A Novel Global Sensitivity Analysis Technique for an Enhanced Multi-criteria Analysis of the Behavior of Complex Environmental Models
Abstract
Complex physically-based environmental models are being increasingly used as the primary tool for watershed planning and management due to advances in technologies for heavy computation and data acquisition. An improved decision needs a proper approach to modelling. This implies that sensitivity analysis (SA) should be considered as an integral part of modelling, since it plays a crucial role for understanding the behavior of these complex models and improving their performance. Local sensitivity analysis approaches are helpful in this context but insufficient to thoroughly characterize model sensitivity. This is mainly due to non-linear behaviour of complex environmental systems and interactions within them. Therefore, a global sensitivity analysis (GSA) should be adopted to provide a comprehensive understanding of model behavior in these cases. One of the main challenges associated with the GSA methods is their substantial computational demand to generate robust sensitivity metrics over the entire factor space. Accordingly, a novel GSA technique, Variogram Analysis of Response Surfaces (VARS) is recently developed. VARS uses the Variogram concept to efficiently provide a variety of global sensitivity indices across a range of scales within the parameter space; a feature that is unique to VARS. In this work, for an enhanced understanding of the model behavior, we adopted a multi-criteria multi-model approach to conduct a thorough GSA. We applied VARS to a couple of environmental models, with various levels of complexity, and used various metrics to measure sensitivity of model response (streamflow) to model parameters. These metrics measure model performance for simulating high flows, low flows, and flow volume. It is indicated that VARS can be used efficiently to provide a thorough and unique GSA for models, and that the metric choice has a great influence on the assessment of model sensitivity.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2016
- Bibcode:
- 2016AGUFM.H34B..08H
- Keywords:
-
- 1847 Modeling;
- HYDROLOGYDE: 1871 Surface water quality;
- HYDROLOGYDE: 1879 Watershed;
- HYDROLOGYDE: 1880 Water management;
- HYDROLOGY