The IRI's Climate Forecast System, and the Precipitation Outlook for the Forthcoming 3-Month Season

摘要

Each month, the International Research Institute for Climate Prediction (IRI) makes forecasts for the global temperature and precipitation for several overlapping 3-month periods extending to 6 months in advance. The forecasts are based mainly on the indications of five dynamical atmospheric general circulation models, whose behaviors are governed mostly by predicted global sea surface temperature (SST) anomalies. The IRI's forecast procedure begins with the prediction of SST anomalies, which are then used as a basis for prediction of the climate over land. The historical skill of the atmospheric models in hindcast mode, and of the real-time forecasts over 4 recent years, are discussed. Due to the large uncertainty in most climate forecasts, the IRI's forecasts are expressed as probabilities of pre-defined tercile categories (below, near, and above normal). The real-time precipitation forecast for an upcoming 3-month period is used as an example for illustrative purposes. The meaning and utility of such forecasts will be discussed in the context of the users' cost/loss contingencies, combined with the fact that most forecasts do not deviate drastically from the climatological distribution. Forecast formats that would be most suitable for the largest proportion of the user communities will be explored, together with prospects for their production.