Abstract: Mangroves, located in the intertidal zones where the sea meets the land, exhibit a strong capacity for carbon sequestration. They represent a typical coastal blue carbon ecosystem, playing a significant role in the global carbon cycle. As a critical indicator for assessing the ecological environment and carbon stock, mangrove biomass holds paramount importance for both mangrove conservation and carbon stock estimation. Targeting the Shankou Mangrove Nature Reserve in Guangxi, this study extracted a total of 124 features using three data combinations unmanned aerial vehicle-light detection and ranging (UAV-LiDAR) data, Sentinel-2 data, and integrated data of both. Then, it conducted an above-ground biomass (AGB) inversion in mangroves using machine learning-based regression models with essential variables, followed by precision evaluation for each model. The primary conclusions can be drawn as follows. Compared to UAV-LiDAR or Sentinel-2 data, the integrated data yielded a superior model for predicting the mangrove AGB, with its verification accuracy evinced by a coefficient of determination (R²) of 0.77 and a root mean square error (RMSE) of 19 t/hm². In the variable importance ranking of the model constructed from the integrated data, the top 25 variables include 20 from the UAV-LiDAR features and 5 from the Sentinel-2 features. The top 5 variables were zq99 (99th percentile height), MDI2 (mangrove discrimination index 2), NDVIre1 (red-edge normalized difference vegetation index 1), zmax (maximum height), and zq20 (20th percentile height). This demonstrates that the UAV-LiDAR data serve as the dominant contributors to the AGB inversion. The total AGB of mangroves in the Shankou Mangrove Nature Reserve within Guangxi was estimated to be 3 855 602 t, with an average AGB of 41.7 t/hm². The estimated AGB exhibited an overall spatial distribution pattern of lower in the west and higher in the east, which is consistent with results from previous research.