The purpose of this study was to estimate the difficult-to-evacuate tsunami population caused by the Chishima Trench Earthquake, assuming that the road network would be disrupted in the event of a landslide. The study used walking speeds to zone difficult-to-evacuate locations during non-snowfall, snowfall, freezing, and heavy snowfall. It proposed a new estimation method for calculating the tsunami inundation population and the difficult-to-evacuate population using micro geodata (high-definition and micro spatial data) such as road networks and buildings in GIS. As a result of the analysis on the Pacific coast region in eastern Hokkaido, it was found that the estimated population of the tsunami inundation zone was smaller than that of the existing research results. The population that was difficult to evacuate increased as the roads became snow-covered and the walking speed decreased. Furthermore, the difference between day and night as well as the difference caused by the presence or absence of road network interruption shrunk. The reason for this was that the mileage decreased as the road surface condition deteriorated, and even if the road network was interrupted, detours were not possible. Since the effects of road network disruption differed due to landform and population distribution, it was considered impractical to implement universal evacuation methods. Consequently, the method proposed in this study provided detailed and highly accurate results, and it was possible to elucidate regional tsunami evacuation problems.