- Email: [email protected]
Alternative evacuation strategies for nuclear power accidents
Reliability Engineering and System Safety 135 (2015) 9–14
Contents lists available at ScienceDirect
Reliability Engineering and System Safety journal homepage: www.elsevier.com/locate/ress
Alternative evacuation strategies for nuclear power accidents Gregory D. Hammond n,1, Vicki M. Bier University of Wisconsin-Madison, Department of Industrial and Systems Engineering, 3270A Mechanical Engineering Building, 1513 University Ave., Madison, WI 53706, USA
art ic l e i nf o
a b s t r a c t
Article history: Received 21 March 2014 Received in revised form 4 October 2014 Accepted 12 October 2014 Available online 27 October 2014
In the U.S., current protective-action strategies to safeguard the public following a nuclear power accident have remained largely unchanged since their implementation in the early 1980s. In the past thirty years, new technologies have been introduced, allowing faster computations, better modeling of predicted radiological consequences, and improved accident mapping using geographic information systems (GIS). Utilizing these new technologies, we evaluate the efficacy of alternative strategies, called adaptive protective action zones (APAZs), that use site-specific and event-specific data to dynamically determine evacuation boundaries with simple heuristics in order to better inform protective action decisions (rather than relying on pre-event regulatory bright lines). Several candidate APAZs were developed and then compared to the Nuclear Regulatory Commission’s keyhole evacuation strategy (and full evacuation of the emergency planning zone). Two of the APAZs were better on average than existing NRC strategies at reducing either the radiological exposure, the population evacuated, or both. These APAZs are especially effective for larger radioactive plumes and at high population sites; one of them is better at reducing radiation exposure, while the other is better at reducing the size of the population evacuated. Published by Elsevier Ltd.
Keywords: Nuclear incidents Evacuation Nuclear power plants Emergency preparedness Protective action recommendations
1. Introduction Current strategies used to determine who to evacuate following a nuclear-power plant accident have not changed significantly since the emergency planning guidelines were established in the early 1980s. While plans and studies have been modified and updated, this has been done under the constraint of a roughly constant evacuation area. Consequently, changes to protective actions have focused on issues such as in which order people should be evacuated, or in which direction they should evacuate [1,19]. Yet in the past thirty years, the task has radically changed; new technologies have been introduced, allowing faster computation, better modeling of predicted radiological consequences, improved accident mapping using geographic information systems (GIS), and new means to communicate. Additionally, the populations surrounding nuclear-power plants are denser; more people live closer to reactors than ever before. In the past 30 years, the average population living within
n
Corresponding author. E-mail addresses: gregory.hammond@afit.edu (G.D. Hammond), [email protected] (V.M. Bier). 1 Present address: Air Force Institute of Technology, Department of Systems Engineering and Management, 2950 Hobson Way, Wright-Patterson AFB, OH, 45433, USA. http://dx.doi.org/10.1016/j.ress.2014.10.016 0951-8320/Published by Elsevier Ltd.
16 km of these plants has increased by 62%, from approximately 40,000 to almost 65,000 per site. Furthermore, at 12 of the 65 reactor sites in the U.S., populations have more than doubled [2]. In the wake of the Fukushima Daiichi nuclear accident – considering the range of new capabilities and the greater population at risk – this study sought to reexamine the U.S. nuclear-power plant evacuation strategy by removing the constraint of a constant evacuation area or predetermined evacuation zones. This research is a proof of concept; its purpose was to develop alternative evacuation strategies for use during the early phase of nuclear-power plant accidents in order to take advantage of some of the recent technological advances. The early phase is defined by the U.S. Environmental Protection Agency (EPA) as “the period at the beginning of a nuclear incident when immediate decisions for effective use of protective actions are required and must therefore usually be based primarily on the status of the nuclear facility and the prognosis for worsening conditions” ([25], p. 5). Thus, the early phase is filled with uncertainty. The plant operators and emergency response officials know only that the situation at the reactor is a cause for concern and that an off-site radiological release is possible, so they can only guess at the extent of the problem. Despite this imperfect knowledge, officials must act and make decisions to protect the public from potential radiation exposure, generally in the form of evacuations (because distance is the best
10
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
protection) and/or sheltering in place. In the earliest periods of this phase, decisions are made based on predictions of the radiological release [25]. Notwithstanding the likely discrepancies between early phase projected doses and actual off-site doses that will be observed later, protective action recommendations must be made in the early phase, because evacuation in advance of the plume (ideally at least 1 h before the plume’s passage) is the best way to reduce dose [12,23]. This research explored alternative methods to determine who should be evacuated during the early phase. An ideal strategy would be able to perfectly evacuate the at-risk population before the radioactive plume passes. At present, of course, neither the APAZs developed in this research nor the NRC’s method can achieve this standard; however, as will be shown, the APAZs demonstrate progress towards meeting that standard. Because no method is perfect, the comparison between APAZs and the NRC’s method was framed as a multi-attribute decision problem using the objectives of minimizing the population to be evacuated and maximizing the total radiological dose avoided. These objectives are based on the current regulatory position of the NRC, which focuses evacuation efforts on high-risk areas [27], as well as the EPA guideline that informs the NRC’s policy, which states that evacuation risk should not exceed the risk from the avoided dose [11]. While there is general acceptance that avoiding radiological dose is beneficial, some might argue that instead of minimizing the population evacuated, the goal should be to maximize the size of evacuation. Because distance is a highly effective defense against radiation [12,23], some might argue that if the entire population surrounding a nuclear power plant could be evacuated prior to passage of the plume, that population could be guaranteed safety, suggesting that the objectives of high total dose avoided and high population evacuated would yield a desirable (if conservative) outcome. This logic is flawed, however, because maximizing the population evacuated can expose many people to risks greater than the risk of the radiological release, and ignores the risks and costs of evacuations. Evacuation can have adverse health impacts. Evacuation risks include travel, events in which travel is the contributing cause, and activities other than travel (i.e., preparation or reception activities) [3,30]. Witzig and Weerakkody have estimated travel risk to be 6 10 8 fatalities per vehicle-km; this risk is considered to be an upper bound as the actual risk is expected to be lower than normal automobile travel due to conditions of heavier traffic and lower travel speeds [3,30]. Injuries or fatalities in which travel contributed to their occurrence is the second category of risk. An example is an individual who evacuates the wrong direction and drives into a radioactive plume; it is believed to be an order of magnitude greater than travel risk [30]. The last evacuation risk, estimated to be 5 10 6 per person, is due to evacuation preparations and the arrival at the reception center ([3,30]). These three risks collectively form evacuation risk. For a given emergency, the evacuation risk is a function of the number of individuals that leave. When larger numbers of people evacuate who are not required to evacuate (i.e., shadow evacuations), the collective risk to the population will significantly increase [3]. The EPA evacuation risk estimate (for fatalities) corresponds to Witzig and Weerakkody’s upper bound estimate meaning that for radiation doses of less than 3 mSv, the evacuation risk is greater than the radiation risk [11]. Maximizing the evacuation area relocates many people who would receive doses less than 3 mSv, exposing them to needless risk. As noted by Aumonler and Morrey, “evacuation risks constitute a harm which should be considered in a decision as to whether to evacuate a population put at risk by a radiological incident” ([3], p. 290). For this reason, a safer course of action would evacuate those whose radiation risk is greater than their evacuation risk, but not those whose evacuation risk is greater than their radiation risk.
Minimizing the population evacuated and maximizing the total dose avoided embodies the EPA’s position that the protective actions should not be “higher than justified on the basis of optimization of cost and the collective risk of effects on health” ([11], p. 135). Thus, for this research, a high dose avoided and low population evacuated are assumed to be preferred. Using these two objectives, APAZs were compared to the NRC’s strategy using a concept called the “efficient frontier,” to allow decision makers to evaluate alternatives using their own value systems. Alternatives are plotted on the basis of the decision objectives (i.e., dose avoided and population evacuated). Dominated options can be excluded from consideration; the decision maker can then select a preferred option based on his or her preferences from among the non-dominated strategies on the efficient frontier. Current U.S. response protocols have been previously well documented. The interested reader is encouraged to review these earlier articles for a more in-depth understanding [11,21,23,24,28]. While the regulations that dictate emergency response have been updated (such as the EPA’s Protective Action Guides and Planning Guidance for Radiological Incidents [25] and the NRC’s guidance for protective action strategies [27]), as noted earlier, the fundamental initial evacuation strategy has remained constant. In the event of a nuclear-power plant accident, plant operators would determine evacuation areas using the NRC’s guidance for protective action strategies [27]. Based on the postulated source term and forecast meteorological conditions, a projected radiological plume is calculated and then fit to pre-established evacuation zones; this is the NRC’s keyhole strategy. (This strategy has been criticized because the plume model provides a simplified view of a complex process that may not correlate with the observed plume causing the wrong people to evacuate [23]). This research proposes an alternative method to determine the evacuation area. In this approach, instead of fitting the projected plume to pre-established zones, the evacuation area would be determined by applying a heuristic enlargement strategy directly to the forecast plume.
2. Calculations 2.1. Development and testing of APAZs Candidate APAZs (described subsequently) were tested using weather data from five nuclear power plants: Limerick; Catawba; Turkey Point; Pilgrim; and Arkansas Nuclear One. These plants were selected based on their proximity to National Weather Service (NWS) data-collection sites. The forecast and hindcast weather data (i.e., predicted and observed weather conditions) used in this analysis was produced by the National Oceanic and Atmospheric Agency (NOAA) and the NWS. The source term varied depending on the plant’s output power and reactor type. The source term for each nuclear-power plant was calculated using the time that the core was assumed to be uncovered in an unmitigated short-term station blackout (STSBO) scenario, described in the State-of-the-Art Consequence Analysis (SOARCA) [9]. The total release ranged from 3.3 1018 Bq to 1.2 1019 Bq. These postulated releases are of the same order of magnitude as that from the Fukushima Daiichi accident [26]. Forecast and hindcast plumes used in this research were generated with NRC’s Radiological Assessment System for Consequence AnaLysis (RASCAL) [6]. Nineteen candidate APAZ strategies were tested using 120 weather observations from summer 2012 and winter 2012–2103. Protective action zones, formed by enlarging a forecast plume in accordance with a given heuristic, were compared to the hindcast plumes.
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
The forecast plume generated from RASCAL was used to identify a mandated evacuation zone in which the 4-day total effective dose equivalent (TEDE) was predicted to be greater than or equal to 0.01 Sv, based on established policy [11]. Each of the 19 enlargement strategies was then applied to the mandated evacuation zone. The APAZ enlargement strategies consisted of varied offsets around the forecast evacuation zone (meaning that an additional X km was added to the size of the evacuation zone). Some of them had a uniform offset of varying distances (e.g., 4 km) around the forecast evacuation zone. Other strategies applied an offset in either a perpendicular or lateral direction to the prevailing wind direction. Other strategies applied a radial offset around the reactor in combination with an offset around the evacuation area. Furthermore, some strategies applied a multiplication factor to the predicted TEDE and then applied an offset. Many different strategies were tested in order to represent a wide range of policy options. The candidate APAZ heuristics were evaluated based the goals of minimizing the evacuation area and maximizing the adequacy of protective actions, as follows. First, the forecast plume was determined. It was then enlarged according to the relevant APAZ heuristic to determine the evacuation area. Next, the hindcast plume was determined, as a basis for assessing the adequacy of the APAZ. Area was measured as the ratio of the APAZ area to the hindcast evacuation area. Adequacy of the APAZ was measured by the fraction of the hindcast evacuation area that was also included in the APAZ evacuation area. The average area evacuated and average adequacy were compared for all nineteen heuristics. The comparison indicated that most heuristics fell on an efficient frontier. Results are shown in Fig. 1. The frontier extends from APAZ I (with small evacuation area but low adequacy) to APAZ T (with a large evacuation area and high adequacy). That the different strategies form a coherent efficient frontier indicates that no strategy can perfectly predict the location of the hindcast plume. Instead, the various strategies represent different trade-offs between the competing objectives of a small evacuation area and high adequacy. Three APAZs (B, D, and E) were retained for further comparison, because they are simple to implement and represent a range of policy choices.
Fig. 1. Ratio of APAZ to hindcast area as a function of APAZ adequacy.
11
APAZs B, D, and E were subjected to additional sensitivity analyses. Using a second set of 40 weather predictions and observations, the effect of dispersion model, source term, and evacuation timing were evaluated. The first sensitivity analysis compared our primary dispersion model (RASCAL), which is a 2-D plume model, with the 3-D HYSPLIT transport model [15]. A paired t-test was performed to check for adequacy differences between RASCAL and HYSPLIT. The t-test suggested that APAZ adequacy using the two different models was not statistically distinguishable. This result means that regardless of the hindcast model used, on average, APAZ adequacy will remain roughly the same. The second analysis considered the effect of alternative source terms. The initial screening was conducted using an unmitigated STSBO source term. To ensure that APAZs are robust with respect to the source term, APAZ adequacy and evacuation area were compared for the STSBO and a: long-term station black-out (LTSBO), STSBO with reactor core isolation cooling (RCIC) for boiling water reactors (BWR); and an LTSBO and thermally induced steam generator tube rupture (SGTR) for pressurized water reactors (PWR). For the alternative source terms, the four APAZs remained in their relative locations along the efficient frontier. The third analysis investigated the effect of forecast age. It was used to determine the effect of timing of the evacuation order. When any predictive PAR strategy is used (i.e., APAZs or keyhole), emergency response officials must make a determination of when to generate the forecast plume. Thus, the purpose of this analysis was to determine the effective time horizon for APAZs to understand their reliability in regards to the actual time of release. In the sensitivity analysis, forecast age was allowed to increase from zero hours (as had been used previously), up to 24 h before the release. This analysis showed that neither APAZ adequacy nor its area varied significantly as the forecast age increased. There is some variation, but that variation appears to be due to statistical noise rather than a significant difference. In all of the above analyses, APAZs B, D, and E all remained on the efficient frontier, indicating that they are all reasonably robust candidate evacuation strategies.
2.2. Comparison of APAZs to NRC evacuation strategies This phase of the research compared the three selected APAZ strategies to other protective action strategies. The APAZs included in this comparison were: (1) APAZ B (with a 5-km radial area around the reactor, plus a 2-km offset around the 0.01-Sv boundary); (2) APAZ D (with a 4-km offset around the 0.01-Sv boundary); and (3) APAZ E (with a 6-km offset around the 0.01-Sv boundary). The comparison PAR strategies were: (1) evacuation of the entire EPZ [9]; and (2) the keyhole strategy (utilizing the predefined NRC sectors [10]). The comparison was based on both the total radiological dose avoided (defined as the sum of TEDE for all persons evacuated) and the total population evacuated. Preferred strategies were assumed to be those with a high total dose avoided and a low population evacuated. Six nuclear-power plants were selected for this phase of the analysis: Surry; Peach Bottom; Grand Gulf; Kewaunee; San Onofre; and Braidwood; they are broadly representative of U.S. plants in terms of geographic region, surrounding population, and reactor type (BWR or PWR). We considered eight different weather observations per plant, chosen to include both summer and winter, days with and without precipitation, and differing times of day (i.e., mid-day or evening). Source terms were derived from the SOARCA report: long-term station black-out, STSBO, and STSBO with reactor-core isolation cooling for BWRs; long-term station black-out, STSBO, and steam generator tube rupture for PWRs. In total, 182 different cases were
12
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
analyzed (due to some weather conditions not being observed at some plants during the study period). The evaluation was performed using ArcGIS mapping software. We began with the geographic boundaries of census block groups, as obtained from the U.S. Census Bureau’s TIGER/Line Shapefiles ([7]). The geographic boundaries of the census blocks groups were then overlaid with 2010 census data ([8]) to determine how many people lived in each block group. Next, the hindcast plume (exported from RASCAL as a shapefile) was overlaid on the population data. The information from these two layers (population and hindcast plume) was used to compute the potential radiation dose that would be incurred if any given block group was not evacuated. The forecast plume (also exported from RASCAL as a shapefile) was then used to determine which geographic areas would be evacuated using APAZs B, D, and E, and the keyhole evacuation area. (It should be noted that the plumes were generally complicated in shape, and spanned multiple sectors.) In all cases, the keyhole was taken to be the union of all sectors in which the forecast plume was projected to fall, plus two additional sectors, one on either side of that region. Occasionally, the dose from the forecast plume exceeded 0.01 Sv beyond the outer boundary of the EPZ. In the keyhole strategy, evacuations past the EPZ were limited to areas where dose projections indicated that protective actions would be necessary [9], since NRC strategies are not defined beyond the EPZ. This strategy was implemented by evacuating the relevant keyhole area, plus all RASCAL program grid elements beyond the EPZ with doses exceeding 0.01 Sv. With the population, hindcast dose, and plume location known, for each evacuation strategy, the analysis then generated two statistics of interest: total dose avoided, given in TEDE (person-Sv); and total population evacuated. Total population evacuated was computed as the sum of the populations of all evacuated census blocks; total dose avoided was based on the populations and corresponding hindcast doses of all evacuated census blocks (i.e., all areas with doses greater than 0.01 Sv TEDE). The total dose avoided and population evacuated for each strategy were then compared to the total dose avoided and population evacuated that could have been achieved with perfect advance knowledge of the hindcast plume.
3. Results Fig. 2 below shows that APAZ B achieved comparable or better dose avoidance than could have been achieved by evacuating only the relevant portions of the hindcast plume on 77% of all cases, but evacuated significantly more people than the hindcast plume area in 41% of all cases. (The ideal is to achieve a comparable or better dose avoided in 100% of cases, and over-evacuate in 0% of cases.) We see that APAZs B (0.77, 0.41), D (0.88, 0.59), and E (0.92, 0.82) still form an efficient frontier. The keyhole strategy (0.67, 0.52) is dominated by APAZ B, and evacuation of the entire EPZ (0.88, 0.88) is dominated by APAZ D. Thus, it appears that APAZs are more efficient than keyhole evacuation or evacuation of the entire EPZ. On average, APAZ B is better at both avoiding radiological dose and minimizing over-evacuation than the keyhole strategy. In practical terms, when APAZ B is better than the keyhole strategy, it avoids about two more latent cancer fatalities (LCF) and evacuates about 4000 fewer people. Likewise, APAZ D is better than evacuation of the entire EPZ at minimizing over-evacuation. Compared to the keyhole strategy, APAZ D is better at avoiding dose (achieving approximately five fewer LCF), but evacuates approximately 6000 more people.
Fig. 2. Comparison of results.
Fig. 3. Effect of plume size on performance.
Breaking down the results further, as shown in Fig. 3, we found that for small accidents (i.e., those where the radioactive plume did not extend more than 8 km from the reactor, those where the plume travelled over a large body of water, or those with a population of less than 60,000 near the reactor), APAZs B and D and the keyhole strategy were comparable. However, for large accidents (i.e., those with plumes that traveled more than 8 km from the reactor, or at locations with more than 60,000 people within the EPZ), APAZs B and D were preferred to the keyhole strategy. APAZ B reduced the evacuated population by as much as 23,000 for large accidents (averaged over multiple sites, weather conditions, and source terms), and avoided between three and eight LCF. APAZ D was comparable to the keyhole strategy with regard to the size of the evacuated population, but avoided on average from 6 to 12 additional LCF in large accidents.
4. Discussion As described in the introduction, there is considerable uncertainty during the early phase of emergency response. Discussion is warranted to describe how we addressed this uncertainty.
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
McKenna has characterized the uncertainty in three domains: transport; individual dose response; and source term [23]. Transport consists of variability in both weather patterns throughout the release and the atmospheric transport model. As noted in Section 2.1, both elements were examined in this study. The meteorological effects were addressed by comparing forecast to hindcast plumes, holding other effects constant. Then, through sensitivity analysis, other effects were evaluated to ensure that the chosen heuristic was sufficiently robust to account for different release mechanisms, the timing of evacuation decisions, and RASCAL model effects. (The plume model effects were explored by comparing results from RASCAL with the HYSPLIT 3-D plume model [15]) None of these effects were found to be significant. This research did not directly address dosimetery uncertainty. However, the impact of this uncertainty was minimized by using the EPA assumption [11] of no shielding. As a result, projected doses were likely higher than those that would be expected following an actual release. While not entirely satisfying, this conservatism is consistent with current emergency-response protocols. Furthermore, while the source term uncertainty was not directly modeled, the results provided yield some insight into its implications. Source term uncertainty is a significant challenge; at best it is a factor of 10, but it could be a factor of 100–10,000 [23]. However, source terms were likely over-estimated because RASCAL assumes a near worst case scenario in which the radionuclides leave the reactor core quickly [26]. Moreover, any errors in source terms would apply equally to the APAZs and the NRC strategies. Finally, as was shown in Fig. 3, if the predicted source terms are too large, then we would expect to see little difference between the APAZs and the NRC strategies, because the true plumes will be small. However, if the predicted source terms are too small, then a distinguishable effect would be observed, because the true plume will be larger than thought. The assumptions and sensitivity analyses do not eliminate the inherent uncertainty in source terms, but it is not a threat to the validity of this research. Additionally, this research did not address the means and methods needed to communicate these alternative evacuation strategies with the public. In the event of a nuclear power accident, many residents are likely to ignore the protective action guidelines. Previous studies have demonstrated that compliance to a given directive during nuclear power emergencies is highly dependent on the nature of the directive. Individuals are willing to evacuate, or even shelter in place; however, when asked to shelter in place while others leave (as is the case of selective evacuations such as these strategies), they are much less willing to stay, with approximately 50% to 60% of respondents indicating that they would ignore the directive and leave ([16,18]; [20]). Evacuation compliance is further complicated by the tendency of individuals to not follow prescribed evacuation routes [22] as well as misperceptions regarding one’s location relative to the plant [17]. To overcome these challenges, these strategies would need to be effectively communicated in simple, understandable terms. The application of risk communication to these alternative strategies should be further investigated. The study is also limited because it assumed perfect evacuation in advance of the plume. Using the baseline assumption of 100% evacuation of the target population before passage of the plume, the APAZ strategies can be compared to other strategies without the use of a transportation model. Not using a transportation model simplifies the calculations significantly. While 100% evacuation is likely unrealistic, it allows for a fair comparison. Based on these results the total populations evacuated for the keyhole strategy and APAZ B were roughly equivalent; likewise, evacuation of the entire EPZ was roughly equivalent to APAZ E. If anything, the assumption of 100% evacuation in advance of the plume benefits
13
those strategies (APAZ E and the entire EPZ) that evacuated larger areas, since these large evacuations are not penalized for excessive road congestion, even though in reality individuals may get stuck in traffic jams while attempting to evacuate. Lastly, the findings of this study is limited to the comparison of protection levels offered by different evacuation strategies. Research comparing the protective strategies of sheltering in place and evacuation in the event of nuclear terrorism suggests that sheltering in place followed by a delayed evacuation provides the greatest level of protection during the early response phase [4,5,29]. Additional research by Denning et al. [13] extends this finding to nuclear power accidents. The purpose of this paper was not to compare sheltering in place and evacuation; rather it sought to present alternative evacuation strategies. Future research should consider the impact of these alternative strategies as applied to delayed evacuation and sheltering in place.
5. Conclusion The objective of this research was to identify alternative evacuation strategies as a proof of concept. In this, the research was successful; it identified candidate evacuation methods that perform better than those currently employed. Should these potential gains be considered significant enough by policy makers, future research should further develop the framework into a concept of operations that addresses potential implementation issues and assumed administrative burdens. Indeed, the main significance of this work is not the choice of one APAZ or another, but rather the methodology that was used to test and evaluate alternatives. We found current strategies to be adequate for smaller accidents; however, we found a distinguishable difference for larger accidents. Therefore, policy makers should seriously consider implementing the findings of this research. Authorities that value balancing evacuation and radiological should consider adopting either APAZ B or APAZ D. Likewise, APAZ E is a viable option for decision makers who to prefer to maximize radiological risk avoidance. Such an action would strengthen the overall defensein-depth of nuclear-power plants. Note, however, that altering the current evacuation strategy is likely to have significant costs and consequences. Indeed, such an action would directly affect the emergency response network, requiring evacuation areas at all levels of government to be redrawn. Moreover, the human response to these alternative protocols must also be investigated. Previous studies have also found benefits with alternative strategies [14,20], but the solutions were dismissed due to concerns of impracticality. Accordingly, the practical importance of the improvements achievable using APAZs should most appropriately be seen as a value judgment. In other words, the APAZ approach, which can improve readiness for larger radioactive releases (and also minimize unnecessary evacuations), must be evaluated in the context of its implementation costs and administrative burdens. The NRC, in its statutory role of ensuring the safe use of radioactive material and protection of the population, should assess the importance of the improvements we identified, through further study, as they relate to public safety and policy.
Acknowledgements The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and the Nuclear Regulatory Commission (NRC) for the provision of the RASCAL consequence management software
14
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
used in this publication. They would also like to thank the reviewers for their insightful comments and recommendations. References [1] Agrawal B, Cunningham M, Denning R, Glynn J, Johnson J, Niyogi P, VanderMolen H. Severe accident risks: an assessment for five U.S. Nuclear Power Plants (no. NUREG-1150). U.S. Nuclear Regulatory Commission, Washington, DC, 1990. [2] AP AP: Populations around U.S. nuclear plants soar. USA Today, 27 June, Available at: 〈http://www.usatoday.com/news/nation/2011-06-27-Nuclearplants-population-evacuation_n.htm〉 (last accessed: 18 Mar 14), 2011. [3] Aumonier S, Morrey M. Non-radiological risks of evacuation. J Radiol Prot 1990;10:287–90. [4] Buddemeier BR. Reducing the consequences of a nuclear detonation: recent research. The Bridge. Summer, 28-38. Available at: 〈https://www.nae.edu/ Publications/Bridge/19804/19920.aspx〉; 2010. [5] Buddemeier BR, Valentine JE, Millage KK, Brandt LD. National Capital Region: Key response planning factors for the aftermath of nuclear terrorism. Technical Report LLNL-TR-512111, Lawrence Livermore National Lab, 2011. [6] Brandon L, Athey GF. RASCAL v. 4.1, software. Washington, DC: U.S. Nuclear Regulatory Commission; 2011. [7] Census Bureau US. 2012 TIGER/Line Shapefiles [WWW Document]. URL 〈http://www.census.gov/cgi-bin/geo/shapefiles2012/main〉 (accessed 2 March 13), 2011. [8] Census Bureau US. American Factfinder [WWW Document]. Available at: 〈http://factfinder2.census.gov〉 (accessed 2 March 13), 2013. [9] Chang R, Schaperow J, Ghosh T, Barr J, Tinkler C, Stutzke M. State-of-the-art reactor consequence analyses (SOARCA) report ( no. NUREG-1935). U.S. Nuclear Regulatory Commission, Washington, DC, 2012. [10] Congel F, Kantor F, McKenna T, Mohseni A, Hepler M, Wingert VL. Criteria for preparation and evaluation of radiological emergency response plans and preparedness in support of nuclear power plants: criteria for protective action recommendations for severe accidents (no. Supp. 3 to NUREG-0654/FEMAREP-1, Rev. 1). U.S. Nuclear Regulatory Commission, Washington, DC, 1996. [11] Conklin C, Edwards J. Selection of protective action guides for nuclear incidents. J Hazard Mater 2000;75:131–44. [12] Cutter SL, Barnes K. Evacuation behavior and three mile island. Disasters 1982;6:116–24. [13] Denning R, Osborn D, Brunnett A, Aldemir T. Nov. relative importance of mitigation, early phase, and late phase response. Paper presented at the American Nuclear Society Annual Meeting, embedded topical meeting: risk management for complex socio-technical systems, Washington, DC, 2013.
[14] Dotson LJ, Jones J, Schneck D, Sullivan R. Identification and analysis of factors affecting emergency evacuation (no. NUREG/CR-6864, SAND2004-5901 v.1). U. S. Nuclear Regulatory Commission, Washington, DC, 2005. [15] Draxler RR, Rolph GD. HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) model access via NOAA ARL READY Website (〈http://www.arl.noaa. gov/HYSPLIT.php〉). NOAA Air Resources Laboratory, College Park, MD, 2013. [16] Giordano A. The people and the plan: intended reactions to a nuclear emergency in New York State. Risk Manage 2005;7:41–56. [17] Giordano A, Anderson S, He X. How near is near? The distance perceptions of residents of a nuclear emergency planning zone Environ Hazards 2010;9:167–82. [18] Johnson JH, Zeigler DJ. Distinguishing human responses to radiological emergencies. Econ Geogr 1983;59:386–402. [19] Jones JA, Walton F, Smith JD, Wolshon B. Assessment of emergency response planning and implementation for large scale evacuations (no. NUREG/CR6981). U.S. Nuclear Regulatory Commission, Washington DC, 2008. [20] Jones JA, Walton F, Sullivan RL. Review of NUREG-0654, supplement 3, “Criteria for protective action recommendations for severe accidents” focus groups and telephone survey (No. NUREG/CR-6953 v.2). U.S. Nuclear Regulatory Commission, Washington, DC, 2008. [21] Lindell MK. An overview of protective action decision-making for a nuclear power plant emergency. J Hazard Mater 2000;75:113–29. [22] Lindell MK, Kang JE, Prater CS. The logistics of household evacuation in Hurricane Lili. Nat Hazards 2011;58:1093–109. [23] McKenna TJ. Protective action recommendations based upon plant conditions. J Hazard Mater 2000;75:145–64. [24] Mileti DS, Peek L. The social psychology of public response to warnings of a nuclear power plant accident. J Hazard Mater 2000;75:181–94. [25] Protective action guides and planning guidance for radiological incidents (no. EPA-HQ-OAR-2007-0268) U.S. Environmental Protection Agency, Washington, DC, 2013. [26] Ramsdell Jr. JV, Athey GF, McGuire SA, Brandon LK. RASCAL 4: description of models and methods (no. NUREG-1940). U.S. Nuclear Regulatory Commission, Washington, DC, 2012. [27] Sullivan R. Criteria for preparation and evaluation of radiological emergency response plans and preparedness in support of nuclear power plants: guidance for protective action strategies (no. revision 1 of Supp. 3 to NUREG-0654/FEMAREP-1). U.S. Nuclear Regulatory Commission, Washington, DC, 2011. [28] Urbanik T. Evacuation time estimates for nuclear power plants. J Hazard Mater 2000;75:165–80. [29] Wein LM, Youngsoo C, Denuit S. Analyzing evacuation versus shelter-in-place strategies after a terrorist nuclear detonation. Risk Anal Int J 2010;30 (9):1315–27. http://dx.doi.org/10.1111/j.1539-6924.2010.01430.x. [30] Witzig WG, Weerakkody WG. Evacuation risks: quantification and application to evacuation scenarios of nuclear power plants. Nucl Technol 1987;78:24–33.
Contents lists available at ScienceDirect
Reliability Engineering and System Safety journal homepage: www.elsevier.com/locate/ress
Alternative evacuation strategies for nuclear power accidents Gregory D. Hammond n,1, Vicki M. Bier University of Wisconsin-Madison, Department of Industrial and Systems Engineering, 3270A Mechanical Engineering Building, 1513 University Ave., Madison, WI 53706, USA
art ic l e i nf o
a b s t r a c t
Article history: Received 21 March 2014 Received in revised form 4 October 2014 Accepted 12 October 2014 Available online 27 October 2014
In the U.S., current protective-action strategies to safeguard the public following a nuclear power accident have remained largely unchanged since their implementation in the early 1980s. In the past thirty years, new technologies have been introduced, allowing faster computations, better modeling of predicted radiological consequences, and improved accident mapping using geographic information systems (GIS). Utilizing these new technologies, we evaluate the efficacy of alternative strategies, called adaptive protective action zones (APAZs), that use site-specific and event-specific data to dynamically determine evacuation boundaries with simple heuristics in order to better inform protective action decisions (rather than relying on pre-event regulatory bright lines). Several candidate APAZs were developed and then compared to the Nuclear Regulatory Commission’s keyhole evacuation strategy (and full evacuation of the emergency planning zone). Two of the APAZs were better on average than existing NRC strategies at reducing either the radiological exposure, the population evacuated, or both. These APAZs are especially effective for larger radioactive plumes and at high population sites; one of them is better at reducing radiation exposure, while the other is better at reducing the size of the population evacuated. Published by Elsevier Ltd.
Keywords: Nuclear incidents Evacuation Nuclear power plants Emergency preparedness Protective action recommendations
1. Introduction Current strategies used to determine who to evacuate following a nuclear-power plant accident have not changed significantly since the emergency planning guidelines were established in the early 1980s. While plans and studies have been modified and updated, this has been done under the constraint of a roughly constant evacuation area. Consequently, changes to protective actions have focused on issues such as in which order people should be evacuated, or in which direction they should evacuate [1,19]. Yet in the past thirty years, the task has radically changed; new technologies have been introduced, allowing faster computation, better modeling of predicted radiological consequences, improved accident mapping using geographic information systems (GIS), and new means to communicate. Additionally, the populations surrounding nuclear-power plants are denser; more people live closer to reactors than ever before. In the past 30 years, the average population living within
n
Corresponding author. E-mail addresses: gregory.hammond@afit.edu (G.D. Hammond), [email protected] (V.M. Bier). 1 Present address: Air Force Institute of Technology, Department of Systems Engineering and Management, 2950 Hobson Way, Wright-Patterson AFB, OH, 45433, USA. http://dx.doi.org/10.1016/j.ress.2014.10.016 0951-8320/Published by Elsevier Ltd.
16 km of these plants has increased by 62%, from approximately 40,000 to almost 65,000 per site. Furthermore, at 12 of the 65 reactor sites in the U.S., populations have more than doubled [2]. In the wake of the Fukushima Daiichi nuclear accident – considering the range of new capabilities and the greater population at risk – this study sought to reexamine the U.S. nuclear-power plant evacuation strategy by removing the constraint of a constant evacuation area or predetermined evacuation zones. This research is a proof of concept; its purpose was to develop alternative evacuation strategies for use during the early phase of nuclear-power plant accidents in order to take advantage of some of the recent technological advances. The early phase is defined by the U.S. Environmental Protection Agency (EPA) as “the period at the beginning of a nuclear incident when immediate decisions for effective use of protective actions are required and must therefore usually be based primarily on the status of the nuclear facility and the prognosis for worsening conditions” ([25], p. 5). Thus, the early phase is filled with uncertainty. The plant operators and emergency response officials know only that the situation at the reactor is a cause for concern and that an off-site radiological release is possible, so they can only guess at the extent of the problem. Despite this imperfect knowledge, officials must act and make decisions to protect the public from potential radiation exposure, generally in the form of evacuations (because distance is the best
10
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
protection) and/or sheltering in place. In the earliest periods of this phase, decisions are made based on predictions of the radiological release [25]. Notwithstanding the likely discrepancies between early phase projected doses and actual off-site doses that will be observed later, protective action recommendations must be made in the early phase, because evacuation in advance of the plume (ideally at least 1 h before the plume’s passage) is the best way to reduce dose [12,23]. This research explored alternative methods to determine who should be evacuated during the early phase. An ideal strategy would be able to perfectly evacuate the at-risk population before the radioactive plume passes. At present, of course, neither the APAZs developed in this research nor the NRC’s method can achieve this standard; however, as will be shown, the APAZs demonstrate progress towards meeting that standard. Because no method is perfect, the comparison between APAZs and the NRC’s method was framed as a multi-attribute decision problem using the objectives of minimizing the population to be evacuated and maximizing the total radiological dose avoided. These objectives are based on the current regulatory position of the NRC, which focuses evacuation efforts on high-risk areas [27], as well as the EPA guideline that informs the NRC’s policy, which states that evacuation risk should not exceed the risk from the avoided dose [11]. While there is general acceptance that avoiding radiological dose is beneficial, some might argue that instead of minimizing the population evacuated, the goal should be to maximize the size of evacuation. Because distance is a highly effective defense against radiation [12,23], some might argue that if the entire population surrounding a nuclear power plant could be evacuated prior to passage of the plume, that population could be guaranteed safety, suggesting that the objectives of high total dose avoided and high population evacuated would yield a desirable (if conservative) outcome. This logic is flawed, however, because maximizing the population evacuated can expose many people to risks greater than the risk of the radiological release, and ignores the risks and costs of evacuations. Evacuation can have adverse health impacts. Evacuation risks include travel, events in which travel is the contributing cause, and activities other than travel (i.e., preparation or reception activities) [3,30]. Witzig and Weerakkody have estimated travel risk to be 6 10 8 fatalities per vehicle-km; this risk is considered to be an upper bound as the actual risk is expected to be lower than normal automobile travel due to conditions of heavier traffic and lower travel speeds [3,30]. Injuries or fatalities in which travel contributed to their occurrence is the second category of risk. An example is an individual who evacuates the wrong direction and drives into a radioactive plume; it is believed to be an order of magnitude greater than travel risk [30]. The last evacuation risk, estimated to be 5 10 6 per person, is due to evacuation preparations and the arrival at the reception center ([3,30]). These three risks collectively form evacuation risk. For a given emergency, the evacuation risk is a function of the number of individuals that leave. When larger numbers of people evacuate who are not required to evacuate (i.e., shadow evacuations), the collective risk to the population will significantly increase [3]. The EPA evacuation risk estimate (for fatalities) corresponds to Witzig and Weerakkody’s upper bound estimate meaning that for radiation doses of less than 3 mSv, the evacuation risk is greater than the radiation risk [11]. Maximizing the evacuation area relocates many people who would receive doses less than 3 mSv, exposing them to needless risk. As noted by Aumonler and Morrey, “evacuation risks constitute a harm which should be considered in a decision as to whether to evacuate a population put at risk by a radiological incident” ([3], p. 290). For this reason, a safer course of action would evacuate those whose radiation risk is greater than their evacuation risk, but not those whose evacuation risk is greater than their radiation risk.
Minimizing the population evacuated and maximizing the total dose avoided embodies the EPA’s position that the protective actions should not be “higher than justified on the basis of optimization of cost and the collective risk of effects on health” ([11], p. 135). Thus, for this research, a high dose avoided and low population evacuated are assumed to be preferred. Using these two objectives, APAZs were compared to the NRC’s strategy using a concept called the “efficient frontier,” to allow decision makers to evaluate alternatives using their own value systems. Alternatives are plotted on the basis of the decision objectives (i.e., dose avoided and population evacuated). Dominated options can be excluded from consideration; the decision maker can then select a preferred option based on his or her preferences from among the non-dominated strategies on the efficient frontier. Current U.S. response protocols have been previously well documented. The interested reader is encouraged to review these earlier articles for a more in-depth understanding [11,21,23,24,28]. While the regulations that dictate emergency response have been updated (such as the EPA’s Protective Action Guides and Planning Guidance for Radiological Incidents [25] and the NRC’s guidance for protective action strategies [27]), as noted earlier, the fundamental initial evacuation strategy has remained constant. In the event of a nuclear-power plant accident, plant operators would determine evacuation areas using the NRC’s guidance for protective action strategies [27]. Based on the postulated source term and forecast meteorological conditions, a projected radiological plume is calculated and then fit to pre-established evacuation zones; this is the NRC’s keyhole strategy. (This strategy has been criticized because the plume model provides a simplified view of a complex process that may not correlate with the observed plume causing the wrong people to evacuate [23]). This research proposes an alternative method to determine the evacuation area. In this approach, instead of fitting the projected plume to pre-established zones, the evacuation area would be determined by applying a heuristic enlargement strategy directly to the forecast plume.
2. Calculations 2.1. Development and testing of APAZs Candidate APAZs (described subsequently) were tested using weather data from five nuclear power plants: Limerick; Catawba; Turkey Point; Pilgrim; and Arkansas Nuclear One. These plants were selected based on their proximity to National Weather Service (NWS) data-collection sites. The forecast and hindcast weather data (i.e., predicted and observed weather conditions) used in this analysis was produced by the National Oceanic and Atmospheric Agency (NOAA) and the NWS. The source term varied depending on the plant’s output power and reactor type. The source term for each nuclear-power plant was calculated using the time that the core was assumed to be uncovered in an unmitigated short-term station blackout (STSBO) scenario, described in the State-of-the-Art Consequence Analysis (SOARCA) [9]. The total release ranged from 3.3 1018 Bq to 1.2 1019 Bq. These postulated releases are of the same order of magnitude as that from the Fukushima Daiichi accident [26]. Forecast and hindcast plumes used in this research were generated with NRC’s Radiological Assessment System for Consequence AnaLysis (RASCAL) [6]. Nineteen candidate APAZ strategies were tested using 120 weather observations from summer 2012 and winter 2012–2103. Protective action zones, formed by enlarging a forecast plume in accordance with a given heuristic, were compared to the hindcast plumes.
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
The forecast plume generated from RASCAL was used to identify a mandated evacuation zone in which the 4-day total effective dose equivalent (TEDE) was predicted to be greater than or equal to 0.01 Sv, based on established policy [11]. Each of the 19 enlargement strategies was then applied to the mandated evacuation zone. The APAZ enlargement strategies consisted of varied offsets around the forecast evacuation zone (meaning that an additional X km was added to the size of the evacuation zone). Some of them had a uniform offset of varying distances (e.g., 4 km) around the forecast evacuation zone. Other strategies applied an offset in either a perpendicular or lateral direction to the prevailing wind direction. Other strategies applied a radial offset around the reactor in combination with an offset around the evacuation area. Furthermore, some strategies applied a multiplication factor to the predicted TEDE and then applied an offset. Many different strategies were tested in order to represent a wide range of policy options. The candidate APAZ heuristics were evaluated based the goals of minimizing the evacuation area and maximizing the adequacy of protective actions, as follows. First, the forecast plume was determined. It was then enlarged according to the relevant APAZ heuristic to determine the evacuation area. Next, the hindcast plume was determined, as a basis for assessing the adequacy of the APAZ. Area was measured as the ratio of the APAZ area to the hindcast evacuation area. Adequacy of the APAZ was measured by the fraction of the hindcast evacuation area that was also included in the APAZ evacuation area. The average area evacuated and average adequacy were compared for all nineteen heuristics. The comparison indicated that most heuristics fell on an efficient frontier. Results are shown in Fig. 1. The frontier extends from APAZ I (with small evacuation area but low adequacy) to APAZ T (with a large evacuation area and high adequacy). That the different strategies form a coherent efficient frontier indicates that no strategy can perfectly predict the location of the hindcast plume. Instead, the various strategies represent different trade-offs between the competing objectives of a small evacuation area and high adequacy. Three APAZs (B, D, and E) were retained for further comparison, because they are simple to implement and represent a range of policy choices.
Fig. 1. Ratio of APAZ to hindcast area as a function of APAZ adequacy.
11
APAZs B, D, and E were subjected to additional sensitivity analyses. Using a second set of 40 weather predictions and observations, the effect of dispersion model, source term, and evacuation timing were evaluated. The first sensitivity analysis compared our primary dispersion model (RASCAL), which is a 2-D plume model, with the 3-D HYSPLIT transport model [15]. A paired t-test was performed to check for adequacy differences between RASCAL and HYSPLIT. The t-test suggested that APAZ adequacy using the two different models was not statistically distinguishable. This result means that regardless of the hindcast model used, on average, APAZ adequacy will remain roughly the same. The second analysis considered the effect of alternative source terms. The initial screening was conducted using an unmitigated STSBO source term. To ensure that APAZs are robust with respect to the source term, APAZ adequacy and evacuation area were compared for the STSBO and a: long-term station black-out (LTSBO), STSBO with reactor core isolation cooling (RCIC) for boiling water reactors (BWR); and an LTSBO and thermally induced steam generator tube rupture (SGTR) for pressurized water reactors (PWR). For the alternative source terms, the four APAZs remained in their relative locations along the efficient frontier. The third analysis investigated the effect of forecast age. It was used to determine the effect of timing of the evacuation order. When any predictive PAR strategy is used (i.e., APAZs or keyhole), emergency response officials must make a determination of when to generate the forecast plume. Thus, the purpose of this analysis was to determine the effective time horizon for APAZs to understand their reliability in regards to the actual time of release. In the sensitivity analysis, forecast age was allowed to increase from zero hours (as had been used previously), up to 24 h before the release. This analysis showed that neither APAZ adequacy nor its area varied significantly as the forecast age increased. There is some variation, but that variation appears to be due to statistical noise rather than a significant difference. In all of the above analyses, APAZs B, D, and E all remained on the efficient frontier, indicating that they are all reasonably robust candidate evacuation strategies.
2.2. Comparison of APAZs to NRC evacuation strategies This phase of the research compared the three selected APAZ strategies to other protective action strategies. The APAZs included in this comparison were: (1) APAZ B (with a 5-km radial area around the reactor, plus a 2-km offset around the 0.01-Sv boundary); (2) APAZ D (with a 4-km offset around the 0.01-Sv boundary); and (3) APAZ E (with a 6-km offset around the 0.01-Sv boundary). The comparison PAR strategies were: (1) evacuation of the entire EPZ [9]; and (2) the keyhole strategy (utilizing the predefined NRC sectors [10]). The comparison was based on both the total radiological dose avoided (defined as the sum of TEDE for all persons evacuated) and the total population evacuated. Preferred strategies were assumed to be those with a high total dose avoided and a low population evacuated. Six nuclear-power plants were selected for this phase of the analysis: Surry; Peach Bottom; Grand Gulf; Kewaunee; San Onofre; and Braidwood; they are broadly representative of U.S. plants in terms of geographic region, surrounding population, and reactor type (BWR or PWR). We considered eight different weather observations per plant, chosen to include both summer and winter, days with and without precipitation, and differing times of day (i.e., mid-day or evening). Source terms were derived from the SOARCA report: long-term station black-out, STSBO, and STSBO with reactor-core isolation cooling for BWRs; long-term station black-out, STSBO, and steam generator tube rupture for PWRs. In total, 182 different cases were
12
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
analyzed (due to some weather conditions not being observed at some plants during the study period). The evaluation was performed using ArcGIS mapping software. We began with the geographic boundaries of census block groups, as obtained from the U.S. Census Bureau’s TIGER/Line Shapefiles ([7]). The geographic boundaries of the census blocks groups were then overlaid with 2010 census data ([8]) to determine how many people lived in each block group. Next, the hindcast plume (exported from RASCAL as a shapefile) was overlaid on the population data. The information from these two layers (population and hindcast plume) was used to compute the potential radiation dose that would be incurred if any given block group was not evacuated. The forecast plume (also exported from RASCAL as a shapefile) was then used to determine which geographic areas would be evacuated using APAZs B, D, and E, and the keyhole evacuation area. (It should be noted that the plumes were generally complicated in shape, and spanned multiple sectors.) In all cases, the keyhole was taken to be the union of all sectors in which the forecast plume was projected to fall, plus two additional sectors, one on either side of that region. Occasionally, the dose from the forecast plume exceeded 0.01 Sv beyond the outer boundary of the EPZ. In the keyhole strategy, evacuations past the EPZ were limited to areas where dose projections indicated that protective actions would be necessary [9], since NRC strategies are not defined beyond the EPZ. This strategy was implemented by evacuating the relevant keyhole area, plus all RASCAL program grid elements beyond the EPZ with doses exceeding 0.01 Sv. With the population, hindcast dose, and plume location known, for each evacuation strategy, the analysis then generated two statistics of interest: total dose avoided, given in TEDE (person-Sv); and total population evacuated. Total population evacuated was computed as the sum of the populations of all evacuated census blocks; total dose avoided was based on the populations and corresponding hindcast doses of all evacuated census blocks (i.e., all areas with doses greater than 0.01 Sv TEDE). The total dose avoided and population evacuated for each strategy were then compared to the total dose avoided and population evacuated that could have been achieved with perfect advance knowledge of the hindcast plume.
3. Results Fig. 2 below shows that APAZ B achieved comparable or better dose avoidance than could have been achieved by evacuating only the relevant portions of the hindcast plume on 77% of all cases, but evacuated significantly more people than the hindcast plume area in 41% of all cases. (The ideal is to achieve a comparable or better dose avoided in 100% of cases, and over-evacuate in 0% of cases.) We see that APAZs B (0.77, 0.41), D (0.88, 0.59), and E (0.92, 0.82) still form an efficient frontier. The keyhole strategy (0.67, 0.52) is dominated by APAZ B, and evacuation of the entire EPZ (0.88, 0.88) is dominated by APAZ D. Thus, it appears that APAZs are more efficient than keyhole evacuation or evacuation of the entire EPZ. On average, APAZ B is better at both avoiding radiological dose and minimizing over-evacuation than the keyhole strategy. In practical terms, when APAZ B is better than the keyhole strategy, it avoids about two more latent cancer fatalities (LCF) and evacuates about 4000 fewer people. Likewise, APAZ D is better than evacuation of the entire EPZ at minimizing over-evacuation. Compared to the keyhole strategy, APAZ D is better at avoiding dose (achieving approximately five fewer LCF), but evacuates approximately 6000 more people.
Fig. 2. Comparison of results.
Fig. 3. Effect of plume size on performance.
Breaking down the results further, as shown in Fig. 3, we found that for small accidents (i.e., those where the radioactive plume did not extend more than 8 km from the reactor, those where the plume travelled over a large body of water, or those with a population of less than 60,000 near the reactor), APAZs B and D and the keyhole strategy were comparable. However, for large accidents (i.e., those with plumes that traveled more than 8 km from the reactor, or at locations with more than 60,000 people within the EPZ), APAZs B and D were preferred to the keyhole strategy. APAZ B reduced the evacuated population by as much as 23,000 for large accidents (averaged over multiple sites, weather conditions, and source terms), and avoided between three and eight LCF. APAZ D was comparable to the keyhole strategy with regard to the size of the evacuated population, but avoided on average from 6 to 12 additional LCF in large accidents.
4. Discussion As described in the introduction, there is considerable uncertainty during the early phase of emergency response. Discussion is warranted to describe how we addressed this uncertainty.
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
McKenna has characterized the uncertainty in three domains: transport; individual dose response; and source term [23]. Transport consists of variability in both weather patterns throughout the release and the atmospheric transport model. As noted in Section 2.1, both elements were examined in this study. The meteorological effects were addressed by comparing forecast to hindcast plumes, holding other effects constant. Then, through sensitivity analysis, other effects were evaluated to ensure that the chosen heuristic was sufficiently robust to account for different release mechanisms, the timing of evacuation decisions, and RASCAL model effects. (The plume model effects were explored by comparing results from RASCAL with the HYSPLIT 3-D plume model [15]) None of these effects were found to be significant. This research did not directly address dosimetery uncertainty. However, the impact of this uncertainty was minimized by using the EPA assumption [11] of no shielding. As a result, projected doses were likely higher than those that would be expected following an actual release. While not entirely satisfying, this conservatism is consistent with current emergency-response protocols. Furthermore, while the source term uncertainty was not directly modeled, the results provided yield some insight into its implications. Source term uncertainty is a significant challenge; at best it is a factor of 10, but it could be a factor of 100–10,000 [23]. However, source terms were likely over-estimated because RASCAL assumes a near worst case scenario in which the radionuclides leave the reactor core quickly [26]. Moreover, any errors in source terms would apply equally to the APAZs and the NRC strategies. Finally, as was shown in Fig. 3, if the predicted source terms are too large, then we would expect to see little difference between the APAZs and the NRC strategies, because the true plumes will be small. However, if the predicted source terms are too small, then a distinguishable effect would be observed, because the true plume will be larger than thought. The assumptions and sensitivity analyses do not eliminate the inherent uncertainty in source terms, but it is not a threat to the validity of this research. Additionally, this research did not address the means and methods needed to communicate these alternative evacuation strategies with the public. In the event of a nuclear power accident, many residents are likely to ignore the protective action guidelines. Previous studies have demonstrated that compliance to a given directive during nuclear power emergencies is highly dependent on the nature of the directive. Individuals are willing to evacuate, or even shelter in place; however, when asked to shelter in place while others leave (as is the case of selective evacuations such as these strategies), they are much less willing to stay, with approximately 50% to 60% of respondents indicating that they would ignore the directive and leave ([16,18]; [20]). Evacuation compliance is further complicated by the tendency of individuals to not follow prescribed evacuation routes [22] as well as misperceptions regarding one’s location relative to the plant [17]. To overcome these challenges, these strategies would need to be effectively communicated in simple, understandable terms. The application of risk communication to these alternative strategies should be further investigated. The study is also limited because it assumed perfect evacuation in advance of the plume. Using the baseline assumption of 100% evacuation of the target population before passage of the plume, the APAZ strategies can be compared to other strategies without the use of a transportation model. Not using a transportation model simplifies the calculations significantly. While 100% evacuation is likely unrealistic, it allows for a fair comparison. Based on these results the total populations evacuated for the keyhole strategy and APAZ B were roughly equivalent; likewise, evacuation of the entire EPZ was roughly equivalent to APAZ E. If anything, the assumption of 100% evacuation in advance of the plume benefits
13
those strategies (APAZ E and the entire EPZ) that evacuated larger areas, since these large evacuations are not penalized for excessive road congestion, even though in reality individuals may get stuck in traffic jams while attempting to evacuate. Lastly, the findings of this study is limited to the comparison of protection levels offered by different evacuation strategies. Research comparing the protective strategies of sheltering in place and evacuation in the event of nuclear terrorism suggests that sheltering in place followed by a delayed evacuation provides the greatest level of protection during the early response phase [4,5,29]. Additional research by Denning et al. [13] extends this finding to nuclear power accidents. The purpose of this paper was not to compare sheltering in place and evacuation; rather it sought to present alternative evacuation strategies. Future research should consider the impact of these alternative strategies as applied to delayed evacuation and sheltering in place.
5. Conclusion The objective of this research was to identify alternative evacuation strategies as a proof of concept. In this, the research was successful; it identified candidate evacuation methods that perform better than those currently employed. Should these potential gains be considered significant enough by policy makers, future research should further develop the framework into a concept of operations that addresses potential implementation issues and assumed administrative burdens. Indeed, the main significance of this work is not the choice of one APAZ or another, but rather the methodology that was used to test and evaluate alternatives. We found current strategies to be adequate for smaller accidents; however, we found a distinguishable difference for larger accidents. Therefore, policy makers should seriously consider implementing the findings of this research. Authorities that value balancing evacuation and radiological should consider adopting either APAZ B or APAZ D. Likewise, APAZ E is a viable option for decision makers who to prefer to maximize radiological risk avoidance. Such an action would strengthen the overall defensein-depth of nuclear-power plants. Note, however, that altering the current evacuation strategy is likely to have significant costs and consequences. Indeed, such an action would directly affect the emergency response network, requiring evacuation areas at all levels of government to be redrawn. Moreover, the human response to these alternative protocols must also be investigated. Previous studies have also found benefits with alternative strategies [14,20], but the solutions were dismissed due to concerns of impracticality. Accordingly, the practical importance of the improvements achievable using APAZs should most appropriately be seen as a value judgment. In other words, the APAZ approach, which can improve readiness for larger radioactive releases (and also minimize unnecessary evacuations), must be evaluated in the context of its implementation costs and administrative burdens. The NRC, in its statutory role of ensuring the safe use of radioactive material and protection of the population, should assess the importance of the improvements we identified, through further study, as they relate to public safety and policy.
Acknowledgements The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and the Nuclear Regulatory Commission (NRC) for the provision of the RASCAL consequence management software
14
G.D. Hammond, V.M. Bier / Reliability Engineering and System Safety 135 (2015) 9–14
used in this publication. They would also like to thank the reviewers for their insightful comments and recommendations. References [1] Agrawal B, Cunningham M, Denning R, Glynn J, Johnson J, Niyogi P, VanderMolen H. Severe accident risks: an assessment for five U.S. Nuclear Power Plants (no. NUREG-1150). U.S. Nuclear Regulatory Commission, Washington, DC, 1990. [2] AP AP: Populations around U.S. nuclear plants soar. USA Today, 27 June, Available at: 〈http://www.usatoday.com/news/nation/2011-06-27-Nuclearplants-population-evacuation_n.htm〉 (last accessed: 18 Mar 14), 2011. [3] Aumonier S, Morrey M. Non-radiological risks of evacuation. J Radiol Prot 1990;10:287–90. [4] Buddemeier BR. Reducing the consequences of a nuclear detonation: recent research. The Bridge. Summer, 28-38. Available at: 〈https://www.nae.edu/ Publications/Bridge/19804/19920.aspx〉; 2010. [5] Buddemeier BR, Valentine JE, Millage KK, Brandt LD. National Capital Region: Key response planning factors for the aftermath of nuclear terrorism. Technical Report LLNL-TR-512111, Lawrence Livermore National Lab, 2011. [6] Brandon L, Athey GF. RASCAL v. 4.1, software. Washington, DC: U.S. Nuclear Regulatory Commission; 2011. [7] Census Bureau US. 2012 TIGER/Line Shapefiles [WWW Document]. URL 〈http://www.census.gov/cgi-bin/geo/shapefiles2012/main〉 (accessed 2 March 13), 2011. [8] Census Bureau US. American Factfinder [WWW Document]. Available at: 〈http://factfinder2.census.gov〉 (accessed 2 March 13), 2013. [9] Chang R, Schaperow J, Ghosh T, Barr J, Tinkler C, Stutzke M. State-of-the-art reactor consequence analyses (SOARCA) report ( no. NUREG-1935). U.S. Nuclear Regulatory Commission, Washington, DC, 2012. [10] Congel F, Kantor F, McKenna T, Mohseni A, Hepler M, Wingert VL. Criteria for preparation and evaluation of radiological emergency response plans and preparedness in support of nuclear power plants: criteria for protective action recommendations for severe accidents (no. Supp. 3 to NUREG-0654/FEMAREP-1, Rev. 1). U.S. Nuclear Regulatory Commission, Washington, DC, 1996. [11] Conklin C, Edwards J. Selection of protective action guides for nuclear incidents. J Hazard Mater 2000;75:131–44. [12] Cutter SL, Barnes K. Evacuation behavior and three mile island. Disasters 1982;6:116–24. [13] Denning R, Osborn D, Brunnett A, Aldemir T. Nov. relative importance of mitigation, early phase, and late phase response. Paper presented at the American Nuclear Society Annual Meeting, embedded topical meeting: risk management for complex socio-technical systems, Washington, DC, 2013.
[14] Dotson LJ, Jones J, Schneck D, Sullivan R. Identification and analysis of factors affecting emergency evacuation (no. NUREG/CR-6864, SAND2004-5901 v.1). U. S. Nuclear Regulatory Commission, Washington, DC, 2005. [15] Draxler RR, Rolph GD. HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) model access via NOAA ARL READY Website (〈http://www.arl.noaa. gov/HYSPLIT.php〉). NOAA Air Resources Laboratory, College Park, MD, 2013. [16] Giordano A. The people and the plan: intended reactions to a nuclear emergency in New York State. Risk Manage 2005;7:41–56. [17] Giordano A, Anderson S, He X. How near is near? The distance perceptions of residents of a nuclear emergency planning zone Environ Hazards 2010;9:167–82. [18] Johnson JH, Zeigler DJ. Distinguishing human responses to radiological emergencies. Econ Geogr 1983;59:386–402. [19] Jones JA, Walton F, Smith JD, Wolshon B. Assessment of emergency response planning and implementation for large scale evacuations (no. NUREG/CR6981). U.S. Nuclear Regulatory Commission, Washington DC, 2008. [20] Jones JA, Walton F, Sullivan RL. Review of NUREG-0654, supplement 3, “Criteria for protective action recommendations for severe accidents” focus groups and telephone survey (No. NUREG/CR-6953 v.2). U.S. Nuclear Regulatory Commission, Washington, DC, 2008. [21] Lindell MK. An overview of protective action decision-making for a nuclear power plant emergency. J Hazard Mater 2000;75:113–29. [22] Lindell MK, Kang JE, Prater CS. The logistics of household evacuation in Hurricane Lili. Nat Hazards 2011;58:1093–109. [23] McKenna TJ. Protective action recommendations based upon plant conditions. J Hazard Mater 2000;75:145–64. [24] Mileti DS, Peek L. The social psychology of public response to warnings of a nuclear power plant accident. J Hazard Mater 2000;75:181–94. [25] Protective action guides and planning guidance for radiological incidents (no. EPA-HQ-OAR-2007-0268) U.S. Environmental Protection Agency, Washington, DC, 2013. [26] Ramsdell Jr. JV, Athey GF, McGuire SA, Brandon LK. RASCAL 4: description of models and methods (no. NUREG-1940). U.S. Nuclear Regulatory Commission, Washington, DC, 2012. [27] Sullivan R. Criteria for preparation and evaluation of radiological emergency response plans and preparedness in support of nuclear power plants: guidance for protective action strategies (no. revision 1 of Supp. 3 to NUREG-0654/FEMAREP-1). U.S. Nuclear Regulatory Commission, Washington, DC, 2011. [28] Urbanik T. Evacuation time estimates for nuclear power plants. J Hazard Mater 2000;75:165–80. [29] Wein LM, Youngsoo C, Denuit S. Analyzing evacuation versus shelter-in-place strategies after a terrorist nuclear detonation. Risk Anal Int J 2010;30 (9):1315–27. http://dx.doi.org/10.1111/j.1539-6924.2010.01430.x. [30] Witzig WG, Weerakkody WG. Evacuation risks: quantification and application to evacuation scenarios of nuclear power plants. Nucl Technol 1987;78:24–33.