Statistical Downscaling of New Zealand Climate

A statistical downscaling system has been developed for New Zealand daily climate variables, based on the perfect prognosis approach. Variables were daily precipitation, maximum, minimum and 10cm earth temperature, solar radiation, and wind run. A number of binary event-probability variables were also developed from the continuous temperature and rainfall values. A set of diagnostic regression equations was developed for 114 New Zealand sites, using as predictors 1000 hPa height and 850 hPa temperature fields from a long time series of reanalyses. A useful level of diagnostic skill was found in most cases. Explained variances, for anomalies from the seasonal cycle, were around 50% for temperatures and wind run, and 30-40% for precipitation and solar radiation. Tests on an independent data set suggested the downscaling system is relatively robust and reliable. Statistics of the fit reflect the geography of New Zealand and the strong forcing of its climate by orographic influences and interplay with the mean westerly circulation. The resulting downscaling system can be applied to a variety of model output, including climate model simulations and largescale numerical weather prediction model runs.