摘要
遥感反演的前提需对模型的输入参数进行敏感性分析。该文选取冬小麦返青期、拔节期、孕穗期和抽穗期,考虑输入参数之间的相关性,以辐射传输模型(PROSAIL)为研究对象,对比分析了局部敏感性和全局敏感性方法在模型不确定性和LAI敏感性上的差异。结果表明,随着冬小麦的生长,模拟光谱与实测光谱吻合度提高,模型模拟的不确定性降低;与局部敏感性方法相比,全局敏感性方法的模型不确定性降低,模拟精度提高,冠层光谱对LAI的敏感性有明显变化,植被指数对LAI的敏感性则相对稳定;与NDVI相比,TGDVI对LAI更敏感。
Sensitivity analysis is the first step of remote sensing inversion.The sensitivity of canopy spectra to canopy structural parameters under local and global methods was calculated based on multi-temporal data in the paper.To demonstrate how the relationship between input parameters influences the model output and sensitivity results,the experiment was carried out in winter wheat in Beijing suburb during 2005-2006 and the measuring data were acquired on 7th April,19th April,27th April and 10th May 2006 individually. Two sensitivity methods were employed to analyze the uncertainty and sensitivity of radiative transfer model （PROSAIL） ：local and global sensitivity analysis. First, the relationship between input parameters was investigated according to observational field data. Then, the canopy spectra were simulated using PROSAIL model under local and global methods and were compared with measuring canopy spectral. Last, based on the uncertainty and sensitivity analysis matrix results, the sensitivity of PROSAIL model to LAI was analyzed under the two methods during the four phonological stages. The results showed that as winter wheat grows, the accuracy of simulated spectra is improved, meaning the model uncertainty decreases, and the sensitivity of vegetation indexes to LAI also decreases. Compared with results under local sensitivity analysis method, the model uncertainty decreases and the sensitivity of spectra to LAI obviously changes,but that of vegetation indexes is relatively stable. Compared with NDVI,TGDVI is more sensitive to LAI.
出处
《地理与地理信息科学》
CSCD
北大核心
2009年第6期17-21,25,共6页
Geography and Geo-Information Science
基金
国家自然科学基金项目(40701119)
国家973项目(2007CB714406)
国家863项目(2007AA10Z201)
遥感科学国家重点实验室科研基金项目
关键词
辐射传输模型
LAI
多时相
局部敏感性分析
全局敏感性分析
不确定性
radiative transfer model
LAI
multi-temporal
local sensitivity analysis
global sensitivity analysis
uncertainty
