Assessment of the Impact of Coastal Wetland Saltmarsh Vegetation Types on Aboveground Biomass Inversion

Coastal wetlands play an important carbon sequestration role in China’s “carbon peaking” and “carbon neutrality” goals. Monitoring aboveground biomass (AGB) is crucial for wetland management. Satellite remote sensing enables efficient retrieval of AGB. However, a variety of statistical models can be used for biomass inversion, depending on factors such as the vegetation type and inversion method. In this study, Landsat 8 Operational Land Imager (OLI) images were preprocessed in the study area through radiation calibration and atmospheric correction for modeling. In terms of model selection, 13 different models, including the univariate regression model, multiple regression model, and machine learning regression model, were compared in terms of their accuracy in estimating the biomass of various wetland vegetation types under their respective optimal parameters. The findings revealed that: (1) the regression models varied across vegetation types, with the accuracy of the biomass estimates decreasing in the order of Scirpus spp. > Spartina alterniflora > Phragmites australis; (2) overall modeling, without distinguishing vegetation types, addressed the challenges of limited samples availability and sampling difficulty. Among them, the random forest regression model outperformed the others in estimating wet and dry AGB with R2 values of 0.806 and 0.839, respectively. (3) Comparatively, individual modeling of vegetation types can better reflect the biomass of each wetland vegetation type, especially the dry AGB of Scirpus spp., whose R2 and RMSE values increased by 0.248 and 11.470 g/m2, respectively. This study evaluates the impact of coastal saltmarsh vegetation types on biomass estimation, providing insights into biomass dynamics and valuable support for wetland conservation and restoration, with potential contributions to global habitat assessment models and international policies like the 30x30 Conservation Agenda.