| |
Hydrologists and water management agencies require the accurate prediction of precipitation on a local spatial scale (10-20 km) that cannot be directly produced from Global Circulation Models (GCMs) due to their coarse grid resolution, typically on the order of several hundred kilometers. This study was conducted to improve short-term temperature and precipitation forecasts for use by Tampa Bay Water in their urban water demand model block of the decision making process. Two statistical downscaling and bias correction techniques were evaluated, logistic regression (LR) and an analog method (AM), using a retrospective forecast (re-forecast) data set. Reforecasting identifies a data set of retrospective numerical forecasts using the same model to generate real-time forecasts. Due to the large computational requirements reforecast data sets are not commonly produced. However, when available, the reforecast data set can be used to improve medium range weather forecast products, improve probabilistic forecasts of extreme events, and aid in diagnosing model errors. The reforecast data set used in this study was created using a fixed version of the National Center for Environmental Prediction’s (NCEP) Global Forecast System (GFS) model. The resulting reforecast data set has a 2.5 degree resolution (approx. 250km grid spacing) and covers 30 years of historical record (1/1/1979 to Present). The forecast skill of the two methods were evaluated with regard to seasonal variation, potential improvements in predictor selection, spatial scale of the application, variability in the selection of user defined parameters and different lead days. |