Compare and contrast: A better way to use satellite images to save lives after tremors

WHEN a big earthquake strikes, it does not do equal harm everywhere. Places resting on unstable sediment will shift around a lot and are thus likely to be damaged badly. Those resting on bedrock are normally better off—though not if they are stuck at the end of a rocky promontory that amplifies a quake’s vibrations in the manner of a tuning fork. Finding the areas most badly damaged, and therefore most urgently in need of assistance, in an area whose geology is not already well understood is thus a high-stakes game of hide-and-seek. It involves experts both on the ground where the earthquake happened, and in faraway laboratories, studying satellite photographs.

Sang-Ho Yun of the Jet Propulsion Laboratory—NASA’s outpost in Pasadena, California—hopes to help those seekers by using such photographs more effectively. These days, pretty much all of the Earth’s surface has been mapped by a technique called satellite-born synthetic-aperture radar. Crucially for disaster-relief work, radar can see through cloud, so does not require clear skies. Equally crucially, its images include information on altitude, accurate to within a few centimetres. Dr Yun’s plan is to compare, automatically, the “before” and “after” shots of a stricken area, to work out which parts have risen or fallen the most, and are thus likely to have suffered most damage. A suitably programmed computer would then colour these in (see image), making them obvious to human users.

On April 25th 2015, as he reports in Seismological Research Letters, he got a chance to test his ideas out. An earthquake of magnitude 7.8, the most powerful in the region since 1934, hit central Nepal. It claimed over 8,000 lives and caused widespread damage. Four days after it struck, an Italian satellite called COSMO-SkyMed, which is equipped with a synthetic-aperture radar, flew over the area. Dr Yun and his colleagues fed the information COSMO-SkyMed’s radar collected into their computers and compared it with radar images taken before the disaster.

Their labours produced wide-area colour-coded maps (see above) that showed where the ground had risen or fallen during the earthquake in ways that might damage buildings. Higher-resolution examination of these high-risk areas was then able to pick out buildings that looked as if they had changed in some substantial way. In some cases these buildings had simply gone askew. In others they had collapsed completely.

To double check the accuracy of their conclusions Dr Yun’s team collaborated with one at the United States Geological Survey, led by Kenneth Hudnut. This let them compare their maps with those created independently, after the disaster, by the National Geospatial Intelligence Agency, one of America’s groups of spies, and by the United Nations. Both these sets of maps were made by people inspecting high-resolution satellite photographs for damage—a process that took three days. Dr Yun’s maps, the comparison showed, contained almost all of the same information.

This test, then, proved that the method works. But Dr Yun and his colleagues know they still have a long way to go before they can provide such maps quickly enough to be useful. The four days it took an appropriately equipped satellite to come by in the case of the Nepalese earthquake were four days too many for effective disaster relief. Such delays are, though, expected to shorten in coming years, as more satellites equipped with synthetic-aperture radar make it into orbit, and more of the agencies operating them realise the value of keeping that radar running all the time, and also of sharing their data as widely and quickly as possible. By 2020, Dr Yun reckons, the average wait should have dropped to between four and seven hours. That is still a long time to be stuck under a collapsed building. But it is a lot better than the current alternative.

http://www.economist.com/