Bagher Heidarpour; Bahram Saghafian; Abolfazl Shamsai; Saeed Golian
Abstract
In present research, the exceptional floods of study area at Golestan Dam watershed were identified using one of the standard tests of identifying outlier values, e.g. Dixon, Grubbs and Grubbs and Beck.In order to determine the probability distribution function and the effect of direct application of ...
Read More
In present research, the exceptional floods of study area at Golestan Dam watershed were identified using one of the standard tests of identifying outlier values, e.g. Dixon, Grubbs and Grubbs and Beck.In order to determine the probability distribution function and the effect of direct application of the floods in flood frequency analysis, two sets of analysis were performed one with the whole data series and the other with deleting the outliers. At this stage, 15 different probability distribution functions were applied with biased and unbiased estimates of the parameters using three estimation methods namely method of moments, maximum likelihood and Probability Weighted Moments methods. According to Kolmogorov-Smirnov and Chi-square goodness-of-fit tests and index error calculation, the log Pearson Type III distribution was determined as the best distribution for both complete data series and data without exceptional floods. Results of the Flood frequency analysis at this stage showed that the probability distribution did not change after removing the exceptional floods, but was highly influential in the magnitude of design flood.In the next step, the method of Water Resources Committee of the United States was used to combine the exceptional floods with other observational data. Using the sensitivity analysis, the historical period for exceptional floods at Tangrah, Tamer and Galikesh stations were calculated to be 300, 60 and 80 years, respectively. The derived historical period were considered in calculation of the flood values for different return periods. For example, 1000-year design flood for the complete data series were calculated to be 14946, 1639 and 2635 cms for the aformentioned stations respectively and reduced to 1434, 1423 and 1296 for the complete data series with modification of the parameters for the probability distribution functions. Therefore, the technique used in this study effectively reduced the cost of the designed hydraulic structures.
