Obtaining sub‐grid‐scale information from coarse‐resolution general circulation model output

The relationships between local temperature and precipitation and large‐scale climate are explored using regression analysis. The motivation for this study is the need of the impact analyst for small‐scale information given only coarser resolution General Circulation Model output. The predictor variables employed are area averages (over ∼2.5 × 10 6 km 2 ) of temperature and precipitation and propinquitous grid point values of mean sea level pressure and 700 mbar height, together with the zonal and meridional gradients of these two variables. Regression analyses are performed using monthly‐mean data from Oregon, with separate analyses for each month. In independent verification, spatial‐mean explained variances range from 58 to 87% for temperature and from 39 to 76% for precipitation. Most of the variance explained arises from the area average of the variable which is the predictand: in other words, if the temperature, say, at point x is to be estimated, the best predictor is generally the area average temperature. There are large spatial differences in the amount of local climate variance that can be explained by large‐scale data. Examples are given which show how site‐specific changes can differ markedly from those at the grid point scale.