The BART Experiment

The BART(Bay Area Regional Temperature) Experiment attempts to objectively forecast temperatures at seven locales in the San Francisco Bay Area by statistical means. Also, the experiment demonstrates the value of AFOS in providing local forecast aids. The method of least squares and a type of screening regression were used to create sets of seasonal forecast equations. Forecasts of daily maximum and minimum temperature at seven stations in northern and central California that are forecast points for the MOS (Model Output Statistics) technique plus the date/time-group-of the-forecast serve as predictors for the BART equations. The average reduction of predict and variance by the BART equations was 85 per- cent for maximum temperature forecasts and 67 Percent for minimum temperature forecasts. The mean standard error of estimate of the predictand was 3 degrees Fahrenheit. Appendices containing FORTRAN IV programs developed during the experiment and the AFOS procedure which obtains the temperature forecasts are also presented. Morris S. Webb, Jr. Includes bibliographical references (pages 7-8). 1979 NWS (National Weather Service) WR (Western Region) Library Public Domain 1931