A study co-authored by researchers from EPFL, ETH Zurich and the University of Zurich shows how satellite data can support humanitarian conflict analysis, helping provide a clearer picture of the consequences of war and reduce the toll on civilians.
Armed conflict not only costs lives but also causes the widespread destruction of infrastructure and forces people from their homes, with ramifications that can last for years. Researchers generally study these effects using text-based records of individual acts of violence, often sourced from media reports. But such records are imperfect, not least because they depend on witnesses or journalists being present to document what happens.
A new study appearing in Nature shows how satellite data can help fill some of those gaps and give us a broader understanding of conflict dynamics. The study was conducted by researchers from EPFL, ETH Zurich and the University of Zurich as part of the Engineering Humanitarian Action initiative.
Prof. Devis Tuia, who heads EPFL's Environmental Computational Science and Earth Observation Laboratory (ECEO) and is a co-author of the study, has spent several years developing new ways of extracting useful information from vast datasets from diverse sources, including satellites.
Building a more complete picture of conflict
"By observing conflicts from space using high-resolution satellite imagery captured at frequent intervals, we can assess the damage that's been done, track population dynamics and monitor the condition of infrastructure across large areas," says Tuia. "Combined with other sources, these data can help us build a more complete picture of what's happening on the ground."
But satellites cannot tell the whole story on their own. With little reference data available in conflict zones, interpreting the satellite images can be a challenge, and some events are difficult or even impossible to detect from space.
In the Nature study, the researchers show how combining satellite observations with text-based data can help address these shortcomings and provide a more comprehensive view of a conflict situation.
Three approaches to combining data
The study sets out three ways in which satellite data can be used alongside other sources: improvement, enrichment and fusion. Fusing different types of data can unlock synergies, whether by allowing one source to enrich another or by using artificial intelligence (AI) to analyze multiple sources simultaneously. "AI lets us harness the synergies among different types of data rather than simply treating them as separate sources that complement one another," says Tuia. For example, this approach can uncover associations between ethnicity, building typologies and targeting strategies. The researchers used case studies from Ukraine and Myanmar to test their data integration models.
For humanitarian organizations working in areas affected by armed conflict, one of the greatest challenges is understanding what's happening in places that are simply too dangerous to reach.
For the Ukraine case study, the researchers combined satellite observations of damaged buildings with text-based records showing how control of the territory had changed over time. Their analysis revealed significantly more damage (in terms of numbers of buildings) in areas captured by Russian forces than in those subsequently retaken by Ukraine.
Combining these two types of data can also reveal violence that might otherwise go undocumented. "In Myanmar, the targeted destruction of Rohingya homes continued long after the massacres had ended," says Corinne Bara, another co-author of the study. "Satellite data lets us see this later phase of the campaign."
Supporting evidence-based humanitarian action
The researchers hope that their approach will prove useful to international organizations. "We're developing groundbreaking technology that draws on the expertise of conflict specialists and brings together knowledge from different fields," says Tuia, who adds that, throughout their research, the team is keeping the wellbeing of conflict-affected populations firmly in mind and taking an open, transparent approach to software design.
"For humanitarian organizations working in areas affected by armed conflict, one of the greatest challenges is understanding what's happening in places that are simply too dangerous to reach," says Klaus Schönenberger, the director of EPFL's EssentialTech Centre. "Satellite data can tell us where conditions on the ground are changing and on what scale, while text-based sources such as field reports and local intelligence can help us understand what those changes mean for the people affected. Combining information from these independent sources in an intelligent way turns isolated signals into actionable information, allowing humanitarian organizations to prioritize assessments, anticipate needs and start planning their response even before they can safely reach an area. This pioneering study marks an important step toward that goal."
