Tag Archives: early warning

Scanning the Internet for Ecological Early Warnings

If Google Flu Trends can, why can’t we? The possibility to mine large amounts of individual reports and local news posted on the Internet as early warning signs of pending epidemic outbreaks has been a part of global epidemic governance for quite some time. The question is; could we do the same for ecological crises? A couple of years ago, a couple of colleagues and I wrote a conceptual piece in Frontiers entitled “Can webcrawlers revolutionize ecological monitoring?” where we elaborated issue. Until today however, the idea hasn’t moved much from its conceptual phase. Luckily, analysts and GIS-experts at the USDA Forest Service, now have begun to test the concept with real world data. In a new paper entitled “Internet Map Services: New portal for global ecological monitoring, or geodata junkyard?”, Alan Ager and colleagues, present initial results from runs with a geodata webcrawler . They report:
At the USDA Forest Service’s Western Wildland Environmental Threat Assessment Center (WWETAC), we are exploring webcrawlers to facilitate wildland threat assessments. The Threat Center was established by Congress in 2005 to facilitate the development of tools and methods for the assessment of multiple interacting threats (wildfire, insects, disease, invasive species, climate change, land use change)
The Threat News Explorer (see image) visualizes some of the results.

However, they also note that
much of the online data is stored in large institutional data warehouses (Natureserve, Geodata.gov, etc.) that have their own catalog and searching systems and are not open to webcrawlers like ours.  In fact, most federal land management agencies do not allow services to their data, but allow downloading and in-house viewers (i.e. FHTET 2006). This policy does not simplify the problem of integrated threat assessments for federal land management agencies.
The group is now developing a more powerful webcrawler. You can find and search the database for geospatial data and map here. Still a long way to go it seems, but a very important first step!

Google Flu á la Sweden

The fact that the spread of flu could be predicted by tapping into searches on Google, gained much attention during 20098-9. (See however here). The idea to tap into web searches to find early warnings of disease outbreaks seems to be spreading, this time however, applied on something that Swedes know far too well: the extremely infectious norovirus. The virus is known to give nausea, vomiting, diarrhea, and abdominal pain.

A research team at the Swedish Institute for Infectious Disease Control, recently published data showing that queries for *vomit* (asterisks denote any prefix or suffix) submitted to the search engine on a medical website in Sweden (www.vardguiden.se), match the number of laboratory-verified cases almost perfectly (see figure).

Figure. Number of queries for *vomit* submitted to a medical Web site (blue), number of laboratory-verified norovirus samples (red), with baselines and 99% prediction intervals, and number of media articles about winter vomiting disease (black) in Sweden, 2005–2010. From: Hulth A, Andersson Y, Hedlund K-O, Andersson M. Eye-opening approach to norovirus surveillance [letter]. Emerg Infect Dis. 2010 Aug: http://www.cdc.gov/EID/content/16/8/1319.htm

The original article in Emerging Infectious Diseases can be found here.

Are Epidemic Early Warnings, Really “Early” Warnings?


Information technological innovations seem to have played quite an important role in detecting early warnings of the current “new flu”, “swine flu” or H1N1.  This topic is elaborated in today’s issue of New York Times. Apparently, WHO received the first warning already on April 10th through its web-crawler based monitoring system. This again proves the usefulness of mining unofficial data for monitoring.

One point missing in the debate however, is the fact that other emerging and re-emerging infectious diseases (EIDs) – such as avian influenza (H5N1), Ebola hemorrhagic fever, and West Nile viral encephalitis – emerge not only as the result of changes in host dynamics or in the pathogen. On the contrary, a range of underlying social- ecological changes such as land use change, deforestation and biodiversity loss seem to contribute to the rise of EIDs globally. Durell Kapan and colleagues article on the social-ecological dimensions of avian influenza is a nice synthesis of how land-use change contributes to increases in H5N1.

So, even if ICT innovations such as Google Flu or GPHIN provide the first signals of pending epidemic outbreaks, they are really not designed to capture changes in underlying social-ecological interactions that induce EIDs. For example, if you want to predict novel outbreaks of Hantavirus pulmonary syndrome (HPS) in Brazil, you might want to keep an eye on deforestation patterns and increases in sugarcane production. Or if you want to stay ahead of increasing risks of Ebola hemorrhagic fever outbreaks in Central West Africa, you might want to track coastal fish stock decrease in the region. These are known to increase “bush-meat” hunting and hence the risk of Ebola outbreaks.

The question is what to call such a system. If field epidemiologist Nathan Wolfe is doing early warning, maybe this approach should be called ecological “early-early warning”?