Cyclone Idai: Analyzing Population Movement in Mozambique with Facebook Disaster Maps

Density maps show the percentage change in population, measuring how people move during an emergency.


Cyclone Idai

Cyclone Idai will make landfall potentially as the equivalent of a Category 4 storm, striking coastal Mozambique sometime between 9 p.m. and 10 p.m., East Africa Time, on Thursday, just to the north of the densely populated city of Beira. Information sources like UNOSAT, the United Nations satellite mapping agency, show that the populations exposed to the wind swathe of this storm are quite large, exceeding 1.5 million people in the most at-risk area. Likewise, baseline studies of the economic geography of Mozambique show that as much as three-fourths of the population living in communities throughout the potentially affected area lives below the poverty headcount line. While much remains unknown about the contours of the event, what is known is that this storm will strike an area which is very poor and socially vulnerable by global standards, and reasonably densely populated, with potentially catastrophic results.

The map above plots areas of Mozambique where the population is higher or lower than baseline levels. Red bars signify areas where a population higher than baseline have moved to over the past 24 hours, and blue bars indicated where people have moved from.

One of the key unknowns is how people are likely to behave in the face of disaster. How are communities throughout this area responding to this severe threat to their lives and homes? Are they able to move out of harm’s way? And is there any evidence of them already doing so?

Facebook disaster maps allows a certain kind of window into these questions. Through aggregated and anonymized data on the locations of Facebook users over 8-hour time intervals during crisis, Direct Relief and others can look at how at least a subset of crisis affected populations may behave in near-real-time. Given Facebook’s exceptionally large global user base, even relatively poor areas like coastal Mozambique display some signal of movement during crisis events.

This map shows population density on March 12, 2019, at 7 p.m.
This map shows population density on March 12, 2019, at 7 p.m.

In the case of Cyclone Idai, density maps can show the percentage change in population, measuring how a certain fraction of the population is moving. On March 12 at 7 p.m., the well-traveled route between the cities of Beira and Mutare display an expected high-density pattern, with high percentage change as compared to the baseline measure of three months ago.

On March 13, however, that density pattern shifts dramatically, with the traffic corridor between Beira and Mutare showing much lower density but the areas further towards the outskirts of the wind swathe displaying upward shifts in population density. There is, in other words, some evidence detectable within the Facebook disaster maps data of an evacuation pattern taking place away from Beira towards outlying areas.

This dataset shows population movement on March 13, 2019.
This dataset shows population density on March 13, 2019.

In this case certain significant caveats do apply. First, the population sample represented in the Facebook data is somewhat small, containing just shy of 6,000 individuals out of a potential baseline population numbering in the hundreds of thousands. Second, given the poverty and inequality of this area it is probable that the population being measured as Facebook users is more likely to be a more mobile population. Therefore, the population density shifts displayed in these maps should not be understood as necessarily indicative of the population movement patterns of the entire affected area.

Nevertheless, as is so often the case, there is no other meaningful source of information available to humanitarian responders, especially in areas like coastal Mozambique, regarding the movement dynamics of populations during times of crisis. As Direct Relief continues to monitor this situation and reach out to partners across the region for possible emergency health care support, the Facebook disaster maps data is an invaluable guide among other core datasets to how Direct Relief may need to adapt the focus of assistance to meet the needs of the most vulnerable.

Giving is Good Medicine

You don't have to donate. That's why it's so extraordinary if you do.