Search Results

Understanding the continuous evolution of environmental systems increasingly relies on tracking large-scale, heterogeneous data across shifting timelines, commonly structured as temporal multigraphs. While such representations are highly expressive, their temporal density and structural heterogeneity make the automated analysis of long-term transitions particularly challenging. In this paper, we propose a unified process for extracting and abstracting recurrent evolutionary patterns within time-varying multigraphs. The proposed pipeline articulates four successive stages: (1) a spatio-temporal modeling stage based on RCC-8 relations to capture both topological structures and the chronological evolution of geographic entities across consecutive time steps; (2) a frequent pattern mining stage designed to discover recurring temporal sequences within multigraph representations; (3) a pattern matching stage that enables efficient, multi-scale localization of these evolutionary signatures; and (4) a hierarchical abstraction stage relying on super-nodes to produce a compact, condensed representation of the graph’s historical progression. An interactive visual environment is additionally provided as a complementary exploration tool to inspect the timeline outputs of the pipeline. We report on the results obtained from historical graph analyses of CORINE Land Cover data across multiple European cities, demonstrating the process’s capacity to reveal recurrent environmental behaviors and meaningful, time-dependent trajectories of change.