Minimizing the uncertainties of RCMs climate data by using spatio-temporal geostatistical modeling

The spatio-temporal riging approach by using five different covariance models, has been applied into Regional Climate Model (RCM) simulated precipitation and temperature dataset in a coastal area. The results of the spatio-temporal technique were evaluated against the ERA-Interim reanalysis data during the period from 1981 to 2000. The reliability of the spatio-temporal interpolation results were estimated by using both the judgment of the wireframe plots, between the sample and the fitted covariance models, and the statistic metrics. Thus, Taylor diagrams were used and the Mean Square Error (MSE) was calculated. The analysis demonstrates that the sum-metric covariance model is highly superior to the other four covariance models as it is closer to the reanalysis data, having the highest correlation coefficient, as well as, the smallest standard deviation, resulting in the smallest Root Mean Square Error. The spatio-temporal interpolation approach improved the MPI and HadGEM2 climate model dataset. The largest enhancement is pointed out in the interpolated RCM precipitation during winter and autumn. Concerning the temperature, the interpolated MPI temperature data is negligibly improved, whereas the interpolated HadGEM2 temperature is particularly optimized during winter and autumn. The spatio-temporal interpolation technique led to the minimization of the uncertainties of the Regional Climate Models, (RCMs) simulations, and also to the best agreement between them and the ERA-Interim reanalysis data during the period from 1981 to 2000. Nevertheless, the MPI climate model is more reasonable compared to the HADGEM2 for the research area.