CR 34:195-210 (2007)  -  DOI: https://doi.org/10.3354/cr00693

ABSTRACT: Empirical studies show that observed frequencies n_P for a daily precipitation amount P can be approximated by n_P ∝ e^mP, with negative values for the exponential coefficient m. The parameter m describes the mean (μ) or any quantile for variables following such exponential distributions. The coefficient m varies from location to location, and exhibits a systematic relationship with local mean temperature and precipitation as well as other geographical parameters. A general linear model can be utilised to model μ directly from local climate conditions and geography. Estimates of m and μ from local climatic conditions allow an estimation of extreme values in the form of high percentiles. Given changes in the mean local climate, it is possible to infer changes in the upper percentiles. A new multi-model ensemble of the most recent climate simulations, carried out for the Intergovernmental Panel on Climate Change fourth assessment report (IPCC AR4), has been subjected to empirical-statistical downscaling, and provides best estimates for the continuing trends in mean temperatures and precipitation in northern Europe. These scenarios are used in conjunction with (1) the established relationship between the exponential coefficient m on the distribution function, and (2) local mean temperature and precipitation for 2050, to infer changes in the 95th percentiles of the rainfall for 2050. Two new independent analyses point towards an increase in the number of extreme precipitation events and a slight change in the number of rainy days over large parts of the Nordic countries. The projected increase was found to be sensitive to to the choice of predictors used to model the geographical dependency, rather than to the choice of method.

KEY WORDS: Climate scenarios · Empirical-statistical downscaling · Probability density function · PDF · Extremes