In this paper, we investigate applying ensemble methods to spatial data stream classification within the context of emergency services, specifically severe weather events. Emergencies demand swift, accurate decisions to mitigate impacts and protect lives. Ensemble methods improve accuracy predictions by combining outputs from multiple neural networks, each trained on diverse aspects of the data, including geographic coordinates, weather data, spatial data, and logistical factors. These models collectively contribute to more precise decision-making, particularly in assessing evacuation priorities. We collected and generated relevant data for affected regions and evacuation centres pertinent to severe weather events. To apply class labels, we utilize various clustering techniques. We find that the system architecture achieves high classification accuracy of spatial stream data, potentially leading to more effective emergency responses.