Extreme value analysis and forecasting of maximum precipitation amounts in the western Black Sea subregion of Turkey

Abstract

Monthly maximum precipitation amounts for the period 1950–2010 were modelled for seven climatological stations in the western Black Sea subregion of Turkey using a distributional and time series analysis approach. First, the generalized extreme value (GEV) distribution was fitted using the location parameter of the GEV distribution as a function of several explanatory variables that affect the maximum precipitation. We quantified the change in extreme precipitation for each location and derived estimates of return levels for monthly precipitation amounts. We also considered a multivariate time series model with exogenous variables using the selected locations for a short‐term forecast of the maximum precipitation amount. It was discovered that the time series model that takes into account the autocorrelation structure of the series performed better than a probabilistic approach using the extreme value theory. The obtained forecast results are promising in terms of accurately defining future precipitation amounts.