Abstract: Coal mine water disasters are one of the key risk factors restricting coal mine safety production, and accurate early warning is the core link of water disaster prevention and control. In view of the problems of insufficient multi-source data linkage and limited early warning accuracy and timeliness caused by threshold staticization in the traditional early warning system, an integrated early warning system of surface water - groundwater - mine water inflow is constructed. The system integrates real-time monitoring data such as surface water level, groundwater pressure, mine water inflow and others, and uses spatiotemporal correlation analysis model to carry out multi-source data fusion and trend prediction. By dynamically correcting early warning thresholds, real-time perception of dynamic changes in mine water inflow is achieved. The practical application verification indicates that the system can accurately determine the early warning range of the water disaster personnel evacuation to the operating point level, with an early warning advance time of more than 5 min, significantly improving the accuracy and response efficiency of the early warning. The research results break through the limitations of the traditional early warning mode, providing scientific decision-making basis for water disaster risk prevention and control in coal mines, and have important practical value for ensuring coal mine safety production.