Xiaoxiang Zhu
2015, Germany
X-raying the Earth
Computer Science

Xiaoxiang Zhu has developed a method to measure changes on our planet more accurately than ever before.

Xiaoxiang Zhu is uncertain – even if her data is clear enough: A golf course in Las Vegas is sinking at a rate of around three centimetres a year. Maybe it’s harmless, reasons Zhu, a young researcher from the German Aerospace Center (DLR) and the Technical University of Munich: there was simply too much groundwater wasted watering the greens. But what if there is something rumbling away secretly underground? The conference centre nearby might be at risk.

Using radar images from space, Zhu developed a method to detect minute land deformations, starting at a rate of one millimetre a year – making the 29-year-old researcher the current world record holder. But these abstract records have very practical implications: With urbanization steadily progressing, “monitoring physical infrastructures is becoming increasingly important,” Zhu says. “But that’s just too expensive using ground-based GPS measurements.”

Ground settlements can threaten buildings, bridges, steel structures, self-supporting roofs, and especially dams. “In rapidly developing countries in particular, often there is just not enough time given to preliminary geological examinations,” criticizes the researcher from Changsha, in Central China.

In order to detect these tiny deformations of the Earth, Zhu uses images from the most precise radar satellite in the world: TerraSAR-X. It uses microwaves to measure the ups and downs of the Earth from a distance of 514 kilometres. But the TerraSAR-X images are of course static. To capture dynamic deformations over time, you typically take 20 to 50 images of the same area, over the course of several months. If you pile all these images exactly on top of each other on the computer, you can analyse tiny movements. Until now, however, this stacking method has only worked for a few pixels at a time. With more pixels included, different objects overlap too much to still be distinguishable from others.

So Zhu worked around the problem: for the first time, she adapted the so-called “Compressive Sensing” method for 4D modelling. Her trick was to combine the scant information from the radar images with “prior knowledge”, e.g. the information that the satellite is looking at the ground from a 45° angle. From this perspective, a house looks as if it were tilted onto the ground. So the radar reflections give the false impression that the ground, walls and roof are all the same height. But if you take that knowledge into account, the available data is sufficient to straighten out these three points, and then plot tiny motions in great details. “Even if one part of a bridge is sinking by a few millimetres, and another part is rising slightly, you will be able to see it,” Zhu says.

So it’s hardly surprising that not just geologists and building companies are interested in the data, but also a growing number of municipalities around the world. In highly populated areas, it is advisable to recognize threats arising from land subsidence well in advance – to locate areas where too much oil, gas or groundwater was extracted, for instance. Whether the sinking groundwater is causing Las Vegas to sink, remains unclear even with Zhu’s method. In any case, the city’s leaders have been warned. But so far, no one has responded to her publications. The gambling metropolis obviously prefers to rely on good luck instead.