Multi-scale dynamic population estimation: an Adaptive Inverse Distance Weighting (AIDW) model incorporating spatial characteristics

Abstract

Dynamic population data are essential for public health, urban planning, and disaster management. However, time-series models have been required for estimating dynamic populations due to limited data availability. The relationship between dynamic populations and spatial characteristics, including residential, employment, and points of interest (POI) density, were examined. Despite these efforts, a comprehensive model integrating spatial characteristics for dynamic population estimation remains absent. A multi-scale approach using Adaptive Inverse Distance Weighting (AIDW) for transient population estimation is introduced to address this. This model incorporates spatial characteristics such as residential, employment, and POI density, and daily mobility patterns, significantly improving accuracy over the traditional Inverse Distance Weighting (IDW) model. Additionally, the estimated dynamic population was disaggregated into smaller spatial units dissemination blocks (DB), providing finer spatial insights for urban planning. Demonstrated in three diverse Montreal neighborhoods, the AIDW model improves dynamic population estimation accuracy by 11.38, 9.23, and 7.19 in neighborhoods A, B, and C. Key findings include notable differences in population distribution between working and non-working hours, particularly in residential and mixed-use areas. However, the model’s reliance on footfall camera data presents a limitation, and future improvements could include integrating additional data sources like smart cards or GPS. © 2025 Cartography and Geographic Information Society.