Assessment of the Utility of the Regional Climate Simulations in the Prediction of Drought

Regional model simulations have been completed for grid-spacing of 108 km, 50 km and 25 km for three periods: the winter season of 1986-87, the 1988 summer drought, and the 1993 summer flood. The simulations at coarse grid-spacing used the Kuo cumulus parameterization with implicit water processes. Both the 25 km and 50 km simulations used the Grell cumulus parameterization with explicit accountability of water processes. The ENSO simulation began on November 1, 1986 and ended on March 1, 1987 for each simulation. The boundary conditions were supplied by the ECMWF analyzes at 12 hour intervals. Both the summer 1988 and 1993 simulations began on April 1 and ended on September 1 of 1998 and 1993, respectively. Subsequently, we used the precipitation fields simulated by the RegCM to investigate whether the model data could be used in a drought forecasting model previously developed for the Ohio River Basin (Brennan and Barros, 1998). The model is an adaptive performance-feedback signal processing system, which uses the rate of change of the monthly anomalies of zonal order (East-West correlation) of precipitation fields to predict a basin-scale PDSI three to nine months in advance. The model was developed using observed precipiattion fields. For this application, the RegCM precipitation at 25 km resolution is used to augment the existing observations. For example, assuming that the current time is the last day of October in 1986, we add to the existing time-series of precipitation fields the four months of simulations generated by the RegCM. Thus, instead of being able to issue PDSI forecasts from January through July of 1987 using the observations alone, the forecast lead-time can be extended to November of 1987. The same procedure was repeated using the 1998 simulations, therefore allowing us to forecast the timing and magnitude of drought recovery up to 14 months in advance. Overall, the predicted PDSI series are very consistent with the actual values [absolute errors do not exceed 6 percent of the maximum range of PDSI values], with better results for the shorter lead times as expected. For the three months lead-time forecasts, no significant differences were found between the predictive skill of the drought model with, and without the RegCM precipitation fields were found. Furthermore, the PDSI forecasts are in good agreement with the estimated PDSI values, and capture well the trends even even for lead-times of fourteen months. Despite the large differences between observed and simulated precipitation amounts, this exploratory analysis suggests that the regional climate model can replicate well the observed time rate of change of the spatial variability of precipitation. The implications of these results for evaluating seasonal-to- interannual climate variability impacts on water resources using climate models will be discussed.

Brennan, K.E., and Barros, A.P., 1998: The Utility of Seasonal to Interannual Climate Predictions for Water Management: A Drought Forecasting Model for the Ohio River Basin. 2nd International Conference on Climate and Water, Espoo, Finland, Vol. 1, 333-342.