Item

Abstract

ABSTRACT: With the proposal of “seamless forecasting”, it is a key problem for meteorologists to improve the forecasting skills of subseasonal forecast. Since the launch of the S2S plan by WMO, the model has been further developed, but in the practical application of the Yangtze River Delta in recent years, we found that there is a large dispersion between members. Compared with the actual data, Only some members predicted well, while others were wrong. In order to improve the accuracy of sub seasonal forecast, the dispersion needs to be reduced. In this paper, the Climate Forecast System Version 2 (CFSv2) model and other modes are used as multi-members，and an improved deep learning model(on-line) is used to train the multi-members to obtain weighted average results of temperature and precipitation for subseasonal forecast in the Yangtze River Delta . The results show that the weighted ensemble model used in this paper is better than original model. Compared with equal-weight average results, the weighted ensemble model has a good ability to depict the temperature and precipitation in the Yangtze River Delta region. In order to compare the prediction performance, we also calculated the average climatic results. In particular, although the prediction of climate average have a better performance than weighted ensemble average through 20-30 days, the weighted ensemble model is substantially superior to climate average through 10-30 days. Thus, the weighted ensemble method( On-Line model) may shed light on improving the skill of sub-seasonal forecasts.