Mangrove ecosystem is one of the most productive ecosystems on the earth, and it is one of the important components of the coastal "blue carbon" ecosystem. As an important part of mangrove blue carbon, obtaining mangrove aboveground biomass accurately and quickly has become one of the hot issues in mangrove ecosystem research. Analyzing spatial distribution pattern and magnitude of Aboveground biomass (AGB) of mangroves in the Maowei Sea of Beibu Gulf can provide the scientific basis for the protection of the mangrove ecological environment and the ecological restoration of “South Red and North Willow” in this area. As a domestic civil hyperspectral satellite independently developed by China, the hyperspectral data of ZY-1-02D provides a new opportunity to research mangrove aboveground biomass. Because of its high performance and efficiency, machine learning algorithms are increasingly used in mangrove-related research. It has become an important means to obtain mangrove parameter information. How accurate is the retrieval of hyperspectral data in mangrove aboveground biomass, whether the domestic hyperspectral satellite data and machine learning algorithm can be applied to the estimation of mangrove aboveground biomass needs further verification. Based on ZY-1-02D Satellite hyperspectral data, three different machine learning algorithms, eXtreme Gradient Boosting (XGBoost), Random Forest Regression (RFR) and k-nearest neighbor regression (KNNR), were used to estimate the biomass of mangrove forests in the Maowei Sea. On this basis, the performance of different machine learning algorithms was compared. The results showed that: (1) The average aboveground biomass of Sonneratia apetala mangrove was the highest (90.93 Mg·ha-1), followed by Aegiceras corniculatum mangrove(52.63 Mg·ha-1), and Kandelia candel mangrove was the lowest (20.27 Mg·ha-1). (2) XGBoost, RF and KNN machine learning algorithms are used to model mangrove aboveground biomass and mangrove spectral variables. The XGBoost model based on log reciprocal first-order transformation has the highest accuracy and is the best machine learning model. In the testing phase, R2=0.751 5, RMSE=27.494 8 Mg·ha-2. (3) Based on the ZY-1-02D Satellite hyperspectral data, the XGBoost algorithm is used to retrieve the aboveground biomass of mangroves in Maowei Sea, which is between 4.58 and 208.35 Mg·ha-2, with an average value of 51.92 Mg·ha-2. The aboveground biomass shows a spatial distribution pattern of low in the middle and high on both sides. In a word, this paper demonstrates that the combination of domestic hyperspectral data and XGBoost machine learning algorithm has a good application prospect in the estimation of mangrove biomass, which can provide scientific basis and technical support for the ecological restoration and protection of Maowei Sea mangroves.