A hybrid approach to improving the skills of seasonal climate outlook at the regional scale

Abstract

A hybrid seasonal forecasting approach was generated by the National Centers for Environmental Prediction operational Climate Forecast System (CFS) and its nesting Climate extension of Weather Research and Forecasting (CWRF) model to improve forecasting skill over the United States. Skills for the three summers of 2011–2013 were evaluated regarding location, timing, magnitude, and frequency. Higher spatial pattern correlation coefficients showed that the hybrid approach substantially improved summer mean precipitation and 2-m temperature geographical distributions compared with the results of the CFS and CWRF models. The area mean temporal correlation coefficients demonstrated that the hybrid approach also consistently improved the timing prediction skills for both variables. In general, the smaller root mean square errors indicated that the hybrid approach reduced the magnitude of the biases for both precipitation and temperature. The greatest improvements were achieved when the individual models had similar skills. The comparison with a North American multi-model ensemble further proved the feasibility of improving real-time seasonal forecast skill by using the hybrid approach, especially for heavy rain forecasting. Based on the complementary advantages of CFS the global model and CWRF the nesting regional model, the hybrid approach showed a substantial enhancement over CFS real-time forecasts during the summer. Future works are needed for further improving the quality of the hybrid approach through CWRF’s optimized physics ensemble, which has been proven to be feasible and reliable.