Predictions in Public: understanding the design, communication and dissemination of fire spread prediction maps to the public. Work Package 8: Focus Groups

Research in Australia has been conducted on the public’s response to risk and warning communication (Dootson et al. 2019, 2021). However, less research effort has focused exclusively on maps, and even less has focused on fire spread prediction maps and the Australian context. The purpose of the research reported here is to explore how community members comprehend, use, perceive, and intend to act upon fire spread prediction maps, in bushfire events. Building on previous work in the Predictions in Public research project, including interviews with communities impacted by bushfires and their use of maps (Work Package 4), a nationwide survey of maps currently used in Australia, across all States and Territories (Work Package 5), research that co-designed evidence-informed map concepts for fire spread prediction maps (Work Package 7), and a nationwide survey examining comprehension of those fire spread prediction maps (Work Package 9), a series of focus groups were conducted to explore participants’ underlying perceptions of risk, uncertainty, their comprehension of the maps, and what actions they would intend to take.

Six focus groups were run across three States in peri urban and regional areas of each State: Western Australia (Roleystone), South Australia (Mount Barker), and Queensland (Esk), with a total of 34 participants. Each focus group ran for 90 minutes, was audio recorded, and was professionally transcribed. The moderator’s guide used in each focus group was divided into four parts. In Part One, moderators provided a welcome, introduction, and background about the funded project. Part Two involved testing the first set of stimuli, including a video introducing the fire spread prediction map and the map itself. The first video and map seen by the focus group participants was based on a simulated bushfire scenario in the location where the focus group was being held. This constituted the ‘familiar’ map. Participants were then asked questions about their comprehension of the familiar map, attitudes, risk perceptions, emotions, protective action intentions, and expectations and preferences. Part Three replicated the process and questions of Part Two but used a video and map of a simulated bushfire scenario in another location. This constituted the ‘unfamiliar’ map. Part Four involved researchers providing an explanation of next steps with the data collected.

The set of stimuli used in the focus groups was co-designed by the research team and the project’s Steering Committee members. The video involved a local public information officer or agency staff member reading a co-designed script to camera, introducing the fire spread prediction map, and contextualising the bushfire scenario. This replicated what might be seen in a press conference or what might be disseminated via social media. The information provided on the fire spread prediction maps used in the focus groups varied slightly depending on the operational context in which the scenario was located. Two of the fire spread prediction maps tested overlaying warning areas onto the predicted fire spread, while one map displayed only the predicted fire spread. The predictions on all the maps involved predictions over three time frames (e.g., predictions for 0-3 hours, 3-6 hours, and 6-9 hours for South Australia map and predictions for 0-6 hours, 6-12 hours, and 12-24 hours for the Queensland and Western Australia maps). The script and maps used in the study are provided in Appendices 2-5.

The sample comprised participants over 18 years old, of which, 44% were female, and 56% were living in an area on the map used in the focus groups, with the remaining 44% residing in a nearby area. Sixty-two percent of the sample were employed (versus homemaker, student, unemployed, or retired). At various points during the focus groups, participants described having direct or vicarious experience with bushfires. A summary of the findings is provided below.

Attention. Participants attention was first drawn to the use of the colour red on the map, leading to some initial inaccurate interpretations of red as a representation of where the fire is currently burning. Attention was also paid to (in no particular order): the threat of a bushfire (i.e., its scale, the danger, the imminency of it), the evacuation routes and centres and thoughts about how to navigate away from the affected area, the presence or absence of road closure information that could support navigation away from affected areas, the locations where participants resided if the map was local to them, and the temporal elements of the map and bushfire scenario (i.e., time of map issuance, time to impact for three fire spread predictions). The map element that first drew their attention typically anchored how they interpreted the rest of the map, for example, the colour red inaccurately representing an incident area. As some participants missed noticing the fire front and/or the burnt area on the maps, participants raised the importance of better highlighting where the fire has been, and where it is now, to support community interpretation of the predicted fire spread depicted on the map.

Comprehension. Broader understanding of the intent of the map first required participants to comprehend numerous elements on the map. Elements such as evacuation routes, evacuation centres, and base map elements like roads, suburbs, rivers, and vegetation, were typically understood across the focus groups. Road closures were only understood in one map that labelled them ‘outbound traffic only’. The technical terms used for ‘incident area’ and ‘fire incident’ were not as easily understood as ‘burnt area’ or ‘active fire front’. Technical elements of the base map such as topography and other geographical features were not universally understood by the participants on familiar and unfamiliar maps, which had a secondary impact on their interpretation of the shape and direction of the predicted fire spread, for example, does it make sense the fire would move in this direction based on the hills present/absent on the map. Finally, the warning areas and fire spread predictions were not always seen as distinct map elements with separate purposes. While there was understanding of what the warning areas were communicating as participants were familiar with them from past bushfires, sometimes this familiarity was incorrectly transposed to the predicted fire paths. For example, sometimes the predictions were referred to as “essentially a Watch and Act area”. Some participants understood the intent of layering the warning areas over the predictions, while others believed it was too much information, making the map hard to understand.

Uncertainty. When exploring whether the participants believed the fire would spread as predicted, participants acknowledged that the modelling used the best available science, was prepared by experts in the fire agencies, and in the absence of conflicting information, the model was trusted. However, the certainty that the fire would spread as modelled on the map was dependent entirely on the participants’ understanding of how inherently unpredictable the weather can be, especially changes in the wind. Wind was a critical component to assessing the certainty of the predicted fire spread. The spread of the fire during the initial time frame depicted in each map (e.g., prediction for 0-3 hours/0-6 hours) was perceived as more likely to occur as modelled, than the predicted spread in the second- and third-time frames (e.g., predictions for 3-6 hours/6-12 hours, 6-9 hours/12-24 hours). The difference in certainty of the fire moving as predicted across those three time frames was because participants believed there was less time for changing weather conditions in that initial prediction, especially the wind, to alter the trajectory of the fire. There was more uncertainty around whether the spread would occur as predicted during the second- and third-time frames because of greater potential for weather (and wind) changes. Participants expected there would be updates as weather conditions evolved and subsequently changed the shape and direction of the fire’s spread during the second- and third-time frames. Further, the more specific the prediction appeared to be (e.g., more jagged lines, peninsula shapes), the more accurate and certain the prediction was perceived to be.

Risk perceptions. Explorations of how much risk people felt based on where the research team had located participants on the map surfaced a complex suite of insights. Risk perceptions were very personal and shaped by multiple combinations of their experience of bushfire previously, their understanding of bushfire behaviour, their confidence in map reading, their location on the map, whether the map was of a familiar area, and some design elements of the map. Cumulatively, these factors shaped perceptions of uncertainty, which led to a perception of risk. For example, where there was higher perceived certainty of the fire progressing as predicted, there were also higher levels of perceived risk. All participants acknowledged that there was a risk of bushfire across all the maps they viewed, however, the difference was in the degree of risk, which was typically related to time frames. Where there was a shorter time frame to predicted impact, there was a higher level of perceived risk. In these shorter time frame scenarios, the question was about when the fire would reach the participants. Participants raised the same question (i.e., when the fire would arrive) when they were located on the map inside the fire’s predicted spread. Whereas slightly longer time frames typically led to lower levels of perceived risk as the question became not when the fire would reach them but if the fire would reach them, before weather conditions changed the direction of the fire or the response teams would get the fire under control. These same insights were held when participants were located on the map at the edge or just outside of the predicted fire spread with a question of if the fire would reach them, weakening risk perceptions for some participants.  

Protective action. For maps that overlayed the warning areas on the fire spread predictions, intended actions discussed by participants included those aligned with the ‘Watch and Act’ and ‘Emergency Warning’ warning levels as outlined by the Australian Warnings System. However, when viewing a map without the warning areas, participants indicated taking actions similar to those under a ‘Watch and Act’ warning. This reflected some misunderstanding of the intent of the fire spread prediction map, which is arguably to trigger further information seeking. When the focus group moderator pointed to different locations on the map for participants to comment on, participants were more likely to evacuate when the risk was more imminent, which typically meant when they were in a ‘red’ area (i.e. the first predicted fire spread or Emergency Warning area overlayed on the first predicted fire spread). Some participants were more likely to be preparing (to defend or evacuate) when they were located inside the predicted spread of the fire during the third time frame or just outside it. While others, at the same locations, were waiting. When exploring what would motivate those waiting to then act, some were waiting for environmental cues (e.g., smoke, flames), others were waiting for an update to the map, and some were waiting for a direct engagement with emergency services. When participants viewed a map of a location unfamiliar to them, they were more likely to state that they would prepare to leave or just leave straight away, with fewer caveats than if they were located on a map familiar to them. The unfamiliar location typically meant participants felt little reason to stay and believed they should mobilise faster to get out of the way of locals who may need support during the bushfire.

Map expectations and use. There were mixed intentions to seek out and use a fire spread prediction map, which typically reflected participants’ existing use of incident warning maps. Some found the fire spread prediction maps more helpful than the incident warning maps (i.e., maps that only showed the warning areas), as they preferred the temporal nature of the map, using time to guide their preparation and evacuation/defence. Participants all expected to find these maps on official agency channels while acknowledging that digital access cannot be the only way to access it, and TV or radio stations had an obligation to share the information for certain members of the community. All expected the map to be updated frequently and as the situation changes. None expected to receive the map at a community meeting, citing the situation modelled on the map as being too late in an event to hold a community meeting[1]. The participants indicated the value of having someone explain the map when it is released to the public (e.g., agency representative explaining it on TV or social media channels). There were also calls for making the map dynamic, with the ability to see wind and toggle between information that is relevant to them to meet their individual informational needs.

The results from this study will be combined with the other work packages in the Predictions in Public research project to cumulatively underpin the future design of maps for use in the public information and warnings milieu in Australia, under the Australian Warning System. This report should not be read in isolation to other work packages in the research project.

[1] Also, no one indicated they would attend a community meeting during a fire, but they might do so in preparation season. However, in a previous piece of work (Work Package 4), interviewees indicated they were interested in attending community meetings.