Improving a method for the determination of outliers in the hydroligical data and applying on the Euphrates and Tigris basins

Flood frequency analyses have a major importance for designing and projecting the bridge, dam, weir, culvert, channel and the other water structures. For a given return period estimation of flood flows by statistical methods is done by flood frequency analysis.In recent years, because of the changes caused by global warming on seasons and the conventional rains outliers are encountered between the measured data. As the sample data?s standard deviation and average are being raised by these outliers, the flood flows found as a result of the flood frequency analysis made by classical methods are very high and meaningless. In fact, the statistical values such as mean, standard deviation and skewness coefficient of the station data of the Euphrates and Tigris basins are calculated and it has been seen that these values are high in most of the stations due to the presence of outliers. In addition, using these stations? data, although the differences between the results of the flood discharges calculated by the most appropriate probability density functions designated by the goodness of fit tests and the proposed method are usually relatively low until 1000 year period, the differences reach to critical levels in 10000 and 20000 year periods. However, at the method we have developed the sample data?s standard deviation and average aren?t used. For the sample data ?-T graph is drawn and then with the help of this graph, outliers detected and cleaned out from the sample. Then, the flood flows were calculated in accordance with the natural slope of the drawn graph. As a result there seen the flood flows calculated by our recommended method are more reasonable and logical values. The water structures will be designed according to these calculated values will prevent to be unnecessary largeness size, and thus to make an economical design will be provided.Also in this study, from the classical distribution functions we have used Gumbel Max. and Log Pearson 3 distributions give close results to our recommended method?s, at least seen the effect of outliers and determined that this distributions were the most appropriate for the Fırat and Dicle basins.Keywords: Flood Frequency Analyses, Goodness of Fit Tests, Probability Density Functions, Outliers.