News•December 12, 2012

Sans Polar Satellites, Sandy Forecasts Would've Suffered

Five days before Hurricane Sandy plowed into coastal New Jersey — driving a wall of water over the iconic Jersey Shore boardwalks and deep into portions of New York City — a computer model run by a European weather modeling center accurately predicted its track and strength, causing weather forecasters to sound the alarm. Now that weather center, as well as the U.S. National Oceanic and Atmospheric Administration (NOAA), are warning that without an operational fleet of polar-orbiting satellites, the European model would have missed that forecast, instead predicting that Sandy would have headed out to sea well east of New Jersey.

The Hurricane Sandy forecast from ECMWF with and without the polar-orbiting satellite data.
Click to enlarge the image. Credit: NOAA

In an analysis NOAA released on Tuesday, but which was first reported in November by the Washington Post’s Capital Weather Gang blog, meteorologists at the European Center for Medium-Range Weather Forecasts (ECMWF) in Reading, England, re-ran their computer model after depriving it of the data that comes from the current fleet of polar-orbiting satellites.

The test was done because cost overruns, launch setbacks, and bureaucratic delays in the U.S. satellite program threaten to cause a gap of a year or more between when one of the current polar-orbiting satellites — known as the Suomi NPP — reaches the end of its design lifetime in 2016, and when the next polar-orbiting satellite is ready for launch in 2017, at the earliest.

Even with the full compliment of today's operational polar-orbiting satellites allowing for advanced warning of the storm, not everyone in the path of Hurricane Sandy was able to get to safety. The storm killed at least 125 people in the U.S., many of them drowning in the massive storm surge. The death toll would likely have been higher if the warnings had not been issued so far in advance. Sandy caused damage that is expected to exceed $100 billion, becoming the second most expensive hurricane to strike the U.S. on record, behind 2005's Hurricane Katrina.

“This study shows the value of polar-orbiting satellites in developing life-saving forecasts with longer lead times,” said Kathryn Sullivan, deputy NOAA administrator and a former astronaut, in a press release. “Had we thought the brunt of the storm was going to stay out in the Atlantic, or if residents had only a day to prepare or evacuate, the results would have been even more devastating.”

Facing the likelihood of a gap in polar-orbiting satellite coverage, weather forecasters in the U.S. and Europe are trying to determine what the consequences will be of losing data that they have come to take for granted.

For the analysis of the Hurricane Sandy forecast, meteorologists deprived the European model of data derived from all of the polar-orbiting satellites, which is a worst-case scenario. It is far more likely that the data gap will involve a reduction from two to one U.S. operated polar-orbiting satellites, rather than a complete loss of such data.

The analysis showed that the ECMWF model’s forecast of Hurricane Sandy’s track would have been off by hundreds of miles without the satellite information. Instead of being within 30 miles of the actual landfall location four-days before landfall, the model would have shown Sandy remaining at sea.

Aerial photo of coastal flooding in Mantaloking, N.J. as a result of Hurricane Sandy.
Credit: Scott Anema, NJNG.

Instruments aboard polar-orbiting satellites gather data that is fed into sophisticated computer models that are used to make weather predictions. In particular, the polar-orbiting satellites gather information on winds and moisture in the upper atmosphere, which compliments information coming from weather balloons that are launched twice daily across the country. The satellites have the advantage of scanning the atmosphere over the oceans, whick lack weather-balloon coverage.

In addition, polar-orbiting satellites, which are different than the satellites that rotate above a fixed point on Earth, also carry instruments used for monitoring volcanic eruptions, gathering sea-surface temperature data, and locating emergency beacons from aviators and mariners in distress.

Not everyone sees the Hurricane Sandy study, and others like it, as clear evidence that weather forecasts will be much less reliable sans one polar-orbiting satellite. Although the Hurricane Sandy study appears to show that a significant decline in weather forecast accuracy may occur during the period of a polar-orbiting satellite gap, its lessons are unclear, said Peter Bauer, the head of the model division in ECMWF’s research department. Bauer and others said there are limits to such single-event studies, and that more research is needed to see how computer models might be affected by a loss of satellite data in other situations as well.

“Hurricane Sandy is a single case and the evaluation of forecast skill should be based on a sufficiently large sample to be able to make statements on cause-effect relationships,” Bauer said in an email conversation. “Sandy has been a single case for which the ECMWF forecasts performed exceptionally well, but it is difficult to defend or attack a future satellite scenario based on this case.” Bauer said that losing data from just one of the current fleet of polar-orbiting satellites constitutes a potentially significant threat to forecast accuracy.

This isn’t the first time that NOAA has publicized a satellite gap analysis centering on a major storm event, without providing a long-term perspective. An earlier analysis run in the wake of the 2010 “Snowmageddon” blizzard in D.C., which was publicized by NOAA officials in testimony on Capitol Hill, showed that computer models would have underestimated snowfall totals in their projections for that event. However, that study also involved one extreme event, rather than data that had been analyzed over a longer period of time.

An image provided by ECMWF, showing the computer model's 5-day forecast for Hurricane Sandy on the left, the forecast without polar-orbiting satellite data in the middle picture, and the actual conditions that occurred in the right hand side.
Click to enlarge the image. Credit: ECMWF.

Some critics see a desire for securing more funding from Congress as the driving force behind these extreme-event studies, since Snowmageddon and Hurricane Sandy were costly and deadly events that are sure to get more attention on Capitol Hill than say, a weather forecast for a sunny spring day, for example.

Other critics of NOAA’s National Weather Service, such as meteorologist Cliff Mass of the University of Washington, have said that NOAA is focusing too much on funding expensive satellite platforms and that it is neglecting other needs, such as maintaining basic weather and climate observation networks and keeping up with other countries in terms of computing capabilities. The NWS' own GFS computer model failed to accurately predict the path of Hurricane Sandy five days in advance, a marked contrast to the European model's performance.

Ryan Maue, a meteorologist with the private forecasting firm WeatherBELL Analytics, said that NOAA should be devoting resources to figuring out why its model failed to accurately predict Hurricane Sandy.

“A fair experiment would require a much longer time period, e.g. month(s) to adequately judge performance under various weather patterns or regimes,” Maue said in an email conversation. “It would be poor practice to base long-range satellite procurement on a singular experiment of an extreme event such as Sandy, which was poorly forecast five days ahead of time by NCEP GFS 'with all satellite data' (in real-time).”

Mitch Goldberg, program scientist for NOAA’s Joint Polar Satellite System, said that maintaining at least two polar-orbiting satellites at all times is “really critical” for the accuracy of weather forecasts (more than two such satellites are needed to act in a backup role in case of any malfunctions). He said that even though the Snowmageddon study and the Sandy analysis looked at individual cases, those were both extreme events that had major societal impacts, making them worth studying. “You really want to be prepared for those cases and you want to have all the information at your disposal,” he said.

Goldberg said he and his colleagues are working to undertake a suite of studies to better understand how computer models might be affected by a loss of some polar-orbiting satellite data.