Figure 6.5.2 Explained variance (%) of monthly PET from multiple linear regression estimates
Return to the point in the document where this figure is located.
First five panels show effect of replacing station windrun with a pressure index.
Text description of figure
The figure consists of 6 panels, each of which contains three line graphs, all showing explained variance in potential evapotranspiration against the month of the year. The explained variance is calculated from multiple linear regression estimates with predictors of monthly temperature, wind and radiation. The first 5 graphs show how the PET explained variance changes with the predictor used to represent wind. Three different wind measures are used: windrun at the site, a Z1 pressure difference index, and Z1 and M1 indices together. Each panel shows a different site: Napier, Masterton, Blenheim, Lincoln and Dunedin. Explained variance ranges between 20%-80% through the year. Except for the short-record Dunedin site, there is no systematic advantage in using the site-specific windrun over the New Zealand-wide Z1 pressure index. Adding M1 as an additional wind predictor gives very little improvement to the explained variance.
The sixth panel applies to Lincoln and shows how the PET explained variance changes with the predictor used to represent radiation. Three measures are used: radiation at the site, rainfall, or "none" (i.e., only use default predictors for temperature and wind). As expected, using radiation as a predictor is best, but the improvement over other measures as predictors is generally smaller than the seasonal variation in explained variance.
