| |
Accurate estimation of reference evapotranspiration (RET) is needed for determining agricultural water demand, reservoir losses, and driving hydrologic simulation models. This study was conducted to explore the application of downscaled NCEP’s Global Forecast System (GFS) reforecast dataset combined with NCEP-DOE Reanalysis 2 dataset to forecast RET over the states of Alabama, Georgia, Florida, North Carolina, and South Carolina in the southeast United States. Since only 12-hour temperature, wind speed, and relative humidity are available in the GFS reforecast dataset, six approaches of estimating RET using the Penman-Monteith (PM) and Thornthwaite equations were evaluated by substituting or adding the climatological mean values of variables including temperature, solar radiation, and wind speed from the Reanalysis 2 dataset. Both GFS and Reanalysis 2 datasets have coarse resolution with roughly 200-km grid spacing. Forecasts were downscaled using forecast analogs and the North American Regional Reanalysis (NARR) dataset (approximately 32-km per grid cell). Two evaluation criterion: Linear Error in Probability Space (LEPS) score and Brier Skill Score (BSS), were used to evaluate the overall forecast skill and the categorical forecast skill, respectively. The skill of both terciles and extremes (10th and 90th percentiles) were evaluated. The RET methods that combined Reanalysis 2 solar radiation data with GFS temperature and wind speed data to estimate parameters in the PM equation showed better skill compared to those that estimated these parameters from GFS outputs only. Most of the forecasts are skillful in the first 8 lead days. The upper extreme forecasts in coastal areas are more skillful than in inland; on the contrary, the lower extreme forecasts in inland areas demonstrated better skill than in coastal areas. Although the five categorical forecasts are skillful, the skills of upper and lower terciles forecasts are better than those of lower and upper extreme forecasts and middle terciles forecasts. |