Selection of representative near-future climate simulations by minimizing bias in average monthly temperature and precipitation

Abstract

The bias in the global and regional climate models signi cantly complicates their use in impact studies. A signi cant difference between the observed and model precipitation in the warm months is registered in Odesa for 1970-2005. This difference is probably determined by complex orography and inappropriate parameterization methods for convective processes climate models. In the last fteen years, the average temperature has increased by about 1°C in winter and by 2°C in summer compared with 1970-2005.Considering decreasing precipitation during summer months, it seems that the climate of Odesa is moving towards the Mediterranean climate -warm to hot, dry summers and mild, moderately wet winters. The approach based on selecting representative simulations with minimum average bias and adjusting the choice to the present-day climate is described and applied for Odesa using data from the RCP8.5 scenario simulations of the EURO-CORDEX project and ERA5-Land reanalysis. The approach can be applied separately for monthly near-surface temperature and total precipitation, as well as jointly for these variables, and provides the satisfactory ability to select models for use then in impact studies. The output variables of simulations selected are close to observed ones in recent years and are well to coincide with the ensemble-mean values in the near future, 2021-2050. On the other hand, the scatter of output variables in the selected simulations adequately describes the uncertainty of the future climate.