Retrieving Canopy Closure Of Forests By Stochastic Radiative Transfer Model

Remote sensing is one of most important methods to estimate forest canopy closure at large.There are mainly three kinds of remote sensing algorithms for canopy closure retrieval:statistical models.physical models and mixed models.Most practices used statistical models,which are lacking physical explanation and limited in local areas.The physical models are with clearer understanding on mechanism,which can be used in large areas,but are less applied due to the higher complexity.The Stochastic Radiative Transfer(SRT)model is applicable in simulating forests with horizontally distributed heterogeneity,which may stand for different canopy closure.However,no report on inversion research based on Stochastic Radiative Transfer model has been found.In this paper,based on the SRT model,an inversion method has been proposed on canopy closure retrieval of Yunnan Pine forests and of forests mixed by dahurian larch and white birch.The fundamental 1s the quantitative relationship between the canopy closure and the probability of finding foliage elements in SRT Model.Then,a look-up-table was constructed to inverse the canopy closure to reflectances from GF-1 and Landsat-8 satellite images.The probability of finding foliage elements and leaf area index are deterlined in the case of a minimum difference between simulated reflectances and satellite observations,in order to calculate the canopy closure based on the stochastic Beer-Lambert-Bouguer law.To correct the cylinder shape assumption,an equivalent model was proposed to match the Yunnan Pine crown shape as an example.The plots of field data were used to assess the inversion accuracy and a statistical inversion method based on NDVI 1s conducted for comparison.Results show that the inversion can map canopy closure of forests accurately in the two study area.For the pure forests of Yunnan Pine,the inversion accuracy is higher(R2=0.8345,RMSE=0.0688).For mixed forests of dahurian larch and white birch,the average treatment in parameter input might produce some errors(R2=0.7042,RMSE=0.1448).The equivalent shape correction model is reasonable and the algorithm is flexible in different crown cases.This study can provide supports on both forward models and inversion methods for forest canopy closure retrieval in large scale or long time series.