User Interpretation of Hurricane Forecast Graphics

In summer 2012, forty participants were tested on their ability to understand a graphical hurricane forecast depicting a fictitious hurricane that was forecast to make landfall along the United States East Coast. The graphic combined potential for damaging winds and a cone of error on a single image. That sample consisted of 20 members of the local public, 9 experts, and 11 students. This initial testing revealed that even users with meteorological training did not always interpret the graphic correctly. In summer 2015, additional testing began with similar forecast graphics and an expanded set of questions. In this case, the cone of error was presented first along with four understanding questions. Next, participants were shown one of three possible forecast graphics. The first option displayed potential for damaging winds and a current storm location. The second option displayed potential for damaging winds and forecasted locations for the next 5 days. A final option presented potential for damaging winds, the cone of error, and forecasted locations for the next 5 days. Participants responded to three understanding questions and two questions about perceived risk associated with the hurricane, and made three judgments regarding a hypothetical event during the forecast period. For each graphic, participants were asked to rate its helpfulness. In a final section, participants responded to six questions regarding their hurricane and hurricane-safety knowledge, their demographics and their experience with weather information.

Preliminary results represent a sample with a high level of weather experience, but do hint at some differences. Early analysis indicated that perceived “helpfulness” tended to increase with additional information provided. A map with just the potential for damaging winds displayed was perceived as least helpful, followed by the option with potential for damaging winds, and forecast track points but no cone. The most helpful image was the graphic that showed all the information. Participants' judgments about whether a friend should hold an event during the forecast period also appeared to be related to whether or not they viewed a graphic with forecasted locations. Early responses indicate a high level of accuracy whether based on forecast graphics or in hurricane and hurricane-safety questions. This was anticipated given the sample's high level of weather experience. Data collection is on-going and further analysis will be conducted as the sample expands to a more general population.