Unravelling the Complexity of Urban Functional Organization Based on Explainable GeoAI

Abstract

Urban expansion and population diversity have created increasingly complex functional organizations, challenging traditional land use strategies and highlighting the need for advanced analytical approaches. This study leverages explainable GeoAI, specifically a Multi-level Receptive Field Graph Convolutional Network, to decipher underlying patterns in urban functional organization across different spatial scales. The explainable AI framework enables interpretable analysis of spatial-functional interactions among urban functions across varying distance thresholds. Comparing five metropolitan five representative cities, the model achieves 92% accuracy while providing transparent insights into scale-specific spatial interactions. Results reveal distinctive distance-dependent patterns: Chinese cities exhibit stronger long-range dependencies, Western cities show balanced multi-scale interactions, while Paris demonstrates short-range relationships. This explainable GeoAI approach advances both methodological transparency and cross-cultural understanding of urban functional complexity.