Canopy Gap Mapping from Airborne Laser Scanning: An Assessment of the Positional and Geometrical Accuracy

Remote Sensing

Stéphanie Bonnet³  Rachel Gaulton¹  François Lehaire³  Philippe Lejeune³  Nicolas Baghdadi²  Heiko Balzter² 

[1] School of Civil Engineering and Geosciences, Newcastle University, Newcastle upon Tyne NE1-7RU, UK; E-Mail:Unit of Forest Management, Department of Biosystem Engineering, University of Liège, Gembloux Agro-Bio Tech, 2 Passage des déportés, 5030 Gembloux, Belgium;;Unit of Forest Management, Department of Biosystem Engineering, University of Liège, Gembloux Agro-Bio Tech, 2 Passage des déportés, 5030 Gembloux, Belgium; E-Mails:

关键词: forest canopy gaps;    ALS;    raster-based approach;    segmentation;    random forest;    geometric accuracy;    stand types;   

DOI  :  10.3390/rs70911267

来源: mdpi

【 摘 要 】

Canopy gaps are small-scale openings in forest canopies which offer suitable micro-climatic conditions for tree regeneration. Field mapping of gaps is complex and time-consuming. Several studies have used Canopy Height Models (CHM) derived from airborne laser scanning (ALS) to delineate gaps but limited accuracy assessment has been carried out, especially regarding the gap geometry. In this study, we investigate three mapping methods based on raster layers produced from ALS leaf-off and leaf-on datasets: thresholding, per-pixel and per-object supervised classifications with Random Forest. In addition to the CHM, other metrics related to the canopy porosity are tested. The gap detection is good, with a global accuracy up to 82% and consumer’s accuracy often exceeding 90%. The Geometric Accuracy (GAc) was analyzed with the gap area, main orientation, gap shape-complexity index and a quantitative assessment index of the matching with reference gaps polygons. The GAc assessment shows difficulties in identifying a method which properly delineates gaps. The performance of CHM-based thresholding was exceeded by that of other methods, especially thresholding of canopy porosity rasters and the per-pixel supervised classification. Beyond assessing the methods performance, we argue the critical need for future ALS-based gap studies to consider the geometric accuracy of results.

【 授权许可】

CC BY   
© 2015 by the authors; licensee MDPI, Basel, Switzerland.