Regional mangrove vegetation carbon stocks predicted integrating UAV-LiDAR and satellite data

Using satellite RS data predicting mangrove vegetation carbon stock (MVC) is the popular and efficient approach at a large scale to protect mangroves and promote carbon trading. Satellite data have performed poorly in predicting MVC due to saturation issues. UAV-LiDAR data overcomes these limitations by providing detailed structural vegetation information. However, how to cross-scale integration of UAV-LiDAR and satellite RS data and the selection of features and machine learning methods hampered the practitioner in making a lightweight but efficient model to predict the MVC. Our study integrated UAV-LiDAR, Sentinel-1, and Sentinel-2 to extract spectral, structural, and textural features at the regional scale. We estimated the influences of different combinations between three vegetation features and machine learning methods (Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting Regression Tree (GBDT), and Extreme Gradient Regression Tree (XGBOOST)) on the results of MVC prediction, and constructed a framework for estimating mangrove vegetation aboveground (ACG) and belowground (BCG) carbon storage in Zhanjiang, the largest mangrove area of China. Our research shows: 1) Compared to using satellite remote sensing (RS), integrating UAV and satellite RS data and fusing multiple vegetation features significantly improved the accuracy of mangrove vegetation carbon stock (MVC) predictions. 2) Structural features, particularly canopy height retrieved from UAV and satellite RS, are essential indicators for predicting MVC. Combined with spectral and structural features, regional MVC was precisely predicted. 3)Although the influence of different machine learning methods on MVC prediction was not significant, XGBOOST demonstrated relatively high precision. We recommend that mangrove practitioners integrate UAV and satellite RS data to predict MVC at a regional scale. Importantly, governments should prioritize the application of UAV-LiDAR in forestry monitoring and establish a long-term mangrove monitoring database to aid in estimating blue carbon resources and promoting blue carbon trading.