"Identify Important and Influential Processes of Complex Environmental Systems under Model and Parametric Uncertainty"
Jan 11, 2023, Schedule:
- Nespresso & Teatime 417 DSL - Commons
- 03:00 to 03:30 PM Eastern Time (US and Canada)
- Colloquium - F2F 499 DSL / Virtual Zoom
- 03:30 to 04:30 PM Eastern Time (US and Canada)
Abstract:
Sensitivity analysis is a vital tool in the modeling community to identify important and influential parameters for model development and improvement, and variance-based global sensitivity analysis has gained popularity. However, the conventional global sensitivity indices are defined with consideration of only parametric uncertainty, but not model uncertainty that arises when a system’s process can be represented by multiple conceptual-mathematical models. Multi-model sensitivity analysis has gained increasing attention for advancing our understanding of complex Earth and environmental systems with interacting physical, chemical, and biological processes. Based on a hierarchical structure of parameter and model uncertainties and on recently developed techniques of model averaging, we developed two new process sensitivity indices for identifying important and influential processes. The indices are designed to answer the following question: how can we identify important and influential processes for the explicitly proposed process models and the probabilistically defined random parameters? A computationally efficient algorithm was also developed to reduce computational cost for the indices. To further reduce computational cost, we developed a new global sensitivity analysis method, called multi-model difference-based sensitivity analysis (MMDS), which can screen noninfluential system process from further investigation such as model calibration. In this seminar, I will present the three methods in a context of environmental modeling with numerical implementation and evaluation. The methods are mathematically and computationally general, and can be applied to a wide range of problems of numerical modeling.