AI-Enabled multi-Objective calibration of a wflow_sbm hydrological model of the Nile Basin

Abstract: The water resources of the Nile Basin are under mounting pressures due to population growth, climate change, and growing transboundary tensions. These pressures intensify upstream–downstream trade-offs, making adaptive, data-driven planning essential to meet demands. wflow_sbm is a well-suited tool for understanding and modelling large-basin hydrological processes, yet it lacks an automated multi-objective calibration workflow. To overcome this limitation, we develop Optiverse, an AI-driven, multi-objective Python framework to calibrate wflow_sbm. Optiverse is designed as a modular, general-purpose package for multi-objective optimization of simulator workflows. Building on the Python package Platypus, Optiverse implements multi-objective evolutionary algorithms to search for Pareto-optimal solutions using NSGA-II and NSGA-III. This study presents a case study of calibrating a wflow_sbm model of the Blue Nile Basin for the period 1991-2020. The calibration was run on high-performance computing resources to meet the computational demands of iterative calibration, enabling reliable convergence to Pareto-optimal solutions. Early results indicate promising improvements across multiple calibration objectives for wflow_sbm, considering multiple calibration locations within the same optimization formulation. This framework provides a practical pathway for AI-enabled calibration of wflow_sbm and, for the Nile, provides a practical tool for decision support under increasing variability and risk.

Keywords: Distributed hydrology; wflow_sbm; optiverse; AI optimization; NSGA-II; calibration; evolutionary algorithms