Dynamic Social Vulnerability Mapping Using Facebook Data

Assessing populations exposed to climate change impacts traditionally relies upon census data estimations. Yet, these only provide a static picture of risk since censuses are often undertaken and released over long periods and thus cannot be updated regularly. In this study, we investigate how to leverage multi-temporal geolocated social media data from Meta-Facebook and assess spatio-temporal variations of population exposure and vulnerability to climate-related risks. Building upon advanced spatial analytical methods, we address the selection bias of social media datasets and further analyse how population exposure varies daily, weekly, and seasonally during a 4-month typhoon-free period in the Philippines in 2021. Results show how changes in population density combined with varying levels of social vulnerability can increase the size of the population exposed to hazard events at specific periods and places, even in scenarios where population movements are constrained. When comparing daytime with nighttime exposure, less vulnerable areas presented a decrease in population density, while areas with higher social vulnerability showed a population increase. An opposite trend, however, was observed during the weekend and holiday periods, with an increase in population in less vulnerable areas. While limitations remain regarding the study period and the representativeness of social media data, our findings contribute to guiding disaster risk reduction strategies and support climate-resilient pathways in complementarity with traditional data sources and field-based practices.