Final Report: A National Assessment of the Impact of Climate Change on Water Resources

EPA Grant Number: R824992
Title: A National Assessment of the Impact of Climate Change on Water Resources
Investigators: Vogel, Richard , Kirshen, Paul , Moomaw, William
Institution: Tufts University
EPA Project Officer: Chung, Serena
Project Period: October 15, 1996 through October 14, 1999
Project Amount: $1,325,371
RFA: Global Climate (1996) RFA Text |  Recipients Lists
Research Category: Ecological Indicators/Assessment/Restoration , Global Climate Change , Climate Change


Water availability under possible future climate change is a subject of great interest and importance. National, regional, and even local water resource system managers must determine whether to adapt incrementally to climate change as it occurs, or to develop an anticipatory strategy based upon very uncertain impacts and societal responses to those impacts. The decision to adapt incrementally to climate change as it occurs, or to develop an anticipatory strategy depends upon the vulnerability of water resource systems under both current and future climate conditions. One objective of this research project was to develop and implement a methodology for evaluating the vulnerability of U.S. surface and groundwater resources under both existing and future climate/demand scenarios. New indices of vulnerability of water resource systems to climate change were introduced, which capture the effect of modifications in the balance among climate, streamflow, aquifer storage, and water use on the behavior of various water resource systems including: hydropower systems, surface water supply reservoirs, irrigation systems, groundwater aquifers, and navigation systems. Our goal was to distinguish which regions of the continental United States and which water use sectors are most vulnerable to potential future modifications in climate.

One of the unique features of this research was its application of existing historical sources of climatic, hydrologic, and other databases. An effort was made to use existing sources of climate and streamflow to derive a general understanding of the climate sensitivity of streamflow using observations of historical fluctuations in climate and streamflow, rather than through use of output from general circulation models and hydrologic models alone.

Assessments of the regional vulnerability of ground and surface water resources to climate change were performed at various temporal scales ranging from seasonal to annual. An important goal was to determine the largest temporal and spatial scales that still capture the impacts of various climate change scenarios, including all relevant agricultural, industrial, and municipal uses of both ground and surface water resources. Few studies have examined the impact of spatial and temporal modeling scales on the accuracy of resulting impact assessments. Our methodology includes comparisons with other studies to determine the level of modeling detail (temporal and spatial scales) necessary to draw definitive conclusions regarding the impact of climate on water resource system behavior.

A general methodology was developed that exploits readily available information on regional climate, hydrology, water demand, and water resource infrastructure. Our methodology is truly a regional approach because it uses regional hydrologic methods to extend site-specific watershed information on climate, hydrology, and water use for use at regional scales. Using recently released national databases of streamflow, climate, water demand, and water control infrastructure, along with recent advances in regional hydroclimatology, and the behavior of water supply systems, generalized hydroclimatologic models of 18 continental U.S. river basins were developed at time scales ranging from the seasonal to annual levels. The hydroclimatologic models provide projections of likely surface and groundwater availability for the entire United States under various climate change scenarios. This is accomplished by using regional hydrologic methods to develop a hydroclimatologic model for each large river basin from the hundreds of individual watershed models in that basin. The individual watershed models are continuous water balance models that capture the gross hydrologic budget including both ground and surface water as well as water control infrastructure such as reservoirs, lakes, and diversions.

Summary/Accomplishments (Outputs/Outcomes):

This final executive summary report outlines the major findings of this project. Each of the following sections outline the major conclusions derived from the various publications which are listed at the end of this summary. Only those conclusions which have relevance to the challenges posed by a national assessment of the impact of climate change on water resources are reviewed in this brief summary. Key findings that should have important policy implications are denoted using bold text.

The Impact of Climate Change on U.S. Water Resources

The Sensitivity of Streamflow to Climate Change in the United States. We introduced a generalized framework for evaluating the sensitivity of watersheds and water resource systems to climate change. Precipitation elasticity of streamflow, p, can be defined by the proportional change in streamflow divided by the proportional change in precipitation. The sensitivity of streamflow to climate change was investigated by calculating precipitation and potential evapotranspiration elasticity of streamflow for 1,360 watersheds in the United States. A unique national dataset of streamflow and climate time-series was constructed that accounts for the complex spatial variations in climate across the United States. The following conclusions were reached:

1. Both model form and model calibration play an important role in determining the sensitivity of climate to streamflow. Estimates of precipitation elasticity of streamflow, p, were shown to depend on both the choice and calibration of a watershed model. Therefore, it is difficult if not impossible to estimate the sensitivity of climate to streamflow using a single watershed model.

2. A contour map based on observations of streamflow and precipitation data was constructed which illustrates the spatial variability of precipitation elasticity of streamflow, . This map should provide a framework for comparing and evaluating all future investigations of the impact of climate change on streamflow within the United States. Values of p generally range from 1.0-2.5, across the United States, which implies that a 10 percent increase in precipitation would result in between a 10 percent and 25 percent increase in streamflow. The highest values of > 2 occur primarily in the arid and semi-arid regions of the Midwest and southwest.

3. The nonparametric estimator of introduced here is shown to be a useful validation statistic for future (and past) climate change investigations. Hopefully future climate change investigations will compare their results with the contour maps of presented in our study.

4. Comparison of climate elasticity derived from different investigators on the same river basin reveals that different hydrologists (for example, Schaake, 1990; and Nash and Gleick, 1991) can obtain very different climate sensitivities using the same model on the same basin. Apparently, in this context, differences between hydrologists can be more significant than differences between models and data!

Indicators of Impacts of Global Climate Change on U.S. Water Resources

A variety of environmental and socioeconomic indicators were introduced for evaluating the impacts of global warming on the water resources of the United States. One of the indicators, regional reservoir storage vulnerability, was found to be a particularly useful index for summarizing the effectiveness of regional water supply systems to meet future demands. A variety of indicator display and evaluation methods are compared and applied under current and one possible future climate change scenario assuming economic growth. We found that the primary impacts of global warming occur in western regions of the United States and include: (1) fewer relative stresses on hydroelectric systems due to an increase in energy supply from other sources, and (2) more stresses on available water resources due to increases in total withdrawals and, in some cases, decreases in flows. We have shown that the wise use of multivariate display methods for understanding tradeoffs among indicators may be preferred to mathematical aggregation of indicators into indices. Mathematical aggregation of indicators tends to result in a loss of detail and information and implies that an objective system of evaluating multiobjective tradeoffs exists, when in fact it does not. Weighting and/or aggregation of indicators into indices tends to add more complexity than value to the results.

The Impact of Climate Change on Floods and Droughts in the United States

Record Flood Events. We have completed regional evaluations of the impact of climate change on the record-breaking properties of floods. A record-breaking flood event is simply an event that exceeds all previous events. The first four moments of the number of record-breaking events in an n-year period are derived, assuming an independent process. The sampling properties of estimates of the mean and variance of the number of record-breaking events also are derived. Using these results, in combination with 1,556 flood records in the United States, it appears that the number of record-breaking flood events in the mid-western and western portions of the Untied States are significantly greater than would be expected if annual flood series were not influenced by climate change. However, subsequent work has shown that if one accounts for the spatial correlation structure of observed streamflow records, the number of record-breaking floods is consistent with a stationary climate. Using over 20 million daily streamflow observations across the United States, we have found no discernable evidence of the influence of climate change on the record-breaking properties of historical floods in the United States. Our work examining the record-breaking properties of floods, led us to examine whether or not discernable trends exist in historical records of U.S. floods and droughts.

Trends in Floods and Droughts. Other investigators have concluded that significant trends in floods and droughts exist throughout the United States, however most previous work has ignored the spatial correlation structure of observed streamflow records. Whenever one ignores spatial correlation of the records, one tends to find more trends than really exist. Trends in floods and droughts were evaluated using a regional average Kendall's S trend test. A bootstrap and an analytical methodology were introduced for accounting for the impact of spatial correlation of the flow records. We found no evidence of trends in historical flood flows but did find evidence of some upward trends in low flows in the midwestern, north central, and upper midwestern regions of the United States. A dramatically different set of conclusions would have been reached if cross-correlation of flow records had been ignored, in which case many more trends would have been incorrectly identified.

The Impact of Climate Change on the Behavior of Storage Reservoirs

The behavior of thousands of individual storage reservoirs for all regions of the United States was evaluated by combining a national inventory of dams with regional hydroclimatologic models developed in this project. Comparisons of the resilience, reliability, yield, and vulnerability of individual storage reservoirs were made under existing and future climate conditions. These comparisons should enable water supply managers to evaluate which regions of the country are most vulnerable to future climate change. Our results indicate that the western and southwestern regions of the United States are much more vulnerable to climate change than all other regions of the United States largely because those regions depend so heavily upon surface water storage combined with the extreme natural variability in climate in those regions.

The Impact of Climate Change on the Cost of U.S. Public Water Supply

Relations were explored between the annual cost of municipal water supply and numerous variables including quantity of water delivered, climatic, hydrologic, infrastructure, and economic variables. Multivariate statistical analyses suggest that present water supply costs are only related to the quantity of water supplied with other variables, particularly regional climate, capital costs, and operations, playing a negligible role. Current capital expenses are a relatively small component of present annual costs because: (1) most projects were publically financed in the past with costs not included in the rate base, (2) repayment of capital costs was completed several decades ago, and/or (3) there are fewer capital projects now due to recent emphasis on demand management. Therefore, present water supply costs cannot be used as a guide to future costs under either existing or future climate conditions.

The Influence of Climate on the Demand for Domestic Water

Relations were explored between the regional demand for domestic water and explanatory variables such as household wealth, combined price of water and wastewater services, and climate variables. When the dataset for the continental United States was analyzed, there were no strong relationships between domestic demand and the possible explanatory variables. If the continental United States is divided into regions of greater than or less than or equal to 20 inches of annual precipitation (i.e., the eastern and western United States), a significant relation among domestic demand, price, and standard deviation of precipitation was found for the western United States, the higher water using region. It is hypothesized that a certain minimum amount of water is used by both the eastern and western United States regardless of price or climate, but above that amount (primarily for outdoor use), there is price and climate sensitivity.

Regional Assessment of Water Allocation under Climate Change

A prototype regional water allocation model was developed using the southeastern region of the United States as an example. The model uses a linear programming algorithm to determine the influence of climate change on the allocation of ground and surface water resources within this region. One of the unique features of the model is that it uses a linearized version of the ?abcd' water balance model to explicitly model the interaction between surface and ground waters.

The Development of Regional Hydroclimatologic Models

The Spatial Scaling Properties of Streamflow

A log-linear relationship exists between the moments of annual streamflow and watershed drainage area, for all regions of the United States. Interestingly, the coefficients of these simple log-linear scaling models exhibit a near perfect linear correlation with moment order for all regions of the United States. This rather general, yet simple, scaling structure of annual streamflow occurs in spite of the fact that scale alone plays only a small role in explaining the moments of annual streamflow in the arid and semiarid regions of the west and southwestern United States. Scale plays a very different role in terms of its ability to explain regional differences in streamflow. For example, in the more temperate eastern regions of the United States, drainage area explains between 90 and 99 percent of the regional variability of annual streamflow, whereas in the more arid western and Midwestern regions, scale only explains a small fraction of the regional variability in streamflow.

The Variability and Persistence of Streamflow

Using a large database of annual streamflows we showed that the observed variability in streamflow persistence in each of 18 major water resource regions could easily be due to sampling alone. This result implies that the large differences one observes in the persistence of streamflow does not imply differences in the physical processes that give rise to persistence as is often assumed. Instead, the large differences in at-site estimates of persistence are probably due to sampling alone. In other words, broad regions of the United States exhibit homogeneous streamflow persistence structure. These results emphasize the importance of exploiting regional information regarding persistence, rather than trusting at-site data alone.

Although 18 broad geographic regions of the United States were found to be homogeneous in terms of streamflow persistence, those regions are not homogeneous in terms of the interannual variability streamflow. To assure homogeneity in terms of the year to year variability of streamflow, we document that the nation must be broken into more than 200 regions. This result is quite useful for studies that seek to perform national or regional hydrologic assessments. Regional assessments based on only 18 water resource regions may not be useful because most of those regions are heterogeneous in terms of hydrologic variability. Instead, at least 106 assessment subregions are required as in the second national assessment, or better yet the 222 planning subregions introduced by the U.S. Water Resources Council are required to assure hydrologic homogeneity. Further experiments confirm our inability to discern the complex long-term persistence structure of natural flow records due to the fact that only short-records are available. Assessment of the long-term persistence structure of actual flow records is further confounded by nonstationarity and nonhomogeneity of those flow records.

Annual Hydroclimatologic Models

Hydrologic, geomorphic, and climatic characteristics of 1,553 undeveloped watersheds across the United States were used to develop regional regression equations that relate the first two moments of annual streamflow to readily measured basin and climate characteristics. These relations are summarized for each of 18 major U.S. water resource regions. The relationships are remarkably precise with adjusted R2 values that range from 90.2 percent to 99.8 percent with an average value of 96.2 percent across the entire U.S. continent. These simple hydroclimatologic models should prove useful in future studies that seek to relate annual variations in climate to annual variations in streamflow. The usefulness of these relationships is evaluated by deriving their information content in terms of equivalent record length. These hydroclimatologic relationships also provide regional estimates of the climate elasticity of streamflow for all regions of the conterminous United States.

Regionalization Methods for Monthly Hydroclimatologic Models

The most fundamental hydrologic aspect of this research project involved the development of regional monthly water balance models for all regions of the United States. The challenge was to obtain a credible and sensible approach for estimating regional water balance models where data are sparse. This is the fundamental challenge behind this research grant, because any national assessment involves modeling hydroclimatologic responses in regions where data are sparse. A new methodology was developed for the regional calibration of monthly water balance models that is both credible and useful for performing regional and national assessments. The idea is to calibrate all water balance models within a region together attempting to achieve good agreement between modeled and observed streamflows at all sites in the region, while simultaneously attempting to obtain good regional relationships between watershed model parameters and watershed characteristics. We found that such a regional calibration approach can reproduce the behavior of observed streamflow throughout a region, while simultaneously providing remarkable improvements in the relationship between watershed model parameters and basin characteristics. These basin characteristics can then be subsequently used to estimate model parameters for ungauged basins.

Validation of Regional Hydroclimatologic Modeling Approach

A review of the literature reveals that many previous hydroclimatologic investigations have not performed adequate validation of the models, causing resulting conclusions to be suspect. To address this problem, we have evaluated the climatic elasticity of streamflow across the United States in an effort to provide a validation metric for past and future hydroclimatologic studies and to allow us to evaluate the climate sensitivity of streamflow across all regions of the United States. In particular, we were able to show that several previous climate change investigations produce questionable results when compared with our own results. To enable us to be certain of whether our elasticity results were correct, rigorous Monte-Carlo experiments were designed and implemented to evaluate alternate methods of estimating climate sensitivity to streamflow. Because our methodology was tested using a rigorous Monte Carlo investigation, we expect that our results will be useful for the validation of future hydroclimatological investigations within the continental United States.

Journal Articles on this Report : 9 Displayed | Download in RIS Format

Other project views: All 59 publications 11 publications in selected types All 11 journal articles
Type Citation Project Document Sources
Journal Article Douglas EM, Vogel RM, Kroll CN. Trends in floods and low flows in the United States: impact of spatial correlation. Journal of Hydrology 2000;240(1-2):90-105. R824992 (Final)
R826888 (1999)
R826888 (2000)
  • Full-text: Science Direct - Full Text HTML
  • Abstract: Science Direct - Abstract
  • Other: Science Direct - Full Text PDF
  • Journal Article Fernandez W, Vogel RM, Sankarasubramanian A. Regional calibration of a watershed model. Hydrological Sciences Journal 2000;45(5):689-707. R824992 (Final)
  • Full-text: Hydrological Sciences Journal PDF
  • Abstract: Hydrological Sciences Journal
  • Journal Article Kirshen PH, Larsen AL, Vogel RM, Moomaw W. Lack of influence of climate on present cost of water supply in the USA. Water Policy 2004;6(4):269-279. R824992 (Final)
  • Full-text: Water Policy PDF
  • Abstract: Water Policy
  • Journal Article Sankarasubramanian A, Vogel RM, Limbrunner JF. Climate elasticity of streamflow in the United States. Water Resources Research 2001;37(6):1771-1781. R824992 (Final)
  • Full-text: North Carolina State University PDF
  • Abstract: AGU
  • Journal Article Vogel RM, Bell CJ, Fennessey NM. Climate, streamflow and water supply in the northeastern United States. Journal of Hydrology 1997;198(1-4):42-68. R824992 (Final)
  • Full-text: Science Direct PDF
  • Abstract: Science Direct
  • Journal Article Vogel RM, Tsai Y, Limbrunner JF. The regional persistence and variability of annual streamflow in the United States. Water Resources Research 1998;34(12):3445-3459. R824992 (1998)
    R824992 (1999)
    R824992 (Final)
  • Full-text: Tufts University PDF
  • Abstract: Water Resources Research
  • Journal Article Vogel RM, Wilson I, Daly C. Regional regression models of annual streamflow for the United States. Journal of Irrigation and Drainage Engineering 1999;125(3):148-157. R824992 (1998)
    R824992 (1999)
    R824992 (Final)
  • Full-text: Tufts University PDF
  • Abstract: Journal of Irrigation and Drainage Engineering
  • Journal Article Vogel RM, Lane M, Ravindiran RS, Kirshen P. Storage reservoir behavior in the United States. Journal of Water Resources Planning and Management 1999:125(5):245-254. R824992 (1998)
    R824992 (Final)
  • Full-text: Journal of Water Resources Planning and Management PDF
  • Abstract: Journal of Water Resources Planning and Management
  • Journal Article Vogel RM, Sankarasubramanian A. Spatial scaling properties of annual streamflow in the United States. Hydrological Sciences Journal 2000;45(3):465-476. R824992 (Final)
  • Full-text: Hydrological Sciences Journal PDF
  • Abstract: Hydrological Sciences Journal PDF
  • Supplemental Keywords:

    water, watersheds, global climate, precipitation, ecosystem, indicators, regionalization, scaling, hydrology, engineering, hydroclimatology, water resources, national, United States, Regional, EPA Regions 1 through 10, streamflow, reservoirs, water supply, rainfall-runoff model, watershed model, cost, demand, trends, persistence, variability, floods, droughts, nonstationary, climate change, elasticity;, RFA, Scientific Discipline, Air, Water, Hydrology, Water & Watershed, climate change, Atmospheric Sciences, Ecological Risk Assessment, Watersheds, water resources, ecosystem models, National assessment, environmental monitoring, watershed, regional hydrologic vulnerability, hydrologic models, regional ecosystems, climate models, aquatic ecosystems, environmental stressors, climate variability, groundwater

    Progress and Final Reports:

    Original Abstract
  • 1997
  • 1998 Progress Report