Leaf Area Index Estimation Basec On Hyperion Data

Leaf Area Index (LAI), as one of the most fundamental indices characterizing the plant canopy structures, is an important structural parameter in land surface process studies. Hyperspectral sensors imaged the earth with narrow-bands, multi-channels simultaneously, therefore, continuous and fine spectral information of different features can be acquired, which favors land-cover identification vegetation parameters estimation greatly.Estimation of the leaf area index taking SR,NDVI,MNDVI,SAVI,TGDVI。ARVI based on hyperspectral image of Culaishan forest farm in ShanDong province as estimation factor was conducted in two different approaches. One was analyzing correlation between vegetation index and field LAI data of entire study area without classifying, and find out optimal statistical model. The other was distinguish the area as conifer and broad-leaved forests, respectively conduct inversion LAI modeling, and estimate LAI data of the entire study area. Linear regression and exponential regression on these two approaches were carried out to figure out the best fit model.Optimal model for the first approach is as follows: LAI=0.9371×SR-1.2886, R~2=0.7947. In second approach, Optimal model for conifer forest is as follows: LAI=5.6189×SAVI-1.8102, in which L =0.1, R~2=0.9734; Optimal model for broadleaved forest is as follows: LAI=0.4428×EXP(SAVI×1.906), in which L=0.5, R~2=0.8723.The precision of first approach was 81.02%, the precision of second approach was 92.78%. It can be drawn that establishing different statistical models for different types of vegetation can achieve higher accuracy in LAI estimation.