Mid-range atmospheric dispersion modelling. Intercomparison of simple models in EMRAS-2 project

Mid-range atmospheric dispersion modelling. Intercomparison of simple models in EMRAS-2 project Raúl Peria ~ nez a , * , Kathleen M. Thiessen b , Sohan L. Chouhan c , Francesco Mancini d , Emilie Navarro e , Gert Sdouz f , 1 , Dejan Trifunovic  g a University of Seville, ETSIA, Ctra Utrera km 1, 41013, Sevilla, Spain b Oak Ridge Center for Risk Analysis, 102 Donner Drive, 37830, Oak Ridge, Tennessee, USA c Canadian Nuclear Laboratories, K0J 1J0 Chalk River, Ontario, Canada d SOGIN S.p.A., Via Torino, 6, I-00184, Rome, Italy e Institut de Radioprotection et de Sûrete  Nucleaire, 31 Avenue de la division Leclerc BP17, F-922625, Fontenay-aux-Roses, France f Austrian Institute of Technology, A-2444, Seibersdorf, Austria g Federal Authority for Nuclear Regulation, Sheikh Zayed First Street, P.O. Box 112021, Abu Dhabi, United Arab Emirates Keywords: Atmospheric dispersion Nuclear power plant Model 137-Cs 131-I abstract An intercomparison of atmospheric dispersion models has been carried out for a hypothetical accident occurring in a nuclear power plant in the center of Spain. The accident consisted of a steam generator tube rupture, and two radionuclides have been considered for the exercise: 137-Cs and 131-I. Meteo-rological conditions and radionuclide release rates were supplied. Models provided deposition maps, timeintegrated concentrations in air and arrival times of the plumes to specific locations. The effect of the meteorological conditions used in the modelling was clear, with different behavior of the plume with neutral stability vs. stable conditions. The predicted arrival times of the plume at specific locations showed much less variability than deposition and air concentrations. This variability in part reflects the uncertainties inherent in atmospheric dispersion modelling and in the selection of parameter values, such as deposition velocities or diffusivities. 1. Introduction Environmental assessment models are used for evaluating the radiological impact of actual and potential releases of radionuclides to the environment. They are essential tools for use in the regulatory control of routine discharges to the environment and also in planning measures to be taken in the event of accidental releases. They are also used for predicting the impact of releases which may occur far into the future, for example, from underground radioactive waste repositories. It is important to check, to the extent possible, the reliability of the predictions of such models by comparison with measured values in the environment or by comparing with the predictions of other models. The International Atomic Energy Agency (IAEA) has been organizing programmes of international model testing since the 1980s. The programmes have contributed to a general improvement in models, in transfer data and in the capabilities of modellers in Member States. The possible benefits of carrying out model validation and testing at an international level were recognized by the Swedish Radiation Protection Institute, which sponsored the Biospheric Model Validation Study (BIOMOVS) and BIOMOVS II programmes starting in 1985 (BIOMOVS II,1996). BIOMOVS was the first international exercise aimed at the testing and validation of models for the prediction of radionuclide transfer through the environment to humans. The Chernobyl accident in 1986 created a renewed need for reliable assessments in many countries and provided an increased impetus for work in this area. It also originated new data sets that could be put to use for model testing. As a consequence, the IAEA was prompted to start a programme on the Validation of Model Predictions (VAMP) in 1988, which concluded in 1996 (IAEA, 2000). More recently, Environmental Modelling for Radiation Safety (EMRAS) program, running from 2003 to 2007, was launched (IAEA, 2012). From 2009 to 2011, the second stage of EMRAS was organized (EMRAS II). The Urban Areas Working Group was organized *Corresponding author. E-mail address: [email protected] (R. Peri a~ nez). 1 Present address: Hockegasse 24/23, A-1180, Vienna, Austria. within this last programme, as part of a theme entitled “Approaches for Assessing Emergency Situations”. The Working Group has been built on the work done by the Urban Remediation Working Group of the first phase of the EMRAS Programme (IAEA, 2012). The goal of the Urban Areas Working Group is to test and improve the capabilities of models used in assessment of radioactive contamination in urban settings, including dispersion and deposition events, shortand long-term contaminant redistribution following deposition events, and potential countermeasures or remediation efforts for reducing human exposures and doses. A mid-range atmospheric dispersion exercise was carried out by the group. It is based on a hypothetical accident at a nuclear power plant and the resulting predicted deposition in urban environments up to 70 km downwind. The scenario assumed a 1-h release from a rupture of a steam generator tube, based on an accident scenario developed by the Institut de Radioprotection et de Sûret e Nucl eaire (IRSN), and uses actual geographic and meteorological information for the Trillo nuclear power plant in central Spain. It should be commented that complex atmospheric dispersion models exist now. For instance, WSPEED I-II (Terada et al., 2012) and LADAS (Suh et al., 2009) were applied in the case of Fukushima accident to evaluate deposition on the sea surface to later compute marine dispersion of these released radionuclides (Peri a~ nez et al., 2015). Also, atmospheric dispersion models were applied to evaluate the source term from Fukushima (Chino et al., 2011; Kobayashi et al., 2013). These models consist of a meteorological prediction model coupled with an atmospheric dispersion model. Nevertheless, the objective of this work consisted of testing simple atmospheric dispersion models, which do not require the meteorological sub-model, and which can provide a very fast answer in case of an accident. Since the final purpose of the models generally is a dose estimation to the public, models are based on conservative approaches. The purpose of this paper is to present the main results from this exercise. The exercise is described in Section 2. Results are presented in Section 3. 2. Methods The nuclear power plant chosen for the exercise is Trillo (TNPP), Fig. 1. Top: General localization of the study area. Bottom: Topography of the domain (elevations in m above sea level). in the central part of Spain, about 70 km northeast from Madrid metropolitan area and 46 km from Guadalajara, which is a smaller town in central Spain located between TNPP and Madrid (Fig. 1). TNPP started operation in 1987. The power is 1043 MW, and the reactor is PW type. Cooling is carried out through two towers. Specific meteorological conditions, which would be representative of a worst-case scenario, were used for the simulations. Simulated wind fields 10 m above the ground were provided. Two situations were considered: one with a stable atmosphere, and one with neutral stability. This allowed assessment of the effects of stability conditions on radionuclide dispersion. In both cases, the same geostrophic wind direction is considered (northeast). Geostrophic wind speeds of 3.0 m/s and 6.0 m/s were used for the stable and neutral conditions, respectively. The boundary layer height was 1000 m for stable and 1500 m for neutral stability. Wind fields 10 m above the ground were obtained from WINMOD model, developed at the University of North Wales (Jones, 1998). WINMOD calculates such wind fields from the geostrophic wind and atmosphere stability. Essentially, the model diagnoses the local modification of the wind field in regions of complex topography. A geostrophic wind speed and direction, as well as the atmospheric lapse rate and boundary layer height, are specified and the model iterates the horizontal momentum and temperature equations at the surface towards a steady state. Wind fields 10 m above the ground are presented in Fig. 2 for both atmospheric conditions. −3.8 −3.6 −3.4 −3.2 −3−2.8 −2.6 40.2 40.3 40.4 40.5 40.6 40.7 40.8 Longitude Latitude • TNPP MADRID • Guadalajara 500 600 700 800 900 1000 1100 1200 −3.8 −3.6 −3.4 −3.2 −3−2.8 −2.6 40.2 40.3 40.4 40.5 40.6 40.7 40.8 Longitude Latitude • TNPP MADRID • Guadalajara 500 600 700 800 900 1000 1100 1200 Fig. 2. Wind fields 10 m above the ground provided by WINMOD model for stable (top) and neutral (bottom) stability conditions. Only one of each 16 provided vectors is drawn for clarity. Land elevations (color scale) in m above sea level. The same hypothetical accident was considered for both meteorological situations. The radionuclides are assumed to be released as gas, and only dry deposition is considered. Two radionuclides with different half-lives are considered, 137 Cs and 131 I, with all of the latter in molecular form. Use of these two radionuclides with different half-lives was considered enough for modelling intercomparison purposes, although of course many more radionuclides and in different chemical forms would be released during a real accident. The hypothetical accident considered here consists of a steam generator tube rupture, a scenario which was developed by the Institut de Radioprotection et de Sûret e Nucl eaire (IRSN) of France. The duration of the release is 1 h, and the release rate is variable for both radionuclides. Modellers were provided with release data over the 1 h period, shown in Fig. 3. Temporal resolution of the release data is 60 s. Total (integrated) releases are 6.42 10 11 Bq and 3.69 10 12 Bq for 137 Cs and 131 I, respectively. An effective release height of 50 m was considered. Modellers were asked to carry out a simulation over 10 h. Endpoints of simulations were contour maps of deposited activity on the ground and of time-integrated activity concentrations in air at ground level, at the end of the simulation. Modellers were also asked to provide a time series of activity concentrations in air at four selected points (two intermediate point between TNPP and Guadalajara town, denoted as IP1 and IP2, Guadalajara and Madrid downtown). These points are shown in Fig. 1. Table 1provides a summary of the five models used in the exercise. More information about individual models is provided in Appendix A-E and in the references included in the Table. The models represent several different purposes (e.g., emergency assessment and research) and two major types of modelling approaches (Gaussian and Lagrangian). One model (Hotspot) provided results in terms of the distance down the plume centerline, while the others were able to provide results with reference to the local geography. For one model (RASCAL), the range of the model predictions did not extend to the distance of Madrid in this exercise. Four models were used to provide results for both stable atmospheric conditions and neutral stability; the fifth (JRODOS) was 0 10 20 30 40 50 60 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2x 109 Time (minutes) Release rate (Bq/s) 137Cs 131I Fig. 3. Radionuclide release rates for the steam generator tube rupture accident. The origin of time is arbitrary. Table 1 Summary of main model characteristics. ADDAM RASCAL USev JRODOS HOTSPOT Type of model Gaussian Gaussian þLagrangian puff Lagrangian Gaussian þsimplified puff Gaussian Calculation range user specified radius 80 km <100 km <100 km >10 m, <100 km Release height 50 m 50 m 50 m 10 m 50 m Stability class E (stable), D (neutral) E, D E, D E E, D Wind vectors summed outside the code Limited number used As provided As provided E: 3 m/2; D: 6 m/s Diffusion Horizontal: 60 m 2 /s High roughness: Karlsruhe-Jülich c coefficients Briggs a Pasquill-Gifford curves b Vertical: 30 m 2 /s Moderate roughness: Mol d Pasquill-Gifford Dry deposition Cs: 0.01 m/s Cs: 0.04 m/s velocity I: 0.008 m/s 0.003 m/s Not applicable Internally determined I: 0.22 m/s Release time-step 1 h 1 h 1 min 0.5 h 1 h Calculation time-step One single step 15 min 10 s 1 h One single step Simulation time Not applicable 10 h 10 h 7 h Not applicable Topography Flat Not used As provided Not applicable Rugosity 0.40 m 0.20 m 0.40 m City terrain Reference Scheier and Chouhan, 2009 Ramsdell, 2012 NP e Ievdin et al., 2012 Homann and Aluzzi, 2014 a Briggs (1971). b Gifford (1968). c Geiss et al. (1981). d Panitz et al. (1989). e Not published. used only for stable conditions. One model (USev) used timedependent source term information as provided, in 1-min increments. Three models considered the entire release in a single 1h time step, while the other model (JRODOS) considered it in two half-hour time steps. The models also differed in their handling of the information about wind speed and direction and in the dry deposition velocities that were used. 3. Results and discussion Contour maps for predicted deposition of 137 Cs are shown in Figs. 4e8. Two sets of predictions (USev and ADDAM, Figs. 4 and 5) clearly show the effect of meteorological conditions: under stable conditions, the predicted plumes intersected Madrid, but with neutral stability, the plume bypassed Madrid. The JRODOS plume for stable conditions also intersected Madrid (Fig. 6). Predictions with RASCAL did not extend as far as Madrid, but the plots suggest that the plume would have intersected Madrid with stable conditions and bypassed Madrid with neutral stability (Fig. 7). Hotspot provided results in terms of the plume centerline rather than the local geography, but the results do show higher deposition farther downwind for the stable conditions (Fig. 8). Predicted values for deposition of 137 Cs and 131 I at specific locations are provided in Tables 2 and 3. As expected solely from the distances, the general tendency was for the highest deposition at Intermediate Point-1 (IP1), then IP2, Guadalajara, and Madrid. Predicted differences between both radionuclides reflect both the different source terms and (for some models) different deposition velocities for the two radionuclides. For 137 Cs deposition in Madrid, predicted values for stable Fig. 4. Calculated 137 Cs deposition by USev model for stable (top) and neutral (bottom) stability conditions. conditions (4 models) differed by a factor of 28. For neutral stability, all predicted values for Madrid were 0 (USev) or a factor of about 10e100 lower than for stable conditions. For 131 I, predicted values in Madrid for stable conditions differed by a factor of about 60. As with 137 Cs, predicted values for 131 I for neutral stability were 0 or a factor of about 10e100 lower than for stable conditions. These values are consistent with the contour maps, which showed the plume bypassing Madrid in the case of neutral stability. For Guadalajara (5 models), predicted values of deposition with stable conditions were within a factor of 12 for 137 Cs and 24 for 131 I. For neutral stability, predicted values were within a factor of 165 for 137 Cs and 26 for 131 I, indicating more variability among the models for neutral stability. ADDAM and USev predicted higher deposition in Guadalajara for neutral stability than for stable conditions, while Hotspot and RASCAL predicted higher deposition with stable conditions. USev model plume bypassed IP2 with neutral stability. ADDAM predicted deposition at both IP1 and IP2 for both stable conditions and neutral stability, with a decrease from IP1 to IP2 of about a factor of 9 for stable conditions and a factor of 3 for neutral stability. Surprisingly, RASCAL model plume bypassed IP1 (0 deposition) under stable conditions, and there is not deposition immediately downwind TNPP (Fig. 7). This was attributed to the meteorological preprocessor included in this model (Appendix B): only a part of the provided wind vectors were used. This probably does not allow to describe appropriately the plume behavior in the initial stages of dispersion. Time-integrated concentrations in air are not shown. However, the general tendency was a decrease with distance, as expected. Thus the highest air concentrations were found at Intermediate Point-1 (IP1), then IP2, Guadalajara, and Madrid. One exception was JRODOS, which predicted slightly higher concentrations for parts of Madrid. Also, RASCAL’s predicted plume bypassed IP1 with stable conditions and predicted higher concentrations of 131 I in Guadalajara than for IP2 for both sets of meteorological conditions. For Madrid, the results again show a large decrease (a factor of ~10 or Fig. 5. Calculated 137 Cs deposition by ADDAM model for stable (top) and neutral (bottom) stability conditions. more) from the values predicted for stable conditions to the values predicted for neutral stability, consistent with a predicted plume bypassing Madrid in the case of neutral stability. An important output from atmospheric dispersion models, related to emergency management, is the arrival time of the radioactive plume to a given point. Predicted values for the approximate time to arrival of the predicted plume at specific locations are provided in Table 4. For any given location and stability class, predicted times to arrival are within a factor of about 2 of each other, for those plumes predicted to reach the location. As described previously, RASCAL predicted that the plume would bypass IP1 under stable conditions, and USev predicted that the plume would bypass both IP2 and Madrid for the case of neutral stability. For plumes predicted to reach a given location, times to arrival were shorter for neutral stability than for stable conditions. In general, ADDAM predicted the longest times to arrival for a given Fig. 6. Calculated 137 Cs deposition (Bq/m 2 ) by JRODOS model for stable conditions. Fig. 7. 137 Cs deposition contours calculated by RASCAL model for stable (top) and neutral (bottom) stability conditions. Fig. 8. Calculated 137 Cs deposition contours by Hotspot model for stable (top) and neutral (bottom) stability conditions. Table 2 Comparison of predictions for deposition of 137 Cs (Bq/m 2 ). NR: not reported. Model IP1 IP2 Guadalajara Madrid Stable conditions (E) ADDAM 9030 964 44.8 1.03 Hotspot NR NR 20 13 JRODOS NR NR 14 28 RASCAL 0 174 166 NR USev 4348 233.5 56.86 10.14 Neutral conditions (D) ADDAM 4970 1490 381 0.0165 Hotspot NR NR 2.3 1.4 RASCAL 267 69.8 122 NR USev 7414 0 144.3 0 Table 3 Comparison of predictions for deposition of 131 I (Bq/m 2 ). NR: not reported. Model IP1 IP2 Guadalajara Madrid Stable conditions (E) ADDAM 43,800 4920 256 6.66 Hotspot NR NR 610 390 JRODOS NR NR 110 72 RASCAL 0 964 922 NR USev 4312 149.7 38.98 11.45 Neutral conditions (D) ADDAM 23, 400 7120 1880 0.0837 Hotspot NR NR 71 45 RASCAL 1480 389 677 NR USev 7395 0 146 0 location and USev and RASCAL the shortest times. 4. Conclusions Participants in this test exercise started with the same information, and while there is general agreement in the results, there are also some obvious differences. It is important to identify and explain the reasons for these differences. As described above, participants varied in their handling of the source term (time-dependent or all at once) and wind field data, and in their selection of parameter values such as deposition velocity (for those models that used a given parameter). The effect of the meteorological conditions used in the modelling is obvious, with different behavior of the plume with neutral stability vs. stable conditions. In this exercise, the most important difference was that the predicted plumes tended to bypass Madrid (the largest city in the test region) with neutral stability but intersected it with stable conditions. Predictions varied as to whether the plume reached or bypassed some of the intermediate points. It is important to note that even when all modellers used the same stability class, the path of the predicted plume could vary among participants. Variability between model predictions was higher for deposition and air concentrations (not shown) than for other endpoints. This variability in part reflects the uncertainties inherent in atmospheric dispersion modelling and in selection of parameter values. For example, for the three models requiring the user to select values for deposition velocity, the selected values ranged from 0.003 to 0.04 m/s for 137 Cs and from 0.008 to 0.22 m/s for 131 I(Table 1). Models may also have varied in the height used for calculation of the air concentrations. Although uncertainties in individual model predictions were not estimated, the results give an idea of the level of uncertainty that should be acknowledged in dealing with modelling results in important assessments. In this sense, a multimodel approach, as described by Monte et al. (2008), may be of interest when environmental processes are complex. Through this approach, the conclusions that obtain the greatest degree of consensus among modellers are made evident and the aspects that are subject to dispute and which should therefore be handled carefully also become clear. The predicted arrival times of the plume at specific locations showed much less variability, usually less than a factor of 2 among three or four sets of model predictions. In practice, this is an important endpoint, in that it provides an estimate of the time available for evacuating an area or getting people to shelter. The results also show the dependence of this endpoint on meteorological conditions. Acknowledgement Work carried out in the frame of IAEA EMRAS-2 project (Environmental Modelling for Radiation Safety-2). Appendix A ADDAM ADDAM (Ballyk et al., 2003) is a Gaussian plume model which considers the following atmospheric dispersion phenomena (all have been validated): plume rise, downwash, and entrainment (effective release height); fumigation; reflection from an elevated inversion; transport and dispersion (plume broadening and plume diffusion); wet and dry deposition and plume depletion; radioactive decay and build-up; external exposure due to cloudshine (including a finite cloud model) and groundshine; and internal exposure due to inhalation. The model runs stochastically with respect to meteorological data, buoyant releases can be considered and the lateral dispersion coefficient can be calculated from the standard deviation in the wind direction. Meteorological conditions are constant and the terrain is flat. The domain of interest is broken down into 16 sectors and a user-specified number of radial distances. Then, for a given release, dilution factors (concentration divided by release rate) and doses are calculated at each distance in the affected sectors. Some limitations are that the model is not suitable for puff releases (release duration should be greater than the travel time) and that calm winds cannot be treated. B RASCAL RASCAL (Ramsdell, 2012) is a radiological assessment tool for use in emergency response applications. It consists of modules that estimate accident source terms for nuclear power plants and other nuclear fuel cycle facilities; transport, dispersion and deposition of radionuclides; and doses. It also includes a meteorological preprocessor that prepares meteorological data for use by the atmospheric transport modules. RASCAL is a Lagrangian trajectory Gaussian-puff dispersion model. There are two atmospheric stability fields. One consists of Pasquill-Gifford stability classes (Gifford, 1968), and the other consists of the inverse Monin-Obukhov length (Monin and Obukhov, 1954). The Monin-Obukhov length is estimated from the Pasquill-Gifford stability class using a graphical relationship between Monin-Obukhov length, stability class, and surface roughness. RASCAL accepts three precipitation conditions: no precipitation, rain, and snow. Every hour, the precipitation grid is updated using hourly observations. The plume is represented by a series of puffs released at 5-min intervals. Each puff contains the activity released during a 5-min period. The height of release is the sum of the actual release height and final plume rise. The dispersion parameters are a function of distance travelled and atmospheric stability using numerical approximations to the Pasquill-Gifford dispersion curves. Deposition is calculated using a source depletion model with a constant dry deposition velocity. Wet deposition is calculated using a simple washout model with constant washout coefficients. C USev The model, not published, is based on the approach by Elliott (1999).Essentially, it is a Lagrangian three-dimensional particletracking dispersion model. Radioactive decay and turbulent diffusion are simulated using stochastic methods. A horizontal and a vertical diffusion coefficient must be specified. Particles are assumed to be deposited when their height is smaller than 10 cm over the ground level, thus a deposition velocity is not required. Table 4 Comparison of predictions of plume arrival times (min). NR: not reported. Model IP1 IP2 Guadalajara Madrid Stable conditions (E) ADDAM 67 150 417 850 JRODOS NR 120 240 420 RASCAL bypassed 90 210 NR USev 40 100 210 460 Neutral conditions (D) ADDAM 33 83 233 483 RASCAL 15 60 150 NR USev 20 bypassed 130 bypassed