Grantee Research Project Results
2000 Progress Report: A National Assessment of Low-Streamflow Estimation Using a Physically Based Statistical Methodology
EPA Grant Number: R826888Title: A National Assessment of Low-Streamflow Estimation Using a Physically Based Statistical Methodology
Investigators: Kroll, Charles , Vogel, Richard
Institution: The State University of New York , Tufts University
EPA Project Officer: Hahn, Intaek
Project Period: September 1, 1998 through August 31, 2001 (Extended to December 31, 2003)
Project Period Covered by this Report: September 1, 1999 through August 31, 2000
Project Amount: $363,500
RFA: Environmental Statistics (1998) RFA Text | Recipients Lists
Research Category: Human Health , Aquatic Ecosystems , Environmental Statistics
Objective:
Low streamflow estimates are crucial for water quality management, issuing and/or renewing National Pollution Discharge Elimination System (NPDES) permits, planning water supplies, hydropower, and irrigation systems, and for assessing the impact of prolonged droughts on aquatic ecosystems. Unfortunately, there is no agreed upon methodology for estimating low streamflow statistics in the United States. The current methods for estimating low streamflow statistics are based on techniques recommended for flood frequency analyses. By combining recent advances and new methods in both physical and statistical hydrology with Geographic Information System (GIS) mapping techniques, we will improve our ability to estimate low flow statistics in riverways throughout the entire United States. The methodology developed will provide water resource planners with a scientifically based procedure to estimate low streamflow statistics at both gauged and ungauged river sites throughout the United States. Physical-Statistical models developed will also aid in understanding the potential impact of global climate fluctuations on low streamflows.Progress Summary:
We currently are finishing the first segment of this research, which examined issues of low streamflow series at gauged river sites across the United States. This has focused on four major areas: trends in low streamflow series, the impact of low streamflow serial correlation (persistence) on quantile estimators, the appropriateness of different probability distributions in describing low streamflow series at intermittent and non-intermittent streamflow sites, and the estimation of low streamflow statistics at intermittent sites. Each of these topics are discussed below.
1. An Analysis of Trends in Low Streamflows in the United
States
With much discourse involving global climate fluctuation, an
important issue is whether we are observing any significant trends in
streamflows series. We have investigated the importance of accounting for both
spatial and temporal correlation on a Mann-Kendall trend test applied in all
regions of the United States, and the problems with employing trend tests with
correlated data. The resulting paper (referenced below) provides some startling
results. Most importantly, when the cross-correlation of low streamflow (and
flood flow) series is ignored, one concludes that there are significant trends
in nearly all regions across the United States. Using a bootstrap resampling
procedure to account for the cross-correlation of these series dramatically
changed this result, with significant upward trends observed only in the Ohio,
north central and upper Midwest regions.
2. The Impact of Serial Correlation on the Return Period and Failure Risk of Hydrologic Design
Conventional methods for estimating the average return period, E(T), and failure risk, R, generally ignore serial correlation (persistence) in streamflow. For commonly observed ranges of persistence, E(T) can be up to 65 percent greater and R up to 27 percent lower than conventional estimates meaning that the expected design life of a system is longer when persistence is taken into account. Also, by ignoring persistence, low flow quantiles estimators may be underestimated by a half order of magnitude or more. This analysis is most pertinent in the upper Midwest region of the United States, where significant trends in low streamflow series were observed.
3. The Probability Distribution of Low Streamflow Series in the United States
The use of L-moment ratio diagrams to discriminate between competing probability distributions for describing streamflow series is now common. In this analysis L-moment diagrams were used to analyze the goodness-of-fit of various probability distributions to low streamflow series across the United States. Using nearly 1500 sites spatially distributed across the United States, it was still difficult to ascertain the "best" probability distribution to fit low streamflow series in various regions. Using a weighted-distance metric, different conclusions were drawn when intermittent river sites (where streamflow is sometimes reported as zero) were removed from the analysis. It was shown that the L-moment ratios of censored probability distributions differs from those of uncensored probability distributions. An analytical experiment was developed to examine the goodness-of-fit of probability distributions at intermittent river sites. The conclusions were that at non-intermittent river sites, the 3-parameter lognormal distribution should be employed, while at intermittent river sites, the 3-parameter Pearson distribution should be used. More importantly, we have developed a framework to investigate goodness-of-fit procedures for censored data sets, which may have far reaching applications in other areas, especially water quality where many censored data sets are observed.
4. A Comparison of Low Streamflow Quantile Estimators for
We currently are finishing the first segment of this research, which examined issues of low streamflow series at gauged river sites across the United States. This has focused on four major areas: trends in low streamflow series, the impact of low streamflow serial correlation (persistence) on quantile estimators, the appropriateness of different probability distributions in describing low streamflow series at intermittent and non-intermittent streamflow sites, and the estimation of low streamflow statistics at intermittent sites. Each of these topics are discussed below.
1. An Analysis of Trends in Low Streamflows in the United
States
With much discourse involving global climate fluctuation, an
important issue is whether we are observing any significant trends in
streamflows series. We have investigated the importance of accounting for both
spatial and temporal correlation on a Mann-Kendall trend test applied in all
regions of the United States, and the problems with employing trend tests with
correlated data. The resulting paper (referenced below) provides some startling
results. Most importantly, when the cross-correlation of low streamflow (and
flood flow) series is ignored, one concludes that there are significant trends
in nearly all regions across the United States. Using a bootstrap resampling
procedure to account for the cross-correlation of these series dramatically
changed this result, with significant upward trends observed only in the Ohio,
north central and upper Midwest regions.
2. The Impact of Serial Correlation on the Return Period and Failure Risk of Hydrologic Design
Conventional methods for estimating the average return period, E(T), and failure risk, R, generally ignore serial correlation (persistence) in streamflow. For commonly observed ranges of persistence, E(T) can be up to 65 percent greater and R up to 27 percent lower than conventional estimates meaning that the expected design life of a system is longer when persistence is taken into account. Also, by ignoring persistence, low flow quantiles estimators may be underestimated by a half order of magnitude or more. This analysis is most pertinent in the upper Midwest region of the United States, where significant trends in low streamflow series were observed.
3. The Probability Distribution of Low Streamflow Series in the United States
The use of L-moment ratio diagrams to discriminate between competing probability distributions for describing streamflow series is now common. In this analysis L-moment diagrams were used to analyze the goodness-of-fit of various probability distributions to low streamflow series across the United States. Using nearly 1500 sites spatially distributed across the United States, it was still difficult to ascertain the "best" probability distribution to fit low streamflow series in various regions. Using a weighted-distance metric, different conclusions were drawn when intermittent river sites (where streamflow is sometimes reported as zero) were removed from the analysis. It was shown that the L-moment ratios of censored probability distributions differs from those of uncensored probability distributions. An analytical experiment was developed to examine the goodness-of-fit of probability distributions at intermittent river sites. The conclusions were that at non-intermittent river sites, the 3-parameter lognormal distribution should be employed, while at intermittent river sites, the 3-parameter Pearson distribution should be used. More importantly, we have developed a framework to investigate goodness-of-fit procedures for censored data sets, which may have far reaching applications in other areas, especially water quality where many censored data sets are observed.
4. A Comparison of Low Streamflow Quantile Estimators for
Future Activities:
The planned future activities for this project include:
- An experiment examining the use of baseflow correlation techniques to estimate low streamflow quantiles. This will involve various regional analyses throughout the conterminous United States. Initial results from this experiment show baseflow correlation to be a poor method for low streamflow estimation, though these results may be due to experimental design.
- Improving and expanding our database of watershed characteristics at over 1500 watershed with coincident long-term discharge records. These sites are spatially distributed throughout the United States. As new digital coverages become available, these are processed and summary statistics are included in the database. We are also investigating the impact of digital scale on this database and the resulting models developed using the database.
- Investigating the estimation of baseflow recession constants and the baseflow index at ungauged river sites. The success of this study will have a far reaching impact, influencing low streamflow predictions, drought forecasting models, and rainfall-runoff models with require these parameters.
- A comprehensive comparison of a variety of low streamflow estimation techniques at ungauged, and partially gauged river sites throughout the United States. This will involve a cross comparison of various methodology, with an analysis of the tradeoffs between streamflow data and estimator uncertainty.
Journal Articles on this Report : 5 Displayed | Download in RIS Format
Other project views: | All 13 publications | 5 publications in selected types | All 5 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
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. |
R826888 (1999) R826888 (2000) R824992 (Final) |
Exit Exit Exit |
|
Douglas EM, Vogel RM, Kroll CN. Impact of streamflow persistence on hydrologic design. Journal of Hydrologic Engineering 2002;7(3):220-227. |
R826888 (2000) |
Exit |
|
Kroll CN, Vogel RM. Probability distribution of low streamflow series in the United States. Journal of Hydrologic Engineering 2002;7(2):137-146. |
R826888 (2000) |
Exit |
|
Kroll C, Luz J, Allen B, Vogel RM. Developing a watershed characteristics database to improve low streamflow prediction. ASCE Journal of Hydrologic Engineering 2004;9(2):116-125. |
R826888 (2000) |
Exit |
|
Reilly CF, Kroll CN. Estimation of 7-day, 10-year low-streamflow statistics using baseflow correlation. Water Resources Research 2001;39(9):Art. No. 1236. |
R826888 (2000) |
Exit Exit |
Supplemental Keywords:
low streamflow statistics, droughts, stochastic hydrology, surface water, groundwater, risk assessment, GIS, geographic information systems, watershed hydrology, statistical hydrology, regional hydrology., RFA, Scientific Discipline, Economic, Social, & Behavioral Science Research Program, Air, Ecosystem Protection/Environmental Exposure & Risk, Hydrology, Ecosystem/Assessment/Indicators, Ecosystem Protection, exploratory research environmental biology, Chemical Mixtures - Environmental Exposure & Risk, climate change, Ecological Effects - Environmental Exposure & Risk, Ecological Effects - Human Health, Ecology and Ecosystems, Atmospheric Sciences, Ecological Risk Assessment, Environmental Statistics, Ecological Indicators, ecological exposure, flood frequency analysis, risk assessment, estimating low stream flow, global environmental data, meteorology, physical environmental statistics research, streams, environmental risks, drinking water supplies, low stream flow estimation model, water resource planners, mitigation strategies, physically based statistical methodology, hydropower, statistical models, aquatic ecosystems, ecosystem impacts, stream flow, data analysis, data collection, spatial-temporal methods, water quality, data models, global warming, innovative statistical models, statistical methodology, land useProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.