Comparison of Different Sampling Density Data for Detecting and Measuring Individual-trees in a Mountainous Coniferous Forest using Small-footprint Airborne LiDAR(<Special Issue>Silvilaser)

この論文をさがす

抄録

This study investigated the effects of laser-sampling density on individual-tree detection and tree height estimation changing the sampling density by overlapping three flight data in a mountainous coniferous forest. The LiDAR system used in this study was mounted on a fixed-wing aircraft. The study area was closed-canopy, middle-aged Japanese cedar (Cryptomeria japonica) plantation in Japan. We prepared three sets of single flight data (3.2 points/m^2), three sets of double-overlapping data (6.5 points/m^2) consisted of two single flight data, and one set of triple-overlapping data (9.7 points/m^2) consisted of three single flight data within this study plot. Namely, a total of seven datasets were used in the analysis. The numbers of detected same trees among same laser-sampling density datasets were different and increased with the increase of the density. The detection rate of same trees among all datasets was approximately 55%, and the detected trees belonged to dominant and co-dominant trees within the plot. In all datasets, we found that if a given field tree has relatively lower treetop-elevation and smaller crown radius than that of the nearest field tree, and these trees are close to each other, the lower tree is difficult to detect in mountainous coniferous forest. But the number of detected small trees between 10m and 18m height increased with the increase of laser-sampling density. LiDAR-derived median and mean tree heights were slightly greater than that of field measured tree height in this study site. Although there were significant differences between field measured and LiDAR-derived tree heights for all datasets (p<0.01), the difference between maximum and minimum RMSE for tree height estimates was only 0.17m and the maximum RMSE was 1.02m. All results of this study indicate that although greater laser-sampling density data can provide information of more varying tree size, 3 or 4 points/m^2 of laser-sampling density data would provide accurate individual-tree detection of upper-storey trees and tree height estimates, given as RMSE, is approximately 1m in middle-aged Japanese cedar forests in mountainous areas.

収録刊行物

参考文献 (32)*注記

もっと見る

詳細情報 詳細情報について

問題の指摘

ページトップへ