The Limits of Risk Communication: What Influenced Evacuation and Preparation Behaviors in North Carolina during Florence and Michael?

In less than a month, North Carolina experienced two significant storm events resulting in widespread damage across communities within the coastal plain, including oceanfront, soundside and upriver, inland locations. Hurricane Florence made landfall at Wilmington, NC on September 14, 2018 and Tropical Storm Michael compounded impacts on both those communities impacted by Hurricane Florence as well as additional ones when it swept through the state on October 11, 2018. Hurricane Florence caused catastrophic damage across the coast and inland due primarily to its stalling for several days and delivering extreme amounts of rain both prior to and after landfall. In addition to extended riverine flooding, flash flooding and long-term, widespread power outages throughout impacted areas compounded impacts in these communities. At least 53 deaths have been attributed to Florence and initial property damage assessments are currently estimated to be $17-$22 billion. Environmental impacts, including water quality conditions post-storm, are yet to be fully understood. Significant damages were incurred across poultry and swine Confined Area Feeding Operations including breaching waste lagoons and over 3.4 million poultry and 5500 hogs killed during the event. Less than a month later, Tropical Storm Michael advanced through the central part of North Carolina and then into Central Coast and Outer Banks, bringing yet another round of high winds, precipitation and additional damaging soundside storm surge. Hurricane Florence impacts covered central and southern coastal areas as well as inland portions of North Carolina while Tropical Storm Michael brought additional impacts to a suite of central inland communities as well as to the state’s northern coastal and soundside communities. Unfortunately, many communities impacted from these storms were also impacted during Hurricane Matthew’s significant precipitation and related flash flooding events of October 2016. These communities have yet to fully recover even dating back to flooding of Hurricane Floyd in 1999. They are now addressing recovery challenges from these multiple extreme weather events with limited on-the-ground municipal capacity to meet significant community needs.

In particular with the long lead time in advance of Hurricane Florence, much national attention focused on the decisions of local and state authorities in North Carolina to order evacuations and the subsequent compliance or failure to comply with evacuation orders by North Carolina residents. Governor Roy Cooper reported that over 5,000 rescues were needed because of catastrophic flooding from storm surge in areas like New Bern and Downeast Carteret County, and rainfall ranging from 10-15” in the southeastern Sandhills to 30-50” in the lower Cape Fear Basin. Many in the meteorological community, particularly on social media, speculated that the large number of rescues indicated either that forecasts were not received or believed. However, this information deficit model approach – the concept that more, or more clearly communicated, meteorological information would improve preparation and evacuation compliance, is incomplete and does not account for the numerous additional factors involved beyond the perception of meteorological information in making decisions about evacuation. Additionally, the inland approach of Michael, coupled with a shorter lead time relevant to Florence, in contrast left many in North Carolina reporting they felt caught off guard despite watch and warning products issued in advance by the National Hurricane Center and the local National Weather Service offices. The two storms provide a rapid opportunity to examine the limits of effective risk communication in advance of hurricanes and the how much influence other information, including perceived or actual barriers, has on an individual’s preparedness and evacuation behavior.

Recent work by Meyer et al. (2018) suggests that storm category by wind, response efficacy bias (the belief that a home is safe) and having a job requiring a person to report to work have been factors in perceived intent to evacuate in Louisiana. McCaffrey et al. (2018) use the Protective Action Decision Model for wildfire evacuation and further classify that individuals who are not inclined to evacuate have a strong tolerance for risk and believe they know how to prepare their properties. Additionally, past North Carolina Sea Grant work with Hatteras Island communities indicates that residents feel a strong commitment to serving community in immediate recovery to limit further damage. Other community partners report fearing additional vulnerability to inland flooding during evacuation based on rescue experiences during evacuation in Floyd and Matthew which we hypothesize could be additional factors constraining evacuation behavior in North Carolina. North Carolina Sea Grant is using interviews to engage our stakeholders and collect qualitative data that supports the development of research hypotheses and outreach efforts to identify which of these factors may have influenced preparedness and evacuation behavior in Florence and Michael. This initial work will be critical in designing effective studies to quantify the limits to how much forecast improvements and messaging can further improve preparation and evacuation behavior. In this presentation, we will review the important factors in addition to forecast information and watch/warning dissemination which went into North Carolina residents’ preparation and evacuation decisions. We will discuss how further social science research can be structured to examine how non-meteorological information, including capacity needed to enable highly vulnerable populations to evacuate, can be better incorporated as storms approach to improve evacuation compliance. Such further studies, both qualitative and quantitative, beyond this rapid assessment phase will enable the development of new messaging and capacity needed beyond the communication of meteorological information to facilitate preparation and evacuation behaviors among coastal residents.

McCaffrey, S., R. Wilson, and A. Konar, 2018. Should I stay or should I go now? Or should I wait and see? Influences on wildfire evacuation decisions. Risk Analysis 38: 1390-1404, https://doi.org/10.1111/risa.12944.

Meyer, M.A., B. Mitchell, J.C. Purdum, K. Breen, and R.L. Iles, 2018. Previous hurricane evacuation decision sna dfuture evacuation intentions among residents of southeast Louisiana. International Journal of Disaster Risk Reduction 31: 1231-1244, https://doi.org/10.1016/j.ijdrr.2018.01.003.