A new method for estimating forest aboveground biomass by integrating sUAV and satellite remote sensing

Abstract

Aboveground biomass (AGB) is a fundamental indicator for assessing forest ecosystem productivity, biodiversity, and carbon sequestration levels. In AGB remote sensing estimation, traditional methods face challenges such as the limited number of plots and issues with matching plots to remote sensing pixels. The insufficient number of plots restricts spatial coverage, while difficulties in plot-to-pixel matching constrain the application of remote sensing data. The use of small Unmanned Aerial Vehicle (sUAV) Light Detection and Ranging (LiDAR) in forest AGB estimation provides a new solution for high-precision and efficient forest resource monitoring. sUAV LiDAR enables the acquisition of detailed point cloud data at a local scale. Compared to traditional remote sensing, sUAV LiDAR offers higher accuracy at smaller spatial scales and eliminates the need for extensive plot establishment. Furthermore, due to the high spatial resolution of LiDAR data, we can extract surface data obtained by the sUAV and cut it using satellite pixels, thereby avoiding the plot-to-pixel matching issue and achieving precise integration from data acquisition to AGB estimation. The main findings are as follows: (1) Using sUAV LiDAR data, features such as canopy closure, leaf area index, and height-related variables were extracted. A Random Forest (RF) algorithm was used to construct an AGB model based on data from 65 in-situ plots (R2 = 0.91, RMSE = 10.94 Mg/ha), and an AGB distribution map of sUAV plots was generated. (2) A total of 1522 samples consistent with the spatial resolution of Sentinel-2 imagery were extracted from the sUAV plot AGB distribution map, and an sUAV-Sentinel-2 (S2) model was established using the RF algorithm (R2 = 0.76, RMSE = 12.44 Mg/ha). (3) Based on the sUAV-S2 model, the forest AGB in the Helan Mountains was successfully estimated, and an AGB map with a spatial resolution of 30 m × 30 m was generated. Compared to the AGB value of 660,360.85 Mg published by the Helan Mountains National Nature Reserve Administration, the sUAV-S2 model achieved a prediction accuracy of 95.68%. We also constructed a plot-S2 model that directly connects ground data to satellite data, which achieved an R2 of 0.69 in the testing phase, further validating the superiority of the sUAV-S2 model. The innovation of this study lies in the integration of sUAV LiDAR with satellite remote sensing, reducing the reliance on traditional ground truth data and establishing a "ground-air-space" integrated research framework. This innovative approach overcomes the limitations of traditional remote sensing technologies in forest resource monitoring, providing new data support and technical pathways for future satellite-based AGB estimation models.