Flood Frequency Analysis of River Swat Using Easyfit Model & Statistical Approach

Abstract

Pakistan has faced a tragic and massive flood in 2010. The climate change is considered a major factor for such a devastating and severe monsoon. The widespread precipitation generated very high runoff in Indus, Kabul, Swat Chenab and Jhelum rivers. Swat River is a major tributary of the Indus Basin River system, located in between the foothills of Hindukush mountain range also known for its snowcapped peaks. The contribution of snowmelt, average ground water and average rainfall in the basin is 65%, 19% and 16% respectively. Average annual rainfall in this region is more than 1000 mm. The dominant sources of rainfall are westerlies and monsoon which contribute 45% and 55% respectively. The sharp flow peaks are generated due to lack of surface storage capacity and non-absorption of runoff in the catchment area of River Swat. Therefore floods are common in this basin, which portrays threat both to infrastructure & humans. The unprecedented flooding in 2010 destroyed Munda & Amandara Headworks at River Swat due to much higher peak flood than the design discharge. High flow velocities and flow energy instigated erosion and also damaged infrastructure such as roads, houses and bridges. Due to undulated terrain and accessibility, there is deficiency of precipitation and discharge recording stations. Consequently measures to avert losses from such events cannot be taken well in time. In this study, flood frequency analysis of River Swat at Chakdara Station was carried out using Gumbel’s Extreme Value (Type-1), Normal, Log Normal and Log Pearson Type III distributions against 2, 5, 10, 20, 25, 50, 100, 500, 1000, and 10000 year return periods to estimate the occurrence of such flood events. Significance tests, Anderson Darling, Kolmogrov Smirnov and Chi-Squared were applied in order to assess the most effective Probability Distribution for the study area. It was observed that statistical distributions helped considerably in estimation of floods at sites of homogeneous regions with less or no data. The results of the study, based on the applied significance tests i.e. Kolmogrov Smirnov, Anderson Darling and Chi-Square, exhibited that Gen Extreme Value [Type-1] ranked superior. Therefore it is the best-fit distribution among the other applied distributions for the Study area. Considering the changes in the behavior of streams and patterns of flooding, it is recommended to review the criteria and design limits for structures in urban & rural areas as well as of river training works.