Statistical downscaling of multi-decadal climate change projections: Bridging the gap between climate models and the end-user

Abstract

Multi-decadal regional climate projections are assimilated into a statistical model in order to produce an ensemble of mid-summer maximum temperature for southern Africa. The statistical model uses atmospheric thickness fields (geopotential height differences between the 500 and 850 hPa levels) from high-resolution reanalysis data as predictors in a perfect prognosis approach in order to develop linear equations which represent the relationship between atmospheric thickness fields and gridded maximum temperatures across the region. The statistical model is found to be able to replicate the increasing maximum temperature trends of the driving regional climate model. Since dry-land crops are not explicitly produced by climate models but are sensitive to temperature extremes, the effect of these projected maximum temperature trends is assessed on dry-land crops over multiple decades by employing a statistical approach similar to the one introduced for maximum temperatures.