Full text of DOI:10.1093/oxfordjournals.rpd.a033158

Radiation Protection Dosimetry
Vol. 90, No. 3, pp. 345–352 (2000)
Nuclear Technology Publishing
PROBABILISTIC ACCIDENT CONSEQUENCE UNCERTAINTY
ANALYSIS OF THE FOOD CHAIN MODULE IN THE
COSYMA PACKAGE
J. Brown and J. A. Jones
National Radiological Protection Board
Chilton, Didcot, Oxon, UK
INVITED PAPER
Abstract — This paper describes the uncertainty analysis of the food chain module of COSYMA and the uncertainty distributions
on the input parameter values for the food chain model provided by the expert panels that were used for the analysis. Two expert
panels were convened, covering the areas of soil and plant transfer processes and transfer to and through animals. The aggregated
uncertainty distributions from the experts for the elicited variables were used in an uncertainty analysis of the food chain module
of COSYMA. The main aim of the module analysis was to identify those parameters whose uncertainty makes large contributions
to the overall uncertainty and so should be included in the overall analysis.
INTRODUCTION
FOOD CHAIN MODELS USED IN COSYMA
COSYMA requires information on the activity con-
centrations in foods as a function of time after an acci-
dent. It does not include a food chain model, but uses
the results of such models in data libraries. Two data
libraries have been produced for COSYMA using the
UK NRPB food chain models, FARMLAND(5) and the
model of GSF, the Research Centre for Environment
and Health in Germany, ECOSYS(6). The models
require a considerable input of basic food chain data,
for example, the interception of activity by crops, root
uptake concentration factors and transfer factors from
animal feed to animal products. The libraries have been
created using agreed values for the food chain para-
meters used in the models for application within the
European Union (EU)(7). There are, however, differ-
ences between the libraries in the predicted activity con-
centrations due to the different foods considered in the
models and some differences in modelling approach.
The foods that can be considered within COSYMA for
both models are green vegetables, grain products,
potatoes and other root vegetables, milk and meat from
cattle and pork. In addition, the FARMLAND data
library contains activity concentrations for meat and
liver from sheep and liver from cattle and the ECOSYS
data library contains data for non-leafy vegetables. The
activity concentrations in foods depend on the time of
year at which deposition occurs. Both FARMLAND and
ECOSYS take into account agricultural practices in the
modelling of the transfer of radionuclides to foods. Data
libraries are produced for deposition occurring at two
times of the year, reflecting winter when relatively few
crops are grown and livestock is generally housed
indoors, and summer when it is the height of the grow-
ing season prior to harvest and animals are typically
Computer codes developed in the European Union
(COSYMA)(1) and the USA (MACCS)(2) are used in
probabilistic accident consequence assessments to
predict the consequences of radionuclide releases
from nuclear installations in terms of doses and risks
to individuals and populations and the effect of
countermeasures. Both the European Commission
(EC) and the United States Nuclear Regulatory Com-
mission (USNRC) recognised the importance of
quantifying uncertainties in these predictions and
have undertaken a joint analysis.
Formal expert judgement techniques were applied to
obtain probability distributions on parameter values for
the codes. Separate panels of experts were convened to
address uncertainties in atmospheric dispersion and
deposition, external doses, food chain transfer, internal
dosimetry, and early and late health effects.
Probability distributions on code input parameters
were used to determine overall uncertainties in the
COSYMA code predictions in two stages. First, sensi-
tivity analyses were undertaken on the four main
modules of the code to identify those model para-
meters whose uncertainty makes a major contribution
to the uncertainties on the code endpoints. Second,
the identified dominating uncertainties were used in
an overall analysis of the uncertainty in the whole
system.
A description is provided of the expert panels con-
vened to elicit uncertainties on the parameters rep-
resenting transfer through the terrestrial food chain(3), a
summary of the results from the panels and a summary
of the analysis of the uncertainty on the food chain
transfer module within COSYMA(4)
.
345

J. BROWN and J. A. JONES
grazing outdoors. Agreed agricultural practices are used The experts were asked to take a number of factors
in the two models that are appropriate for the EU(7)
into account when determining the uncertainties on the
.
COSYMA contains both data libraries and, within a run elicitation variables. Within the soil and plant processes
of the system, the use of the ‘summer’ or ‘winter’ data panel, the experts were asked to consider the compo-
library is chosen depending on the date of the year in sition of both a generic soil and specific soil types, aver-
the meteorological data file being used. The aim of the age weather conditions (including rainfall patterns and
study was to use the COSYMA food chain model to intensities), the physical form of the deposited material,
represent the uncertainty given by the experts. The
results of the study will be similar if either the ECOSYS
or FARMLAND model is used. For convenience, the
study used the FARMLAND model.
Table 2. Summary of elicitation variables for which
distributions were obtained from the expert panels.
Soil and plant processes panel
EXPERT PANELS FOR ELICITATION OF FOOD
CHAIN PARAMETERS
(1) Soil migration: time taken for 50% of initial deposit to
migrate below depths of 1, 5, 15 and 30 cm for a range
of soil types.
(2) Fixation of caesium and strontium in soil: fraction that
becomes unavailable for uptake by plants at a series of
times following deposition.
Two panels of experts were used to obtain uncertainty
distributions on food chain transfer covering the areas
of soil and plant transfer processes and transfer to and
through animals. The members of the two panels are (3) Root uptake concentration factors: activity concentrations
of caesium and strontium in a range of crops at maturity
which are grown on a range of soils containing 1
Bq.kg−1. Elicited as a function of time following
deposition.
listed in Table 1 and were drawn from Europe and the
US. Within the panels, the most important processes in
determining the transfer of radionuclides through
terrestrial foodchains were considered for the elements
strontium, caesium and iodine. The quantities for which
the experts gave distributions (elicitation variables) are
summarised in Table 2.
Due to the diverse nature of agricultural areas and
practices and the diversity of land types, generic distri-
butions obtained for food chain parameters will not
necessarily be applicable to all European and US
regions. The regions of interest were therefore restricted
to those typical of warm temperate climates, for
example north-western Europe and the north-eastern
and south-eastern areas of the US. Conditions specific
to Mediterranean countries, arid areas of the US and
areas subject to arctic conditions were excluded. The
(4) Interception factor: fraction of total deposition that is
intercepted by the plant surface for a range of crops at
maturity.
(5) Resuspension factor: resuspension factor for typical
surface crop and pasture grass from resuspension of a
fresh deposit to soil by wind driven processes.
(6) Retention times: time taken for the original activity on
the plant to be reduced by 50% for a range of crops.
(7) Concentration in grain at harvest: activity concentration
in grain at harvest following deposition to the crop at a
number of times before harvest.
(8) Concentration in root vegetables at harvest: activity
concentration in root crops at harvest following
deposition to a number of times before harvest.
experts were asked to provide estimates of uncertainty Animal transfer processes panel
(1) Animals’ consumption rates: daily consumption rates of
applicable to the main agricultural production areas in
the regions of interest. Areas such as semi-natural
environments were only to be considered in so far as
they contributed to the food production in these regions.
a range of animal feeds by animals considered.
(2) Consumption rate of soil: daily consumption rate of soil
for animals considered.
(3) Availability of ingested feed: fraction of activity
associated with pasture grass available for transfer across
the gut for activity freshly deposited, biologically
incorporated and associated with soil.
Table 1. Food chain panel experts.
(4) Transfer to meat: fraction of the daily intake transferred
to a kg of meat in the animals considered at equilibrium
under constant feeding of activity for Cs and Sr.
(5) Transfer to eggs: fraction of the daily intake transferred
to a kg of eggs at equilibrium under constant feeding of
activity for Cs and Sr.
(6) Transfer to milk: the fraction of the daily intake
transferred to a litre of milk in the animals considered at
equilibrium under constant feeding of activity for Cs, Sr
and I.
Soil and plant processes
Animal processes
Expert
Expert
Country
Country
M. Frissel
J. Garland
Netherlands P. Coughtrey
UK
UK
France
US
UK
UK
F. Daburon
O. Hoffman
B. Howard
J. Pearce
R. Kirchmann Belgium
G. Proehl
G. Shaw
Germany
UK
W. Whicker
L. Anspaugh
US
US
P. Strand
C. Vandecasteele
G. Voigt
Norway
Belgium
Germany
US
(7) Biological half-life in animals: weighted average
residence time of Sr, Cs and I in meat of the animals
considered after equilibrium in the animal has been
reached.
G. Ward
346

THE FOOD CHAIN MODULE
and also any influence of crop types and varieties, agri- percentile is close to unity, the 50th percentile values
cultural practices and crop yields. Within the animal are close to 0.4, and the 5th percentile values close to
processes panel, the experts were asked to consider a 0.06. The aggregated resuspension factors give rise to
number of factors which may affect the consumption large uncertainty bands with a range factor of about
rate of feeds by animals and the subsequent transfer of 10,000, with the 50th percentile relatively close to the
activity to the animal products such as the age and 5th percentile. For the retention times, range factors of
weight of animals, milk yields and stages of lactation, the order of 20 are found for all crops. The aggregated
quality of animal feeds, physical form of the deposited data for the concentrations in grain at harvest show
material and soil types. The experts were then allowed similar results for both strontium and caesium with
to use any information sources to assist them in range factors varying from 90 to 500. The smaller range
developing their distributions, such as experimental factors are seen for deposition at 30 days and 60 days
data, published literature and analytical techniques.
before harvest for caesium and at 15 days and 30 days
The experts used a wide range of information, draw- before harvest for strontium. For concentrations in root
ing on an international pool of published data on trans- crops at harvest the range factors are much larger than
fer in terrestrial food chains, their own experimental those seen for grain, with the range factors for strontium
data and personal experience. In general, the experts did significantly smaller than those observed for caesium;
not use models to obtain the uncertainty distributions, the range factors are smallest for deposition occurring
with the exception of those on soil migration. Here, a 60 days and 90 days before harvest for both elements.
number of analytical approaches were adopted based on
For the animals’ consumption rates, many of the
published methods or those developed by the experts experts provided uncertainty estimates based on experi-
based on their own experimental data and field obser- ence in their own countries. Aggregation of these data,
vations. Further details of the rationales adopted by the in most cases, was considered appropriate to give gen-
experts can be found in Reference 3.
eric values for European conditions. In general, the
range factors observed across the range of animals and
feeds were less than 20, varying from a factor of 4 for
the consumption of pasture grass by dairy cows and
RESULTS OF EXPERT JUDGEMENT EXERCISE
This section presents a selection of the uncertainty sheep to 20 for consumption of cereals by pigs. The
distributions obtained from the expert panels. Following range factor on the consumption of soil by animals
the expert elicitation, the 5th, 50th and 95th percentiles varied between 40 and 200, depending on the animal.
of the distributions of each elicited variable were
For the transfer to meat, milk and eggs, the largest
obtained from each expert. For each elicitation variable, range factors were generally observed for transfer to
these distributions were then combined into single dis- lamb, pork and chicken. The experts were least uncer-
tributions, with each expert’s distributions being given tain for caesium with range factors from 20 to 80. The
equal weight(3). The distributions discussed in this paper range factors for strontium were larger, particularly for
are these aggregations of the individual elicited distri- meat, where they varied from 200 for dairy cows to
butions for each elicited parameter. Most of the results 2000 for poultry. The transfer of iodine to eggs and
are discussed in terms of a range factor, which is the sheep milk is also very uncertain with range factors of
ratio between the 95th and 5th percentiles of the aggre- about 1100 and 600, respectively, being observed. For
gated distributions.
the biological half-lives in meat, for the animal species
For soil migration parameters, the aggregated range considered, the aggregated range factors were smallest
factors vary from about 20 to 80, increasing with for caesium, ranging from 10 to 30, and highest for
migration below greater depths, and are smaller for cae- strontium, ranging from 500 to 800. The range for iod-
sium than for strontium, typically by about a factor of ine was 200–500.
2 for generic soils. For fixation in soil, the range factors
A comparison of the aggregated distributions
for caesium and strontium are similar, ranging from 3 obtained from expert elicitation has been made with the
to 60, being largest for organic soils, but are typically default values currently used in COSYMA in the
less than 10 for times greater than 1 year following FARMLAND model(5). Also, where possible, uncer-
deposition. The range factors observed for root uptake tainty distributions used in past uncertainty studies
concentrations for caesium and strontium for all crops carried out with FARMLAND and ECOSYS have
were broadly similar and typically in the range of 50– been made(4)
200. The exceptions were those for caesium uptake by
.
In general, the default FARMLAND values lie fairly
root vegetables, which were between 400 and 4000, close to the 50th percentile of the expert aggregated dis-
depending on the time after deposition. The overall tribution, typically between the 35th and 65th percen-
trend was that the range factors for root uptake of stron- tiles. As an example, Table 3 shows the comparison for
tium were smaller than those for caesium for all crops, the root uptake concentration factors for cereals, where
typically by a factor of 2. The aggregation of data for it can be seen that there is good agreement between the
interception factors results in range factors of about 50th percentile values for a generic soil and the model
10–40 for all crops. In all cases, the aggregated 95th default. Notable exceptions are the transfer of strontium
347

J. BROWN and J. A. JONES
to beef, as illustrated in Table 4, where the default value 50th percentiles of the aggregated expert data, the
is near the 5th percentile of the distribution, and the maximum value and 99.9th percentile value for the
consumption rate of cereals by pigs and the subsequent ECOSYS and FARMLAND studies, respectively, also
transfer of strontium to pork, where the defaults again both fall below the 50th percentile of the aggregated
lie near the 5th percentiles of the aggregated distri- expert distribution.
butions.
In comparing the aggregated distributions with uncer-
tainty distributions used in past uncertainty studies, only
PROCESSING OF ELICITED VARIABLES TO
general observations can be made, as the percentiles of
MODEL INPUT PARAMETERS
the distributions used in the three studies are different.
In general, the uncertainty factors in both of the past
The distributions on the important parameters govern-
studies are smaller than those observed from the aggre- ing transfer of iodine, caesium and strontium through
gated distribution, particularly in the ECOSYS study. the food chain have been derived from the experts’
For illustration, two examples are given here. For root aggregated parameter distributions. Distributions on
uptake concentration factors for cereals (Table 3), the other parameters used within the FARMLAND model
distributions from the earlier ECOSYS and FARM- have been determined by project staff based on reviews
LAND studies both lie within the 5th and 95th percen- of the literature and previous uncertainty analyses on
tiles of those from this study. However, for strontium food chain parameters. Although unlikely to contribute
transfer to beef, Table 4 shows that, as well as the significantly to the overall uncertainty, these parameters
default values of the models lying between the 5th and are included for completeness and to ensure that the
Table 3. Comparison of values for soil–plant concentration ratio for cereals from aggregated expert distributions with
those used in COSYMA and from past uncertainty studies.
Element
Concentration ratio (Bq.kg−1 fresh mass plant/Bq.kg−1 dry mass soil)
Aggregated expert distributions(a)
FARMLAND
Default(b)
ECOSYS
study(c)
FARMLAND study(d)
5%
50%
95%
Sr
Cs
1.1E-2
7.5E-4
1.4E-1
1.6E-2
1.2
1.8E-1
2E-1
1E-2
6E-2–6E-1
8E-3–5E-2
2E-2–6.3E-2–2E-1
3E-3–1.2E-2–5E-2
Notes:
(a)
Value for a generic soil, 3 years following deposition.
Default value for a generic soil used in model.
Minimum and maximum values of log-uniform distribution.
0.1–50–99.9 percentiles of log-normal distribution.
(b)
(c)
(d)
Table 4. Comparison of values for F(fa) for beef cattle and dairy cows from aggregated expert distributions with those
used in COSYMA and from past uncertainty studies.
(a)
Element
F(fa) (d.kg−1
)
Aggregated expert distributions
FARMLAND
default(b)
ECOSYS study(c) FARMLAND study(d)
5%
50%
95%
Sr: dairy cows
Sr: beef cattle
Cs: dairy cows
Cs: beef cattle
3.8E-5
1.6E-4
1.1E-3
3.1E-3
2.4E-3
4.8E-3
2.1E-2
4.0E-2
9.1E-3
6.2E-2
7.6E-2
9.1E-2
3E-4
3E-4
3E-2
3E-2
1E-4–5E-4
1E-4–5E-4
5E-3–5E-2
2E-2–6E-2
6E-5–1.9E-4–6E-4
7E-3–2.5E-2–9E-2
Notes:
(a)
Ff is the fraction of the daily intake transferred to a kg of meat at equilibrium under constant feeding of activity.
Default value used in model.
Minimum and maximum values of log-uniform distribution.
0.1–50–99.9 percentiles of log-normal distribution.
(b)
(c)
(d)
348

THE FOOD CHAIN MODULE
overall uncertainty in the predicted activity concen- ous quantities, where the probability reflects the occur-
trations is captured.
rence of different atmospheric conditions at the time of
Some of the required distributions for input the release. In this study, the CCDF is characterised by
parameters to FARMLAND could not be determined its mean value and the 95th and 99th percentiles. The
directly from the elicited variables. In this case methods uncertainty distributions resulting from parameter
were adopted to fit the required distributions on model uncertainty have been determined for these three charac-
parameters to distributions obtained from the expert teristic values of the CCDFs. The uncertainty distri-
elicitation, a method called probabilistic inversion in butions are represented by an ‘uncertainty factor’,
this study(8). Model parameters obtained in this way defined as the ratio of the 95th and 5th percentiles of
were the transfers within different depths of soil, and the uncertainty distribution. Also of interest is the ‘refer-
the transfer of activity deposited on cereal and root veg- ence uncertainty coefficient’, defined as the ratio of the
etable plants to the edible portion of the crop, namely 95th percentile of the uncertainty distribution to the cor-
the grain seed and the tuber. A comparison was made responding value of the CCDF predicted using the
between the distributions provided by the experts and default values for the parameters in the model. The
those implied by the distributions determined for the quantities considered are described in more detail in
model parameters. For soil migration, the model used in Reference 8.
the inversion process provided a good fit to the experts’
distributions for the times taken for activity to migrate
down the soil column. For the transfer of deposited
Extent of countermeasures
activity within cereals and root vegetables, a reasonable
The results for the uncertainty on the extent and dur-
fit was obtained for some of the percentiles and times ation of food restrictions are summarised in Table 5.
of deposition before harvest. However, in other cases For the CB2 source term, the uncertainty in the initial
the fit was poor; the model used substantially underesti- area restricted is similar for the different percentiles of
mated the values of activity concentrations at harvest the distribution. For milk, green vegetables and beef, the
provided by the experts, particularly for cereals. These uncertainty factor for the initial area restricted is
fits, however, represent the best that can be achieved between 5 and 70 with the lowest uncertainty associated
with the current methodology and have been used in the with green vegetables and milk. The uncertainty factors
study. A fuller description of the fit obtained using this are larger for the time integrated areas. For cereals, the
process is given in Reference 4.
uncertainty factor for an initial area is about 20,000;
this reflects the much larger uncertainty associated with
activity concentrations in cereals. For the DBA source
term, the uncertainties tend to be larger than those seen
FOOD CHAIN MODULE ANALYSIS
COSYMA gives information on a wide variety of for the CB2 source term. This is due to the activity con-
consequences of an accident. Uncertainty on the values centrations in the foods being close to the threshold at
of the parameters in the food chain model do not affect which restrictions are imposed, as would be expected
all of the COSYMA endpoints; the ones considered in based on the size of the release and subsequent lower
this module analysis can be summarised as follows:
activity concentrations in foods. The uncertainty factor
for the area with cereal restrictions for the DBA source
term is infinite, as the 5th percentiles of the uncertainty
distributions are zero. The areas affected are small, even
at the higher percentiles of the uncertainty distribution;
for example the 95th percentile of the uncertainty distri-
bution on the mean value of the initial area is 0.37 km2.
The reference uncertainty coefficient (not shown in the
Table) for the extent of milk restrictions for the CB2
(1) Individual and collective long term effective dose
and doses in bone marrow and thyroid.
(2) Individual and collective risks of fatal cancers (total
and from thyroid) and of leukaemia.
(3) The areas and their time integrals affected by food
restrictions, for meat, milk, green vegetables and
grain.
Two source terms, named CB2 and DBA, taken from source term, and for green vegetable restrictions for
a level 3 probabilistic risk assessment of the proposed both source terms, are greater than the uncertainty fac-
Hinkley Point PWR in the UK, were chosen for the food tor, showing that the reference CCDF is generally lower
chain module analysis. Full details of the source terms than the 5% envelope of the uncertainty distribution. For
are given elsewhere(9). CB2 is a medium sized release both source terms, the reference uncertainty coefficient
that makes a major contribution to the overall risk of for the extent of beef restrictions is much lower than
late health effects. DBA is a design basis accident.
the uncertainty factor, ie. the reference CCDF is near
the centre of the uncertainty band. For grain restrictions
the reference uncertainty coefficient is small (generally
less than 2) indicating the reference CCDF is very near
Results
In the absence of any uncertainty, the results of the upper end of the uncertainty range for this endpoint.
COSYMA are presented in terms of complementary The parameters whose uncertainties make major con-
cumulative distribution functions (CCDFs) of the vari- tributions to the overall uncertainty on the areas of food
349

J. BROWN and J. A. JONES
restrictions are similar for the two source terms. How- taken and for the situation where no countermeasures
ever, parameters for caesium are more important than were considered. In general, the uncertainty observed in
those for iodine for the CB2 source term, while the the numbers of health effects reflects the uncertainty in
reverse is the case for the DBA source term (this reflects the individual doses. The uncertainty factors for
the relative amounts of each released in the two source numbers of fatal cancers can be taken as indicative of
terms). The important contributors to the uncertainty for those obtained for all the health effects considered and
milk and beef are the interception onto pasture and the are given in Table 6. The table shows that, for the CB2
transfer of caesium to milk and meat, respectively. Also, source term, the uncertainty with countermeasures
the half-life of retention on hay and silage is important implemented is much lower than that observed when no
for the time integral of the area for beef. For cereals, countermeasures are considered. This reflects the fact
the parameters contributing significantly to the uncer- that most of the food is banned and so the influence of
tainty are the interception and retention of caesium on the uncertainty on the activity concentrations in food is
the cereal plant and the effectiveness of processing of small. The uncertainty factors for the DBA source term,
the cereal seed into flour. For green vegetables, soil con- if countermeasures are implemented, are larger than
tamination of the vegetable is the most important con- those observed for the CB2 source term. This is because
tributor to the uncertainty for food restriction areas, the extent of countermeasures is much reduced and so
reflecting the assumption in COSYMA that food restric- ingestion of contaminated foods contributes signifi-
tions would be imposed on food as produced rather than cantly to the longer term doses. The uncertainty
after preparation prior to marketing. For the DBA presented therefore largely reflects the uncertainty asso-
release, the uncertainty on parameters for iodine transfer ciated with the ingestion dose. For the CB2 source term,
to milk also contribute significantly to the overall uncer- the reference uncertainty coefficients for the numbers of
tainty to the extent of restrictions for milk.
health effects are about a factor of 2 smaller than the
uncertainty factors, as shown in Table 6, indicating that
the default values lie within the uncertainty range for
these endpoints. The study did not consider the uncer-
Individual doses and numbers of health effects
The uncertainty on individual doses and health effects tainty on the numbers of health effects for the DBA
was considered both for when countermeasures are source term in the absence of countermeasures.
Table 5. Uncertainty factors for the extent and duration of food restrictions.
Initial area
Time integral of area
Mean value 95th percentile 99th percentile
Mean value 95th percentile 99th percentile
Bans for CB2:
Milk
Grain
Green vegetables
Beef
7.3
25000
5
9.1
22000
7.2
8.5
15000
7.6
26
7710
56
26
5500
71
23
3600
68
37
40
47
58
53
53
Bans for DBA:
Milk
Green vegetables
Beef
580
74
450
510
62
390
290
47
220
180
74
310
130
62
240
110
47
220
Table 6. Uncertainty factors and reference uncertainty coefficients for numbers of fatal cancers.
Number of fatal cancers
Uncertainty factor
Reference uncertainty coefficient
Mean 95th percentile 99th percentile
value
Mean 95th percentile 99th percentile
value
CB2: With countermeasures
CB2: No countermeasures
considered
3.7
10
6.2
23
6.5
25
1.5
10
2.1
10
2.0
11
DBA: With countermeasures
6.3
13
14
6.3
7.1
7.2
350

THE FOOD CHAIN MODULE
The parameters whose uncertainties make large con- butions on the input parameters for a food chain model
tributions to the overall uncertainty on the numbers of were obtained using formal techniques of expert judge-
health effects, both for DBA and for CB2 when counter- ment. Two panels of experts were convened to address
measures are not considered, are the interception factor uncertainties associated with transfer within the terres-
for pasture, the retention time on hay and silage and the trial food chain. Elicitation questions were posed in such
transfer to meat for caesium and iodine. Rather different a way that the resulting data library of uncertainty distri-
parameter uncertainties are identified for the CB2 source butions can be used for many different uncertainty stud-
term if countermeasures are considered; here the most ies in the future, using different food chain models.
important uncertainties are those on the transfer to meat
The distributions have been used in an analysis of the
for caesium, the resuspension factor to crops and some uncertainty in the food chain module of COSYMA, the
of the parameters describing the transfer of caesium in main aim of which was to identify those uncertainties
soil and its root uptake to pasture.
which make major contributions to the overall uncer-
tainty on the predicted consequences. Results for two
source terms are presented here. In the absence of
results for a much wider range of source terms, care
Parameters included in the overall analysis
The parameters included in the final analysis were should be taken in generalising these results, particularly
those that were placed in the first or second ranks those associated with threshold criteria, such as food
according to the value of their partial rank correlation restrictions.
coefficient or those that make more than 15% contri-
For the source terms considered in the study, the most
bution to the normalised R2 value, for at least one important uncertainties are those related to the concen-
endpoint and one source term, as described in Reference trations of iodine and caesium, particularly in milk. The
8. The parameters are listed in Table 7. The inclusion uncertainties on the areas affected by food bans range
of the transfer of silver to liver of dairy cows was not from about a factor of 5 to more than 20,000 for restric-
expected and the reason for its importance is undergoing tions on cereals due to the large uncertainty associated
investigation. Since some of the identified parameters with activity concentrations in cereals. The uncertainty
influencing activity concentrations in food were compo- factors on the numbers of late health effects are typi-
nents of compartment models, all the parameters of the cally less than 10.
models involved were included in the overall analysis.
This ensured the distributions on the model outputs
were as consistent as possible with the distributions
ACKNOWLEDGEMENTS
specified by the experts.
The expert judgement parts of this project were partly
funded by the European Commission under contract
number FI3P-CT92-0023. The uncertainty analysis of
COSYMA was partly funded by the European Com-
CONCLUSIONS
This paper has described an exercise in which distri- mission under contract number FI4P-CT95-0006.
Table 7. Parameters included in the overall uncertainty analysis.
Selected input parameters
Biological half-life, dairy cows, I
Daily intake of hay/silage, dairy cows
Ff transfer to meat, beef cattle, Cs
Ff transfer to meat, dairy cows, Cs
Interception factor, cereals
Interception factor, hay/silage
Interception factor, pasture
Interception factor, potatoes
Retention time, green vegetables
Retention time, hay/silage
Root uptake, pasture, Cs
Soil migration, pasture, 3 migration rate
constants, Cs
Soil fixation pasture, Cs
Processing loss, cereals
Soil contamination, green vegetables
Translocation, cereals, Cs
Translocation, potatoes, Sr
Ff transfer to liver, dairy cows, Ag
Fm transfer to milk, dairy cows, Cs
Fm transfer to milk, dairy cows, I
Processing loss, green vegetables
Resuspension factor, pasture
Retention time, cereals, Cs
REFERENCES
1. COSYMA: A New Programme Package for Accident Consequence Assessment. Report EUR 13028 (Luxembourg: EC) (1991).
2. Chanin, D. I., Sprung, J. L., Ritchie, L. T. and Jow, H.-N. MELCOR Accident Consequence Code System (MACCS): User’s
Guide. Report NUREG/CR-4691 (Albuquerque, NM, USA) (1990).
3. Brown, J. and 10 others. Probabilistic Accident Consequence Uncertainty Analysis. Food Chain Uncertainty Assessment.
NUREG/CR-6523, EUR 16771, SAND97-0335 (Washington, DC, USA, and Brussels-Luxembourg) (1997).
351

J. BROWN and J. A. JONES
4. Brown, J., Jones, J. A., Fischer, F., Hasemann, I., Goossens, L. H. J., Kraan, B. C. P. and Cooke, R. M. Probabilistic Accident
Consequence Uncertainty Assessment Using COSYMA, Uncertainty from the Food Chain Module. Report EUR 18823
(Luxembourg: EC) (2000).
5. Brown, J. and Simmonds, J. R. FARMLAND: A Dynamic Model for the Transfer of Radionuclides through Terrestrial Food-
chains. NRPB-R273 (London: HMSO) (1995).
6. Muller, H. and Proehl, G. ECOSYS-87: A Dynamic Model for Assessing Radiological Consequences of Nuclear Accidents.
Health Phys. 64, 232–252 (1991).
7. CEC. Underlying Data for Derived Emergency Reference Levels, Post Chernobyl Action. Eds J. Sinnaeve and G. Gerber
(Luxembourg: CEC) EUR 12553 (1991).
8. Jones, J. A., Kraan, B. C. P., Cooke, R. M., Goossens, L. H. J., Fischer, F. and Hasemann, I. Probabilistic Accident Conse-
quence Uncertainty Assessment Using COSYMA, Methodology and Processing Techniques. Report EUR 18827 (Luxembourg:
EC) (2000).
9. Jones, J. A., Fischer, F., Hasemann, I., Goossens, L. H. J., Kraan, B. C. P. and Cooke, R. M. Probabilistic Accident Conse-
quence Uncertainty Assessment Using COSYMA, Overall Uncertainty Analysis. Report EUR 18826 (Luxembourg: EC) (2000).
352