Study On The Leaf Area Index Quantitative Retrieval For Bamboo Forest Based On The Remote Sensing And Ground Survey Data

As an important part of the forest ecosystem,bamboo forest plays an extremely important role in mitigating global climate change and the carbon cycle of forest ecosystems.The leaf area index(LAI)is defined as the half of the sum of all green leaf areas on a unit ground surface.It is a key parameter that drives the forest ecosystem model to model the carbon cycle and quantify the exchange of matter and energy between the land and the atmosphere.The unique physiological and ecological characteristics and management measures of bamboo forest make the spatial and temporal variation of leaf area index of bamboo forest complex and the inversion of remote sensing uncertain.The existing leaf area index products seriously underestimate the leaf area index of bamboo forest,and the difference of canopy spectrum of bamboo forest in on-and off-years brings challenges to LAI remote sensing inversion of bamboo forest.In this paper,two typical types of bamboo forests,namely Lei Bamboo forest and Moso Bamboo Forest were used as research objects.At the site scale,based on the MODIS surface reflectance data,stepwise regression and correlation analysis were used to perform variable screening,and two models of stepwise regression and BP neural network were constructed in combination with LAI measured data,which were used to retrieve LAI time series data from January 2014 to March 2017;At county scale,based on Landsat 8 OLI remote sensing data,Moso bamboo forest distribution information was extracted.Based on this,combined with Landsat7ETM+ data,the Random Forest Model was used to classify on-and off-years bamboo forests,and the distribution of bamboo forests in on-and off-years was obtained.Analyze the difference in spectral information between the annual size bamboo forests and adopt stepwise regression analysis method to construct the inverted leaf area index index of Moso bamboo forest.Finally,obtain the leaf area index of Moso bamboo in Anji County.The study mainly obtained the following conclusions:Firstly,based on MODIS surface reflectance data and combining BP neural network model to reverse the LAI of Lei Bamboo forest,the accuracy is higher than that of linear regression model,and the problem that LAI underestimated by MODIS LAI product is solved.Secondly,based on the time series analysis of remote sensing information of bamboo forest in on-and off-years,combined with random forest model,the distribution information of bamboo forest in on-and off-years can be extracted effectively with high classification accuracy.Thirdly,the LAI inversion model of bamboo forest in can be constructed by distinguishing on-and off-years of Moso bamboo.The inversion accuracy of LAI in bamboo forest can be improved.