Khwairakpam et al. (2026) Enhanced flood quantile estimation and its implications in rainfall–discharge relationship during flood events in Brahmani-Baitarani River basin, India

Identification

• Journal: Water Practice & Technology
• Year: 2026
• Date: 2026-01-01
• Authors: Robindro Singh Khwairakpam, Sananda Kundu
• DOI: 10.2166/wpt.2026.190

Research Groups

• Department of Geography, Manipur University, Canchipur, India

Short Summary

This study aimed to identify suitable probability distributions for flood quantile estimation and analyze rainfall-discharge relationships in the Brahmani-Baitarani River basin, India. It found that multi-parameter distributions like Wakeby, Log-Pearson 3, and Generalised Extreme Value (GEV) provided the most reliable flood quantile estimates, revealing strong but non-linear rainfall-discharge correlations during flood events.

Objective

• To select the most suitable probability distributions for flood quantile estimation at each gauged location within the Brahmani-Baitarani River basin.
• To analyze the recurrence intervals (RIs) and frequency curves of observed discharge.
• To establish a relationship between heavy rainfall and discharge during flood events in the Brahmani-Baitarani River basin.

Study Configuration

• Spatial Scale: Brahmani-Baitarani River basin, eastern India, spanning 51,822 square kilometers (Brahmani sub-basin: 39,033 square kilometers, Baitarani sub-basin: 12,789 square kilometers). The study focused on seven hydrological gauge stations across these sub-basins.
• Temporal Scale: Daily flow data from 1973 to 2018 (duration varied by station, e.g., Altuma: 1996–2018, Anandapur: 1973–2018). Daily rainfall gridded datasets for 120 years (for Jenapur and Anandapur).

Methodology and Data

• Models used:
  • Probability Distributions: Lognormal (two parameters), Gumbel Max (two parameters), Log-Pearson 3 (three parameters), Log-Logistics 3P (three parameters), Generalised Pareto (three parameters), Generalised Extreme Value (GEV) (three parameters), Generalised Gamma (four parameters), and Wakeby (five parameters).
  • Parameter Estimation Methods: Method of L-moments, Maximum Likelihood Method, Method of Moments.
  • Goodness-of-Fit (GOF) Tests: Kolmogorov–Smirnov (K-S) test, Anderson–Darling (A-D) test (at a significance level of α = 0.05).
  • Accuracy Tests: Nash–Sutcliffe efficiency (NSE), mean absolute error (MAE), normalised root mean squared error (NRMSE), Kling–Gupta efficiency (KGE), and coefficient of determination (R²).
  • Outlier Detection: Multiple Grubbs-Beck test.
  • Selection Method: Relative scoring method for basin-wise and site-wise best-fit distribution selection.
  • Rainfall-Discharge Analysis: Time-series analysis, polynomial curve fitting, and coefficient of determination (R²) calculation.
• Data sources:
  • Daily river discharge data: Central Water Commission, India.
  • Daily rainfall gridded datasets: National Data Centre, Indian Meteorological Department (IMD), Pune (for Jenapur and Anandapur).
  • Flood-related information: Odisha State Disaster Management Authority, Govt. of Odisha, and Centre for Research on the Epidemiology of Disasters (CRED), University of Louvain, Belgium.

Main Results

• The annual maximum series (AMS) of discharges exhibited medium to high interannual variability, with the coefficient of variation (C_v) ranging from 45% at Jenapur to 82% at Jarikela.
• Goodness-of-fit tests showed that the Kolmogorov–Smirnov test did not reject any distribution, while the Anderson–Darling test rejected Generalised Pareto for most gauges, Generalised Gamma for three gauges, and Wakeby for one gauge.
• A basin-wide relative scoring method identified the Wakeby distribution as the most suitable (total score: 542), with Log-Pearson 3 as a reliable alternative (total score: 529).
• Distributions with more parameters (Wakeby, Log-Pearson 3, GEV) generally outperformed those with fewer parameters in estimating flood quantiles.
• Site-specific analysis revealed varying best-fit distributions: GEV for Altuma and Anandapur, Log-Logistics (3P) for Champua, Wakeby for Gomlai and Jenapur, and Log-Pearson 3 for Jarikela and Tilga.
• A strong correlation (R² values ranging from 0.97 to 0.99) was observed between predicted and observed discharge values using the best-fit distributions.
• Estimated recurrence intervals (RIs) for mean discharge ranged from 2 to 2.8 years, for large floods from 5 to 7.7 years, and for maximum discharges were 7 to 33 times higher than mean discharges.
• Design flood recommendations include a 50-year flood for the lower Brahmani catchment, a 100-year flood for the upper Brahmani, and floods with RIs greater than 100 years for the lower Baitarani.
• Analysis of rainfall-discharge relationships showed strong correlations (moderate to strong R² values) during flood events. Notably, recent floods (2018, 2020, 2022) occurred with below-average AMS rainfall, indicating a non-linear rainfall-runoff relationship in the basin.
• Recurrence intervals of peak rainfall during recent flood events were often short (e.g., 1.01 to 1.03 years in August 2022), suggesting that less rainfall can trigger floods in vulnerable regions like lower Jenapur.

Contributions

• Successfully identified robust, advanced basin-wise and site-wise probability distributions for reliable flood quantile estimates in a hydrologically heterogeneous river basin.
• Introduced an innovative approach by applying multivariate probability distribution models for improved and more realistic flood estimation in complex, heterogeneous basins.
• Provided new insights into rainfall-discharge dynamics, confirming a non-linear relationship where recent floods occurred despite below-average rainfall.
• Established a method to determine the rainfall quantity responsible for flood occurrence, aiding in early flood warning systems.
• Offered specific design flood recommendations (50-year, 100-year, and >100-year RIs) for hydraulic structure design and flood management in different reaches of the Brahmani-Baitarani basin.

Funding

The authors did not receive any funding for this present study.

Citation

@article{Khwairakpam2026Enhanced,
  author = {Khwairakpam, Robindro Singh and Kundu, Sananda},
  title = {Enhanced flood quantile estimation and its implications in rainfall–discharge relationship during flood events in Brahmani-Baitarani River basin, India},
  journal = {Water Practice & Technology},
  year = {2026},
  doi = {10.2166/wpt.2026.190},
  url = {https://doi.org/10.2166/wpt.2026.190}
}