ESTIMATION OF MAXIMUM ANNUAL DISCHARGES BY FREQUENCY ANALYSIS WITH FOUR PROBABILITY DISTRIBUTIONS IN CASE OF NON-HOMOGENEOUS TIME SERIES (KAZANI KARST SPRING IN REPUBLIC OF MACEDONIA)

Frequency analysis is usually used in hydrology for the possibility and scale of discharge extremes, especially low flows or flood features of concrete gauge. The paper presents a frequency analysis carried out on a 51 year period (1961-2011) series of maximum annual discharge, on Kazani karst spring gauge station. The time series were non-homogenous according to Cumulative deviations test (Q = 1.531) and Standard normal homogeneity test ( = 10.543). After adjusting of annual maximum discharges (AMD) before the year of change (1995), with using AnClim software, the frequency analysis was made using four probability distributions. Obtained results show possibility of karst spring flood occurrence for different return periods. After their calculation the best fitting distribution was chosen using graphical and statistical testing between empirical and theoretical data distribution. According to the spring significance for nature and humanity, it was crucial to choose a right probability distribution for annual maximum discharges of Kazani karst spring. Therefore, a comparison using graphical and statistical testing (Kolmogorov-Smirnov, chi(2) and Anderson Darling test) was made. The best fit according to three tests was the Gumbel distribution with AMD of 4.54 m(3)/s for return period of 10 years, 6.06 m(3)/s for return period of 100 years and 7.56 m(3)/s for return period of 1000 years, respectively.