Remote Sensing Inversion Study Of Vegetation Canopy Leaf Area Index Based On ALI Image

The Leaf Area Index (LAI) is an important parameter in the structure ofvegetation canopy, for it not only directly control several vegetation biophysicalprocesses, such as photosynthesis, plant respiration, plant transpiration andprecipitation intercepted, but also provide structured quantitative information fordescription of the initial energy exchanges of vegetation canopy surface. Moreover,LAI usually plays an important role in some subject studies and applications, such asagriculture, forestry, ecology and agricultural meteorology. However, it’s impossibleto obtain LAI fast in large area for the limitation of several factors such as thetraditional methods, data analysis and instrument error, which is unable to meet thegrowing needs. With the development of science technology and increasinglywidespread of the application of remote sensing technology, fast and accuratelyconduct large scale LAI dynamic monitoring becomes possible. In the Non-pointsource pollution simulation study, getting the LAI of study area fast and accurately ismeaningful for modifying the plant growth mode of SWAT distributed model aboutthe question if LAI should adopt mean value.This study used the Meijiang River watershed as the study area and used theAdvance Land Image (ALI) as the remote sensing data. After using the software ofENVI to conduct several basic pretreatment, different forms of Vegetation Index (VI)was derived from the ALI. According to the rule of vegetation vertical distribution andhorizontal distribution, chosen some sample plots in study area and obtained themeasured LAI value by the Plant Canopy Analyzer LAI-2000. After this, a spatialrelationship between the measured LAI and the VI value which was calculated by theremote sensing data, and then based on the sample plot information, derived thecorresponding VI by the software of ArcGIS. Then get the retrieval model by fittingthe linear and nonlinear relationship between LAI value and VI value, and accordingto the correlation coefficient (R2) to find out the best retrieval model for LAIinversion. The primary conclusions of this study as follows:(1) By establishing the linear regression model of the LAI-VIs, find that theLAI-MSAVI model is the best model among the single linear models; its R2reachedto the0.5601.and then is the R2of the SAVI series. For the multiple linear regressionmodel, the R2reached to the0.673. (2) By establishing the non-linear regression model, find that the quadraticpolynomial model was the best model among all the non-linear models, the range ofits R2from0.2897to0.6314, which is reached to the best. Next is the logarithmicmodel; the range of R2is from0.233to0.5838.(3) By comparative analysis of linear model and non-linear model, find that themultiple linear model was the best, the quadratic polynomial model was the second.Generally speaking, the result of non-linear models is better than the linear model.(4) By a comprehensive analysis and comparison, finally, choose the linearmodel of LAI-MSAVI, multiple linear model, and quadratic polynomial model ofLAI-RDVI, exponential model of LAI-RDVI, Logarithmic model of LAI-RDVI andPower function model of LAI-RDVI as the models to conduct accuracy test.and afterthe accuracy test find that the accuracy of the exponential model ofLAI-RDVI(y=0.255e1.8741x) is the highest, which is reached to70.02%.