Modeling and analysis of geographic events supported by multi-source geographic big data

Geographic events, as a basic construct in geographic process description, have become a core content of geographic information system (GIS). Due to the limitation of acquiring human activity data, GIS modeling and analysis of geographic events has long been focused on the event-induced changes of geospatial objects and the interaction between the objects. However, in recent years, with the explosive growth of location-based service data and the rapid development of quantitative depiction of human activities, the impact of geographic events on human activities and online social participation in geographic events have aroused wide concern in many fields, which poses great challenges to the space-time cognition, modeling methods and analysis framework of geographic events. In this regard, this study discussed the conceptualization and categorization of geographic events in the context of big data, and then introduced the space-time semantics and graph-based data model for geographic events. The "node-edge" graph data structure is used to establish event ontology, the secondary or cascading events, the evolution process, and the "cause-effect" interaction. The spatiotemporal data mining approaches for geographical events were also summarized, which are limited to conventional event detection in "physical space". Integrating "virtual space" event discovery and propagation simulation ideas into data mining approaches is essential for recognizing multi-scale spatiotemporal responses and understanding regional difference of human activities under diverse geographic events. Finally, the study used urban rainstorm events as an example to examine the conceptualization and modeling method of geographic events. Social responses to urban rainstorms and regional differences were examined at inter-urban and intra-urban scales. The case study proved the concept and verified the feasibility and practicability of the proposed framework. © 2021, Science Press. All right reserved.