Incorporating population dynamics in the context of earthquake shelter location-allocation analysis

Location-allocation is a widely used approach to optimally select earthquake shelters and efficiently allocate the population in case of an emergency. A significant limitation often ignored by the vast majority of studies is the utilization of static aggregations of residential population, which may lead to sub-optimal location decisions and inefficient allocations. To overcome this limitation, in this article, an attempt to spatially refine population data, as well as, to capture population fluctuations throughout the day, using areal interpolation methods along with open spatial information, is undertaken. Then, its influence on the shelter location selection and population allocation process is examined. The city of Mytilini, Lesvos, Greece is used as the case study to further investigate three location-allocation scenarios using block-level census population, building-level night and day estimations as input. The results indicate that using spatially refined population data provide reduced distances, better shelter selection and capacity optimization, and finally, more efficient allocations. Moreover, using building-level day estimations of the population distribution reveals significant shifts in sheltering demand from residential areas to mixed and commercial zones. The use of detailed population dynamics data can give insights about the adequacy of shelter provision under different scenarios, and therefore, help civil protection authorities to make much more informed decisions.