State-of-the-Art Statistical Approaches for Estimating Flood Events

Area of Science:

• Hydrology and Water Resources Engineering
• Statistical Hydrology
• Flood Risk Management

Background:

• Accurate annual peak flow discharge (APFD) estimates are essential for hydraulic infrastructure design, operation, and flood risk management.
• Various statistical methods exist for quantile estimation, each with potential impacts on water resource applications.

Purpose of the Study:

• To evaluate and compare different quantile estimation methods for APFDs.
• To assess the performance of linear higher order-moments (LH-moments) and nonparametric kernel functions.
• To determine the most suitable methods for water resources management and engineering in Punjab, Pakistan.

Main Methods:

• Application of LH-moments (η = 0, 1, 2) and nonparametric kernel functions to APFD data from 18 stream gauge stations.
• Fitting of generalized logistic (GLO), generalized extreme value (GEV), and generalized Pareto (GPA) distributions.
• Performance evaluation using Anderson-Darling, Kolmogorov-Smirnov, and Cramér-Von Mises tests, and LH-moment ratio diagrams.

Main Results:

• GEV and GPA distributions showed the best fit for most stations, followed by GLO.
• LH-moments (η = 0, 1, 2) yielded lower relative absolute errors, particularly for higher return periods.
• Gaussian kernel function provided comparable estimates for small return periods.

Conclusions:

• LH-moments (η = 0, 1, 2) are highly valuable for accurate APFD quantile estimation, especially for critical high return periods.
• Nonparametric kernel functions, particularly Gaussian, offer competitive estimates for smaller return periods.
• Both LH-moments and kernel functions are important tools for water resources management and engineering projects.