Monitoring war destruction from space using machine learning

Abstract

Existing data on building destruction in conflict zones rely on eyewitness reports or manual detection, which makes it generally scarce, incomplete, and potentially biased. This lack of reliable data imposes severe limitations for media reporting, humanitarian relief efforts, human-rights monitoring, reconstruction initiatives, and academic studies of violent conflict. This article introduces an automated method of measuring destruction in high-resolution satellite images using deep-learning techniques combined with label augmentation and spatial and temporal smoothing, which exploit the underlying spatial and temporal structure of destruction. As a proof of concept, we apply this method to the Syrian civil war and reconstruct the evolution of damage in major cities across the country. Our approach allows generating destruction data with unprecedented scope, resolution, and frequency—and makes use of the ever-higher frequency at which satellite imagery becomes available.

Mapping artisanal and small-scale mines at large scale from space with deep learning

“…For instance, with very high resolution images (1m) brick kilns are detected with a precision of 88% and with a very low probability to be missed [ 27 ], as well as trees in Sahel (0,5m) [ 28 ] that are detected up to a size of 3 m 2 with a very high recall of 95%. Last, our classifier achieves an Area Under Curve ( AUC ) of 0.86 on the test set ( S4 Fig ), which is similar and coherent with respect to the score of 0.84 obtained in [ 29 ] when detecting buildings’ destruction in conflict zones. To conclude, drawing a comparison to existing work highlights the necessity of making trade-offs between the area covered, the satellite resolution, the accuracy of the detection and the minimum size that can be detected.…”

Monitoring war‐generated environmental security using remote sensing: A review

“…Images of Gaza were used to assess building damage from the conflict in 2014. Similar research has been conducted for Syria using MAXAR imagery (Mueller et al, 2021). Even though individual building destruction may not directly and immediately impact environmental degradation, resources and space are needed to build new buildings and get rid of the ruins.…”

Quantifying war-induced crop losses in Ukraine in near real time to strengthen local and global food security

“…Easier access to imagery and cloud computing platform to store data and provide computing power has greatly increased extent, scope, and granularity with which satellite imagery is used to assess damages due to conflict since the review of studies on this topic by Witmer (2015). Documenting conflict effects in urban areas requires very high resolution imagery as illustrated by Mueller et al (2021) who develop a machine learning algorithm applied to Syria. 2 Effects of war on agricultural production can rely on lower resolution imagery as illustrated by studies such as Alix-Garcia et al (2013) for Darfur and a number of studies covering areas of the former Soviet Union including Ukraine to assess 1 As exposure to violence often leads the poorest to revert to subsistence as in Colombia (Arias et al 2019), levels of food consumption may not decline linearly with conflict exposure as shown in Afghanistan (D'Souza & Jolliffe 2013).…”