Abstract:

Objective: The advantages and disadvantages of classification information based on normalized point cloud, first return based on normalized point cloud and canopy height model(CHM) in estimating forest canopy cover in the core area of Winter Olympic Games were studied, and the estimation accuracies of the three algorithms were also analyzed. The purpose of this paper was to explore the optimal estimation method of tree canopy cover, so as to provide technical supports for accurately grasping the information of tree crown canopy in the core area of Winter Olympic Games and to promote the sustainable forest management. Method: Based on the airborne LiDAR data and the ground field data of 67 sample plots in the core area of the Winter Olympics Games, linear regression was used to fit the measured and estimated values of crown canopy, and the determination coefficient(R²) and root mean square error(RMSE) were calculated. The estimation accuracies of crown canopy through classification information based on normalized point cloud, first return based on normalized point cloud and canopy height model were compared. The correlations between sample site canopy cover, sample site laser point cloud density and canopy cover estimation error were analyzed, and the effects of canopy height model resolution on the stability of the canopy cover estimation method were also discussed. Result: 1) The estimation accuracy of canopy cover classification information based on normalized point cloud was the highest(R²=0.790 1, RMSE=0.124 3), and the estimation error was the lowest with an average overestimation of 1.17%. The second was the algorithm based on canopy height model(R²=0.763 8, RMSE=0.134 9), and the accuracy of first return based on normalized point cloud was the lowest(R²=0.758 2, RMSE=0.149 1). 2) There was no significant correlation between the sample site canopy cover and the estimation error. The sample sites with less than 40% canopy cover were generally underestimated by three algorithms, there was similar as for overestimation and underestimation results in plots with crown coverage of 0.4-0.8, wherea the sample sites with more than 80% canopy cover were generally overestimated. There was no correlation between laser point cloud density and estimation error, and the increase in laser point cloud density didn't improve the estimation accuracy of tree crown canopy. 3) The formal stability of the algorithm based on the canopy height model was the highest, and there was no significant difference in the result of the canopy cover estimation for the ten different resolutions of the raster, with R² ranging from 0.755 1 to 0.762 2 and RMSE ranging from 0.150 7 to 0.153 9. The best CHM resolution for canopy cover estimation in the core area of the Winter Olympics Games was 0.8 m×0.8 m. Conclusion: Through the estimation of canopy canopy of 67 plots in the core area of the Winter Olympic Games, it showed that three algorithms, classification information based on normalized point cloud, first return based on normalized point cloud and canopy height model might be suitable, and the estimation accuracy of crown canopy classification information based on normalized point cloud could be the highest. The advantages and disadvantages of the three algorithms were comprehensively analyzed by combining the tree crown canopy of the sample plots, the laser point cloud density of the sample plot and the resolution of the CHM, which could provide technical supports for the investigation of large-scale forest canopy cover.

Key words: laser LiDAR, canopy cover, Winter Olympic Games, normalized point cloud, first return, canopy height model(CHM)