Every year, natural disasters displace millions of people from their homes. However, humanitarian organizations often lack accurate data to quantify the proportion of people who have been displaced, as well as where these populations ended up after a crisis. As climate change increases the frequency and severity of natural disasters, response organizations require improved data to better understand the dynamics of weather-based displacement. By providing improved data and tools on this subject, Facebook can help increase the effectiveness of humanitarian response of food, medical, and housing aid while preserving people’s privacy.
Today, we’re announcing the launch of a new version of Facebook Displacement Maps as part of the Disaster Maps product suite. These maps will help our nonprofit and research partners better understand how many people were displaced because of a natural disaster, as well as where they ended up. The methodology for these maps was cocreated by Facebook and the Internal Displacement Monitoring Centre, which works on measuring internally displaced people who are forced to flee their homes but remain within their country of origin. Partners such as SEEDS India and the Harvard T.H. Chan School of Public Health have used previous versions of our displacement maps for disaster response and public health research, and we hope that our full list of over 100 NGO and research partners will be able to leverage these new maps to fulfill their missions.
Over a year ago, we launched an initial version of Displacement Maps, which aimed to understand the number of people displaced long-term after a natural disaster. However, we heard from our partners that the trends highlighted in our maps did not always correspond to accounts from the field. After analyzing this further, we found out that our old methodology struggled to distinguish between people who were displaced as a consequence of the natural disaster and those traveling after a disaster for work or leisure purposes.
Our improved Displacement Maps use aggregated and de-identified data from people using Facebook on their devices who have opted into location history. We analyze the patterns of displaced people in the area affected by a natural disaster that have abrupt changes in their usual movement patterns as a consequence of the crisis, aggregated to a city level. To establish a baseline, we first analyze people’s normal movement patterns in the 30-day period before the crisis. Then, we analyze movement patterns in the two-week period after the crisis and compare them with the precrisis movement patterns. For both time periods, we calculate and compare people’s home location and their typical distance traveled away from home. Populations are defined as “displaced” when they reside at least two kilometers away from their precrisis home locations and their typical distance traveled away from the precrisis home has doubled. Populations are considered “never displaced” if either of the above conditions is not met and “unknown” if people have not connected to Facebook at least three days in the two-week period after the crisis.
Starting on day 15 after the crisis, we produce daily updates of the population status to count the number of people displaced and returned, within and across cities, aggregating to a country level when the city count is too low. Populations that were originally classified as displaced are considered “returned” once they are observed for three days in a row less than two kilometers away from their home location.
Cyclone Fani hit India and Bangladesh on May 2019, causing at least 89 fatalities and more than $8 billion in damages. The Odisha region in India was the most affected and was hit harder by the cyclone than it had been by any other natural disaster in 20 years. Using the new displacement methodology, we can estimate not only the proportion of people who have been displaced, but also which cities experienced the largest number of internally displaced people, as well as displaced populations from other areas. Through our analysis we saw that cities were affected very differently, with areas like Bhubaneswar experiencing 15% displacement, to cities near the Kolkata area experiencing only 1% to 2% displacement.
The map below shows in red the cities with the highest percentage of displacement and in blue those cities with the lowest percentage of displacement. Additionally, it includes the trajectory of the cyclone and the severity of its winds. As the map shows, the cities closer to the most severe stretches of the cyclone trajectory had the greatest levels of displacement. Once the cyclone lost strength, displacement tapered off.
Typhoon Hagibis hit Japan in October 2019 and was the most devastating typhoon to hit the Kanto region in over 60 years. Given the severity of the storm, government officials issued evacuation orders to a significant proportion of the population. In the figure below, our new displacement maps show large numbers of people leaving the Tokyo area.
Image: Displacement trends from Tokyo after Typhoon Hagibis.
Outside of Tokyo, which had the largest number of people displaced, Aichi and Kanagawa also saw high levels of weather-based displacement. Other prefectures like Tottori and Shimane had much lower numbers of displacement.
In addition to showing the volume of displacement at a moment in time, our maps are able to calculate displacement levels daily, enabling partner organizations to quantify the number of people displaced and returned on a daily basis. This is central to understanding how long a disaster has affected a given area and why some cities recover faster than others, which our partners identified as a core data gap for their efforts. Furthermore, because we generate disaster maps for a wide variety of natural disasters — not just big ones — our displacement data can fill gaps where official statistics do not exist.
By working with Facebook, we have a better sense of how many people have been displaced, where they have been displaced from, where they’ve been displaced to, and for how long. This is especially important for accounting for displaced people who are not being counted in official estimates. [Through this work], we may be able to expand the scope of our monitoring and understand the different displacement dynamics for all kinds of events — not just the major disasters that everyone tracks.
—Justin Ginnetti, Internal Displacement Monitoring Centre
We hope that our new displacement maps are a valuable resource for our partners in disaster response and that this work helps improve the delivery of services to populations in need around the world.
New partners interested in using displacement data maps can reach out to firstname.lastname@example.org. To learn more about our Data for Good program, visit our website.