LiDAR mapping of canopy gaps in continuous cover forests: A comparison of canopy height model and point cloud based techniques

Abstract

In continuous cover forest systems, canopy gaps are created by management activities with an aim of encouraging natural regeneration and of increasing structural heterogeneity. Light Detection and Ranging (LiDAR) may provide a more accurate means to assess gap distribution than ground survey, allowing more effective monitoring. This paper presents a new approach to gap delineation, based on identifying gaps directly from the point cloud and avoiding the need for interpolation of returns to a canopy height model (CHM). Areas of canopy are identified through local maxima identification, filtering and clustering of the point cloud, with gaps subsequently delineated in a GIS environment.

When compared to field surveyed gap outlines, the algorithm has an overall accuracy of 88% for data with a high LiDAR point density (11.4 returns per m²) and accuracy of up to 77% for lower density data (1.2 returns per m²). The method provides an increase in overall and Producer's accuracy of 4 and 8% respectively, over a method based on the use of a CHM. The estimation of total gap area is improved by, on average, 16% over the CHM based approach. Results indicate that LiDAR data can be used accurately to delineate gaps in managed forests, potentially allowing more accurate and spatially explicit modelling of understorey light conditions.

Acknowledgements

This work was funded by a NERC CASE studentship with the Forestry Commission (NER/S/A/2004/12732) and was completed whilst the corresponding author was based at the University of Edinburgh (School of GeoSciences). The NERC Airborne Research and Survey Facility supplied LiDAR data for Glasfyndd and Clocaneog whilst LiDAR for Aberfoyle was acquired by the Environment Agency and supplied by the Forestry Commission. We would also like to thank the NERC FSF and GEF for providing ground instrumentation and assistance. We are very grateful to Sophie Hale (Forest Research) and the Tyfiant Coed Project at the University of Wales, Bangor for access to field data from previous surveys and to all those who assisted during the fieldwork, especially V. Barron and P. Wieland. Particular thanks are due to J.C. Suárez, I.H. Woodhouse and A. Pommerening for advice and support and to the anonymous reviewers who provided valuable suggestions for improvements.