A NUMERICAL METHOD FOR FORECASTING RAINFALL IN THE LOS ANGELES AREA

Abstract

The application of modern statistical methods to the forecasting of rainfall in Los Angeles is discussed. Forecasts are made by graphical integration of a number of objective meteorological variables and the results presented in terms of the probability of rainfall occurring in each of several amount categories. The accuracy of this technique is discussed and compared with that obtained by current conventional forecasting methods, while the precision of the probability estimates is compared with a subjective evaluation of the probability distribution. Both comparisons show a slight, but statistically nonsignificant bias in favor of the numerical method. The probability forecasts are shown to provide additional information regarding the reliability of each prediction which, by applying the principle of calculated risk, may be used to minimize the cost of carrying on any repetitive operation. An example of the use of this type of forecast is given, showing the saving which would result in a typical industrial operation in Los Angeles during the winter season of 1949?50.