This seasonal forecasting system consists of probabilistic precipitation seasonal forecasts based on the full estimate of the probability distribution.
Probabilistic seasonal forecasts from multi-model ensembles through the use of statistical recalibration, based on the historical performance of those models, provide reliable information to a wide range of climate risk and decision making communities, as well as the forecast community. The flexibility of the full probability distributions allows to deliver interactive maps and point-wise distributions that become relevant to user-determined needs.
The default map shows globally the seasonal precipitation forecast probability (colors between 0 and 1) of exceeding the 50th percentile of the distribution from historical climatology. The historical climatology used is 1982-2010 up to the forecast issued in August 2021, and is 1991-2020 from the forecast issued in September 2021. The forecast shown is the latest forecast made (e.g. Dec 2017) for the next season to come (e.g. Jan-Mar 2018). Four different seasons are forecasted and it is also possible to consult forecasts made previously. The forecasts are directly computed from the extended logistic regression model as probabilities of exceeding (or non-exceeding) of every 5th percentile of the climatological distribution. The specific quantile (in steps of 5 percentile points) can then be selected. The user can also specify a quantitative value in physical units (here seasonal total precipitation in mm) for probability of exceeding or non-exceeding. The final probability maps are smoothed spatially with a 9×9 point Gaussian smoother.
Clicking on a point on the map will show the local probability of exceeding and probability distribution of the forecast (green) together with the climatological distribution (black).
The distribution of seasonal rainfall may often rightfully be approximated by a normal distribution, especially when considering only years when it actually rained. But there can be a significant number of years in the sample when rainfall is 0. This translates in situations where the probability of rainfall to be just above 0 is less than 100%, which are represented in by the vertical lines at 0mm in the distributions graphs.
IRI’s official forecasts since 1997 can be consulted on IRI website.
Colors Scales
Color scales are colors indicating that the distribution of the forecast tends towards drier (shades of brown) or wetter (shades of blue) conditions than normal (moccasin).
Older Forecasts
Older and retrospective forecasts can be found under the last tabbed entries of the Seasonal Forecasts Maproom. The flexible forecast Maprooms shown for start times between January 2011 to March 2017 were made retrospectively using the revised methodology introduced in April 2017. The Flexible Forecasts issued from July 2012 to March 2017 are discontinued and used a different methodology to obtain the full distribution of the forecasts.
Release dates
Our seasonal forecasts are released on the 15th of each month. If the 15th falls on a weekend or holiday, they are released on the closest workday, earlier or later.
Forecast ELR coefficients: 1˚ spatial resolution multi-model ensemble mean and ELR coefficients available here.
Historical precipitation: ELR coefficients derived from the CPC-CMAP-URD dataset available here at 1˚ spatial resolution.
Contact help@iri.columbia.edu with any technical questions or problems with this Map Room.