Performance of bias corrected monthly CMIP6 climate projections with diferent reference period data in Turkey

Decisions that are based on the future climate data, and its consequences are significantly important for many sectors such as water, agriculture, built environment, however, the performance of model outputs have direct influence on the accuracy of these decisions. This study has focused on the performance of three bias correction methods, Delta, Quantile Mapping (QM) and Empirical Quantile Mapping (EQM) with two reference data sets (ERA and station-based observations) of precipitation for 5 single CMIP6 GCM models (ACCESS-CM2, CNRM-CM6-1-HR, GFDL-ESM4, MIROC6, MRI-ESM2-0) and ensemble mean approach over Turkey. Performance of model-bias correction method-reference data set combinations was assessed on monthly basis for every single station and regionally. It was shown that performance of GCM models mostly affected by the region and the reference data set. Bias correction methods were not detected as effective as the reference data set over the performance. Moreover, Delta method outperformed among the other bias correction techniques for the computation that used observation as reference data while the difference between bias correction methods was not significant for the ERA based computations. Besides ensemble approach, MIROC6 and MRI-ESM2-0 models were selected as the best performing models over the region. In addition, selection of the reference data sets also found to be a dominant factor for the prediction accuracy, 65% of the consistent performance at the stations achieved by the ERA reference used bias correction approaches.

---

Opracowanie rekordu ze środków MEiN, umowa nr SONP/SP/546092/2022 w ramach programu "Społeczna odpowiedzialność nauki" - moduł: Popularyzacja nauki i promocja sportu (2022-2023).