CoLM模拟土壤温度和湿度最敏感参数的研究

摘要

合理的参数估计是提高陆面模式模拟能力的关键,而其过高的维数极大地增加了合理估计的难度.参数的敏感性分析,旨在针对目标变量找出最敏感的参数,从而实现在有限计算机资源条件下,对参数进行合理估计.本文以Common Land Model (CoLM)为研究对象,利用Morris方法定性地从40个参数中筛选出影响土壤温度和土壤湿度的敏感参数,并通过Sobol’方法从敏感性顺序和各敏感参数的累积贡献率两个方面,对Morris方法分析结果进行验证.在此基础上,本研究还利用Sobol’方法对已筛选的参数做定量敏感性分析,最终确定参数的主效应、交互效应和总效应.研究结果表明,Morris方法可以基于少量样本实现复杂的陆面模式的参数筛选,而Sobol'方法的结果又从定量的角度描述了每个敏感参数对模型响应的影响程度,并且两种方法结论一致.%Proper parameter estimation is critical for enhancing the simulation capability of land surface models,yet the high dimensionality of model parameters significantly increases the difficulty of parameter estimation.Sensitivity analysis of model parameters aims at identifying the most sensitive parameters for further analysis in order to estimate the parameters with finite computer resources.In this study,we conducted sensitivity analysis of forty adjustable parameters selected from the Common Land Model (CoLM) using soil temperature and soil moisture data.The Morris method was used to qualitatively screen out the sensitive parameters from the insensitive ones.We also adopted the Sobol' sensitivity method to quantitatively estimate the sensitivity of the screened parameters.The main effect,interaction effect,and total effect of the parameters were computed.The results showed that the Morris method could provide reliable parameter sensitivity analysis using a relatively small sample size,and the Sobol' results provided a quantitative measure how each parameter contributes to the total variance of the model response,which was consistent with the results obtained by the Morris method.