Weather conditions could affect widely both energy supply and demand balance in Europe. To explore relevant and sufficiently contrasted scenarios for Generation Adequacy Studies, Meteo-France operated ARPEGE-Climat climate model to create a reference framework including 3 scenarios of 200 yearly simulations of weather conditions at current climate (fixed ~2000), and at future climate (fixed ~2050) according to the hypothesis of RCP 4.5 and RCP8.5 concentration evolution. The simulations are available on a 20 km resolution grid covering Europe and North Africa. A bias adjustment post-treatment, based on HIRLAM re-analysis1, was applied to all the considered parameters to suppress some biases and get relevant extreme values. The available data include especially temperature, wind, precipitation and solar radiation. Each simulated year might be interpreted as a possible realization of the target climate.
For wind, precipitation and solar radiation, bias adjustment was done thanks to a classical quantile-mapping or anamorphosis method. A particular attention was given to the 2m temperature, to remove model bias with innovative statistical approach concerning the extreme values. This approach is based on the estimation of absolute maximum and minimum temperature thresholds that cannot be overshot at current climate, for each grid point.
Bias adjustment function for current climate is done for every day (1 to 366) at 00, 06, 12 and 18 UTC, considering a 15 days sliding window centered on the considered date. The 30 years HIRLAM distribution function is stretched (extension of tail of distribution) to reach estimated 200 years temperature HIRLAM limits. Then a transfer function (anamorphosis) is defined from simulation to reanalysis and applied to the simulated data for bias adjustment.
For the correction of future climate simulation bias, for each grid point of the current climate simulation, a multiple linear regression between the corrected simulation and the values of the neighbouring grid points of the raw simulation is adjusted. The obtained estimator is corrected for its error from the knowledge of the residuals of the regression (which gives exactly the corrected simulation). This statistical model is applied to the raw values of future climate simulation.
Based on this corrected data, a derived temperature dataset for current and future climate was elaborated for European cities location, including a regression between the observations and the HIRLAM analysis .
1 http://www.euro4m.eu/downloads/Factsheets/EURO4M_Factsheet_D2.03_SMHI_HIRLAM_3D