Remote Sensing Aboveground Biomass Inversion of Four Vegetation Types in the Nanji Wetland

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Aboveground biomass (AGB) serves as a crucial indicator for assessing vegetation carbon sequestration capacity. While AGB levels vary significantly across different vegetation types and regions, the spatial distribution of AGB for specific wetland communities remains poorly characterized. To address this, we integrated field-collected data with Sentinel-2 spectral bands and remote sensing indices, employing random forest (RF) regression and Backpropagation Neural Network (BPNN) for AGB modeling. Through comparative evaluation of their inversion performance, the optimal model was selected to estimate vegetation AGB in the Nanji Wetland. By incorporating wetland classification data, we further generated spatial distribution maps of AGB for four dominant vegetation types during the dry season. The main findings are as follows. Important variables for the RF model included spectral bands B12, B11, B3, B2, B9, B1, B8, B6, and B4 and the Modified Normalized Difference Water Index (MNDWI), Normalized Difference Water Index (NDWI), Kernel Normalized Difference Vegetation Index (KNDVI), and Simple Ratio Index (SR). RF demonstrated significantly higher predictive accuracy (R2 = 0.945, RMSE = 109.205 g·m−2) compared to the BPNN (R2 = 0.821, RMSE = 176.025 g·m−2). The total estimated AGB reached 4.03 × 109 g; Carex spp. dominated AGB accumulation (1.49 × 109 g), followed by P. australis spp. (6.69 × 108 g), M. lutarioriparius spp. (4.60 × 108 g), and Polygonum spp. (3.61 × 108 g). The AGB exhibited a clear spatial gradient, decreasing from higher-elevation lakeshore areas towards the central lake. The results provide detailed spatial quantification of AGB stocks across dominant vegetation types, revealing distinct spatial characteristics and interspecies variations in AGB. This study offers a valuable baseline and methodological framework for monitoring wetland carbon dynamics.
Funding Information
  • National Key Research and Development Program of China (2020YFD1100603, 82-Y50G22-9001-22/23, 41361049, ZRKJ20242512)
  • National Natural Science Foundation of China (2020YFD1100603, 82-Y50G22-9001-22/23, 41361049, ZRKJ20242512)
  • Science and Technology Innovation Project of Jiangxi Provincial Department of Natural Resources (2020YFD1100603, 82-Y50G22-9001-22/23, 41361049, ZRKJ20242512)

This publication has 48 references indexed in Scilit: