PNNL

Abstract

A simple method was developed to forecast 3-month and 6-month Standardized Precipitation Indices (SPIs) for the meteorological drought prediction over the United States based on the precipitation (P) seasonal forecasts from the NCEP’s Climate Forecast System (CFS). Before predicting SPI, the P forecasts from the coarse resolution CFS global model were bias-corrected and downscaled to a regional grid of 50 km. The downscaled CFS P forecasts out to 9 months were appended to the precipitation analyses to form an extended P data set. The SPIs were calculated from this new time series. Five downscaling methods were tested: (1) bilinear interpolation (BI), (2) a bias correction and spatial downscaling method based on the probability distribution functions (BCSD), (3) the Schaake method, (4) the Bayesian method developed by the Princeton University group, and (5) multi-method ensemble as the equally weighted mean of the BCSD, Schaake and Bayesian outputs. Hindcasts initialized in November, February, May and August were tested. For initial conditions from April – May, statistical downscaling methods are compared with dynamic downscaling based on the NCEP Regional Spectral model (RSM) and forecasts from a high resolution CFS T382 model. The skill is regionally and seasonally dependent. Overall, the 6-month SPI is skillful out to 3 - 4 months. For the first 3-month lead times, there is no statistical significant difference among different methods of downscaling because forecast skill comes from the precipitation analyses prior to the forecast time. After 3 month, the multi-method ensemble has small advantages, but forecast skill may be too low to be useful in practice. The skill of the BCSD method is comparable with hindcasts from the RSM and T382. That suggests the systematic errors are largely terrain related.