Assessment of approaches for monitoring forest structure dynamics using bi-temporal digital aerial photogrammetry point clouds

• Published in: Remote sensing of environment 2021-03, Vol.255, p.112300, Article 112300
• Main Authors: Fu, Xiaoyao, Zhang, Zhengnan, Cao, Lin, Coops, Nicholas C., Goodbody, Tristan R.H., Liu, Hao, Shen, Xin, Wu, Xiangqian
• Format: Article
• Language: English
• Subjects: Aerial photography; Bi-temporal; Canopies; Carbon cycle; Digital aerial photogrammetry; Dynamic structural analysis; Forest ecosystems; Forest growth; Forest management; Forest resources; Forest structural parameters; Forestry; Forests; Inventories; Mathematical models; Monitoring; Parameter estimation; Parameter sensitivity; Photogrammetry; Resource management; Spatial data; Spatial discrimination; Spatial resolution; Subtropical forest; Sustainable forestry; Terrestrial environments; Three dimensional models; Unmanned aerial vehicle; Unmanned aerial vehicles
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2021.112300

Assessing changes in forest structure over time is crucial for monitoring forest resources, supporting sustainable forest management practices, and providing key insights into changes in the terrestrial carbon cycle. Recent research interest and rapid growth of unmanned aerial vehicle (UAV)-based digital aerial photogrammetry (DAP) technology principally due to its low cost and timeliness, is providing high-spatial resolution data for enhancing forest inventories and forest dynamics monitoring. The increasing prevalence of these UAV acquired datasets in forestry promotes a need for better understanding of how DAP-based point clouds change over time, and how these changes may relate to changes in forest structure. In this study, we utilized bi-temporal DAP data to investigate changes in forest structure over a 3-year period in subtropical planted forest stands throughout Dongtai Yellow Sea National Forest Park, Jiangsu Province, China. To do so, we evaluated both direct (i.e., structural parameter changes estimated using the differences between bi-temporal DAP metrics) and indirect (i.e., structural parameters were modelled for each date, and their changes predicted as their differences) methods to estimate the changes in forest structure. In addition, once models were developed, changes in Lorey's mean height and volume were extrapolated across the entire study site and examined related to the forest type and age. Our assessments of the different approaches showed that the direct approach (R2 = 0.54–0.78) resulted in improved accuracy compared to the indirect method (R2 = 0.51–0.73). The distributional metrics, namely, height percentiles (e.g., H75 and H95) and canopy return density (e.g., D7 and D5), and the Weibull-fitted metrics (e.g., α) were found to be sensitive to changes in structural parameters, whereas canopy volume-related metrics had relatively low predictive capabilities. Overall, the predicted changes in Lorey's mean height and volume were mapped over the entire study area, and indicated that Lorey's mean height increased mostly in middle-aged and young forest stands. Over-mature stands showed the lowest height increment. This study proved the capability of using bi-temporal point clouds from UAV-based DAP for enhancing forest inventories and promoting sustainable forest management. •Using bi-temporal DAP point clouds data to assess forest structure dynamics.•Development of direct and indirect approaches to assess changes in forest structure.