Limits...
Assessing and correcting topographic effects on forest canopy height retrieval using airborne LiDAR data.

Duan Z, Zhao D, Zeng Y, Zhao Y, Wu B, Zhu J - Sensors (Basel) (2015)

Bottom Line: Finally, a height weighted correction method is applied to correct the topological effects.The method is applied to LiDAR data acquired in South China, and its effectiveness is tested using 41 field survey plots.The results show that the terrain impacts the canopy height of individual trees in that the downslope side of the tree trunk is elevated and the upslope side is depressed.

View Article: PubMed Central - PubMed

Affiliation: Key Laboratory of Digital Earth Science, Institute of Remote Sensing and Digital Earth (RADI), Chinese Academy of Science, Haidian District, Beijing 100094, China. zjj@mail.csu.edu.cn.

ABSTRACT
Topography affects forest canopy height retrieval based on airborne Light Detection and Ranging (LiDAR) data a lot. This paper proposes a method for correcting deviations caused by topography based on individual tree crown segmentation. The point cloud of an individual tree was extracted according to crown boundaries of isolated individual trees from digital orthophoto maps (DOMs). Normalized canopy height was calculated by subtracting the elevation of centres of gravity from the elevation of point cloud. First, individual tree crown boundaries are obtained by carrying out segmentation on the DOM. Second, point clouds of the individual trees are extracted based on the boundaries. Third, precise DEM is derived from the point cloud which is classified by a multi-scale curvature classification algorithm. Finally, a height weighted correction method is applied to correct the topological effects. The method is applied to LiDAR data acquired in South China, and its effectiveness is tested using 41 field survey plots. The results show that the terrain impacts the canopy height of individual trees in that the downslope side of the tree trunk is elevated and the upslope side is depressed. This further affects the extraction of the location and crown of individual trees. A strong correlation was detected between the slope gradient and the proportions of returns with height differences more than 0.3, 0.5 and 0.8 m in the total returns, with coefficient of determination R2 of 0.83, 0.76, and 0.60 (n = 41), respectively.

No MeSH data available.


Related in: MedlinePlus

Correlation and residual error of the estimated Lorey’s height before and after terrain correction. (a) Correlation of the estimated Lorey’s height before terrain correction; (b) Residual error of the estimated Lorey’s height before terrain correction; (c) Correlation of the estimated Lorey’s height after terrain correction; (d) Residual error of the estimated Lorey’s height after terrain correction.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4507651&req=5

sensors-15-12133-f013: Correlation and residual error of the estimated Lorey’s height before and after terrain correction. (a) Correlation of the estimated Lorey’s height before terrain correction; (b) Residual error of the estimated Lorey’s height before terrain correction; (c) Correlation of the estimated Lorey’s height after terrain correction; (d) Residual error of the estimated Lorey’s height after terrain correction.

Mentions: To assess the consequence of correcting topographic effects on forest canopy height, the two best regression models were chosen and built according to the maximum R-square stopping rule from 10-fold cross validation, namely Model I and Model II (Table 3). The height of the canopy at the 10th, 40th, 60th, 90th percentile and average height were selected in Model I, the determination coefficient R2 is 0.61, corresponding adjusted R2 is 0.58, RMSE is 2.24 m, the response mean is 12.72 m, and the maximum K-fold R2 is 0.52. The height of the canopy at the 10th, 40th, 50th, 70th, 80th, 90th percentile and average height were selected in Model II, the determination coefficient R2 is 0.77, corresponding adjusted R2 is 0.71, RMSE is 1.86 m, and the response mean is 12.72 m, and the maximum K-fold R2 is 0.62 respectively (Figure 13). Experimental results show that the correlation between Lorey’s height calculated by filed survey and canopy height quantiles after terrain correcting is better than before terrain correcting, which reveals that normalized canopy height point cloud after the terrain correction is closer to the natural formation of forest canopy. It can demonstrate that the method of terrain correction restores natural forms of forest canopy.


Assessing and correcting topographic effects on forest canopy height retrieval using airborne LiDAR data.

Duan Z, Zhao D, Zeng Y, Zhao Y, Wu B, Zhu J - Sensors (Basel) (2015)

Correlation and residual error of the estimated Lorey’s height before and after terrain correction. (a) Correlation of the estimated Lorey’s height before terrain correction; (b) Residual error of the estimated Lorey’s height before terrain correction; (c) Correlation of the estimated Lorey’s height after terrain correction; (d) Residual error of the estimated Lorey’s height after terrain correction.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4507651&req=5

sensors-15-12133-f013: Correlation and residual error of the estimated Lorey’s height before and after terrain correction. (a) Correlation of the estimated Lorey’s height before terrain correction; (b) Residual error of the estimated Lorey’s height before terrain correction; (c) Correlation of the estimated Lorey’s height after terrain correction; (d) Residual error of the estimated Lorey’s height after terrain correction.
Mentions: To assess the consequence of correcting topographic effects on forest canopy height, the two best regression models were chosen and built according to the maximum R-square stopping rule from 10-fold cross validation, namely Model I and Model II (Table 3). The height of the canopy at the 10th, 40th, 60th, 90th percentile and average height were selected in Model I, the determination coefficient R2 is 0.61, corresponding adjusted R2 is 0.58, RMSE is 2.24 m, the response mean is 12.72 m, and the maximum K-fold R2 is 0.52. The height of the canopy at the 10th, 40th, 50th, 70th, 80th, 90th percentile and average height were selected in Model II, the determination coefficient R2 is 0.77, corresponding adjusted R2 is 0.71, RMSE is 1.86 m, and the response mean is 12.72 m, and the maximum K-fold R2 is 0.62 respectively (Figure 13). Experimental results show that the correlation between Lorey’s height calculated by filed survey and canopy height quantiles after terrain correcting is better than before terrain correcting, which reveals that normalized canopy height point cloud after the terrain correction is closer to the natural formation of forest canopy. It can demonstrate that the method of terrain correction restores natural forms of forest canopy.

Bottom Line: Finally, a height weighted correction method is applied to correct the topological effects.The method is applied to LiDAR data acquired in South China, and its effectiveness is tested using 41 field survey plots.The results show that the terrain impacts the canopy height of individual trees in that the downslope side of the tree trunk is elevated and the upslope side is depressed.

View Article: PubMed Central - PubMed

Affiliation: Key Laboratory of Digital Earth Science, Institute of Remote Sensing and Digital Earth (RADI), Chinese Academy of Science, Haidian District, Beijing 100094, China. zjj@mail.csu.edu.cn.

ABSTRACT
Topography affects forest canopy height retrieval based on airborne Light Detection and Ranging (LiDAR) data a lot. This paper proposes a method for correcting deviations caused by topography based on individual tree crown segmentation. The point cloud of an individual tree was extracted according to crown boundaries of isolated individual trees from digital orthophoto maps (DOMs). Normalized canopy height was calculated by subtracting the elevation of centres of gravity from the elevation of point cloud. First, individual tree crown boundaries are obtained by carrying out segmentation on the DOM. Second, point clouds of the individual trees are extracted based on the boundaries. Third, precise DEM is derived from the point cloud which is classified by a multi-scale curvature classification algorithm. Finally, a height weighted correction method is applied to correct the topological effects. The method is applied to LiDAR data acquired in South China, and its effectiveness is tested using 41 field survey plots. The results show that the terrain impacts the canopy height of individual trees in that the downslope side of the tree trunk is elevated and the upslope side is depressed. This further affects the extraction of the location and crown of individual trees. A strong correlation was detected between the slope gradient and the proportions of returns with height differences more than 0.3, 0.5 and 0.8 m in the total returns, with coefficient of determination R2 of 0.83, 0.76, and 0.60 (n = 41), respectively.

No MeSH data available.


Related in: MedlinePlus