This abstract describes the development of a “Hurricane Risk Calculator” to provide actionable information about potential hurricane impacts at a user’s specific location. A significant goal is to contextualize the potential risks and translate these into easily understandable forms to guide effective evacuation decisions and optimize the timing of other protective actions. To this end, we established a “research collective” comprised of experts spanning the disciplines of meteorology, numerical modeling, verification, structural engineering, cloud computing, user design, social science, and human vulnerability. We aim to develop a framework that intersects real-time probabilistic predictions of hurricane hazards with probabilistic descriptions of vulnerability. The probabilistic risk obtained can then be used to guide risk-informed decisions through comparison of the risks of various alternatives.
We focus first on risks posed by hurricane winds. While wind hazard only accounts for 14% of historical direct hurricane deaths in the U.S., threats posed by extreme winds can still be a powerful motivating factor in evacuation decisions (optimal or not). As an example, 2.5 million people participated in a traffic-choked evacuation in advance of Hurricane Rita (2005). The ensuing chaos resulted in at least 80 deaths, more than the direct deaths from the storm. Many evacuated because they did not know how their houses would perform in the anticipated major hurricane winds. The risk calculator for wind risk will be driven by a new probabilistic prediction framework that accounts for flow-dependent uncertainty by leveraging solutions of global ensemble prediction systems. By generating a very large ensemble of synthetic storm scenarios, the uncertainty inherent in the interrelated predictions of track, intensity, and size can be simulated. Then, point-wise probabilities of exceedance can be computed for any wind speed thresholds. The next step involves intersecting the predicted probabilistic winds with the structure-specific vulnerability to compute risk. At a basic level, this can be accomplished by considering the class and age of the structure and the design wind speed that it was built to. Better yet would be to describe the building’s vulnerability through probabilistic fragility curves or a component failure analysis. In that case, probabilistic risk can be obtained, allowing estimation of the likelihood that the structure will lose its ability to protect the life and safety of its occupants. This also allows calculation of the range of damage that may be expected.
We are developing a user-friendly, public-facing version of the tool, in which a user will be able to enter their street address into a web page or mobile app and then view a dashboard-like interface with graphical and textual products that translate the potential wind impacts for the specific structure at that address. To better inform assessment of structural vulnerability, the user may enter building-specific information via a short questionnaire. To keep the tool simple and understandable, its output will use a three-color categorical rating system that conveys the potential safety of the structure during the storm, as well as the habitability of the structure after the storm:
- “Green tag condition likely”: no significant structural damage expected; low risk to life and safety; structure is expected to be habitable following the storm; short-term interruptions of utilities possible.
- “Yellow tag condition likely”: some structural damage possible; some loss of contents likely; elevated risk of injury and possible death to occupants during the storm; structure may be uninhabitable following the storm due to water damage, mold, or loss of utility services; extended loss of power is likely.
- “Red tag condition likely”: significant damage likely up to a total loss of the structure and its contents; significant risk of injury and/or death to occupants during the storm; structure expected to be uninhabitable following the storm due to extensive damage; long-term outages of utility services are expected.
As a next step, we will analyze the accuracy of the probabilistic risk framework by comparing retrospectively-predicted damage states to actual damage states obtained by on-the-ground assessments for several recent high-impact hurricane landfalls.
Future work will refine the framework, extend it to additional hazards such as storm surge, and estimate utility restoration times. Social science studies are needed to understand how these new forms of information affect people’s decision-making process for evacuations and other protective actions and to determine the best ways to communicate the risk outputs of the tool. Other work is needed to improve the modeling of wind over land, including fetch-dependent influences of land use, terrain, and the environment, as well as to incorporate the effect of terrain speed-up.
This research collective is open to all researchers and practitioners interested in contributing. We especially welcome participation from the emergency management community, the forecaster community, and industry. We envision that this research will lead to a broad range of tools and applications, which, when coupled with next-generation emergency management practices, will better enable the most at-risk populations to take protective actions that enhance life safety, while allowing those at low risk to remain in place. Achieving these goals will substantially enhance our nation’s readiness, responsiveness, and resilience in the face of hurricane threats.