Hurricane Sandy, one of the most destructive storms in US history, struck New Jersey four years ago this weekend. It left hundreds dead, thousands homeless and millions without power. Around $75bn of damage was inflicted on homes, roads and offices.
The devastation and death toll were grim reminders of human vulnerability in the face of the elements. However, Hurricane Sandy’s impact could have been much worse, forecasters in the UK revealed last week. Scientists at the European Centre for Medium-Range Weather Forecasts in Reading were the first to warn there was a significant chance of Sandy making a sudden “left-hook” as it swept up the east coast of the US in late October 2012, striking Manhattan and New Jersey. Other centres were suggesting that the storm would peter out at sea.
The forecast – which was made five days before Hurricane Sandy hit the US mainland – was unexpected and confirmed only much later by other forecasters. Forewarned, New Jersey was able to protect itself from the worst effects of the 169km/h (105mph) hurricane.
It was a key success and now this forecasting prowess is to be further exploited by the Reading centre – rated one of the world’s top forecasting organisations – in order to make further improvements in its medium-range forecasting, which typically covers weather events over three to 10 days. “The crucial point about our approach to forecasting how Hurricane Sandy would behave once it had formed was a technique called ensemble forecasting,” said Florence Rabier, the centre’s director general.
“Instead of running just a single forecast, our computers ran a number of forecasts – around 50 – using slightly different starting conditions. This gave us a much better idea of what might occur at a particular time.”
Thanks to ensemble computing –which has been pioneered at meteorological centres around the world – it is now possible to make weather predictions for, on average, the coming seven days. By honing the technique, however, scientists at the Reading centre hope to extend that to two weeks. “We are going to concentrate on major improvement to our software to try to push ensemble forecasts so they are more exact and reach further into the future,” said Rabier.
Another example of ensemble forecasting was provided by this September’s storms around Kalamata in southern Greece. Several people died when torrential rain caused flash flooding that washed cars down narrow streets and left them piled on the sea shore. Again, the use of multiple forecasts from slightly altered starting points was able to provide early warnings.
At the centre’s headquarters,which has 34 member nations, two massive Cray super-computers digest data streamed from a global array of measuring devices: robot buoys floating in oceans; balloon-born sensors; instrument packages on civil and military aircraft; automatic land-based weather centres; and satellites.
“These are some of the most powerful computers in the world and we have two, to make sure there is no break in their highly complex analyses,” said the centre’s Umberto Modigliani. “The forecasts we make are passed to individual countries’ meteorological offices.”
The wide array of data available to the centre is crucial. However, the most important is that from weather satellites – distant geostationary probes that hover over the same spot on Earth at a height of 35,786km – and polar orbiting satellites, which sweep over the poles at about 850km, taking several days to cover the surface of the Earth. The former provide constant but limited data because of their distance from the Earth, and cannot properly study weather patterns at latitudes above 60 degrees. By contrast polar satellites provide masses of information, but pass over the same location only once or twice a day.
It was data from these satellites, combined with the use of ensemble computing, that provided the key insight into Hurricane Sandy’s sudden left-hook. Tony McNally, of the centre’s satellite data section, said: “We ran a post-mortem of the storm by re-running our computer models using data stored in our computers. However, we removed all the satellite data and used only data from buoys, aircraft and other non-orbiting sources. The computer then predicted that the hurricane would not make landfall. In other words, without data from satellites, we’d have got it wrong. With it, and with ensemble computing, we got it spot-on. ”
