Acero Triana, J. S., Ajami, H., Razavi, S. (2022) Uncertainty and Sensitivity Analyses of a Semi-Distributed Non-Parsimonious Hydrologic Model Across Spatial and Temporal Scales Using New and Traditional Global Sensitivity Analysis Approaches. In AGU Fall Meeting Abstracts (Vol. 2022, pp. H15F-05). https://ui.adsabs.harvard.edu/abs/2022AGUFM.H15F..05A/abstract#:~:text=Bibcode%3A-,2022AGUFM.H15F..05A,-Feedback/Corrections%3F

Hydrologic models improve our understanding of the terrestrial hydrologic system under past conditions and are used for scenario analysis to assess environmental impacts on water resources. Semi-distributed hybrid hydrologic models (SHHMs) have been increasingly used in recent years, overcoming the limitations of lumped conceptual and distributed physically based models in simulating hydrologic fluxes across highly managed basins. This growth in popularity is due to the availability of GUI-GIS tools, relatively fast model execution time, and their flexibility in implementing water and land management practices. Being non-parsimonious, SHHMs require rigorous sensitivity analysis (SA) before calibration to identify parameters that mainly affect the hydrologic response and generate realistic results for the right reasons. However, recent studies have suggested that traditional SA methods, such as those rooted in elementary effects and variance analysis of model outputs, may present a relatively narrow view of sensitivities as they do not integrate global sensitivity information across multiple perturbation scales. Given the complexity of constraining SHHMs, in this study, we used the variogram analysis of response surfaces tool (VARS-TOOL) to assess parameter sensitivity and uncertainty of a variance-based SA method (Sobol) against a new variogram-based analysis method (VARS). Our discussion addresses whether VARS is more informative than Sobol and how parameter sensitivity is impacted by model spatiotemporal resolution using single- and multi-objective functions versus assessments independent of observations. To address these questions, we used the Soil and Water Assessment Tool (SWAT) to simulate hydrologic fluxes across three small catchments (<200 km2) with different physiographic features in the USA. Our results suggest no significant differences between the Sobol and objective and non-objective VARS. At the same time, SWAT is highly dependent on the initial runoff curve number for moisture condition II, resulting in large equifinality and prediction uncertainty, and constraining parameter ranges is limited even when a multi-objective function is used. Our results are expected to provide more insights into the reliability of hydrologic models for scenario-based assessment.

https://ui.adsabs.harvard.edu/abs/2022AGUFM.H15F..05A/abstract#:~:text=Bibcode%3A-,2022AGUFM.H15F..05A,-Feedback/Corrections%3F