Statistical Analysis of the Maximum Annual Rainfall Data at Al-Shuwaija Marshes in Wasit Governorate, Iraq

Five statistical distributions were used in this study to model the maximum annual rainfall at five metrological stations in and around Al-Shuwaija Marshes in Wasit governorate, Iraq. These stations are Badra, Sumar, Kut, Ilam, and Mehran. Selecting the best probability distribution is based on the test of goodness-of-fit using the Kolmogorov-Smirnov test. Data was collected from 2000 to 2020 from Iraqi and Iranian metrology authorities. This data is subdivided into two sub-samples: the first is from 2000 to 2013 for building the model, and the second is from 2014 to 2020 to validate the model. The Kolmogorov-Smirnov test results showed that for some stations, there was more than one frequency distribution that fit the data as they had the same minimum SIG value. In these cases, the t-test and F-test results were used as criteria to aid in selecting the best-fit distribution. For the Badra station, the Log-Normal frequency distribution (61.663, 0.502) was the best-fit distribution with the SIG value of 0.998. For the Sumar station, three distributions had the same SIG value (0.998) Normal, Log-Normal, and Gamma, with the best-fit for Log-Normal (69.008, 0.593) selected according to the t-test 0.146, 0.024, and 0.093 between a three generated data using each of these distributions and the observed one. The same observation was found for the Kut station, except that the best-fit distribution was the Normal distribution (55.746, 30.7323) with a SIG value of 0.879. As for the Ilam station, the results indicated that the Normal and the Gamma distributions had the same SIG value; the last was selected as the best one according to the t-test results 0.081, 0.010. and 0.263. for the Mehran station, the Normal distribution was the best one that fitted the maximum rainfall data. For each station, three series were generated using the model. Each generated series was tested against the observed using the validation data period. The Kolmogorov-Smirnov test indicated that the distribution could generate data drawn from the same population as the observed data with respect to frequency.