How Reliable Are Selected Methods Of Projections Of Future Thermal Conditions? A Case From Poland

The aim of the paper was to assess the robustness of four bias correction techniques: simple bias correction, distribution based bias correction, delta change and distribution based delta change. Data from nine RCM simulations of CORDEX project and 41 Polish weather stations were used. The methods were calibrated in the period 1971–1985 and evaluated in 1991–2005. The improvement in mean, 10^th and 90^th percentiles was shown, without significant differences among methods. For 1^st and 99^th percentiles the improvement was generally weaker and simple methods seem to be more robust than the distribution based ones. Strong differences between individual models were found, so the use of model ensemble is recommended.