The Waldo Canyon fire started on June 23, 2012, three miles west of Colorado Springs. Three days later, on June 26, it exploded eastward toward the city, engulfing several neighborhoods. Over 32,000 residents from Colorado Springs, Manitou Springs and Woodland Park, as well as several small mountain communities along Highway 24, were forced to evacuate because of the fire. More than 346 homes were destroyed by the fire. U.S. Highway 24 was closed in both directions for much of the event. The Waldo Canyon Fire is the most expensive fire in Colorado State history with insurance claims totaling more than US$352.6 million dollars, according to The Gazette.
Online Response by the Public:
Data collected includes public tweets from Twitter containing the #waldocanyonfire hashtag posted between June 25, 2012 and July 10, 2012. Over 100,000 messages from more then 25,000 unique Twitter users were collected in this initial period of collection. This data allows us to estimate an hourly usage rate of the hashtag #waldocanyonfire for this period. We show the time series of the posting rate below. Night and day panels are distinguished by the background color (gray covers the hours of 8:00 pm MDT to 8:00 am MDT). Hourly seasonality is clearly visible as the rate of posting drops drastically during nighttime hours. Peak activity occurs around 9:00 pm MDT on June 26th, 2012 — coinciding with the time the fire moved towards residential areas.
Online Response by Official Government Organizations:
In addition to collected public tweets, we enumerated a set of regional and local official government entities involved in the response. These 16 accounts were initially identified as key players in communicating official information with members of the general public. We begin by considering the effect of the fire event on the numbers of individuals following these government accounts. For each of the government accounts in the figure below, we show the percentage increase in follower count and number of accounts being followed during the time period of interest. Many of the accounts, local organizations in particular, gain large numbers of followers during the initial onset of the fire. However, none of the government accounts seems to add many new following relationships over the observation peiod, as seen in the right panel below. Government accounts tend not to change their online relationships during the wildfire, while general public users actively seek information from official accounts by following them.
We also consider the posting behavior of the government organizations. In the figure below, we illustrate the posting behavior and underlying follower network among our set of official accounts. Each node in the network represents a single official organization. Nodes grow in size over time as they gain followers. Each directed tie represents a following relationship on Twitter. We find that these ties do not change over the course of the event. Nodes are highlighted in red when they post a message to Twitter. As tweets are automatically delivered to each of that account’s followers, we highlight these information exchange pathways in blue at the time of posting.
Posting activity is somewhat concentrated in the cluster in the upper right — these tend to be local organizations. Highly active government organizations also gain the most followers, relative to their initial follower count, during this period. We observe time of day seasonality in activity, as we did with general public posting. More tweets are posted during business hours and accounts see little activity at night.
One might also ask about the content of the message posted to Twitter. What kinds of information do these government accounts share with their followers? To the left, we show a graphic illustrating the most commonly used words in the tweets posted by this set of government accounts. We have removed the hashtag (#waldoconyonfire), since it is present in a large majority of posts. We observe emphasis on words related to evacuation and shelter. We also see named places including Interstate 25 (i25). Other words likely refer to requests for donations, instructions, and sources of information.
In addition to Tweet content, we consider elements specific to Twitter included in messages posted. We look at four, what we call “microstructure,” elements: (1) presence of an “RT” to indicate this post is a retweet, (2) presence of an @username directing the content to a specific user, (3) inclusion of external URLs, and (4) inclusion of a hashtag (#keyword). In the figure below, we show the proportion of messages posted by each account that contain each of these four elements of interest respectively. Accounts from regional government entities are shown in dark blue and local entities in light blue. These official government accounts vary greatly in their behaviors related to each of microstructure elements. Very few accounts direct tweets at others. Regional organizations tend to retweet others’ posts at a higher rate than local organizations. A large majority of the tweets posted by government organizations related to the wildfire in Colorado include URLs and hashtags in their content. Hastags are likely useful for organizing content and denoting specific channels that might be of interest to users. Inclusion of URLs indicates that much of the content posted by these accounts provides additional information resources. This combination of findings suggests that these accounts generally use Twitter as a platform for information transmission.
Information-seeking is one likely motivation for Twitter users to follow government organizations during the wildfire. Twitter facilitates information exchange through these following relationships. Another mechanism for information diffusion on this platform involves passing content from one individual to the next. “Retweeting”, as it is called, allows users to re-post information they have been exposed to so that each of their followers also has the opportunity to see the original message. We consider the probability that any given message posted by one of the government accounts is retweeted at least once. We compare the chance of being retweeted before and during the wildfire for each organization, as seen below.
Many government accounts have high chances of being retweeted on average before the wildfire. FEMA has a retweet probability of almost 1. In other words, almost every Tweet posted by FEMA will be retweeted at least once by another user. Though most organizations see their retweet probability increase during the wildfire, this increase is largest for the organizations on the right of the figure. In general, these are organizations local to Colorado Springs.
- When an event occurs local organization gain large numbers of followers.
- Establishing a social media strategy pre-event is important. Organizations should not judge attention demand for social media during non-event periods.
- Content generation on Twitter varies in a predictable way based on the time of day. Interpreting changes in attention needs to take this diurnal cycle into account.
- Original content tends to be produced by local organizations, while retweeted content tends to come from non-locals.
- Low rates of directed messaging indicate a trend to use Twitter as a broadcast channel more than a conversational channel.
- Inclusion of a URLs may show that these organizations recognize the limitations of information shared via Twitter, perhaps due to the character lengths, requiring links out to additional information.
- Hashtag use indicates these organizations are developing a sophistication in how to participate effectively during a disaster event.
This material is based on research supported by the National Science Foundation under awards CMMI-1031853 and CMMI-1031779 and by the Office of Naval Research under award N00014-08-1-1015.
The above analysis was done by HEROIC PIs Jeannette Sutton and Carter Butts and team members Emma Spiro, Britta Johnson, and Sean Fitzhugh. Please cite as follows.
Spiro, E., Sutton, J., Johnson, B., Fitzhugh, S., and Butts, C. (2012). “HEROIC Team Explores Waldo Canyon Wildfire in Colorado.” Online Research Highlight. http://heroicproject.org