Non-Expert Understanding of Hazard Maps: An Eye-Tracking Study

Maps are the most commonly used means of visualizing and communicating natural hazard information to support decisions about risk mitigation. They are a product of hazard assessment studies which involve different input parameters with uncertainties relevant to decision making. This process is further complicated by the subjective uncertainties that arise in the audience when confronted with the visualization of hazard information. 

Natural hazard maps are primarily designed to be used by experts, but they are also used in their unaltered form to communicate with non-experts, many of whom are unfamiliar with the map’s scientific background and implications. Previous studies focus mainly on evaluating such maps with expert groups (e.g., directly involved stakeholders and authorities), with less attention on non-experts (e.g., the public audiences) who are confronted with these maps before purchasing a house, getting insurance or making other critical decisions. 

To address this gap, our study investigates how well hazard maps are understood and interpreted by non-expert audiences. We tested two earthquake hazard maps of Germany that differ in color palettes (rainbow vs. colorblind-friendly and perception-optimized yellow-orange-red-brown color palettes) and data classification schemes (algorithmic Fisher vs. quasi-logarithmic classification schemes). We showed both maps to 20 non-expert participants during the 2024 Science & Innovation Days (a public engagement event) in Tübingen, Germany. Participants answered map-reading and hazard perception questions (e.g., participants were asked to read off the hazard level for a given city, and to compare hazard levels between for a pair of cities) while their eye movements were monitored with eye-tracking software. 

To identify if either map improved map reading and hazard perception, participants’ responses were scored, analyzed and compared using a two-sample Mann–Whitney U and Fisher’s Exact tests. In general, the differences detected in participants’ responses were not statistically significant, perhaps due to the small sample size. Still, we observed that nearly all participants who used the redesigned map (8 out of 9) correctly read the hazard level for a city while only 33% (3 out of 9 participants) who used the rainbow color map responded correctly.

Eye-tracking data were used to analyze focus-metrics. Composite heatmaps accumulating the duration of eye fixations of all participants indicate that their eye movements were focused more on the high hazard zones and the corresponding values shown on map legend when answering questions using a hazard map redesigned to use best practices for hazard perception.

To quantify these differences, the ratio of fixations on high-hazard zones to total fixations on the map were calculated for both map versions. The data were tested for normality and the statistical significance of the differences were evaluated using Independent Samples t-tests for equal variances. While the results were not statistically significant, participants viewing the redesigned map showed a greater number of fixations on high-hazard zones compared to the participants viewing the original map, with a moderate effect size. We note tendencies in the data that encourage the repetition of the experiment with a larger sample size.