This study investigates dynamically different data-driven methods,specifically a statistical downscaling model (SDSM), a time laggedfeedforward neural network (TLFN), and an evolutionary polynomial regression(EPR) technique for downscaling numerical weather ensemble forecastsgenerated by a medium range forecast (MRF) model. Given the coarseresolution (about 200-km grid spacing) of the MRF model, an optimal use ofthe weather forecasts at the local or watershed scale, requires appropriatedownscaling techniques. The selected methods are applied for downscalingensemble daily precipitation and temperature series for the Chute-du-Diablebasin located in northeastern Canada. The downscaling results show that theTLFN and EPR have similar performance in downscaling ensemble dailyprecipitation as well as daily maximum and minimum temperature serieswhatever the season. Both the TLFN and EPR are more efficient downscalingtechniques than SDSM for both the ensemble daily precipitation andtemperature.
展开▼
