资源环境科技发展态势分析平台

The North American Multi-Model Ensemble (NMME) provides hindcasts and real-time predictions for monthly mean climate fields at lead times of up to a year. These global climate model outputs can be useful in constructing improved seasonal forecasts. Here, several simple methods are developed and evaluated for forecasting monthly temperatures up to a year in advance based on either unweighted or weighted NMME outputs, and compared to previously developed statistical forecast methods that use only time series of past observations. It is found that the NMME-based methods produce forecast temperature probability distributions that are appropriately shifted toward the warm end of past experience and also show skill at representing interannual variability. NMME-based methods clearly outperformed purely statistical methods for forecasting temperatures over ocean, though over land this improvement is less clear over the evaluation period tested. The NMME seasonal forecasts may be particularly useful for giving early warning of heat waves, given their societal significance and higher conditional skill under those conditions.