Government Social Media Messages across Disaster Phases

Government Social Media Messages across Disaster Phases Clayton Wukich* *Department of Political Science, Sam Houston State University, SHSU Box 2149, Huntsville, TX 77341, USA. Email: [email protected] Abstract Social media platforms allow emergency managers to augment traditional approaches to crisis communication. Research on government messaging, however, disproportionately addresses large-scale disaster response efforts, neglecting smaller-scale incidents and activities across other disaster phases (e.g., prevention, mitigation, preparedness, and recovery). This article offers a more complete analysis of messaging strategies by integrating existing typologies and analyzing state-level emergency management agencies in the United States over a one-year period. Findings illustrate a range of messages, with response and preparedness being most prevalent. While all agencies disseminated situational information, protective action messages, and preparedness guidance, fewer engaged in more interactive tactics that facilitate public participation and interagency collaboration. More work, therefore, is needed to pursue social media’s full potential in promoting risk reduction.

1

1. Introduction Emergency managers face variation in terms of the size, scope, and nature of potential hazards (Mileti, 1999; Rosenthal, Boin, & Comfort, 2001). They must, therefore, promote and coordinate a variety of risk reduction efforts among different audiences; this requires a multifaceted approach to crisis communication (Falkheimer & Heide, 2006; Sellnow & Seeger, 2013). Communicating to large audiences represents a challenge because the failure to disseminate critical pieces of information prevents constituents from making informed decisions which results in suboptimal outcomes (Quarantelli, 1988; Comfort, 2007). Social media websites (i.e., networking platforms through which users generate and share content) provide an emerging set of tools that augment existing communication outlets and make possible the rapid dissemination of information (Sutton et al. 2014; Olteanu, Vieweg, & Castillo, 2015). A key challenge for emergency managers is to develop effective messaging strategies that harness social media’s potential. Research on social media messaging has concentrated on response and initial recovery efforts for large-scale incidents including natural disasters (Hughes, St. Denis, Palen, & Anderson, 2014), technological disasters (Sutton et al., 2013a), and terrorist attacks (Sutton et al., 2013b). Large-scale incidents, however, are not the only situations in which emergency managers operate. In practice, emergency managers use social media platforms across all disaster phases (e.g., prevention, mitigation, preparedness, response, and recovery) and many incident types. While practitioner-based training resources acknowledge this fact (FEMA, 2013; CDC, 2014), the academic literature largely omits the diversity of messages generated outside of large-scale response operations. Furthermore, while researchers and practitioners prescribe interactive tactics to engage citizens in conversations and coordinated action to reduce risk (Veil, Buehner,

2

& Palenchar, 2011; Crowe, 2012; Dufty, 2012; Sievers, 2015), limited research exists that systematically addresses the extent to which emergency managers adopt and use these approaches (Atkinson & Wald, 2007; St. Denis, Hughes, & Palen, 2012; Wukich & Mergel, 2015; Reuter, Ludwig, Kaufold, & Spielhofer, 2016). This article provides a more comprehensive analysis of messages disseminated by government agencies; specifically, it accounts for each phase of disaster management, several incident types, and a variety of administrative activities. In doing so, it presents a more complete message typology than previous studies. Messages disseminated by state-level emergency management agencies in the United States over a twelve-month period are analyzed and the following research questions are examined: 

To what extent do emergency managers disseminate messages across disaster phases?



What message types are employed most frequently?



What range of hazards are addressed?

These exploratory questions address the willingness of organizations to accept new technologies and develop more interactive modes of communication with citizens and other organizations. While findings demonstrate that agencies relied on a variety of response-related messages, preparedness content was also prevalent. This message repertoire provides a diverse set of tactics for emergency managers who seek to promote risk reduction across a range of activities. However, while all agencies disseminated messages such as (a) protective action information, (b) situational information, and (c) preparedness guidance, fewer engaged in more interactive tactics to promote public participation and interagency collaboration. This finding suggests that more work is needed in order to pursue social media’s full potential in promoting risk reduction.

3

2. Types of messages across disaster phases and social media’s potential Emergency managers craft messages across a variety of operational conditions (Sellnow & Seeger, 2013). Content varies based on the situation and the audience (Falkheimer & Heide, 2006). Pre- and post-disaster impact offers an initial distinction between situations, and during these periods, different types of information are needed to make risk reduction decisions (Mileti, 1999; Reynolds & Seeger, 2005). The phases of disaster management (e.g., prevention, mitigation, preparedness, response, and recovery) further differentiate those operational perspectives. During response operations, for example, protective action messages offer advice to citizens regarding how they can reduce their immediate exposure to risk (Mileti & Sorensen, 1990; Lindell & Perry, 2012), and situational information provides additional insight while also contributing to an interagency common operating picture upon which officials draw to inform their decision making (Comfort, 2007). Reynolds and Seeger (2005) noted the importance of communicating risk prior to a crisis in order to increase public understanding and engender interagency cooperation. Those messages can be planned and structured over time to educate and persuade audiences regarding the necessity to prevent, mitigate, and/or prepare for shared risk. Reynolds and Seeger (2005) pointed out that these messages should be principally persuasive in nature and address known probabilities of potential hazards as opposed to the principally informative nature of messaging during an actual disaster. Many emergency managers have implemented those previously offline activities online through social media messaging, particularly protective action and situational information (Hughes et al., 2014; St. Denis, Palen, & Andeson, 2014; Sutton et al., 2014; Olteanu et al., 2015). Government agencies can use social media to quickly and directly reach large and diverse

4

audiences rather than rely only on intermediaries such as traditional news media for conduits (Hughes & Palen, 2012), and they can proactively encourage citizen participation during critical situations to elicit intelligence, cultivate situational awareness, and promote the further distribution of key pieces of information (Chatfield, Scholl, & Brajawidagda, 2013; Meijer 2014; Sutton et al., 2014). Using social media to share information with other agencies can also improve operational effectiveness (Jung & Park, 2016; Sutton et al., 2013a). Despite its growing popularity, however, not all emergency managers use social media (see Helsloot & Groenendaal, 2013; Hunt, Smith, Hamerton, & Sargisson, 2014; Su et al., 2013) and those who do demonstrate varied preferences ranging from one-way messages to two-way and multi-way communication strategies (Mergel, 2014; Reuters et al., 2016; Wukich & Mergel, 2015). Government agencies face challenges in using social media to improve the effectiveness of their communication strategies. One limitation is the brevity of messages; Twitter, for example, limits messages to only 140 characters, and these terse messages may lead to ineffective communication efforts. For example, Bean et al (2016) conducted focus groups with individuals to analyze message effectiveness. Their findings demonstrate how commonly-used message styles were often considered to be uninformative, confusing, and fear inducing. Another challenge relates to the openness facilitated by social media (Hughes & Palen, 2012), the public’s expectation of government omnipresence, and the resulting demands on personnel’s time (Zavattaro & Sementelli, 2014). Whereas public information officers once maintained greater control over disaster-related messages (Hughes & Palen, 2012), competing voices now generate considerable noise, much of which is inaccurate (Starbird, Maddock, Orand, Achterman, & Mason, 2014). This compels emergency managers to dedicate time to scan posts for rumor control purposes. Furthermore, emergency managers face increasing pressure from

5

constituents to more quickly disseminate official messages and guidance (Hughes & Palen, 2012). These commitments might impede social media adoption and use especially in organizations operating in conditions characterized by resource scarcity. For example, Su et al. (2013) reported that 75% of state and local emergency managers that they surveyed in the U.S. indicated that they had insufficient staffing to support social media monitoring and the remaining 25% indicated that their available staff had additional responsibilities that often detracted from such efforts. The following sections assess the literature on social media messages in emergency management to illustrate the variety of message types employed. In the growing literature on social media use, the response and initial recovery phases have received the most attention by far; however, some studies have addressed potential uses during other phases particularly preparedness (Crowe, 2012; Dufty, 2012; Sievers, 2015).

2.1. Protective action messages Officials involved in disaster response craft protective action messages to empower community members so that they may reduce their exposure to risk in the event of a disaster (Mileti & Sorensen, 1990; Lindell & Perry, 2012). Drawing from the existing literature, Sutton et al. (2014) outlined the characteristics of effective protective action messages, which ideally include (a) explicit guidance on what citizens should do to reduce their exposure to risk; (b) a specific timeline of prescribed actions; (c) the location of those affected; and (d) a description of the impending hazard and its consequences. Government agencies have disseminated protective action information via social media in the form of warnings and watches (Chatfield et al., 2013; Sutton et al., 2014), advisories (Bruns, Burgess, Crawford, & Shaw, 2012; Imran, Elbassuoni, Castillo, Diaz, & Meier, 2013; Sutton et

6

al., 2014), and evacuation orders (White, 2012; Sutton et al., 2014). Chatfield et al. (2013), for example, reported on a tsunami early warning program in Indonesia and the role citizens play in amplifying those social media messages. Sutton et al. (2014) analyzed advisory messages during a major wildfire in Colorado (USA) that contained suggestions on “how to respond, what actions to take, and what to refrain from doing during the disaster” (p. 775). Bruns et al. (2012) reported that advisories were commonly posted during the 2011 large-scale flooding in Queensland, Australia. And Reuters et al. (2016) document the willingness of emergency service personnel across Europe to use social media to disseminate similar pieces of information during disasters. The abundance of examples suggests that this type of messaging is frequent and widespread.

2.2. Situational information Emergency managers depend on critical pieces of information to develop situational awareness in order to (a) better understand the nature of a situation; (b) interpret key indicators for change; and (c) project that situation’s future status (Endsley, 1995; Comfort, 2007). Indicators are often monitored during both disaster and non-disaster conditions. During non-disaster conditions, organizations may monitor risk indicators such as weather forecasts and 911 emergency calls (Comfort, 2005). During an emergency, responders seek out information from a variety of sources. Baseline components of situational information include an incident’s type, location, and time. Responders are also interested in hazard impact and implications regarding vulnerable populations. They typically receive this information from 911 call centers and other official sources (Comfort, 2005), however, social media now provide alternative sources (St. Denis et al., 2012; Hughes et al., 2014).

7

Government agencies represent key sources of situational information via social media (Bruns et al., 2012). For example, Hughes et al. (2014) identified a range of situational information disseminated by fire and police departments during Hurricane Sandy, including weather updates, debris removal status, and the availability of utilities; they also observed that information on changing storm conditions was one of the most frequently disseminated types of messages. Information on damage, casualties, and other consequences of an incident is also commonly shared (Hughes et al., 2014; Olteanu et al., 2015). For example, analysis of social media messages in the immediate aftermath of the Haiti earthquake revealed a focus on collapsed buildings, injuries, and subsequent medical needs (Caragea et al., 2011). Other consequences reported might include road hazards, the unavailability of goods and services, and the closure of institutions and critical infrastructure. During the 2013 Colorado floods, St. Denis et al. (2014) reported closure messages as the most frequently disseminated messages. Hughes et al. (2014) noted a similar finding in their Hurricane Sandy study.

2.3. Resource provision Social media has been used to inform the public and other agencies about the distribution of key resources (Bennett, 2014; Hughes et al., 2014). Following the Haiti earthquake, the availability of medical services and food was disseminated widely (Caragea et al., 2011). During the 2011 Queensland floods, messages contained advice regarding how to locate and acquire other public goods and services (Bruns et al. 2012). Lindsay (2011) noted the applicability of social media to communicate “how to apply for assistance, announcing application deadlines and providing information and links to other agencies and organizations that provide recovery assistance” for recovery purposes (p. 22). And Bennett (2014) noted that this type of recovery information was

8

disseminated following a 2011 tornado in Alabama, (USA) and enabled agencies to distribute resources to those in need.

2.4. Crowdsourcing and other direct interaction with the public To this point in the review, only passive messaging requiring input from the originator has been evaluated. Ideally, social media can be used across disaster phases to promote online conversations and coordinate risk reduction activities. This approach contrasts with traditional one-way government communication strategies and limits the extent to which government agencies and the public interact (Veil et al., 2011; Mergel, 2014). Wukich and Mergel (2015) documented a set of messaging strategies in which agencies sought to directly engage the public and other agencies; those strategies reject one-way communication modes and instead seek out two and multi-way interactions. Crowdsourcing, a practice in which government agencies request public input to solve a problem (see Brabham, 2013), and other resource seeking requests represent key examples. Some organizations use social media to explicitly request situational information from the public. Atkinson and Wald (2007) documented this type of intelligence gathering in their analysis of the U.S. Geological Survey’s “Did You Feel It?” campaign in which the agency elicited first-person reports of seismic activity around the world. Another example of this type of intelligence gathering is Jarkarta, Indonesia’s flood identification program in which city officials identify potential eyewitness accountsand then engage individuals in conversations to elicit more information (Meier, 2015). In addition to intelligence gathering, government agencies may recruit volunteers, attempt to raise relief funds, and make requests for additional resources such as food and clothing (Crowe, 2012; Hughes et al., 2014). While these requests are not limited to response operations, Wukich

9

and Mergel (2015) reported a relatively low number of requests in their study of U.S.-based agencies across all phases. Furthermore, two-way message strategies, in general, appear to be used infrequently across both the United States (Su, Wardell, & Thorkildsen, 2013) and Europe (Reuter et al., 2016).

2.5. Prevention, mitigation, and preparedness Veil et al. (2011) speculated that “the midst of a crisis is not the time to suddenly try a hand at social media. To build partnerships and build trust, the discussion with the public should already be taking place” (p. 119). Daily communication can help to establish the credibility so important to develop large online followings (Veil et al., 2011; Mergel, 2014). Through social media, government agencies now have a direct and ready conduit to engage constituents regarding prevention, mitigation, and preparedness guidance (Crowe, 2012; Dufty, 2012; FEMA, 2013; CDC, 2014; Sievers, 2015). While social media platforms offer vehicles to disseminate education-related messages, promote interagency exercises, and engage citizens in conversations about risk reduction, scant research has addressed the extent to which these activities take place. Wukich and Mergel (2015) reported that over a third of the government agency messages that they analyzed pertained to prevention, mitigation, and preparedness. Most were one-way, education-related messages that promoted household risk reduction. Some emergency management agencies employed more interactive tactics in the form of conversation starting prompts to engage the public on risk reduction. Some incentivized participation by offering prizes or recognition for taking part (Wukich & Mergel, 2015). This practice, referred to as gamification or “the use of game design elements in non-game contexts” (see Deterding, Dixon, Khaled, & Nacke, 2011, p. 1), rewards

10

risk reduction conversations and efforts. Wukich and Mergel (2015) reported a limited amount of such messages posted across the U.S., however.

2.6. Administration and operations Government agencies across different policy domains use social media to report their day-to-day activities and thus demonstrate a certain degree of transparency (Mergel, 2012). Emergency managers are able to do the same. “Examples include announcing a director’s public schedule, posting employment opportunities, and trumpeting agency accomplishments. Again, these messages are not directly related to field operations, but instead focus on administrative accomplishments and public relations” (Wukich, 2015, p. 285). Messages regarding agency operations also promote transparency. They include descriptions of the actions taken by personnel in pursuit of their organization’s mission; they also include executive actions taken such as emergency declarations and preparedness proclamations. These messages offer situational awareness for others. While Hughes et al. (2014) noted that those types of messages were prevalent during Hurricane Sandy, less is known about the extent to which agencies post similar messages across other disaster phases.

2.7. Rumor management Much of what is posted online by individuals and by some organizations is inaccurate (Starbird et al., 2014), and in order to ensure that the public has accurate information, some government agencies monitor social media for rumor management purposes (Veil et al., 2011; St. Denis et al., 2014; Wukich & Mergel, 2015). During the 2011 Queensland floods, Bird, Ling, and Hayes (2012) reported that local police scanned social media for inaccurate information and posted corrections. After Hurricane Sandy, FEMA’s social media team created a “rumor control” initiative in which they identified false rumors, worked with other staffers to gather clarifying

11

information, and shared those messages via their social media accounts (U.S. Department of Homeland Security, 2013). Clarifying widely circulated false rumors can be difficult and time intensive, however. Analyzing tweets during the Boston Marathon bombing, Starbird et al. (2014) found that correction messages received much less attention than the original rumors. Furthermore, Sutton et al. (2013) observed that rumor management and corrections messages were far less frequently disseminated by government agencies than other types. The labor-intensive nature of this type of message, therefore, creates barriers to implementation (Wukich, 2015).

2.8. Emotional messages, opinions, and commentary Whether directly impacted by an incident or not, citizens use social media to engage in emotional exchanges in attempts to make sense of those uncertain and stressful situations (Bruns et al., 2012; Heverin & Zach, 2012; Olteanu et al., 2015). Just like the average citizen, emergency managers are people with emotions and opinions and, at times, may express their emotions during extreme events via social media. Research demonstrates that emergency managers post messages of support and well wishes to those affected by disasters and gratitude to those who have provided support (Hughes et al., 2014). Some might use social media to appraise the actions or inactions of others, although as bureaucrats, emergency managers have fewer incentives to engage in such activities and perhaps more to lose than others outside of government (see Boin, t’Hart, Stern, & Sundelius, 2005).

3. Methodology 3.1. Case selection

12

The existing research concentrates predominantly on response and initial recovery efforts to large-scale incidents (Olteanu et al., 2015). This focus has illuminated a range of practices, but simultaneously limits our ability to understand the full repertoire available to emergency managers across all disaster phases as well as offering insight into the administrative work required. This article examines state-level emergency management agencies operating in the United States over a twelve-month period of observation (April 2013 through March 2014), an approach that captures not only a number of large-scale incidents (e.g., the Boston Marathon bombing, the West, Texas fertilizer plant explosion, the Colorado floods, the Oso, Washington landslide, and 56 other federally declared disasters), but also discusses a larger number of more routine emergencies (e.g., smaller storms and accidents). In addition, the case selection facilitates the analysis of other disaster phases and administrative work while accounting for variation across geographic regions and exposure to various hazards. Analysis at the state level, rather than at the local level, is appropriate because state agencies generally dedicate more personnel and resources to development and implementation of their social media policies (Su et al., 2013), thus creating a potentially larger and more diverse set of messages per organization.

3.2. Data collection State emergency management agencies’ websites were first reviewed to identify 51 accounts for 48 agencies.i This approach has been used previously to reduce the risk of selecting unofficial accounts (Sutton et al., 2014; Wukich & Mergel, 2015). The resulting account names were then used to collect data directly from Twitter. Through its application programming interface (API), Twitter makes users’ messages publicly available for download. Facebook does not offer the same data accessibility. By means of a program designed in-house to interface with Twitter’s API, 38,682 time and date-stamped messages, including text, were downloaded in an Excel

13

spreadsheet format for analysis. These messages represented the complete record of publication, not a sample, from the 48 agencies analyzed.

3.3. Data analysis One goal of this article was to understand the extent to which emergency managers used social media across all disaster phases and hazard types. Accordingly, each message was manually assigned two additional sets of attributes: disaster phase and incident type. Disaster phase was operationalized to include prevention, mitigation, preparedness, response, and recovery. “Not applicable” was included to capture administrative activities, unrelated content, and nonemergency management operations. This approach conforms to previous research (Wukich & Mergel, 2015). In addition, each message was assigned an incident type. This attribute included natural disasters, technological and accidental hazards, and terrorist attacks, all of which are commonly used attributes in disaster research (Mileti, 1999). Additional categories–public health, household safety, crime, all-hazards, preplanned events, and not applicable–were added to offer a more complete list. Two analysts working in concert manually coded each message. They reviewed each other’s work and discussed and resolved discrepancies to promote intercoder reliability. Frequency distributions for both attributes–disaster phase and incident type–were analyzed via a cross tabulation. A chi-square test of independence was used to determine whether there was a relationship between those variables. Messages were also coded by whether they occurred during federally declared disasters to identify those associated with large-scale incidents. Sixty such incidents occurred during the period of observation and were identified via FEMA’s website; FEMA posts federally declared disaster declarations at https://www.fema.gov/disasters. Count data offers an initial estimate

14

regarding the extent to which emergency managers use social media only for large-scale disasters. Another goal of this article was to identify the most frequently disseminated message types. Each message was coded to determine whether it belonged to one or more of several message types defined in Table 1. Message types were identified from both academic research (e.g., Hughes et al., 2014; Sutton et al., 2014; Olteanu et al., 2015; Wukich & Mergel, 2015) and practitioner-based training resources (e.g., FEMA, 2013) to provide a comprehensive typology. Many messages were associated with response and initial recovery operations such as protective action and situational information, the provision of resources, intelligence gathering, and rumor management. Other types were associated with prevention, mitigation, and preparedness, including educational messages intended to empower individuals and external agencies to reduce their exposure to risk. Other messages sought to engage the public directly in conversations about risk (see conversation starters); while others sought to engage the public through games and contests to promote preparedness (see gamification). Message types were not mutually exclusive; therefore, messages that represented multiple types were coded accordingly. Data were analyzed using frequency counts and measures of central tendency and dispersion to identify the extent to which each message type was used.

4. Findings 4.1. Disaster phase and hazards Most messages pertained to response operations during natural hazard-related incidents. Table 2 presents a cross tabulation that reports message frequency by disaster phase and hazard type. A chi-square test of independence was conducted, and a significant interaction was found (χ²(35) = 15

64209.8, p < .0005) between the two variables. Response (55.2%) and recovery-related messages (6.8%) comprised the majority of all distributed messages (62.0%) over the twelve-month period of observation. Most of those messages (e.g., response 89.2%; recovery 88.4%) apply to natural hazards, including both routine incidents and large-scale disasters such as tropical storms, tornadoes, and blizzards. In all, 67.9% of all messages related to natural disasters; 16.4% pertained to all-hazards messaging; and 3.5% dealt with public health and household safety. Less attention was paid, however, to technological and accidental hazards (3.2%), crime (2.9%), and terrorism (1.1%). Preparedness messages (29.9%) constituted a sizable portion of all content. While many of those tweets addressed natural hazards (39.3%), the majority (50.3%) took an all-hazards approach to encourage the public to adopt risk reduction practices applicable across an array of incident types. For example, Michigan tweeted, “Create a family emergency plan so that you know what to do, where to go & how to communicate with your family during a disaster. #NATLPREP.” While preparedness tweets were widespread, there were fewer prevention (2.7%) and mitigation (1.2%) messages. Prevention tweets generally pertained to risks related to public health, household safety, and crime. Mitigation messages often referenced insurance. For example, Massachusetts tweeted: “Statistics indicate majority of Americans have inadequate insurance, though the cost of additional coverage is often affordable.” This type of tweet ideally raised awareness regarding the underinsured population and prompted individuals to evaluate their current situations. Also frequent were messages not addressing risk reduction at all. Labelled “not applicable,” these tweets (4.3%) included commentary and conversations not related to emergency management. Examples addressed operations outside of emergency

16

management activities such as regularly scheduled blood drives, and some included other pieces of news and information not applicable to the policy domain.

4.2. Additional Messages Types 4.2.1. Protective action and situational information In aggregate, Table 3 reports that most messages pertained to response-related themes, and Table 4 presents their measures of centrality and dispersion per agency. Situational awareness comprised 24.5% of all messages (M = 196.9; SD = 255.5). Additionally, advisories (15.7%; M =126.6; SD = 153.4), warnings and watches (10.4%; M = 83.8; SD = 167.5), and hazard impact (9.6%; M = 77.0; SD = 91.7) were commonly posted. In the midst of severe flooding, Colorado tweeted the following advisory, “Just 6 inches of moving water can knock you off your feet in a flash flood. Turn around, don’t drown. #WaldoFlood #COwx.” However, only 11.2% of all messages were posted during large-scale, federally-declared disasters; therefore, the vast majority of messages addressed (a) smaller-scale, more routine incidents; (b) agency reaction to risk indicators that failed to materialize into incidents; and (c) education-related messages, among others. Wisconsin, for example, tweeted the following situational information, “The heavy snow continues in northwestern Wisconsin... http://t.co/E5VSCIkFqy.” This type of message made the public and other organizations aware of the continued snowfall, but did not address a major disaster and was a quite common form. In addition, many agencies regularly retweeted warnings and watches posted by their regional National Weather Service offices. Vermont, for example, posted, “RT @NWSBurlington: Flood Advisory thru 11:30pm for an ice jam on the Mad River near Rt 100B. Don’t drive on water-covered roads! #VTwx http…”

17

This example included situational as well as protective action information. However, evacuations, closures, and openings were posted with less frequency, most likely due to the fact that those messages are more associated more commonly with large-scale disasters. Of the 246 evacuation messages, for example, 136 (55.3%) were posted during federally declared disasters. While less common, these types of protective action messages possessed immediate, life-ordeath implications. For example, Colorado tweeted the following during widespread flash flooding, “@wigginsfire: The town of Orchard is in immediate danger. EVACUATE NOW.” Not all messages were so dire and time sensitive. Twitter provides emergency managers with a constant potential audience and all agencies posted preparedness guidance. The next section investigates that type of content. 4.2.2. Education-related, training, and drills All agencies posted preparedness-related content; three associated types were education-related messages that appealed to individuals to reduce their exposure to risk, tweets about interagency training opportunities, and messages promoting community-wide preparedness drills. Education messages (23.8%; M = 196.9; SD = 245.2) promoted risk reduction activities at the household, organizational, and community-levels. While most related to preparedness (86.1%), a notable amount (10.5%) referenced prevention and mitigation (2.1%). In addition, some agencies used existing incidents as focusing events to raise awareness of preparedness activities. For example, Colorado tweeted, “With the fires raging around CO, now may be a good time to consider how your family can prepare, and about... http://t.co/PxiOWxo9df.”

18

Other agencies used recovery periods to educate their communities regarding preparedness for the next incident. Following a large-scale blizzard, South Dakota posted, “Surveys from after winter storms show 49% of respondents were glad they had extra food, water, beer, coffee and pet food. #sdwx.” Other common hazards addressed included flooding, hurricanes, earthquakes, and cyber crime. However, most education messages were framed along all-hazard lines and promoted preparedness tips applicable along a range of risk factors and disasters. Another preparedness-related message type had to do with training opportunities for emergency managers. In all, 3.8% (M = 30.5; SD = 39.4) of all content promoted, described, or coordinated professional trainings, conferences, or exercises. Vermont posted, “Social media for emergency response training hosted by DEMHS #ndptc http://t.co/KnQPvulJ8B.” The audience for those types of messages were other agencies; therefore, this is one example in which tweets clearly were intended to promote interagency collaboration. Promoting and coordinating community-wide drills was also accomplished through preparedness messages. The Great American ShakeOut Earthquake drills, for example, have been implemented across the U.S. to improve preparedness by practicing how to react during earthquakes. Oregon promoted their version of the event with the following, “Get ready to Drop! Cover! Hold! with over 260,000 #OregonShakeOut win prizes #ShakeOutSelfie @RedCrossPDX http://t.co/WNKD7HLyb4.” All of those messages–education, training, and drills–encouraged others to engage in specific risk reduction activities. Not all messages analyzed specifically appealed to individuals to act, however. Many were crafted to describe agency operations. 4.2.3. Operations and administration

19

Across all phases, agencies posted messages regarding their actions and the actions of other agencies. This type accounted for 11.2% of all messages posted (M = 83.8; SD = 167.5), and many provided situational awareness for others. For example, Nevada tweeted messages regarding the implementation of control burns to mitigate wildfire risk, “The Nevada Division of Forestry will be conducting a controlled burn at the Cole Ranch in Lincoln County.” This type of message informs other agencies of planned operations and ideally reduces the anxiety of residents who then expect to see burning in their areas. While only 3.0% of operational messages reported mitigation activities, 26.7% of all messages related to preparedness (see Table 3). These messages generally described an agency’s preparedness activities and at times included educational elements as well. Some agencies reported on the work of others. Missouri posted the following regarding a U.S. Geological Survey operation to acquire additional seismic risk data, “USGS will fly low-level flights over New Madrid fault zone to locate concealed faults and gain new info about area. http://t.co/8Y6dsM1K40.” Although this type of data is useful for mitigation and preparedness activities, the majority of operational messages (51.9%) addressed response activities. In the event of a risk indicator such as a severe weather forecast, emergency managers may respond by standing up their emergency operations centers (EOCs) and readying their response operations. Ohio, for example, posted a message detailing such activities, “The State EOC stands ready as we assess & monitor forecasted systems as they move into Ohio #ohwx http://t.co/cvyLMeUi7b.” This tweet communicated the agency’s response posture to other agencies while also communicating the specific storm warning to the public.

20

Other messages addressed operations during slower, longer-lasting crises. California, for example, tweeted about their response to ongoing drought conditions, “Gov’s Drought Task Force meets with Mendocino officials on #cadrought in Ukiah http://t.co/JtQLXkOcAi.” This message let constituents know that state personnel were allocating time for the problem. Other messages contained more time sensitive information. Many agencies posted when disaster declarations were requested and declared. Following the Oso landslide disaster, Washington State tweeted, “Governor Inslee requests federal disaster declaration for Oso Landslide damage http://t.co/TulmsijwNN.” Fewer messages (1.3%; M = 8.5; SD = 17.6) of all posted dealt specifically with nonoperational administrative matters. This type included news regarding job postings, awards won or awarded, and staff promotions and retirements. In all, both operational and administrative tweets provided the public with information regarding what agencies were doing. 4.2.4. Opinions and commentary Of all tweets, 4.3% (M = 34.6; SD = 69.2) included some type of opinion or commentary by the author. These types of messages offered a more personal tone. Florida, for example, whose personnel participated in a large-community event posted, “Having a lot of fun @radiodisney holiday concert in @CelebrationFLA. Lots of smiles tonight! @IDEASorlando http://t.co/RIDYiSxVmk.” During a holiday, Mississippi’s public information officer tweeted, “I

want to wish everyone a safe and Happy Thanksgiving. We are all so thankful for a relatively quiet hurricane season http://t.co/pakfpQWMD3.” The most prevalent type of message, though, communicated thanks to community partners and other agencies for their contributions across all disaster management phases. Others offered congratulations and well-wishes to others for professional achievements such as appointments

21

and promotions. These messages appear to have promoted good will and online acknowledgements of accomplishments. 4.2.5. Rumor management, crowdsourcing, and other interactive strategies While protective action, situational information, and preparedness messages were frequently disseminated by all agencies, more interactive message types such as rumor management and crowdsourcing management were less widely adopted. Table 4 illustrates the lack of implementation of those message types. While most agencies posted at least one crowdsourcing message (e.g. intelligence gathering, volunteer recruitment, fundraising, and others), only 22 agencies posted a rumor management tweet. Rumor management involves monitoring social media feeds to identify false rumors and then provide corrected information either to the source or through a mass distribution. For example, New Jersey tweeted, “Rumor Control: Despite media reports, NO tornado touchdown in Butler, Morris Co. last night. Info confirmed by NWS. Possible microbust.” More agencies engaged in crowdsourcing by directly asking the public for information, action, and/or resources. Intelligence gathering was one such tactic with 38 agencies engaged, averaging 4.5 messages over the course of the year. Wyoming, following seismic activity, replicated USGS’s “Did You Feel It” campaign, “Did any of you feel the 4.9 earthquake 12 miles west of Fort Washakie? It happened just a little after 7 a.m.... http://t.co/AnELc0BmsT.” More agencies engaged in volunteer recruitment (42 agencies; M = 5.1; SD = 20.9), fundraising (39 agencies; M = 3.5; SD = 5.9), yet fewer requested additional resources from other parties (28 agencies; M = 1.6; SD = 4.6). The most common interactive tact employed the use of conversation starters (44 agencies; M = 13.0; SD = 18.1). Florida, for example, posted,

22

“What’s in your disaster supply kit?! It is always good to be prepared! Tweet us your “must haves”! #whatsinyourkit.” Prizes and other rewards were used by 36 agencies to incentivize those conversations. Arizona created a contest, the Emergency Kit Cook-Off, that rewarded citizens for creating recipes using dry and canned good products. The goal was to promote family preparedness through the stockpiling of food. They posted, “Have you submitted a #KitCookOff recipe yet? Share your culinary creativity & get rewarded for it @ http://t.co/r0Vje282uB #NatlPrep #smem.” But again, this type of interactive behavior was far less common than the dissemination of situational information, protective action messages, and general preparedness guidance.

5. Discussion The existing literature on social media messaging is almost entirely comprised of response and recovery case studies. Those studies illustrate innovative practices to quickly reach the public post-disaster when attention is high (Hughes et al., 2014; St. Denis et al., 2014; Sutton et al., 2014). Other studies, however, illustrate (a) the underused nature of social media in emergency management (Helsloot & Groenendaal, 2013) and (b) the complete lack of use in certain critical situations (Hunt, Smith, Hamerton, & Sargisson, 2014). Indeed, survey research indicates a range of social media adoption and use preferences in both the United States (Su et al., 2013) and in Europe (Reuters et al., 2016). Yet little research systemically examines the actual messaging patterns of agencies across multiple disaster phases. To address this gaps, this article examines the behavior of U.S.-based agencies over an entire year to capture a range of messaging strategies that include each disaster phase. As a result, it presents a more comprehensive message typology than previously published and finds that many agencies employ a diverse set of messaging strategies.

23

The vast majority of messages analzyed were posted during situations other than large-scale disasters (e.g. routine incidents and other phases). This article reports that all agencies frequently posted situational awareness information and protective action messages often in the lead up to and during smaller-scale incidents. This finding conforms to previous research government message preferences during large-scale disasters (Bruns et al. 2012; Sutton et al., 2013b). This article also contributes to our understanding of how social media is used outside of response operations. Pre-event messages, for example, have not received much scholarly attention despite their application by practitioners (see FEMA, 2013; CDC, 2014). Through social media, agencies can now directly reach audiences with prevention, mitigation, and preparedness messages, whereas in the past they depended on earned media to relay information piecemeal and at the discretion of traditional media outlets. This article reports that education, training, and drill-related messages comprised a notable percentage of all content and offer strategies for emergency managers to communicate non-response guidance. These findings provide empirical support for previous research that speculated about social media’s utility (see Crowe, 2012; Dufty, 2012; Sievers, 2015). This article also highlights the irregularity of more interactive communication strategies to engage citizens and other organizations. Crowdsourcing and other interactive strategies were employed by the majority of agencies, but less frequently than other message types. Rumor management was used even more sparingly, which conforms to previous research (Sutton et al., 2013b). These strategies require additional time and effort to implement, which appears to present barriers to use (see Hughes & Palen, 2012; Su et al., 2013). When done effectively, though, these more labor-intensive practices may build trust between government agencies and their constituents (Veil et al., 2011; Mergel, 2014) and provide value to risk reduction efforts

24

(Atkinson & Wald, 2007; Chatfield et al., 2013; Meier, 2015). These more interactive types of messages should be considered by emergency managers. Despite the fact that all agencies employed multiple message types, the high standard deviation scores reported in this article signal that some agencies posted more frequently than others and with more regularity. While this might be explained by the number of incidents faced, it might also be a product of the time and effort required to implement certain strategies. Some agencies undoubtedly have more personnel and resources to dedicate to social media than others. This is particularly germane to crowdsourcing and rumor management messages. Also, government agencies are still developing their standard operating procedures regarding social media (see Mergel & Bretschneider, 2013); many may not yet have developed their full repertoire of tactics, and other agencies still may not possess organizational cultures that value social media or support from their leadership (see Reddick & Norris, 2013). Further research is needed to examine those issues.

5.1. Future research Researchers may expand this analysis to other levels of government and other regions of the world. U.S.-based state agencies and those from other developed economies possess the resources to dedicate to social media adoption; however, other regions may be less inclined to do so (see Wukich & Khemka, 2016). Understanding the factors that limit social media adoption and use is critical in developing strategies to maximize its potential. A limitation of this article is that it fails to evaluate the effectiveness and reach of government social media messages. Despite the optimism of many researchers, Helsloot and Groenendaal (2013) suggested that the praise given to social media’s potential may not match its utility in practice. Future research on efficacy and reach can address those concerns. Bean et al.

25

(2016) evaluated potential message efficacy for post-extreme event response operations. Other researchers may consider evaluating the extent to which agencies actually impact the behavior of their constituents and the extent to which those behavioral changes influence disaster outcomes. Ripberger, Jenkins-Smith, Silva, Carlson, and Henderson (2014), for example, analyzed public attentiveness to tornado watches and warnings on Twitter, reporting that people paid greater attention to the more severe warnings than they did watches. Future research should advance this critical line by addressing how the public reacts to messages across disaster phases and the extent to which they then act to reduce their exposure to risk. Gaining a better understanding of reach (i.e., the ability of an agency to reach its intended audience) is related and also critical, for if community members and other agencies fail to receive key messages, they likely will not make informed decisions (see Quarantelli, 1988; Comfort, 2007). Some research addresses reach; Sutton et al. (2014), for example, presented a set of characteristics that increase the likelihood of protective action messages being passed along such as the inclusion of hashtags and hyperlinks. However, other means can be used to extend reach such as (a) increasing the number of followers; (b) investing in paid online advertising; (c) targeting specific communities via relevant messaging and hashtags; and (d) organizing chats with high profile officials. Also, reach for preparedness purposes may be extended by addressing ongoing events in other regions when the public’s attention is elevated and more receptive.

5.2. Conclusion In summary, U.S.-based agencies have implemented a variety of messaging strategies that address actions across all disaster phases. This article demonstrates the diversity of messaging strategies available to emergency managers and reports that the vast majority of state-level

26

agencies in the U.S. implement these forms of communication. Fewer agencies, however, regularly employ more interactive tactics, which suggests that more work is needed in order to pursue social media’s full potential in promoting risk reduction.

Acknowledgments: The author thanks Dr. Alan Steinberg and Mr. Johnny Nguyen for their work in developing the data collection tool, and Dr. Steinberg for his substantial contribution to the data coding process. The author also wishes to thank the journal’s reviewers for their critiques. This project was supported by Sam Houston State University’s College of Humanities and Social Sciences, the Department of Political Science, and the Center for the Study of Disasters and Emergency Management.

27

References Atkinson, G. M., & Wald, D. J. (2007). ‘“Did You Feel It?” Intensity Data: A Surprisingly Good Measure of Earthquake Ground Motion’, Seismological Research Letters, Volume 78, Number 3, pp. 362-368. Bird, D., Ling, M., & Haynes, K. (2012). ‘Flooding Facebook - The use of social media during the Queensland and Victorian floods.’ Australian Journal of Emergency Management, Volume 27, Number 1, pp. 27-33. Bean, H., Liu, B. F., Madden, S., Sutton, J., Wood, M. M., & Mileti, D. S. (2016). ‘Disaster Warnings in Your Pocket: How Audiences Interpret Mobile Alerts for an Unfamiliar Hazard’, Journal of Contingencies and Crisis Management. Bennett, D. M. (2014). ‘How do emergency managers use social media platforms?’ Journal of Emergency Management, Volume 12, Number 3, pp. 251-256. Boin, A., t’Hart, P., Stern, E., & Sundelius, B. (2005). The Politics of Crisis Management: Public Leadership under Pressure, Cambridge University Press, New York. Brabham, D. C. (2013). ‘The Four Urban Governance Problem Types Suitable for Crowdsourcing Citizen Participation’, In C. N. Silva (Ed.), Citizen E-Participation in Urban Governance: Crowdsourcing and Collaborative Creativity, IGI Global, Hershey, PA. pp. 5068. Bruns, A., Burgess, J., Crawford, K., & Shaw, F. (2012). #qldfloods and @QPSMedia: Crisis Communication on Twitter in the 2011 South East Queensland Floods. http://www.cci.edu.au/floodsreport.pdf. Kelvin Grove, QLD (Australia): (accessed 5 May 2014). Caragea, C., McNeese, N., Jaiswal, A., Traylor, G., Kim, H.-W., Mitra, P., Wu, D., Tapia, A., Giles, L., Jansen, B. (2011). Classifying text messages for the Haiti earthquake. Paper presented at the Proceedings of the 8th International Conference on Information Systems for Crisis Response and Management (ISCRAM2011), (8-11 May 2011). Centers for Disease Control (CDC). (2014). Social Media: Zombie Apocalypse, http://emergency.cdc.gov/socialmedia/zombies.asp (accessed 10 March 2016). Chatfield, A. T., Scholl, H. J., & Brajawidagda, U. (2013). ‘Tsunami early warnings via Twitter in government: Net-savvy citizens' co-production of time-critical public information services’, Government Information Quarterly, Volume 30, Number 4, pp. 377-386. Comfort, L. K. (2005). ‘Risk, Security, and Disaster Management’, Annual Review of Political Science, Volume 8, pp. 335-356. Comfort, L. K. (2007). ‘Crisis Management in Hindsight: Cognition, Communication, Coordination, and Control’, Public Administration Review, Volume 67, Number 1, pp. 189197. 28

Crowe, A. (2012). Disasters 2.0: The application of social media systems for modern emergency management, CRC Press, Boca Raton. Deterding, S., Dixon, D., Khaled, R., & Nacke, L. (2011). From game design elements to gamefulness: defining "gamification". Paper presented at the Proceedings of the 15th International Academic MindTrek Conference: Envisioning Future Media Environments, Tampere, Finland. Dufty, N. (2012). ‘Using social media to build community disaster resilience’, The Australian Journal of Emergency Management, Volume 27, Number 1, pp. 40-45. Endsley, M. R. (1995). ‘Toward a Theory of Situation Awareness in Dynamic Systems’, Human Factors: The Journal of the Human Factors and Ergonomics Society, Volume 37, Number 1, pp. 32-64. Falkheimer, J., & Heide, M. (2006). ‘Multicultural Crisis Communication: Towards a Social Constructionist Perspective’, Journal of Contingencies and Crisis Management, Volume 14, Number 4, pp. 180-189. Federal Emergency Management Agency (FEMA). (2013). ‘Social Media in Emergency Management’, Washington, DC: Federal Emergency Management Agency, https://emilms.fema.gov/is42/index.htm (accessed 14 January 2014). Helsloot, I., & Groenendaal, J. (2013). ‘Twitter: An Underutilized Potential during Sudden Crises?’ Journal of Contingencies and Crisis Management, Volume 21, Number 3, pp. 178183. Heverin, T., & Zach, L. (2012). ‘Use of microblogging for collective sense-making during violent crises: A study of three campus shootings’, Journal of the American Society for Information Science and Technology, Volume 63, Number 1, pp. 34-47. Hughes, A. L., & Palen, L. (2012). ‘The Evolving Role of the Public Information Officer: An Examination of Social Media in Emergency Management’, Journal of Homeland Security & Emergency Management, Volume 9, Number 1, pp. 1-20. Hughes, A. L., St. Denis, L. A., Palen, L., & Anderson, K. M. (2014). Online Public Communications by Police & Fire Services during the 2012 Hurricane Sandy. Paper presented at the 2014 International Conference on Human Factors in Computing Systems (CHI 2014), New York, NY. Hunt, S., Smith, K., Hamerton, H., & Sargisson, R. J. (2014). ‘An Incident Control Centre in Action: Response to the Rena Oil Spill in New Zealand’, Journal of Contingencies and Crisis Management, Volume 22, Number 1, pp. 63-66. Imran, M., Elbassuoni, S., Castillo, C., Diaz, F., & Meier, P. (2013). Extracting Information Nuggets from Disaster-Related Messages in Social Media. Paper presented at the 10th International ISCRAM Conference, Baden-Baden, Germany.

29

Jung, K., & Park, H. W. (2016). ‘Tracing Interorganizational Information Networks during Emergency Response Period: A Webometric Approach to the 2012 Gumi Chemical Spill in South Korea’, Government Information Quarterly, Volume 33, Number 1, pp. 133-141. Lindell, M. K., & Perry, R. W. (2012). ‘The Protective Action Decision Model: Theoretical Modifications and Additional Evidence’, Risk Analysis, Volume 32, Number 4, pp. 616-632. Lindsay, B. R. (2011). Social Media and Disasters: Current Uses, Future Options, and Policy Considerations, Washington, DC, https://www.fas.org/sgp/crs/homesec/R41987.pdf (accessed 20 September 2013). Meier, P. (2015). ‘Social Media for Disaster Response – Done Right!’ http://irevolution.net/ category/crowdsourcing/ (accessed 10 December 2015). Meijer, A. J. (2014). ‘New Media and the Coproduction of Safety: An Empirical Analysis of Dutch Practices’, The American Review of Public Administration, Volume 44, Number 1, pp. 17-34. Mergel, I. (2012), Social Media in the Public Sector: A Guide to Participation, Collaboration, and Transparency in the Networked World, Jossey-Bass/Wiley, San Francisco. Mergel, I. (2014), Social Media Practices in Local Emergency Management: Results from Central New York, http://sotechem.syr.edu/wp-content/uploads/2014/11/SoTechEM_ Executive_Report-mwedit111914.pdf (accessed 20 October 2015). Mergel, I., & Bretschneider, S. I. (2013), ‘A Three-Stage Adoption Process for Social Media Use in Government’, Public Administration Review, Volume 73, Number 3, pp. 390-400. Mileti, D. S. (1999), Disasters by Design: A Reassessment of Natural Hazards in the United States, Joseph Henry Press, Washington, D.C. Mileti, D. S., & Sorensen, J. H. (1990), Communication of emergency public warnings: A social science perspective and state-of-the-art assessment, http://www.cires.org.mx/ docs_info/CIRES_003.pdf (accessed 9 September 2015). Olteanu, A., Vieweg, S., & Castillo, C. (2015). What to Expect When the Unexpected Happens: Social Media Communications across Crises. Paper presented at the Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing, Vancouver, BC, Canada. Quarantelli, E. L. (1988). ‘Disaster Crisis Management: A Summary of Research Findings’, Journal of Management Studies, Volume 25, Number 4, pp. 373-385. Reddick, C. G., & Norris, D. F. (2013). ‘Social media adoption at the American grass roots: Web 2.0 or 1.5?’ Government Information Quarterly, Volume 30, Number 4, pp. 498-507. Reuter, C., Ludwig, T., Kaufhold, M.-A., & Spielhofer, T. (2016) ‘Emergency Services’ Attitudes towards Social Media: A Quantitative and Qualitative Survey across Europe.’ 30

International Journal of Human-Computer Studies. Reynolds, B., & Seeger, M. W. (2005). ‘Crisis and Emergency Risk Communication as an Integrative Model.’ Journal of Health Communication, Volume 10, Number 1, pp. 43-55. Ripberger, J. T., Jenkins-Smith, H. C., Silva, C. L., Carlson, D. E., & Henderson, M. (2014). ‘Social Media and Severe Weather: Do Tweets Provide a Valid Indicator of Public Attention to Severe Weather Risk Communication?’ Weather, Climate, and Society, Volume 6, Number 4, pp. 520–530. Rosenthal, U., Boin, A., & Comfort, L. K. (Eds.) (2001), Managing Crises: Threats, Dilemmas, Opportunities, Charles C Thomas, Springfield, IL Sellnow, T. L., & Seeger, M. W. (2013), Theorizing crisis communication. John Wiley & Sons. Sievers, J. A. (2015). ‘Embracing Crowdsourcing: A Strategy for State and Local Governments Approaching “Whole Community” Emergency Planning.’ State and Local Government Review, Volume 47, Number 1, pp. 57-67. St. Denis, L. A., Hughes, A. L., & Palen, L. (2012), Trial by Fire: The Deployment of Trusted Digital Volunteers in the 2011 Shadow Lake Fire, Paper presented at the 9th International ISCRAM Conference Vancouver, BC, Canada. St. Denis, L. A., Palen, L., & Anderson, K. M. (2014), Mastering Social Media: An Analysis of Jefferson County’s Communications during the 2013 Colorado Floods, Paper presented at the Information Systems for Crisis Response and Management Conference (ISCRAM 2014), State College, PA. Starbird, K., Maddock, J., Orand, M., Achterman, P., & Mason, R. M. (2014), Rumors, false flags, and digital vigilantes: misinformation on twitter after the 2013 Boston marathon bombing, Paper presented at the iConference 2014, Newport Beach, CA. Su, Y. S., Wardell III, C., & Thorkildsen, Z. (2013), Social Media in the Emergency Management Field: 2012 Survey Results, CNA. Arlington, VA Sutton, J., Spiro, E., Butts, C., Fitzhugh, S., Johnson, B., & Greczek, M. (2013a), ‘Tweeting the Spill: Online Informal Communications, Social Networks, and Conversational Microstructures during the Deepwater Horizon Oilspill’, International Journal of Information Systems for Crisis Response and Management (IJISCRAM), Volume 5, Number 1, pp. 58-76. Sutton, J., Johnson, B., Spiro, E., & Butts, C. (2013b), Tweeting What Matters: Information, Advisories, and Alerts Following the Boston Marathon Events. Online Research Highlight, http://heroicproject.org. (accessed 10 June 2015). Sutton, J., Spiro, E. S., Johnson, B., Fitzhugh, S., Gibson, B., & Butts, C. T. (2014), ‘Warning Tweets: Serial Transmission of Messages during the Warning Phase of a Disaster Event’, Information, Communication & Society, Volume 17, Number 6, pp. 765-787. 31

U.S. Department of Homeland Security. (2013). ‘National Response Framework: ESF#15 SOP, Annex R, R-5’, DHS, Washington, DC http://www.fema.gov/media-librarydata/965d87d8c5ffc4bcccb01979913e01fc/ESF15_SOP_08-30-2013-02.pdf (accessed 2 February 2015). Veil, S. R., Buehner, T., & Palenchar, M. J. (2011), ‘A Work-In-Process Literature Review: Incorporating Social Media in Risk and Crisis Communication’, Journal of Contingencies and Crisis Management, Volume 19, Number 2, pp. 110-122. White, C. (2012), Social Media, Crisis Communication, and Emergency Management: Leveraging Web 2.0 Technologies. CRC Press, Boca Raton. Wukich, C. (2015), ‘Social Media Use in Emergency Management’, Journal of Emergency Management, Volume 13, Number 4, pp. 281-295. Wukich, C., & Khemka, A. (2016), ‘Social Media Adoption, Message Content, and Reach: An Examination of Red Cross and Red Crescent National Societies’, International Journal of Emergency Management. Wukich, C., & Mergel, I. (2015), ‘Closing the Citizen-Government Communication Gap: Content, Audience, and Network Analysis of Government Tweets’, Journal of Homeland Security & Emergency Management, Volume 12, Number 3, pp. 707-735. Zavattaro, S. M., & Sementelli, A. J. (2014), A Critical Examination of Social Media Adoption in Government: Introducing Omnipresence’, Government Information Quarterly, Volume 31, Number 2, pp. 257-264.

32

Table 1. Message Types for Emergency Managers Type

Definition

Example

Previous Studies

Warnings & Watches

Alerts and indications of danger

AlabamaEMA: Tornado Warning Fayette and Lamar Issued: April 11 at 1:08PM CDT Expiring: April 11 at 2:15PM CDT Urgency: Immediate

Chatfield et al., 2013; FEMA, 2013; Sutton et al., 2014

Advisories

Guidance on actions to take in the lead up to, during, and after incident

AR_Emergencies: With active flash flood warnings across the state please remember to Turn Around, Don't Drown!

Bruns et al., 2012; Imran et al. 2013; Sutton et al., 2014

Evacuations

Information regarding evacuation orders, routes, and related efforts

COEmergency: @wigginsfire: The town of Orchard is in immediate danger. EVACUATE NOW.

Situational Information

Reports on the location, nature, and changing conditions of an incident

ReadyWisconsin: The heavy snow continues in northwestern Wisconsin... http://t.co/E5VSCIkFqy

Hazard Impact

Information on damage, casualties, and other consequences of an incident

SCEMD: Utilities reporting approx 3700 power outages statewide, #scwx #sctweets

Closures

Reports on the closure of facilities and other infrastructure

IDHS: I-65 is closed again between US 30 and IN-26 as road conditions are deteriorating. Drivers should avoid this area and find another route.

Openings

Reports on the opening of facilities and other infrastructure

SDemergencyMgmt: A downed power line has been cleared from I-29 south of Sioux Falls, and the interstate is open again in that area, @SDHighwayPatrol says.

White, 2012; Sutton et al., 2014

Descriptions of agency operations (e.g., personnel in the office and in the field) and executive actions taken (e.g., emergency declarations and preparedness proclamations); Can include descriptions of other agencies' actions

KansasEmergency: Kansas sends help to Oklahoma http://t.co/U1TChsGkYr

Hughes et al., 2014

Information regarding the provision of resources

AlaskaDHSEM: Disaster Unemployment Assistance is available. Deadline is 7/29. http://t.co/J2eeIKtTFC #AK13Flood

Lindsay, 2011; Bennett, 2014; Hughes et al., 2014

Corrections and/or clarifications on misinformation and rumors originating from an external source

ReadyNJ: Rumor Control: Despite media reports, NO tornado touchdown in Butler, Morris Co. last night. Info confirmed by NWS. Possible microbust.

Veil et al., 2011; Bird et al., 2012; St. Denis et al., 2014; Wukich & Mergel, 2015

Corrections and/or clarifications on agencyoriginated information

KYEMPIO: Clarification for those receiving the short version of prev message. Fatalities reported from tornado in MS - not KY.

Veil et al., 2011; Wukich & Mergel, 2015

Operations

Resource Provision Rumor Management Corrections

33

White, 2012; Sutton et al., 2014 Bruns et al., 2012; Hughes et al., 2014; St. Denis et al., 2014 Hughes et al., 2014; Imran et al., 2013; Olteanu et al., 2015 Sutton et al., 2014; St. Denis et al., 2014; Hughes et al., 2014

Intelligence Gathering

Requests for information to increase situational awareness and improve effectiveness during response and recovery operations

WyOHS: Did any of you feel the 4.9 earthquake 12 miles west of Fort Washakie? It happened just a little after 7 a.m.... http://t.co/AnELc0BmsT.

Atkinson & Wald, 2007; Veil et al., 2011; Chatfield et al., 2014; Wukich & Mergel, 2015

Volunteer Recruitment & Coordination

Volunteer recruitment and coordination efforts

ReadyIllinois: Volunteers needed for sandbagging on Sunday and Monday in Nunda Township in Crystal Lake http://t.co/G1yWI555eP

Crowe, 2012; Hughes et al., 2014; Wukich & Mergel, 2015

Financial donation requests

AlabamaEMA: Best way to help tornado victims is to donate to the Red Cross at http://t.co/fM4PUpn8rP or text REDCROSS to 90999. #okwx

Bruns et al., 2012; Crowe, 2012; Sutton et al., 2013; Hughes et al., 2014; Wukich & Mergel, 2015

Resource Seeking & Coordination

Requests for resources other than volunteers and funds

FLSERT: To donate food items @RedCross recommends contacting your local food bank. For clothing/household items visit http://t.co/xch04WBlPo

Crowe, 2012; Hughes et al., 2014; Wukich & Mergel, 2015

Education

Promotion of risk reduction activities (i.e., prevention, mitigation, and preparedness advice)

MichEMHS: Create a family emergency plan so that you know what to do, where to go & how to communicate with your family during a disaster. #NATLPREP

Training & Exercises

Promotion, description, and/or coordination of trainings, conferences, and exercises

vemvt: Social media for emergency response training hosted by DEMHS #ndptc http://t.co/KnQPvulJ8B

Drills

Promotion and/or coordination of community preparedness drills

MoStormAware: Will You Drop, Cover, Hold On? Video spreads the word about #MO's ShakeOut earthquake drill on Oct. 17 at 10:17 AM: http://t.co/N3I4SZ0VMS

Wukich & Mergel, 2015

Conversation Starters

Requests for information/dialogue with constituents regarding risk reduction not during response or recovery operations

MDMEMA: How prepared are you for the next disaster? Tell us all about it by sharing your story at [email protected]. #MDPrepares.

Wukich & Mergel, 2015

Gamification

Promotion of risk awareness and reduction efforts through games and contests

AzEIN: Have you submitted a #KitCookOff recipe yet? Share your culinary creativity & get rewarded for it @ http://t.co/r0Vje282uB #NatlPrep #smem

Wukich, 2015; Wukich & Mergel, 2015

Administrative News

Promotion of administrative activities and accomplishments such as announcing awards won, accreditations received, request for proposal (RFP) solicitations, job postings, staff promotions, and retirements

SCEMD: SCEMD wins a national award for our #SC #Earthquake Guide! 2013 Blue Pencil & Gold Screen Award from #nagc2013! http://t.co/xTagXTwfY8

Crowe, 2013; Wukich, 2015

Expression of support, well-wishes, gratitude, sympathy, frustration, and/or other emotions; Appraisal of actions and inaction

FLSERT: Excited to have the British Consulate at the State EOC today!

Heverin & Zach, 2012; Hughes et al., 2014; Olteanu et al., 2015

Fundraising

Opinion & Commentary

34

Crowe, 2012; FEMA, 2013; Wukich & Mergel, 2015 Wukich, 2015

Table 2. Messages by Hazard Type and Disaster Phase

Natural Technological Terrorism Health & Safety Crime All-Hazards Preplanned Events NA Total

Prevention N % 59 5.7 77 7.4 21 2 519 50 358 34.5 0 0 4 0.4 0 0 1038 2.7

Mitigation N % 291 62.9 23 5 0 0 36 7.8 20 4.3 93 20.1 0 0 0 0 463 1.2

Preparedness N % 4541 39.3 161 1.4 110 1 354 3.1 475 4.1 5808 50.3 89 0.8 4 0 11542 29.9

Response N % 19034 89.2 888 4.2 255 1.2 374 1.8 171 0.8 397 1.9 209 1 0 0 21328 55.2

35

Recovery N % 2313 88.4 68 2.6 27 1 63 2.4 88 3.4 53 2 2 0.1 2 0.1 2616 6.8

NA N 0 0 0 1 1 0 0 1643 1645

Total % 0 0 0 0.1 0.1 0 0 99.9 4.3

N 26238 1217 413 1347 1113 6351 304 1649 38632

% 67.9 3.2 1.1 3.5 2.9 16.4 0.8 4.3 100

Table 3. Message Types by Disaster Phase Total* FDD** Prevention Mitigation Preparedness Response Recovery Type N % N % N % N % N % N % N % Situational Information 9450 24.5 1021 10.8 1 65 0.7 24 0.3 8983 95.1 377 4.0 3.3 964 Education 9191 23.8 299 10.5 191 2.1 7918 86.1 22 0.2 67 0.7 Advisories 6076 15.7 923 15.2 2 5 0.1 5683 93.5 386 6.4 16.2 Operations 4337 11.2 703 36 0.8 128 3 1157 26.7 2253 51.9 610 14.1 5.9 Warnings & Watches 4021 10.4 236 - 3980 99.0 41 1.0 16.9 Hazard Impact 3696 9.6 625 - 3324 89.9 372 10.1 Resource Provision 1872 4.8 512 27.4 15 0.8 33 1.8 186 9.9 907 48.5 718 38.4 9.6 Commentary 1659 4.3 159 11 0.7 11 0.7 492 29.7 567 34.2 145 8.7 Training & Exercises 40 2.7 1466 3.8 9 0.6 39 2.7 1389 94.7 13 0.9 Closures 935 2.4 266 28.4 1 0.1 5 0.5 859 91.9 70 7.5 3.5 Drills 653 1.7 23 638 97.7 15 2.3 6.3 Conversation Starters 624 1.6 39 5 0.8 14 2.2 472 75.6 90 14.4 17 2.7 Administration News 9 1.8 500 1.3 94 18.8 216 43.2 30 6.0 28.1 Openings 409 1.1 115 277 67.7 132 32.3 1.9 Gamification 359 0.9 7 4 1.1 2 0.6 353 98.3 18.8 Volunteer Recruitment 255 0.7 48 3 1.2 3 1.2 126 49.4 57 22.4 51 20.0 Evacuations 136 55.3 246 0.6 238 96.7 8 3.3 Intelligence Gathering 32 14.8 216 0.6 192 88.9 24 11.1 28.7 Fundraising 167 0.4 48 10 6 98 58.7 57 34.1 Corrections 114 0.3 21 18.4 2 1.8 22 19.3 72 63.2 10 8.8 Rumor Management 27 32.5 83 0.2 1 1.2 10 12 55 66.3 16 19.3 Resource Seeking 77 0.2 22 28.6 11 14.3 34 44.2 24 31.2 *Percentage of all tweets (38,682). **FDD = tweets disseminated by states during federally declared disasters. Percent calculated as percentage of message type.

36

NA N

%

29 0 153 13 433 16 26 160 15 2 8 1 8

0.3 0 3.5 0.7 26.1 1.1 4.2 32 5.9 1.2 7 1.2 10.4

Table 4. Measures of Centrality and Dispersion

Type Situational Information Education Advisories Operations Warnings & Watches Hazard Impact Resource Provision Commentary Training & Exercises Closures Drills Conversation Starters Administration News Openings Gamification Volunteer Recruitment Evacuations Intelligence Gathering Fundraising Corrections Rumor Management Resource Seeking

Organizations N % 48 100.0 48 100.0 48 100.0 48 100.0 47 97.9 47 97.9 48 100.0 46 95.8 47 97.9 42 87.5 40 83.3 44 91.7 29 60.4 41 85.4 36 75.0 42 87.5 28 58.3 38 79.2 39 81.3 29 60.4 22 45.8 28 58.3

Mean 196.9 191.5 126.6 90.4 83.8 77 39 34.6 30.5 19.5 13.6 13 10.4 8.5 7.5 5.3 5.1 4.5 3.5 2.4 1.7 1.6

i

Median 129.5 72 66 50.5 35.5 59.5 19.5 16 21.5 10 4 6 3 4 2 3 1 2.5 2 1 0 1

Max 1301 922 574 780 1024 530 519 468 199 143 179 91 264 112 61 52 142 30 38 21 18 32

Min 2 2 2 5 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

SD 255.5 245.2 153.4 123 167.5 91.7 76 69.2 39.4 28 27.8 18.1 37.7 17.6 12.2 8.6 20.9 5.5 5.9 4.5 3.7 4.6

Idaho and Pennsylvania did not possess official Twitter accounts during the period of observation. Colorado, Delaware, and Missouri possessed two official accounts each. Google and Twitter searches were conducted as well to ensure no other official accounts existed.

37