Monthly Dynamical Extended-Range Ensemble Forecasts with Different SST Forcing

A monthly dynamical extended-range ensemble forecast system is established with a global spectral model (T106L19) and an initial ensemble method of breeding of growing modes (BGM). Forty-eight examples of monthly integrations are performed with this system, and within each example, there are three sets of monthly ensemble forecasting experiments. The first set, named the CSST experiment, is forced by climatic sea surface temperatures (SSTs), and the second set, named the FSST experiment, is forced by predicted SSTs, while the last set, named the AVE30 experiment, is the mean of CSST and FSST. Then, the model results are verified and analyzed. Verifications show that the initial ensemble forecasts, based on BGM, can improve the monthly prediction skill remarkably when compared to a single control run, and it is especially precise for the Pacific and Northern America (PNA) region and the predicted SSTs have a positive contribution to the forecasts skill, particularly. The AVE30 experiment, that takes impacts of both initial uncertainties and boundary uncertainties into consideration, has a higher prediction skill than both the CSST and FSST experiments and suggests that we should attach more of an importance to roles of both initial conditions (ICs) and boundary conditions (BCs) to improve the monthly extended-range forecast skill. Meanwhile, analyzing the results indicate that atmospheric responses to SST forcing become significant after the tenth model integration day and the influences of SST forcing on the second and third ten-day period have a direct effect on the prediction skill of the monthly mean circulation. The influences of SST forcing on the second and third ten-day period relate to SSTs variation as well as ICs and the influences differ under different ICs. The monthly ensemble prediction system exhibits a strong capability of forecasting the monthly mean air temperature in China during the summer. The results of FSST forced by predicted SSTs prevail against those of CSST forced by climatic SSTs in general.