Summary of Meeting Paper

The 1996 Annual Meeting of the Society for Risk Analysis-Europe

Options for Decision Makers--Recovery Actions After a Nuclear Accident. N. A. Higgins and K. A. Jones, National Radiological Protection Board, Chilton, Didcot, Oxfordshire, OX11 ORQ UK

INTRODUCTION

Following the unlikely event of a nuclear accident, actions would be taken to ensure that the public was adequately protected. In considering the actions that might be taken it is convenient to identify two time phases, the emergency phase and the recovery phase. Broadly speaking the emergency phase lasts from the time when the release of radioactivity is first anticipated or occurs until it is judged that there is no further threat of release. The recovery phase then begins and continues until all those affected have resumed 'normal' lifestyles. During the recovery phase less urgent countermeasures can be performed which will protect the public from longer term, chronic risks. In addition actions may be required to aid the long term recovery of the affected population. This continuing study explores the definition and role of factors which characterise the post emergency situation and which could be used to establish what, if any, recovery actions would be required. This work is providing input to the production of criteria which will establish a generic framework for post-emergency decision making after an accident.

The main pathways of exposure during the recovery phase of an accident are external irradiation from radionuclides deposited in the environment, inhalation of resuspended radionuclides and ingestion of contaminated foods. Contaminated foods are dealt with separately from other exposure pathways, and are subject to Regulations issued by the Council of the European Communities. Countermeasures to protect consumers from contaminated food were therefore not considered in this study. The countermeasures which may be applied to protect the public from exposure via the external irradiation and inhalation pathways are composed of restrictions on access and decontamination. However, accidents severe enough to require measures such as permanent relocation and long term restrictions on the movement of the population are extremely unlikely. This paper therefore considers more limited interventions which are more likely to be appropriate, such as small scale decontamination procedures and short term relocation. A series of interlinked scenarios are explored to examine the grounds on which decisions to take these actions might be made and how the decisions might change with the severity of the accident.

SCENARIOS

The postulated accident scenarios were based on a Magnox 'design basis accident' source term. An important feature of this source term for the current discussion is that the released radionuclides are mainly short-lived. Four closely related scenarios were developed, which differed in the amount of deposition from the plume that was assumed to occur. The first scenario assumed deposition in dry weather conditions whilst the other three assumed levels of deposition which varied as a result of different levels of rainfall, from light rain to a heavy downpour. It was postulated for all the scenarios that the release traveled in the direction of a village a short distance downwind of the site and it was for this village that the implementation of recovery actions was considered.

Decontamination techniques range over the entire gamut from cheap to expensive, slow to fast implementation, and highly effective in reducing dose to only weakly dose-effective. Complex intrusive techniques are unlikely to be required and were therefore not considered. Any technique applied to deal with a release in which the majority of the dose causing concern would come from short-lived radionuclides must be applied quickly to achieve its full effectiveness. Therefore decontamination techniques that need to be applied over extended periods or would take a long time to organise were not considered.

Estimates were made of the likely dose reductions which could be achieved using one of two techniques applied to similar surfaces but at opposite ends of the cost scale. One, fire hosing of paved surfaces, (roads, driveways etc.) represents a procedure which must be applied rapidly to achieve its maximum effectiveness and which is comparatively cheap, quick, and moderately dose-effective. The other technique, road planing, is a more substantial procedure which, in general, does not need to be applied as quickly (although it is assumed to be started quickly in this case to be effective against short-lived radionuclides), takes much longer, has high dose-effectiveness, and is considerably more expensive. There are a great many techniques that lie somewhere between these two extremes in cost but rarely match them in the effectiveness of their dose reduction. In practice techniques which act on different surfaces may be combined but for tractability this additional complication has not been considered at this stage.

It was assumed that the population had been evacuated during the emergency phase and that they would stay out until any decontamination measures were completed. In addition, because of the importance of short-lived radionuclides in the release, temporary relocation was considered separately for time periods which matched the times the population would be absent if one of the two decontamination techniques were applied. This enabled the benefits of the decontamination options to be assessed, and the implicit contribution to these benefits produced by removing the population from the affected area for the required length of time. Table 1 indicates the estimated effectiveness of these countermeasures for the particular nuclide mix released. It demonstrates that the short lifetime of a high proportion of the radionuclides in the postulated release is the dominant factor contributing to the very high effectiveness of the options.

METHODOLOGY

To enable the important factors in a decision to decontaminate to be clearly identified the technique of multi-attribute value analysis (MAVA) was applied to each of the four scenarios. This enables subjective valuations of individual factors or attributes, acted on by the selected options, to be combined to give a preference order.

Given below is a list of the attributes which have been considered. This list is not meant to be prescriptive but rather to highlight issues of obvious concern to the decision maker.

In general the scores of the above attributes against the selected options can be estimated directly. However, this is difficult for the very important reassurance attribute which acts as a measure of the public acceptability of the doses that may result from the selection of a particular countermeasure. A surrogate measure, using the maximum external irradiation dose that may be received in the first year, was therefore used. The attributes were scored against the set of options given in Table 1 and in addition the option of 'No action'. The 'No action' option means that no decontamination or temporary relocation has taken place, i.e. the population returned to their homes at the end of the emergency phase.

Linear value functions were adopted initially for all the attributes and equal weighting was assumed for the three main arms under light rain conditions. Subsequently, variations in the weighting given to the arms and also the consequences of a nonlinear value function for reassurance were investigated. The reassurance derived by the local population is likely to be affected by many complex and subtle issues, e.g. how much trust the population places in the authorities, how much information they receive and of what type. A simple linear value function is unlikely to be able to represent these requirements in all circumstances.

RESULTS AND DISCUSSION

The estimated attribute scores associated with each option for each of the postulated scenarios were calculated and valued. It is important to note that the scores of two of the attributes (cost and time to return to normal living) are independent of the particular scenario and therefore their value functions are also scenario independent. In addition, if linear value functions are assumed for the other attributes, although the scores differ between scenarios, the value of the options for the scenarios with rain are identical. This latter effect is a consequence of a lack of experimental information on the variation in the dose effectiveness of decontamination techniques under different wet conditions, i.e. the dose effectiveness of road planing is assumed to be the same whether there was a light downpour or heavy rain at the time of deposition.

The objective of this multi-scenario approach is to enable the implications of choices for one accident scenario to be easily assessed under distinct but closely related conditions. In this way a consistent and satisfactory choice can be made across all the scenarios and governing criteria extracted for further study. Clearly in order to compare value estimates across scenarios a consistent weighting scheme must be developed. The scheme used required the same dose tradeoff with the scenario independent attributes in all the scenarios. For example, there was a consistent variation in the relative weighting of time to return and reassurance. Therefore the weighting scheme diminished the importance of the cost and time to return attributes as the rainfall and hence the level of dose considered increased. The dose ranges over all the scenarios were such as to indicate little need for action in the lowest dose range and a likely need for more significant and potentially invasive actions in the highest dose range. The difficulty is in deciding the appropriate level of response under conditions which give intermediate levels of dose. In this case there is no clear a priori position to adopt and it is likely that different individuals or groups would have differing views on the options and attributes. Weighting schemes were investigated which were consistent with the expected response in the low and high dose scenarios and provided insights into the scale of action appropriate at intermediate levels of dose. Results were found to be moderately robust with respect to variations in the weightings employed. A consistent weighting scheme links the scenarios by reducing the number of independent weights. A sensitivity analysis across all the scenarios indicates that if the independent weights were chosen to be in a low dose scenario, the preference order becomes progressively less susceptible to changes in these primary weights the higher the dose scenario considered. This behaviour is to be expected as the range of generally acceptable options must reduce as the level of dose to which a population might be exposed increases. Conversely it indicates that at lower ambient levels of dose, views representing only slightly different weightings will result in different options being favoured. This indicates that an important aim of future work should be a greater understanding of the factors and views influencing the range of options that may be selected to mitigate the effects of low and intermediate doses.

The effect of adopting a nonlinear variation in value between the levels of reassurance produced by different options was explored. Comparatively simple rounded step functions were used. The distinction produced by the consistent weighting scheme between the best and worst options is reinforced by this form of function. However, considering only the best reassurance options, it also decreased the overall differences in value between these options. Thus, using a nonlinear value function of this form decreases the influence of the attribute in determining the preference order of the best options.

This work has provided a good foundation but in order to confirm and broaden the understanding gained and further aid the production of generic criteria for recovery countermeasures it is essential to consider other situations. Different scenarios will be considered in order to elucidate more fully factors that are likely to be relevant to all accident situations. In addition the application of a greater range of decontamination options both singly and in combination will be explored and the effect of uncertainty in the outcome of any applied option considered. This knowledge will strengthen the case used in the development of basic criteria and provide the opportunity for more detailed secondary criteria to be developed.