Evaluation of selected bias correction methods for the development of local climate projections in Slovakia

Regional and local climate projections are important for climate change impact research. However, the outputs of global and regional climate models show deviations from observations on regional and local scales. Generally, the deviations increase with a larger difference between model and observation scales. These deviations hinder the usage of climate model output for practical applications. To improve agreement of the model outputs to observations without the necessity of running a higher resolution climate model, statistical downscaling methods are utilized.

One widely used group of methods are bias correction methods, which are relatively simple, computationally undemanding, and only adjust the model output. However, both bias correction in general, and individual bias correction methods have limitations, and their application may not always be appropriate. Therefore before using any bias correction method to obtain climate projections, it is first necessary to evaluate the performance of the method, and also to determine the effect it has on the simulated climate change signal.

In this work we evaluated four existing bias correction methods. Two of these methods perform a correction by a factor (additive for temperature, scaling for precipitation). The other two are quantile mapping methods, one of which was specifically designed to preserve the simulated climate signal. We applied the methods to daily average, maximum and minimum air temperatures, and daily precipitation totals, which were simulated by an ensemble of regional climate models from the EURO-CORDEX framework. Before applying the methods, the model outputs were interpolated to locations corresponding to meteorological stations.

Based on the evaluation, we selected methods for air temperature and for precipitation that achieved the best results in terms of both 1) effective reduction of bias, and 2) low modification of the simulated climate signal. Using these methods, we then created local projections of air temperature and precipitation until the year 2100 for selected meteorological stations in Slovakia.