Assessing the accuracy of forest above-ground biomass and carbon storage estimation by meta-analysis based close-range remote sensing.

Abstract

The swift progress of close-range remote sensing necessitates a quantitative evaluation of its accuracy in estimating forest above-ground biomass (AGB) across diverse scales, forest types, methodologies, and variables. These evaluations will enhance the effectiveness of remote sensing in forest monitoring, reveal the carbon sequestration capability of forest vegetation, and underscore the critical function of forests as terrestrial carbon sinks. In this study, we designated R ² as the effect size for the meta-analysis given that it is commonly regarded as a measure for estimating accuracy in AGB research, which indicates the explanatory capacity of independent variables. Utilizing 187 global investigations and 233 datasets, this research systematically assessed the accuracy (R ²) of ground light detection and ranging (LiDAR), unmanned aerial vehicles (UAVs), spectra, and red-green-blue (RGB) sensors across the single-tree, plot, and stand scales. The discrepancies in accuracy across the various research methods and the independent variables in the allometric growth equation were also assessed. The research indicated that ground lidar exhibited the best accuracy across all studies and was the most effective approach at both the single-tree and plot scales. Nonetheless, as the scale of the research broadened, both accuracy and sample size diminished. Furthermore, the variations from different approaches among different forest types were substantial; therefore, it was necessary to model these forest types explicitly. By integrating diameter at breast height (DBH or D) and tree height (H) as independent variables in the allometric growth equation, the method showed improved estimation accuracy. The estimation of AGB must address the issue of accumulated error arising from the interconversion of DBH and H, single-tree segmentation, and specific allometric growth equations, which are subsequently compounded at the plot and stand levels. Close-range remote sensing is currently the most efficient method for estimating forest AGB, surpassing conventional measurement techniques. Yet, due to sensor limitations, no single sensor achieved optimal results independently. The integration of multi-source data and scale adaptation strategies further enhanced the efficacy of close-range remote sensing, surpassing the conventional survey methods. Moving forward, efforts should prioritize cross-platform data standardization, deep learning model refinement, and the establishment of non-destructive validation systems to support high-precision forest carbon monitoring, in alignment with carbon management goals.