A Multiple Regression Technique of Estimating Mean Monthly Temperature Using Sea-Level Pressure

摘要

A simple point to point stepwise linear regression model which predicts mean monthly temperature using mean monthly sea-level pressure data is shown to be comparable in skill to a model which uses the coefficients of the principal components of the sea-level pressure as predictors. Regression equations are formed using as dependent data the pressure records from individual grid points in an area centered over North America for the period 1899 to 1960. Forecasts are then made from the equations for an independent record from 1961 to 1977. These predictions are shown to be less accurate than the forecasts made using the coefficients of the principal components. However, they display identical skill in forecasting above or below the long term monthly mean. Limited skill is demonstrated in predicting mean monthly temperature for January based on an actual long range prediction. (Author)