Improving Evacuation Strategies through Adaptive Planning

Principal Investigator: 
Performance Period: 
July 2014 to June 2015
Commercialization Status: 
N/A
Project Keywords: 
Risk management
emergency planning
Abstract: 
Traditional approaches to emergency planning and evacuation often rely on predefined emergency-planning zones that are specified in advance of the disaster (e.g., based on jurisdictional boundaries). However, with modern technological advancements (including faster computational tools, more accurate dispersion models, geographic information systems, and mobile communications), there is no longer a compelling reason for this approach, since dispersion models can be run in real time, the results displayed on geographic information systems, and evacuation instructions relayed to emergency responders in the field (or even to those citizens who have devices with global-positioning capabilities). A recent doctoral dissertation by Hammond (2013) has demonstrated that such adaptive approaches to evacuation perform better on average than the use of predefined evacuation zones at reducing either the total dose to which citizens are exposed in a disaster, the size of the population to be evacuated, or both. Moreover, adaptive evacuation zones were shown to be especially effective for larger disasters and at high-population sites, both of which make emergency response more important.   Note that experience with nuclear disasters in both the Soviet Union and Japan has shown that the societal disruption from large-scale evacuations can be severe. Morris (2013) described the phenomenon of “relocation trauma” after Chernobyl: “Of the old people who relocated, one Chernobyl medical technician…said: ‘Quite simply, they die of anguish’.” Moreover, evacuation can create significant health risks in and of itself, especially for vulnerable populations (such as the elderly).  For example, 90 people died in Texas in the evacuation in advance of Hurricane Rita, due to factors such as hyperthermia and/or chronic health conditions. Similarly, after the nuclear accident in Japan, dozens of people who had been evacuated (some of whom had been relocated multiple times) died for reasons unrelated to radiation exposure.  Finally, evacuations during a 2003 forest fire in California resulted in five fatalities due to heart attacks caused by the stress of the event. Thus, reducing the size of the population to be evacuated (in the short term) or relocated (in the longer term) is a non-trivial concern.   However, knowledge of public evacuation behavior is critical to estimating the benefits of reduced evacuations, since compliance with evacuation orders will never be perfect. In fact, past studies have estimated the rate of compliance with official directives for nuclear-plant disasters in the U.S. to range from approximately 17% to 82% (Hammond 2013). In particular, “shadow evacuation” by people not ordered to evacuate could significantly reduce the benefits of reduced evacuation zones. For example, after the accident at Three Mile Island, Zeigler et al. (1981) argued that the governor’s directive affecting people living within 10 miles of the plant “seemed to establish the critical evacuation boundary in the minds of area residents.” If people living near nuclear plants have internalized the 10-mile emergency-planning zone as “normative,” smaller evacuations may be perceived as inherently unsafe. Flynn (1979) also indicated that females (especially pregnant females) and households with small children were more likely to evacuate, again potentially leading to large shadow evacuations. More recently, Dotson and Jones (2004) studied 50 evacuations between 1990 and 2003, and found that both shadow evacuations and refusal to evacuate diminished the effectiveness of organized evacuation efforts. Nishino et al. (2012) likewise distinguish between “voluntary” and “involuntary” evacuations after Fukushima, and note that in some areas far from the power plant, up to half of all evacuees may have evacuated voluntarily, before an evacuation order for that jurisdiction had been issued. Thus, estimates of the benefits of alternative evacuation strategies must contend with the fact that evacuation behavior cannot be perfectly predicted or controlled, but only influenced by clear evacuation messaging.   This project extends the work of Hammond (2013). In particular, we quantify the extent of societal disruption that can be expected in response to accidents at various nuclear-power plants around the U.S., identify tradeoffs between health effects and population evacuated (since large evacuations can keep health effects to a minimum, but increase societal disruption), and explore the potential of non-mandatory shadow evacuations to diminish benefits of adaptive planning zones.  We also explored the use of adaptive evacuation planning to non-radiological disasters (e.g., chemical and biological attacks).   Our analysis of radiological emergencies began with a literature review of past nuclear disasters, addressing four main topics: health effects; non-health concerns; evacuation; and political reactions.  The main conclusion of that literature review is that the magnitude of the accidents in the former Soviet Union and Japan (including the attendant societal disruption) makes clear the value of preventing such accidents, and minimizing their overall societal consequences.  The analysis of tradeoffs between health effects and population evacuated involves comparing the number of people relocated and the number of latent cancer fatalities under different accident scenarios, weather conditions, and evacuation thresholds at five different nuclear plants.  The tradeoff between number of people evacuated and number of lives saved for different dose thresholds is remarkably stable, providing a basis for determining whether the dose threshold for relocation should be revised.  Disutility functions were developed to help evaluate that tradeoff taking into account risk aversion for large relocation events, and also possible interaction effects between size of relocation and number of lives saved (e.g., if a large relocation is considered especially undesirable if the number of cancer fatalities will also be large).  These disutility functions are informative in evaluating different possible dose thresholds for relocation.