Oregon District Hydrologic Studies
Using Advanced Statistical Techniques to Improve Streamflow Forecasts in the Upper Klamath Basin, Oregon
Project Chief: John C. Risley
Cooperator: Bureau of Reclamation
BackgroundUpper Klamath Lake is the primary source of water to the Bureau of Reclamation's (BOR) Klamath Project. Accurate water-supply forecasts of inflows to the lake, and for four other locations in the Upper Klamath Basin, are essential to manage the allocation of limited water supplies throughout the summer. From January to June forecasts for these five locations are provided by the United States Department of Agriculture, Natural Resources Conservation Service (NRCS). The forecasted streamflow volumes are determined from a combination of multiple regression model output, manually collected data, some subjective judgement, and coordination with the National Weather Service California-Nevada River Forecast Center. The uncertainty of the current regression models is relatively high. As a consequence of the high model uncertainty, forecasts may be too high or too low in any given year. Because of the intense demand for water in the Klamath Basin, critical decisions must be made by water managers and users early in the season based on these forecasts. Because many of these decisions have serious economic implications, accurate forecasts are of paramount importance.
Improving streamflow prediction and water management in the Upper Klamath Basin involves three major elements. The most obvious one is improving the accuracy and sophistication of the forecast models, which will manifest itself in a narrowing of the confidence limits around the median forecast. For example, a major hydrologic process, not accounted for in the current models, is the interannual effect of ground-water inflow to streams. The second element is an improved understanding and communication of forecast uncertainty so that the public can more clearly appreciate the range of possible streamflow outcomes that risk-based decisions can be based upon. The third element is to understand climate variability better, both short-term (year to year) fluctuations as well as longer-term cycles (decadal scale and longer).
Relevance and Benefits
The study will assist and improve water management decisions in the Upper Klamath Basin and develop a better physical understanding of the basin hydrology. Specific benefits provided by the work are opportunities to:
The project will have two phases. Phase 1 will specifically pertain to improving streamflow forecast methods. The work in phase 2 will concern improving the communication of forecast uncertainty to other agencies and the public and short- and long-term regional climate analyses.
In phase 1, an initial task in improving the streamflow forecasting methods will be data compilation. Available historic daily streamflow, precipitation, snow water equivalent (SWE), air temperature, and lake-level data from within and near the Upper Klamath Basin will be assembled for the study. Next, new ground-water index variables that influence interannual baseflow contributions to streamflow will be developed possibly from observation well data and climate time series data. The current NRCS models, which are based on principal components regression (PCR), will be updated using the ground-water index variables. In addition to revising the NRCS regression models, new forecast models using artificial neural network (ANN) technology will be developed. Daily precipitation, SWE, and air temperature time series will be used as input to predict flow volumes at a single forecast location. Sensitivity analyses and analyses of output from the ANN models using visualization techniques will be used to provide valuable insight into the dynamics of the surface- and ground-water system. A comparison analysis of the existing PCR, improved PCR, and ANN forecast models will be made. Model performance will be assessed on minimized residual error. Research findings from the study will be published in a journal article authored by USGS and NRCS personnel. A short manual describing the operation of the ANN models will be published as a USGS Water-Resources Investigations Report. During the course of the study, NRCS personnel will be given complete training in use of ANN modeling software. In future years, as additional years of climate data are collected, NRCS will have the option of updating and developing new ANN models.
Work in phase 2 will include improving the communication and understanding of forecast uncertainty to water managers and the public. To accomplish this, new graphical ways of displaying forecast distributions, including the progression of forecasts throughout the winter and spring, will be explored. Guidance on evaluating the risk and opportunity costs of basing water management and allocation decisions on streamflow forecasts, other than the median forecast, will also be developed.
Another activity in phase 2 will be an investigation of short- and long-term climate variability in the Upper Klamath Basin. Using climate time series, the relationship between the SOI and the PDO will be analyzed. Tree-ring chronology literature and other literature on long-term climate fluctuations relevant to the UKB region to better understand decadal-scale wet and dry oscillations will be also be compiled and analyzed.
U.S. Department of the Interior |
U.S. Geological Survey|
Maintainer: Oregon Webteam
Last modified Wednesday - Jan 9, 2013 at 18:56:14 EST
Privacy Statement · Disclaimer · FOIA · Accessibility