>
SVniAUHUl
lAtt new
I
I
rtwutr BA« new
>K>ur
Figure 2.4. Soil-moisture accounting model schematic.
1-14
-------
Lettenmaier
distribution of tension water storage. The upper tension zone provides ET before the lower zone. The sensitivity
of ET changes to climate variations is discussed in Section 3.1.4.
Total runoff is the sum of direct runoff from impervious and water surfaces, surface runoff and interflow
from the upper zone free water, and primary and secondary baseflows from the lower zone free waters. The
model does not relate soil moisture explicitly to vegetation characteristics. The equations that describe the
linkage between the tension zone contents, potential evapotranspiration, and actual evapotranspiration act as a
surrogate for vegetation effects.
2.43 Spatial Disaggregation Model
The spatial disaggregation model consists of two stages. The first stage relates the monthly flows at the
six water resource system model nodes (see Sheer and Randall (1968) for details) to the monthly flows at the
four study catchments described in Section 23. These six nodes are termed "primary sites." The second stage
relates the remaining seven water resource system model nodes used by Sheer and Randall (termed "secondary
sites") to a selected primary site. Table 23 identifies the primary sites, the secondary sites, and, in the case of
the secondary sites, the primary site to which they are indexed.
Monthly flow data were approximately log normally distributed, so data were transformed logarithmically
to the normal domain. The model used to relate the primary site monthly flows to the monthly flows for the four
study catchments is of the form
Yk ' AA *
-------
Lettenmaier
The parameter matrices Afc and Gk can be expressed in terms of the covariances of the Yk and Xk as
and G = Sg, - AE^1 (12b)
where S^ - E(YkYkT)
Once the monthly streamflow logarithms at the four sites were disaggregated to the six primary sites,
the disaggregated streamflow logarithms were exponentiated to provide the corresponding streamflow in real
space. The disaggregation procedure explicitly preserves the means, variances, and covariances of the logarithms
of streamflows (both between the primary sites, and between the study catchments and the primary basins) in
each month. The model does not preserve these same statistics explicitly in real space; however, it was verified
using the historic flows that these moments were preserved quite well in real space as well (Table 2.4 gives
historic and simulated correlation coefficients for the six primary sites for January and July). More importantly,
the model does not preserve the lagged covariances in either real or log space (the correlation between the flows
in month k and month k-1 are not preserved). A more complex model of the form
explicitly preserves the lagged correlations. This model was tested, and it was found that for some months, no
feasible parameter matrices existed. The possibility of using the more complex model in months where
parameter solutions could be found, and the simpler model (Eq. 2.1) in the remaining months, was considered.
We concluded it was better that any distortions in the covariance structure of the simulated flows relative to the
historical flows be consistent through all months, so the simpler model (Eq. 2.1) was used.
Once the primary site monthly flows were estimated, the secondary site monthly flows were estimated
using a simple linear regression,
ln(QWj) = awq|( UJ + 0W + i7Wj (2.4)
where Qk
least
which i
model explicitly preserves only the first two moments (niean and variance) of the historic flows; it does not
explicitly preserve any of the off-diagonal covariances. The secondary sites were selected because their flows are
much smaller than the primary sites, so failure to preserve statistics other than the mean and the variance at the
secondary sites is of considerably less importance than at the primary sites.
2~5 Model Implementation
This section describes the selection of precipitation-temperature stations to provide representative
climatic data for the four study catchments, as well as the method used to subdivide each study catchment into
elevation zones. In addition, the procedure used for calibration (parameter estimation) of the Snowmelt and
Soil-Moisture Accounting models is described. Characteristics of the selected precipitation-temperature stations,
as well as final parameter estimates for each study catchment, are given.
25.1 Precipitation-Temperature Stations and Data Quality
Since precipitation input errors are among the most important source of runoff simulation errors,
selection of meteorological stations is an important step in the modeling process. Few records are available for
mountainous areas. For this reason, most of the meteorological stations used lie at relatively low elevations.
1-16
-------
Lettenmaier
Table 2.4 Simulated and Historic Correlation Coefficients for the Six Primary Sites (base case)
January Historic
1.000
.950
.817
.855
.740
.907
.950
1.000
.872
.918
.812
.949
.817
.872
1.000
.989
.944
.955
.855
.918
.989
1.000
.932
.978
.740
.812
.944
.932
1.000
.900
.907
.949
.955
.978
.900
1.000
July Historic
.000
.827
.773
.826
.746
.875
.827
1.000
.727
.750
.672
.887
.773
.727
1.000
.949
.905
.897
.826
.750
.949
1.000
.833
.897
.746
.672
.905
.833
1.000
.832
.875
.887
.897
.897
.832
1.000
January Simulated
1.000
.918
.797
.836
.613
.905
.918
1.000
.844
.895
.729
.922
.797
.844
1.000
.984
.884
.936
.836
.895
.984
1.000
.852
.955
.613
.729
.884
.852
1.000
.785
.905
.922
.936
.955
.785
1.000
July Simulated
1.000
.824
.774
.787
.688
.884
.824
1.000
.698
.676
.566
.904
.774
.698
1.000
.916
.808
.865
.787
.676
.916
1.000
.719
.829
.688
.566
.808
.719
1.000
.737
.884
.904
.865
.829
.737
1.000
1-17
-------
Lettenmaier
The selection criteria used were data record length, quality of record (in terms of missing data), geographical
locations of stations with reference to the basin topography, and distance of the stations from the study
catchments. The precipitation-temperature stations selected for each of the study catchments are given in Table
^•J«
To reduce the data handling effort, only one precipitation-temperature gage station was chosen for each
basin. In mountainous basins, orographic effects, which generally cause precipitation to increase with elevation,
dominate the local climatology. Establishing the proper precipitation-elevation relationships is made difficult by
the paucity of high-elevation meteorological stations and their susceptibility to recording errors. The approach
we used was to define an elevation-dependent adjustment factor to relate precipitation at a given elevation (e.g.,
the midpoint of a snowmelt model elevation band) to gage precipitation as follows:
Pe - [1 + P,(e-eg)/100]Pg GL5a)
where Pe is precipitation at elevation e, P is gage precipitation at gage elevation e . The (nonlinear) form of
the factor P((e-e_) was determined by triaTand error based on annual water balance considerations. In general,
P, is a monotomc function of e-e , but its second derivative is negative, reflecting the fact that the rate of
increase of mean precipitation witn elevation decreases at high elevations.
The relationship of temperature to elevation was described by a constant lapse rate. In fact the lapse
rate is a function of meteorological conditions, e.g^ wet or dry day, storm type, and other factors. However, the
available data were insufficient to support more complex relationships.
25.2 Snowmelt Model Parameter Estimation
It is essential that mountainous catchments be divided into elevation zones, and that the snowmelt model
be applied to each zone separately because low elevations might be receiving rain while higher elevations receive
snow from the same storm. The weighted mean of the pseudo-precipitation from all zones was treated as the
mean area! precipitation, which is the input to the Soil-Moisture Accounting Model. In general, the weighting
factors used were equal to the radios of the elevation zone subareas to the total basin area
The elevation bands were delineated as follows. First, hypsometric curves (elevations versus area
fractions) were developed. Each basin was then divided into three or four zones of equal area, depending on
the elevation range, and the elevation of the midpoint of each band was identified. The McCloud and North
Fork American River catchments were divided into three zones while the Merced River and Thomes Creek
catchments were divided into four zones (Table 2.6).
Initial values for the precipitation-elevation adjustment factors P,(°) for each elevation band were
estimated from the change in mean annual precipitation with elevation of selected nearby precipitation stations.
Initially, Pf(°) was assumed to be a linear function of the elevation difference. Subsequently, refinements were
made through a trial and error approach, which was carried out concurrently with the calibration of the
soil-moisture accounting model The final precipitation-elevation adjustment factors are given in Table 2.7.
The snowmelt model was manually calibrated for all the elevation zones. Ideally, observed snow course
data could be used for calibration but in practice snow course observations are sparse in time and space, and they
represent point realizations which usually are not representative of the elevation band average snow water
equivalent predicted by the model. In practice, the calibration procedure involves adjusting the most important
parameters to ensure that the model predicts the initiation of snow accumulation in the fall and the gradual
melting of the snowpack in the late winter, and spring. The other parameters were assigned nominal values
which have been used in previous studies.
1-18
-------
Lettenmaier
Table 2.5. Predpitation/temperature Gage Stations Selected for McCloud River, Merced River, North Fork
American River, and Thomes Creek Study Catchments
Basin
McCloud
Merced
North
Fork
American
Thomes
Station ID
4-5449
4-9855
4-1912
4-2081
Station Name
NcCloud
Yosemite
Park HQ
Col fax
Covela
Elevation
/«Aa+\
lieeij
3280
3966
2410
1430
Years of Record
Temp
76
81
114
46
Preclp
76
81
117
68
1-19
-------
Lettenmaier
Table 2,6.
Elevation Zones for McCloud River, Merced River, North Ford American River, and Thomes
Creek Study Catchments
Basin
McCloud
Merced
North
Fork
American
Thomes
Zone
1
2
3
1
2
3
4
1
2
3
1
2
3
4
Elevation
Range (ft)
2800-4300
4300-5300
5300-12000
3900-7500
7500-8350
8350-9400
9400-12000
750-3350
3350-5250
5250-8600
900-3100
3100-4200
4200-5500
5500-9500
Median
Elv. (ft)
3900
4600
6500
5750
7900
8800
10050
2550
4100
6300
2000
3750
4700
6350
Percent
Basin
Area
50
30
20
25
25
25
25
34
32
34
25
25
25
25
Basin
Area
(Square Miles)
358
321
342
203
1-20
-------
Lettenmaier
Table 2.7. Precipitation Adjustment Factors by Elevation Zone for McCloud River, Merced River, North
Fork American River, and Thomes Creek Study Catchments
Basin
McCl oud
Merced
North
Fork
American
Thomes
Precipitation Adjustment Factors
Zone 1
0.304
0.544
0.025
0.123
Zone 2
0.443
0.839
0.283
0.357
Zone 3
0.786
0.884
0.474
0.402
Zone 4
0.927
0.450
1-21
-------
Lettenmaier
The lapse rate, or rate of temperature decreases with elevation, is a critical parameter. As with the
function P/°X the lapse rate is usually nonlinear. It was found that average lapse rates for four periods (each
of length 6 hours) of each day for all basins vary from -O^C/lOOm to -O8*C/100m. Together with the Pf
function, lapse rates were estimated concurrently with the calibration of the Soil-Moisture Accounting Model (see
Section 153).
Seasonal melt factors were interpolated between MFMAX (maximum nonrain melt factor, which occurs
on June 21) and MFMIN (minimum nonrain melt factor, which occurs on Dec. 21) and were estimated to vary
between 0.9 to L2 and 0.2 to 0.5 mm/°C/6 hours, respectively. PXTEMP, the temperature above which
precipitation was assumed to be rain OF) was taken to be 32*F. SI, the mean area! water-equivalent above
which 100% area! snow cover always exists was assumed to vary between 75 and 150mm. Tables 2A and 2.9 give
descriptions and calibrated values of the snowmelt model parameters.
2J>3 Soil Moisture Accounting Model Parameter Estimation
Parameter estimation for the soil moisture accounting model was based on a process of initial parameter
estimation suggested by Peck (1976X sensitivity analysis to determine those parameters deserving further
attention, followed by automated parameter estimation using a simplex search procedure (Nelder and Mead,
1965) for the most sensitive parameters. The objective function in the search procedure was the sum of the
squared difference between the logarithms of the predicted and observed (daily) streamflow over the calibration
period.
The parameters estimated using the search procedure were LZFPM, LZFSM, LZPK, LZSK, UZK,
UZFVVM, ZPERC, REXP, UZTWM, LZTWM, and PXADJ (parameters are defined in Table 2.10). The
parameters which were assigned nominal values were PCTIM, SSOUT, ADIMP, SIDE, IMPRT, and SARVA.
All parameter estimates were based on calibration periods of length four years. The calibration periods
were selected to include dry, medium, and wet years so that during calibration, the model was subjected to a
broad range of changes in conceptual storages.
Because of its deep and highly permeable sofls, the McCkmd River catchment is dominated by
subsurface flow, even during intense storms. For this reason, the McCloud River catchment was calibrated
manually. Calibrations for all catchments were complicated by errors in the rain and melt data produced by the
snowmelt model, which itself was calibrated through trial and error. The initial and optimized values, upper and
lower bounds for the sensitive parameters are reported in Table 2.10. To test whether the model had been
overfit, that is, whether the simulation errors were consistent between the calibration periods and an independent
verification period consisting of those years in the period 195140 not used for calibration, a seasonal Wilcoxon
test (see Hindi, 1988; Hettmansperger, 1984) was applied to the monthly sum of tog flow differences squared.
The results for all four study catchments fell within the 95% critical region for a two-tailed test, which confirmed
that the performance of the model in the calibration and verification periods was comparable.
2.6 Model Input Characterization
This section describes the relationship between the upper left and right quadrants of Figure 2.1, that is,
the relationship between the historic input data for the hydroiogk model, and the input corresponding to the
GCM alternatives. The historic input data were adjusted to reflect the altered climate predicted by each of the
seven alternatives. This approach was taken in preference to using the GCM output directly, or attempting to
develop a stochastic model to predict the space-time meteorological structure associated with the GCM output.
The reasons for using adjustments to the historic record as input to the hydroiogk models were first, in terms
of the project time constraint, that it was a straightforward approach that required no new model development,
and second, that the sequences might be considered a feasible realization of the (present) natural process since
the historic record has actually occurred In the remainder of this section, the specific approach used to provide
the precipitation, temperature, and potential evaporation inputs b described
1-22
-------
Leltenmaier
Table 2.8. Snowmelt Model Parameter Description
Parameter
Description
DAYGM
Average dally ground melt at the snow-soil Interface In mm
EFC
Area over which evapotransplration can take place when there
Is complete areal snow cover
ELEV
Mean height of the basin elevation band
MBASE
Melt factor base temperature (assumed to be 0*C)
MFMAX(MFMIN) Maximum and mimimum non-rain melt factors which occur on
June 21 and Dec. 21 respectively
NMF
Maximum negative melt factor
PLWHC
PXTEMP
SCF
Percent liquid-water holding capacity of ripe snow
Temperature In *C to divide rain from snow
A multiplying factor to correct for gage catch deficiency in
the case of snowfall
SI
Areal water quivalent in mm above which there is always
complete areal snow cover
TALR
TIPM
UDAJ
Lapse rate CC/lOOm)
Antecedent snow temperature Index parameter
Mean wind function value during raln-on-snow periods
1-2)
-------
Lettenmaier
Table 2.9. Calibrated Values of Snowmelt Model Parameters
PAYGM
EFC
MBASE
MFMAX
MFMIN
NMF
PLWHC
PXTEMP
SCF
SI
TALR
TIPM
UOAJ
McCloud
River
0.4
0.9
0.0
0.90
0.40
0.12
0.07
0.5
1.03
200
-0.5
0.3 i
i
0.10 '
Merced
River
0.4
0.9
0.0
1.10
0.40
0.12
0.07
0.5
1.03
150
-0.7
0.3
0.10
North Fork
American River
0.4
0.9
0.0
1.10
0.20
0.12
0.07
0.5
1.03
100
-0.8
0.3
0.10
Thomes
Creek
0.4
0.9
0.0
1.20
0.20
0.12
0.07
0.5
1.03
120
-0.7
0.3
0.10
1-24
-------
Lettenmaier
Table 2.10, Initial and Optimized Final Values and Upper and Lower Bounds for the Soil-moisture
Accounting Model Parameters
Thomes Creek
LZFPM LZFSM LZPK LZSK UZK UZFWM ZPERC REXP ZTWM LZTWM PXADJ
Initial 6 4 .009 .02 .10 2.0 20 1.80 3 4 0.008
Optimized 1.75 1.75 .005 .053 .09 1.67 33.4 1.55 2.7 7.10 .006
Upper
Bound 7.0 5.0 .03 .08 .3 3.5 43 3.0 7.0 11.0 .02
Lower
Bound 1.1 1.0 .005 .01 .05 .05 10.0 1.1 1.5 2.0 .004
HcCToud River Basin
LZFPM LZFSM LZPK LZSK UZK UZFWM ZPERC REXP UZTWM LZTWM PXADJ
Initial 20 5.6 .003 .026 0.06 3.0 43.3 1.60 4.0 5.0 1.1
Optimized 53.0 14.7 .0022 .004 .07 2.0 81.3 1.03 5.2 6.4 0.86
Upper
Bound 77.0 25.0 .03 .08 .3 9.5 98 3.0 11.0 21 1.7
Lower
Bound 1.1 1.0 .0005 .003 .05 .05 10.0 1.01 1.0 2.0 0.6
Merced River Basin
LZFPM LZFSM LZPK LZSK UZK UZFWM ZPERC REXP UZTWM LZTWM PXADJ
Initial 6.0 2.9 .006 .01 .11 1.9 34 1.8 4.8 7.2 0.90
Optimized 3.7 3.0 .016 .05 .08 2.80 49.5 2.30 1.2 8.1 0.98
Upper
Bound 27.0 10.0 .03 .08 .3 9.5 68 3.0 11.0 21.0 1.7
Lower
Bound 1.1 1.0 .0005 .004 .05 .05 10.0 1.01 1.0 2.0 .05
American River Basin
LZFPM LZFSM LZPK LZSK UZK UZFWM ZPERC REXP UZTWM LZTWM PXADJ
Initial 2.17 1.88 .009 .055 0.09 1.60 36 1.10 1.68 7.27 .91
Optimized 2.70 2.6 .006 .056 .13 1.4 24.3 1.11 1.47 6.03 .92
Upper
Bound 7.0 5.0 .03 .08 .3 3.5 43 3.0 7.0 11.0 1.7
Lower
Bound 1.1 1.0 .005 .01 .05 .05 10 1.1 1.0 2.0 0.5
1-25
-------
Lettenmaier
2.6.1 Predpitatkm
For each GCM alternative, the National Center for Atmospheric Research (NCAR) provided a disk file
of the output corresponding to the center of each grid cell used by the given model, along with a program to read
and print the output This program provided, among other variables, the predicted GCM precipitation
corresponding to a base case (nominally, present conditions) and the alternative climate. For all cases except
Case 3 (GISS transient X the base cases and alternative climates were represented as long-term monthly averages.
NCAR computed the ratio of the GCM alternative long-term precipitation to the base case long-term
precipitation (or, in the case of the GISS transient, the ratio was computed for each decade over the 80-year
transient run). The NCAR program was modified slightly to interpolate the results to longitude 120°W, latitude
40°N, which is approximately the centroid of that part of the Sacramento-San Joaquin basin that contributes most
of the runoff. For all but the transient run, the computed precipitation ratio was then applied to all the raw input
precipitation records.
In the case of the transient run, a statistical test (Spearman's rho; see for example, Conover, 1971) was
applied to the 80-year sequence of decadal precipitation factors to determine whether there was a statistically
significant trend For those cases (four months) where the trend was significant, a linear regression was fit to
the decadal precipitation Table 2.11, soil moisture accounting model parameter description factors, and the
resulting "ramp" was used to adjust the historical precipitation records. In those cases (eight months) where the
trend was not significant, the average precipitation factor (over the eight decades) was computed, and this
average precipitation adjustment was used in the same way as were the NCAR-computed factors for the steady-
state runs.
2.62 Temperature
For the steady-state runs, a temperature shift was computed as the difference between the 2xCO, and
base condition. These monthly differences were then applied directly to the historic data. In the case of the
GISS transient run, the sequences of eight decadal shifts were tested for trend, in the same manner as were the
precipitation factors. An months were found to have statistically significant uptrends, so a linear regression was
fit to all months, and the resulting ramps were used as input to the hydrologic models.
263 Potential Evapotranspiration
Potential evapotranspiration (PET) was computed from the Penman equation, which is given in
Veihmeyer (1964). Penman's equation is based on a theoretical energy balance approach. It predicts PET as
a function of temperature, average wind speed, humidity, mean solar radiation, and the ratio of duration of bright
sunshine to maximum possible duration of bright sunshine.
Penman's equation was applied on a monthly basis, using average values of the input variables. Some
of the input variables (wind speed) are not well known, so they were adjusted to obtain a total annual
evaporation estimate that was roughly consistent with observed pan evaporation at stations throughout the
Central Valley (usually on the order of 5545 inches of pan evaporation on the valley floor).
Once an adequate "fit" was obtained, the input values were compared with selected station values to
make sure they were physically realistic The most sensitive input value was wind speed, which can be strongly
affected by local factors, so the trial and error approach seemed justified Nonetheless, the assumed (seasonally
constant) value of 200 miks per day that was used was remarkably dose to the observed long-term mean at Red
Bluff, which has one of the longest records in the Central Valley.
The Penman PET was then recomputed for each GCM using two sets of temperature data: the base
condition for that GCM, and the predicted temperature corresponding to the CO, doubling. The monthly
differences in the Penman PET were computed and these differences were then applied to the historic data as
input to the hydrologic models. In the GISS transient case, the differences corresponding to the base and altered
climate temperatures at the beginning and end of the temperature "ramp" were computed, and the resulting
Penman PET differences were used to define a PET "ramp."
1-26
-------
Lettenmaier
Table 2.11. Soil Moisture Accounting Model Parameter Description
Soil
Moisture
Phase
Direct
runoff
Upper
zone
Perco-
lation
Lower
zone
Initial
water
Climatic
Index
No
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Parameters
PCTIM
ADIMP
RIVA
EFC
UZTWM
UZFWM
UZK
ZPERC
REXP
LZTHH
LZFSH
LZFPN
LZSK
LZPK
PFREE
RSERV
SIDE
UZTHC
UZFHC
LZTHC
LZFSC
LZFPC
ADIHC
PXADJ
PEADJ
Description
Minimum impervious area (percent)
Additional impervious area (percent)
Riparian vegetation area (percent)
Effective forest cover (percent)
Upper zone tension water capacity (inch)
Upper zone free water capacity (inch)
Daily upper zone free water withdrawal rate
(ZPERC+l)xPBASE is the maximum percolation rate
Exponent for the percolation equation
Lower zone tension water capacity (inch)
Lower zone' supplemental freewater capacity (inch)
Lower zone primary freewater capacity (inch)
Daily supplemental withdrawal rate
Daily primary withdrawal rate
Fraction of percolation water passing directly
to LZFN storages
Fraction of lower zone free water cannot be
transferred to LZTVJ
Ratio of non-channel baseflow to channel baseflow
Upper zone tension water content (inch)
Upper zone freewater content (Inch)
Lower zone tension water content (Inch)
Lower zone supplemental free water content (inch)
Lower zone primary free water content (Inch)
Tension water contents of the ASIHP area (Inch)
Precipitation adjustment factor
ET-demand adjustment factor
1-27
-------
Lettenmaier
CHAPTERS
RESULTS
As described in Chapter 2, the long-term hydrologic response of the study catchments was simulated for
climates associated with a base case (nominally, present conditions) as well as three sets of GCM predictions
(Cases 1, 2, and 4) that were based on steady state climate following a doubling of current atmospheric CO2
concentrations. In addition, the hydrologies of the study catchments associated with a transient climate resulting
from doubling of atmospheric CO2 over an 80-year period (Case 3) were also evaluated.
Three alternative climates (Cases 5-7) were used to test the sensitivity of the hydrologies of the study
catchments to selected aspects of the GCM climate predictions. These included a scenario (Case 5) designed
to explore the relative effect of predicted precipitation and temperature change, a scenario (Case 6) designed
to explore the effect of different interpretations of the predicted geographic distribution of temperature and
precipitation changes, and a scenario (Case 7) in which the long-term climate was assumed to be similar to that
experienced in the 1930*5. Cases 0-4 are referred to as "initial scenarios," and are discussed in Section 3.1. Cases
5-7 are referred to as "sensitivity scenarios," and are discussed in Section 3.2. The temperature shifts and
precipitation scaling factors associated with Cases 1-7 are given in Tables 3.1 and 3.2, respectively.
Because the snowmelt and soil moisture accounting models operate on daily or shorter time steps, and
all cases involved running the snowmelt and soil moisture accounting models for 100 years, large amounts of
computer output were generated. To simplify the analysis of the results, we selected the following model
(simulated) variables to summarize the alternative hydrologies: 1) average snow water equivalent over each
study catchment; 2) monthly average study catchment runoff; 3) monthly average study catchment
evapotranspiration; 4) average end of month study catchment soil moisture storage in selected zones (see Section
2.4.2 for description of the model soil moisture zones); 5) predicted (disaggregation model) monthly primary
node streamflow; and 6) predicted (linearly regressed) monthly secondary node streamflow. In all cases, the
mean of each variable was computed for the 100-year simulation period, as well as the standard deviation
(expressed as a coefficient of variation in some cases). In the interest of brevity, only selected mean values are
reported in this chapter; Appendix B contains a complete graphical summary of the results.
3.1 Study Catchment Results for Initial Scenarios
The hydrology of the Sacramento-San Joaquin basin in general, and the study catchments in particular,
is dominated by high-elevation snow accumulation in the winter and snowmelt in the spring and early summer
months. Snow water storage is especially important because of the disproportionate amount of precipitation that
occurs at high elevations, and because snow water storage shifts the peak of the annual runoff hydrograph from
the high precipitation winter months toward the spring and summer. Increasing winter temperature decreases
the amount of precipitation falling as snow, and causes any snow that accumulates in the winter to melt earlier.
This is the principal mechanism by which the alternate climates affect the hydrology of the Sacramento-San
Joaquin basin. For this reason, the order of presentation of the results is as follows: predicted average snow
water equivalents, monthly average runoff, monthly average evapotranspiration, and monthly average soil
moisture. Before presenting the results, however, a brief discussion of the climate alternatives is in order.
All four of the initial scenarios (Case 1: GFDL 2xCOy Case 2: GISS 2xCO2, Case 3: GISS transient,
and Case 4: QSU 2xCO2) predicted increasing average temperature for all months (see Figure 3.1). Generally,
the largest temperature changes were predicted by the GFDL and GISS models, and the smallest by the QSU
model. Figure 3.1 also shows the beginning and ending values of the "ramp" fit to the Case 3 temperature
transients. In general, the final values in the transient case are slightly less than the Case 1 and Case 2 steady-
state values, but are larger than Case 4. Figure 3.1 shows that there was no consistency in the predicted
long-term precipitation changes. The GISS model (Case 2) generally predicted an increase in precipitation in
the winter months. Because the precipitation regime is highly seasonaX summer precipitation changes for the
catchments are of much less importance hydrologically. The GFDL model predicted increased fall precipitation
and generally decreased winter and spring precipitation. The GISS transient and the OSU model had less
obvious patterns, with predicted increases in some months and decreases in others.
1-28
-------
Lettenmaier
Table 3.1. Temperature Shifts in °C for GCM Cases 1-7
Case
Jan Feb Mar Apr Hay June July Aug Sep Oct Nov Dec
1 3.5 4.35 4.5 4.55 5.70 6.10 4.35 3.90 4.90 4.30 4.10 3.45
2 5.9 4.6 4.6 5.4 3.1 4.1 3.9 5.50 7.20 5.3 3.4 4.4
3a -.02 .33 .10 -.03 -.33 -.05 -.02 -.65 .87 -.46 .48 -1.61
4.13 3.35 3.0 4.22 2.29 2.68 4.10 3.28 4.55 4.14 4.72 3.26
4 0.55 2.03 1.31 2.08 1.97 2.68 2.12 3.12 2.39 1.58 3.08 2.57
5 5.9 4.6 4.6 5.4 3.1 4.1 3.9 5.50 7.20 5.3 3.4 4.4
6b 6.60 5.06 4.78 5.15 3.14 4.0 3.31 5.98 8.55 5.33 3.47 4.69
4.82 3.84 4.24 5.77 3.18 4.15 4.80 4.88 5.17 5.19 3.24 4.00
7 -0.16 -0.54 0.97 0.98 0.43 0.04 0.36 0.90 0.65 0.53 0.62 0.66
3 The upper and lower entries represnt the beginning and end transient temperature shifts for the 80-year
period ramp used for Case 3.
The first row represents temperature shifts for the 3 nothern catchments and the second row the sourthern
catchment.
1-29
-------
Lettenmaier
»
Table 3.2 Precipitation Scaling Factors for GCM Cases 1-7
Case
Jan Feb Mar Apr Nay June July Aug Sep Oct Nov Dec
1 .936 1.13 .924 .911 .710 .947 .11 .889 .818 1.084 1.189 1.096
2 1.21 1.06 1.34 .764 .976 1.28 .655 1.26 .765 1.30 1.24 1.15
3a .954 1.082 1.123 .836 1.112 .864 .908 1.542 2.157 1.096 1.182 .783
1.171 1.383 1.564 1-070
4 1.03 1.13 1.04 1.36 .93 1.05 1.01 .88 1.07 0.89 0.76 0.89
5 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
6b 1.227 1.156 1.525 1.019 1.106 1.439 1.061 1.289 0.615 1.369 1.397 1.110
1.181 0.911 1.045 .363 .771 1.038 .017 1.220 0.999 1.184 .989 1.216
7 0.854 0.994 0.845 0.934 0.879 1.15 0.622 0.249 0.67 0.807 .593 .949
3 Only months with an upper and lower entry had a statistically significant transient. For months with a
significant transient, the upper and lower entries represent the beginning and end precipitation scaling factors
used in the 80-year ramp for Case 3.
b The upper entries represent precipitation scaling factors for the three northern catchments and the lower
entries represent the southern catchment (Merced River).
1-30
-------
Lettenmaier
8.0
o
CO
g 6-0
i—
o
GC
case
2 i
+ 2 GISS
T 3 GISS Transient
(dotted lii
+ * osu '
lines)
OCT DEC FEB RPR
MONTH
JUN
RUG
OCT
2-5
OCT
DEC
FEB
RPR
MONTH
JUN
RUG
OCT
Figure 3.1 Monthly temperature shifts (1) and precipitation scaling
factors (2) predicted by general circulation models for
climate scenario Cases 0-4
1-31
-------
Lettenmaier
Figures 3.2a-d show long-term study catchment-average snow water equivalent, runoff,
evapptranspiration, and soil moisture storage, respectively, by month for Cases 0-4. Figure 33a shows the
predicted average annual runoff by decades, while Figure 33b shows the catchment monthly mean flows of four
successive 25-year periods for the transient case (Case 3). The results for Case 3 shown in Figures 3.2a-d
represent averages over the 100-year simulation period and do not represent the transient nature of the Case 3
response. In Figure 3.2d, the soil moisture zone to which the model performance was most sensitive is indicated.
Because of the differences in the geologic and hydrologic characteristics of the catchments, the particular
subsurface zone varies by catchment. For this reason, the relative values are of greater importance than the
numerical values of the predicted storages.
The results of the simulations are reported in the remainder of this chapter. In most cases, the results
are reported as averages over the 100-year simulation period described in Chapter 2. For selected model
variables, standard deviations over the 100 year simulation period were computed. Brief interpretations of the
results for snow water equivalent, runoff, evapotranspiration, and soil moisture are given hi Sections 3.1.1-3.1.4,
respectively.
3.1.1 Snow Water Equivalent
Figure 3.2a shows long-term study catchment-average snow water equivalent by month for Cases 0-4.
There was a marked reduction in average snow water equivalent for all of the study catchments for all of the
alternative climates (Cases 1-4). The predicted changes, when expressed as a proportion of the base case average
snow water equivalent, were the greatest for the North Fork American River, which has the lowest mean
elevation, and the least for the Merced River, which is the highest basin. In Cases 1 and 2, snow storage is
virtually eliminated in the lower elevation study catchments.
Although the proportionate reduction in mean snow water equivalent was the least for the Merced River
because it is influenced more by snowfall and snowmelt than the other basins, the magnitude of the reduction
in snow water equivalent was larger than for any of the other study catchments. The Merced lost more snow
water equivalent than the North Fork American River or Thomes Creek had in the base case. The Case 1 and
2 temperature increases are large enough to eliminate almost all the influence of snow water storage from all
study catchments but the Merced.
3.1.2 Runoff
Figure 3.2b shows the predicted changes in the seasonal distribution of study catchment runoff. The
effect of reduced snow storage is immediately apparent; in all cases, the annual hydrograph peak shifted earlier
hi the year because of a decrease in the amount of snowfall in relation to rainfall. Runoff increased markedly
in all cases for the winter months and decreased substantially in the late spring and summer. The predicted
changes in annual runoff relative to the base case are reported in Table 33. Table 3.4 summarizes the months
in which predicted runoff is reduced from present conditions.
Although the general shift in the annual streamflow hydrograph was consistent in all of the catchments,
site-specific effects were observed as well. For instance, for the Merced River, which has high snowmelt-runoff
between March and July, the effect of increased precipitation in Case 2 was overshadowed by the temperature
increase. The North Fork American River has a mixed rainfall-runoff and snowmelt-runoff regime for the base
case. This changes to a complete rainfall-runoff situation for Cases 1 and 2. The phase shift is less for Case 4
because the temperature increase is more modest. Because the North Fork American River has a large
rainfall-runoff component for the base case, the increase in whiter flows was not proportionally as large as for
the Merced, which is snow-dominated in the base case. For Thomes Creek there was relatively less snowmelt,
and the base flow in the summer approached zero for all cases including the Base Case. Case 2 resulted in the
largest runoff changes, doubling the December and January mean flows and increasing February by about
two-thirds. Although the mean flow and standard deviation increased, the coefficient of variation for the Thomes
Creek runoff was generally reduced
1-32
-------
Merced
RPR
MONTH
= 400
North Fork American
DEC
FEB
RPR
MONTH
JUN
RUG
OCT
;400
Thomes
FEB flPR
flONTH
£550
McCloud
OCT
DEC
FEB
flPR
MONTH
JUN
flUO
OCT
Figure 3.2a Study catchment monthly mean weighted snow water equivalent for climate scenario Cases 0-4
e.
n
-------
Merced
150
onsc
CRSE l
CRSE 2
CRSE 3
CRS£ 4
flPR
MONTH
200
North Fork American
DEC
FEB flPR
MONTH
JUN
PUG
OCT
Thomes
McCloud
DEC FEB flPR JUN RUG OCT
MONTH
OCT DEC FEB RPR JUN RUG OCT
MONTH
Figure 3.2b Study catchment monthly mean streamflow for climate scenario Cases 0-4
-------
Mercecf
O
A CflSE I
CASE 2
CASE 3
cnse 4
North Fork American
UJ
DEC
FEB OPR
MONTH
JUN
RUG
OCT
o
CRSE t
+ CRSE 2
CASE a
cnsE 4
OCT DEC FEB flPR JUN RUG OCT
Thomes
OCT DEC FEB flPR JUN RUG OCT
McCloud
OCT DEC FEB
JUN flUG OCT
MONTH
£
£
Figure 3.2c Study catchment monthly mean evapotranspiration for climate scenario Cases 0-4
S.
n
-------
Merced
0 onsc
CASE
4- CRSE 2
CRSE 3
CRSE 4
FEB flPR JUN RUG OCT
MONTH
,3-00
North Fork American
o
I CASE I
+ CRSE z
CRSE 3
CRSE 4
FEB flPR JUN RUG OCT
.J .50
OCT
• 1.75
Thomes
OCT DEC FEB flPR
JUN RUG OCT
McCloud
OCT DEC FEB flPR
Figure 3.2d Study catchnent monthly near, soil moisture for climate scenario Cases 0-4
-------
1000
t—
$600
8
I600
:
, 200
Merced
0 10 20 30 40 SO 60 70 80 90 100
YEAR
1000
vu
it! BOO
600
x
„ 400
1
I
200
North Fork American
0 10 20 30 40 SO 60 70 60 90 100
YEfW
1000
800
600
Thomes
X
„ 400
200
0 10 20 30 40 50 60 70 80 90 100
YEflR
1000
tfeoo
600
X
- 400
200
McCloud
0 I0 20 30 <0 SO 60 70 60 90 100
YEPR
Figure 3.3a Study catchment decadal mean flows for transient climate scenario (Case 3)
n
S.
n
-------
Merced
North Fork American
FEB RPR
YERR
OCT DEC FEB RPR JUN RUG OCT
Thomes
McCloud
flUC OCT
20
OCT DEC FEB flPR
JUN RUG
OCT
Study catchment monthly flows for four successive 25-periods for transient climate
scenario (Case 3)
-------
Lettenmaier
Table 33. Simulated Mean Annual Study Catchment Runoff for GCM Cases 0-4.
Case
0
1
2
3
4
Annual Runoff in Thousands of Acre- feet
Merced
River
518.0
509.3
596.5
551.2
500.6
North Fork
American River
679.0
680.0
795.1
733.3
669.1
Thomes Creek
267.7
304.4
367.7
297.1
269.1
McCloud
River
448.2
427.7
591.6
521.7
406.0
1-39
-------
Lettemnaier
Table 3.4. Months with predicted reduction in streamflow, based on the climate change predicted by the GCMs.
Basin
Merced
NF American
Thomes
McCloud
Case
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
Months
May to September
May to September
May to August
May to November
April to September
April to September
May to September
May to November
April to August
April to August
April to August
May to November
May to November
June to September
None or negligible
May to January
The importance of groundwater for the McCloud River is shown in Figure 3.2b. Although all scenarios
shifted the flows earlier, only Case 2 overwhelmed the soil moisture storage capacity of the model. Case 2
produced peaks in January and March, with slightly lower flows in February, but the entire winter had a
predominance of rainfall runoff. The standard deviation of flows generally increased also, as did the coefficient
of variation.
The shift in the annual distribution of runoff is critical for water resources management; the seasonal
distribution of runoff, and its variability, are more important than the annual runoff change. The implications
of these changes in seasonal runoff distribution, and variability, are addressed in detail by Sheer and Randall
(1988). For most of the catchments and climate change scenarios, runoff variability was substantially increased
in the winter months and reduced slightly in the summer months.
3.13 Evapotranspiration
As described in Chapter 2, potential evapotranspiration (PET) was estimated for the alternative climate
scenarios using the Penman equation with the GCM-predicted temperature changes. In all cases for all basins,
the maximum actual evapotranspiration simulated by the soO moisture accounting model occurred earlier in the
year (See Figure 3.2c). Case 1 generally resulted in a flatter crest for the monthly mean ET than was predicted
for the base case. For Thomes Creek, Cases 1 and 2 were flatter, lower, and peaked earlier than Cases 3 and
4. ET depends on soil moisture as well as PET; therefore, although PET increased for all months for Cases 1-4
because of the increased temperature, the direction of changes in actual ET varied by month.
The soil moisture-accounting model assumes that ET depends on the moisture contents of the conceptual
tension zones. The rate of ET declines as the soil dries. Therefore, the shift of the flow from spring to winter
shifts the ET similarly. ET also depends upon temperature, so that wet winter soils dp not yield as much ET
as similarly wet spring soils. The net result, despite the change in seasonal distribution, was relatively little
change in annual total ET. Although the changes in seasonal distribution of ET were similar over all the study
catchments, there were differences among the cases. Cases 1 and 2 have the greatest increase in mean
temperature, and they show the greatest shift in timing of ET. Case 4 predicts the least temperature change,
and it shows the least change in ET. The variability of the ET increased more or less in proportion to the mean.
1-40
-------
Lettenmaier
The increased spring ET suggests that agricultural irrigation demand might be increased over the present
situation.
3.1.4 Soil Moisture Storage
Figure 3.2d shows long-term catchment-average soil moisture storage by month for Cases 0-4. As
described in Chapter 2, the soil moisture accounting model has five conceptual storage zones. The capacity of
the soil moisture zones strongly affects the response of the model to altered inputs. The soil moisture zone
capacities were estimated through the automated calibration process described in Chapter 2. While neither their
capacities nor contents are measurable physical quantities as are, for instance, snow water equivalent or runoff,
the soil moisture storages do reflect, in a general sense, the physical soil moisture storage capacities of the
catchments.
Figure 3.2d summarizes the results for the soil moisture zone to which the model performance was most
sensitive hi each study catchment Because of the differences in the geologic and hydrologic characteristics of
the study catchments, the particular subsurface zone varies by catchment. For this reason, for comparisons
between catchments, emphasis is placed on the relative values.
There are few distinctive trends observed between the basins, but some general climate-related changes
in most of the catchments can be seen. Some of the generalizations that follow do not apply to the McCloud
catchment, which behaves differently than the others because of its extremely large soil moisture storage capacity.
The warmer and generally wetter climates predicted by the GCMs cause increased rainfall relative to snowfall,
making more moisture available during winter and early spring, at the expense of late spring and summer.
Therefore, there is a definite phase shift in practically all storages (see Figure 32d). This trend is stronger
among the free water zones than the tension zones, and during summers than winters. In addition, Cases 1 and
2 exhibit larger phase shifts than 3 and 4 since temperature and precipitation changes for the former are higher.
The moisture content of the upper tension zone is virtually unaffected by climate change during the wet
October-March period because it has the first priority to absorb moisture and has small capacity (usually less
than 2 inches). It is filled in all the climate scenarios. During the spring and early summer, the decrease in
snowmelt gives rise to more severe moisture shortages, which are reflected in reduction of tension water storage.
Because Cases 1 and 2 reflect the greatest warming, they result in larger phase shifts than Cases 3 and 4. The
lower tension zone, which is supplied after its upper counterpart, shews larger phase shifts than the upper tension
zones, in particular during the October-March period. This is partly because after supplying the upper zone, the
net moisture available is reduced Moreover, partly due to the generally larger capacities of the lower tension
zones in all the catchments, the lower tension zones are more affected by the availability of moisture.
The free water zones' contents are influenced by climate change. Not only do they exhibit larger phase
shifts, but the changes are also more erratic between basins and among the various cases. The upper free water
zone shows larger fluctuations in moisture content than the lower zones (Figure 3.2d). The upper free water
zones are comparatively more sensitive to modest changes in precipitation than are the other zones (Figure 3.1),
especially for Case 2 in the McCloud River catchment
32 Sensitivity Scenarios
The sensitivity areiMrfof were undertaken for three purposes: 1) to determine the extent to which
precipitation (as opposed to temperature) change drives the simulations for the various climate scenarios (Case
5); 2) to determine the relative effect of an interpretation of the GCM predictions that provides differential
input to the southern (Merced) study catchment relative to the three northern study catchments (Case 6); and
3) to evaluate the general character of the hydrologic scenarios associated with the GCM predictions relative
to the historic 1930's drought (Case 7). The first issue was addressed by using the GISS GCM predictions (Case
2) with the precipitation adjustment factors set to 1.0, but with the Case 2 temperature shifts retained (this
formed the new Case S). The second issue was addressed by assigning temperature shifts and precipitation
1-41
-------
Lettenmaier
factors associated with the GISS 2x00, model ceU (Case 2) centered at 35-22"N, 120°W to the Merced River
study catchment, and the GISS grid cellcentered at 43.040N, 120°W to the three northern study catchments (as
opposed to an interpolated grid cell centered at 40.0°N, 120°W which was assigned to all study catchments in
Cases 1-4). This formed the new Case 6. Case 7 was formed by computing the ratio of the average monthly
precipitation for selected Northern California stations for the period 1931-40 to the long-term (1890-1980)
precipitation at the same stations. Temperature shifts for the 1931-40 decade relative to the 1890-1980 means
at the same stations were also computed. For each month, the median precipitation ratio, and temperature shift,
were computed. These medians were used to define a climate scenario (Case 7) which was analogous to the
GCM scenarios.
The sensitivity scenarios are summarized in Sections 3.2.1-3.2.4.
The Case 6 (geographic sensitivity to inputs) results showed that the hydrologic simulations were
qualitatively the same (at least for the GISS model) regardless of whether the grid cells were interpolated as
described in Chapter 2, or a nearest cell center approach was used. This result held for all variables (snow water
equivalent, runoff, evapotranspiration, and soil moisture). Therefore, Sections 3.2^-3.2.4 are devoted to a
discussion of the temperature sensitivity (Case 5) and 1930's analog (Case 7).
3.2.1 Snow Water Equivalent
The difference between Case 2 and Case 5 indicates the sensitivity of the snow water equivalent to
GCM-predicted temperature changes only. There was relatively little difference in this respect between Case
2 and Case 5. This confirms the earlier interpretation of cases 1-4 that changes in winter snow accumulation
were primarily temperature dependent The Case 2-predicted temperature changes were quite large; hence,
there would be relatively little snow in any of the study catchments except the Merced River under the CO2
doubling scenarios.
The Case 7 (1930*5 analog) results emphasize how different the 1930*5 conditions were from the GCM
predictions; under these conditions the simulated snow water equivalents were slightly less than the base case,
but were much larger than any of the GCM climate scenarios. This reflects the fact that the 1930's drought was
caused primarily by a reduction in precipitation; winter temperatures were relatively little changed from the
base case.
3.2,2 Runoff
For the precipitation sensitivity scenario (Case 5), the Merced River had lower mean monthly flows and
standard deviations due to the reduction in winter rainfall relative to Case 2, but the phase shift in runoff was
more or less the same. This is no surprise because temperature, which determines whether precipitation
occurred as rainfall or as snowfall, was the same as in Case 2. Case 7 represents the precipitation and
temperature analog of the relatively dry decade of 1930's. For Case 7, the temperature change was minor
relative to the base case, an increase of less than 1°C. Therefore, the seasonal distribution of runoff was
comparable to the base case. However, all the months had lower flows and standard deviations. This reflects
the dry conditions (most precipitation factors less than 1.0) experienced in the 1930's.
For the North Fork American River, the November to March flows (mean and standard deviations) were
significantly lower in Case 5 than in Case 2. Due to minimal changes in temperature, Case 7 had a similar mix
of rainfall runoff and snowmelt runoff as the base case. The changes in runoff variability (standard deviations)
were generally larger than the changes hi runoff volumes for the 1930's analog.
The Thomes Creek responses for Cases 5 and 7 showed qualitatively similar patterns to those observed
for the North Fork of the American River. This is to be expected because the hydrologies of these catchments
are somewhat similar.
The McCloud River simulations showed more marked impact of the reduced precipitation in Case 5
relative to Case 2. On average, runoff occurred earlier, with increased ET and reduced soil moisture contents.
1-42
-------
Lettenmaier
The groundwater recession began about two months earlier. Again, the McCloud catchment behaved somewhat
differently than the other four catchments; the Case 5 results show that the interaction of changed temperature
and precipitation was stronger in this catchment than in the other three. This is primarily a result of the damped
storm response, which heightened the importance of between-storm dynamics.
3.23 Evapotranspiration
The Merced River is the only catchment for which there was any substantial change in ET for Case 5
compared to Case 2. This can be traced to the greater importance of snow water storage in this catchment under
the alternative (warmer) climates. Reduction of precipitation in Case 5, relative to Case 2, reduced the average
snowpack, and in consequence the spring and summer soil moisture. The amount of moisture available for
meeting the daily ET demands was relatively unaffected by reductions in moisture supply, until the tension water
storages were reduced to near zero. This was the case only for the Merced catchment. For the three catchments
which are rainfall-driven in the wanner climate scenarios, ET was relatively little affected by the Case 5 reduction
in winter precipitation.
For Case 7 (1930's analog) there was a slight phase shift in Case 7 ET for all the basins relative to the
base case. The shift patterns was similar for all the study catchments. The 1930's analog climate had little effect
on the mean ET except during summers (June to September) when the soil moisture contents of tension zones
are reduced to near zero.
3.2.4 Soil Moisture Storage
More of the soil moisture reduction associated with the reduced precipitation in Cases 5 and 7 came
from the upper free water zone than the lower free water zones in both Cases 5 and 7. The McCloud catchment,
however, had a much larger secondary free water capacity in the lower zone than the other study catchments
(about 10 inches storage capacity) to maintain its high baseflow. Therefore the reduction in precipitation in both
Cases 5 and 7 gave rise to relatively higher reductions in moisture storages in the McCloud River than in the
other catchments.
Generally, the lower free primary zone had larger capacities than the secondary zone. The moisture
content of the McCloud catchment was as high as 40 inches, while the maxima for the other catchments were
in the vicinity of only three inches. For both lower free zones, the phase shifts during summers and early
autumns were more consistent than the other seasons because these seasons experience larger changes in
moisture supply.
33 Spatial Disaggregation: Primary Nodes
Section 2.43 described the spatial disaggregation model used to generate monthly streamflows at the
six primary sites given in Table 22. For the base case, the disaggregation model explicitly preserved the mean
and variance of the flows at the primary nodes. Higher moments (e.g., skewness) were not preserved. The
model also explicitly preserved the correlation between the sites in each month, but not the lagged correlation
at a given site or between sites (for instance, the January correlations between node 1 and node 2 were
preserved, but the January-February correlations at node 1 were not preserved). The model implicitly assumes
that the statistical structure (moments) of the node flows relative to the catchment flows would remain the same
under the alternative climates scenarios.
Figure 3.4a shows the monthly mean simulated flows for each of the six primary sites for Cases 0-4, and
5-7, respectively. These figures confirm that the disaggregated (primary node) flows were qualitatively consistent
with those of the catchments. The phase shifts between various cases, as well as the high flows and low flows
at the primary nodes, were generally comparable to those of the catchments. There was, of course, a substantial
difference in runoff volumes, because the nodes represent much larger drainage areas than the catchments. For
example, Site 2 (Sacramento River at Red bluff) had monthly runoff as high as 15 x 106 acre-feet, which is 10
1-43
-------
Lettenmaier
400-0
OCT DEC FEB BPR JUN flUC OCT
.2700.0
o
2 cnse t
+ CRSE 2 •
o OWE s
CRSE 4
FEB RPR
WWTH
JUN PUG OCT
OCT DEC FEB APR JUN RUG OCT
,1000-0
DEC FEB RPR JUN flUG OCT
.1600-0
OCT DEC FEB RPR JUN flUC OCT
.1500-0
OCT DEC FEB RPR JUN BUG OCT
Figure 3.4a
Water resources system primary node mean monthly flows for
climate scenario Cases 0-4.
1-44
-------
Lettenmaier
times higher than the highest study catchment mean flow. On a relative basis, however, the primary site annual
runoff hydrographs were similar to those of the catchments.
3.4 Spatial Disaggregation: Secondary Nodes
The seven secondary nodes selected for this study are listed in Table 2.1. The secondary nodes generally
represent smaller, lower elevation drainages than do the primary nodes. Because they lie at lower elevations,
some of the secondary nodes have a large number of zero flows (and in some cases zero mean flow) in the
historic record for the summer and fall months. Generally, the influence of the secondary site flows on the water
resources management model described by Sheer and Randall (1988) is considerably less than that of the primary
nodes. This consideration supported use of the simple regression model to simulate the secondary site flows
conditioned on the (simulated) flow at a specified primary node (Eq. 2.4).
Figure 3.4b shows the mean simulated flows at the secondary sites for Cases (M. For those nodes where
summer flows were nearly zero under the base condition, the net effect of the warmer climates was to increase
the runoff in nearly every month. Because summer runoff was zero under the base case, no reduction was
possible. Site 1 provides a good example of this effect For those sites (for example, site 7) where there was
some summer runoff for the base case, the change in the annual runoff hydrograph was more similar to those
of the study catchments and the primary nodes. In Cases 1-4, the simulated increase in winter runoff exceeded
the decrease in summer and fall. Cases 5-7 show similar results. Case 7 (193ffs analog) showed the expected
reduction in the mean flows during the winter months. This is the only case in which there was a significant
reduction in the secondary site flows.
1-45
-------
5001-
300
RPR
MONTH
RPR
MONTH
300
OCT
flPR
MONTH
JUN
DEC
FEB RPR
MONTH
JUN
RUG
OCT
RUG
OCT
Figure 3.4b Water resource system secondary node mean monthly streamflow for climate scenario Cases 0-4
-------
300
700 r-
(O
DEC
FEB
flPR
MONTH
JUN
PUG
OCT
flPR
MONTH
JUN
PUG
OCT
650
OCT DEC FEB RPR JUN RUG OCT
MONTH
Figure 3.4b (Continued)
e.
n
-------
Letteamaier
CHAPTER 4
SUMMARY AND CONCLUSIONS
The primary objectives of this work were to develop a methodology to provide a descriptive
interpretation of the hydrologic effects of global climate change as predicted by selected GCMs, and to apply the
methodology to the Sacramento-San Joaquin basin as a case study. This study is preliminary; the simulation
results discussed in Chapter 3 should be interpreted in a "what if?" and not a predictive sense. Furthermore,
existing hydrologic models (in particular, the National Weather Service snowmelt and soil moisture accounting
models) were used, and the assumptions and simplifications incorporated in the models are reflected in the
results. The following assumptions were made either implicitly or explicitly:
1) The altered climate scenarios were identical to current climate except that all precipitation was adjusted
by a factor equal to the ratio of the selected GCM average monthly precipitation to the base case average
monthly precipitation, and temperatures were shifted by an amount equal to the difference between the selected
GCM scenario and the base case. This has the following implications for the hydrologic simulations:
o Hydrologic systems are strongly affected by the variability of the driving variables, as well as their mean.
In the case of precipitation, adjustment by a fixed factor implies that the coefficient of variation (standard
deviation divided by the mean) is the same for the altered climate scenarios as for the base case. For
precipitation factors greater than one (as for Cases 1-4 in most of the winter months) this means that
the standard deviation of the inputs will increase. This particular assumption regarding the form of the
altered input affects the stochastic structure of the output, which may well have significant implications
for water resources management.
o The spatial variability of the inputs was assumed to be exactly the same as in the base case. The
performance of large multiple site water resource systems, such as the California State Water Project
and the Central Valley Project, can be strongly affected by the spatial correlation of streamflows. If
streamflows at the different sites (rivers) are highly correlated, droughts are likely to occur
simultaneously at all, or many, sites. If the spatial correlation is less, and the storage locations are
dispersed, the required storage will be smaller. It is unlikely that the spatial correlation of the inputs
would remain the same under substantially warmer climates, but the GCM grid spacing is too coarse
to allow alternative inferences to be made at present
o The precipitation arrival process (that is, the probabilistic structure of wet and dry day sequences) was
assumed to be unchanged from the historic record While the GCMs provide precipitation predictions
at time scales of one day or less, the interpretation of these predictions in terms of point precipitation
(as recorded at a precipitation gage) is difficult. The GCMs provide grid cell average values, but spatial
averaging over an area the size of a grid cell removes most of the information regarding the point arrival
processes. Further work is needed to verify the relationship of GCM-predicted short-term precipitation
to observable quantities (e.g., gage or gage-averaged precipitation corresponding to the base case).
Changes in the precipitation arrival process affects catchment runoff response, even in the absence of
changes in the longer term (e.g., monthly) statistics. For example, fewer storms of increased rainfall
intensity are likely to lead to increased runoff, reduced soil moisture, and decreased ET in the long run.
2) The hydrologic models provide an adequate description of the catchment dynamics under the altered
climate. Two major issues arise in this respect. The first regards the appropriateness of the hydrologic models
to the base case. The National Weather Service River Forecast System, of which both the soil moisture
accounting model and snowmelt model are part, has been widely used and verified operationally for a range of
hydrologies. The soil moisture accounting model, in particular, was originally developed for use in the
Sacramento basin. Although other hydrologic models might have been used, we believe the NWS models contain
about the right level of detail for medium sized catchments, and can be expected to capture the essential
elements of the long-term (as opposed to event) hydrologic response. The issue of applicability under alternative
1-48
-------
Lettenmaier
climate scenarios is more difficult to address. One major problem is that the soil moisture model cannot
explicitly account for long-term changes in vegetation. At a minimum, the implicit assumption that the model
relationship between PET and ET will hold in warmer climates needs to be verified. In the longer term,
hydrologic models capable of simulating long-term runoff need to be explored.
3) The spatial disaggregation model provides an adequate description of the relationship between the study
catchment flows and the flows at the water resource system nodes. Two key limitations related to the spatial
disaggregation model must be considered:
o As described in Chapter 2, the spatial disaggregation model is unable to preserve the lagged correlations
in the primary basin flows. Generally, the simulated primary basin flows for the alternative climate
scenarios (including the base case) were found to have lower lag-one correlations, especially during
the high runoff winter months, than the historic flows. This result is of concern because correlations
affect the estimated reliability of the water resource system. These correlations do not affect any
conclusions regarding the hydrologic response of the study catchments to the climate alternatives. The
undersimulation of the lagged correlations is the result, in part, of groundwater inflow to the water
resources system nodes that is not present in the higher elevation study catchments. Selection of a
different set of catchments might allow use of a more complex disaggregation model (see Chapter 2)
that could resolve this problem. However, it appears likely that none of the potential catchments would
be strongly enough affected by groundwater in the winter months to allow feasible parameter estimates
for this model. An alternative would be to model the groundwater effect directly. This would likely be
a time-consuming undertaking. In terms of the relative importance of this problem, it is probably less
than changes in the stochastic structure discussed below, or any of the issues relating to the hydrologic
model discussed above. One reason for this is that the study design provided that comparisons be made
with a base case, rather than historic streamflows; therefore the reduced winter correlations are
evidenced in the base case, as well as the alternative climate scenarios.
o The structure of the stochastic relationship between the primary water resource system node streamflows
and the study catchment flows is constant under the climate alternatives. One problem with this
assumption is that the study catchments lie at high elevations and are affected by changes in snow
accumulation patterns. Some of the contributing areas to the water resources system nodes lie at lower
elevation and are rain-affected under present conditions. Therefore, the effect on the water resources
system nodes of a general warming would be different than in the catchments, resulting in a likely
overestimation of the effect of the altered climate scenarios. Again, this problem affects only the water
resource system node runoff predictions and not the interpretation of the catchment results (Chapter
3). In addition, because the Sacramento-San Joaquin hydrology is snow-affected, the nodes that
contribute most of the inflow to the water resource system will not be much affected by this problem.
The problem is likely to be greatest for the low-lying secondary nodes, whose flows under the climate
alternatives will likely be somewhat overestimated Each of these assumptions, and the related
limitations imposed, suggests a direction for future research.
Recognizing the preliminary nature of the work and the limitations imposed by the assumptions, the
following general conclusions can be made:
o The general warming associated with all the GCMs would result in substantial decreases in average snow
accumulations in all four of the study catchments.
o Reduction in the amount of precipitation occurring as snow would increase winter runoff and decrease
spring and summer runoff.
o Increased precipitation occurring as rainfall in the winter months would increase winter soil moisture
storage, and would make more moisture available for ET in the early spring. Increased temperatures
would increase spring ET.
1-49
-------
Lettenmaier
The reduction in moisture supply as snowmelt in the spring, coupled with increased spring ET, would
reduce late spring, summer, and fall soil moisture, which would in turn reduce runoff during those
periods.
Although the above points suggest the general character of the changes that would occur under a general
global wanning for any given catchment, the specific nature of the hydrologic change would depend on
physiographic characteristics (notably, the area-elevation distribution) of the catchment, as well as the
geologic and topographic features that control the precipitation-runoff response. Substantial hydrologic
diversity existed between catchments, especially the McCloud River, which drains an area of deep
volcanic ash in the vicinity of Mount Shasta and has exceptionally persistent baseflow.
1-50
-------
Lettenmaier
REFERENCES
Anderson, EA., "National Weather Service river forecast system - Snow accumulation and ablation model,"
NOAA Technical Memorandum NWS HYDRO-17, November, 1973.
Burnash, RJ.C., Ferral, R.L. and Mcquire, RA., "A generalized streamflow simulation system Conceptual
modeling for digital computers," UJS. National Weather Service, Sacramento, CA, 1973.
Conover, WJ., "Practical nonparametric statistics, Wiley, New York, 1971.
Foufoula-Georgiou, E., "Discrete-time point process models for daily rainfall", Water Resources Series Technical
Report No. 93, Department of Civil Engineering, University of Washington, March, 1985.
Crotch, S.L., "Regional intercomparisons of General Circulation Model predictions and historical climate data",
Technical Report DOE/NBB-0084, Atmospheric and Geophysical Sciences Division, Lawrence Livermore
National Laboratory, April, 1988.
Hansen, J., Russell, G., Rind, D., Stone, P., Laus, A., Lebedeff, S.,
Ruedy, R. and Travis, L., "Efficient three-dimensional global models for climate studies: Models I and II",
Monthly Weather Review, 111(4), 609-662,1983.
Hansen, J., Fung, I., Lacis, A., Rind, D., Lebedeff, S., Rueddy, R.,
Russell, G. and Stone.P., "Global climate changes as forecast by Goddard Institute for space studies
three-dimensional model," Journal of Geophysical Research, 93(D8), 9341-9364,1988.
Hettmansperger, T.P., Statistical inference based on ranks, Wiley, New York, 1984.
Hirsch, R.M., "Statistical methods and sampling design for estimating step trends in surface-water quality,", Water
Resources Bulletin, 24(3), 493-504,1988.
Klemes, V., "The Hurst phenomenon: A puzzle?", Water Resources Research, 10(4), 675-688,1974.
Levtham, K.M., "Physical considerations in the analysis and synthesis of hydrologic sequences", C.W. Harris
Hydraulics Laboratory Technical Report No. 76, Department of Civil Engineering, University of Washington,
1982.
Lettenmaier D.P. and Surges, SJ., "Climate Change: Detection and its impact on hydrologic design", Water
Resources Research, 14(4), 679-687,1978.
Linsley, R.K., Kohler, MA. and Paulhus J.L.H., Hydrology for Engineers,
2nd Ed., McGraw Hill, New York, 1975.
Nelder, J A. and Mead, R., "A simplex method for functional minimization", Computer Journal, 9,308-313,1965.
Peck, E.L., "Catchment modeling and initial parameter estimation for the national weather service river forecast
system", NOAA, Technical Memoandum, NWS HYDRO-31,1976.
Rind, D., Goldberg, R.R. and Ruedy, R., "Change in climate variability in the 21st century", unpublished
document, Goddard Space Flight Center, 1988.
1-51
-------
Lettenmaier
Sheer, DJP. and Randall, D.R, "Methods for evaluating the potential impact of global climate change on water
related activities: Case studies from Atlanta, Georgia and the State of California", report to U.S. Environmental
Protection Agency Office of Policy, Planning, and Evaluation, May, 1988.
Veihmeyer, FJ., "Evapotrasnspiration," Chapter 11 in Handbook of applied hydrology (V.T. Chow, ed), McGraw
Hill, New York, 1964.
Wallis, JH. (ed) "Climate, climatic change, and water supply," National Academy of Sciences, Panel on Water
and Climate, 1977.
1-52
-------
METHODS FOR EVALUATING THE POTENTIAL IMPACTS OF GLOBAL CLIMATE CHANGE:
CASE STUDIES OF THE STATE OF CALIFORNIA AND ATLANTA, GEORGIA
by
Daniel P. Sheer, PhJX, P£.
Dean Randall, Ph.D., P.E.
Water Resources Management Inc.
Columbia, MD 21045
Contract No: CR814637
-------
CONTENTS
FINDINGS 2-1
CHAPTER 1: INTRODUCTION 2-2
USING CIRCULATION MODEL OUTPUT TO ASSESS THE POTENTIAL IMPACTS OF
GLOBAL CLIMATE CHANGE 2-2
CHAPTERS CALIFORNIA CASE STUDY 2-4
METHODS 2-4
RESULTS AND INTERPRETATION 2-6
SUMMARY 2-16
CHAPTER 3: ATLANTA CASE STUDY 2-17
METHODS 2-17
RESULTS AND INTERPRETATION 2-17
SUMMARY 2-25
REFERENCES 2-28
11
-------
Sheer
1
FINDINGS
Case studies to evaluate the potential impacts of global climate change on water supply were performed
for the Central Valley in California and the Atlanta, Georgia, metropolitan area. Streamflow traces, which were
generated from the GCM models' outputs, were routed through hydrologic mass balance operation simulation
models of the two systems.
In California, the total volume of streamflow for the system increased under each of the climate change
scenarios. However, the seasonably of the flows changed such that the winters are wetter and the already dry
summers are drier. This results in an increased probability of spring flooding and reduced water deliveries to
the consumers. It is doubtful that changing operating policies for the system will significantly improve this
situation.
The streamflows of the Chattahoochee River in Atlanta show the same shift in seasonality as occurs in
California However, one scenario shows an increase in total streamflow while the other two show a decrease.
Under the GISS 2xCO2 scenario, there is more flow, and Lake Lanier levels tend to be higher. Under the
GFDL 2xCO2 scenario, there is less flow, and the lake levels drop significantly. The effect on recreational use
of Lake Lanier under this scenario would be disastrous.
'Although the information in this report has been funded wholly or in part by the US. Environmental
Protection Agency under contract no. CR814637, it does not necessarily reflect the Agency's views, and no official
endorsement should be inferred from ft.
2-1
-------
Sheer
CHAPTER 1
INTRODUCTION
One of the most important goals of research on the potential for global climate change is to prepare society
to adapt to such changes. Many of the most dramatic impacts of climate change, should it occur, will be caused
by changes in the availability and reliability of water supplies. California, in particular, is likely to be affected by
climate change because of the relatively high rate of utilization of the water resources in the state. Georgia and
the remainder of the southeastern United States are continuing to feel the impacts of prolonged below-normal
rainfall through the decade of the 1980s.
Unfortunately, there has been little work on the development of techniques for assessing the potential
impacts of global climate change on water resources related activities. In fact, even long-term historical
climatological data are neither widely nor well used in the planning or operation of water resources systems in
the United States. Most facility and operations planning is based almost entirely on hydrologic records, either
historical or generated. While recent advances in the analysis of water use data and rainfall runoff modeling
make it desirable to use climatological data, the techniques are just beginning to find their way into practice. This
lack of commonly used techniques makes it quite difficult to estimate the impacts of either potential or predicted
changes in long-term climate (such as information on global warming) on water resource-related activities.
The goal of the research performed for this study is to determine how global climate changes might affect
water supply availibility. Case studies were performed for the Central Valley of California and metropolitan
Atlanta, Georgia. This is done by using streamflow traces generated from global circulation model output as
input to hydrologic models that simulate the operation of the systems. This method was used because it allows
one to analyze each area as an integrated system to show how water use might change under given climate
change scenarios. It also allows the operating policies of the systems to be changed to see if operations can be
improved to help counteract the climate change.
USING CIRCULATION MODEL OUTPUT TO ASSESS THE POTENTIAL IMPACTS OF GLOBAL
CLIMATE CHANGE
The process of assessing the impacts of global climate change and developing options to mitigate those
impacts is essentially similar to the process of using climatological data in planning. There are substantial
complications, however. Most of these deal with the lack of calibration of techniques for predicting climate
change (in particular global circulation models), the problems of converting output from those techniques to
hydrologic traces, and finally, problems of assessing the likely changes in water use that result directly from
climate change.
The basic tool for converting climatological data to hydrology is a rainfall runoff model Such models use
precipitation, temperature, and sometimes wind data at a scale comparable to that available from common
sources, typically the National Weather Service. Usually there are several gauging stations proximate to the basin
of interest, and data from these stations are interpolated to produce a trace of climatological data suitable for
running a model. Unfortunately, the scale of techniques suitable for assessing potential global climate change
is much larger than the scale of most rainfall runoff models, e.g., the area represented by a single grid point in
a global circulation model may encompass several basins, which each might be calibrated using tens (or even
hundreds) of weather stations. Orographic effects cause large local variations in climatology, and these effects
are often magnified in the local hydrology. The level of detail of techniques for assessing future climate is simply
not yet adequate to address the issues of local changes so critical to hydrology.
2-2
-------
Sheer
The calibration of techniques that convert climatological data to hydrology involves setting parameters that
either directly or implicitly represent the evapotranspiration of water by local ecosystems. Techniques are
generally quite sensitive to these parameters. Substantial changes in climate are likely to produce corresponding
changes in ecosystems, thus invalidating calibrations based on historical data. And estimating the biological
changes that might occur and affect hydrology is made more difficult by substantial impact of the hydrology itself
on the kinds of ecosystems that might be supported. In preparing the hydrology for use in the California and
Atlanta case studies, these issues were addressed directly (Lettenmaier, this volume; Hains, this volume),
although their results show how much more research is needed in this infant field.
Just as difficult is the task of predicting water use under conditions of changed climate. Climate has direct
influence on many economic variables including water requirements for cooling, cropping patterns, and
population density. All of these economic factors will help determine the water use requirements. To estimate
the overall impact of climate change, full consideration must be given to these items. The relationship between
demand and climate was addressed in the Atlanta case study, and the results are included as an appendix. For
lack of time, the results were not included in the simulation models, however.
It is most important, in assessing the impacts of global climate change, to realize that water resources systems
continually adapt to new information concerning existing climate, and that if climate change happens slowly
enough, the systems will adapt to the change incrementally. Indeed most predictions conclude that climate, if it
does change, will change slowly, i.e., the annual rate of change in climatological norms will be much smaller than
the normal interannual variation.
As a result, responses to climate change are likely to occur in the same manner as responses to more
information about current climate. That is, actual changes and implementation of alternatives will be driven
primarily by extreme events. Such changes can be either operational or structural, and if the best possible
assessment of impacts and alternatives is to be done, the techniques used must be capable of examining both
kinds of responses.
In the case studies below, we attempted to implement techniques, particularly simulation models, that could
account for the factors mentioned above. This method can be used to solicit local input regarding the nature
of the incremental adaptations that might occur in the face of a gradual change in the climate. In both California
and Atlanta, local water managers were exposed to the results of the studies using these techniques, and the
results are reported below.
It should be noted that the traces of precipitation and temperature from the global circulation models is at
best a guess. The simulation models used in this study are likely far more accurate than the streamflows (which
were generated from the global circulation model output) that were used as inputs.
2-3
-------
Sheer
CHAPTER 2
CALIFORNIA CASE STUDY
The State of California and the federal government have enormous investments in facilities designed to deal
with the vagaries of existing climate variation in California. The federally operated Central Valley Project (CVP),
and the state-operated State Water Project (SWP) (as well as other state, federal, and local projects) move water
from the wet and sparsely populated northern part of the state to the dry and densely populated southern part,
and hold water during wet years or seasons for use in drier periods. This water is used for irrigated agriculture
and for municipal and industrial supply. In addition, under current law much of the storage in the two systems
is committed to maintaining minimum levels of outflow to San Francisco Bay. Despite the investments, in recent
years the demand for water has exceeded the supply. With or without climate change, this is likely to occur
even more frequently in the future.
The distribution of water shortages in California is a function of the water rights of individual users under
the state's water law. California is an appropriation doctrine state, so generally those who began using water first
have by far the most reliable supplies. Complex arrangements have evolved, based on water rights, for water
exchanges and sharing, which dictate the shortages borne by various users. Simulation models are used to test
these arrangements, based on historical hydrology modified to account for future development.
These models can also be used to evaluate the impacts a climate-induced change in hydrology might have
on water users and on instream flows. Before they can be used in this way, however, suitable hydrologic traces
must be prepared. These traces must be consistent not only with the given climate change scenario, but also with
the host of other assumptions made in developing the water resource simulation model. Finally, care must be
taken in the manner in which the simulations are run and the results reported.
Operational strategies for large-scale water resource systems can have an enormous influence on the overall
performance of the system. Such strategies are based, either implicitly or explicitly, on the statistics of historical
hydrology. This is particularly important in California. If climate changes, it is likely to change rather gradually,
and streamflow statistics will change gradually as well. Over time, operations have changed, and the changes have
been driven in large part by the occurrence of occasional extreme drought or flood. Recent floods and droughts
in California have, indeed, affected operations in exactly this manner. Therefore, it is vital that assessments of
the potential impacts of climate change consider the opportunities available to react to those changes through
changes in operating policies, including adjustment of drought management schemes and changes in reservoir
rule curves.
METHODS
Existing models and previous analysis of water use in California made it possible to estimate the potential
impacts of climate change on water deliveries in California, and on outflow to San Francisco Bay. The model
used was developed by Water Resources Management Inc. (WRMI) for the Metropolitan Water District of
Southern California (MWD), and was used with their kind permission (Sheer and Baeck, 1987). (MWD
currently uses the model in its planning efforts.) The version of the model used emulates the State of
California's Department of Water Resources (DWR) Planning Simulation Model (CA DWR, 1986).
The WRMI model does a mass balance simulation. It uses adjusted historical hydrologic inputs, projected
water use demands, instream and Delta outflow requirements, and operating policies for determining levels of
water deliveries to users in both the CVP and SWP. All of the major hydrologic features of the Central Valley
are included. A schematic of that model is shown in Figure 1. The model was designed to simulate the
operating policies that are currently used by both the CVP and SWP. The results of the simulation include time
traces of reservoir levels, deliveries to CVP and SWP water users, flows at points in the Sacramento River and
its major tributaries, and Delta outflow.
2-4
-------
Trinity
River
Upper
Sacramento
Feather
River
American
River
Ml
Cleor JjWhlskeytown
• V Creek
Tunnel
SIMPLIFIED SWP/CVP
SIMULATION MODEL
SCHEMATIC
Folsom South Service Area
Silverwcod
&
Perrlj
Lakes
Delta
Surplus
Son Luis
CVP
• I «11 01
Figure 1. Simplified SWP/CVP simulation model schematic.
-------
Sheer ,
The hydrology used as a basis for the California case study, hereafter called the "scenario flows," was
obtained from Lettenmaier (this volume). The flows he developed correspond to "virgin flows," unaltered by
changes in land use. As discussed above, the flows used by the DWR model, and therefore the WRMI model,
have to be adjusted to account for the projected levels of upstream water use and land development.
Five different traces have been run through the model:
1) Base run. This trace was generated using GISS output for lxCO2- It represents current climate and
is nearly identical to historical impaired flows.
2) GFDL2xCO2. This trace was generated using GFDL output for 2xCO2.
3) GISS 2xCO2. This trace was generated using GISS output for 2xCO2.
4) Oregon State University 2xCO2. This trace was generated using OSU 2xCO2 output.
5) GISS Transient. This trace is intermediate between the base run and the GISS 2xCO2 run. It
represents a gradual change in atmospheric CO2 from current levels to 2xCO2 over a period of
about 80 years. For this scenario it is impossible to differentiate between changes in flows due to
local, short-term weather changes and long-term climate changes. Therefore, the conclusions that
can be drawn from these runs are not clear, and no plots are provided.
A rather complex process for adjusting historical flows to future conditions has evolved for use by both the
CVP and the DWR in California, and those will not be discussed here. Flows which correspond to the inflow
at nodes in the DWR and WRMI models for 1990 conditions have been produced using this method. They are
called "impaired flows." The process is also run in reverse to produced what are called "unimpaired," or "full
natural" flows (CA DWR, 1987). These flows nominally represent the flows that would occur in the absence of
development. To correct the scenario flows for projected "impairments," the 1990 impaired flows were subtracted
from the 1990 unimpaired flows, and the resulting "impairments" were subtracted from the scenario flows to get
the hydrologic traces actually used in the simulation modeling. In equation format:
Scenario - (1990 unimpaired -1990 impaired) * new trace
Each of these scenarios was run with a monthly time step for the 57-year period from 1951 to 2007. The first
30 years of the record were derived directly from Lettenmaier's adjustment of the historical flows from 1951 to
1980. The remaining 27 years were taken by a random sampling (with replacement) for the first 30 years. A
period of record longer than 30 years was developed to allow the development of adjusted operating policies that
were not entirely tailored to the 30-year period being evaluated. If the adjustments to the rules were based solely
on the 30 years, the rules would then be "optimal" for that particular period, and would not reflect any
uncertainty as to the longer term. Fifty seven years was an arbitrary choice for period of record; it corresponds
to the length of the historical record normally used in developing operating policies for the SWP. This series
of runs is summarized in Figures 3 through 9, which show the response of the system to the change in inflows.
RESULTS AND INTERPRETATION
Figure 2 shows a comparison of the total monthly inflow into the Delta for the different scenarios. This
graph indicates that while the total streamflow entering the Delta is greater, the seasonality of the flows for each
scenario increases, with wetter falls and winters and drier springs and summers. (It should be noted that for the
current climate in the Central Valley most of the precipitation falls in the winter, and the summers are very dry.)
The increased seasonality is particularly strong for the GISS 2xCO2 case. The OSU 2xCO2 scenario is most like
current streamflows, although there is still a seasonality shift.
2-6
-------
K>
base
gfdl-2
FIGURE 2
Mean Monthly Delta Difference
giss-2 osu-2
XXV ///
(GCM-hist) Maf
10 -
7.5 -
5 -
2.5 -
n
u —
fi 1
rH , J Mn J
'
^
N
^
1
kJ
/
/
/
^
LH
s
/
U
4
1
J -|JLM ^^
J
s
n '
\
'
/
oct nov dec jan feb mar apr may jun jul aug sep yr
month
-------
Sheer
Figure 3 shows the deliveries to SWP under the four scenarios. The 2xCO- scenarios all show significantly
lower deliveries, especially GISS and GFDL, despite their increased annual flow. This dramatic decrease is
caused by the increased seasonally of the flows. A reasonable interpretation of this is as follows:
Increased temperatures lead to more winter precipitation falling as rain, thus reducing snowpack. Higher
winter temperatures also lead to the snowpack melting earlier. The early snowmelt spills from the
reservoirs instead of going into storage and being available to augment streamflow during the early part
of the irrigation season. The net effect is the loss of the large reservoir currently represented by the late
spring snowpack, and a corresponding decrease in the water that can be reliably supplied by the system.
The effect of the increased seasonality of the GFDL run seems to be offset by its larger runoff volume.
Figure 4 shows the mean annual volume of SWP deliveries for the four scenarios. For example, the GFDL
2xCO2 scenario results in about a 16% reduction in deliveries from the base case.
Deliveries to the CVP are not reduced under the scenarios. The U.S. Bureau of Reclamation estimates
that it can sell an additional 1 million acre-feet of water annually, which indicates that the CVP has a surplus
of storage. The SWP has much less storage, so deliveries suffer more during dry periods.
Figures 5, 6, and 7 show how the Oroville Reservoir storage, the primary storage facility for the SWP, is
affected by the different scenarios. At the end of March (Figure 5), which is the last of the wet months, the
storages for the scenarios are nearly identical. Storages in March are limited by the need to preserve free
volume in the reservoirs for flood control, so the storage for each scenario is at the flood control rule. Figures
6 and 7 show clearly that the reservoir storage is diminished owing to the reduced spring and summer flows.
Oroville reservoir is substantially lower by the end of May.
Total outflow from the Delta is shown in Figure 8. As one would expect, each of the 2xCO2 scenarios
shows an increase in total outflow.
Carriage water is an environmental constraint that is built into the model The H.O. Banks (SWP) and
Tracy (CVP) pump plants pump water from the south side of the Delta into the California Aqueduct and the
Delta Mendota Canal, respectively. At high discharges this pumping causes a southward current across the
Delta. During these times of high pumpage, a minimum amount of flow must be maintained through the Delta
During times of low flow, water must be released from the reservoirs to meet this requirement This release is
called "carriage water."
One of the impacts of a warming climate is a rise in sea level, which will affect the Delta Philip Williams,
who is studying the effects of the climate change scenarios on the Delta, suggested that the Delta carriage water
requirement should be doubled to counteract both the sea level rise and increasing salinity in San Francisco Bay.
The base and GFDL 2xCO2 scenarios were done with the doubled carriage water requirement A comparison
of SWP deliveries for those two scenarios with current and doubled carriage water is shown in Figure 9.
Base,lCW indicates the base run. Base, 2CW is the base case with double the doubled carriage water. GFDL2,
1CW and GFDL2, 2CW indicate the GFDL 2xCO2 case with current and doubled carriage water, respectively.
This shows that increasing the carriage water requirement would have little bearing on the deliveries for years
corresponding to the years from 1950 to 1980. It must be noted that the period from 1950 to 1980 does not
include the drought of record for the Central Valley of California (1927-1934). An analysis using those years
would show a substantial impact on deliveries during the drought. The carriage water will have a stronger effect
over a long, sustained drought than a short, intense drought as was experienced in 1977.
Several changes in operating rules might improve the performance of the system in terms of deliveries.
Reducing flood pools by increasing rule curve storage in April might reduce the relative drawdown at the end
of the year, and thus increase deliveries. However, this cannot be done without increasing the storage levels in
March (the wettest month), increasing the risk of a flood. With the higher flows in March, one might expect that
2-8
-------
s
t
1.95
base
FIGURE 3
Annual SWP Deliveries
1.97
(Thousands)
1.99
gfdl-2x
year
2.01
giss
-2x
osu-2x
-------
volume in KAF
4000 -
FIGURE 4
Mean Annual SWP Delivery
-i
3000 -
i
2000 H
1000 -
0
2834
2391
7
base
gfdl-2x
2430
2632
giss-2x
osu-2x
-------
LL
<^^-
*-£
II
oC
3.5 H
2.5 H
1.5 -
1 -
0.5 -
1.95
D * base
FIGURE 5
Oroville Storage (March)
1.99
(Thousands)
gfdl-2x
year
giss-2x
2.01
A osu-2x
-------
Ǥ
3.5 H
2.5 -
1.5 -
0.5 -
FIGURE 6
Orovllle Storage (May)
D base
gfdi-2x
(Thousands)
year
O giss—2x
osu-2x
-------
a-
3.5 H
2.5 H
1.5 H
0.5 H
1.95
O * base
FIGURE 7
Oroville Storage (Sep)
1.97
+ gfdi-2x
(Thousands)
year
O giss-2x
1.99
CA
2-01
A osu-2x
-------
tS
1
1.95
D base
FIGURE 8
Total Annual Outflow
1.97
-f gfdi-2x
(Thousands)
year
O giss-2x
1.99
2.01
osu-2x
-------
II
3 O
It
2.8 -
2.6 -
2.4 -
2.2 -
2 -
1.8 -
1.6 -
1.4 -
1.2 -
1 -
0.8 -
0.6
0.4 -
0.2 -
0
1951
FIGURE 9
Annual SWP Deliveries for 1,2xCW
ll i I
1961
1971
D base,1cw
O gfdl-2,1cw
1981
base,2cw
gfdj-2,2cw
1991
2001
CA
-------
Sheer
March storages might have to be further reduced to accommodate potentially larger floods. This would only
exacerbate the water supply impacts of climate change. Changing the rule curve at San Luis to increase water
held in March and April might also help.
A meeting was held on March 4, 1988, with representatives of the State DWR and the Bureau of
Reclamation to discuss these results. At that meeting the question of potential changes in operations to
accommodate changed climate was raised. The consensus among all involved, including the current operators
of the projects, was that the magnitude of the change in seasonality was such that operational changes alone
would make little if any improvement in system performance. This seemed especially true in light of the potential
for increased need for flood control storage.
Assuming that the scenarios are representative of the future climate, water deliveries from the CVP/SWP
will decrease unless storage is added to the system. Also, unless the current operating rules for flood control
storage change, there will be an increased threat of flooding, particularly in the Sacramento and Stockton
metropolitan areas.
SUMMARY
Four climate change scenarios were routed through WRMTs CVP/SWP simulation model to evaluate their
impacts on the California water supply system. Though the GISS and GFDL 2xCO2 scenarios produce more
runoff, the increased seasonality of the flows will result in a substantial reduction in water delivered. One can
speculate on what would be required to offset the loss in deliveries. Because the drought of record is 7 years
long, it would require about 7 acre-feet of storage to support an average annual delivery of 1 acre-foot at the
current reliability with the current climate, and more under a less favorable climate. Average deliveries fell some
440 thousand acre-feet (KAF) from the base run to the GFDL 2xCO2 run. Assuming that the less favorable
climate would raise the ratio from 7:1 to 10:1, some 4.4 million acre feet of additional storage would be required
to maintain current deliveries and reliability. The authors believe this estimate is low. Although sites for
additional storage of this magnitude of storage are available (by enlarging Shasta Reservoir or building Auburn
DamX their political feasibility and social desirability is problematic.
2-16
-------
Sheer
CHAPTERS
ATLANTA CASE STUDY
The Chattahoochee River is the main water supply source for the City of Atlanta and its environs. Lake
Lanier is formed behind Buford Dam, located about 30 miles upstream of Atlanta on the Chattahoochee. The
lake is used to maintain Chattahoochee flows, and thus reliable water supply for the region. Lake Lanier,
however, is a U.S. Army Corps of Engineers (COE) reservoir authorized for flood control, navigation, and
hydropower generation. No storage in the lake is specifically authorized for the purpose of water supply. Nor
is any storage allocated to recreation, despite the fact that the lake is the most heavily used of any COE lake in
the United States.
Minimum flows are maintained to enhance navigation on the Chattahoochee and Appalachicola Rivers
downstream of Columbus, Georgia, which is about 110 miles downstream from Atlanta. Navigation releases of
1000 cfs from May through October are required. Hydropower generated at Buford Dam is sold by the
Southeastern Power Administration. The minimum hydropower release is 500 fr/sec (cfs). The Chattahoochee
is the primary water supply source for about 74% of the 22 million people in the seven-county metropolitan
area, while about 9% of the people use water directly from the lake. The Atlanta Regional Commission (ARC)
does not expect those percentages to change.
The wet nature of the 1960s and 1970s kept lake levels high. However, droughts in the 1980s have had
significant impacts on recreation in the lake. Lake levels have dropped to historical lows, ranging problems for
boaters in particular. Power generation has also been affected, and releases to maintain instream flows have
been reduced. Despite these impacts, there has always been substantial storage remaining in the lake, and water
supply has never been seriously threatened.
METHODS
A mass balance model of Lake Lanier was developed for this simulation. A schematic of that model is
shown in Figure 10. Current water demands from the Chattahoochee were provided by staff at the ARC
Historical inflows for the initial development were provided by the COE. Simplified rule curves for Lake Lanier
operations were also provided by the COE, as weD as functions for converting releases to power generation, and
release requirements for navigation.
After the model was complete, a meeting was held in Atlanta involving the Georgia Environmental
Protection Division, the ARC, and the COE to review the reasonableness of the operating rules to be used for
the historical runs. All suggestions made at that meeting were incorporated, and the results of the model runs
seemed reasonable to all concerned.
The model produces traces of lake levels, power generation, and flows below Atlanta, a reach of the river
where water quality is critical A 30-year record (1950-80) with a monthly time step was used for aD the runs.
Inflows for the five scenarios, which are the same at those described in the California Case Study section of
thu report, were obtained from Haint (this votumeX Gains were computed by multiplying the ratio of historical
gams to inflows by the scenario inflows. AH demands and operating rule curves remained the same for the
scenarios.
RESULTS AND INTERPRETATION
Five dhTerent traces have been run through the model, which are indicated below. Figure 11 shows a
comparison of the total monthly inflows into Lake Lanier. The mean annual flow for the base case, which
represents current cfimate, is about 9% higher dun historical The GISS 2x00. scenario has stronger
SMMcmaBry than the base case with wetter winters and drier summers. The mean annual flow is about 18%
higher th» historical TheGI^feOOjrunexbiMtsthesanKtypeof seatonafyu
2-17
-------
Sheer
Lake
Lan1er
gain
Chottahoochee
Rl ver
evap
return
I
r
return
Chattahoochee
River below
Atlanta
demand
demand
Figure 10. Lake Lanier model schematic.
2-18
-------
FIGURE 11
Mean Inflow Difference - Lake Lanier
base
giss-2
gfdI-2
osu-2
\\Y
//
/
/
/
\\XVV"
(GCM-hist)/1000 Kaf
150
\\j\j
100 -
? 50 -
0
-50 -
-100 -
-150 -
?nn -
I
L
u n i ^j ^™ n n n
"LI J-J f"
n y
u
1
T-^n
\
-fl
\
s
$
s
\
\
/
'K
\
\
/
-, ^n
1
RJ r
/
/
si
S
S
|i
K
L
J^
/
/
ki
s
\
S
s
\
/
/
^>
V
\
\
1 k
IN
4_
^JJ
d
n
N
\
\
\
x
/
/
/
x
-
\
\
\
\
\
s
\
oct nov dec jan feb mar apr may jun jul aug sep yr
-------
Sheer
with noticeably less mean annual flow. The OSU 2xCO, run shows the same general characteristics as the
GFDL ZxCOj, with a less pronounced reduction in mean annual flow. This scenario is not included in the plots
because the GISS and GFDL 2xCO2 scenarios represent the extremes, so all OSU scenario traces will lie
between those two, and to reduce the clutter in the graphs.
Figures 12 through 15 show the results of these runs. They include Lake Lanier elevation, power generation,
and outflow below Atlanta for the base, GISS IxCOy and GFDL 2xCOr All plots were done using monthly
values. The COE operating rules have been used for all of the projected scenarios. It is likely that the operating
rules could be changed to provide better operations under the projected scenarios. The following conclusions
are drawn from these plots.
Lake Lanier elevations for September are shown in Figure 12. The target trace is the storage target for
recreation, elevation 1061.32 ft The highest target for the lake is 1067.08 ft (June) and the lowest is 1060.0 ft
(October - December). The lake elevations for base and 2xCO2 tend to be higher than the target The base
and 2xCX>2 traces fall below the targets 21 and 7 times in 360 months, respectively. One would expect higher
lake elevations from GISS 2xCO^ because of the increased inflow. The levels from the GFDL scenario are
noticeably lower with many more violations (261) of the lower target elevation; differences between the trace and
the target of 5 to IS ft are common. Under the GFDL scenario, the lake levels would have been continuously
under the targets from 1954 through 1961. The combination of the frequency and severity of low lake levels
would have a disastrous effect on recreation.
Mean annual outflows below Atlanta are shown in Figure 13. Two minimum flow requirements are used
downstream of Atlanta, 750 and 650 cfs. The higher value is used under normal conditions, while the lower value
is used when Lake Lanier falls below specified levels. (The rules for reducing downstream flows are explicit, but
are not followed exactly in practice. In 1988, flows downstream were reduced well before the reservoir fell to
the specified target level.) The outflows for GISS 2XCO2 are higher than base, as one would expect Neither
base nor GISS 2xCO2 violates the lower requirement, but they violate the upper requirement 29 and 43 times,
respectively. However, these violations are slight, typically about 10 cfs. Outflows for the GFDL run violate the
upper low flow requirement 117 times and the lower requirement 21 times. Again, given the lower magnitude
of inflows for GFDL, this should be expected.
Power generation is shown in Figures 14 and 15. The patterns indicate that the GFDL scenario yields less
energy. Figure 15 indicates that the total energy generated for the base and GISS scenarios is reasonably dose,
while GFDL yields about 38% less energy than base. Using a price of $0.03 to $0.10 per kilowatt hour (KWH),
this results in an annual loss of between $127,000 and $424,000 in energy generation. It should be noted that the
value of the power produced at Lake Lanier is not adequately reflected by a unit cost per KWH. This is because
the generators at the dam produce peaking power, and if they are not reliably available, additional fossil fuel or
nuclear generating units would be required at enormous capital cost. Calculating those values is beyond the scope
of this study.
Two additional runs of the model were made with different operating policies. The first was done to reduce
water supply withdrawals (deliveries) when lake levels are low because the COE has not authorized the reservoir
for water supply. This was done by specifying the recreation target levels as the water supply constraints. When
the lake level falls below the target level, no deliveries are allowed. The purpose of this model run is not to
suggest that that would actually happen, but to show the effect of giving recreation lake level targets precedence
over water supply. This model run is summarized in Table 1 and Figure 16.
The data shown in Table 1 indicate that the water delivery reduction policy has some effect for the base and
GISS scenarios, but that the results for GFDL are disatrous in terms of water shortages. About 40% of the
months have zero deliveries. One would expect that there would be fewer violations of level targets and low flow
requirements as less water is being withdrawn and consumed from the system. Figure 16 shows that this policy
does in fact lead to significantly higher lake levels, particularly for the GFDL scenario.
2-20
-------
base
FIGURE 12
Lake Lanier Elevation (September)
giss-2x gfdl-2x target
elev. (ft)
1075
1065 r-
1055
1045
50
55
60
65
year
70
75
80
-------
base
giss-2x
FIGURE 13
Outflow below Atlanta
gfdl-2x
flow - cfs
5000,
4000 H
3000
2000
1000
Ql 1—I—I
50
I I I I
55
J I I L
j—i i i L
J I J L
J 1 I L
60
65
year
70
75
80
-------
base
MWH per year
20000
15000 -
10000
5000^-
giss-2
FIGURE 14
Lake Lanier Power Generation
gfdl-2
Ql 1 I I I
50
55
60
65
year
70
75
80
-------
MWH
15000
FIGURE 15
Mean Annual Power Generation
12500 -
10000 -
7500 -
5000 -
2500 -
0
12163
11171
6935
base
giss-2
gfdl-2
-------
Sheer
Table 1. Comparison of Standard, Water Supply Reduction and Navigation Reduction Model Runs
mean annual energy
(MWH)
mean demand supplied
(cfs)
# times delivery - 0
# times level < target
# times outf.< 750 cfs
# times outf.< 650 cfs
scenario
Base
GISS-2
GFDL-2
Base
GISS-2
GFDL-2
Base
GISS-2
GFDL-2
Base
GISS-2
GFDL-2
Base
GISS-2
GFDL-2
Base
GISS-2
GFDL-2
none
11200
12200
6900
621
621
621
0
0
0
21
7
261
29
43
117
0
0
21
reduction
water
11200
12200
7300
608
620
379
8
1
143
16
7
222
26
42
10
0
0
0
navi-
gation
11200
12200
7300
612
620
449
0
0
0
16
7
204
27
42
60
0
0
0
The second modified operating policy was done to see the effect of eliminating the navigation flow
requirement, with the belief that there is little economic benefit in providing these releases. In this case, when
the lake levels fall below the target levels, no navigation releases are made. Deliveries were handled as they
were for the standard policy. The lake level graph of this policy is shown in Figure 17. Comparing this with
Figure 14 again shows that this policy significantly increases lake levels.
SUMMARY
A hydrologic mass balance model was built to simulate the operation of Lake Lanier and the Chattahoochee
River above Atlanta, Georgia, to address how a global climate change might affect the water supply of
metropolitan Atlanta. Four climate change scenarios were routed through the model Each scenario indicates
a shift in seasonality of flows, with drier summers and wetter winters. Under the GISS 2xCO- scenario, levels
in Lake Lanier are similar to the base case, while GFDL 2xCO2 shows a definite decrease in levels.
2-25
-------
base
elev. (ft)
1075
1070 p-
y^
1065 -
1060
1055
1050 f-
1045
50
FIGURE 16
Lake Lanier Elevation (September)
giss-2x gfdl-2x target
55
60
65
year
70
75
80
-------
base
elev. (ft)
1075
1070 P-
1065;-
1060
1055^
1050 -
FIGURE 17
Lake Lanier Elevation (September)
giss-2x gfdl-2x target
1045
50
55
60
65
year
70
75
80
-------
Sheer
REFERENCES
California Dept of Water Resources, "California Central Valley Unimpaired Flow Data," Sacramento, CA,
February 1987.
California Dept of Water Resources, "Operations Criteria Applied in DWR Planning Simulation Model,"
Memorandum Report. Sacramento, CA, February 1986.
Sheer D.P., and M.L. Baeck, Documentation of the CVP/SWP Sjunvlatjpp Models developed bv WRML Water
Resources Management, Ino, Columbia, MD, September 1987.
2-28
-------
THE IMPACTS OF CLIMATE CHANGE
ON THE SALINITY OF SAN FRANCISCO BAY
Philip B. Williams
Philip William* and Associates
Pier 35, The Embarcadero
San Francisco, CA 94133
Contract University of Washington #503011
-------
CONTENTS
FINDINGS 3-1
CHAPTER 1: INTRODUCTION 3-3
STATEMENT OF PROBLEM 3-3
SAN FRANCISCO BAY MORPHOLOGY 3-3
SAN FRANCISCO BAY HYDRODYNAMICS 3-6
SAN FRANCISCO BAY HYDROLOGY 3-6
CHAPTER 2: METHODOLOGY 3-8
OVERALL APPROACH 3-8
SAN FRANCISCO BAY MORPHOMETRY 3-8
TIDAL EXCHANGE CHARACTERISTICS 3-9
SALINITY RESPONSE 3-9
CHAPTER 3: RESULTS AND INTERPRETATION 3-12
MORPHOMETRY 3-12
HYDRODYNAMICS 3-12
SALINITY 3-12
LIMITATIONS AND UNCERTAINTY 3-20
CHAPTER 4: POLICY IMPLICATIONS 3-28
THE FUTURE OF THE DELTA 3-28
THE ESTUARINE ECOSYSTEM 3-28
EXPORT OF WATER FROM THE DELTA 3-28
REDUCING THE RANGE OF UNCERTAINTY 3-28
REFERENCES 3-29
-------
Williams
FINDINGS1
With future predicted climate change due to doubling of carbon dioxide levels, there would be major
changes in the hydrology, morphology, tidal hydraulics, and salinity of the San Francisco Bay Estuary.
The major change in hydrology would be the alteration of freshwater inflow, primarily from the Sacramento
River, which would affect estuarine circulation and salinity distribution.
Irrespective of hydrologic modifications, the San Francisco Bay Estuary would experience major changes
in morphology, circulation, and salinity due to climate-induced accelerated sea-level rise.
For a sea-level rise of 1 meter (33 feet) associated with a doubled carbon dioxide level, the area and
volume of the estuary would increase. The extent of the increase would depend on whether grating, often
substandard, levees were strengthened and maintained. If they were not, and large areas were flooded, the area
of the estuary could triple from approximately 1,100 to 3,500 square kilometers, and its volume double from 7
to 14 cubic kilometers. If all the existing levees were unproved and maintained, the increase in area would be
about 30% and volume 15%.
The exact amount of sea-level rise in the range of 1/2 to 2 meters is less important hi determining the
future physical characteristics of the estuary than whether or not levees are allowed to fail
The most vulnerable part of the estuary to inundation by rising sea level is the Delta, an agricultural area
whose fragile levees are susceptible to failure with the current sea level Failure of the Delta levees in
themselves would add about 1,500 square kilometers to the estuary and create a vast inland arm of the San
Francisco Bay.
These morphological changes could greatly affect the movement of sediment hi the Bay by capturing more
sediment in the upper Bays and particularly the Delta. Secondary effects could include:
- Erosion of mudflats and salt marshes fringing the Bay, aggravating losses due to sea-level rise.
- Increased clarity of water, possibly increasing algal production.
- Increased wave energy within the Bay, causing erosion of levees.
The effects of morphologic changes could be counterbalanced by hydrologic changes. Higher winter flood flows
could increase sediment delivery to the Bay.
The morphological changes would greatly alter the movement of water hi the Bay. Such effects would
include:
• Movement of the estuarine-induced circulation system upstream.
Changes hi tidal characteristics.
- Increased tidal current velocities.
'Although die mformation in tin report has been funded wholly or partly by die UJS. Environmental
Protection Agency under Contract #503011, it does not necessarily reflect die Agency's views, and no official
endorsement should be inferred from it
3-1
-------
Williams
The change in hydrodynamics due to sea-level rise effects would greatly alter the salinity distribution in
the estuary, causing a given average salinity level to migrate roughly 15 kilometers upstream.
Increased runoff predicted by climate models tends to partially compensate for the increase in average
annual salinity caused by sea-level rise. The net result is for a given average annual salinity level to migrate up
to 10 kilometers upstream.
All the climate models indicate significant average increases in salinity in the estuary in the spring and
summer. Salinities formerly exceeded in 20% of years would be exceeded in about 50% of years.
Increases in salinity will greatly affect the transfer of water across the Delta for irrigation supply. In order
to maintain the present low salinity levels of exports from the Delta, the amount of water released from
reservoirs for salinity repulsion (carriage water) would have to be doubled and/or physical modifications made
to Delta channels.
The following major policy decisions would be required to minimize economic, social, and environmental
costs due to climate change:
• Determination of which low-lying areas surrounding the estuary would be protected and which
abandoned.
- Identification of institutional and financial mechanisms to provide for protection of low-lying areas.
- Alteration of present water allocation and water management practices to incorporate sufficient
freshwater inflow to the estuary to maintain a viable estuarine ecosystem.
- Alteration of water allocation and water management practices in order to optimize beneficial uses of
water diversions from the Delta.
3-2
-------
Williams
CHAPTER 1
INTRODUCTION
STATEMENT OF PROBLEM
San Francisco Bay is the largest estuary on the U.S. Pacific Coast. Its estuarine ecosystem is dependent on
the amount and timing of freshwater inflows and the resulting salinity distribution throughout the Bay. The
upper part of the estuary, the Sacramento-San Joaquin Delta, acts as a conduit for water supply transfers from
the northern part of California to the arid San Joaquin Valley and Southern California Consequently, salinity
management in the estuary is vitally important to protecting the estuarine ecosystem and the quality of water
exported for agriculture and urban water users.
Projected climate change due to the greenhouse effect will greatly alter the salinity distribution in the
estuary. There are two primary impacts:
o The timing and amount of freshwater inflow will change.
o Sea-level rise will alter the tidal characteristics of the Bay, with dramatic changes taking place if levees
protecting low-lying areas are allowed to fail.
This investigation is a reconnaissance-level study intended to provide an indication of the possible extent of
these impacts. Because it was prepared in a short timeframe, and with a limited budget, simple analytic
techniques were used. In addition, existing water management practices and physical facilities in place were
assumed unchanged for all scenarios. It is anticipated that future studies will carry out more refined analysis to
produce results that can be used for planning purposes.
SAN FRANCISCO BAY MORPHOLOGY
The San Francisco Bay Estuary was formed in the last 10,000 years as rising sea level inundated the lower
valley of the Sacramento River. It has a complex morphology of interconnected bays separated by straits (see
Figure 1). Hydrodynamically, the estuary is considered in two parts:
o The Northern Reach, consisting of the Central Bay, linked with San Pablo Bay by San Pablo Straits,
connected to Suisun Bay by the Carquinez Straits, which in turn are linked with the tidally influenced
Delta through the narrows downstream of Sherman Island. The rivers that drain the 153,000 sq.
kilometers. Central Valley discharge into the Delta. The northern reach usually functions as a partially
mixed to well-mixed estuary, except during very high river flows.
o South San Francisco Bay, which connects with the Central Bay through the narrows at Yerba Buena
Island. Because this Bay receives little direct freshwater inflow, it usually acts as an estuarine-influenced
lagoon with isohaline conditions reflecting salinity conditions in the Central Bay.
The morphology of San Francisco Bay has been greatly altered by man's activities. Almost all of the once
extensive 2^00 sq. kilometers of tidal marshes have been converted primarily to agricultural land, but also to salt
ponds, port development, military, and urban uses (see Figure 2). Most of the former tidal marsh area has
subsided owing to compaction and in some areas owing to groundwater or natural gas extraction. Typically, 1
to 2 meters of subsidence has occurred, except in the Delta. In the central and western Delta, the freshwater
tidal marshes created deep peat soils that are excellent for agriculture. However, when tilled, these soils oxidize
and blow away, lowering the land elevation up to 80 mm (3 inches) per year. The Delta is now composed of
3-3
-------
Approximate location of the
10.0 foot NGVD contour
indicates maximum area affected
by 100-yr. high tide with
SAN FRANCISCO BAY ESTUARY
Numbers indicate miles upstream
of Golden Gate
SCALE OF MILES
048
3-4
FIGURE 1
-------
GENERALIZED AND APPROXIMATE HISTORIC CHANGES
IN AEREAL DISTRIBUTION OF TIDAL MARSHES
Moray pristine since
dots of erection
Source: Atwater et al., 1979
Mostly leveed or
filled by people
during the indicated '
period of time
APPROXIMATE
DATE OF CREATION
OF MARSH
Before or after 86O
After I860
1860-1970
1940-1970
1900-1940 11111 \ Before I860
k. I860 -1900 P|jU
FIGURE 2
3-5
-------
Williams
approximately 50 islands at elevations as low as 6 meters below mean sea level (NGVD)2 (DWR, 1987). Many
of these islands are separated by narrow tidal channels and are protected by fragile peat levees. Even with the
existing sea level, levee failures are frequent and the islands are costly to maintain (Josselyn and Atwater, 1982).
The typical 100-year high-tide flood elevation is approximately +2 meters NGVD in the Delta (DWR, 1987).
SAN FRANCISCO BAY HYDRODYNAMICS
The salinity distribution within San Francisco Bay is dictated by mixing processes governed by tidal ebbs
and flows, wave action, and the freshwater flows entering the Delta. With its present configuration, the mixed
semi-diurnal tides propagate up the northern reach from the Golden Gate as a progressive wave, with the tidal
lag increasing and the tidal range decreasing inland to Sacramento. In the South Bay, the tides create an
oscillating standing wave that tends to amplify the tidal range.
The freshwater inflow at the upstream end of the estuary creates an estuarine circulation cell upon which
many estuarine species depend. During periods of high winter floods, a stratified circulation can be created with
low surface salinity as far as the Golden Gate. As flows drop, the estuary rapidly mixes and at low and moderate
flows becomes well to partially mixed.
During high-flow periods, lower-salinity water from the Central Bay stratifies the South Bay. This greatly
reduces residence time of pollutants in the heavily urbanized South Bay and is also important in ecosystem
productivity.
SAN FRANCISCO BAY HYDROLOGY
Most of the freshwater inflow to the estuary is contributed by the Sacramento River, with a lesser amount
from the San Joaquin River. Except during high winter flood flows, inflows are controlled by releases from the
extensive system of reservoirs upstream. Some of the inflow to the Delta is exported by the pumps of the State
Water Project and Central Valley Project, some is used for irrigation on the Delta islands, and the remainder
flows downstream into Suisun Bay. This amount of water is referred to as "Delta outflow." Because of the
limited capacity of the Delta channels, during periods of high-export pumping some of the water that flows into
the Delta from the Sacramento River flows back up the San Joaquin River. In order to prevent too much salt
mixing with this flow (around Sherman Island at the western edge of the Delta), a sufficient amount of water
has to be released from upstream reservoirs to ensure enough Delta outflow to repel salinity intrusion from
Suisun Bay. This amount of Delta outflow is referred to as "carriage water" and is illustrated in Figure 3.
In addition to carriage water, Delta outflows are managed to maintain salinity standards for irrigation in the
Western Delta, for fish migration, and to provide low-salinity water for waterfowl habitat in Suisun Marsh.
These latter requirements are stated in Decision 1485 of the State Water Resources Control Board (SWRCB,
1978). However, much of the time carriage water requirements govern Delta outflow.
Much of the freshwater discharging to the estuary now occurs as winter runoff, and so salinity levels in
different parts of the estuary can vary considerably throughout the year. The spring and summer freshwater
flows and salinity are particularly important in maintaining the estuary productivity (Williams and Josselyn, 1987).
Year-to-year flow variation is also high, and the cumulative effect of successive years of drought and
consequently low runoff can greatly affect the estuary.
2NGVD means National Geodetic Vertical Datum, formerly referred to as "1929 mean sea level datum."
3-6
-------
Williams
FLOW DISTRIBUTION WITH EXPORTS ILLUSTRATING CARRIAGE WATER
INFLOW
x OUTFLOW REQUIRED TO
/ MAINTAIN WATER QUALITY
/ AND FLOW STANDARDS
(
I
o Lodi
^CHANNEL DEPLETIONS
\ EXPORT
'CARRIAGE WATER NEEDED
TO MAINTAIN WATER
QUALITY WITH EXPORTS
EXPORT PUMPS
£/EXPORT
SUMMER FLOW DISTRIBUTION
UNDER PRESENT CONDITIONS
Source-. DWR Exhibit 62 San Francisco Bay/ Delta Hearing Process
3-7
FIGURE 3
-------
Williams
CHAPTER 2
METHODOLOGY
OVERALL APPROACH
This study is being carried out as one component of an integrated series of studies on the effects of climate
change in California. Lettenmaier et al. (Volume A) has simulated the runoff to the Central Valley for the
precipitation and temperature regimes predicted by three different general circulation models run at double
current carbon dioxide levels. Sheer (this volume) used these runoff predictions to simulate reservoir operation
and irrigation deliveries in the Central Valley.
The main task of this study was to develop a simulation of the seasonal and annual salinities in different
parts of San Francisco Bay for different future climate change and sea-level rise scenarios with the present water
management system. A further study by Josselyn (Volume D) [Romberg Tiburon Center for Environmental
Studies] used this analysis of salinity changes to analyze their effects on the estuarine ecosystem.
Three steps were required to simulate seasonal and annual salinities:
o Define the future morphometry (shape) of San Francisco Bay with sea-level rise.
o Determine the tidal exchange characteristics for San Francisco Bay for the future morphometry.
o Determine the salinity response of future tidal characteristics to future Delta outflow scenarios.
In addition, because carriage water requirements affect future Delta outflows, new carriage water
requirements for future tidal characteristics were developed With these new carriage water requirements, the
monthly Delta outflows for three different doubled carbon dioxide climate change scenarios were modeled.
Two separate scenarios of the morphometry of the estuary were used to investigate the range of possible
responses of the system. Both of these were for doubled carbon dioxide levels and assume a 1-meter rise in
mean sea level
The two scenarios are:
o 1-meter sea-level rise, no levee breaks.
o 1-meter sea-level rise with all levees failed.
A 1-meter rise was selected, as it is in the mid-range of predictions for the effect of doubled carbon dioxide
levels on global sea level (National Research Council, 1987). All levees were assumed failed for the second
scenario, as this is the "worst-case" condition for this analysis. Actually, it is likely that almost all levees
protecting developed areas would be strengthened However, this would only protect less than 10% of the total
area
SAN FRANCISCO BAY MORPHOMETRY
Unfortunately, no accurate analysis has been carried out of the morphometry of San Francisco Bay.
Therefore, for the purposes of this study, a separate analysis of the elevation/area and elevation/volume
relationships had to be carried out for all areas below +3 meters (+10.0 ft.) NGVD. This elevation was selected
as it would represent the approximate elevation of the 100-year high tide after 1 meter of sea-level rise. USGS
3-8
-------
Williams
topographic maps and NOAA bathymetric charts were used to develop the morphometric relationships for each
segment of the estuary shown in Figure 1. These maps varied in age from the 1930s to 1970s and consequently
may slightly underestimate areas and volumes in locations where rapid subsidence is taking place.
Using existing morphometiy for future sea-level rise conditions necessarily ignores the effect of geomorphic
response to sea-level rise. This would probably occur as sedimentation in the Delta and in shallower areas
around the Bay, and as scouring of the deeper channels in response to increased tidal action. These variables
have not been investigated.
TIDAL EXCHANGE CHARACTERISTICS
The changes in estuarine morphometiy with an increase in sea level have a significant effect on tidal
circulation, which in turn is a major determinant of salinity distribution.
Fischer's one-dimensional finite element tidal hydrodynamic model (Fischer, 1970) was used to simulate
the tidal exchange characteristics at various points in the estuary. The estuary was represented by a nine-segment
model that simulated existing morphometiy and was calibrated to simulate the existing tidal range, using channel
roughness as the calibration variable. The calibrated values are shown for existing conditions in Table 1 and are
within approximately 0.1 meter of the predicted tidal range.
The model calculates the average ebb and flood tide flow rates at each of the straits separating the different
segments of the Bay shown in Figure 1. The model was then run for the two sea-level rise scenarios using the
future morphometiy of the Bay.
SALINITY RESPONSE
In order to simulate the response of salinity hi the estuary to different Delta outflows for future conditions,
the mixing model developed by Denton (Denton and Hunt, 1986) was used.
As the model was originally formulated, the estuary was simulated as four interconnected bays separated
from each other by three straits, and separated from the ocean by the straits of the Golden Gate (see Figure 4a).
The model calculates the exchange of water for each bay for each tidal cycle, taking into account the volume
of freshwater entering the estuary. At the end of each tidal cycle, it is assumed that each Bay is completely
mixed uniformly. Because much of the water entering a bay on the flood tide leaves it on the ebb, only a portion
of the water volume is exchanged in a tidal cycle from one bay to another as "new water" (Figure 4b). The "new
water" ratio (explained on Figure 4b) is used as a calibration coefficient for each strait. Calibrating this simple
model in this way is convenient and allows for reproducible results. The effect of many other complex mixing
processes is lumped together in the calibration coefficient
Denton's model was run for steady-state (constant freshwater outflow) conditions for the two sea-level rise
scenarios, as well as a scenario with levee failure at existing sea level The steady-state condition is a theoretical
calculation of salinity that would seldom actually occur because of the large seasonal change in runoff, but is
useful for systematically analyzing sea-level rise effects. For each of these runs, the volumes and tidal exchanges
of each bay obtained from Fischer's model were changed to account for sea-level rise and levee failure. New
water ratios were kept constant
The change in salinity in Suisun Bay for nine different steady-state Delta outflows was determined using
the model From this, new carriage and D1485 water requirements for the sea-level rise scenarios were
estimated. This was done by analyzing the additional flow required to maintain the same salinity as occurs in
the existing condition immediately downstream of the Delta in Suisun Bay.
3-9
-------
DIAGRAMS ILLUSTRATING DENTON'S MIXING MODEL
Williams
OCEAN
eD
4a
Sketch showing the system of four interconnected bays used to model flushing
in San Francisco Bay. The main flushing mechanisms are the exchange of
water through the four connecting straits and the net through flow
new
resulting from the Delta outflow Qf. Pollutant discharges to each of the bays
are also included.
OCEAN
BAY A
Ratio
Ratio R,
4b
Definition sketch of the tidal exchange through strait 1 during the ebb tide.
Strait 1 connects bay A (Central Bay) with the ocean. The water passing
through the strait on the flood tide has a salinity Sm. A portion of this water
(Volnew/Volflood « RO) is "new" water from the ocean that has not previously
been in the bay. Similarly, a portion (R|) of the water exiting the bay through
strait 1 during the following ebb is "new" water from bay A. The shaded area
shows the water from the flood tide that remains in bay A at the end of the ti-
dal cycle. The average salinity S'm of the ebb flow results from the mixture of
the new water from bay A and the water from the previous flood tide.
Source: Denton and Hunt. UCB/HEL-86/01.1986
3-10
FIGURE 4a-b
-------
Williams
Using the new carriage water requirement in its reservoir operation simulation, Sheer (Volume A) simulated
the monthly Delta outflow for the three climate change scenarios for the base period 1951 to 1980. The three
scenarios were developed from the following General Circulation Models (GCMs):
o Geophysical Fluid Dynamics Laboratory (GFDL)
o Goddard Institute of Space Science (GISS)
o Oregon State University (OSU)
All of these scenarios have greater annual freshwater inflow to the estuary than occurs at present. In
addition, all have a significant seasonal shift in runoff from the spring to the winter.
Denton's model was run again using the simulated monthly Delta outflows as input. The output included
average monthly and average annual salinities in different bays within the estuary under the different scenarios.
3-11
-------
Williams
CHAPTERS
RESULTS AND INTERPRETATION
MORPHOMETRY
The elevation/area and elevation/volume relationships for the whole estuary, with and without complete
levee failure, are shown in Figures 5 and 6 and summarized in Table 2. They illustrate the dramatic effect that
failure of perimeter levees would have on increasing the area and volume of the estuary. As is shown in
Table 2, most of this increase would occur from failure of Delta island levees. These levees are the most fragile
and poorly maintained, and are susceptible to failure even at the existing sea level The exact amount of future
sea-level rise - 0.5 meter, 1 meter, or 2 meters - is less important to the physical characteristics of San
Francisco Bay than whether or not the Delta levees are preserved
The morphpmetric plots (Figures 5 and 6) illustrate that without levee failure there would be a net loss of
intertidal area, including mudflats and salt marsh. This effect could be aggravated by changes in sediment
transport mechanisms within the estuary. For example, possible reduced sediment concentrations entering San
Pablo Bay could allow for increased wave erosion of mudflats and salt marsh. In addition, the deeper water
could increase wave energy, leading to further reduction of intertidal areas by wave erosion.
The curves shown in Figures 5 and 6 include the probable effect of morphological responses to rise in sea
level Such responses would include sfltation in protected areas (such as inundated Delta islands) and scouring
of deeper tidal channels hi response to higher tidal velocities.
HYDRODYNAMICS
The tidal transport or average tidal velocity increases dramatically with increasing sea level (Table 1). If
levees are maintained, the deeper water tends to make the tidal channel hydraulicaDy more efficient, and this
tends to compensate for the increased energy losses due to the larger velocities. The net result of maintaining
the levees is that tidal ranges do not appear to change dramatically from the existing condition and may even be
amplified in the South Bay.
If an levees fafl, the increase in tidal prism (the volume of water between high and low tides) in the Delta
increases the velocities and energy losses in downstream constrictions in Carquinez Strait and Chipps Island
and greatly reduce the tidal range. This reduction in tidal range largely compensates for the increased tidal
volume. However, h is important to note that tins analysis assumes rigid boundary channels. In fact, there
would be a considerable erosion in response to the increased tidal velocities in the lower part of the Delta,
causing deepening and widening of channels. These changes could substantially affect the tidal hydrodynamics
and allow for an increased tidal transport in the future.
SALINITY
The steady-state salinity analysis for the two sea-Ie^ The change
in steady-state salinity in Srisun Bay is shown in Figure 7. The changes msahnttyu^itNighc^u^northeni reach
of the estuary for a fixed Delta outflow of 110 or/sec. (4000 ds) is shown in Figure 8. For the 110 m3 sec
flow, there b a landward migration of given salinity levels of about 15 kilometers (Figure 8). For Suisun Bay,
it can be seen that approximately double the Delta outflow is required to maintain the same salinity as
present-day conditions. Consequently, the carriage water requirement to maintain tow-salinity water for export
would be approximately doubled
3-12
-------
SAN FRANCISCO BAY ESTUARY
STAGE/AREA
ONE METER RISE
EXISTING SEA LEVEL
-50.0
0
| | I I I I | I I I I I I I | I M I I I I | I I I I I I I | I I I I I I I | I I I I I I I |
200
400 600 800
AREA (X 1000 ACRES)
1000
1200
DATE: 3/18/88
_l DMTMQ CONDITIONS
Philip Williams k Associates
Pier 35, The Emborcadero
San Francisco', California 94133
-------
SAN FRANCISCO BAY ESTUARY
STAGE/VOLUME
10.0 -•
5.0 -E_
o.o -E-
Q* -5.0 -E
o -10.0 -E-
. -15.0 -E
B -20.0 -E-
Lu j
g -3ao i •
t< -35.0 -E
U -40.0-E-
^ -45.0 -E
j
S
ONE METER RISE
SEA
IIMnmlHllHlll|lilHlllllHllllll|lllllHlllHIMMIl|llUlimlMHHUl|H
0 2 4 6 8 10 12 14 16 18 20
VOLUME (X MILLION ACRE-FEET)
a
R
DATE: 3/02/88
D05THW
o --- o COMPITTE UVCE FAILURE
Philip Williams it Associates
Pier 35, The Embarcadero
San Francisco, California 94133
-------
TABLE 1
CHANGE IN TIDAL CHARACTERISTICS
WITH 1 METER SEA LEVEL RISE
Williams
Location
Diurnal Tidal Range (m)
No All
Levee Levees
Existing Break Failed
Avg. Tidal Transport1
m3/sec x 103
No All
Levee Levees
Existing Break Failed
Golden Gate
Pt. San Pablo
Benicia
Collinsville
Bay Bridge
1.73
1.68
1.66
1.40
1.84
1.73
1.62
1.57
1.21
1.87
1.73
1.46
1.25
0.30
1.59
34.0
9.3
4.0
2.3
17.6
53.8
17.3
6.8
3.4
28.3
51.0
17.3
7.1
6.2
28.3
Note:
1 Average of ebb and flood tide flow rates
3-15
-------
Williams
TABLE 2
CHANGES IN SAN FRANCISCO BAY AREA AND VOLUME
WITH SEA LEVEL RISE OF 1 METER
Area km2
Volume (m9 x 10')
Location
Central Bay
San Pablo Bay
Carquinez
Straits
Suisun Bay
Delta
South Bay
Existing*
206
210
15
85
166
437
Levees2
Intact
218
352
16
186
170
498
All2
Levees
Failed
218
558
16
384
1,619
741
Existing1
2,368
900
185
401
839
2,319
Levees2
Intact
2,529
1,209
202
506
1,011
2,775
All2
Levees
Failed
2,529
1,665
202
913
5,242
3,281
Total Bay:
1,119
1,440 3,536
7,012
8.232 13,832
Notes:
1 Area and volume at 0.0 meter NGVD.
2 Area and volume at 1.0 meter NGVD.
3-16
-------
Location
TABLE 3
SUMMARY OF STEADY-STATE SALINITIES
Salinity (ppt) for Steady-State Delta Outflows
1,500 cfs (40m3/s) 4,000 cfs (110m3/s) 20,000 cfs (560mJ/s)
Exist. 1m Sea Exist. 1m Sea Exist. 1m Sea
Sea Level Rise Sea Level Rise Sea Level Rise
Williams
Central
San Pablo
Carquinez
Suisun
Delta*
South
i^evej.
32
30
27
24
18
32
No
Levee
Fail.
32
31
29
27
22
32
With
Levee
Fail.
32
31
29
27
(24)'
32
Lrtsvei.
30
27
20
15
8
30
No
Levee
Fail.
31
29
25
21
13
32
With
Levee
Fail.
31
29
25
21
(15)a
32
i^evej.
24
14
5
2
0.2
24
No
Levee
Fail.
26
20
10
5
1.0
26
With
Levee
Fail.
26
19
10
6
(1.7)2
26
Notes
1. Delta is calibrated for Western Delta salinity (see Appendix A).
2. These Delta salinity values have little meaning because of changes in
channel morphology with levee failure.
347
-------
35.0 n
30.0 -_
25.0 -_
20.0 -_
_ 15.0 :
z
$ 10.0 T
5.0 -_
0.0
STEADY STATE SALINITY
SUISUN BAY
II II
0
I I 1 I
10000
-* EXISTING CONDITIONS
•a 1 M SEA LEVEL RISE: NO LEVEE FAILURE
-A 1 M SEA LEVEL RISE AND LEVEE FAILURE
"***fl
30000
20000
DELTA OUTFLOW (CFS)
40000
50000
DATE: 04/26/88
BY: L. FISHBAIN
Philip Williams Associates
Pier 35, The Embarcadero
San Francisco, California 94133
-------
35.0 n
30.0 -
OH
O, _
— 20.0 1
g 15.0
10.0
5.0
^
0.0
STEADY STATE SALINITY
VS DISTANCE FROM THE GOLDEN GATE
FOR DELTA OUTFLOW OF 4,000 CFS
EXISTING CONDITIONS
B- D 1 M SEA LEVEL RISE: NO LEVEE FAILURE
A * ! M SEA LEVEL RISE: AND LEVEE FAILURE
0
I I I II I I 11 I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
10 20 30 40 50 60
DISTANCE FROM GOLDEN GATE IN MILES
70
DATE: 03/20/88
BY: L. FISHBAIN
Philip Williams & Associates
Pier 35, The Embarcadero
San Francisco, California 94133
8
-------
Williams
The modeling analysis carried out indicates that there would be little difference in the salinity regimes with
sea-level rise, with or without levee failure. However, the steady-state levee failure scenario may underestimate
salinity, as it does not incorporate the probable scouring of tidal channels that could allow a greater migration
of salinity inland.
The steady-state analysis indicates the salinity changes due solely to sea-level rise effects. The monthly
simulations of the climate change scenarios incorporate the combined effects of altered hydrology and sea-level
rise as well as the seasonal changes in runoff obtained from Sheer (this volume).
The monthly simulation scenarios that were analyzed and their results are summarized in Table 4. Figure
9 shows the average annual salinity in the northern reach of the estuary for the three different hydrology
scenarios. The comparatively low freshwater runoff of the OSU scenario causes the greatest increase in salinity
over existing conditions. Comparison of Figure 9 with the steady-state results shown in Figure 8 indicate that
the higher runoff for all three climate-change scenarios partially compensates for the sea-level rise effects. The
combined hydrologic and sea-level rise effect of climate change results in a net inland migration of the average
annual salinity of up to about 10 kilometers (6 miles). This shift in salinity is approximately the same magnitude
as that which has already occurred due to upstream water development in the last century (Williams and
Fishbain, 1987).
A closer examination of the effect on salinity of levee failure illustrates the buffering effect of the greatly
increased volume of the estuary due to sea-level rise effects. The average monthly salinity m Suisun Bay with
and without levee failure for the GISS hydrology scenario is shown in Figure 10. The seasonal fluctuation in
average monthly salinity is 3 to 4 ppt less for the levee failure scenario (run 6) than without levee failure (run
5), even though the average annual salinity is approximately the same.
The average monthly salinity variation in Suisun Bay for the three hydrology scenarios is shown in Figure
11. In all cases, the average spring and summer salinities are higher, due to the combined effects of sea-level
rise and reduced runoff. Only in the winter is the greatly increased runoff of the GISS scenario sufficient to
compensate for the effect of sea-level rise.
Figures 12 and 13 illustrate the changes in frequency of a particular value in March and June in Suisun
Bay. While the salinity distribution resulting from the GISS scenario shows little change from existing conditions
in March, the lower runoff of the OSU scenario indicates a significant shift, with salinities that were formerly
exceeded in 20% of years now occurring 50% of the time. In June, there is a significant change in frequency
for both scenarios, with salinities that were formerly exceeded in 50% of years now occurring in 80% of years.
LIMITATIONS AND UNCERTAINTY
This study was a reconnaissance-level analysis carried out in a short time frame with limited resources.
There are substantial limitations and uncertainties in the analysis. These are:
- Morphometric analysis. Systematic up-to-date detailed surveys could indicate errors of up to plus or
minus 20% in the bays' volume.
- Sea-level rise analysis. There is a range of uncertainty in predictions of sea-level rise after doubling
of COj, varying from 0.5 to about 2 meters.
- Geomorphic response. Siltation and erosion of estuarine sediments caused by changing tidal
characteristics would alter the morphometry of the estuary and its tidal hydraulics.
- Extent of levee failure. It is likely that with sea-level rise, some levees, including those protecting urban
areas, will be upgraded and the extent of inundation not as great as indicated
3-20
-------
Williams
TABLE 4
SUMMARY OF SALINITY RESULTS
Run
No.
1
2
3
4
5
6
7
8
Sea Levee
Level Fail-
Rise ure
(m) Condi-
tion
0 No
0 Yes
1.0 Yes
1.0 Yes
1.0 No
1.0 Yes
1.0 Yes
1.0 Yes
Carr-
iage
Water
Exist
Exist
Exist
Exist
Double
Double
Double
Double
Hydro-
logy
Scen-
ario
Exist1
Exist1
Exist'
Exist1
GISS
GISS
GFDL
OSU
Avg.
Ann
Delta
Out-
flow2
(km')
19.
19.
19.
19.
33
33
24
20
7
7
7
7
.3
.3
.7
.1
Average Salinity (ppt)
Suisun Bay
Ann.
7.2
7.1
8.9
8.7
8.1
7.9
8.7
9.5
Mar.
2.2
2.9
4.5
4.6
1.4
2.1
3.0
4.5
Sep.
12.5
11.2
12.7
11.9
13.5
12.0
12.4
12.5
San
Ann.
19.3
20.1
21.4
21.3
19.1
19.3
20.6
21.7
Pablo Bay
Mar.
12.2
13.8
16.1
16.2
8.3
9.0
12.3
15.0
Sep.
25.1
25.1
25.4
25.0
25.8
25.1
25.3
25.4
Note:
Existing hydrology is DWR 1990 scenario
Base period 1951-1980
3-21
-------
H
O,
O,
30.0 -i
25.0 -
20.0 -
15.0 -
3 10.0 H
5.0 -
0.0 -
AVERAGE ANNUAL SALINITY
VS DISTANCE FROM THE GOLDEN GATE
COMPARISON OF DIFFERENT HYDROLOGY SCENARIOS
* EXISTING CONDITIONS (BASE CASE) (l)
— A GIS3 HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (a)
o a GFDI. HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (7)
Or £ QSU HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (8)
I I 1 1 I I I I I | I 1 I I I I I I I | I I I I I I I I 1 | I I I III pi I | I I I I I I I I I | I I I I I I I I I | I I I I I I I I I |
0 10 20 30 40 50 60 70
DISTANCE FROM GOLDEN GATE IN MILES
DATE; 04/18/88
BY: L. FISHBAIN
Philip Williams Associates
Pier 35, The Emharnadero
San Francisco, California 04133
-------
AVERAGE MONTHLY SALINITY IN SUISUN BAY
0,
30.0. n
25.0 -
20.0 -
15.0 -
10.0 -
5.0 -
0.0
EFFECT OF LEVEE FAILURE
* * EXISTING CONDITIONS (BASE CASE) (1)
A A GISS HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (fi)
o- -a GISS HYDROLOGY WITH SEA LEVEL RISE, NO LEVEE FAILURE (5)
OCT ' NOV ' DEC ' JAN ' PEB ' MAR" APR ' KAY ' JUH ' JUL
MONTH
AUG ' SEP
DATE: 04/18/88
BY: L. FISIIBAIN
Philip Williams Associates
Pier 35, The Emharcadero
San Francisco, California 84133
10
-------
AVERAGE MONTHLY SALINITY IN SUISUN BAY
COMPARISON OF DIFFERENT HYDROLOGY SCENARIOS
&4
6
30.0 -i
25.0 H
20.0 H
15.0 H
10.0 ^
5.0 H
0.0
EXISTING CONDITIONS (BASE CASE) (1)
— — * GISS HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (6)
a GPDL HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (7)
« OSU HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (B)
OCT ' NOV ' DEC ' JAN ' FEB ' MAR ' APR ' MAY ' JUN ' JUL ' AUG ' SEP '
MONTH
DATE: 04/12/88
BY: L. FISHBAIN
Philip Williams Associates
Pier 35, The Embarcadero
San Francisco, California 04133
11
-------
SUISUN BAY SALINITY - MARCH
FREQUENCY RELATIONSHIP FOR THREE OUTFLOW SCENARIOS
0,
30.0 n
25.0 -
20.0 -
15.0 -
10.0 -
S.'O -
0.0 -
* EXISTING CONDITIONS (BASE CASE) (1)
A GISS HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (6)
a QSU HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (8)
- Q -Q
-A «-—-
11II1111 111III M I I 111 M I 11 I N 11111 111 H | 111 | M I ITTTTrpTr
0
••MM
DATE: 04/18/88
BY: L. FISHBAIN
10
20 30 40 50 60 70 8
EXCEEDENCE FREQUENCY (PERCENT)
90 100
Philip Williams Associates
Pier 35, The Embarcadero
San Francisco, California 94133
12
-------
SUISUN BAY SALINITY - JUNE
FREQUENCY RELATIONSHIP FOR THREE OUTFLOW SCENARIOS
30.0 n
25.0 -
0,
20.0 -
15.0 -
10.0 -
5.0 -
0.0
* EXISTING CONDITIONS (BASE CASE) (1)
A GISS HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (6)
a OSU HYDROLOGY WITH SEA LEVEL RISE AND LEVEE FAILURE (8)
1 1 H 1 1 1 1 1 [1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 M 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 fl I lTlTnTT]
10
20 30 40 50 60 70 80
EXCEEDENCE FREQUENCY (PERCENT)
90 100
9
DATE: 04/25/88
BY: L. FISHBAIN
Philip Williams Associates
Pier 35, The Emharoadero
San Francisco, California 94133
13
-------
Williams
Modeling of tidal hydrodynamics. A simple model was used to achieve reasonable results in the budget
and time frame available. More sophisticated physical or numerical modeling of the tidal hydrodynamics
would give a better representation of tidal ranges and exchanges, particularly in the complex geometry
of the Delta. The salinity distribution is particularly sensitive to the tidal exchange determined by the
hydrodynamic model.
Modeling of salinity. A simple model was used to achieve reasonable results in the budget and time
frame available. A more sophisticated mixing model would give greater accuracy and more closely
reflect the specific physical characteristics of each bay.
Hvdrologic base period. The hydrologic base period selected (1951-1980) is wetter than average, 20
cubic kilometers (16.0 million acre-feet) Delta outflow as compared to 16 cubic kilometers (or 13.0
million acre-feet) for the 1922-1978 average. This may tend to underestimate the long-term increases
in salinity by not accounting for extended dry periods.
Delta salinity. The salinity in the Delta at the intake to the export pumps is very sensitive to the physical
characteristics of the Delta channels and islands. Although average monthly salinities in the Delta were
calculated in this analysis, they are not presented in this report, because the physical significance of
these results from such a simple model is unclear, considering the complexity of the geometry hydraulics
of the Delta.
Carriage water requirements. As for existing water management planning, these have been based on
steady-state analysis. Complex changes in the hydraulics of the Delta and Suisun Bay such as scouring
of tidal channels could increase salinity repulsion requirements.
3-27
-------
Williams
CHAPTER 4
POLICY IMPLICATIONS
THE FUTURE OF THE DELTA
The question of the influence of the failure of the Delta island levees on salinity intrusion is already crucial
in determining the justification for public expenditures to improve the existing levees. An estimate to improve
these levees to protect against flooding from existing sea level is approximately $4 billion (DWR, 1982). The
total valuation of property in the Delta is about $2 billion (DWR, 1987).
With accelerated sea-level rise causing major increases in salinity, the policy question becomes the following:
Does the incremental value of maintaining levees in the Delta to prevent further increases of salinity in addition
to protecting farmland and infrastructure justify the cost?
THE ESTUARINE ECOSYSTEM
Only in the last decade have the freshwater inflow requirements for maintaining the entire estuarine
ecosystem been quantified. These requirements have not yet been incorporated into the water allocation and
reservoir operating criteria of California's water system. With the major changes in estuarine hydrology,
morphology, and hydrodynamics, it will become even more important to modify water resources management
to incorporate protection of the estuarine ecosystem.
At the same time, climate change may allow an opportunity to restore lost environmental resources. For
example, this could be done by increasing areas of tidal wetland. In addition, increased reservoir releases for
salinity repulsion may also have substantial ecological benefits for San Francisco Bay.
EXPORT OF WATER FROM THE DELTA
The vulnerability of the State Water Project and Central Valley Project to salinity intrusion into the Delta
was the genesis of the controversial and now-abandoned Peripheral Canal plan. The high cost of large diversion
schemes that bypass the Delta probably means that salinity will continue to be managed by releases from
reservoirs. Sea level rise and climate change will require that these releases be increased. The larger releases
will require reallocation of water contracts and revision of present water management practices. These revisions
may lead to consideration of structural modifications to the physical system such as channel modifications or
off-stream storage.
REDUCING THE RANGE OF UNCERTAINTY
Because of the large economic, social, and environmental costs of these policy implications, the value of a
greater degree of certainty is high. The next level of analysis for predicting the salinity response of the estuary
to sea-level rise will require the following:
- Detailed morpbometric description of the estuary.
- Utilization of detailed hydrodynamic and mixing models of the estuary. At present, it appears that
the Corps of Engineers' Physical Model of San Francisco Bay (three-dimensional) and a version of
Fischer's numerical model (pseudo-two-dimensional) are available.
- Development of a model for morphologic changes in tidal channels with changing tidal prism.
- Analysis of sediment budget and sediment dynamics in the estuary.
3-28
-------
Williams
REFERENCES
Denton, Richard A. and James R. Hunt. "Currents in San Francisco Bay. Final Report." University of
California, Berkeley, California, 1986.
Department of Water Resources. Delta Levees Investigation. Bulletin 192-82. State of California. 1982. 213
pp.
Department of Water Resources. Sacramento-San Joaquin Delta Atlas. Sacramento, California, 1987. 71 pp.
Fischer, Hugo B. "A Method for Predicting Pollutant Transport in Tidal Waters." University of California,
Berkeley, Water Resources Center, Contribution No. 132. 1970. 143 pp.
Josselyn, N., and Atwater, B.F. 1982. "Physical and biological constraints as man's use of the shore zone of the
San Francisco Bay estuary." In: Kockelman, WJ., TJ. Conomos, and A.E. Leviton, eds. San Francisco Bay.
TJse and Protection. San Francisco, CA: American Association for the Advancement of Science, p. 57-84.
Kockelman, William J., T. John Conomos, and Alan E. Leviton. San Francisco Bay: Use and Protection.
American Association for the Advancement of Science. San Francisco, California, 1981 310 pp.
National Research Council. Responding to Changes in Sea Level: Engineering Implications. National Academy
Press, Washington, D.C., 1987. 148 pp.
State Water Resources Control Board. "Water Quality Control Plan: Sacramento-San Joaquin Delta and Suisun
Marsh." Sacramento, California, 1978.
Williams, Philip B. and Larry Fishbain. "Analysis of Changes in Suisun Bay Salinity Due to Existing and Future
Water Development." Report 412-2. San Francisco, California, 1987.
Williams, Philip B. and Michael Josselyn. "An Overview of Flow and Salinity Standards Required to Protect
the Ecosystem of San Francisco Bay." San Francisco, California, i987. S3 pp.
3-29
-------
EFFECTS OF CLIMATE CHANGES ON THE LAURENTIAN GREAT LAKES LEVELS
Thomas E. Croley n
Holly C. Hartmann
National Oceanic and Atmospheric Administration
Great Lakes Environmental Research Laboratory
2205 Commonwealth Boulevard
Ann Arbor, MI 48105-1593
Interagency Agreement Identification Number DW13932631-01-0
-------
CONTENTS
FINDINGS 4-1
CHAPTER I: INTRODUCTION 4-3
THE PHYSICAL SYSTEM 4-3
RELATED WORK 4-4
STUDY APPROACH 4-4
CHAPTER H: COMPONENT PROCESS MODELS 4-6
RUNOFF MODELING 4-6
OVERLAKE PRECIPITATION 4-7
OVERLAKE EVAPORATION 4-7
CONNECTING CHANNEL FLOWS 4-9
MODELS APPLICABILITY 4-9
CHAPTER ffl: SIMULATION METHODOLOGY 4-11
STEADY-STATE CLIMATE CHANGE ASSESSMENTS 4-11
TRANSIENT-CASE CLIMATE CHANGE ASSESSMENTS 4-12
CHAPTER IV: RESULTS 4-13
CLIMATE CHANGE STEADY-STATE IMPACTS 4-13
Basin Meteorology 4-13
Basin Hydrology 4-13
Over-Water Meteorology 4-15
Lake Heat Balance 4-16
Net Supply Components 4-19
Lake Water Balance 4-21
CLIMATE CHANGE TRANSIENT IMPACTS 4-23
SENSITIVITIES 4-25
CHAPTER V: IMPLICATIONS OF RESULTS 4-26
ENVIRONMENTAL IMPLICATIONS 4-26
SOCIOECONOMIC IMPLICATIONS 4-26
Power Production 4-26
Navigation 4-28
Industrial Operations 4-28
Commercial Operations 4-28
Commercial Fishing 4-28
Agriculture 4-29
Recreation Interests 4-29
Riparians 4-29
POLICY IMPLICATIONS 4-29
REFERENCES 4-31
n
-------
Croley
FINDINGS1
The Great Lakes Environmental Research Laboratory has developed conceptual model-based techniques
for simulating moisture storages and runoff from the 121 watersheds draining into the Laurentian Great Lakes,
overtake precipitation into each lake, and the heat storages and evaporation from each lake. We model each of
these components separately and then use them in conjunction with operational regulation plans and hydraulic
routing models of outlet and connecting channel flows to estimate water levels on Lakes Superior, Michigan,
Huron, St. Clair, Erie, and Ontario to consider the existing basin and lake storages and possible meteorology.
Integration of the models allows the consideration of climate change scenarios developed from atmospheric
circulation models through linkages on air temperature, precipitation, humidity, wind speed, and cloud cover.
Alternate scenarios are considered by abstracting changes in linkages, making these changes in historical data,
observing the impact of the changed data in the model out puts, and comparing these to model results using
unchanged data.
Three steady-state climate change scenarios corresponding to atmospheric modeling of a doubling of
CO2 in the atmosphere are considered here with the hydrologic models; they are compared to a steady-state
simulation obtained with historical data representing an unchanged atmosphere. One transient climate change
scenario representing atmospheric modeling of the transition from present conditions to increased CO, content
in the atmosphere is considered here with the hydrologic models and compared to a transient simulation with
historical data to assess impacts. The study results should be received with caution as they are of course
dependent on the atmospheric circulation model output with large uncertainties and are only possibilities for a
future with increased atmospheric CO2 content.
The higher air temperatures under the 2xCO2 scenarios lead to higher overland evapotranspiration and
lower runoff to the lakes with earlier runoff peaks, since snowpack is reduced up to 100% and the snow season
is shortened from two to four weeks. This also results in more than a 50% reduction in available soil moisture.
Water surface temperatures peak earlier on Lake Superior; since the climate becomes similar to present-day
climates on the southern lakes, the lake temperature behaves similar to present-day southern deep lakes. There
are larger amounts of heat resident in the deep lakes throughout the year. As a result, buoyancy-driven turn-
overs of the water column do not occur many times on four of the six lakes. Currently, they occur twice a year
on all lakes. Ice formation also will be greatly reduced over winter on the deep Great Lakes and lake
evaporation increases on all lakes. The average steady-state lake levels are seen to drop between 1/2 to 11/3
meters for the GISS 2xCO2 climate scenario, drop 2 to 2 1/2 meters for the GFDL, and drop 1/2 to 1 meters
for the OSU scenario. Transient analyses of the GISS 80-yr transient scenario indicate that the lakes drop
between 13 and 93 mm per decade on the average. The Lake Ontario regulation plan fails in all steady-state
and transient climate change analyses, reflecting its design for regulation of current ranges of Lake Ontario water
levels and St. Lawrence Seaway flows.
The climate change effects modeled herein, if they occur, will require new paradigms in water
management in the Great Lakes Basin. Allocation conflicts between users of the Great Lakes will likely result.
Lowered lake levels could produce large reductions in wetland areas and lower hydropower production. While
reduced lake ice formation could lengthen the shipping season, lower lake levels could also increase waterborne
shipping costs via lower vessel load limits, traffic backups at the Welland Canal and Sault Ste. Marie, and
dredging of sediments highly contaminated with toxics. Dredging and disposal of toxic-contaminated harbor
sediments may pose critical problems for municipal and private marinas and create conflicts between the many
governments having jurisdiction over the lakes. To manage potential allocation conflicts, the Boundary Waters
Treaty of 1909 may have to be modified to consider commercial, industrial, riparian, recreational, and ecological
'Although the information in this report has been funded partly by the U.S. Environmental Protection
Agency under Interagency Agreement no. DW13932631-01-0 with the Office of Policy Analysis, it does not
necessarily reflect the Agency's views, and no official endorsement should be inferred from it.
4-1
-------
Croley
interests in addition to presently considered domestic and sanitary water supply, navigation, hydropower, and
irrigation interests. The Lake Superior and Ontario regulation plans would probably have to be revised to handle
persistently low water supplies.
Climate change also poses concerns unrelated to lake levels. Lake ecologies will change if the lakes do
not experience their typical spring and fall water column turnovers. Changes in fish community and population
structures will likely benefit some species but reduce ranges and sustainable yields for others. Whiter tourism
and recreation operations that require dependable snowcover may suffer total collapse throughout much of the
Great Lakes region. In addition, soil moisture shortages may prove critical for agricultural operations.
4-2
-------
Croley
CHAPTER I
INTRODUCTION
Climate change will impact on many aspects of the hydrologjc cycle with consequences for mankind that
are interrelated and often tunes difficult to discern. Impacts of climate change on water supplies have been the
focus of several recent studies. The U.S. Environmental Protection Agency (EPA, 1984) and Rind (personal
communication, 1988) used the hydrologic components of general circulation models to assess changes in water
availability in several regions throughout North America, but the regions were very large; Rind used only four
for the entire continent and indicated that assessments with smaller regions were needed Because the
Laurentian Great Lakes possess tremendous water and heat storage capacities coupled with constricted lake
outlets, they respond slowly to changed meteorologic inputs. This "memory results in a filtering or dampening
of most short-term climate fluctuations with lake levels reacting primarily to longer-period fluctuations
characteristic of climate change. The large Great Lakes system thus is ideal for studying regional effects of
climate changes.
THE PHYSICAL SYSTEM
The Laurentian Great Lakes contain 23,000 km3 of water (about 20% of the world's fresh surface water)
and, with their surrounding basins, cover 770,000 km in the United States and Canada The lakes' surface areas
comprise about one-third of the total basin area. The basin extends over 3200 km from the western edge of
Lake Superior to the Moses-Saunders Power Dam on the St Lawrence River. The water surface drops in a
cascade over this distance over 182 meters to sea level. The most upstream, largest, and deepest lake is Lake
Superior. Lake Superior outflows are controlled according to regulation Plan 1977, under the auspices of the
International Joint Commission. The lake has two interbasin diversions of water into the system from the
Hudson Bay basin, the Ogoki and Long Lake diversions. Lake Superior waters flow through the lock and
compensating works at Sault Ste. Marie, Michigan, and down the St. Mary's River into Lake Huron where it is
joined by waier flowing from Lake Michigan.
Lakes Michigan and Huron are considered to be one lake hydraulically, because of their connection
through the deep Straits of Mackinac. The second interbasin diversion takes place from Lake Michigan at
Chicago. Here water is diverted from the Great Lakes to the Mississippi River basin. The water from Lake
Huron flows through the St. Clair River, Lake St. Clair, and Detroit River system into Lake Erie. The drop in
water surface between Lakes Michigan-Huron and Erie is only about 2.4 meters. This results in a large
backwater effect between Lakes Erie, St Clair, and Michigan-Huron; changes in Lakes St. Clair and Erie levels
are transmitted upstream to Lakes Michigan and Huron. From Lake Erie the flow continues through the
Niagara River and Welland Canal diversion into Lake Ontario. The major drop over Niagara Falls precludes
changes on Lake Ontario from being transmitted to the upstream lakes. The Welland Canal diversion is an
intrabasin diversion bypassing Niagara Falls and is used for navigation and hydropower production. There is also
a small diversion into the New York State Barge Canal system, which is ultimately discharged into Lake Ontario.
Lake Ontario outflows are controlled in accordance with Plan 1958-D. From Lake Ontario, the water flows
through the St Lawrence River to the Gulf of St. Lawrence and the Atlantic Ocean.
There are three primary types of fluctuations of Great Lakes levels: annual, seasonal, and short-term
variations due to wind setup and storm surge. Annual fluctuations result in most of the variability leading to
record high or low lake levels. There is an overall range of about 1.8 meters in the annual levels. Superimposed
on the annual levels are seasonal cycles, which range in magnitude from about 038 meters on Lake Ontario to
about 030 meters on Lake Michigan-Huron. In general, the seasonal cycles have a minimum hi the winter,
usually January or February. The levels then rise due to increasing water supplies from snowmelt and spring
precipitation until they reach a maximum in June for the smaller lakes (e.g., Erie and Ontario) or in September
4-3
-------
Croley
in the case of Lake Superior. The lakes begin their seasonal declines in the late summer and fall. The final type
of fluctuation, storm surge or wind setup, is relatively short-lived, lasting only several hours. While sometimes
large (Lake Erie can experience differences between levels on the eastern and western ends of the lake as large
as 4.9 meters), they are too transitory to be considered by the model applications herein and are not discussed
further.
The hydrologic cycle of the Great Lakes basin determines the lake levels. Runoff constitutes a
significant part of the Great Lakes' water supplies, particularly during the snowmelt season, late March through
early June. Because the lakes are so large, lake precipitation and evaporation are of the same order of mag-
nitude as runoff. On a monthly scale, precipitation is fairly uniformly distributed throughout the year. Lake
evaporation typically has the greatest effect on water supplies during the winter months (Derecki, 1980); cool dry
air and warm water result in massive evaporation then. Condensation on the cool lake surface from the wet
overlying air occurs in the summer. Net groundwater flows to each of the Great Lakes are generally negligible
(DeCooke and Witherspoon, 1981). Net supplies typically reach a maximum in the late spring and a minimum
in late fall. The imbalance between supplies and outflows from the lake results in rising or falling lake levels.
Although the net basin supply components are of similar magnitude, their estimations are not equally sophistica-
ted.
RELATED WORK
Preliminary estimates of the impact of climate warming on the Great Lakes water resources have been
undertaken. Quinn and Croley (1983) estimated that a 3°C rise coupled with a 65% increase in precipitation
would decrease the net basin supply to Lake Superior by 10%. A similar 3°C rise with no precipitation increase
would decrease the Lake Erie net basin supply by 33%. Cohen (1986) indicated 4 to 21% decreases in net basin
supply for various double-CO2 scenarios with annual temperature changes varying between 3.1 and 4.8°C and
varying projections of wind speed and precipitation changes. His computations are very sensitive to changes in
precipitation, wind speed, and humidity. To examine the potential response of the unregulated Great Lakes to
decreased water supplies, Quinn (1988) used two hydrometeorologic scenarios with GLERL's Hydrologic
Response Model A 15% decrease in net basin supplies reduced water levels on Lakes Michigan-Huron, St.
Clair, and Erie by 77,60, and 51 cm, respectively. A 30% decrease in supplies reduced the lakes levels by 156,
125, and 106 cm, respectively.
Environment Canada began, in 1984, a major study to assess the impacts of climate change on the
province of Ontario (Allsopp and Cohen, 1986). Goddard Institute of Space Sciences double carbon dioxide
climate scenarios were modified and used with models or estimators of net basin supply components. However,
the water supply models used only monthly data, they assumed water temperatures rose equal to air tempera-
tures, and, for the most part, they used a fragmented approach in that consistency between modeled processes
was not preserved. While their air temperature and precipitation changes were consistent with atmospheric
scenarios, dewpoint was increased arbitrarily the same as air temperature and sunshine was decreased
independently; the original 4xCO2 scenario was used to estimate 2xCO2 changes by dividing monthly anomalies
in temperature and wind speed by two; and alternate scenarios were used where wind speeds were unchanged
or reduced an arbitrary 20%. Overall water supplies to the Great Lakes system under these conditions were
about 15% lower than the base case; water levels were routinely expected to drop below the drought-period
levels of the 1960s.
STUDY APPROACH
The Great Lakes Environmental Research Laboratory (GLERL) has developed conceptual model-based
techniques for simulating moisture storages and runoff from the 121 subbasins draining into the Great Lakes,
overtake precipitation into each of the Great Lakes and Lake St Clair (hereafter included as a Great Lake), and
the heat storages and evaporation from each of the lakes. We model each of these components separately and
then use them in conjunction with operational regulation plans and hydraulic routing models of outlet and
-------
Croley
connecting channel flows to estimate water levels on Lakes Superior, Michigan, Huron, St. Clair, Erie, and
Ontario to consider the existing basin and lake storages and possible meteorology. Integration of the models
allows the consideration of climate change scenarios developed from atmospheric circulation models through
linkages on air temperature, precipitation, humidity, windspeed, and cloud cover. Alternate scenarios are
considered by abstracting changes in linkages, making these changes in historic data, observing the impact of the
changed data in the model outputs, and comparing it to model results using unchanged data. Scenarios, supplied
by the UJS. Environmental Protection Agency (EPA) and obtained from three different general circulation
models, are used in this manner to understand how climate warming affects lake levels through interrelated chan-
ges in the atmospheric-hydrologic linkages.
This report describes the models and methodology used and their limitations, presents and interprets
the results, and addresses the potential environmental, scoioeconomic, and policy implications of the results.
4-5
-------
Croley
CHAPTER H
COMPONENT PROCESS MODELS
RUNOFF MODELING
The GLERL Large Basin Runoff Model (LBRM) consists of moisture storages arranged as a serial and
parallel cascade of "tanks" (Croley, 1983a,b); water flows from the snowpack to the upper soil zone tank, from
the upper to the lower soil zone and surface storage tanks, from the lower to the groundwater and surface tanks,
from the groundwater to the surface tank, and from the surface tank out of the watershed It makes use of
physical concepts for snowmelt and net supply to the watershed surface, infiltration, heat available for evapo-
transpiration, actual evapotranspiration, and mass conservation. As a conceptual model, the LBRM is useful not
only for predicting basin runoff, but for facilitating our understanding of watershed response to natural forces
as well. The main mathematical feature of the LBRM is that it may be described by strictly continuous equa-
tions; none of the complexities associated with inter-tank flow rate dependence on partial filling are introduced.
For a sufficiently large watershed, these nuances are not observed owing to the spatial integration of rainfall,
snowmelt, and evapotranspiration processes.
Daily precipitation, temperature, and insolation (the latter available from climatological summaries as
a function of location) may be used to determine snowpack accumulations and net surface supply based on
degree-day determinations of snowmelt. The net surface supply is divided into infiltration to the upper soil zone
and surface runoff by taking infiltration proportional to the net supply rate and to the areal extent of the unsat-
urated portion of the upper soil zone. Outflow from each storage within the watershed is proportional to the
moisture in storage. The evapotranspiration rate from the upper and lower soil zones is proportional to avail-
able moistures there and to the heat rate available for evapotranspiration; it also reduces the heat available for
subsequent evapotranspiration. The total amount of heat in a day is split between that used for and that still
available for evapotranspiration by empirical functions of air temperature based on a long-term heat balance.
Mass continuity yields a first-order linear differential equation for each of the tanks (Croley, 1982) that are
tractable analytically, and the model is applied to daily data
The Great Lakes basin is divided into 121 watersheds draining directly to a lake. The meteorologic data
from about 1800 stations about and in the watersheds are combined through Thiessen weighting to produce
areally-averaged daily time series of precipitation and maximum and minimum air temperatures for each
watershed (Croley and Hartmann, 1985b). Records for all "most-downstream" flow stations are combined by
aggregating and extrapolating for ungauged areas to estimate the daily runoff to the lake from each watershed.
The LBRM is calibrated to determine the set of parameters resulting in the smallest sum-of-squared-errors
between model and actual daily flow volumes for the calibration period (Croley, 1983b; Croley and Hartmann,
1984, 1985a). Calibrations are repeated with initial storages equal to observed long-term averages until there
is no change in the averages to avoid arbitrary initial conditions in case their calibration effects do not diminish
rapidly. However, the simulation effect of the initial values greatly diminishes with the length of the period and
after 1 year of simulation, the effects are nil from a practical point of view. After the LBRM is calibrated for
each watershed (subbasin), the model outflows are combined to represent each Great Lake basin; this
distributed-parameter model integration filters individual subbasin model errors. The LBRM calibration periods
generally cover 1965-1982 depending upon flow data availability. Table 1 presents calibration results for the
distributed-parameter applications. While the calibrated parameters are used for all scenarios, the statistics in
Table 1 apply only to the historical calibration and verification periods.
4-6
-------
Croley
Table 1. Large Basin Runoff Model Calibration Statistics1.
Lake
Superior
Michigan
Huron
St. Clan-
Erie
Ontario
No.
of
Sub-
basins
22
29
27
7
21
15
Mean
1-day
Flow
(mm)2
1.12
0.89
1.06
0.90
1.01
1.41
Flow
Std.
Dev.
(mm)2
0.67
0.47
0.69
136
1.28
1.13
Root
Mean
Square
Error
(mm)2
0.25
0.18
0.26
0.62
0.54
0.43
Correlation
Calib.
0.93
0.93
0.92
0.89
0.91
0.93
Verif.
0.77
0.86
0.69
0.87
0.90
0.89
1 Calibrations and statistics generally cover 1966-83;
verification correlation generally covers 1956-63.
2Equivalent depth over the land portion of the basin.
OVERLAKE PRECIPITATION
The lack of overtake precipitation measurements means that estimates typically depend on land-based
measurements and there may be differences between land and lake meteorology. Although gage exposures may
significantly influence the results of lake-land precipitation studies (Bolsenga, 1977,1979), Wilson (1977) found
that Lake Ontario precipitation estimates based on only near-shore stations averaged 5.6% more during the
warm season and 2.1% less during the cold season than estimates based on stations situated in the lake. Using
a network that also included stations somewhat removed from the Lake Ontario shoreline, Bolsenga and Hagman
(1975) found that eliminating several gages not immediately in the vicinity of the shoreline increased overtake
precipitation estimates during the warm season and decreased them during the cold season. Thus, for the Great
Lakes, where lake ef fects on near-shore meteorology are significant and the drainage basins have relatively low
relief, the use of all available meteorologic stations throughout the basin is probably less biased than the use of
only near-shore stations.
OVERLAKE EVAPORATION
Great Lakes evaporation studies typically used mass transfer formulations from the classic Lake Hefner
study (UJS. Geological Survey, 1954,1958); see Richards and Irbe (1969) and Derecki (1976). More recently,
Phillips (1978) and Quinn (1979) included atmospheric stability effects on Great Lakes evaporation bulk transfer
coefficients; the latter approach is used presently by both Canadian and U.S. agencies applied to monthly data
for water surface temperatures, windspeed, humidity, and air temperatures (Derecki, 1976,1979,1981a,b; Quinn
and Kelley, 1983; Atmospheric Environment Service, 1988). The present study uses that approach, applied to
daily data combined with models for over-water meteorology, ice cover, and lake heat storage and with a lumped
representation of a lake's heat balance (Croley, 1988). As over-water data are not available generally, over-land
data are used by adjusting for over-water conditions. Phillips' and Irbe's (1978) regressions for over-water
corrections are used directly by replacing the fetch (and derived quantities) with averages. Air temperatures
and specific humidities over ice are used for over-ice evaporation calculations and over water for the over-water
calculations; the two estimates are combined by weighting for the fraction of the surface covered in ice. Existing
data on ice cover (Assel, 1983) were used to determine empirical relations between ice cover extent and air
temperatures, similar to other efforts (Derecki, 1978,1981a).
4-7
-------
Croley
Spring turnover occurs around June for Superior and around May for Ontario (when water temperatures
reach 3.98°C, at which water is most dense). As surface temperature begins increasing above 3.98°C, a stable
profile develops, with water surface temperature increasing faster than temperatures at depth, until a well-defined
epilemnion layer is present at the end of the summer. As the net heat flux to the surface then changes to
negative, water surface temperature drops and the temperatures at depth first grow and then recede, keeping
the upper part of the profile vertical. The mixed-layer (where the temperature profile is vertical) deepens until
the profile again approaches a vertical line throughout at 3.98°C (representing fall turnover). Then a symmetical
behaviour is observed with temperatures less than 3.98°C as the lake continues to lose heat; the surface
temperature changes fastest until the net heat flux at the surface changes to positive again. Water surface tem-
perature then increases toward 3.98°C and the temperatures at depth first decrease and then increase as the
profile again approaches a vertical line (representing spring turn over). These observations enable the exten-
sion of the mixed-layer thermal structure developed for oceans (Gill and Turner, 1976; Kraus and Turner, 1967)
to the Great Lakes to allow the determination of a simple one-dimensional model for surface temperature incre-
ments or decrements from past heat additions or losses, respectively (Croley, 1988). The effects of past
additions or losses are superimposed to determine the surface temperature on any day as a function of heat in
storage; each past addition or loss is parameterized by its age as a proxy for the wind history or accumulated
mixing. Turnovers can occur as a fundamental behaviour of this superposition model, and hysteresis between
heat in storage and surface temperature, observed during the heating and cooling cycles on the lakes, is
preserved.
Heat in storage in the lake at the end of each day is evaluated from a daily conservation of energy by
taking the change in storage equal to the sum of the fluxes integrated over the day. Short-wave radiation is
interpolated from generalized maps of Canadian and northern U.S. mid-monthly dear-sky values and adjusted
for daily cloudcover and average short-wave reflection is taken simply as one-tenth of the incident (Gray et al.,
1973). Net long-wave radiation exchange is estimated for each day with the water and atmosphere behaving as
gray bodies with cloud-cover correction only to atmospheric radiation (Keijman, 1974). Sensible heat transfers
are taken as the latent heat of evaporation times Bowen's ratio, evaluated from daily air and water temperatures
and associated humidities, (Gray et aL, 1973) and added to evaporative advection and latent heat transfers.
Energy adverted with precipitation is adjusted if the precipitation is snow to account for the heat required in
snowmelt The energies adverted into and out of the lake with other water flows are determined from the water
surface temperature and the mass flow rates. The heat delivered to the ice pack each day is used in a heat
balance over the ice to adjust the ice pack mass for accumulation, aggradation, and ablation; it is given by a
simple account of energy fluxes also. Reflected short-wave radiation is taken here (Gray et al., 1973) as a
function of the ice covet. Net long-wave radiation exchange is computed each day by using ice surface
temperature equal to the air temperature or zero °C, whichever is smaller. Evaporative heat transfers from ice
include the heat of fusion as well as vaporization at the temperature of the ice surface. Sensible heat transfers
also are calculated from the daily Bowen ratio by using ice surface temperature. Energy adverted with
precipitation onto the ice surface is uncorrected for melt since that is taken as occurring with the ice pack.
As both water surface temperature and evaporation over water and ice are unknown and must be
determined each day, an iterative approach is used The water surface temperature at the beginning of the day
is determined inversely from the heat storage at the beginning of the day (which is equal to the heat storage at
the end of the previous day). The water surface temperature at the end of the day is initially set equal to that
at the beginning. Then 1) the beginning and end water surface temperatures are averaged as the water surface
temperature during the day; 2) it is used with the day's meteorology to compute evaporation over water and ice;
3) stored heat at the end of the day is found from the heat balances over water and ice with fluxes determined
as above; and 4) an improved water surface temperature at the end of the day is computed again. These four
steps are repeated until the water surface temperature at the end of the day converges to within 0.001° C. If the
water surface temperature passes through 3.98°C, turnover is considered to have occurred. Any time that either
turnover occurs or the mixed layer begins developing, the age of the mixed layer is reset.
Unfortunately, there are no really good independent evaporation data to calibrate and verify evaporation
models on the Great Lakes. Water balances are insufficient owing to the large errors induced by subtracting
4-8
-------
Croley
nearly equal large inflows and outflows to each Great Lake or owing to errors of estimate in the water balance
components. However, with the joint heat balance and evaporation model described here, it is possible to
compare water surface temperatures with available data, now available from the National Oceanic and
Atmospheric Administration's Polar Orbiting Satellite Advanced Very High Resolution Radiometer (Irbe and
Saulesleja, 1982; Irbe et al., 1982; AES, 1988) and reduced for all lakes except Michigan.
The heat balance model was calibrated to determine values of the seven model parameters that give the
smallest sum-of-squared-errors between model and actual daily water surface temperatures observed by satellite
during a calibration period. Daily meteorological over-land data at from five to ten near-shore stations about
each Great Lake are assembled and averaged for correction to over-lake data. The water surface temperature
and the number of days since the last turnover must be initialized prior to modeling or calibration but for
calibration or long simulations the initial values are generally unimportant. The effect of the initial values
diminishes with the length of the simulation and after 1 year of simulation, the effects are nil from a practical
point of view as with the runoff model. Calibrations start on 1 January 1979 and end on 31 December 1985 for
all Great Lakes and start on 1 January 1979 and end on 31 December 1983 for Lake St. Clair. Nearly two years
at the beginning of these periods are without surface temperature data, allowing sufficient initialization prior to
computation of comparison statistics.
There is good agreement between the actual and calibrated-model water surface temperatures;
correlations are greater than 0.98, means and variances both differ only by a few percent, and the root mean
square error is between 1.2 and 15°C on the deep lakes. The period 1966-78 was used as a verification period,
and statistics were computed comparing water surface temperatures to data outside of the calibration period to
assess independently how well it performs; root mean square errors were 13 to 1.6 °C on the deep lakes (Cro-
ley, 1988); the model is judged to perform well in describing deep lake evaporation with less confidence placed
in shallow lake evaporation. Lake Michigan parameters were set from the other lakes.
CONNECTING CHANNEL FLOWS
Lake Superior outflows are found by using "Plan 1977" as implemented by the VS. Army Corps of
Engineers (USACE) for simulation studies (International Lake Superior Board of Control, 1981,1982). It tries
to balance Lakes Superior and Michigan-Huron relative to their long-term levels while considering their normal
variability. It requires a hydraulic routing model for the connecting channels to determine projected water levels
for Lake Michigan-Huron, which then affect the control of Lake Superior outflows. The GLERL Hydrologic
Response Model (HRM) (Hartmann, 1988; Quinn, 1978) uses reservoir routing concepts and discharge equations
to reflect present conditions; it appropriately relates lake storage changes to lake level fluctuations and consi-
ders consumptive use rates (Hartmann, 1987). The Ogoki diversion comprises part of the gaged river flows into
Lake Superior and thus is modeled by the LBRM. Constant diversions of 40, 91, and 261 cubic meters per
second are used for the Long Lake, Chicago, and Welland Canal diversions, respectively. Consumptive use rates
of 7, 56,62, and 15 cubic meters per second are used for Lakes Superior, Michigan-Huron, Erie, and Ontario,
respectively (International Great Lakes Diversions and Consumptive Uses Study Board, 1981). Long-term
average rates of ice retardation of flows over 1937-1981 are used for the St Clair and Detroit Rivers. Lake
Ontario levels and outflows are determined using the USACE implementation of Plan 1958-D (International St.
Lawrence River Board of Control, 1963). The plan attempts to satisfy many, often conflicting, interests, including
riparian, shipping, and hydropower concerns, both upstream and downstream of the lake outlet; it is suspended
often during times of non-normal water supplies.
MODELS APPLICABILITY
Since we have daily models derived for other purposes, we use a daily resolution of data with our
models. However, basin-wide processes of runoff, lake evaporation, channel routing, and lake level change are
often described with weekly or monthly models for lake-level simulation (this ignores short-term fluctuations
associated with storm movement which are not addressed in this study). Likewise, spatial resolution finer than
4-9
-------
Croley
about 1000-5000 km2 (the present average resolution of our models and their applications) is unnecessary, and
much can be done in assessing lake level changes at resolutions of 100,000-1,000,000 km2 with lumped versions
of our models. This coarse spatial resolution is still much finer than the present GCM grids.
There is some indication of model applicability outside the time periods over which the models were
calibrated. The LBRM is used in forecasts of Lake Superior water levels (Croley and Hartmann, 1987) and
comparisons with climate outlooks showed the runoff model was very close to actual runoff (monthly correla-
tions of water supply were on the order of 0.99) for the period August 1982-December 1984, which is outside
of and wetter than the calibration period (Croley and Hartmann, 1986). The model also was used to simulate
flows for the time period 1956-63, outside the period of calibration. The correlation of monthly flow volumes
between the model and observed during this verification period are also contained hi Table 1. They are a little
lower than the calibration correlations but quite good except for Lakes Superior and Huron (there were less than
two-thirds as many flow gages available for 1956-63 as for the calibration period for these basins). Likewise, lake
evaporation models were verified over time periods independent of the calibration period (Croley, 1988). To
assess the applicability of the process models to a climate warmer than the one under which they were calibrated
and verified requires access to meteorologic data and process outputs for the warmer climate which unfortunately
do not exist. Warm periods early in this century are not sufficiently documented for the Great Lakes. In
particular, data are lacking on watershed runoff to the lakes, water surface temperatures, windspeed, humidity,
cloudcover, and solar insolation.
It is entirely possible that the models are tied somewhat to the present climate; empiricism is employed
in the evapotranspiration component of the LBRM and in some of the heat flux terms in die heat balance and
lake evaporation model. Coefficients were determined or selected in accordance with the present climate. The
models are all based on physical concepts that should be good under any climate; however, the assumption is
made that they represent processes under a changed climate that are the same as the present ones. These
include the rainfall-runoff concepts of linear reservoir moisture storages and partial-area infiltration as well as
lake heat-storage relations with surface temperature and gray-body radiation. However, the calibration and
verification periods for the component process models include a range of air temperatures, precipitation, and
other meteorological variables that encompass much of the changes in these variables predicted for a changed
climate. Even though the changes are transitory in the calibration and verification period data sets, the models
appear to work well under these conditions.
4-10
-------
Croley
CHAPTER ID
SIMULATION METHODOLOGY
We constructed a master computer procedure that integrates the Large Basin Runoff Model, overtake
precipitation estimates, lake evaporation models, the Lake Superior regulation plan, the Hydrologic Response
Model, and the Lake Ontario regulation plan to provide a model of the entire Great Lakes system. We develop-
ed it specifically to look at the impact of changed climate by doing simulations with changed meteorology that
represent scenarios of changed climate and comparing with simulations based on historical meteorology
(representing an unchanged climate). Inputs are areal-average daily precipitation and maximum and minimum
air temperatures for each of the 121 watersheds about the Great Lakes and areal-average daily air temperature,
doudcover, humidity, and windspeed for each of the five Great Lakes and Lake St. Clair. Outputs are organized
into three groups corresponding to basin hydrology and runoff, overtake meteorology and evaporation, and net
supplies to the lake and water levels.
STEADY-STATE CLIMATE CHANGE ASSESSMENTS
As instructed by EPA, our general procedure for the investigations of steady-state behaviour under a
changed climate required that we first simulate 30 years of "present" lake levels by using historical daily maximum
and minimum air temperatures, precipitation, windspeed, humidity, and cloud cover data for the 1951-80 period
and present diversions and channel conditions; this is called the "base case" scenario. The initial conditions were
arbitrarily set but an initialization simulation period of 1 January 1948 through 31 December 1950 was used to
allow the models to converge to conditions (basin moisture storages, water surface temperatures, and lake levels)
initial to the 1 January 1951 through 31 December 1980 period. We also then repeated the simulation with initial
lake levels set equal to their averages over the simulation period until they were unchanging to facilitate
investigation of "steady-state" conditions. Then we conducted simulations with adjusted data sets.
Ratios of "future" to "present" absolute air temperature, specific humidity, cloud cover, and precipitation,
and differences of "future" and "present" windspeed were supplied by EPA as GISS (Goddard Institute of Space
Sciences), GFDL (Geophysical Fluid Dynamics Laboratory), and OSU (Oregon State University) atmospheric
model predictions, at grid points spaced 7.83 degrees latitude by 10 degrees longitude, 4.44 by 15, and 4 by 5
respectively, for a "future" atmosphere with twice the CO, content of the "present" atmosphere. We used these
ratios or differences with the historic data to estimate 30-year sequences of atmospheric conditions associated
with a changed climate, referred to as the "2xCCL" scenario(s). We inspected each of the 770,000 square
kilometers within tile Great Lakes Basin to see which of the GISS or GFDL or OSU model grid points it was
closest to and applied the monthly adjustment at that grid point to data representing that square kilometer. By
combining all square kilometers representing a water shed or the lake surface, we derived an areally averaged
adjustment to apply to our areally averaged data sets for the watershed or lake surface, respectively. We then
used each 2xCO2 scenario in simulations similar to the base case scenario. As for the base case scenario, we
repeated the simulation until the lake level averages, also used as initial conditions, were unchanging. We then
interpreted differences between the 2xCO2 scenarios and the base case scenario as resulting from the changed
climate.
Transfer of information between the general circulation models and our hydrologic models in the
manner described involved several assumptions. Solar insolation at the top of the atmosphere and through the
atmosphere on a clear day were assumed to be unchanged, modified only by cloud cover changes. Over-water
corrections were made in the same way, albeit with changed air temperatures, water surface temperatures,
humidities, and windspeeds, which presumed that over-water/over-land atmospheric relationships were
unchanged Our procedure for transfering information from the GCM grid to our spatial data is an objective
approach but simple in concept. It ignores interdependencies in the various meteorologic variables as all were
averaged in the same manner. Of secondary importance, the spatial averaging of meteorologic values over a
GCM grid box filters all variability that exists in the GCM output over that grid box. If GCM output at the grid
4-11
-------
Croley
box corners were supplied and interpolation between these point values were allowed, then at least some of the
patial variability might be preserved. Of course, little is known about toe validity of spatial linear interpolation
and, for highly variable spatial data, it may be inappropriate. However, the same is true for the spatial averaging
that was used to supply GCM results to us.
TRANSIENT-CASE CLIMATE CHANGE ASSESSMENTS
We simulated one transient scenario supplied by EPA: "GISS transient A." As instructed by EPA, our
general procedure for the investigation of this transient-case behaviour under a changing climate required that
we first simulate 80 years of "present" lake levels and component processes over the period 1981-2060. We did
this by using historical daily maximum and minimum air temperatures, precipitation, windspeed, humidity, and
cloud cover data for the 1951-80 period, repeated three times, and applied to initial conditions (basin storages,
water surface temperatures, and lake levels) observed 1 January 1981; only the 80 years from 1 January 1981
through 31 December 2060 are of interest. We repeated three times the procedure developed for steady-state
climate change investigations (reported above). The first simulation used initial conditions observed 1 January
1981; the second used the end-of-run conditions from the first simulation as initial conditions; and the third used
end-of-run conditions from the second. The three simulations were combined to represent the entire period of
interest. After this "base case" scenario was completed, we conducted simulations with adjusted data sets.
The EPA-supplied GISS transient A scenario consists of nine sets of monthly ratios for precipitation,
air temperature, humidity, and cloud cover, and monthly differences for windspeed, one set for each decade from
1970-9 through 2050-9. These ratios or differences were between "present" and "future" values for each of the
data and represent atmospheric model predictions for an increasing atmospheric-cog-content over the period
1971-2059. We used these ratios or differences by interpolating between decadal averages to obtain adjustments
for each month of each year of the period 1981-2059 and by applying them in three simulations as for the base
case: 1981-2010 adjustments to 1951-80 data for the 1981-2010 period simulation, 2011-40 adjustments to 1951-80
data for the 2011-40 simulation, and 2041-59 adjustments to 1951-1969 data for the 2041-59 simulation. We took
the 2060 adjustment as the same as the 2059 adjustment, since the GISS scenario ended in 2059, and applied
it to 1970 data for the 2060 simulation. Discerning the 2xCOg signal from the historical variations in the adjusted
data sets is enhanced by comparing values 30 years apart, thus eliminating the (repetitive) historical variations.
A differencing approach that does this is described in the presentation of results. We combined ratios for each
month of each year of the simulation from the nearest atmospheric model gridpoint for all square kilometers
representing a watershed or the lake surface to derive an areally averaged adjustment to apply to our areally
averaged data sets for each watershed or lake surface. We then used the transient scenario segments in
simulations as we did with the original historical data, combining them to represent the entire period of interest
and then interpreted differences between the transient scenario and the base case scenario as resulting from the
changing climate.
4-12
-------
Croley
CHAPTER IV
RESULTS
Steady-state behaviours in all aspects of the hydrological cycle are exemplified here in figures for the
Lake Erie basin for the GISS and base case comparisons and summarized for all lakes and all climate-change
scenarios for the entire period in tables. Transient behaviour simulations also are summarized here in tables.
CLIMATE CHANGE STEADY-STATE IMPACTS
Basin Meteorology
The 2xCO2 climate (GISS) air temperatures are higher throughout the annual cycle than the lxCO2
climate (base case); the difference is smaller during the summer than during the winter for all lakes but smallest
for the northernmost lakes (Superior, Michigan, and Huron); see Figure 1. The average steady-state GISS air
temperatures are 4.2 to 4.7°C higher, depending on the basin; see Table 2 for this and other scenarios. The
2xCO2 climate precipitation is similar in the winter, spring, and summer to the IxCO, climate precipitation over
all of the Great Lake basins; see Figure 1. In the fall, the 2xCO2 climate precipitation is significantly higher than
in the IxCO, climate on the Superior basin but significantly lower on the other basins. The average steady-state
GISS annual precipitation is 148 mm higher over the Superior basin (expressed as a depth over the land por-
tion of the basin) to 66 mm lower over the Ontario basin with a fairly smooth change with longitude; see Table
2. Precipitation changes are much less consistent than air temperature changes between the various GCMs and
the different lakes as illustrated in Table 2.
Basin Hydrology
The resulting average steady-state GISS evapotranspiration from the land portion of the basin is higher
for the 2xCO2 climate in all cases, with a fairly smooth change with latitude from 153 mm higher over the
Superior basin to 65 mm higher over the Huron basin to 87 mm higher over the Ontario basin; see Figure 1 and
Table 2. Runoff from the land portion of the basin is reduced by the 2xCO2 climate in all basins, changing for
the GISS scenario from only 7 mm lower over the Superior basin to 153 mm lower over the Ontario basin in a
smooth variation with longitude; see Table 2. The average annual cycle of runoff, depicted in Figure 1, has
changed as well; run off peaks slightly earlier and with smaller magnitude under the 2xCO2 climate than under
the lxCO2 climate. This is largely the result of the very big changes observed in the snowpack accumulation and
ablation as well as other basin moisture storages.
In the GFDL climate scenario, evapotranspiration sometimes appears limited by available water. Note
in Table 2 on the Superior and Michigan basins that since precipitation decreases under the GFDL scenario and
increases under the GISS or OSU scenarios, less water is available in these basins for evapotranspiration in the
GFDL scenario than in either the GISS or OSU scenarios. Thus, even though air temperatures are higher under
the GFDL scenario than under the others (and more heat is available to drive evapotranspiration), less actual
evapotranspiration occurs under the GFDL scenario. On the other basins, precipitation increases under the
GFDL scenario enough so that water availability is not so limiting, and evapotranspiration is greater for the
GFDL scenario than for either the GISS or OSU scenarios. In comparing the OSU and GISS scenarios, water
availability is not as limiting under the GISS scenario for the Superior and Michigan basins and more
evapotranspiration occurs. For the other basins, GISS water availability is more limiting and less
evapotranspiration occurs under the GISS scenario than under the OSU scenario.
4-13
-------
Croley 30 p
-10
5
Lake Erie Basin (GISS)
Over-land Air Temperature (°C)
OVerflanci Precipitation (rpm/cjjay)
* i i • i i j
Evapotranspiration (mm/day)
Riinoff (mm/day)
i ! ; *
Show! Waiter
Zone Moisture (mm)
Grouridwater
Storjage
Total Storage
J F M A M J
S O N D
Figure 1. Steady-state GISS Lake Erie over-basin model outputs.
4-14
-------
Croley
Table 2. Average Annual Steady-State Basin Hydrology Differences.
Basin
Air
Temperature
CC)
GISS GFDLOSU
Superior
Michigan
Huron
St. Clair
Erie
Ontario
43
4.7
4.6
4.7
4.7
4.6
72
62
6.4
5.7
5.7
5.9
3.4
3.5
33
3.4
3.4
32
Precipitation
(mm)1
GISSGFDL OSU
148
16
-43
-53
-53
-66
-36
-8
28
56
48
27
58
42
46
51
54
74
Evapo-
transpiration
(mm)1
GISS GFDLOSU
153
92
65
78
85
87
66
76
99
119
121
111
88
85
81
113
119
104
Runoff
(mm)1
GISS GFDL OSU
-7 -102
-76 -84
-109 -71
-132 -64
-139 -74
-153 -84
-31
-43
-35
-63
-65
-30
1 Expressed as a depth over the land portion of the basin.
On the Superior basin, the average steady-state snowpack storage is reduced by more than half; on the
other basins, more to the south, the snowpack is almost entirely absent under the GISS 2xCO2 climate; see
Figure 1 and Table 3. This reduction in snowpack accumulation results from the higher air temperatures,
especially during the winter, that accompany the changed climate. The snow season (period of freezing air
temperatures) is shortened also two weeks to one month under the GISS scenario. The effects on the snowpack
are felt throughout the basin in terms of the derived moisture storages in the soil zone, groundwater, and surface
zones. Figure 1 illustrates the general impact on all Great Lake basins of generally lower moisture storages that
peak earlier in the 2xCO^ climate than in the lxCO2 climate scenarios. This general lowering of moisture in
storage in each of the basins is summarized in Table 3 and in some cases represents greater than a 50%
reduction in available moisture (see "Total Basin Storage" column).
Over-Water Meteorology
The over-lake air temperature, humidity, and wind speed differ from over-land since the lower atmospheric layer
is affected by the water surface over which it lies. The model corrections to over-land meteorology observa-
tions for over-water conditions depend heavily on the water surface and over-land air temperature which in turn
are a function of the over-lake meteorology and heat balance at the surface of the lake. Figure 2 illustrates the
GISS and base case annual cycles for Lake Erie in over-lake meteorology and heat balance effects. Note that
monthly changes from the GISS scenario superimposed on a daily time series creates discontinuities between
months that are reflected in some outputs in Figure 2 (e.g., November-December cloud cover and July-August
wind speed). In general, for all 2xCO2 scenarios, the synergjstic relationship that exists between air and water
temperature yields a general increase in both that follows the lxCO2 climate patterns, similar to over-land
behaviour in Figure 1. Table 4 shows that the average steady-state air temperature difference between the GISS
and base cases varies from 43°C on Lakes Michigan, Erie, and Ontario to 5.5°C on Lake Superior. An increase
with latitude is more pronounced than variation with size of the lake in terms of volume or heat capacity,
although Lake Superior not only has the largest rise in over-lake air temperatures but also has the largest rise
relative to over-land air temperature rise, probably reflecting the large heat storage capacity influence on the air
layer over the lake. Absolute humidities over the lakes have increased appreciably for the 2xCO2 climate, while
cloud cover and over-water wind speed have dropped after adjustment of over-land values for over-water con-
ditions at increased water temperatures.
4-15
-------
Croley
Table 3. Average Annual Steady-State Basin Storages Differences.
Basin
Snow Pack
Water
Equivalent
(mm)1
Soil
Moisture
(mm)'
Groundwater
Moisture
(mm)'
Total
Basin
Storage
(mm)'
GISS GFDL OSU GISS GFDL OSU GISS GFDL OSU GISS GFDL OSU
Superior
Michigan
Huron
St. Clair
Erie
Ontario
-31
-11
-25
-8
-5
-15
-39
-11
-26
-8
-5
-16
-19
-9
-20
-7
-5
-13
-1
-8
-14
-3
-4
-6
-9
-9
-10
-2
-3
-4
-3
-4
-4
-2
-2
-2
-1
-13
-3
-4
-4
-4
-26
-14
-2
-2
-2
-2
-11
-6
-1
-2
-2
-1
-29
-34
-44
-16
-13
-30
-86
-36
-40
-12
-10
-24
-37
-20
-25
-11
-9
-17
'Expressed as depths over the land portion of the basin.
Lake Heat Balance
The heat budget gives rise to increased water surface temperatures as seen in Figure 2 and summarized
in Table 5. Since Lake Erie is a very shallow lake, the annual cycle of the 2xCO2 climate water surface
temperatures follows a pattern very similar to the IxCO, climate but several degrees higher. The average
steady-state water surface temperature increase for the GBS scenario ranges from 43 °C on Lake St. Clair to
5.6°C on Lake Superior, reflecting again the influence of heat storage capacity in a lake. The heat storage
capacity of a lake has another influence on the increase in water surface temperatures that cannot be seen in
Table 5. The higher heat content of Lake Superior earlier in the year allows the 2xCO2 climate water surface
temperatures to peak earlier than the lxCO2 climate; as over-lake air temperatures are affected by the water
temperatures, they also peak ahead of the base case for Lake Superior. Lake Superior then behaves under the
warmer climate as do the deep lakes to the south under the current climate; average surface temperatures peak
in August (currently September for Superior) as the climate for Superior becomes more like the warmer
southern climates. Large amounts of heat now reside in the deep lakes throughout the year. All of the deep
lakes (Superior, Michigan, Huron, and Ontario) show water surface temperatures that stay above 3.98°C (at
which water density is maximum) throughout the average annual cycle for all scenarios (actually the OSU
scenario on Lake Superior did result in two out of thirty years when surface temperatures dipped below 3.98
degrees). This means that buoyancy-driven turnovers of the water column would not occur. It also means that
ice formation will be greatly reduced over winter on the deep lakes.
Note that as the average air temperature increases between scenarios, the average water surface
temperature increases at a reduced rate (this rate further decreases at higher air temperature rises). For
example on Lake Superior, the OSU scenario increases over-land air temperatures 3.4°C over the base case,
the GISS scenario increases them further an additional 0.9°C to 43°C over the base case), and the GFDL
scenario increases them another 2,9°C (to 7-2°C over the base case, see Table 2). The rises between
scenarios of 0.9 to 2.9°C result in rises in water surface temperatures, repsectively, of 0.8 and L8°C; see Table
5. (This is true on all lakes for all air temperature increases between scenarios with the minor exception of Lake
Michigan between the OSU and GISS scenarios.) Part of the reason for this is that water sureface tempertures
are lower bounded at the freezing point while air temperatures are not. However, it is also true that the deep
lakes show smaller rises in water surface temperatures during the fall (including the August peak in surface
temperatures) and winter than do the air temperatures, between the base case and each of the scenarios. This
4-16
-------
Croley
Lake Erie (GISS)
Over-lake Air TemperatureTC)
r^zP—En:
100
JFMAMJJASOND
Figure 2. Steady-state GISS Lake Erie over-lake model outputs.
4-17
-------
Croley
Table 4. Average Annual Steady-State Over-Lake Meteorology Differences.
Basin
Air
Temperature
(Deg C)
Absolute
Humidity
(mb)
Cloud Cover Wind Speed
(%) (m/s)
GISS GFDL OSU GISS GFDL OSU GISS GFDL OSU GISS GFDL OSU
Superior
Michigan
Huron
St. Clair
Erie
Ontario
5.5
4.9
5.1
5.1
4.9
4.9
7.0
6.0
6.6
5.9
5.6
5.9
4.5
3.5
3.7
3.8
3.4
3.5
4.4
4.4
4.1
5.4
5.1
4.6
6.1
5.7
5.5
6.4
5.8
5.8
3.7
3.2
3.0
3.7
3.5
3.3
-1.1
-0.1
-4.4
-3.7
-4.9
-4.8
-5.8
-2.0
-2.8
-2.9
-4.1
-5.4
-4.2
-2.6
-3.4
-2.6
-3.6
-3.2
-0.1
-0.2
-0.1
-0.1
-0.1
-0.1
-0.2 0.0
-0.4 -0.1
-0.2 0.0
-0.3 -0.1
-0.3 -0.1
-0.3 0.0
Table 5. Average Annual Steady-State
Lake Surface Temperature and
Evaporation Differences.
Basin
Water Surface
Temperature
(Deg C)
Overlake
Evaporation
(mm1)
GISS GFDL OSU GISS GFDL OSU
Superior
Michigan
Huron
St. Clair
Erie
Ontario
5.6
4.7
4.7
4.3
4.4
4.9
7.4
5.5
6.0
5.0
5.0
5.9
4.8
3.4
3.6
3.0
3.0
3.6
152
176
199
297
290
213
284
279
297
408
414
281
173
179
166
262
232
172
'Expressed as a depth over the lake.
is not illustrated well in Figure 2 for Lake Erie, but is very prominent on all deep lakes. Even in comparing cool
and warm years from the historical record, where there are comparable rises in spring and summer air and water
temperatures, the fall and winter rises of surface temperature are smaller than the air temperature increases.
The point is that air temperature rise is only a partial indication of what can be expected of water surface
temperatures. Water temperatures depend on the total heat balance of the lake with the atmosphere which, in
turn, depends on changes in humidity, wind speed, and cloud cover in addition to air temperatures. As average
air temperatures increase, the average water surface temperatures (in particular, the surface temperatures during
the evaporation season of fall and winter) generally increase at lower rates and the rate decreases as air
temperatures rise. This is contrary to standard assumptions in other works that do not perform heat balances
to determine surface temperatures but set surface temperature rises equal to air temperature rises (Cohen, 1986,
1987). However, the evaporation computations are found to be very sensitive to this assumption since the fall
4-18
-------
Croley
and winter are when most of the evaporation occurs on all of the lakes. At high air-water temperature
differences, the effects of humidity and wind speed changes are secondary to the temperature (and hence vapor
pressure) differences, and large evaporation may be calculated even with offsetting humidity and wind speed
changes as evident in Tables 4 and 5. At lower air-water temperatures, the effects of the humidity and wind-
speed changes become primary in determining the relative magnitudes of evaporation between the various scen-
arios.
The higher water surface temperatures under the 2xCO2 climate result in increased annual lake
evaporation ranging from 152 mm over Lake Superior to 297 mm over Lake St. Clair for GISS; see Table 5.
The shallow lakes have the highest increase in evaporation under all 2xCO2 scenarios while the deep lakes have
similar (but smaller) increases. Note that the Lake Michigan evaporation may be suspect because no data were
used in the calibration of the Lake Michigan heat balance; calibrated parameters from Lakes Huron, Erie, and
Ontario were adapted for use on Lake Michigan. Note that while average humidities are up and average
windspeeds are down in Table 4 (by themselves suggesting that evaporation drops), evaporation in Table 5 is
higher. This is because the water surface temperature (and associated saturated vapor pressure at the surface)
has increased sufficiently to over-compensate the rise in atmospheric humidity (die average vapor pressure
difference between the water and the air has increased) and drop in wind speed. Interestingly, the Lake
Superior evaporation under the GISS scenario is less than that under the OSU scenario even though GISS water
surface temperatures rise more than OSU (compare 5.6 to 4.8 degrees, respectively, in Table 5). However, in-
spection shows that with the increase in air temperatures consequent with the GISS scenario, the over-land
air-water temperature difference (see Tables 2 and 5) drops less than with the OSU scenario (43 - 5.6 = -13
for GISS and 3.4 - 4.8 - -1.4 for OSU); this means that air stability over the lake is larger generally with the
GISS scenario than with the OSU scenario, resulting in larger over-lake air temperature and dewpoint
corrections toward the water surface temperatures. Resulting air-water vapor pressure differences and
windspeeds are lower (the latter is reflected in Table 4) with consequent lower evaporation.
Net Supply Components
Over-lake precipitation, runoff, and lake evaporation sum algebraically as the net basin supply and are
presented again in Figure 3 for convenience. Since over-lake precipitation is taken here as the same as
over-land, Figure 3 and Table 6 show the same relations for GISS vs. base case precipitation as does Figure 1
and Table 2. Runoff in Figure 3 is the same as in Figure 1 except it is expressed as a depth rate over the lake
rather than over the basin. Net basin supply in the GISS scenario in Figure 3 is seen to be uniformly less under
the 2xCO2 climate than under the IxCO, climate; this is true for Lakes St. Clair and Erie. It is nearly true on
Lakes Huron (only January supplies are nigher) and Ontario (only January and February supplies are higher);
Lake Michigan supplies are smaller 75% of the time and Lake Superior supplies are smaller only about half of
the time. Lake Superior experiences increased net basin supplies during the fall, and winter under the 2xCO2
climate and Lake Michigan has increased net basin supplies during the winter. This leads to a reduced drop
in the average steady-state net basin supply on Lakes Superior and Michigan for the GISS scenario; it sig-
nificantly drops for all other Great Lakes and all other scenarios as shown in Table 6. The GISS scenario in
Table 6 results in a larger precipitation rise and a smaller drop in basin runoff for Lake Superior than do either
the GFDL or OSU scenarios, resulting in higher net basin supply (and lower difference with the base case) than
for these other scenarios. Lake Michigan shows a larger precipitation rise for the GISS scenario than for either
Figure 3. Steady-state GISS Lake Erie lake-level model outputs.of the other two and a smaller drop in basin
runoff for the GISS scenario than for the GFDL. On the remaining lakes, the GISS scenario results in drops
in over-lake precipitation (instead of rises as with the GFDL and OSU scenarios) and larger drops in basin
runoff than do the other scenarios, resulting in larger drops in net basin supplies for the GISS scenario. Table
7 summarizes the changes in the net basin supply components for the entire Great Lakes basin; they were
computed by converting the equivalent depths of Table 6 to annual flow rates on each lake and adding them over
all the lakes. The changes from the base case are also expressed relatively in Table 7. Even though more heat
is available under the GFDL scenario than under the GISS or OSU scenarios, evapotranspiration is lower
because less water Is available, as seen by inspection of the average precipitation in Table 7. In the OSU and
GISS scenarios, water availability is not as limiting and the higher air temperatures of the GISS scenario lead
4-19
-------
Croley
6 r
Lake Erie (GISS)
Over-lake Precipitation (mm/day)
M'
.
jnoff (aim/day)
aporatijon (him/day
(mm/day)
Inflow (mm/day)
172.6
JFMAMJJASOND
Figure 3. Steady-state GISS Lake Erie lake-level model outputs.
4-20
-------
Croley
Table 6. Average Annual Steady-State Net Supply Components Differences.
Basin
Over -Lake
Precipitation
(mm)1
Basin
Runoff
(mm)'
Over -Lake
Evaporation
(mm)'
Net Basin
Supply
(mm)'
6ISS GFDL OSU GISS GFDL OSU GISS GFDL OSU GISS GFDL OSU
Superior 148 -36 58 -11 -159 -48
Michigan 16 -8 42 -152 -168 -87
Huron -43 28 46 -241 -158 -78
St. Clair -53 56 51 -1870 -901 -891
Erie -53 48 54 -328 -173 -153
Ontario -66 27 74 -565 -311 -112
152 284 173 -16 -479 -163
176 279 179 -312 -455 -223
199 297 166 -483 -427 -197
297 408 262 -2219-1252-1102
290 414 232 -671 -540 -331
213 281 172 -844 -564 -210
'Expressed as depths over the lake.
Table 7. Average Annual Steady-State Great Lakes Basin Hydrology
and Net Basin Supply Components.
Scen-
ario
BASE
GISS
GFDL
OSU
Over
Land
Precip-
itation
(cms)
13637
13871 +2%
13725 +1%
14483 +6%
Evapo-
trans -
plration
(cms)
7727
9317 -21%
9176 -19%
9204 -19%
Basin
Runoff
(cms)
6090
4658 -24%
4714 -23%
5438 -11%
Over
Lake
Precip-
itation
(cms)
6499
6747 +4%
6501 +0%
6903 +6%
Over
Lake
Evap-
oration
(cms)
5352
6821 +27%
7685 +44%
6745 +26%
Net
Basin
Supply
(cms)
7237
4584 -37%
3530 -51%
5596 -23%
to higher evapotranspiration than in the OSU scenario even though more water is available under the OSU
scenario. Note also that in Table 6, the lesser drops in net basin supplies for the GISS scenario compared to
the GFDL scenario for Lakes Superior and Michigan offset the greater GISS drops for all other lakes, so that
the total GISS net supply drop over all basins is smaller than the overall GFDL drop and the summed net basin
supplies in Table 7 are lower under the GFDL scenario than under the GISS scenario.
Lake Water Balance
Results in Figure 3 reveal that river inflow and outflow make up a large part of the water budget of
Lake Erie, and drops in Lake Erie inflows are accompanied by drops in outflows. The drop in lake levels comes
from the drop in net basin supplies and the net drop in inflows and outflows is only partly offsetting. Table 8
shows that Lake Erie's average GISS steady-state inflow minus its outflow rises 656 mm (compare to the drop
in net basin supplies of 671 mm) while lake levels drop about 1.16 m. Similar drops in net flows are observed
on all lakes under all scenarios and the lake levels drop on all of the Great Lakes. Some of the planned results
were not obtainable because of failure of the regulation plan(s) under the climate change scenarios. As earlier
4-21
-------
Croley
Table 8. Average Annual Steady-State Flows Differences.
Basin
Inflow
(mm1)
Outflow
(mm')
GISS
GFDL OSU
GISS
GFDL OSU
Superior
Michigan-Huron
St. Clair2
Erie
Ontario
-10
-40628
-1857
-3407
-113
- -32892
-1473
-2432
-14
-386
-42831
-2513
-
-162
-312
-33980
-1794
-
'Expressed as a depth over the lake.
'Lake St. Clair inflows and outflows are relatively large
since it is such a small lake (wide spot on the river).
Table 9. Average Annual Steady-State Lake Levels
Differences.
Basin
Superior
Michigan-Huron
St. Clair
Erie
Ontario
GISS
(m)
-0.46
-1.31
-1.21
-1.16
-
GFDL
(m)
.
-2.48
-2.12
-1.91
-
OSU
(m)
-0.47
-0.99
-0.87
-0.79
-
climate change and consumptive use studies by others have found, the Lake Ontario regulation plan's mathemati-
cal algorithms behave erratically under extreme low water supply conditions (RH. Quinn, GLERL, personnel
communication, 1988; C. Southam, Environment Canada, personal communication, 1988).
The average steady-state lake-level difference summary for all 2xCO2 scenarios is given in Table 9. The
GFDL scenario produced such low water supplies to Lakes Superior and Michigan that the rule curves of both
regulation plans (Lakes Superior and Ontario) failed. The Lake Superior regulation plan uses a balancing
approach that attempts to consider the impacts of regulation as well on Lakes Michigan and Huron. However,
the balancing equation fails when Lake Superior's levels become too low. To get some idea of system behaviour
under the GFDL scenario without using many different regulation rules to avoid ambiguity, Lake Superior
outflows were taken equal to Lake Superior net supplies and other inputs (e.&, Long Lac diversions) on an
annual basis. This probably is correct over long periods (actual Lake Superior net basin supplies plus diversions
differ from the St. Mary's River outflows by less than 1% over 1951-80) and, for the 30-yr period used herein,
gives an estimation of long-term differences in lake levels. No attempt was made to determine what Lake
Ontario levels might be. The Lake Superior net supplies and Long Lac diversions (Ogoki diversions are already
included in the net basin supplies) were used as the St Marys River flows with the GLERL Hydrologic Response
Model The lakes drop more under the GFDL and GISS scenarios than under the OSU scenario because of
their much larger drops in net supplies to all lakes but Superior than is observed under the GISS scenario. Lake
4-22
-------
Croley
Superior drops more tinder the OSU scenario since GISS supplies are not reduced as much, as discussed above.
GFDL has smaller drops in net basin supplies than GISS for Lakes Huron, St Clair, Erie, and Ontario yet larger
drops in the water levels on these lakes because of its larger drop in supplies to Lakes Superior and Michigan
which affect inflows to lower lakes and offset the reduced drops in supplies there. Because the GISS scenario
shows larger drops in net basin supply on the remaining downstream lakes than the GFDL and OSU scenarios,
the drops in lake levels in Table 9 under the three scenarios approach each other somewhat on the downstream
lakes. The lake levels drop 40 to 130 cm for the GISS scenario, 190 to 250 cm for the GFDL scenario, and 50
to 100 cm for the OSU scenario.
CLIMATE CHANGE TRANSIENT IMPACTS
There are serious difficulties in analyzing transient impacts of climate change with short historical data
sets as outlined under "Methodology." In simulating 80 years (1961-2060) by using 30 years of historical data
(1951-80) repeated three times (for the 1981-2010,2011-2040, and 2041-2060 periodsX the variations contained
in the historical record repeat, of course. As corrections to the historical data are made to reflect the transient
GISS climate changes, it is found that the repeating fluctuations of the historical record so completely dominate
the superimposed climate changes that it is difficult to see the effect of climate change. Thirty-year averages
could be used to filter historical variations for presentation of results; however, only two complete 30-yr periods
are contained in the simulation and a 30-yr average filters some of the climate change as well A differencing
approach is used here to give a very general idea of transient behaviour; it works by comparing transient
behaviours 30 years apart to eliminate the effect of the historical variations. By comparing the GISS and base
cases transient simulation changes for decades 1,4, and 7 (1981-90, 2011-20, and 2041-50, which are all based
on the same 1951-60 data period), decades 2,5, and 8 (1991-00,2021-30, and 2051-60, which are all based on the
same 1961-70 data period), and decades 3 and 6 (2001-10 and 2031-40, which are both based on the same
1971-80 data period) in the tables following, the effect of the repetitive natural variations is eliminated since the
same underlying historical data segments are used in each grouping. The climate change trends then can be
identified.
As an example of applying this differencing approach, decadal-average annual changes in air temperature
between the 80-year base case and transient simulations are given in Table 10 for Lake Superior; air temperature
is increasing on the order of about 1.6°C per 30-yr period. This is found by subtracting GISS-base changes for
decade 1 from 4,4 from 7,2 from 5,5 from 8, and 3 from 6 and averaging to get, respectively. 0.9 + 1.9 + 13
+ 23 + 1.8 • 8.2, and dividing by 5 (the number of comparisons) to get 1.64 per 30 years. This is more
conveniently expressed as 0.5*C per decade. Other hydrometeorological quantity variations are computed in the
same way and are summarized in Table 11.
Table 10. Lake Superior Basin Transient GISS Decadal-Aver age Annual Over-
Land Air Temperature Changes from Base Case (°C).
Decade 1981-90 1991-00 2001-10 2011-20 2021-30 2031-40 2041-50 2051-60
Number 12345678
0.7 0.9 1.2 1.6 2.2 3.0 3.5 4.5
On Lake Superior, precipitation increases are partially offset by evapotranspiration increases so that
annual runoff, expressed as a depth over the land portion of the basin, increases about 4 mm each decade. All
other lakes show a drop in annual runoff of 7-15 mm/decade. The average snowpack accumulates about 1-4 mm
less each decade, and average soil zone moisture and groundwater generally drop 03-1 mm/decade (except on
Superior); the resultant effect is a lowering of total basin moisture storage about 0.1-5 mm/decade. Over-lake
4-23
-------
Croley
air temperatures increase at about the same rate as over-land average air temperatures. Over-lake humidity
generally increases slightly (but negligibly) each decade; note however that while it is corrected for over-water
conditions in the simulations, it is not computed by taking increased evaporation into account. Rather, over-land
humidity is supplied in the GISS atmospheric model outputs (inputs to this simulation).
Cloud cover generally decreases slightly (but negligibly) and is not influenced by these simulations; it
is strictly an input here from the GISS atmospheric model Over-lake wind speed is almost not affected and the
water surface temperature increases by about 0 J-0.7°C each decade. Resultant annual lake evaporation increases
18-40 mm/decade. Net basin supplies are highly variable but generally drop on all lakes except Superior. These
changes in net supplies are reflected in changes in outflows and lake levels. The increased supplies to Lake
Superior are more than offset by increases in net outflow, which results in a small decline in Superior water
levels of about 13 mm/decade. The associated rise in the combined Michigan and Huron inflows allows the
annual net outflow in Table 11 to drop only 31 mm/decade; their combined net supplies drop more allowing lake
levels to fall about 59 mm/decade. This is a consequence of the Lake Superior regulation plan, which endeavors
to balance water levels on Lakes Superior, Michigan, and Huron about their long-term mean values. On the
lower lakes, there are drops in net basin supplies, net outflows, and lake levels; Lake St. Clair drops about 64
mm/decade, while Lake Erie drops about 66 mm/decade and Lake Ontario drops about 93 mm/decade. Only
the first seven decades were used in computing the average decadal difference in net outflows and lake levels
for Lake Ontario, since the regulation plan failed in the eighth decade. Note that while there are similar patterns
in the behaviour of the lake levels between this transient analysis and the GISS steady-state analysis (see Table
9), the magnitudes of the drops are dissimilar (while of the same general order of magnitude).
Table 11. GISS Transient Changes Summary.
Hydrologlcal Variable
Basin Air Temperature
Annual Basin Precipita.1
Annual Basin Evapo trans.'
Annual Basin Runoff
Snowpack1
Soil Moisture1
Groundwater1
Total Basin Moisture1
Lake Air Temperature
Lake Humidity
Lake Cloud Cover
Lake Wind Speed
Surface Temperature
Annual Lake Evaporation2
Annual Net Basin Supply2
Annual Net Outflow2
Lake Level
Units
(0C/dec):
(mm/dec) :
(mm/dec) :
(mm/dec) :
(mm/dec) :
(mm/dec) :
(mm/dec) :
(mm/dec) :
(°C/dec):
(mb/dec) :
(%/dec) :
(m/s/dec) :
(0C/dec):
(mm/dec) :
(mm/dec) :
(mm/dec) :
(mm/dec) :
Sup
+0.5
+29
+25
+4
-4
+0.5
+2
-0.1
+0.7
+0.6
+0.1
+0.0
+0.7
+18
+17
+20
-13
. Mic
+0.6
+7
+14
-7
-1
-0.8
-1
-4
+0.6
+0.5
-0.2
+0.0
+0.5
+19
-27
-3
_ i
. Hur
+0.6
+2
+12
-9
-4
-1
-0.3
-5
+0.6
+0.5
-0.5
+0.0
+0.6
+22
-41
1
59
Basin
. StC
+0.6
-0
+15
-15
-1
-0.4
-0.4
-2
+0.6
+0.7
-0.4
+0.0
+0.5
+38
-245
-241
-64
. Eri
+0.6
+3
+16
-15
-0.7
-0.4
-0.4
-2
+0.6
+0.7
-0.5
+0.0
+0.6
+40
-75
-70
-66
. Ont.
+0.6
+1
+16
-14
-2
-0.6
-0.3
-4
+0.6
+0.6
-0.5
-0.0
+0.6
+24
-75
-573
-93J
'Expressed as a depth over the basin.
^Expressed as a depth over the lake.
'Computed over first 7 decades since Ontario regulation plan fails in eighth.
4-24
-------
Croley
SENSITIVITIES
Without temperatures below freezing, the snowpack is insensitive to precipitation. Although the GISS,
GFDL, and OSU scenarios show conflicting estimates of possible precipitation changes, each shows increases
in air temperatures that significantly reduce snowpack storage, especially in the southern Great Lakes basin.
Thus, even if precipitation increases are larger than any suggested by the GCMs, snowpack will be much reduced
under wanner winters. Similarly, regardless of actual changes in precipitation, the Great Lakes basin will
experience reductions in soil moisture storage and runoff. Soil moisture and runoff peak shortly after snowmelt
and then drop throughout the summer and fall due to high evapotranspiration demands; each climate scenario
produces earlier snowmelt and a longer period of evapotranspiration. Although soil moisture and runoff
certainty vary with precipitation, they are most sensitive to it in midsummer when at their annual minimums.
Thus, within the limits of precipitation produced by the GCMs, soil moisture and runoff scenarios are fairly
insensitive to precipitation.
Of the meteorological variables that affect lake evaporation (air temperature, humidity, cloud cover, and
wind speed) wind speed is probably the most critical, although air temperature and humidity are also important.
Over-lake evaporation is very sensitive to the air temperature increase inherent in the scenario, which governs
the sensitivity to humidity and wind speed changes. In the range of air temperature increases considered for the
steady-state scenarios, the higher temperature rises associated with the Lake Superior GISS scenario compared
to the OSU scenario increased over-lake air stability to such an extent that evaportion actually decreased
compared to the OSU scenario, and this was not true with smaller air temperature rises. For the smaller air
temperature rises, there was more opportunity for increases in evaporation since the rise in water surface
temperatures (and vapor pressure difference with the atmosphere) compensated for the slighter drop in wind
speed and rise in atmospheric humidity. That this turn around point in the behaviour of evaporation occurred
in the range of climate variations considered under the three climate change scenarios, showing the uncertainty
that may be associated with climate change evaporation estimates.
Because net basin supplies are a simple addition of lake evaporation, runoff, and precipitation, they are
equally sensitive to changes in any of the components. Inspection of Table 7 reveals much greater variability in
the estimates of the net basin supplies across the different scenarios than in the components. Partly, this is
because limits were reached (as in the case of limiting water availability for evapotranspiration under the GFDL
scenario or in the case of air temperature increases resulting in increases in stability and drops in lake
evaporation under the GISS Superior scenario), but this also illustrates the potential for uncertainty in changes
in estimated net supplies (and lake levels).
4-25
-------
Croley
CHAPTER V
IMPLICATIONS OF RESULTS
Climate change has the potential to alter many traditional activities in the Great Lakes region. Although
changes in lake levels have major implications for Great Lakes users, altered precipitation, evaporation, and
runoff patterns will be important as well. Environmental implications and socioeconomic implications are
addressed separately herein; however, these concerns are often intimately intertwined since a healthy ecosystem
is vital for the social or economic well-being of a number of interests (e.g., riparian, commerical fishing,
recreation, commercial interests) and some issues (e.g^ disposal of toxic dredge spoil) combine environmental,
social, and economic dimensions.
ENVIRONMENTAL IMPLICATIONS
Lowering of Great Lakes levels could dramatically affect the Great Lakes' ecosystem production through
dependence on the consistent availability of marshes and wetlands that serve as breeding and nursery areas for
fish and wildlife. Even a 20-cm lowering of Lake Michigan-Huron levels could affect 64% of all Great Lakes
wetlands in the U.S. (Manny, 1984). Although confined wetlands may be especially vulnerable to disruption from
lake level declines (Wall, 1988), even open shoreline wetland extents could be permanently reduced due to their
direct lake level dependence, unsuitable offshore substrates, and steep offshore dropoffs, combined with a
resulting reduction in seeds and rhizomes for colonization (Manny, 1984). On the other hand, the total areas
of different wetland types may remain nearly unchanged if water levels drop so slowly that the shoreline can
adapt (Meisner et al., 1987).
Decreased wetland extents could significantly reduce Great Lakes fisheries production, even in deeper
waters; over half of all Great Lakes fish species use wetlands for spawning and nursery habitat (Goodyear et al.,
1982). Meisner et al. (1987) provide an initial assessment of the potential impacts of increased water
temperatures on Great Lakes fishes. Among their general expectations are northward shifts in the geographical
distribution of warm and cold water species, changes in relative abundance of species within fish communities,
and changes in yields of different species. Table 12 summarizes some of these anticipated impacts.
Increased water temperatures may result in substantial changes in the Great Lakes ecosystem. With
water temperatures remaining above 4°C throughout the year, buoyancy-induced turnovers in the fall and spring
may not occur. Without turnover, hypolimnion chemistry may be altered; oxygen may be depleted, releasing
nutrients and metals from lake sediments. On the other hand, the lakes may experience a single winter turnover
even with water temperatures above 4°C if temperature gradients are small and winds are strong enough to
induce turbulent mixing (Hutchinson, 1957).
SOCIOECONOMIC IMPLICATIONS
Power Production
The waters of the Great Lakes are extensively used for hydropower production. Facilities range from
low-head plants on the St. Mary's River to high-head facilities in the Niagara and St. Lawrence Rivers. A climate
warming would result in decreased flows and water surface elevations, which would contribute to lower
hydropower productioa This could be especially important since hydropower is inexpensive and nonpolluting
when compared to the primary alternatives, fossil fuel or nuclear power facilities.
4-26
-------
Croley
Table 12. Anticipated Impacts of Increased Water Temperatures On
Selected Great Lakes Basin Fishes1.
Species
smallmouth bass
largemouth bass
bigmouth buffalo
lake trout
lake whitefish
brook trout
whitefish
yellow perch (south)
walleye (south)
alewife
yellow perch (north)
walleye (north)
lake whitefish
northern pike
walleye
Impacts
- northward extension of range
- northward contraction of range
contraction of range to stream headwaters
• reduced populations due to competition
with other trout for remaining habitat
decreased populations due to increased
egg and larval mortality or inhibited
reproduction
increased populations due to increased
reproduction and reduced mortality
- decreased sustainable yield
'Meisner et al., 1987
Fossil fueled and nuclear power plants sited around the lakes use lake water for cooling; conversion to these
modes of power production would increase the consumptive use of water for cooling (via evaporation), which
would further exacerbate the anticipated low lake levels. Coal-fired power plants additionally require the
economic efficiencies provided by waterborne transportation of coal; with lower water levels, higher trans-
portation costs would directly affect power production costs.
The full impact of climate change on power production interests depends not only on the water supplies
available for hydropower, cooling, or transportation, but on the changes in peak power demands that result from
the increased air temperatures. In much of the U5. portion of the Great Lakes basin, peak power demands
occur in the summertime for cooling (R. Crissman, New York Power Authority, personal communication, 1988);
climate warming could increase peak power demands, making the loss of hydropower production even more
critical. On the other hand, climate warming may substantially reduce the peak power demands for winter
heating that occur in Canada, making replacement of hydropower facilities nonproblematic (J. Eaton, Ontario
Hydro, personal communication, 1988). However, impacts on peak power demands are difficult to predict since
they are so closely tied to population levels, and continued growth in the use of air conditioners in Canada could
raise summer peak demands above winter levels.
4-27
-------
Croley
Navigation
The Great Lakes/St. Lawrence Seaway is a major freshwater transportation system. This system
depends upon adequate depths in the connecting channels and harbors to function at full capacity. During
conditions of low levels, more trips must be made to move the same amount of cargo; this increases shipping
costs, and the increased traffic could cause backups at recognized bottlenecks in the system (e.g., Welland Canal).
Any impacts on the shipping industry will have direct impacts on industries that depend upon the Great Lakes
for transport of production inputs or final products (e.g^ iron, steel, grain).
Shipping interests suggest that industry contraction in the 1960s was related to the record low water
levels of that period (Marchand et al., 1988). Under the climate change scenarios, those low levels will be typi-
cal. Climate change impacts on the shipping industry could be made even more severe if the trend toward
increased vessel size continues; for dense cargos (e.g., iron ore) those ships' capacities are extremely sensitive
to changes in draft. Decreased channel depths will likely require extensive dredging in both the connecting
channels and harbors to maintain present shipping capabilities. However, bottom sediments in the Great Lakes
are highly contaminated with toxics, creating a problem with dredge spoil disposal.
Industrial Operations
Industrial interests typically use the Great Lakes as part of the production process (e.g., water supply,
waste disposal) and/or as a mode of transportation for inputs or products. These interests include companies
involved in a wide range of activities, including grain shipment, food processing, pulp and paper processing,
petroleum refining, organic chemicals and synthetics, inorganic chemicals, industrial minerals, metal mining and
refining, iron and steel production, metal casting, metal plating, and plastics fabrication. Thus, lowered Great
Lakes levels would likely have adverse effects that pervade regional and national industrial operations and
economies.
Commercial Operations
For many businesses in the Great Lakes region (e.g., marinas, hotels, resorts, restaurants), economic
success is intimately tied to their shoreline location. As lake levels fall, these businesses will experience prob-
lems similar to those of the 1960s: reduced scenic views, inaccessible docking facilities, and unusable water
intakes or waste disposal outlets. However, these adverse effects would likely be only transitory; although
individual businesses may suffer under the steady-state climate change conditions, the industry as a whole will
likely be able to adapt by simply moving with the lakeshore. There may be notable exceptions, however. The
lower lake levels and connecting channel water levels would greatly reduce the areas accessible to small craft,
including passenger vessels. This could require extensive private dredging and the rebuilding of ramps. As in
the major harbors used by the navigation industry, bottom sediments are highly contaminated with toxics. Many
municipalities and private harbors would likely have to cease operation if faced with the costs associated with
dredging and disposal of contaminated spoil (P. Keillor, Sea Grant Institute, personal communication, 1988).
Other types of businesses may be affected more by hydrometeorologic changes than lowered lake levels;
thus, these impacts will not be limited to lakeshore businesses. Commerce that depends upon reliable snow
cover (e.g., skiing, snowmobiling, ice fishing) may suffer total collapse throughout much of the Great Lakes
region, especially in areas that are only climatically marginal for winter recreation at present; other areas (e.g.,
the more northern portions of the Lake Superior and Huron basins that are also subject to lake effect snows)
may still have sufficient snow during their peak tourist season (Wall, 1988). Riverine canoe rental operations
may become more seasonal as river flows become too low except during peak runoff periods.
Commercial Fishing
Commercial fishing operations (e.g., commercial anglers, fish packers, processors, exporters) use the
Great Lakes to provide an essential production input. Harbor access problems will occur due to the lowered
lake levels, but changes in fisheries population structures may be even more important. Fisheries production is
4-28
-------
Croley
intimately tied to wetland extent, which may be irretrievably lost as lake levels are lowered (Manny, 1984). Even
if stocking programs are used to maintain fish populations, the industry may have to adjust to use of different
species since increased water temperatures and the absence of semi-annual lake turnovers may affect what fish
can be harvested at marketable levels. Additionally, climate change could indirectly affect commercial fishing
operations; if dredging of channels and harbors makes toxic contaminants available for uptake by fish, their
market ability may be limited.
Agriculture
Agricultural interests near Great Lakes shorelines use the lakes for water supply and take advantage of
the soil fertility and climate moderation provided by lakeshore locations. Climate change will likely provide both
benefits and costs to agriculture. More moderate winter temperatures may provide suitable conditions for
significant increases in insect and disease problems. In northern portions of the Great Lakes basin, increased
air temperatures would improve growing conditions and enable major agricultural expansion (Smit, 1987).
Agricultural producers located in areas without substantial moisture reserves would experience harsh growing
conditions during the summer as soil moisture drops below levels suitable for traditional crops (Smit, 1987).
Lakeshore producers could mitigate any short falls in soil mositure by increased use of Great Lakes water for
irrigation, although possible exacerbation of lower lake levels could create conflicts with other interest groups.
Under lower lake level conditions, many low-lying areas will have improved drainage and additional fertile
bottom-lands will be available; however, those rich lands may also contain exceedingly high levels of toxics,
limiting farming potential Also, as agricultural products are often shipped via waterborne transport, lowered
lake levels will increase transportation costs for Great Lakes farming products.
Recreation Interests
Recreation interests use the Great Lakes in such a wide variety of ways that climate change is certain
to provide both benefits and costs. Recreation activities that depend upon the ecosystem production of the Great
Lakes (e.g., fishing, hunting, birding) will suffer if productivity is reduced due to loss of wetlands or lack of lake
turnovers. For recreation uses dependent on other lake amenities (e.&, shoreline location), adverse effects
would likely be only transitory; although recreation at specific sites may suffer under the steady-state climate
change conditions, the recreational activities (e.g., beach use, water sports, camping) as a whole will likely be able
to adapt by simply moving with the lakeshore. There may be notable exceptions, however. Long-term loss of
marina and harbor access due to lower lake levels and high costs of toxic dredge spoil disposal could affect
recreational boating. Beach use could be affected if public access to the shoreline is not maintained. Winter
recreation (e.g., siding, ice fishing) would be negatively affected by the warmer winter temperatures. Riverine
canoeing opportunities could be reduced as river flows become too low during all but peak runoff periods.
Riparians
Great Lakes riparian interests derive many benefits from their shoreline location, including water
supplies, shore access for beach use and other recreation, and scenic vistas. Individual benefits can be very sensi-
tive to lake levels as their property becomes inundated or far removed from the waterfront. Under the climate
change scenarios examined, transitory effects for riparians may be severe. Under steady-state conditions of cli-
mate change, however, although individual property owners may suffer, riparians as a whole will likely adapt by
simply moving with the lakeshore.
POLICY IMPLICATIONS
Climate warming may require new paradigms of how the Great Lakes will be viewed from social,
economic, and ecological perspectives. Reduced water supplies and lake levels likely will require management
strategies to resolve conflicts over the allocation of Great Lakes water. Management will be made especially
difficult because different interest groups are affected differently by climate change; while some groups may
experience severe negative effects, others may actually benefit. In addition, Great Lakes uses are often
4-29
-------
Croley
conflicting; optimizing for one interest may adversely affect another. System interdependences (i.e., changes in
levels or flows in one part of the Great Lakes system affect levels and flows throughout the remainder of the
system) add to the complexity of management. A major policy decision will be the distribution of benefits
between commercial, riparian, recreational, and ecological interests, between upstream and downstream interests,
and finally, between the many jurisdictional (e.g., U.S., Canadian, federal, state, provincial, municipal) interests.
It is likely that the US. and Canadian governments will react based upon the relative strengths of the interest
groups and the pressures they can exert.
Presently, the Boundary Waters Treaty between the VS. and Canada mandates a specific hierarchy of
Great Lakes interests to be protected or enhanced; the primary interest is domestic and sanitary uses, followed
by navigation, and power and irrigation, in that order (International Joint Commission, 1965). The treaty may
have to be modified to reflect the changed relative importance of those interests compared to other commercial
and industrial interests, riparian, recreational, and ecological interests.
No historical analog, comparable to conditions suggested by the climate change scenarios, exists to
provide insight as to what the management response will be to a prolonged period of extremely low lake levels.
During the relatively mild and short-term low levels of the mid-1960s, there was an increased emphasis on
bringing additional water into the system, improved regulation, and on further system-wide water level
regulations to counteract lake level lowerings that resulted from historical dredging and mining operations. A
major thrust of water management under a warmer climate will probably be to keep water in die system. This
will require extensive revision of the existing Lake Superior and Ontario regulation plans as well as the possi-
ble regulation of Lakes Michigan-Huron and Erie. The existing regulation plans were not designed for the low
net supplies expected with climate change and failed in the assessment simulations. The debate over interbasin
diversions of water will also likely intensify. There will probably be demands to increase the amount of water
brought into the Great Lakes through existing diversions into Lake Superior as well as the consideration of new
incoming diversions. In addition, efforts will likely be made to curtail the water diversion out of Lake Michigan
at Chicago. Presently, an informal agreement, "The Great Lakes Charter," between governors and premiers in
the Great Lakes region exists to forestall new diversions out of the Great Lakes basin (McAvoy, 1986). With
probable increased demands for water from outside the basin, that agreement will require greater authority to
remain effective.
4-30
-------
Croley
REFERENCES
Allsopp, T. R., and S. J. Cohen. "CO2-Induced Climate Change and Its Potential Impact on the Province of
Ontario." In: Preprint Volume. Conference on C|fanflte and Water Management - A Critical Era. American
Meteorological Society, Boston, Massachusetts, 1986. pp. 285-290.
Assel, R. A. "A Computerized Ice Concentration Data Base for the Great Lakes." NOAA Data Report ERL
GLERL-24, Natl. Tech. Inf. Ser., Springfield, Va. 22161, 1983.
Atmospheric Enviroment Service. "Monthly and Annual Evaporation from the Great Lakes Bordering Canada"
Environ. Canada, Atmos. Environ. Serv., Downsview, Ont, 1988.
Bolsenga, S. J. "Lake-Land Precipitation Relationships Using Northern Lake Michigan Data." J. ADD!. Met..
16(11):1158-1164,1977.
Bolsenga, S J. "Determining Overwater Precipitation from Overland Data: The Methodological Controversy
Analyzed." J. Great Lakes Res, 5(3- 4*301-311.1979.
Bolsenga, S. J., and J. C. Hagman. "On the Selection of Representative Stations for Thiessen Polygon Networks
to Estimate Lake Ontario Overwater Precipitation." International Field Year for the Great Lakes Bulletin 16.
National Oceanic and Atmospheric Administration, Rockville, Maryland, 1975. pp. 57-62.
Cohen, S. J. "Impacts of CO2-Induced Climatic Change on Water Resources in the Great Lakes Basin."
Climatic Change. 8:135-153,1986.
Cohen, SJ. "Sensitivity of Water Resources in the Great Lakes Region to Changes in Temperature,
Precipitation, Humidity, and Wind Speed." In: Proceeding^, The Influence of CJjtnqte Change ^yl (Climatic
Variability on the Hvdrologic Regime and Water Resources. International Association of Hydrologic Sciences
publication no. 168,1987. pp. 489-499.
Croley, T. E., n. "Great Lakes Basins Runoff Modeling." NOAA Tech. Memo. ERL GLERL-39. Natl. Tech.
Inf. Ser., Springfield, Va. 22161,1981 % pp.
Croley, T.E., II. "Great Lakes Basins (U^A.-Canada) Runoff Modeling." J. Hvdrol. 64:135-158,1983a
Croley, T.E., II. "Lake Ontario Basin (U.S A.-Canada) Runoff Modeling. J. Hvdrol.. 66:101-121,1983b.
Croley, TJB., II. "Verifiable Evaporation Modeling on the Laurentian Great Lakes." NOAA Tech. Memo. ERL
GLERL-XX. Environmental Research Laboratories, Boulder, Colorado, 1988. (in prep.)
Croley, T. E., D, and Hartmann, H. C. "Lake Superior Basin Runoff Modeling." NOAA Tech. Memo. ERL
GLERL-SQ. NatL Tech. Inf. Ser.. Springfield Va. 22161,1984. 284pp.
Croley, T.E., n, and Hartmann, H.C. "Lake Champlain Water Supply Forecasting." GLERL Open File Report,
Contribution No. 450, Great Lakes Environmental Research Laboratory, Ann Arbor, Michigan, 1985a. 56 pp.
Croley, T.E., H, and Hartmann, H.C. "Resolving Thiessen Polygons." J. Hvdrol.. 76:363-379,1985b.
Croley, T.E, II, and Hartmann, H.C. "Near-Real-Time Forecasting of Large-Lake Water Supplies; A User's
Manual" NOAA Tech. Memo. ERL GLERL-61. Environmental Research Laboratories, Boulder, Colorado,
1986. 82pp.
4-31
-------
Croley
Croley, T.E., II, and Hartmann, H.C. "Near Real-Time Forecasting of Large Lake Supplies." J. Wat. Res. Plan.
Manag. Div.. ASCE, 113(6):810-823, 1987.
DeCooke, B. G., and D. F. Witherspoon. Terrestrial Water Balance.* la- IFGYL - The International Field
Year for the Great Lakes. EJ. Aubert and T.L. Richards, eds. Great Lakes Environmental Research
Laboratory, Ann Arbor, Michigan, 1981. pp. 210*219.
Derecki, J. A. "Multiple Estimates of Lake Erie Evaporation." J. Great Lakes Res.. 2(1). 124-149,1976.
Derecki, J A. "Evaporation from Lake St. Clair." NOAA Tech. Memo. ERL GLERL-23. NatL Tech. Inf. Sen,
Springfield, Va. 22161,1978.
Derecki, JA. "Estimates of Lake St. Clair Evaporation." J. Great Lakes Res.. 5(2). 216-220, 1979.
Derecki, JA. "Evaporation from Lake Superior." NOAA Tech. Memo. ERL GLERL-29. Environmental
Research Laboratories, Boulder, Colorado, 1980.
Derecki, JA. "Stability Effects on Great Lakes Evaporation." J. Great Lakes Res.. 7(4). 357-362, 1981a
Derecki, J A. "Operational Estimates of Lake Superior Evaporation Based on IFYGL Findings." Water Resour.
Res.. 17(5). 1453-1462,1981b.
Gill, A. E., and J. S. Turner. "A Comparison of Seasonal Thermodine Models with Observation." Deep-Sea
Res.. 23. 391-401, 1976.
Goodyear, C. D., T. A. Edsall, D. M. Dempsey, G. D. Moss, and P. E. Polanskl "Atlas of the Spawning and
Nursery Areas of Great Lakes Fishes, Volume 13: Reproductive Characteristics of Great Lakes Fishes." Report
No. FWS/OBS-82/52, Fish and Wildlife Service, Washington, D.C., 1981
Gray, D. M., G. A. McKay, and J. M. Wigham. "Energy, Evaporation, and Evapotranspiration." Handbook on
the Principles of Hydrology. D. M. Gray, ed. Water Information Center, New York, 1973. pp. 3.1-3.66.
Gray, D.M., GA. McKay, and J.M. Wigham. "Potential Variation of Great Lakes Levels: A Hydrologic
Response Analysis." NOAA Tech. Memo. ERL GLERL. Environmental Research Laboratories, Boulder,
Colorado, 1988. (in prep.)
Hartmann, H. C. "An Evaluation of Great Lakes Hydraulic Routing Models." NOAA Tech. Memo. ERL
GLERL-66. Environmental Research Laboratories, Boulder, Colorado, 1987.
Hutchinson, G. E. A Treatise on Limnology. Volume I. Part 1 - Geography and Phvsics of Lakes. John Wiley
and Sons, New York, New York, 1957.
International Great Lakes Diversions and Consumptives Uses Study Board. "Great Lakes Diversions and
Consumptive Uses." International Joint Commission, Washington, D.C., 1981.
International Joint Commission. "Rules of Procedure and Text of Treaty." International Joint Commission,
Washington, D.C, 1965.
International Lake Superior Board of Control. Regulation of Lake Superior: Plan 1977 Development.
Description, and Testing. International Joint Commission, Washington, D.G, 1981.
International Lake Superior Board of Control Regulation of Lake Superior; Operational Guides for Plan 1977.
International Joint Commission, Washington, D.C, 1982.
4-32
-------
Croley
International St Lawrence River Board of Control Regulation of Lake Ontario: Plan 1958-D. International
Joint Commission, Washington, D.C., 1963.
Irbe, J. G., and A. Saulesleja. "An Operational Program for Monitoring Surface Temperatures of Lakes and
Coastal-Zone Waters in Canada from Polar-Orbiting Satellite Infrared Data." Actes du symposium internation-
al de la Commission VII de la Societe Internationale de photogrammetrie et teledetection, 13-17 September,
Toulouse, France, International Archives of Int. Soc. Photoprapy flnd Rem. Sens. 24(Vn-l):717-724.1982.
Irbe, J. G., R. K. Cross, and A. Saulesleja "Remote Sensing of Surface Water Temperature of the Great Lakes
and Off the Canadian East Coast." Northwest Atlantic Fisheries Organization Scientific Council Studies No. 4,
Special Session on Remote Sensing, September, Dartmouth, Canada, 1982. pp. 31-39.
Keijman, J. Q. "The Estimation of the Energy Balance of a Lake from Simple Weather Data." Boundarv-Laver
Meteorol. 7:399-407. 1974.
Kraus, E. B., and J. S. Turner. "A One-Dimensional Model of the Seasonal Thermocline II; The General Theory
and Its Consequences." Tellus. 19:98-105.1967.
Manny, B. A. "Potential Impacts of Water Diversions on Fishery Resources in the Great Lakes." Fisheries
9(5):19-23,1984.
Marchand, D., M. Sanderson, D. Howe, and C. Alpaugh. "Climatic Change and Great Lakes Levels: The Impact
on Shipping." Climatic Change. 1988. (in press)
McAvoy, P. V. "The Great Lakes Charter: Toward a Basin-Wide Strategy for Managing the Great Lakes." In:
Proceedings of the Great Lakes Legal Seminar: Diversion and Consumptive Use. The Center for the Great
Lakes, Chicago, 1986.
Meisner, J. D., J. L. Goodier, H. A, Regier, B. J. Shuter, and W. J. Christie. "An Assessment of the Effects of
Climate Wanning on Great Lakes Basin Fishes." J. Great Lakes Res. 13(3):340-352,1987.
Phillips, D. W. "Evaluation of Evaporation from Lake Ontario During IFYGL by a Modified Mass Transfer
Equation." Water Resour. Res.. 14(2). 197- 205,1978.
Phillips, D. W., and J. G. Irbe. "Land-To-Lake Comparison of Wind, Temperature, and Humidity on Lake
Ontario During the International Field Year for the Great Lakes (IFYGL)." Rep. CLI-2-77. Environ. Canada,
Atmos. Environ. Serv., Downsview, Ont, 1978.
Quinn, F. H. "Hydrologic Response Model of the North American Great Lakes." J. Hvdrol.. 37:295-307,1978.
Quinn, F.H. "An Improved Aerodynamic Evaporation Technique for Large Lakes with Application to the
International Field Year for the Great Lakes." Water Resour. Res.. 15(4). 935-940,1979.
Quinn, FJi. "Likely Effects of Climate Changes on Water Levels in the Great Lakes." In: Proceedings. Fust
Nnrtfr flpifriran Conference on Preparing for Climate Change. Climate Change Institute, Washington, D.C.,
1988. pp. 481-487.
Quinn, F. Hn and T. E. Croley n. "Climatic Water Balance Models for Great Lakes Forecasting and
Simulation." In: Preprint Volume: Fifth Conference on Hvdrometeorology. American Meteorological Society,
Boston, Massachusetts, 1983. pp. 218-223.
Quinn, F.H, and R.K-Kelley. "Great Lakes Monthly Hydrologic Data." NOAA Data Report ERL GLERL-26.
Nad. Tech. Inf. Ser., Springfield, Va. 22161,1983.
4-33
-------
Croley
Richards, T. L, and J. G. Irbe. "Estimates of Monthly Evaporation Losses from the Great Lakes, 1950 to 1968,
Based on the Mass Transfer Technique." paper presented at the 12th Conference on Great Lakes Research,
International Association of Great Lakes Research, Ann Arbor, Michigan, May, 1969.
Smit, B. "Implications of climatic change for agriculture in Ontario." fflpate Change Digest CCD 87-02.
Environ. Canada, Atmos. Enviroa Serv., Donwsview, Ont, 1987.
VS. Environmental Protection Agency. Potential Qjp?t't' Impacts of Increasing Atmospheric CO^ with
Emphasis on Water Availability and Hydrology in the United States. EPA Office of Policy, Planning, and
Evaluation, Washington, D.C., 1984.
U.S. Geological Survey. "Water Loss Investigations, VoL 1, Lake Hefner Studies. Geol. Surv. Prof. Pan. 269
IL&1954.
UJS. Geological Survey. "Water Loss Investigations: Lake Mead Studies." Geol. Surv. Prof. Pap. 298 U.S.. 1958.
Wall, G. "Implications of climatic change for tourism and recreation in Ontario." C/limqte Change Digest CCD
88-05. Environ. Canada, Atmos. Enviroa Serv., Downsview, Ont., 1988.
Wilson, J. W. "Effect of Lake Ontario on Predpitatioa* Monthly Weather Review 105:207-214,1977.
4-34
-------
IMPACT OF GLOBAL WARMING ON GREAT LAKES ICE CYCLES
by
Raymond A. Assel
National Oceanic and Atmospheric Administration
Great Lakes Environmental Research Laboratory
2205 Commonwealth Blvd.
Ann Arbor, MI 48105-1593
(GLERL Contribution No. 620)
Interagency Agreement Identification Number DW13932631-01-0
-------
CONTENTS
Page
FINDINGS 5-1
CHAPTER 1: INTRODUCTION 5-2
GREAT LAKES ICE COVER AND ICE CYCLES - A BRIEF REVIEW 5-2
ICE CYCLE MODELS 5-2
CHAPTER 2: METHODOLOGY 5-4
MODEL DEVELOPMENT AND ADJUSTMENTS 5-4
MODEL LIMITATIONS 5-7
SCENARIOS USED 5-8
CHAPTER 3: RESULTS 5-11
WINTERS WITHOUT ICE COVER 5-11
DATES OF FIRST/LAST ICE COVER AND ICE COVER DURATION 5-11
The 1951-80 Base Period 5-11
Doubled CO, Scenarios 5-11
The Transient Scenario 5-11
The 1930-39 Analog Scenario 5-12
DAILY AVERAGED BASIN MEAN ICE CONCENTRATION 5-12
CHAPTER 4: INTERPRETATION AND LIMITATIONS OF RESULTS 5-24
INTERPRETATION 5-24
LIMITATIONS 5-24
CHAPTER 5: IMPLICATIONS OF RESULTS 5-26
ENVIRONMENTAL IMPLICATIONS 5-26
SOCIO-ECONOMIC IMPLICATIONS 5-26
CHAPTER 6: POLICY IMPLICATIONS 5-27
GLOSSARY 5-28
REFERENCES 5-29
u
-------
Assel
FINDINGS1
Lake basin mean depth is an index of thermal inertia, and freezing degree-days (FDD) are an index for
energy loss at the water surface. Daily basin mean FDD are correlated with daily basin mean ice concentration
during winter. Annual ice cover duration and daily ice concentration are simulated with FDD regression
equations for Lake Erie's West, Center, and East Basins and for Lake Superior's West, East, and Whitefish Bay
Basins. Models are calibrated over a 20-year (1960-79) period. Models are applied to an independent four-
winter period (1980-83) to evaluate simulation error. Standard errors on the independent winters ranged from
20 to 30 percent for annual maximum ice concentration, and standard errors were 2 to 4 weeks in magnitude for
annual ice cover duration.
Average monthly air temperature for doubled CO, scenario winters is above-freezing for Lake Erie
basins. Ice cover still forms during some doubled CO, winters because of a sufficient number of consecutive
days with below-freezing air temperatures. Ice concentration and duration trends followed the air temperature
trends of the three doubled CO, scenarios; that is, the coldest scenario (Oregon State University) had the
greatest ice concentration with the longest ice cover duration, and the warmest scenario (Geophysical Fluid
Dynamics Laboratory) had the smallest ice concentration with the shortest ice season duration. Daily average
ice cover under all of the doubled CO, scenarios is limited to the shore area and shallows of each lake basin.
Doubled CO, scenario winters without ice cover occur 37 to 83 percent of the time in the East and Central
Basins of Lake Erie and up to 17 percent of the time for the West Basin of Lake Erie and the East Basin of
Lake Superior. Average winter duration for the 1951-80 base period was 13 to 16 weeks. Under the doubled
CO2 scenarios, the average winter duration is 5 to 13 weeks shorter. The shorter duration and less extensive ice
covers would affect lake ecology, and loss of some cold water fish species such as lake whitefish may occur. The
shipping season, which traditionally stops during the winter months, would likely be extended, perhaps to a year-
round season.
Under the 79-year transient CO, scenario (1981-2059), only Lake Erie basins have winters without ice
cover. During the last three decades of the transient (2030-59) 30 to 80 percent of the winters for Lake Erie's
Center and East Basins are without ice cover. Transient scenario daily ice concentration, averaged for the years
2010-39, is significantly less than the base period for all lake basins. However, extensive ice cover will occur
under many transient CO, winters, particularly during the first 29 years (1981-2009). Average ice cover duration
is 3 to 7 weeks shorter during the next 30 years of the transient 2010-39 relative to the 1951-80 base period.
During the last decade of the transient scenario (2050-59), decadal-averaged ice concentration and ice cover
duration is similar to the doubled CO2 scenarios.
Ice cover simulation of an analog climatic wanning period (1930-39) shows that average ice cover duration
was 1/2 to 1 week shorter than the 1951-80 base period for Lake Superior and about 3 to 4 weeks shorter for
the Center and East Lake Erie Basins. Annual maximum ice concentration was less than the base period but
greater than the doubled CO2 scenario.
'Although the information in this report has been funded partly by the UJS. Environmental Protection
Agency under contract DW13932631-01-0, it does not necessarily reflect the Agency's views, and no official
endorsement should be inferred from it.
5-1
-------
Assel
CHAPTER 1
INTRODUCTION
GREAT LAKES ICE COVER AND ICE CYCLES - A BRIEF REVIEW
The Great Lakes ice cycle is divided into three periods: fall cooling, ice formation, and ice loss. During fall
cooling, thermal stratification is lost as the entire water column cools to the temperature of maximum density
near 4 degrees C. Subsequent cooling results in the formation of less dense surface water, winter restratification,
and the start of the ice formation period. Ice loss occurs in spring owing to increasing solar radiation and above-
freezing air temperature.
During the winter ice formation period, vertical ice accretion occurs as a result of heat transfer from the ice-
water interface through the ice to the atmosphere. As the ice cover thickens its vertical extent reduces the rate
of heat loss from the ice-water boundary and retards additional ice accretion. Under the current climate regime,
the upper limit of thermodynamic ice growth appears to be about 50 to 100 cm for bay and harbor sites in the
Great Lakes (Sleator, 1978). Local factors such as air temperature, water depth, winds, and snowfall are
responsible for the variation of ice thickness. Winds can cause portions of an ice cover to compact and override
or submerge under the remaining ice cover; the result is rafted, ridged, or jammed ice, depending upon the
amount and vertical extent of ice rubble formed. The U.S. Coast Guard has reported wind-induced ice thickness
of up to 8 m in the Great Lakes.
The Great Lakes' ice cover forms in the shallow bays and harbors in December and in deeper bays and
along the perimeter of the Great Lakes during January. Lake Erie, with its shallow depth, forms midlake ice
cover in January; midlake areas of the other Great Lakes usually form extensive ice cover in February. Annual
maximum ice coverage occurs in February and early March, but even then, some areas tend to remain open
water. Normal maximum ice cover expressed as a percentage of total surface area is 90% for Lake Erie, 75%
for Lake Superior, 68% for Lake Huron, 45% for Lake Michigan, and only 24% for Lake Ontario (Assel et al.,
1983). Lake Ontario's small annual maximum ice coverage results from the combination of (1) its large thermal
inertia (mean depth of 86 m), (2) average water surface temperatures during winter near but above 0 degrees
C, which make air-water temperature differences primarily a function of air temperatures, and (3) its relatively
mild winter air temperatures (-4.4 degrees C, compared with -9.8 degrees C for Lake Superior). Air
temperature is the single most important atmospheric climate variable affecting ice cover. The average winter
temperature for all five Great Lakes in 1979 was -6.8 degrees C, and the annual maximum ice extent was nearly
100%. The average winter temperature for 1983 was -2.2 degrees C and the maximum ice extent was
approximately 23% of the combined Great Lakes surface area (Assel et aL, 1985).
In spring, ice loss results from melting caused by solar radiation and above-freezing air temperature. Solar
radiation penetrates and is absorbed within the ice, reducing the structural strength of the ice due to preferential
melting at ice crystal boundaries. The weakened ice cover can then be easily broken by winds and wave action,
and melted or transported to windward (eastern) lake shores. In March, areas of open water and low ice
concentration expand from the deeper, more exposed midlake areas toward the perimeter and eastern shores.
By mid-April, any remaining ice is usually located in the shore zone; however, during some years, ice cover lasts
into May.
ICE CYCLE MODELS
Ice cover models that lend themselves most easily to climate analysis are empirical and statistical in nature
because of the availability of input data needed to calibrate and evaluate them. For this reason and because of
a need to complete this analysis in a timely manner, only relatively simple empirical statistical models using
cumulative air temperature in the form of degree-days are considered in this paper.
5-2
-------
Assel
Freeze-up dates on shallow inland lakes in Canada were found to correlate with weighted mean daily air
temperatures (Bilello, 1964) and accumulated freezing degree-days (FDD) (Williams, 1965). Later Williams
(1971) also correlated breakup dates with previous dates of breakup and air temperatures. And Tramoni et al.
(1985) and Barry (1986) found that freeze-up on shallow inland lakes hi Canada and Finland are an index of fall
air temperatures; an increase of 1 degree C in average fall air temperature corresponded with a 3- to 10-day
delay in date of freeze-up. One of the earliest empirical ice cover studies on the Great Lakes was made by Oak
and Myers (1953); they used February air temperature to forecast spring opening dates for navigation at various
bays and harbors. Systematic aerial ice reconnaissance observations of Great Lakes ice cover began in the late
1950's; this established a data base for making empirical and statistical studies of large areas of the Great Lakes
ice cover. Richards (1963) correlated antecedent heating degree-days for the previous summer (index of summer
heating) and freezing degree-days for late fall and winter (index of water cooling and ice formation) with
observations of lake-averaged Great Lakes ice cover. Snider (1971) developed threshold FDD accumulations
for navigationally significant ice in specific areas of the Great Lakes, based on average heat storage in the water.
Rogers (1976) used FDD and thawing degree-days to develop regression models of annual maximum lake-
averaged ice concentration. More recently, Assel et al. (1985) developed a regression model of regional annual
maximum ice cover of the combined area of the five Great Lakes, based on regional average winter temperature.
Their model implies that a 4 to 5 degree C increase in regional average winter temperature from the 1951-80
base period average would result in regional annual maximum ice cover between 0 and 9% for the combined
area of the Great Lakes. Howe et al. (1986) developed empirical models for each lake relating monthly average
air temperature on the perimeter to annual maximum ice cover. They estimated the climatically "normal" annual
maximum ice cover based on 1951-80 mean air temperatures and the expected annual maximum ice cover for
a doubled CO- warming. Their results indicate that, except for Lake Erie, the expected annual maximum ice
cover under a doubled CO2 warming is nil.
The studies noted in the previous paragraphs developed models of different parts of the annual ice cycle.
hi this present study empirical models were developed that simulate daily mean basin ice concentration for the
entire ice cycle, using some of the methods from the earlier studies. This initial analysis is limited to Lakes Erie
and Superior - the two lakes that represent the extremes in mean lake depth and air temperatures for the Great
Lakes. A summary of development methodology and an assessment of model limitations are described. The
range and expected values of ice concentration and ice cycle duration are presented and briefly interpreted
relative to eavironmental and socio-economic implications.
5-3
-------
Assel
CHAPTER!
METHODOLOGY
MODEL DEVELOPMENT AND ADJUSTMENTS
Two important variables affecting ice formation are (1) beat stored in the water mass and (2) the energy
budget at the air-water boundary. Mean lake basin depth (volume divided by surface area ) is an index that
reflects thermal inertia, and accumulated freezing degree-days (FDD) are an index of energy loss from the
surface mixing layer. Lakes Superior and Erie are divided into basins using lake bathymetry: East, Center, and
West Basins for Lake Erie, and East and West Basins for Lake Superior (Figure 1 and Table 1). In addition,
Whitefish Bay is included as a separate basin of Lake Superior because of its importance to navigation.
Table 1. Lake Basin Parameters (approximate values)
Lake Superior
Mean Depth (m)
Area(km2)
Volume (km3)
Max FDD (C)#
WB
135
21971
2966
1299
EB
152
58947
8960
1255
WFB
41
1182
48
1255
Lake Erie
WB
9
5135
46
318
CB
19
14635
278
342
EB
27
5909
159
368
# Averaged annual maximum FDD for the 1951-80 base period.
Mean basin ice concentration is calculated for each lake basin from synoptic ice charts (Assel, 1983) for the
years 1960-79. These data are used to develop regression equations between synoptic mean basin ice
concentration and daily mean basin FDD accumulation through the date of the synoptic ice concentration
observation (Table 2). Equation 1 was developed for Lake Superior's East and West Basins, and Equation 2 for
Whitefish Bay and the three Lake Erie basin. Equations 3 and 4 differentiate between initial ice formation in
the shallow and deep areas of each Lake Superior basins. Equation 3 is used for initial ice formation in the
shallows of Whitefish Bay and, Equation 4 is used for initial ice formation in shallow areas of Lake Superior's
East and West Basins. Equation 5 is an ice reduction factor that is part of Equation 1 and Equation 2.
The three periods of the annual ice cycle are simulated by the system of equations in Table 2. Initial ice
formation is a function of threshold FDD accumulations given as part of the constraints in Equations 2,3, and
4. Threshold FDD values are estimated from analysis of historic ice charts and corresponding FDD
accumulations from November 1 to the date given on the ice chart. If these threshold FDD values are not
accumulated, ice cover will not form. Threshold FDD values for Lake Erie's West, Center, and East Basins are
27, 75, and 110 FDD, respectively. Initial ice extent at these threshold values is 45%, 37%, and 30%,
respectively. If less than 430 FDD accumulate for the East and West Basins of Lake Superior, ice cover will not
form. The threshold FDD value needed for initial ice formation on Whitefish Bay is 350.
Ice growth is simulated by the expressions (FDD-BFDD) in Equations 1 and 2. These expressions are an
index of heat loss at the air-water boundary and they also incorporate the hysteresis effect of antecedent FDD
accumulations on increasing ice extent. The hysteresis effect in Equation 1 (exponential term) is a function
of the number of days past the date of end-of-fall overturn because the maximum ice cover (which is usually
below 100%) is related to the number of days past the date of end-of-fall overturn on Lake Superior's East and
5-4
-------
Assel
RGURE 1
LAKE BASINS AND TEMPERATURE STATIONS
Thunder Bay
K i lomet• r »
Stall Ste. Marie
LAKE SUPERIOR
WEST BASIN
Toledo
> Cleveland
K i tome i»r»
LAKE ERIE
5-5
-------
Assel
Table 2. Annual Ice Cycle Simulation Model Equations
:========================»===========»=========
100
EO 1 IfE — ---™_—————.................................................—............
1 + Cl*exp[C2*((FDD-BFDD)/JD)J + C3*(FDD-BFDD) + THFAC
EQ. 1 constraints: 1. 730 < FDD < FDDCRIT
2. JD = 1 when FDD=BFDD and JD increments
by 1 every day after that date.
3. ICE-100 if FDD > FDDCRIT
100
EQ. 2 ICE
1 + FZFAC + Cl*exp[C2*(FDD-BFDD)] + THFAC
EQ. 2 constraints: 1. 450 < FDD (for Whitefish Bay only)
2. FZFAC-1000 when FDD < BFDD, FZFAC-0 when FDD > BFDD
3. ICE=0 if FDD < BFDD
EQ. 3 ICE * 5*(FDD-350)/100
EQ. 3 constraints: 1. 350 < FDD < 450
HCE-OifFDD <350
3. JD < DMFDD
EQ. 4a ICE - 5 + 10*(FDD-580)/150
EQ. 4a constraints: 1. 580 < FDD < 730
2, JD < DMFDD
EQ. 4b ICE - 5*(FDD-430)/150
EQ, 4b constraints: 1. 430 < FDD < 580
2. ICE«0 if FDD <430
3. JD < DMFDD
EQ. 5 THFAC
I(JD-DMFDD)/{Sqrt(MFDD-BFDD)/MELT}J
EQ. 5 constraints: 1. THFAC=0 if JD < DMFDD
2. THFAC-9999 if JD-DMFDD > a. or b. below.
a Sqrt[(MFDp-BFI)D)/MELT]
b. maximum historic observed value of days past
annual maximum FDD to date of last observed ice.
SKttVatV&SSB«S5SCSSSaBVaB3BXVK3B««SB:VS»SSB9BBS3BB«K3SV:SSMI
See glossary for definition of terms.
5-6
-------
Assel
West Basins. The hysteresis effect of antecedent FDD accumulation in Equation 2 is related to the FDD
accumulation necessary for virtually 100% ice coverage because the ice cover on Lake Erie basins and Lake
Superior's Whitefish Bay approaches 100% most winters. Basin ice concentration generally decreases by the
date FDD accumulations reach their annual maximum value - average air temperatures are usually above
freezing after that date. Consequently, the date of annual maximum FDD accumulation defines the end of the
ice growth period and the beginning of the ice decay period. A temporal ice concentration reduction factor,
Equation 5, is activated in Equations 1 and 2 after the date of annual maximum FDD accumulation. Ice loss
in Equation 5 is a function of average daily ice melt rate, average ice thickness on the date of maximum FDD
accumulation, and number of days past the date of maximum FDD accumulation. The average ice melt rate
(cm/day) during the ice loss period was optimized by trial and error during the regression analysis. The rate
is 1250 (cm/day) for Lake Superior's East and West Basins, and Whitefish Bay, and 0.889,0.667, and 0333, for
Lake Erie's West, Center, and East Basins, respectively. Lake-averaged ice ablation rates for bays and harbors
of Lakes Superior and Erie are 1.4 and 1.1 cm/day, respectively (Bolsenga, 1988). Lower ablation rates for Lake
Erie's Center and East Basins apparently reflect the eastward advection of ice into these basins during the ice
loss period, which in effect, prolongs the period of ice loss and reduces the average daily rate of ice ablation.
A Stefan ice growth expression (the square root of the accumulated FDD) was used to estimate ice thickness.
Complete ice loss occurs when the number of days past the date of annual maximum FDD exceeds the number
of days needed to melt the ice at the average melt rate. The largest observed number of days between the date
of annual maximum FDD accumulation and the date of ice cover loss was estimated from historic ice charts for
the Lake Erie basins, and this was made a constraint for Equations 5 and 2 for simulating ice loss hi Lake Erie
basins.
MODEL LIMITATIONS
Model prediction error for each basin was evaluated by the cross validation method; that is, the data were
divided hi half and model coefficients were generated for each subset of the original data. Coefficients for one
subset were used to simulate ice concentration for the second subset. The model prediction error for each data
subset ranged from 15 to 28% (Table 3).
Table 3. Model Cross Validation for Root Mean Square Error (RMSE)
RMSE
First half of data
Second half of data
Lake Superior
WB EB WFB
25.9
27.8
20.9
18.8
14.7
22.8
WB
17.8
22.1
Lake Erie
CB EB
24.9
27.0
19.7
18.1
Model prediction error over the entire data base was evaluated by simulation of annual maximum lake-
averaged ice concentration during four winters outside the model calibration period. Estimates of lake-averaged
annual maximum ice cover and date of occurrence for the early 1980s have been provided by the UJS. Coast
Guard (personal communication, United States Coast Guard, Ninth District Headquarters, Cleveland, Ohio).
Basin mean ice concentrations from simulation models for Lakes Superior and Erie were areally weighted and
summed to obtain the lake-averaged ice concentration for the dates of annual maximum ice concentration
provided by the Coast Guard. The four "test winters" contained both extremely high and extremely low annual
maximum ice concentrations (Table 4). Models did well for both winter extremes; standard errors are 30% for
Lake Superior and 20% for Lake Erie. It is significant that the models did well during the mild 1982-83 winter,
since that winter is ranked as the 10th warmest winter in the Great Lakes during the 200-year period 1783-1983
(Assel et aL, 1985X and thus it likely approaches conditions of some of the CO2 global wanning scenarios.
5-7
-------
Asset
Table 4. Simulation Error for Annual Maximum Ice Extent
Lake
Superior
Erie
Winter
Season
1979-80
1980-81
1981-82
1982-83
1979-80
1980-81
1981-82
1982-83
Maximum
Observed
75
92
97
21
95
100
100
25
Ice Concentration
Simulated
34
62
%
13
86
85
74
40
Error
(Obs-Sim)
41
31
1
8
9
15
26
15
Errors in simulated ice cover duration were also analyzed. Ice cover duration is defined here as the
difference in days between the dates of initial ice cover greater than 0.5% concentration and final and complete
loss of ice cover. Dates of first- and last-ice were estimated from Canadian composite ice charts and from
National Ocean Service water level gauge ice reports. Standard error in simulated ice cover duration ranged
from 3 to 4 weeks for Lake Superior basins and from 2 to 3-1/2 weeks for Lake Erie basins (Table 5).
Determining ice cover duration during mild winters can be difficult. The first and last dates of ice were used
to determine season duration, but in mild winters this method may over-estimate the actual number of days with
ice cover because of alternating periods of ice formation and ice cover loss. For example, an error of 40 days
for West Basin of Lake Erie during the mild 1982-83 winter resulted from intermittent ice cover during two days
in December. If those spurious data were omitted, the initial ice cover would have been on January 18th and
the season duration would have been 45 days, producing an error of only 3 days.
It is difficult to quantify the magnitude of error in ice cover models caused by not considering wind effects
because wind effects are highly nonlinear and depend upon physical processes that occur on time scales of a few
hours to a few days. Analysis of data from Lewis (1987) indicates that for the 28-year period 1957-1985, the
annual average number of storms with wind speeds of 88 km/hour or greater during the 3-month period
January, February, and March is about one severe storm per winter. If winter storms of this magnitude were
to increase in frequency during the climate warming scenarios, this would contribute to a reduction in both ice
duration and ice concentration; a reduction in storm frequency from the present climate average of one per
season would have the opposite effect. High winds during the early winter would tend to retard ice formation,
particularly in the deeper lake basins; thus, models would predict initial ice formation too early and ice extent
would be over-estimated. An extended period of calm conditions in early winter would have the opposite effect;
initial ice covers would form earlier and be more extensive. High winds during the winter after an extensive ice
cover exists could reduce ice covers temporarily, and models would over-estimate ice extent. High winds in
spring could move the ice out of a basin or melt it by upweUing, resulting in an over-estimated ice coverage
during part of the ice loss period and a tardy simulated date of last ice cover.
SCENARIOS USED
The 1951-80 dimatological period used to define the current normal value for climate elements was used
as a base period in this study. The decade of the 1930's had several years of much above normal temperatures
and is used as a historic analog that approaches CO, global warming scenarios. Cumulative departures from
long-term monthly mean temperature (1895-1977) for the contiguous United States show a increasing
temperature trend from 1921 to 1954 and a decreasing trend after that (Diaz and Quayle, 1980). Mild winters
5-8
-------
Assel
Table 5. Simulation Error for Ice Cycle Duration
Lake Superior
winter season
1979-80
1980-81
1981-82
1982-83
Lake Erie
winter season
1979-80
1980-81
1981-82
1982-83
Ice Cycle Duration in Day
West Basin East Basin
obs sim error obs sim error
105 87 18
134 101 33
IB 119 4
84 114 -30
West Basin
obs sim error
118 106 12
79 81 -3
95 107 -12
42 82 -40
108 88 20
125 103 22
139 118 21
91 92 -1
Center Basin
obs sim error
69 72 -3
94 80 14
92 89 3
37 16 21
Whitefish Bay
obs sim error
108 101 7
129 110 19
150 121 29
83 97 -14
East Basin
obs sim error
76 86 10
121 103 18
113 111 2
28 17 11
obs - observation; sim - simulation
occurred in the Great Lakes during the early 1950's but not during the rest of the 1951-80 base period when
severe winters were prevalent (Assel, 1986).
Monthly mean air temperatures ratios (doubled CO2/1CO2) for the GISS, GFDL, and OSU scenarios and
for the GISS-A scenario were multiplied by daily mean air temperatures for stations on the perimeter of each
basin for the 1951-80 base period data (Table 6). The closest grid point for a given global circulation model
(GISS, GFDL, OSU) was determined for each meteorological station (Table 6), and the air temperature ratios
at that grid were used for simulating the doubled CO2 and transient CO2 temperature time series. The daily
observed station temperatures were converted to degrees Kelvin, multiplied by the appropriate ratio (year and
month) for a given scenario, and then converted back to degrees C, following EPA instructions. Mean daily air
temperatures for the 1930-39 historic analog scenario and the 1951-80 base period were abstracted from a Great
Lakes Environmental Research Laboratory air temperature data base (Assel, 1986) for analysis. Mean basin
daily air temperatures and FDD accumulations were calculated and the daily FDD data were used to drive the
ice cycle simulation models. Winter seasons were started in November prior to the year shown as the first year
of the scenario. For example, the 1951-80 base period includes air temperatures and FDD for November and
December, 1950. The GISS-A scenario began November 1980; the average ice cover statistics for the first
decade of these scenarios was only nine winters, 1981-89. Thirty-year monthly averaged temperatures for the
1951-80 base period, the doubled CO, scenarios, and decadal monthly averages for the 1930-39 analog and
transient CO2 scenario are given in Table 7 for East and West Basin of Lake Erie and East and West Basins
of Lake Superior. Monthly air temperatures for the doubled CO, scenarios are above 0 degrees C for Lake
Erie, but only during November and occasionally during March for Lake Superior. This is somewhat misleading
since daily average temperatures are occasionally below 0 degrees C; consequently, global circulation model
output statistics should concern a daily rather than a monthly time period. Daily values may produce
significantly different results relative to ice cover simulation. The transient decadal average monthly
temperatures for the first three decades are in general similar to the 1951-80 base period. November-through-
March averaged temperatures for the last decade of the transient is similar to GISS doubled CO, scenario
averaged temperatures. The 1930-39 analog temperatures are less then 2 degrees C warmer than the 1951-80
base period in most of the winter months.
5-9
-------
Assel
Table 6. Air Temperature Stations used in FDD Analysis
Basins Temperature Stations
Eri WB Detroit, MI., Toledo, OH., Cleveland, OH.
Eri CB Cleveland, OH., Port Dover, ONT.
Eri EB Port Dover, ONT., Erie, PA., Buffalo, NY.
Sup WB Duluth, MR, Houghton, MI., Thunder Bay, ONT.
Sup EB Sault St. Marie, MI., Thunder Bay ONT.
Sup WFB Sault St. Marie, MI., Thunder Bay, ONT.
Table 7. Monthly Average Temperatures (°C) for Base Case, Doubled COy and Decadal
Average for the Analog and Transient CO2 Scenarios
Lake Erie - West Basin
Nov. Dec. Jan. Feb. Mar. Nov.
-3.0 2.1 4.6
13 6.9 8.8
1.7 7.1 9.6
0.9 63 6.4
-1.9 2.0 4.7
-1.0 23 4.6
-33 1.8 6.1
-18 33 5.9
0.9 3.7 6.6
-0.8 4.0 6.9
-0.1 55 72
10 6.4 8.5
23 63 10.7
Lake Erie - East Basin
Dec. Jan. Feb. Mar.
1951 - 80
GISS 2CO2
GFDL 2CO2
OSU 2CO2
1930 - 39
1981-89
1990-99
2000-09
2010 - 19
2020-29
2030-39
2040-49
2050-59
4.9
9.1
9.8
6.7
53
4.6
6.4
6.4
6.7
72
7.7
8.5
11.0
-13
5.5
5.2
1.6
-03
-0.8
-L2
0.4
23
0.4
2.4
3.6
3.8
-42
2.0
15
13
-12
-15
-3.4
-4.9
-2.0
-18
-16
-03
12
Lake Superior - West Basin
Nov. Dec Jan. Feb. Mar.
1951 - 80
GISS 2CO2
GFDL 2CO2
OSU 2CO2
1930 - 39
1981-89
1990-99
2000-09
2010 - 19
2020-29
2030-39
2040-49
2050-59
-1.9
4.1
4.4
0.5
-13
-3.9
-12
-1.6
-L8
-1.0
-0.6
0.1
32
-9.4
-4.0
-1.9
-6.7
-7.6
-93
-8.9
-8.8
-6.7
•65
-65
-5.4
-4.0
-1.4
53
5.0
12
-1.0
-0.9
-12
0.0
1.9
0.4
10
32
3.8
1.6
11
0.7
-12
-11
-3.7
-5.1
-16
•32
-18
-0.9
0.8
-4.1
02
0.6
-0.2
-33
-11
-4.5
-3.9
-0.2
-10
-12
1.0
1.1
0.6
5.4
5.6
4.7
0.4
0.8
03
13
23
15
3.6
5.0
4.8
Lake Superior - East Basin
Nov. Dec Jan. Feb. Mar.
-133 -113 -52 -1.1 -8.9
-8.1 -52 -0.5 3.9 -3.0
-6.9 -4.1 1.9 5.0 -1.6
-8.4 -7.4 -12 13 -62
-11.0 -10.4 -4.9 -1.1 -7.4
-115 -9.6 -4.4 -10 -8.9
-114 -112 -4.4 -03 -83
-13.7 -9.8 -4.1 -0.4 -82
-10.8 -8.4 -4.5 0.0 -63
-10.8 -95 -18 0.5 -63
-10.7 -8.2 -1.0 0.7 -6.1
-8.6 -6.6 -0.8 L8 -4.7
-7.4 -7.1 -03 4.4 -3.4
-119
-7.4
-5.8
•S2
-10.7
-11.6
-111
-132
-10.7
-11.0
-10.4
-8.5
-7.1
-11.4 -5.2
-6.1 05
-4.1 1.8
-7.4 -12
-11.0 -5.6
-9.5 -4.4
-11.6 -4.6
-10.5 -4.4
-7.9 -4.0
-9.4
-8.6
-6.5
-6.7
-2.8
-1.6
-0.7
-0.7
5-10
-------
Assel
CHAPTERS
RESULTS
Daily mean basin ice concentration was simulated for the 30 winter seasons of the 1951-80 base period, the
30 winters of each doubled CO, scenario, the 79 winters of the transient CO, scenario, and for the winters of
the 1930-39 analog scenario. These data and monthly statistics for all scenarios (average, median, maximum,
minimum, and standard deviation) and decadal averages of monthly statistics for the transient CO, scenario are
available at the Great Lakes Environmental Research Laboratory.
WINTERS WITHOUT ICE COVER
The whiter of 19S2-S3 was extremely mild, and ice cover did not form hi Lake Erie's East and Central
Basins (International Niagara Working Committee, 1983). The ice cycle models for these basins accurately
simulated the lack of ice cover that whiter. This was the only winter during the base period that Lake Superior
or Lake Erie lacked ice cover. Under the doubled CO2 scenarios up to 7% of the waiters for Lake Superior
basins and up to 17% of the winters for the West Basin of Lake Erie lack ice. From 37 to 83% of the winters
for Center and East Basins of Lake Erie also are without ice cover (Table 8). Under the transient scenario, only
Lake Erie basins have winters without ice cover. During the first five decades, no more than 20% of the winters
for Lake Erie basins are without ice cover. Because of the greater mean depth of the Center and East Basins
of Lake Erie (compared to the West Basin), they have more winters without ice cover hi the latter decades of
the transient and for the doubled CO, scenarios. During the last three decades of the transient, 30 to 80% of
the winters for Central and East Basins of Lake Erie (but no more than 10% of winters for West Basin) are
without ice cover. During the 1930-39 analog decade, only the East Basin of Lake Erie had winters without ice
cover, and then, for only 20% of the tune.
DATES OF FIRST/LAST ICE COVER AND ICE COVER DURATION
The 1951-80 Base Period
Ice covers began forming hi shore areas of Lake Superior during the first half of January and were lost near
the end of April The average annual duration of the ice cycle on Lake Superior basins was 15 to 16 weeks.
Initial ice formation on Lake Erie occurred hi the shallow West Basin the third week of December and was
usually lost during the thud week in March. On average, the West Basin of Lake Erie had ice cover for about
13 weeks. In Lake Erie's Central and East Basins, first-ice occurred early hi January and it was completely lost
by late March (Center Basin) or mid-April (East Basin). Average ice cover duration was about 12 weeks for
the Center Basin and dose to 14 weeks for the East Basin of Lake Erie.
Doubled CO2 Scenarios
The average dates of first- and last-ice were based only on winters with ice cover, and all 30 winters were
used hi calculating the average annual duration of ice cover. Under the doubled CO2 scenarios, Lake Superior
ice formation starts 2-1/2 to 6-1/2 weeks later and ends 2-1/2 to 6 weeks earlier than it did hi the 1951-80 base
period; Lake Erie ice formation starts 3 to 4-1/2 weeks later and ends 4 to 6 weeks earlier than the 1951-80 base
period (Table 9). The average duration of ice cover is 5 to 13 weeks shorter for Lake Superior and 8 to 13
weeks shorter for Lake Erie.
The Transient Scenario
GISS-A decade averages show a general trend of later first-ice and earlier last-ice dates. Because of the
built-in bias of lower temperatures in the 1951-80 base temperature data used to construct the transient scenario
5-11
-------
Assel
Table 8. Percentage of Winters Without Ice Cover
Lake Superior Lake Erie
Scenarios WB EB WFB WB CB EB
1CO2 1951-80
2CO2 GISS
2CO2 GFDL
2CO, OSU
GISS-A 1981-89
GISS-A 1990-99
GISS-A 20(XM)9
GISS-A 2010-19
GISS-A 2020-29
GISS-A 2030-39
GISS-A 2040-49
GISS-A 2050-59
Analog 1930-39
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7
0
0
0
0
0
0
0
0
0
0
0
0
7
0
0
0
0
0
0
0
0
0
0
0
10
17
7
0
0
0
10
0
0
10
10
0
3
67
73
37
10
0
0
10
0
30
40
70
0
3
80
83
60
10
0
0
20
0
30
70
80
20
temperatures, dates of first-ice are earlier and dates of last-ice are later than the 1951-80 base period in some
of the early decades of GISS-A. A 29-year (1981-2009) average and a 30-year (2010-39) average for dates of
first- and last-ice and ice cover duration were calculated to filter out the effects of the bias of lower temperatures
in the 1951-80 base period (Table 9). Average dates of first-ice, last-ice, and ice cover duration for 1981-2009
are virtually the same as the 1951-80 base period, with the exception of the Center and East Basins of Lake Erie.
Average date of first-ice for 2010-39 is 1-1/2 to 2 weeks later and average date of last-ice is 2 to 4 weeks earlier
than during the 1951-80 base period. Average ice cover duration for 2010-2039 is 3 to 4 weeks shorter for Lake
Superior basins and 5-1/2 to 7-1/2 weeks shorter for Lake Erie basins. The decadal average ice cover duration
for the last two decades of the transient are 6-1/2 to 9 weeks shorter for Lake Superior basins and 9 to 13 weeks
shorter for Lake Erie basins, compared to the 1951-80 base period (Table 9). Average ice cover duration during
the 2040-49 decade is similar to the doubled CO2 OSU scenario, and average ice cover duration for 2050-59 is
similar to the doubled CO2 GISS scenario.
The 1930-39 Analog Scenario
The 1930-39 analog has a shift to both later first- and later last-ice dates for Lake Superior basins and the
West Basin of Lake Erie, relative to the 1951-80 base period. Average ice cover duration was 1/2 to 1 week
shorter than the base period for Lake Superior basins and for the West Basin of Lake Erie. Analog decadal
average ice duration was 3 to 4 weeks shorter then the base period for the Center and East Basins of Lake Erie.
DAILY AVERAGED BASIN MEAN ICE CONCENTRATION
Daily averaged ice concentration was calculated for the 30-year base period, for each of the doubled CO2
scenarios, and for consecutive 29-year and 30-year non-overlapping periods of the GISS-A transient scenario
(that is, 1981-2009 and 2010-39). The standard deviation of the base period daily average ice concentration was
also calculated as an estimate of ice cover variability under the current climate regime. Selected results are
shown in Figures 2 through 10. The daily averaged ice concentration portrayed in these figures was set to zero
before the average date of first-ice and after the average date of last-ice (Table 9). This modification was done
to be consistent with the dates given in Table 9 and does not significantly alter any of the findings.
A t-test (Brownlee, 1967) at the 99% probability level was performed between the daily averaged ice
concentration of the base period and each of the two non-overlapping periods of the transient, and for each of
5-12
-------
Assel
>led CO2 scenarios. During the 1981-2009 period, average daily ice cover was significantly different than
period during late March and early April for Lake Superior basins and during the last three weeks of
the doubled CO,
the base
February for the Center Basin of Lake Erie. The daily average ice cover for the period 2010-39 was significantly
different than the base period daily average most of the winter for all lake basins. The t-test analysis for the
doubled CO2 scenarios showed significant differences between the daily average ice concentration of the 1951-80
base period and the doubled CO2 scenarios. It is probable ice cover will be restricted to shoal areas and to the
shore zone of each lake basin during an average doubled CO2 winter (Figures 5-10). However, it will still be
possible, although not very likely, to have extensive ice cover formation during some doubled CO, scenario
winters. Extensive ice cover will form during many of the transient scenario winters, particularly during the
1981-2009 year period of the transient (Figures 2-4).
Table 9. Average First-Ice/Last-Ice Dates and Ice Duration
Lake Basin
SupWB
SupEB
SupWFB
EriWB
EriCB
EriEB
1951-80
1CO2
Jan 6
Jan 9
Jan 2
Dec 17
Jan 6
Jan 12
Average First D
GISS
Feb 14
Feb 18
Feb 6
Jan 18
Feb 4
Feb 3
late of Ice
led CO2 -
GFDL
Feb 20
Feb 22
Feb 15
Jan 17
Feb 4
Feb 7
Average Last Date of Ice
Lake Basin
SupWB
SupEB
SupWFB
EriWB
EriCB
EriEB
Lake Basin
SupWB
SupEB
SupWFB
EriWB
EriCB
EriEB
1951-80
1CO2
Apr 27
Apr 26
Apr 26
Mar 19
Mar 29
Apr 18
1951-80
1CO2
112
108
115
93
83
97
..-.-...-»-»pouo
GISS
Mar 31
Apr 1
Apr 1
Feb 17
Feb 28
Mar 2
Average Ice
led CO2 -
GFDL
Mar 15
Mar 14
Mar 14
Feb 13
Feb 28
Mar 7
Duration
Cover *
OSU
Jan 25
Jan 30
Jan 19
Jan 7
Jan 27
Feb 2
Cover •
OSU
Apr 8
Apr 8
Apr 8
Feb 14
Feb 28
Mar 7
(Days) **
Doubled CO,
GISS
46
43
55
26
8
6
GFDL
24
19
26
23
6
5
OSU
75
69
80
35
19
13
1981-2009
GISS-A
Jan 5
Jan 10
Jan 4
Dec 19
Jan 10
Jan 15
1981-2009
GISS-A
Apr 24
Apr 22
Apr 22
Mar 13
Mar 24
Apr 9
1981-2009
GISS-A
108
103
109
84
71
82
ra.
GISS-A
Jan 16
Jan 21
Jan 12
Dec 31
Jan 16
Jan 26
2010-39
GISS-A
Apr 14
Apr 13
Apr 13
Feb 25
Mar 4
Mar 18
2010-39
GISS-A
88
84
92
54
41
43
5-13
-------
Assel
Table 9. (continued)
Decadal average ice cover duration for transient and analog
Lake
Basin
Sup
Sup
Sup
Eri
Eri
Eri
WB
EB
WFB
WB
CB
EB
1930
106
103
112
85
61
70
1981
110
103
110
72
62
75
Decade Starting Year
1990 2000 2010 2020 2030 2040 2050
107
100
107
%
83
92
109
105
111
83
68
79
93
86
95
45
34
31
86
82
91
66
53
52
86
83
91
51
37
46
66
61
70
31
14
5
46
44
56
29
8
7
* Average dates of first- and last-ice for winters with ice cover.
** Average season length for all winters in each scenario.
5-14
-------
Q
UJ
OC
UJ
o
UJ
o
O)
<
03
UJ
S
UJ
o
OC
UJ
Q.
RQURE 2
DAILY AVERAGED ICE COVER 1951-80 BASE & GISS-A SCENARIO
LAKE ERIE CENTER BASIN
QI38-A 1981-2009
QI88-A 2010-2039
DEC
JAN
MAR
APR
MAY
MONTH
-------
FIQURE 3
DAILY AVERAGED ICE COVER 1951-80 BASE & GISS-A SCENARIO
LAKE SUPERIOR WEST BASIN
GIS3-A 1981-2009
GISS-A 2010-2039
DEC
JAN
FEB MAR
MONTH
APR
MAY
-------
FIGURE 4
DAILY AVERAGED ICE COVER 1951-80 BASE & QISS-A SCENARIO
LAKE SUPERIOR AT WHITEFISH BAY
100
o
111
cc
111
§
o
HI
o
CO
<
CD
LL
O
UJ
UJ
o
DC
UJ
Q.
80
QIS3-A 1981-2009
QISS-A 2010-2039
DEC
JAN
FEB
MAR
APR
MAY
MONTH
-------
Q
UJ
OC
UJ
O
O
UJ
O
CO
<
00
LLJ
O
UJ
O
DC
UJ
CL
RGURE 5
DAILY AVERAGED ICE COVER 1951-80 BASE & 2CO2 SCENARIOS
LAKE ERIE WEST BASIN
100
80H
60
40
20 H
LEGEND
BASE
BASE +/- SD
GISS 2CO2
QFDL 2CO2
OSU 2CO2
a
DEC
JAN
FEB MAR APR
MONTH
MAY
-------
RGURE 6
DAILY AVERAGED ICE COVER 1951-80 BASE & 2CO2 SCENARIOS
LAKE ERIE CENTER BASIN
100
0
UJ
oc
UJ
o
o
1AJ
o
CO
<
CD
U.
O
UJ
I
UJ
o
cc
UJ
Q.
80
60-
40-
20
LEGEND
BASE
BASE +/- SO
GISS 2CO2
QFDL 2CO2
OSU 2CO2
/ 7
K/ , , f-\/£y,
I r^'^^l
DEC
JAN
FEB MAR
MONTH
APR
MAY
-------
4>
FIGURE 7
DAILY AVERAGED ICE COVER 1951-80 BASE & 2CO2 SCENARIOS
LAKE ERIE EAST BASIN
100
Q 80
111
QC
HI
5
O
UJ
O 60
CO
<
CD
LU
O
LU
CD
LU
o
QC
iii
Q.
40
20-
LEQEND
BASE
BASE +/- SD
QISS 2C02
QFDL 2C02
OSU 2CO2
l
DEC
JAN
FEB MAR
MONTH
APR
MAY
-------
o
UJ
tr
UJ
s
o
UJ
O
CO
<
CD
UJ
O
UJ
o
cc
UJ
0.
FIGURE 8
DAILY AVERAGED ICE COVER 1951-80 BASE & 2CO2 SCENARIOS
LAKE SUPERIOR WEST BASIN
100
80
60
DEC
JAN
FEB MAR
MONTH
APR
MAY
-------
HI
I
w
o
oc
UJ
0.
FIGURE 9
DAILY AVERAGED ICE COVER 1951-80 BASE & 2CO2 SCENARIOS
LAKE SUPERIOR EAST BASIN
100
Q 80
UJ
OC
UJ
O
O
UJ
Q 60
CO
<
03
U.
O
40 H
20 H
LEGEND
BASE
BASE ->•/- SD
QISS 2CO2
QFDL 2CO2
OSU 2CO2
a
DEC
JAN
FEB MAR
MONTH
APR
MAY
-------
Q
UJ
OC
UJ
8
O
UJ
O
CO
<
CD
UJ
I
s
cc
UJ
Q.
FIGURE 10
DAILY AVERAGED ICE COVER 1951-80 BASE & 2CO2 SCENARIOS
LAKE SUPERIOR AT WHITEFISH BAY
100
BASE +/- 8D
GI33 2CO2
QFDL 2CO2
OSU 2CO2
.,, \\
r »' » V :
20-
DEC
JAN
FEB MAR
MONTH
APR
MAY
-------
Assel
CHAPTER 4
INTERPRETATION AND LIMITATIONS OF RESULTS
INTERPRETATION
Winter duration is much shorter, ice cover less extensive, and frequency of winters without ice cover much
greater under doubled CO, warming scenario winters relative to the 1951-80 base period. Average winter
duration for the 1951-80 base period was 13 to 16 weeks; under the doubled CO, scenarios, the average winter
duration is 5 to 13 weeks shorter. Daily average ice cover under all of the doubled CO2 scenarios is likely to
be limited to the shore area and shallows of each lake basin because these are the first areas to form ice and
because maximum average daily ice concentration (30 year average) is less than 30% for all basins except
Whitefish Bay. The frequency of winters without ice cover is much greater in the East and Central Basins of
Lake Erie relative to the West Basin because of their greater mean depth. Ice concentration and duration trends
followed the air temperature trends of the three doubled CO2 scenarios; that is, the coldest scenario (OSU) had
the greatest ice concentration with the longest ice cover duration, and the warmest scenario (GFDL) had the
smallest ice concentration with the shortest ice season duration. Under the 79-year transient CO2 scenario
(1981-2059), only Lake Erie basins have winters without ice cover; during the last three decades of the transient
(2030-59), 30 to 80% of the winters for Lake Erie's Center and East Basins are without ice cover. Transient
scenario daily ice concentration, averaged for the years 2010-39, is significantly less than the base period for all
lake basins. However, extensive ice cover will occur under many transient CO2 winters, particularly during the
first 29 years (1981-2009). Average ice cover duration is 3 to 7 weeks shorter during the next 30 years of the
transient 2010-39 relative to the 1951-80 base period. During the last decade of the transient scenario (2050-59),
decadal-averaged ice concentration and ice cover duration is similar to the doubled CO2 scenarios. Ice cover
simulation of an analog climate warming period (1930-39) shows that average ice cover duration was 1/2 to 1
week shorter than the 1951-80 base period for Lake Superior and about 3 to 4 weeks shorter for the Center and
East Lake Erie basins. Annual maximum ice concentration was less then the base period but greater than the
doubled CO2 scenario.
LIMITATIONS
The ice cycle models do not account for the effects of wind, nor do they consider the entire annual heat
budget of Lakes Superior and Erie. If the climate warming scenarios have significantly higher winds or greater
frequency of severe storms, then models may over-estimate ice concentration; if wind speeds are lower, models
may under-estimate ice concentration. Lake heat storage increases under global wanning may result in threshold
FDD developed for the 1960-79 base period no longer being representative of time of ice formation. Two points
should be kept in mind regarding this limitation: (1) there were winters in the 1960-79 calibration period for the
ice cycle analysis in which only shore ice formed in Lakes Erie and Superior so that the threshold FDD
accumulations approximate winters in which no ice cover forms, and (2) the threshold FDD values verified well
over independent data for a winter without ice cover and for a winter with only shore ice. The threshold FDD
simulated a no ice cover condition for mild 1952-53 winter, no ice cover formed in the East and Central Bases
of Lake Erie that winter, and the models verified well for the mild 1982-83 winter season in which ice cover
formed later than normal. Another potential limitation is associated with the average daily ice melt rates
developed during the regression analysis. The melt rates used to estimate ice loss may under-estimate melt rates
during CO2 warming scenarios. But the threshold FDD values and ice melt rates represent at the very least an
upper limit of potential ice concentration and ice duration under global warming, and they are considered a good
first approximation to ice conditions under doubled CO2 wanning scenarios.
5-24
-------
Assel
The uncertainty in simulated ice concentration and ice duration (standard error analysis) is estimated to
range from 20 to 30% for ice concentration and from 2 to 4 weeks for ice cover duration. The models are
sensitive to the magnitude and number of consecutive days with air temperature below freezing. If FDD
accumulations fluctuate about the value needed for initial ice formation all winter, the models show intermittent
periods of ice formation and loss. In such cases, ice cover and ice duration can be either over- or under-
estimated.
Information on the spatial and temporal distribution of the ice cover is limited by the calculation of basin
mean ice concentration. This is a significant limitation in studies where spatial details on ice concentration are
needed, as in ecological studies where the date and extent of initial ice cover and duration of shore-fast ice is
important.
5-25
-------
Assel
CHAPTERS
IMPLICATIONS OF RESULTS
ENVIRONMENTAL IMPLICATIONS
We are just beginning to understand the importance of ice cover to lake ecology. Freeberg and Taylor (in
press) observed that year-class strength of lake whitefish is related to winter severity. Under the doubled CO2
scenarios, the Great Lakes may not have ice cover some winters. If ice cover is missing, whitefish and perhaps
other cold water fish species may vanish from the Great Lakes. Bolsenga (in press) has observed that some
biological activity actually increases under the protection of the ice cover in the shore zones of the Great Lakes;
the loss of the ice cover therefore may result in a reduction in the annual abundance of some micro-organisms
and perhaps significantly affect larger life forms that prey on them. The ice cover also protects some shore areas
against the impact of high-energy waves that might otherwise cause shore erosion (Zumburge and Wilson, 1953).
SOCIO-ECONOMIC IMPLICATIONS
The ice cover impedes and eventually stops most navigation in the Great Lakes during the winter months.
Aids to navigation that would be damaged by ice are removed in late fall and reinstalled the following spring.
Ice booms, which help prevent ice jams, are installed at the head of the St Marys and Niagara Rivers to aid the
formation of stable ice cover lakeward of the head of these rivers (International Niagara Working Committee,
1983). The U.S. and Canadian Coast Guard and hydropower authorities are involved in this activity; the Coast
Guard also assists ships beset in the ice. The results shown here indicate the navigation season could be
extended to 10 months or perhaps even 12 months under a doubled CO2 climate warming. Thus a considerable
cost savings may be associated with reduced Coast Guard and hydropower authority activity and increased
shipping activity in the winter months. The greatly reduced extent and duration of ice cover will likely result in
higher evaporation from Lake Superior and lower lake levels during the winter months (Croley and Hartmann,
in press). The higher lake evaporation during winter implies an increase in snowfall in the Snow Belt regions
of the Great Lakes. There is a considerable amount of winter recreational activity on ice-covered bays and
harbors of the Great Lakes - ice boating, ice fishing, snowmobile racing. Much if not all of this activity would
be reduced or discontinued completely with reduced ice cover.
5-26
-------
Assel
CHAPTER 6
POLICY IMPLICATIONS
The management of the Great Lakes fishery should be reviewed relative to commercial and sports fishing
since there may be loss or reduction of some fish species and increases or introduction of other fish species.
Port and harbor facilities may need to be upgraded to support increased ocean-going and local shipping activities
that would become possible with year-round navigation. New regulation plans may need to be developed for
controlling flows through the St Marys, Niagara, and St Lawrence Rivers.
5-27
-------
Assel
GLOSSARY
LAKE & BASIN ABBREVIATIONS
Sup » Lake Superior
Eri - Lake Erie
WB = West Basin
CB = Center Basin
EB - East Basin
WFB = Whitefish Bay
SCENARIO AND AGENCY ABBREVIATIONS
UNITS ABBREVIATIONS
C - Celsius
cm » centimeters
m = meters
km = kilometers
EPA
2C02
1CO2
GCM
GISS
2CO2 GCM GISS-A
CO2 GCM GFDL
2CO2 GCM OSU
= United States Environmental Protection Agency
» double carbon dioxide scenario
= single carbon dioxide scenario
= Global Circulation Model
» Goddard Institute of Space Science
= Goddard institute of Space Science transient
- Geophysical Fluid Dynamics Laboratory
= Oregon State University 2CO2 GCM
ADDITIONAL TABLE. FIGURE. AND TEXT ABBREVIATIONS
Eq. • equation
C1,C2,C3 = coefficients of regression in Table 2
FDD * the accumulated freezing degree-days (C) on a given date
FDDCRIT = a critical FDD accumulation for a given day in the annual
Great Lakes ice cycle; if FDD exceeds this value ice cover remains at 100 percent.
BFDD - a threshold value of FDD representing (1) the date of the end of fall overturn in Eq. 1 and (2)
the number of FDD needed to cool the near shore water to 0 degrees C for Lake Erie basins, Eq.
2.
FZFAC = a on/off switch for ice formation in Eq. 2, ice is not permitted to form until FDD equals
BFDD. Before that time FZFAC=1000, after that time FZFAC-0.
JD • a day counter, the ice cycle starts (JD=1) the first day FDD is greater than BFDD
MFDD * the annual maximum FDD accumulation
DMFDD » the date of the annual maximum FDD accumulation
MELT » the average daily ice melt rate (cm/day)
ARITHMETIC OPERATORS
sqrt
exp
multiplication
addition
greater than
square root
exponential function (base e)
/ - division
- » subtraction
< - less than
STATISTICAL ABBREVIATIONS
RMSE - Root Mean Square Error
SD = Standard Deviation
5-28
-------
Assel
REFERENCES
Assel, RA. "A Computerized Data Base of Ice Concentration for the Great Lakes.* NOAA Data Report ERL
GLERL-24, NOAA Great Lakes Environmental Research Laboratory, Ann Arbor, Michigan, 1983. 26 pp.
Assel, RA., "Great Lakes Degree-day and Winter Severity Index Update: 1897-1983." NOAA Data Report
ERL GLERL-29, NOAA Great Lakes Environmental Research Laboratory, Ann Arbor, Michigan, 1986. 54 pp.
Assel, RA., RH. Quinn, GA. Leshkevich, and SJ. Bolsenga. "NOAA Great Lakes Ice Atlas." PB84160811,
National Technical Information Service, Springfield, Virginia 1983. 115 pp.
Assel, RA., C. R. Sinder, and R. Lawrence. "Comparison of 1982-83 Winter Weather and Ice Conditions with
Previous Years." Moa Wea. Rev. 113(3):291-303, 1985.
Bilello, MA^ "Method for Predicting River and Lake Ice Formation." J. Applied Meteorology 3(l):38-44, 1964.
Bolsenga, S J. "Nearshore Great Lakes Ice Cover." J. Cold Regions Sci. and Tech. 15:99-105, 1988.
Bolsenga, SJ. "An Under Ice Ecology Pilot Program, Operations and Preliminary Scientific Results." J. Great
Lakes Res. (in press).
Brownlee, KA. "Statistical Theory and Methodology in Science and Engineering", second edition. John Wiley
& Sons, Inc. New York, New York, 1967. pp. 295-305.
Croley T£.A and H.C. Hartmann. "Effects of Climatic Changes on the Laurentian Great Lakes Levels."
NOAA Great Lakes Environmental Research Laboratory, Ann Arbor, Michigan (in press).
Diaz H.F, and Quayle R.G. "The Climate of the United States Since 1895: Spatial and Temporal Changes."
Moa Wea Rev. 108(3):249-266, 1980.
Freeberg M.H, and W. Taylor. "The Impact of Egg and Larval Mortality on Year Class Strength of Lake
Whitefish." Department of Fisheries and Wildlife, Michigan State University, East Lansing, Michigan (in press).
Lewis, PJ. "Severe Storms Over the Great Lakes: A Catalogue summary for the Period 1957-1985." Canadian
Climate Center Report No. 87-13, Atmospheric Environment Service, Downsview, Ontario, Canada.
Howe, DA-, D.S. Marchand, and C. Alpaugh. "Socio-economic Assessment of the Implications of Climatic
Change for Commercial Navigation and Hydro-electric Power Generation in the Great Lakes St. Lawrence River
System." Great Lakes Institute, University of Windsor, Windsor, Ontario, Canada. 1986. 118 pp.
International Niagara Working Committee. "1982-83 Operations of the Lake Erie-Niagara River Ice Boom."
UJS. Army Corps of Engineers. Buffalo District, Buffalo, New York. 1983. 7 pp.
Oak, W.W., and H.V. Myers. "Ice Reporting on The Great Lakes." Weatherwise 6(1):7-10, 1953.
Palecki M A, and Barry, R.G. "Freeze-up and Break-up of Lakes as an Index of Temperature Changes during
the Transition Seasons: A Case Study for Finland." J. Climate and Applied Meteorology 25:893-902, 1986.
5-29
-------
Assel
Richards, T.L. "Meteorological Factors Affecting Ice Cover on the Great Lakes." In: Proceedings of the Sixth
Conf. on Great Lakes Research. International Association for Great Lakes Research, Ann Arbor, Michigan,
1963. pp. 204-215.
Rogers, J.C. "Long-range Forecasting of Maximum Ice Extent of the Great Lakes." NOAA Tech. Memo. ERL
GLERL-7. National Technical Information Service, Springfield, Virginia. 1976.15 pp.
Sleator, F.E. "Ice Thickness and Stratigraphy at Nearshore Locations on the Great Lakes." NOAA Data Report
ERL GLERL-1-2. NOAA Great Lakes Environmental Research Laboratory, Ann Arbor, Michigan, 1978. 434
pp.
Snider, R.C. "Great Lakes Ice Forecasting." NOAA Tech. Memo. NWS OSD 1. National Technical
Information Service, Springfield, Virginia, 1971.106 pp.
Tramoni F., R.G. Barry, and J. Key. "Lake Ice Cover As A Temperature Index for Monitoring Climate
Perturbations." Zeit. Gletscherkunde 21:43-49,1985.
Williams, G.P. "Correlating Freeze-up and Break-up with Weather Conditions." Canad. Geotech. J.
11(4)313-326,1965.
Williams, G J». "Predicting the Date of Lake Ice Breakup." Water Resour. Res. 7(2):323-333,1971.
Zumburge, J.H., and J.T. Wilson. "The Effects of Ice on Shore Development." In: Proceeding of the Fourth
Conf. Coastal Engineering, Chicago I1L, J. W. Johnson, ED. Council on Wave Research, The Engineering
Foundation, University of California, Berkeley, California, 1953. pp. 201-205.
5-30
-------
POTENTIAL CLIMATE CHANGES TO THE
LAKE MICHIGAN THERMAL STRUCTURE
by
Michael J. McCormick
NOAA/GLERL
2205 Commonwealth Blvd.
Ann Arbor, MI 48105
Interagency Agreement No. DW13932957-01-0
-------
CONTENTS
FINDINGS 6-1
CHAPTER 1: INTRODUCTION 6-2
CHAPTER 2: METHODS 6-4
The Mixed Layer Model 6-4
Model Development 6-4
The Scenarios 6-7
CHAPTER 3: RESULTS 6-8
Base Climatology Simulation 6-8
Impacts of Climate Change 6-11
CHAPTER 4: DISCUSSION, CONCLUSIONS, AND SPECULATIONS 6-23
Interpretation of Results 6-23
Speculations 6-24
REFERENCES 6-25
11
-------
McCormick
FINDINGS1
A one-dimensional numerical model after Garwood (1977) was used to estimate the vertical climatological
temperature structure in Lake Michigan. The climatology was based on the model output from simulations of
the 1981-1984 offshore temperature field. Quasi-two-dimensional effects were also accounted for by the model
by prescribing a weak upwelling velocity during the winter months. Once the climatology was estimated, several
different global circulation model (GCM) scenarios were examined. Three different GCM scenarios of doubled
CO, (2xCO,) and one transient scenario were simulated. They were (1) Goddard Institute of Space Sciences
(GISS) 2xCOg, (2) the Geophysical Fluid Dynamics Laboratory (GFDL) 2xCO2 scenario, (3) the Oregon State
University (OSU) 2xCO2 scenario, and (4) GISS Transient scenario A corresponding to the 2010-19 decade
(GISS A).
In general, the GISS, GFDL, and OSU simulations suggest the following impact on Lake Michigan. The
winter heat content of the lake will be significantly higher than under current climate estimates. The summer
heat content will, in general, be higher than the current climate too but not to the same extent as seen during
the winter months. The higher winter heat content will cause an earlier setup for thermal stratification by as
much as two months and thus a much longer stratified season will result The earlier onset of stratification
coupled with little change hi the wind stress pattern will yield stronger stratification. Thus, the greatest
differences between the 2xCO2 and the present climatology are for an earlier, longer duration, and stronger
stratification.
The monthly averaged mixed layer depth (mid) may be deeper in the whiter and shallower during the
summer than current seasonal averages. However, for the winter months this is not true at less than monthly
time scales. At higher than monthly frequencies, the present mlds may penetrate to the lake bottom any time
during late fall through spring in response to storms and strong surface cooling, while the 2xCO2 calculated mlds
do not. In general, the 2xCO2-derived mlds are restricted from penetrating deep waters because of the
persistence of higher than present water column temperatures ( > 4°C), which results in less potential energy being
converted into mechanical energy to aid the mixing and deepening process. Thus, the true range in mixed layer
depth may well be severely decreased in the future, with only infrequent to rare episodes where the surface mixed
layer encroaches on the deep lake bottom, Le., no turnover. The shallow summer mixed layer will be warmer
and more buoyant than presently observed, making it more difficult for entrainment and/or mixing to occur.
The most critical parameter controlling the thermal structure is the wind stress. Calculations of the potential
climate impacts were made using uncertain future scenario winds that differ little from the present climate.
Should future windspeeds be reduced from those used here, then sensitivity analyses suggest that all of the
previously described impacts may underestimate the true impact on the annual thermal cycle.
Simulation results based on the GISS A scenario suggests that some of these effects may be evident 20 to
30 years from now.
1 Although the information in this report has been funded wholly or in part by the U.S. Environmental
Protection Agency under Interagency Agreement No. DW13932957-01-0, it does not necessarily reflect the
Agency's views, and no official endorsement should be inferred from it
6-1
-------
McConnick
CHAPTER 1
INTRODUCTION
Lake Michigan is a large lake covering over 57,000 km2 with a maximum depth of 281 meters and a mean
depth of 85 meters. Because of its large size, the dominant controlling physics of the lake environment is more
similar to oceanic situations than it is to "small" lakes. Water temperature is one of the most fundamental
physical properties, and accurate knowledge of its distribution is often critical to oceanographic and limnological
problems. Despite the importance of temperature, little is known of its climatology in the Great Lakes and in
particular that of Lake Michigan. Feit and Goldenberg (1976) determined surface water temperature
climatologies for Lakes Superior, Huron, Erie, and Ontario, but the record lengths were short, ranging from only
4 to 10 years. While water intake temperature records of much longer duration exist for many locales in the
Great Lakes, their nearshore proximity makes them a poor candidate for constructing meaningful climatologies.
Other temperature data sets from more favorable locations exist, but their poor temporal coverage precludes
their usage as well.
The most comprehensive data set to date, describing Lake Michigan temperatures in offshore waters, was
obtained by the National Data Buoy Center (NDBC) in the central southern basin of the lake from 1981-1984.
However, these data too do not fulfill all of the needs for generating a water temperature climatology because
of gaps in their temporal coverage and limited spatial coverage as well. In particular, these data only cover the
top third of the water column (Le., 50 meters of the 150-meter mooring depth), and since the mooring is
deployed only during the ice-free season, no data exist for the winter months. Consequently, the only alternative
is to estimate the climatology by modeling the temperature field, and to use the NDBC data for model testing.
The Garwood (1977) model is used herein to estimate the water temperature climatology for Lake Michigan
and potential changes to it that may occur should the climate change. Intermodel comparisons by McConnick
and Meadows (1988) and Martin (1985) found the Garwood model to be successful for simulating the seasonal
temperature cycle for inland seas and in open ocean applications, respectively. Figure 1 shows the study location
and an idealized temperature profile with a shallow surface mixed layer.
The remainder of this report will describe the model development, results, conclusions, and speculations on
the Lake Michigan simulations under various future climate scenarios hypothesized by several GCMs.
6-2
-------
McConnick
Depth
1
Temperature—.
Mixed Layer
Figure I. Lake Michigan study location (top) and an idealized temperature profile showing a shallow surface
mixed layer (bottom).
6-3
-------
McCormick
CHAPTER 2
METHODS
The Mixed Layer Model
The Garwood model version used here is described in McCormick and Meadows (1988). First though, a
brief description of the model is in order. The model is one-dimensional in the vertical and is based on the
turbulence kinetic energy (TKE) budget. During the summer months, the thermal structure at any given time
in general is dependent on the dynamic balance between the wind stress, which tends to destabilize the water
column and mix it, and a positive surface heat flux, which warms the surface waters and tends to stabilize the
water column and retard mixing. During the winter months when the lake is cooling and the surface heat flux
is negative, the wind stress effects on mixing are aided in their efforts by convective mixing, generated from
gravitational instabilities due to the surface cooling. These and other important processes are expressed in the
TKE equation.
Two of the processes in the TKE budget which also affect the vertical distribution of temperature are viscous
dissipation and entrainment due to shear instabilities at the mixed layer base. This version of the Garwood
model allows for mixed layer deepening due to turbulent erosion and shear mechanisms. McCormick and
Meadows (1988) added the shear instability source to Garwood's model and assumed the major shear source
to be from wind-generated pure inertia! oscillations. Observations on the frequency distribution of kinetic energy
in the Great Lakes support this interpretation (e.g., Savior et al., 1980; Boyce and Chiocchio, 1987). To estimate
this contribution to the TKE budget, the shear strength was estimated from one-dimensional momentum
equations after Thompson (1976).
If mixed layer deepening is to be realistically described for all possible forcing conditions, then energy
dissipation must be explicitly included in the TKE budget. Garwood parameterizes dissipation on two scales.
First, energy is removed in proportion to the magnitude of the total TKE, and in the second dissipation process,
energy loss is proportional to both the TKE and the depth of the mixed layer. This parameterization of
dissipation is advantageous to long-term simulation by avoiding the possible carry-over and buildup of potential
energy over annual time scales. Thus, under well-behaved forcing conditions, cyclic solutions are possible.
Model Development
No process-oriented models have been used to do multi-year simulations of temperature in Lake Michigan.
Making the transition from seasonal to annual length simulations has been problematic and shortcomings still
remain. The time and place chosen for testing and enhancing the Garwood model was 1981-1984 at the site of
the NDBC meteorological buoy in the center of the southern basin with a depth of 150 meters. During the ice-
free months of 1981-1984, the NDBC hung a thermistor string from their buoy. Nine thermistors were positioned
at approximately 5 meters spacing covering the top SO meters of the water column. Temperatures were recorded
at hourly intervals, but the data return and quality were less than ideal. At no time during the 1981-1984 period
were all thermistors operational. At various times, as few as two thermistors and as many as seven were
recording useful data. Furthermore, analyses of the low frequency response of the data suggests that their
accuracy is no better than 0-5°C. Nonetheless, it is the best available data set for this study.
Hourly meteorological data were assembled for a period spanning 30359 hours from 16 July 1981 through
31 December 1984. (The 16 July date was the date of the first NDBC temperature record) No offshore water
temperature records are available for the 1951-80 period The meteorological data were obtained from the
NDBC buoy and from airport meteorological stations at Milwaukee, Wisconsin, and Muskegon, Michigan. The
airport data were averaged with respect to each other and were used whenever buoy meteorological data were
missing. Airport meteorological data were used for December in 1981, for January through March and October
through December in 1982, and for 1983 and 1984 January through March and for the month of December.
Both the airport windspeeds and directions were adjusted for overwater conditions following Schwab (1983).
6-4
-------
McCormick
Hourly observations of windspeed and direction, air temperature, dew point temperature, and total cloud
cover were used to force the model Data on the shortwave global radiation (290-4000 mm) were unavailable
and thus were estimated from an empirical model after Cotton (1979). The NDBC meteorological buoy had no
provisions for measuring the dew point temperature, essential for estimating the latent heat flux, so all dew point
data were taken from shore-based data and were corrected for over-water conditions after Philips and Irbe
(1978). Once the meteorological data were assembled, the model testing and development began.
The Garwood model was forced with hourly meteorological data, and numerically integrated over one-
hundred-fifty 1-meter-thick grid points at one-hour time steps. The initial conditions were estimated from the
16 July 1981 data and the solution was marched in time for 30359 hours to the end of 1984. Simulation of the
winter regime revealed the need to reevaluate the model physics.
First, the cold winter temperatures of 1982 drove the surface water temperature to freezing conditions on
several days. No faculties are included in the model to properly account for ice formation, buildup, and decay.
Therefore, the surface temperature was artificially constrained to always be greater than or equal to zero. The
modeled surface heat flux during these episodes was set equal to the surface irradiance only and thus does not
represent the true surface heat flux. Fortunately, these episodes were infrequent enough in their occurrence so
as to not seriously bias the monthly averaged surface heat flux estimate. Second, Farmer and Carmack (1982)
noted the importance of the nonlinear pressure temperature term on the density when temperatures are near
the temperature of maximum density. The interaction between pressure and temperature has a strong influence
on the mixed layer depth in deep lakes, like Lake Michigan, during the winter months. Hence, in contrast to
most mixed layer modeling efforts, it was deemed necessary that pressure effects be explicitly accounted for by
the equation of state. The equation of state after Pickett and Herche (1984) was used and is shown here in the
following equation,
p - 999.968 - .00773T02 + •526xlO-4T(3 + .0492P - .00021PT0 (1)
where, p - density (kg/m3), TQ = temperature °C (T-3.98), and P = pressure (bar).
Simulations of the 1982 Lake Michigan springtime transition to thermal stratification with mixed layer models
after Denman (1973), Garwood (1977), McCormick and Scavia (1981), and Thompson (1976) suggested the need
for yet additional physics. Each model was premature in its timing of the spring transition. Studies of the
velocity profile at several locations in the benthic boundary layer of the southern basin by James Saylor of the
Great Lakes Environmental Research Laboratory (GLERL) in Ann Arbor, Michigan, has revealed the presence
of an Ekman boundary layer. Mass balance calculations suggest that significant upwelling velocities, We, (due
to a convergence of the Ekman boundary layer) should persist during the winter months in the region where this
study was made, and thus it may be an important source/sink of heat to surface waters which must be accounted
for by the mixed layer model. The steady-state Ekman pumping velocity, We, is given as
We - (pfy'CurKr,,) (2)
where, f » Coriolis force and TV = bottom stress vector. The Curl of the bottom stress vector was estimated
from current meter data from four moorings surrounding the central portion of the southern basin. This final
model modification was implemented by prescribing We for January through May of each year. Constant
monthly values were used with a peak velocity of 1 m/day used for March. The monthly velocities are listed in
Table 1. The upwelling details will be described in a forthcoming paper.
Specifically, the heat flux due to upwelling, when it occurs, is handled at each time step by first calculating
the temperature profile without any consideration of upwelling. Then the temperature change, ATi, at level i,
is calculated by equation (3),
AT,-(T,+ 1-T,)W0At/AZ (3)
where, At is the time step and AZ is the grid size. Equation (3) is applied from the surface (i» 1) to near bottom
(i-n-1). Thus, when surface waters are colder/warmer than those at depth, the upwelling heat flux is
positive/negative. Although this is a coarse approximation of the true upwelling structure, it nonetheless has
6-5
-------
McConnick
TABLE 1. Model Inputs
Model
(1)
(2)
(3)
(4)
Model
(1)
(2)
(3)
(4)
Model
(1)
(2)
(3)
(4)
Jan
-.4
-.3
-.5
-.1
Jan
8
7
6
3
Jan
1.82
1.56
1.13
1.16
Feb
-0.5
•1.0
-0.3
-0.2
Feb
8
7
4
3
Feb
1.49
1.44
1.06
1.20
WIND SPEED (m/s) (2xC02 - GCM Control)
Mar Apr May Jun Jul Aug Sep Oct Nov Dec
-1.7 -0.6 -0.8 -0.3 -0.1 0.0 0.1 1.0 1.1 -1.2
0.3 -0.4 -1.6 -1.6 1.6 0.9 -0.7 -0.6 -0.1 -0.7
0.3 0.4 -0.2 -0.2 -0.6 0.2 -0.3 -0.4 -0.1 -0.1
1.0 -0.4 0.1 -0.5 0.1 1.0 0.6 0.1 -0.2 0.1
OVER-WATER AIR TEMPERATURE (°C)
(GCM - Base Climatology)
Mar Apr May Jun Jul Aug Sep Oct Nov Dec
6533236478
6548956678
3333332333
0300000003
HUMIDITY RATIOS ^xCOj/GCM Control)
Mar Apr May Jun Jul Aug Sep Oct Nov Dec
1.64 1.40 1.30 1.28 1.07 1.28 1.39 1.23 1.54 1.50
1.43 1.37 1.25 1.18 0.97 1.16 1.12 1.31 1.56 1.58
1.05 1.04 1.12 1.17 1.11 1.18 1.25 1.18 1.09 1.14
1.13 1.17 1.06 1.09 1.13 1.14 1.09 1.01 1.05 1.29
SHORTWAVE SOLAR RADIATION RATIOS (2xC02/GCM Control)
Model
(1)
(2)
(3)
(4)
Jan
0.92
2.05
1.05
0.97
Feb
1.04
1.15
1.04
0.90
Mar Apr May Jun Jul Aug Sep Oct Nov Dec
0.98 1.03 1.00 0.99 0.98 1.04 1.04 1.12 1.03 0.99
1.15 0.93 1.05 1.05 1.02 1.01 1.01 1.01 1.07 1.74
1.07 1.09 1.03 0.99 1.01 0.98 1.00 1.02 1.00 1.01
1.01 1.01 0.99 0.96 0.93 0.97 1.01 0.99 1.03 0.86
FRACTIONAL CLOUD COVER RATIOS (2xC02/GCM Control)
Model
(1)
(2)
(3)
(4)
Jan
1.03
1.13
0.76
1.00
Feb
0.93
1.18
0.83
1.05
Mar Apr May Jun Jul Aug Sep Oct Nov Dec
.97 0.98 1.07 1.09 1.08 1.00 1.06 0.85 1.00 0.93
.92 1.09 0.90 0.83 0.95 0.82 0.90 1.00 1.09 0.92
.74 0.61 0.68 0.94 0.85 1.17 0.83 0.81 0.90 0.91
.97 1.04 0.98 1.19 1.44 1.11 0.92 1.04 1.05 1.03
UPWELLING VELOCITY (m/day) (For All Simulations)
Jan Feb Mar Apr May June through December
0.2 .55 1.0 .45 0.3 0.
Model (1) - GISS 2xC02
Model (2) - GFDL 2xC02
Model (3) - OSU 2xCO,
Model (4) - GISS A (Transcient Scenario for 2010-19)
6-6
-------
McConnick
enabled more accurate simulation of the upper water column thermal structure when tested over short time
periods. An ongoing effort to understand the offshore upwelling structure is presently being addressed at
GLERL using a three-dimensional circulation model.
The remaining processes are included in the local heat budget: sensible, latent, net longwave, and shortwave
global radiation. The sensible and latent heat fluxes are calculated after bulk aerodynamic formulas with
atmospheric stablity-dependent exchange coefficients. The stability dependence is based on the work of Businger
et al. (1971), and the program is documented in Schwab et aL (1981). The net longwave radiation is calculated
after Wyrtki (1965), and the penetrating components of the solar irradiance are approximated after Ivanoff
(1977). The extinction coefficients for the visible and infrared radiation bands were 021 m-1 and 2J&S m-1,
respectively.
The Scenarios
Three different GCM results and one transient result were used and compared against the Base climatology,
as estimated by the previously described simulation. The three GCM simulations, corresponding to a climate
with an effective doubling of atmospheric CO2 concentration, were made with the following models: (1) GISS,
(2) GFDL, and (3) OSU. The transient run was made using a decadal average corresponding to 2010-19 with
scenario A. This run is identified in the tables and figures as GISS A.
The 2xCO, meteorology used to drive the Lake Michigan simulations was estimated from model output from
the IxCO, and 2xCO2 GCM simulations. These data were formed into a (^COg/lxCC^) ratio and then used
to adjust (he Base climatology as described below. The lxCO2 GCM simulations were run for a 30-year period
corresponding to 1951-1980. The 2xCO2 general circulation model simulations were also done for a 30-year
period, but with a doubling of the atmospheric concentration of CO, and other greenhouse gases. Monthly
averages were formed for each meteorological parameter, at each grid point, for each simulation. A
ZxCOj/lxCO, ratio for each parameter was formed by dividing the monthly averaged quantity from the 2xCO,
simulation by The lxCO2 one. The transient (GISS A) simulation was handled in the same manner. The GCM
model output from the grid point closest to central southern Lake Michigan was used to represent the future
climate inputs for Lake Michigan. Five different parameters were used from the GCM output: (1) windspeed,
(2) air temperature, (3) humidity, (4) incident solar radiation at ground level, and (5) fractional cloud cover. The
hourly base meteorological data from 1981 to 1984 were adjusted by multiplication with the applicable
(2xCO2/lxCO2) GCM ratios. The GCM ratios were held constant on monthly time scales.
The windspeed adjustments were made differently. The supplied GCM monthly windspeed estimates were
made by vector averaging rather than by scalar averaging the GCM winds. Thus, when the ^COj/lxCO, GCM
wind ratios were formed, the calculated ratios were often very large. If the Base climatology winds were
multiplied with these ratios, then hurricane force winds would have occurred for at least 2 months out of every
simulation year. Therefore, to avoid potentially disastrous results and yet still salvage some of the information
in the GCM winds, the differences between the monthly averaged 2xCO2 and lxCO2 windspeeds were used in
place of their ratio. These differences were then added to the Base climatology winds. The resulting changes
to the wind stress in the Base meteorology were small and more consistent with expectations from other studies
(Cohen, 1986). The monthly averaged GCM inputs are shown in Table 1.
6-7
-------
McCormick
CHAPTERS
RESULTS
Base Climatology Simulation
McCormick and Meadows (1988) simulation of Lake Erie temperatures with the Garwood model found the
optimal model constants to be identical to those found by Martin (1985) in his simulation of North Pacific Ocean
data The success of this model in such diverse environments instills confidence in the model parameterizations
of the governing physics. Thus, all of the model simulations for the Base and future climate scenarios were made
without altering the model coefficients.
The Base climatology simulations are shown against surface water temperature data in Figures 2 and 3. The
surface temperatures are illustrated for the entire tune periods for which the NDBC data were available. For
clarity, less than 5% of the approximately 19,000 observations are depicted in the figures.
The lack of time series data throughout the water column and winter temperature data limits the ability to
fully evaluate model performance. Nonetheless, from Figures 2 and 3 some of the effects of offshore upweUing
and model sensitivity to windspeed are made clear. In Figure 2, low-pass filtered surface temperatures generated
by the mixed layer model with no upwelling (i.e., 0.0 meter/day) are shown against data. The effects of a 10%
decrease (top of Figure 2) and 10% increase (bottom of Figure 2) in windspeed (WS) are shown as well. The
rms error for the 0.0-meter/day simulation was over 4°C for surface temperatures. It is evident that much of
this error is contributed by the poor simulation of the 1982 data
Significant improvement in the surface temperature simulation was made by using a weak upweUing
velocity (i.e., "Variable UpweUing" in Figure 3), which was held constant on monthly time scales and operational
from January through May (Table 1 and Figure 3). The rms error was approximately 3°C overall, with the 1982
data once again being the most difficult to simulate. Additional improvement in the surface temperature
simulation was made in the 0.0-meter/day case with a 10% reduction in windspeed. The overaU rms error for
these simulations was approximately 2°C, with the major difference between this and the other simulations
occurring in 1982. The 1981, 1983, and 1984 rms errors were either similar or slightly worse than the 0.0-
meter/day or variable upwelling simulations with unaltered winds. The 2°C rms error was half of the error seen
in the 0.0-meter/day simulation and better than 1°C in rms error compared to the simulation with upwelling.
The improved rms error occurred because of significant improvements in simulating the spring 1982 data That
year was the coldest winter in this study, and the reduced winds compensated for model shortcomings by
reducing lake heat losses during the winter and thus enabled better agreement between model and data during
the spring transition period.
If the objective were to solely fit surface temperature data, then the represenative choice for our Base
climatology would have been obvious. However, there is no physical justification for arbitrarily reducing the
windspeeds. And although the rms errors with variable upweUing were larger than the 0.0-m/day case with
reduced winds, there is mounting evidence, as described earlier, to justify the use and necessity of upwelling to
properly describe the offshore heat budget. Therefore, the simulation with variable upweUing was judged to be
the most represenative of the region under study and consequently became the "Base" climatology referred to
throughout this work.
Again it is important to note here that in terms of surface water temperature simulation of the effects of
the "no upwelling" versus the "variable upwelling" cases, either one could be made to mimic the other by adding
either a positive or negative 10% bias to the windspeed data This illustrates that the windspeed is most critical
for accurate determination of the Base climatology, and for estimating any possible future alterations to it as weU.
6-8
-------
McCbnnick
Surface Water Temp vs Time
0.0 (- 10% WS)
1981
1982
1983
1984
30 r
0.0 (+ 10% WS)
1981
1982
1983
1984
Figure 2. Low-pass filtered (336-hr cutoff) surface temperatures with no upweUing and under different wind
conditions. Each curve pattern is duplicated by the line joining it to its label.
6-9
-------
McCormick
Surface Water Temp vs Time
PH
30 Variable Upwelling (-10% WS)
Variable Upwelling
25
20
15
10
1981
1982
1983
1984
30
25
20
15
o
Variable Upwelling (+10% WS)
0.0 to/day
1981
1982
1983
1984
Figure 3.
Low-pass filtered (336-hr cutoff) surface temperatures with variable upwelling and under different
wind conditions. Each curve pattern is duplicated by the line joining it to its label.
6-10
-------
McCormick
Impacts of Climate Change
Table 2 and Figures 4-11 summarize the model results. Each figure shows the full simulation and its yearly
average. Each figure has also been low-pass filtered to better identify possible trends.
Figures 4 and 5 show the net surface heat flux as calculated by the various model simulations. These plots
have been low-pass filtered with a 720-hour cutoff period. When upweUing is nonexistent in the simulations
(June through December), the net surface heat flux is the only source/sink of heat for the water column. An
accurate accounting of the change in heat content from January to June, however, requires that the upweUing
heat flux be included in the budget. Table 2 shows monthly estimates of this term, and it should be noted that
the method used to estimate it, for Table 2, is subject to error since it involves multiplication of small differences
between large numbers.
Two points are of interest in these figures. First, the dose similarity between the surface heat fluxes
throughout much of the year, particularly during late summer and early fall. And second, the very large heat
losses seen in December and January of the first and third winters in the Base simulation (top of Figures 4 and
5). The monthly averaged heat flux components are listed in Table 2 and are discussed below.
Of the five heat flux components, the net longwave and shortwave global radiation terms appear to be the
most consistent in their phase and magnitude from model to model (Table 2). The remaining three processes,
sensible, latent, and upweUing, show less model to model agreement The sensible heat flux loss is greatest
during January for each model except OSU, where it occurs during December. The latent heat loss reaches its
maximum during January for Base and GISS A, while it occurs significantly earlier in the other three models,
i.e., October for GISS, September for GFDL, and November for OSU. The averaged upweUing flux shows large
model to model differences. In the Base simulation, the upweUing flux, in general, represents a heat source,
while in GISS, GFDL, and OSU it is a heat sink. Only in GISS A is the upweUing term positive when averaged
over the 5-month period in which it is operational. However, when all the flux terms are summed the net heating
rate that results suggests that the annual averaged net heat flux is within 10 W/m2 of being zero for all cases.
More importantly, the maximum difference between the annual net beat flux under a GCM scenario and Base
is only 8 W/m2. This emphasizes how the persistence of smaU changes in the net heat flux can lead to dramatic
changes to the environment, and because of the uncertainties surrounding these estimates, why they often lead
to controversy.
Figures 6 and 7 show the low-pass filtered surface water temperature. In Figure 6 the Base, GISS, GFDL,
and OSU climatologies are shown. These GCM results suggest higher surface temperatures throughout the year.
Comparison of the transient scenario for 2010-19 (Figure 7) suggests that higher surface temperatures will prevail
from January through July. During the remainder of the year, there is little difference in temperature between
GISS A and Base.
As we proceed down in the water column we can begin to estimate more and more of the potential climate
impact on Lake Michigan. The mixed layer depth comparisons (Figures 8 and 9) together with Figures 6 and
7 suggest how the heat content of the upper water column may behave in the future. Under the GCM scenarios
the mixed layer depth will in general be deeper in the early winter months and will shaUow in spring much
sooner than in the Base results. This suggests that thermal stratification will begin much earlier than is presently
observed. If the interannual variability seen in the top of Figure 8 is truly representative of the Lake Michigan
climatology, then the transition to summer stratification under the GCM scenarios may occur two or more
months earlier than under the present climate.
This is weU Ulustrated after the first cold winter (Figure 8). In that case, the GCM results (GISS, GFDL,
and OSU) suggest that thermal stratification wiU begin in April rather than the late June date seen in Base. The
GISS A results (Figure 9) shows this same tendency but not to the same degree.
6-11
-------
McCormick
TABLE 2. Monthly and annual averaged heat flux components from the Base
climatology and GCM simulations
SENSIBLE
Model
(0)
(1)
(2)
(3)
(4)
Jan
-231
-88
-117
-79
-152
Feb
-80
-9
-25
-42
-33
Mar
-54
0
5
-20
-73
Apr
0
6
9
11
17
May
9
-10
-8
-5
5
HEAT
Jun
-2
-16
-12
-22
-9
FLUX (W/m2)
Jul Aug Sep
-2 -2 -23
-8 -4 -1
19 3 -5
-10 -4 -25
-4 -4 -31
Oct
-26
-15
-4
-20
-30
Nov
-70
6
-22
-45
-72
Dec
-169
-43
-41
-106
-96
Annual
-54
-15
-16
-30
-40
LATENT HEAT FLUX (U/m2)
Model
(0)
(1)
(2)
(3)
(4)
Jan
-141
-86
-125
-127
-126
Feb
-63
-56
-51
-71
-49
Mar
-65
-32
-66
-82
-69
Apr
-24
-23
-23
-36
-14
May
0
-11
-17
-21
0
Jun
0
-16
-49
-27
1
NET LONGWAVE
Model
(0)
(1)
(2)
(3)
(4)
Jan
-81
-57
-60
-72
-68
Feb
-62
-48
-45
-65
-49
Mar
-61
-47
-50
-66
-65
Apr
-48
-44
-39
-61
-35
May
-41
-58
-67
-70
-47
SHORTWAVE
Model
(0)
(1)
(2)
(3)
(4)
Jan
58
52
109
84
56
Feb
86
102
91
116
75
Mar
134
138
177
193
143
Apr
185
197
165
276
183
May
212
205
252
292
217
Jun
-53
-63
-70
-73
-59
Jul Aug Sep
-32 -62 -104
-56 -80 -124
-116 -126 -145
-61 -84 -112
-18 -54 -108
Oct
-89
-136
-112
-105
-96
Nov
-107
-91
-96
-128
-100
Dec
-130
-90
-97
-124
-92
Annual
-68
-66
-85
-81
-60
RADIATION (W/m2)
Jul Aug Sep
-32 -40 -55
-49 -37 -30
-17 -31 -37
-57 -36 -59
-41 -43 -61
Oct
-60
-57
-44
-64
-60
Nov
-64
-40
-47
-64
-64
Dec
-72
-52
-52
-67
-57
Annual
-55
-48
-46
-62
-54
GLOBAL RADIATION (W/m2)
Jun
254
241
297
264
223
Jul Aug Sep
248 218 170
235 226 170
264 247 188
275 198 194
193 200 185
Oct
112
148
114
139
108
Nov
70
72
71
82
70
Dec
50
57
101
60
42
Annual
150
153
173
181
141
UFWELLING FLUX (W/m2)
Model
(0)
(1)
(2)
(3)
(4)
Jan
17
-21
-23
-2
9
Feb
100
-30
-33
9
41
Mar
128
-75
-66
6
70
Apr
27
-59
-49
-22
11
May
-13
-90
-109
-91
-24
Jun
0
0
0
0
0
Jul Aug Sep
000
000
000
000
000
Oct
0
0
0
0
0
Nov
0
0
0
0
0
Dec
0
0
0
0
0
Annual
21
-22
-23
-8
8
6-12
-------
McCormick
Table 2. (continued)
NET HEATING
Model
(0)
(1)
(2)
(3)
(4)
Jan
-379
-200
-214
-195
-281
Feb
-20
-41
-62
-52
-14
Mar
83
-16
0
31
5
Apr
140
77
64
168
163
(W/m2) (Equals
May
167
36
52
105
151
Jun
199
146
166
142
156
Jul
183
123
151
147
130
sum of all fluxes)
Aug Sep
113 -12
105 16
93 1
74 -2
100 -15
Oct Nov
-63 -172
-60 -52
-46 -95
-50 -156
-78 -166
Dec
-321
-130
-89
-238
-203
Annual
-7
0
1
-2
-4
Model (0) - Base Climatology; Models (l)-(4) are the same as in Table 1.
6-13
-------
McCormick
Net Surface Heat Flux vs Time
500
250
X
5-250
tt,
HJ
£-500
ffi
-750
GISS
Base
JAN
JAN
JAN
JAN
Averaged Net Surface Heat Flux
500 r
osu
5^250
-750
JFMAMJJASOND
Figure 4. Low-pass filtered (720-hr cutoff) net surface heat flux.
6-14
-------
Net Surface Heat Flux vs Time
McConnick
500
250
X
5-250
g-500
-750
GISS
Base
JAN
JAN
JAN
JAN
Averaged Net Surface Heat Flux
500 r
5-250
-500
•s
.0)
-750
JFMAMJJASOND
Figure 5. Low-pass filtered (720-hr cutoff) net surface heat flux.
6-15
-------
McConnick
Surface Water Temp vs Time
30
25
20
15
10
5
JAN
JAN
JAN
JAN
Averaged Surface Water Temp
osu
JFMAMJJASOND
Figure 6. Low-pass filtered (336-hr cutoff) surface water temperature.
6-16
-------
Surface Water Temp vs Time
30
25
20
g"
10
GISSA
GISS
JAN
JAN
JAN
JAN
Averaged Surface Water Temp
30
25
20
15
10
GISS
Base
JFMAMJJASOND
Figure 7. Low-pass filtered (336-hr cutoff) surface water temperature.
6-17
-------
McCormick
During the cooling season, the mixed layer depths deepen at approximately the same rate as seen in the
current climatology. This reflects the similarity in surface heat content and that changes to the net surface heat
flux are smallest during the summer months. Deepening proceeds until late fall when the GCM results suggest
an overall cessation of further deepening.
Recall that the simulation is performed for a water column of 150 meter depth. The mixed layer depths
in the Base climatology encroach on the bottom for significant periods. Figures 8 and 9 were low-pass filtered
with a 720-hour cutoff period; thus mixed layer depth fluctuations with shorter time periods are lost. The mixed
layer depths under the GCM scenarios do penetrate to the bottom, on occasion, but not anywhere nearly as often
as in the Base simulation.
Figures 10 and 11 show potential climate effects on the heat content of the entire water column. This is
shown in terms of the vertically averaged water temperature. The GISS and GFDL scenarios show a consistently
greater heat content than Base. The biggest differences occur during the winter months when the vertically
averaged temperature is significantly higher than that seen in Base. During summer and early fall, however, the
heat content increase is less pronounced, with smaller relative increases over the present climatology.
The OSU and GISS A scenarios depart from GISS and GFDL. GISS A (Figure 11) while showing a
general warming during the winter and spring, also shows a possible slight decrease in heat storage during the
summer months. However, the decreased heat content is small enough to not merit speculation.
The OSU simulation is more similar to the GISS A results than it is to the GFDL and GISS simulations.
The OSU run tends to mimic that depicted by GISS A but is displaced to slightly warmer temperatures such that
the yearly averaged heat content (Figure 10) shows only zero to positive increases in heat content over Base at
all times.
6-18
-------
0)
Q
45
90
135
180
Mixed Layer Depth vs Time
Base GISS GPDL OSU
JAN
JAN
JAN
JAN
Averaged Mixed Layer Depth
Base
GISS
GFDL
OSU
45
5 90
P-i
a
135
180
JFMAMJJASOND
Figure 8. Low-pass Filtered (720-hr cutoff) mixed layer depth.
6-19
-------
McConnick
45
90
a,
0)
Q
135
180
Mixed Layer Depth vs Time
Base
GISS
GISSA
JAN
JAN
JAN
Averaged Mixed Layer Depth
B_se
0 r
135 -
180
GISSA
JFMAMJJASOND
Hgure 9. Low-pass filtered (720-hr cutoff) mixed layer depth.
6-20
-------
Heat Content vs Time
McConnick
,0
OSU
GFDL
GISS
Base
Jan
Jan
Jan
Jan
Averaged Heat Content
10
I
t!
OSU
GFDL
GISS
Base
J FMAMJJASOND
Figure 10. Low-pass filtered (168 hr cutoff) heat content.
6-21
-------
McCormick
Heat Content vs Time
o
10
GISSA
GISS
Base
Jan
Jan
Jan
Jan
Averaged Heat Content
GISSA
JFMAMJJASOND
Figure 11. Low-pass filtered (168 hr cutoff) heat content.
6-22
-------
McCormick
CHAPTER 4
DISCUSSION, CONCLUSIONS, AND SPECULATIONS
Interpretation of Results
Tables 1 and 2 suggest that the primary action driving the changes in the GCM simulations is the increased
air temperature. Indeed, on a yearly averaged basis the sensible heat flux showed the greatest absolute change
from the Base simulation. The large monthly increases in air temperature were up to 8°C (see Table 1) and
were responsible for the change in the sensible heat flux. The additional air temperature increase (relative to
the GCM output) resulted from over-water modification of the land-based temperature, which was mandated
in the Base climatology simulations to avoid excessively large surface heat losses. For consistency purposes, the
GCM inputs were treated in an identical fashion.
In general, the GCM results suggest that the mixed layer depth will be shallower than Base. There is one
exception to this. In the early winter season (see January and February in Figures 6 and 7), the GCM mixed
layer depths are deeper than those in Base, yet still far above the bottom. The reason for the deeper mixed layer
during this time period stems from weaker "reverse" stratification in the GCM simulations. In Base, the
temperature contrast between surface and bottom waters is much greater than in the GCMs. The strong surface
cooling in Base during early winter results in stronger early winter stratification and a shallower surface mixed
layer. However, this scene quickly reverses as the lake gains heat. The GCM-simulated mixed layer begins to
shallow while the Base generated one deepens. This interesting behavior occurs because of the relationship
between the total heat content, as generated by the various simulations, and the temperature of maximum
density.
Wind stress has been shown (Adamec and Elsberry, 1984) to be the most sensitive term in controlling
thermal structure. For example, in Figure 2 a negative 10% bias in windspeed could cause a 50% improvement
in the rms error of surface temperature. In these simulations, however, the wind plays no greater role in the
GCM calculations than in the Base simulation. This was supported by comparing the GCM simulations using
GCM winds, versus GCM simulations using the Base climatology winds. Differences in the model results were
insignificant However, this by no means suggests that future wind fields are unimportant with respect to present
conditions. All it does say is that the monthly vector averaged windspeeds are unimportant and inappropriate
for assessing GCM wind sensitivity.
An additional area of concern is the use of monthly averaged data. McCormick and Meadows (1988) have
shown that over 90% of the energy associated with mixed layer deepening occurs at daily and higher frequencies.
Thus if an accurate assessment of mixing impacts on water quality or other limnological problems is to be made,
then the spectral distribution of the wind stress must be well represented There are numerous examples in the
literature where the distribution of physical and chemical tracers is strongly influenced by the frequency and
severity of storm events. Therefore, if the physics is to be described through process-oriented models, as used
in this study, then the episodic nature of mixing requires high frequency information on all the driving forces,
particularly the wind. Although this information was lacking, it does not invalidate this study so much as it points
out the need for further study.
Of the heat flux components, the net longwave radiation was the least sensitive to change in the GCM
scenarios. This is a consequence more of the empirical formulation used to estimate it than it is a confident
estimate of the true response. In fact, there is a growing body of literature suggesting that most empirical
longwave radiation formulations are not accurate enough for climatological applications (Frouin et al., 1988; Fung
et al., 1984).
The net longwave radiation term is not the only term subject to uncertainty. The surface heat flux can be
expected to be in error by as much as 20-30 W/m2 on monthly time scales (Wyrtki and Uhrich, 1982). This
6-23
-------
McConnick
uncertainty could mask the effects of climate change. The present state of knowledge is too uncertain for
quantifying climate change. Yet by using verified, process-oriented models and by referencing the GCM
simulations to the Base simulation, the model and data uncertainties are minimized such that greater confidence
can be placed on the relative changes. Thus the direction, and not magnitude, of change has been the focus
throughout this study.
In conclusion, in each of the GCM scenarios the change is in the direction of significantly higher heat
content, particularly during the winter, a deeper mixed layer depth in early winter followed by a shallower one
in summer, an earlier onset to density stratification, a longer stratified season, a more buoyant surface mixed
layer resulting in less energy available for mixing, and, in general, higher surface water temperatures. The
transient scenario suggests that some of these effects may be evident 20-30 years from now.
Speculations
If the GISS or GFDL scenarios are realized, then surface temperatures in offshore waters may never
decrease below 4°C. In other words, the lake may not fully overturn during mild winters and thus bottom waters
may remain isolated from surface exposure for significant lengths of time. It is possible that the deeper regions
of the Great Lakes (i.e., > 100 meters deep) may experience a permanent thermocline with a shallower seasonal
one occurring in surface waters, just like much of the world's oceans. In areas where the bottom depths are
deep enough for this to occur, and if these regions are polluted, then the reduction in large-
scale vertical mixing, as implied by the GCM simulations, may result in anoxic environments being formed where
they have never before existed.
Wherever temperature effects are important, impacts will be felt. For example, the earlier warming of
surface waters may result in changes to fish recruitment. Undoubtably, there too must be a reduction in the
amount and duration of ice cover. Reducing the ice cover may result in less shoreline protection and increased
erosion. And finally, profound changes may occur in the biota through changes in the composition of the food
chain to those species which would gain a competitive advantage from changes to the seasonal thermal structure.
6-24
-------
McConnick
REFERENCES
Adamec, D. and R. L. Elsberry. 1984. Sensitivity of mixed layer predictions at ocean station PAPA to
atmospheric forcing parameters. J. Phys. Oceanogr. 14:769-780.
Boyce, F. M. and F. Chiocchio. 1987. Inertia! frequency osculations in the central basin of Lake Erie. J. Great
Lakes Res. 13:542-558.
Businger, J. A., Wyngaard, J. C., Izumi, Y. and E. F. Bradley. 1971. Flux-profile measurements in the
atmospheric surface layer. J. Atmos. Sti. 28:181-189.
Cohen, S. J. 1986. Impacts of CCyinduced climatic change on water resources in the Great Lakes basin.
Climatic Change. 8:135-153.
Cotton, G. F. 1979. ARL models of global solar radiation, p. 165-184. National Climatic Center. In: Hourly
solar radiation - surface meteorological observations. Solmet VoL 2. - Final Report Dept. of Energy. 184 pp.
Denman, K. L. 1973. A time-dependent model of the upper ocean. J. Phys. Oceanogr. 3:173-184.
Farmer, D. M. and E. Carmack. 1982. Wind mixing and restratification in a lake near the temperature of
maximum density. J. Phys. Oceanogr. 11:1516-1533.
Feit, D. M: and D. S. Goldenberg. 1976. Climatology of surface temperatures of Lakes Superior, Huron, Erie
and Ontario. U. S. Dept. of Commerce. NOAA. NWS. TDL Office Note TDL-76-16. Silver Spring, Md. 14
pp.
Frouin, R., Gautier, C. and J. Morcrette. 1988. Downward longwave irradiance at the ocean surface from
satellite data: methodology and in situ validation. J. Geophys. Res. 93:597-619.
Fung, L, Harrison, D. E. and A, A. Lads. 1984. On the variability of the net longwave radiation at the ocean
surface. Rev. Geophys. 22:177-193.
Garwood, R. W., Jr. 1977. An oceanic mixed layer capable of simulating cyclic states. J. Phys. Oceanogr.
7:455-468.
Ivanoff, A. 1977. Oceanic absorption of solar energy, p. 47-72. In: Modelling and prediction of the upper
layers of the ocean. E.B. Kraus, ed. New York: Pergamon Press. 325 pp.
Martin, P. J. 1985. Simulation of the mixed layer at OWS NOVEMBER and PAPA with several models. J.
Geophys. Res. 90:903-916.
McConnick, M. J. and D. Scavia. 1981. Calculation of vertical profiles of lake-averaged temperature and
diffusivity, in lakes Ontario and Washington. Water Resour. Res. 17:305-310.
McConnick, M. J. and G. A. Meadows. 1988. An intercomparison of four mixed layer models in a shallow
inland sea. J. Geophys. Res. 93:6774-6788.
Phillips, D. W. and J. G. Irbe. 1978. Lake to land comparison of wind, temperature and humidity on Lake
Ontario during the International Field Year for the Great Lakes (IFYGL). Fisheries and Environment Canada.
Atmospheric Environment. CLI-2-77. 51 pp.
Pickett, R. L. and L. R. Herche. 1984. A simple density equation for Great Lakes Waters. 27th Conf. on Great
Lakes Res. St Catherines, Ontario: Brock University. May 1-3.
6-25
-------
McCormick
Saylor, J. H., Huang, J. C. K. and R. O. Reid. 1980. Vortex modes in southern Lake Michigan. J. Phys.
Oceanogr. 10:1814-1823.
Schwab, D. J., Bennett, J. R. and A. T. Jessup. 1981. A two-dimensional lake circulation modeling system.
NOAA Tech. Mem. ERL GLERL - 38. Ann Arbor, ML 79pp.
Schwab, D. J. 1983. Numerical simulation of low-frequency current fluctuations in Lake Michigan. J. Phys.
Oceanogr. 13:2213-2224.
Thompson, R. O. R. Y. 1976. Climatological numerical models of the surface mixed layer of the ocean. J. Phys.
Oceanogr. 6(4):496-503.
Wyrtki, K. 1965. The average annual heat balance of the North Pacific Ocean and its relation to ocean
circulation. J. Phys. Oceanogr. 18:4547-4559.
Wyrtki, K. and L. Uhrich. 1982. On the accuracy of heat storage computations. J. Phys. Oceanogr. 12:1411 -
1416.
6-26
-------
THE EFFECTS OF CLIMATE WARMING ON
LAKE ERIE WATER QUALITY
by
Alan F. Blumberg
Dominic M. Di Toro
HydroQual, Inc.
1 Lethbridge Plaza
Mahwah,NJ 07430
Contract No. 68-01-7288
-------
CONTENTS
Page
FINDINGS 7-1
INTRODUCTION 7-3
METHODOLOGY 7-4
Thennocline Model 7-4
Lake Erie Eutrophication Model 7-5
Thermodine/Eutrophication Model Interaction 7-9
Climate Scenarios 7-9
RESULTS 7-13
Verification of the Thennocline Model 7-13
Thermal Response to Scenarios 7-19
Water Quality Response to Scenarios 7-22
REFERENCES 7-27
-------
Blumberg
FINDINGS
1
Simulations of the water quality in the Central Basin of Lake Erie with a coupled hydrodynamic and
water quality model have been used to quantify the response of the system to possible global climate warming
trends. It appears that no matter what die detailed changes of the lake stratification dynamics may be, climate
warming will lead to a degradation in water quality. Losses of 1 mg/L of dissolved oxygen in the upper layers
and losses of 1 to 2 mg/L in the lower layers can be expected. There will also be a concomitant increase in the
area of the lake that is anoxic. Even for the historically oxygen-rich periods which occur during windy years, the
impact of climate warming will be to produce a depletion in the dissolved oxygen levels.
An area-averaged thermocline model constructed from the more general hydrodynamic equations of
motion has been used to estimate the lake temperatures and thermocline variability as forced by surface heating
and winds. The highly variable vertical turbulence mixing processes are parameterized by the use of a second-
moment, turbulence closure submodel with no adjustments from previous applications to its requisite coefficients.
Applications of the thermocline model succeed reasonably well in reproducing the physical behavior of the lake
during two very different periods, 1970 and 1975. An already calibrated Lake Erie eutropbication model using
the vertical mixing information emanating from the thermocline model formed the basis of the water quality
analysis.
Climate scenarios from three different atmospheric general circulation models are used to drive the
coupled thennocline/water quality model The general circulation models estimate the equilibrium climate
induced by a doubling of the atmospheric CO2 concentration. All the models show global warming trends of
18 to 4.2°C in surface air temperature; however, considerable uncertainties exist in the other hydrometeorological
parameters (Cohen, 1986). While there is a need to improve the state-of-the-art in simulating equilibrium
climate change, the range of conditions used in this study encompasses a large part of the expected atmospheric
response.
An analysis of the results from the coupled model forced by the climate warming scenarios and
specifically designed sensitivity experiments suggest that there will be a significant decline in the lake's water
quality. The decline is due to the expected wanner lake temperatures which increase the rate of bacterial
activity in the hypolimnion waters and sediment enough to drive the system to lower dissolved oxygen levels
independent of the depth at which the thermocline becomes established, that is, the dynamics of the lake
stratification. Water quality criteria for ambient dissolved oxygen for fresh water fish (cold water criteria)
specifies a one day mtnimnm of 3.0 mg/L for adult fish life stages and 8.0 mg/L minimum for early life stages,
(EPA, 1985). A 1.0 mg/L reduction in the epilimnion dissolved oxygen as projected by all of the warming
scenarios would not greatly affect fish life. However, the losses of dissolved oxygen in the hypolimnion as
projected by the scenarios could lead to dissolved oxygen levels of 3.0 mg/L or less, which would certainly pose
a threat to adult fish life. For a comprehensive review of the impacts of this predicted loss of dissolved oxygen
and more generally, the impacts of warming on water resources, the reader is referred to Cohen (1986).
The conclusions reached here must be tempered with the following caveats. First, there is the
possibility that the eutrophication model has been exercised beyond the range of data (specifically water
temperature) for which it was calibrated and validated. This should lead to increased uncertainty in the
magnitude of the results for the various climate warming scenarios, but should not affect the direction of change.
In addition, the conclusions apply to the Central Basin only. The Eastern Basin, being much deeper, and the
Western Basin, being much shallower, are certain to have different responses to the climate warming. Indeed,
some caution should be exercised in attempting to extend the conclusions of this study to the Great Lakes in
general. Dissolved oxygen concentrations in lakes are very dependent on site-specific characteristics such as
'Although the information in this report has been funded wholly or partly by the U.S. Environmental
Protection Agency under Contract No. 6841-7288, it does not necessarily reflect the Agency's views, and no
official endorsement should be inferred from it.
7-1
-------
Blumberg
morphometry, prevailing wind conditions, nutrient enrichment, light extinction, and the time history of external
loadings, as reflected in the composition of the sediments. While it is true in general that dissolved oxygen
depletion rates would be expected to increase with increasing temperature, site-specific studies, like the one
conducted here, would probably be necessary to develop actual predictions for a particular lake. Other caveats
to the conclusions include the fact that while the two base years encompass a wide range of baseline anoxic
conditions, they do not represent a full range of climate variability. No change is assumed in nutrient loadings
from the base years, and the analysis does not incorporate the estimated drop in lake levels expected (see Cohen,
1987). Lower lake levels would reduce the volume of the lower layer in lake Erie, increasing eutrophicatioa
The models were not run for the winter, but the sensitivity of results to higher water column temperatures in the
spring was tested and it was found that no significant difference resulted.
In future efforts, it may be possible to estimate the reductions in the total phosphorus loading that
would be required to return the lake to its present conditions. However, simulations with the water quality
model of five years or longer need to be made so that the long-term response of the sediment can be properly
incorporated. These simulations, once conducted, can be analyzed to establish target loadings for phosphorus
that would eliminate the anoxia in the Central Basin.
7-2
-------
Blumberg
INTRODUCTION
Atmospheric accumulations of greenhouse gases, of which carbon dioxide (CO2) is the major
constituent, has led to concern that the earth's climate may be undergoing a global warming in response to these
accumulations. An assessment by Schlesinger and Mitchell (1987) suggests that if current trends in the emissions
of these gases continue, the earth could experience a global mean warming of 2JJ to 4.2°C in the next century
or so. It becomes important to identify and quantify the potential impacts of the possible climate change so that
alternative strategies may be developed to cope with the wanning. This study addresses a very specific issue
within the realm of water resources. It seeks to determine how climate warming will change the present water
quality in Lake Erie, a valuable natural resource. The conclusions of this study are more general and could
perhaps lead to results that have Great Lakes-wide implications, especially for certain embayments and nearshore
areas.
The effect of increases in nutrient inputs to all the Great Lakes, and to Lake Erie in particular, has been
to increase the production of algal biomass. When the lake thermally stratifies with the formation of a
thermodine, the lower part of the lake, the hypolimnion, is isolated and the dissolved oxygen (DO) concentration
begins to decrease owing to the respiratory consumption of oxygen by the algae. If the algae population is
excessive, there may be enough respiration to deplete the oxygen completely causing anoxic conditions that are
fatal to all aerobic organisms including fish. Since the joint U.S.-Canada agreement on Great Lakes Water
Quality chose the elimination of anoxia as the goal for phosphorus control for Lake Erie (International Joint
Commission, 1978), this study will use changes in DO as the measure of changes in overall water quality.
The consequences of increased atmospheric temperature may be dramatic since changes in surface
heating to the lake can lead to different lake stratifications. The extent of oxygen depletion is very sensitive to
the depth at which the thermodine develops. If the thermodine sets up nearer the surface, the hypolimnion
is deep, more oxygen is trapped and the depletion is less intense. If, however, the thermodine sets up nearer
to the bottom, then the volume of the hypolimnion is reduced and depletion will occur sooner. The actual
occurrence of anoxia is a balance between the rate of oxygen depletion and the destratification of the lake in the
fall. Hence, if thermodines set up nearer to the bottom for longer periods of time, anoxia becomes a more
probable event. The warmer lake temperatures could also lead to anoxia or to at least lower levels of DO by
increasing the rate of bacterial activity in the sediments, by increasing the biological productivity and respiration
in the water column, and by decreasing dissolved oxygen saturation values. Thus, the reduction in the area of
anoxia that has been achieved via the phosphorus control measures instituted at (very) large cost may be reversed
by a systematic warming of the lake.
The purpose of this investigation is to examine the relationships between climate changes and the water
quality of the Central Basin of Lake Erie. Data from 1970 and 1975 observations in the Central Basin form the
basis for much of this investigation. These years form a unique set because, in terms of water quality, these
years were quite different. There was considerable anoxia present in the deeper portions of the lake in 1970,
while the DO distribution in 1975 showed no areas where the oxygen had been depleted. The principal
difference between 1975 and 1970 was the shallow depth at which the thermodine developed in 1975. In
addition, the use of these two particular years makes possible the direct application of a Lake Erie eutrophication
model. The model, described in Section 2, has been calibrated and verified for these years. The use of different
years in this analysis probably would not change the condusions of this study because 1970 and 1975 bracket the
range of possible conditions with respect to dissolved oxygen.
7-3
-------
Blumberg
METHODOLOGY
Two existing, well-tested modeling frameworks will be used in the analysis. To estimate the warmer
water temperatures and the thermocline variation produced by climatic changes, a thermocline model of the
Central Basin of Lake Erie has been built. The relevant quantities emanating from this model are then input
to an already calibrated Lake Erie eutrophication model through the use of a consistent model coupling
procedure. Once the various climate change scenarios are identified, simulations and analyses can be
performed.
Thermocline Model
There is little doubt that turbulent vertical mixing of heat is the primary factor in determining the vertical
temperature and pollutant structure of a lake. The discussion that follows describes a thermocline model which,
to good approximation, can predict the vertical distribution of mixing, currents, and temperature in a lake in
response to winds and surface heating/cooling.
The most important cause of motion and mixing in a lake is the wind blowing at the surface. The water
column is usually thermally stratified, and as a result, causes the vertical mixing due to winds to be limited to
the surface layers in the majority of cases. In narrow lakes, the currents in the surface layers are in the same
direction as the wind. Below the surface, wind-mixed layer, there is a compensating and oppositely directed
current pattern established through conservation of volume. In wide lakes, the structure of the surface currents
and deep compensating flows are more complex and considerably modified by depth variations. If the winds are
strong and persistent enough, and/or the vertical stability of the water column is weak enough, the thickness of
the wind-mixed layer will be gradually increased and the wind-induced currents spread out over a larger portion
of the water column. The vertical distribution of currents and turbulent mixing, and the bottom stress, can be
computed fairly well from a knowledge of the surface wind stress in conjunction with a one-dimensional (vertical
direction) thermocline model.
The area-averaged, governing equations, assuming that the Rossby number is sufficiently small so that
the nonlinear acceleration terms can be discarded, are:
(1) A «£ -
A(BP).
(2) A « + fUA - % [A (KM + „) g] +
and
P) * JJ - fc I* "H * »> ft
where z and t are the vertical space coordinate and time, A(z) is the horizontal area U, and V are the velocities
in the east-west and north-south directions, respectively, 9 is the temperature, and pQ is the fluid density. The
surface compensating return flow at depth is produced through the right-hand-most terms in Equations (1) and
(2), where A(BP) is the difference in pressure between the east and west (north and south) ends of the basin
7-4
-------
Blumberg
integrated along their respective north-south (east-west) end boundaries. These pressure terms are obtained by
appealing to the steady form of the vertically integrated continuity equations, which imposes the condition that
the vertical integrals of U and V vanish. The use of the steady form basically assumes that both the surface
gravity and internal waves generated by the end boundaries instantaneously propagate to the center of the basin
and set up the sea surface and pycnocUne. Currents that occur on time scales greater than a day or so are thus
properly accounted for, while higher frequency motions are probably not.
The vertical turbulent mixing coefficients for momentum and heat are K.. and 1C., and are computed
from the turbulence closure submodel of Mellor and Yamada (1982). The closure submodel contains
nondimensional empirical constants that are fixed by reference to a small subset of the available laboratory data
and, as such, are independent of the particular thermocline model application. A background mixing, to
presumably account for internal waves and other unresolved physical processes, is denoted as v and is equal to
0.01 cmz/s for this application. The boundary conditions at the surface are:
(4) 'o (KM + "> I S • If
(» <*. + „) fz - H
[w wl
r , r
« yj
and at the bottom:
(6) *
212 ay
3z ' 32
r b bi
• IV 'yj
(7) fir j. ..\ 5A
(KH * "> S - 0
where rw and 1** denote the wind stress at the surface and the frictional stress at the bottom, H denotes the
surface heat flux, and where the subscripts x and y denote components in the easterly and northerly directions,
respectively.
No heat transfer to the sediments is permitted by the use of Equation (7) although it has been pointed
out by Heinrich et al. (1981) that such a heat transfer can have a significant effect on near-bottom temperatures
of shallow lakes. The process of heat transfer between the sediments and overlying waters is poorly understood
and is therefore not considered here. The bottom stress is obtained by matching the computed near-bottom
velocity to that of the logarithmic law of the wall. This approach for deducing currents, mixing and bottom
stresses has been used successfully many times in the literature (see for example, Kraus, 1977; Martin, 1985).
A variation of this approach, whereby the temperature structure is computed without calculating the details of
the velocity fiefd, has been successfully used in Lake Erie by Lam and Schertzer (1987). For details concerning
the numerical solution technique used here, the reader is referred to Blumberg and Mellor (1983).
Lake Erie Eutrophication Model
The Lake Erie eutrophication model consists of a set of mass balance equations that quantify the mass
transport and kinetic interactions of the biota (phytoplankton and zooplankton), nutrients (phosphorus, nitrogen
and silica), and computes the DO consequences that result from these reactions. The model is fully documented
(Di Toro and Connolly, 1980) and has been employed by others for more detailed examinations of phosphorus
availability (De Pinto et al., 1986). Similar models have been developed and applied to other Great Lakes
settings (Thomann et al., 1975; Bierman, 1976; Lam et al., 1983). In a recent study, Di Toro et al. (1987) have
presented a retrospective analysis of the Lake Erie model performance over a 10-year period. That analysis
further supported the model's predictive capability.
7-5
-------
Blumberg
The model is based upon a segmentation of Lake Erie into volumes that represent the epilimnion,
hypolimnion, and an active sediment layer of the three major basins (Figure 1). The active sediment layer is
explicitly included to more realistically account for the oxygen demand exerted by the sediment (SOD). The
kinetics employed are designed to simulate the annual cycle of phytoplankton production, its relation to the
supply of light and available nutrients, and the effect on DO. The calculation is based on formulating the
kinetics that govern the interactions of the biota and the available and unavailable forms of the nutrients, and
applying these kinetics to the regions of Lake Erie within the context of conservation of mass equations. The
15 variables for which these calculations are performed are:
Phytoplankton
1. Diatom chlorophyll-a
2. Non-Diatom chlorophyll-a
Zooplankton
3. Herbivorous zooplankton carbon
4. Carnivorous zooplankton carbon
Nitrogen
5. Detrital and dissolved organic nitrogen
6. Ammonia nitrogen
7. Nitrate nitrogen
Phosphorus
8. Unavailable phosphorus
9. Soluble reactive phosphorus
Silica
10. Unavailable silica
11. Soluble reactive silica
Carbon, Hydrogen, Oxygen
12. Detrital organic carbon
13. Dissolved inorganic carbon
14. Alkalinity
15. Dissolved oxygen
Comparisons of the model results with extensive field data from 1970 and 1975 demonstrated that the
calculation can reproduce the major features of the seasonal distribution of phytoplankton and nutrients over a
range of observed concentrations. The fact that the Western, Central, and Eastern Basin distributions are all
reasonably well reproduced, using the same kinetic structure and coefficients, suggests that the calculation has
a certain generality and can reproduce conditions as distinct as those in the Western and Eastern Basins.
The dissolved oxygen comparisons for 1970 and 1975 in the Central Basin are shown on Figure 2. The
agreement is quite good with the computation duplicating all the major features of the temporal oxygen
distribution. As can be seen, the epilimnion dissolved oxygen follows the saturation oxygen concentration very
closely during the entire year except for a small oxygen peak of supersaturation appearing in early July. This
is due to photosynthetic production resulting from the increase in phytoplankton chlorophyll taking place at this
7-6
-------
Blumberg
EPILIMNION SEGMENTS
0-BOTTOM
0-17 METERS
HYPOLIMNION SEGMENTS
17 METERS-BOTTOM
..17-22 METERS
u-22 METERS-BOTTOM
SEDIMENT SEGMENTS
ALL 5 CM. DEEP
Figure 1. Lake Erie model segmentation of Western, Central, and Eastern basins: water segments 1-6,
sediment segments 7-10.
7-7
-------
1970
1975
12.00
6.00
0.00
OtSSOt VfD
OXYGfN
- CENTRAL BASIN
EPILIMNION
e>
>-
X
o
o
LJ
J'F'M'A'M'J'J'A'S'O'N'D
12.00
<2 6.00
o.oo
_ DISSOL VfO OXYGfN
- CENTRAL BASIN
HYPOLIMNION
J'F'M'A'M'J1 J'A'S'O'N'D1
DISSOLVED
OXYGfN
CENTRAL BASIN
EPILIMNION
JASOND
- CENTRAL BASIN
HYPOLIMNION
DISSOLVED
OXYGfN
J'F'M'A'MV J'A'S'O'N'D
1
Figure 2. Comparison of predicted and observed surface and bottom layer dissolved oxygen distributions
for 1970 and 1975. The data are plotted as the mean over the layer +.1 standard deviation.
-------
Blumberg
time. In the hypolimnion, the low oxygen values during the summer months of 1970 are caused by a combination
of phytoplankton decay within the hypolimnion and the oxygen demand exerted by the sediment. The increasing
temperature, which results in increasing reaction rates, the sulking of the phytoplankton, which have grown in
the epilimnion, and stratification, which inhibits transfer of oxygen from the epilimnion to the hypolimnion, all
contribute to the oxygen decline. The decline continues until overturn in early fall At this time, there is a rapid
increase in the hypolimnion dissolved oxygen, which continues to the end of the year. The higher oxygen
concentrations in 1975 are basically due to the larger volume of the hypolimnion in that year. As shown by Di
Toro and Connolly (1980), the sediment oxygen demand, which is a constant on an areal basis, would result in
a smaller volumetric depletion rate in 1975. The maximum anoxic area of the Central Basin for 1970 is reported
to be 6600 km , and this agrees well with the September maximum in the model of 6800 km2.
Thermocline/Eutrophication Model Interaction
To use the eutrophication model in a predictive mode for the analysis of the various climate scenarios,
three ingredients are required. The first is the expected water temperatures since they affect all the kinetic rates
in the eutrophication model. The second is the depth of the thermocline and hence, the volumes of the
epilimnion and hypolimnion. The third ingredient is the bulk vertical dispersive exchange between the epilimnion
and the hypolimnion. The first two ingredients can be directly extracted from the thermocline model results.
The dispersive exchange, on the other hand, needs to be computed using the temperature fields from the
thermocline model run, averaged over the epilimnion and the hypolimnion, in conjunction with a temperature
balance equation similar to Equation (3) structured for a two layer system.
The area integration of Equation (3) over the hypolimnion gives,
80 80
<8> VH 5T ' Ai Ki «T
where VH and 0., are the volume and volume-averaged temperature of the hypolimnion, respectively, and where
i denotes the interface between the epilimnion and the hypolimnion. The interfacial dispersion coefficient, K,,
is readily computed from Equation (8) given the results from a thermocline model run.
Climate Scenarios
Three simulations of the changes in the equilibrium climate induced by a doubling of the atmospheric
CO2 (2xCO2) concentration have been performed using atmospheric general circulation models (GCMs). These
simulations were performed with the Goddard Institute for Space Studies (GISS) GCM by Hansen et al. (1984),
the Geophysical Fluid Dynamics Laboratory (GFDL) GCM by Manabe and Wetherald (1986), and the Oregon
State University (OSU) GCM by Schlesinger and Zhao (1988). Climate scenarios in the form of monthly
averaged data over an annual cycle have been obtained for this study from the three GCM simulations. The
relevant data are selected for the grid point closest to Lake Erie. The results of control runs using the present-
day COu concentration (IxCCX) were also provided as a means of quantifying the predicted climate changes.
A transient scenario using the uISS model was also included and involved the results of two 100-year runs in
the form of decadal mean value differences (for each month) for the last 80 years of the simulations.
To use the GCM results in this study, the results were expressed as changes in surface air temperature
and changes in surface windspeed. Figure 3 and Table 1 contain these values for each of the 2xCO, and
transient scenarios investigated. The values represent the deviation from the control, or base runs (lxCO2).
For the 2xCO, scenarios the surface air temperature changes were calculated by subtracting the control run
values from the 2xCO2 run values. The transient scenario decadal changes were calculated from the ratio
between the temperature (in °K) of the scenario and the temperature (in °K) of the base run. The temperature
differences from the GISS and GFDL 2xCO, scenarios shown on Figure 3 are similar with the exception of June
and July when the GFDL values are double those of GISS. Both 2xCO2 model runs show an increase in average
monthly surface temperature of 3 to 7°C. The OSU 2xCO2 scenario, on the other hand, exhibits much less
7-9
-------
Blumberg
Change in Air Temperature
GIBS 2xC02
6FDL 2xC02
6.00
OSU 2xC02
Change in Wind Speed
GISS 2xC02
6FOL 2xC02
a.oo
i.OO
0.00
0)
» » » »
XXX
OSU 2xC02
Figure 3. The monthly changes in surface air temperature and wind speed resulting from the three climate
model simulations.
7-10
-------
Table 1. Change in Air Temperature and Wind Speed for the Transient Scenario A
Chang* in Air Temperature (*C)
Scenario/Month
Scenario A (1960-1989)
Scenario A (1990-1999)
Scenario A (2000-2009)
Scenario A (2010-2019)
Scenario A (2020-2029)
Scenario A (2030-2039)
Scenario A (2040-20*9)
Scenario A (2050-2059)
Scenario/Month
Scenario A (1980-1989)
Scenario A (1990-1999)
Scenario A (2000-2009)
Scenario A (2010-2019)
Scenario A (2020-2029)
Scenario A (2030-2039)
Scenario A (20*0-20*9)
Scenario A (2050-2059)
January
1.35
0.81
0.27
1.35
1.08
2.96
2.42
5.38
January
-0.67
0.14
0.26
-0.46
0.38
0.19
0.19
0.20
February
0.27
-0.27
1.08
2.98
2.17
4.07
3.25
5.69
February
-0.83
-O.S9
-1.02
-1.41
-0.74
-0.11
-0.56
-0.42
March
0.27
0.00
0.27
1.6S
2.47
2.47
4.39
4.67
March
-0.60
0.09
0.64
0.16
0.03
0.35
-0.28
0.2S
April
0.84
0.00
1.12
1.68
2. 52
3.37
3.65
3.65
April
-0.29
-0.21
-0.54
0.07
-0.45
0.29
0.04
-0.51
May
0.85
0.85
0.85
1.13
1.70
2.55
3.40
4.53
Change
May
-0.23
-0.09
0.14
-0.20
-0.04
-0.46
-0.46
0.08
June
-0.29
0.87
0.58
O.SB
1.15
2.31
2.31
3.17
in Wind
June
0.19
0.46
0.16
0.17
0.52
0.19
-0.25
0.06
July
1.17
1.46
1.46
1.75
1.75
2.92
2.92
4.09
Auituat
0.29
0.29
0.29
0.59
2.36
1.47
2.06
5.30
September
0.00
0.00
0.29
0.87
1.74
3.19
4.06
4.35
October
MMM^HMMM
-0.84
0.56
0.28
0.84
0.84
1.40
3.09
4.21
Hovember
0.00
1.37
1.09
2.46
1.91
2.46
3.83
5.74
December
-0.54
1.35
1.08
3.24
2.16
3.24
4.05
5.94
Average
0.28
0.61
0.72
1.59
1.82
2.70
3.29
4.73
Speed (ni/s)
July
-0.37
-0.34
0.30
-0.26
0.08
-0.45
-0.57
-0.08
August
0.11
-0.16
0.00
0.10
-0.21
-0.11
0.18
-0.06
September
0.74
0.46
0.54
0.51
0.45
0.85
0.58
1.16
Octobe^
0.27
-0.06
-0.38
-0.40
0.05
-0.05
-0.2*
0.11
Mov amber
1.66
0.71
0.77
0.34
0.81
0.83
0.65
0.83
December
0.57
0.65
0.94
0.09
0.94
0.84
1.08
0.46
Average
0.05
0.09
0.15
-0.11
0.15
0.20
0.03
0.17
5
-------
Blumberg
temperature change, about half of the GISS and GFDL cases. Temperature changes of 2 to 4°C are thus
predicted. The transient scenario average monthly changes in air temperature increase from the 1980-1989
decade to the 2050-2059 decade. At the end of the eighth decade (2050-2059X temperatures increased between
4 and 6°C from the control run.
The wind data available from the GCM data files are vector mean wind speeds instead of the more
appropriate scalar mean wind speeds. Vector mean speeds are often significantly less than the true mean wind
speed and will thus underestimate wind-induced mixing. This error is somewhat reduced here because the winds
in the GCMs exhibit little directional variability within each scenario. The changes in wind speed for the 2xCO2
scenarios shown on Figure 3 are small and typically negative, suggesting that the winds will become weaker
under the various climate scenarios. The OSU 2xCO, scenario, it should be noted, has larger winds occurring
in the critical spring and summer periods; increases of 1 to 1.5 m/sec are indicated. The largest increase in wind
speed is about 1.7 m/sec, which occurs in November of Transient Scenario A (1980-1981) as shown in Table 1.
The heat fluxes and wind stresses required as input by the thermodine model were computed according
to:
(9) q.q *£«
O 01
and
do) r - r0 + m AW
where Q is the heat flux, r is the wind stress, T is the air-lake temperature difference, and W is the wind speed.
The o subscript denotes the base case, either 1970 or 1975, and A denotes the 2xCO,-lxOO, difference. The
rate of change of heat flux Q from the lake with respect to the air-lake temperature difference has been
computed by Gill (1982) based on the ideas of Haney (1971) for the world ocean. It appears that a value of 32
Watts m'2 is appropriate for the latitude of Lake Erie. The rate of change of wind stress with respect to wind
speed is readily estimated to be 3 x 10"3 gm cm""
7-12
-------
Blumberg
RESULTS
Before proceeding to an assessment of the thermal and water quality response of Lake Erie to the
various climate warming scenarios identified in the previous section, it becomes important to assess the skill of
the thennocline model itself. The water quality model as utilized by Di Toro and Connolly (1980) used
temperatures, thennocline depths, and bulk dispersion coefficients deduced from observations. Projections to
future conditions are thus not possible without some predictive capability for these important inputs. In this
section, the skill of the thennocline model is demonstrated, and then the model is used to estimate what the
thermal response of Lake Erie would be under the various scenarios. Water quality projections follow using the
computed thermal regimes for the various climate warming scenarios.
Verification of the Thermocline Model
The thennocline model presented above was applied to Lake Erie for the years 1970 and 1975. All
simulations started on April 1 and terminated on December 31 to avoid considering the extensive ice cover which
normally occurs during January through March. There are 40 vertical levels in the model with finer spacing near
the surface than in the deeper regions. A time step of 30 minutes is used The model is forced by hourly
estimates of the surface wind stress and by monthly estimates of the surface heat flux. A continuous series of
hourly wind speed and direction observations for 1970 and 1975 were obtained at the Erie, Pennsylvania, and
Cleveland, Ohio, airports. These observations were corrected from overland to overwater winds (see Schwab
and Morton, 1984) and turned into wind stresses using a quadratic drag law and a drag coefficient based on the
work of Large and Pond (1981). The surface heat fluxes have been calculated by Di Toro and Connolly (1980)
for 1970 and 1975 on a monthly basis using estimates of the heat storage. These values, illustrated on Figure
4, were linearly interpolated to form a set of daily values and are consistent with those computed by Derecki
(1976) and Schertzer (1987) using hydro-meteorological data. The initial temperature profile was a vertically
homogeneous 2.0°C.
A comparison of the predicted and observed surface and near-bottom temperatures for each of the
two years is shown on Figure 5. The predicted and observed temperatures agree fairly well throughout the
annual cycle. It appears that the model misses the onset of stratification by a month or so in 1970, while the
correct onset is predicted in 1975. The peak in the hypolimnion temperatures, which occurs in
September/October, well after the surface peak temperatures, is modeled quite well. The rapid fall deepening
of the epilimnion due to surface cooling and convection is also captured by the model However, the tuning of
this overturn event is off by two to three weeks with the model being later than the observations. The model
appears to reproduce the maximum stratification in both years. In 1970, the data for August (typically the most
stratified month) show bottom and surface temperatures ranging from 11 to 22°C for a temperature difference
of AT =11°C. The 1970 model values are 12 to 24°C for a temperature difference of AT «12°C. In 1975, the
data for the same monthly time period shows bottom and surface temperatures ranging from 13 to 22°C, for a
AT « 9°C. The 1975 model values are 13 to 20°C, for a AT - TC. Thus, while the model is not perfect, it does
capture most of the large observed top to bottom difference and it does so at the proper time of the year.
Another measure of model skill is to compare vertical profiles between prediction and observation. Such
a comparison is made on Figure 6 where the average temperature in two meter sections of the Central Basin
are shown for July cruises in 1970 and June cruises in 1975. The model reproduces the vertical structure
observed in 1970 very well However, the model has a tendency to produce slightly wanner bottom temperatures,
perhaps because there is no mechanism in the model to transfer heat to the sediment. The comparison of the
1975 case is not as good The month of June is a transitional period when the lake goes from vertically mixed
to stratified. Errors in the timing of the onset of stratification can lead to the kinds of comparisons indicated
for 1975. The seasonal variation of the vertical structure of the lake is shown in the upper portions of Figure
7 for 1970 and Figure 8 for 1975. The mixed layer is shallow in summer owing to intense solar heating and
typically weak winds, but deepens rapidly in the fall The isotherms deepen slowly in the spring and summer
owing to downward diffusion and rise rapidly in fall owing to convection. It has been observed that 1970, with
its stronger winds and larger surface heating, has a greater thennocline depth than 1975.
7-13
-------
Blumberg
COMPARISON OF CALCULATED HEAT FLUX FUNCTIONS
FOR 1970 AND 1975
LAKE ERIE CENTRAL BASIN
400 -
CALCULATID
• •— fSTIMATCD
1975
—~ CALCULATED
Figure 4. Calculated heat flux (cal/cm2-day) to the central basin using observed temperatures for each
cruise. During the winter period, the net heat flux shown is estimated.
7-14
-------
Blumberg
« - Epilimnion
- Hypolimnion
JFMAMJJASOND
- Epilimnion
a - Hypollmnion
Figure 5. Comparison of predicted and observed surface and bottom temperatuers for 1970 and 1975 over
the annual cycle. The data are plotted as the mean over the layer ±1 standard deviation; lines are
the computations.
7-15
-------
Blumberg
TEMPERATURE PROFILES
1970
1975
TEMPERATURE (°C)
o(
5
10
20
25
) 10 20 w
1 1 D
X
X
X
?
D J
—
c
5
10
15
20
25
) 10 20
X
HDH
D
X
X
h"~C3r~l
1 — OM
- HOHX*
F^J> fc^
I^Jt V
|
- HCH
D-DATA: 7/28-7/29
X-MODEL
D-DATA: 6/24-6/29
X-MODEL
Figure 6. Comparison of the predicted and observed thermal vertical structure for July 1970 and June 1975.
The data are plotted as the mean over a two-meter layer ±1 standard deviation. Much of the June
1975 error can be attributed to errors in the timing of the onset of stratification.
7-16
-------
-5
io
-20
-25
1970
BASE CASE
Blumberg
M
A M J J A
MONTH OF YEAR
N
Figure 7. The seasonal variation of the thermal vertical structure for the 1970 cases. Contour in °C. Dashed
line is the mixed layer depth for the base case.
7-17
-------
Blumberg
-25
M
M J J A
MONTH OF YEAR
N
Figure 8.
The seasonal variation of the thermal vertical structure for the 1975 cases. Contour in °C. Dashed
line is the mixed layer depth for the base case.
7-18
-------
Blumberg
To explain the difference between the observed and computed thermal structure, it is important to
realize that there are no coefficients available in the thennocline model that can be adjusted in order to improve
model performance for these applications. One must look toward improving model physics or assembling better
model forcing data While some improvements in the exchange processes between the water column and the
sediment can be envisioned, the model physics based on areally averaged dynamics appear well established for
basin-wide problems. On the other hand, the data available for model forcing, especially the wind fields, are
subject to considerable uncertainty. Mid-basin wind stresses are estimated from observations of wind speed and
direction at weather stations around the lake. The model error in this study is probably due to the use of only
two stations. Since the timing of the formation and subsequent destruction of the thermocline are dependent
on details in the wind field, the use of perhaps six stations around the lake would be necessary to incorporate
the spatial variability of the wind fields into the mid-basin average.
Thermal Response to
The calibrated and skill assessed thermocline model was run using the climate warming scenarios
presented earlier. The evolution of the vertical structure over an annual cycle is illustrated on Figures 7 and 8
for the GISS, GFDL, and OSU 2xCO^ scenarios. It is evident from these results that the 2xCO2 scenarios
produce a warmer, longer lasting warming regime in the lake. The heating season, as defined by the presence
of surface water with temperatures greater than or equal to 18°C, lengthens about 40 days using 1970 as the base
case and almost doubles to between 64 and 135 days using 1975 as the base. When comparing scenarios to then-
base cases, the climate scenarios using the 1975 base case begin heating some four weeks earlier and heat six
weeks longer than the base, while those using the 1970 base case begin heating about two weeks earlier and heat
approximately two weeks longer than the base. The maximum surface temperature increases by about 5°C in
all scenarios.
The thermocline, defined as the depth of maximum vertical temperature gradient in August (the
warmest month), becomes shallower by about 2 m for the GISS and GFDL 2xCO2 cases. The OSU 2xCO2
scenario results in a slightly deeper thermocline in spite of its smaller air temperature change. Apparently the
increase in wind speeds can mix the extra surface heating deeper into the water column than can the lesser wind
speeds and greater surface heating of the GISS and GFDL 2xCO2 scenarios. The change in volume of the
hypolimnion for the various scenarios is tabulated in Table 2. Because of the shape of the Central Basin, a 1-
to 2-meter rise in the thermocline leads to 50 to 100 percent increases in volume. The 1975 based OSU 2xCO2
scenario with its deeper thermocline actually has a 20 percent smaller hypolimnion volume. Table 2 provides
a summary of the results from all the scenarios. The use of the transient scenarios, it will be noted, produces
decadal mean values that exhibit much variability. The trend, however, is clearly toward warmer waters
throughout the April to December period
The change in epilimnion and hypolimnion temperature is shown on Figure 9. Most of the extra heat
is confined to the epilimnion where changes are on the order of 4°C. The waters of the hvpolimnion increase
about 1 to 2°C with an even greater increase of about 3 to 4°C occurring in the later part of the year. The
temperature difference across the epilimnion-hypolimnion interface, therefore, increases by 2 to 3°C, as shown
in Table 2. This difference is certain to reduce the dispersive exchange between the upper and lower layers of
the lake.
The results from a number of sensitivity runs have also been analyzed. These runs included running the
base cases with different combinations of forcing. For example, experiments were run where the 1970 winds were
used in conjunction with the 1975 heating and vice versa. As illustrated in Table 2, the runs showed that the
location of the thermocline is a delicate balance between wind stirring and surface heating. Increasing the wind
stirring by having stronger winds causes the thermocline to form deeper and at a later time. Conversely, weaker
winds or larger surface heating will produce a shallower thermocline at an earlier time. These ideas are further
substantiated by the differences between the OSU and GISS/GFDL 2xCO2 results. Other cases have been
conducted to address the findings of Assel (1988) who showed that under the 2xCO2 scenarios it is likely that
the central basin of Lake Erie will be void of its now characteristic winter ice cover. Results from simulations
with initial temperature profiles of 4°C and 8°C indicated that the summertime conditions for these cases were
identical to those of the benchmark case, which used a 2°C initial profile.
7-19
-------
Table 2. Summary of Climate Scenario Calculations
Mixed3
Layer
Run Depth
1970 Base Case
1975 Base Case
1970 GISS (2xC02)
1970 GFDL (2xC02)
1970 OSU (2xC02)
1975 GISS (2xC02)
1975 GFDL (2xO>2)
1975 OSU (2xC02)
1970 Winds/1975 Heating
1975 Winds/1970 Heating
1970 Transient
Scenario A (1980-1989)
Scenario A (1990-1999)
Scenario A (2000-2009)
Scenario A (2010-2019)
Scenario A (2020-2029)
Scenario A (2030-2039)
Scenario A (2040-2049)
Scenario A (2050-2059)
14
13
10
12
14
10
11
14
15
12
12
13
14
13
14
12
12
12
Length
Start
Date
6/23
7/10
6/05
5/30
6/04
6/14
5/31
6/15
7/14
6/19
6/21
6/20
6/20
6/18
6/17
6/11
6/07
6/05
of Heatine Season0
End
Date
10/14
09/12
10/31
11/05
10/27
10/20
10/31
10/14
09/09
10/14
10/10
10/14
10/15
10/16
10/19
10/22
10/27
11/03
Duration
Davs
113
64
148
159
145
128
153
121
57
117
111
116
117
120
124
133
142
151
Temperature
Maximum
ro
25
20
28
30
27
24
27
23
19
26
25
25
25
25
26
26
26
29
Temperature0 Vertical** Increase ine
Difference Dispersion Hypolimnion
(°C) (cm^s'l) Volume (X)
9 0.02
5 0.12
10 0.02 90
12 0.02 50
9 0.02 10
7 0.06 40
11 0.07 30
6 0.09 -20
aBased on depth of the maximum vertical temperature gradient during August.
°Based on presence of water with temperatures > 18"C at the surface.
cBased on averages over the epilimnion and the hypolimnion.
dfiased on August values.
eRelative to base case.
CO
-------
1970
EPILIMNION
•IBS
1970
HYPOLIMNION
•ra. 2x0)9
OFDL
g a
I!?
09U
osu
— •.!
— *.m
1925.
MSS
s:,
!
•FDL
OM toCOj
09U 2XC02
JFMAMJJASOND
tH-
• -**1 • • • *
Figure 9. The monthly difference in temperature (°C) between the climate scenario and its base case for the
epilimnion (left) and the hypolimnion (right).
sr
-------
Blumberg
Water Quality Response to Scenarios
Given the epilimnion and hypolimnion temperatures resulting from the climate warming scenarios, it
is a straightforward task to estimate how water quality responds to the warming. The dispersion coefficients
computed from Equation (8) that are consistent with the new temperatures are shown in Table 2. There is little
scenario to scenario variation in the coefficients for the 1970-based cases. The temperature difference between
the epilimnion and hypolimnion is great enough in every 1970-based case to prevent mixing. On the other hand,
the 1975-based cases show much scenario to scenario variation. The increase in the temperature difference
reduces the vertical exchange by a factor of 2 or so, from 0.09 cm2/s to 0.06 cm /s, during times of maximum
stratification.
The projected2 epilimnion and hypolimnion DO for the GISS, GFDL, and OSU scenarios are shown
on Figure 10 along with the observations for 1970 and 1975. The epilimnion DO worsens by about 1 mg/L
in every case with respect to the observed values. This decrease is primarily due to the change in the saturation
level for DO. For example, a 4°C increase in epilimnion temperature produces an approximately 0.6 mg/L
decrease in the level of DO for each scenario. The hypolimnion DO, as shown on Figure 10, decreases
dramatically, losses of 1 to 2 mg/L being typical under the various scenarios. The 1975-based cases, for which
the observation did not reach the anoxic level, show losses of up to 5 mg/L of DO. The 1970 observations
have very low levels of DO already, so a small loss leads immediately to anoxia. It is interesting to note that the
GISS 2xCO2 scenario using 1970 as the base year leads to a three-month period of higher levels of DO.
The overall decrease in DO can be explained by noting that the sinks of oxygen in the hypolimnion, the
benthic respiration rate due to the decay of organic material, and bacterial metabolism of organisms in the water
column are strong functions of temperature. In fact, the increase in hypolimnion temperatures due to the
climate warming increases the SOD, the largest part of the oxygen sinks, considerably overwhelming any decrease
in dissolved oxygen saturation. On the other hand, the volume of the hypolimnion has increased (see Table 2),
and there is now a greater volumetric reservoir of dissolved oxygen available to meet water column and benthic
oxygen demand. The results of the scenario simulations indicate that the volumetric depletion rate becomes
greater despite increases in the volume of the hypolimnion, which in the absence of any change in SOD, would
result in a smaller volumetric depletion rate and higher levels of DO. Figure 11 illustrates the volumetric
depletion rates for the model scenarios and the base cases. An experiment, whereby the thermal structure of
the 1970-based GFDL 2xCO, scenario was used in conjunction with the observed, and somewhat cooler, 1970
temperatures showed that hign levels of DO would be present. This suggests that the major decline in the lake's
water quality is due to the wanner lake temperatures, which increase the rate of bacterial activity in the sediment
enough to produce apparently significant increases in the SOD. The OSU 2xCO2 scenario produces the worst
DO distributions because the thermocline depth changed little while the SOD became much greater, again due
to the elevated temperatures. It appears that no matter at what depth the thermocline forms, the increased lake
temperatures will lead to degradation in the water quality.
The change in the area! extent of anoxia, defined here as 0 mg/L, in the Central Basin can be
computed using the empirical method developed by Di Toro and Connolly (1980). Figure 12 shows, as monthly
averages, the percentage of the Central Basin that observationally is anoxic and the percentage that is projected
to become anoxic under each of the scenarios. The 1970-based scenarios produce areas of anoxia that
encompass considerable portions of the lake. The areas of anoxia also occur two weeks or so sooner than those
of the base case. The 1970 observations indicate only a month of severe anoxia, and only about 65 percent of
the basin is affected. The OSU 2xCO2 scenario, with the smallest change in air temperature, produces the worst
case in terms of extra anoxia for both the 1970- and 1975-based cases. It should be remembered that under the
present climate conditions 1975 had ample DO, and now with the warming both the GFDL and OSU 2xCO2
1975 based scenarios are showing large regions (20 to 50 percent) of anoxia for periods of a month and greater.
2Tbe projected values are obtained by adding the difference between those values computed using the
2xCO2 scenarios and the lxCO2 scenarios to the original observations.
7-22
-------
1970
EPILIMNION
oi ss
GFOL
osu
OFDL
OSU
197O
e.o
n.e
HYPOHMNtON
GISS
6FDL
OSU
GFOL
OSU
Funire 10 The monthly averaged dissolved oxygen (mg/L) or the epilimnion and hypolimnion of the Central
Basin. The hatched bars denote the projections, while the clear bars are observations. The
difference between these bars is the climate warming-induced change.
2
!
-------
Blumberg
o:
70
60
50
40
30
20
10
BASE CASE
6ISS 2xC02
GFDL2xC02
OSU 2xC02
i i i i i
JFM AMJJ ASOND
70
60
50
40
30
20
10
i i i
i i
1975
BASE CASE
GISS 2*C02
GFDL 2xC02
OSU 2xCOe
J I
i l I
J FMAMJJASOND
MONTH
Figure 11. A comparison of the annual cycle of volumetric hypolimnion SOD between base case and the
various climate warming scenarios.
7-24
-------
Blumberg
1970
JULY
AUGUST
BASE CASE
9.8 X
40.6%
0.0%
6XSS
11.7%
80.5%
0.0%
GFDL
22.8%
94.4%
5.9%
OSU
55.3%
100%
28.8%
Figure 12. A comparison of the monthly average percentage of anoxia in the Central
base years and the projections from various climate warming scenarios.
represent the regions of anoxia
Basin between the
The hatched areas
7-25
-------
Blumberg
Since the lake turns over each year, there appears to be little thermal carryover from year-to-year. The
transient scenarios thus become a series of steady-state ones similar to the 2xCO, cases. It is possible then to
make some comments about the resulting water quality for each decade in the Transient series. As Table 2
illustrates, all of the transient scenarios have thermocline depths and heating season durations that are bounded
by the 1970-based 2xCO2 cases. The resulting DO distributions and changes in area! extent of any anoxia are
thus also bounded by the 1970 based cases and the conclusions reached previously apply here. The transient
scenario cases were not run hi the thermocline model for the 1975 base case and, as such, no additional
comments can be provided.
7-26
-------
Blumberg
REFERENCES
Assel, R A., Impact of Global Wanning on Great Lakes Ice Cycles. NOAA/ERL/GLERL Report, Ann Arbor,
Mich., 1988.
Bierman, VJ., Mathematical model of the selective enhancement of blue green algae by nutrient enrichment.
IK Modeling Biochemical Processes in Aquatic Ecosystem. Ed. R. Canale, Ann Arbor, Mich: Ann Arbor
Science, 1-32, 1976.
Blumberg, AJ7., and Mellor, G.L~, Diagnostic and prognostic numerical circulation studies of the South Atlantic
Bight J. Geophvs. Res.. 88. 4579-459^
Cohen, S J. Impacts of CO.-induced climatic change on water resources in the Great Lakes basin. Climatic
CJaogs, 8, 1986, 135-153.
Cohen, S J. Influences of Past and Future Climates on the Great Lakes Region of North America. Water
International 12, 163-169, 1987.
De Pinto, J.V., Young, T.G, and Salisbury, DJL, Impact of phosphorus availability on modeling phytoplankton
dynamics. Hvdrobioloeical Bulletin. 20, 225-243, 1986.
Derecki, JJL, Heat Storage and Advection in Lake Erie. Water Resources Res.. 12, 1144-1150, 1976.
Di Toro, DM., and Connolly, J.P., Mathematical Models of Water Quality in Large Lakes. Part 2: Lake Erie.
USEPA EPA-600/3-80-065, Duluth, Minnesota, 230 pp., 1980.
Di Toro, DM., Thomas, NA., Herdendorf, C.E, Winfield, R.P^ and Connolly, J.P., A Post Audit of a Lake
Erie Eutrophication Model J. Great Lakes Res.. 13, 801-825, 1987.
Gill, A.E., Atmosphere-Ocean Dynai^ic^ Academic Press, New York, New York, 1982.
Haney, R.L., Surface Thermal Boundary Conditions for Ocean Circulation Models. J. Phvs. Oceanogr.. 1,
241-248, 1971.
Hansen, J., Lads, A., Rind, D., Russell, G., Stone, P., Fund, I., Ruedy, R., and Lerner, J., Climate Sensitivity:
Analysis of Feedback Mechanisms. In: CliTiate Processes and ffimate Sensitivity. Qeophvs. Monoer. Ser.. 29,
Eds. J.E. Hansen and T. Takahashi, AGU, Washington, DC, 130-163, 1984.
Heinrich, J., Lick, W., and Paul, J., Temperatures and Currents in a Stratified Lake: A Two-Dimensional
Analysis. LGjejLLakSLBsS., 7, 264-275, 198L
International Joint Commission, Great Lakes Water Quality Agreement of 1978, Washington, DC and Ottawa,
November 22, 1978. International Joint Commission, Windsor, Ontario, 1978.
Kraus, EU, Modelling and Prediction of the Upper Layers of the Ocean. Pergamon Press, Ed. E.B. Kraus,
p. 325, 1977.
Lam, D.C.L., and Schertzer, W.M., Lake Erie Thermocline Model Results: Comparison with 1967-1982 Data and
Relation to Anoxic Occurrences. J. Great Lakes Res- 13, 757-769, 1987.
7-27
-------
Blumberg
Lam, D.C.L., Schertzer, W.M., and Fraser, AS., Simulation of Lake Erie Water Quality Responses to Loading
and Weather Variations. Scientific Series No. 134, National Water Research Institute, Canada Centre for Inland
Waters, Burlington, Ontario, 1983.
Large, W.G., and Pond, S., Open Ocean Momentum Flux Measurements in Moderate to Strong Winds. J.
Geoohvs. Res.. 11, 324-336,1981.
Manabe, S., and Wetherald, R.T., Reduction in Summer Soil Wetness Induced by an Increase in Atmospheric
Carbon Dioxide. Science. 232. 626-628.1986.
Martin, P J., Simulation of the Mixed Layer at OWS November and Papa with Several Models. J. of Geophvs.
Res.. 90, 903-916,1985.
Mellor, G.L., and Yamada, T., Development of a turbulence closure model for geophysical fluid problems. Rev.
Geophvs. and Space Phvs.. 20, 851-875, 1982.
Schertzer, W.M., Heat Balance and Heat Storage Estimates for Lake Erie, 1967 to 1982.
J. Great Lakes Res.. 13, 454-467, 1987.
Schlesinger, M.E., and Mitchell, J.F.B., Climate Model Simulations of the Equilibrium Climatic Response to
Increased Carbon Dioxide. Rev. Geophvs.. 25, 760-798,1987.
Schlesinger, M.E., and Zhao, Z.-C, Seasonal Climate Changes Induced by Doubled COL as Simulated by the
OSU Atmospheric GCM/Mixed-Layer Ocean Model. Climate Research Institute Report, Oregon State
University, Corvallis, Oregon, 84 pp., 1988.
Schwab, D J., and Morton, J A., Estimation of Overtake Wind Speed from Overland Wind Speed: A Comparison
of Three Methods. J. Great Lakes Res.. 10, 68-72, 1984.
Thomann, R.V., Di Tore, D.M., Winfield, R., and O'Connor, D J., Mathematical Modeling of Phytoplankton
in Lake Ontario, 1. Model Development and Verification. Environmental Protection Agency, EPA
660/3-75-005. Corvallis, Oregon, 177 pp., 1975.
U.S. Environmental Protection Agency, Ambient water quality criterion for dissolved oxygen: freshwater aquatic
life. Federal Register 50,15634-15668,1985.
7-28
-------
IMPACTS OF GLOBAL WARMING
ON RUNOFF IN THE UPPER
CHATTAHOOCHEE RIVER BASIN
by
David K. Hains
C. F. Hains, Hydrologjst, Inc.
Northport, AL 35476
Contract No. CR814637
-------
CONTENTS
FINDINGS 8-1
CHAPTER 1: INTRODUCTION 8-3
HYDROLOGY OF A SOUTHEASTERN HEADWATER BASIN 8-3
RELATED REPORTS .8-3
ORGANIZATION OF THIS REPORT 8-3
CHAPTER 2: METHODOLOGY 8-6
THE SACRAMENTO MODEL 8-6
Development of the Model 8-6
Limitations Resulting from the Model 8-8
THE SCENARIOS 8-9
Scenarios Used , 8-9
Issues Resulting from the Scenarios 8-9
CHAPTER 3: RESULTS 8-10
IMPACTS OF CLIMATE CHANGE 8-10
Direction and Magnitude of Changes 8-10
Transient A Scenario 8-13
CHAPTER 4: INTERPRETATION OF RESULTS 8-22
CHAPTER 5: IMPLICATIONS OF RESULTS 8-24
ENVIRONMENTAL IMPLICATIONS 8-24
LIST OF ABBREVIATIONS 8-25
REFERENCES 8-26
APPENDIX A 8-27
-------
Hains
FINDINGS1
This report presents the results of modeling studies of the hydrology and the effects of climate change on
the Apalachicola-Chattahoochee-Flint (ACF) River system in the southeastern United States. Two separate
studies were performed on the system. A detailed hydrologic study was performed on the upper Chattahoochee
River, a small headwater portion of the system. A more general statistical study was done on the entire river
system near its outlet to the Gulf of Mexico to provide flow estimates needed by other researchers.
For several years, atmospheric scientists have been studying the effects of carbon dioxide (CO2) on climate
change using general circulation models (GCMs). For this report results in the form of weather variables from
three of these modeling groups have been used as inputs to a hydrologic analysis to determine the range of
effects which can be predicted in the hydrology of a major southeastern river system as a result of global warming
and climate change driven by increasing CO2 concentrations. Three different GCMs were used to develop
steady-state scenarios, each based on doubling of CO, concentrations in the atmosphere. The three models used
were the Goddard Institute for Space Studies (GBS) model, developed by NASA; the Geophysical Fluid
Dynamics Laboratory (GFDL) model, developed at Princeton University; and the Oregon State University (OSU)
model. One transient scenario was used in the GISS model based on a gradually increasing concentration of
CO2 following the current trend.
The Sacramento Model, developed by the National Weather Service (NWS), was used to model the
hydrology and the effects of climate change for the Upper Chattahoochee River Basin at Buford Dam on Lake
Lanier in Georgia Although the Sacramento Model is intended to be a continuous simulation model, its usual
application is for flood forecasting by the River Forecast Centers of the National Weather Service. As a result,
existing calibration data for most watersheds have been optimized to give most accurate results for flood peaks.
The version of the Sacramento Model used in this study was obtained from the University of Washington where
parallel studies are being conducted for catchments on the Pacific Coast. This version of the model has been
set up for computer-assisted calibration to minimize the percent error in all ranges of flow. The Sacramento
Model takes rainfall and pan evapotranspiration as time series inputs and applies them to the watershed being
modeled using equations to describe the process yhat are partially based on the physical nature of the process
and partially based on empirical data.
Although this study was originally intended to stop with the computation of inflow to Lake Sidney Lanier
on the upper Chattahoochee River, other investigators participating in this report to Congress required an
estimate of the effects of climate change on the freshwater flows in the Apalachicola River at Blountstown,
Florida. An autoregressive, multiple correlation model was applied to the long record of historical monthly flows
of the Apalachicola River at Chattahoochee, Florida, along with historical monthly rainfall and pan evaporation
data. This autoregressive model was used along with climate change data to estimate the effects on the flow of
the Apalachicola River at Blountstown.
The results were quite mixed In general, we found that the various GCM outputs indicated a decrease
in rainfall and an increase in temperature, and thus pan evaporation, when CO2 is increased. For these cases,
the hydrologic model showed a decrease in average annual runoff ranging from 14 to 27%. However, the
GISS2X GCM output showed an increase in precipitation and resulted in a decrease in pan evaporation, with
the result that the hydrologic model used on the upper Chattahoochee showed an increase in average annual
runoff while the regression model used on the Apalachicola showed a small decrease under nearly identical
scenarios. These differences appear to be due to differences in the hydrologic settings of the two study areas
rather than differences in the behavior of the models themselves. Transient A GCM output had some
unexpected fluctuations in the first decade before the CO2 was changed very much. However, in the later
1 Although the information in this report has been funded wholly or in part by the VS. Environmental
Protection Agency under contract no. CR814637, it does not necessarily reflect the Agency's views, and no
official endorsement should be inferred from it
8-1
-------
Hains
decades of this run, the precipitation tends to decrease slightly and evaporation shows robust increases of up to
25%, and the hydrologic model shows a gradual but obvious decrease in runoff of as much as 30%.
Although any reduction in runoff seems severe during a drought of record proportions such as that
experienced in the Southeast in 1988, the climate change scenarios are not particularly severe from a historical
and hydrologic point of view. We have experienced dry periods as severe as any predicted by the climate change
scenarios. In addition to the impacts of global climate change, changes in land use over a long period of time
will also affect runoff trends ~ especially on the entire ACF river system. It is only on systems whose long term
effects are measured in decades - such as forests, swamps, and fish populations, systems with very long memories
— that the dominant reductions in runoff predicted here will be felt By that time it will be too late to reverse
the damage.
8-2
-------
Mains
CHAPTER 1
INTRODUCTION
HYDROLOGY OF A SOUTHEASTERN HEADWATER BASIN
The Chattahoochee River is part of a three-river system (Chattahoochee, Flint, and Apalachicola Rivers)
that arises in Georgia and Alabama and empties into the Gulf of Mexico in the Florida Panhandle as shown in
Figure 1. The detailed hydrologic study is limited to the headwaters of the Chattahoochee River down to the
location of the first reservoir at Buford Dam on Lake Lanier in Georgia as shown in Figure 2 and includes
about 2700 square kilometers. The study area is limited to this catchment to reduce the complications of
modeling the management of multiple reservoirs and because of the particular significance of Lake Lanier as the
water supply for the city of Atlanta.
The Chattahoochee River arises in the Blue Ridge physiographic province in the mountains of
northeastern Georgia near the South Carolina border and flows in a southwesterly direction though the hills of
the Piedmont province surrounding Atlanta. Approximately 15% of the study catchment is in mountainous
terrain with the highest elevations around 1370 meters above sea level. The rest of the basin is at elevations
between 300 and 450 meters above sea level. The Lake Lanier catchment normally has a humid climate
throughout the year. The annual precipitation is about 1400 millimeters and snowfall is rare enough to not be
included in this study. There are seasonal variations in precipitation, but periods as long as a month without
precipitation are extremely rare. The area is mostly used for crop and timber production. The catchment itself
has very little effect from urbanization. The Southeastern States experienced a major drought in the summer
of 1986 and most of the area experienced record low flows. Northeastern Georgia and the upper Chattahoochee
River were the most severely affected part of the state.
RELATED REPORTS
The Corps of Engineers has produced several reports for drought management on the Apalachicola-
Chattahoochee-Flint (ACF) River system (US. Army Corps of Engineers, 1985 and 1986) and they have funded
reports on the effects of weather extremes on the hydrologic system of the Upper Chattahoochee River (Raney
et. al. 1984) and the Mississippi-Alabama Sea Grant Consortium has funded analyses of the climatology of the
Apalachicola, Chattahoochee, and Flint river basins (Nichols and Raney, 1984); although, none of these has
considered the effect of climate change.
ORGANIZATION OF THIS REPORT
Because the modeling for the Apalachicola River at Blountstown was not originally intended to be a part
of this study and because deterministic hydrologic modeling was not used there, the methods used for that study
will be documented in Appendix A.
The rest of this report will describe the hydrologic model used in the detailed hydrologic study along with
the methods used to develop the precipitation traces and pan evaporation data. Caveats will be given where
appropriate both concerning the use of the model and the input data transformations required for the various
CO2 scenarios. The report will present and interpret the results of the detailed hydrologic study and will address
the environmental implications.
8-3
-------
Mains
ARAL
RIDGE
VALLE
REGION
EAST GULF
COASTAL PLAIN
EAST GULF
COASTAL PLAIN
Figure 1. Location of the study area in Georgia, Alabama, and Florida
8-4
-------
BUFORD DAM
T°J.
PEACHTRBE CREEK
VEST POINT
TO
JIM WOODRUFF DAM*-^t
CREEK
TO'HEADWATERS OF
WEST POINT LAKE
LAKE
LANIER
APA LAC HI CO LA RIVER
AND
APALACHICOLA BAY
Figure 2. Drainage area of the Apalachicola
-Chattahoochee-Flint River System.
8-5
-------
Mains
CHAPTER 2
METHODOLOGY
THE SACRAMENTO MODEL
Development of the Model
The National Weather Service (NWS) has adopted the Sacramento Model (Burnash et. al., 1973) for
hydrologic simulation of floods at many of the River Forecast Centers throughout the United States. This model
is also used for extended streamflow prediction by NWS. Although the model is used extensively on a daily basis
by NWS, there is almost no public literature related to the use and applicability of the model. Although the
model is well discussed in hydrologic literature (Burnash et al., 1973; Burnash and Ferral, 1981), information on
the application of the model is limited to unpublished NWS documentation and personal training by NWS at the
centers where it is used.
The Sacramento Model is a deterministic hydrologic model, which attempts to parameterize soil moisture
characteristics in a manner that will logically distribute applied moisture in various conditions and locations in
the soil so that percolation characteristics will be reasonable when streamflow is effectively simulated.
The output of the model is streamflow runoff resulting from processing precipitation and pan evaporation
through an algorithm representing the upper soil zone and the lower soils. As shown in Figure 3, the different
sources of runoff are impervious runoff, surface runoff, interflow, supplementary base flow, and primary base
flow. The five basic soil moisture storages in the model are upper zone and lower zone tension water storages,
which are filled by infiltrated water, and the three free water storages. Upper zone free water is used for
interflow and percolation to lower zones. The two lower zone free water storages receive water at the same time
from percolated water but they drain independently. Potential evaporation from the basin is satisfied by the
model based on availability of water on the surface and in capillary tension.
The version of the model used in this study was obtained from the University of Washington (UW) where
similar studies are under way for this same report. This version is basically the same as that currently in use by
NWS although it is much better suited to the research nature of this study, whereas the current NWS version
is part of a very large real-time streamflow forecasting system used throughout the country. The UW version
of the program includes computer-assisted parameter calibration which is specially adapted to optimize general
streamflow rather than peak flow alone.
The Sacramento model was used almost as received from the University of Washington. Minor
conversion was needed to install the program on a Definicon DSI020 co-processor board in an IBM AT. Two
other modifications were made to the model. One was made to get it to accept pan evaporation data by months
and years for the transient scenarios. As received, the model only accepted a single list of 12 monthly pan
evaporation values for the complete simulation period The other modification was to extend the model storage
capacity from 20 years to 101 years. These modifications were tested to verify that they did not introduce any
change in model behavior by themselves.
Precipitation data used for the calibration and the simulation were obtained from the Dawsonville,
Georgia recording raingage. Other raingage locations in the catchment would have been better for this study
except that they had numerous large periods of missing record. It was decided to use the more consistent record
rather than the nearest record because calibration with a very intermittent record supplemented from another
gage would have been like calibrating with a continually moving raingage. The Dawsonville raingage had a fairly
good record although even it had a few missing periods - some as large as three months. These deficiencies
were made up without adjustment from other nearby records such as the Clermont raingage, which is in the
center of the study catchment.
8-6
-------
« i t 4
I ET DEMAND
J L
-
MPBMMMEft
E T
E T
E T
. "
ET
1
PI
»
[ PERVIOUS AREA
UPPER
1 KIM
J IMPERVIOUS
4NN>
ZONE
CftCCtt
TENSION WATER ^'"
uifw
^
-^
1 . 1
«x-
fKE W
ul
1
IHKOUatOH
IfllK.KV
X
s
\ '-^ 1
1
LCWEJR
TENSION VMWER
itlw
i
--^.
^
1 FfREE
MER
r«r
ow ,
SAftVA
~1
ZONE
FREE WATER
p
\.Vf
RSERV
s
art
LISK ,
SURFACE
RUNOFF
•kftnciou/
SUmEMENUl
•ASEHOW
1
iLim
MtMART RASE FlOW
TOIAl
CHANNEl
INHOW
RASE
FlOW
mSTRMJTlON
FUNCTION
STREAM
FlOW
ssour
SURSURFACE
DISCHARGE
Figure 3. Flow Diagram of the Sacramento Model (Peck, 1976)
-------
Hains
Pan evaporation data were obtained for Athens, Georgia. These data cover the period from 1947 to 1970,
which is the longest period of record near the site. Long-term monthly average pan data were used in the
calibration and the simulations because the data did not cover the whole period of the study.
Initial calibration parameters were obtained from the Atlanta River Forecast Center (RFC). Initial runs
using the Atlanta RFC parameters showed that, while they produced good simulation of the peaks, the
parameters were not optimized for low flows. Because the GCM output is presented as monthly averages to be
used to modify historical meteorological data, it is not likely that the GCM effects on short-term, intense events
will be preserved. The most meaningful effects from the GCM output will be on long-term events such as long-
term wet or dry periods, which would have a major impact on water supply reservoirs such as Lake Lanier. As
a result, the parameters received from the Atlanta RFC were optimized for general streamflow which weights
low flows as well as floods. In this case, flood flows may not model quite as well but low flow events improve
greatly.
The model was calibrated and verified using VS. Geological Survey daily data for station 02334500
Chattahoochee River near Buford, Georgia for the years 1949 through 19SS. These were the only years for which
data were available for this location before the dam on Lake Lanier was dosed. The years 1945 through 1955
were in the warmest decade of record as recorded at Atlanta, which is about 100 kilometers to the southwest of
the center of the study catchment. The calibration-verification years were representative of the period of record
for pan evaporation data as recorded at Athens, Georgia. Precipitation during the calibration-verification years
included both the wettest year of record the two driest years prior to 1981. As a result, these years are a good
sample of the hydrologic extremes experienced in this part of Georgia.
The calibration itself was performed for the years 1952-1954 using a logarithmic objective function.
Average flow for these years came within one standard deviation of the long-term average flow. However, they
include the third highest and the third lowest monthly flow of record up to 1984 prior to the beginning of the
drought of 1988. As a result, these years represent the hydrologic extremes for this basin with the exception of
prolonged drought.
Although there seems to be some difference of opinion in the hydrologic community over the amount of
data required for adequate calibration, several trends seem to stand out. Donigian (1984) and Bourne (1978)
suggest that 3-5 years of data should be used for calibration. Donigian (1984) says that wet years are better for
calibration because they usually have more uniform rainfall over the watershed Hassett and Lumb (1974) show
on a basin near the Chattahoochee River that, as the calibration period is extended, very little improvement in
calibration is achieved after 2-3 years. Johanson and Crawford (1979) explained that, although many years might
be required to collect a significant number of meteorological/hydrologic events in an arid climate, 2-3 years
should provide a good sampling of the hydrologic spectrum of a watershed in a subhumid climate.
Limitations Resulting from the Model
After calibration on the years 1952-1954, the period 1949-1955 was simulated and the years 1949-1951 and
1955 were used for validation. The hydrologic model did not calibrate as well as expected. Donigian (1984) gives
calibration guidelines for hydrology/hydraulics modeling as: Very Good <10% difference; Good 10-15%
difference; and Fair 15-25% difference between simulation and recorded values on an annual and monthly basis.
In a few cases, monthly water balance errors were as high as 35% although most months were within 10 to 15%
of the observed runoff. The months showing poor comparisons with observed runoff were not limited to either
wet or dry months, so they probably are caused by unrepresentative rainfall. These problems with calibration
would be lessened if more meteorological stations were available.
The Sacramento model does not consider CO2 directly at all. All of the CO2 effects are modeled only
as reflected in the various GCM outputs.
8-8
-------
Hains
. Because the GCM modelers seem to place little confidence in their evaporation data, adjustments to
historical pan evaporation data for the hydrologic model were made on the basis of the Penman equation (Linsley
et at, 1982), which models daily pan evaporation on the basis of air temperature, wind movement, vapor pressure,
and solar radiation. Monthly values for these variables were obtained from the GCM outputs, and monthly pan
evaporation was computed for each scenario using the Penman equation. The same computation was done for
the GCM control runs. The ratio of scenario to control Penman-computed pan evaporation data was then
applied to the historical pan evaporation data on a monthly basis following the standard EPA procedure for
development of the scenarios.
THE SCENARIOS
Scenarios Used
The 2X GISS, GFDL, and OSU scenarios were run as well as the GISS Transient A scenario. Time did
not permit running the analogue and transient B scenarios. Precipitation ratios were used directly from the
NCAR gridded data files and applied to daily historical precipitation data Actual GCM values for wind, solar
radiation, air temperature, and either specific humidity or mixing ratio were used in the Penman equation to
compute pan evaporation values, which were then converted to evaporation ratios and applied to historical data
monthly pan evaporation data.
The EPA standard procedure was followed for development of the scenarios. The upper Chattahoochee
catchment falls within only one grid cell for each scenario. Historical daily precipitation data for Dawsonville
for the years 1951 through 1980 were used along with long-term monthly averages for pan evaporation data
collected during a major part of the base period. These were modified by the ratios received from NCAR for
the appropriate grid cell and month for the steady-state scenarios. The 1951 data were duplicated as 1950 in the
model runs to provide one year of spin up data.
Historical dairy precipitation data for 1951 through 1980 were applied for the years 1981 through 2010 and
repeated for 2011 to 2040 and 2041 to 2059 to develop precipitation inputs for the transient scenario. The years
1970 through 1980 historical data were applied to the years 1970 to 1980 for 11 years of spin-up time in the
transient model
Issues Resulting from the Scenarios
No particular problems arise from the use of the 1951 to 1980 historical period for the steady-state
scenarios. However, several issues arise from the application of these data to the transient scenario. The use
of a 30-year historical record for the 90-year period means that, except for the 30 years from 2011 to 2040, the
leap years will not match. Therefore any rainfall occurring on February 29 will be ignored, and the leap year
Februaries in the model runs will have one more day to runoff in base flow recession than they would have if
the leap years had matched up properly. This introduces short term variations in monthly runof,f which are
based on calendar irregularities rather than the climate scenarios. These will tend to reduce the long-term
annual runoff very slightly, probably on the order of 1 or 2%.
The process of folding the 30-year record and reusing it introduces a 14% flow anomaly in the first month
after the fold. This reduces to 1% within 4 months although because of the long persistence of base flow in this
basin, it takes 3 years for this anomaly to completely disappear. This is compounded by the leap year shift
problems mentioned above. The overall effect is less than 0.5% on an annual basis.
Because the GCM output is presented as monthly averages, short-term intense events are not preserved.
These have been introduced into this study from the historical record. This procedure loses any variation in
storm frequency, duration, or intensity resulting from climate change. If storm events were to increase in
intensity, even without an increase in annual or monthly precipitation, more water would run off as surface flow
instead of as base flow. This would cause increased annual runoff because less water would infiltrate and there
would be less evapotranspiration.
8-9
-------
Hains
CHAPTERS
RESULTS
IMPACTS OF CLIMATE CHANGE
Direction and M*
Qbar Std +1 ctd -1 «td Mas SOXlle Mln Zcb Q Xch Std
Bace 61.84 11.95 73.79 49.89 88.26 65.19 38.82
TransA
1980
1990
2000
2010
2020
2030
2040
2050
2X->
GISS
CFDL
OSU
41.99
67.00
66.20
49.33
53.01
59.89
43.38
49.95
69.57
45.08
49.92
8.52
10.62
9.26
10.62
9.80
9.88
8.98
11.19
13.39
10.59
10.96
50.52
77.62
75.46
59.95
62.81
69.77
52.36
61.14
82.97
55.67
60.88
33.47
56.38
56.95
38.71
43.21
50.01
34.40
38.77
56.18
34.49
38.96
56.10
78.92
85.77
71.27
62.35
77.28
61.14
65.95
97.44
68.30
74.02
41.03
70.03
65.13
44.49
55.73
60.26
39.36
52.90
71.42
43.98
54.68
27.78
52.13
50.66
38.74
35.88
43.78
33.33
32.28
47.74
28.77
27.58
13. 9X 13. 2X
-26. 21 -10. 5X
-18. 3X -7.4X
XBASE 61.08 11.84 72.92 49.24 88.26 64.85 38.82
TRANSA 54.34 12.54 66.88 41.80 85.77 55.05 32.28 -12.IX 5.OX
At first glance, it appears unusual that the GISS 2X scenario would cause a 14% increase in long-term
annual flow while the GFDL and OSU 2X scenarios show decreases in long-term annual flow of 26% and
18%, respectively. However, average precipitation for the 2X GISS scenario had an increase of 12%, while
the GFDL and OSU 2X scenarios had decreases of 13% and 3.6% respectively. The other input to the
hydrologic model, pan evaporation, was higher in all three scenarios with the GFDL being the highest at
28% followed by the OSU at 12% and the GISS at 9%. Keep in mind that increased precipitation will cause
an increase in runoff if other factors remain constant, while increased pan evaporation will normally decrease
runoff.
Figure 5 presents long-term seasonal flow values for the three steady-state scenarios on the upper
Chattahoochee study along with observed flow. The GISS and GFDL scenarios seem to parallel the observed
flows, with the GISS intensifying both the highs and the lows and the GFDL generally decreasing the
8-10
-------
S
o
o
90
80 -
70 -
60 -
50 -
40 -
30 -
20 -
10 -
Variability of Annual Flow
Upper Chattahoochee River
^
X
AVA
T
T
T
A VA VA
T T
Base TransA 1980 1990 2000 2010 2020 2030 2040 2050 2X=> GISS GFDL OSU
\/ry\ -i
Q bar
Std
Figure A
-------
o
V
00
b
0)
A
CJ
Upper Chattahoochee--2XC02
Seasonal Flows
140
130 -
jan
n 1 1 r
feb mar apr may jun jul
aug sep
year
•a
x
OBSERVED
GISS2X
GFDL2X
OSU2X
Figure 5
-------
Hains
high-flow months. The OSU scenario shows decreases for all monthly flows, however, the flows in the wetter
spring months are decreased sufficiently more such that the wet season is shifted a month earlier. Figure 6
shows the ratio of the scenario flow/observed flow by season for the steady-state scenarios. This figure points
out the strong differences between the results of the three climate models. Some have their greatest effect
in the wet season, others in the dry. Some are wetter, others are drier than those observed. The largest
difference between the climate models can be seen by comparing the GFDL and the OSU models, which are
complete opposites.
These two effects of climate change are not the same for all months, figures 7 through 9 present the
ratio of modeled values/observed values by months averaged over the 30-year scenario period. On each graph,
the climate input ratios (precipitation and pan evaporation) are shown along with the hydrologic output (flow).
Figure 8 shows that even in the GISS scenario, which has an overall increase in precipitation, there are
months which have reduced precipitation. In general, the months that have the greatest rainfall occur in the
winter and spring months when pan evaporation is at a minimum As a result, increases in precipitation
accompanied by increases in pan evaporation will cause higher high flows and lower low flows.
The overall effect on annual runoff depends not only on the percent increase or decrease in precipitation
and pan evaporation but also on the magnitude of the base precipitation and pan evaporation which is being
changed For instance, a small percentage increase in precipitation in a wet month can have more effect on
annual runoff than a much larger increase in precipitation in a dry month.
The effects of the 2X scenarios are quite significant. The GISS scenario has the least effect representing
an increase in annual flow of about 0.7 standard deviation. The GFDL and OSU 2X scenarios show decreases
in average annual flow of about 1.4 and 1.0 standard deviation, respectively. No one climate variable seems
to dominate all effects, although precipitation is assumed to have a greater effect pending further investigation.
Even though the effects are large, the period of the year in which the increases or decreases in
precipitation and pan evaporation occur is significant The hydrologic effects are extremely nonlinear, so the
wetness or dryness of the catchment during an increase or decrease in precipitation or pan evaporation will
alter the strength of the effect.
Changes in precipitation appear to dominate the effects of climate change on runoff. Figures 7 through
9 display the parallel between change in precipitation and change in runoff. This parallelism tends to lag
about one month, so that the month with the lowest precipitation ratio is followed a month later by a month
with the lowest flow ratio. This type of effect is quite consistent with the short-term memory of water storage
in the system. Changes in pan evaporation have a more general effect -- bringing the annual average ratio
down when the pan evaporation ratio is high. This general effect is consistent with the long-term memory of
groundwater storage, which has a significant persistence in the system for several years.
Transient A Scenario
The transient scenario must be viewed from a different perspective than the steady-state scenarios. For
the transient scenario, carbon dioxide in the atmosphere has been assumed to follow current trends up through
2059. As a result, the climate is changing all the time. Decadal variability of the transient A scenario is
presented in Figure 4 and Table 1. Because the base period has an increasing flow trend, the results are
sawtoothed at the point where the base period is repeated. As a result, it is not easy to see the GCM effects
without careful consideration.
Figure 10 shows the annual ratios of transient/baseline climate inputs (precipitation and pan evaporation)
and hydrologic outputs (flow) as 5-year running averages. The ratio data are easier to interpret than the flow
data alone because the ratio has the particular effects of the baseline period removed. There is no clear effect
of climate change on precipitation for the transient run. Precipitation ratios seem to oscillate between wet
and dry trends about every 20 years, with the long-term ratio being dose to one. The climate change effects
for the transient run seem to be mainly associated with increases in pan evaporation which are robust and
obvious.
8-13
-------
V
n
«J
£>
•c
cfl
fi
V
o
m
•¥>
OB
0.3
Upper Chattahoochee--2XC02
Seasonal Flow Ratios
oct nov dec jan feb mar apr may jun jul aug sep
year
I
X
D
OBSERVED
GISS2X
GFDL2X
Figure 6
OSU2X
-------
m
o
n
a)
tf
0.3
Upper Chattahoochee~-GFDL2X
Seasonal Input Output Ratios
oct nov dec jan
1 1 1 1 T
feb mar apr may jun
aug sep
year
Figure 7
-------
n
id
,0
oJ
C
V
o
m
K
Upper Chattahoochee—GISS2X
Seasonal Input Output Ratios
oct nov dec jan feb mar apr may jun jul aug sep
0.7
A
year
1
X
Figure 8
-------
V
m
tS
jo
a
V
o
n
K
Upper Chattahoochee--OSU2X
Seasonal Input Output Ratios
oct
T 1 1 1 \ 1 1 i r
nov dec jan feb mar apr may jun jul
T T
aug sep
year
Figure 9
-------
0)
m
cd
ft
0
cd
fi
V
o
m
s_^
o
**4
cd
0.7
Upper Chattahoochee--TRANSA
5 year average Input Output Ratios
S3
ofc
2060
•a
O PAN + RAIN
O FLOW
Figure 10
-------
Hains
During the first decade of the transient A scenario, when there was a decrease in precipitation as
compared with the base period, pan evaporation was already elevated with a resulting 10% decrease in runoff.
This is followed by a decade with sufficiently more precipitation than the base period so that the runoff
exceeds that of the base period for several years. This seems unusual at the beginning of a transient run of
the GCM when it is assumed that the CO2 effects are minimal. The rest of the transient run follows an
expected pattern of an accelerating reduction in runoff driven primarily by increases in pan evaporation. The
positive excursions in rainfall appear to be less each time throughout the transient run and may possibly show
a general decline in precipitation if the transient scenario is extended over a longer period
To present the effects of the transient scenario in the dearest comparison with the identical period from
the baseline data, the individual months were accumulated for both the base period and the transient scenario.
These were plotted on the same graph versus time as shown in Figure 11. The effects discussed above can be
seen very dearly in the departures of the transient A curve from the observed curve. The initial departure
in the 1980s, the decreasing departure in 1990s and 2000s, and the robust departure beginning about 2020 are
very dear.
This same behavior is displayed in Figure 12, where the two accumulated monthly curves are plotted
against each other. The straight line is the slope that the curve would have if there were no effects. Where
the slope is less than the straight line, the transient runoff is less than the base period runoff. Where the slope
is greater than the straight line, the transient is greater than the base period runoff.
8-19
-------
d
o
55
I
CO
fcx-v
OB
•*
|
O
2
1.9
1.8
1.7
1.5
1.4
1.2 H
0.9 -
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
0
Upper Chattahoochee—TRANSA
Cumulative Mass Plot
Ob! erved
Transient A
$
1980 1990 2000 2010 2020 2030 2040 2050
Figure 11
-------
Upper Chattahoochee—TRANSA
Double Mass Plot
0
S
I ~
£S
S3
. F*
•4! ••*
«8
V
•M
n
C
ti
h
H
2.1 -
2 -
1.9 -
1.8 -
1.7 -
1.6 -
1.5 -
1.4 -
1.3 -
1.2 -
1.1 -
1 -
0.9 -
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
0
0
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 r
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(Millions)
Observed (CFS-Months)
-------
Mains
CHAPTER 4
INTERPRETATION OF RESULTS
The changes in the hydrology of the Upper Chattahoochee River for the GCM scenarios were the result
of changes in precipitation and pan evaporation. The sensitivity of the hydrology to each of these variables
is complex. When the system is wet, changes in precipitation and pan evaporation do not have as much effect
on the streamflow as they do when the system is fairly dry. There seems to be roughly twice as much
sensitivity to changes in precipitation as there is to changes in pan evaporation. The sensitivity to precipitation
changes is much more immediate because the short-term moisture storages are affected. Because evaporation
effects the long-term moisture storage in the system more strongly, sensitivity to this climate variable is more
gradual.
Temperature, humidity, solar radiation, and wind movement have been combined using the Penman
equation for pan evaporation as described above. Increases in temperature, wind movement, and solar
radiation will increase evaporation, while increases in humidity will decrease evaporation. Table 2 and
Figure 13 show the relative effect of each variable from the steady-state scenarios on pan evaporation. In each
case, temperature has the greatest effect on evaporation. Changes in humidity have about half as much effect
as changes in temperature; changes in wind movement have less than one-fifth as much effect as changes in
temperature; and changes in solar radiation have less than a tenth as much effect as changes in temperature.
No value is given for the effect of humidity alone on the GFDL evaporation because the GFDL2X humidity
is so high that it cannot occur without the associated rise in temperature.
Table 2. Relative Effect of Each Climate Variable on Changes in Penman Evaporation
GFDL GISS OSU
*
Partial PET change due to change in:
Temperature 35% 19% 16%
Humidity -15%* -8% -3%
Wind Movement 5% -4% -3%
Solar Radiation 3% 1% 1%
Total PET change 28% 9% 12%
Estimated from combined temperature and humidity effects.
Gradual changes in CO2 along the lines of those used in the transient A scenario might not be noticed
in the hydrology of the Upper Chattahoochee River over the next 20-30 years. If the effects reported by the
transient GCM run for the first 3 decades are accepted, the first effects would be a temporary decrease in
precipitation with less effects on pan evaporation. This would be followed by a slight increase in precipitation
followed by a long, steady increase in pan evaporation resulting in a steady decline in runoff.
8-22
-------
w
IX
C
(0
c
V
o
h
V
IX
Effect of Climate Variables on ET
40%
Upper Chattahoochee
bFDL
30* -
20915 -
10* -
058
-20*
Temperature
Humidity Wind Movement Solar Radiation Total PET change
* Est. from combined Temp, and Humidity
Figure 13
-------
Mains
CHAPTERS
IMPLICATIONS OF RESULTS
ENVIRONMENTAL IMPLICATIONS
Because of the conflicting nature of the results of the steady-state GCM scenarios, the environmental
implications of this study depend on which scenario is accepted. However, all of the scenarios show increasing
pan evaporation and, with die exception of the GISS2X scenario, they show decreasing precipitation. Even
the GISS model seems to suggest decreasing precipitation in the transient scenario after the first 30 years.
This suggests that there will be less water in the southeastern United States. But where does it go? The
GCMs operate by balancing the energy and mass fluxes in the atmosphere for each grid cell of the model.
With increased temperature, some of the moisture will simply remain in the atmosphere. The rest must show
up elsewhere on the earth if the GCMs are to be given any credence at all.
It is reasonable to draw the conclusion that the current trend in CO, around the world will result in less
runoff for the Upper Chattahoochee River Basin and for other areas nearby. This will have prominent effects
on all aspects of the environment. Studies by others are in progress to assess the impact of the hydrologic
effects on water supply, agriculture, forestry, biology, and other areas.
The decrease in runoff will not be dearly distinguishable from short-term trends already present in the
present normal climate. The year 1988 may be the worst drought in history in the southeast United States -
- drier by many times than any year in the 30-year base period used in these studies. Within a few years, we
should be once again in a wet cycle. Under such conditions, who will be able to identify the current drought
as the beginning of a new trend as opposed to an extreme event that is simply a statistical feature of the
current normal climate? It is important to do and improve studies of the present type to assess the
possibilities of global warming driven by increased carbon dioxide in the atmosphere. However, it is not likely
that the effects of climate change will be first observed from its hydrologic effects. By the time hydrologic
effects are dearly evident, other aspects of the environment will already have experienced irrevocable change.
8-24
-------
Mains
LIST OF ABBREVIATIONS
ACF - Apalachicola-Cbattahoochee-Flint River System
CO, -- Carbon Dioxide
GCM ~ General Circulation Model
GFDL ~ Geophysical Fluid Dynamics Laboratory
GISS - Goddard Institute for Space Studies
NASA - National Aeronautic and Space Administration
NCAR - National Center for Atmospheric Research
NWS - National Weather Service
OSU - Oregon State University
RFC - River Forecast Center
8-25
-------
Hains
REFERENCES
Bourne, R. Gregory, Gerald N. Day, and Thomas N. Debo. " Water Quality Modeling using Hydrocomp
Simulation Programming (HSP)." In: Proceedings of the 26th Annual Hydraulics Division Specialty
Conference. Am. Soc. of Civil Engineers.Collepe Park, Maryland, 1978. pp 358-362.
Burnash, Robert J. C, R. Larry Ferral, and Richard A. Mcguire. "A Generalized Streamflow Simulation
System, Conceptual Modeling for Digital Computers." National Weather Sendee, and California Department
of Water Resources, Sacramento, California, 1973. 204 pp.
Burnash, Robert J. C, R. Larry Ferral. "A Systems Approach to Real Time Runoff Analysis with a
Deterministic Rainfall-Runoff Model." Proceedings of the International Symposium on Rainfall-Runoff
Modeling. Mississippi State.Mississippi, 1981. pp. 107-120.
Donigjan, Anthony S., John C. Imhoff, and others. "Application Guide for Hydrological Simulation Program -
FORTRAN (HSPF)." EPA-600/3-84-065, US. Environmental Protection Agency, Athens, Georgia, 1984.
177pp.
Hassett, Timothy, and Allen Lumb. "Use of Watershed Models to Predict Low Flows." In: Proceedings:
Contribution of Irrigation and Drainage to the World Food Supply. Specialty Conference. Am. Soc. of Civil
Engineers, Biloxi, Mississippi, 1974. pp. 203-217.
Johanson, Robert, Norman Crawford and others. "Hydrologic Simulation Program FORTRAN (HSPF)
Workshops." Workshop sponsored by: VS. Environmental Protection Agency and Hydrocomp, Athens,
Georgia, 1979.
Linsley, Ray K., Max A. Kohler, and Joseph L. H. Paulhus. Hydrology for Engineers. Mcgraw-Hill, New
York. 1982. 508pp.
Nichols, William G., and Donald C. Raney. "1984 Water Assessment for the Apalachicola-Chattahoochee-
Flint River Basin Water Management Study." UJS. Army Corps of Engineers, Mobile District, Mobile,
Alabama, 1984. 114 pp.
Peck, E. L. "Catchment Modeling and Initial Parameter Estimation for the National Weather Service River
Forecast System." NOAA Technical Memorandum NWS HYDRO-31. US. Department of Commerce, Silver
Spring, Maryland, 1976.
Raney, Donald C., William G. Nichols, and Pamela S. Brandes. "An Investigation of Drought Related Factors
Affecting Pool Elevation o Lake Lanier, Georgia and the Development of a Reservoir Water Availability
Index." U.S. Army Corps of Engineers, Mobile District, Mobile, Alabama, 1984.
SAS Institute, Inc. "SAS/ETS User's Guide, Version 5 Edition." Gary, North Carolina, 1984. 738 pp.
U.S. Army Corps of Engineers, Mobile District. "Drought Water Management Strategy for the Apalachicola-
Chattahoochee-Flint Basin." 1986.
U.S. Army Corps of Engineers, Mobile District. "Interim Drought Management Plan for the Apalachicola-
Chattahoochee-Flint River Basin." 1985.
8-26
-------
Hains
APPENDIX A
The Apalachicola River Study
The original study area for this report only included the upper Chattahoochee River down to Buford Dam
on Lake Lanier. The possibility of modeling the entire Apalachicola-Chattahoochee-Flint River system (ACF)
was raised prior to the beginning of these studies; however, the detailed hydrologic study was limited to the
upper Chattahoochee because calibration of as many as 30 subareas and data retrieval of non digital records
for 13 reservoirs would have been prohibitive in both time and expense. It is unlikely that a satisfactory model
of the entire ACF system could have been produced in the 3-5 months available for the research portion of
these studies.
However, other researchers involved in the current Report to Congress have important models of
environmental systems that require hydrologic inputs and more specifically require assessments of the change
of hydrologic inputs based on the results of the global circulation model (GCM) scenarios. The GCMs report
a value for hydrologk runoff in their outputs. Unfortunately, this is one of the GCM variables for which there
is no confidence within the current state of the art. The lack of confidence in the runoff outputs is best
illustrated by the fact that, with the exception of the GISS model, the GCMs predict zero runoff for most
months of each year for both the base case and the changed CO, scenarios.
In order to supply the runoff estimates required by these researchers, an autoregressive multiple
correlation technique was used to develop a statistical model of the entire ACF system which would yield flow
estimates as outputs from meteorological inputs and which would be responsive to the changed inputs of the
GCM scenarios.
HYDROLOGY OF THE APALACHICOLA-CHATTAHOOCHEE-FLINT RIVER SYSTEM
Figure 2 in the main report shows the study area which includes the entire Apalachicola-Chattahoochee-
Flint River system down to Blountstown, Florida. This system includes the headwater basin of the upper
Chattahoochee River, which is the subject of the main report The upper Chattahoochee study encompasses
about 2700 square kilometers whereas this study includes about 45600 square kilometers. The ACF system
includes 13 reservoirs with a combined impoundment in excess of 5 cubic kilometers. Figure 1 in the main
report shows that the ACF system arises in the Blue Ridge mountains, flows through the Peidmont, across
the Fall Line into the Gulf Coastal Plain. The ACF system crosses five degrees of latitude and goes over 800
kilometers from the inland mountains to the Gulf Coast. For such a large area, the climate is not uniform
(Nichols and Raney, 1984). Average monthly temperatures decrease in the interior of the basin, particularly
at higher altitudes, by as much as 5 degrees Celsius in the winter months and 3 degrees Celsius in the summer
months. Figure A.1 shows the average annual precipitation within the ACF system which varies from about
47 inches (120 centimeters) per year to over 60 inches (150 centimeters) per year. The variation in
precipitation in the basin can be attributed largely to the proximity to the mountains and the Gulf Coast and
to the prevailing weather circulation patterns.
For a system the size of the ACF, even the unregulated surface water bodies will act as a filter with a long
delay. With the addition of manmade impoundments and regulated flows, the memory of the system is long
indeed-on the order of several months. The impacts of small storms on any part of the system will be
completely lost after being stored and passed through several reservoirs.
A-27
-------
Hains
Figure A.I.
Average Precipitation (inches/year)
for Apalachicola, Chattahoochee and
Flint River System
A-28
-------
Hains
METHODOLOGY
Several meteorological stations in and around the ACF basin were chosen to represent the climate
variation of the region in the analysis. The stations used were Tallahassee, Florida, Macon and Atlanta,
Georgia, and Montgomery, Alabama, for precipitation and Tifton and Experiment, Georgia, for pan
evaporation. Data from these stations were compiled into a monthly average time series for the 38 year
period from 1948 to 1985.
The IDENTIFY portion of the SAS (SAS, 1984) ARIMA procedure was used initially to find the lag
correlation of each of these stations with the flow on the Apalachicola River at Chattahoochee, Florida (USGS
Gaging Station 02358000). This same procedure was used to find the auto correlation of the flows with
themselves. The strongest correlation was the autocorrelation for the flow station itself lagged by one month.
This was followed by a negative correlation with the Tifton evaporation data lagged by two months. Following
these variables were the Macon and Atlanta lagged by 0 and 1 month. The Montgomery and Tallahassee
rainfalls did not meet the test for significance in the regression after the other variables were entered. This
is not surprising. In a system with so much delay, modeled at the gross time scale of one month, the cross-
correlation of the variables will be high enough that some of them would not add any new information to the
regressioa
The regressions were initially performed on the untransformed monthly average data and a fairly high
correlation was achieved. However, because of the logarithmic skew of hydrologjc extremes, the high flow
months were dominating the correlation and the low flow months were poorly modeled. To correct for the
skew, the final regression was performed on the logarithms of the monthly data to give equal weight to both
ends of the hydrologic extremes. The log correlation coefficient was not quite as high as the non-log
correlation, although the simulation of the base period was slightly better correlated and the simulation of low
flows was greatly improved. Table A.1 lists the variables included in the final multiple regression along with
their regression coefficients, partial correlation and the sigrlcance of each variable in the regression. The
correlation of the resulting modeled flow with observed f /* s 0.83.
THE SCENARIOS
The GISS, GFDL, and OSU steady-state scenarios were .nn as well as the GISS Transient A scenario.
The EPA standard procedure was followed for development of the scenarios. Because of its size, the ACF
basin falls in more than one grid cell for each scenario. For the GISS model, both the steady-state and the
transient scenario the ACF basin falls into 4 grid cells. Ratios from the GCM results were weighted on a
drainage area basis for each cell to produce a composite precipitation ratio which was then applied to each
historical rainfall record on a monthly basis. For potential evaporation, the Penman equation was applied
here just as it was for the upper Chattahoochee study. After the evaporation ratios were determined, they
were weighted on a drainage area basis for each cell to produce a composite evaporation ratio which was
applied to the historical Tifton pan evaporation record on a monthly basis.
Data for the year 1950 were used to provide a year of spin up for the 1951 to 1980 steady-state scenarios.
Precipitation and evaporation data for 1951 to 1980 were applied for the years 1981 to 2010 and repeated for
2011 to 2040 and 2041 to 2059 to develop inputs for the Transient A scenario. Data for the years 1970 to 1980
were applied to the year 1970 to 1980 for 11 years of spin up time in the transient scenario.
A-29
-------
Mains
Tmbl. A.I. STEPWISE REGRESSION PROCEDURE FOR DEPENDENT VARIABLE LOGCHAT
STEP 7 VARIABLE LOGCHAT2 ENTERED R SQUARE • 0.76610469 C(P)
OF SUM OF SQUARES MEAN SQUARE
REGRESSION
ERROR
TOTAL
INTERCEPT
LOGATL
LOGMAC
LOGATL1
LOGMAC1
LOGTIFT2
LOGCBAT1
LOGCHAT2
7
442
449
B VALUE
2.47638548
0.11209612
0.16853160
0.08658081
0.10063262
-0.43678937
0.36745177
0.06408580
21.67404930
6.61718772
28.29123702
STD ERROR
0.02355991
0.02310643
0.02411703
0.02445807
0.03542957
0.04683286
0.03769756
3.09629276
0.01497101
TYPE II SS
0.33891059
0.79643100
0.19295093
0.25344546
2.27543022
0.92161705
0.04326615
8.00000000
F PROB>F
206.82 0.0001
F PROB>F
22.64
53.20
12.89
16.93
151.99
61.56
2.89
0.0001
0.0001
0.0004
0.0001
0.0001
0.0001
0.0898
BOUNDS ON CONDITION NUMBER:
4.180124,
105.5716
STATISTICS FOR REMOVAL: STEP 8
DF - 1,442
PARTIAL MODEL
Table A.I (continued)
VARIABLE
LOGATL
LOGMAC
LOGATL1
LOGMAC1
LOGTIFT2
LOGCHAT1
LOGCHAT2
R**2
0.0120
0.0282
0.0068
0.0090
0.0804
0.0326
0.0015
R**2
0.7541
0.7380
0.7593
0.7571
0.6857
0.7335
0.7646
SUMMARY OF STEPWISE REGRESSION PROCEDURE FOR DEPENDENT VARIABLE LOGCHAT
STEP
VARIABLE
ENTERED REMOVED
1 LOGCHAT1
2 LOGTIFT2
3 LOGMAC
4 LOGMAC1
5 LOGATL
6 LOCATL1
7 LOGCHAT2
NUMBER
IN
1
2
3
4
5
6
7
PARTIAL
R**2
0.5252
0.1166
0.0820
0.0234
0.0114
0.0060
0.0015
MODEL
R**2
0.5252
0.6419
0.7238
0.7472
0.7586
0.7646
0.7661
C(P)
451.158
232.767
79.901
37.725
18.181
8.890
8.000
F
495.6473
145.5673
132.3429
41.1504
20.9687
11.2427
2.8900
PROB>F
0.0001
0.0001
0.0001
0.0001
0.0001
0.0009
0.0898
RESULTS
With the exception of the GISS scenario, the results are quite similar to those for the upper
Chattahoochee study. This is encouraging because the hydrology was modeled with a much simpler method
for the ACF basin than it was for the upper Chattahoochee. However, especially for the GFDL where 91%
of the basin is in the same grid cell as the upper Chattahoochee river, the meteorological input ratios were
very similar to those for the upper Chattahoochee, so the results should be similar if the hydrologic models
are both reasonably reliable.
The results will be presented here with little discussion because most of what was said in the main report
applies here as well.
A-30
-------
Hains
Steady-State Scenario^
Figure A.2 is a graph showing one standard deviation on each side of the long-term average annual flow
at Blountstown for the base period observed flows and the GCM scenario flows. The data for Figure A.2
appear in Table A2. These results are similar to those for the upper Chattahoochee except for the GISS
scenario which shows a 6% decrease in long-term average flow here as compared to a 14% increase for the
upper Chattahoochee study. This difference is the result of the influence of the three additional grid cells
which were used to modify the climate inputs to the model
Figure A3 presents long-term seasonal flow values for the 3 steady-state scenarios at Blountstown along
with observed flow. Figure A.4 shows the ratio of scenario flow/observed flow by season for the steady-state
scenarios. Figures A3 through A.7 present the ratio of modeled values/observed values by months averaged
over the 30 year scenario period. On each graph, the climate input ratios (precipitation and pan evaporation)
are shown along with the hydrologic output (flow) ratio.
Transient A Scenario
The transient scenario must be viewed from a different perspective than the steady- state scenarios. For
the transient scenario, CO, in the atmosphere has been assumed to follow current trends up through 2059.
As a result, the climate is changing all the time. Decadal variability of the transient A scenario is presented
in Figure A.2 and Table A.2. Because the base period has an increasing flow trend, the results are sawtoothed
at the point where the base period is repeated. As a result, it is not easy to see the GCM effects without
careful consideration.
Table A.2. Annual Variability of Hydrologic Model Scenarios—ApalaehlcoU River
< Cubic Meters per Second >
Qbar Std -1-1 »td -1 «td Max SOX lie Mln Xch Q Zch Std
Base 639.71 171.88 811.59 467.82 1003.72 6*2.25 326.83
TransA
1980
1990
2000
2010
2020
2030
2040
2050
2X->
CISS
CFDL
OSU
482.52
647.38
728.05
530.57
595.36
629.68
462.95
536.38
619.40
512.31
567.95
133.54
185.31
108.71
124.42
174.88
93.90
109.53
156.76
156.56
132.83
142.09
616.06
832.69
836.76
655.00
770.25
723.58
572.48
693.14
775.97
645.14
710.04
348.98
462.07
619.35
406.15
420.48
535.78
353.42
379.62
462.84
379.47
425.86
671.31
959.57
921.97
690.34
901.13
803.52
61.14
65.95
949.75
800.94
856.75
513.81
608.16
710.58
580.30
552.77
603.77
39.36
505.91
614.02
513.95
565.63
290.59
444.12
623.17
328.59
395.16
530.97
289.06
803.07
311.23
247.94
282.91
-6.IX -6.0Z
-22.42 -20.31
-13.91 -14.7X
XBASE 659.81 166.64 826.46 493.17 1003.72 648.17 326.83
TRANSA 575.57 157.22 732.78 418.35 959.57 574.55 289.06 -10.OX -8.5X
A-31
-------
O
V
m
6
s_>>
0
E
h
V
*
K
900
800 -
700 -
600 -
500
400 -
300 -
200 -
100 -
\
T
Variability of Annual Flow
Apalachicola River at Blountstown
12
\
T
\
T
/ //
\
~l ' ( ' " I
Base TransA 1980 1990 2000 2010 2020 2030 2040 2050 2X=> GISS GFDL OSU
1
ZZ1 -i std
(V\l Q bar
Figure A.2
+ 1 Std
-------
o
v
m
2 B
* "^
fi C
C id
^^ 00
0) ?
bfl 0
tZA
a) H
43^
o
on
Apalachicola—2XC02 SCENARIOS
SEASONAL FLOW
oct nov dec jan feb mar apr may jun jul aug sep
OBSERVED
GISS2X
GFDL2X
Figure A.3
-------
w
5
m
z
w
u
1.2
1.1 -
Apalachicola—2XC02 SCENARIOS
SEASONAL FLOW RATIOS
oct nov dec jan feb mar apt may jun jul aug sep
D OBSERVED
+ GISS2X
o GFDL2X
•a
Figure A.4
-------
V
n
-------
GO
od
(9
a
V
0
m
0.7
Apalachicola GISS2X
Seasonal Input Output Ratios
oct nov dec jan feb mar apr may jun jul aug sep
year
1
Figure A.6
-------
v
n
aJ
fi
v
o
n
Apalachicola—OSU2X
Seasonal Input Output Ratios
oct nov dec jan feb mar apr may jun jul aug sep
0.7 -
0.6
year
Figure A.7
-------
Hains
Figure A.8 shows the annual ratios of transient/baseline climate inputs (precipitation and pan evaporation)
and hydrologic outputs (flow) as a five year running average. The ratio data are easier to interpret than the flow
data alone because the ratio has the particular effects of the baseline period removed. There is no dear effect
of climate change on precipitation for the transient run. Precipitation ratios seem to fluctuate between wet and
dry trends with the long-term ratio being close to one. The climate change effects for the transient run seem
to be mainly associated with increases in pan evaporation which are robust and obvious.
During the first decade of the transient A scenario, when there was a decrease in precipitation as
compared with the base period, pan evaporation was already elevated with a resulting 10% decrease in runoff.
This is followed by a decade with enough more precipitation than the base period that the runoff almost equals
that of the base period for several years. This seems unusual at the beginning of a transient run of the GCM
when it is assumed that the CO, effects are minimal. In the fourth decade of the transient scenario the
precipitation ratio reaches its highest level and then begins to decline. The rest of the transient run follows an
expected pattern of an accelerating reduction in runoff driven primarily by increases in pan evaporation.
In order to present the effects of the transient scenario in the clearest comparison with the identical
period from the baseline data, the individual months were accumulated for both the base period and the transient
scenario. These were plotted on the same graph versus time as shown in Figure A.9. The effects discussed above
can be seen very clearly in the departures of the transient A curve from the observed curve. The initial
departure in the 1980's, the decreasing departure in 1990's and 2000's and the robust departure beginning about
2020 are vejy clear.
This same behavior is displayed in Figure A.10 where the two accumulated monthly curves are plotted
against each other. The straight line is the slope the curve would have if there were no effects. Where the slope
is less than the straight line, the transient runoff is less than the base period runoff. Where the slope is greater
than the straight line, the transient is greater than the base period.
A-38
-------
V
m
d
cd
a
«
o
m
a
K
1.3
1.2
0.9
0.8
0.7
Apalachicola~-TRANSA
5 year average Input Output Ratios
1980 1990 2000 2010 2020 2030 2040 2050 2060
+ ATLANTA PRECIP
O MACON PRECIP
Figure A.8
-------
m
A
v
C
0
S
I
CO
fc^
U m
~ ti
o S
* "
ed
PM
?
e
3
O
22
20
18
16
14
12
10
8
6
4
2 -
0 -
1980
Apalachicola—TRANSA
Cumulative Mass Plot
Obse rved
Trans
ent A
1990 2000 2010 2020 2030 2040 2050
Figure A.9
-------
m
A
+>
fi
o
S
I ~
w 2
b C!
u o
~3
. »"N
*
v
•p«
m
ti
cd
0
Apalachicola—TRANSA
Double Mass Plot
T
4
T
8 12
(Millions)
Observed (CFS-Months)
S
e.
Figure A.10
-------
POTENTIAL IMPACTS OF CLIMATE CHANGE ON THE
TENNESSEE VALLEY AUTHORITY RESERVOIR SYSTEM
by
Barbara A. Miller
W. Gary Brock
Tennessee Valley Authority
Engineering Laboratory
Norris,TN 37828
Interagency Agreement #DW64932639-01-0
-------
CONTENTS
EXECUTIVE SUMMARY 9-1
CHAPTER 1: INTRODUCTION 9-3
GLOBAL CLIMATE CHANGE 9-3
PROJECT OBJECTIVES 9-3
CHAPTER 2: BACKGROUND 9-4
THE TVA SYSTEM 9-4
Elements of the Integrated Reservoir System 9-4
Managing the Reservoir System 9-9
Sensitivity of the TVA Reservoir System to Climate Change 9-9
HYDROLOGIC OVERVIEW OF THE TENNESSEE RIVER BASIN 9-11
Historical Precipitation 9-11
Historical Runoff 9-11
CHAPTER 3: METHODOLOGY 9-13
THE WEEKLY SCHEDULING MODEL 9-13
Background and Development of the Model 9-13
Application of the Model 9-14
SUPPLEMENTARY MODELS 9-14
CLIMATE CHANGE SCENARIOS 9-14
LIMITATIONS OF THE METHODOLOGY 9-19
CHAPTER 4: IMPACTS OF CLIMATE CHANGE ON THE TVA RESERVOIR SY9-22
RESULTS OF THE WEEKLY SCHEDULING MODEL 9-22
Reservoir Operations 9-22
Flooding at Chattanooga 9-24
Power Operations 9-29
IMPLICATIONS FOR TVA SYSTEM OPERATIONS 9-31
Reservoir Operations 9-31
Navigation 9-31
Flood Control 9-32
Dam Safety 9-32
Power Operations 9-33
Water Quality 9-33
Recreation 9-35
Water Supplies 9-35
CHAPTER 5: SUMMARY AND CONCLUSIONS 9-37
METHODOLOGY 9-37
IMPACTS OF CLIMATE CHANGE ON THE TVA SYSTEM 9-38
CONCLUSIONS 9-45
LIST OF ABBREVIATIONS, SYMBOLS, AND DEFINITIONS 9-46
REFERENCES 9-47
11
-------
Miller
EXECUTIVE SUMMARY1
The objective of this project was to identify the sensitivity of the Tennessee Valley Authority (TVA)
reservoir system to global climate change. The range of potential impacts on reservoir operations, navigation,
flood control, hydropower production, water availability, water quality, and recreation were evaluated for the
Tennessee River Basin. Implications of these changes and potential adaptive strategies were also outlined for
the reservoir system and related TVA programs.
The Weekly Scheduling Model (WSM), a reservoir planning model, was used to assess climate change
impacts on the TVA reservoir system. Monthly ratios for surface runoff (2xCO2/CONTROL) provided by the
Environmental Protection Agency (EPA) for two climate scenarios were used to adjust historical local inflows
into each project for the 30-year study period (1951-1980). The weather scenarios, based on data generated by
NASA's Goddard Institute for Space Studies (GISS) general circulation model (GCM), represent potential
climate changes resulting from a doubling in atmospheric CO2 concentrations (possible by the mid-21st century).
The first GISS scenario predicts a warmer and wetter climate for the Tennessee Valley region. East
of Chattanooga, runoff increases 31% annually. In this eastern basin, which contains all the large tributary
storage reservoirs, flows are exaggerated during the traditional flood season (75% in March) and decreased
during a dry time of the year (-28% in November). West of Chattanooga, monthly ratios are more variable,
but the integrated effect of changes in runoff to the Tennessee River is to increase local inflows most of the year.
The second scenario, GISS Inverse, simulates a warmer and significantly drier climate for the TVA
region. In the eastern basin, runoff is substantially reduced during the traditional flood season and only modestly
increased diiring the dry period of year. The net effect is a 31% reduction in average annual runoff and low
inflow rates throughout the year. West of Chattanooga, runoff ratios are more variable, but the general trend
is to reduce runoff from current levels.
The increased runoff predicted by the GISS scenario results in higher reservoir elevations throughout
the year at all major projects in the TVA system. At the large storage projects in the eastern basin, normal
operating levels are met or exceeded in the fall; normal maximum levels are exceeded in wet years during the
traditional flood season; and summer full pools are maintained for an extended period of time. Exceedence of
normal maximum reservoir levels would likely result in spillage from dams and increased flooding in the
Tennessee Valley. Flood-prone urban areas in Tennessee such as Chattanooga and low-lying agricultural areas
could be particularly vulnerable to flood damage. Safety issues at dams and nuclear power plants would also
need to be reevaluated.
Primary benefits of the GISS scenario include a 16% average annual increase in hydropower production,
valued at $54 million; enhanced recreational opportunities; and improved water availability for water supplies and
minimum flow requirements. Water quality impacts would be variable and site specific, depending on the
relative influence of increased flows and assimilative capacity versus increased nonpoint source pollution.
Under the GISS Inverse scenario, decreased runoff would result in an overall decline in storage and
water availability at major projects in the TVA system. At the tributary storage reservoirs, lake levels are
lowered throughout the year, with median levels up to 9 meters (30 feet) below the base case and minimum
levels often below normal minimum pool levels during dry years. Due to constraints in the WSM, mainstem
reservoirs are filled to normal operating levels and minimum downstream flow and navigation requirements are
met, but at the expense of severely reducing water levels at the tributary projects.
'Although the information in this report has been funded wholly or in part by the US. Environmental
Protection Agency under Interagency Agreement #DW64932639-01-0, it does not necessarily reflect the Agency's
views, and no official endorsement should be inferred from it.
9-1
-------
Miller
Adverse impacts of the GISS Inverse scenario include a 24% decrease in annual average hydropower
generation with a replacement value of $87 million dollars; seriously impaired water quality; degraded
recreational opportunities; and decreased water availability for water supplies. The ability of many TVA projects
to fulfill their present multipurpose functions could be threatened The benefits of this scenario are primarily
related to reduced flood potential and the reduced probability of dam failure.
Both climate scenarios could significantly impact the operation of the TVA reservoir system. Substantial
changes in the reservoir guide curves and operating philosophy, as well as potential structural changes and/or
additions to the system, would be required to respond to an altered climate. Under the wetter GISS scenario,
flood control and safety issues would predominate with the need to create additional flood capacity through
operational changes, dam modification, and/or flood protection works. Added turbine capacity could be justified,
while more pumped-storage could be needed for peaking purposes. Nonpoint source pollution control programs
would need to be expanded; while enhanced recreational opportunities would encourage economic development.
Under the drier GISS Inverse scenario, drought-related issues would increase in significance. Difficulty
in satisfying project purposes would necessitate a reordering of TVA priorities. Alternative sources of energy
would need to replace lost hydropower potential, while fossil and nuclear plants may be unable to meet thermal
and/or safety standards owing to decreased flows and elevated water temperatures. Industrial and municipal
treatment facilities could be subject to more stringent waste standards. Increased power costs, decreased
recreation revenues, and increased industrial restrictions could have significant adverse impacts on the Tennessee
Valley economy.
The project objective was to assess the sensitivity of the TVA reservoir system to extreme climate
changes and to identify the implications of these changes. The study was not intended to predict the future
climate in the TVA region. Given the noted sensitivity of the TVA reservoir system to climate change, the
recent extended drought in the Tennessee Valley, and the general scientific consensus that atmospheric changes
are highly probable, this issue should be investigated in more detail. Climate change issues should have an
increasing role in TVA long-range planning and capital expenditures.
9-2
-------
Miller
CHAPTER 1
INTRODUCTION
GLOBAL CLIMATE CHANGE
Concern over changes in global climate caused by rising concentrations of CO2 and other greenhouse gases
has increased in recent years. A doubling in CO, concentrations, possible as early as 2030, could increase global
air temperatures by LS to 45°C (National Academy of Sciences, 1987). Increases in temperature are likely to
be accompanied by changes in the frequency and distribution of other climate variables such as precipitation,
wind, evapotranspiration, and runoff. Consequent alterations in regional hydrologic cycles would necessitate a
revaluation of the priorities and operating philosophies of water resources systems to meet the competing
demands of energy production, flood control, water supply, navigation, recreation, and environmental quality.
To assess the potential range and sensitivity of regional effects of climate change, the Environmental
Protection Agency (EPA) contracted with TVA to evaluate the impact of two climate scenarios on the TVA
reservoir system. This project was one of approximately SO case studies sponsored by EPA to evaluate potential
climate change impacts in the United States.
PROJECT OBJECTIVES
The Tennessee Valley was selected by the EPA for a case study of potential climate change impacts because
it represents a large, comprehensively managed water resources system. Based on hypothetical climate scenarios
provided by EPA, the TVA Engineering Laboratory proposed to evaluate the sensitivity of the TVA reservoir
system to extreme climate changes. The range of potential impacts to reservoir operations, flood control,
navigation, hydropower production, water availability, water quality, and recreation are evaluated for the
Tennessee River Basin. In the assessment of climate change impacts, model results are extrapolated where
feasible to determine potential impacts on the need for capital expenditures and/or other system alterations.
The results of this study are not intended to predict precise changes, but to identify the elements of the
TVA reservoir system and related programs most sensitive to climate change. A more detailed version of this
report is provided by Miller and Brock (1988).
9-3
-------
Miller
CHAPTER 2
BACKGROUND
THE TVA SYSTEM
The Tennessee River drainage basin encompasses a seven-state area in the southeastern United States,
including parts of Tennessee, Mississippi, Alabama, Georgia, North Carolina, Virginia, and Kentucky. The
105,931 square kilometer (40,900 square mile) area includes a population of approximately 5 million people. The
TVA, a multiple-purpose Federal Agency, operates a wide variety of programs within the region, including water
resources and reservoir management; power production; regional agricultural and industrial development; and
natural resource conservation and protection. A map of the Tennessee River Basin and the TVA power service
area, which is almost twice the size of the basin, is shown in Figure 1. Summary statistics of major TVA activities
are given in Table 1.
The TVA reservoir system, which is comprehensively managed as an integrated system, includes 42 major
dams and reservoirs. The TVA Act of 1933 directs that TVA operate these dams and reservoirs primarily for
the purposes of navigation, flood control, and hydroelectric power generation. Additional considerations include
recreation, water supply, resource conservation, and environmental management. Major climate changes could
have profound effects on the operation of such a multipurpose system.
Elements of the Integrated Reservoir System
Thirty-three of the major dams and reservoirs operated by TVA are shown schematically in Figure 2.
Although all dams and reservoirs within the TVA system serve multiple functions, major projects can be
categorized based on their primary purpose: single-purpose power projects; tributary multipurpose reservoirs;
or mainstem multipurpose reservoirs.
Twelve reservoirs within the system were built or acquired strictly as hydroelectric generating facilities.
These reservoirs are generally relatively small with minimal flood storage capacity.
The 21 tributary multipurpose reservoirs in the Tennessee Valley were built to ensure adequate flows for
mainstem navigation, flood control, and power production. These reservoirs are deep and provide key flood
storage capacity for the system. Flood control dominates the annual operating cycle, and the tributary reservoirs
are operated to fit the historical annual streamflow cycle in the region. As illustrated in Figure 3, tributary
reservoirs are lowered to flood control levels by January 1 to provide storage for the heavy flows of winter and
early spring. Normally, spring rains are allowed to fill the reservoirs from mid-March until full-pool levels are
reached by the end of May or early June. During the summer, when streamflows are typically low and power
demands high, lake levels are lowered significantly. Reservoir levels continue to recede through the fall until
winter flood control levels are once again reached by January 1.
The nine major reservoirs on the mainstem of the Tennessee River constitute the third category ~ mainstem
multipurpose reservoirs. In addition to hydroelectric faculties, each of these nine projects has navigation locks.
The locks and dams create a navigation channel with a minimum depth of 33 meters (11 feet). This inland
waterway, which extends 1,046 kilometers (650 miles) from Knoxville, Tennessee, to Paducah, Kentucky, links
the Tennessee River with the Ohio and Mississippi Rivers.
Navigation constraints affect the operation of these mainstem reservoirs. Pool fluctuations are severely
limited to maintain the minimum navigation channel depth. The typical mainstem drawdown range between
summer "full pool" and winter flood minimum levels is only 0.6 to 2 meters (2 to 7 feet). In contrast, some of
the deeper tributary reservoirs are drawn down by as much as 23 to 30 meters (75 to 100 feet) (see Figure 3).
9-4
-------
Figure 1
Map of the Tennessee River Valley
/ VA
LEGEND1
TVA Power Service Area
Tennessee River Watershed
ENG LAB 1988
-------
Miller
Table I
Summary of Major TVA Programs*
PROGRAM
GENERAL INFORMATION
Reservoir Operations Major Dams and Reservoirs:
2. Navigation
3. Flood Control
4. Power Production
5. Recreation
6. Fisheries (Sport)
31 TVA
4 Alcoa
8 Cumberlands
Water Surface: 259,590 hectares (641,455 acres)
Shoreline: 18,012 km (11,195 mi)
Length of Waterway: 1,050 km (676 mi)
Channel Depth: 3.3 m (II ft)
Traffic (1983): 25.4 million metric tons (28 million tons)
Private Investment (1933-1984): $4.8 billion
System Detention Capacity:
Jan I: 14.7 billion m5. (11,908,860 Ac-ft)
Mar 15: 12.9 bi11 ion m5 (10,548,260 Ac-ft)
Sunnier: 3.3 billion m5 (2,671,300 Ac-ft)
Total Generating Capacity (1966): 32,092 MW
55% Coal-Fired
18.4% Nuclear
14% Hydro
7.8% Combustion Turbine
4.8% Pumped-Storage
Power Service Area:
Area: 202,000 sq km (77,992 sq mi)
Population: 7.1 million
Tennessee River Watershed:
Area: (05,960 sq km (40,910 sq mi)
Population: 4.9 mi 11 ion
Public Access and Recreational Lands: 92,130 hectares (227,643 acres)
Recreation Visits (1986): 74.7 million
Value of Development and Equipment (1978): $630 million
Fishing Trips (1986): 17.1 million/year
Catch: 6.8 - 9.1 million kilograms/year (15.0-20.1 million Ib/year)
Cost of Goods and Services: $400 million/year
*From: TVAHandbook, 1986
9-6
-------
Miller
Figure 2
Schematic Diagram of the TVA Water Control System
{ft " ' I J ^A
t- "•* t /-£^ \ ^r^
jfr-f-sHe sjf^r1
i Comoonr =< Aimnco oom« avoiM by [A),
Corp* ol En^KMn «im otnona oy >Q.
DIAGRAM OF
WATER CONTROL
SYSTEM
ENG LAB 1988
9-7
-------
Miller
Figure 3
Typical Reservoir Operating Levels
sso
I" »+o
»30
tfO
i.e vet
*ULTI*J*K3Se LfVFLS OVIHH9
^jress sea om.r
'LOOO COHTKX. MMTfOHS,
r * i * j j t s o *
TYPICAL rXIBUTA*r ffSfflVONf &t*XTIM tfVfLS
On tributary ruirooirt.
lakt levels provide Storage
capacity for the winter flood
teason, then normally rite to
hi$h levels by the end of spring.
The stored water is gradually
released to augment the smaller
river flows of summer attd fall
for flower production, water
Supply, and other needs.
sia
so*
§ so*
j£ 3OO
5-
I.«
tiemnt.
I I
* it j j * s o
uMMgntKHn testxvoi* anjnurmf Ltvns
Water levels on mainstream
reservoirs eon fall only a frto
feet for winter flood eetftnol.
beeaute adequate water derpfru
must be maintained in the
navigation chormti to the next
dam upstream.
ENG LAB 1988
-------
Miller
Reservoir Sstem
Scheduling water releases from the TVA reservoirs is a complex process. Daily and seasonal operations
must account for the following: the amount of water in storage; travel time through the system; unregulated local
inflows; and daily weather, streamflow, and power demand variations. River control managers must determine
on a daily basis the rate and quantity of water to be released from each dam to obtain maximum benefits while
meeting the seasonal operating goal -- the orderly filling of reservoirs during the spring and the lowering of the
reservoirs during the summer and fall. This transition from minimum winter flood levels to full pool summer
levels, ideally hi a manner such that minimal water is spilled and hydroelectric energy potential lost, influences
operational guidelines for each reservoir in the TVA system.
Consistent with providing flood control and navigation, TVA gives emphasis to maximum production of
hydroelectric power. Hydropower is a low-cost source of electricity that is unaffected by increases in fuel and
construction costs. It is also the most immediately responsive source of power, which makes it ideal for peaking
power, supplying additional power quickly during those times of the day when power demands are highest.
Operation of the reservoir system for maximum hydroelectric production, however, involves unavoidable conflicts
with other public needs and concerns such as recreation and water quality issues. TVA is presently Devaluating
the Operation of the reservoir system to resolve some of these conflicts in light of today's needs and changing
priorities.
Sensitivity of the TVA Reservoir System To Climate Change
Major elements of the TVA reservoir system are presented schematically in Figure 4. Overland and
tributary inflow rates resulting from runoff over the Tennessee River Basin directly affect reservoir pool levels
and dam flow releases. Pool levels and discharge rates, in turn, directly impact the primary functions of the TVA
reservoir system — flood control, navigation, and power production — as well as other important functions --
water supplies, recreation, the protection of wildlife and fisheries, and the enhancement of water and
environmental quality.
The effects of the occurrence of extreme weather events, such as floods or droughts, under the present
climate regime are illustrative of the sensitivity of the TVA reservoir system to climate shifts. Flood-producing
storms force the rapid movement of huge volumes of water through the system, generally resulting in the spillage
of water and the bypassing of hydroelectric generating facilities with consequent losses in power production.
Under flood conditions, flood-prone areas can be inundated resulting in economic losses to urban and
agricultural regions.
Conversely, droughts substantially reduce the streamflow throughout the system. Summer full pool levels
become impossible to meet in some reservoirs, often resulting in the deterioration of water quality through
stagnation, excess algae growth, and lowered assimilative capacity due to reduced dissolved oxygen content.
During an extreme drought, water supplies could be threatened, and wildlife and fisheries adversely affected.
Hie current reservoir operation strategy involves the balancing of daily weather and streamflow variations
against seasonal operating goals. These goals are based on SO years of operating and flood experience
determined by the historical annual streamflow cycle. Major changes in the climate regime and streamflow cycle
would probably necessitate the development of alternative operational strategies to respond to new weather
patterns. If altered operational strategies and policies were insufficient to meet the operating goals of the system,
then alternative methodologies would need to be explored. For example, under an extremely wet climate
scenario, inadequate reservoir storage could be supplemented by flood control works, such as levees, or
modifications to existing dams to increase flood storage capacity. At the same time, new capacity could be added
to existing hydropower facilities to take advantage of increased flows. Conversely, a significantly drier climate
could mean the construction of additional storage reservoirs, more stringent environmental codes, a greater
emphasis on nuclear power, and/or the reduced use of stream flow for cooling purposes at existing fossil plants.
It must be emphasized, however, that the frequency and distribution of extreme events is as important as changes
in long-term average conditions in determining the overall impact on the TVA system.
9-9
-------
Miller
Figure 4
Primary Functions of the TVA Reservoir System
CLIMATE
Meteorology/ Hydrology
RESERVOIR
OPERATING
STRATEGY
RESERVOIR
POOL LEVELS
a
PROJECT RELEASES
WATER
SUPPLIE
WILDLIFE
a
FISHERIES
WffTER
CUAUTVC RECREATION
1988
9-ID
-------
Miller
HYDROLOGIC OVERVIEW OF THE TENNESSEE RIVER BASIN
Historical Precipitation
The TVA region is one of the wetter regions in the United States. Long-term (1890-1986) average annual
precipitation in the basin is 130 centimeters (51 inches) per year, although average annual precipitation may vary
in the eastern mountainous regions from 254 centimeters (100 inches) to 102 centimeters (40 niches) in more
sheltered areas. March is generally the wettest month, while September and October are likely to be the driest.
Hydrologically, the TVA basin can be divided into two distinct regions: the mountainous region to the east
of, or above, Chattanooga; and the relatively flatter region to the west of, or below, Chattanooga. While the two
regions presently receive approximately the same total annual precipitation, the annual distribution varies.
Intense summer storms are more likely to occur in the eastern region, while the western basin generally receives
higher precipitation in the winter and spring.
Historical Runflff
Long-term (1903-1986) mean annual runoff in the Tennessee River Basin is 56 centimeters (22 inches).
Annual amounts, however, have varied from 28 centimeters (11 inches) in an extremely dry year to 84 centimeters
(33 inches) in the wettest year. For the 1951 to 1980 study period, approximately 58 centimeters (23 inches) of
mean annual runoff was observed in the Tennessee River Basin. The annual variation in runoff for the historical
and study periods is shown in Figure 5. The study period typifies the long-term record; however, it does not
incorporate the extremely dry periods during 1904,1941,1981, or the current drought beginning in 1984.
The annual distribution of runoff is influenced by rainfall, soil conditions, and evapotranspiration patterns.
Runoff is heaviest in the winter and early spring (December-May) when the vegetation is dormant and the
ground is saturated. As the growing season commences, infiltration and evapotranspiration increase, and runoff
decreases substantially through the summer and early fall.
9-11
-------
30
X
2
20 —
i
ENG LAB I960
Figure 5
Annua] Runoff in the Tennessee River Basin from 1903-1986
TENNESSE RIVER BASIN - ANNUAL RUNOFF
1 1
1903 - 1986 ANNUAL
1303 - 1986 AVERAGE
4-
4-
\
STUDY PERIOD
} 1951-1980
4-
1910 1920
1930 1940
1950 1960
1970
1980
-------
Miller
CHAPTER 3
METHODOLOGY
THE WEEKLY SCHEDULING MODEL
Background and Development of the Model
The Weekly Scheduling Model (WSM), an operational planning model, was the primary model used to
assess climate change impacts on the TVA reservoir system. This model was developed by TVA to simulate
long-term, week-to-week variations in water level, discharge, and power production for 42 reservoirs operated
within the system. Typical uses of the model include power system scheduling; reservoir operations planning;
and assessment of new or modified operating policy. A brief description of WSM follows; more detailed
information concerning the development and application of the model has been provided by Shane and Gilbert
(1981), Gilbert and Shane (1982), Shane (1984), and Waffel (1985).
The model is based upon a linear programming algorithm, which is used to schedule the reservoir system
in weekly time steps. This algorithm selects a weekly reservoir schedule for each reservoir in the TVA system
by sequentially satisfying a series of operating objectives in a prescribed order of priority. Each objective is
satisfied to the greatest extent possible without causing a deterioration in higher level objectives. Current
operating policy, as defined by the TVA Act, is simulated to mlnimiTe deviation from historically normal
operating levels, subject to reservoir release and level constraints imposed to meet navigation, flood control, water
supply, power generation, water quality, and recreational requirements. These operating constraints and
long-range guides are summarized in the model as weekly ideal operating bounds on project release, headwater
elevation, and power generation. Deviations from current policy are modeled by eliminating individual objectives,
changing priorities, or adding special objectives.
For each week, local inflows for the week are specified at all reservoirs. At 23 reservoirs, storage at the end
of the week is predetermined based on headwater elevation guides. These 23 prescribed reservoirs are primarily
mainstem reservoirs or tributary power projects with minimal storage capacity, which have restricted operating
curves. In the remaining 19 reservoirs, which are primarily larger, multipurpose tributary reservoirs, the end-
of-the-week storage is simulated based on satisfying model constraints and objectives. The model selects release
schedules that satisfy operating objectives to the extent possible while satisfying simple continuity expressions.
The 19 storage values computed for the end of the week are used as starting water elevations for the following
week and the process is repeated
Based on the end-of-week storage, reservoir headwater elevations are computed based on mathematical
formulas derived for level water surface conditions. Average weekly hydrogeneration, spill, plant capability, and
power production costs can be derived from storage and release information. This output data can be
interactively plotted to examine the operations at individual reservoirs or to analyze system performance data
The WSM will model the reservoir system operation for each hydrologic year selected by the user from an
unbroken hydrologic record of local inflows. In the TVA system, this record extends from 1903 to the present.
Generally, the entire hydrologic record is used for model simulation to develop a range of probable operations
based on the historical streamflow pattern. The resulting envelope of curves, referred to as probabilistic pool
level forecasts, is used to evaluate weather and/or policy impacts on the operations of a specific reservoir. These
curves are described in the List of Abbreviations, Symbols, and Definitions.
Application of the Model
A primary input into the WSM is historical local flow into each reservoir in the TVA system. Local inflows
result from unregulated overland and tributary flows into each project. It was assumed that changes in runoff
9-13
-------
Miller
are directly proportional to changes in local flows. Consequently, in the application of the WSM, monthly ratios
(iZxCOj/CONTROL) for surface runoff provided by EPA for two climate scenarios were used directly to adjust
historical local inflows into the TVA system for the 30-year study period (1951-1980). As dictated by EPA, it
was also assumed that the monthly ratios applied equally to each week within a month in the WSM.
Results of the WSM are presented for a base case and the two climate scenarios. Historical inflows for the
study period are used for the base case, while appropriately adjusted inflows are used for the climate scenario
runs. Due to the integrated management of the TVA reservoir system, the model must be applied to the entire
system.
In the application of the WSM it was also assumed that existing guide curves, or normal operations, defined
the target or preferred operating conditions for each reservoir. Furthermore, the set of constraints used with
the WSM were based upon current operating policy.
SUPPLEMENTARY MODELS
Supplementary models to the WSM were used to perform power cost analysis and to assess flooding
potential at Chattanooga, an urban area with a high damage potential. In the power cost analysis, results of the
WSM were used to determine average weekly peak and offpeak hydrogeneration cost differences for the base
case and climate change scenario usmg projected cost data for 1988.
Estimated changes m flood stages at Chattanooga were determined using a model (FLDRT) developed to
estimate weekly tributary release constraints for the WSM. The model u not an operational flood routing model
used for day-to-day operations. The model uses historical daily average inflows (unregulated local inflows to Ft.
Loudoun-Teffico, Watts Bar, and Chickamaoga, the three main river reservoirs above Chattanooga), as weO as
fixed generating rule curves and balanced storage principles, to estimate how much tributary release from seven
upstream tributaries could be accepted during these historical events without exceeding flood stage criteria at
Chattanooga. These constraints are then used in the WSM for the long-term simulations, and the WSM observes
these constraints to the extent possible without violating higher priority constraints (e.g., the WSM will not fill
tributary reservoirs to the top of gates to avoid the mception of minor flood damage at Chattanooga). After the
WSM simulation has been performed, the resuHag tributary releases for the historical flood events can be routed
back through the FLOUT model to determine what the resulting peak flood stage at Chattanooga would be.
For both the GBS and GISS Inverse scenarios, historical runoffs were multiplied by the ratios referenced
earlier. Whether these "average" ratios are applicable for larger flood events is still a matter of speculation that
cannot be adequately addressed within the scope of this study. For the GISS scenario, it is possible that just the
increase in unregulated main river local inflows is sufficient to cause appreciable increases in flood stage at
Chattanooga as compared to base case events. Since the tributaries are also receiving increased inflow, there
are occasions when they cannot store afl the extra runoff, so they are forced to release water to the main river
reservoirs, which compounds the problem and increases flood damage even more.
CLIMATE CHANGE SCENARIOS
Data for the climate change scenarios were generated by NASA's Goddard Institute for Space Studies
(GISS) general circulation model The model ihnuhto the physical processes of the atmosphere and oceans
to estimate global climate. The GBS model has global coverage on a 10* longitude by 8* latitude grid The
GISS data provide monthly average (tafcea over a 10-year period) values for key climate parameters, including
surface temperature, Dretipttation, ruaofl; mmwfty, solar radiation, doud cover, and wind speed For each
parameter, monthly average values for a control run, a 2xCO2 run and the ratio (2xCOVcontrol)were provided
The control run was based on a IxCOu level of 315 mg/L, which was valid around 1958. The 2xCO2 levels
represent the endpoint effects of a doubling in COj atmospheric concentrations (£30 mg/L). The computed
9-14
-------
Miller
ratios, which were applied to the TVA historic data for the study period 1951-1980, reflect potential climate
changes on climate variables due to this doubling of CO2 concentrations.
The TVA region extends from approximately 34°-37°N latitude and from 81°-89°W longitude.
Consequently, the TVA region falls within two of the GISS grids, which are defined by 35.2TN latitude-90° W
longitude and by 3522°N latitude-80°W longitude. The grid point at 35.22°N defines an area extending from
31 JO*N to 39.13°N, which encompasses the entire TVA region in the north-south direction. In the east-west
direction, the 80°W grid point defines a region extending from 75" to 8S°W, which essentially includes the
portion of the Tennessee River Valley east of Chattanooga. The 90°W grid point, extending from 85° to 95°W,
includes the portion of the Valley west of Chattanooga. This east-west division at Chattanooga closely parallels
the two major hydrologic units in the Tennessee Valley.
In interpreting the GISS data it is important to note that the grid sizes are large and incorporate a wide
range of topography and local climate variation. The GCM model also has a relatively smooth terrain and does
not account for sharp hills and mountains like the real earth. Furthermore, while the 90°W grid has 100% land
coverage, the 80°W grid extends into the Atlantic Ocean and has only 76% land coverage.
GISS Scenario. The GISS scenario, provided by EPA, simulates a warmer and wetter climate for the Tennessee
Valley region. Based on the given ratios for temperature, precipitation and runoff, the effects of the GISS
scenario data on TVA historical data are presented in Figures 6 and 7. In the eastern portion of the Valley
(80°W), the GISS model predicts a 4.(TC (7.1°F) increase in average annual temperature. Monthly temperature
increases range from 2.1°C (3.7°F) in August to 6.2°C (1L2*F) in March. Precipitation increases in the
winter through summer months, with a slight decrease occurring in the fall. The greatest monthly precipitation
increases occur in Jury (41%) and in February and March (24%), while the largest decrease (-25%) occurs in
November. Average annual precipitation increases by 12% in the eastern basin.
Runoff follows a similar pattern in the eastern region, with a general increase in the winter through summer
months and decrease in the fall The largest increase in runoff occurs in March (73%), which coincides with
the traditional flood season. The largest decrease in runoff occurs in November (-28%), currently one of the
driest times of the year. The net effect in the eastern basin of the GISS runoff ratios is to exaggerate peak
flows during the wettest period of the year and further decrease flows during the driest period. Average annual
runoff increases in the eastern basin by 31%. These effects are illustrated in Figure 8, which compares median
natural streamflows for current and postulated 2xCO, conditions. The flows at Cherokee Dam are indicative
of the effects on the Upper Holston River Basin, while Chickamauga Dam illustrates the effects on the entire
eastern basin.
West of Chattanooga (90° W), the monthly ratios reveal some variability and more extreme ratios. On a
seasonal basis, the general trends in the west are similar, though less pronounced, to those for the eastern valley.
The GISS model predicts an average annual temperature increase of 53*C (95T). Monthly temperature
increases range from 3-5°C (63°F) in May to 8-5°C (153°F) in November. Minimal change in the average
annual precipitation is predicted in the western basin. On a seasonal basis, precipitation increases through the
spring and summer, and slightly decreases in the fall and winter. Average annual runoff in the western basin
increases by 10%. The greatest monthly increases in runoff occur in February (126%) and Jury (148%), while
the largest decreases occur in October (-25%) and November (-27%).
Effects of the GISS runoff ratios on median natural flows at Kentucky Dam are presented in Figure 8. As
Kentucky Dam is located at the mouth of the Tennessee River, these flows reflect the integrated influence of the
GISS data on the entire Tennessee River Basin. Increased peak flows in the early spring and decreased peak
flows in the fall result from the influence of GISS data in the eastern basin, while the variability in the data
results from the influence of the GISS data in the western basin.
GISS Inverse Scenario. The original intention of this project was to evaluate potential impacts of two different
climate scenarios based on runoff data provided by the EPA. The runoff data for the second scenario, generated
by the Princeton University Geophysical Fluid Dynamics Laboratory (GFDL) global circulation model, appeared
9-15
-------
Figure 6
Effect of GISS Scenario on Average Surface Temperature and
Precipitation in the Western (90°W) and Eastern (SCTM) Basins
78
60
SB
40
30
AVERAGE TEMPERATURE 30° U
1 1 I 1 1 1 1 1 1 1—f
HISTORICAL .-' -
-- GISS
SB
25
28
IS
18
AVERAGE PRECIPITATION 30° M
V
3
.38
.24
.18
.12
.86
.08
i—i—i—i—i—i—i—i—i—i—r
— HISTORICAL
•- GISS
.68
.51
.34
o
1C
u
.17
.88
90
88
78
58 -
48
30
AVERAGE TEMPERATURE 80° U
1 1 1 1 1 T
T I I I I
HISTORICAL
-- GISS
31
25
28
15
18
LJ
O
* F M
I F I M I * I H I J I * I * I s I ° l-!LUd
.38
.24
.18
.12
.86
.88
AVERAGE PRECIPITATION 80° W
1 1 1 1 1 1 1-
HISTORICAL
- CISS
.66
u
.34 "
.17
.00
ENG LAB 1988
-------
>-
<
o
\
2
Figure 7
Effect of GISS Scenarios on Average Runoff in the
Western (90°W) and Eastern (80°W) Basins
AVERAGE RUNOFF 30°W
Miller
.10
.85
HISTORICAL
GISS
GISS INVERSE -4
.52
- .39
a
\
z:
LJ
M A
M
D '
AVERAGE RUNOFF 80°W
.25
HISTORICAL
GISS
GISS INVERSE -4
.52
--.39
--.26
0
\
x
u
-------
Miller
Figure 8
Effect of GISS Scenarios on Median Natural Streamflow
at Cherokee, Chlckamauga and Kentucky Dams
CHEROKEE
i — i — i — i — i — i — i-
- HISTORICAL
--- GISS
— - GISS INVERSE
.5
. 4
.3
,2
. 1
.0
CHICKAMAUGA
1—i—r
HISTORICAL
--- GISS
— - GISS INVERSE
en
s:
u
-• 2
\ i—i r~T~r 4
a
0
KENTUCKY
200
1—i—i—r
—i—i—i—r
HISTORICAL
--- GISS
— - GISS INVERSE
5
4
3
2
1
0
9-18
-------
Miller
(treasonable for the TVA region. Although ample precipitation was apparent throughout the year, the runoff
values Indicated op to seven months of zero runoff in both the control and 2xCO2 model runs.
Experiments with the GISS 2xCO, model indicate that predicted water availability in the United States is
highly variable and influenced bymodefsensitiviry, model resolution, and sea surface temperature gradient (Rind,
1988). Consequently, to meet EPA project objectives and time limitations, EPA and the model developers
recommended utilizing the inverse of Che GISS runoff values. For example, if GISS runoff values indicated a
24% increase tot runoff in January for the eastern basin, then GISS Inverse runoff ratios would be computed on
the basis of a 24% decrease in runoff. While this approach does not show impacts of a particular model run,
the runoff values are within the range of the sensitivity analysis of the GISS model. Furthermore, the approach
illustrates the potential effects of a warmer and significantly drier climate, which has been predicted by other
GCM and hydrologic models used hi the southeast (Mains, 1988). The two scenarios together, therefore,
represent the full spectrum of impacts from the extremes of a wetter vs. drier climate.
The runoff ratios for the GISS Inverse scenario for the eastern and western basins ore presented in
Table 2. In the eastern basin, runoff decreases from January through October, with the largest decrease (-75%)
occurring in March. Runoff increases by 18 and 28% during November and December, respectively. The effect
of these changes in the eastern basin on median natural streamflow are illustrated in Figure 8 for Chickamauga
Reservoir. Flews during March, the traditional flood season, are significantly reduced Flows during November
and the beginning of December, while somewhat increased, remain at relatively low levels. Under the GLSS
Inverse scenario, the peak flows are reduced over the base case and shifted from the winter through spring to
the winter. The net effect of GISS Inverse in the eastern basin is to produce relatively low inflows throughout
the year.
In the western basin, Changes in runoff are more variable. The largest decreases in runoff (-100%) occur
in Wbroary and Juht, resulting in no runoff during those months. The greatest increase in runoff (52%) occurs
in March. The impact of these changes on median natural streamflow at Kentucky Reservoir, the last reservoir
in the TVA system, is shown in Figure 8. Flows are reduced during February, May, and July, are similar to the
base case from August through October; and are slightly increased in November. Peak flows, which are reduced
from the base case, occur in January and from March through April The overall effect is to reduce runoff and
natural streamflow for most of the year.
LIMITATIONS OF THE METHODOLOGY
Results of the WSM provide an excellent overview of the response of the TVA reservoir system, as an
integrated system, to changes in historical inflow or runoff. As noted previously, however, the application of the
model was based upon use of existing guide curves and operating policies. Furthermore, analysis of the results
of the WSM were based on current water use levels and seasonal patterns of power demand. In the event of
major di™«te shifts, current operating guides would most probably be adjusted to reflect changes in weather
patterns; model constraints would probably be modified to reflect changes in operating policies and priorities;
and temperature changes would alter water use patterns and power demands.
In this «f"pii^g level study, which was conducted to assess sensitivity of TVA operations to climate change,
GCM runoff ratios provided by EPA were used directly to adjust historical inflow data in the WSM. This
methodology assumes that changes in runoff are directly proportional to changes in local inflow. This approach
is reasonable to assess general trends in a large, complex system such as the TVA reservoir system. However,
in a more detailed analysis of climate impacts, precipitation, temperature, and other climate variables should be
used togenerate more site-specific runoff data using hydrologic models. Furthermore, groundwater effects, which
were not addressed in this study, should be evaluated.
In the application of the monthly runoff ratios there is also an implicit assumption that the distribution of
historical runoff remains the same. It is highly probable that, in a major climate shift, the distribution and
frequency of precipitation and runoff would be altered. The operation of a reservoir system such as TVA is very
9-19
-------
Miller
Table 2
Runoff Ratios for the GISS Scenarios
WESTERN BASIN RUNOFF
90"W LONGITUDE - 35.22'N LATITUDE
GISS GISS INVERSE
RATIO DIFFERENCE RATIO DIFFERENCE
JAN 0.600 -40.0 1.400 40.0
FEB 2.264 126.4 0.000 -100.0
MAR 0.482 -51.8 1.518 51.B
APR 0.850 -15.0 1.150 15.0
MAY 1.723 72.3 0.277 -72.3
JUN 1.060 6.0 0.940 -6.0
JUL 2.477 147.7 0.000 -lOO.O
AUG 0.692 -30.8 1.308 30.8
SEP 1.299 29.9 0.701 -29.9
OCT 0.748 -25.2 1.252 25.2
NOV 0.727 -?7.3 1.273 27.3
DEC 1.171 17.1 0.829 -17.1
EASTERN BASIN RUNOFF
80"W LONGITUDE - 35.22°N LATITUDE
GISS GISS INVERSE
% %
RATIO DIFFERENCE RATIO DIFFERENCE
JAN 1.237 23.7 0.7$3 -23.7
FEB 1.233 23.3 0.767 -23.3
MAR 1.750 75.0 0.250 -75.0
APR 1.392 39.2 0.608 -39.2
NAY 1.353 35.3 0.647 -35.3
JUN 1.436 43.6 0.564 -43.6
JUL 1.407 40.7 0.593 -40.7
AUG 1.272 27.2 0.728 -27.2
SEP 1.228 22.8 0.772 -22.8
OCT 1.186 18.6 0.814 -18.6
NOV 0.719 -28.1 1.281 28.1
DEC 0.813 -18.7 1.187 18.7
9-20
-------
Milk-1-
sensitive to the spatial and temporal distribution of precipitation and runoff events, particularly under extreme
drought or flood conditions. For example, in regard to flood control and dam safety issues, an evenly distributed
increase In runoff for each week of the month is easier to accommodate than a significant increase concentrated
in a short duration, such as the 3- or 9-day critical flood period. Similarly, it is easier to manage large storm
volumes emptied over multipurpose tributary reservoirs with significant flood storage capacity than the same
volume emptied over mainstem reservoirs with more limited storage capacity. These types of issues are beyond
the scope of this study.
Finally, direct use of the G1SS runoff ratios implies some degree of confidence in general circulation model
predictions of climate variables. These models, which are based on the fundamental physical laws governing
conservation of mass, momentum, and energy, incorporate a great range of time and space scales and must
approximate numerous, complex physical processes on a global basis. At present, the available general
circulation models provide a reasonable qualitative representation of large-scale climate features, but are less
reliable in representing the detailed regional and seasonal evolution of climate variables. Temperatures can be
predicted with a greater degree of confidence than more complex variables such as rainfall or runoff. The
current generation of general circulation models are useful for predicting perturbations to long-term average
conditions; they cannot, however, predict detailed weather 50 to 100 years in the future (MacCracken, 1986;
Crotch, 1988).
Some of the problems associated with using general circulation model results on a regional basis are evident
in the western basin (90°W) runoff data The monthly ratios are highly variable and may suggest osculations in
the GISS model lor that grid point. In the TVA region, all of the large multipurpose tributary reservoirs are
located in the eastern basin (80*W) above Chattanooga, where the runoff ratios appear to be more reasonable.
These tributary reservoirs have the greatest flexibility in operation and display the largest impacts to the climate
change scenarios. The eastern basin reservoirs also have the most direct impact on flood control at Chattanooga.
Consequently, in this analysis, the variability in the western basin runoff data had minimal effect on the overall
results and conclusions of the project. These problems emphasize, however, that this report represents a "first
cut" assessment of the sensitivity of the TVA reservoir system to climate change and does not attempt to predict
the furare climate for the Tennessee Valley.
9-21
-------
Miller
CHAPTER 4
IMPACTS OF CLIMATE CHANGE ON THE TVA RESERVOIR SYSTEM
RESULTS OF THE WEEKLY SCHEDULING MODEL
The Weekly Scheduling Model (WSM) was used to evaluate TVA systemwide impacts on reservoir
operations due to changes in runoff predicted by the GISS and GISS Inverse scenarios. Model application and
methodology are explained in Chapter 3, while the results of the model runs are described below.
Reservoir Operations
In the WSM, operations on the mainstem reservoirs and single-purpose projects are restricted and operated
according to fixed guide curves. Operations at the remaining 19 multipurpose tributary reservoirs are simulated
by the WSM to account for system constraints and runoff patterns, using normal operating curves as target
elevations. Operations at these 19 tributary reservoirs, however, are flexible and consequently reveal the largest
impacts due to climate change.
For illustrative purposes, the results of the GISS scenarios at Norris Reservoir are presented in the
probabilistic pool level forecast in Figure 9. Norris Reservoir, a large tributary storage reservoir located on the
Clinch River in eastern Tennessee, is typical of the multipurpose tributary reservoirs in the Tennessee Valley.
GISS Scenario. Runoff predicted by the GISS scenario results in higher projected elevations at Norris Reservoir
for all probabilistic plots for most of the year. In the eastern grid (80*W), although the GISS scenario predicts
slightly decreased runoff in November and December, the increased runoff and subsequent storage earlier in the
year enable Norris to meet its discharge requirements and still maintain slightly higher elevations during the late
faH
The projected median curve is higher throughout the year and at times more than 3 meters (10 feet) higher
than observed in the base case. It is also significant that the projected median elevations remain at or near the
normal maximum level providing full summer pool levels from the beginning of April into Jury. The upper
envelope curve, showing maximum elevations, indicates the reservoir will often be above the normal maximum
level during the traditional flood season and during very wet periods will fill to its total capacity. These higher
maximum levels from February through May greatly increase the probability of spill at Norris and downstream
projects, with likely flooding and possible dam safety implications.
The lower envelope curve indicates that minimum elevations at Norris generally will be higher, by up to 3
meters (10 feet), under the GISS scenario. Only during November and December, when GISS predicts decreased
runoff in the eastern basin, are the minimum elevations similar for the base and climate change scenario. The
significant drops in minimum elevations in April, noted in both the base and GISS scenarios, are caused
by the annual spring filling of mainstem reservoirs from tributary flows.
The results at the other multipurpose tributary reservoirs, most of which lie east of Chattanooga in the 80°W
grid, are similar in nature. In general, the higher runoff predicted by GISS in the winter through summer months
in this area results in the following:
Higher median projected reservoir elevations for most of the year;
Extended operations at higher levels during the summer, with median elevations often at or near
normal maximum levels;
9-22
-------
Figure 9
Probabilistic Pool Level Forecasts for Norris Reservoir
Miller
ItN
Base Case
ion
MM.
ION
1(00
1011.
1010
100*
5 ,000.
I m
i **°
a MI
M»
m
tn
Ml
GISS Scenario
.
10M
IOM
IOM
ion
1011
1010
IOM
MO
*;>
Ml
NO
GISS Inverse Scenario
LEGEND:
Median Projection -—Normal Operations "-Upper & Lower Envelope
ENG LAB 1988 ' 943
-------
Miller
Higher reservoir elevations in the fall, where the additional water can be used to satisfy downstream
minimum flow requirements;
Higher maximum elevations, particularly during the traditional flood season, resulting in increased
probability of spill at tributary and mainstem projects; and
Higher minimum reservoir elevations throughout most of the year.
As indicated by the generally higher reservoir elevations, the decreased runoff predicted hi the eastern portion
of the basin in the late fall is more than compensated for by the increased runoff and storage earlier hi the year.
The variable runoff predicted for the western basin does not appear to significantly affect the eastern tributary
reservoirs.
GISS Inverse Scenario. Runoff predicted by the GISS Inverse scenario results hi lower projected reservoir
elevations for Norris Reservoir, as illustrated in Figure 9. The probabilistic plots indicate reservoir levels would
be lower throughout the year, except during late fall and early winter when the GISS Inverse scenario predicts
increased runoff.
The projected median curve is lower most of the year and at times is approximately 5 meters (15 feet) lower
than observed in the base case. From November into January, the GISS Inverse median curve is projected to
be approximately the same as the base case. The upper envelope indicates that Norris did not exceed its normal
maximum pool level during this analysis and often approached the base case normal levels. During late fall and
early winter, the maximum curve is predicted to exceed base case levels, at times up to 3 meters (9 feet). The
lower envelope curve indicates that minimum elevations at Norris would be lower throughout the year, by up to
8 meters (25 feet), under the GISS Inverse scenario.
The results at the other tributary reservoirs are similar in nature. The runoff into these reservoirs is
predicted to be lower from January to November and results in the following:
Lower projected median reservoir elevations for most of the year;
Maximum reservoir elevations never exceeding normal maximum pool levels, often approaching
current normal levels; and
Much lower minimum elevations, with most reservoirs being drawn to or below normal minimum
pool levels.
in this analysis, Norris Reservoir was not drawn to minimum pool levels because of constraints in the WSM.
In actual operation it could also be drawn to its normal miniiiiiim level under this scenario.
GISS Scenario. Under the GISS scenario, during exceptionally wet years, storage at the tributary reservoirs is
inadequate during the flood season to provide the present level of control, and spillage from dams is likely to
occur, [hiring these flood periods, local flows into the mainstem reservoirs would also be increased, resulting
in an increased probability of flooding at Chattanooga, Tennessee, a major city on the Tennessee River with the
greatest urban damage potential in the VaBey. Under present conditions, the protection of Chattanooga against
excessive flooding is a major priority of die flood control system in the eastern basin. There is currently about
a 20% annual chance of minor flood damage and a 5% chance of substantial flood damage at Chattanooga.
9-24
-------
Miller
To assess the increased flood potential at Chattanooga, estimated daily regulated inflows from the upstream
reservoirs, as well as historical dafly unregulated local inflows into the mainstem reservoirs Ft. Loudoun-Tellico,
Watts. Bar, and Chickamauga, were routed past Chattanooga for both the base case and climate change scenarios
(see Chapter 3 for more detailed methodology). The results of this analysis are summarized in Figure 10 and
Table 3. These results, which are based on simplified models, show relative effects between, the current base case
and the climate change scenarios, and may not agree exactly with historical flood stages.
The flow duration curve presented in Figure 10 indicates that under the GISS climate change scenario, large-
magnitude flows would occur more frequently near Chattanooga on the Tennessee River at Chickamauga Dam.
The figure can also be interpreted as implying that flow magnitudes increase at each duration.
The potential effects of these increased flow magnitudes on river stages at Chattanooga are summarized in
Table 3 In a comparison of estimated flood peaks for five historic floods. Floods exceeding a 9.1 meters (30 feet)
stage cause serious flood damage hi the city of Chattanooga. At a 7.6 meters (25 feet) flood stage, barge
terminals and docking facilities are flooded, resulting in some economic losses. Under the climate change
scenario, flood stages are increased for four of these floods. During the estimated historic floods, river stages,
though above 7j6 meters (25 feet), were still less than 9.1 meters (30 feet) and were contained within the banks
of the Tennessee River. Under the GISS scenario, during two of these floods, river stages exceed 9.1 meters (30
feet) and severe flooding would occur in Chattanooga
The most dramatic effect occurs with the 1973 March flood where estimated historic river peak stages at
Chattanooga would increase from 9.0 to 17.1 meters (29.6 to 563 feet) and peak flows from 5,093 cms (181,900
cfs) to 14,700 cms (525,500 cfs). This represents an 8.1 meter (26.7 foot) increase in stage and 9,621 cms
(343,600 cfs) increase in flow. The largest runoff ratio for the GISS model occurs in March, which in this
instance coincides with a major flood event. This estimated GISS flood is greater than the largest recorded
flood in Chattanooga, which occurred hi 1867 with a peak flow of 12,850 cms (459,000 cfs). It should also be
noted that the maximum probable flood (regulated) for Chattanooga under current climate conditions has an
estimated flow of 17,000 cms (610,000 cfsX which is only 16% greater than the 14,700 cms (525,500 cfs) flow
estimated with the GISS scenario. A map of Chattanooga showing the maximum probable floodplain is
presented in Figure 11.
Based on a flood damage curve for Chattanooga, the following shows the range of potential damages:
Potential Flood Damage at Chattanooga
Stage Estimated Damage
(feet) (1988 dollars)
34 > 1 million
40 -100 million
56 -1 billion
Consequently, for the two largest floods shown in Table 3, damages to Chattanooga, in 1984 dollars, could range
from slightly under $100 million to $1 billion.
Actual peak flood stage for the March 1973 flood exceeded 9.1 meters (30 feet) and caused serious damage
in Chattanooga.
9-25
-------
Miller
Figure 10
Flow Duration Curves for the Tennessee River at Chlckamauga Darn
LCOCMDi
tfctc c*se
BUS
«t»SINVFft9E
6
I
r ,.
4O (K> «O YO
OF TIME rOUALLCO 0* EXCEEDED
100
F.NG LAB 1988
9-26
-------
Table 3
Potential Effects of GISS Scenarios on Flood Stages at Chattanooga
YEAR
1912
1936
1957
1973
1975
DATE
APR 2-APR B
MAR 26 -APR 8
JAN 29-FEB II
MAR 12 -MAR 18
MAR 26-APR 1
ESTIMATED*
HISTORIC FIOOO
STAGE* » FLOW
(FT) (1000 cfs)
26.9 155.0
26.8 154.5
26.9 155.0
29.6 181.9
26.9 155.0
ESTIMATED*
GISS FLOOD
STAGE" FLOW
(FT) (1000 cfs)
29.4 180.2
26.9 155.0
26.9 155.0
56.3 525.5
46.1 384.5
ESTIMATED CHANGE
DUE TO GISS
CLIMATE CHANGE
S1AGE«» FLOW
(FT) (1000 cfs)
2.5 25.2
O.I 0.5
0.0 0.0
26.7 343.6
19.2 229.5
ESTIMATED*
GISS
INVERSE ROOD
STAGE* » FLOW
(FT) (1000 cfs)
18.6 76.4
18.6 77.2
26.9 155.0
20.2 92.1
16.0 52.1
ESTIMATED CHANGE DUE
TO GISS INVERSE
CLIMATE CHANGE
STAGE** FLOW
(FT) (1000 cfs)
-8.3 -78.6
-8.2 -77.3
0.0 0.0
-9.4 -89.8
-10.9 -102.9
•Flood stage Is confuted by a dally model and Is Intended for order of magnitude comparative purposes only.
The three main river pools are assumed to be at the bottom of their winter operating zones at the beginning of
each event.
••Bank overflows at 30 ft stage.
Damage to docking facilities at 25-ft stage.
£
-------
Figure 11
Maximum Probable and 1867 Floodplains in the Chattanooga Area
•
FLOODED AREAS FOR
REGULATED
AND
M PRC.IB.JLBI.E FLOOD
REGULATED
-------
Miller
GISS Inverse Scerjiflrin. Under this climate change, even during wet periods, flood storage in the tributary
reMiyoira and local inflows into the main stem are reduced sufficiently to greatly decrease the probability of
flooding at Chattanooga. The flow duration curve presented in Figure 10 indicates that the magnitude of flows
at each duration would decrease at Crrickamauga Dam near Chattanooga.
The potential effects of these decreased flows on river stages at Chattanooga are summarized in Table 3.
Under this scenario, flood stages would be reduced for four of the five simulated floods, with die other flood
having the same stage as the estimated historic flood. The estimated stages for all five of these simulated GISS
Inverse floods are below 30 feet, the stage where serious flood damage begins.
Power Operations
(f ISS Sfffnario- For the GISS scenario, a major impact of the increased runoff predicted for most of the year
is an Increase in system hydropower generation. As indicated in Figure 12, the average weekly total system
generation would increase substantially from January through October, and decrease slightly in November and
early December. Weekly average offpeak and peak generation follow a similar pattern, except peak generation
also declines slightly in February. Maximum system generation also increases substantially throughout most of
the year, approaching system capability from January through June.
The results of the power cost analysis indicate that the enhancement in system generation under the GISS
scenario represents an average annual energy gain of 16%, or approximately 3,100,000 MWh (3,100 gigawalt
hours (GWh)). This includes an increase of 503 GWh during peak hours and 2,630 GWh during offpeak hours.
This energy gain produces an average annual power benefit of about $54.7 million in 1988 dollars with $11.4
million attributable to peak hour generation increases and $433 million to offpeak generation increases. Each
year of modeled hydrology experienced a power benefit, ranging from a low of $35.8 million for 1975 hydrology
to a high of $66.1 million for 1976.
Ahfanugfr the GISS scenario has positive effects on system power operations, the increases in generation
would have been greater if the mainstem projects were able to utilize more of the predicted flows for power
generation. During the winter and spring, the GISS flows often exceed the turbine discharge capacities of the
mainstem projects; therefore, the excess flows cannot be used for power production. The increased spill volume
and tatiwater levels also decrease the operating head and therefore the power output. Increased spill frequency
also affects the operating flexibility of the mainstem projects, causing more operation at plant capacity. This
•decreases the usefulness of these projects to satisfy peak load demands.
GISS inverse Scenario. The decreased runoff predicted for this scenario results in a decrease in system
hydropower generation. As indicated in Figure 12, the average weekly total system generation would decrease
mhjtafflitfcaiiy fjam, January through November. Weekly average offpeak and peak generation follow a similar
pattern. It should be noted that a substantial loss in dependable hydrosystem capacity also occurred in this
iario because of the reduced inflows and reduced operating heads. Expected seasonal capacity losses were
observed for wtually all years, with summer losses about four times higher than winter losses. Summer capacity
i xeached more than 1,700 MW for 1966 hydrology.
The results of the power cost analysis for this scenario indicate that the reduced system generation
represents an average annual energy loss of approximately 4,697,000 MWh (4,697 GWh), or 24%. This includes
a deecease of 1,980 GWh during peak hours and 2,717 GWh during offpeak hours. These decreases in energy
production result in benefit losses of $87.2 million in 1988 dollars with $42.9 million attributable to peak
generation losses and $443 million to offpeak generation losses. Each year of modeled hydrology experienced
a Joss of power benefit, ranging from a loss of $69.1 million for 1957 hydrology to a high of $106.8 million for
1975 hydrology.
9-29
-------
Miller
Figure 12
Climate Change Impacts on System Generation
2 « 6 • 10 18 14 l« 18 20 tl 24 2« It 30 32 34 38 38 40 48 44 4« 48 90 98
6ISS
LEGEND:
•PEAK
OFFPEAK
• TOTAL
2 4 • • M 12 14 W !• M t2 {4 M !• SO S2 94 96 M 40 42 44 46
48 SO 92
GtSS INVERSE
ENG LAB 1988
-------
Miller
IMPLICATIONS FOR TVA SYSTEM OPERATIONS
The remits of the WSM and associated analysis reveal that the increased flows predicted by the GISS
climate scenario and the decreased flows predicted by the GISS Inverse scenario could have major effects on
TVA operations. The programs or areas of activity which would experience the greatest impacts, based on
current operating policy, can be summarized as follows:
Reservoir Operations
The reservoir guide curves presently used to operate the reservoir system were developed based on historical
weather patterns and flood experience. The curves, which allocate storage space to be held in reserve during
the flood season (except when used for flood regulation), provide target reservoir elevations for different times
of the year to meet navigation, flood control, and power demand constraints.
Scenario. In the GISS scenario, the increased runoff predicted in the eastern region during the winter
through rammer months resulted in higher iftmimum, median, and maximum elevations throughout most of the
year. The higher maximum elevations indicate a high probability of spillage at all major tributary reservoirs
during exceptionally wet years. During the fall months when the GISS model predicted reduced runoff, the
increased storage from earlier months generally enabled the reservoirs to maintain higher fall reservoir levels.
These trends indicate that the current reservoir guide curves and operating philosophy would need to be
reevaluated* to provide more storage capacity during the flood season; take advantage of the opportunity for
longer summer full pool levels; adjust to changes in power load demands due to seasonal changes in temperature .
etc. Any reevaluation of the guide curves would also need to take into consideration the likelihood of increased
variability in the occurrence of extreme events and potential changes in the Probable Maximum Precipitation.
GISS Inverse Scenario. In this climate change scenario, the decreased runoff predicted in the eastern basin
generally results in lower minimum, median, and maximum elevations throughout the year. These lower
elevations indicate more reservoir storage would be available for flood control operations than in the base case
and should reduce the probability of downstream flooding. However, satisfying the other project purposes would
be more difficult to accomplish. During very dry periods it probably would be impossible to satisfy all the
competing purpose*. and the very low reservoir levels would create adverse public reaction.
The current operating philosophy and guides should be reevaluated in order to determine appropriate
normalminimum and normal maximum levels; determine the required flood storage for the GISS Inverse floods
and reallocate flood storages if necessary; and develop operating strategies that would minimize the adverse
impacts of this scenario.
A total reevaluation of the operation of a complex reservoir system that includes 42 major dams and
reservoirs and serves multiple functions is a major undertaking. A comprehensive review of TVA reservoir
operations and planning is currently in progress to provide policy guidelines for operation of the agency into the
21st century. The purpose of the study is to evaluate operational priorities, such as flood control, navigation, and
power vs. water quality and recreational needs in order to ensure that the agency is responsive to the changing
needs and values of the region. Major changes in the climate regime, and whether these changes are
incorporated into the project, could have an impact on the findings and policy implications of the reservoir review
study.
Navigation
Increased high-flow periods with greater flow magnitudes in the mainstem reservoirs could result
in more instances of flooding of industrial and docking facilities. The higher flow velocities associated with high-
flow periods could necessitate the suspension of navigation during certain times of the year. Under median-
flow conditions, however, the flow needs of the mainstem reservoirs for navigation purposes would be easily met
or exceeded under the climate change scenario.
9-31
-------
Miller
GISS Inverse Scenario. This climate change scenario would have little effect on navigation. Minimum pool levels
on the mainstem reservoirs, which were maintained in this scenario, provide ample depth for navigation.
However, as illustrated in Table 3, the decreased runoff of the GISS Inverse scenario generally results in reduced
flow magnitudes during flood events. This will reduce the frequency of flooding of docking facilities and the
interruption of navigation due to very high flow velocities.
Flood Control
GISS Scenario. The results of the WSM and the analysis of major flood events at Chattanooga indicate that
increased flooding could be a major impact of the GISS climate scenario. As indicated in the previous section,
flood damages at Chattanooga for a major flood event could range from slightly under $100 million to $1 billion
in 1984 dollars. In addition to Chattanooga, other flood-prone areas in Tennessee, such as Lenoir City, Knoxville,
Kingsport, Clinton, Charleston-Calhoun, and the agricultural area surrounding Savannah, would be vulnerable
to flood damages.
Increased flood control could be addressed by adjusting the reservoir guide curves or reservoir operating
elevations to create more seasonal flood storage; modifying existing dams to provide flood storage above existing
levels; or building additional dams to regulate flood flows. The cost of modifying existing dams, depending on
the type of dam and alteration required, can range from $5 to $25 million in 1987 dollars.
Alternatively, flood control structures, such as dikes or levees, could be constructed to protect flood-prone
areas such as Chattanooga. The cost of levee construction is approximately $1.5 million dollars per mile (1988
dollars) for a 7.6-meter (25-foot) levee.
A major impact of a significantly wetter climate such as the GISS scenario, would be the need to reevaluate
the Probable Maximum Precipitation (PMP) and Probable Maximum Flood (PMF). All recently constructed
major structures in the TVA system, including the nuclear power plants, were designed to withstand the PMF.
Any significant increase in this flood would entail a reevaluation of the safety of these projects.
GISS Inverse Scenario. In general this climate change scenario has favorable impacts on flood control The
results of the WSM and the analysis of major flood events at Chattanooga indicate that the frequency and
magnitude of floods would be reduced. This could result in opportunities for additional development along
floodplain areas of the Tennessee River Valley. However, before such activities are undertaken new floodplain
management studies should be made.
ain Safet
.Scenario. The increased likelihood of the major dams operating at or above normal maximum levels for
extended periods of time would necessitate a reevaluation of dam safety at these projects. Individual projects
could need increased storage capacity; increased capacity of existing spillways; additional spillways; or
strengthening and/or raising of the dam structures. Dam safety analysis would also need to take into account
possible changes in the PMF.
A complete analysis of dam safety is presently under way at TVA, and a number of dams are being
retrofitted to meet current safety standards. Safety modifications have been completed or are in the construction
phase at 11 dams, with project cost ranging from $300,000 to $11 million. Over the next decade, an additional
15 dams will be upgraded Given the higher flows predicted and increased hydropower potential predicted by
the GISS scenario, the question arises whether present dam rehabilitation work and capital expenditures should
be based upon historical hydrology or take into account the possibility of future climate changes.
, Inverse Scenario. The reduced runoff predicted in this scenario decreases the likelihood of operations at
or above maximum pool levels and reduces the probability of dam failure or of incurring major damage to the
9-32
-------
Miller
TVA dams. The current dam safety program is more than adequate to ensure continued safe operation of the
dams.
Power Operations
GISS Scenario. Under the GISS scenario, the estimated 16% increase in average annual system generation
(3,100 GWh), valued at $54.7 million (1988 dollars), suggests that many projects could justify added turbine
capacity. Given present faculties, during much of the year under the climate change scenario many hydropower
projects would need to operate at full capacity to keep the reservoir levels at or below normal maximum levels.
While this would increase the use of hydropower to satisfy the base power load, it would decrease the flexibility
to use hydropower for peak load demands. This could mean the need for more pumped-storage faculties to
supply power quickly during periods of peak demand and/or adding additional capacity at existing hydro plants.
In the TVA region the largest seasonal power demands are for summer cooling. The GISS scenario predicts
higher surface temperatures throughout the year. The largest increases in temperature occur in the fall, thereby
extending the period of warm weather. These changes, coupled with changes in hydrology, could necessitate the
need to reevaluate future power demands and the relative mix of hydro, fossil, and nuclear power to meet these
demands.
GISS Inverse Scenario. The estimated 24% loss in average annual system hydrogeneration (4,600 GWh), valued
at $87.2 million, would have to be replaced with alternative sources of generation, such as coal-fired, oil, gas, or
nuclear generation. Additional capacity would not be required at present to provide this generation. However,
increased power demand or retirement of power plants in the future may result in the need to build additional
capacity to supply this loss in hydro production. These large losses in generation and capacity on the hydro
system will make it more difficult to meet peak load demands. As in the GISS scenario, additional
pumped-storage plants or special fossil plants may be required to quickly supply peak power demands.
The reduced flows of the scenario could also have negative impacts on the fossil and nuclear plants. Water
quality constraints on discharge water temperatures were not considered in the WSM analysis. Even though
adequate volumes of water were supplied to satisfy the water quality constraints, the water may be too warm for
use as cooling water and cause suspension of operations at projects such as Bull Run, Colbert, and John Sevier
Fossil Plants. Although cooling towers are available at Sequoyah and Watts Bar Nuclear Plants, elevated water
temperatures could exceed safety standards for Essential Raw Cooling Water (ERCW) intake.
Water Quality
GISS Scenario. The increased runoff predicted by the GISS scenario could have both positive and negative
impacts on water quality. Most projects would be able to generally exceed present downstream minimum flow
requirements and, through much of the year, would have greatly increased flow releases. Increased dam releases,
resulting in higher flows or tailwater levels below the dams, would increase assimilative capacity of
these reaches.
Tailwater dissolved oxygen (DO) levels, a key parameter in determining the biological health of a river, are
largely determined by the quality of water released from the upstream dam. Most hydropower dams release
water from deep in the reservoir. Consequently, during the critical summer months when many storage
reservoirs in the TVA system are stratified, the DO level of the hypolimnic layer (or cold bottom layer) of the
reservoir determines the DO content of the releases.
Hypolimnic DO levels result from the complex interaction of a number of physical, biological, and chemical
factors including the temperature and quantity of inflows to the reservoir; the residence time through the
reservoir; and the organic loading rate from internal and external sources. Generally, increases in organic and
nutrient loadings and warmer hypolimnic temperatures increase the rate and extent of hypolimnic DO depletion.
Conversely, shorter residence times, or more throughflow (inflow/outflow) through the reservoir, elevate
hypolimnic DO levels. The increased runoff predicted by the GISS scenario will increase inflow quantities and
9-33
-------
Miller
most probably erosion rates, thereby raising inflow temperatures, increasing organic loading rates, and decreasing
residence times through the reservoir. These effects will have contradictory impacts on hypolimnic DO levels,
and the overall net impact will vary with specific reservoir geometry, time of year, and the relative balance
between opposing factors.
Due to the complexity of reservoir dynamics, it is difficult to predict the impact of the GISS model on overall
reservoir water quality without detailed site-specific investigations to study the dynamics of each reservoir. It is
highly probable, however, that the increased runoff predicted by the GISS scenario would increase nonpoint
source pollution, or the sediment, nutrient, and other chemical loads, into the reservoirs. The effect of these
higher loading rates could mean lower DO levels, higher turbidity levels, and/or excessive algae growth. On the
other hand, increased inflows could improve overall reservoir water quality by increasing reservoir mixing,
shortening the residence time through the reservoir, and/or providing a larger volume of water for dilution of
negative impacts.
GISS Inverse Scenario. The reduced runoff predicted by the GISS Inverse scenario is likely to have an adverse
impact on the water quality of TVA streams and reservoirs. In a typical TVA reservoir, reduced inflow volumes,
increased residence times, and low lake levels will result in a deterioration of the water quality. Owing to
increased air temperatures and solar radiation predicted by GISS Inverse, the normally warm upper layer of
water, the epilimnion, is likely to have increased water temperatures. As cold water in the lower level of the
reservoir (hypolimnion) is withdrawn for downstream releases, the warmer epilimnion will become larger and
deeper. Increased residence times allow natural processes to deplete more of the available oxygen resources,
resulting hi a reduction of the DO content of the hypolimnion.
The reduced runoff will likely decrease erosion, initially producing improved water clarity in the reservoirs.
Ultimately, however, the increased light penetration may promote luxuriant algal growths, causing decreases in
clarity and large diurnal variations in DO and pH. Increased clarity could also promote the growth of aquatic
weeds at locations not previously infested.
The severity of these water quality changes may be magnified by the normal assimilation of treated municipal
and industrial waste, particularly if waste and heat loads remain at current levels. The lower flows, higher
temperatures, and lower DO levels will reduce the ability of the reservoirs to assimilate permitted wasteloads.
The situation could be further complicated by the addition of waste pollutants such as color, dissolved solids,
and toxins.
The tailwaters, or riverine reaches below dams, will also be adversely impacted by the increased temperature,
reduced DO content, and decreased magnitude of the released flow. Many of the streams also provide for the
assimilation of municipal and industrial wastewater after appropriate treatment. Although treatment levels are
theoretically stringent enough to protect beneficial water uses under low-flow conditions, the extremely reduced
runoff produced by the GISS Inverse scenario may diminish streamflow below the critical levels normally used
in allocating wastes. The situation may become particularly severe at facilities that are incapable of achieving
satisfactory levels of treatment under present conditions.
Several papermill plants in the TVA region depend on dilution of their highly colored wastewater to avoid
aesthetically objectionable conditions downstream of their discharges. Following treatment of paper wastes, the
plants coordinate waste discharges from large storage lagoons with releases from TVA dams. Under low-flow
conditions, the plants must reduce the volume of discharged wastewater and subsequently increase storage in the
treatment lagoons. Under the GISS Inverse scenario, during dry periods, prolonged
low-flow conditions may force these plants to cut back on production, increase the storage capacity of the
treatment lagoons, or be in noncompliance with color requirements for the Tennessee River.
For water quality purposes, TVA has agreed to maintain minimum daily flows of 56 cms (2,000 cfs) at
Knoxville, 17 cms (600 cfs) at Charleston, 168 cms (6,000 cfs) at Chattanooga, 39 cms (1,400 cfs) below
Chilhowee dam, and 10 cms (350 cfs) or 1/3 of the plant intake at John Sevier Fossil Plant (JSF) for the bypass
requirement These constraints have high priority in the WSM, and the required flows are maintained under
9-34
-------
Miller
the GISS Inverse scenario. As discussed previously, however, during dry periods, reservoirs are drawn as
necessary (sometimes below acceptable levels) to meet these requirements.
Recreation
GISS Scenario. Under present operating policy, full-pool reservoir levels are generally reached by May or early
June. Through the summer months, when streamflows are typically low and power demands high, lake levels
are significantly lowered Under the GISS scenario, the results of the WSM indicate that the increased runoff
in the winter through summer months would produce extended operations at higher levels during the summer.
The ability to maintain summer full-pool levels for longer periods of time would greatly enhance recreation
benefits, particularly on the eastern tributary reservoirs. Extended summer pool operations increase opportunities
for boating, fishing, and other water-based activities, which is important in promoting tourism and economic
development. The priority assigned, and who should pay, to maintaining summer full-pool levels is at present
a major issue in reservoir operations. Recreation proponents are interested in maintaining summer pool levels
from April through October.
Under current operating conditions, the TVA reservoir system received approximately 74.7 million recreation
visits in 1986. Recreational development and equipment in the reservoir system were valued at greater than $630
million in 1978. Extended summer pool levels could increase these numbers substantially.
The increased flows predicted by the WSM would not only increase recreational uses of the reservoir, but
could also enhance recreational benefits in the tailwaters below the dams. Increased flows could increase
boating, fishing, canoeing, and/or rafting opportunities. At present, special weekend recreational releases are
made at several dams, such as Ocoee and Chatuge Dam, to enhance downstream floating activities. Under the
climate change scenario these types of releases would be more commonplace.
GISS Inverse Scenario. Assuming the reduced lake levels and streamflows produced by the GISS Inverse
scenario are similar to conditions evaluated by the 1986 drought plan (Clark et al., 1986), adverse impacts on
recreation in the Tennessee Valley can be expected. The reduced assimilative capacity of streams and lakes may
result in higher concentrations of sewage bacteria, causing increased health hazards in high recreational use areas.
Higher water temperatures will increase the survival time of sewage bacteria resulting in further potential for
contamination. Increased color levels, particularly below industrial discharges, and algal growths may produce
aesthetically objectionable conditions and discourage swimming and recreational boating.
As TVA reservoir levels decline, exposing mudflats and extensive shoreline, the attractiveness of the
reservoirs will also decline. Recreational users of the reservoirs may also be discouraged by increased turbidity
and odor levels. The incidence of navigation hazards on tributary and some mainstem reservoirs is likely to
increase if reservoir levels are significantly below normal summer full-pool levels.
Fish and aquatic life will also be adversely impacted by the lower lake levels and associated negative impacts
on water quality. As aquatic habitat is reduced, fish stocks are likely to become depressed, particularly in
tailwaters and tributary reservoirs. Reduced fish stocks, along with loss of aesthetic appeal of the reservoirs, will
probably result in large losses in recreational fishing.
Commercial recreation operators and other commercial services that depend on recreation visitors to TVA
reservoirs could be severely impacted by decreased recreational opportunities. Facilities such as boat launches
will be impaired by lower lake levels. The $630 million (1978 dollars) investment in recreational development
and equipment in the Tennessee Valley could be threatened.
Water Supplies
GISS Scenario. In general, the higher runoff predicted by the GISS scenario would benefit water supplies by
increasing reservoir levels and flow magnitudes below dams, and recharging groundwater supplies.
9-35
-------
Miller
GISS Inverse Scenario. Based on the results of drought studies conducted in the TVA area (Clark et al., 1986),
it can be extrapolated that the reduced streamflows and reservoir levels associated with the GISS Inverse scenario
will have negative impacts on the quality and quantity of water supplies in the Tennessee Valley. Water supply
studies in the Tennessee Valley portion of Tennessee, Alabama, and Kentucky identified 49 public and 40
self-supplied commercial and industrial water users that would be adversely effected by prolonged drought
conditions. All of these systems, which would primarily experience water shortage problems, are served by
groundwater sources or small tributary streams.
Water supplies located in TVA reservoirs or below TVA dams would probably have sufficient water to meet
demands, but could experience operational difficulties and customer dissatisfaction due to degraded water quality
conditions. Taste and odor problems and complaints about higher water temperatures are likely to increase.
Increased dissolved solids and algae growths may clog treatment plant filters. Several water intakes in TVA
reservoirs are located immediately above waste discharges. During extended low flow conditions there are
increased opportunities for wastes to backflow upstream to water supply intakes.
9-36
-------
Miller
CHAPTERS
SUMMARY AND CONCLUSIONS
The objective of this project was to identify the sensitivity of the TVA reservoir system to global climate
change. Potential changes in reservoir operations, navigation, flood control, hydropower production, water
availability, water quality, and recreation were evaluated for the Tennessee River Basin. Implications of these
changes, as well as potential adaptive strategies, were also outlined for the reservoir system and related TVA
programs.
METHODOLOGY
The Weekly Scheduling Model (WSM) was used to assess climate change impacts on the TVA reservoir
system. WSM is a planning model used to simulate long-term, week-to-week variations in water level, discharge,
and power production for the 42 reservoirs operated within the system. Based on a linear programming
algorithm, the model selects a weekly reservoir schedule for each reservoir by sequentially satisfying a series of
operating objectives in a prescribed order of priority. The primary objective is to minimise deviation from
historical normal operating levels, subject to reservoir release and level constraints imposed to meet navigation,
flood control, water supply, power generation, water quality, and recreational requirements.
A primary input into the WSM is historical local flow into each reservoir in the TVA system. Local inflows
result from unregulated overland and tributary flows into each project It was assumed that changes in local
flows were directly proportional to changes in runoff. Consequently, in the application of the WSM, monthly
ratios (IxCOj/CONTROL) for surface runoff provided by EPA for two climate scenarios were used directly to
adjust historical local inflows to each project for the 30-year study period (1951-1980). It was also assumed that
monthly runoff ratios applied equally to each week within the month; existing guide curves defined the target
operating conditions for each reservoir; and current operating policy defined the set of model constraints.
The two weather scenarios provided by EPA are based on the NASA Goddard Institute for Space Studies
(GISS) general circulation model The GISS model grid divides the TVA basin into two regions that coincide
with major hydrologic units: the portion of the Tennessee Valley east of Chattanooga (80°W grid); and the
portion west of Chattanooga (90°W) grid. For each grid, GISS provides monthly average values for key climate
variables for a control run, a 2xCO, run, and the ratio (fcCOj/CONTROL). The 2xCO2 levels, and therefore
ratios, represent the endpoint effecb of a doubling in CO2 atmospheric concentrations.
The first scenario is based on GISS model data directly provided by EPA. This GISS scenario predicts a
warmer and wetter climate for the Tennessee Valley. In the eastern basin average annual runoff increases 31%.
On a seasonal basis, runoff increases in the winter through summer months and decreases in the late fall. The
largest predicted increase occurs in March (73%) and the largest decrease in November (-28%). The net effect
in the eastern basin, which contains all of the large tributary storage reservoirs, is to exaggerate peak flows
during the traditional flood season and further decrease flows during a dry period of year. West of Chattanooga,
runoff ratios are more variable. On a seasonal basis, runoff increases in the winter and summer and decreases
slightly in the spring and fall. The integrated effect of eastern and western basin runoff on the Tennessee River
is to increase local inflows throughout the year except during November and December.
Runoff data initially provided by EPA for the second weather scenario were generated by Princeton
University's Geophysical Fluid Dynamics Laboratory (GFDL) general circulation model. The GFDL runoff data,
however, appeared unreasonable for the Tennessee Valley. To meet EPA project objectives and time constraints,
EPA and the GISS model developers recommended utilizing the inverse of the GISS runoff values. This
approach, referred to as GISS Inverse, uses runoff values in the range of model sensitivity and illustrates the
potential impacts of a warmer and significantly drier climate. The two scenarios together, therefore, represent
9-37
-------
Miller
the full spectrum of impacts from the extremes of a wetter versus drier climate. Both of these extremes have
been predicted for the southeast by other GCM and hydrologic models (Mains, 1988).
In the eastern basin, average annual runoff decreases 31% under the GISS Inverse scenario. On a seasonal
basis, runoff decreases from January through October, with the largest decrease (75%) in March, while runoff
increases during November (18%) and December (28%). As runoff is reduced substantially during the
traditional flood season and only modestly increased during a dry period of the year, the net effect in the eastern
basin is to create tow inflows throughout the year. In the western basin, the runoff ratios are more variable, but
the net effect is also to reduce runoff from current levels.
The methodology applied in this project provides an overview of the response of the TV A reservoir system,
as an integrated system, to changes in historical local inflow. There are, however, several important limitations
to the methodology, including the following: existing guide curves and current operating policy were used in the
WSM; runoff ratios provided by the general circulation model were applied directly; current water use and
power demand patterns were assumed in the analysis; the historical distribution of runoff was assumed to remain
the same; and the spatial and temporal distributions of runoff and precipitation events, particularly under extreme
drought or flood conditions, were not considered. In a more detailed analysis of climate impacts model
constraints should be modified to reflect changes in operating policies and priorities; temperature effects on
power and water use patterns should be evaluated; temperature, precipitation, humidity, and other climate
variables should be used with hydrologic models to generate more site-specific runoff data; groundwater issues
should be addressed; and the distribution of rainfall and runoff events should be considered.
IMPACTS OF CLIMATE CHANGE ON THE TVA SYSTEM
flfSS Scenario. The impacts of the GISS scenario on the TVA system are summarized in Table 4 and Figure
13. The generally increased runoff predicted by GISS results in higher reservoir elevations throughout the year
at all major projects in the Tennessee Valley. However, the impacts of this scenario are most apparent in the
large storage reservoirs located in the eastern portion of the Tennessee Valley. Although runoff is slightly
decreased in November and December, the additional storage earlier in the year enables the reservoirs to meet
or exceed normal operating levels during the fall. During UK traditional flood season (December-March),
normal maximum levels are exceeded several times during wet years at the tributary projects. During the
summer, full-pool levels are generally maintained for an extended period of time.
The major adverse impact of the GISS scenario is to increase jpill, particularly during wet years, at tributary
and mainstem projects during the traditional flood season. This would result in increased flood potential
throughout the Valley. At Chattanooga, Tennessee, a major city on the Tennessee River with the greatest urban
damage potential in the Valley, the river stage of four out of five major historical floods is increased. During
two of these flood events, stages would exceed the Tennessee River banks and could cause damages in the range
of $100 million to $1 billion (1984 dollars).
Extended operations at or above normal maximum levels would also necessitate the revaluation of dam
safety at TVA projects. Additionally, potential changes in the Probable Maximum Precipitation and Probable
Maximum Flood resulting from a major climate shift could have significant safety implications at dams and
nuclear power plants.
Primary benefits of the GISS scenario include increased power production; enhanced recreational
opportunities; and improved water availability for water supplies and minimum flow requirements. Average
annual system generation would increase by 3.1 million MWh, or 16%, at an annual value of $54.7 million (1988
dollars). The increased opportunity to maintain extended fuO summer pools would greatly enhance recreational
opportunities, particularly on the tributary storage reservoirs. The general increase in water availability and
storage would improve water supplies, the assimilative capacity of lakes and streams, and the abfltty to exceed
minimum flow requirements for water quality. The overall effect on water quality, however, would be
site-specific depending on the relative influence of increased inflows versus the potential for increased nonpoint
source pollution (NFS).
9-38
-------
Table 4. Potential Impacts of the GISS Scenario on the TVA Reservoir System
Activity
Impact
Imp!1catIons
Potential Alternatives
and/or Modifications
Cost
Information
Reservoir - Higher levels throughout
Operations the year
- Elevations above normal
maximum during traditional
flood season
- Extended full summer pool
- Higher fall reservoir levels
Increased storage and
and wafer availability
Increased likelihood of
spi11 and downstream
flooding
Enhanced recreational
oppor funities
Improved abiIity to moat
minimum flow requirement's
Roevaluate reservoir guides
and operating philosophy to
better utilize available water
and provide additional storage
capacity during fIood season
Navigation - Minimum navigation levels
exceeded throughout the
year
- Increased periods of high
flow
Minimal overall effect
Increased incidences of
flooding dock facilities
and suspension of
navigation
$4.8 billion
private investment
(1933-1984)
Flood Control -
Increased spi11 at
tributary and malnstem
reservoirs during
flood season
Increased flood peaks at
Chattanooga (March 1973
peak flood stage increased
by 26.7 ft); overflow
banks during major floods
Increased flooding - Adjust reservoir guide
potential throughout Valley curves to provide more
Increased flood damage seasonal flood storage
at Chattanooga - Modify existing dams
- Construct additional dams
- Construct flood control
structures (lovoes) in
flood-prone areas
(Chattanooga)
$100 million -
$1 billion (1984
dollars) potential
damage at
Chattanooga for
major event
$5-27 million
(1987 dollars)
to modify existing
dams
$1.5 mi 11 ion/mile
(1988 dollars) to
construct Ievoos
Dam Safety - Dams operate at or above
normal maximum levels for
extended periods of time
Roevaluate dam safety
Potential impacts on
Probable Maximum Precip
(PMP) and Probable
Maximum Flood (PMF)
Adjust reservoir guide curves
Increase reservoir storage
capac i ty
Increase spillway capacity
Construct additional spillways
Strengthen or raise existing dams
$5-27 million
(1987 dellars)
to mod i f y
existing dams
-------
Table 4. Potential Impacts of the 6ISS Scenario on the TVA Reservoir System
(continued)
Activity
Impact
Implications
Potential Alternatives
and/or Modifications
Cost
Information
Power - 3.1 mil I Ion MWh or 16%
Operations increase In average annual
system generation
- Full capacity operation
required at many plants
for extended periods
- Increased spiII
Improved hydropower
generation
Increase use to satisfy
base load, but loss of
flexibility for peaking
operations
Decreased operating heads
and power output during
periods of spiII
Justify added capacity
Construct more pumped-storage
facilities for peaking purposes
Reevaluate future power needs
and relative mix of hydro,
fossil and nuclear
Reevaluate maintenance
schedules
- $54 mi 11 ion annual
value of increased
energy (1988
doIlars)
Water Quality -
Variable impact on
00, temperature, and
general water quality
Improved assimilative
capacity In tailwaters
Exceed minimum flow
requirements
Increased nonpoint source
pollution (NPS)
Relative influence of
increased Inflows vs
increased NPS wiII
determine general
water quality
Site specific impacts
Recreation - Extended full summer pools
- Increased tailwater flows
Enhanced reservoir recreation
(boating, fishing, swimming)
Enhanced tal(water recreation
(canoeing, rafting, etc)
Potential for increased
tourism & economic
development
Reevaluate reservoir
guide curves to provide
full summer pools April
through September
Encourage economic
development of recreation
opportunities
- $630 million
(1978 dollars)
current value of
recreational
development &
equipment
Water Supplies - Increased water
availability &
storage
- Adequate water supplies
-------
Figure 13
Summary of Probabilistic Pool Level Forecasts for Morris and Watauga Reservoirs
IOJ5
M>50
W1S
IOIO
roia
IOIO
1009
«oo
•9»
»M
•IS
NORMAL MAXIMUM
V..-'"
MAXIMUM ENVELOPE
*•••••••
NORMAL MAXIMUM
AMJJAIOND
i r M
NORRIS
BASE CASE
LEGEND'
•- GISS
ENG LAB 1988
I»TO
MM
I»9O
l»40
U <•*<>
MEDIAN
s ~s.
__ _ __ __ U'l^i
HBMAtTHNiiniB ^
111
1990
I94O
1910
I9ZO
1910
ISOO
1890
i« so-
ia 70
M«0-
l«90-
1140-
NORMAL MAXIMUM
NORMAL MINIMUM
•••*"""
MINIMUM ENVELOPE
t ' r ' M ' A ' M
J J A I 0
WATAUGA
N 0
- GISS INVERSE
-------
Miller
A climate change similar to GISS would likely result in a revaluation of the current reservoir operating
philosophy and guide curves to better utilize the increased water availability and to provide additional storage
capacity during the flood season. Added turbine capacity could be justified, while more pumped-storage or
special fossil plants could be needed for peaking purposes. It is likely that major dams would need to be
modified to meet safety standards and/or create additional flood storage capacity. Finally, programs to control
or mitigate the adverse impacts of increased nonpoint source pollution would need to be expanded.
GISS Inverse Scenario. The results of the GISS Inverse scenario are summarized in Table 5 and Figure 13.
Under the GISS Inverse scenario, runoff is decreased throughout most of the year resulting in a general decrease
in storage and water availability. At the tributary storage reservoirs, lake levels are lowered throughout the year.
Median reservoir levels are up to 9 meters (30 feet) lower than the base case and, during dry years, minimum
reservoir levels are often below normal minimum pool levels.
Based on current operating policy, the WSM is constrained to fill the mainstem reservoir to normal
operating levels and to meet minimum downstream flow requirements for water quality. Consequently, these
specific objectives are met under the GISS Inverse scenario; however, this is accomplished at the expense of
severely reducing reservoir levels in the tributary storage reservoirs. The adverse impacts of these reduced levels
include reduced power generation; impaired water quality; degraded recreational opportunities; and decreased
water availability for water supplies.
Under the GISS Inverse scenario, average annual system power generation is decreased by 4.7 million MWh,
or 24%. As hydropower, the least cost source of energy, must be replaced by more expensive sources of energy
such as fossil or nuclear power, this loss in energy has an $87.2 million (1988 dollars) replacement value. Not
only is system generation reduced, but owing to the reduced heads and flows, system hydropower capacity would
also be reduced. Additionally, owing to decreased flows and elevated water temperatures, operations at several
fossil and nuclear plants could be restricted because of thermal and/or safety limits.
The reduced reservoir levels and tailwater flows would likely result in significant deterioration of the water
quality in TVA lakes and streams. The reduced DO levels, increased temperatures, and reduced assimilative
capacity would adversely affect aquatic biota, fish, and wildlife, as well as recreational uses. The reduced water
availability is likely to result in water shortages for groundwater water supply systems, while operational
difficulties and customer dissatisfaction are likely to be experienced by supply systems withdrawing reservoir
water.
The benefits of the GISS Inverse scenario are related to reduced flood potential and associated damages.
Current normal maximum levels would rarely be exceeded and flood stages at Chattanooga would be reduced.
Increased development of current floodplain areas may be possible. The probability of dam failure or major
damage would be substantially reduced
To deal .with the significant reduction in water availability predicted by GISS Inverse, current operating
philosophy and reservoir guides would need to be reevaluated to increase storage during the wet periods and
conserve water during the extended dry periods of the year. Drought-related issues would increase in significance
as compared to management of the reservoir system for flood control It would be difficult to satisfy project
purposes at many reservoirs, and a reordering of TVA priorities would probably be required Alternative sources
of energy would need to replace lost hydropower potential, while it is likely that industrial and municipal
treatment plants would need to adhere to more stringent waste standards. Adverse economic impacts on the
Tennessee Valley could be significant resulting from increased power costs, decreased recreational revenues, and
increased industrial restrictions.
9-42
-------
Miller
Table 5
Potential Impacts of the GISS Inverse Scenario on the TVA Reservoir System
Activity
Impact
Implication!
Potential Alternatives
•nd/or Modi flections
Cost
Information
Reservoir - Lower levels throughout
Operations the year
- Maximum levels never
exceed existing normal
maximum pool levels,
often resemble current
norm*I I eve Is
- Median levels up to 30
feet lover than base
case levels
- Minimum levels often at
or below existing normal
minimum pool levels
Decreased storage and
water availability
Decreased likelihood of
uncontrolled spill and
downstream flooding
More difficult to
satisfy project purposes
• Create adverse pubIic
reaction
Reevaluate reservoir
guides, normal maximum
and minimum levels, and
operating philosophy to
better utiIize the
decreased runoff
Navigation
Minimum pool levels
maintained on mainstem
reservoirs
Minimum navigation levels
maintained
$4.8 bi11 ion private
investment (1953-1984)
Flood Control
Decreased maximum
reservoir levels
Decreased flood stages at
Chattanooga
Decreased flooding
throughout valley
Minimal flood damage
at Chattanooga
Reevaluate reservoir
guide curves to reallocate
flood storage (increase
normal minimum levels)
Increased development of
current floodplain areas
Dam Safety
Decreased likelihood of
operations at or above
maximum pool levels
Reduced probabiIity
of dam failure or
major damage
• Potential impacts on PMP
and PMF
Evaluate current dam
safety program and need to
retrofit older dams
Power . 4.7 million MWh or 241
Operations decrease in average
annual system generation
- Reduced flows and
Increased water
temperatures
Reduced hydropower
generation and capacity
Loss of hydropower
for reliable
peaking operations
Potential problems
with meeting thermal
and/or safety Iimi ts
at some foss iI and
nuclear plants
Increased reliance on
nuclear and fossiI
plants
Construct alternative
means to satisfy peaking
demands (pumped-storaga
and fossil plants with
peaking capability)
> Increased use and/or
construction of closed-
cycle cooling systems
S87 mi 11 ion annual
loss in benefits
(1988 $)
$75 million (1985 $)
to retrofit John Sevier
Fossil Plant with
cooling towers (costs
at other plants would
vary)
9-43
-------
Miller
Table 5
Potential Impacts of the GISS Inverse Scenario on the TVA Reservoir System
(continued)
Act i v i ty
Impact
Implications
Potent!*! Alternatives
•nd/or Modifications
Cost
Infornution
Water Quality
Reduced reservoir
levels
Reduced tall water flows
Reduced nonpoint source
pellution
Deteriorated reservoir
water quality (reduced
00 and increased
temperatures)
Oegraoad water quality
in taiI waters
Reduced assimilative
capacity in lakes and
streams
Adverse impacts to fish
and wildlife
Increased health hazards
from rtduced assimilative
capacity
More stringent waste
treatment standards
and/or cut-backs in
industrialized
production
> Improved situation
assessment and
monitoring techniques
to Identify critical
problem
Recreation
Reduced reservoir levels
Reduced taiIwater flows
Degraded recreational
opportunities in lakes
and stream
Adverse economic effects
on coomercial operations
and recreational
developments
Reevaluate reservoir
guide curves to minimize
adverse effects
Target specific reservoirs
for recreational use
- $630 million (I97B $)
current value of
recreational
development and
equi praent
Mater Supplies - Reduced reservoir levels - Water shortages
experienced by
groundwater or
small tributary
stream systems
- Operational
difficulties and
customer
dissatisfaction for
systems on large
reservoirs
- Increased waste
treatment costs
Encourage conservation
through education and
economic incentives
Improve water conservation/
leak detection in
distribution systems
Recycle industrial process
end cool ing water
Total metering of public
water supplias
9-44
-------
Miller
CONCLUSIONS
Results of this study indicate that a major shift in the global climate could have significant impacts on the
TVA reservoir system. The warmer and wetter climate suggested by the GISS scenario could benefit the
Tennessee Valley through increased power production, enhanced recreational opportunities, and increased water
availability. However, flood control, as well as dam and nuclear plant safety, could become major issues.
Conversely, the significantly reduced water availability resulting from the GISS Inverse warmer and drier climate
could threaten the ability of many TVA projects to fulfill their current operating goals and purposes. Under both
climate scenarios, significant changes in the reservoir guide curves, as well as potential structural changes and/or
additions to the system, would be required to respond to an altered climate.
The results of this study are based on runoff data generated by a general circulation model and represent
potential climate changes from a doubling of atmospheric CO2 (possible by the year 2030). The study is intended
to assess the sensitivity of the TVA reservoir system to extreme climate changes and to identify the implications
of these changes. The project objective was not to predict the future climate in the TVA region. Scientists concur
that significant global climate changes -- increased global mean surface temperatures and precipitation - are
highly probable (NAS, 1987). It is, however, currently impossible to accurately predict the magnitude of the
change or climate impacts on a regional basis. Furthermore, given the normal cycles of wet and dry periods in
the historical record, it is difficult to distinguish long-term changes from natural weather variations.
The Tennessee Valley is presently experiencing the worst drought on record. Precipitation and runoff have
been significantly below average for more than four years. The question arises, "How many years of abnormal
weather constitute a long-term change in climate?" Given the noted sensitivity of the TVA reservoir system to
climate change, the recent weather patterns in the Tennessee Valley, and the general scientific consensus that
atmospheric changes are highly probable, this issue should be investigated in more detail
Future climate change studies should include an in-depth hydrologic study of the TVA region based on
temperature, precipitation, and other climate variables generated by general circulation models. Based on
potential changes in hydrology and temperature, detailed assessments should be conducted to address TVA
reservoir and power system impacts; Tennessee Valley social and economic development; and institutional and
legal ramifications. Finally, potential adaptation strategies need to be identified and methodologies need to be
developed for incorporating climate change issues and uncertainties into TVA's long-range planning and
decision-making process.
9-45
-------
Miller
LIST OF ABBREVIATIONS, SYMBOLS, AND DEFINITIONS
ABBREVIATIONS
ALCOA
cfs
cms
DO
EPA
ft
GFDL
GISS
GWh
m
MWh
NPDES
PMF
PMP
TVA
WSM
Aluminum Company of America
cubic feet per second
cubic meters per second
dissolved oxygen
Environmental Protection Agency
feet
Geophysical Fluid Dynamics Laboratory
Goddard Institute of Space Studies
gigawatt hours
meters
megawatt hours
National Pollutant Discharge Elimination System
Probable Maximum Flood
Probable Maximum Precipitation
Tennessee Valley Authority
Weekly Scheduling Model
SYMBOLS
carbon dioxide
nitrogen oxide
-- chlorofluorocarbons
DEFINITIONS
lower envelope—minimum weekly reservoir elevations for given period of record
median projection-median of WSM projected operations based on given period of record
normal operations-normal reservoir operations based on 15 years of operation experience (1972-86); target
elevation for each year of simulation
normal maximum level—elevation above which a reservoir would not be
operated except during periods of high flow; extended operations above normal maximum levels often result in
spillage
normal minimum level-elevation below which a reservoir would not be operated except under extreme drought
conditions
upper envelope-maximum weekly reservoir elevations for given period of record
Definition of curves used in Probabilistic Pool Level Forecasts
9-46
-------
Miller
REFERENCES
Clark, L. R. et al., "TVA's Response to Drought Related Water Quality Impacts in the Tennessee River System,"
Water Quality Drought Task Force. TVA, Knoxville, Tennessee, 1986.
Gilbert, K.C., and R.M. Shane, "TVA Hydro Scheduling Model: Theoretical Aspects," ASCE Journal of the
Water Resources Planning and Management Division. VoL 108, No. WR1, pp 21-36, March 1982.
Crotch, S.L., "Regional Intercomparisons of General Circulation Model Predictions and Historical Climate Data,"
Department of Energy Report No. NBB-0084, Dist. Category UC-11, Washington, D.C., 1988.
Hains, D.K., "Impacts of Runoff in the Upper Chattahoochee River," (EPA National Assessment of Climate
Change Effects), C.F. Hains, Hydrologist, Inc., 1988.
MacCracken, M.C., The Reality of the Greenhouse Effect," Lawrence
Livermore National Laboratory, Livermore, California, Preprint UCRL-95580,1986.
Miller, B A., and W.G. Brock, "Sensitivity of the Tennessee River Valley to Global Climate Change," Report No.
WR28-1-680-101, Engineering Laboratory, Morris, Tennessee, 1988.
National Academy of Sciences, "Current Issues in Atmospheric Change: Summary and Conclusions of a
Workshop, October 30-31,1986," National Academy Press, 1987.
Rind, D., "The Doubled CO, Climate and the Sensitivity of the Modeled Hydrologic Cycle," Journal of
Geophysical Research (in print), 1988.
Shane, R.M., "Weekly Scheduling Model for the TVA Reservoir System," Report No. WR28-1-500-126, TVA
Engineering Laboratory, Norris, Tennessee, 1984.
Shane, R.M., and K.C. Gilbert, "A Weekly Time Step Scheduling Model for the TVA Operated Reservoir
System," Water Resources Publication, Littleton, Colorado, 1981.
Waffel, Heinz-Dieter, "Operation Objectives in TVA's Weekly Scheduling Model," Report No. WR28-2-590-119,
TVA Engineering Laboratory, Norris, Tennessee, 1985.
------- |