Modern weather forecasting techniques may help scientists develop regional forecast models for seasonal flu outbreaks in the not too distant future.
Alicia Karspeck, a research scientist at the Boulder-based National Center for Atmospheric Research is the co-author of a study on the future of flu forecasting, in which she partnered with lead author Jeffrey Shaman, an assistant professor of environmental health sciences at Columbia's Mailman School of Public Health.
Karspeck cautioned that widespread application of the flu forecasting methodology may be several years away.
"These are very exploratory forecasts," she said Tuesday. "At this point, you could more refer to this as a pilot study trying to demonstrate the capability of doing this."
The seed for the new study sprung from a conversation several years ago in which Karspeck suggested to Shaman that data assimilation techniques used in weather forecasting to build a prediction system could also be applied to forecast the appearance of flu.
Karspeck explained, "The analogy here is that with weather prediction, in order to predict what is going to happen in the weather tomorrow or the next day, one essential is to know what the weather is today." In other words, the data of today informs the data of tomorrow.
In forecasting flu, researchers have a new tool to employ to collect today's data. Near-real-time data is now available from Google Flu Trends, which estimates outbreaks based on the number of flu-related queries through the search engine from a specific region.
"Why that is so important is that it provides us the capability to initialize a model of disease transmission in a population," Karspeck said. "We can take this new type of data, which didn't exist five years ago, and use this to initialize these forecasts."
Flu season arrives in temperate areas of the Northern Hemisphere as early as October and as late as April, and annually kills about 35,000 Americans. Improved regional forecasting might lead people to obtain vaccines, and be more careful about their contact with others who are noticeably coughing or sneezing.
In previous work together, Shaman and Karspeck studied Web-based estimates of flu-related illness in New York City from the winters of 2003-04 to 2008-09. They found that retroactively generated flu forecasts showed the timing of a flu outbreak could be predicted seven weeks in advance of the peak.
Contact Camera Staff Writer Charlie Brennan at 303-473-1327 or firstname.lastname@example.org.