Evaluation of Probability Distributions for Flood Frequency Analysis Using PDS

Flood Frequency Analysis (FFA) is a key statistical tool used to estimate the magnitude and probability of flood events associated with different return periods. Traditionally, FFA relies on the Annual Maximum Series (AMS), which considers only the largest flood event recorded each year. However, in regions that experience multiple flood events annually, such as Attanagalu Oya in Sri Lanka, the AMS approach may fail to capture important secondary or tertiary floods. To overcome this limitation, the present study applies the Partial Duration Series (PDS) method, which includes all flood events exceeding a predefined threshold and therefore provides a more comprehensive representation of flood characteristics. A threshold range between 3.5 m and 5.0 m, with increments of 0.1 m, was systematically evaluated to identify the most suitable level for event extraction. To ensure statistical independence among flood events, a declustering technique was applied to remove dependent peaks. Five probability distributions, Gumbel, Generalized Extreme Value (GEV), Lognormal, Weibull, and Generalized Pareto Distribution (GPD), were fitted to the PDS dataset. Parameter estimation was performed using established techniques, including the Method of Moments, Maximum Likelihood Estimation, and L-moments, depending on the specific distribution. Model performance was assessed using Nash–Sutcliffe Efficiency (NSE), Percent Bias (PBIAS), Root Mean Square Error (RMSE), and the RSR index. Results indicated that the GEV distribution provided the best overall performance (NSE = 0.9714, PBIAS = 0.1101, RMSE = 0.0252, and RSR = 0.1691), highlighting its suitability for design flood estimation and risk assessment in the study area. The Gumbel and Weibull models showed acceptable performance. The findings emphasize the importance of improved threshold selection and parameter estimation methods, as well as the potential integration of multivariate approaches to better capture flood dynamics and support effective hydrological planning and disaster risk management.