Assessment of Salt Marsh Above-Ground Biomass Using UAV Multispectral Imagery and Open-Access Canopy Height Models

This study aims to identify and classify three dominant estuarine species (Juncus maritimus, Phragmites australis, and Spartina patens) in the Lima River estuary, Portugal, and develop species-specific AGB regression models utilising UAV multispectral imagery and dronederived canopy height models (CHM). A Random Forest classifier mapped the landscape with a 91.9% overall accuracy. Utilising 75 destructive field samples, 504 predictive models combining vegetation indices and height variables were evaluated, revealing that model performance was heavily conditioned by species. J. maritimus yielded an additive-exponential model (R² = 0.834) driven by NDVI and CHM height. In contrast, models for P. australis (R² = 0.498) and the mat-forming S. patens (R² = 0.453) exhibited weaker fits and structural limitations resulting in extrapolation artefacts. Spatially extrapolating these models across 40.20 hectares demonstrated that while the UAV framework is highly reproducible, structurally complex species will necessitate higherresolution 3-D structural metrics for accurate AGB prediction.