" ; , * l t+ 'E i ih f f tEH|+J F ,158f i t2 i 51 63 lL '+ i ' i i t l +121 j

Papers in Nleteorologl and Geophl'sics \rol. l i . No. 2, pp 51-63. December l99J

Numerical Simulation of Regional Scale Dispersion

of Radioactive Pollutants from the Accident at the

Chernobyl Nuclear Power Plant

by

Takehiko Satomura, Fujio Kimurax, Hidetaka Sasaki,

Tomoaki Yoshikawa

Meteorological Research Institute, Tsukuba, Ibaraki, 305 Japatt

and

foshitaka Muraji**

International Meteorological Oceanographic Consultant, Shintomi,

Chuou-ku, Tokyo, 104 Japan

(Received \{arch 31, 1994 ; Revised November 15, 1994)

Abstract

Air concentrat ion and deposit ion of radioactive pol lutants in Europe after the Chernobl ' lnuclear pou.er plant accident is simulated using a long-range transport model developed in

the MRI. This model is composed of the rveather forecasting part u,hich rvas the routineregional weather forecasting model of the JN{A and the Lagrangian advection-dif fusion part.

The source data distr ibuted b5, ths ATN'IES project is used to determine the init ial releaserate from the pou'er plant.

Calculated concentrat ions of Cs-137 and I-131 in the surface level atmosphere agree rvel lu, i th observation. Surface deposit ion of Cs-137 has. however, poor correlat ion r.r, i th theobserr-ed deposit ion. Bad precipitat ion forecast of the r 'veather forecasting part of the model

and the dif ference betu'een the horizontal scale represented by observation and b1. simrrlat ionare considered to be responsible for this poor correlat ion.

*) Present aff i l iat ion: Inst i tute of Geoscience, Univ. of Tsukuba, Tsukuba, Ibaraki, 305 Japan**) Present aff i l iat ion: Energl ' Sharing Co., Ltd., Arakan'aoki-Nishi, Tsuchiura, 300-11 Japan

el99a by the Meteorological Research Inst i tute

5 2 Satomura T . , F . K imura , l l . Sasak i . T . \ 'osh ika i ra and Y. \ lu ra i i \ -o l . 15 . No. 2

i tation. This means that the wet depositionseems to be dominant for Cs-i37. On the otherhand, distribution of deposited I-131 rvas quitedifferent from that of Cs 137, so that the drl 'deposition can be expected to be more importantfor I -131. On the g lobal scale, Pudykiewicz (1990)reported the distribution of the radioactivepollutant simulated b-v a global scale predictivetracer model. He calculated the pollutant dis-tribution by both diagnostic mode and fullypredictive mode of the model and compared theresults with the observation. The model includesdry and wet deposition, but the amount ofdeposited radioactivit-v rvas not compared rvithobservation.

The mechanism of transport and removal ofthe radioactive pollutant rvould be similar to thatof the long-range transport of sulphate andnitrate, w'hich are the main precltrsors of acidrain. The accident gave Lts a chance to evaluatethe accuracy of the long-range transport modelfor acid rain such as the model used here.

In this study, the N{RI long-range transportmodel is evaluated in the case. of the Chernobylaccident. The emission and observational dataon air and deposited concentration of radioac-tive pollutants used in this studr' ' ,r 'ere the datadistributed in a project narned ATMES (JointIAEA ,/ WN{O CEC Atmospheric Transportl lodel Evaluation Studv). This project wasstarted a ferv months after the accident. and theaim of the project \\'as to ret,ie'nv and inter-calibrate atmospheric transport models usingthe observational data of the Chernobyl accident(Klug, e la l . . 1991a). In t l . r is pro ject , a t ime ser iesof the released anrount of radioactive pollutantsu.ith their effective heights and time series ofsurface air concentration and of depositedamounts of rzrdioacti,,-e pollutants obsen ed overEr-rrope u'ere distributed. Though the amountand the fornr of the emitted substances are notprecisell ' determined (Petro,, ', private communt-cation) and up to 3 hours ambiguity remains inthe obsen.ation time (Raes e/ a/., 1989, 1990),these distributed data are believed to be the bestdata set over Europe on the Chernobyl accident.

1. Introduction

Recently, regional and international con-cern on air pollution has increased, and u'ith itneeds for tools to evaluate long-range transportof pollutants have augmented. In the Meteor-ological Research Institute, a Lagrangian airpollution transport model has been developed tosimulate air pollutants related with acid rainover Eastern Asia. Although databases ofreleased and concentrated pollutants averagedover the time scales of months are wellestablished (for example, Johnson, 1983; Air andEnergy Engineering Research Lab., 1989), obser-vation of air concentration and deposition ofpollutants released from a u'ell-defined source ishardly available; this kind of data is indis-pensable to evaluate and improve the long-range transport model.

On April 25, 1986, a serious accident occur-red at the Chernobyl nuclear pov,'er plant, USSR,and a large amount of radioactive substanceswere released into the atmosphere. Because ofits serious impact over Europe, studies of themovement and deposition of radioactive sub-stances have been completed (for example,Kimura and Yoshikawa, 1988; Hass e/ al., 1990 :Puhakka et al., 1990). These studies haveestablished the following: the radioactive cloudsreleased by the main explosion of the plant u'astransported northward very fast in the u'armconveyor belt and deposited over northernEurope in the first several days after theexplosion. On 30 April, the radioactive cloudn-rarched to central Europe. At the same time, apart of the pollutants transported to the middletroposphere moved eastr,vard at high speed andreached Japan on 4 May 1986.

Deposited radioactive pollutants have serr-ous effects on l ives on the surface and, therefore,they should also be analyzed and predictedprecisely. h.r the case of Chernobyl, wetdeposition is believed to be important in thenorthern and central European areas (Hass e1al,1990 ; Puhakka et al., 1990). In the UnitedKingdom, Clark and Smith (1988) investigateddistribution of dry and wet deposition of I 131and Cs-137 after the accident. The amount ofdeposited Cs 137 correlated rnell with precip-

Numerical Sinulation of Regional Scale I)ispersion of Radioactive pollutants 53

2. Numerical Model

The model consists of tu'o parts: a weatherforecasting part r,vhich predicts meteorologicalvariables, and an advection-diffusion part ofpollutants u,'hich uses meteorological variablesforecasted by the first part as input data.

2.1. Weather Forecasting Part

This part of the model is almost the sameas the old version of the routine weatherforecasting model of the Japan MeteorologicalAgency (JMA). Deta i ls of th is model aredescribed in a technical report issued by theNumerical Prediction Division of JMA (1986).We modified the topography and physicalparameterization in the original forecastingmodel to apply it to the Chernobyl accident. Inthis section, a brief description of the model isg iven for rcaders ' convenience.

2. 1. 1. Couerning Equations

We use the primitive equations in o-co-ordinates and the polar stereographic projectionat 60'N. The equations are:

o I n \ o d , * , d l ; r o \ f ti l \ , , , , t ' ) ' d . r J t t ^ t t ) ' ; r . \ v ^ t t ) ' o n \ , , , r " 1 , , , , r t , l )/ i t t r ou t \ t Jd ' ; r 0 d l ' ^ . . l i n ' r \\/ '

,)., ' t '

d.r ' ) ttr ?,..t ' ",,, i .r '

' f" ,,r.(;;,J.

o t : \ o . ; : ' - - '

; , \ , : , , ' t ) i r ' t r* ' ) --" ( '* ' ' #( ' , ' ,1, ' ) - ,1, , , , , . , ,I 6 n t d t n \ ; r 6 6 t r 1 d P ^ ^ ( i i r r\ ' '

d . r ' ' "

d j ) - t t t 6 t ' ' o

t t r , ) . Y ' ' - -

, , , t \ j o ). l r - r i : ^ ' - , '

i ( " , u \ i \ t t *o ) i 1 t *u ) - i ( ' ' " , u )o l \ )1 t ' I o .Y o- \ ' oO \ i l t ' /

n f r . , c o t H ( 3 )

' o t t t t o t ' o t t t 2 P ' 6 0

o t : r \ a ; - d r t ro to l \ t i , '

( l ) ' 6 x \ t r ^ q

) ' 6 : . \ , * q t

' 6 n \ , , , r ' t 1

. ; " " p J I : . { - l l

I a , l r rl n ' m ' 0 6

o t t \ 6 t f 6 > * 6 / t r o ti , \ , , , ' / ; ; ; ; ;o( , ; , ; , ) r ) (5)

A, l ' . . , , d l )^- L D o ' - ' ( 6 )()o 06

rvhere the symbols are as follows

pressure at the model top,surface pressure,

r u/ mand n ui m,respectively,water vapor mixing ratio,vertical eddy flux of watervapor,vertical eddy flux of sensibleheat,amount of decrease in watervapor by condensation orn r p c i n i f a t i n n

Reynolds stress by verticalwind shear,horizontal eddy diffusion of u,

1* sir-r 60"t t I

1 + S 1 n p

9is the latitude,zr: P, -Pt ,pr

pr

P : p i p a ,

u ano u'

q

E

N{

T

F,, F", FA, FN

u, d, and q, respectively,and the other symbols are conventional.

2.1.2. Grid Systems and DiJferenceSchemes

The calculation domain and the topographyin the model are shown in Fig. 1. The domain iscovered horizontally by Arakawa's B grid (73 X55). The grid size d is \27 km at 60'N. Thedomain is divided into 16 layers in the verticaldirection as shown in Fig. 2. Finite differenceschemes in space are based on Arakawa andLamb (Igi7), and time difference scherrreis Tatsumi's economical explicit scheme"(Tatsumi, 1983).

2. 1.:1. Bottndary Conditions

At the upper and the lower boundaries, weassume ri - 0. At the lateral boundaries, we applythe Hovermale condition by adding the followingterm to the right-hand sides of eqs. (1) (5):

r<( f ) 'v l re Ae) , ( 8 )

q / Satomura ' f . , F . K imr r ra , I l . Sasak i , T . \ 'osh ikawa and Y. N ' lu ra t i

Fig .1 . Domainof in tegra t ion .Coast l inesare inc l i ca tedb l 'do ts .Conto l r r fo r te r ra ine leva t ion(so l id l ine) beg ins a t 200 n i , a t in te rva ls o f 200 m. Po in ts labe led

"S" . ' I lu " . \ l " ,

"A" , "B i "

and "Be"

are Stockhohn, Budapest, \ lol . Att ikis. Bi l thoven and l ler l in. respectively.

V o l . 1 5 . \ o . 2

u,.here A is a ph-vsical r.ariable, A is a prescribed

boundary value, / is the distance from theboundarl ',

K = 4 x 1 0 " s e c t ,

L : 6 d ,

and,

1 2 I I I rv n . r , . \ ' , . f ' f , . d . . \ d

- - 1 , r . / . . \ . l d , . ) J , - , r

A r . r , a . : ' a t 4 r l r r , . r , ( 9 )

2. 1. 4. Parameterization of PkysiculProcesses

a) Horizontal and vertical diffusion

As the vertical eddy' diffusion, we

emplol ' the c losure model of level 2 (Mel lor

and Yan'rada. 197,1). At the lowest layer,Monin-Obukor-'s similarity theory is appliedto determine vertical f luxes from the lorverboundarl '. The surface temperature ispredicted. As the horizontal eddy diffusion,an artif icial , lth order diffusion is applied asfollorvs :

F - , , - - I ( r . \ ' { rV (V ?a ) } Ka r " rV lV , ' (V a ) } '

t r , - K r .V { ; rV (V ?b ) } , ( 10 ;

u'here rz is horizontal velocity u or r., b isscalar variable 0 or q,

-\unterical Sintulation of Regirnal Scale Dispersion of Radioaclir.e It)l lgtants

0

o = 0 . 0 5

0 . 1

0 . 1 5

0 . 2 5

N ? R

0 . 4 5

; = 0

l l r l. 1 t , - . / ) . r . { , . ' d L r r ' , . . .

I i h . . 5 x 1 ( ) r m 2 l s e c .

K o , . - 1 0 x 1 0 4 m 2 i s e c .

These tern.rs are ir.rtroduced to dissipatenoise rvith a ver]' short r',,ar.e1ength, andhave no ph1's ica l meaning.

b) Conr,ection and Condensation\Ve use t"vo t1'pes of parameterizati<n

for precipitation process in the model : thelarge-scale condensatior-r parameterizationand the moist conr,'ective adjustment. Theformer prevents supersaturat iorr and thela t t e r keeps the a tn rosphe r i c l apse ra tel re r u een rhe d r r , ad iaba t i c and the mo is tadiabatic lapse rate.

o.ss-- / - - ' - - - l

0 . 6 4

o . 7 2

o . 7 9

o . 8 5

0 . 9o . 9 40 . 9 71

Fig.2. \ 'ert ical grid s1'stem in a.coordinate.

2.1.5. Input Data

At the time of the Chernob-vl et'ent, a globalobjectir,e anal-vsis rnas executed everl ' 12 hoursb1- lhe JMA. It has 2.5 degrees resolution both rnlatitude and longitude on 16 pressure levels.From this objective analysis, \\ 'e prepare init ialand boundarl ' data for our model bf inter-polat ion.

Sea sr-rrf ace temperature is interpolatedfrom global clirnatological data rvith 1 degreeresolution in latitude and longitude prepared byAlexander and N{oble1' (19i6).

2.2. Advection-Diffusion Part

Advection and diffusion of radioactivepollutants are described b1- a random-u'alkmodel in the same coordinate s)'stems as in ther.veather forecasting part. Winds, precipitationand diffusion coefficients used in the randomr'valk n-iodel are predicted bl' the 'nveather

forecasting model. The variables are stored at

56 Satomura 1 ' . . l ' . K i rnura , I l . Sasak i

every one hour integration of the forecast model,but the data of the first 12 hours are not usedbecause the so-called spin-up time of the f orecastmodel is about 12 hours. Therefore, w.e integrate36 hours in the forecasting part to create inputdata for 2,1 hours in tegrat ion of the advec-l : ^ - l : f f . . ^ : ^ - ^ ^ - +L l ( r l l _ u r l I u ) l u l r p a r r .

2. 2. 1. Gouermng Equations

The equations of the random rvalk modelare :

t d xI ; , - - r t .

I o.t

) d j '

) ; n ' '

I

| lo - a ' t t .l d t

u'here X, Y, ar.rd o are positions of a particle rnthe three-dimensional space. A random variableR is defined as

T. Yosh ikar i 'a and Y. \ lu ra j i Vo l . t5 . \ ' o . :

are satisfied: the height of the particle islor ier than a prescr ibed cr i t ica l height / l -

0.99 in a-coordinate, and a number givenrandon.rl-v for each particle and each timestep is smaller than a value Pat,1, defined as

D * 2 K lr \ - r . \ 6 /

( 1 2 )

rvhere Kz is the verticai diffusion coefficientderived from the forecast model, D/ is the timeinter.u-al in the random r,r 'alk model, and the signof the right-hand side is chosen randoml-v foreach particle ar.rd each time step.

The ef fect of hor izonta l subgr id-scalediffusion is not included. The init ial plume isbroadened bl' the horizontal and the verticalu'ind shear.

2.2.2. Time Difference Scheme

A s imp l i f i ed Runge -Ku t ta me thod t semployed for t ime integration of three-dimensional advection terms u.ith a long timestep dl :10 min. We enrplol.the Euler-backvn ardscheme for random vertical diffusion terms rvithshorter t ime step Dr-2 n.rin in order to avoidartif icial convergence or divergence of particlesat the la1'er where Kz varies abruptl) '.

2. 2.3. Parameterization for Depositiott

a) Dr) 'deposi t ionA particle is assumed to be deposited by.

a drv process if the following tu'o conditior.rs

( L i )

where I/a2 is the dry deposition velociry11a,n-- 100m. In this study, I/a*, :1 X 10 3 mr sis used for I 131. Based on the measurementin the United Kingdom, Clark and Smith(1988) est imated Vt1,-3X10 : rmr 's for I -131.They also shou-ed, hovn'ever, that the valuesof deposition velocities \\ ' 'ere lvidelydistributed. Therefore, W,1,-yX 10 3 mr/s forI-131 used in this studf is not an artif iciallysmall r, 'alue rvhile this value is three timessmaller than that thel' estimated. ForCs 137, Vt1,:5X10 a m,is is used and is thesame as the estimate b1' Clark and Smith( 1 988) .

b) Wet depositior.rWet deposition uf the pollutants n-as

estimated every one hour. The particles aredeposited on the surferce u,ith a probabil it l 'defined b1'

Ltuao, Ca,ro ' d l 'RI t .

rvhere Cr.,, is the n.e,t deposition rate, D/ theestimated duration cif u-et deposition (one hour)and RR is a prec ip i ta t ion factor .

The deposition rates for pcll lutants arediff icult to estimate. since the1. may vary u'iththe phase , , f thc nrater ia l and meteorologicalconditions. For SO: and SOr ', n'et depositionrates areest in . rated tobe intherange of 3 - - 50x1 0 o s ' a n c l ( ) . 5 - 2 x 1 0 t s ' ( E l i a s s e n , 1 9 8 0 ) .Pudl'kien'icz (1990) assumed Ca,.p-3.5X10 ss 1

for Chernobl' l radioactive pollr"rtants. In thisstr-rdy', 5 x 10 ''s r is assumed for the u,etdeposi t ion rate of I -131 and Cs 137.

The precip i ta t ion factor RR is def ined as:

o for particles at grid pointsRR:1 if the predicted precipitation rate

is greater than 0.05 mm,,hour,RR:0 otherl.,, ise,

o f or particles between grid pointsRR is interpolated horizontally using RRs

, , I ; " p d tI d d e p

I laep

( 1 1 )

1 9 9 1

2.2..1. Source Terms

a) Release heightThe release heights described in the

ATMES technical specification documentincluded in Klug et al. (I991b) are convertedto o ler,'el as :

300m a-0.965600m-- o:0.9541500m-_- a:0.853

b) Other source term characteristicsWe assume the normal distribution rn

both horizontal and vertical directions. 'fhe

standard deviation used in this model isI27kmi2:63.5km in the hor izonta l d i rect iorrand 0.05 in the a level (vertical directior-r).

Release rates of radioactive aerosol,specified in the ATN,{ES Protocols, areshown in F ig.3. As a correct ion ofd i s i n teg ra t i on ,8 .02 da1 ' s i s used fo r ahal f - l i fe da-v of I 131.

2. 2. 5. Concentration calctthtiott

The number of released particles is 20000 tneach for the cases of I 131 and for Cs-137.

- fo

obtain the surface air concentration at alocation, particles rvithin a box of '2 horizontalgrid lengths 1--250 km) fronr the location and 0.1r-er t ica l length in a coordinate ( -1 km) arecoLlnted and the number of particles is convertedinto the concentration. To obtain the depositedconcentration at a location, the particles lvithinan area of 2 horizontal grid lengths from thelocation are counted and converted into theamoLrnt of deposition.

\ u n r e r t c a l S i m u l a t i o n o f R e g i o n a i S c a l e D i s p c r s i o n o I R a d i o a c t i y e f o l l u t a n t s

a r n ' , n d f h o n a r f i n l a

l f lq1Day

2 . 0 1 A

1 . 5 l 0

I . o 1 0 '

5 - 0 t 0

0 . 0 1 0

; ::;]] ' Source Dara

2 6 2 1 2 A 2 e . o o l n 2 3 4 s 6

Fig.3. Time change of Cs-137 (black bar)and I-13l(gre-v bar) release (unit in TerraBq/da1) b1' the Chernobyl accident. Sol idl ine u' i th asterisk is the effect ir ,e ini t ialplume center of mass height (unit in meters).

3. Results

3.1. Movement of pol lutants

Figure '1 shows forecasted r,veather mapsand radioactive clouds. The rnodel predicts then 'ea t he r na t t e rn r r e l l comna red r r i t h t he

objecti,, 'e ana11,5is of the JMA : centers ofc1-clones and antic5'clones are predicted u.itherrors less than about 500 km and 4--8 hPaercept t$'o cases. The trvo exceptional cases areor,er the Adriatic Sea on April 29, u-here themodel predicts a too strong c-vclone, and oversor"rth England on NIa.v 3, u.here the model'predicts a too n eak cy clone (not shown)., \ l t ho r rgh the s t ru r rg ( r r ' eak t cvc lone o \e r t heAdr ia t i c 5ea ( so r r t h E r rg la r rd r ma \ a t t r ac t t heraclioactive ckrud stronger (rieaker) than theobservation, the model is expected to representthe or.-erall movement of the radioactive cloudin good accurac-v.

The radioactive cloud released at Chernobl'1on April 26 moved northr,vestu.ard and reachedthe Srvedish coast of the Baltic sea in only trvodays. Thus, the speed of the pollutants averagedin these first two days is about 10 m,,'sec. Thisrapid movement indicates that a warm conveyorbelt formed b], a lorv pressure zone o\rer \',,estern

ffit\il"r!1\

58 Satomura ' l ' . , F . K inura , H. Sasak i , T . \ 'osh ikau 'a anc l \ ' . \1u-a t i

l

Fig .4 . Forecas ted sur face pressure map ( le f t co lumn)and pos i t ion o f Cs- i37 par t i c lesin the air (r ight column) at 00 UTC (a) Zf i Apri l ; (b) 1 N,Ial ' ; (c) .1 N{al ' : (d) 6 N{at ' .Contorlr interval of surface pressure is I hl,a.

\ ' r r ] . 1 r . \o . l

( a )

( b )

( c )

l,,''',i : '

( d )

191.1

Europe and the nurth Atlantic Ocean trans-ported the pollutants. This rnovement coincidesr,vith the trajectory anall.sis by Puhakka e/ a/.(1 990) .

After April 30, the pollutants released early-on entered the cold section of the c1'clone oversouthern Europe and thel' u'ere transpurtedsouthwestn ard b1 a' cold conveS.or belt tt iAustria ar.rd Sli'iss. On the other hand thepollutants released later moved in the southerl-vdirection and reached or,er Greek, Turkey andthe Mid-East countries.

Po l l u l an l s re leased a f t e r ) l a1 ' J m ig ra tedover Eastern Europe in a fer",. da1's and graduallr-moved further to the East.

3.2. Comparison with surfaceobservations of air concentration

Obserr. 'ed data of radioactir,e pollutants atabout ninetl ' stations for air concentration andabout thirt].for deposition \\rere distributed b1'the ATMES steering committee (Raes c/ a1,1989) and stored in the data bank named RENIData Bank (Raes el al., 7990). \Ve selected fourstations Stockholm, Mol, I3udapest and Attikrsfrom the distributed l ist of obserr.ation stationsand compared observed pollutants rn'ith calcu-lated values. \Ve chose the pcrints because thelocations of these four stations ranges oyer the."vhole European Continent (Fig. I ) and the)' areexpected to be t1'pical in the area over rvhichdense pollutant cloud flor,r,ed. Figr-rre 5 shor','sobserved and calculated surface air concen-t rat ion of Cs-137 at Stockholnr , N{ol , Budapest^ . - r A + l : 1 - : ^ T L : . r : - - . . - . ) s h o , , l s a g o o d c o r r e l a t i o nd t l u n L t 1 K t 5 . r r 1 r 5 r r g u t c

betrveen simulated concentration and obser-vation : init ial arrir.al day of pollutants (exceptat Budapest) and the tendencl ' to increase ordecrease in time at each location agrees rvithobservation. Looking int<t the detail. horver,er,lve notice some discrepancies bet',veen obser-vation and the model results, especiall l . under-estimation of simulated concentration on N{ay'2 4 at Stockholm and after \Ia1. 7 at \{ol. Wealso notice the delal- of the init ial arrival t inrein the simulatior.t at Budapest as mentionedabot'e.

The surface air concentration of I-131 rs

\ rn rer ica l S i l t r r la t ion o f I leg lona i Sca le D ispers i r rn o f l lac l ioac t ivc l ,o l lu tan ts

shon'n in Fig. 6. Obserr,ed concentration patternsare almost the same as in F-ig. 5, rvhile theconcentration of I-131 rvas denser than that ofCs-137 (note that the maximum value of theordinate in Fig. 6 is one order larger than thatin Fig. 5). Our rnodel simulates the air con-cer.rtration of I-131 rvith almost the same accu-rac.v as that of Cs-137, and the discrepanciesmentioned abor.e sti l l exist.

The effect of r.ertical subgrid"scale diffu-sion is included in the model b1' employing arandom-'"r'alk method as eq. (12). The effect ofhorizontal subgrid-scale diffusion is, hou'ever,not included in the diffusion part nor in ther", 'eather forecasting part. The init ial plume isbroadened onll 'b1'the horizontal and the vertical',vind shear.

' fhe pollutant cloud is, therefore.much narrolver than the other investigations (forexample, Puhakka et al., 1990, and Hass e/ al,1990). We sr-rspected that the discrepancy of theair concentration betu'een the model and theobservation mentioned above resulted from thislack of horizontal diffusion. Son-re additionalsimulations were, therefore, performed to studl't he e f f ec t s o l subg r i d - sca l c ho r i zon ta l d i f f u -s i on . Recause no re l i ab le pa rame te r i za t i onscheme for models using 8x -- 100 km is known,u'e used constant horizontal diffusion coef-f i c i en t s o f 2x105 , 5x105 , 1x10 ' r , 2X106 m2 isec .Ho',vever the results did not indicate clearimpror-ement : the results '"1'ere improved atsome points but had no effect at other points (notshor,r, 'n). Probabl-v, more sophisticated para-rneterization for the subgrid-scale horizontaldiffusion is required to improl'e the modelperforniance. A long-range transport exper-iment performed after careful preparation u'i l lbe great help for choosing the scheme of thehorizontal subgrid-scale diffusion.

Trvo other factors also affect the horizontalrv idth of the pol lu tant c loud: the in i t ia lhorizclntal and r,ertical u.idth of the source andthe number of released particles. If the init ialrvidth is broadened or if the number of particlesis increased, the sir.nulated pollutant plumer.r.ould be broadend.

'fhis broadening has a

potential to dissolr,e some part of the abovediscrepancies in general. In this paper, however.the effects of r-ariation of these parameters are

Satomura ' l ' . .

F . I ( imura . I l . Sasah i . T . Yosh ikau a and l

t o '

t o '

d'-'i

ro - '

t o - t

V o l . 4 i , N o . 2

t o '

t o t

- t o o

) ro - '

] ro - 'I

to - '

t o t

1o '

1 o '

i ' , ^-'

'; 1 0 _

t o - o

t o t

1 o '

t o o

; , " - r

- 10- '

t o '

t o - t

t o '

r o o

- t o _ '

I; 1 0 - '

t o _ '

t o - t

a / 2 5 / 4

( b )l\{oL

1-131 Arr Concenlrat ion

I

I

I

I

4 / 2 4 4 / 3 0 5 / 2 5 / 4 5 / 6 5 / 8

1 0 -

t o t

- 1 0 '

E'o- '

i

1 0 '

1 o _ 5

Fig.5. Air concentrat ion of Cs- 137 at (a) Stockholm, Su.eden , tUi f t" f . Belgir.rrn :(c) Rudapest, Hungarr. ; (d) Att ikis, ( ]reece. Sol id l ine is the obsen'at ion anddashed l ine is the rnodel result.

1 0 _

t o o

. - to_ t

; r o '

t o '

t o - t

t o '

t o o

;- l o ' �

,i ro-'

r o '

t o - 55 / 2 5 / 4 5 / 6 4 / 2 0 4 / 3 0 5 / 2

BUDAPESTI-131 Air C0ncentnlion

4I

I

I

I

4 / 2 4 4 / 3 0 5 / 2 5 / 4 5 / 6 5 / 8

ATTlKISa s- I J l Arr G)nccnlral ion

Fig.6. Same as Fig. 5, except for I-131 air concentrat ion

Numericii l -Sjrrulation of Regional Scale Dispersion oi Radioactivt, Pollutants 61

not discussed in detail for the f<tllorving reasons:thesignifir:ance of source-term specification rvouldbe lost b1'broadening the init ial source u'idth,and the discrepancy found at Stockholm rvouldnot be improved b1- the increase of the releasedparticle number as realized in Fig. ,1c.

3.3. Comparison with observations ofdeposition

Simulated and observed daily depositions ofCs-i37 at Bilthoven and at Berlin are shou.n rnFig. 7a and b, respectivell.. The simulateddeposition does not correspond u'ell w.ith theobservation, although our model sirnulated u,elithe air concentration at these tr,vo points asshown in Fig. 7c and d. This discrepancv rsprobablf induced by the follor,ving two factors:the poor precipitation prediction in the modeland the difference in the horizontal scalebetweer.r obsenation and the model. The dailyprecipitation at Bilthoven and Berlin predictedb-v the model shor,ved litt le correlation with theobservation as sho\\.n in Fig. 8. Because mostdeposition of Cs-137 occurred in u'et process( l Iass e/a/ . , 1990 ;Puhakka et aL. .1990), th is poorprediction of the precipitation is undoubtedilone cause of the poor prediction of the depo-s-it ion. On the other hand, the amount of r,verdeposition of the radioactir, 'e pollutants strongll 'depends on the place as u.el1 as precipitation:models must simulate the right position oflocation and time of precipitation and pollutants'arrival to predict the r,t,et deposition accuratell ' .If the precipitation and the lvet deposition havea rather small scale, good prediction of rvetdeposition b-v the mesoscale model $.ith ax = 100km becomes diff icult even in the case rvhere themodel predicts averaged precipitation rvell. Inthis sense, observed values do not represent l i 'etdeposition in the area u'ith the horizontal scaleof model 2 grid lengths (- 250 km).

To improve the precipitation forecast, anerv long-range transport model rvhich uses aner,v routine 'uveather frtrecasting n-rodel of theJMA as the u'eather forecasting part is beingdeveloped. In this model, both the horizontal andthe vertical resolutions are improt ed to simulatefine-scale phenomena.

4. Summary

As a case studl', u,'e simulated the mor,e-ments and deposition of radioactive pollutantsreleased by the Chernobyl accident. The model\\,as composed of the weather forecasting partthat was the routine regional n'eather fore-casting model of the JMA and the Lagrangianadvection-diffusion part. The source data dis-tributed by the ATMES project was used. Inspite of there being no horizontal dif iusion in ourdiffusion part, the simulated surface air con-centration of Cs-137 and I-i31 showed goodcorrespondence u'ith the observation. Theeffects of subgrid-scale horizontal diffusion wereexamined by extra simulations and it was foundthat the results r,",'ere improved at some pointsbut deteriorated at the other points.

Simulated deposition had poor correlationrvith the observed deposition. We consider poorprecipitation forecast and the locality of ob-servation are responsible for this discrepancy.

References

Air and Energy Engineering Research Lab., 1989:'Tlte

1985 IVAPAP emissions inuentory(Versiott 2) : Deuelofment of the annualdata and modelers' tapes. F.nvironmentProtection Agencl', Research TrianglePark, pp. '2- ' ) -2 3.

Alexander, R. C. and R. L. \ Ioble l ' , 1976:Monthlyaverage sea-surface temperature andice pack l imits on a 1'global grtd, Mon.Wea. Ret,., 104, 1.13-104.

Arakaq,a, A. and V. R. Lamb, 1977 : Compu-tational design of the basic dynamicalprocess of the IICLA general clr-culatior.r model, Metltods in Comfut-tntionul Physics, 17, 173 265.

Clark, M. J. and F. I l. Sniith, 1988 : \\ret and drydepos i t i on o f Che rnoby l r e l eases ,lVature, 332,245-249.

Eliassen, A., 1980: A revier,v of long-rangetransport modeling, I. Appl. Meteor.,19, 237'�2,\0.

l lass, If., NI. Nlemmesheimer, H. GeiF, H. J.Jakobs, N{ . Laube and A. Ebel , 1990:Simulation of the Chen-robyl radio-acti'n-e cloud over Europe using the

B i l t h o v c nCs-1.37 Dajh Deposi l ion

Satomura ' l ' . ,

F . K imura . FL Sasak i . T . \ 'osh ikau a and \ - . N lu ra j i

1 0 -

t c '

- 1 o - '

; 1 0 - '

i o _ n

t o - t

Fig.7. Daily deposit ion of Cs-137 at (a) Ri l thor.en, Netherlands, (b) Berltn,Gerrnanl ' , air concentrat ion of Cs-137 at (c) l l i l thor.en and (d) Rerl in. Sol idl ine is the obsen.ation and dashed l ine is the model result.

Fig.8. Daily precipitat ion at (a) Bi l thor.en and (b) tserl in.Sol id l ine is the observation and dashed l ine is the rnodel result.

\ L , l . 1 i r . N , r . l

BILTHOVENCs l3J A i r Concent ra l ion

B E R L I NCs l -1? Air Concentrat ion

Numerical Sintulation of l legional Scalt Dispersion o{ Radioactir.e Pollutants

EURAD mode l , A tmos . Enu . , 244 ,673 692.

Johnson, \V. B., 19811 : Interregional exchangesof air pollution : ^v'Iodel t1'pes andappi icat ions. J . Ai r Pol lu t . Contro lAssoc., 33, 563 571.

Kimura, F. and T. Yoshikau'a, 1988 : Numer icalsimulation of gkrbal scale dispersion ofradioact ive pol lu tants f rom the acci -dent at the Chernob5.l nuclear powerplant,./. Mett,or. Soc. -Jafan,66, 489-495.

Klug, W., G. Craziani. G. Grippa, D. Pierce andC. Tassone. 1991a: Attnospheric long.range transport ntodel eualuation study,Ispra.

Klug, W., G. Graziani , G. Gr ippa, D. Pierce and Raes.C. Tassone, 1991b : Atmctspheric long.range trnnsport model eualtution stttdl, ,Appe ndix t, o lunte, Ispr a.

Mellor, G. L. and T. Yan-rada, 197,1 : A hierarchlof turbulence closure models for plan-etarl ' boundarl ' la-vers, /. Atmos. Scz., 31,1 7 9 1 . 1 8 0 7 .

Numerical Prediction Di,, ' ision, 1986: Outline ofOperotio nal N'umerical Weather Predic

tion at Japan Meteorological Agency,Japan Meteorological Agency, Tokyo,93 p.

Pudvkier.r ' icz, J., 1990 : A predictive atmospherlctracer model, J. Meteor. Soc. Japan, 68,2 t3 .225 .

Puhakka, T. , K. Jv lhd P. Saar ik iv i and J.Koistinen, 1990: Meteorological fac-tors influencing the radioactive depos-it ion in Finland after the Chernobylaccident, I. APPI. Meteor.,29, 813 829.

Raes, F., G. Graziani, D. Stanners and F. Girardi,1990 : Radioactivit l ' measurements inai r over Europe af ter the Chernobylaccident, Atmos. Enu., 24A,909 916.

F. , G. Graziani , L . Grossi , L . Marc iano, D.Peirs, B. Pedersen, D. Stanners and N.Zarimpas, 1989: Rndioactiuity meas-urements in Europe after the Chernobylaccident Part I : 4ir, Commision of theEuropean Communities, EUR 12269EN, Ispra, 247 p.

Tatsumi, Y., 1983 : An economical explicit t imeintegration scheme for primitive model,

J. Meteor. Soc. Japan,61, 269 288.

7 t )t' '/ ) 4 t) R+h#g.FfrEFf&t: t 6trtN**E'*+h*aEEEEEfrffid)&IE, i : tz- 2 = 2

11ltt i#e. 4rf. l i -* I :rf l* . lae 4.TiI: . , , ' , .Jt i .&fi . f f i i i6Ev,. 'on

1- - t v t 1 "4 t ) lE . f ' ) J j t 'Ep ) i ' + i t r . nw l : / , t i l : l r i : F t4 ' 11 :1 . i r : l : t h tH .o , a -e . , , . l r t j i i 6+ t i + , i e l g f j t

f iE_ l : - r v ' - 1 , i i 5 i i f t ' f t . r ' , i 7 ) J i f l t rEF+d i r15 : t7 t s ! f { l r . i ; ' f i 1 .L / : . : , . q .7 )v t J , \ , " *< t j . : n f t t ) v -+>+r ' ) vaA?.*+:n7-* f i t j ' r 'v . , - , ' ->) : t?) i t t t i .na.<+7tva,r ; * ' lh t 'H.a iYly t ; IH- f b. , . , t f t i t i \n ' i , t i t++hETttb\ i [EEl r - a . ATN,IES-z ' r : ) t ? l - ( . | lCfr : l t_ l t1 f , l . i t f .7- , +H\ \ / : .

; fe .o{ f r+, ,E7"ru/ )cs-11t7! t - l31n. ; r .AtFi f f i ly iJaq, I l r l } " r ( a j - ! .h" f :11t , / : . r - r .L , i tE€lJ&E. i l l {L hb/; r,: t I l ttt '-, /:. i iat ' i | ]r l€ i lLt l!ft7K7-+ft/)Jt't irt0'tE: r, ftBiHrliF_ L c_i'tv :n:? ft?lt|\f,<t4 7 J - tv OtEv t tr;, i t Ei A n#n lF.lr|1 L +; z- ) ' It 6

4 rflr;ft|f, : iiiiEjt?jru:J<{,J.i1fi'r"r' i f'&) trL+t1.aifui+

r f lp t td: i f ik l a+ ) z :_7 ) >.? '