On 29 October 2024, devastating floods struck the province of Valencia in Spain, resulting in at least 221 deaths, injuries, and severe damage to infrastructure, including roads, streets, homes, and personal belongings. Extreme weather events cause large-scale population displacements, but these movements are often temporary (Hoffmann et al., 2020; Mugabe et al., 2021). Populations generally return days after the event. Estimating the scale of these movements in real time is challenging due to a lack of appropriate data. Such information is essential for implementing rapid humanitarian responses, including the provision of temporary shelter, food, and services (Rowe, 2022; Vestby et al., 2024). Digital footprints from online platforms have emerged as a novel source of information, providing valuable insights into people’s movements in near real time with high temporal and spatial granularity. This granularity holds significant potential for the efficient delivery of humanitarian aid to the places that need it most (Alexander et al., 2019; González-Leonardo et al., 2024).
In this Policy Insight, we use active user count data from the Facebook app to analyse population distribution patterns in the region of Valencia following the flood disaster, focusing on two specific points in time: 31 October 2024 (immediately after the floods) and 13 November 2024 (two weeks later). Specifically, we compare the number of users in each location at these points to a baseline period (i.e., a 45-day period before the floods) and calculate the percentage change in user counts. However, it is known that digital footprint data are based on only a share of the population and may not be representative of the full local population. To address this, Facebook active user counts have been adjusted to better represent the overall population.
Towns west of Valencia recorded a great decline in user counts after floods, suggesting population outflow
On 31 October 2024, Facebook app user data reported the largest decline (over 50%) in towns west of Valencia heavily impacted by floods, including Chiva, Cheste, Requena, Utiel, and Montroy, as well as the coastal town of Cullera (Figure 1). In contrast, no change was recorded in the number of active Facebook users in the suburbs of southern Valencia, such as Catarroja and Paiporta, which were also severely affected by flooding. Similarly, towns in southern Valencia, including Almussafes and Algemesí, reported no changes in the number of Facebook users. These patterns suggest that western towns experienced significant population outflows following the floods, while suburban and southern towns showed no changes, likely reflecting differences in evacuation patterns, with more people potentially staying in the latter areas.
The data show an increase in the number of Facebook users in the city of Valencia, which was not affected by the flood. They also show significant increases in the number of Facebook users in towns around flooded areas, particularly Sueca, Ribarroja, and Alzira. Flooding in these areas primarily impacted agricultural fields and forests rather than residential areas. The increase in user counts may indicate temporary population displacements from flooded areas to safer and unaffected locations, such as the city of Valencia and areas bordering the flood zone.
Post-flood rise in user counts suggests population inflow
On 13 November 2024, a general increase in the number of Facebook users is observed. This pattern includes flooded areas within the suburbs of Valencia and western and southern towns in relation to the number of users during the baseline period and on 31 October 2024. Certain areas within the flooded western towns of Cheste and Chiva, as well as the coastal town of Cullera, seem to have been exceptions, reporting continued but smaller declines in the number of Facebook users. Additionally, large increases were recorded in the city of Valencia and other towns near flooded areas, such as Sueca, Ribarroja, and Alzira. These results suggest that a share of the displaced population may have returned home two weeks after flooding ceased and recovery work started, alongside inflows of volunteers and emergency staff.
Figure 1. Percentage change of Facebook users between crisis (1st and 13th November) and baseline periods.
Note: Flood extent corresponds to Copernicus satellite imagery on 2nd November 2024; flood extent in Chiva and Cheste was not included due to the lack of data, although this area was heavily affected by food.
Urgent need for humanitarian assistance in areas with increasing numbers of users
The increase in the number of active Facebook app users in flood-affected areas of Valencia suggests a concentration of local populations who may have stayed behind, along with a potential arrival of volunteers and emergency staff. This rise in the number of users indicates a potential increase in demand for temporary shelter, food, and services, as well as the supply of equipment and workers to support cleaning and reconstruction efforts. The rise in user numbers in the city of Valencia and in towns bordering the flooded area also indicates a potential increase in demand for temporary shelter, services, and food to support displaced populations from flooded areas, as well as arriving volunteers and emergency staff. The persistent decline in users in certain flooded areas within the towns of Chiva and Cheste suggests that residents could not return or that the arrival of volunteers or emergency staff was less significant. These areas may have been particularly affected by the floods and could require additional equipment, workers, or volunteers for reconstruction.
This analysis considers Facebook app user count data to infer changes in population distribution, but it is important to acknowledge the limitations of such data. User activity may not fully represent actual populations, as it is influenced by factors such as digital engagement patterns and access to technology. Furthermore, more precise insights into population displacement patterns could be derived by incorporating origin-destination mobility data. In summary, while this analysis offers a useful signal of changes in population distribution, integrating mobility data and addressing data biases are key steps in drawing a more accurate picture of the true extent of population shifts.
References
Alexander, M., Polimis, K., & Zagheni, E. (2019). The impact of Hurricane Maria on out-migration from Puerto Rico: Evidence from Facebook data. Population and Development Review, 45(3),617-630. https://doi.org/10.1111/padr.12289.
González‐Leonardo, M., Neville, R., Gil‐Clavel, S., & Rowe, F. (2024). Where have Ukrainian refugees gone? Identifying potential settlement areas across European regions integrating digital and traditional geographic data. Population, Space and Place, 30(8), e2790. https://doi.org/10.1002/psp.2790.
Hoffmann, R., Dimitrova, A., Muttarak, R., Crespo Cuaresma, J., & Peisker, J. (2020). A meta-analysis of country-level studies on environmental change and migration. Nature Climate Change, 10(10), 904-912. https://doi.org/10.1038/s41558-020-0898-6.
Mugabe, V. A., Gudo, E. S., Inlamea, O. F., Kitron, U., & Ribeiro, G. S. (2021). Natural disasters, population displacement and health emergencies: multiple public health threats in Mozambique. BMJ Global Health, 6(9), e006778. https://doi.org/10.1136/bmjgh-2021-006778.
Rowe, F. (2022). Using digital footprint data to monitor human mobility and support rapid humanitarian responses. Regional Studies, Regional Science, 9(1), 665-668. https://doi.org/10.1080/21681376.2022.2135458.
Vestby, J., Schutte, S., Tollefsen, A. F., & Buhaug, H. (2024). Societal determinants of flood-induced displacement. Proceedings of the National Academy of Sciences, 121(3), e2206188120. https://doi.org/10.1073/pnas.220618812.