Non-expert interpretations of hurricane forecast uncertainty visualizations

Abstract

Uncertainty represented in visualizations is often ignored or misunderstood by the non-expert user. The National Hurricane Center displays hurricane forecasts using a track forecast cone, depicting the expected track of the storm and the uncertainty in the forecast. Our goal was to test whether different graphical displays of a hurricane forecast containing uncertainty would influence a decision about storm characteristics. Participants viewed one of five different visualization types. Three varied the currently used forecast cone, one presented a track with no uncertainty, and one presented an ensemble of multiple possible hurricane tracks. Results show that individuals make different decisions using uncertainty visualizations with different visual properties, demonstrating that basic visual properties must be considered in visualization design and communication.

Keywords::

• decision making
• hurricane prediction
• uncertainty
• visualization

Notes

^1. Timepoint was dummy coded such that the 24-hour timepoint was coded as 0 and 48-hour timepoint was coded as 1. In addition, visualization type was dummy coded such that the cone-centerline was the reference group. For distance, given the relatively large distance between oil rig locations (average 63.76 km and 118.07 km for the 24-hour and 48-hour timepoints, respectively), we divided distance by 10 prior to analysis so that the coefficient would correspond to a 10 km change (rather than a 1 km change). The analysis collapsed over the six hurricane forecasts.We used the following model:Model reliability estimate = .992, σ² = 1.06667, τ = 0.87893

One of the hurricane stimuli used for the study was later found to begin at a slightly different apex for the ensemble visualization. The model was re-run excluding all trials with this hurricane from each condition (df = 23783), however results did not differ from the analysis including all trials. Analysis including all trials is reported.

^2. Although choosing a different reference group for statistical comparison would change the specific results of the planned contrasts, the overall take-home message of the analyses would not change.

^3. Two additional participants were excluded from these analyses due to experimenter error in data collection, so that these and the analyses of questionnaire data were based on 198 participants.