Importance of uncertainties in dose assessment to prove compliance with radiation protection standards Manfred TSCHURLOVITS, Atominstitute of Austrian Universities, Vienna University of Technology, Stadionallee 2, A-1020 Vienna, Austria E-mail: tschurlo@ati.ac.at Abstract. Quantification of radiation protection measures is based upon prospective limitation and retrospective assessment in terms of dose, but executed often in auxiliary quantities as dose rate or activity concentration. One key issue is, however, not considered at all in standards: how to handle uncertainties resulting from assessment. Limiting quantities (as annual dose of an individual) are expressed as a single even number and hence without any explicit uncertainty and give the impression that they are most precise, although they are rounded numbers or default values. Results of measurements are mostly in terms of auxiliary quantities and have to be converted to a limiting quantity and expressed by a number and a confidence interval, which is often very large because of technical reasons. In international standards, limiting quantities and operational quantities were coined. There is a large number of possible definitions of dose, but clarity and applicability is not proportional to the number of different quantities already created. This is influenced by the fact that a quantity is defined to fulfil a certain well quantitites are used to demonstrate compliance with limits, no only the number, but also the uncertainty has to be considered in an appropriate manner. The hierarchy of limiting quantities has to be taken into account in accepting a given uncertainty. The acceptable uncertainty is, however, dependent on the relation of the figure to the limit. When derived quantities are used to prove compliance, biasing optimisation by applying overconservative as well as disregarding uncertainties contributing factors. For these reasons, standards have to include information to which extent an uncertainty is acceptable in dose assessment to prove compliance with limits, where the hierarchy of limits has to be taken into account. The paper discusses some possible approaches. 1. Introduction Limits in the dose range which is characterized in Radiation protection by normal operation must not be considered as a sharp boundary between safe and not safe. This is because only stochastic effects are to be expected and are hence the underlying issue. The risk factors underlying the limits are both variable and uncertain [1]. Therefore, an execution like a precise threshold is not justified. This question was in most cases of less importance since the observed dose range was well below limits. In recent recommendations and standards, however, the dose limits are lowered, and hence the question how to prove compliance may become acute as numbers are more approaching the limits than previously. Limits are rather a measure to satisfy administrators in applying radiation protection standards, and this is mirrored by the definition[2]: Limit: The value of a quantity used in certain specified activities or circumstances that must not be exceeded. Assessing of the quantities subject of limitation by measurements, one has the problem that the measured quantities include an inherent uncertainty being often in size comparable with the quantity itself, where the preset limits have the uncertainty of zero by definition. It will be shown that a uniform consideration of the question of acceptable uncertainty and its relation to different (see below) limits is neither necessary nor reasonable. This is because the consideration has to take into account the measured number in relation to the limit, the properties of the assessment technique and the position of the considered quantity in the hierarchy of limits. The rank in the hierarchy is directed whether the quantity is a primary limit or a derived quantity, as well described in [3], where the following distinction is made. The major characteristics are added in parenthesis. 1
Primary limits (effective dose apply to an individual, no access by measurement) Secondary limits (in terms of quantities in a well defined relation to the primary limit as operational quantities) Derived limits (related to a primary or secondary by a defined model) Authorized and Operational limits are not different in nature to derived limits This hierarchy seems of formal nature only and has apparently no consequences in most cases. However, one application is important because this hierarchy is not actually taken into account yet. This is when compliance with the limit is to be demonstratred by measurements, as always required by the authorities. As only for a primary limit the rather strict condition that a value must not exceeded apply, the question of the relation of the result and his uncertainty with the limit has will be discussed here. In all other cases, the (secondary or derived, lower than primary in the hierarchy) limit is related with the primary limit by a model, where an example for the most simple form is dose [µsv] = dose rate[µsv/h]. time [h] (1) In this case, the quantity time, to be understood as occupancy time at the site of interest has the same importance than the radiological quantity dose rate for the result. This implies that the data quality of the figure (i.e. mean and uncertainty) have to be equivalent. In addition, a conservatism in applying this quantity time is not appropriate, as the risk factors are based upon the LNT- Hypothesis, and hence the limit numbers are already conservative. Another example is shown in equ. (2) for internal exposure, where intake is the product of activity concentration and consumption. Radiation protection standards set in force relations between limiting issues as dose limits and the recommended application of assessed issues as results of direct dosimetric measurements, but also as auxiliary quantities as activity concentration, dose rate, etc. Standards are used to give figures as limits in a very strict form, eg. as an even number with one significant digit and an uncertainty zero by definition. No information on acceptable uncertainty is given, but if one follows common scientific practice, the number of significant digits can be used as a hint for acceptable uncertainty. The annual dose limit is given as 1 msv/a and not as 1,00 msv/a. Both expressions are identical in number, but different in acceptable uncertainty. Examples of this kind are the annual dose limit in terms of effective dose, of occupationally exposed persons of 20 msv in a year, and 1 msv in a year for the population. International standards [2] address the question of verification, but this is done only in terms of numbers, but not in uncertainty. 2. Dose Assessment, Verification of compliance with a limit Compliance with the limit is proved [1] by comparison of the total effective dose ET from all pathways by E T = H p (d)+ h(g ) I j,ing + h(g ) I j, inh (2) j j,ing j j,inh where: Hp(d) is the personal dose equivalent from penetrating radiation during the year; h(g)j,ing and h(g)j,inh are the committed effective dose per ingested or inhaled unit intake for radionuclide j by the group of age g; and Ij,ing and Ij,inh respectively are the intake via ingestion or inhalation of radionuclide j during the same period. This relation is linking the limiting quantity E T and operational quantities to be used to estimates done by assessment and measurements. Numbers in the limiting quantity has by definition no uncertainty, numbers assessed by operational quantities have unavoidable uncertainties. As multiple assessments change the total uncertainty, the number of measurements has to be taken into account. As reference time interval, all primary parameters refer to a time interval of one year. No information what uncertainty might be acceptable and specification of assessment are included. 2
3. Acceptable uncertainty and choice of quantities Some problems arise when inappropriate quantities are used and uncertainty is not considered for proving compliance. The question what uncertainty is deemed acceptable can not be solved in general, as limits are different in hierarchy as shown above. Additional confusion was introduced by the use of different kinds of quantities in limitation of dose. Based upon [4], [2] and [5], it is defined for members of the public: The estimated average doses to the relevant critical group of members of the public shall not exceed the limit of 1 msv/a, effective dose. This implies that the dose clearly refers to individuals. In technical Standards [6], however a term Ortsdosis (dose at a certain site) is coined and defined as: Equivalent dose, measured at a certain site" [7]. This quantity is clearly a secondary quantity and adopted in [6] and used as limiting quantity to prove compliance with the primary limit. The Austrian Radiation Protection ordinance will probably also follow the procedure. The development is because previous standards made no distinction between operational (i.e. secondary) and limiting (primary) quantities, and this is maintained. In addition, also used since many years, the limiting quantity as used is no primary limit (annual dose) but a derived limit (dose per week). The development is not in line with recent standards[5], because the effective dose is used correctly expressed as primary limit, comprising of external exposure, and internal exposure by inhalation and ingestion. The effective dose applies clearly to an individual. Technical standards, however, use an operational quantities, which apply by definition a secondary limit for limitation. This lead to the situation that a figure to be measured (operational quantity, by definition assessed retrospectivly) is used for prospective design (limiting quantity) of protective measures of an installation a dose of a individual is changed to a dose at a certain site, which in turn is biasing optimisation (see below) Technical standards do therefore apply quantities in some cases in another sense then initially intended 4. Possible relations between measured figures and limiting numbers 4.1.1 General The following issues have be checked before assessement: is the figure subject of comparison identical in nature to the figure to be compared? is the reference interval in assessment identical with the reference interval in standards? When the operational quantity is in a proved relation with the limiting quantity and the reference interval is identical (e.g. a measuring interval of one year and annual dose), a direct comparison is possible. If different conditions apply (multiple measurement as monthly in personal dose assessment, measuring of a derived quantity as activity concentration), exposure models have to be taken into account. 4.1.2 Possible relations of M with the limiting number L are It is assumed that a given measured figure, firstly unspecified and indicated by the symbol M has an uncertainty of ± σ. Possible relations of M with the limiting number L are: M + σ L (3) M L, M + σ L (4) M > L, M - σ L (5) 3
When condition (5) apply, compliance is not proved. This condition is therefore not considered below. 4.1.3 Importance of model When only deterministic models apply, the following basic operations have to be taken into account. The models may include * multiplication (example: dose rate and time) * addition (Σ of monthly doses ) * subtraction ( background) The total uncertainty is, applying error propagation, coined by the largest individual uncertainty. A reduction of the uncertainty of a single parameter does not improve the total accuracy. Table 1 shows some examples. Nature of limitation A Primary B Secondary Quantity subject of assessment Effective dose E L Operational quantity Property Annual individual dose not direct measurable by definition proved relation to E subject of assessment examples * individual dose of occupationally exposed persons * Dose in critical group of the population H p (10) C Derived Quantity related to A via model and default parameters dose at a certain site H * (10) in a defined interval H p (10) per month Dose rate activity concentration of radionuclide i D Authorized E Operational F Guidance level D < A,B,C E < D Quantity, when exceeded, appropriate actions should be considered [2] Different definitions for medical exposure and general application Table 1: some examples for different limits One can easily see that with increasing departure from the primary limit model assumptions are becoming more prominent for the properties of the result than assessed or measured figures. 4.2 Examples 4.2.1 External individual dose- measurement of the dose of an occupationally exposed person. The primary limit is expressed as E (effective dose ) of 20 msv per year. Compliance is proved by E T = H p (d)+ j h(g ) j,ing I j,ing + j h(g ) j,inh I j, inh where I j,ing and I j,inh are zero. 4
The secondary limit for penetrating radiation is expressed as H p (10). The measurements are usually carried out as with a measuring interval of one month, and the acceptable uncertainty σ is limited by standardisation procedure of the measuring procedure[8]. Compliance is proved when apply a) as number by 20 msv as H p (10) Σ H p (10) i, where H p (10) i dose per month, and i. 1 to 12 b) as uncertainty of the annual dose by σ = (H p (10) 2 1 +...+ H p (10) 2 12) 1/2. This is because the individual uncertainties of the different monthly measurements are distributed at random under normal conditions. When the limit is approached, it has to be taken into account that both the number and the uncertainty of the measurement are further enlarged by the natural background. There is compliance with the limit when condition (4) apply: Σ H p (10) i E even when Σ H p (10) i + {H p (10) 2 1 +...+ H p (10) 2 12} 1/2 E 4.2.2 Assessment of ambient equivalent dose at a certain site by dose rate measurements The limitation is proved by H*(10)[µSv/h]. residence time t [h] H*(10) E σ of H*(10) is limited by standardisation of the equipment [8]. As the residence time t of a certain individual at the site under consideration in one year is not known prospectively, σ(t) is also not known. The most conservative approach is one year. Regarding the importance of the uncertainty of the individual parameters for the total uncertainty of the annual dose, the uncertainty of the dose rate measurement can be neglected. Standards, however, are limiting secondary derived quantity dose measured H*(10) at a certain site in terms of dose rate. This implies that, disregarding there uncertainty of time, the condition shown below apply. If this model is applied to prove the dose limit (e.g. the annual dose, being a primary limit) by accurate assessment of the derived limit: dose rate[µsv/h], or dose per week, the uncertainty of the result is governed by the prediction of time say 1 hour or 8766 h per year. To adopt a residence time of more then 8766 h per year is over- conservative, but even to take a full year is biasing optimisation. These facts implies that improvement of the uncertainty of the dose rate measurement will not improve the uncertainty of the annual dose at all. For these reason, the condition (8) apply for proving compliance: H*(10) E also even when H*(10) + σ H*(10) E 4.2.3 Assessment of activity concentration 4.2.3.1 General At assessment of activity concentration in certain substances where samples (of air, water, soil etc have to be taken), two major contributors are to the total uncertainty of the result: the uncertainty of sampling and the uncertainty of measurement. This is discussed below. 5
4.2.3.2 Uncertainty of measurement At counting of the situation is different than above. Even with a standardised and well calibrated equipment (where this contribution in calibration to the total uncertainty can be disregarded), the counting interval is a parameter subject of selection. Therefore, the standard deviation σ of the measurement is not limited by the equipment as above, but specified by the counting conditions. Although the quantity activity concentration of a radionuclide i is again a derived quantity, it is frequently used for limitation. This might be irrational if a certain limit is taken for foodstuffs with different consumption (as spice or potatoes). Although the conditions are similar to 4.2.2, condition (3) applies [9] : A c,i + σ(a c,i ) L(A c,i ) In addition, an important implicit property of this condition is that a simple measurment with a large uncertainty is sufficent to prove compliance with the limit when the result of measurement is well below the limit, and the quality of the measurment has to be high if the limit is approached [9]. 4.2.3.3 Uncertainty of sampling Uncertainty of sampling is still not sufficiently considered in the development. It looks interesting that even very complete and well recognised manuals on measuring techniques for environmental radioactivity as [10] pay little attention to sampling.. a sample must be representative of the... material that is to be analysed. However, obtaining a..representative sample of environmental materials is often not straightforward (para. 1.6.1). However, multiple sampling is suggested (para. 2.1). The estimation of the variance of sampling is still difficult. The variance of sampling depends on the spatial and temporal distribution of the activity in the bulk material during the sampling period. For example, if one takes one liter of water every day from a river with a mean flow rate of 1000 m 3 /s, the monthly sample contains of about 10-11 of the total. This sample, being a very small fraction of the total will be representative only when the condition of spatial and temporal equilibrium applies in full. The assessment of the total uncertainty requires the assessment of the uncertainty of sampling. This can be done by consideration of the following scenarios, where two influences have to be taken into account: Spatial distribution of the radioactive material in both the sampling material and over the sampling area ( e.g. by different retention in sediments of low activity, different radionuclide deposition, very low activity in inhomogeneous retaining material ) Temporal distribution of the radioactive material ( e.g. by inhomogeneous releases or inhomogeneous dispersion of radionuclides in air). As it is not the purpose of radiation protection measurements to investigate details of radionuclide transport and retention, rather a statistical "black box" approach seems promising instead of detailed investigations. These conditions lead to the requirement that a certain number of independent samples has to investigated to take into account for possible spatial and temporal distribution of the radioactivity in the bulk material. The number of samples required is directed by the variation of the results and the acceptable uncertainty of the result (mean), the acceptable regression coefficient to assess variance and shape of the distribution. The acceptable uncertainty in turn depends on the purpose of the assessment. For example, it can be large of the purpose of the measurement is only to check the compliance with a limit and the measured figure is well below the limit. On the other hand, the uncertainty has to be small when the result is approaching the limit or the result has to be used for parameter assessment. In any case, if the result is based on a single sample, no total uncertainty can be given. If the uncertainty of sampling has to be considered, multiple sampling has to be carried out. The system of taking one single sample and to declare that this sample to be "representative" will no longer fulfil recent requirement of monitoring for dose assessment. Even in very simple cases, the minimum 6
number of samples to fit a distribution has to be proved. Dependent on circumstances, it might be assumed that in the order of about five samples should be a compromise to permit fitting of a distribution to express reasonable uncertainties. Multiple sampling and estimation of the distribution of the results rather than reporting a single figure will improve data quality and explain environmental variations. The capacity of doing multiple independent sampling and measurement is usually easily available by optimising sample preparation and counting conditions by choice a carefully selected and reasonable Lower Limit of Detection instead of using unjustified large sample volumes and long measuring intervals leading to data of no practical relevance in terms of dose. This question can be refined by taking into account the purpose of an assessment, see table 2. Purpose of the assessment Compliance with a derived limit c i [Bq/m 3 ] Compliance with a primary dose limit [µsv/a] Transfer parameter assessment e.g.[bq/kg per Bq/m 3 ] Variance of sampling dependent on Not Applicable Number of samples Importance of pathway and magnitude of transfer factor Variance of measurement dependent on Number of counts and LLD Number of counts, LLD, number of samples, assumptions on consumption Number of counts, LLD i number of samples Result Applying equ.(5) Applying equ.(4) Relation of activity concentration in both compartments Validity of the result For this very sample only Corresponding to area and period covered by samples Considered compartments and area and period covered by samples Table 2: Examples for different purposes of assessment It can be seen that only in the first case one sample is sufficient, and the result is valid for this very sample only. In all other cases, multiple sampling is required. Otherwise, all effort on quality control in measurements will be overruled by samplind uncertainties. 5. Conclusions The are still some problems in executing recent general developments and in radiation protection standards. Some are associated with performing measurements to demonstrate compliance in adequate form und to take into account uncertainty. The development in appropriate handling of new quantities is not yet terminated. This is because the new quantitites are simply replacing old quantitite without taking into account their different nature and properties. This is occasionally not in line with the conceptual basis [2] of radiological protection: * For administrative reasons, primary quantities are replaced by secondary or derived quantities because of easier assessment and monitoring. This has no conceptual basis in the dose range of stochastic effects. The compliance with primary quantities is often replaced by compliance with derived quantities. * Requirement for the accuracy of the assessment of other than primary quantities (dose per week, activity concentration) are enhanced although both prospective but also and retrospective estimates of contributing parameters (residence time, consumption) include large uncertainties by definition * It is not often distinguished between limiting and operational quantities. * A consistent appraoch to include uncertainties in standards is required to take full advantage of the development of the new quantitites * At present, the presentation of a result without uncertainty does not correspond recent developments. 7
References 1. Tschurlovits M., Taghizadegan R.,and Engelbrecht R.: Handling uncertainty and variability in risk communication, this proceedings 2. FAO/I/AEA/ILO/NEA/PAHO/WHO International Basic Safety Standards for Protection against ionizing Radiation and for the Safety of Radiation Sources Safety Series No. 115, IAEA 1996 3. IAEA Safety Series 76 Radiation Protection Glossary 1986 4. International Commission on Radiological Protection (ICRP) 1990 recommendations 5. Council directive 96/29/ Euratom 13.5.96 laying down basic safety standards for the protection of the health of workers and the general public against the dangers of ionizing radiation 6. Austrian Standard ÖNORM S 5214 7. German Standard DIN 6814 8. Austrian Standard ÖNORM S 5255 9. Austrian Standard ÖNORM S 5250 10. N.A.Chieko et al, Eds: EML Procedures Manual, (known as HASL 300) 8