Management of water resources poses difficult scientific and engineering challenges world-wide. Loss of life and property due to too much (flooding) and too little (drought) water is a recurrent theme in both the developed and underdeveloped world. In the U.S., the Mississippi River floods of 1993, California's lengthy drought of the late 80s and 90s, and recent flooding in the Pacific Northwest are only a few among many examples. Water issues are particularly troublesome in the western U.S., where population growth and the scarcity of natural supplies are forcing reassessments of water management protocols. In the face of such conflicts, options for more efficient water management bear particular scrutiny. One area that has long been of concern in western water management is the accuracy of short to intermediate range streamflow forecasts. In most of the west, streamflow is derived largely from spring snowmelt, so fairly accurate forecasts can be made at the time of maximum snow accumulation (typically early spring) of the subsequent spring and summer runoff. Forecasts with longer lead times and for other forecast dates are usually much less accurate. Improvements in coupled modeling of the ocean-atmosphere system, such as El Nino-Southern Oscillation (ENSO) teleconnections, now appear to offer the potential for increasing the accuracy of seasonal-to-annual (mid-range) climate forecasts, which should translate to improved streamflow forecasts. At the same time, modernization of the National Weather Service, including state-of-the-art weather observing systems, advanced weather and climate modeling, and more sophisticated telecommunications and distributed computing systems at field offices offer the potential for improved long range (up to one year or more) streamflow forecasts, and better integration of such forecasts by water management agencies. Among the issues this session will discuss are a) what is the potential for improved intermediate range climate forecasting in western North American, and where and why is that potential the greatest; b) how do (or might) improved climate forecasts translate to improved streamflow forecasts, and at what lead time; c) how might water management efficiency be improved by improved streamflow forecasts, and are existing operating algorithms equal to the task?

As noted above, there is a great deal of interest in the application of intermediate range climate forecasts for water resources management. To date, however, there are no known applications in which quantitative intermediate range climate forecasts, e.g., precipitation and temperature forecasts for lead times of months to a year or more, have been used for water resources operation. As part of a JISAO project being sponsored by NOAA/OGP’s Human Dimensions Program ("Climate variability, impacts, and response strategies", Ed Miles, PI) the University of Washington group will soon be developing experimental long-range streamflow forecasts for the Columbia River system using output from NOAA’s global long-range climate forecast model. Other groups elsewhere in the U.S. are undertaking similar projects. This work is of great interest to NOAA, and potentially to water management agencies in the West, and elsewhere in the U.S.

The goal of this workshop was to bring together a small group representing the water resources and climate forecast communities, to explore how the needs, models, and constraints of the two communities can be meshed. At present, there are sizable challenges to making seasonal-to-interannual streamflow forecasts at the watershed level. Present day climate prediction efforts mainly employ global O(100km) coarse-resolution numerical and/or empirical models. The Columbia River forecast work underway at JISAO served as a case study, but the workshop focused on more general issues that arise in the application of output from weather and climate models to produce hydrologic forecasts. For instance, linking the high-resolution needs of the watershed models to the relatively low-resolution products of climate prediction tools poses a non-trivial set of problems. Alternative methods for downscaling the NCEP GCM output (for instance, using a nested high resolution mesoscale weather prediction model within the GCM) is among the various possibilities that will be explored.