Predictions in public: using qualitative data to understand the design, communication and dissemination of predictive maps to the public

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 in an Australian context. The purpose of this research is to assess the extent to which community members use, comprehend, perceive, and act upon maps, including incident and fire spread prediction maps in bushfires.

In collaboration with the project’s Steering Committee, we identified three locations across Australia that recently experienced a bushfire event and were willing to work with us on this project. These locations were: Cardinia Shire, Victoria (the 2019 Bunyip Complex fires); Southern ACT and Snowy Monaro Regional Council, New South Wales (NSW) (2019-2020 Black Summer fires); and the Huon Valley Council, Tasmania (2019 Riveaux Road fire). Across the three locations, we interviewed 94 participants using a semi-structured interview technique. The interviews explored community residents’ use, comprehension, perceptions, and actions in response to bushfire maps, including fire spread prediction maps, in two scenarios: 1) during their previous experiences in actual fire events; and 2) when shown two types of maps and associated warning messages relevant to their state/location during the interview. All recorded interviews were transcribed, and we used qualitative data analysis methods to organise and analyse the interview data (i.e., the transcripts).

Part 1: Previous Experience with Maps

In the first part of the interview, residents were asked about a recent experience with a bushfire, including the types of information they received and what they did in response to that information and how bushfire maps played a role in their decision-making (if at all).  See Appendix A for the pre-interview questionnaire and Appendix C for the interview guide.  The findings of Part 1 are summarised, below.

Across all three locations, people were first made aware of the fire event via a number of means, including both physical and social cues. However, first awareness was complicated by the many fires to which they were exposed over the course of that fire season.

Throughout the event, residents used different types of maps on a fairly frequent basis, including maps from:

• Local fire agencies (accessed via the agency’s app or website),
• The Bureau of Meteorology,
• Google, and
• Third-party weather or hazard mapping platforms, like the Windy app, DEA hotspots map, and bushfire.io.

People checked maps more often during certain periods, including when the fire hadn’t spread to them yet (for early information) or during the event, e.g., when the fire was moving faster. Many participants used maps on a frequent basis, even mentioning using them between 20 to 50 times per day. However, at some point, constant checking like this could restrict people from doing other necessary actions and therefore, may be reduced by clearly communicating when map updates are likely to occur.

Maps were only one tool in their information ‘toolbox’, since they often used maps in combination with other information sources provided via different sources and channels. One such example were community meetings, where residents had an opportunity to view the maps and listen to fire agency experts explain the maps in more detail.

Participants used maps for different purposes. These included:

• To self-localise, or identify where they were in relation to the fire event;
• Gather information about the fire event (i.e., fire and weather conditions) and what to do next;
• Monitor the extent or rate of spread using the burnt area;
• Cross-reference map information with other sources;
• Confirm or explain the physical cues that they were seeing around them (however, residents also consulted trusted sources to confirm what they saw on the map);
• Make judgements about fire spread predictions and risk levels;
• Inform or warn others who may be at risk; and
• Monitor the impact of the fire on their or others’ properties (e.g., after they evacuated).

Participants encountered several challenges with bushfire maps during their experiences. These included issues with:

• Timely information (e.g., maps that were slow to update or didn’t change for longer periods of time);
• Missing information;
• Inconsistent information across sources, platforms, and geographical boundaries;
• Inaccurate information; and
• Inaccessibility of map information (either due to internet/coverage, device, usability, or comprehension issues).

Those who constantly checked the maps were searching for updates, noting that they understood how fast fires can move and therefore, wanted to receive regularly updated information. A lack of timely information can leave participants confused about the state of the event, and often, when lacking information, respondents looked for additional information themselves. 

Also, inconsistency of information across multiple map sources, platforms and geographical borders can create confusion and decisions to be postponed. In these instances, some maps provided different information (and sometimes more up-to-date information) than other map sources or platforms, making it difficult to understand which one was more accurate. Where participants could decipher which source or platform was timelier than others, they lost confidence in the maps displaying old or potentially inaccurate information.

A number of improvements were suggested by residents. Among these, participants expressed a strong interest in fire spread predictions and their associated confidence intervals being communicated in future bushfire maps (i.e., how the fire might spread over time). People also wanted information on:

• The current locations of the fire, including the fire front;
• Wind speed and direction;
• The type of fire and its intensity levels;
• The burnt areas and their spatial accuracy;
• Road closures; and
• The time of the last update and/or expiry time for the map and how often the maps would update.

Residents responded in different ways to protect themselves and their loved ones, based on the information provided to them, including maps. Those who evacuated mentioned the one cue or piece of information that prompted them to go. In addition to being told to evacuate by emergency officials and members of their social network, people were also prompted to leave by witnessing intense physical cues (e.g., heavy smoke or ember showers) and receiving or inferring information about the bushfire’s movement (likely provided by maps).

Part 2: Example maps used in Victoria, NSW/ACT, and Tasmania

In Part 2 of the interview, participants examined two different types of bushfire maps. The interviewer asked them to describe what they saw; what captured their attention; what they thought the map was telling them; which areas were at the highest risk of harm and what people in those high-risk locations should do; and how useful the map would be in a bushfire event, whether they had confidence in the information provided, and what additional information or improvements they would suggest for the map. Appendix B shows the maps that were used in the study for each of the three locations, with the accompanying warning information. The findings from Part 2 are summarised, below.

Regarding map awareness, participants were mostly familiar with the maps shown to them during this part of the interview. Additionally, various types of map features captured their attention first, including warning polygons or red shaded areas and the grey or blackened areas. Respondents also interpreted the maps’ symbols and features in different ways across all maps and locations. Comprehension issues included confusion about the triangle symbology (e.g., NSW/ACT Map 1), the meaning of the triangle location (across all study areas), the risks associated with grey/burnt areas (e.g., NSW/ACT and Victoria Map 2), the meaning of the warning polygons (e.g., Tasmanian maps), and the risks across the entire ‘fire spread’ or ‘impact zone’ areas (e.g., Victoria Map 2).

In terms of risk assessments, none of the maps facilitated participants in unanimously identifying one location of highest risk. Additionally, some respondents even noted that the maps did not provide them with sufficient information to make that assessment. Overall, the types of areas identified as risky often included warning areas or polygons (with the highest risk level), areas of potential fire spread, and areas with the highest number of fires in one location. Also, based on their risk assessments but regardless of map type, participants often identified a similar set of behavioural responses for those located in the risky areas: 1) evacuating the area if it was safe to do so and routes were available, 2) staying and defending, or 3) in lieu of a specific response, following the actions specified in their household bushfire plan.

Participants stated the reasons that they would use maps (regardless of map type – i.e., incident or prediction map). These included the following:

• to identify risk levels and/or where the risk was located (including where they were in relation to those locations, where possible),

• to make decisions about protective actions (e.g., whether to stay/go or to avoid the area if located elsewhere), and
• to identify possible routes out of the area and the safe areas to travel.

For fire spread prediction maps in particular, participants also stated that they would use them to monitor the fire spread over time (i.e., NSW/ACT Map 2) or were uncertain how they would use them since they covered such a large area and contained less local detail (i.e., Victoria Map 2). Some residents noted that they could use the prediction maps in conjunction with the local-area incident maps to make decisions on how to protect themselves. They also highlighted the importance of the base map information for context, and in particular, the ability for map users to locate themselves and understand their risk or the risk to others based on the mapped fire event.

In terms of feedback on the maps, participants expressed both higher and lower levels of confidence in specific maps based on a number of criteria. For example, residents linked higher levels of confidence with higher levels of trust in the map source and a higher understanding of the inherent uncertainty in bushfire maps products. On the other hand, participants expressed lower levels of confidence when they perceived the map information as out-dated, the maps lacked sufficient information or detail, information on the map was confusing, or the maps were too general in scale (and did not provide localised information). Finally, regardless of location, participants recognised the need to confirm the map’s message with other sources of information before they could trust the information. This need for confirmation falls in line with warning research in hazards and disasters.

Across all maps, participants wanted more information. The two main types of information that residents requested were: wayfinding and navigation information and information on environmental conditions. Within wayfinding and navigation, participants requested the following:

• Information to assist them with self-localisation (e.g., city/town names, names of landmarks, topographic information, parks, road names, etc.),
• Traffic information and road closures,
• Evacuation options and safe refuge locations.

Information on environmental conditions included:

• Fire size/scale, intensity, activity (including burnt areas), location, spread and direction;
• Weather forecasts;
• Emergency response information.

While additional information can provide clarity, participants also requested map features that would improve use and comprehension, including a scale bar, compass, legend, increased resolution, and clear colouring. Colouring on maps can become problematic if the colours are too close in shade to other colours, they interfere with (or are indistinguishable from) other areas on the map, and/or are inconsistent with the other colour/rating systems for bushfires and other emergencies.

However, the inclusion of new information in future bushfire maps may be dependent on the map type and purpose. Wayfinding information like route status, evacuation options and safe areas for refuge might be better suited on a more localised incident map (e.g., Victoria Map 1) rather than a larger-scale, state-wide prediction map (e.g., NSW/ACT Map 2). The same can be said for emergency response information. It will be important for residents to understand the purpose of the landscape-wide prediction maps (i.e., to communicate risk on a broader scale), and to look to other mapping platforms (e.g., state agency hazard incident maps) for localised information to make appropriate protective action decisions.

Considering when and how to include this new information is important, since participants were overwhelmingly observed to infer the missing information themselves when it wasn’t provided to them. Participants made inferences about future fire spread predictions in order to understand their fire risk, including the direction of fire spread; fire locations, including the location of the fire front; weather or wind conditions; and the scale of the event, the extent of damage and the resources required. In some cases, these assumptions may not be accurate. Overall, when desired information is not provided to participants, they will look for it elsewhere. In such cases, they will fill in the gaps with their own knowledge and/or experience, which may not always lead to accurate conclusions.

Our findings have important implications for fire map design. In particular, for fire spread prediction maps, these findings highlight the importance of ensuring that map readers can understand their location in relation to the risk (i.e., self-localise). Additionally, participants provided a number of suggestions on ways to communicate the direction of fire spread in simple ways and include important information on the fire front and locations of fire activity (to assist them in understanding the risk and making the best protective action decisions for themselves and their loved ones). Participants also provided suggestions on different ways that fire spread prediction maps could clearly communicate risk and uncertainty, to be tested in Phase 2 of this project via laboratory experiments, surveys and focus groups.