Strengthening Conflict Research with Satellite Data

Wars not only claim lives but also cause widespread destruction and displacement, with long-term consequences for entire countries. These dynamics are receiving increasing attention in conflict research. Until now, however, the field has relied primarily on text-based data that document individual incidents of violence and are largely based on media reports. Such data depend on people reporting acts of violence, resulting in systematic gaps.

Valerie Sticher, senior researcher at the Department of Political Science at UZH, is aware of this challenge. For several years, her research has therefore focused on how satellite data can support conflict research by complementing existing text-based data. She has now brought together the key insights from this work with researchers from ETH Zurich, EPFL and international partners in the article “Advancing conflict research and mitigation with satellite-based data”, published in the scientific journal Nature. Sticher began her research on this topic while she was still at the Center for Security Studies at ETH Zurich.

“Satellite data complement text-based data, providing a more comprehensive picture of armed conflicts,” says Valerie Sticher. This is because they can reveal developments that often receive only limited attention in reporting, such as destroyed crops or damaged buildings. According to Sticher, another key advantage of satellite data is that they can provide information even in places beyond the media spotlight, where little text-based data is available. However, satellite-based data also have limitations. Challenges include, for example, a lack of reference data from conflict areas. Certain events are also difficult or impossible to detect in satellite imagery. In their Nature study, the researchers show how these limitations can be addressed so that satellite-based data, when combined with text-based data, can provide as comprehensive a picture of a conflict as possible.

Using empirical examples from Ukraine and Myanmar, the study’s authors show how satellite-derived and text-based conflict data can be combined. One data source can provide information that the other lacks. In Ukraine, the researchers link satellite-derived data on damaged buildings with text-based data on which conflict party controlled a particular area and when. Only by combining these two data sources does it become apparent that significantly more buildings were damaged in areas captured by Russian forces than in areas recaptured by Ukrainian forces. Satellite data also provide a more comprehensive picture of violence in Myanmar, showing that the targeted destruction of Rohingya homes continued long after the massacres had ended – a phase of violence that is largely unreported in text-based data. “In the longer term, the two data sources could be modelled jointly to identify patterns of violence that would otherwise remain hidden from us,” says co-author Jan Dirk Wegner from the Department of Mathematical Modeling and Machine Learning at UZH.

The researchers hope that the new research approach could be of considerable value to international organizations. “Combining satellite-derived and text-based data allows international organizations such as the UN or the ICRC to better assess the consequences of armed conflicts, identify particularly affected regions more quickly, and target humanitarian assistance more effectively,” says Karsten Donnay, professor of digital democracy at UZH.