By Brian Vastag
Washington Post Staff Writer
Monday, February 14, 2011; 4:49 AM
Last July, a set of troubling numbers appeared in a computer at a key weather-forecasting center west of London.
Expect massive, unprecedented rainfall in northwest Pakistan, the numbers said.
But nobody at the center noticed, and officials in Pakistan failed to interpret the signal as a warning.
So the data went unheeded, exposing deadly gaps in the world's loose network of weather prediction systems.
On the day of the heaviest downpour in northern Pakistan, July 29, ten inches of monsoon rain pounded the largest city in the area, Peshawar. The subsequent floods overwhelmed the country, inundating an area about the size of Italy. Two thousand people drowned, and 20 million more - 12 percent of Pakistan's population - were displaced. The disaster also destroyed billions of dollars' worth of infrastructure, crops and livestock.
If data from the European weather center had been properly analyzed, Pakistan could have been warned more than a week ahead of the calamity, an independent flood-prediction expert is now saying. There was "80 percent probability [of severe rain] eight to nine days out," said Peter Webster of the Georgia Institute of Technology, who examined the European center's data. Webster will publish his analysis, which he performed months after the floods, in an upcoming issue of Geophysical Research Letters.
The incident highlights both the growing sophistication of computer programs to predict extreme weather and the difficulties in communicating that information to an anxious public.
The multinational center whose computers foresaw the extreme rain - the European Center for Medium-Range Weather Forecasting - simply does not have the resources to provide specific weather forecasts to individual nations, said Manfred Kloeppel, scientific and technical assistant to the center's director. No one at the center saw the danger signal in advance, Kloeppel said, but "even if someone had noticed this, it is simply not our role to take the phone and call someone."
Instead, the center runs continuous global weather simulations on its computers. Daily summaries are posted online for members of the World Meteorological Organization, which includes Pakistan, to use as they see fit.'Missed' opportunity
The head of Pakistan's weather service said that his agency did forecast heavy rains four day ahead of the deluge, in part by using the European data. But the agency "missed" on predicting that the rain would be so extreme, said Arif Mahmood, director general of the Pakistan Meteorological Department, in an interview from Islamabad. Never in memory had the summer monsoons that soak much of South Asia veered so far west.
Mahmood lamented the information gap that left potentially lifesaving data sitting unnoticed on computers in England.
"Nobody informed us," he said. "If there is the capacity to predict this, they must inform the World Meteorological Organization or the government of Pakistan that this type of emergency is coming."
In the 1970s, forecasters could predict extreme weather two to three days out, said Louis Uccellini, director of the National Centers for Environmental Prediction, operated by the National Oceanic and Atmospheric Administration (NOAA). Now, thanks to satellite monitoring, weather balloons and networks of ocean buoys, powerful computer programs can anticipate the large outlines of extreme weather events seven to 10 days ahead, he said.
But because weather is inherently unpredictable - with tiny initial variations triggering huge events days later - even the most powerful weather simulations provide only a range of probabilities. "No event is 100 percent, no event is zero percent," said Jon Gottschalck, head of forecast operations at the Climate Prediction Center, also part of NOAA. "There's still an element of human experience in forecasting."
That's the tricky part.
"It's not easy to know when to send out alerts," Gottschalck said, because forecasters don't want to "cry wolf" and warn of storms that fail to materialize. Hence, forecasters tend to be conservative when announcing warnings.
In the United States, the Climate Prediction Center highlights potential trouble spots about a week ahead, as it did for the giant Midwestern blizzard in early February. Local weather offices, which gather high-quality weather data for small regions, then issue all alerts and warnings, typically one to two days before a storm hits.
Sorting out how to better predict extreme weather and inform the public is a top priority for meteorologists and climate change scientists as most simulations of our warming world anticipate more extreme weather events.
A 2009 European Central Bank report calculated that extreme weather saps up to 1.1 percent of global economic output each year - with developing countries hit the hardest. That's why weather predictors are racing to improve their capabilities.Help for developing world
Webster has already shown that effective flood-warning systems can be built in the developing world. After extreme flooding devastated low-lying Bangladesh in 1998, Webster developed a flood-warning system for that country. Data from the European center flow to Webster's group, which passes it on to non-governmental organizations in Bangladesh. From there, warnings propagate along cellphone calling trees to reach remote areas, Webster said. Public education programs, which include religious leaders, have primed the population to move themselves and their livestock to higher ground.
The system got its first robust test in 2007, when heavy floods inundated much of the country again.
By all accounts, the system excelled.
"Nobody was drowned, and the savings [in the flooded regions] was $400 to $500 per household," said Webster. "That's about the annual salary there."
Webster said that governments and non-governmental groups in Asia have recently banded together to improve their extreme weather prediction systems. Also, in November, President Obama announced that the United States will provide computer programs and training to meteorologists in India to help them better anticipate the outlines of their annual monsoon.
But when Webster approached the U.S. Agency for International Development - which provided seed funding for his Bangladesh project - to build a flood warning system for Pakistan, the agency turned him down.
"These modern prediction systems can lower the treadmill of poverty a little bit," Webster said. "You can get through a flood and still have something."
A spokeswoman for USAID, Gina Jackson, said, "We receive a lot of unsolicited proposals and we can't fund all of them." She said no official was available to explain the decision.
Mahmood, Pakistan's meteorology head, said his agency would embrace any technical assistance from the United States to help with future flood prediction efforts, but that USAID has not contacted his agency.