Evaluation of various regionalization techniques for the seasonal precipitation IDF estimates of Baden Württemberg

This research addresses a strong need to precisely improve the statewide seasonal precipitation intensity duration frequency (IDF) estimates at ungauged locations. In order to obtain IDFs at unvisited sites, IDFs observation locations are interpolated. Therefore, different deterministic and geostatistical approaches include Inverse Distance Weighted (IDW), Ordinary kriging (OK), Regression Kriging (RK), Co-Kriging (CoK), Kriging with External Drift (KED), and Functional Kriging (FK) have been taken into account for comparison. Apart from visual assessment, a cross-validation approach is used to compare these methods to judge their prediction accuracy.

Annual or intra-annual IDF calculations across the state is not well correlated with other variables except elevation, thus directionally smoothed altitude is only considered as a covariate that offered a significant reduction in bias.  All results indicate that IDW interpolation is incapable of improving the regional point IDF approximations provided by kriging algorithms except in the case of annual IDF predictions at shorter scales where its performance is more or less similar to OK.  Whereas summer IDF observations are well predicted by KED that also exhibits good behavior for longer duration extremes of all seasons. Moreover, the shorter duration winter IDF guesstimates are best achieved with CoK. From now, it can be noticed that the accuracy of the interpolator changes according to the hydrological seasons and storm durations.

Overall, this study ensures to design of a well-planned map in advance for the entire state of Baden Wurttemberg on the basis of accurate forecasting of seasonal IDF estimates of precipitation extremes at unsampled sites. Hence, this crucial step will surely help us to tackle the natural disasters due to climate change before time.