Subseasonal Forecasts

Subseasonal forecasts of precipitation.

This section is dedicated to subseasonal forecasts, i.e. that bridge the gap between medium range weather forecasts (up to 10 days) and seasonal climate predictions (above a month). They are issued at different frequencies (from daily to once or twice a week) forecasting daily values with lead times from 1 to about 40 days, depending on the Global Producing Center (GPC). The availability of forecast products in the subseasonal-to-seasonal time range offers an unprecedented opportunity to develop intra-seasonal forecast information that other forecasts can't, in association with increased lead time compared to medium range weather forecasts, and with higher temporal resolution than seasonal forecasts that give an overview of an upcoming seasons (3 months). For instance, subseasonal forecasts may allow delivering relevant information about key climate characteristics such as the timing of the onset of a rainy season for agriculture, the risk of extreme rainfall events or heat waves in regards to public health.

At the moment, we propose experimental subseasonal forecasts of 2-week precipitation terciles based on the multi-model ensemble of individual forecasts issued every Saturdays through the SubX real-time database and every Thursday through the delayed S2S database. The forecasts are presented in the form of tercile precipitation probabilities for the weeks #2-3 and #3-4 averages, i.e. the 2-week periods 8-to-21 days and 15-to-28 days from the forecast issue date. For the delayed S2S forecasts, a Skill Maproom provides measures of historical skill for individual model forecasts and their multi-model ensemble.

Experimental (SubX) Forecasts
Calibrated Subseasonal Tercile categories precipitation experimental forecasts.
Near-Real Time Experimental (S2S) Forecasts
Calibrated Subseasonal Tercile categories precipitation experimental forecasts issued 1-2 months behind real time.
Subseasonal skill score based on the historical performance of each model and their multi-model ensemble.