Evaluation of the Effect of Bias Correction Methods on the Skill of Seasonal Precipitation Forecasts of CFSv2 Climate Model

Bias correction methods are one of the most common statistical post-processing methods which are utilized on the output of climate models. This study evaluates the effect of five bias correction methods on the skill of seasonal precipitation forecast (fall season) from the CFSv2 climate model based on 12 stations located in Gorganrud basin in Iran. Bias correction methods that have been used in this study consists of two non-parametric methods (Linear Scaling (LS), Empirical Quantile Mapping (EQM)), one parametric method (Power Transformation (Ptr)), and two parametric methods based on the statistical distribution (Parametric Quantile Mapping (PQM), Generalized Parametric Quantile Mapping (GPQM)). Various metrics have been used for evaluating the effects of these methods on the skill of seasonal precipitation forecast which consists of bias, Pearson correlation coefficient, ranked probability skill score (RPSS), and the relative operating curve skill score (ROCSS). The Results of this study revealed that most of bias correction methods decreased the biases of the raw forecasts. The effect of each bias correction method on the RPSS and ROCSS (below and above normal events) scores may vary based on location and time, and each method can improve or worsen these two scores based on location and time. The results of this study suggest that the evaluation of various bias correction methods and distinguishing the most suitable method based on the goal of each study would be helpful in the improvement of seasonal precipitation forecast skill.