Evaluation and ranking of restoration strategies for radioactively contaminated sites

Download Evaluation and ranking of restoration strategies for radioactively contaminated sites

Post on 03-Jul-2016




3 download

Embed Size (px)


<ul><li><p>Journal of</p><p>Environmental Radioactivity 56 (2001) 3350</p><p>Evaluation and ranking of restoration strategiesfor radioactively contaminated sites</p><p>Th. Zeevaerta,*, A. Bousherb, V. Brendlerc,P. Hedemann Jensend, S. Nordlindere</p><p>aSCK CEN, Boeretang 200, 2400 Mol, BelgiumbWestlakes Scientic Consulting, Moor Row, Cumbria CA24 3LN, UK</p><p>cForschungszentrum Rossendorf, Institut f .ur Radiochemie, 01314 Dresden, GermanydRisoe National Laboratory, Dpt. of Nuclear Safety Research, 4000 Roskilde, Denmark</p><p>eStudsvik Eco &amp; Safety AB, 61182 Nyk .oping, Sweden</p><p>Received 28 February 2000; accepted 12 July 2000</p><p>Abstract</p><p>An international project, whose aim was the development of a transparent and robustmethod for evaluating and ranking restoration strategies for radioactively contaminated sites</p><p>(RESTRAT), was carried out under the Fourth Framework of the Nuclear Fission SafetyProgramme of the EU. The evaluation and ranking procedure used was based on theprinciples of justication and optimisation for radiation protection. A multi-attribute utility</p><p>analysis was applied to allow for the inclusion of radiological health eects, economic costsand social factors. Values of these attributes were converted into utility values by applyinglinear utility functions and weighting factors, derived from scaling constants and expertjudgement. The uncertainties and variabilities associated with these utility functions and</p><p>weighting factors were dealt with by a probabilistic approach which utilised a LatinHypercube Sampling technique. Potentially relevant restoration techniques were identied andtheir characteristics determined through a literature review. The methodology developed by</p><p>this project has been illustrated by application to representative examples of dierentcategories of contaminated sites; a waste disposal site, a uranium tailing site and acontaminated freshwater river. # 2001 Elsevier Science Ltd. All rights reserved.</p><p>Keywords: Restoration; Radioactivity; Ranking; Evaluation; Optimisation</p><p>*Corresponding author. Tel.: +32-14-33-28-68; fax: +32-14-32-10-56.</p><p>E-mail address: tzeevaer@sckcen.be (T. Zeevaert).</p><p>0265-931X/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.PII: S 0 2 6 5 - 9 3 1 X ( 0 1 ) 0 0 0 4 6 - 7</p></li><li><p>1. Introduction</p><p>A number of nuclear facilities in Europe began operation in the 1950s and 1960sand, as a consequence, are now reaching the end of their life expectancy. The maintechnical part of the installations will be subject to a controlled decommissioning.However, the sites themselves may require restoration where the contamination hasbeen dispersed or where radioactive residues are contained by methods which may beunreliable for long-term storage.Restoration of such sites would appear necessary both for the sake of protecting</p><p>the public and for relieving otherwise costly control of the sites. However, clean-upby conventional techniques can be very expensive and may, therefore, not oer theoptimal solution based on cost-benet evaluations. The US experience at formermilitary sites shows that the application of alternative techniques, e.g., in-siturestoration, is hampered by the lack of transparent risk assessments in consideringthe exposure of present and future populations and of restoration workers. Anothershortcoming is the lack of relevant data from on-site experiences.An international project partly funded by the EU under the Fourth Framework of</p><p>the Nuclear Fission Safety Programme has been addressing these problems. In thisproject, a transparent, generic decision-aiding methodology has been established,which is capable of evaluating and ranking restoration strategies for radioactivelycontaminated sites and their close surroundings (RESTRAT). The study has nowbeen concluded and a manual (Zeevaert, Bousher, Brendler, Nordlinder, &amp;Hedemann, 1999a) has been produced to explain the methodology and to apply itto typical example cases. A summary of the progress made at the mid-term wasreported by Jackson et al. (1999). The nal report for this project is also available(Zeevaert, Bousher, Brendler, Nordlinder, &amp; Hedemann, 1999b).</p><p>2. Method</p><p>The method proposed for the ranking of restoration options is based on theradiation protection principles of justication and optimisation, as recommended inICRP 60 (International Commission on Radiological Protection, 1991). In ICRP 60recommendations, distinction is made between practices and interventions. Thisdistinction may not always be clear for clean-up of radioactively contaminated land.However, in cases where there is existing exposure of a population from sitescontaminated with the residues of past, or old, practices or work activities, theprinciples of protection for intervention are applicable, and dose limits do not apply.According to the justication principle, the intervention or restoration should domore good than harm, which means that the net benet (benet of the reduction inradiation detriment less the harm and costs of the restoration) should be positive.According to the optimisation principle, among the justied restoration options, thestrategy with the highest net benet should be selected.Quantitative decision-aiding techniques for radiological optimisation analysis</p><p>purposes are available. The principal ones are described in detail in ICRP 55</p><p>T. Zeevaert et al. / J. Environ. Radioactivity 56 (2001) 335034</p></li><li><p>(International Commission on Radiological Protection, 1989) and by Stokell, Croft,Lochard, and Lombard (1991).Decisions in site restoration require multiple criteria to be taken into account. The</p><p>major attributes, to be considered, include:</p><p>* Health attributes;* Economic attributes;* Social attributes.</p><p>Social factors are dicult to quantify. Therefore, a multi-attribute utility (MAU)analysis is the obvious technique to be applied for optimisation. In order to carry outthe evaluations in this way a hierarchy of attributes must be set up. Themajor attributes are divided into sub-attributes according to the scheme as shownin Fig. 1.The values for the attributes of each option are determined and converted into</p><p>utility values through the use of utility functions. Each attribute is assigned aweighting factor, expressing the relative importance of that attribute with respect tothe other attributes of the same group. Each alternative strategy can then be assigneda score which is equal to the sum of the products of utility values and weightingfactors for all its attributes. The option with the highest score is considered to beoptimal.Utility functions can be linear, concave or convex. They can either be of an</p><p>increasing type or of a decreasing type. Risk-neutral, risk-averse and risk-proneutility functions of the decreasing type are shown in Fig. 2.In this study, risk-neutral, linear utility functions are applied. These are expressed</p><p>as</p><p>Ux 100 1xmin x</p><p>xmax xmin</p><p> ;</p><p>where (xmin; xmax) is the value range of the attribute considered.The restoration option which does best for a particular attribute is assigned a</p><p>utility value of 100, and the option which does worst a utility value of zero. Otheralternatives are assigned intermediate utility values according to the utility functionshown above.Determination of the weighting factors of the attributes is a subjective task; each</p><p>decision-maker may well come up with a dierent set of weighting factors.Therefore, a simple scaling method has been proposed to establish conversionconstants between the weighting factors for attributes of the same hierarchy level.For the major attributes, health attributes, economic costs and social factors, these</p><p>constants can be expressed as</p><p>weconomicwhealth</p><p> C1;wsocialwhealth</p><p> C2</p><p>and the sum of the weighting factors should be equal to one, i.e.</p><p>weconomic wsocial whealth 1:</p><p>T. Zeevaert et al. / J. Environ. Radioactivity 56 (2001) 3350 35</p></li><li><p>This would determine the weighting factors as</p><p>whealth 1</p><p>1 C1 C2; weconomic </p><p>C11 C1 C2</p><p>; wsocial C2</p><p>1 C1 C2:</p><p>The value of C1 can be determined as the ratio of the ranges of economic costs to theranges of collective doses, multiplied by the monetary value of the man. Sievert(taken here as 100,000 EUR man Sv1) (Nordic Radiation Protection Authorities,1991).The ratio C2 is much more dicult to quantify. Intuitively, it would be expected to</p><p>be less than one and, for non-accidental situations such as restoration ofcontaminated sites with small exposures, signicantly less than one. In this study avalue of 0.25 has been proposed.The weighting factors for the sub-attributes of a same group have been determined</p><p>in a similar manner to that described above. For sub-attributes, expressed in thesame units, conversion constants have been taken equal to the ratios of the ranges ofthe values of the sub-attributes. For the social sub-attributes, reassurance is given ahigher weight (57 times higher) than disturbance because of its more permanentnature. However, further research studies are needed before the weighting factors for</p><p>Fig. 1. Attibute hierarchy for restoration of a contaminated site.</p><p>T. Zeevaert et al. / J. Environ. Radioactivity 56 (2001) 335036</p></li><li><p>social sub-attributes can be properly quantied. Further details may be found inHedemann Jensen (1999).</p><p>3. Assessment of attributes</p><p>3.1. Radiological health</p><p>The radiological impact was assessed in terms of eective doses to the public andto the restoration workers. Collective dose was considered as a measure for theradiological health detriment. For the public, the collective doses have beentruncated at 100 and 500 years, in accordance with international recommendations(Barraclough, Robb, Robinson, Smith, &amp; Cooper, 1996; United Nations ScienticCommittee on the Eects of Atomic Radiation, 1982). The doses are calculated bothwith and without the restoration measures implemented, so that the doses avertedthrough those measures could be determined.Maximum annual individual doses to the critical group (before restoration) have</p><p>been assessed for the sake of comparison with the recommendations of ICRP(International Commission on Radiological Protection, 1998) and of IAEA(International Atomic Energy Agency, 1997) on the clean-up of contaminated land.For restoration workers, doses can be readily calculated from the contamination</p><p>levels at the site and the labour volumes required. Exposure will only take placethrough external irradiation and inhalation on the site.For the public, a more complicated impact assessment model is needed for</p><p>assessing the doses. Ingestion of contaminated food and water and exposure o-site(through use of contaminated water for irrigation purposes for instance) must alsobe taken into account. Therefore, a compartmental type of model has been applied.This is based on the BIOPATH model developed by Studsvik (Bergstr .om, Edlund,Evans, &amp; R .ojder, 1982), which has been validated and veried in severalinternational studies such as PSACOIN (Klos et al., 1993). The compartmentalscheme applied has been adapted to each of the major categories of contaminatedsites considered. Fig. 3 shows its application to a contaminated fresh water river(Molse Nete).</p><p>Fig. 2. Examples of utility functions of the decreasing type : risk-neutral (linear), risk-averse and risk-</p><p>prone utility functions.</p><p>T. Zeevaert et al. / J. Environ. Radioactivity 56 (2001) 3350 37</p></li><li><p>In the impact assessment model, two parts can be distinguished:</p><p>* the transport model (the proper compartmental part),* the exposure model.</p><p>The transport model describes the exchange or transfer of the contaminants(radionuclides) between biosphere compartments. Processes that are responsible forthose transfers can be identied. They may be caused by man, such as dredging ofsediment, irrigation, or they may be of a natural origin, such as water ow,advection, diusion, sedimentation, bioturbation, etc. Mathematically, thesetransfers are expressed by a set of rst-order dierential equations with constantor time-varying transfer coecients (rate constants, expressing the number ofturnovers per unit of time). The application of restoration techniques are modelledeither by adapting the transfer coecient values for the processes that are inuencedor, by adapting the value of the source term in the cases where sources are removed(with or without subsequent separation).The exposure model is the part of the assessment model, which converts</p><p>radionuclide (contaminant) concentrations, in the relevant biosphere media, intodose values to the exposed population. Exposure pathways to the population mustrst be identied. They may include:</p><p>* external irradiation: on contaminated elds or river banks,* inhalation due to resuspension,* ingestion of contaminated drinking water,* ingestion of sh from contaminated surface water,* ingestion of food crops, contaminated through irrigation, or grown on soil</p><p>contaminated through application of amendments,</p><p>Fig. 3. Compartmental scheme for a freshwater river (Molse Nete).</p><p>T. Zeevaert et al. / J. Environ. Radioactivity 56 (2001) 335038</p></li><li><p>* ingestion of milk, meat contaminated through the watering of cattle, or throughcattle grazing on contaminated pasture.</p><p>Uncertainty and sensitivity analyses were also carried out with respect tothe collective doses, using the PRISM program (Gardner, R .ojder, &amp; Bergstr .om,1983). This is a general tool for addressing the uncertainties in any modelarising from the uncertainty or variability in parameter values. The uncertainty isevaluated by generating random parameter values applying a systematicLatin Hypercube sampling method, and executing the model for each set ofinput parameter values. The joint set of model parameters and results are thenstatistically evaluated by considering means, percentiles, standard deviations. Forthe sensitivity analysis, correlations between model parameters and model resultsare examined, using the simple Pearson and the Spearman rank correlationcoecients.An important issue recognized in the RESTRAT project has been the inuence of</p><p>physico-chemical phenomena, or processes, on the source term evolution and on themigration of radionuclides in the environment. Relevant physico-chemical phenom-ena have been identied from a literature review and the parameters, for describingthe processes in a quantitative way, have been determined.Distribution coecients (Kd) play a very important role in this respect. In</p><p>common with most impact assessment models, the BIOPATH model also uses singleKd values taken from the literature. This concept is, however, too simplistic. Manydierent basic physico-chemical phenomena (hydrolysis and complexations, redoxreactions, mineral precipitation and dissolution, adsorption and ion exchange)determine the Kd value and even small changes in the physico-chemical parameters,Eh, pH, concentrations, mineral composition, temperature, etc, can inuence itsvalue. This assigns very large uncertainties to Kd values which constitute importantsources of uncertainty to the dose results.A better strategy than using single Kd values is to unfold the Kd into a parameter</p><p>vector; decomposing the Kd into its underlyin...</p></li></ul>