A comparison of methods to estimate daily global solar irradiation from other climatic variables on the Canadian prairies

Abstract

Historic estimates of daily global solar irradiation are often required for climatic impact studies. Regression equations with daily global solar irradiation, H, as the dependent variable and other climatic variables as the independent variables provide a practical way to estimate H at locations where it is not measured. They may also have potential to estimate H before 1953, the year of the first routine H measurements in Canada. This study compares several regression equations for calculating H on the Canadian prairies. Simple linear regression with daily bright sunshine duration as the dependent variable accounted for 90% of the variation of H in summer and 75% of the variation of H in winter. Linear regression with the daily air temperature range as the dependent variable accounted for 45% of the variation of H in summer and only 6% of the variation of H in winter. Linear regression with precipitation status (wet or dry) as the dependent variable accounted for only 35% of the summer-time variation in H, but stratifying other regression analyses into wet and dry days reduced their root-mean-squared errors. For periods with sufficiently dense bright sunshine observations (i.e. after 1960), however, H was more accurately estimated from spatially interpolated bright sunshine duration than from locally observed air temperature range or precipitation status. The daily air temperature range and precipitation status may have utility for estimating H for periods before 1953, when they are the only widely available climatic data on the Canadian prairies. Between 1953 and 1989, a period of large climatic variation, the regression coefficients did not vary significantly between contrasting years with cool-wet, intermediate and warm-dry summers. They should apply equally well earlier in the century.