Abstract: Aimed at the problems existing in the organization and management of multi-source heterogeneous Earth observation data, in the report, a requirement-oriented adaptive tile data cube （RATD） based on the Beidou grid was proposed, which aimed to achieve dynamic organization and efficient management of multi-source heterogeneous Earth observation data. The finite state machine （FSM） was introduced for the dynamic update modeling of tile data. Based on the requirements, the tile data was generated and updated. The surface cover classification of Fujian Province was used as a case, the feasibility of the model was verified. The results suggested that the overall classification accuracy and Kappa coefficient of the updated RATD were both above 0.93, in addition, the efficiency of data acquisition process and the required time of RATD was better than that of the traditional methods significantly. Our findings provide the new ideas and methods for the efficient management and application of earth observation（EO） data.