Recent improvements in weather observation and monitoring have increased the precision of tornado warnings. The National Weather Service currently issues storm-based tornado warnings, and even more geographically specific warnings that include probability information are under development. At the same time, the widespread proliferation of smartphone and mobile computing technology supports the rapid dissemination of graphical weather warning information. Some broadcasters and private companies have already begun using probabilistic-style tornado warning graphics. However, the development of these new types of warnings has occurred with limited research on how users interpret probabilistic visualizations. This study begins filling this void by examining responses to color scheme and relative position using probabilistic tornado warning designs. A survey of university students is used to measure the level of perceived fear and likelihood of protective action for a series of hypothetical warning scenarios. Central research questions investigate 1) differences in responses across warning designs, 2) clustering of extreme responses in each design, 3) trends in responses with respect to probability levels, 4) differences in responses inside versus outside the warnings, and 5) differences in responses near the edges of the warning designs. Results suggest a variety of trade-offs in viewer responses to tornado warnings based on visual design choices. These findings underscore the need for more comprehensive research on visualizations in weather hazard communication that can aid meteorologists in effectively warning the public and spur appropriate tornado protection behaviors in a timely manner.

No abstract

No abstract

Recent advancements in severe weather detection and warning dissemination technologies have reduced, but not eliminated, large-casualty tornado hazards in the United States. Research on warning cognition and behavioral response by the public has the potential to further reduce tornado-related deaths and injuries; however, less research has been conducted in this area compared to tornado research in the physical sciences. Extant research in this vein tends to bifurcate. One branch of studies derives from classic risk perception, which investigates cognitive, affective, and sociocultural factors in relation to concern and preparation for uncertain risks. Another branch focuses on psychological, social, and cultural factors implicated in warning response for rapid onset hazards, with attention paid to previous experience and message design. Few studies link risk perceptions with cognition and response as elicited by specific examples of warnings. The present study unites risk perception, cognition, and response approaches by testing the contributions of hypothesized warning response drivers in one set of path models. Warning response is approximated by perceived fear and intended protective action as reported by survey respondents when exposed to hypothetical tornado warning scenarios. This study considers the roles of hazard knowledge acquisition, information-seeking behaviors, previous experience, and sociodemographic factors while controlling for the effects of the visual warning graphic. Findings from the study indicate the primacy of a user's visual interpretation of a warning graphic in shaping tornado warning response. Results also suggest that information-seeking habits, previous tornado experience, and local disaster culture play strong influencing roles in warning response.

Hurricane Sandy's 80 mph wind speeds did not achieve major hurricane status on the Saffir-Simpson scale, yet the storm had extreme consequences for the New York metropolitan area. Post-event recovery has been quite variable across the region, especially in New Jersey. This paper examines the progression of recovery at two time intervals — 6 months and 1 year after the storm made landfall. Based on in situ observations of reconstruction and repair within the surge damage zone along New Jersey's coast (N = 765 locations or points), we hypothesize that the timing and distribution of recovery is based on exposure (water depth, prior flood experience, building heights), vulnerability (social vulnerability, seasonal occupancy), and government support (FEMA assistance, Small Business Administration loans). Predictions of recovery (recovered, not recovered) based on these input measures is quite good at 6 months (80 percent correctly predicted) with slight improvements at 1 year (82 percent correctly predicted). The model is much better at predicting recovered points than recovering. In terms of the individual factors predicting recovery the level of exposure is the most significant, statewide. Within the two most affect counties — Monmouth and Ocean — only water depth and building height were significant predictors. Housing tenure (seasonal occupancy) was an important predictor of the location and timing of recovery for the first 6 months, especially in Monmouth County, but became insignificant thereafter. Government support in the form of FEMA individual assistance grants was a significant predictor statewide in both time periods. Clearly other factors are influencing the pace and location of recovery including confusion surrounding flood hazard zones and insurance. Those factors influencing the lack of recovery for the roughly 20 percent of our observations awaits further, more detailed analyses.