期刊
ENVIRONMENTAL MODELLING & SOFTWARE
卷 109, 期 -, 页码 293-305出版社
ELSEVIER SCI LTD
DOI: 10.1016/j.envsoft.2018.08.022
关键词
Component-based modeling; Uncertainty propagation; Nonparametric hierarchical model; Cokriging; Data-driven emulators
类别
资金
- National Science Foundation [1504728, 1632824]
- National Institute of Food and Agriculture (NIFA)/USDA [2017-67017-26167]
- NIFA [914699, 2017-67017-26167] Funding Source: Federal RePORTER
- Division Of Undergraduate Education
- Direct For Education and Human Resources [1504728] Funding Source: National Science Foundation
- Office Of The Director
- Office of Integrative Activities [1632824] Funding Source: National Science Foundation
A network of independently trained Gaussian processes (StackedGP) is introduced to obtain predictions of geospatial quantities of interest (model outputs) with quantified uncertainties. The uncertain nature of model outputs is due to model inadequacy, parametric uncertainty, and measurement noise. StackedGP framework supports component-based modeling in environmental science, enhances predictions of quantities of interest through a cascade of intermediate predictions usually addressed by cokriging, and propagates uncertainties through emulated dynamical systems driven by uncertain forcing variables. By using analytical first and second-order moments of a Gaussian process with uncertain inputs using squared exponential and polynomial kernels, approximated expectations of model outputs that require an arbitrary composition of functions can be obtained. The performance of the proposed nonparametric stacked model in model composition and cascading predictions is measured in a wildfire and mineral resource problem using real data, and its application to time-series prediction is demonstrated in a 2D puff advection problem.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据