Parametric or non-parametric: analysis of rainfall time series at a Hungarian meteorological station

Parametric methods (linear trend, t-test for slope) for analyzing time series are the simplest methods to get insight to the changes in a variable over time. These methods have a requirement for normal distribution of the population that can be a limit for application. Non-parametric methods are distribution-free methods, and investigators can have a more sophisticated view to the variable tendencies in lime series. 144-year-long time series of precipitation data measured at the meteorological station in Keszthely, Hungary (latitude: 46 degrees 44', longitude: 17 degrees 14', elevation: 124 m above Baltic sea level) were analyzed by Mann-Kendall trend test for detecting tendencies in the time series. Sen's slope estimator was applied to estimate the slope of the linear changes. In average. 44 mm decline can be shown for 100 years in the annual sum, 29.7 mm and 25.7 mm in the precipitation sum of spring and autumn (in 100 years), respectively. The rainfall sum of winter increased by 15.4 mm. Sums of April, May, and October declined by 10.8 mm, 13 mm, and 20.9 mm, respectively, according to one-tailed Mann-Kendall tests. These results were compared to the previous results of the authors carried out by parametric methods. Results of two-tailed tests of parametric and non-parametric methods are easily comparable. Parametric method (linear trend) proved significant decreasing tendencies for spring, April, and October. Nonparametric Mann-Kendall tests show significant declining tendencies for spring, autumn, and October.