Global and regional skill of the seasonal predictions by WMO Lead Centre for Long-Range Forecast Multi-Model Ensemble

The World Meteorological Organization (WMO) Lead Centre for Long-Range Forecast Multi-Model Ensemble (WMO LC-LRFMME) has been established to collect and share long-range forecasts from the WMO designated Global Producing Centres (GPC). In this study, the seasonal skill of the deterministic multi-model prediction of GPCs in WMO LC-LRFMME is investigated. The GPC models included in the analysis cover 30 years of common hindcast period from 1981 to 2010 and real-time forecast for the period from DJF2011/2012 to SON2014. The equal-weighted multi-model ensemble (MME) method is used to produce the MME forecast. We show that the GPC models generally capture the observed climatological patterns and seasonal variations in temperature and precipitation. However, some systematic biases/errors in simulation of the climatological mean patterns and zonal mean profiles are also found, most of which are located in mid-latitudes or high latitudes. The temporal correlation coefficients both of 2 m temperature and precipitation in the tropical region (especially over the ocean) exceed 95%, but drop gradually towards high latitudes and are even negative in the polar region for precipitation. The prediction skills of individual models and the MME over 13 regional climate outlook forum (RCOF) regions for four calendar seasons are also assessed. The prediction skills vary with season and region, with the highest skill being demonstrated by the MME forecasts for the regions of the tropical RCOFs. These predictions are strongly affected by the ENSO over Pacific Islands, Southeast Asia and Central America. Additionally, Southeast of South America and North Eurasian regions show relatively low skills for all seasons when compared to other regions.