中国森林与农田遥感叶面积指数产品精度分析

发布时间：2018-05-30 07:07

本文选题：叶面积指数 + MODIS　；参考：《南京信息工程大学》2015年硕士论文

【摘要】：由于全球气候变化、碳源汇变化驱动机制等研究的需要,叶面积指数(Leaf Area Index, LAI)作为气候模式、碳循环模式等动态过程模型的关键输入,LAI的研究和应用更应需要在区域及全球尺度上进行,因此利用遥感卫星数据进行LAI反演,进而生产全球遥感LAI产品数据得到越来越广泛应用。然而,由于不同的测量方法和仪器及冠层结构的不同因素定义叶面积指数,LAI产品能显著变化,目前还不具有准确性和一致性的全球和区域应用产品。因此在应用遥感LAI产品时,对LAI产品的精度评价显得尤为重要。本研究以中国东北大兴安岭加格达奇林区、江苏省南京市典型农田为研究区。基于环境卫星遥感数据获得典型植被指数,利用环境卫星植被指数和实测LAI构建回归分析模型分别反演林地、小麦和水稻LAI。最后,利用环境卫星反演LAI数据通过尺度转换对MODIS LAI产品和GLOBCARBON LAI产品进行验证,并分析LAI产品误差来源,再对农田区MODIS LAI产品和GLOBCARBON LAI产品进行订正。本研究的主要结论如下：(1)在中国东北大兴安岭加格达奇林区,三种LAI数据在植被区域LAI值域范围最大的为GLOBCARBON LAI数据,其值在0.93～4.91, HJ-1kmLAI数据与MODIS LAI数据值域基本相同,GLOBCARBON LAI均值最高,其值比HJ-1kmLAI高0.29,误差为11%,而MODIS LAI数据均值则比HJ-1kmLAI均值低0.28,误差为11.8%,两种遥感LAI数据产品在研究区精度误差均在20%左右,但GLOBCARBON LAI存在高估现象,而MODIS LAI数据则为低于实测反演值。(2)在南京农作物研究区,小麦的GLOBCARBON LAI比HJ-30mLAI均值低1.18,误差为44%,MODIS LAI比HJ-30mLAI均值低1.75,误差为66%；在水稻区,GLOBCARBON LAI比HJ-30mLAI均值低0.84,误差为25%, MODISLAI比HJ-30mLAI均值低1.47,误差为43%。通过结果分析可以看出,研究区小麦和水稻MODIS LAI、GLOBCARBON LAI的均值明显低于环境卫星反演得到的LAI值,存在严重低估现象。根据分析,由于南京农田呈零星状分布,地表异质性严重,导致低分辨率GLOBCARBON LAI和MODIS LAI产品存在混合像元。(3)再对南京农作物研究区GLOBCARBON LAI和MODIS LAI产品混合像元进行分解,得出在小麦区,GLOBCARBON LAI均值高估HJ-30mLAI为0.25,误差从44%降到8.6%,MODIS LAI比HJ-30mLAI均值低0.29,误差从66%降到10.9%；在水稻区,GLOBCARBON LAI比HJ-30mLAI均值高0.28,误差从25%降到7.6%,MODIS LAI比HJ-30mLAI均值低0.23,误差从43%降到6.7%。从订正后数据来看,订正后的MODIS LAI和GLOBCARBON LAI极大的改善了混合像元的问题。但GLOBCARBON LAI存在高估现象,而MODIS LAI数据则为低于实测反演值。
[Abstract]:The Leaf Area Index (LAI) is the key input of the dynamic process model such as the climate model and the carbon cycle model as a result of the global climate change and the driving mechanism of carbon source and sink change. The research and application of LAI should be carried out on the regional and global scales. Therefore, the remote sensing satellite data is used for the LAI inversion, and then the remote sensing satellite data is used to inverse the LAI. The production of global remote sensing LAI product data is becoming more and more widely used. However, because of the different measurement methods and instruments and the definition of the leaf area index of the different factors of the canopy structure, the LAI product can change significantly. At present, the accuracy and consistency of the global and regional application products are not yet accurate. Therefore, when using remote sensing LAI products, the LAI products are applied to the products. This study takes the Jiagedaqi forest area of Greater Khingan Range in Northeast China and the typical farmland of Nanjing city of Jiangsu Province as the research area. Based on the environmental satellite remote sensing data, the typical vegetation index is obtained. The regression analysis model of the environmental satellite vegetation index and the measured LAI is used to reconstruct the woodland, and the wheat and rice LAI. are last, The LAI data of MODIS LAI products and GLOBCARBON LAI products are verified by scale conversion, and the error sources of LAI products are analyzed, and the MODIS LAI products and GLOBCARBON LAI products in farmland are revised. The main conclusions of this study are as follows: (1) three LAI data in the Jiagedaqi forest area of Greater Khingan Range, Northeast China. The maximum range of LAI range in the vegetation area is GLOBCARBON LAI data, its value is 0.93 ~ 4.91, HJ-1kmLAI data and MODIS LAI data range are basically the same, the mean value of GLOBCARBON LAI is highest, its value is 0.29 higher than HJ-1kmLAI, and the error is 11%, while the MODIS LAI data mean is 0.28 lower than the HJ-1kmLAI mean, and the error is 11.8%, two kinds of remote sensing data products. The accuracy error in the study area is around 20%, but the GLOBCARBON LAI is overestimated, while the MODIS LAI data is lower than the measured inversion value. (2) in the Nanjing crop research area, the GLOBCARBON LAI of wheat is 1.18 lower than the HJ-30mLAI mean, the error is 44%, MODIS LAI is 1.75 lower than HJ-30mLAI, and the error is 66%. In the rice region, GLOBCARBON LAI is compared to those in the rice region. The mean value of 0mLAI is 0.84, the error is 25%, and the MODISLAI is 1.47 lower than the HJ-30mLAI mean. The error is 43%. through the result analysis. The mean value of MODIS LAI and GLOBCARBON LAI in the study area is obviously lower than the LAI value obtained by the environmental satellite, and there is a serious underestimation. According to the analysis, the farmland is scattered in Nanjing and the ground surface is different. The quality is serious, resulting in a mixed pixel of low resolution GLOBCARBON LAI and MODIS LAI products. (3) then the mixed pixel of GLOBCARBON LAI and MODIS LAI products in the Nanjing crop research area is decomposed, and the average value of GLOBCARBON LAI is 0.25, the error is reduced from 44% to 8.6% in the wheat region, and the error is 0.29 lower than that of the average. From 66% to 10.9%, in the rice area, GLOBCARBON LAI is 0.28 higher than that of HJ-30mLAI, the error is reduced from 25% to 7.6%, MODIS LAI is 0.23 lower than the mean of HJ-30mLAI, and the error is reduced from 43% to 6.7%. from the revised data. The revised MODIS LAI and GLOBCARBON LAI greatly improve the problem of the mixed image element. The IS LAI data is lower than the measured inversion value.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S771.8;S127

【参考文献】