Chairman F. LANGE
PROBABILISTIC SAFETY ASSESSMENT FOR THE TRANSPORT OF RADIOACTIVE WASTE
TO A UK REPOSITORY AT SELLAFIELD P.R. APPLETON
AEA Consultancy Services, Risley, United Kingdom
Abstract
A study has been undertaken to provide a detailed understanding of the radiological and non-radiological risks associated with the transport of radioactive waste from the sites at which waste is produced in the UK to a proposed deep repository at Sellafleld, and to ensure that these risks meet the design targets specified by Nirex. The routine transport collective dose to members of the public was assessed to be 0.2 man Sv per year, which is only about 0.004% of the natural background dose. Accident frequencies were calculated using event tree methodology. The radiological consequences of accidents were assessed using the probabilistic computer code CONDOR. The risk expectation value was calculated to be
1.5 x 10s - 8.6 x 10* latent cancer fatalities per year (depending on the transport mode scenario). These values are significantly lower than the corresponding predictions for non-radiological accident fatality rates, 0.05 - 0.035 fatalities per year. The non-radiological accident risk for the most exposed individual member of the public was assessed to be 5 x 10" - 1.7 x 10" per year, very much less than the Nirex target of S x 1O7 per year.
Plots of societal risk were shown to lie in the region of 'negligible risk', as defined by the UK Health and Safety Commission for non-radioactive dangerous goods transport.
1. INTRODUCTION
UK Nirex Ltd (Nirex) has been established to develop and operate a deep repository for the disposal of intermediate and low level waste (ILW and LLW) arising in the United Kingdom. Nirex is also responsible for producing standards for the design of waste packages, and for developing integrated transport arrangements for the movement of packaged ILW and LLW from UK waste producing sites to the repository. Nirex is concentrating its investigations on Sellafield in West Cumbria as a potential location for the deep repository.
This paper describes a study undertaken to provide a detailed understanding of the radiological and non-radiological risks associated with the transport of ILW and LLW to the proposed repository at Sellafield, and to ensure that these risks meet the design targets specified by Nirex as part of company policy.
Two other papers presented at this seminar describe the integrated transport arrangements [1] and the package designs [2] being planned by Nirex.
All waste must be transported in conformity with the IAEA Transport Regulations [3]
applicable at the time. This study and its subsequent phases will provide a useful basis when considering the safety aspects of any future changes in the Regulations.
The following sections of this paper describe:
• The input data (including the selection of representative waste streams)
• The calculation of routine transport doses
• The assumed response of the packages to impact and fire accident conditions
• The assessment of accident frequencies and radiological consequences
• Risks presented in several forms
« Planned future development work
• The main conclusions of the study.
2. INPUT DATA
Three classes of waste package were included in the assessment:
• Type B packages containing ILW (reusable shielded transport containers holding waste in four 500 litre drums)
• Industrial packages containing ILW
• Industrial packages containing LLW.
This represents a simplification of the set of packages currently envisaged [2], but the principal packages are covered. Also, accident response information is not yet available for all package designs, some of which are at an early conceptual stage.
Nirex policy is that rail transport shall be used wherever practical for the transport of waste to the repository, although it is not possible to utilise rail transport for all waste transport. Two transport mode scenarios were considered for assessment purposes:
• Rail-only: all packages would be transported by rail.
« Road/rail: all packages below the European Union (EU) weight limit of 40 tonnes for unrestricted road transport would be transported by road, and all heavier packages would be transported by rail.
Even in the rail-only scenario, some transport by special road vehicles will be required where there is no on-site railhead for the direct loading of heavy packages on to rail wagons. The risks of these short road journeys are not included in the assessment.
However, doses to workers involved in the road-to-rail trans-shipment operation have been included in the results of the rail-only scenario under 'Handling'.
A maximum waste disposal volume scenario was assumed, corresponding to 1240 ILW and 480 LLW packages to be transported per year [4]. These packages originate at nearly 30 sites in the UK and the corresponding annual transport distances are:
N)
ILW package km per year Road Rail
LLW package km per year Road Rail
Table 1 Routine Transport Collective Doses
Road/rail
In excess of 350 separate waste streams were identified [4] so it was necessary to group these and identify representative streams for detailed analysis. Two methods were adopted.
First, an approximate risk ranking parameter was computed for each waste stream.
This parameter was defined to include the important factors affecting transport risk, but was sufficiently simple to be evaluated for all the waste streams. The factors included were the package km, the quantity of activity (in Bq) per package, the fractional release in defined accident conditions, and a hazard index for the waste stream (based on the A? value [3]).
The absolute value of the parameter has no meaning, but the relative values can be used to rank the streams in terms of their importance as contributors to the total transport risk. It was found that the computed values of the parameter spanned eleven orders of magnitude for the 350 waste streams. About 30 streams gave rise to 99% of the total hazard index, so these formed the focus for further study.
Second, the streams were formed into groups on the basis of their origin and general characteristics. The following groups were identified:
Fuel dement debris
Plutonium contaminated material Ion exchange resins
Sludges and floes
Special wastes and miscellaneous contaminated items
Uncategorised (a small number of special, high activity wastes) Other ILW
LLW.
A single waste stream was selected to be representative of each of these groups, using the ranking data. Initially three streams were selected to represent all LLW, with one being further divided into combustible and non-combustible material. However, the differences in radionuclide composition and in the radiological consequences of releases for these streams were so small that finally only one stream was selected to represent all LLW, the stream giving the worst predicted consequences.
3. ROUTINE TRANSPORT DOSES
Routine transport doses were calculated using a methodology similar to that employed in the IAEA INTERTRAN computer code [5], but with changes to make the algorithms more appropriate for UK road and rail transport conditions.
Scenario,
An average ILW package external dose rate of 31 /iSv h ' at 2m from the surface was obtained from Nirex shielding calculations. An average external dose rate of 80 ^Sv h"' on the surface of LLW packages was obtained from available Nirex inventory data.
The resulting collective doses are shown in Table 1. The collective dose for members of the public was assessed to be about 0.2 man Sv per year for both road/rail and rail-only scenarios. Using a risk factor of 0.05 Sv1 this corresponds to an expectation value of 0.01 fatalities per year. For comparison, the collective dose to members of the public due to naturally-occurring sources of radiation along the transport routes was calculated to be 5500 man Sv per year. The additional collective dose due to the waste package transport therefore represents an increase of only 0.004%.
In addition, estimates of the maximum individual dose were made. Three hypothetical individuals were considered: a rail commuter regularly positioned on a station platform while waste packages passed by; a person living near traffic lights on the road approach to the repository; and a person living near a road-to-rail trans-shipment point. The maximum individual doses in all cases were estimated to be less than the Nirex target dose for members of the public of 0.05 mSv y1. However, it was recognised that further work is desirable to identify more ciosely the exposure times and distances for the critical groups.
4. PACKAGE ACCIDENT RESPONSE
The accident response of the waste packages was based on the IAEA Transport Regulations [3]. It was pessimistically assumed that the package containment would fail completely in accident conditions marginally more severe than those of the IAEA package tests (ie an impact more severe than 2.4 m s~' against an unyielding target or a non-trivial fire for an industrial package, and an impact more severe than 1 3 m s1 against an unyielding target and a 30 minute fully engulfing fire for a Type B package). This is a conservative assumption, for the design process is likely to result in packages with margins beyond these limits.
5. ACCIDENT FREQUENCIES
Accidents were identified which have the potential to exceed the IAEA Transport Regulation test conditions and therefore result in radiological consequences. These included:
Fall from a high bridge
Impact with a lineside or roadside object (tunnel abutment etc) Collision with a second train or road vehicle
Railhead transfer accident
Major fire (involving another vehicle carrying flammable material) Minor fire.
Eight accident condition categories were defined using impact and fire severity parameters. These categories covered all identified accident scenarios, including very low probability extreme conditions. The two most severe conditions considered were a fire in a tunnel involving a second train of flammable goods tankers, and impact against an unyielding target at 40 m s"'.
The historical lecord for the world-wide transport of radioactive materials is very good, so there are few instances of transport accidents with radiological consequences.
However, that presents difficulties for an assessment such as this. For example, no meaningful estimate of the probability of a high-speed impact of a waste package on a tunnel abutment can be derived from the fact that such an event has never occurred. Fortunately, event tree methodology can be used to estimate such probabilities, as described in the following paragraphs.
Event tree methodology involves dividing the accident development into a number of steps, beginning with an initiating event (such as derailment) and assigning probabilities for the severities of conditions (such as the speed) which are relevant to further steps.
Historical data for UK transport were used to derive initiating event probabilities as follows:
« A rail wagon derailment probability of 1.54 x 107 per wagon km for bogie freight wagons
• A rail (same-line) collision probability of 2.4 x 10"' per train km e A fatal or serious road accident probability of 7 x 10-3 per vehicle km
for motorways and 2.1 x 10-7 per vehicle km for major roads.
Additional information (speed distributions, fire probabilities, etc) was obtained from published and unpublished British Rail and UK Department of Transport sources. Data concerning specific hazards along the routes (eg location and heights of bridges, and nature of underlying surface) were obtained from detailed route and map surveys.
Probabilities for all the identified accident scenarios were developed in this way, and an example is shown below. Where data were uncertain, pessimistic values were adopted.
Probability of impact al 13-27 m s' with an unyielding tunnel abutment
= derailment probability (1.54 x 10"7 per wagon km)
x probability of sufficient wagon displacement from line of travel to strike the abutment (0.16)
x probability of presence of tunnel abutment at derailment location (0.0031)
x probability of abutment being effectively unyielding (0.72) x probability of wagon speed at derailment being in the range
13-27 m s'(0.11)
= 6 x Iff12 per wagon km.
Accident frequencies were simply obtained by multiplying these probabilities by the total distances travelled per year.
6. RADIOLOGICAL CONSEQUENCES OF ACCIDENTS 6.1 Release Fractions
The industrial packages and reusable Type B transport containers will form the outer containment boundary. In accidents slightly more severe than the IAEA Transport Regulation test requirements it was assumed that this boundary will fail. However, for the very strong Type B packages in particular, it is virtually inconceivable that the packages will burst open completely under accident conditions. Damage is likely to be confined to lid-body interface gaps opening up, except in the most extreme cases. Since the retention of the radioactive contents in an accident cannot be quantified at the present stage of package development and testing, no retention was assumed for the purpose of the assessment.
Therefore the results are probably quite pessimistic in this respect.
Within the ouler container containment boundary is the wasteform itself. Nirex has conducted programmes of work to investigate the impact response of immobilized ILW, including experiments, theoretical studies and literature reviews. Several different designs of 500 litre drum, all meeting the Nirex waste package specification but containing various inactive simulated wasteforms, have been dropped in different orientations from a range of heights. The drums were then subjected to detailed examination. Particle size analysis of the loose debris was undertaken using sieving, laser particle siring and aerosol analyser techniques. A distribution of results was obtained and the worst-case respirable release fraction (1.8 x 10"5 for a 9 m drop) was pessimistically applied to all the example ILW streams.
Nirex has also conducted a major programme of work to investigate the effects of heat on immobilized ILW. Small-scale and full-scale inactive simulant samples, and small-scale active samples, have been heated to 300°C and 1000°C. A computer heat transfer model to predict the temperature distribution in a 500 litre drum of immobilized waste has been developed and verified using the experimental data. In general it was found that the release fractions were dependent both on the radionuclide and on the type of wasteform. Example release fractions derived from this work for a two-hour 1000^ fire and employed in the safety assessment study are:
e 1.5 x Ifr" for Co-60 in fuel element debris
• 3 ? x 10* for Pu-239 in plutonium contaminated material
» 1.66 x 10° for Cs-137 in sludges.
For LLW, a respirable release fraction of 10"3 was adopted under impact conditions, based on a literature search and flowing-air entrainment data. These (hta did not include supercompacted waste; it is expected that most or all LLW will be supercompacted before transport to the repository, and that the resulting wasteform will have a much lower impact release fraction than assumed here.
For fires involving LLW, pessimistic respirable release factions were adopted for caesium, for other nuclides in non-combustible material and for other nuclides in combustible material, based on data in the literature. Since this study was completed Nirex has investigated the behaviour of supercompacted LLW in fires. The preliminary results indicate that the assumed release fractions are likely to overestimate the releases from supercompacted LLW by at least one or two orders of magnitude.
6.2 Dispersion and Health Effects Calculations
The radiological consequences of release were evaluated using the CONDOR computer code [6], This is a probabilistic consequence assessment code which was developed jointly in the UK by AEA Technology (SRD), Nuclear Electric and the National Radiological Protection Board. It models the downwind dispersion of released activity, taking account of dry and wet deposition processes, radioactive decay and any variation in the meteorological conditions. From the resulting distribution of the released material in the environment, the code evaluates the radiation doses to man via a number of exposure pathways: cloudshine, groundshine, inhalation, resuspension and ingestion of contaminated food.
CONDOR evaluates the radiological consequences for a large number of different meteorological sequences, in which the weather conditions change hourly, and calculates a consequence probability distribution. Hourly meteorological data from a representative site over an eight-year period were sampled to derive the input data for the assessment.
CONDOR calculations were run using the population distribution around three representative sites. The population distribution data were taken from the UK census, The three sites were chosen to be representative of 'urban', 'intermediate', and 'rural' locations along the waste transport routes. The population densities (in people km'2) for the respective sites were:
• 4165, 1433 and 118 averaged out to 1 km
« 5181, 1026, and 62 averaged out to 10km.
It was pessimistically assumed that no countermeasures would be taken, such as evacuation and food bans.
The calculations included both individual doses (doses which would be received by an individual who was located at a specified distance from the release) and societal doses (probabilistic distributions of frequency against dose for the exposed population).
For all the releases assessed, the results indicated no early fatalities, only risks of latent cancer deaths.
If a release were to occur, the wastestreams selected to be representative of plutonium contaminated material and uncategorised material (see Section 2) would give rise to the most significant radiological consequences. The inhalation exposure pathway led to nearly 99%
of the total predicted dose for the plutonium contaminated material (plutonium radionuclides being by far the most important in this stream). For the uncategorised material stream the groundshine pathway resulted in about 63% of the total dose, with 30% resulting from the ingestion pathway and almost all the remainder from inhalation. Caesium and cerium radionuclides were the most important in this stream.
Averaged over all weather conditions, the predicted individual jt a distance of 0.1 km from a release ranged rrom 9 mSv to 1 nSv for the different representative waste streams.
The expectation values of the societal consequence distributions, assuming a release to have occurred, were all less than one latent cancer fatality. They ranged from about 0.3 to less than 10"* fatalities for the different representative waste streams, the largest values being associated with releases in urban areas of high population density. For each stream and release location, the probability of ten or more latent cancer fatalities was less than If/', if a release were to occm.
7. RISKS
7.1 Introduction
The frequencies of potential accidents (see Section 5) and radiological consequences which would result if the accident occurred (see Section 6) were combined to evaluate risks.
Three separate risk measures were evaluated and compared with available criteria and levels of acceptance:
• The risk expectation value (the nsk derived from the average value of the societal consequence distribution)
• The individual risk (the risk to the most exposed hypothetical individual)
• The societal nsk (the probability distribution of frequency against number of fatal cancers in the exposed population).
7.2 Risk Expectation Value
The risk expectation value RB was calculated as follows:
RE = ^Ej^F.^N^
where F, j = frequency of each accident category i for each waste group j Pt = probability of population distribution in region k
N1Jt = number of latent cancer fatalities conditional upon accident category i for example wastestream j in region k.
Hence RH = 1.5 x 10~s latent cancer fatalities per year for the road/rail transport scenario, and 8.6 x 1&* for the rail-only scenario.
The plutonium contaminated material group provided the largest single contribution to the total predicted value of R„.
There is no UK or Nirex risk criterion for comparison with RB. However, comparisons with routine transport dose risks (see Section 3) and non-radiological transport accident risks help to place the radiological accident risk RB in perspective. Accidente will inevitably occur during the transport of any commodity, and UK statistics for fatal accidents in general freight were analysed to derive fatality frequencies for waste transport accidents in which radiation exposure is not a factor.
Road/rail scenario
Rail-only scenario Expected number of fatalities per year Non-radiological transport accidents 0.05 0.035 Routine transport radiation exposure 0.01 0.01 Radiation exposure in accidents 0.000015 0.0000086 RE provides a useful measure of the average risk. However, it does not differentiate between the contributions to the total risk from the higher consequence, lower frequency events and those from tower consequence, higher frequency events. Therefore other risk measures were also evaluated, as described in the following sections.
7.3 Individual Risk
The transport system was approximated by a straight line running the length of the
The transport system was approximated by a straight line running the length of the