An Improvement of NAEFS SPP – 2nd Moment Adjustment

The NAEFS Statistical Post Process (SPP) has been implemented at NCEP in 2006, upgraded in 2007, and extended in 2010. The current NAEFS SPP mainly consists of two major parts which are bias correction and downscaling. The NAEFS SPP is first operational application of multi-model, multi-center, multi-process ensemble forecast. There are many successes since NAEFS SPP was introduced into daily operation. However, for the bias correction, especially for NCEP GEFS, a same bias is considered for all ensemble perturbed members because single model is used for ensemble forecast. In general, there is no second moment adjustment or correction. In past years, a high expectation for 2nd moment improvement is strongly requested from users to enhance probabilistic forecasts. Recently, a method to hybrid current bias correction method (Kalman Filter) and Bayesian Model Average (BMA) is considered to effectively improve both of first moment and 2nd moment by using limited samples. Further, in order to improve longer lead-time forecast, such as week-2 forecast, 25-year ensemble reforecast data will be tested to expect improving NAEFS SPP for week-2 forecast.