Addressing Curse of Dimensionality in Sensitivity Analysis: How Can We Handle High-Dimensional Problems?
Abstract
Complex environmental models are now the primary tool to inform decision makers for the current or future management of environmental resources under the climate and environmental changes. These complex models often contain a large number of parameters that need to be determined by a computationally intensive calibration procedure. Sensitivity analysis (SA) is a very useful tool that not only allows for understanding the model behavior, but also helps in reducing the number of calibration parameters by identifying unimportant ones. The issue is that most global sensitivity techniques are highly computationally demanding themselves for generating robust and stable sensitivity metrics over the entire model response surface. Recently, a novel global sensitivity analysis method, Variogram Analysis of Response Surfaces (VARS), is introduced that can efficiently provide a comprehensive assessment of global sensitivity using the Variogram concept. In this work, we aim to evaluate the effectiveness of this highly efficient GSA method in saving computational burden, when applied to systems with extra-large number of input factors ( 100). We use a test function and a hydrological modelling case study to demonstrate the capability of VARS method in reducing problem dimensionality by identifying important vs unimportant input factors.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2016
- Bibcode:
- 2016AGUFM.H11A1289S
- Keywords:
-
- 1805 Computational hydrology;
- HYDROLOGYDE: 1846 Model calibration;
- HYDROLOGYDE: 1847 Modeling;
- HYDROLOGYDE: 1873 Uncertainty assessment;
- HYDROLOGY