___________________

¹ More information about the flooding impacts of Hurricane Hillary in Death Valley, California, is available at https://www.nps.gov/deva/learn/nature/hilary.htm.

extreme events is another challenge, she explained. Berg described two challenges that stand in tension with one another: the problem of “noise” (i.e., many messages from multiple sources such as social media and traditional media) and how noise can overwhelm people and overfocus the message on information less relevant to staying safe (e.g., landfall time, possibility of breaking records) while obscuring messaging from the NHC and other institutions about the risks that such storms pose. Lamers added that, although media often focus on factors such as atmospheric pressure and windspeed, the deadliest hazard in the United States in the past decade has been rainfall-induced flooding. Lamers and Schoenfeld both spoke about the challenge of relaying the hazards of intense rain events when the public might perceive these events to be just “bad weather.”

Schumacher’s second question to the group introduced a theme that surfaced repeatedly throughout the workshop: the importance of strong relationships between various partners. In their responses to Schumacher’s question of how decision makers, including emergency managers (EMs), responded to warnings from the various institutions represented, each panelist emphasized the benefits of having strong long-term, working relationships between their national institution and local EMs and other local decision-makers. Berg noted that as weather changes, relationships between the NHC and certain locations, particularly concerning unusual types of events, may need to strengthen: “We don’t see tropical cyclones hitting Southern California that frequently, so we don’t have that pre-established relationship with many of those emergency managers on tropical cyclones themselves.” Schoenfeld shared the hope that partnerships around “more routine” hazards would extend in the case of unprecedented events. Lamers noted that this extension seemed to be successful in the case of Hurricane Hilary, which was rare and costly but resulted in no casualties. He added that plans for targeted evacuations in New York City, developed after Hurricanes Henri and Ida, and for evacuations targeting vulnerable populations during Hurricane Hilary in California, revealed that decision-makers have a high level of confidence in the rainfall forecast and other information provided at the national level. Echoing another common theme—improving communication through better understanding of specific recipients’ needs and wants for this information —Lamers said, “I think understanding a little bit more of how they are making these decisions . . . could really help inform us as we try to improve these forecasts over time.” Berg highlighted the importance of local specificity: even though a region might be frequently exposed to a certain type of event, a particular locality within that region may not have direct experience and, therefore, not the same level of familiarity. This specificity makes a difference in how populations understand risk and when and how they take action.

The discussion then turned to novel or innovative communication approaches during rare events. Schoenfeld shared how, in anticipation of Hilary, her office “pushed to get the timeline for products issuance accelerated,” knowing that a tropical storm watch or warning had never been issued for the region. Berg also mentioned that accelerated timing was part of the NHC’s messaging around

Hurricane Hilary, and Lamers described how the Excessive Rainfall Outlook tool can help visualize the extent of risk and influence (i.e., accelerate) the timing of messaging around unprecedented events.² Berg highlighted two other novel approaches at the NHC. The first is “Key Messages,” which are designed and used to highlight essential points about hazards and forecast uncertainty for select tropical cyclones.³ Key Messages enable consistent messaging through local channels. The second is livestreaming on social media and other platforms (e.g., Facebook, NHC YouTube channel).

Schumacher then asked the panel to reflect on what they might have done differently during recent events. Lamers emphasized the importance of talking to partners and taking their feedback seriously. “Rainfall rates” in urban areas, as a specific aspect of vulnerability and exposure, was of profound importance after Hurricanes Henri and Ida, he noted.⁴ Therefore, the National Weather Service is developing an Urban Rain Rate Dashboard, a tool that will involve hydrologic modeling of a city to help decision-makers in large cities understand when rainfall rates are spiking in their areas. Lamers noted that “several dozen cities” will be mapped and modeled initially, with perhaps more to follow.

Schoenfeld described challenges to communicating uncertainty and associated concerns about losing credibility with the public. She noted that in response to feedback from the public, her office now more actively shapes and narrows the focus of warnings as the event unfolds. Berg returned to the concept of noise in messaging and the obstacles to helping people focus on the most relevant hazard in a changing situation with multiple hazards; he described, specifically, how people might hear “hurricane” and miss the danger of hazards such as rainfall, tornadoes, and rip currents, among others.

DISCUSSION

Questions posed by audience members spanned a range of topics. Sara McBride, Research Social Scientist, U.S. Geological Survey (USGS), asked about the usefulness of ARKStorm in risk communication, a worse-case hypothetical storm

___________________

² The Excessive Rainfall Outlook tool is “a graphical map issued by the Weather Prediction Center (WPC) that forecasts the probability that rainfall will exceed flash flood guidance (FFG) within 25 miles (40 kilometers) of a point across the contiguous United States (CONUS)” (National Weather Service, 2023). More information about the tool is available at https://www.weather.gov/media/notification/PDDs/PDD_ERO_Days_4_5_T2O.pdf.

³ The Key Messages graphic on the NHC website and NHC social media pages “includes the text of the Key Messages and relevant tropical cyclone graphics, which can include the cone graphic, the 34-kt cumulative wind speed probability graphic, or a rainfall forecast graphic provided by the Weather Prediction Center.” More information about Key Messages and their associated graphics in the context of Hurricane Ian are available at https://www.nhc.noaa.gov/aboutnhcgraphics.shtml#KEYMESS.

⁴ Rainfall rates “measure the intensity of rain within a certain period of time” (Weather Nation, 2021). More information about rainfall rates is available at https://www.weathernationtv.com/news/understanding-rainfall-rates.

scenario model that is used as a learning tool to provide EMs, the public, and other groups with an assessment of what is historically possible.⁵ Schoenfeld and Lamers both noted that messaging the true worst-case scenario is not always ideal in the realm of public communication, because it can distract and obscure information relevant to particular hazards or locations. However, they noted that worst-case scenario exercises, such as ARKStorm, can be used to develop preparedness protocols, helping EMs and officials to envision what atypical events might look like in their area or to stress-test emergency response systems. Karen Florini, Vice President for Strategic Impact, Climate Central, asked about the way that rapid intensification events affect warnings, such as a tropical depression rapidly and unexpectedly intensifying into a high-category (e.g., 3-5) hurricane. Berg acknowledged that such events are occurring more frequently but that statistics indicate that forecasting of such events is improving.

Questions from the audience then turned to best practices for risk communication in partnership with other countries and continents. The strength and importance of partnerships were again highlighted, as exemplified by the NHC’s coordination with countries in other regions (e.g., the Caribbean). However, Berg stressed that, although the United States can and does make recommendations to other countries, each individual country is responsible for deciding whether and how it issues warnings and other messages. He added that the Caribbean and Central America are the only regions in the world where such coordination and partnership occur between the country responsible for tropical cyclone forecasts (in this case the United States) and other countries in that region. National hurricane centers exist in Japan, Australia, Fiji, and other places, “but they don’t have the framework that we have in our part of the world where they are coordinating with other countries around them quite as closely as we are,” and those centers perhaps should consider strengthening international partnerships “in order to get those risk messages out to other countries that aren’t actually making the forecasts themselves,” acknowledged Lamers. In addition, Lamers noted as a best practice that the NHC in Miami includes all hazards in its official public advisory products, which garner widespread attention. Official advisories from the NHC in Miami include hazards stemming from the potential for rainfall and flooding, storm surge, tornadoes, and wind, in addition to the more standard forecast track and intensity. Such an approach could be adopted by other regional warning centers around the world.

In response to the final question, the panel described how strategies change as lead-time changes.⁶ Schoenfeld emphasized the importance of finding ways

___________________

⁵ ARKStorm was developed by NOAA, USGS, Scripps Institute of Oceanography, State of California, California Geological Survey, University of Colorado, Federal Emergency Management Agency, NCAR, California Department of Water Resources, California Emergency Management Agency, and other organizations. It addresses “massive U.S. West Coast storms analogous to those that devastated California in 1861-62 and with magnitudes projected to become more frequent and intense as a result of climate change.” More information is available at: https://www.usgs.gov/programs/science-application-for-risk-reduction/science/arkstorm-scenario.

⁶ Lead-time is “the difference in time between the onset of an observed event, and the issuance of a forecast that is associated with the observed event” (Lough et al., 2008., p. 5). More information is available at https://gsl.noaa.gov/fiqas/publications/articles/TAF_Leadtime_Metric_FULL_Description.pdf.

to communicate uncertainty, both to the public and in conversation with partners, while Lamers noted that decisions about lead-times should account for the rhythms of daily life—for example, timing a message so that people receive it before embarking on weekend travel.

RESEARCH PERSPECTIVES ON RISK COMMUNICATION

The second panel of the session complemented the first, with researchers sharing their perspectives on the same topic of risk communication around atypical storms. Ann Bostrom moderated this panel.

Roxane Cohen Silver, Distinguished Professor of Psychological Science, University of California, Irvine, spoke first about her work to determine what makes research “ideal” from a methodological point of view, and then on her findings from research using new methodologies on risk perception around evacuation zones. Silver and colleagues conducted a literature review of “decades of research on how people evacuate from natural disasters, whether or not they do, and the characteristics of this body of research” (see Thompson et al., 2017). The research team noted that the high number of conflicting results was due to methodological limitations and deemed “ideal research” to be research that identifies at-risk population samples before an event, takes a longitudinal approach with immediate and then repeated post-event assessments, and uses representative samples of the subject population.

The research team then designed a study based on these ideal characteristics to explore the psychological impact of Hurricane Irma (Garfin et al 2022). The team collected data at two points: (1) about 60 hours before landfall, from about 1,600 people, in a representative sample and (2) 1 month later, in a follow-up assessment, with a representative sample of 1,500 people. Findings showed that about 40 percent of participants did not correctly identify their evacuation zone status, about 6 percent were unaware of evacuation orders, and about 17 percent evacuated unnecessarily (Ibid.). The strongest predictor of whether people evacuated is whether they had evacuated for a previous storm, independent of whether they were in an evacuation zone. Similarly, pre-hurricane risk perceptions were a strong predictor of whether a person evacuated. However, Silver asked, “What predicts the pre-hurricane risk perceptions?” Media exposure and clarity of messaging—particularly visual messaging—were the strongest predictor, said Silver. She highlighted a mixed public message wherein the text instructed readers to evacuate only if they were in an evacuation zone, but the image showed a cone across the entire state.

Silver concluded her remarks with three messages: (1) it is important to make changes to communication practices as well as research; (2) risk perceptions and perceptions of evacuation zone status appear to shift over time, and perceptions of zone status may not be accurate; and (3) partnering with media during the event to craft and update evacuation messages that accurately reflect changing risks is critically important, and debriefing after the storms with both the media and the general public can increase understanding and preparation in anticipation of future storms.

Recalling the reference to noise during an earlier panel, Spiro explained that this noise is, in a sense, her primary research topic: the flow of information within social media environments and what people do with the information they encounter in these environments. In the midst of a crisis event, people “come together to try and make sense of what is going on around them” and to find information that might help them make decisions—a “collective sense-making process,” as she called it. Commonly, rumors emerge in these situations, Spiro explained, defining rumor as “information unverified at the time it’s being talked about.” Rumors can be useful when they bring people together, reduce anxiety, and spur people to take protective action. Challenges arise, however, when the collective sense-making process is strategically manipulated through mis- or dis-information.

Spiro described findings from an ongoing study that she co-leads at the University of Washington that analyzes millions of messages sent on select social media platforms for message content, format, design, and how people engage with information in order to identify and better understand the “rules” that govern information flow. These rules include levels of uncertainty and trust of the social media platform, emotional valence, the actual words, and larger narratives and worldviews within the different communities affected by the event. Spiro stated that the study has helped to identify “the features of different information narratives of rumors that contribute to their virality in online systems” and has shown that people engage with these systems and spread information largely because they want to help each other.

The social media information environment poses widespread challenges for researchers, EMs, and the public, Spiro noted. The system is participatory by nature—people have to participate in order for information to be spread online—and therefore vulnerable to manipulation by bad actors who use participants to spread mis- or disinformation. The emergence of artificial intelligence (AI) and deepfake images increases the complexity of information and vulnerability of the system: “We have challenges around not only the increased quantity and quality of information and a very low barrier to doing that [i.e., creating a false image] but also increased personalization of information, persuasion of information, and other open pathways for these tools to involuntarily produce false or misleading information.” Spiro also noted that these manipulations are not entirely new, referencing an image of a shark swimming down a highway that tends to pop up on social media during a tropical storm or hurricane.

Researchers face challenges as well. Although critical to understanding information flow during a crisis, data on social media use are becoming more difficult to collect as social media platforms increasingly limit access options for researchers, Spiro explained. Fragmentation of the social media environment, with multiple platforms that users move across, also complicates observation of behavior and information flow. Similarly, “multimodal information” that combines audio, video, and text is becoming more common and presents new challenges. Spiro concluded by emphasizing that trust in information and institutions is down, which hampers

research and undermines trust in other people more generally.

Bostrom, as moderator, posed the opening question for the group discussion session. She asked how the experiences of the broadcasters and EMs described during the first panel might connect with their own research agendas. Demuth highlighted the need to establish a common understanding of the term “risk communication,” adding that it could be thought of as a frame that connects different research areas, as well as the research space to forecasting work. This broad concept encompasses not only the messages to the public, she remarked, but also the tools available to forecasters, the challenges they face in predicting and timing, and the ways that researchers might help deepen forecasters’ understanding of information as they make decisions about risk and generate new messages for partners and the public.

In her response, Silver emphasized the need for partnerships, including among EMs and people in the weather community who “help us figure out what kinds of questions we should ask.” Shaping and improving research questions, methodologies, and communication strategies by intentionally and carefully attending to the particularities, preferences, and feedback of the community on the receiving end of research and messaging is important. Silver cited an example from her current research in Lake County, California, a community at a high risk of wildfires. Although 60 percent of the community has suffered wildfire damage since 2015, 60 percent of respondents in the research sample said they did not know their evacuation zone, while 40 percent of those who said they did were mistaken. Her research team is sharing these data with EMs and is exploring other ways to support risk awareness.

Spiro noted the important but challenging work of lending expertise in real time, working with journalists, for example, to ensure high-quality information as events unfold. She noted that such partnerships push researchers in “thinking about how research might have to operate in a very rapidly changing environment.”

Returning to the topic of evacuation zone awareness, Richard Allen, Director, Berkeley Seismology Lab, University of California, Berkeley, asked whether it was possible to develop targeted messages to the cell phones of people in affected evacuation zones, similar to early warning earthquake messages. Silver noted that EMs working to improve evacuation zone awareness have noted information retention as an issue. Another challenge is that the evacuation zone is represented by a complex set of numbers and letters that are unfamiliar to the public. Christina Finch, an audience member, added that evacuation zones are developed at local levels. Therefore, there is neither a national standard for developing the zones nor a “consistently available resource to find all zones.” She wondered whether an individual’s understanding of their evacuation zone and ability to remember this information are influenced by how zones are defined and the types of information and messaging (e.g., maps, apps) that are available in particular jurisdictions. Silver responded that a website was designed during Hurricane Irma to show the zone for any specific address entered, but she could not get useful information from it.

Her research showed that people do not know their zone, but there is a lot of information about how and why this issue exists, and how it can be addressed. Demuth added that her research similarly shows people are incorrect in their perceptions about evacuation zones. However, she also urged the expansion of knowledge around evacuation zones, including a more nuanced understanding of the multiple hazards in a complex weather event in the context of evacuation. “How do we map these impacts with some of these extreme events or these atypical events when we might not have evacuation zones that help us understand people’s actual risk?”

The last question looked to the future, with an audience member asking the panelists to describe the types of new research they might employ to improve risk communication. Spiro noted that one important step is to explore the most effective ways to share information across platforms, while developing better ways to observe how information moves across platforms. She also mentioned the need to better understand how information from official and unofficial sources can be used and the potential benefit of leveraging AI tools. Silver commented on the difficulty and expense of conducting pre-disaster, real-time studies, with the pressures of timing making pre-disaster funding critical, along with the need for quick assessment and approval of projects from the institution’s review board. Demuth said she hopes for more social science research of this kind and echoed Spiro in calling for further development of real-time transfer of knowledge from researchers to the “operational community” in the midst of an event.

This page intentionally left blank.