A. Markopoulos, Statistical analysis of floods on a global scale, Diploma thesis, 115 pages, Department of Water Resources and Environmental Engineering – National Technical University of Athens, Athens, July 2017.
Major scientific interest is raised in statistical analysis of large floods on global or/and regional scale, especially in the framework of a changing climate. A global data base of flood discharges (World’s Catalogue of Maximum Observed Floods, IAHS, 2013) is processed. Based on length and quality criteria, 125 stations worldwide are chosen for the analysis. Emphasis was given to trend detection, and therefore a thorough research was carried out to each station individually to locate major dams and reservoirs upstream of the stations. For the trend analysis two methods have been applied, linear regression and Mann-Kendall trend test. For stations that were strongly correlated, the Mann-Kendall trend test was modified with a specific methodology that is being proposed. Autocorrelation and Hurst parameters were used to quantify the correlation structure of the data set. Hurst parameter was calculated by applying the aggregated variance method. Four different probability distribution functions, belonging to the extreme value family, were fitted to all timeseries. The goodness-of-fit test that were used, were the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC) and a modification of the well-known Anderson-Darling Criterion (ADC).