Characterizing and modeling seasonality in extreme rainfall

A comprehensive understanding of seasonality in extreme rainfall is essential for climate studies, flood prediction and various hydrological applications such as scheduling season‐specific engineering works, intra‐annual management of reservoirs, seasonal flood risk mitigation and stormwater management. To identify seasonality in extreme rainfall and quantify its impact in a theoretically consistent yet practically appealing manner, we investigate a dataset of 27 daily rainfall records spanning at least 150 years. We aim to objectively identify periods within the year with distinct seasonal properties of extreme rainfall by employing the Akaike Information Criterion (AIC). Optimal partitioning of seasons is identified by minimizing the within‐season variability of extremes. The statistics of annual and seasonal extremes are evaluated by fitting a generalized extreme value (GEV) distribution to the annual and seasonal block maxima series. The results indicate that seasonal properties of rainfall extremes mainly affect the average values of seasonal maxima and their variability, while the shape of their probability distribution and its tail do not substantially vary from season to season. Uncertainty in the estimation of the GEV parameters is quantified by employing three different estimation methods (Maximum Likelihood, Method of Moments and Least Squares) and the opportunity for joint parameter estimation of seasonal and annual probability distributions of extremes is discussed. The effectiveness of the proposed scheme for seasonal characterization and modeling is highlighted when contrasted to results obtained from the conventional approach of using fixed climatological seasons.