Enhancing Flood Risk Management: A Comparative Study of Regional Frequency Models in the Upper Meghna River, Bangladesh

Flood risk management is essential in Bangladesh, frequently affected by severe flooding due to its location at the confluence of the Ganges, Brahmaputra, and Meghna rivers. This study assesses the effectiveness of Gumbel and Log-Pearson Type III (LP3) probability distributions for flood frequency analysis at the Bhairab Bazar station in the Upper Meghna River. Using 32 years (1990-2021) of annual peak discharge data from the Bangladesh Water Development Board, flood magnitudes were predicted for various return periods. The Gumbel distribution predicted discharges from 10,709.71 m³/s for a 2-year return period to 24,519.62 m³/s for a 200-year return period, while LP3 estimates ranged from 10,701.51 m³/s to 22,911.84 m³/s for the same periods. The peak over threshold (POT) approach yielded higher discharge estimates, showing its sensitivity to extreme events. For a 200-year return period, the Gumbel-POT and LP3-POT estimates were 22,117.40 m³/s and 21,964.07 m³/s, respectively. Goodness-of-fit tests, including Kolmogorov-Smirnov, Anderson-Darling, and Chi-squared, favored the LP3 distribution for both extreme value series (EVS) and POT data, especially in critical tail regions. A rating curve was also developed using the generalized reduced gradient algorithm to better understand the river's hydraulic behavior. These findings are crucial for local flood management strategies. Discharges exceeding critical thresholds, like the 5.8-m danger level and 6.8-m severe flood level, highlight the need for robust measures. This analysis offers essential insights for designing hydraulic structures, planning flood mitigation, and improving prediction models to enhance flood risk assessments in the Upper Meghna River basin.