Advancing Physical Realism in Hydrological Modelling: Selection and Integration — A Review and Synthesis

Hydrological models play a key role in understanding, predicting, and managing anthropogenically altered hydro-environments. They support assessment of water quantity and quality, simulation of water and material transport, prediction of risks and disasters, and decision support for management. However, fragmented applications and limited interoperability constrain their utility. Following PRISMA 2020 guidelines, we systematically searched the Web of Science Core Collection and selected 30 representative hydrological models and 186 peer-reviewed studies for evidence synthesis (case-specific details provided in Supplementary Appendices). We classify these models by structural and functional characteristics and propose a practitioner-focused, decision-oriented synthesis framework (with practical criteria) that maps model capabilities to fit-for-purpose research and management tasks. Pathways for model integration, enhancement of physical realism, and applications to extreme events and ungauged basins are outlined. The framework highlights persistent challenges, including uncertainty in key hydrological and environmental processes, prediction of compound extremes, and reliable simulation in ungauged settings. We further outline pathways for ungauged and data-scarce settings via physics-guided machine learning and multi-source remote sensing. Process-level constraints, physically informed pre-parameterization, and transparent uncertainty treatment are emphasized to strengthen physical realism. The concept of “hydrological limits” is positioned as a policy-relevant lens to quantify thresholds and feedback within a planetary-boundaries context. This review advances adaptive water resources management in increasingly complex water environments. Maps 30 hydrological models to research and water-management tasks to guide model choice. Summarizes 24 integration modes and a reusable model-selection workflow with typical couplings. Physical-realism checklist: finer resolution, pre-parameterization, multi-process checks. Stresses long-term modelling for slow feedback and sub-daily calibration for extremes. Physics-guided ML with remote sensing for ungauged/data-scarce basins to enable decision support.