Design And Implementation Of Adaptive Sampling For Sensitivity Analysis Of Earth System Model

With global warming,ecology and environment are seriously damaged,the frequent occurrence of extreme weather makes it urgent to study the mechanism of global climate change and predict the trend of future change.The earth system model is an important tool for the study of the mechanism,a computer based complex system model.The complex model often includes many parameters,and the values of these parameters have a great influence on the performance of the model.How to correctly estimate these parameters is a complex scientific problem.Parameter sensitivity analysis is a commonly used method to identify important parameters,to understand the influence of input parameters on output variables,and provide a scientific basis for parameter optimization.This paper first introduces 3 traditional sampling methods,commonly used for sensitivity analysis,and 4 adaptive sampling methods based on Markov Monte Carlo algorithm.There are 7 kinds of sampling methods.Then illustrating sensitivity analysis and 2 other evaluation methods,homogeneity and posteriori probability distribution.By comparing sample methods in the models,the adaptive sampling methods based on Markov chain Monte Carlo algorithm provide parameter convergence interval and parameters posterior distribution for the sensitivity analysis of parameters,and parameters in different interval showed different sensitivity.In order to make full use of the advantages and disadvantages of traditional sampling and adaptive sampling,an adaptive sampling scheme,which includes three sampling processes and two sensitivity analysis processes,is designed for the sensitivity analysis of the earth system pattern.