生态模型的灵敏度分析

灵敏度分析用于定性或定量地评价模型参数误差对模型结果产生的影响，是模型参数化过程和模型校正过程中的有用工具，具有重要的生态学意义．灵敏度分析包括局部灵敏度分析和全局灵敏度分析．局部灵敏度分析只检验单个参数的变化对模型结果的影响程度；全局灵敏度分析则检验多个参数的变化对模型运行结果总的影响，并分析每一个参数及其参数之间相互作用对模型结果的影响．目前，在对生态模型的灵敏度分析中，越来越倾向于使用全局灵敏度分析的方法．但国内仍多采用局部灵敏度分析方法，很少采用全局灵敏度分析方法．文中详细论述了局部灵敏分析和全局灵敏度分析的主要方法(一次变换法、多元回归法、Morris法、Sobol’法、傅里叶幅度灵敏度检验法和傅里叶幅度灵敏度检验扩展法)，希望能为国内生态模型的发展提供一个比较完善的灵敏度分析方法库．结合国内外的灵敏度分析发展现状，指出联合灵敏度研究、灵敏度共性研究及空间直观景观模型的灵敏度分析将为生态模型灵敏度分析研究中的热点和难点．

Sensitivity analysis in ecological modeling

Sensitivity analysis is used to qualitatively or quantitatively apportion the variation of model output to different source of variation.It is a very useful tool in model parameterization and calibration,and has important ecological significance by identifying the governing factors for a certain ecological process simulated.There are two schools of sensitivity analysis, local sensitivity analysis and global sensitivity analysis.The former examines the local response of the output( s ) by varying input parameters one at a time,holding other parameters to a central value; and the latter examines the global response (averaged over the variation of all the parameters) of model output( s ) by exploring a finite (or even an infinite) region.Since it is very easy to conduct local sensitivity analysis,it is very popular in ecological models.However,local sensitivity analysis is not computationally effective,because it can only get the sensitivity of a single parameter at a time.It can not take into consideration the effect of interaction of different parameters.Additionally,the value of other parameters will affect the sensitivity of the parameter specified.In view of this,global sensitivity analysis is increasingly preferred to local sensitivity in recent years.However,for most of the ecological modeling study published in Chinese,only local sensitivity analysis is conducted.To provide a toolbox of alternative sensitivity analysis algorithms for ecological model development in future study,we reviewed the main methods of both local sensitivity analysis and global sensitivity analysis,including one at a time method,multivariate regression,Morris' method,Sobol's method,Fourier Amplitude Sensitivity Analysis,and Extended Fourier Amplitude Sensitivity Analysis.Based on the state of the art research on sensitivity analysis,the sensitivity of the interaction between two or more than two model parameters,the sensitivity of common model parameters in a set of models,and the sensitivity analysis in spatially explicit landscape model simulation were identified as the key areas and difficulties of future study on sensitivity analysis in ecological modeling.