Spatio-temporal data-driven analysis of mobile network availability during natural disasters

The accurate assessment of mobile network availability during large-scale natural disasters is essential for ensuring effective preparation and fast response. However, traditional network availability assessment models are ideal and cannot effectively take into account the spatio-temporal dynamics of mobile network failures in a disaster scenario. Therefore, their evaluation results are generally inaccurate and of coarse granularity, thus not meeting the strict requirements for disaster preparation and response. In this paper, we propose a data-driven analysis framework for the accurate assessment of mobile network availability by integrating essential geographical features from various sources, e.g., seismic intensity data, buildings and land usage data, base station location data, and many other data in related studies. Furthermore, we explore the spatio-temporal inter-correlations and dynamics of several key factors of network failures and their impacts on network availability by associating them with corresponding geographical features in a disaster scenario. We demonstrate our analysis framework with a synthetic earthquake scenario in the Tokyo area and validate our analysis by comparing to existing studies.