vvEPA
United States
Environmental Protection
Agency
Policy, Planning
And Evaluation
(2122)
EPA230-R-93-009
December 1993
The Colorado River Basin
And Climatic Change
The Sensitivity Of
Streamflow And Water Supply
To Variations In Temperature
And Precipitation
WYOMING
COLORADO
RacyctetfRecyclatote
Printed on paper Oiat contains
at least 50% reofded ffbor
-------
-------
The Colorado River Basin
and Climatic Change
The Sensitivity of Streamflow and Water
Supply to Variations in Temperature
and Precipitation
Linda L. Nash
Peter H. Gleick
Pacific Institute for Studies in
Development, Environment, and Security
Oakland, California
A Report Prepared for
The United States Environmental Protection Agency
Office of Policy, Planning, and Evaluation
Climate Change Division
EPA230-R-93-009
December 1993
-------
-------
ACKNOWLEDGEMENTS
We would like to acknowledge the participation of the U.S. Bureau of Reclamation in this
study. David Westnedge and Gerald Williams of the National Weather Service River
Forecasting Service in Salt Lake City provided us with model runs, advice, and comments.
Roy Jenne and Dennis Joseph of the National Center for Atmospheric Research provided
GCM data. In addition, we would also like to thank several reviewers for their comments
and suggestions, including Joel Smith of the U.S. EPA, the Metropolitan Water District of
Southern California, and the U.S. Bureau of Reclamation. Any errors or omissions, of
course, remain the responsibility of the authors. This work does not necessarily reflect
the opinions of the National Weather Service, the U.S. Bureau of Reclamation, or the U.S.
EPA. This work was supported by the U.S. Environmental Protection Agency, grant
#CR816045-01.
-------
-------
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ii
EXECUTIVE SUMMARY vii
INTRODUCTION 1
Background 1
Scenarios of Climate Change for Impact Assessment 5
METHODS OF ANALYSIS I: HYDROLOGIC MODELING 9
Background 9
Description of the Model 11
Model Calibration 14
Application of Climate Scenarios to the NWSRFS Model 17
RESULTS OF HYDROLOGIC MODELING 21
Annual Runoff 21
Seasonal Runoff 28
Transient Scenario 32
GCM Runoff Scenarios 33
Discussion of Hydrologic Modeling Results 34
METHODS OF ANALYSIS II: WATER-SUPPLY MODELING 43
Description of the Model 43
Modeling Assumptions 46
RESULTS OF WATER-SUPPLY MODELING 51
Runoff 51
Reservoir Storage 54
Depletions and Deliveries 62
Hydroelectricity Production 67
Uncontrolled Spills 67
Salinity 68
Time-Shifted Scenario 71
Summary and Discussion of Water-Supply Modeling Results 73
STUDY CONCLUSIONS 80
Future Work 88
REFERENCES 80
APPENDIX A: CALIBRATION RESULTS FROM THE NWSRFS MODEL A-1
APPENDIX B: THE LAW OF THE RIVER AND CRSS OPERATING PROCEDURES B-1
APPENDIX C: ADDITIONAL RESULTS FROM THE CRSS MODEL C-1
-------
LIST OF FIGURES
Figure 1:
Figure 2:
Figure 3:
Figure 4:
Figure 5:
Figure 6:
Figure 7:
Figure 8:
Figure 9:
Figure 10:
Figure 11:
Figure 12:
Figure 13:
Figure 14:
Figure 15:
Figure 16:
Figure 17:
Figure 18:
Figure 19:
Figure 20:
Schematic of study 2
Map of the Upper Colorado River Basin 19
Change in runoff as a function of change in precipitation for the White
River model 26
Distribution of annual runoff for the White River model 27
Distribution of annual runoff for the Animas River model 27
Point estimates of annual flow for the White River, with approximate 90%
confidence regions 29
Effect of temperature increases on the average hydrograph 30
Distribution of January runoff for the Animas River model 31
Distribution of June runoff for the Animas River model 31
Mean annual runoff, mean spring runoff, and mean fall runoff for the
White River at Meeker 32
Map of the Colorado River Basin showing the location of selected
CRSS stations and major reservoirs 52
Annual runoff at Green River in the base case and the ±20% runoff
scenarios
56
Annual runoff at Lees Ferry in the base case and the ± 10% runoff
scenarios 56
Cumulative frequency of annual runoff at Lees Ferry for all scenarios 58
Upper basin storage on August 1 plotted as a function of year 62
Lower basin storage on August 1 plotted as a function of year. 62
Minimum, mean, and maximum annual depletions in the upper basin,
lower basin, and Mexico 66
Minimum, mean, and maximum hydropower generation In the upper
and lower basins , 69
Frequency and approximate annual volume of uncontrolled spills
which occur in the upper basin during a simulation run of 78 years 69
Salinity as a function of year at Davis and Imperial Dams 70
IV
-------
Figure 21:
Figure 22:
Impact of the time-shifted scenario on storage in the upper basin 72
Relationship between storage in Lake Mead and annual deliveries to
CAP
76
-------
LIST OF TABLES
Table 1:
Table 2:
Table 3:
Table 4:
Table 5:
Table 6:
Table 7:
Table 8:
Table 9:
Table 10:
Table 11:
Table 12:
Table 13:
Table 14:
Table 15:
Table 16:
Table 17:
Table 18:
Table 19:
Table 20:
Table 21:
Table 22:
Hypothetical climate scenarios used in regional hydrologic studies-
Changes in temperature and precipitation in the Colorado River Basin
predicted by general circulation models
Summary of calibration results for the NWSRFS model
Climate change scenarios used in the NWSRFS model
Annual inflow into Lake Powell (Two-elevation model) for all scenarios
Annual streamflow of the White River for all scenarios
Annual streamflow of the East River for all scenarios
Annual streamflow of the Animas River for all scenarios
Changes in runoff generated by GCMs and the NWSRFS hydrologic model....
Impacts of climatic change on runoff in semi-arid basins
Scheduled demands used by the Bureau of Reclamation in the CRSS model.
Description of input sequences used in the CRSS model
Annual runoff of the Green River at Green River, Wyoming
Annual runoff of the Colorado River at Lees Ferry
Annual runoff of the Colorado River above Imperial Dam
Major reservoirs in the Colorado River Basin
Storage in Flaming Gorge reservoir on August 1 for various scenarios
Storage in Lake Powell on August 1 for various scenarios
Storage in Lake Mead on August 1 for various scenarios
Percent frequency with which scheduled deliveries to MWD, CAP, and Mexico
are met •
Annual runoff at various points for the base case and the time-shifted scenario-
Sensitivity of water-supply variables to changes in natural flow in the Colorado
River Basin
10
15
20
.22
23
24
25
,33
.36
.45
50
54
.55
55
.57
.59
.59
. 60
. 66
. 72
. 76
VI
-------
THE SENSITIVITY OF STREAMFLOW AND WATER SUPPLY
IN THE COLORADO RIVER BASIN TO CLIMATIC CHANGES
EXECUTIVE SUMMARY
Linda L. Nash
Peter H. Gleick
June 1993
Pacific Institute for Studies in
Development, Environment, and Security
1204 Preservation Park Way
Oakland, California 94612 1
(510) 251-1600
Growing international concern about the greenhouse effect has led to increased interest in the
regional implications of changes in temperature and precipitation patterns for a wide range of societal and
natural systems, including agriculture, sea level, biodiversity, and water resources. The accumulation of
greenhouse gases in the atmosphere due to human activities are likely to have significant, though still poorly
understood, impacts on water quality and availability. One method developed over the last several years
for determining how regional water resources might be affected by climatic change is to develop scenarios
of changes in temperature and precipitation and to use hydrologic simulation models to study the impacts
of these scenarios on runoff and water supply. In this paper we present the results of a multi-year study of
the sensitivity of the hydrology and water resources systems in the Colorado River Basin to plausible climatic
changes.
The Colorado River is one of the most important river systems in the western United States. It is
the principal source of water in a semi-arid basin that covers approximately 243,000 square miles, parts of
seven states, and reaches into Mexico (Figure ES-1). The study was conducted in two parts: the first part
evaluated the effects of changes in temperature and precipitation on runoff using a conceptual hydrologic
model developed and operated by the National Weather Service. Among the impacts studied were changes
in streamflow into Lake Powell and on three important tributaries of the Upper Colorado River: the White
1 Final Report. This work was supported by the U.S. Environmental Protection Agency, Grant #
CR816045-01.
vii
-------
SELECTED CRSS STREAMFLOW STATIONS
1. Green River near Green River, Wyoming
2. Colorado River near Cisco, Utah
3. San Juan River near Bluff, Utah
4. Colorado River at Lee Ferry, Arizona
5. Colorado River below Davis Dam, Arizona/Nevada
6. Colorado River below Parker Dam, Arizona/California
7. Colorado River above Imperial Dam, Arizona
WY
NEVADA
boundary between upper
and lower basins
NEW MEXICO
Figure ES-1: Map of the Colorado River basin (excluding Mexico) showing the location of
selected CRSS stations and major reservoirs. (Source: redrawn from USDOI, 1987.)
viii
-------
River, the East River, and the Animas River. The second phase of the project then evaluated how these
hydrologic changes might affect water supply, salinity, and hydroelectricity production throughout the entire
Colorado River Basin using the Colorado River Simulation System (CRSS), a reservoir-simulation model
developed and operated by the U.S. Bureau of Reclamation.
Two types of climate scenarios were used for these sensitivity studies: hypothetical temperature and
precipitation scenarios, and scenarios generated by general circulation models (GCMs) of the climate. The
hypothetical scenarios included increases in average temperatures of 2° to 4°C and increases and decreases
in precipitation of 10 and 20 percent. The regional changes in temperature and precipitation from three
GCMs were also evaluated. The scenarios chosen reflected both the best understanding and the uncertainty
about the expected magnitude of regional climatic changes when the study began.
Our results suggest that certain aspects of the hydrology and water-supply system of the Colorado
River Basin are extremely sensitive to climatic changes that could occur over the next several decades. Not
only are significant changes in runoff possible, but the ability of the existing water supply system to mitigate
the worst effects is limited. For example, the major reservoirs of the Colorado Basin lessen the impacts of
reduced flows, but only for a short period of time. Under conditions of long-term flow reductions and current
operating rules, these reservoirs are drawn almost completely dry, hydroelectricity production drops
dramatically, and salinity in the Colorado River increases to the point where it fails to meet legal standards
almost all of the time. The results strongly suggest that the current approaches to water management in
the basin will have to be modified to balance the many competing demands and priorities under conditions
of altered climate, and that current water allocations may well be threatened.
Changes in Colorado River Basin Hydrology
The principal impacts of changes in temperature and precipitation on runoff in the Colorado Basin
are summarized below.
• Increases in temperature of 2°C alone, with no change in precipitation, cause mean annual runoff
in the Colorado River Basin to decline by 4 to 12 percent.
• A temperature increase of 4°C causes mean annual runoff to decrease by 9 to 21 percent.
• Increases or decreases in annual precipitation of 10 to 20 percent result in corresponding changes
in mean annual runoff of approximately 10 to 20 percent.
• A temperature increase of 4°C would require an increase in precipitation of 15 to 20 percent merely
to maintain annual runoff at historical levels.
• Temperature increases shift the seasonally of runoff in the Colorado Basin, causing a distinct
increase in winter runoff and a decrease in spring runoff. This is the result of a decrease In winter
snowfall and snowpack, an increase in winter rain, and a faster and earlier spring snowmelt These
temperature-driven changes could increase the potential for winter and spring flooding in some
regions.
• GCM temperature and precipitation scenarios modeled as part of this study suggest that
ix
-------
precipitation increases would be offset by increased evapotranspiration, with the net effect being a
reduction in runoff ranging from 8 percent to 20 percent.
• Of the three GCMs used to develop climate scenarios in this study, the GFDL model results in the
most extreme decreases in runoff for all the sub-basins studied (-10 to -24 percent) because it
predicts a relatively large regional temperature increase and no change in precipitation. The least
extreme effects are generated by using either the UKMO or the GISS grid points, which incorporate
respective increases in precipitation of 30 and 20 percent and lead to increases in runoff of 0 to 10
percent.
• High-elevation basins appear to be more sensitive to changes in temperature and precipitation than
low-elevation basins. Of the three sub-basins studied, the East River near Almont, Colorado is the
most sensitive to changes in temperature and precipitation because of its higher elevation.
• In general, runoff in the Upper Colorado River basin is slightly more sensitive to a 10 percent change
in precipitation than to a 2°C change in temperature. Thus, while increased temperatures will cause
significant decreases in runoff, the overall response of the basin will ultimately depend upon the
direction and magnitude of changes in precipitation.
In summary, the hydrologic modeling results suggest that large changes in streamflow may occur
in the Colorado River basin as a result of plausible climatic changes. GCM scenarios indicate that runoff
in the basin is likely to decrease. The impacts of these potential changes in streamflow would be felt
throughout the basin as changes in water deliveries, reservoir storage, and hydroelectricity production.
Changes in the Colorado River Water Supply System
The changes in runoff determined in the first part of the project were then used to evaluate impacts
on several water-supply parameters, including salinity, reservoir levels, deliveries to users, and
hydroelectricity generation. Some quite severe effects were seen, assuming no changes in the operating
rules of the basin. For example, a 20 percent reduction in natural runoff would cause mean annual
reductions in storage of 60 to 70 percent, reductions in power generation of 60 percent, and an increase
in salinity of 15 to 20 percent. In contrast, a moderate increase in temperature (2°C) and a large increase
in precipitation (20 percent) would result in roughly a 20 percent increase in mean annual runoff, a 30 to 60
percent increase in storage, a 40 percent increase in power production, and a 13-15 percent decrease in
salinity. The principal impacts on water supply identified with the CRSS model include the following:
• Changes in mean annual actual streamflow along the River range from -31 percent to +32 percent
for the scenarios studied. Decreases in runoff are relatively smaller in magnitude in the Lower Basin
because they are cushioned by additional reservoir releases. For example, a decrease in natural
flow of 20 percent causes a 31 percent decrease in mean annual streamflow at the Upper Basin
station of Green River, but only an 11 percent decrease at Imperial Dam near the Mexican border.
• Decreases in natural runoff cause severe changes in minimum runoff. For example, the -10%
scenario causes mean annual runoff in the Upper Basin to decline by about 15%, but minimum flows
at Lees Ferry drop 86%.
• In the base case (i.e., under current hydrology), annual releases from Lake Powell never drop below
the objective minimum of 8.23 million acre-feet per year (maf/yr); however a runoff decrease of 10%
causes releases from Lake Powell to fall below 8.23 maf/yr in several years.
-------
• Reservoir storage and power generation are the variables most sensitive to changes in runoff.
Changes in long-term mean storage in Lake Mead on August 1 are on the order of -70 percent, or -
8,700 thousand acre-feet (taf) for the -20 percent runoff scenario, to +60 percent, or +7,400 taf for
the +20 percent runoff scenario.
• Lake Powell falls below minimum power pool 20 percent of the time when runoff drops by 5 percent;
this frequency rises to nearly 60 percent when runoff decreases by 20 percent. The -20 percent
(runoff) scenario causes Lake Mead to go completely dry roughly 25 percent of the time.
• The sensitivity of storage to changes in runoff suggests how carefully the system is currently
managed and that consequently there may be little room for error in forecasting seasonal flows
should the hydrologic regime undergo any significant changes.
• High salinity levels, already a critical concern for the Lower Basin, would be severely exacerbated
by any decreases in runoff.
• While the runoff scenarios modeled in this study may appear extreme, streamflow in the region may
have a much higher variability than is commonly recognized. For instance, the most extreme
scenario modeled in this study, a 20 percent decrease in mean annual runoff, may not even be
incompatible with the current (non-greenhouse) hydrologic regime. Tree-ring reconstructions
suggest that over the last 500 years, the lowest 80-year mean at Lee Ferry is less than 11 maf, which
corresponds to a 27 percent decrease in natural flow, compared to the 1906-83 instrumental record.
The impact of changes in natural runoff on several water-supply parameters is summarized in Table
ES-1 and in the sections below.
Table ES-1: Sensitivity of water-supply variables to changes in natural flow in the Colorado
River Basin [1].
Change in
Natural
Flow
-20
-10
-5
5
10
20
Change in
Actual
Flow [2]
(10-30)
(7-15)
(4-7)
5-7
11-16
30
Change
in
Storage [3]
(61)
(30)
(14)
14
28
38
Change in
Power
Generation [4]
(57)
(31)
(15)
11
21
39
Change
in
Depletions [5]
(11)
(6)
(3)
3
5
8
Change
in
Salinity [6]
15-20
6-7
3
(3)
(6-7)
(13-15)
Notes: [1] Average change compared to the base case over a 78-year simulation run. Numbers in parentheses represent
DECREASES.
[2] Changes in flow represent the range of changes at five points: Green River, Cisco, Bluff, Lee Ferry, and Imperial
Dam.
[3] Mean storage throughout the basin on August 1.
[4] Mean annual power generation throughout the basin.
J5] Depletions are summarized over the entire basin, although depletions are defined differently in the upper and low
basins. See Hundley (1975) for details.
[6] Changes in salinity represent the range of changes at three points: Davis, Parker, and Imperial Dams.
xi
-------
Water Deliveries to Users
Delivery of water to different users are affected dramatically by different scenarios, depending on
streamflow changes and the application of the law of the river. For example, in the base case, deliveries
to the Central Arizona Project would ordinarily fall to their minimum level 20 percent of the time and
scheduled deliveries are met or exceeded 40 percent of the time. If runoff drops 5 percent, our results
suggest that full scheduled deliveries will be met in only 25 percent of the years and that in half of the years,
only minimum levels are delivered.
Although the delivery data suggest that Mexico is affected only in extreme cases, the quality of
Mexican water decreases significantly. In fact, all Lower Basin users would suffer a significant decline in
water quality (see Salinity).
Hydroelectricltv
Under current operating rules, hydroelectricity production, like reservoir storage, is extremely
sensitive to changes in runoff. If flows in the Upper Basin were to decrease by 10 percent, average annual
storage decreases by 30 percent and power production drops by 26 percent. A decrease in flows of 20
percent would reduce storage by 63 percent and power production by nearly 50 percent. An increase in
flows of 10 percent would increase storage by 28 percent and power generation by 21 percent.
In the Lower Basin, a 10 percent decrease in runoff reduces storage by 30 percent and power
production by 36 percent. A drop in runoff of 20 percent reduces Lower Basin storage by 50 percent and
power production by 65 percent.
Salinity
The most critical concern for the Lower Basin is salinity and salinity is the only water-quality
parameter studied. Even in the base-case scenario salinity criteria are consistently exceeded at all points
in the Lower Basin for most years. Decreases in runoff of only 5 percent cause salinity criteria to be
exceeded in virtually all years. Even if average flows were to increase by 20 percent, salinity criteria are
exceeded continuously for long periods.
Under almost no climate-change circumstances can existing water-quality criteria be met given
projected demands and operating constraints. Our results suggest that at least a 20 percent increase in
natural runoff would be necessary to bring the salinity levels in the Lower Basin into compliance with existing
criteria, In the absence of other activities to reduce salinity in the river.
Seasonal Timing of Runoff
A variety of recent hydrologic analyses have suggested that changes in the seasonality of runoff may
xli
-------
be a major impact of climate change in hydrologic basins dependent on snowfall and snowmelt. One
scenario was run to study the effects of shifts in the seasonality of runoff. The results suggest that an
increase in temperature of only 2°C would shift peak runoff one month earlier, to May, in the Upper Basin.
Under current operating conditions, such a shift in timing reduces the overall efficiency with which the
system is operated, reducing effective storage and deliveries, and increasing the average annual salinity.
We recommend that changes in operations to account for changes in the timing of runoff should be
evaluated.
Summary and Discussion
The results of this assessment suggest that violations of the Colorado River Compact are likely to
occur under all scenarios of decreased runoff, assuming that no changes in the operating parameters of the
system occur. For instance, storage strategies and targets work extremely well in the base case scenarios
but are substantially less effective under alternative scenarios. Thus, violations of the Compact would
potentially occur even if runoff dropped only 5 percent. The sensitivity of storage to changes in runoff reflect
how carefully the current system is operated and how little room there is for forecast error if water supply
is to be maximized without resulting in damaging flood-control releases or uncontrolled spills.
As might be expected, the reservoir simulation results presented here suggest that many of the
procedures and inputs used in the Bureau of Reclamation model are closely tuned to the historic hydrologic
record. While it is likely that many of the severe impacts noted here could be avoided under different
operating conditions and rules, we were constrained in the current study from evaluating any alternative
operating criteria.
The problem of planning water management in the face of a high degree of climate and hydrological
uncertainty cannot be easily resolved; nonetheless, it may be possible to increase flexibility in water
management. This flexibility will need to be reflected in technical and operational decisions, as well as in
the legal and economic institutions that govern water use in the basin.
The problem of planning is compounded by the fact that we cannot say with certainty whether runoff
in the basin will increase or decrease. Most people with an interest in the basin have focused on the
prospect of long-term decreases in runoff and the shortages that would result, which is a logical reflection
of the region's preoccupation with drought. The fact that average temperatures in the region will almost
certainly increase suggests that, if we assume no knowledge about changes In precipitation, we would
expect runoff to decrease as a result of increases in evaporation and vegetative water use. This may be
reason enough to plan for supply shortages; but increased water storage must be traded off against the
need for flood-control space. The greatest risk of climatic change Is the potential for streamflow variability
to increase substantially, increasing the frequency of both sustained drought events and high-flow events.
xiii
-------
Beyond the scope of this study were several important issues that policymakers and water-supply
managers will have to consider. First, the environmental and ecological impacts of changes in water supply
have not been addressed here. In general ecosystems are more sensitive to seasonal, monthly, daily, and
even hourly changes in streamflow and water quality than to long-term changes. Unlike water supply, the
impacts on the environment cannot be adequately assessed using aggregated time periods or large-scale
models. Undoubtedly, however, given the predicted rate of climatic change and the potential magnitude of
runoff changes examined here, serious ecological problems would occur.
This study has also not taken projected future economic developments nor some future demands
into account. Currently the issue of reserved water rights and Native American claims have obscured future
demand scenarios in the basin. Because of the large amounts of water involved, these unresolved claims
could have dramatic impacts on water allocation throughout the region and thus add to the uncertainty that
the basin faces.
Finally, while this study has suggested what the impacts of climate change could be on water
supply, it has not addressed the impacts of climate change on water demand. In fact, demands will change
both in time and space. Obviously, agricultural water demand will vary as crops and production patterns
are altered in response to climatic changes. Ecosystem water requirements will also vary, both in response
to increased temperature and as a result of ecological and environmental changes. Urban and industrial
usage will change as a result of both changes in climate and changes in population. It is quite possible
that changes in demand over the next 50 to 100 years will equal or exceed changes in supply. In all
likelihood, the greatest possibilities for adapting to climatic change lie in the area of demand management,
particularly in the agricultural and urban sectors, and the potential for conservation and water transfers needs
to be assessed from both a quantitative and an institutional perspective. If we are to plan adaptation
strategies, future research must address the integrated impacts of climatic change on demand and supply
across sectors.
Given the prospect of future climatic changes, it is imperative that we consider how we can increase
the resiliency of our existing water-management systems and minimize the social and environmental impacts
of changes in water availability. We need to identify those responses that will provide us with the greatest
flexibility in the coming decades and to develop management schemes that recognize both the variability
and the dynamic nature of our climate.
xiv
-------
THE SENSITIVITY OF STREAMFLOW AND WATER SUPPLY
IN THE COLORADO RIVER BASIN TO CLIMATIC CHANGES
INTRODUCTION
Background
Human activities are substantially increasing the atmospheric concentration of greenhouse
gases. These gases, in turn, are expected to increase the overall average temperature of the Earth's surface
and alter precipitation patterns worldwide. The magnitude of increases in global average temperature is
predicted to range from LffCto 4.5CC over the next century (IPCC, 1990). The regional impacts of these
changes will vary and cannot yet be predicted with much confidence; however, existing global climate
models indicate that temperature increases in central North America will exceed the increase in the global
mean, and will be accompanied on average by reduced summer precipitation and soil moisture (IPCC, 1990;
Manabe and Wetherald, 1980; Rind, et al., 1990).
Such global climatic changes may have substantial impacts on water resources. Higher
temperatures, new precipitation patterns, rising sea level, and changes in storm frequency and intensity will
alter water availability, quality, and demand. Despite recent advances in modeling the atmosphere, large
uncertainties remain about the details of regional hydrological changes. Until large-scale climate models
improve both their spatial resolution and their hydrologic parameterizations, information on the effects of
global climatic changes on hydrologic sub-basins can best be produced using detailed, basin-specific
hydrologic models. In this study, we analyze the potential impacts of climatic change on the hydrology and
water resources of the Colorado River Basin. First, we use a regional hydrologic model to study the effect
of changes in temperature and precipitation on runoff in several sub-basins of the Upper Colorado.
Subsequently, we analyze the impact of changes in runoff on water supply, water deliveries, and water
-------
quality using the Colorado River Simulation System (CRSS), a reservoir-simulation model developed and
operated by the U.S. Bureau of Reclamation (Figure 1).
The Colorado River is one of the most important river systems in the United States. Although
not a large river, even in comparison to other rivers in the US, the Colorado flows through some of the most
arid regions of the country and is the primary source of water for a region with extensive agriculture, large
cities, and a diverse ecosystem. The Colorado River Basin covers approximately 243,000 square miles, parts
of seven states, and reaches into Mexico. Annual unimpaired runoff of the Colorado River at Lee Ferry has
ranged from 5.6 (million acre-feet) maf to 24.0 maf since regular streamflow recording was initiated in the
early part of this century.2 Over the same period, mean annual unimpaired runoff has been about 15.1 maf;
however, tree-ring analyses dating back to 1512 have suggested that the long-term mean may be closer to
13.5 maf (Stockton and Jacoby, 1976).
The apportionment of the Colorado River has been more complete than that of the waters of
any other river through many hard-fought lawsuits, negotiations, political battles, and an international treaty.
The Colorado River Compact of 1922 divided the basin into two sections. The upper basin, in which most
of the region's runoff originates, includes those parts of Wyoming, Colorado, Utah, New Mexico, and Arizona
that drain into the Colorado River above Lee Ferry, Arizona.3 The more arid lower basin encompasses most
of Arizona, southeastern Nevada, southeastern Utah, western New Mexico and portions of southern
California. The lower basin states were guaranteed that the upper basin states would deliver an annual
average of 7.5 maf of water (over a ten year period) to Lee Ferry, a point on the river approximately on the
Arizona-Utah border. The upper basin states received a right to use an equivalent amount of water (if it was
2
For convenience to US water managers, water volumes are presented here in acre-feet, the standard
unit of measurement in the western United States. One acre-foot is equivalent to 1,233 cubic meters. A flow
of one cubic meter per second (cms) is equal to 70.02 acre-feet per day.
g
Lee Ferry, Arizona, also known as the "compact point" is the point at which the Colorado River passes
from the upper to the lower basins as established by the Colorado River Compact of 1922. It is located
approximately 16 miles downstream of Lake Powell and one mile downstream of the Paria River. It should
not be confused with Lees Ferry, which is a point further upstream on the river, near Glen Canyon Dam.
-------
Climate-change
scenarios
Hydrologic
model
Water-supply
model
NWSRFS
Two-elevation
White River
East River
Aniraas River
Changes in
runoff
CRSS
Reservoir
storage
Hydroelectric! ty
Deliveries
Uncontrolled
spills
Salinity
Figure 1: Schematic of study showing the relationship among various models.
available). The parties contemplated each basin eventually using equal quantities of water (7.5 mat), plus
up to another one million acre-feet for the lower basin. Subsequently, the 1944 Treaty signed by Mexico
and the United States guaranteed an annual flow into Mexico of not less than 1.5 maf, except in times of
severe shortage. Under the Compact, the upper basin in not actually required to deliver a fixed quantity of
water at Lee Ferry in any particular year, though current operating criteria adopted by the Bureau of
Reclamation provide for releases of 8.23 maf from Lake Powell annually.lf the Mexican Treaty obligation is
assumed to be shared equally by both basins (although this remains a disputed point), then the required
3
-------
delivery from the upper to the lower basin is 82.5 maf in every 10-year period, except in those periods when
Mexican Treaty obligations are reduced.
The water apportioned between the basins has also been rather precisely divided among the
states within each basin by the Boulder Canyon Project Act (1928), the Upper Colorado River Basin
Compact (1949), and several court decisions handed down in Arizona v. California. In addition, water
delivered to California is divided among users by the Seven Party Agreement. (These agreements and
allocations are also discussed in Appendix B.)
Water allocation continues, however, to be a contentious issue in the basin. Future demands
for Colorado River water are predicted to outstrip supplies. The population of the region is more than 19
million; and, despite the fact that the area is approaching the limits of its water supply, population and
economic activity have continued to expand. Although severe shortages have not yet been felt in the basin,
there is growing concern that the pressures of increased demand and the potential for periodic supply
shortages will create problems in the future.
Droughts in the Colorado River Basin have generally been considered as isolated, temporary
events that can be overcome through storage and short-term conservation strategies. The validity of this
assumption is challenged by paleoclimatic data which indicate that the region has experienced much more
severe and sustained droughts in previous centuries than in our own (Stockton, et al., 1991). Now the
prospect of anthropogenically induced climatic change offers the unsettling prospect that the region may
face both permanent and more extreme changes in its climate than previously considered. Enhanced
greenhouse warming will almost certainly cause increases in the region's average temperature, and could
cause either increases or decreases in average annual precipitation (IPCC, 1990; Mitchell and Qingcan,
1991). As a result, the basin could experience changes in the likelihood and severity of prolonged droughts
or extreme floods. In any case, the storage and supply facilities and institutions that have evolved in the
4
-------
basin are predicated on streamflow data gathered within the last 80 years. In fact, the Colorado River
Compact of 1922 was based upon less than 20 years of data, and as a result allocated more water than is
likely to be available in an average year. The ability of this system to function under altered climatic
conditions has not been seriously considered.
Scenarios of Climate Change for Impact Assessment
To assess the implications of global warming for water resources, regional-scale details of future
changes are needed for temperature, precipitation, evaporation, wind speed, and other hydroclimatological
variables. Because our ability to predict these details is limited, climate-impact analysis must rely upon the
development of scenarios. Such scenarios can be either hypothetical or derived from General Circulation
Models (GCMs), paleoclimatic reconstructions, or recent historical climate analogues (WMO, 1987; USEPA,
1989).
Hypothetical scenarios are simple combinations of changes in variables (usually temperature,
precipitation, and potential evapotranspiration) that are consistent with global changes expected as a result
of the greenhouse warming. While such scenarios are limited by the fact that they may not be internally
consistent, they provide a very useful means of testing hydrologic vulnerabilities. If constructed
systematically, hypothetical scenarios can be used to develop sensitivity studies that delineate the relative
importance of changes in temperature and precipitation to changes in runoff. Subsequently, as estimates
of future temperature and precipitation improve, the impacts on water resources can be easily estimated.
Table 1 lists the range of hypothetical scenarios used in a variety of studies. The values chosen
typically reflect best estimates of changes in important climatic variables, although extreme values are
occasionally chosen to explore where a system might fail to perform as expected or designed. Thus, the
practice of using hypothetical temperature increases of 1, 2, 3, or 4? Celsius reflects the consensus that
greenhouse warming will produce temperature rises in this range, given an equivalent doubling of
-------
atmospheric carbon dioxide (IPCC, 1990).4 Given the greater uncertainty about both the magnitude and
the direction of regional precipitation changes, both increases and decreases in precipitation are frequently
modeled.
Much of the effort to understand climate has focused on the development of computer models
that simulate many of the intricate and intertwined phenomena that make up the climate. The most complex
of these models, GCMs, are detailed, time-dependent, three-dimensional, numerical simulations that include
atmospheric motions, heat exchanges, and important land-ocean-ice interactions (IPCC, 1990). Climate
models, however, are still simple when compared with the complexities of the real climate system. For
Instance, current GCMs handle cloud formation and ocean currents quite primitively, although these are
important climatic processes (Ramanathan, 1981). Oceans are generally modeled as simple slabs, and only
some of the GCMs take heat transport by currents and circulation into account. In addition, the models use
a smoothed topographic profile that precludes an accurate representation of regional orographic effects.
Despite these limitations, general circulation models currently provide the best information available on the
response of the atmosphere to increasing concentrations of greenhouse gases, as well as valuable insights
Into the potential impacts across broad regions (IPCC, 1990).
In theory, GCM estimates of changes in hydrologic variables, such as runoff, could be used
directly to estimate changes in water resources (see, for example, USEPA, 1984). In practice, however,
GCM-generated hydrologic data suffer from two major limitations. First, the spatial resolution of GCMs is
too coarse to provide hydrologic information on a scale typically of interest to hydrologists.5 Present
Regional temperature changes, however, may be higher or lower.
GCM resolution Is unlikely to dramatically Improve for many years because of the extreme cost of high-
speed computer time-a factor of two increase in resolution requires approximately a factor of eight increase
In computer time [Somerville, 1987]. With a typical model resolution of 4.5 degrees latitude by 7.5 degrees
longitude and nine vertical layers in the atmosphere, computing one year of weather at 30-minute intervals
takes 10 hours of computer time on a Cray XMP computer-one of the fastest in the world.
-------
Table 1: Hypothetical climate scenarios used in regional hydrologic studies.
Study [1]
Stockton and
Boggess[1979]
Nemec and
Schaake[1982]
Revelle and
Waggoner [1983]
Flaschka et al.
[1987]
Gleick
[1986, 1987a,b]
Fitzgerald and
Walsh [1987]
Schaake [1990]
This study
Temperature PET [2] Precipitation
±2°C ±10%
+1°C, +3°C ±10,25%
+2°C, +4°C -10%
±2°C ±10,25%
+2°C,+4°C ±0,10,20%
±0,5,15% ±0,10,20%
+2°C +10% +10%
+2°C, +4°C ±0,10,20%
Notes: [1 ] All studies use different methods and assumptions. Please refer to individual sources for details.
[2] Potential evapotransplration.
resolutions are usually between 4 to 7.5 degrees latitude by 5 to 10 degrees longitude - grid areas of
hundreds of thousands of square kilometers. Yet, hydrologists are often interested in climatic events that
7
-------
occur on the scale of tens or hundreds of square kilometers - a scale several orders of magnitude finer than
current GCM resolution.6
Second, hydrologic parameterizations in GCMs are very simple and often do not provide the
detailed information necessary for water-resource planning (WMO, 1987). For example, the GCM soil-
moisture budget is typically computed by the so-called "bucket method", in which the field capacity of the
soil is assumed to be uniform everywhere (Manabe, 1969a,b). Runoff occurs when the soil moisture
exceeds this capacity, and the rate of evaporation is determined as a simple function of the soil moisture
and the potential evaporation rate (Manabe and Wetherald, 1985). Efforts are being made to improve GCM
hydrology (Dickinson, 1984; IPCC, 1990), including improvements in vegetation parameterizations and the
behavior of soils. Until such improvements occur, however, other methods must be used to evaluate
hydrologic impacts.
Temperature predictions are considered to be the most reliable GCM output relative to
precipitation, and other climatic variables (IPCC, 1990). More generally, GCM predictions of changes in
temperature, precipitation, and other climatological variables are considered much more reliable than
predictions of runoff or soil moisture (IPCC, 1990; WMO, 1987). Consequently, several investigators have
emphasized using temperature and precipitation estimates for a doubled-COj environment as inputs to more
detailed regional models (e.g., USEPA, 1990; Lettenmaier and Gan, 1990; Bultot, et al., 1988; Gleick,
1987a,b).
Under the guidance of the U.S. Environmental Protection Agency, a set of climate-change
scenarios was developed for use in evaluating the impacts of the greenhouse effect on water availability in
mis is not meant to imply that increasing GCM resolution alone will resolve the bulk of the problems
with GCMs, which suffer from several other limitations. Nonetheless, the resolution problem is critical for
hydrologic analysis, particularly in regions where hydrologic processes are dominated by orography.
-------
the Colorado River. These include several combinations of hypothetical changes in temperature and
precipitation and scenarios derived from three state-of-the-art GCMs were used to develop inputs for use
in modeling the Colorado River Basin. The use of more than one GCM has two advantages: first, reliance
on one GCM may give a false impression of accuracy; and second, the use of more than one GCM
highlights model differences and similarities and permits a broader analysis of outcomes and sensitivities.
The data on temperature and precipitation changes due to a doubling of carbon dioxide come from the
Goddard Institute for Space Studies (GISS) model, the Geophysical Fluid Dynamics Laboratory (GFDL) Q-
flux model, and the United Kingdom Meteorological Office (UKMO) model (Hansen, et al., 1983, 1988;
Manabe and Stouffer, 1980; Manabe and Wetherald, 1987; Wilson and Mitchell, 1987). Each of these models
is an equilibrium run, i.e. carbon dioxide is doubled all at once in the models and a new equilibrium climate
is established.
In addition, data from a GISS transient run were incorporated into our analysis. In the transient
run, the GISS model was started with same amount of greenhouse gases in the atmosphere as measured
in 1958, and the concentration of gases was gradually increased. We developed a climate scenario that
reflected the decadal average of temperature and precipitation changes that occur in the years 2030 to 2039.
These data were presumed to provide an indication of how much change will occur over the next 40 years,
given the assumptions in the GISS model concerning the future rate of greenhouse-gas emissions (Hansen,
et al., 1988). The changes in temperature and precipitation predicted for the Colorado River Basin by each
of these GCM runs is given in Table 2.
METHODS OF ANALYSIS I: HYDROLOGIC MODELING
Background
Once scenarios of climate change are developed, hydrologic models can be used to estimate
impacts on water resources. If accurate estimates of future water availability are to be calculated, regional
-------
hydrologic evaluations need to incorporate the complexities of snowfall and snowmelt, topography, soil
characteristics, natural and artificial storage, and monthly or seasonal variations.
Table 2: Changes in temperature and precipitation in the Colorado River
Basin predicted by general circulation models (GCMs). [1]
Equilibrium [2]
GISS1
GISS2
GFDL
UKMO1
UKMO2
Transient [3]
GISS1
GISS2
A Temperature (°C)
+4.8
+4.9
+4.7
+6.8
+6.9
+3.2
+2.5
A Precipitation (%)
+20
+10
0
+30
+10
+10
+20
Notes: [1] For the GISSand UKMO GCMs, the upper Colorado River basin was interesected by two
different grid points. The more northern grid point is labeled "1"; the more southern is labeled "2".
[2] Equilibrium GCM runs, in which greenhouse gas concentrations have stabilized at roughly twice
current levels.
[3] The GISS transient run, in which greenhouse gases are increasing gradually. The numbers
presented here represent the avearge over the decade 2030 to 2039.
The use of hydrologic models, rather than GCMs, for assessing the regional impacts of climatic
changes has several attractive characteristics. First, diverse modeling techniques exist. This permits
flexibility in Identifying and choosing the most appropriate approach for evaluating any specific region.
Second, hydrologic models can be chosen to fit the characteristics of the available data. Third, hydrologic
models are regional in scale and are far easier to manipulate and modify than are GCMs. Fourth, regional
models can be used to evaluate the sensitivity of specific watersheds to both hypothetical changes in climate
and to changes predicted by large-scale GCMs or climatic analogues. And finally, methods that incorporate
both detailed regional characteristics and output from GCMs can take advantage of the continuing
Improvements In the resolution, regional geography, and hydrology of global climate models (Glelck, 1989).
10
-------
Past studies of the hydrologic impacts of climatic change can be divided into two categories:
(1) stochastic methods that rely primarily on statistical techniques for evaluating the hydrologic
characteristics of a region or for extending the existing hydrologic record (such as Schwarz [1977], Revelle
and Waggoner [1983], and Stockton and Boggess [1979]); and (2) deterministic or conceptual methods that
use physically based, mathematical descriptions of hydrologic phenomena (Nemec and Schaake, 1982;
Gleick, 1986,1987a,b; Mather and Feddema, 1986; Cohen, 1986; Flaschka, et_iL, 1987; Bultot et al.. 1988;
Lettenmaier and Gan, 1990). To date, climate-impact studies of the Colorado River Basin have been limited
to stochastic methods (Revelle and Waggoner, 1983; Stockton and Boggess, 1979). These studies
necessarily assume, however, that the relationships among temperature, precipitation, and streamflow will
remain unchanged under future climatic conditions. In contrast, this study used a conceptual hydrologic
model to study the sensitivity of the basin to greenhouse warming. A recent attempt to use a deterministic
model to study climatic impacts on a small sub-basin of the Colorado River is presented in Schaake (1990).
In this project we expand upon that work by incorporating additional climate scenarios and modeling
additional sub-basins. By modeling actual hydrologic processes (e.g. percolation, soil-moisture storage,
snowmelt, etc.), deterministic techniques incorporate an additional level of complexity. So long as these
hydrologic processes do not change significantly under a CC^-altered climate, deterministic models should
be more robust than derived statistical relationships between meteorologic variables and streamflow. In fact,
however, all attempts to study the impacts of climatic change using hydrologic models are limited by their
dependency on historic data, which may not be applicable to future conditions.
Description of the Model
The large size of the Colorado River Basin complicates the development of a physically based
hydrologic model; indeed, no completely satisfactory basin model exists. As a result, we modeled several
sub-basins In the Upper Colorado River Basin, using a conceptual hydrologic model developed and operated
by the National Weather Service River Forecasting Service (NWSRFS) in Salt Lake City, Utah. These models
11
-------
simulate the hydrologic processes important for river forecasting, including soil moisture, snowfall, and
snowmelt.
The NWSRFS is comprised of two linked models: a soil-moisture accounting model that
calculates gains and losses of water in the soil through various processes (e.g. evaporation, transpiration,
infiltration); and a snow accumulation and ablation model that calculates the accumulation of snow and the
contribution of snowmelt to soil moisture and runoff. The soil-moisture accounting model is a modified
version of the Sacramento Model described in Burnash et al. (1973). The Sacramento Model is widely used
and generally accepted as one of the most reliable in varied climatic conditions on several continents,
including both arid and humid regions (Nemec and Schaake, 1982). The model distributes soil moisture into
an upper and lower zone. Movement between zones is controlled by a physically based percolation
equation whose parameters are controlled by the free water in the upper zone and the soil-moisture
deficiency in the lower zone. The snowmelt model uses air temperature as the sole index to energy
exchange at the snow-air interface and is described in detail in Anderson (1976). The inputs to the model
are areal temperature and precipitation data; the output is streamflow (runoff) on a 6-hourly basis.
The NWSRFS models the Upper Colorado River Basin as a series of approximately 50 small
sub-basins that are linked together. For forecasting purposes, all of the sub-basins are modeled
simultaneously. For calibration purposes, however, each of these sub-basins is modeled separately. In this
study, we modeled three sub-basins which were selected based upon: (1) the existence of an adequate
historical streamflow record (at least 35 years), (2) a relatively high volume of streamflow, (3) streamflow
records classified as "good" or better by the U.S. Geological Survey (USGS), and (4) the presence of only
limited withdrawals and upstream regulation. These three basins are the White River at Meeker, the East
River at Almont, and the Animas River at Durango. In addition, the NWS has developed a composite model
(referred to here as the 'Two-elevation model") that divides the entire Upper Colorado River Basin into two
elevation zones and uses a limited number of data stations to predict inflow into Lake Powell. Given the
12
-------
constraints of this study, it was not possible to study all of the Upper Colorado River sub-basins. Yet by
studying smaller, detailed sub-basin models, only limited information on the entire basin could be generated.
Thus, we used the composite Two-elevation model to obtain an overview of the impacts on the entire upper
basin. The Two-elevation model has an additional advantage of being highly correlated with streamflow
nodes in the CRSS water-supply model.
All three sub-basins are high-elevation, snowmelt-driven watersheds, with no significant rainfall
showing up in the average hydrograph. Streamflow measurements for the White River model come from
the USGS gauging station located 2.5 miles east of Meeker at an elevation of 6300 feet. The drainage area
of the White River covers approximately 770 square miles. The period of record for the White River dates
from 1909. Mean annual discharge computed over the period 1949-1983 is about 435 thousand acre-feet
(taf). East River measurements are made at the Almont station, which has an elevation of 8006 feet. The
period of record dates from October, 1934. Streamflow measurements for the Animas River are made at
an elevation of 6502 feet at the station of Durango. Records date from 1912. In all cases, monthly and
annual streamflow records are classified as "good".7 Streamflow into Lake Powell, which is used to calibrate
the Two-elevation model, is calculated by the Bureau of Reclamation based on reservoir outflow, changes
in reservoir storages, and evaporative losses, and is checked against the combined flows of three upstream
USGS gauging stations (the Colorado River at Cisco, the Green River at Green River [Utah], and the San
Juan River at Bluff).
As stated above, the NWSRFS is a forecasting model that was developed for the short-term
forecasting of streamflows. For this purpose snow-pack conditions, daily observations of temperature and
USGS classifications are defined as follows:
Excellent - 95% of daily discharges are within 5% of their true value.
Good - discharges are within 10% of their true value.
Fair - discharges are within 15% of their true value.
Poor - discharges do not fall within 15% of their true value.
13
-------
precipitation, and present streamflow information are used as inputs into the model, and future streamflow
forecasts are produced as outputs. For the purposes of this study, however, the model was run in
calibration (or simulation) rather than forecasting mode. To calibrate the model, past records of temperature
and precipitation are correlated with concurrent streamflows. Independent parameters (associated with soil
moisture accounting, snow ablation and snowmelt, and streamflow routing) are subsequently modified to
improve the fit of simulated to observed data. By altering historic temperature and precipitation data, future
climate scenarios and their resulting streamflows can also be simulated. The comparison of simulations
obtained from actual historic data and altered data provides information about the changes in streamflow
that might be expected from changes in climatic conditions.
Model Calibration
The standard test for credibility of a given hydrologic simulation model is verification with data
not used in model calibration. In many cases, however, the data set is too limited to permit this type of
testing. Because the model used in this study is a forecasting model used daily for operational purposes,
ail model testing and calibration has been done by the National Weather Service in Salt Lake City. The
entire 35-year record (1949 to 1983, inclusive) was used to calibrate each of the sub-basins.
The World Meteorological Organization model intercomparison program suggests that various
criteria be used to test general purpose streamflow models, including differences between simulated and
observed flows, mean flow, characteristics of maximum and minimum flows, and seasonal characteristics
(WMO, 1985; WMO, 1987). A set of these criteria are evaluated for the NWSRFS model calibration runs.
The results are summarized in Table 3 and are presented in detail in Appendix A. In all cases, the model
has a fairly good fit. The analysis of daily streamflow data for all models shows a consistent bias of
overpredicting low flows and underpredicting high flows. In general, however, the model appears to perform
satisfactorily so long as predicted flows are within about 20% to 25% of the mean.
14
-------
Because the entire streamflow record was used to calibrate the NWSRFS model, independent
tests of validation could not be undertaken as part of this study. The success of the NWSRFS as a
forecasting tool, however, suggests that the model has the capability to simulate the effects of changes in
temperature and precipitation. In addition, a critical assumption of this research is that the NWSRFS model
is able to simulate adequately runoff under climatic conditions different from those for which the model has
been calibrated. While there are reasons for believing that the model possesses this capability for moderate
climatic changes, the use of this model (or any model) may be problematic if simulated conditions differ
significantly from calibrated conditions. For example, changes may occur in plant-transpiration rates and
in vegetative cover under a CO, -altered climate. These types of changes and their effect on streamflow are
Table 3: Summary of calibration results for the NWSRFS model
Model
Two-elevation
White River
East River
Animas River
r2
Daily
Rows
0.94
0.92
0.93
0.93
1*
Monthly
Flows
0.92
0.88
0.91
0.93
Mean Annual
Flow
% Bias
-1.25
-0.36
1.05
1.14
Monthly Volume
RMS Error
(taf)
3.62
7.98
6.98
10.9
15
-------
not accounted for in a model calibrated on current climatic conditions. Nevertheless, the short time-step
used (6-hourly) implies that the model's storage behavior beyond calibrated conditions is only for limited
periods and should have a relatively minimal impact on average annual runoff outputs. And, to the extent
that studies focus on relatively short-term and "moderate" changes in climate, significant changes in model
parameters would not be expected (Nemec and Schaake, 1982).
Another assumption of the model is that water withdrawals are not significantly affecting runoff.
Because withdrawals are not accounted for in the model directly/they are implicit in the values chosen for
other parameters. Thus, as withdrawals increase in a particular basin, the calibration of ail parameters for
that basin change to account for the decrease in streamflow. So long as withdrawals remain a relatively
small factor in basin streamflow, this omission should not be critical to the model's ability to simulate
different climate scenarios. To minimize this problem, sub-basins were selected in which withdrawals were
known to be relatively minor.8
A further weakness of the Two-elevation model is that model parameters have been averaged
spatially. In general, the strength of the NWSRFS model is its use of physically based parameters to
describe hydrologic processes. Thus, while the exact value of parameter may not be known, a reasonable
range of values can be determined from existing data. This becomes increasingly difficult as the scale of
the model is increased. For example, it is much more problematic to choose infiltration parameters for the
entire Upper Colorado River Basin than for a small (and presumably more homogenous) sub-basin. Thus,
while the Two-elevation model may "fit" the data as well as any sub-basin model, these results should be
treated more skeptically. Nonetheless, because of the time and resources required to study the more than
50 sub-basins, the Two-elevation model was included in this study because it provides the only means of
assessing the potential impacts of climate change on the entire Upper Colorado River Basin.
inability to account for withdrawals explicitly is of greater concern for the Two-elevation model
because substantial withdrawals are occurring.
16
-------
Application of Climate Scenarios to the NWSRFS Model
The hypothetical scenarios used in each of the model runs are shown in Table 4. In the
absence of information on the distribution of annual changes throughout the year, mean annual changes
were applied uniformly to all the historical data. Temperature changes were applied as absolute amounts,
while precipitation changes were interpreted as percent differences:
*T = Tnew-Told (1)
(2)
rold
Potential evapotranspiration (PET) rates were assumed to follow the general relationship to
temperature of 4 percent per degree Celsius as derived by Budyko (1982:119). Wetherald and Manabe
(1975) found that global evaporation increases by 3 percent when temperature increases by 1° C.
Accordingly, for the Two-elevation model, additional sensitivity runs were done using a potential
evapotranspiration rate of 3% per degree Celsius. As expected, the potential evapotranspiration rate is most
important for temperature-dependent scenarios (i.e. increases in temperature with no net change in
precipitation). For a temperature increase of 4?C and no net change in precipitation, the use of a 4% per
degree potential evapotranspiration rate rather than a 3% per degree rate decreases mean monthly runoff
by an additional 3%. For other scenarios, the effect of the potential evapotranspiration rate was much less
important.
Temperature data in the model were altered by changing the mean elevation of the basin
relative to the existing station data using an appropriate lapse rate. For standard calibration runs, the model
normalizes temperature station data to the mean elevation of the basin being modeled. To convert this
station data, the model uses minimum and maximum lapse rates (to convert minimum and maximum
temperature data, respectively) For climate change runs, the elevation of the sub-basin was altered using
an average lapse rate, usually between 0.5 and 0.7° C per 100 meters. It is important to note that model
17
-------
results are very sensitive to the lapse rates used for modifying temperature data. The use of higher (lower)
lapse rates would reduce (increase) the effect of temperature changes on runoff.
The GCM scenarios used in the model runs are also listed in Table 2. In all cases at least two
GCM grid points intersect the Colorado River Basin and, at the same time, include vast areas outside of the
basin. Figure 2 shows the approximate location of grid points and the modeled hydrologic sub-basins. The
grid points represent spatially averaged data and, as such, misrepresent any particular point within the box.
In selecting GCM grid-point data for use in hydrologic modeling, we chose not to modify the data in any
way (i.e. through interpolation) because we found little justification for doing so.
Each of the sub-basins (White, Animas, and East Rivers) fell well within a specific GCM grid
box, although not always the same grid box, depending upon the GCM. The Two-elevation model, on the
other hand, was spread across two different grid boxes in each GCM. In the case of the GISS and GFDL
models, there was little difference in the scenarios generated by the adjacent grid points, and thus only one
point from each model was used. In the case of the UKMO model, however, the adjacent grid points yielded
substantially different scenarios so that data from both points (labeled UKMO 1 and UKMO 2) were applied
to the Two-elevation model.
The available GCM data consist of mean monthly changes in temperature and precipitation
developed from a historical baseline that encompasses years 1951 through 1980. These data were averaged
to obtain mean annual changes in temperature and precipitation and then applied uniformly to the long-term
historical data. As in the case of the hypothetical scenarios described previously, changes in temperature
18
-------
White River at Meeker
•East River at Almont
Lake Powell
(Two-Basin Aggregated Model)
Animas River at Durangn
# Gissa
UPPER COLORADO RIVER BASIN
Figure 2: Map of the Upper Colorado River basin showing the location of modeled sub-basins
and GCM grid points. (Source: redrawn from Upper Colorado Region Comprehensive
Framework Study, Main Report, June 1971.)
19
-------
Table 4: Climate-change scenarios used in the NWSRFS model.
Hypothetical
T+2°C, P-20%
T+2°C, P-10%
T+2°C, P+0
T+2°C, P+10%
T+2°C, P+20%
T+4°C, P-20
T+4°C, P-10%
T+4°C, P+0
T+4°C, P+10%
T+4°C, P+20%
Two-
Elevation
«
X
X
X
—
X
X
X
X
X
White
River
X
X
X
X
X
X
X
X
X
X
East
River
X
X
X
X
X
X
X
X
X
X
Animas
River
X
X
X
X
X
X
X
X
X
X
GCM [1]
GISS1:
GISS 2:
GFDL:
UKMO1:
UKMO 2:
T +4.8°C, P+20%,
T+4.9°C, P+10%
T +4.7°C, P+0
T +6.8°C, P+30%
T+6.9°C, P+10%
—
X
X
X
X
X
—
X
X
X
__
X
X
—
X
— «
X
X
X
Note:
(11 All GCM scenarios represent annual average changes for an equilibrium (2XCO2) run.
20
-------
were applied as absolutes (i.e. +2> C), while changes in precipitation were applied as percentages (i.e. +10%
of precipitation in the base case).9
RESULTS OF HYDROLOGIC MODELING
Annual Runoff
For the three Colorado River sub-basins, the magnitude of changes in mean annual runoff
induced by the hypothetical scenarios ranged from decreases of 33% to increases of 19%. The greatest
decrease in runoff was seen in the East River for a 4" C increase in temperature in conjunction with a 20%
decrease in precipitation. The greatest increase was seen in the White River basin when a 2? C increase was
combined with a 20% increase in precipitation. In all cases, at least a 10% increase in precipitation was
required to offset the effect on annual runoff of a 2fC temperature rise. A 20% increase in precipitation
caused runoff to increase in every case. For the Two-elevation model, mean annual runoff decreased by
12% and 21 % when the respective hypothetical scenarios of T+2° C and T+4? C were appl ied with no change
in precipitation. Tables 5 through 8 show these results. In general, the Two-elevation model was more
sensitive to increases in temperature than the three sub-basin models. While this may be an artifact of the
Two-elevation model itself, it may also be explained by the increased importance of evaporation in the lower
elevation zones that the model encompasses.
For the Animas and East rivers, all GCM scenarios led to decreases in runoff, ranging from -8%
to -20%, which reflects the dominant effect of increased evaporation. For the White River, two out of the
four GCM scenarios showed increases in runoff (of 10% to 12%), while the other two scenarios resulted in
9 Mean monthly changes (rather than mean annual changes) cannot be used in the NWSRFS without
modifications to the model. All historical temperature and precipitation data are stored in data files that are
called upon by the calibration program. The program then normalizes these data for the basin being
modeled using a single coefficient. Mean annual temperature and precipitation data can therefore be easily
modified by altering these coefficients. In order to incorporate monthly changes, however, it would be
necessary to alter the data associated with particular months by different amounts. While this can be done,
it requires access to the actual program files, which were not available for this study.
21
-------
decreases in runoff (of -8% to -10%); this is related to the grid point used. Using the Two-elevation model,
three of the four GCM scenarios resulted in decreases in mean annual runoff ranging of -14% and -24%. The
fourth scenario resulted in an increase of less than 1%.
Table 5: Annual inflow (taf) into Lake Powell (Two-elevation model) for all scenarios.
Scenario
Mean [1]
SD
CV
Minimum
Maximum
Base
10940
2983
0.27
4481
Note: J1] Numbers in parentheses represent percent change from the base case.
17040
T+2° P-10%
T+2° P+0
T+2° P+10%
T+4° P-20%
T+4° P-10%
T+4° P+0
T+4° P+10%
T+4° P+20%
GISS2
GFDL
UKMO1
UKMO2
8386 (-23.3%)
9656 (-11.7%)
11000 (0.6%)
6447 (-41.0%)
7522 (-31.2%)
8668 (-20.7%)
9879 (-9.7%)
11150 (2.0%)
9444 (-13.6%)
8369 (-23.5%)
10950 (0.2%)
8639 (-21.0%)
2418
2727
3046
1970
2260
2554
2854
3162
2804
2514
3240
2693
0.29
0.28
0.28
0.31
0.30
0.30
0.29
0.28
0.30
0.30
0.30
0.31
3357 (-25.1%)
3924 (-12.4%)
4504 (0.5%)
2520 (-43.8%)
2892 (-35.5%)
3373 (-24.0%)
3911 (-12.7o/o)
4443 (-0.9%)
3624 (-19.1%)
3180 (-29.0%)
4107 (-8.3%)
3173 (-29.2%)
12940 (-24.1%)
14330 (-15.5%)
16350 (-4.0%)
11480 (-32.6%)
12480 (-26.8%)
13490 (-20.8%)
14530 (-14.8%)
16180 (-5.1%)
14220 (-16.5%)
13270 (-22.1%)
16070 (-5.7%)
13926 (-18.3%)
22
-------
Table 6: Annual flow (taf) of the White River for all scenarios.
Scenario
Base
T+2°
T+2°
T+2°
T+2°
T+2°
T+4°
T+4°
T+4°
T+4°
T+4°
GISS
Mean [1]
434.9
P-20°/o
P-10%
P+0
P+10%
P+20%
P-20%
P-10%
P+0
P+10%
P+20%
1
GFDL
UKMO1
UKMO2
335.
374.
417.
465.
515.
1
6
0
1
7
320.9
357.6
396.
9
440.4
487,
476.
389.
488.
401.
9
2
7
5
3
(-22.9%)
(-13.9%)
(-4.1%)
(7.0%)
(18.6%)
(-26.2%)
(-17.8%)
(-8.70/0)
(1.3%)
(12.2%)
(9.6%)
(-10.4%)
(12.3%)
(-7.7%)
SD
104.5
70.6
82.9
97.5
114.8
132.9
70.0
80.6
92.9
107.9
126.2
122.9
91.7
128.3
97.4
cv
0.24
0.21
0.22
0.23
0.25
0.26
0.22
0.23
0.23
0.24
0.26
0.26
0.24
0.26
0.24
Minimum
242.
193.
214.
234.
255.
279.
180.
200.
221.
241.
264.
8
6
6
7
0
0
7
6
5
7
0
252.9
214.
250.
211.
1
2
8
(-20.3%)
(-11.6%)
(-3.4%)
(5.0%)
(14.9%)
(-25.6%)
(-17.4%)
(-8.8%)
(-0.5%)
(8.7%)
(4.2%)
(-11.8%)
(3.0%)
(-12.8%)
Maximum
670.5
474.7
541.1
608.7
697.1
788.6
468
532.4
599.7
666.9
756.2
746.2
599.7
790.1
640.4
(-29.2%)
(-19.3%)
(-9.2%)
(4.0%)
(17.6%)
(-30.2%)
(-20.6%)
(-10.6%)
(-0.5%)
(12.8%)
(11.3%)
(-10.6%)
(17.8%)
(-4.5%)
Note: [1] Numbers in parentheses represent percent change from the base case.
23
-------
Table 7: Annual flow (taf) of the East River for all scenarios.
Scenario
Base
T+2° P-20%
T+2° P-10%
T+2° P+0
T+2° P+10%
T+2° P+20%
T+4° P-20%
T+4° P-10%
T+4° P+0
T+4° P+10%
T+4° P+20%
GISS2
GFDL
UKMO2
Mean [1]
230.7
165.8 (-27.6%)
186.9 (-18.7%)
209.4 (-9.10/0)
233.5 (1.3%)
258.7 (12.3%)
153.8 (-33.1%)
172.8 (-25.0%)
192.8 (-16.5%)
223.4 (-3.4%)
246.4 (6.6%)
205.6 (-11.2%)
187.0 (-19.1%)
187.6 (-19.0%)
SD
84.9
60.6
69.1
77.8
86.2
94.3
58.8
66.9
74.9
86.3
93.8
80.9
73.4
76.2
CV
0.37
0.36
0.37
0.37
0.37
0.36
0.38
0.39
0.39
0.37
0.38
0.39
0.39
0.41
Minimum
76.9
60.2 (-22.8%)
66.4 (-14.0%)
72.5 (-5.8%)
79.1 (2.8%)
86.4 (12.2%)
54.4 (-29.3%)
61.6 (-19.9%)
68.8 (-10.6%)
77.6 (0.8%)
84.7 (10.1%)
70.2 (-8.8%)
64.6 (-16.1%)
64.2 (-16.6%)
Maximum
477.0
358.6 (-24.8%)
401.8 (-15.8%)
446.1 (-6.5%)
490.5 (2.8%)
535.0 (12.2%)
348.9 (-26.8%)
388.4 (-18.6%)
428.6 (-10.2%)
487.0 (2.1%)
528.3 (10.8%)
456.2 (-4.4%)
420.2 (-11.9%)
438.9 (-8.0%)
Note: [1] Numbers in parentheses represent percent change from the base case,
24
-------
Table 8: Annual flow (taf) of the Animas River for all scenarios.
Scenario
Base
T+2°
T+2°
T+2°
T+2°
T+2°
T+4°
T+4°
T+4°
T+4°
T+4°
P-20%
P-10°/o
P+0
P+10%
P+20%
P-20%
P-10%
P+0
P+10%
P+20%
GISS2
GFDL
UKMO2
Mean [1]
550.6
406.6
458.6
512.3
568.4
628.2
376.8
424.3
473.3
525.0
578.9
505.5
459.3
465.3
(-26.1%)
(-16.7%)
(-7.0%)
(3.2%)
(14.1%)
(-31.5%)
(-22.9%)
(-14.1%)
' (-4.7%)
(5.1%)
(-8.4%)
(-16.7%)
(-15.7%)
SD
192.
143.
162.
181.
200.
5
5
3
6
8
220.5
133.
150.
168.
187.
2
8
8
1
205.5
182.4
165.7
169.2
CV
0.35
0.35
0.35
0.35
0.35
0.35
0.35
0.36
0.36
0.36
0.35
0.36
0.36
0.36
Minimum
240.4
165.
188.
212.
238.
264.
150.
170.
191.
214.
240.
9
8
3
0
4
5
6
5
6
2
205.0
184.8
182.1
(-31.0%)
(-21.5%)
(-11.7%)
(-1.0%)
(1.0%)
(-37.40/0)
(-29.0%)
(-20.3%)
(-10.7%)
(-0.1%)
(-14.7%)
(-23.1%)
(-24.2%)
Maximum
941.7
682.6
762.2
853.0
947.8
1051.5
640.1
715.8
791.8
874.2
961.5
847.2
775.1
798.8
(-27.5%)
(-19.1%)
(-9.4%)
(0.6%)
(11.7%)
(-32.0%)
(-24.0%)
(-15.9%)
(-7.2%)
(2.0%)
(-10.0%)
(-17.7%)
(-15.2%)
Note: [1 ] Numbers in parentheses represent percent change from the base case.
25
-------
All relationships between runoff and precipitation are nearly linear for the range of scenarios
studied (Figure 3), with the exception of the T+4°C scenarios on the East River. In this case, runoff
increases more slowly than precipitation. Model biases undoubtedly affect this relationship. Percent
changes in runoff are dominated by low-flow years, which are generally underpredicted; thus percent
increases in runoff are probably underestimated and percent decreases are overestimated. If this is in fact
the case, the actual relationship is somewhat curvilinear and concave up, and runoff is still more sensitive
to increases in precipitation than these results indicate.
Annual flows are normally distributed in the Two-elevation and East River models and
approximately log-normally distributed in the White and Animas.River models. In all cases, the climate
o
C
cr
0)
en
c
ID
U
CD
CJ
CD
Q.
30
20
10
0
-10
-20
-30
-40 -
White River at Meeker
18.6
1.3
-30
-20
-10
0
10
20
30
Percent Change in Precipitation
Figure 3: Change in runoff as a function of change in precipitation for the White River model.
The relationship is nearly linear for the range of hypothetical scenarios modeled here.
26
-------
.3
.15
0
.3
0
•i-i
t; -15
to
c.
L.
0
.3
.15
0
(
l
EL L .
F
I;
i
1 Base
i
i
111
i
|T+2 C, P-20%
i
T+4 C, P-20%
White River at
Distribution of Total
by scenari
i
I/JT+2 C, P+0
J
J;i
iflk
T+4 C. P+0
|
Meeker
Annual Flow
0
iT+2 C, P+20%
nf'
A
lULfl
T+4 C, P+20%
hi
3 400 BOO 0 400 800 0 400 800
Annual Flow [thousand acre-feet]
Figure 4: Distribution of annual runoff (taf) for the White River model for selected hypothetical
scenarios.
.5 i
.25-
0
.5 H
c
o
-l-l
ro -25
C-
LL.
o
.5
oc;
0
• Animas River at Durango
1 ase Distribution of Total Annual Flow
by scenario
JLji
rffjftmJlih
i
| T+2 C, P-20%
i
ITT
""*
J !
imij-]
I
i T+4 C, P-20%
1
E£
i r
| T+2 C, P+0
i
fh i
rJtterlhi
t
' T+4 C, P+0
i
rl m N
rfl 1 1 H |H I'l
T+2 C, P+20%
n
In
JltoiMi
T+4 C, P+20%
PI
J'rTh jn Ji
rfl II THl Wl'hl
0 600 1200 0 600 1200 6 600 1200
Annual Flow [thousand acre-feet]
Figure 5: Distribution of annual runoff (taf) for the Animas River model for selected
hypothetical scenarios.
27
-------
change scenarios result in distributions of annual streamflow that are roughly log-normal (Figures 4-5).
Temperature increases cause annual flows to decrease and to consolidate, i.e. the distribution narrows, and
low-flow years become more frequent. Precipitation increases of 20% spread the distribution at the upper
end. This result is also evident in the coefficient of variation, which increases in most of the scenarios that
incorporate a 20% precipitation increase (Tables 5-8). The implication is that increased flows are likely to
increase variability on an annual basis.
The statistical significance of these results was estimated following the method used by Klemes
(1985: App. B). For each scenario, the mean and standard deviation (u,a) of the annual streamflow series
were treated as perfect estimates of the true mean and standard deviation for the distribution of annual
flows. Subsequently, 125 series of 35-year flows were randomly generated from a log-normal distribution
defined by/i and a. The mean and standard deviation of each 35-year series were then plotted (7 versus
n), and the 90% confidence region was defined to be the ellipse that contained 90% of these points. These
confidence regions are illustrated for the White River model in Figure 6.
Using the above method, only three scenarios were significant for all basins at the 90%
confidence level: T+4°C, P-20%; T+4'C, P-10%; and T+2*C, P-20%. For the White River, one additional
scenario, T+2? C and P+20%, was also significant. None of the GCM scenarios were significant at the 90%
level. The statistically significant scenarios correspond to a minimum change in mean annual streamflow
of 18% on the White River, 25% on the East River and 22% on the Animas River (Nash and Gleick, 1991).
Seasonal Runoff
Temperature increases cause peak runoff to occur earlier in the year. A temperature increase
of £? C shifts peak runoff from June to May for the White and Animas rivers. For the East River, peak runoff
still occurs in June, although it is not nearly so exaggerated. For all three basins, the 2" C rise creates a
double peak, with high runoff occurring in both May and June. When temperature is increased by 4° C, the
28
-------
o>
u
ID
•a
10
en
o
c
o
•rH
4J
(0
-iH
0)
a
•0
(D
TJ
to
4-1
en
400 -
300 -
^ 200 -
100 -
0 -
200
300 400 500
Mean [thousand acre-feet]
600
Figure 6: Point estimates of annual flow (mean and standard deviation) for the White River,
with approximate 90% confidence regions for the base case and selected hypothetical
scenarios.
East River also undergoes a distinct shift in the timing of peak runoff, from June to May. The UKMO
scenario for the Animas and White rivers shifts peak runoff from June to April, which reflects the 6.ffC
temperature rise. Figure 7 illustrates the general effect of temperature on the timing of peak runoff for the
East River. In all cases, the sub-basins remain snowmelt-driven, although peak runoff is occurring earlier
in the year.
Histograms of January and June runoff are presented for the Animas River in Figures 8 and
9. The distribution of January runoff becomes much more flat as a result of increases in temperature and/or
precipitation. This is indicative of the higher flows which are occurring during the winter, as more
precipitation falls as rain rather than snow. Still, flows in January are very low compared to typical spring
or summer flows. The impact of dimate-change scenarios on June runoff is the opposite. Increases in
29
-------
o
o
as
E
r-4
O
100 -
75-
50 -
25-
0 -
East River at Almont
D J F
A M
Figure 7: Effect of temperature increases on the average hydrograph (East River model). A
temperature increase of 4°C shifts peak runoff from June to May.
temperature cause the distribution to narrow. Whereas in the base case, June runoff ranges from
approximately zero to 400 thousand acre-feet (taf), a temperature increase of 4" C cuts this range in half,
from zero to 200 taf.
Figure 10 illustrates mean runoff as it varies between high- and low-flow seasons for the White
River. Spring runoff is averaged over three months of high runoff (April, May, June) and fall runoff over three
months of low runoff (October, November, December). These results suggest less extreme seasonal flows
as a result of climate change in most cases. The effect of an evenly applied increase in temperature is to
reduce the seasonal variation in runoff, primarily as a result of reduced streamflow in the spring. In the
Animas River model, climate scenarios diminish the difference between spring and fall flows because spring
runoff decreases in all cases. When substantial precipitation increases are incorporated into the model,
however, seasonality becomes more pronounced. In the White and East River models, climate scenarios
do not decrease spring runoff as dramatically, while scenarios that incorporate precipitation increases of
20% augment spring runoff substantially.
30
-------
.5 •
.25
0
.5
o
•H „-
-M .25
CJ
fO
U.
.25
0
i
"~ rlT
Base
H
J
IE)
T+2 C, P-20%
Jia'
1
| T+4 C, P-20%
i
it
Animas Riven at
Distribution of Ja
by scenar
i
| T+2 C, P+0
n '
i 1
i T+4 C, P+0
liL
Dunango
nuary Flow
io
T+2 C, P+20%
T+4 C, P+20%
0 20 40 0-20 40 0 20 40
Volume of Flow [thousand acre-feet]
Figure 8: Distribution of January runoff (taf) for the Animas River model for selected
hypothetical scenarios.
.5 -1
.25
0 i
.5-j
c
o
Tl OK
£ -25 -
CD
C_
U. '
0.
.5 4
0 •
i
1 Base
'
i
jiyidMj
| T+2 C, P-20%
^
%
' I "!
9
* r!
iflmrvi-i
- T+4 C, P-20%
m^,
Animas Riven at Dunango
Distnibution of June Flow
by scenario
T+2 C, P+0
Jlftl n^i FT-H dl
-
i
0 200 400 0
T+4 C, P+0
ro-ifc]
T+2 C, P+20%
n
I'vl
HLJ'i
JHlLn ECB dli d3
T+4 C, P+20X
m
>
^
200 400 0 200 400
Volume of Flow [thousand acre-feet]
Figure 9: Distribution of June runoff (taf) for the Animas River model for selected
hypothetical scenarios.
31
-------
500
• Annual
H Fall
0 Spring
White River at Meeker
T+4 C Scenarios
Base
-20% -10% 0 +10%
Precipitation Scenario
+20%
Figure 10: Mean annual runoff, mean spring (April, May, June) runoff, and mean fall (October,
November, December) runoff for the White River at Meeker. The base case and T+4°C
scenarios are shown.
Transient Scenario
The changes in temperature and precipitation generated by the GISS transient scenario for the
year 2030 fall within the range established by the hypothetical scenarios in which runoff varies linearly with
changes in precipitation. Thus, using the data generated by the hypothetical scenarios, we interpolated to
find corresponding changes in runoff for the transient scenario. For the more northern GISS grid point,
which encompasses the White River basin, temperature increases by 3.2° C and precipitation increases by
10%. This corresponds to an increase in mean annual streamflow of about 4% on the White River at Meeker
and a significant shift in seasonality. For the southern grid point, which encompasses the Animas and East
river basins as well as the Lake Powell inflow (Two-elevation model), temperature rises by 2.5" C and
precipitation increases by 20%. This corresponds to an increase in mean annual runoff of 12% on the
Animas River, 11% on the East River, and 9% in the Two-elevation model (inflow into Lake Powell) (Table
9).
32
-------
Table 9: Changes in runoff generated by GCMs and the NWSRFS
hydrologic model
Transient [2]
GISS1
GISS2
A Runoff (%)
GCM
NWSRFS
Equilibrium [1]
GISS1
GISS2
GFDL
+20
+5
+5
+10
-8 to -14
-13 to -16
-5
+30
+4
+10 to+12
Notes: (1] Equilibrium GCM runs, in which greenhouse gas concentrations have stabilized at
roughly twice current levels.
[2] The GISS transient run, in which greenhouse gases are increasing gradually. The
numbers presented here represent the avearge over the decade 2030 to 2039.
GCM Runoff Scenarios
GCM runoff scenarios are compared with the NWSRFS modeling results in Table 9. GCM
runoff predictions do not necessarily agree even in direction with those suggested by the hydrologic
modeling of GCM changes in temperature and precipitation. In the GISS equilibrium runs, runoff increases
by 20% at the more northern grid point (GISS 1) and by 5% at the more southern grid point (GISS 2).
Hydrologic modeling results that used the GISS temperature and precipitation inputs suggest that runoff
would increase by 10% in the White River basin (GISS 1) and decrease between 8 and 14% in the GISS 2
region. For the GFDL model, the runoff outputs indicate a increase of 5%, while hydrologic modeling
suggests runoff decreases between 16 and 23%. For the GISS transient scenario, GCM runoff decreases
by 5% at the more northern grid point, while the White River model suggests that equivalent temperature
and precipitation changes would result in a 4% increase in runoff. In the lower basin, represented by the
33
-------
GISS 2 grid point, GCM runoff increases by 30%. Hydrologic modeling using temperature and precipitation
inputs from the same grid point indicate that runoff would increase only between 10 and 12%. In general,
GCMs underestimate decreases in runoff and overestimate increases when compared to corresponding
outputs from the NWSRFS hydrologic model.
Discussion of Hvdroloaic Modeling Results
In the first study to analyze the impacts of climatic change on the Colorado River, Stockton and
Boggess (1979) used Langbein's relationships (Langbein and others, 1949) to estimate the effects of a 2*C
temperature rise and a 10% decrease in precipitation. Their results suggested that streamflow in the upper
basin would decline by about 44%. Following up on that work, Revelle and Waggoner (1983) developed
a linear regression model of runoff, using precipitation and temperature as independent variables. Their
results indicated that a 2? C temperature increase would decrease mean annual streamflow by 29%, while
a 10% decrease in precipitation would decrease runoff by about 11%. In combination, these changes would
result in a 40% decrease in runoff, in close agreement with Stockton and Boggess's earlier result.
In contrast, our studies with a conceptual model suggest less severe impacts on runoff and a
relatively greater sensitivity of annual runoff to precipitation rather than temperature changes. A 2°C
temperature rise decreases mean annual runoff by less than 10% in the three sub-basins studied. When
combined with a 10% decrease in precipitation, runoff decreases are on the order of 20%. These results
are comparable to other studies of arid and semi-arid basins that have used conceptual hydrologic models
(e.g. Gleick, 1987b; Flaschka, et_aj., 1987), supporting Karl and Riebsame's (1989) conclusion that the
Langbein relationships overstate the role of evaporation.
In a recent study, Schaake (1990) modeled the Animas River altering temperature, precipitation,
and potential evapotranspiration independently. (In contrast, in this study, changes in PET were linked to
changes in temperature.) Schaake found that a ? C temperature rise and a 10% increase in PET resulted
34
-------
in a 9% decrease in mean annual runoff. Our results show a 7% decrease in mean annual runoff for a 2? C
temperature rise and an 8% increase in potential evapotranspiration (refer to Table 8), which is in close
agreement with the results from Schaake. For the range of scenarios presented here, mean annual runoff
changes nearly linearly with precipitation, although this relationship begins to break down as precipitation
increases by 20% at which point runoff begins to increase relatively faster. Results from Schaake indicate
that, in the absence of temperature and potential evapotranspiration increases, this non-linearity occurs for
a precipitation increase of only 10%, which causes a corresponding increase in runoff of 19%. Overall, our
results are within the range reported by other investigators for semi-arid river basins (Table 10).
The results derived from GCM scenarios fall within the range established by the hypothetical
scenarios. Of the three GCMs, the GFDL model (T+4.9°C, P+0) results in the most extreme decreases in
runoff for all basins (-10% to -24%) because it predicts a relatively large regional temperature increase and
no change in precipitation. The least extreme effects are generated by either the UKMO 1 or the GISS 1
grid point, which incorporate respective increases in precipitation of 30% and 20% and lead to increases
in runoff of 0 to 10%. Overall, however, the GCM scenarios suggest that decreases in runoff are much more
likely than increases in this region. This is consistent with the work of Rind, et al. (1990), who have
analyzed the frequency of droughts using GCM outputs other than soil moisture and have found increased
drying. Moreover, it is only the GCM grid points which incorporate large increases in precipitation (20 to
30%) in which runoff does not decrease. The greater uncertainty associated with precipitation changes
should be kept in mind. All the GCM scenarios suggest large regional increases in temperature, which
would lead to decreased runoff, unless offset by precipitation increases of 20% or more.
The GISS transient scenario implies increases in runoff in all three sub-basins and in the Two-
elevation model. These range from 4% in the White River basin to 12% in the Animas River basin. In
contrast, the GISS equilibrium scenarios imply decreases in runoff of -8% to -14%, except on the White River
where runoff increases by 10%. This suggests the potential for short-term increases in runoff (due to
35
-------
*^^
*_•
co
1
JD
O
>C
E
,c
a=
o
2
•S
3
CO
CO
E
o
co
CD
0)
CO
XI
.0
to
"o
0
s
a
E
o
.52
rt
O
o
^
0
o
4*
£ h
1
CD
Q.
0)
C
0)
g>
5
O
O
o
2
O
o
T—
C
O) CO
C AS
CO .£ CL
5 'g
ol
£ S K
CM CO T-
1 1 1
.i a
CC CD
13^
i s 1
rt 5 CO 5- CO
CM i— CM T-
1 1 1 1
2
N.
r-
1
•—• , «2.
2 tr i
Q) QJ O
J> ,3> Oir
if ic o •
il|i
m e 2 cc
•*-* nj ^ r>
i i
. 1
jg CC
r ra
S 'E
U <
I
§ -2
c? I
.
>
ir
o
Eg
$£$
' T- 1
^
f <£
^
^I^SiE
?*l*-8
iver
2 ^2
l
ill
§g§^
-------
changed precipitation patterns) that may obscure a long-term trend towards decreases in runoff for some
sub-basins.
Runoff results taken directly from GCMs show poor correspondence with results generated by
the NWSRFS model using GCM temperature and precipitation scenarios. In general, runoff and soil moisture
outputs from GCMs suggest less drying than the NWSRFS model, despite increased air temperatures and
PET. Rind, et al. (1990) have concluded that soil moisture deficits and vegetation dessication are
understated in the GCM simulations because of their lack of realistic land surface models. Thus, even
though GCM estimates of PET may be quite high (reflecting higher temperatures), actual evapotranspiration
remains quite low in the models due to inadequate assumptions about evapotranspiration efficiency.
Overall, GCM predictions of runoff should be considered less reliable on a regional basis than those results
obtained by hydrologic modeling (WMO, 1987).
The statistical significance of these results cannot be assessed in a definitive manner. On the
one hand, because data generated by the sensitivity runs are highly correlated with data generated by the
base runs, sensitivity estimates of changes in the mean and standard deviation would be expected to be
reasonably accurate and statistically significant with respect to one another. At the same time, however,
the streamflows generated by the scenarios may not be significantly different from values compatible with
the historic streamflow series. Using the method put forth by Klemes (1985, App. B), our analysis suggests
that precipitation changes of more than 10% would be necessary before changes in runoff would be
significantly different from the historic streamflow series, even if the streamflow distribution were to remain
stationary. Moreover, temperature changes of 4? C would not produce a statistically observable impact on
runoff, unless accompanied by precipitation decreases. This is consistent with the finding of Klemes (1985)
that precipitation changes of 15 to 20% would be required to generate statistically significant changes in
runoff in the Pease River (Texas) and the Leaf River (Missouri). This conclusion does not imply that the
impacts of climatic change are insignificant but does suggest the difficulty inherent in detecting the impacts
37
-------
of climatic change on runoff, given a relatively short and variable streamflow record. Thus, it is likely that
long-term changes in the hydrologic regime on the Colorado River attributable to climatic change would be
interpreted as extreme events (e.g. as droughts) for some time and may delay adaptation as a result.
Although all the scenarios studied alter the annual and monthly distribution of flows, annual
variability is not strongly affected. This is as we expected, given that we did not alter the distribution of the
model inputs, but merely transposed them. In addition, the differential effect of the scenarios on high- and
low-flow years is relatively moderate. While the percent change in mean annual runoff with respect to the
base case is higher for low-flow years than it is for high-flow years, in all cases these differences are within
10 percent. Of potentially greater concern is the increased frequency of extreme events; however, better
information is needed from GCMs before changes in interannual variability can be properly evaluated
(Mearns, etal., 1990).
The analysis of seasonal impacts is constrained by the fact that changes in temperature and
precipitation were applied uniformly to all daily data. Actually these annual changes would be distributed
unevenly throughout the year. While GCM results provide some insights into seasonal changes, they are
not definitive. The GISS and UKMO models suggest that absolute temperature increases in the Colorado
River Basin are greater in winter, while the GFDL model indicates that temperature increases are greatest
In the summer and fall months. All three GCMs are in agreement with respect to the prediction that
percentage increases in precipitation are likely to be greatest in the winter and spring. Because these are
the seasons with the greatest precipitation under current conditions and because there is likely to be a
considerable loss of snowmelt storage due to higher temperatures, a relative increase in winter and spring
precipitation could substantially increase the probability of flooding, particularly if operational procedures
are not rapidly adjusted.
38
-------
Our results suggest that an increase in temperature will shift the seasonally of runoff, with peak
runoff occurring in May rather than June. This change reflects the fact that under higher temperatures more
precipitation falls as rain rather than snow, and snowmelt runoff occurs earlier in the year. This result has
been seen in several other regional studies (e.g. Gleick, 1986; Bultot. et al.. 1988). Moreover, because this
seasonal result is induced by changes in temperature, rather than more uncertain changes in precipitation,
the authors believe it is fairly robust. Temperature increases had a much smaller effect on the White River
than on the other basins, which is due to the lower elevation of the White River basin. The NWSRFS model
reduces evapotranspiration when snow is on the ground by an amount proportional to the area! snow cover.
Because a rise in temperature causes less ground to be covered with snow for fewer days out of the year,
evapotranspiration increases while runoff decreases. We would expect this effect to be most significant in
higher elevation basins which have proportionately more snow cover. This is in fact the case for the three
sub-basins modeled here. The highest elevation basin, the East River at Almont, also shows the greatest
sensitivity to temperature increases. Overall, the Two-elevation model showed an even greater sensitivity
to changes in temperature, which may reflect a greater sensitivity to evapotranspiration, although it is difficult
to draw a comparison because of the vastly different scale of the Two-elevation model. On a percentage
basis, the sensitivity of runoff to temperature in the White River was less than one-half that in the Two-
elevation model. All four models showed nearly an equal sensitivity to changes in precipitation. Relative
seasonal changes are most significant for the East River, in which 10% and 20% increases in precipitation
increase the absolute variation in runoff between spring and fall months. The interpretation of NWSRFS
model results in this study must be tempered by three principal caveats. First, as described above, the
ability of the NWSRFS model to accurately simulate runoff under conditions of altered climate is subject to
some question. Secondly, all climate scenarios were applied on an annual basis, which may be a
reasonable approximation for temperature increases but undoubtedly skews seasonal precipitation patterns
which are likely to change dramatically under conditions of altered climate. Finally, the historical record was
limited to 35 years, which is too short to allow a substantive analysis of natural (non-greenhouse) variation.
39
-------
Yet notwithstanding these limitations, the authors believe that the NWSRFS results provide the best
information currently available on the sensitivity of runoff in the basin to climatic changes.
In summary, the hydrologic modeling results suggest that significant decreases in runoff are
a likely impact of climatic change in the Upper Colorado River Basin. These results are consistent with
similar studies of semi-arid basins. The potential water-supply implications of these changes are evaluated
in the following section.
40
-------
[Blank]
41
-------
[Blank]
42
-------
METHODS OF ANALYSIS II: WATER-SUPPLY MODELING
Description of the Model
The impacts of changes in runoff on water supply and delivery were analyzed using the U.S.
Bureau of Reclamation's Colorado River Simulation System (CRSS). The CRSS is a reservoir-simulation
model that tracks streamflow, reservoir storage, and water supply throughout the Colorado River Basin using
a monthly time-step. It uses adjusted, historical hydrologic inputs ("natural streamflow"), projected water
demands, reservoir characteristics (e.g., area-capacity relationships), and operating policies (e.g., scheduled
releases, reservoir target storages) to determine levels of water deliveries to various users. All the major
hydrologic and storage features of the Colorado River Basin are modeled. The model was designed to
simulate the operating policies that are currently used by the Bureau of Reclamation. The outputs of the
model are actual streamflow and salinity, reservoir levels, hydroelectricity generation, uncontrolled spills, and
water deliveries on a monthly basis. The CRSS serves as the Bureau of Reclamation's primary tool for
studying the operation of the river and the impact of projected developments in the basin. The model is
documented in USDOI (1987). By changing either inputs (e.g., natural streamflow) or operating parameters
(e.g., reservoir target storages), modelers can study the response of the whole system. In no sense does
the model "predict" future shortages or surpluses, but it does portray the sensitivity of those outcomes to
changes in inputs or operating parameters.
The hydrologic inputs to the model are natural streamflow and salinity data, which are defined
as historical data adjusted to remove the effects of human development. Historical streamflow data for most
stations on the Colorado River exist from 1906. Gaps in the data base have been filled by regression
estimates. To derive natural streamflow data, changes in river flow and water use due to human demands,
changes in vegetation, and changes in basin evaporation are calculated, and historic flows are adjusted
accordingly. Historical salinity data were developed by the USGS using a regression procedure that
calculates salt load as a function of historical streamflow and several variables representing development,
including upstream adjustments to streamflow, consumptive use, diversions, and irrigated acreage. Adjusted
43
-------
results presented for the model indicate that errors tend to be systematic rather than random: low flows are
underpredicted, and high flows are overpredicted.12
.Modeling Assumptions
For this study, hydrologic inputs were developed using the Index Sequential Method (ISM), in
which the historic record is wrapped around itself and run through the model using different starting dates.
The existing record can be thought of as a piece of tape in which the year 1906 appears on one edge and
1983 appears on the other. In the ISM method, the ends of the tape are connected and the record becomes
continuous, with year 1983 immediately preceding year 1906. The starting point for modeling purposes can
now be chosen from among any of the years. Every year in the record may be used as a separate starting
point, or, for convenience, some limited set may be selected, such as every fifth year. The use of historic
data in the ISM rests on the assumption that past streamflows are indicative of the future, i.e. that the
geophysical processes governing streamflow are both stationary and well-described by existing data.
Accordingly, the past record is assumed to provide reliable information about the statistical properties of
future flows, including mean, variance, and skewness, even though the sequence of future flows will
undoubtedly be different from the past. The ISM allows the historic data to maintain its statistical
characteristics (e.g., mean, variance, and skewness); but it also introduces some uncertainty with respect
to the timing of specific streamflow sequences, allowing an analysis of the effects of the hydrologic starting
point on results, e.g., the effect of having the 1920s' "wet period" early or late in the simulation period. In
this study, the hydrologic record was staggered by 5 years, and 15 sets or "sequences" of data were
simulated. Trace 1 begins with data from 1906, trace 2 with data from 1911, and so on. The Index
Sequential Method is frequently used to generate probabilities of occurrence in any single year or set of
12See USDOI (1987), Section IV, "Validation", especially plot no. 2.
46
-------
-years. Thus, if the information of interest is the probability of water shortages in the year 2020, potential.
flows in the year 2020 are generated by a set of historical traces. The traces are then treated as
independent observations, and the probability that a shortage will occur is inferred.13
In contrast, in this study we were interested not in the performance of the system at a particular
date, but in how it would function over the long-term under scenarios of climate change at some unspecified
time in the future. Because climatic change is an incremental, but not necessarily linear, process that will
occur gradually over the next century and beyond, the timing of its occurrence cannot be predicted with
any accuracy. Thus, our aims for this study were to compare a limited number of scenarios under
hypothetical "normal" conditions and under conditions of altered streamflow in order to ascertain the
sensitivity of the system to possible climatic changes.
We selected three historical sequences that were analyzed independently in order to: (1) assess
the impact of different trace starting points on the statistics of interest (i.e. the difference in results among
sequences 1, 2, and 3); (2) bound the plausible results that might be generated by different historical data
sequences; and (3) analyze the impact of changes in runoff inputs on seasonal and annual streamflow
statistics (e.g. how a 10% decrease in natural streamflow inputs compares to the base case for a given
historical sequence). The results presented here should thus be interpreted not as probabilities but as
sensitivities. They suggest how a number of water-supply variables would change if a given historical data
sequence were altered in the manner specified in each scenario; they say nothing about the likelihood of
occurrence.
13
In actuality, the historic record cannot be used to develop probabilities of future events, given that
the distribution of future streamflows is unknown. Despite this fact, the term "probability" is commonly used
in such studies.
47
-------
The CRSS is capable of running up to 150 years in single simulation. In this study, we chose
to analyze 78 years of data because we felt that it provided a long enough sequence of years for our
analyses without forcing us to selectively repeat some, but not all, of the historic data. Thus, our base
period for this portion of the study consists of the historical hydrology from 1906 to 1983. In order to
alleviate the inconsistency created by varying demands during the operation of the model, we elected to
analyze only those years in which demand is constant. There are two reasons for this: (1) varying demands
obfuscate the effect of the trace starting pojnt on model results; and (2) the climate-change scenarios refer
to an equilibrium condition (e.g. in the case of the GCM scenarios, a point at which atmospheric COfe has
doubled) at some unspecified future date, thus we did not want the analysis to be dependent on how
demands might vary in the period 1990 to 2040. We report our results as monthly or annual frequencies
derived from a 78-year model run ("Years 1 to 78"), with demands constant at year 2040 predicted levels.14
We were also constrained in this study to use October 1989 reservoir levels as our starting
point for each simulation run. Because of the large storage-to-annual-flow ratio on the River (approximately
four-to-one), starting storage levels can have a significant effect on results. After 50 years of simulation,
different sequences produced very different reservoir levels. Thus, by choosing to analyze only the last 78
years of a 128-year run, starting storages were varied implicitly by sequence.15 Of the 15 sequences
14
Because the model is run in "real-time" mode (i.e. 1989 was equal to Year 1 in our model runs), in
order to maintain demands at constant levels, the model was run for 130 years (1989-2119). Demands are
scheduled to become constant in Year 2040. Thus, we analyzed the last 78 years of a 130-year simulation
(2041-2119).
Ideally we would have preferred to run the exact same sequence of historic data with different starting
storages in order to analyze independently the impact of initial reservoir storage levels. This would allow
us to see explicitly how water-supply variables are affected by initial storage levels. This was not possible
for this study. Although the method used here allowed us to vary starting storage levels, it did not allow us
to analyze their impact because starting storages are implicitly related to the starting point of each historic
data trace (i.e., a high level of initial storage results from the wetter periods having occurred recently in the
model run, and thus these very wet sequences will not occur again for several decades.) Thus, as we note
later, the difference in results among sequences was not great.
48
-------
produced by the CRSS model, we selected three to analyze, which correspond to low, medium, and high
starting storage levels. A description of these sequences is given in Table 12.
For this part of the project, hypothetical scenarios of runoff were constructed as percent
changes. The hypothetical scenarios analyzed include changes in natural runoff of ±5%, ±10%, and ±20%.
The magnitude of these changes corresponds roughly to the results generated by the NWSRFS model,
which suggested that changes in runoff in the higher elevations of the upper basin were likely to range from
-30% to +10%. Because the model generates extreme results for the -20% scenario, we did not attempt
to model a decrease in runoff of -30%. We chose to vary streamflow inputs systematically in order to
t
generate information about the sensitivity of the system to variations in runoff inputs. Percent changes were
applied uniformly to all the input data used in the model, e.g. natural (historic) streamflows were decreased
by 10% at all points and then run through the model.16 This resulted in a new set of natural runoff
numbers in which the mean was altered by a specific percentage and the variance was altered in proportion
to the mean (i.e., the coefficient of variation remains unchanged). Although the variability of climate and
runoff may change as a result of the greenhouse affect, at present, very little is known about how future
climatic changes will affect variability. Neither GCMs nor historical data give a clear indication of how
variability will change, nor is there any reason to expect a homogenous response to warming in terms of
changes in variability (ICF, 1989; IPCC, 1990; Mearns, et al., 1990).
In addition, in order to assess the effect of a shift in the timing of runoff, a time-shifted scenario
was modeled in which runoff inputs were shifted backward by one month; thus, historic flows for February
were fed into the model as January runoff. This simulates the seasonal effects of increases in temperature
on snowfall and snowmelt as discussed above. (See discussion of seasonal runoff under Results of
Hydrologic Modeling, above).
16Percent changes in runoff were applied to years 53 through 130 (i.e. Years 2042-2119). The model
was run for the first 50 years without any alteration in inputs. (See FN #14 above.)
49
-------
Table 12: Description of input sequences.
Sequence
Number
1
2
3
Starting
Storage (taf)[1]
20,995
36,482
54,647
Historic
Input Data [2]
1967-1983; 1906-1966
1944-1983; 1906-1943
1929-1983:1906-1928
Notes: [1] Total system storage (Upper and Lower basins) at beginnning of period of analysis.
[2] This shows the order in which historic hydrology was run through the model for each
sequence. Year 1 in the model runs uses natural flow data from 1967 for sequence 1
1944 for sequence 2, and 1929 for sequence 3.
50
-------
RESULTS OF WATER-SUPPLY MODELING
Runoff
Changes in runoff were analyzed at five points in the system: Green River, at Green River,
Wyoming ("Green River"); the Colorado River at Cisco ("Cisco"); the San Juan River at Bluff ("Bluff'); the
Colorado River at Lees Ferry ("Lees Ferry"); and the Colorado River below Imperial Dam ("Imperial"). Green
River, Cisco, and Bluff are all upper basin points. Lees Ferry is located near Glen Canyon dam, about 16
miles upstream of the Compact Point. Imperial Dam is located in the lower basin (Figure 11).
Changes in the mean, standard deviation, maxima, and minima of annual runoff at Green River,
Lees Ferry, and Imperial Dam are summarized in Tables 13-15. Generally the differences in annual statistics
generated by different sequences were not significant, in part because starting storage levels and hydrologic
trace were not varied independently.17 Thus, those sequences that had low starting storages also had
relatively high flows early in the simulation run. Because of the small differences generated by the different
sequences, the results of only one sequence, sequence number two (s2), which represents a middle
scenario, are presented here. (For comparative purposes, the annual statistics at Lees Ferry are given for
all three sequences in Appendix C, Table C2. Differences in the mean among sequences are within 2% for
all scenarios.)
A 20% decrease in natural runoff causes between an 11% to 31% decrease in modeled runoff
among the five points analyzed. A 20% increase in natural runoff causes a 31% increase in modeled runoff
at each of the five points analyzed. For the upper basin points, a 5% change in natural runoff causes a 7
to 8% change in actual runoff, and the effect of changes in natural runoff is essentially linear over the range
of scenarios examined. This is not true in the lower basin where storage has a greater mitigating effect on
decreases in natural runoff.
17
See FN #15 above.
51
-------
SELECTED CRSS STREAMFLOW STATIONS
1. Green River near Green River, Wyoming
2. Colorado River near Cisco, Utah
3. San Juan River near Bluff, Utah
4. Colorado River at Lee Ferry, Arizona
5. Colorado River below Davis Dam, Arizona/Nevada
6. Colorado River below Parker Dam, Arizona/California
7. Colorado River above Imperial Dam, Arizona
WY
NEVADA
boundary between upper
and lower basins
NEW MEXICO
Figure 11: Map of the Colorado River basin (excluding Mexico) showing the location of
selected CRSS stations and major reservoirs. (Source: redrawn from USDOI, 1987.)
-------
Decreases in natural runoff cause severe changes in annual minimum runoff. For instance, the
-10% scenario causes mean annual runoff in the upper basin to decline by about 15%. Minimum flow,
however, declines by between 32% (at Cisco) and 86% (at Lees Ferry). Even the -5% scenario causes runoff
at Lees Ferry to fall considerably below the objective minimum release of 8.23 maf in 6 years, while the -10%
scenario causes streamflows to fall below this level in 15 of the 78 years. Also interesting is the fact that
increased-flow scenarios do not change the annual minimum streamflow at Lees Ferry and Imperial Dam.
Even in the +20% scenario, annual deliveries at Lees Ferry still fall to 8.23 maf in 14 of the 78 years. The
increased-flow scenarios cause maximum flows in the upper basin to increase by up to 27% (in the +20%
scenario). The +20% scenario causes the maximum annual runoff at Lees Ferry to jump by 35%, from 17
maf to nearly 23 maf. At Imperial Dam, this same scenario raises the maximum annual runoff to 17.8 maf.
Model outputs are closely correlated with patterns in the historical data that are used as model
inputs. In Figures 12 and 13, annual runoff at Green River and Lees Ferry has been smoothed (using 3-year
moving averages) and plotted as a function of time (year). At the upstream point of Green River some
extremes are evident. A sequence of low-flow years occurs between years 9 and 20 and again between
years 63-68. When correlated with model inputs, these periods correspond to the actual years of 1953-1964
and 1929-1933, respectively. Similarly a high-flow period is obvious between years 38-50, which correspond
to the historical years 1983 and 1906-1917 and which, in fact, were the highest runoff periods in the existing
instrumental record. These patterns are even more obvious at Lees Ferry, where annual flows are tightly
controlled (see Figure 13). In the base case, annual releases from Lake Powell never drop below the
objective minimum of 8.23 maf/year; however, a runoff decrease of 10% causes releases from Lake Powell
to fall below 8.23 maf in years 9-20, resulting in shortages to lower basin users. Historically, this period
(1953-64) is the most critical dry period on record in terms of water supply. Similarly the effect of the "wet
period" that occurred in the early part of the century is also very evident; even streamflows in the -10%
scenario rise above the 8.23 maf level for several years. Thus, when interpreting these results, the historical
hydrology needs to be kept in mind:
53
-------
Table 13: Annual flow (taf) of the Green River at Green River, WY.
Scenario
-20%
-10%
-5%
Base
+5%
+10%
+20 %
Mean
Flow[1]
679
827
902
977
1,051
1,126
1,277
(-30.5 %)
(-15.3%)
(-7.6 %)
(7.7 %)
(15.4 %)
(30.8 %)
Standard
Deviation
303
353
378
404
429
454
503
Minimum
Flow
91
151
197
252
282
287
304
(-63.9 %)
(-40.1 %)
(-21.8%)
(11.9%)
(13.9%)
(20.6 %)
Maximum
Flow
1,424
1,693
1,826
1,964
2,098
2,231
2,502
(-27.5 %)
(-13.8%)
(-7.0 %)
(6.8 %)
(13.6%)
(27.4 %)
Note: [1] Numbers in parentheses represent percent change compared to the base case.
A more meaningful way to look at annual runoff is to consider how runoff frequency changes
across scenarios. Figure 14 shows the cumulative frequencies of annual runoff at Lees Ferry. The
cumulative frequency diagram shows a sharp discontinuity at 8.23 maf, which represents the objective
minimum release from Lake Powell. In the base case scenario, JTQ years have a streamflow less than 8.23
maf, but in the -5% scenario about 6% of the years fall below 8.23 maf; in the -10% scenario, this increases
to 17%; and in the -20% scenario, 36% of the years fall below this targeted level.
Reservoir Storage
Much of the difference in runoff generated by the climate-change scenarios, rather than being
passed through the system, is being cushioned through increased water storage or increased releases.
While the natural streamflow data that are input into the model refer to a condition in which no storage
54
-------
Table 14: Annual flow (taf) of the Colorado River at Lees Ferry (below Glen Canyon Dam).
Scenario
Mean
Flow[1]
-20 %
-10%
-5%
Base
+5%
+10%
+20 %
6,929
8,205
8,801
9,372
10,037
10,774
12,289
(-26.1 %)
(-12.5%)
(-6.1 %)
(7.1 %)
(15.0%)
(31.1 %)
Standard
Deviation
2,024
1,784
1,693
2,089
2,572
3,023
3,549
Minimum
Flow
832
1,143
3,710
8,230
8,230
8,230
8,230
(-89.9 %)
(-86.1 %)
(-54.9%)
(0)
(0)
(0)
Maximum
Flow
8.230
15,790
14,514
16,869
18,671
20,307
22,756
(-51.2%)
(-10.0%)
(-14.0 %)
(10.7%)
(20.4 %)
(34.9 %)
Note: [1] Numbers in parentheses represent percent change compared to the base case.
Table 15: Annual flow (taf) of the Colorado River at Imperial Dam-
Scenario
-20 %
-10%
-5%
Base
+5%
+10%
+20 %
Mean
Flow[1]
5,381
5,605
5,818
6,053
6,366
6,742
7,954
(-11.1 %)
(-7.4 %)
(-3.9 %)
(5.2 %)
(11.4%)
(31.4%)
Standard
Deviation
511
279
611
1,112
1,527
2,013
2,873
Minimum
Flow
2,565
3,524
5,650
5,650
5,650
5,650
5,650
(-54.6 %)
(-37.6 %)
(0)
(0)
(0)
(0)
Maximum
Flow
5,656
6,270
6,057
11,241
13,646
15,186
17,773
(-49.7 %)
(-44.2 %)
(-19.4%0
(21.4%)
(35.1 %)
(58.1 %)
Note: [1 ] Numbers in parentheses represent percent change compared to the base case.
55
-------
2500 -
2000 -
Q)
CD
H-
0)
8,
T3
to
m
o
o
1000 -
500 H
0 -
.Annual Flow at Green-River
(3-year moving average)
Year
~7B
Figure 12: Annual runoff (taf) at Green River in the base case and the ±20% runoff scenarios.
Runoff is plotted as a three-year moving average.
17000 -
r~i
-*J
S
H-
1
m
§
TJ
m 8230 —
§
o
j— < • • •
•tj
L_l
.'*'• '
O
i-H
LL.
o -
*
Annual Runoff at Lees Ferry
C3-year moving average)
' •-". •• ..-•••- ; - /:•":-,.• - . • '• T-I\, ••- • .-,..-- • -.
+ 10% ^. l\ /A / \ _
• • - • ' — ^^ i \A ' \ / ^ ( / i
' \ / * ' A' ' \
' -N /VAV/ V / \
/^ /M !^fm A ' '
-Y-nrV / \,J/< / J \V/\ ;J '
..- V /x/^\ ,N-' \base -
^7 it- -10% ' ;
i ' ' ' ' ' ^B
Year
ngure is: Annual runoff (taf) at Lees Ferry in the base case and ±10% runoff scenarios.
Runoff is plotted as a three-year moving average.
-------
Table 16: Major reservoirs in the Colorado River Basin.
Reservoir
Blue Mesa
Fontenelle
Flaming Gorge
Navajo
Lake Powell
Lake Mead
Lake Mohave
Lake Havasu
Source: USDOI,
Notes: [1]Livei
Live
Capacity [1]
(taf)
830
345
3,724
1,642
24,454
27,019
1,810
619
1987; Weatherford, 1990:61.
capacity is the volume of wati
Annual
Evaporation {2]
(feet)
1.05
2.27
2.10
1.80
3.96
6.50
7.31
7.39
sr that can be withdrawn by
Bank
Storage
(%) [3]
—
3.30
8.00
6.50
gravity.
Power '
Generating
Capacity
(MW)
60
10
108
950
1,345
240
120
(2J Evaporation is calculated on a monthly basis by multiplying a monthly evaporation coefficient by the surface area
of the reservoir. The numbers given here represent the average of 12 monthly evaporation coefficients and are in
units of feet.
[3] Bank storage is calculated as a percent of monthly storage.
exists, actual storage throughout the entire Colorado River system is about 60 mat, or approximately four
times the average annual streamflow of the river at Lee Ferry. It is this storage capacity that is cushioning
annual changes in streamflow, particularly in the lower basin. The system's major reservoirs are summarized
in Table 16. While the upper and lower basins have nearly equal storage capacities, because the major
upper basin reservoir -Lake Powell- is located so far downstream, its releases primarily serve lower basin
water users. For this project, we elected to analyze changes in three reservoirs as well as in overall storage
changes in the upper and lower basins. The reservoirs selected include one upper reach reservoir, Flaming
Gorge; the major upper basin reservoir, Lake Powell; and the major lower basin reservoir, Lake Mead.
The effect of hypothetical changes in runoff on reservoir storage is shown in Tables 17 through
19. Reservoir storage is reported as storage on August 1, which corresponds to the end of the spring runoff
season and is roughly when peak storage occurs in the Colorado system. In the upper basin, decreases
in runoff of 5,10, and 20% generate respective decreases in mean storage on August 1 of 16,30, and 65%.
57
-------
c:
o
•rH
4-»
a
03
c_
CD
>
•4-1
ra
a
1 -
.75 -
.5 -
.25 -
0 -
8230 12000 18000
Annual Runoff [thousand acre-feet]
24000
Figure 14: Cumulative frequency of annual runoff at Lees Ferry for all scenarios. The plot
shows the frequency (y-axis) with which annual runoff is equal to or less than a given
volume (x-axis).
While less likely under scenarios of climate change, increases in runoff of 5,10, and 20% generate respective
increases in mean storage on August 1 of approximately 18, 25, and 30%. For Lake Mead, the major lower
basin reservoir, these figures are comparable (see Table 19). Decreases in natural runoff of 20% reduce
mean storage on August 1 in Lakes Powell and Mead to less than 25% and 15% of their respective
capacities. In both the -10% and the -20% scenarios, Lake Mead is completely drained in some years. For
both the upper and lower basins, a 20% increase in natural runoff generates completely full reservoirs.
A rough water-balance of the lower basin indicates that decreases in flow/storage are being
partially offset by decreases in evaporation and bank storage (i.e. water that is stored in the surrounding
soils). A 10% decrease in natural runoff causes average annual storage in the lower basin to decrease by
58
-------
Table 17: Storage (taf) in Flaming Gorge reservoir on August 1 for various scenarios.
Scenario
-20%
-10%
-5 %
Base
+5%
+10%
+20 %
Mean
Storage [1]
757 (-70.0 %)
1,689 (-33.0%)
2,085 (-17.3%)
2,522
2,963 (17.5%)
3,150 (24.9%)
3,306 (31.1%)
SD[2]
629
1,134
1,063
780
486
368
282
Minimum
Storage
77 (-92.7%)
97 (-90.8%)
142 (-86.5%)
1,055
1,946 (84.5%)
2,119 (100.9%)
2,348 (122.6%)
Maximum
Storage
2,640 (-27.2 %)
3,545 (-2.3 %)
3.544 (-2.3%)
3.627
3.627 (0)
3.627 (0)
3.627 (0)
Notes [1] Numbers in parentheses represent percent change compared to the base case.
[2] Standard deviation.
Table 18: Storage (taf) in Lake Powell
Scenario
-20%
-10%
-5%
Base
+5 %
+10%
+20 %
Mean
Storage [1]
5,915 (-62.9%)
11,260 (-29.4%)
13,434 (-15.8%)
15,949
18,790 (17.8%)
19,978 (25.3%)
20,873 (30.9 %)
on August 1
SD[2]
3,614
6.684
6,628
5,046
3,045
2,188
1 ,533
for various scenarios.
Minimum
Storage
1,904 (-73.8%)
2,627 (-63.8%)
2,736 (-62.3%)
7.265
12,145 (67.2%)
14,193 (95.4%)
1 6,1 37 Y* 22.7%;
Maximum
Storage
19,312 (- 13.3%)
21.326 (-4,3%)
21,800 (-2.1%)
22.277
22.509 (7.0%;
22,885 (2.7%)
22,970 (a 7 %;
Notes [1] Numbers in parentheses represent percent change compared to the base case.
[2] Standard deviation.
-------
Table 19: Storage, (taf) in Lake Mead on August 1 for various scenarios.
Scenario
• -20% ••
• • : -10% .
-5%
Base .
+5 % '
+10%
; ' , +20 %
Mean •
Storage [1]
3,674
8,071
10,545
12,366
14,166
17,211
19,808
(-70.3%)
,(-3,4.7%)
(-14.7%)
(14:6%)
(39.2%)
(60.2 %)
SD[2]
2,853
5,317
4,889
5,027
• 5,068
3,678
2,512
Minimum
Storage
0
0
2,888
5,975
7,688
9,258
10,597
(-100.0%)
(-100,0%)
(-51.7%)
(28.7%)
(54.9 %)
(77.4 %)
Maximum
Storage
8,385
19,687
21,891
22,170
22,426
22,716
23,623
(-59.5 %)
(-8.5%)
(-0.4 %)
(0.2%)
(1.8%)
(3.0 %)
Notes [1] Numbers in parentheses represent percent change compared to the base case.
[2] Standard deviation.
4348 taf (30%). In the absence of changes in evaporation and bank storage, runoff decreases of that
magnitude would be expected to cause substantially greater decreases in storage. In fact, decreases in
bank storage and evaporation of approximately 500 taf/year occur as a result of a 10% decrease in runoff
(See Appendix C, Table C5). Evaporation effects, however, are underestimated here because evaporation
rates will increase in a warmer climate. This would be reflected in higher evaporation coefficients for each
of the reservoirs and still greater decreases in water availability.
Figures 15 and 16 show plots of August 1 storage as a function of time. Most obviously these
plots reveal how the variability of the flow-input data affects storage. Also they suggest how lesser
quantities of runoff could result in extended shortages if we assume the same historical variability of runoff.
In the upper basin, the -5% scenarios causes storage to fall below 10 maf for a period of nearly 20 years.
In the -10% scenario, this period extends to 30 years. In the -20% scenario, nearly all years have less than
10 maf of storage. In the lower basin, a 10% decrease in runoff causes lower basin storage to fall below
6 maf for a period of 20 years. The exceedingly high flows that follow this period (corresponding to the
60
-------
historical period of the 1920s), however, allow the reservoirs to recover quickly and to reach near maximum
capacity. Only in the -20% scenario do reservoirs fail to recover to functional levels. The -5% scenario
takes storage in the lower basin to new low levels, although reservoirs recover to median levels within a
few years. The -10% scenario causes extended periods of very low storage, and recovery takes 15 to 20
years. In the -20% scenario, reservoirs are unable to recover to average levels over the modeled period.
In fact, the -20% scenario causes Lake Mead to run completely dry roughly 25% of the time.
These figures also show the impact of reservoir sedimentation over time. While not specifically
related to climatic change, sedimentation is likely to have an impact on system operations over the next
several decades. For the largest reservoirs, the CRSS calculates loss of storage capacity due to
sedimentation as an absolute amount per month. Over a 78-year run, Lake Powell loses 4760 taf or nearly
20% of its capacity, and Lake Mead loses 3000 taf or 11% of capacity. This represents a significant loss of
storage capacity that needs to be considered when assessing the system's future effectiveness.
More interesting than average changes in storage is how frequently critical storage levels are
reached under various scenarios. For instance, in the base case, Lake Powell never falls below minimum
power pool (the minimum volume necessary to generate hydroelectricity). Cumulative frequency diagrams
for Lakes Powell and Mead are presented in Appendix C (Figures C6 and 07) and are summarized here.
The -5% scenario causes Powell to fall below its minimum power pool (4.1 maf) roughly 20% of the time;
this frequency increases to nearly 60% under the -20% scenario. Similarly, in the base case, the frequency
with which Lake Powell contains two or more years worth of storage (roughly 16.5 maf) is just under 50%.
This frequency rises to 70% under the +5% scenario, and to 90% under the +20% scenario. Lake Mead
has an active storage capacity of roughly 26 maf and a minimum power pool of 10 maf. When storage falls
below the minimum power pool, deliveries to downstream users are reduced to their minimum allowable
levels. Even in the base case, monthly storage falls below minimum power pool 50% of the time. And with
a 5% increase in flow, releases are still reduced to their minimum level in 30% of the months. A decrease
61
-------
Storage on August 1
30 1 -20%
20 -
CD
O
tO
C
O
03
cn
£
0
-M
cn
-10%
-5%
+5%
T 1 1 r
-i +10%
78
1 ' ' 78
30-
20 -
10-
0 -
+20%
(dashed line represents median-
storage on August 1 -for the
base case.)
7B
Yean
Figure 15: Upper basin storage on August 1 plotted as a function of year, for all scenarios.
in runoff of 20% leaves the reservoir essentially empty in about 30% of the years, while the minimum
storage level required for power generation is never attained.
Depletions and Deliveries
Consumptive water use in the basin is reported in terms of depletions and deliveries to major
users. Reservoir evaporation Is modeled explicitly by the CRSS and is not considered a depletion.
Scheduled depletions are those shown in Table 11. In addition, for some users, deliveries are constrained
so that they never fall below a minimum level. In this study, the minimum deliveries for the Central Arizona
Project (CAP) and the Metropolitan Water District of Southern California (MWD) were 451 taf and 500 taf,
62
-------
Storage on August 1
301 -20%
20 -
\A^^vvV^ V
-------
O)
03
01
O
CO
•a
(D
en
o
ui
o
•H
a.
Q)
a
r-l
CD
C
c
•<
14DOO-,
10500-
7000
3500
14000
10500
7000
3500-
UPPER BASIN
-10X base +10X
14000
10500
3500-
LOWER BASIN
-10% base +iox
-20* -5X +5X +20X
-20X
-5X +5X +20X
H minimum
E3 mean
g maximum
(dashed line represents demand
in each region)
-20X -5X +5% +20X
scenario
Figure 17: Minimum, mean, and maximum annual depletions in the upper basin lower
basin, and Mexico for all scenarios.
Table 20: Percent frequency with which CRSS scheduled deliveries
to MWD, CAP, and Mexico are met or exceeded.
Runoff ~~~~
Scenario MWD CAP Mexico
CRSS Scheduled
delivery (taf) 500 1467 1515
-20%
-10%
-5%
Base
+5%
+10%
+20%
100
100
100
100
100
100
100
0
28
35
59
77
95
97
64
94
100
100
100
100
100
-------
Hvdroelectricity Production
Hydroelectricity production, like reservoir storage, is extremely sensitive to changes in runoff.
Changes in power production are more sensitive to the historical sequence than the other variables analyzed
in this study. Although differences among sequences in the base case are insignificant, in the -20%
scenario, different sequences generate as much as a 10% difference in results 0"able C6).
In the upper basin, power production does not decline as rapidly as storage on an average
annual basis. The -10% scenario causes average annual storage to decrease by 30% while power
production decreases by 26% (Figure 18). In the -20% scenario, power production drops by 49% compared
to a decline in storage of 63%. Storage increases, however, tend to exceed power, increases on a
percentage basis. In the +5% scenario, overall power generation jumps by 1 thousand gigawatt-hours
(GWh) per year, or 11%, while storage increases by 14%. In the +10% scenario, power generation
increases by 21%, compared to an increase in storage of 28%.
In the lower basin, power production reductions are on par with, or slightly greater than,
reductions in storage, largely because Lake Mead has a relatively high minimum power pool (10 maf). Even
though the CRSS reduces deliveries to minimum levels in order to maintain some power-generating capacity,
the magnitude of runoff decreases modeled in this study still reduce power production in the lower basin
substantially. Although the -10% scenario causes a 12% reduction in runoff at Lees Ferry and a 30% decline
in lower basin storage, it causes a 36% decline in lower basin power production. Similarly, the -20%
scenario causes a 50% decline in lower basin storage and a 65% decline in power production.
Uncontrolled Spills
In this study, no uncontrolled spills occurred in the lower basin except in the +20% scenario,
in which spills occur in 2 out of the 78 years. The total volume of spills for these years is 1.5 maf and 8 maf.
For the upper basin, the base-case scenario generates uncontrolled spills in 7 years out of a total of 78 (9%),
with the maximum volume of spills in any one year equal to 1.5 maf (Figure 19). When natural runoff is
67
-------
increased by 5%, uncontrolled spills occur in 11 years, with a maximum annual volume of i.7 maf. A 10%
increase in natural runoff results in 16 years that experience uncontrolled spills, with a maximum annual
volume of 3 rriaf. In the +20% scenario, uncontrolled spills are occurring in more than one-third of the years
(33). In 8 of these years, spills exceed 1.5 maf; and in 4 of these years, spills exceed 3 maf. The maximum
annual volume spills in this scenario is 4.5 maf. Even though spills occur under scenarios of increased flow,
the existing flood control criteria for the reservoirs, which require that 5.35 maf of storage space be available
in Lake Mead or upper basin reservoirs on January 1, are never violated.
Salinity
Salinity is the only water-quality parameter estimated by the CRSS model. It is defined as total
dissolved solids (IDS) and reported in units of milligrams/liter (mg/l). The model assumes uniform salinity
In reservoirs, but does take into account the effects of evaporation. Existing, but not projected, salinity
control projects are incorporated into the model.
Even in the base-case scenario salinity criteria are consistently exceeded at all points (Figure 20).20
In the base case, salinity concentrations are within the criteria at all points in less than 15 years. Decreases
in runoff of only 5% cause essentially all years to exceed the criteria. Moreover, even in the +20% scenario,
salinity criteria are exceeded continuously for long periods, roughly 20 years. Differences in absolute salinity
between stations increase as runoff decreases. For example, in the base-case scenario, salinity below Davis
Dam measures 858 mg/l on an average annual basis, increasing to 1019 mg/l at Imperial Dam, a difference
of 161 mg/l. In the -20% scenario, this difference increases to 208 mg/l, with salinity values at Davis and
Imperial reaching 1010 mg/l and 1218 mg/l respectively.
20 Numeric criteria for salinity on an annual, flow-weighted basis were established in 1972 for three
locations along the River:
(1) Below Hoover Dam: 723 mg/l
(2) Below Parker Dam: 747 mg/l
(3) At Imperial Dam: 879 mg/l
In addition, Minute 242 establishes a relative standard for water delivered to Mexico, which is not to exceed
the salinity level measured at Imperial Dam by more than 130 ±30 mg/l.
68
-------
15000 -i
[P 10000
cu
QJ
CD
c_
CD
O 5000
Q.
(O
UPPER BASIN
+ 10%
base
-10%
I
-20% -5%
+5%
+20%
15000 -i
10000-
5000-
LOWER BASIN
-20%
-5%
I minimum i mean | maximum
+5%
+20%
Figure 18: Minimum, mean, and maximum hydropower generation (annual) in the upper and
lower basins for all scenarios.
40 -
to
3 30 -
fe &
•S. 3000 taf
H Volume >1500 taf IHI
H Volume >50
F— ^
-20%
/-•• - -
-10%
-
-5%
^^m
taf U v ->.|
-^ ,^ ~ ^
^
base
v
.
+5%
>/ ^ ^' ^
', f * ^*"
>
*
^" 1
^
+10% +20%
Figure 19: Frequency and approximate annual volume of uncontrolled spills which occur in the
upper basin during a simulation run of 78 years.
-------
Davis Dam
1500 -
OI
£ 1200 -
CD
cn
TJ
09
-M
f
f
o
723 -
400 -
standard
Year
i
78
2000 -
1500 -
to
cn
1000
879-
o
I-H
U.
500 -
Imperial Dam
w.q. criteria
Yean
—r~
78
Figure 20: Salinity as a function of year at Davis and Imperial Dams. The base case and the
±20% runoff scenarios are shown. Water-quality criteria are continually exceeded in all but the
+20% scenario.
-------
Time-Shifted Scenario
In addition to quantitative changes in runoff inputs, we also ran one time-shifted scenario to study
the effects of shifts in the timing (seasonality), but not quantity, of runoff. The results obtained from the
NWSRFS hydrologic model suggest that increases in temperature of 2" C would shift peak runoff to the
month of May rather than June in the upper basin. In general, the quantity of monthly runoff appears to shift
backward by one month (refer to Figure 7). To simulate this effect in the CRSS model, natural-flow inputs
were shifted backwards by one month so that June runoff was input as May runoff, July runoff as June,
January runoff in yearn as December runoff in year n-1, and so on.
The results of this scenario suggest that, even though the natural-flow inputs do not change on an
annual basis, overall system operations are less efficient. Average annual runoff increases marginally at
several points (Table 21) as do upper basin depletions, reflecting the fact that less water is being held in
storage. Mean decreases in August 1 storage range from -11% in the upper basin and -4% in the lower.
The effect of the time-shifted scenario on the cumulative frequency of storage is shown in Figure 21. In the
base case, storage in the upper basin falls below 15 maf only with a 15% frequency; this rises to more than
30% under the time-shifted scenario. Deliveries in the lower basin also suffer somewhat, with average annual
deliveries to CAP and MWD declining by 89 taf (6%) and 21 taf (3%), respectively. Scheduled deliveries to
CAP are met with slightly less frequency, 60% versus'65% of the years. These changes are most likely due
to the model forecasting procedure, which establishes target storages based on reservoir contents and time
of year. Thus, because higher streamflows occur earlier in the year, more water is being released in the
winter and early spring. Subsequent months, however, have lower-than-predicted runoff, causing most
deliveries to decline on an annual basis.
The time-shifted scenario also causes a slight increase in average annual salinity. Flow-weighted
salinity at Davis, Parker, and Imperial dams increases by about 20 mg/l (2.5%), which is comparable to the
increase of roughly 25 to 35 mg/l (3%) for the -5% scenario. The fact.that salinity increases by 2.5% is
71
-------
Table 21: Annual runoff (taf) at various points for the base case and the time-shifted
scenario.
Mean
Station Base TS[1]
Green 977 997
Cisco 4,522 4.712
Bluff 1,356 1,348
Lees Ferry 9,393 9,557
Imperial 6.098 6,053
SD Minimum Maximum
Base TS Base TS Base TS
404 381 252 281 1,964 1,984
1,678 1,547 1,193 1,247 8,413 8,244
694 727 361 294 3,280 3,480
2,089 2,193 8,238 8,239 16,884 16,889
1,161 1,112 5,650 5.650 11,597 11,241
Note: J11 Time-shifted scenario, in which all historic input data is shifted backwards by one month, e.g.. January of Year "n" is
input as December of Year "n-1 * in order to reflect the shift in seasonality expected as a result of increased temperatures.
o
u_
0)
«o
1 -
.75-
.5-
.25-
0 -
Time-shifted scenario
(938 observations)
5.0 10.0 15.0 20.0 25.0
Storage [million acre-feet]
30.0
Figure 21: impact of the time-shifted scenario on storage in the upper basin. This graph
shows the frequency (y-axis) with which monthly storage is equal to or less than a particular
volume (x-axis).
-------
interesting because streamflow at Imperial decreases by <1% in the time-shifted scenario. The frequency
with which the salinity criterion at Imperial Dam is exceeded increases marginally, from 75% to 78%.
Summary and Discussion of Water-Supply Modeling Results
To date, few studies have attempted to model the impacts of climatic changes on regional water-
supply systems. This reflects both the lack of suitable models and the paucity of regional information on
climate-induced changes in runoff. Two exceptions are the studies of the State Water Project in California
and the Tennessee Valley Authority, both done as part of the US EPA study of climate impacts (USEPA,
1990; Lettenmaier and Sheer, 1991). In these studies, a limited number of GCM scenarios were analyzed
using large-scale water-supply models. In both cases, water-supply systems were found to be sensitive to
GCM-derived scenarios of climatic change. One of the conclusions of the California study was that changes
in operating rules might improve the ability of the system to meet delivery requirements, but only at the
expense of an increased risk of flooding. Both studies noted that climatic changes are likely to increase the
tension between flood control and water supply and/or hydroelectricity production.
Our results from the CRSS model similarly suggest that the water-supply system of the Colorado
River Basin is sensitive to changes in runoff that might be plausibly associated with climatic change, and
that some tradeoffs will be necessary to balance multiple purposes. Looking back at the hydrologic
modeling discussed in Part I of this report, we can relate climate scenarios to the changes in the water
supply variables given in Table 22. Overall, the GCM scenarios suggest decreases in runoff on the order
of 10 to 20%. A 20% reduction in runoff would cause reductions in mean (August 1) storage of 60 to 70%,
reductions in mean annual power generation of 60%, and an increase in mean annual salinity of 15 to 20%.
In contrast, should the region experience only a moderate increase in temperature (2° C) and a large increase
in precipitation (20%), this would result in roughly a 20% increase in runoff, a 30 to 60% increase in mean
storage, a 40% increase in power production, and a 13-15% decrease in salinity. On the other hand, a
73
-------
temperature increase of 4" C coupled with a precipitation decrease of 20% would result in approximately a
30% decrease in runoff, which is more extreme than any of the scenarios modeled with the CRSS.
These CRSS results suggest that Compact violations are likely to odcur under all scenarios of
decreased runoff. This primarily reflects current operating parameters. The CRSS does not impose
shortages on the upper basin but passes them on to the lower basin. Under the terms of the Colorado River
Compact, however, the lower basin could theoretically require the upper basin to curtail usage in order to
meet the Compact requirements during a period of severe drought (Hundley, 1975; Getches, 1991). Thus,
the delivery and depletion results presented here reflect a potentially unlikely scenario in which the lower
basin bears the brunt of any shortage, without resorting to a Compact call. For instance, although CAP
deliveries should be fairly secure under all but the -20% scenario assuming that a Compact call is enforced,
In these simulations some reductions to CAP are occurring even in the base case, probably in order to
maintain the minimum power pool in Lake Mead. This can be seen in Figure 22, which shows that CAP
deliveries fall from their scheduled level of 1467 taf to the minimum level of 451 taf as storage in Lake Mead
declines to 10 maf, which is equivalent to minimum power pool. Under the operating regime modeled here,
CAP would not receive their full allocation in the future without persistent increases in annual runoff.
The reservoir simulation results presented here suggest that many.of the procedures and inputs used
in the model are closely tuned to historic hydrology. For instance, storage strategies and targets work
extremely well in the base case scenario but are substantially less effective under alternative scenarios.
Thus, Compact violations would potentially occur even in the -5% scenario, even though this could most
likely be avoided if the CRSS operating parameters were altered.
74
-------
If operating parameters were altered, the result would be a very different allocation of shortages.
According to Getches (1991:22), the upper basin has present perfected rights21 to only about 2 maf and
in cases of severe shortage, the upper basin could be required to reduce its Usage to that amount so that
the remaining water could flow to the lower basin and Mexico. In the model runs presented here, however,
upper basin consumption never falls below 2.8 maf even though substantial shortages are occurring in the
lower basin. In the -20% scenario, overall shortages to the upper basin are only about 5%; but lower basin
depletions fall by 15%, which represents a 0.9 maf shortfall of lower basin entitlements on an average annual
basis. But under the existing legal framework, the lower basin and Mexico would not be legally forced to
endure shortages until the total water available in the basin for consumptive use fell below 11 maf.22 This
occurs in one year out of 78 in the -10% scenario, and 3 years out of 78 in the -20% scenario. Thus, if the
CRSS modeled Compact calls, the lower basin would rarely suffer shortages. The upper basin, on the other
hand, would suffer much more extreme shortages than those suggested by the modeling runs presented
here. In the base case, upper basin depletions would be limited to 3 maf or less roughly one-third of the
time, a cutback of more than 20% over present levels. In the -5% scenario, this percentage of years in
which consumption would be 3 maf or less rises to 61%. In the -20% case, the upper basin would never
receive more than 3 maf, and would receive only 2 maf in about 10% of the years. Of course, these
frequencies are dependent upon when and how quickly reservoir storage is consumed.
The variables most sensitive to changes in natural runoff are reservoir storage and power generation,
which are particularly sensitive to decreases in runoff (Table 22). For example, changes in mean storage
in Lake Mead on August 1 are on the order of -70% (-20% scenario) to +60% (+20% scenario). It is difficult
to say much about the risks of flooding to the basin based on these scenarios. Unlike water-supply
"Present perfected rights" refer to those water rights that were already established by upper basin
users at the time the Colorado River Compact was signed, in 1 922. These rights are not subject to compact
calls. See Getches (1991).
is includes 2 maf for the Upper Basin, 7.5 maf for the Lower Basin, and 1.5 maf for Mexico. See
Getches (1991).
75
-------
Table 22: Sensitivity of water-supply variables to changes in natural flow in the Colorado
River Basin [1].
Change in
Natural
Row
-20
-10
.5
5
10
20
Change in
Actual
Row [2]
(10-30)
(7-15)
(4-7)
5-7
11-16
30
Change
in
Storage [3]
(61)
(30)
(14)
14
28
38
Change in
Power
Generation [4]
(57)
(31)
(15)
11
21
39
Change
in
Depletions [5]
(11)
(6)
(3)
3
5
8
Change
in
Salinity [6]
15-20
6-7
3
(3)
(6-7)
(13-15)
Notes: (1 ] Average change compared to the base case over a 78-year simulation run. Numbers in parentheses represent
DECREASES.
[2] Changes in flow represent the range of changes at five points: Green River, Cisco, Bluff, Lee Ferry, and Imperial
Dam.
[3] Mean storage throughout the basin on August 1.
J4J Mean annual power generation throughout the basin.
[5] Depletions are summarized over the entire basin, although depletions are defined differently in the upper and low
basins. See Hundley (1975) for details.
[6] Changes in salinity represent the range of changes at three points: Davis, Parker, and Imperial Dams.
20 -
0)1-1
rag
°T
C/J ^
CJ
CO 4 n
iH
in""*
_j E=
o -
base case A i\
Ml* "
M > 1 \ / \
1 v \ ' '
/ 1 1 \
1 \ 1 \
Lake Mead / v »
\ i \
Ni.1
1 \
' \
•\ / \
\ i \
« * \ i i
• V \ 1 \
/ \ J_ J
--; \^-^/\/ (inn n \
iVviil \
/ i n / i
/ \ AAA / If — ' CAP
_/ Uvuu V
~j i i 1 1— '
i 78
Year
Q.
° 4J
s s
0) f
Q) V
fe ro
^
-2177 51
a in
t-t 0
(D j;
-1464 g§
F*
-450
Figure 22: Relationship between storage in Lake Mead and annual delvieries to CAP. In the
base case (and the +5%) scenario, CAP deliveries fall to their minimum level (450 taf) when
Lake Mead falls below minimum power pool (approximately 10 maf).
-------
shortages, which are primarily a function of average annual flows, floods are a function of the duration and
severity of particular snowmelt and precipitation events. While climatic changes will in fact alter snowmelt
and precipitation patterns, these effects cannot be adequately evaluated without more detailed regional
information. In general, the loss of snowpack storage associated with global warming is likely to increase
reservoir spills in some seasons. The Hoover Dam flood control regulations call for releases from Lake
Mead not to exceed 28,000 cfs, in order to avoid damage in the flood plain below Parker Dam (USACOE.
1982). Any uncontrolled spill in the lower basin is cause for concern. Depending on their magnitude and
duration, uncontrolled spills in the upper basin may be an indication of high, but not necessarily
uncontrolled, runoff in the lower basin that may nonetheless be damaging. The volume of upper basin spills
in the +10% scenario (up to 3 maf annually) suggests that flood control would be an issue. More generally,
the sensitivity of storage to changes in runoff illustrates how carefully the current system is operated and
how little is the room for forecast error if water-supply is to be maximized without resulting in damaging
flood-control releases or uncontrolled spills.
The range of basin storage over which the level of power generation shows little variation is very
wide, from about 5 maf to 23 maf. This insensitivity of power production to reservoir levels indicates that
power plant releases are not being adjusted to reflect water-storage levels in the basin. In other words,
power is generated at a relatively constant level until critical (i.e. minimum power pool) reservoir levels are
reached, and then no power is generated. In the -10% and -20% (runoff) scenarios, minimum power pool
is frequently breached (e.g. with respective frequencies of 75% and 100% in lake Mead), and so dramatic
declines occur in hydroelectric output. An alternative, and possibly more efficient, operating strategy might
make power generation more sensitive to reservoir levels, so that lower levels of power were produced over
longer periods of time.
Not surprisingly, the most critical concern for the lower basin is water quality/salinity. Under almost
no circumstances can existing water-quality criteria be met given projected demands and operating
77
-------
constraints. Our results suggest that at least a 20% increase in natural runoff would be necessary to bring
the salinity levels in the lower basin into compliance with existing numeric criteria. Although the scenarios
considered here result in only moderate changes in salinity, the problem is already so severe in the base
case that even moderate declines in water quality are of particular concern.
Increases in salinity are disproportionate to decreases in runoff. A (modeled) runoff decrease of 11 %
at Imperial Dam brings a average annual salinity increase of nearly 20% (200 mg/l), while a runoff increase
of 11% at the same location results in only a 10% (71 mg/l) decrease in salinity. In addition, annual
maximum salinity concentrations increase dramatically. For instance, at Imperial Dam maximum annual
salinity concentrations rise from 1279 mg/l in the base case to 1516 mg/l in the -10% scenario and to 1848
mg/l In the -20% scenario. This represents percentage increases of 19% and 44%, respectively. Although
the model's accuracy with respect to salinity calculations may be questioned, the phenomenon of non-
linearity has been established both empirically and theoretically (Vaux, 1991). These complications imply
the need to continue to develop and to improve water-supply models such as the CRSS so that the multi-
faceted impacts of changes in runoff can be adequately assessed. The results also suggest the importance
of flow-Independent sources of salinity in downstream reaches because differences in absolute salinity
between stations increase as runoff decreases.
Overall, the water-supply modeling results illustrate how carefully the system must be managed in
order to meet the multiple needs of the basin. Of course, this is not a surprising result; it reflects the
historical over-allocation of supply as well as rapidly growing demands.
The water-supply results are unquestionably sensitive to the volume of demands used in the model.
In reality, the numbers used for these runs represent probable supplies rather than actual demands. For
Instance, MWD's demand in the model is set to 500 taf, although MWD regularly takes and uses significantly
greater quantities of Colorado River water (Getches, 1991:18). On the other hand, upper basin demand
78
-------
numbers embody several assumptions about growth and development that have been contested. Upper
basin demands in these runs are nearly 35% greater than current demands. Were these demand projections
altered, they would have substantial impacts on the operation and results of the CRSS model.
Finally, although changes in mean natural runoff of 20% may seem extreme, in fact, changes of this
magnitude over a limited period of time are conceivable even without the advent of enhanced greenhouse
warming. A 20% decrease in natural runoff would lower the annual mean at Lee Ferry from 15 maf to 12
maf. Tree-ring reconstructions suggest that over the last 500 years the lowest 80-year mean is less than 11
maf, which corresponds to a 27% decrease in natural flow. If climatic changes were coupled with such
extreme, non-greenhouse variations, the impacts on the basin would be more severe than even the most
extreme scenarios presented here.
79
-------
STUDY CONCLUSIONS
To date, most hydrologic studies have been limited to analyzing changes in runoff and soil moisture.
These are important parameters to study, but they tell us only a limited amount about how water-supply
systems may be challenged under conditions of climatic change. In order to assess the ability of the
political and water-management infrastructure to distribute water in an equitable and efficient manner under
a greenhouse-affected climate, we need information on the spatial distribution of water. This is the type of
information provided by reservoir-simulation models such as the GRSS. ,. .
The results of this study suggest that the Colorado River Basin would be very vulnerable to potential
climatic changes. Certainly a temperature increase in the range of 2-4° C is well within the range of
plausibility.23 Without any change in precipitation, these temperature changes alone imply decreases in
runoff of 5 to 10 percent. This would result in average annual declines in mean annual reservoir storage and
power generation of 30% to 60%. Average annual depletions would decrease by 3% to 6%, and Compact
calls could potentially occur in several years. Moreover, decreases in runoff would exacerbate an already
severe salinity problem in the lower basin. Should precipitation increase, some or all of these impacts might
be offset; but should precipitation decrease, the impacts may exceed even those presented here. It should
be borne in mind that these results reflect runoff changes of 5 to 20% imposed on the hydrology of the last
80 years. The results would be different if a different hydrologic record had been used. For instance, the
hydrology of the last 400 years suggests that much more severe and sustained droughts have occurred in
the past (Stockton, et al., 1991). If this hydrology were used as a basis for a similar study, decreases in
runoff would have still greater impacts on the Colorado River Basin.
In this study, the current operating system fails to manage adequately long-term decreases in natural
runoff of 20%. Lesser changes challenge the system; however, they do not overwhelm it. Yet over the long-
23
GCM predictions for this region suggest greater increases in temperature, from 4° to 7°C on an
average annual basis.
80
-------
term, the system appears to operate more comfortably under a slightly increased runoff regime (+5%),
although it could probably still operate more efficiently. This reflects the fact that the system is likely to be
over-allocated if all presently scheduled demands come on line in the next 50 years. On the other hand,
relatively moderate decreases in streamflow (i.e. -5%) would pose considerable challenges to the basin.
Given the assumptions that bound this study, it appears likely that any long-term decrease in streamflow
would bring extended periods of drought and water-supply shortages.
Although we were not able to assess the impact of changes in operations as part of this study, our
results suggest that the system would almost certainly benefit from alterations in the operating regime should
the magnitude or persistence of streamflow change. The current operations are, in some sense, an artifact
of historic experience. Management assumptions and the perception of risk are conditioned by recent
hydrologic experience in the basin. An example of this is discussed by Dracup et al. (1985:239) in
connection with the flooding experienced in the lower basin during 1983:
The period of time that the Colorado reservoir system was filling constituted a periqd during which
true exposure to climatic impacts, i.e. precipitation variability, did not exist.... The encroachment into
the flood plain was possible because water was in storage upstream, and also because the period
of filling Lake Powell was drawn out for almost two decades. Two decades are more than sufficient
to affect societal perceptions of climate stability.
Water managers have traditionally relied upon historical hydrologic records and past experience in
order to plan, inferring the probability of future shortages and floods from their frequency of occurrence in
the past. If the existing record on the Colorado River is examined, however, it shows little ability to predict
future conditions. The classic example of this is provided by the 20-year period immediately preceding the
adoption of the Colorado River Compact in 1922. During this period, average annual flows at Lee Ferry were
estimated to be 16.4 maf/year, of which the Compact intended to allocate 15 maf/year (Hundley, 1975).
No period of similar duration and high flows has occurred since then, and the average runoff at Lee Ferry
from 1906 to 1990 has been only about 15 maf/year. Tree-ring analyses suggest that the long-term average
runoff may be as low as 13.5 maf/year and that the most critical period on record may have had a 20-year
81
-------
average runoff of only 11 maf (Stockton and Jacoby, 1976). While this is an extreme example, it nonetheless
illustrates the problem of relying exclusively on the recent instrumental record as a basis for planning, and
suggests that any attempt to model future water supply will be hindered by such a reliance on historic data.
Ultimately the problem is our ignorance of the underlying distribution that governs streamflow.
Current operating procedures, although somewhat flexible, are strongly keyed to the existing historical
record. When viewed from the perspective of climatic change, this becomes a concern. Although the
existing record is now nearly 80 years in length, this is not a long record given the high variability of
streamflow in the basin, our poor understanding of streamflow distributions, and the likelihood of future shifts
In underlying climatic variables. The ability of a system to perform adequately in the past is at best a weak
Indicator of its potential to perform in the future. While certainly a system must be able to address historic
variations and extremes to be effective over the long term, it must be able to address even greater variations
that might reasonably be anticipated in the future. Scenarios derived from GCMs are useful in this respect
because they provide additional information on changes in streamflow that might accompany climatic
changes. Most of the GCM temperature and precipitation scenarios modeled as part of this study suggest
that runoff will decrease even though precipitation may increase, with the magnitude of decrease ranging
from 8% to 24%. The problem of planning water management in the face of a high degree of climate and
hydrological uncertainty cannot be easily resolved; nonetheless, it may be possible to increase flexibility in
water management. This flexibility will need to be reflected in technical and operational decisions, as well
as In the legal and economic institutions that govern water use in the basin.
The problem of planning is compounded by the fact that we cannot say with certainty whether runoff
in the basin will increase or decrease. Most people with an interest in the basin have focused on the
prospect of long-term decreases in runoff and the shortages that would result, which is a logical reflection
of the region's preoccupation with drought. The fact that average temperatures in the region will almost
certainly increase suggests that, if we assume no knowledge about changes in precipitation, we would
82
-------
expect runoff to decrease as a result of increases in evaporation and vegetative water use. This may be
reason enough to plan for supply shortages; but increased water storage must be traded off against the
need for flood-control space. The greatest risk of climatic change is the potential for streamflow variability
to increase substantially, increasing the frequency of both sustained drought events and high-flow events,
and thus complicating management.
In addition to the uncertainty in future hydrology posed by climatic changes, any change in
hydrology may pose additional policy challenges for the region. As hydrology changes, it may well become
more difficult to reconcile the claims of different users and multiple purposes along the river. Institutional
and operational regimes will have to respond to tensions between the upper and lower basins, between
demands for hydroelectricity and water supply, and between water supply and flood control.
Inevitably the discussion of climate-change and water resources leads to the question of storage,
specifically whether increased storage is a reasonable response to climate-induced changes in water supply.
Reservoirs are frequently viewed as a response to supply shortages; however, given the already high levels
of storage available on the Colorado; additional reservoir capacity would do little or nothing to alleviate
potential reductions in flow. Reservoirs serve solely to decrease seasonal and inter-annual variability (over
a limited number of years); they do not increase the volume of water available on a long-term basis. In fact,
additional reservoirs in highly developed regions may actually decrease water supply over the long-term
through evaporative and bank-storage losses (Klemes, 1985; Langbein, 1959). Only if climatic changes were
to increase streamflow variability, without decreasing long-term supply, would additional reservoirs in the
Upper Colorado River Basin have any benefits. The question of change in variability has not been
addressed in this study.
In addition, the development of water resources may inadvertently reduce flexibility in some cases.
For example, the decrease in the interannual variability of streamflow in the Colorado River has been
83
-------
accompanied by an increase in both usage and dependence, and thus the long-term vulnerability of the
region to climatic changes has increased. The low variability of water supply in the lower basin has
encouraged the total use of available resources, thus removing any real drought "cushion". While this
generates economic benefits, it also increases the economic costs of a severe and sustained drought once
storage has been exhausted. Similarly, the perceived invulnerability of flood plains has encouraged
additional development that subsequently reduces operating flexibility (USACOE, 1982; Dracup, etal., 1985).
On the Colorado River, ample flood-control storage exists, but as others have pointed out, the basin's
concern with drought and water storage has resulted in a series of operating rules and customs that
maintain reservoirs nearly full, leaving little room for forecast error or for managing extremely high flows
without damage. The range in which flood control and water supply are balanced is very narrow as the
system is currently operated. This is an issue that would almost certainly be exacerbated by climatic
changes.
Beyond the scope of this study were several important issues that policymakers and water-supply
managers will undoubtedly have to consider. First, the environmental and ecological impacts of changes
In water supply have not been addressed in this study. Part of the problem lies in the lack of information.
In general ecosystems are more sensitive to seasonal, monthly, daily, and even hourly changes in
streamflow and water quality than to long-term changes. Unlike water supply, the impacts on the
environment cannot be adequately assessed using aggregated time periods or large-scale models.
Undoubtedly, however, given the predicted rate of climatic change and the potential magnitude of runoff
changes examined here, serious environmental concerns would be raised.
This study has also not taken projected future developments nor some future demands into account.
Currently the issue of reserved water rights and Native American claims have obscured future demand
scenarios in the basin. Because of the large amounts of water involved, these unresolved claims could have
84
-------
dramatic impacts on water allocation throughout the region and thus add to the uncertainty that the basin
faces.
Finally, while this study has suggested what the impacts of climate change could be on water
supply, it has not addressed the impacts of climate change on water demand. In fact, demands will change
both in time and space. Obviously, agricultural water demand will vary as crops and production patterns
are altered in response to climatic changes. Ecosystem water requirements will also vary, both in response
to increased temperature and as a result of ecological and environmental changes. Urban and industrial
usage will change as a result of both changes in climate and changes in population. In fact, it is quite
possible that changes in demand over the next 50 to 100 years will equal or exceed changes in supply. In
all likelihood, the greatest possibilities for adapting to climatic change lie in the area of demand
management, particularly in the agricultural and urban sectors, and the potential for conservation and water
transfers needs to be assessed from both a quantitative and an institutional perspective. If we are to plan
adaptation strategies, future research must address the integrated impacts of climatic change on demand
and supply across sectors.
Given the uncertainty surrounding potential climatic changes and the problems encountered in trying
to model impacts, care must be taken to view the results presented here in their appropriate context. While
some analysts and planners, when faced with large uncertainties, may prefer to refrain from any attempt to
assess the impact of climatic change on water resources, we believe that it is preferable to see how far one
can get using current information and models even though they might seem inadequate to the task. The
greatest danger, however, is that the numbers will be accepted uncritically or as predictions when, in fact,
they are bounded by considerable uncertainty. Nevertheless, numbers may help us to represent and to
comprehend the sensitivity of the basin to plausible scenarios of climatic change. In particular, the scenarios
of changes in temperature and precipitation derived from GCMs provide the best information currently
available on climatic change. When translated into changes in runoff and water supply, as in this study,
85
-------
these climate scenarios suggest that past assumptions about water-supply reliability may be severely
challenged in the coming decades. By suggesting plausible future scenarios, we may find the impetus to
consider what changes we can make to balance multiple purposes under varying conditions of climate.
Given the prospect of future climatic changes, it is imperative that we consider how we can increase the
resiliency of our existing water-management systems and minimize the social and environmental impacts
of changes in water availability. We need to identify those responses that will provide us withthe greatest
flexibility in the coming decades and to develop management schemes that recognize both the variability
and the dynamic nature of climate.
Future Work
This study has identified the overall sensitivity of the region as well as the rough magnitude of
potential impacts. It has suggested concerns about basin-wide planning mechanisms, potential future
conflicts, and the risks of increased variability. The results generated by the three sub-basin models suggest
that additional modeling of the Upper Colorado River Basin would be useful. An important step would be
to assess the region on a sub-basin-by-sub-basin basis in order to identify and categorize the response of
Individual watersheds. This would provide a more accurate picture of how the overall basin would respond
to climatic changes. Moreover, smaller-scale studies would enable researchers to evaluate the relative
sensitivities of supply .and demand within sub-basins in order to identify critical regions and to focus
adaptation strategies on a sub-regional basis. Potentially better generalizations could be made if the
hydrologic modeling incorporated larger spatial coverage of the basin and additional climate scenarios.
Additional modeling may also allow for more detailed validation of the NWSRFS (or other appropriate) model
and would lend greater confidence to the results presented here.
The results of the reservoir-simulation modeling also suggest numerous opportunities for additional
research. This study was limited to modeling only hypothetical scenarios of changes in natural runoff that
were applied uniformly across the basin. First, this modeling could be extended by disaggregating
86
-------
streamflow scenarios both temporally and spatially using statistical techniques. Potentially, this type of study
could provide a more accurate picture of how the basin would respond to climatic change, and also when
and where critical situations are likely to occur. Secondly, additional and more complex scenarios of
changes in runoff could be developed if additional hydrologic modeling of the Upper Colorado River Basin
were undertaken. Thirdly, operational flexibility could be explored in detail with a modified, and potentially
simpler, version of the CRSS in which operating parameters and assumptions could be more easily adjusted.
This would allow a quantitative assessment of the model's sensitivity to operating assumptions as well as
a more policy-oriented study of operational flexibility and opportunities for improved water management.
87
-------
REFERENCES
Anderson, E.A., National Weather Service River Forecast System-Snow Accumulation and Ablation Model,
NOAA Technical Memorandum NWS HYDRO-17, U.S. Department of Commerce, Silver Spring, Maryland.
217 pp., 1976.
Budyko, M.I., The Earth's Climate: Past and Future. International Geophysics Series, Vol. 29, Academic
Press, New York, 307 pp., 1982.
i
Bultot, F, A. Coppens, G.L. Dupriez, D. Gellens, and F. Meulenberghs, Repercussions of a COfe Doubling on
the Water Cycle and on the Water Balance: A Case Study for Belgium, Journal of Hydrology. 99, 319-347,
1988.
Burnash, R.J., Ferral, R.L, and R.A. McGuire, A Generalized Streamflow Simulation System -- Conceptual
Modeling for Digital Computers. Joint Federal-State River Forecast Center, Sacramento, California. 204 pp.,
1973.
Cohen, S.J., Impacts of CC^-Induced Climatic Change on Water Resources in the Great Lakes Basin,
Climatic Change. 8, 135-153, 1986.
Dickinson, R.E., Modeling Evapotranspiration for Three-dimensional Global Climate Models, in Climate
Processes and Climate Sensitivity (Hansen, J.E. and T.Takahashi, editors), pp. 58-72, American Geophysical
Union Monograph 29, Vol. 5, Maurice Ewing, 1984.
Dracup, J.A., S.L Rhodes, and D. Ely, Conflict Between Flood and Drought Preparedness in the Colorado
River Basin, In Strategies for River basin Management (Lundqvist, J., U. Lohm, and M. Falkenmark, editors),
D. Reide!, Amsterdam, pp. 229-244, 1985.
Flaschka, I.M., C.W. Stockton, and W.R. Boggess, Climatic Variation and Surface Water Resources in the
Great Basin Region. Water Resources Bulletin. 3(1), 47-57, 1987.
Getches, D.H., Water Allocation During Drought in Arizona and Southern California: Legal and Institutional
Responses, University of Colorado, Natural Resources Law Center, Research Report Series, Boulder, CO,
101 pp., January, 1991.
Gleick, P.M., Methods for Evaluating the Regional Hydrologic Impacts of Global Climatic Changes, Journal
of Hydrology. 88, 99-116, 1986.
Gleick, P.H., The Development and Testing of a Water Balance Model for Climate Impacts Assessment:
Modeling the Sacramento Basin. Water Resources Research. 23(6), 1049-1061, 1987a.
Gleick, P.H., Regional Hydrologic Consequences of Increases in Atmospheric CO, and Other Trace Gases,
Climatic Change. 10,137-161, 1987b.
Gleick, P.H., Climate Change, Hydrology, and Water Resources. Review of Geophysics. 27, 329-344,1989.
Hansen, J., Russell, G., Rind, D., Stone, P., Lacis, A., Lebedeff, S., Ruedy, R., and L Travis, Efficient
Three-dimensional Global Models for Climate Studies: Models I and II, Monthly Weather Review. 111 (4),
609-662, 1983.
88
-------
Hansen, J., Fung, I., Lacis, A., Rind, D., Lebedeff, S., Ruedy, R., Russell, G., and P. Stone, Global Climate
Changes as Forecast by Goddard Institute for Space Studies Three-dimensional Model. J. of Geoohvs. Res.
93(D8), 9341-9364, 1988.
Hundley, N., Jr., Water and the West. University of California Press, Berkeley, 395 pp., 1975.
ICF Incorporated, Scenarios Advisory Meeting Summary Report (August 31 - September 1 , 1989), prepared
for the U.S. Environmental Protection Agency, 14 pp., December, 1989.
Intergovernmental Panel on Climate Change (IPCC). Climate Change: The IPCC Scientific Assessment. (J.T.
Houghton, G.J. Jenkins, and J.J. Ephraums, editors), World Meteorological Organization/United Nations
Environment Program, Cambridge University Press, Cambridge, 364 pp., 1990.
Karl, R.R. and W.E. Riebsame, The Impact of Decadal Fluctuations in Mean Precipitation. Climatic Change.
15, 423-448, 1989.
Kendall, D.R. and J.A. Dracup, An Assessment of Severe and Sustained Drought in the Colorado River basin,
Chapter 2 in Severe Sustained Drought in the Southwestern United States (Gregg, F., editor), Phase 1
Report to the U.S. Department of State, Man and Biosphere Program, 1991.
Klemes, V., Sensitivity of Water Resource Systems to Climate Variations, World Climate Applications
Programme, WCP-98, World Meteorological Organization, 142 pp., 1985.
Langbein, W.B., Water Yield and Reservoir Storage in the United States, U.S. Geological Survey Circular No.
409, Department of Interior, Washington, D.C., 1959.
Langbein, W.B. and others, Annual Runoff in the United States, U.S. Geological Survey Circular No. 5, U.S.
Department of Interior, Washington, D.C., 1949 (reprinted 1959). ;
Lettenmaier, D.P. and T.Y. Gan, Hydrologic Sensitivities of the Sacramento-San Joaquin River Basin,
California, to Global Warming, Water Resources Research. 26(1), 69-86, 1990.
Lettenmaier, D.P. and D.P. Sheer, Climatic Sensitivity of California Water Resources, Journal of Water
Resources Planning and Management. 117, 108r125, 1991.
Manabe, S., Climate and the Ocean Circulation I: The Atmospheric Circulation and the Hydrology of the
Earth's Surface, Monthly Weather Review. 97(1 1), 739-774, 1969a.
Manabe, S., Climate and the Ocean Circulation II: The Atmospheric Circulation and the Effect of Heat
Transfer by Ocean Currents, Monthly Weather Review. 97(1 1), 775-805, 1969b.
Manabe, S. and R.J. Stouffer, Sensitivity of a Global Climate Model to an Increase of CQj Concentration in
the Atmosphere. J. Geo. Res.. 85(C10). 5529-5554. 1980.
Manabe, S. and R.T. Wetherald, On the Distribution of Climate Change Resulting from an Increase in
j -content of the Atmosphere, J. Atmos. Sci.. 37, 99-1 18, 1980.
Manabe, S. and R.T. Wetherald, CQj and Hydrology. Advances in Geophysics. 28A, 131-157, 1985.
Manabe, S. and R.T. Wetherald, Large-scale Changes in Soil Wetness Induced by an Increase in Carbon
Dioxide, J. Atmos. Sci.. 44, 1 21 1 -1 235, 1 987.
89
-------
Mather, J.R. and J. Feddema, Hydrologic Consequences of Increases in Trace Gases and CO2 in the
Atmosphere, in Effects of Changes In Stratospheric Ozone and Global Climate. Volume 3, U.S.
Environmental Protection Agency, Washington, D.C., pp. 251-271, 1986.
Meams, L, P.H. Gleick, and S.H. Schneider, Climate Forecasting, in Climate Change and U.S. Water
Resources (Waggoner, P.E., editor), John Wiley and Sons, New York, 1990.
Mitchell, J.F.B. and Z. Qingcun, Climate Change Prediction, in Climate Change: Science. Impacts, and
Policy: Proceedings of the Second World Climate Conference (Jager, J. and H.L. Ferguson, editors),
Cambridge University Press, New York, 59-70, 1991.
Nash, LLand P.H. Gleick, The Sensitivity of Streamflow in the Colorado Basin to Climatic Changes. Journal
of Hydrology. 120, 221-241, 1991.
Nemec, J. and J. Schaake, Sensitivity of Water Resource Systems to Climate Variation, Hvdroloaical
Sciences 27(3), 327-343, 1982.
Ramanathan , V., The Role of Ocean-Atmosphere Interactions in the CO, Climate Problem, J. Atmos. Sci..
38, 918-930, 1981.
Revelle, R.R. and P.E. Waggoner, Effects of a Carbon Dioxide-induced Climatic Change on Water Supplies
In the Western United States, in Changing Climate! National Academy of Sciences, National Academy Press,
Washington, D.C., 1983.
Rind, D., R. Goldberg, J. Hansen, C. Rosenzweig, and R. Ruedy, Potential Evapotranspiration and t he
Likelihood of Future Drought. Journal of Geophysical Research. 95, 9983-10004, 1990.
Schaake, J.C., From Climate to Flow, in Climate Change and U.S. Water Resources (Waggoner, P.E., editor),
John Wiley & Sons, New York, pp. 177-206, 1990.
Schwarz, H.E., Climatic Change and Water Supply: How Sensitive is the Northeast?, in Climate. Climatic
Change, and Water Supply. National Academy of Sciences, Washington, D.C. 1977.
Somerville, R.C.J., The Predictability of Weather and Climate, Climatic Change. 11, 239-246, 1987.
Stockton, C.W. and W.R. Boggess, Geohydrological Implications of Climate Change on Water Resource
Development, U.S. Army Coastal Engineering Research Center, Fort Belvoir, Virginia, 1979.
Stockton, C.W. and G.C. Jacoby, Jr., Long-term Surface-water Supply and Streamflow Trends in the Upper
Colorado River Basin Based on Tree-ring Analyses, Lake Powell Research Project Bulletin No. 18, University
of Arizona, Tucson, 70 pp., 1976.
Stockton, C.W., D.M. Meko, and W.R. Boggess, Drought History and Reconstructions from Tree Rings,
Chapter 1 In Severe Sustained Drought in the Southwestern United States (Gregg, F., editor), Phase 1
Report to U.S. Department of State, Man and Biosphere Program.
U.S. Army Corps of Engineers (USACOE), Colorado River Basin Hoover Dam: Review of Flood Control
Regulation Final Report, USACOE, Los Angeles District, July, 1982.
U.S. Department of Interior (USDOI), Bureau of Reclamation, Colorado River Simulation System: System
Overview. USDOI, Denver, Colorado, 93 pp., 1987.
90
-------
U.S. Department of Interior (USDOI), Bureau of Reclamation. Updating the Hoover Dam Documents 1978.
U.S. GPO, Washington, D.C., 1980.
U.S. Environmental Protection Agency (USEPA), Potential Climatic Impacts of Increasing Atmospheric CO,
with Emphasis on Water Availability and Hydrology in the United States, Strategic Studies Staff, Office of
Policy Analysis, Office of Policy, Planning and Evaluation, 1984.
U.S. Environmental Protection Agency (USEPA), the Potential Effects of Global Climate Change on the
United States, Report to Congress, U.S. Environmental Protection Agency, Office of Policy, Planning, and
Evaluation, December, 1989.
Vaux, H.J., Jr., The Impacts of Drought on Water Qualfty, Chapter 5 in Severe Sustained Drought in the
Southwestern United States. Phase 1 report to the U.S. Department of State, Man and Biosphere Program,
1991.
Wetherald, R.T. and S. Manabe, The Effect of Changing the Solar Constant on the Climate of a General
Circulation Model. J. Atmos. Sci.. 45, 1397-1415, 1975.
Wilson, C.A. and J.F.B. Mitchell, A Doubled CO2 Climate Sensitivity Experiment with a Global Climate Model
Including a Simple Ocean, J. of Geophvs. Res.. 92(011), 13315-13343, 1987.
World Meteorological Organization (WMO), Intercomparison of Models of Snowmelt Runoff, Operational
Hydrology Report, WMO, Geneva, Switzerland, 1985.
World Meteorological Organization (WMO), Water Resources and Climatic Change: Sensitivity of Water-
Resources Systems to Climate Change and Variability, World Meteorological Organization, WCAP-4,
WMO/TD-No. 247, Geneva, Switzerland, 50 pp., 1987.
91
-------
APPENDIX A: CALIBRATION RESULTS FROM THE NWSRFS MODEL
APPENDIX B; THE LAW OF THE RIVER AND CRSS OPERATING PROCEDURES
APPENDIX C: ADDITIONAL RESULTS FROM THE CRSS MODEL
-------
APPENDIX A:
CALIBRATION RESULTS FROM THE NWSRFS MODEL
-------
-------
o
1
0>
0
o
.o
o
1
:§
O
s
»
1
f is
< o •=
II
" p
s
SSf
E
CO
8 To1
t E
H «
«
cs ? ^ evi » CM »-
g
OQcocoi-CMCMoeo
S
III IT-
co r--
§8
o
CO
? « §
A-3
-------
&
c
I
a
ra
E
3
OT
"53
•§
E
'1
JS
*T C3 CXI
S S S g 8
ct S § S «
§»~ O> O) Tf
U> CM CO (VI
C) OO o> CO N^
^ C, ^ CO CO
O> T*- CO N- O
10 CD CM b CO
CO ^ r^ O IO
1
SCO CO CO N>
CM CO i- 10
IO CO CM CO CO
i- CO M-' 00 T-'
f CM
OC) T^ IO IO CO
^ T- *f O
CM Si CM CM S °
"* 5 S r
O CO i- O> O CO
CM 00 10 j-; 0>
c8
V V V V V
A-4
-------
§
if
03
o
1
.g
'•*
O
>.
il»
O (T C
2 -s uj
•*" S
i* ^§ UJ
< O -=
I
lie
D I iS
CO
2" S
2 3 fe
a E 5
11°
s
SS *§ ~o
^ CO S
SCO CM ^* ^~ T"" tO C3 O) *'r" IO CO CO
T~ O) t**> f*» CO CM OJ O) ^" IO OJ C9
^^ ^3 C3 CO CO C^ CO ^^ T~ CM ^f O) ^3
COi— N-lOlOCNJCOlOOOtOCVJ O
lOOf^-CDThF^COlOCOOO^r^ CD
h*CM^*OjT-O>"' Jj
T-
oWi^oocNioidi^csicNjcsi i-1
I I I I I I I I
lO^- O>
r^ocoiniqioOT-^tocot^rs. r~f
c>ooc>oc>Wc6coT^c>c> v-
si
I
§ = It S
-5 -S < CO >
A-5
-------
i
i
co
I
tr
£
i
0)
I
I
o
I
ice
o
pi
1 T
2.
# I
lit
ill
-
6
TO TO O CO CO
•j uj cvi cvi co
TO t CO CO CM
§8822
TO CO CM CM CM
T— TO ^ CO CO
CO t-- O> IO CO
d d cj co •*
CO T- O O TO
c> o o o o
§> 8 « 5 8
3t *r O T— co
TO CO T—
co «* w r«. TO
T^ •* N; O •*
O O O CO N. T- u>
o o o co in
O CM
CO t CO
"
o cS T-_ 5 S S^
O O O O C3 U>
V V V V V
A-6
-------
I
o
? i fi
HI
ii
#
i
UJ
III
3 8 w
i
w
O> O CO t~~ -r- CO OJ
m*#meit^.t^m<6
CVJCVlCMCVICVI-4-ini-
i-CQCVICM
i i i i
i i T
i>^
10
W
a cj
8
O
_ CD O. <
A-7
-------
CO
8
9
1
o>
S
,g
I
S
o
V)
fie
1 = 1
m
3 §
2 S
u. «
8
I*- N. i-
CD Ifl U)
CM
N.
£3
§ R Sj S
S §§ S S
8 2 S IS §
c> i- W ri e\i
q o o q o5 d
O C7> CD t\J •*
i-; i-; O Cvi Tf
i- f CM O <0
i-^ CM t>- CM O>
o o d cj co
Si S Si «
8 S 5| §8 5| ff
O O O O i~ CO
V V V V V
A-8
-------
I
ro
O
I
_>,
O
-5
>
(0
< l-s
?f ui
J 2 S
2 8 I
11*5-
m
< I m
CO
>•
hi
if
= 1
E
33
OO
35
n
tfJcdiricoT^ojtt ri
T-COU>O>I*.^»- o>
9
2 J8
T ~ t
f
c _ £? Q. <
3330) t»
•3 -^ < CO >•
A-9
-------
— •
£
S3
5
5—
•a
V
N
•c
CO
E
3
V)
I
be
i
1
0
•C—
£
.0
1
"5
0
66
CD
B
K-
co
w!
* ui
?i
111
I § -o
'5 lii "ro
S °
» ^J
lii
CO
*&
III
T, ^
^ E re"
"3 S §
1 S
•5 a
2 j
"™" *>%
— S §
"c jS
s ^ s s
jf i-1 uj o>
^t U5 CO CM
CO CO O) rt
od m t*Z c\i
CM TO CM CM
500 N. CO
. N- •* U)
CM •* ui N!
CM ^- T- i-
p O O O
C> C3 C> CJ
1 1
*<1" ^J* CO IO
IO ^" ^ ir-
10 ui o o
OO 1^. O> CM
CO N. CO CM
o o oj N.'
O i- OD T-
TT CO <3 C\J
O O CM t^
O O T- O> CO CM
O> CM CO CM
§LO if)
-------
APPENDIX B:
THE LAW OF THE RIVER
AND CRSS OPERATING ASSUMPTIONS
-------
-------
THE LAW OF THE RIVER AND CRSS OPERATING ASSUMPTIONS
This appendix describes: (1) the major laws and agreements that govern allocation of the waters
of the Colorado River, and (2) the major operating parameters that are modeled by the CRSS model. The
discussion which follows on the "Law of the River" is drawn primarily from Getches (1991) and Hundley
(1977). The discussion of operating procedures is drawn from USDOI (1987).
The Law of the River
The apportionment of the Colorado River has been more complete than that of the waters of any
other river. The seven states along the 1400-mile river entered into the Colorado River Compact of 1922
dividing use of the river's water between the upper basin and the lower basin. The lower basin states of
Arizona, California, and Nevada were guaranteed that the upper basin states of Colorado, Wyoming, Utah,
and New Mexico would deliver an annual average of 7.5 million acre-feet of water to Lee Ferry, a point on
the river approximately on the Arizona-Utah border. The upper basin states received a right to use an
equivalent amount of water (if it was available). The lower basin also secured the right .to increase its
beneficial consumptive uses by another one million acre-feet. The parties contemplated each basin
eventually using equal quantities of water (7.5 million acre-feet), plus up to another one million acre-feet for
the lower basin. The Compact established that future obligations to Mexico would be shared equally by both
basins. A 1944 treaty with Mexico set the obligation for U.S. water deliveries from the Colorado at 1.5 maf
a year.
Under the Compact, the upper basin is not actually required to deliver a fixed quantity of water at
Lee Ferry for the lower basin in any particular year, though current operating criteria adopted by the Bureau
of Reclamation provide for releases of 8.23 million acre-feet annually. The only annual delivery obligation
in the Compact is one-half the Mexican Treaty guarantee of 1.5 maf. The water apportioned between the
basins has also been rather precisely divided among the states within each basin as described below.
B-3
-------
The Colorado River Compact required approval by Congress and ratification by each of the 7 basin
states. Before California agreed to ratify the agreement, it insisted on passage of the Boulder Canyon
Project Act, which authorized the construction of Boulder Canyon Dam (later known as Hoover Dam). In
passing this legislation in 1928, Congress added a suggested allocation of water among the states of the
lower basin, giving 4.4 maf to California, 2.8 maf to Arizona, and 0.3 maf to Nevada. Shortly after the
legislation passed, both California and Utah ratified the Compact. In 1944 Arizona finally approved the
Compact as a means of securing some of the benefits of Hoover Dam and of assuring that the lower basin's
Mexican treaty obligations would be shared among the three lower basin states.
The Upper Colorado River Basin Compact, approved in 1948, divided the upper basin's share of
water among each of the states on a proportional rather than absolute basis, except for Arizona, which
has only a small area in the upper basin and which was allocated 0.05 maf/year. Colorado received 51.75%
of the upper basin's share, Utah 23%, Wyoming 14%, and New Mexico 11.25%.
The lower basin's water was finally allocated among the states by the U.S. Supreme Court decision
In .Arizona v. California (1963), which adopted the apportionment suggested in the Boulder Canyon Project
Act. In addition, this decision recognized the rights of Native American tribes to water required for irrigable
acreage on reservations, although most of these rights have not yet been quantified.
California's rather firm entitlement to 4.4 maf a year, plus any surpluses to which the state is entitled,
has been divided by a 1931 "Seven Party Agreement". This agreement gives the highest priority to several
agricultural irrigation districts for up to 3.85 maf, then to the Metropolitan Water District of Southern California
and the City of Los Angeles for up to 550,000 acre-feet, then (to the extent that water remains unused) to
MWD and to the City and County of San Diego for 550,000 and 112,000 acre-feet respectively, with equal
priority. There are additional allocations and priorities, but these major provisions leave little water for any
other users.
B-4
-------
Two other pieces of federal legislation complete the list of major components of the Law of the
River: the Colorado River Basin Project Act of 1968 and Colorado River Basin Salinity Control Act of 1974.
The Colorado River Basin Project Act authorized the Central Arizona Project. In order to obtain passage
of this legislation, Arizona conceded that any annual shortages would be met from CAP'S allocation before
any reductions were made in the 4.4 maf of water designated for California. The Salinity Control Act requires
limits on the salinity of water entering Mexico and authorizes construction of a desalinization plant at Yuma,
Arizona.
CRSS Operating Assumptions
The CRSS model incorporates the Secretary of Interior's Operating Criteria for the reservoir system
as laid out in the "Criteria for Coordinated Long-Range Operation of Reservoirs" (USDOI, 1980). Among the
provisions which the CRSS models are:
• A minimum objective release from Lake Powell of 8.23 maf/year;
• The Mexican Treaty of 1944, which requires an annual delivery to Mexico of 1500 taf, except
in times of extreme shortage during which the burden is to be shared equally by U.S. and
Mexico. The CRSS model schedules deliveries to Mexico of 1515 taf to account for
unavoidable over-deliveries.
• . Section 602(a) of the Colorado River Basin Project Act, which allows excess water to be
stored in Lake Powell to the extent reasonably necessary to assure deliveries to the Lower
Basin without impairing future consumptive uses in the Upper Basin. The amount of this
storage is calculated based on the length of the most critical historical flow period, projected
demands in the Upper Basin, and the minimum power pool in Upper Basin reservoirs.
Typically, the Bureau assumes a 12-year critical period in which no shortages were imposed
on upper basin users;
• Balancing active storages in Lakes Powell and Mead at the end of the water year;
In addition, the CRSS also simulates the following procedures:
• Flood control provisions, which require that 5.35 maf of storage space be provided by
January 1 of each year in Lake Mead or upstream reservoirs. Between January 1 and July
31, flood control releases are based on forecasted inflow to prevent filling of Lake Mead
beyond its 1.5 maf minimum space to protect against rain floods. Minimum flood control
space is to increase linearly from 1.5 maf on August 1 to 5.35 maf on January 1.
• A surplus strategy, which is input into the model as a probability in order to minimize
unscheduled releases and increase hydroelectric output. For this study, the surplus strategy
was set to 0.7, the level of assurance normally used by the Bureau of Reclamation in its
modeling runs. Based on the historic record, an assurance level of 0.7 causes unscheduled
flood-control releases to be made in not more than 30% of the years;
B-5
-------
• The shortage strategy, which is triggered by the water-surface elevation of Lake Mead.
Level 1 and Level 2 shortages are imposed on the Central Arizona Project (CAP) and the
Southern Nevada Water Project (SNWP). Level 3 shortages are shared proportionately by
Mexico and US users.
The CRSS does not model water-rights priorities. Thus, when shortages occur, they are
implemented at their point of occurrence rather than being passed on to a user with a more junior water
right. In addition, the CRSS does not model Compact calls. Thus, when annual runoff at Lee Ferry falls
below 8.25 maf, shortages are borne primarily by lower basin, rather than upper basin, users.
B-6
-------
APPENDIX C:
ADDITIONAL RESULTS FROM THE CRSS MODEL
-------
-------
Table C1: Calibration Data for the CRSS Model
Mean Annual Bias (%) [1]
Station
Colorado River at Cisco, UT
Green River at Green River, WY
Colorado River at Lees Ferry, AZ
Colorado River below Hoover Dam
Colorado River at Imperial Dam
Flow
0.15
-1.61
-0.67
-1.62
-4.42
Salinity [2]
4.45
-1.78
3.08
-3.15
-5.28
Source: USDOI, 1987: 2.
Note (1 ] Bias is calculated on the basis of 16 observations of total annual flow (1968-1983) and is equal to simulated flow minus
[2] Salinity is calculated on a flow-weighted basis and is equal to total salt load for the year divided by total flow for the year.
Table C2: Mean annual runoff (taf) at Lees Ferry
— Comparison of the results obtained for three different sequences [11
Scenario
S1[2]
S2[3]
S3 [4]
Base
9,348
9,372
9,353
-20 %
-10%
-5 %
+5%
+10%
+20%
6,751
8,105.
8,769
9,959
10,629
12,119
6,926
8,205
8,801
10,038
10,775
12,289
6,843
8,079
8,728
10,045
10,785
12,307
Notes: [1 ] The numbers given here represent the total annual flow averaged over a 78-year record.
12] Sequence 1 has a starting storage level of 20,955 taf; input data begin with the year 1967.
[3] Sequence 2 has a starting storage level of 36,482 taf; input data begin with the year 1944.
[4] Sequence 3 has a starting storage level of 54,647 taf; input data begin with the year 1929.
C-3
-------
Table C3: Annual flow (taf) of the Colorado River at Cisco.
Scenario
-20%
-10%
-5%
Base
+5%
+10%
+20%
Mean
Flow [1]
3.181 (-29.7%)
3,849 (-14.9%)
4,182 (-7.5 %)
4,522
4,868 (7.7 %)
5.214 (15.3%)
5,910 (30.7o/o)
Standard
Deviation
1,227
1,419
1,540
1,678
1.807
1,912
2,117
Minimum
Flow
634 (-46.9 %)
802 (-32.8 %)
1,095 (-8.2"/o)
1,193
1,298 (8.8%)
1,410 (18.2%)
1,658 (39.0%)
Maximum
Flow
6,034 (-28.3 o/o)
6,793 (-19.30A)
7,241 (-13.9%)
8,413
8,985 (6.8 o/o)
9,551 (13.5%)
10,683 (27.0o/o)
Note: [1J Numbers in parentheses represent percent change compared to the base case.
Table C4: Annual
Scenario
-20%
-10%
-5%
Base
+5%
+10%
+20%
flow (taf) of the San
Mean
Flowfl]
983 (-27.5 %)
1,176 (-13.3%)
1,265 (-6.7%)
1.356
1,462 (7.8%)
1,571 (15.9%)
1,789 (31.9%)
Juan River at
Standard
Deviation
603
674
691
694
765
821
931
Bluff
Minimum
Flow
99 (-72.6 o/o)
114 (-68.40/0)
140 (-61.20/0)
361
402 (11.40/0)
423 (17.2o/o)
479 (32.7 %)
Maximum
Flow
2,580 (-21.3%)
3,036 (-7.4 o/o)
3,052 (-7.0 %)
3.280
3,513 (7.1 o/o)
3,755 (14.5 o/o)
4,177 (27.3o/o)
Note: [1] Numbers in parentheses represent percent change compared to the base case.
C-4
-------
Table C5: Effect of changes in runoff on average annual reservoir storage,
evaporation, and bank storage in Lake Powell.
Scenario
-20 %
-10%
-5 %
+5%
+10%
+20 %
Change in
Storage [1]
(taf)
(9,437)
(4,416)
(2,388)
2,751
3,875
4,720
Change in
Evaporation
(taf)
(215)
(95)
(50)
55
77
94
Change in
Bank Storage
(taf)
(755)
(354)
(191)
220
310
377
Note: [1] All numbers refer to difference relative to the base case.
Table C6: Average annual power generation (GWh) in the Upper Basin.
Scenario
S2[2]
S3 [3]
-20 %
-10 %
-5%
2,485
3,914
4,697
2,770
4,040
4,726
2,550
3,714
4,515
Base
5,460
5,471
5,460
+5%
+10%
+20 %
5,953
6,377
7,162
6,028
6,479
7,272
6,042
6,493
7,284
Notes: [1 ] Sequence 1 has a starting storage level of 20,955 taf; input data begin with the year 1967.
[2] Sequence 2 has a starting storage level of 36,482 taf; input data begin with the year 1944.
[3] Sequence 3 has a starting storage level of 54,647 taf; input data begin with the year 1929.
C-5
-------
s
o
QJ
QJ
2£
(D
CM
o cu
I-CD £
^ I
QJ
c_
CJ
CO
"CN
QJ
C35
(0
CX. n
J3:
«*
>> S
El
o»
en
8
UD
To
in
in
in
CM
= O
C-6
-------
CO
03
OJ
Ji^
ro
cu
en
JQ
00
ro
en
_in
CM
-O
CM
CD
C-7
-------
o
o
o
-O r-i
O -M
CD 03
CU
H-
03
O
(O
O T3
-O C
O CO
CD (f)
O
CO
O O
0_
(U
a
ra
o
ID
® o
(0
o
_o
o
in
in
in
in
CM
UOT^OBJJ
C-8
-------
o
-o
in
o
-LO
CM
03
CD
«+-
03
O
CO
-O c:
o 03
•*-i en
ID
O
CO
0]
-in
•i-i
i—i
cu
i-H
ID
o
-o
in
I
in
\
in
o
-in
C\J
in
CM
75
•o
•o
o
1
o
CO
Q
I
*o
•D
O
•o
I
o
*
|S
-- ***
3.1
£o
»l
UL a
C-9
-------
\^
in
in
CM
CM
CD
CD
H-
CD
C_
O
ra
•il
CO
CD
-s •
c_
CD
o
-in
CD
a
CO
13
C
c
-o
in
in
CM
I
"3
TJ
*i
C-10
-------
-^ cu
o
o
o
§ *
in
o
•*-• CD
a
CD
•a
c:
ro
en
o 1
.O 4-«
O "-1
^ en
CD
c_
CD
CD
a
uS 1
s I
-------
-------
|