On the choice of reference database and calibration period of bias-corrected simulations: A case study for Hungary

The aim of the present study is to investigate the accuracy of bias-adjusted regional climate model (RCM) simulations using various calibration periods, demonstrated for the region of Hungary. High-resolution (0.11°) RCM simulations of daily near-surface mean air temperature, daily minimum and maximum air temperature, and daily precipitation provided by the EURO-CORDEX community are analysed. The model ensemble consists of 5 RCM simulations driven by 4 different general circulation models for the historical time period 1976–2005. The publicly available, most accurate, measurement-based and quality-controlled HuClim is used as the reference dataset. The internationally widely used percentile-based quantile mapping method is applied for the bias-correction and it is performed on a monthly level. The novelty of the present study is that we used two different calibration periods to create bias-corrected datasets: an earlier and a more recent 30-year long period, and made these new datasets available in Zenodo. In addition to these HuClim-based bias-corrected databases, another database, containing bias-corrected RCM simulations and produced by the EURO-CORDEX community is also investigated. The assessment is carried out for the period 1993–2005, which is the overlapping time interval of the different calibration periods. According to our results, the accuracy of the bias-correction depends on the chosen calibration period and on the analysed climate index, and the choice of the validation period also affects the results. As next step, we plan to extend our research on projections under RCP4.5 and RCP8.5 scenarios.