Identification of possible risks to hydrological design under non-stationary climate conditions

Traditional practices using probabilistic methods resorted in designing hydraulic structures commonly assume that hydro-meteorological time series are stationary over time. The prevailing thought has been challenged due to the deterioration of the natural function of hydrologic cycle. It is an inevitable fact that the estimation of the design parameters under non-stationary conditions could prevent the expected function of the hydraulic structure from being performed successfully. This study was conducted to come up with how the regional frequency distribution behavior of maximum rainfall with 6-, 12-, and 24-h duration in the Euphrates–Tigris river basin would differ under non-stationary conditions. According to the ITA (innovative trend analysis) approach, a statistically significant trend was detected in approximately 89% (61% increasing and 28% decreasing) of the maximum rainfall sequences of the stations in the region. In order to reveal the variability in the regional frequency distribution of the maximum rainfall series, the L-moments algorithm was applied to the full (F) rainfall datasets and the first and second rainfall sub-series (FH and SH) formed by the ITA method. In this comparison process based on the PITA (probabilistic innovative trend analysis) approach, considerable differences were detected in the L-moment statistics of the F, FH, and SH data sets and the quantile estimates for some risk levels. In particular, the findings associated with the SH data series differed markedly when compared to those of the F and FH datasets.