Government researchers have confirmed that the steep decline in air traffic during the coronavirus pandemic has affected the quality of weather forecasting models by sharply reducing the amount of atmospheric data routinely collected by commercial airliners.
In a study, researchers showed that when a short-term forecasting model received less data on temperature, wind and humidity from aircraft, the forecast skill (the difference between predicted meteorological conditions and what actually occurred) was worse.
The researchers and others had suspected this would be the case because atmospheric observations from passenger and cargo flights are among the most important data used in forecasting models. The observations are made by instruments aboard thousands of airliners, mostly based in North America and Europe, as part of a program in place for decades. They are transmitted in real time to forecasting organizations around the world, including the National Weather Service.
During the first months of the pandemic, when air traffic declined by 75 percent or more worldwide, the number of observations dropped by about the same percentage.
“With every kind of observation that goes into weather models, we know they have some impact on improving accuracy overall,” said one of the researchers, Stan Benjamin, a senior scientist at the Global Systems Laboratory, a part of the National Oceanic and Atmospheric Administration, in Boulder, Colo. “If you’ve really lost a lot of observations of some kind there could be some stepping back in skill overall.”
While the researchers showed that the data loss contributed to making the model less accurate, NOAA said that so far it had not seen an impact on the type of short-term forecasts that companies use to make business decisions or a person might use to decide if they need to take an umbrella when going out.
“We are not directly seeing a readily apparent reduction in forecast accuracy as we continue to receive valuable data from passenger and cargo aircraft along with numerous other data sources,” an agency statement said. Those other sources include satellites, ocean buoys and instruments carried aloft by weather balloons.
The amount of data from aircraft has also increased in recent months as air travel has picked up, the agency said. The daily number of flights by passenger aircraft in the United States is now at about 50 percent of pre-pandemic levels. Flights by cargo aircraft were not as affected.
Dr. Benjamin, along with two colleagues working at the laboratory, Eric P. James of the University of Colorado and Brian D. Jamison of Colorado State University, simulated conditions during the pandemic in April by taking data from 2018 and 2019 and eliminating 80 percent of it before feeding it into a forecasting model developed by NOAA called Rapid Refresh.
They compared the errors that resulted to those if the model included no aircraft data.
“We had to look to see if 80 percent gives 80 percent of impact,” Dr. Benjamin said. “But it’s not quite that much.” They found that eliminating 80 percent of the data produced errors that were 30 to 60 percent of the errors that would have resulted from having no data at all. The study was published in the Journal of Applied Meteorology and Climatology.
Separately, the World Meteorological Organization, which in the spring had expressed concern about the loss of aircraft-based observational data, announced this week that it had signed an agreement with an aviation industry group to expand the observation program to cover parts of the world where little data is currently gathered.
The agreement with the International Air Transport Association calls for adding more airlines and aircraft to the program, including those that have routes in Africa and other less-monitored areas.
Currently about 40 airlines participate in the program, and in all about 3,500 aircraft are have the equipment to make and transmit observations. In the United States, Delta, United, American and Southwest and the cargo carriers United Parcel Service and FedEx are involved.