Regional frequency analysis of maxima rainfalls with different intensity via L-moments aproach

In this study, regional frequency analysis of annual maximum rainfalls (AMR) at standard durations are carried out through methods of L-moments in Euphrates-Tigris Basin. In the first stage, homogeneity status dealing with AMR series from 18 rainfall stations in Euphrates-Tigris Basin were tested by using Mann-Whitney U (MWU) test. Based on the MWU test results, homogeneity condition for the 5% of significance level was fulfilled for AMR data of eight stations in different standard durations. Homogeneity of these data sequences was performed with double mass curve (DMC) method. In the second stage, regional frequency analysis associated with L-moments approach was applied to the AMR data both before and after the homogeneity analysis. In both conditions, it was first analyzed whether all stations fulfilled the homogeneous region condition. After determining that this choice could not be achieved, analysis was carried out on the double and triple alternative subgroups determined by clustering analysis. However, for some AMR series, a single homogeneous region condition has been met. The AMR data 5min, 15min, 24h before DMC application were assumed as a whole homogeneous region, whereas the AMR data dealing with 5min, 15min, 30min, 1h, 6h, 12h, 24h after the DMC application was formed as a single homogeneous region. In the condition unless DMC analysis, the AMR data sequences of 1h, 2h, 3h, 6h, 12h were carried out condition of regional frequency analysis based on L-moments method in the two sub-regions while the AMR data of 2h, 3h in the condition where the DMC is applied produced a satisfactory results for regional frequency analysis in two sub-regions. Then, regional probability distribution functions are defined for each homogeneous region. At the last stage, the best IDF relationship that would represent the region was developed with the Genetic Programmer (GP) in which were taken into considered rainfall intensities for different return period obtained depending on the most suitable regional probability distribution. It was concluded that the IDF relationship developed using the data after the DMC analysis would better represent the region.