UAV-based estimation of aboveground biomass carbon in the Sundarbans mangrove forest of Bangladesh

Abstract

Reliable biomass and carbon assessments are essential for understanding mangrove forest dynamics and their role in global carbon cycling. However, field-based biomass estimation is time-consuming, labor-intensive, and costly. Remote sensing provides a practical and cost-effective alternative, particularly with unmanned aerial vehicles (UAVs) that can capture high-resolution imagery. The assessment of biomass carbon in the Sundarbans Mangrove Forest (SMF) using high-resolution remote sensing remains understudied. This study develops an integrated carbon assessment framework for estimating aboveground biomass carbon (AGBC) in the SMF using UAV-derived canopy metrics and field measurements. Error propagation and feature importance analyses were performed to quantify uncertainty and assess predictor influence, providing a robust and uncertainty-aware approach for mangrove biomass carbon assessment. crown projection area (CPA) was used to estimate proxy diameter at breast height (DBH) using linear, polynomial, logarithmic, power, and exponential regression models. All CPA–DBH models showed comparable performance (R2 = 0.73–0.75). The linear and polynomial models achieved the lowest root mean square error (RMSE) (1.77 cm), while the polynomial model produced the lowest mean absolute error (1.22 cm) and was therefore selected for estimating Proxy DBH. Field-based AGBC was calculated from measured tree height and DBH using species-specific and generic allometric equations applicable to the Sundarbans, whereas UAV-based AGBC was estimated using UAV-derived Mean DSM Height and CPA-derived Proxy DBH. Simple linear regression, multiple linear regression (MLR), and machine-learning (ML) models (Random Forest, Support Vector Machine, Artificial Neural Network, and Extreme Gradient Boosting) were evaluated for AGBC prediction. The MLR model showed the best performance on the validation/model-selection subset (R2 = 0.86, RMSE = 14.70 Mg C ha−1, MedAE = 4.58 Mg C ha−1). UAV-derived AGBC estimates closely matched field measurements and previously reported values for the Sundarbans, demonstrating the potential of high-resolution UAV metrics for reliable mangrove AGBC estimation and supporting carbon monitoring, REDD+ initiatives, and sustainable mangrove management in Bangladesh.