Terrestrial Laser Scanning Remote Sensing Retrieval Study Of Broad-leaved Forest Canopy Clumping Index

Forest canopy clumping index（CI）reflects the deviation of leaves distribution from random distribution and is an important index to describe the radiation transmission process in the canopy.The gap fraction（GF）directly represents the spatial distribution of radiation in the canopy,and is an indispensable canopy structure parameter in the clumping index retrieval model.In recent years,with the development of Li DAR technology,terrestrial laser scanning（TLS）obtains a large number of high-resolution spatial samples in the form of 3-D point cloud data set,which is widely used for retrieving canopy structure parameters.Monte Carlo simulations can be used to obtain approximate solutions to quantitative problems through a large number of random sampling while avoiding complicated mathematical calculations.In order to make full use of the high-resolution canopy information provided by 3-D point cloud data,this thesis proposes a new method to estimate the GF based on Monte Carlo simulations.Aiming at the problem of different path length of light in canopy caused by clumping effect,two methods were proposed to extract the absolute path length distribution of the canopy and realize the retrieval of CI of broad-leave forest canopy.The main research contents and achievements are as follows.（1）Retieving gap fraction based on the Monte Carlo simulations.The proposed method constructs the gap fraction as a probabilistic model based on the law of large numbers and estimates the approximate value of the GF by Monte Carlo simulations.The GF results are obtained by determining the discrimination distance,simulating the laser beam and estimating the GF.We selected four sample plots of broad-leave forest on the campus of UESTC.The results of the GF were compared and analyzed with those estimated by the projection slice method and the digital hemisphere photograph（DHP）method,and it was found that the slice of 0.1°*0.1°in the projection slice method was in better agreement with the results of proposed algorithm,and the determination coefficient reached 0.68.Compared with the projection slicing method,the proposed algorithm does not need to discuss the model parameters and also considers the influence of the leaf point cloud distribution on the results.Compared with the DHP-based GF,the GF estimated by the proposed method is lower,which avoids the defect that hemispherical photos can not capture the vertical structure of canopy.（2）Retrieving clumping index based on the path length distribution model.For the application of the path length distribution model proposed by Hu et al.（2014）^[69]in broadleaf forest canopy,based on the high-resolution 3D ground-based point cloud data,this thesis proposes two methods to retrieve the absolute path length distribution of light through the canopy based on the boundary mask method and the grid projection method.According to the GF and path length distribution,the path length distribution model is used to estimate CI.Comparing the CI results of the two methods of extracting absolute path length,it was found that the trend of CI was consistent with the change of zenith angle,and the correlation between the retrieval results of the two methods was strong,with the coefficient of determination reaching 0.65.It indicates that the two methods can be verified with each other,and both of them show the effectiveness.The results were also compared with the CI results estimated by the finite-length averaging method（LX）,and it was found that the results estimated by LX method were low and insensitive to the change of zenith angle,which indicates that the proposed method overcomes the problem of low CI caused by the overestimation of LAI in the gapless sample line,and also proves that the method of retieval of the CI based on the path length distribution method is feasible and effective.The results also found that both methods are insensitive to the grid size,indicating that the robustness and applicability of the two methods are good,and the grid size can be determined according to the scale of sample plot.