Characteristic Analysis of Sectional and OD Traffic Volumes Under the 2022 Winter Snow Disaster in Sapporo Using Non-negative Tensor Factorization

Abstract This study investigates the impact of the 2022 winter snow disaster on road traffic in the Sapporo metropolitan area by analyzing both cross-sectional and origin–destination (OD) traffic data. Using non-negative Tucker factorization, we extract latent spatiotemporal features from two third-order tensors: one representing sectional traffic volumes across time, and the other capturing OD flows between municipalities. The cross-sectional analysis reveals a broad, city-wide decline in traffic volumes during the winter period, with particularly sharp reductions along arterial roads connecting central Sapporo with peripheral areas. Observation point-related factors enabled classification of locations by severity of disruption, revealing route-specific vulnerability. OD analysis similarly identifies persistent reductions in trip numbers, with further declines during major snowfall events. Core tensor and residual analyses show that inter-regional movements. Especially inbound trips to central Sapporo were significantly suppressed, while some commuter corridors maintained demand, likely through modal shifts. The integrated findings suggest that the snow event led to a prolonged and spatially uneven decline in both traffic volume and demand, especially for longer-distance travel.