Probability Function Estimation for the Maximum Precipitation Model Using Kernel Estimators

Abstract

The distribution of maximum rainfall level is not a homogeneous phenomenon and is often characterized by multimodality, and often the phenomenon of heavy right-hand tail. Modelling this phenomenon using classic probability distributions leads to ignoring multimodality, and thus underestimating or overestimating the predicted values in the tail tails – the most important from the point of view of safe dimen-sioning of drainage systems. To avoid the abovementioned difficulties, a non-parametric kernel estimator method of maximum precipitation density function was used (on the example of rainfall data from a selected station in Poland). The meth-odology proposed in the paper (for use on any rainfall data from other meteorologi-cal stations) will allow the development of more reliable local models of maximum precipitation.