A Preliminary Investigation of Temperature Errors in Operational Forecasting Models

Temperatures taken from model output (FOUS reports) routinely transmitted by the National Centers for Environmental Prediction are tabulated to determine errors during three months in the summer of 1996. These short-term model forecasts are compared to the models’ own initialization over three 1-month periods. The results suggest that the Eta Model consistently forecasted low-level temperatures that were too warm, while the Nested Grid Model consistently forecasted low-level temperatures that were slightly too cold. This paper shows some of these statistics, along with a case example in which the models made very large errors (greater than 10°C). A check of recent model output indicates that changes made recently to the Eta Model have reduced much of the bias found in the summer of 1996. A recent case example shows, however, that large short-term forecast errors still occur, even with the updated Eta Model.