The IRI's forecasts of temperature and precipitation are issued near the middle of each month for the globe, with views by individual continent: View them. These forecasts are for the four overlapping 3-month periods extending out to six months in advance. Past forecasts are also shown, along with their corresponding observations, revealing the amount of accuracy of each forecast. The forecasts provide probabilities that precipitation or temperature will be in the lowest one-third of the climatological distribution, the middle one-third, or the highest one-third. In order to show the probabilities using only one map, the probability of the category having the greatest probability (33.3% or greater) is shown by color shading. On the maps for individual continents, histograms indicate the probabilities of all three categories for some regions. The climatological distribution is determined by the observations for the season in question over a recent 30-year history. The forecasts represent a distillation of information from a number of inputs, the most important being the predictions of several dynamical atmospheric prediction models that respond to the expected patterns in sea surface temperature (SST).

Forecasts of the individual prediction models contributing to the climate forecasts are shown each month, viewable here. Skill estimates based for these models are provided, based on hindcast simulations using prescribed observed SST, either by individual model or across all models for comparison.

Forecasts of the SST, to which the atmospheric models respond, can be viewed here. More is said about the SST forecasts in the entry panel for that product.

The forecasts of the individual models are objectively combined into a multi-model ensemble probability forecast, where the weights assigned to the individual models are based on historical hindcast performance for the season and location. Maps showing these purely objective forecast results are shown. [The year shown in the url may be adjusted]. The IRI's final issued climate forecasts, referenced earlier, are determined almost entirely by this objective multi-model product but sometimes may be adjusted slightly on the basis of factors such as known model biases, uncertainties in aspects of the SST forecast, late-breaking SST or climate information, etc. The most common kind of such subjective, human adjustment is a weakening of the forecasts' deviation from the climatological probability distribution of a 1/3 probability for each of the tercile-based categories.