Investigating Statistical Downscaling Methods and Applications for the NCEP/GEFS Ensemble Precipitation Forecasts

Abstract:

Significant discrepancies exist when coarse resolution model precipitation forecast products on standard output grids are verified against high-resolution analyses, remaining a challenge for NWP model guidance products. To enhance the usefulness of the model products, tremendous efforts with various statistical post-processing techniques are being made to reduce those discrepancies and recover small scale features using observations and a long-term reforecast climatology as the baseline. Among them, downscaling ensemble using forecast analogs (Hamill et al., 2006) and multiplicative downscaling using Parameter-elevation Regressions on Independent Slopes Model (PRISM) Mountain Mapper by WPC show promising improvement in skill of forecasts. This work concentrates on these two commonly used statistical downscaling approaches along with the Frequency Matching Method (FMM, Zhu and Luo, 2015) developed at NCEP/EMC. In this work, these three approaches will be investigated when applied to the standard one degree NCEP Global Ensemble Forecast System (GEFS) ensemble precipitation forecasts based on the 5-km high resolution NCEP Climatology-Calibrated Precipitation Analysis (CCPA) and 18 years ensemble control only reforecast data from the latest version of GEFS (GEFS v11.0). We will explore the effectiveness and feasibility of these approaches and to discover their strengths and weaknesses, with a focus mainly on generation of much higher 5km NDGD grid GEFS ensemble precipitation forecasts over the CONUS. This work is also expected to identify factors that influence the performance for each approach, such as the choice of case matching methods, the sample size, weighting function, regime definition, etc. A summary of the investigations and outcomes will be shown. Suggestions for some possible directions to produce such a high resolution ensemble precipitation forecast products in the future will be provided.