A unified history-future spatiotemporal data model for geographic processes

Le résumé fourni par la source

The representation of geographic processes is essential for revealing their evolutionary patterns and mechanisms over space and time. However, existing spatiotemporal data models mainly focus on historical change and are limited in representing future evolution and the interrelationships among heterogeneous geographic processes, which constrains prospective spatiotemporal analysis. This study presents a novel spatiotemporal data model to describe the historical and future evolution and associations of geographic processes within a geographic system. Unlike historical evolution, which is typically deterministic, the proposed model represents future evolution as inherently uncertain, defined along two dimensions: (1) conditional uncertainty—future states may take multiple possible forms under different driving conditions; and (2) temporal uncertainty—future states evolve continuously as new historical states emerge. An update approach based on deep neural network-based spatiotemporal prediction models is introduced to dynamically generate future states of geographic processes. Finally, the experiment on three interrelated geographic processes in an island system shows that the proposed model can provide a new perspective integrating history and future to analyse the spatiotemporal evolution of both single processes and multiple processes.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.