Long-Range Forecasting of Temperature and Precipitation with Upper Air Parameters and Sea Surface Temperature in a Multiple Regression Approach

Abstract

With a premise that the patterns of the general circulation undergo continuous variations at various time scales, experiments with a multiple regression scheme are conducted in the long-range forecasting of temperature and precipitation using upper air parameters and sea surface temperature for the period 1963-1983. The scheme uses five predictors; it also effectively eliminates the problem of multi-collinearity. The scheme gives a lead time of predictors two to eleven months preceding predictands. The forecast scheme performs the complete regression analysis as an annual updating procedure for the teleconnection of the large-scale flow patterns. In forecast experiments, the data from the forecast year are excluded from the data base of the entire study period in the regression analysis. This also ensures that the forecast experiments are completely independent of the data of the forecast year. The experiments indicate that the predictands properly selected over various sizes of large areas of the United States, Canada and the USSR show considerable forecast skill.