Adaptive algorithm for large scale dtm interpolation from lidar data for forestry applications in steep forested terrain

Light Detection and Ranging (lidar) has become a valuable tool in forest inventory because it yields accurate measurements of tree heights. However, tree height can be accurate only if the height of the ground, i. e., the Digital Terrain Model (dtm) is first accurately established. gh great advances have been made in lidar technology over the past decade, filtering lidar data for Digital Terrain Model (dtm) interpolation is still a challenge, especially in steep and complex terrain with forest cover. Several algorithms proposed in the literature address this challenge but their performance deteriorates with the decreasing point density caused by the presence of forest cover and steep slopes. In this paper, we propose a new adaptive algorithm for dtm interpolation from lidar data in steep terrain with forest cover. The algorithm partitions the input data and estimates a section of the dtm by fitting a linear or quadratic trend surface, or uses cubic spline interpolation depending on the complexity of the section of terrain. The performance of the algorithm is tested in three ways: by visual assessment, by comparison of the tree-height estimates produced using the generated dtm with those obtained using field survey, and by use of International Society for Photogrammetry and Remote Sensing (isprs) test data. Test results show that the algorithm can cope well with steep slopes and low lidar point densities, giving a more accurate estimate of average tree height compared to conventional algorithms. The algorithm can be used for dtm extraction in large scale forest inventory projects in challenging environments–complex terrain and low lidar point densities.