Skip to main content



The project MARIA (Methods for Assessing the Radiological Impact of Accidents) was initiated in 1983 as part of the CEC Radiation Protection Research Programme. The aim of the project was, firstly, to review and then to develop further methods and models for accident consequence assessment (ACA) within the EC. A European ACA code, COSYMA (COde SYstem from MAria), is now available as a result of the many investigations performed within the MARIA project.

There remain a number of areas within the field of ACA modelling where further development and refinements are required, in addition the COSYMA code and its databases also require maintenance and some further development. These are the aims of the current extension of the MARIA project under the 1990 to 1991 Radiation Protection Programme.
It was the essential aim of the study to identify improved atmospheric dispersion models which can be applied in accident consequence assessment (ACA) codes as a substitute of the conventional straight line Gaussian plume model. Additionally, quantitative indications of the implications which different types of atmospheric dispersion models might have on the results of ACAs were given. The model has to be devised with the requirement that it uses real meteorological data extracted from routine observations which are recorded and reported continuously from meteorological stations. The only data needed by Gaussian like models are measurements of wind speed, wind direction, and precipitation intensity, and, in addition, the stability class which can be derived from synoptic observations.

For these purposes probabilistic comparative calculations have been performed with different kinds of atmospheric dispersion models on the basis of artificial source terms. The model types were:
a Gaussian model (MUSEMET) which can be applied to 3 different modes (the trajectory mode (referred to as mode 1) taking into account hourly changes of spatially homogeneous meteorological conditions) the straight line mode (mode 2) ignoring any information on the wind direction and the straight line model taking into account the wind direction prevailing at the beginning of hourly release phases (referred to as the Gaussian mode 3 model);
a Gaussian like trajectory puff model which is able to consider meteorological data which change temporally and spatially (RIMPUFF);
a Eulerian grid point model which solves the diffusion advection equation numerically and which is potentially able to consider explicitly the 3 dimensional inhomogeneous and instationary wind fields and turbulence fields in the atmosphere (TRANSLOC).

The main result of the study is that there are Gaussian like trajectory models available to substitute the straight line Gaussian model in ACAs. Under the assumptions of t he source terms considered in this study and dependent on their characteristics, quantitative differences which were more or less pronounced arose in the accident consequences due to the application of different atmospheric dispersion models. The study showed that using a trajectory model leads to a spatial distribution of activity over larger areas and an increase of the plume's residence time in the area over which it is dispersing combined with enhanced depletion and dilution by dry and wet deposition processes and tubulent diffusion. Therefore, the activity transported to farther distances may be significantly smaller.

It was a logical consequence from this study to apply trajectory models in the new program system UFOMOD for assessing the consequences of nuclear accidents. This led to a completely novel concept of atmospheric dispersion modelling in the new UFOMOD. Different ranges of validity are distinguished due to the facts that following:
strictly, Gaussian plume and puff models are only applicable for travel distances up to a few tens of kilometres and that they become increasingly inaccurate or unreliable at longer distances;
site specific characteristics are only relevant in the near range and vanish at farther distances;
the quality and quantity of consequences in the near range (fast protective measures, early health effects) are different from the far range (long term countermeasures, stochastic health effects);
the near range can be modelled much more in detail than the far range;
many applications of ACA refer to only 1 of both distance ranges.

The ranges of validity are then assigned to respective trajectory models:
the near range (less than or equal to 50 km), where modified versions of the atmospheric dispersion models MUSEMET and RIMPUFF are used;
the far range (greater than or equal to 50 km), where the computer code MESOS is applied. (It is a long range dispersion model simulating the transport of radioactive material over large areas up to thousands of kilometres which combines the requirement of short computing time with the ability to disperse radioactive material along precalculated wind fields derived from synoptic meteorological data measured within whole Europe).

UFOMOD and COSYMA (code system from MARIA (methods of assessing the radiological impact of accidents)): Atmospheric dispersion and deposition:
The application of different trajectory models with ranges of validity near to the site and at far distances, respectively, is a significant step forward to an appropriate treatment of site specific problems and questions arising in connection with the transportation of radioactive material over large land areas up to thousands of kilometres. The new structure of OFOMOD clearly reflects this problem oriented modelling by the division into 3 subsystems each built to assess accident consequences resulting from acute or chronic exposure.

The UFOMOD system contains 4 different models for atmospheric dispersion which are appropriate for different applications.

In the 2 near range versions UFOMOD/NE and UFOMOD/NL the segmented plume model MUSEMET and the puff model RIMPUFF can be applied. Both are Gaussian like trajectory models. MUSEMET, as the standard atmospheric dispersion model of the UFOMOD system, is implemented as a subroutine in the UFOMOD code. RIMPUFF is a stand alone code with an appropriate interface to UFOMOD. MUSEMET assumes that hourly meteorological data of 1 single meteorological station are known and that these data are constant in consecutive hourly time intervals. For the application in the far range where the site specific characteristics become less important, the MESOS model is available. It is a trajectory puff model which simulates the transport, dispersion, and deposition of airborne material up to distances of thousands of kilometers, thereby taking into account the changing local meteorological conditions obtained from a database of standard meteorolog ical observations from synoptic stations and ships every 3 hours across Europe.

In addition, the special straight line Gaussian model ISOLA has been implemented to estimate the spatial concentration distribution for low level long duration (weeks or months) releases of radioactive material.

To define starting times of weather sequences and the probabilities of occurrence of these sequences, it is convenient to apply stratified sampling. Therefore, a preprocessing program package called METSAM has been developed to perform the sampling. METSAM is designed in such a way that all possible weather situations are classified according to the initial wind direction, the time a plume need to leave a predefined area around the source, and the total amount of precipitation fallen during the dispersion. Afterwards, at least 1 weather sequence is sampled randomly from each class. This procedure ensures that the whole spectrum of weather conditions is taken into account.

In 1983, the European project MARIA (methods for assessing the radiological impact of accidents) was initiated as a part of the Commission of European Communities (CEC) Radiation Protection Research Programme with the aim to review and build on the nuclear accident consequence assessment (ACA) methods in use within the European community. The main contractors are the Kernforschungszentrum Karlsruhe (KfK, Germany) and the National Radiological Protection Board (NRPB, United Kingdom).
One of the objectives of the second period of the project (MARIA-2, 1985 to 1989) was to develop a joint accident consequence assessment computer program package, COSYMA (code system from MARIA) for use in the CEC, which should represent a fusion of ideas and models from the KfK program UFOMOD, the NRPB program MARC, of new developments within both institutions and of contributions from other MARIA contractors.

The general structure of the COSYMA program package was taken over from UFOMOD. It consists of 3 independent ACA programs for the assessment of the consequences of acute exposure in the near range (subsystem NE) and of long term exposure in the near range and in the far range, respectively (subsystems NL and FL).

Output from each subsystem gives individual organ doses and health effect risks, collective doses and the projected number of health effects. Special emphasis was given to the development of evaluation programs which allow a multitude of graphical and tabular representations of the results. Furthermore, 3 special evaluation programs were developed, which allow correlation of nuclide specific activity concentrations with the areas affected, individual organ doses with the number of individuals affected, and collective doses with individual doses. Presented graphically, such a form of results may be of considerable help in the analysis and interpretation of accident consequences especially at farther distances or for small source terms.

Health effects models in CO SYMA:
A new Health Effects Model for Nuclear Power Plant Accident Consequence Analysis and a new model for stochastic somatic health effects were developed forming a new health effects model for UFOMOD. In a slightly expanded form, this became the model in COSYMA, which is outlined below.

The models for the nonstochastic health effects are implemented in themodules for assessing individual risks of the subsystem NE. The health effects considered include all those causes for deaths for which a dose response relationship has been specified. However, it does not include all the nonfatal effects of irradiation which have been considered.

With respect to the stochastic somatic effects, both fatal and nonfatal effects are considered. Age dependent, sex dependent and time dependent dose risk coefficients of 10 fatal cancer types have been provided.

For the new program system UFOMOD, the concept of activity risk coefficients (ARC) has been developed to calculate the number of the fatal late health effects; this formalism is also used in COSYMA. These precalculated coefficients are normalised to the initial unit activity concentrations in the air or on ground, and contain all information of the age and lifetime distributions in the population, the time and age dependence of the intake of activity for internal exposure pathways, the time and age dependence of dose accumulations for all exposure pathways, and the time and age dependence of individual risk. During an ACA run, individual risks for fatal cancers are simply calculated by multiplication of these coefficients with the initial activity concentrations; an intermediate dose assessment step is not necessary.

Foodchain modelling in COSYMA:
In the models calculating foodchain related accident consequences in the subsystems NL and FL, 9 foodstuffs are currently taken into account: milk (including milk products), beef (including veal), pork, mutton, grain products, potatoes and 3 types of vegetables (leafy, no nleafy, root). The time dependent foodchain concentrations are implemented in COSYMA as precalculated data.
To account for seasonal effects, 2 such data sets are implemented in UFOMOD and COSYMA, 1 for a release on January 1st to represent an accident in winter (November to March) and 1 for a release on July 1st to represent an accident in summer (April to October). In the calculations, each weather sequence is, depending on it's real date, attributed either to the winter or to the summer period and the corresponding foodchain data set is selected by the program.

Tritium model:
Since 1988, a model is being developed at KfK to assess within the frame of an ACA the atmospheric transport/deposition and the radiation exposure of tritiated water (HTO), gaseous tritium (HT), and fission products from postulated airborne releases from hypothetical fusion reactors. The fission products have the same phycochemical and radiological properties as nuclides from fission reactors and can be treated in UFOMOD/COSYMA by simply adding the corresponding data to the data bases already installed. Tritium, however, is chemically identical to hydrogen and thus interacts directly with water and organic substances, making processes like conversion of HT to HTO, reemission after deposition, and the conversion of HTO into organically bound tritium (OBT) relevant, all of which may modify the total balance of the available HT or HTO inventories. The exposure pathways of significance for tritium are inhalation, ingestion and the radiation exposure from skin absorption, the latter playing no role for conventional radionuclides.

The consequences of a tritium release are commonly estimated in 2 separate ways. On the one hand, Gaussian (trajectory) dispersion models will describe all atmospheric transport and deposition processes which result in radiation exposure from inhalation and skin absorption. On the other hand, and separated from the dispersion process, compartment models will descr ibe the transfer of tritium through the foodchains and the resulting radiation exposure from ingestion.

KfK took a somewhat more realistic approach by coupling directly an atmospheric dispersion model describing the atmospheric transport and deposition processes under consideration of all relevant transfer processes in the environment (soil, plant and animal) for approximately 100 hours after the release event, when the atmospheric transport plays the dominant role, with a first order compartment model describing the transfer of tritium through the foodchains. The resulting computer code, UFOTRI (unfallfolgenmodell fuer tritiumfreisetzungen) will in future be used as a submodule in COSYMA.

Countermeasure models and criteria:
The modelling of countermeasures in COSYMA (code system from MARIA (methods of assessing the radiological impact of accidents)) has been extended in comparison to UFOMOD and MARC programs to allow the user considerable freedom in specifying a wide range of emergency actions and criteria at which these actions will be imposed and withdrawn, so that most of the recommendations and criteria adopted in different European community (EC) countries and some of those which may be suggested in future can be modelled. Some of the countermeasures can be initiated automatically in an area which can be defined by the user. In general, countermeasures are defined in the program on the basis of dose criteria. The corresponding doses are summed over specified exposure pathways, and the user has the option of selecting which of the following routes should be considered:
sheltering (can be initiated as a single countermeasure within a predefined circle or an area determined by an isodose line and the dose can result from any combination of the inhalation, deposited gamma and cloud gamma doses, with the periods over which the doses should be integrated specified by the user);
evacuation (considered to be an action aimed at reducing short term exposure, it can be initiated automatically in a geometrically defined area or on the basis of a dose criterion, and can be preceded by a period of sheltering);
relocation (considered to be an action aimed at reducing long term exposure and initiated only on a criterion based on the effective dose);
return from evacuation or relocation and decontamination (the same criterion is used for returning from evacuation and relocation and decontamination are considered together);
stable iodine tablets (the effect of taking stable iodine tablets on the thyroid dose from radioactive iodine depends on the time delay between inhaling the radioactive iodine and taking the stable iodine tablets);
Foodbans (can be impo sed or withdrawn on the basis of levels for activity levels in food or for individual doses).

Development of an ECONOMICS Module:
A joint model has been developed by the National Radiological Protection Board (NRPB) and Kernforschungszentrum Karlsruhe (KfK) within the MARIA programme to assess the economic consequences of accidental releases of radioactivity. The aim of this model is to assess in detail the economic costs of early and late health effects, as well as of different countermeasures that are considered in order to reduce the number of health effects in theaffected population. These countermeasures include sheltering, evacuation, relocation, food bans and decontamination.

The main aims of the investigations were to understand, apply and refine uncertainty and sensitivity analysis methods with respect to the program UFOMOD subsystem NE. This involves the following:
to get a deeper insight into the propagation of parameter uncertainties through its different modules (the atmospheric dispersion module, the module describing early protective actions, the module calculating short term organ doses, the healths effects module);
to quantify their contribution to the confidence bands of the intermediate and final results of an accident consequence assessment (ACA);
to combine the most important parameters from all submodule uncertainty analyses to a final overall uncertainty analysis.

Uncertainty/sensitivity analysis procedures:
Before starting the uncertainty and sensitivity analysis, a detailed discussion of the parameter uncertainty has to take place.
Having defined ranges and distributions for model parameters it is necessary to select specific values for each of the uncertain model parameters to be used in each run of UFOMOD (ie to have a suitable sampling scheme).
Each UFOMOD run produces 1 complementary cumulative frequency distribution (CCFD). Confidence bands have to be estimated for each set of CCFDs. The width of the bands is an indicator of the sensitivity of model predictions with respect to variations in parameters, which are imprecisely known.
To quantify the relative importance of the uncertain model parameters to the output of the accident consequences model some sensitivity measures are needed to 'rank' the parameters with respect to their influence on the consequences.
The partial (rank) correlation coefficient PCC or PRCC, respectively, are measures that explain the relation between a consequence variable and 1 or more model parameters. When a nonlinear relationship is involved it is often more revealing to calculate PCCs between variable ranks than between the actual values for the variables.

To summarise so me general findings with respect to sampling:
uncertainty (UA) and sensitivity analyses (SA) with random sampling (RS) and Latin hypercube sampling (LHS) do not lead to significantly different results;
sample sizes of 1.5 x model parameters are sufficient for UA;
to get statistically stable results, larger sample sizes are needed for SA;
with respect to distribution effects, for some distribution types there are differences between raw and rank values in the correlation matrices of the LHS code.

To summarise some general findings with respect to the evaluation of sensitive parameters:
not every PRCC value makes sense therefore significance tests have to be used;
a large absolute PRCC value is not in every case an indication for a considerable amount of responsibility for uncertainty in consequences therefore use PRCC values and coefficients of determination R{2};
in most cases the number of PRCCs, which are above the 'white noise level', increases with the sample size;
for the atmospheric dispersion submodule of UFOMOD the most important parameters were stable in their rankings regardless of the distributions (predefined by experts or all uniform, respectively) and for the less important parameters the PRCCs vary, but even then in most cases the ranking is the same.

Uncertainty/sensitivity analysis applications:
In a series of investigations with the mentioned submodules, a great deal of experience was gained with methods and evaluation techniques for uncertainty and sensitivity analyses. The influence on results of different sampling techniques (random, LHS) and sample sizes, parameter distributions and correlations could be quantified and the usefulness of sensitivity measures like PRCC and R{2} for the interpretation of results could be demonstrated.
CCFDs of the following quantities were evaluated:
air and ground activity concentrations of the 2 nuclides iodine-131 and caesium-137 and 3 distances (0.875 km, 4.9 km and 8.75 km);
individual bone marrow and lung doses at the same distances;
acute risks of haematopoietic and pulmonary syndrome at the same distances;
number of early health effects from haematopoietic and pulmonary syndrome.

The accident consequence assessment program package, COSYMA, has been modified following discussions with potential users and analysis of example runs. Special aspects were: the inclusion of hereditary effects and the quantification of loss of life expectancy in a simplified manner; the completion of the ingestion pathway with different options for calculating collective doses and for introduction of food bans; the preparation of the land and sea matrix; and a more flexible coding of counter measures and econonoc modelling. The corresponding documents, in particular the user guide, have been updated. An international code comparison exercise was initiated by the adhoc group on probalistic accident consequence assessment codes. A basic document has been prepared, containing the main objectives, the task specifications and the endpoints of the Benchmark calculations. An additional technical document has been prepared by the project management group with data and instructions needed when doing the code comparison calculations. In particular, it contains detailed descriptions of gridded data, meterological and economic data, source terms and result sheets for each consequence to be calculated.

Investigations are under way to determine the specifications of simplified versions of COSYMA for special purposes without significantly altering the model predicitons in the corresponding area of application.
Topics to be studied under the 1990 to 1991 MARIA project

This summary describes the work being undertaken as a part of the MARIA project by the Nuclear Research Centre Karlsruhe (KfK), FRG, and the National Radiological Protection Board (NRPB), UK.

COSYMA development and maintenance

The COSYMA system is an important tool for research and application, and as such requires continuous updating, testing and improvement. The COSYMA code will be maintained by KfK and NRPB throughout the period covered by the contract, and there will be further development of its databases by both organizations.

The increasing availability of small computers, and the need by inexperienced ACA code users of the COSYMA code as a tool, necessitate the development of a reduced and simpler version of the code. This version will be more limited in its flexibility, but will be consistent in its basic assumptions and data with the full COSYMA code, and will produce compatible results. This work will be undertaken largely at NRPB, in cooperation with KfK.

Under the MARIA contract, NRPB and KfK will participate jointly, with the COSYMA code, in the international ACA code intercomparison exercise being organized by the CEC and NEA. It is anticipated that the participation of the code in this exercise will indicate any weak aspects in coding and in data which will be improved in future releases of COSYMA.

Uncertainty analysis techniques

Continuing research will be undertaken into the application of available uncertainty analysis techniques to ACA code systems. An important task is research into techniques for the reliable and defensible quantification of the variation in model parameters, including the use of expert judgement. Both NRPB and KfK will be involved in these studies, together with other CEC contractors.

Model and code development

A number of areas have been identified where further modelling development is needed; these will be considered during the period of this contract. These include:
at NRPB there will be investigations with COSYMA to examine the effects of using models of different complexity in the area of atmospheric dispersion;
at KfK, wet deposition processes will be included in the foodchain module of COSYMA;
at NRPB, there will be further development of the urban contamination model EXPURT; onsideration will be given to incorporating EXPURT results in a COSYMA database;
at KfK, the economics module will be extended to incorporate specific data on the economic productivity of the region near to the release point;
at KfK, there will be further development of the health effects models used in COSYMA; in particular the method of calculating loss of life expectancy will be improved, and the dependence on dose and dose rate of both the dose/risk relationships for nonstochastic health effects and of the cancer risk factors will be included.

Funding Scheme

CSC - Cost-sharing contracts


Forschungszentrum Karlsruhe Technik und Umwelt GmbH

76021 Karlsruhe

Participants (4)


United Kingdom
OX11 0RQ Didcot