Summary

  • USGS's Did You Feel It service turned visitors' reports of shaking and household effects into community intensity maps, with answers grouped geographically.
  • By 2006, its operators had built filtering, retained corrections and separate server duties around a deceptively simple questionnaire.

A bookshelf has a limited scientific vocabulary. It can, however, offer a fairly concrete answer to a question about things falling off it. In the early years of USGS's Did You Feel It website, that sort of domestic observation became part of a public map. A visitor arrived looking for information and left some behind.

The agency's 2001 account describes the sequence after the 1999 Hector Mine earthquake in southern California. People checked the location and magnitude online, entered their ZIP code and answered questions about their experience: being woken, for example, or objects falling from shelves. A map began taking shape within minutes. Several thousand responses arrived within a couple of hours.

The questionnaire came with a return trip

USGS had long gathered observations through questionnaires mailed to post offices. Sending them, waiting, interpreting the replies and drawing a map could take months. The website brought those steps much closer together. Responses fed an intensity value for each reporting ZIP-code area; the map changed as more arrived. In the 2001 description, grey areas marked places without data.

Intensity describes the effects of shaking at a location. The website could therefore collect something ordinary people were equipped to observe. Its 2005 fact sheet gave visitors a short route: choose a region, select a recent earthquake or an unknown event, complete the questionnaire, then return to see the developing map. After the 2003 San Simeon earthquake, more than 14,000 responses had arrived within a few hours. The form offered its contributors a visible destination for their answers.

The arithmetic had a history too. In their 2006 technical account, David J. Wald, Vincent Quitoriano and James W. Dewey traced the system's informal online start to 1997 and its scoring approach to earlier telephone questionnaires. Numerical values assigned to answers allowed reports to be processed consistently. By the time of their paper, US maps refreshed every five minutes, with reports grouped by postal area. International collection, added in 2004, grouped observations by city.

Keeping the front door open

The June 2005 earthquake near Anza brought nearly 21,000 responses in its first hour. Reading the site generated a separate, larger traffic problem: page visits had reached thousands per second. To keep accepting observations, operators made most pages static, cached frequently requested material and separated page delivery, form processing and map generation across redundant servers. The visitors admiring the map needed to leave some computing room for the visitors helping to make it.

Answers also needed supervision. A mistyped postal code could transport a report far from its intended location. The 2006 paper describes filters for implausible or inconsistent entries, with suspect records flagged and bypassed in calculations while retained in the data. For duplicate entries, the latest timestamp won. Analysts kept final control over quality. The familiar household question opened into a carefully managed scientific record.

Participation supplied reach, and its geography mattered. The authors connected reporting to population and internet access; damaged computers and power cuts could leave gaps in badly affected areas. Behind each coloured patch lay people, connectivity and a set of processing decisions. The early web had persuaded the audience to help draw the picture. Keeping that picture useful required attending to how it was assembled.