For the popular, crowdsourced earthquake monitoring website Did You Feel It?, even not contacting the website can count as valuable data, a new study finds.
Called DYFI for short, the website was started in the late 1990s by scientists at the U.S. Geological Survey (USGS). After an earthquake, residents visit the website, put in their locations, usually as zip codes, and describe what they felt during the event. These reports are then used to map the area around an epicenter and categorize the intensity of the shaking from the event.
In roughly the past 20 years, about a million DYFI responses from affected people have led to quickly generated earthquake intensity maps. However, DYFI has always left out zip codes where the site received no reports from the public.
“The way DYFI is currently set up, they assume if they don’t get any reports that they have no information,” said study author John “Jack” Boatwright of the USGS in Menlo Park, Calif. In a new paper in the March/April issue of Seismological Research Letters, he and colleague Eleyne Phillips, also of the USGS in Menlo Park, have found instead that those nonreports are, in fact, valuable and neglected data.
Given that so many U.S. citizens are familiar with the DYFI site, the researchers contend that the absence of reports from a specific zip code signifies that shaking simply wasn’t felt, rather than that there were “no data.” They were inspired to investigate when they compared DYFI maps, which are generated from people reporting, with ShakeMaps generated by seismic instruments and found that the amounts of shaking detected differed sharply in areas of low intensity. This suggested to them that the instruments were more sensitive than humans to shaking, and they wondered how to improve the results from DYFI.
To transform the silence from nonreporting zip codes into useful data, Boatwright and Phillips devised a scheme of assigning each of those zip codes the lowest intensity rating on what’s known as the modified Mercalli intensity scale, which is a I.
The result? Intensity maps now show ground shaking declining at a more realistic, faster rate with distance from the epicenter than maps generated using the usual DYFI data, the researchers report in their study, which was published online on 1 February.
“Using not-felt responses is important, as they prevent the average intensity in a community being too high at large epicentral distance,” said Koen Van Noten, an earthquake geologist at the Royal Observatory of Belgium in Brussels.
The scientists also came up with ways to minimize errors that can crop up in quake intensity maps when zip codes generate too few felt earthquake reports compared to their population sizes. The USGS team tested these refinements by regenerating the intensity maps of two recent moderate California earthquakes.
Jim Dewey, scientist emeritus with the USGS National Earthquake Information Center in Golden, Colo., said that with the way DYFI currently handles information, an entire zip code might be represented by a smattering of reports. By weighting reports and nonreports, shaking intensity for a zip code is better represented, he added. Dewey took part in the creation of the DYFI website but not in this new research.
Power of Citizen Science
Reporting your experience after an earthquake isn’t new. Seismologists have been interested in the geographic coverage of shaking for more than a century. After the 1906 earthquake in San Francisco, Andrew Lawson put an advertisement in the papers asking for people to send in particulars on the earthquake if they felt it.
Until the 1990s, reports on earthquakes were all submitted by mail. The arrival of Internet reporting helped the DYFI system take off. Today, about 1% of the population reports their observations after an earthquake. “It’s unbelievable,” said Boatwright. He quips that the response might represent better penetration than any website, except maybe for Amazon.
To date, a magnitude 5.8 earthquake in Virginia in 2011, which damaged the Washington Monument and other structures in the U.S. capital, prompted the most DYFI reports: 144,201. Another earthquake, which struck Napa, Calif., in 2014, ranks as that state’s most highly reported earthquake in 6 years, with 46,000 responses. “Napa had a number of buildings in the city center damaged,” said Boatwright. “They are still fighting for their post office to get fixed.”
Earthquake Test Sites
People’s perceptions of shaking offer a great way to quantify an earthquake, especially when there are no seismic instruments nearby, Boatwright told Eos. The Mercalli scale uses 10 roman numerals from I (no shaking at all) to X (extreme shaking) to denote the amount of shaking. In general, shaking intensity decreases as you move away from the epicenter of an earthquake, but it’s not uniform: Changes in geology (something seismologists like to call site effect) can influence how far rumbling might be felt.
Within minutes of an earthquake, anyone can log on to the DYFI website and see a map showing earthquake intensity. Each individual zip code around an earthquake displays a color code corresponding to its reported shaking intensities. The site updates maps as more information arrives from citizen reporters, yielding a portrait of the distribution of shaking severity that changes almost in real time.
The researchers used two California earthquakes to test their new approach: a magnitude 4.5 San Juan Bautista earthquake that was both “felt” and “not felt” in the heavily populated San Francisco Bay Area and the Weitchpec earthquake, a magnitude 5.6 event in low-population Humboldt County.
For both earthquakes, nonreporting zip codes were modeled as “not felt” and given an intensity value of I. In general, shaking intensities decreased faster with distance from the earthquake’s epicenter on a map that included the nonreporting intensity I data, Boatwright and Phillips reported. The authors said that when intensity I data from nonreporting zip codes are omitted, there might be an overestimation of shaking in distant locations.
“The work of Boatwright and Phillips is important to do,” said Dewey. “It helps us understand the unprocessed data.”