Experiments in Temperature and Precipitation Forecasting for Illinois

Abstract

Six years of daily temperature and precipitation forecasting are studied for Urbana, Illinois. Minimum temperature forecast skills, measured against a climatological control, are 57%, 48%, 34% and 20% for the respective forecast ranges of one, two, three, and four days. Maximum temperature skills are comparable. Precipitation probability skills of 29%, 19%, 6% and ?2% are found for the same respective forecast ranges. However, our skill in predicting precipitation amount, given that a measurable quantity occurs, is only 17% at the first day range and negligible thereafter. An examination of objective National Weather Service (NWS) forecasts shows this guidance to be slightly less skillful than our consensus in forecasting temperature and precipitation. Some temporal improvement is found in both the consensus and guidance temperature forecasts, but none can be found in the more difficult problem of forecasting precipitation. Significant warm and dry biases are frequently found in both our consensus and NWS guidance forecasts, especially during the summer season. These biases may be associated with the organized convective character of the precipitation in Illinois. Forecasts often miss these key events and, therefore, will often predict excessively warm maximum temperatures. Finally, the results show that our consensus skill is comparable to the state of the art. Student or faculty individuals usually lose to our consensus, as does the NWS objective forecast guidance. This establishment of the consensus forecast as typically being superior to an individual forecast has been reported by investigators in eastern United States cities.