Seasonal-range forecasting of the ozark climate by a principal component regression scheme with antecedent sea surface temperatures and upper air conditions

JAE-WON LEE, ERNEST C. KUNG

Abstract

On the basis of principal component analysis of long-term climatological records, regression models are formulated and forecast experiments are conducted for monthly temperature and precipitation of the Ozark Highlands area, a large area of low mountains and plateau in the south central midwestern United States. Predictors include global sea surface temperatures, hemispheric upper air fields and the local climate observations. The experiments for all months of the year are performed with the data from continuous 15-year segments of 1961-75 to 1980-94 for those years beyond the respective data segments. Relationships between regional-scale and large-scale climate variables are investigated by cross-correlation analysis to identify useful teleconnections for seasonal-range forecasting. The predictability of the Ozark Highlands climate is examined with the multiple linear regression scheme and the principal component regression scheme. It is shown that the forecast performance by the latter is superior to that of the former. The results of the extensive forecast experiments reveal the useful and stable predictability of the Ozark Highlands climate elements. The validity of the forecasting models is verified for up to 10 years after the data period of regression formulation.

Full Text:

pdf