Hydrologic and water quality model evaluation with global sensitivity analysis: Improvements and applications

Hydrologic and water quality (H/WQ) models can well-assess the impacts of natural and anthropogenic disturbances on natural systems. Today, these models have become extremely important policy making tools for tackling a range of water resources problems e.g. Best Management Practices, Total Maximum Daily Loads, etc. For well-informed decision making, applications of these models should undergo thorough reliability/global sensitivity and uncertainty analyses (GSA/UA) evaluation. Though the H/WQ modeling community has begun to realize the importance of GSA/UA, they are often ignored in practice. This study aimed to understand and improve SA/UA techniques and their applications to H/WQ and drought models. This work is divided into three specific research objectives.;A drought index, the Agricultural Reference Index of Drought (ARID), a model with correlated parameters was used as a test case for the first objective. An innovative approach was proposed to incorporate parameter correlation while performing GSA using a variance based technique. GSA results obtained using this approach were compared to those of a correlation-based SA methodology. Results were comparable. ARID was found to be monotonic and additive in nature and was most sensitive to the root zone depth and soil hydraulic properties.;The second part of this work proposed a new multi-criteria trajectory-based sampling strategy for the screening SA method of Elementary Effects (EEs). The concept of generating uniformly distributed parameter samples was merged with the often-used goal of maximizing sample spread. The evaluation of this strategy with benchmark strategies indicated that it was more than an order of magnitude computationally faster for high dimensional models and that it also somewhat improves parameter screening.;The last objective was aimed at the evaluation of the routing and pollutant attenuation module of the Watershed Assessment Model (WAM). GSA/UA showed that WAM was most sensitive to empirical/conceptual parameters, indicating their importance in the model calibration/validation process and the need for monitoring efforts to verify the selection of these parameters. As the first formal evaluation of WAM by GSA/UA, the results obtained in this work are a valuable contribution to the application of WAM to addressing water quality issues in the state of Florida.