A Comparative Analysis of Ensemble NWP Models for Flood Forecasting

Abstract

This study presents the first application of gamma quantile mapping to bias-correct ensemble precipitation forecasts from seven global NWP models (ECMWF, NCEP, UKMO, CMA, JMA, ECCC, NCMRWF) in the data-scarce Saliyan Basin, Iran. The integration of these models with advanced bias correction techniques significantly improves flood forecasting accuracy. To address systematic biases in the raw forecasts, gamma quantile mapping was applied, significantly enhancing the reliability of the precipitation inputs. These bias-corrected forecasts were then used as inputs for the GR4J hydrological model to simulate river flow and predict flood events. The study period included a major flood event in March 2019, which was used to evaluate the performance of the ensemble forecasting system. Results demonstrated that bias correction using gamma quantile mapping substantially improved the accuracy of flood forecasts, with the ECMWF and UKMO models showing the highest skill scores. The ensemble approach effectively captured the uncertainty in flood predictions, providing valuable insights for risk assessment and decision-making. This research highlights the importance of bias correction in ensemble forecasting and offers a robust framework for flood prediction in data-scarce regions. The findings have significant implications for improving flood early warning systems and mitigating flood-related damages in similar basins worldwide.