Integration of UAS and Backpack-LiDAR to Estimate Aboveground Biomass of Picea crassifolia Forest in Eastern Qinghai, China

Precise aboveground biomass (AGB) estimation of forests is crucial for sustainable carbon management and ecological monitoring. Traditional methods, such as destructive sampling, field measurements of Diameter at Breast Height with height (DBH and H), and optical remote sensing imagery, often fall short in capturing detailed spatial heterogeneity in AGB estimation and are labor-intensive. Recent advancements in remote sensing technologies, predominantly Light Detection and Ranging (LiDAR), offer potential improvements in accurate AGB estimation and ecological monitoring. Nonetheless, there is limited research on the combined use of UAS (Uncrewed Aerial System) and Backpack-LiDAR technologies for detailed forest biomass. Thus, our study aimed to estimate AGB at the plot level for Picea crassifolia forests in eastern Qinghai, China, by integrating UAS-LiDAR and Backpack-LiDAR data. The Comparative Shortest Path (CSP) algorithm was employed to segment the point clouds from the Backpack-LiDAR, detect seed points and calculate the DBH of individual trees. After that, using these initial seed point files, we segmented the individual trees from the UAS-LiDAR data by employing the Point Cloud Segmentation (PCS) method and measured individual tree heights, which enabled the calculation of the observed/measured AGB across three specific areas. Furthermore, advanced regression models, such as Random Forest (RF), Multiple Linear Regression (MLR), and Support Vector Regression (SVR), are used to estimate AGB using integrated data from both sources (UAS and Backpack-LiDAR). Our results show that: (1) Backpack-LiDAR extracted DBH compared to field extracted DBH shows about (R2 = 0.88, RMSE = 0.04 m) whereas UAS-LiDAR extracted height achieved the accuracy (R2 = 0.91, RMSE = 1.68 m), which verifies the reliability of the abstracted DBH and height obtained from the LiDAR data. (2) Individual Tree Segmentation (ITS) using a seed file of X and Y coordinates from Backpack to UAS-LiDAR, attaining a total accuracy F-score of 0.96. (3) Using the allometric equation, we obtained AGB ranges from 9.95-409 (Mg/ha). (4) The RF model demonstrated superior accuracy with a coefficient of determination (R2) of 89%, a relative Root Mean Square Error (rRMSE) of 29.34%, and a Root Mean Square Error (RMSE) of 33.92 Mg/ha compared to the MLR and SVR models in AGB prediction. (5) The combination of Backpack-LiDAR and UAS-LiDAR enhanced the ITS accuracy for the AGB estimation of forests. This work highlights the potential of integrating LiDAR technologies to advance ecological monitoring, which can be very important for climate change mitigation and sustainable environmental management in forest monitoring practices.