Estimating coastal mangrove aboveground biomass of Zhejiang province based on sentinel-2 images and multiple machine learning algorithms

Mangrove forest is a tropical or subtropical woody plant community, which plays an important role in carbon sequestration and oxygen release, improving the bay environment and biodiversity. The aboveground biomass (AGB) of mangrove forests is a key parameter to evaluate and reflect the carbon cycle capacity and productivity level of mangroves, and it is very important to accurately estimate mangrove forest AGB. In recent years, machine learning algorithms have achieved good results in forest AGB estimation, but the performance of mangrove AGB inversion in subtropical regions remains undemonstrated. Based on Sentinel-2 imagery and ground survey data, this study employed five machine learning algorithms—random forest (RF), categorical boosting (CATboost), extreme gradient boosting (XGboost), support vector machine (SVM), and local cascade ensemble (LCE)—to construct an aboveground biomass (AGB) model for artificially introduced mangrove areas along the coast of Zhejiang Province, China. The results show that texture variables contributed the most to coastal mangrove forest AGB. All models perform well in predicting mangrove forest AGB, among which had the highest accuracy with an R 2 of 0.68 and RMSE of 6.85 Mg ha −1 based on XGboost model. The spatial distribution of mangrove AGB was estimated based on the optimal XGboost model. The total mangrove AGB in the region was 5930.28 Mg and the average AGB was 11.35 Mg ha −1 . This study can provide a basis for the management of coastal mangrove resources and the assessment of carbon sink capacity in Zhejiang Province.