Prediction of Maximum Precipitation in Turkey with Polynomial Regression

Abstract

Global warming is increasing the amount of atmospheric water and the magnitude of precipitation in some regions even though the annual total precipitation trend is decreasing. Consequently, the prediction of future maximum precipitation based on existing observations has gained importance. Understanding the expected variability in precipitation events and predicting the location and yearly periods of probable extreme precipitation are important for the efficient prevention of potential natural catastrophes like floods and dam failures. In this study, future extreme precipitation was predicted for each month of year at a 95% confidence level by using the precipitation data of 66 stations in Turkey. In the context of the developed method, the observed precipitations of each station for each month are sorted in ascending order and the best polynomial fitting to the data is determined, then the expected extreme precipitation values higher than all previous observations were determined by extrapolation. The results of the software developed for predicting the expected maximum precipitation values by determining the degrees and coefficients of the best-fitting polynomials for each month at each station show the months and regions with extreme precipitation and risks of flooding. Also, maps showing the predicted monthly extreme precipitation events throughout Turkey are generated for each month. The PolReg software developed as part of the study is freely presented together with the manuscript for readers' use.