Tree extraction and estimation of walnut structure parameters using airborne LiDAR data

Abstract

The development of new tools based on remote sensing data in agriculture contributes to cost reduction, increased production, and greater profitability. Airborne LiDAR (Light Detection and Ranging) data show a significant potential for geometrically characterizing tree plantations. This study aims to develop a methodology to extract walnut (Juglans regia L.) crowns under leafless conditions using airborne LiDAR data. An original approach based on the alpha-shape algorithm, identification of local maxima, and k-means algorithms is developed to extract the crowns of walnut trees in a plot located in Viver (Eastern Spain) with 192 trees. In addition, stem diameter and volume, crown diameter, total height, and crown height were estimated from cloud metrics and other 2D parameters such as crown area, and diameter derived from LiDAR data. A correct identification was made of 178 trees (92.7%). For structure parameters, the most accurate results were obtained for crown diameter, stem diameter, and stem volume with coefficient of determination values (R2) equal to 0.95, 0.87 and 0.83; and RMSE values of 0.43 m (5.70%), 0.02 m (9.35%) and 0.016 m3 (21.55%), respectively. The models that gave the lowest R2 values were 0.69 for total height and 0.70 for crown height, with RMSE values of 0.84 m (12.4%) and 0.83 m (14.5%), respectively. A suitable definition of the central and lower parts of tree canopies was observed. Results of this study generate valuable information, which can be applied for improving the management of walnut plantations.

Keywords

• Airborne laser scanning
• Alpha-shape algorithm
• Precision agriculture
• Cloud metrics
• Dendrometry