Performance of Archimedean copula functions in annual flood estimation, Case study: Qarah-Soo Watershed

Flood is known as one of the most devastating natural hazards which cause great damages to human societies, municipal, industrial and agricultural centers. Flood estimation in confluence points of rivers– for being the location for many infrastructures – due to economic and environmental matters receives a great importance. It is possible to estimate the flooding likelihood by examining the number of floods with specified return period. In this study, according to the statistical data of peak flows between 1358 to 1379 Hijri-Shamsi in the basin of the Qarah-Soo in Kermanshah Province, annual peak flows were estimated using copula in four models of 1) regression 2) summation 3) weight factor and 4) the nearest stations. For this, first, homogeneity and data quality were investigated. The probability distributions were fit to the data series and with regard to marginal distribution functions of upstream stations, joint distribution function at the confluence was obtained by the Archimedean copula functions. Based on the best copula function for upstream stations, Gumbel copula function was selected. Results showed that copula function in the form of regression was superior to the other models with a coefficient of determination equals to 0.711, RMSE equals to 79.387, Kendall's tau correlation coefficient equals to 0.872 and Spearman Rho coefficient equals to 0.677. Eventually, discharge amounts for different return periods were calculated according to the selected model.