©2019 DigitalGlobe, a Maxar company
Doreen Boyd remembers the first time she saw a hint of slavery from space. A satellite image from 2017 of Rajasthan state in India showed a brown oval that looked like a dusty high school track. But it was nothing so innocuous: She knew it was a brick kiln, one of tens of thousands across South Asia that are often run on forced labor. Boyd, director of the data program at the Rights Lab at the University of Nottingham in the United Kingdom, realized such imagery could help her tally the kilns, enabling organizations on the ground to target slaveholders at the sites. “You can’t see slavery directly, but you can infer it,” she says.
A surge in the number of Earth-observing satellites, along with improvements in algorithms that can interpret the deluge of data they provide, are putting modern slavery under a spotlight. This week, at a conference in New York City sponsored by the United Nations University (UNU), computer scientists, slavery experts, and policy strategists presented the latest efforts in their fields and brain-stormed ways to work together. “We’re doing team science,” says Austin Choi-Fitzpatrick, an expert in peace studies at the University of San Diego in California who has interviewed slaveholders at kiln sites like those the Rights Lab studies from space.
Some 40.3 million people are held in bondage today, according to the latest estimates from the Inter-national Labor Organization, headquartered in Geneva, Switzerland. But finding them is hard. “People affected by this are often hidden from the gaze of the state,” says James Cockayne, director of UNU’s Centre for Policy Research in New York City, who helped organize the conference. Boyd estimates, however, that one-third of all slavery is visible from space, whether in the scars of kilns or illegal mines or the outlines of transient fish processing camps.
In a 2015 effort, DigitalGlobe, whose Earth-observing satellites provide much of the data for Google Earth, recruited users to zoom in on images of Ghana’s Lake Volta, where experts suspect children are forced to work in fishing. “We were looking across this massive lake to try to detect boats,” says Rhiannan Price, director of DigitalGlobe’s global development program in Westminster, Colorado. In total, 90,000 users pinned 80,000 boats, buildings, and fish cages.
Boyd is now using artificial intelligence to speed up the search. As a pilot project, she and her col-leagues at the Rights Lab used crowdsourced visual searchers to identify brick kilns. The oval shape of the large ovens, sometimes 150 meters long, and their chimneys are distinctive, even from space. “You cannot mix them up with something else,” Boyd says.
Since then, Boyd has turned to machine-learning algorithms that recognize the kilns after being trained on the human-tagged examples. Last month, in the journal Remote Sensing, she and her colleagues reported that the algorithms could correctly identify 169 of 178 kilns in Google Earth data on one area of Rajasthan, although it also output nine false positives.
Another company, called Planet, has about 150 small satellites that snap images of the globe’s entire landmass daily. The images are lower-resolution than DigitalGlobe’s, but their frequency opens up opportunities to identify changes over time. “Every day, we see every building, every field, every mine, every quarry, every forest,” says Andrew Zolli, Planet’s vice president of global impact initiatives in New York City.
With Planet data, Boyd and the Rights Lab plan to investigate fast moving signatures of slavery. From space, you can watch a cotton harvest in Turkmenistan and, based on how quickly the cotton disappears, you can tell whether machines or hands picked it. In the Sundarbans, an area spanning India and Bangladesh, shrimp farms and fish-processing camps employ slave labor to clear mangrove trees—a process satellites can capture.
Rights Lab also plans to use satellite data in other parts of the spectrum. The European Space Agency’s Sentinel-1 satellite uses radar to measure tiny changes in elevation—which could reveal the ground subsidence of a mining work tunnel with illegal laborers inside. The agency’s Sentinel-2 satellite, which detects frequencies of infrared light, can spotlight mining operations based on the reflections of newly exposed minerals.
Sky-high solutions, even smart ones, aren’t a cure-all. Other organizations need to convert detective work from space into action on the ground. And even before the analysis starts, researchers need to know what they’re searching for. “The imagery itself is dramatically less worthwhile if you don’t have local knowledge of what you’re looking at,” Zolli says.
Source: Science Mag