Alternative model for precipitation probability distribution: application to Spain

Abstract

In climatology, there is difficulty in describing the probability distribution of rainfall because there are many days without precipitation, which usually causes the most likely daily rainfall to be zero. None of the widely used models is able to describe the overall variation of daily precipitation. This article proposes an alternative model for the probability of precipitation. The model, based on 4 parameters, has been applied to daily rainfall throughout all months of the year and for 108 stations in Spain. This alternative model provides better results than the commonly used probability models (Generalized Extreme Value, Pareto and Generalized Pareto Distribution, Gamma, Gumbel, Weibull, Exponential and Log-normal). Our model had a mean absolute error of <10% for most of the stations analyzed. Thus, this alternative model could be used to correct the probability distributions of daily precipitation obtained from weather forecasting and climate models.