Detection of forest understory evergreen trees in a deciduous forest using UAV-LiDAR and RGB data during the deciduous period

Forest ecosystems play a critical role in maintaining key ecosystem services and functions. However, the impacts of climate change have become apparent in recent years. Long-term forest observation data suggest that deciduous forests are being replaced by evergreen tree species, potentially as a result of climate change. Detecting such ecosystem changes at an early stage is crucial for assessing the impacts of climate change and implementing effective management. In this study, we aim to develop a method for identifying evergreen trees in deciduous broadleaf forests where the effects of climate change are becoming apparent. Specifically, we employed cost-effective and efficient UAV-LiDAR technology. By focusing on the deciduous season, we effectively enhanced the detection of evergreen trees, as their presence becomes more distinguishable during this period.

The study was conducted in a 6-hectare plot within the deciduous broadleaf forests of the Ogawa Forest Reserve in Japan, a site where long-term forest monitoring has been conducted. This site harbors Pieris japonica subsp. japonica (hereafter, PJ), an evergreen shrub that has shown an increase in recent years. Other species that retain green leaves during the deciduous season, such as dwarf bamboo (Sasa) and epiphytic plants, are also present. To ensure effective detection of PJ, we first stratified the acquired LiDAR data into different canopy layers (upper canopy trees and multiple understory layers). We then determined the required point density for rational segmentation of PJ. Using RGB data, we extracted "green" points for each canopy layer. This method effectively excluded dwarf bamboo and epiphytic plants, enabling the accurate extraction of PJ. The study demonstrated that combining UAV-LiDAR with RGB data is highly effective for identifying understory evergreen trees. This approach facilitates the extraction of "green" objects by canopy layer in deciduous forests during the deciduous season. This method would be not only efficient for detecting forest changes but also applicable to identifying invasive species and enhancing forest management practices.