Sensitivity analysis of flood frequency to the length of the data recording period

Abstract Floods are the significant natural events that can have profound impacts. This study examines how the length of data records affects the sensitivity of continuous probabilistic models applied in the flood frequency analysis and record periods were analyzed. The FREQ software was utilized to conduct the flood frequency analysis for the Karkheh watershed, Iran. The results revealed that the stations were categorized into two regions. In the first region, as the length of the record period increased, the coefficient of variation, skewness, and kurtosis decreased. In contrast, in the second region, these parameters increased with the extension of the record period. In terms of geomorphology, it was found that the average catchment area in the regions with decreasing parameters is larger, while the elevation in these areas is lower. This suggests that the decreasing areas are situated at the lower latitudes, lower-elevation and downstream portions of the basin, with the influence of dams also playing a significant role. In this part of the study, the 2-parameter and 3-parameter log-normal distributions exhibited a better fit compared to the other distributions. It was concluded that 46.7% of the stations showed the uncertain or unstable trends for the extreme return periods, 40% of the stations displayed a consistent trend across all of the return periods, while 13.3% of them showed an uncertain trend for the return periods greater than 50 years.