The COLA global atmospheric model has been used to produce "forecasts" for the winters (January - March) of 1982 - 1997 in order to quantify the model's forecast skill. The accompanying Figure shows the anomaly correlation coefficient of seasonal average precipitation rate for all winters, JFM 1982 - JFM 1997. This is a map of the correlation in time of the COLA model forecast with the observed seasonal average precipitation rate. The observational data set used for this model validation is the NOAA CPC CAMS-OPI precipitation data set which is based on rain gauge stations over continental North America in combination with satellite estimates of precipitation rate over ocean. We show the correlation only over land points to be consistent with the forecast figure for JFM 1998.

As may be seen in the Figure, the highest values of forecast skill are in southern California and along the coast of the Gulf of Mexico. These areas are coincident with the areas where El Niño is known to have its largest effect in North America, so this map is consistent with other evidence that the COLA model responds fairly accurately to tropical Pacific sea surface temperature anomalies in its Northern Hemisphere winter circulation forecasts. Generally speaking, the areas with higher anomaly correlation coefficient are areas where the current (JFM 1998) forecast may be considered more reliable.