Downscaled Climate Projections
In some studies, it is assumed that one should only use large-scale predictors that are well simulated by the models. This would mean that instead of using precipitation from the climate models (which is relatively poorly simulated) to predict local precipitation, one should use geopotential height or sea level pressure from the climate models to predict local precipitation. We take the view that by far the best predictors to use are the predictors directly related to the downscaled variables. So, for example, one should use modelled precipitation to predict local precipitation and modelled maximum daily temperature to predict local maximum daily temperature. As noted in other studies, one advantage of using the "direct" predictors is that they are usually by far the most skillful predictors for downscaling the current climate. Perhaps an even bigger advantage, however, is that for predicting the future, the direct predictors are least likely to suffer the pitfalls that can easily be encountered when using indirect predictors.
For our downscaling we use "direct" predictors for each of the three downscaled variables: precipitation, maximum daily temperature, and minimum daily temperature. Particularly for temperature, the direct predictors are nearly always the most important predictors. The direct predictors have the least skill for summer precipitation. We also use additional predictors besides the direct predictors to potentially improve skill in processes unresolved by the large-scale climate models. These other predictors are discussed in the next section.