Submitted to: Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 8/9/2003
Publication Date: 11/18/2004
Citation: Schneider, J.M., Garbrecht, J.D. 2004. How good are the seasonal predictions of total precipitation for Oklahoma? In: Proceedings of the Oklahoma Water 2004 Conference, October 29-30, 2003, Stillwater, Oklahoma. 2003&2004 CDROM. Interpretive Summary: Experimental climate forecasts are issued monthly by the NOAA Climate Prediction Center (CPC) for lead times from 0.5 to 12.5 months. The forecasts for 3-month total precipitation appear to offer information that could reduce uncertainty and risk in agricultural management. However, assessments of forecast performance at regional scales are not offered with the forecasts. The potential utility of these forecasts for practical applications is assessed using three measures: usefulness, dependability, and effectiveness. Usefulness measures the frequency and degree of departure of the forecasts from average conditions. Dependability measures the ability of the forecasts to accurately predict either wet or dry departures. Effectiveness is the frequency of dependable forecasts of wet and dry departures. Collectively, these measures indicate that from 1997 through 2002, the utility of these forecasts varies strongly across the Oklahoma and surrounding states, and that dependability is different for forecasts for wet vs dry departures. In the watersheds contributing to streamflow in Oklahoma, the utility of these forecasts has been modest at best, and limited to forecasts for wetter conditions during the late fall, winter, and early spring.
Technical Abstract: Experimental climate forecasts for 3-month total precipitation are issued monthly by the NOAA Climate Prediction Center (CPC) for lead times from 0.5 to 12.5 months. Among these predictions, the CPC probability of exceedance forecasts present information on expected shifts in the probability distribution of precipitation amounts, relative to climatological distributions. The statistical character of these forecasts could support a wide range of applications in water resource management. However, assessments of forecast performance at regional scales are not offered with the forecasts. We have examined forecast utility over the years 1997-2002, for all lead times in seven forecast divisions covering the watersheds contributing to streamflow in Oklahoma. Potential utility is assessed using three simple measures: usefulness (degree and frequency of forecast departures from climatological distributions), dependability (correspondence between direction of forecast departures and actual occurrences), and effectiveness (the frequency of dependable forecasts with useful departures). Using these measures, forecasts during the study period are shown to have widely varying degrees of utility, depending on location, season, and ENSO state. Utility for the Oklahoma has been modest at best, and limited to El Niño-related forecasts for wetter conditions during the late fall, winter, and early spring. Forecasts for drier conditions during the same seasons related to La Niña episodes have not been dependable in this region.