Directory UMM :Data Elmu:jurnal:E:Environmental Management and Health:Vol10.Issue4.1999:

(1)

Real time predictions of transport, dispersion and

deposition from nuclear accidents

Jùrgen Brandt

National Environmental Research Institute, Department of Atmospheric

Environment, Roskilde, Denmark

Jesper H. Christensen

National Environmental Research Institute, Department of Atmospheric

Environment, Roskilde, Denmark

Zahari Zlatev

National Environmental Research Institute, Department of Atmospheric

Environment, Roskilde, Denmark

1. Introduction

The Chernobyl accident, 25 April 1986, 21:23 UTC, emphasized the need for fast and reliable forecasts of transport, dispersion and deposition of radioactive air pollutants. Following the Chernobyl accident, many national and international activities have therefore been initiated to develop reliable models that can be used in connection with accidental releases. A comprehensive, high-resolution, 3D tracer model has been devel-oped. The tracer model is based on a combi-nation of a Lagrangian short-range puff model and a Eulerian long-range transport model. The combined 3D tracer model is called the Danish Rimpuff and Eulerian Accidental release Model (DREAM) (Brandt, 1988; Brandtet al., 1996a, 1996b, 1996c; 1998a, 1998b). The Lagrangian model is used to calculate the initial transport, dispersion and deposition of the plume in an area close to the source and the Eulerian model is used to calculate transport, dispersion and deposi-tion on long range. By coupling a Lagrangian model with a Eulerian model, the idea is to gain the advantages of both kind of models (Brandt, 1998). The combined model has been tested against numerical experiments, using a revised version of the Molenkamp-Crowley rotation test (Brandtet al., 1996a) and vali-dated against measurements from the two ETEX releases in the autumn of 1994 (Nodop, 1997), and the Chernobyl accident (Brandt, 1998; Brandtet al., 1998a, 1998b). For a more precise description of the mean meteorologi-cal fields, the meteorologimeteorologi-cal meso-smeteorologi-cale model MM5V1 (Grellet al., 1995) is used as a meteorological driver for the transport model. A schematic diagram for the whole model system is shown in Figure 1.

2. The DREAM

2.1 The Eulerian model

The Eulerian long-range transport model is applied in the whole model domain. In the Eulerian modelling framework, the advective transport, dispersion, emission, wet deposi-tion and radioactive decay in the atmosphere is described by the equation (see, for

example, Pudykiewicz, 1989)

@C

whereCis the tracer mixing ratio (the concentration c divided by the air density), u;v;_are the wind speed components in the x;y; directions, respectively,Kx;KyandK

are dispersion coefficients, whereKxandKy are assumed constant,E x; y; ; tis the emission,is the scavenging coefficient for wet deposition andkris representing the radioactive decay. The vertical co-ordinate in the model is in-coordinates

PÿPt

PsÿPt

2

wherePis the pressure,Psis the surface pressure andPt is the pressure at the top of the model. Dry deposition is applied in the model as a lower boundary condition for the vertical dispersion with the termÿv_dC, wherevdis the dry deposition velocity.

2.2 The Lagrangian model

The Lagrangian model is a short-range puff-model which simulates a continuous release

The current issue and full text archive of this journal is available at

http://www.emerald-library.com

Unfortunately reproduction of some figures in this article presented problems of clarity. How-ever, they have been included as a demonstration of data presentation as described in the article. Environmental Management

and Health

10/4 [1999] 216±223

#MCB University Press

[ISSN 0956-6163]

Keywords

Models, Pollution, Forecasting, Real time

Abstract

Describes a tracer model, DREAM (the Danish Rimpuff and Eulerian Accidental release Model), devel-oped for studying transport, dis-persion, and deposition of air pollution caused by a single but strong source. The model is based on a combination of a Lagrangian short-range puff model and a Eu-lerian long-range transport model. It has been run and validated against measurements from the two European Tracer Experiment (ETEX) releases and from the Chernobyl accident. An air pollu-tion forecast system, THOR, is under development, to make fore-casts of various air pollutants on a European scale. Some preliminary results are shown. DREAM will be implemented in THOR for calcula-tions of real time prediccalcula-tions of transport, dispersion and deposi-tion of radioactive material from accidental releases (e.g. Cherno-byl). Some applications of the DREAM model and examples of model results are described.

(2)

changing in time by sequentially releasing a series of puffs which are advected and dispersed individually along trajectories. The puffs are Gaussian shaped in the horizontal direction but have been transformed into -co-ordinates in the vertical direction by using the hydrostatic approximation and the ideal gas law (Brandt, 1998). Assuming total reflection from the ground and from the top of the planetary boundary layer (the mixing height), the contribution from allNpuffs to the mixing ratioCx;y; atx; y; is given by

whereMiis the mass of the individual puffi,gis the acceleration due to gravity,Ris the gas

constant for dry air,Tvis the mean virtual temperature,xci; yci; ciare the location of the center of the puffs,xy~ i; ~i are the horizontal and vertical standard deviations of the puffs, is given by:

PsÿPt Pt with height z due to dry deposition in the Lagrangian puff model is assumed to be a sum of the mass losses in the horizontal grid-cells,

i; j, that are covered by the puff, in the lowest the lowest model layer (transformation from ztois carried out using ideal gas law and hydrostatic approximation) andv_d

i;j is the dry deposition velocity in the grid-celli;j.I andJare the total number of horizontal grid-cells, in the lowest model layer, covered by the puff i inxandydirections, respectively. The mass loss due to wet deposition and

Figure 1

Schematic diagram showing main modules for MM5V1 and DREAM and flow chart for the whole model system. TERRAIN/LANDUSE, DATAGRID and INTERP modules are preprocessors for interpolation of terrestrial and meteorological input data. Emission data are given as input to the Lagrangian model. Advanced visualization and animation techniques have been implemented in the model system

[ 217 ]

Jùrgen Brandt,

Jesper H. Christensen and Zahari Zlatev

Real time predictions of transport, dispersion and deposition from nuclear accidents

Environmental Management and Health

(3)

radioactive decay is described in a similar way, but here the summation is performed over all grid-cells in all three directions,

x; y; , that are covered by the puff

M_lX

K

k1

XJ

j1

XI

i1

M_l_;_i_;_j_;_k1ÿeÿt_i_;_j_;_kkr ₇

2.3 Parametrizations of subgrid scale phenomena

Different simple and comprehensive para-metrizations for dispersion, dry and wet deposition and the mixing height have been implemented (in both the Lagrangian model and the Eulerian model) and tested in the model against measurements from Cherno-byl and the two ETEX experiments. Also the

sensitivity to the accuracy of the meteorolo-gical input data has been studied (Brandt, 1998; Brandtet al., 1998b). The best perform-ing parametrizations found from these tests are used in the model. A scheme based on the Monin-Obukhov similarity theory is used for the vertical dispersion in the Eulerian model. The mixing height is parametrized by using a combination of a Zilitinkevich-Mironov scheme for neutral and stable conditions and a bulk Richardson scheme for unstable con-ditions. Dry deposition velocities are para-metrized by using a simple scheme based on the friction velocity, the Monin-Obukhov length and some land use categories. The scavenging coefficient for wet deposition is found from a simple scheme depending on

Figure 2

131

I surface concentrations in Europe on 5 May 1986, 00.00 UTC, nine days after the start of the Chernobyl accident, calculated by using DREAM. The square in the Figure indicates the area where the Lagrangian model is applied

Jùrgen Brandt,

(4)

the relative humidity. Horizontal and vertical dispersion in the Lagrangian model is parametrized based on Monin-Obukhov similarity theory. The horizontal dispersion coefficients in the Eulerian model,Kxand Ky, are set constant 10;000m2=s(Brandt, 1998).

2.4 Numerical implementation

It is difficult to treat the partial differential equation (PDE) (1) directly in the Eulerian model. The Eulerian model has therefore been split, according to rules given in McRae et al.(1982) (see also Brandtet al.(1996c)) into three sub-models containing:

1 3D advection, horizontal dispersion and emissions;

2 vertical dispersion and dry deposition; and

3 wet deposition and radioactive decay.

Sub model 2 is performed by using a finite element algorithm with piecewise linear independent functions; the chapeau functions (Pepperet al., 1979; Brandt, 1998). This algorithm has been tested together with other schemes in Pepperet al. (1979) and Brandtet al. (1996a). Time integration of sub-model 1 is carried out using a predictor-corrector scheme with up to three correctors (Zlatev, 1995).

The three-dimensional advection espe-cially is a very important process in transport modelling, so this submodel re-quires an accurate time integration method. Sub-model 2 is solved using the less

Figure 3

Accumulated total (wet and dry)131I depositions in Europe, 12 days after the start of the Chernobyl accident, calculated by using DREAM

[ 219 ]

Jùrgen Brandt,

(5)

expensive, but more stable implicit-method (Christensen, 1995; Brandt, 1998). The third sub-model is very simple and is solved directly. The combined model is applied on a polar stereographic projection, using an Arakawa A grid. The spatial grid-resolution in the Eulerian model is 25km625km at 60

degrees northern latitude and in the Lagran-gian model 5km65km.

The coupling of the Lagrangian model and the Eulerian model is carried out using the puff-coupling applied and tested in Brandtet al.(1996a). The model has been optimized to run on a CRAY C92A with an efficiency with

Figure 4

Time series of calculated and measured131I concentrations at six different measurement stations, during the Chernobyl episode. The time is in days after the start of the release. The stations are: A07 (Austria), CH01 (Switzerland), D06 (Germany), I03 (Italy), NL06 (The Netherlands) and UK04 (UK)

Jùrgen Brandt,

(6)

respect to peak performance of 50 percent, corresponding to 450 MFLOPS. A typical three-day run takes approximately 20 min-utes on this machine.

3. Some examples of model results

In Figures 2 and 3 some examples of model results calculated by using DREAM are given.

In these examples the mesoscale model MM5V1 has been used as a meteorological driver for the system. Figure 2 shows the surface concentrations of131_{I, nine days after}

start of the release from the Chernobyl accident.

The release period was ten days. In Figure 3 the accumulated total (wet and dry) deposition of131_{I, 12 days after start of the}

release, is given.

The calculated concentrations and total de-positions have been compared with mea-surements at 97 measurement stations in Europe, showing good agreement (Brandt, 1998).

A few examples of measured and calculated time series are given in Figure 4.

4. Real time predictions of air

pollution

The DREAM model is presently being implemented in a weather and air pollution forecasts system, THOR. The goal is to be able to calculate real time forecasts of air pollution episodes in connection with major nuclear accidents or episodes of high levels of air pollution such asO3. For photochemical

air pollution episodes the Danish Eulerian model (DEM) is used (Zlatev, 1995). An

Figure 5

Ozone concentrations in The Netherlands during an episode, 30 July 1994, 17.00 UTC, calculated by using DEM

[ 221 ]

Jùrgen Brandt,

(7)

example of a photochemical episode over The Netherlands in July 1994 is given in Figure 5. The Figure shows theO3concentrations

calculated by using the DEM (Zlatev, 1995). According to the draft of the new EU regulatives, the population should be informed when theO3concentrations exceed

90ppb. Furthermore, the population should be warned when the concentrations exceed 120ppb. As seen in Figure 5 theO3

con-centrations exceeded both critical levels considerably in this period.

The numerical weather forecast model and the air pollution forecast model are run at a powerful workstation, an SGI, Origin 200 with four processors and shared memory. The peak performance of this workstation is 1.44 GFLOPS. The performance of the air pollution forecast model, DEM, is 350 MFLOPS, corresponding to nearly 25 percent of the peak performance. When calculating real time forecasts of air pollution, it is important that all major processes (including data transfers and data processing) are carried out quickly. In Table I the estimated turn-around times for the THOR weather and air pollution system, using the present com-puter, for these tasks are given. A certain time is necessary for data transfer of 3D meteorological data. However, this is carried out automatically twice a day.

5. Conclusions and plans for future

research

Some application of the DREAM model has been discussed in this paper. The model has been developed for handling transport, dis-persion and deposition from accidental releases, both short-range and long-range. The model has been validated against measurements from the European Tracer EXperiment (ETEX) and the Chernobyl accident, with good results (see Brandt, 1998). DREAM, together with the Danish Eulerian model, is presently being applied in a real

time air pollution forecast system, THOR. The models will be further optimized on this system in order to minimize the turnaround time for the whole system. At present, data assimilation of real time measurement data in the model is not included. The model is restarted from concentrations calculated from the previous run. When performing real time predictions of air pollution, the initial conditions in the models are important. Therefore, some work will be carried out in this area in the future.

References

Brandt, J. (1998), ``Modelling transport, dispersion and deposition of passive tracers from accidental releases'', PhD thesis, National Environmental Research Institute, Depart-ment of Atmospheric EnvironDepart-ment, Roskilde, Denmark.

Brandt, J., Bastrup-Birk, A., Christensen, J.H. and Zlatev, Z. (1998a), ``Numerical modelling of transport, dispersion and deposition ± validation against ETEX-1, ETEX-2 and

Chernobyl'',Environmental Modelling and

Software, p. 15.

Brandt, J., Mikkelsen, T., Thykier-Nielsen, S. and Zlatev, Z. (1996a), ``Using a combination of two

models in tracer simulations'',Mathematical

and Computer Modelling, Vol. 23 No. 10, pp. 99-115.

Brandt, J., Mikkelsen, T., Thykier-Nielsen, S. and Zlatev, Z. (1996b), ``The Danish Rimpuff and Eulerian Accidental release Model (The

DREAM)'',Physics and Chemistry of the

Earth, Vol. 21 No. 5-6, pp. 441-4.

Brandt, J., Dimov, I., Georgiev, K., Wasniewski, J. and Zlatev, Z. (1996c), ``Coupling the advection and the chemical parts of large air pollution models'', in Wasniewski, J., Dongarra, J.,

Madsen, K. and Olesen, D. (Eds),Lecture Notes

in Computer Science, Applied Parallel Com-puting, Industrial Computation and Optimi-zation, Proceedings of the Third International Workshop, PARA '96, Vol. 1184, UNI-C, Lyng-by, Denmark, 18-24 August, Springer Verlag, Berlin, pp. 65-76.

Brandt, J., Bastrup-Birk, A., Christensen, J.H., Mikkelsen, T., Thykier-Nielsen, S. and Zlatev, Z. (1998b), ``Testing the importance of accurate meteorological input fields and parametrizations in atmospheric transport modelling using DREAM ± validation against

ETEX-1'',Atmospheric Environment, Vol. 32

No. 24, pp. 4167-86.

Christensen, J.H. (1995), ``Transport of air pollu-tion in the troposphere to the Arctic'', PhD thesis, National Environment Research In-stitute, Department of Atmospheric Environ-ment, Roskilde, Denmark.

Grell, G.A., Dudhia, J. and Stauffer, D.R. (1995), ``A description of the fifth generation Penn State/NCAR Mesoscale Model (MM5)'', NCAR/TN-398+STR, NCAR Technical Note,

Table I

Total time required for a three-day forecast on an SGI Origin 200

Task Time

Data transfer (meteorology) 60 minutes Meteorological model 105 minutes Air pollution model 5_{20 minutes}

Visualization 5_{20 minutes}

Total (approximately) 5_{3.5 hours}

Jùrgen Brandt,

(8)

Mesoscale and Microscale Meteorology Division, National Center for Atmospheric Research, Boulder, CO.

McRae, G.J., Goodin, W.R. and Seinfeld, J.H. (1982), ``Numerical solution of the atmos-pheric diffusion equation for chemically

reacting flows'',Journal of Computational

Physics, Vol. 45, pp. 1-42.

Nodop, K. (Ed.) (1997),Proceedings, ETEX

Symposium on Long-range Atmospheric Transport, Model Verification and Emergency Response, 13-16 May, Vienna, Office for Official Publications of the European Com-munities, Luxembourg, p. 249.

Pepper, D.W., Kern, C.D. and Long, P.E. Jr (1979), ``Modelling the dispersion of atmospheric pollution using cubic splines and chapeau

functions'',Atmospheric Environment,

Vol. 13, pp. 223-37.

Pudykiewicz, J. (1989), ``Simulation of the Chernobyl dispersion with a 3D

hemi-spheric tracer model'',Tellus,Vol. 41B,

pp. 391-412.

Zlatev, Z. (1995), ``Computer treatment of large air

pollution models'',Environmental Science and

Technology Library, Vol. 2, Kluwer Academic Press, Dordrecht, Kingston upon Thames and Norwall, MA.

[ 223 ]

Jùrgen Brandt,