Submitted to: American Meteorological Society Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 9/1/2003
Publication Date: 1/1/2004
Citation: GARBRECHT, J.D., SCHNEIDER, J.M., ZHANG, X.J. DOWNSCALING NOAA'S SEASONAL PRECIPITATION FORECASTS TO PREDICT HYDROLOGIC RESPONSE. CD-ROM. BOSTON, MA: AMERICAN METEOROLOGICAL SOCIETY. 2004.
Interpretive Summary: Recent advances in atmospheric and ocean sciences have made it possible to forecast trends in weather conditions up to a year in advance. In particular, NOAA is providing seasonal precipitation forecasts expressed as total precipitation over 3-month increments. To effectively use these forecasts in investigatins that require for water resources simulations, the forecasts must often be expressed in smaller time steps such as daily precipitation values. A method to convert the forecasted 3-month precipitation into daily values was proposed and tested by use of a hypothetical forecast at Temple, Texas. The test demonstrated that the method produced daily precipitation that were representative of the forecasted total precipitation. It was further recognized that forecasted changes in precipitation should be at least 10% or more of the mean precipitation to produce daily weather outcomes that can clearly be distinguished from average weather conditions. With this method it is now possible to calculate daily weather outcomes associated with seasonal forecasts, which in turn can be used in computer simulation of water resources investigations.
Technical Abstract: The impact of NOAA's seasonal precipitation forecasts on the water resources system must be determined to establish the utility of the forecast for water resources decision making. For a number of water resources applications this can be achieved by modeling the hydrologic system for the range of forecasted conditions. A method to adjust the precipitation parameters of a daily weather generator to reflect NOAA's forecasted seasonal precipitation conditions was proposed and tested. On a monthly basis, the generated precipitation reflects the forecast departures well. For two separate 100-year weather generations with different random numbers, the root-mean-square (RMS) differences between generated and target monthly precipitation were 2.9 and 4.4%, respectively. Maximum monthly difference was 7.7% for both data sets. The monthly differences between the generated and target values are inherent to the stochastic nature of weather generation and other model approximations. It is recommended that the monthly forecast departures be about 10% of the mean or higher to elevate the forecast signal above the noise of the stochastic component of the weather generation process.