Advanced Atmospheric Dispersion Modelling and Probabilistic Consequence Analysis for Radiation Protection Purposes

Transcription

1 Advanced Atmospheric Dispersion Modelling and Probabilistic Consequence Analysis for Radiation Protection Purposes H. Thielen R. Martens W. Brücher Gesellschaft für Anlagen- und Reaktorsicherheit, Köln ABSTRACT : Currently within the licensing procedure of nuclear facilities in Germany atmospheric dispersion calculations are predominantly based on the simple Gaussian plume approach. Advanced model systems consisting of a diagnostic flow model together with a Gaussian puff model or a Monte-Carlo particle simulation model have less restrictions with respect to terrain effects, source configuration and non-stationary conditions. In combination with a probabilistic analysis with takes into account the variability of the weather situations a refined, realistic assessment of the radiological consequences of a incidental or accidental release can be performed. By means of cumulative complementary frequency distributions (CCFD) the probability of reaching or exceeding a specified dose value at a point of interest can be analysed. Only minor modifications to the standard model system for the licensing procedure of non-radioactive pollutant emissions (AUSTAL2000) with respect to γ cloudshine, radioactive decay and wet deposition are necessary to provide a state of the art model system for nuclear regulatory purposes. 1 DISPERSION MODELS For the analysis of possible consequences following an accidental release of airborne radioactive material from a nuclear facility the assessment of the radiation exposure of the population is a necessary prerequisite. To determine the potential doses the contributions of inhalation, γ-cloudshine and γ-ground radiation must be available, either by measurement or calculation. In case of atmospheric dispersion calculations the time-dependent spatial distribution of the airborne radionuclides and their dry and wet deposition to the ground has to be simulated. Corresponding atmospheric dispersion models provide such concentration and deposition fields based on a given source term, topography, and meteorological conditions. 1.1 Atmospheric dispersion models for regulatory purposes in Germany Within the licensing procedure of nuclear facilities in Germany atmospheric dispersion calculations following the release of airborne radionuclides are predominantly based on the Gaussian plume model (GPM) [1, 2]. Applying this model potential doses are calculated for releases due to normal operation as well as for incidental or accidental releases. Based on a 2-dimensional Gaussian concentration distribution (Fig. 2) the Gaussian plume model provides near ground concentrations and related deposition values. The shape of this distribution depends on atmospheric turbulence and is parameterised as a function of distance, 31

2 source height and meteorological conditions. This turbulence parameterisation was derived from tracer experiments with continuous emissions, source heights larger than 50 m in rather rough terrain during stationary meteorological conditions. The corresponding German regulations contain simple procedures to consider building effects, orographic effects as well as thermal plume rise in the dispersion model. Due to the analytical solution of the Gaussian approach this type of dispersion model is fast and only few input parameters are necessary. Hence, the application of a Gaussian plume model is simple. On the other hand, several simplifications inherent to the model limit the applicabilty to certain situations. Turbulence parameters for the GPM are not available for all roughness lengths and do not take into account the vertical profile of turbulence (not applicable for nearground emissions. Non-homogeneuos flow pattern due to structured terrain or surface may be treated only very roughly. Furthermore, the analytical solution of a GPM is Z Y limited to stationary conditions (emission and meteorology) and to dispersion without gravitational settling of aerosol and radioactive decay. wind h X Fig. 2: Gaussian Plume model 1.2 Advanced atmospheric dispersion models In contrast to the Gaussian plume model advanced atmospheric dispersion models like Gaussian puff models or Lagrangian particle simulation models (Monte-Carlo models) do generally account for more complex atmospheric dispersion and deposition conditions. Gaussian puff models, like e.g. the French ICAIR [3], the German GAUWOL [4] of the Society of German Engineers (VDI), or RIMPUFF from the Danish Risø National Laboratory [5] which is part of the Real Time Online Decision SuppOrt System (RODOS [6]), may take into account time dependent emissions, horizontally varying and time dependent meteorological conditions. The French-German Model (FGM) for the treatment of the atmospheric dispersion in case of a nuclear accident [7] applies a Gaussian puff model for concentration calculations, too. Particle simulation models additionally may take into account horizontally and vertically variable meteorological fields, including wind and turbulence, arbitrarily shaped source Ausbreitungsrichtung flow direction geometry, gravitational settling and washout of aerosol particles. In this type of dispersion model the tracks of a large number of particles based on the mean wind and random turbulent wind fluctuations (Fig. 3, green particles) and additional effects as settling (Fig. 3, red particles) are calculated. Typical models of this type are LASAT [8] and PARMOD Fig. 3: Principle of a particle simulation model 32

3 [9]. Together with an appropriate flow model which provides the mean wind components on a 3-dimensional grid in combination with a module supplying turbulence fields like standard deviations of velocity fluctuations and characteristic turbulent time scales according to a guideline of the Society of German Engineers (VDI) [10] the capabilities of the particle simulation model come into effect. In order to achieve a sufficient accuracy the number of released particles has to be high enough. A basic prerequisite for the utilization of the major advantages of Lagrangian dispersion models is the availability of 3-dimensional fields of flow and turbulence. Diagnostic (massconsistent) flow models provide an economic calculation of these datasets based upon suitable profiles of wind and turbulence (from parameterisations or measurements) taking orographic or even building effects into account. Fig. 4 shows a result of this type of model chain in an area with structured terrain and roughness length and a near surface release. The Figure represents the results of a flow and dispersion calculation in a stably stratified atmosphere causing a strong distortion of the flow and the plume due to orography. This concentration pattern cannot be treated with a Gaussian plume model. A Gaussian puff model can only approximately simulate this situation. Compared to more complex and time-consuming prognostic flow models the faster diagnostic models only provide static boundary conditions for dispersion calculations. Hence, sequences of diagnostic flow simulations are needed in order to calculate time-dependent emission and dispersion. Fig. 4: Patterns of near ground flow and dispersion in complex terrain during a stably stratified situation: wind simulation via a diagnostic flow model, concentration distribution via Lagrangian particle simulation model Within the framework of the implementation of the EU air quality directives [11] in Germany the standard model system for the licensing procedure of non-radioactive pollutant emissions (Technical Instructions for Air Quality Management - TA Luft [12]) was upgraded from a 33

4 Gaussian plume model to a combination of a Lagrangian particle simulation model with a diagnostic flow model (AUSTAL2000 [12]). GRS was involved in the process of model development and provides a free graphical user interface for AUSTAL2000 (GO-AUSTAL [13]). With this upgrade in atmospheric dispersion modelling for air quality management applications the state of the art is implemented again and the computation of mandatory probabilistic pollution measures like the frequency of limit value exceedances is now possible. 2 PROBABILISTIC CONSEQUENCE ANALYSIS For a radiological consequence analysis often other values than provided by AUSTAL2000 are of interest. Commonly the so-called deterministic approach is applied. It is based on pre-fixed meteorological conditions which guarantee the calculation of conservative concentration and deposition values, i.e. values which are higher than most of the remaining cases. Within a probabilistic consequence analysis the variability of the weather situations for the specified site is explicitly taken into account. Based on atmospheric dispersion calculations for every possible weather situation in combination with their respective site specific frequency of occurrence the so-called cumulative complementary frequency distributions (CCFD) of near ground air concentrations, deposition levels and the resulting potential doses may be derived. The probability of reaching or exceeding a specified dose at a given distance in downwind direction from the location of a release can be read from the curves. This conditional frequency doesn t take into account the frequency of occurrance of the release, i.e. it is assumed that the release has taken place. The following examples illustrate the added value of a probabilistic conseqence analysis. 2.1 Example: Sabotage Attack on Spent Fuel Casks On the basis of a specified source term calculations of potential radiation exposures of individuals in the vicinity of a breached spent fuel cask have been made [14]. It is assumed that the sabotage attack on a spent fuel cask leads to an almost instantaneous release of radioactive material through a penetration channel. Therefore, an atmospheric dispersion model must be applied which is capable of adequately simulating the following boundary conditions: Near ground-level release of gaseous and particulate material in the size range up to 100 µm; Short term release within a few seconds following the detonation; An initial configuration of released airborne material which is approximated by a rectangular column of 10 m by 10 m cross-section and 20 m height homogeneously filled with the released air-borne material with given particle size distribution. The applied atmospheric dispersion model is the GRS-version of the Monte-Carlo type particle simulation model LASAT which utilizes modern turbulence parameters for the calculation of short term dispersion and deposition processes in the atmospheric boundary layer [10]. This model chain is comparable to the above mentioned AUSTAL2000. Quite elaborate dispersion and deposition calculations have been performed for a multitude of atmospheric dispersion conditions taking into account four adjoining particle sizes. These results were used to derive the CCFDs of ground-level air concentrations and deposition levels in downwind direction from the location of a release as well as the corresponding effective doses from inhalation and γ-ground radiation taking into account the broad spectrum of atmospheric dispersions conditions and their respective frequency of occurrence. Weather 34

5 data typical for large parts of Germany and almost flat terrain have been used for this purpose. As one of the results a CCFD of potential effective dose from inhalation is shown in Fig. 5. Assuming that a release occurred, the probability of certain effective doses of a person residing at a given distance in downwind direction from the location of a release can be read from the curves. By presenting such CCFD for different down-wind distances the decline of potential inhalation doses with distance can easily be visualised msv conditional probability for individual dose D % 99% 50 m 300 m 1000 m 3000 m natural annual radiation exposure 3000 m 1000 m 300 m E effective dose D in direction of plume dispersion [Sv] 50 m Fig. 5: CCFD of inhalation dose - particles AED < 12 µm (adult, ICRP 72 dose coefficients) 2.2 Example: 2-dimensional approach for probabilistic consequence analysis The second example demonstrates a 2-dimensional probabilistic analysis using the German AUSTAL2000 including mesoscale diagnostic flow field calculation and Lagrangian dispersion modelling. Based on a time series of 8760 hourly averaged meteorological data (=1 year) consisting of wind speed, wind direction and stability class atmospheric dispersion calculations were carried out: For the meteorological conditions of each hour of this year a hypothetical release from a volume source (20x20x50 m_) lasting 1 hour was assumed and the resulting near ground concentration field was calculated. As pollutant the inert tracer SO 2 was selected and emitted with 1 g/sec. Fig. 6 shows the result of the 8760 flow field and dispersion calculations after superposition of the corresponding spatial concentration distributions in terms of a 2-dimensional distribution of the mean concentration. Furthermore, the position of the source (Q1), possible points of interest (P1 P4) and the site specific wind direction distribution are represented. 35

6 Fig. 6: 2-dimensional distribution of the near-ground mean concentration in complex terrain derived via AUSTAL2000 (see text) conditional probabilily for concentrations C % 99 % point 1 point 2 point 3 point ground level concentration C in µg/m? Fig. 7: CCFD of ground level concentration for the 4 points of interest 36

7 For the analysis of incidental or accidental releases such time-averaged results are not adequate. However, frequency distributions of concentrations calculated for points of interest contain valuable information as in the first example. Fig. 7 contains the corresponding CCFD s of the near surface concentrations for the 4 points of interest (see Fig. 6). Due to the frequency distribution of the wind direction the concentrations are mostly zero, i.e. 90 % (points P1 P3) or 80 % (P4) of all situations. For some rare and unfavourable conditions concentrations of more than 100 µg/m_ up to 600 µg/m_ are reached at P2, P3 and P4. However, with a probability of 99 % the concentrations are below 130 µg/m_ (P4), 30 µg/m_ (P2 and P3) or 9 µg/m_ (P1). The maximum concentration value is related to the most unfavourable meteorological condition with a respective frequency of 1/8760. The current version of AUSTAL2000 is only applicable for non-radioactive airborne material. Calculations of potential doses are not possible because it does not simulate γ-cloudshine, radioactive decay and wet deposition due to precipitation. Nevertheless, this example gives evidence of the benefit of probabilistic consequence analysis on a high quality level applying a state of the art mesoscale model chain incorporating flow field and atmospheric dispersion models. 3 CONCLUSIONS Currently within the licensing procedure of nuclear facilities in Germany atmospheric dispersion calculations are predominantly based on the rather simple Gaussian plume approach. The guideline for the radiation protection expert in the emergency staff [15] applies a more advanced approach based on the French-German model [7]. Meanwhile in Germany the standard model system for the licensing procedure of nonradioactive pollutant emissions was upgraded from a Gaussian plume model to a state of the art mesoscale model chain comprising a diagnostic flow model together with a Lagrangian particle simulation model (new German TA Luft, AUSTAL2000 [12]). In order to guarantee consistent realistic atmospheric dispersion modelling of both, radioactive and non-radioactive airborne pollutants, further regulatory work should aim at the development of an appropriate atmospheric dispersion model for radioactive airborne material. This task will require only some modifications to AUSTAL2000 with respect to the simulation of γ-cloudshine, radioactive decay and wet deposition due to precipitation. The resulting model chain will be applicable for the calculation of atmospheric dispersion in topographically structured terrain incorporating dry and wet deposition, gravitational settling, radioactive decay, and γ-cloudshine. Building effects may be treated explicitly provided that the corresponding flow field is available. There are neither restrictions with respect to source geometry nor to non-stationary meteorology or variable emission characteristics. Due to increased computer resources restrictions with respect to computing time have decreased. Thuch state of the art mesoscale model chain permits an analysis of radiological consequences following an airborne release of radionuclides on a high quality level. On the one hand deterministic calculations can be carried out on the basis of pre-fixed meteorological conditions which guarantee the determination of conservative concentration and deposition values. On the other hand, also a differentiated probabilistic analysis is possible which takes into account the variability of the weather situations. By means of cumulative complementary frequency distributions (CCFD) the probability of reaching or exceeding a specified dose value at a given distance from the location of a release can be analysed. 37

8 4 LITERATURE [1] Bekanntmachung einer Empfehlung der Strahlenschutzkommission (Neufassung der 'Berechnung der Strahlenexposition') vom 29. Juni 1994, Bundesanzeiger Nr. 222a vom 26. November 1994 [2] Allgemeine Verwaltungsvorschrift zu 45 Strahlenschutzverordnung: Ermittlung der Strahlenexposition durch die Ableitung radioaktiver Stoffe aus kerntechnischen Anlagen oder Einrichtungen. Bundesanzeiger vom [3] Monfort, M.: Présentation des modèles physiques du code de calcul ICAIR4 de dispersion passive des pollutants dans l atmosphère. Note Technique DPEA/SECRI/ [4] Verein Deutscher Ingenieure: Umweltmeteorologie, Atmosphärische Ausbreitungsmodelle, Gauß-Wolken-Modell. Kommission Reinhaltung der Luft im VDI und DIN, VDI 3945, Blatt 1, Beuth Verlag, Berlin, 1996 [5] Thykier-Nielsen S., Mikkelsen T., Nyrén, K.: RIMPUFF USER S GUIDE, RIMDOS Version (Version 47 -PC): Risø(Denmark): Risø National Laboratory, Jan 23., 1995 [6] J. Ehrhardt J. et al.: The RODOS System: Decision Support for Off-Site Emergency Management in Europe. Radiat. Prot. Dosim. 73(1-4), pp (1997) [7] Crabol B. et al.: French German Model (FGM) for the treatment of the atmospheric dispersion in case of a nuclear accident. French-German Commission fort he Safety Problems of French-German of Nuclear Installations, DFK 98/1, Bonn (Germany): Federal Ministry for Environment, Nature Conservation and Nuclear Safety, Paris (France): Ministry of Industry, October 1998 [8] Janicke L.: A random walk model for turbulent diffusion. Ingenieurbüro Janicke, Gesellschaft für Umweltphysik, Dunum, Berichte zur Umweltphysik, Nummer 1, Auflage 1, August 2000, (Print), ISSN , (Internet), ISSN [9] Verein Deutscher Ingenieure: Umweltmeteorologie, Partikelmodell. Kommission Reinhaltung der Luft im VDI und DIN, VDI 3945, Blatt 3, Düsseldorf, September 2000 [10] Verein Deutscher Ingenieure: Umweltmeteorologie, Messwertgestützte Turbulenzparameterisierung für Ausbreitungsmodelle. Kommission Reinhaltung der Luft im VDI und DIN, VDI 3783, Blatt 8, Düsseldorf, Dezember 2002 [11] The Council of the European Union: Council Directive 96/62/EC of 27 September 1996 on ambient air quality assessment and management. Official Journal L 296, 21/11/1996 pp [12] Erste Allgemeine Verwaltungsvorschrift zum Bundes-Immissionsschutzgesetz (Technische Anleitung zur Reinhaltung der Luft TA Luft) vom 30. Juli 2002, GMBl , S. 511/605. Corresponding computer program is available via: [13] Gesellschaft für Anlagen- und Reaktorsicherheit (GRS) mbh: GO-AUSTAL Eine Hilfe zur Anwendung der neuen TA-Luft. Free graphical user interface for AUSTAL [14] Lange F. et al.: Experiments to Quantify Potential Releases and Consequences from Sabotage Attack on Spent Fuel Casks, 13th International Symposium on the Packaging and Transportation of Radioactive Materials (PATRAM), Chicago, U.S.A., September 2001 [15] Strahlenschutz Kommission: Leitfaden für den Fachberater Strahlenschutz der Katastrophenschutzleitung bei kerntechnischen Notfällen. Berichte der SSK, Heft 37, München, Urban und Fischer-Verlag, München, 2003, 38