An operational puff dispersion model

Atmospheric Environment Vol. 26A, No. 17, pp. 3179 3193, 1992. 0004-6981/92 $5.00+0.00 Printed in Great Britain. ((3 1992 Pergamon Press Ltd

A N O P E R A T I O N A L P U F F D I S P E R S I O N M O D E L

G. H. L. VERVER

Royal Netherlands Meteorological Institute, P.O. Box 201, NL 3730 AE, De Bilt, The Netherlands

a n d

F. A. A. M. DE LEEUW

National Institute of Public Health and Environmental Protection, P.O. Box 1, NL 3720 BA,

Bilthoven, The Netherlands

(First received 30 March 1991 and in final form II June 1992)

Abstract--An operational puff dispersion model is described, which is adapted from the work of Van Egmond and Kesseboom (Atmospheric Environment 17, 265-274, 1983). It is intended for operational use at KNMI and RIVM in case of an accidental release of toxic or radioactive gases or particles into the atmosphere. The area covered by the model is restricted by the meteorological input that is available; the operational version will cover Europe. The main processes accounted for by the model are relative diffusion by wind shear and turbulence, advection as a function of space and time, the diurnal cycle of boundary layer height and stability, dry and wet deposition and linear chemical transformation or radioactive decay.

Input fields for the model are horizontal wind (every 3 or 6 h; on two or three levels), precipitation data (every 3 h; optional) and height of the boundary layer (or mixed layer; every 3 h; optional). The meteorological data used may be either analyses or prognostic fields. The source is defined by an effective emission height and a source strength as a function of time. The model calculates fields of concentrations near the surface and dry and wet depositions as a function of time. The model is tested on the cesium- 137 and iodine-131 dispersion over Europe during the Chernobyl episode (April and May 1986). A 10-day period is simulated and results are compared with measurements. Arrival times of the material are simulated very well, in most cases within 6 h accuracy. The observed air concentrations and model outcome show reasonable agreement.

Key word index. Puff model, Chernobyl.

Bq

C

(c)

C', C"

h

H I k~ Kz L P {2 RL R,

tLh

tLv t .

v, W e

17V h

N O T A T I O N

Bequerel; unit of radioactivity (number of dis- integrations per second) concentration (emission units per cubic metre; emu m- 3) mean concentration in the fully mixed boundary layer (emu m- 3) constants mean height of the turbulent boundary layer (m) effective source height (m) precipitation intensity (m h - ~) radioactive or chemical decay rate (s-t) eddy diffusivity (m 2 s- t) Obukhov length scale (m) power of the 'power law' for wind profiles source strength (emu s- t) Lagrangian autocorrelation function Lagrangian autocorrelation function for un- resolved turbulence Lagrangian time scale of horizontal velocities (s) Lagrangian time scale of vertical velocities (s) characteristic time scale of subgrid horizontal diffusion (s) time averaged wind vector (m s-1) dry deposition velocity (m s- t ) entrainment velocity (m s- ~) average vertical velocity at z=h (m s 1)

x , y

Zp

A

~7 r

fix

fly

~7 z

horizontal coordinates relative to the centre of mass of the puff (m) height of the centre of mass of the puff (m) scavenging coefficient (s- ~) scavenging ratio horizontal dispersion coefficient (m) dispersion coefficient in the alongwind direction (m) dispersion coefficient in the crosswind direc. tion (m) vertical dispersion coefficient (m)

I. I N T R O D U C T I O N

On 26 April 1986 a severe accident occurred at the nuclear power plant in Chernobyl . Fo r approximate ly 11 days large amoun t s of radioact ive material were emitted into the a tmosphere. Dur ing the days tha t followed, the con taminan t s were t racked over Europe by analyses of the meteorological s i tuat ion and a small n u m b e r of radioactivi ty measurements . In the following weeks more measurements became available and dispersion models were used to reconstruct the depos- i t ion and concent ra t ion pat terns over Europe (Van Egmond and Wirth , 1986).

3179

3180 G.H.L. VERVER and F. A. A. M. DE LEEUW

The experience during the Chernobyl accident has led to the conclusion that, in an early stage, dispersion models could give valuable estimates of the impact of an accidental release of material into the atmosphere. An emergency response model must have the following attributes:

(a) a realistic description of physical processes that govern the system being modelled;

(b) description of long-range transport on a European scale, but allowing sufficient detail on the national scale;

(c) operationality, that is, the provision, with the shortest possible delay, of adequate estimates of deposition and concentration fields based on analysed and prognostic meteorological data.

The last requirement limits the meteorological input of the transport model to on-line, routinely available data. Obviously, in the forecasting mode the com- puting time of the model should be within acceptable limits.

In the years prior to the accident in Chernobyi, the MESOS model (ApSimon et al., 1985) was developed. In the post-Chernobyl period a large number of models have been developed and applied to simulate the transport of the radioactive cloud (see the ATMES intercomparison study; Klug, 1991). Although on-line applicability was not included in the ATMES study, it is clear that not all models are operational in that sense. Most models will be used afterwards, e.g. for a detailed reconstruction of the dispersion of the cloud and/or for interpretation of measurement data. How- ever, for emergency-response modelling real-time applicability is, in our opinion, a prerequisite. Therefore KNMI and RIVM have developed a puff dispersion model. Recently the model was used for a period of 3 months, to make daily 24, 48 and 72 h forecasts of the dispersion of pollutants from the Kuwait oil fires, at the request of the WMO.

The model is an improved version of the model PUFF developed earlier by Van Egmond and Kesse- boom (1983). The model now handles non-uniform mixing heights that may be provided by a meteorological preprocessor. A new formulation is used for lateral diffusion of material and for estimation of stability and friction velocity. The number of levels of windfields is enhanced to three. The dispersion introduced by wind shear is taken into account by splitting the puffs. The model makes optimal use of the resolved scales of the meteorological parameters by projection and recombination of puffs. This method incorporates the advantages of an Eulerian approach in the La- grangian framework of the model.

The aim of the model is to calculate concentrations and depositions of hazardous material released acci- dently into the atmosphere. The model area may, in principle, be chosen freely. However, the model formulation is not suited for a scale smaller than ~ 25 km. Model results may be used to estimate the pollutant

exposure of the population directly by air or indi- rectly, e.g. through the food chain. The model pro- duces valuable input to methods describing the exposure pathway to estimate the dose.

Processes accounted for by the dispersion model are relative diffusion by turbulence and wind shear, advection of material dependent on height, dry deposition (dependent on stability) and wet deposition, linear chemical transformation or radioactive decay, the diurnal cycle of stability and mixing height.

The quality of the results of the dispersion calculations depends heavily on the quality of the meteorological input data. The dispersion model itself is independent of the origin of meteorological data. Input may be either analysed data or prognostic fields. In the operational setting, the KNMI Fine Mesh Limited Area model and the ECMWF global model will be used to provide the meteorological input.

In most applications, the model will have only a single source of one or two pollutants. With respect to computer resources the Lagrangian description is therefore appropriate. Also, near the source, proper detail is attained and effects of numerical diffusion are avoided. However, nonlinear processes (like more complex chemistry) cannot be accounted for by the model. The emission may either be instantaneous or continuous, with a strength varying in time.

In Section 2 the mathematical description of the model is given. Section 3 presents results of the model as applied to the Chernobyl accident, and a comparison with measurements is made.

2. DESCRIPTION OF THE DISPERSION MODEL

2.1. General

In the model an emission from a point source is divided into a number of Gaussian-shaped puffs con- taining mass M = QAt, where Q is the source strength (emission units per second) and At is the time step of the model, which is 1 h in the version described here. In the case of an instantaneous emission, only one puff is released. The source strength Q is input to the model, and may be updated every hour.

The mass of each puff is distributed over two layers, the mixed layer and the reservoir layer (Fig. 1). The presence of two layers in this model allows descrip- tions of the process of fumigation and the transport of pollutants at higher altitudes, decoupled from the surface. The height of the turbulent mixed layer in the model is a function of time in order to account for the diurnal cycle caused by insolation. Local differences in surface characteristics, like roughness, humidity, albedo and differences in insolation caused by cloud cover, make the mixing height a function of position also. These differences will be most apparent near a land/sea transition. Above land, the height of the mixed layer is typically 100 m during the night and 1500 m during daytime. There is no rigid upper limit to the height of the reservoir layer. It is determined by

An operational puff dispersion model 3181

i ii i i i

- 1 Mixed layer

O0 12 2 4

(a)

i ii iii

J

k O0 12

(b)

1v v

Reservoir layer

I

J

36

iv v

Reservoir

layer

J

2 4 36

Fig. 1. Illustrations of the vertical concentration profiles of a puff at different times. (a) Material is released above the mixed layer, i: Material is directly emitted into the reservoir layer (H>h). The vertical profile is Gaussian. ii: The dispersion coefficient of the mass in the reservoir layer is changed only slightly. Part of tfie material is entrained into the mixed layer due to a rise of the mixing height. In the mixed layer a uniform concentration profile is assumed. Close to the surface a concentration gradient is introduced by dry deposition, iii: At the time the maximum mixing height is reached, all material is entrained and uniformly mixed in the mixed layer. When at the end of the day turbulence decays, this layer is split into a reservoir layer having a uniform concentration profile and a shallow stable nighttime boundary layer, iv: During the night the concentration in the mixed layer will reduce to dry deposition, v: After a break up of the stable boundary layer during the morning hours, material from the reservoir layer is entrained into the mixed layer and rapidly mixed to the ground. (b) Material is released in the mixed layer. The height of the layers is sketched at the position of the puff. i: Material is emitted into the mixed layer (H < h). ii: After a few hours the material is well mixed, and a uniform concentration profile is assumed. Close to the surface concentration gradient is introduced by dry deposition, iii,

iv and v: as in (a).

the maximum height to which material from a single puff is dispersed. Therefore, the upper boundary of the reservoir layer depends on the vertical diffusion coefficient for material in the reservoir layer, and, if the emission takes place above the mixed layer, also on the effective emission height.

Within the mixed layer, turbulence is generated by wind shear and/or buoyancy, and material will be dispersed rapidly. The turbulence intensity within the reservoir layer is zero or very small, and material diffuses only very slowly in the vertical direction (Section 2.3). The effects of wind shear on horizontal puff growth is discussed in Section 2.4.

The advection velocity is determined by interpolation of the wind vectors to the centre of mass of the puff. The diurnal cycle of mixing height causes a material flux between both layers (Section 2.5) which, in conjunction with a different advection in these layers, enhances the horizontal spread of material (e.g. McNider, 1988). To take into account this effect of wind shear, a procedure is developed to split a puff into two separate puffs that will be advected separately, one in the mixed layer, and one in the reservoir layer (Section 2.7).

The mass in the mixed layer is affected by the dry deposition process; wet deposition affects masses in both layers (Sections 2.8 and 2.9). Radioactive decay or chemical transformation affects the mass in both layers (Section 2.10). A detailed description of the model is given by Verver et al. (1990).

Some processes are not taken into account by the model. In the first place these are physical phenomena not resolved by the model, like convective clouds, local effect on the wind fields and convective precipitation. Since advection is calculated using only the horizontal components of the wind, large-scale subsi- dence or upward movements are also not taken into account. However, these large-scale processes do affect the boundary-layer height used in the model, since it is derived from the meteorological model that at least partly resolves these processes.

2.2. Description of the puffs

The size of the puffs is characterized by two quantit- ies, the horizontal dispersion coefficient, a,, and the vertical dispersion coefficient, a=. The horizontal concentration distribution is assumed to be Gaussian and aximetric, the vertical concentration profile is either uniform, when the material is vertically well mixed, or Gaussian with corrections that account for the presence of the ground and the inversion above. Near the ground, the profiles are corrected for dry deposition.

Mixed layer. The concentration within the mixed layer is described by the well-known formula for a Gaussian puff, including reflection terms (e.g. Turner, 1969). After a few hours (the exact time depends on stability and mixing height), the mass-distribution in the mixed layer is considered vertically homogeneous, and only in the horizontal direction is a Gaussian shape prescribed.

Concentration near the surface. The concentration near the surface is affected by dry deposition. Close to the surface a constant flux layer is assumed. The concentration near the ground at a height z ( < 50 m) is estimated from:

c(z)= (c ) Vg(z = 50 m) (1) Vg(zl

where Vjz) is the deposition velocity (described in Section 2.8) and (c ) is the average mixed-layer concentration (see list of symbols in the Notation). It is implicitly assumed that the concentration at z = 50 m is equal to (c) .


Reservoir layer. When the emission takes place above the mixed layer, the Gaussian formula is used, without reflection terms. In this layer there is very little turbulence, so the vertical dispersion, az, will barely change.

When the mixing height rises, mass will be transported from the reservoir layer to the mixed layer. It is assumed that this material is instantly and fully mixed in the mixed layer (see Fig. la, stage ii). When at the end of the day the turbulent boundary layer decays and a shallow stable nocturnal boundary layer is formed, a uniform concentration profile remains in the reservoir layer (see Fig. 1, stage iii).

2.3. Vertical dispersion coefficients

Mixed layer. When a puff is released in the mixed layer, turbulence will spread the material vertically. Vertical puff growth is not described in much detail, since in most cases material becomes well mixed within the first 2 h of transport. Vertical puff growth is estimated by the empirical formula: az = ax b. The values assigned to a and b are taken from Pasquill (1974). The stability class is based on the Obukhov length and the roughness length according to Golder (1972), where L is estimated with the method of Holtslag and Van Ulden (1983) and Van Ulden and Holtslag (1985).

When tr z > h/x//~, the vertical mass distribution is considered homogeneous over the mixed layer.

Reservoir layer. Within the reservoir layer turbulence is almost zero. The vertical dispersion coefficient is calculated from tr 2 = 2Kzt, where Kz is set to a very small value: 0.5 m E s- 1.

Most of the vertical spread in the reservoir layer is due to turbulence during previous time steps, when the air was part of the mixed layer.

2.4. Horizontal dispersion coefficients

The horizontal dimension of a puff increases due to turbulence and due to shear of the horizontal wind direction and wind speed. In this section first the puff growth due to turbulence is described, then the equa- tions describing the disfoersion of material due to wind shear. Very briefly a description is given of horizontal dispersion observations by Gifford (1977, 1985). After this the relatively simple model formulation of horizontal dispersion is given.

Turbulence. Following Taylor's dispersion theory, due to turbulence the size of a puff will grow initially proportional with time t. Wherr the horizontal size of a puff is much larger than the size of the largest eddies, it will grow proportional with t 1/2. The distinction between these regimes is determined by the characteristic time scale of the largest horizontal eddies, the Lagrangian timescale tLia, defined by:

tLh=f;RL(z)d~, (2)

where the Lagrangian autocorrelation function is

usually written as the exponential function RL(t ) = exp ( -- t/tLh ). The horizontal growth rate is written as (e.g. Csanady, 1973):

fo da2 = 2v 2 RL(z)dz. (3) dt

For horizontal dispersion a wide range of eddy sizes contribute to dispersion. Unlike vertical eddies, horizontal eddies are not limited by the height of the boundary layer. Therefore, mesoscale and synoptic- scale turbulence of spatial scales up to several hundred kilometres may contribute to horizontal growth. Gif- ford (1985) cautiously suggests a tLh somewhere between 40 and 80 h to explain observational data from the Mt Isa smelter plume.

Wind shear. A second important mechanism for horizontal puff growth is vertical shear of the horizontal wind vector. In this paper this is referred to simply as "wind shear".

In the contribution of wind shear to puff growth, three stages may be distinguished. During the first stage, immediately after the release from the point source, the cloud will grow in the horizontal direction due to turbulence alone. In stable situations, with strong wind shear, this stage may be absent.

The puff will grow in the vertical direction also, and the shear of the horizontal wind to which it is exposed starts to contribute to its horizontal spread (Fig. 1, stage ii). Expressions have been derived for this contribution by Saffman (1962), Smith (1965), Van Egmond et al. (1984) and Venkatram (1988). They find for a puff growing unbounded in all directions: '

da2 =c'(dfi/dz)2Kzt 2 for t >> tLv (4) dt

where c' is a constant, tLv is the Lagrangian time scale for vertical velocities, d~/dz and the vertical exchange coefficient K z are constant with height. This stage of puff growth is not taken into account by the puff model, because it is restricted to a very limited time range, under most circumstances less than 1 h.

The rapid growth described by Equation (4) will come to an end when the material is fully mixed within the boundary layer (Fig. l, stage iii). The vertical size remains constant, which leads to a diminished growth rate in the horizontal direction. Saffman (1962) derived for this case:

dtr 2 = (d~/dz)2h4/(6OKz). (5)

dt

According to Equation (5), a puff trapped within the mixed layer grows like the square root of time (in a stationary meteorological situation, the right-hand side of Equation (5) is a constant).

Observations. Observations compiled by Gifford (1977) show a 2 proportional to t a up to 3000 s of travel time. An updated set of observational data is compiled by Gifford (1985). It shows that for travel times from


2 is proportional to 1 h up to approximately 4 days ay t 2. This is also found by Hanna (1986).

Model calculations by McNider (1988) show that the prolonged linear growth rate found by observations may also be explained by the diurnal cycle of the boundary layer height in combination with vertical wind shear.

The model. In the RIVM/KNMI puff model a simple formulation of the horizontal spread is chosen, keeping in mind the considerations given above. A nearly linear growth of a~ is found in most observations, even up to 50 kin. It may be explained by either the large range of spatial scales of horizontal turbulence (Gifford, 1985) or the diurnal cycle of wind shear in combination with vertical mixing (e.g. Pasquill and Smith, 1983; McNider, 1988). Beyond the spatial range of these observations, the model at least partly resolves the large-scale turbulence, due to the puff-projection mechanism described in Section 2.7. The model resolves also a part of the effect of wind shear on horizontal dispersion by its two-layer structure. Therefore, a linear growth rate is assumed for puffs smaller than approximately the grid spacing, as is found by observations and explained by subgrid wind shear and turbulence. Larger puffs will disperse main- ly by processes resolved by the model. To account for the contribution of subgrid-scale processes, a growth rate like the square root of time is imposed. The interpolation between these regimes is obtained by an equation similar to Taylor's formulation (Equation 3):

da 2 t

= c"[R,(z)dz (6) dt

where R*(t) = exp( - t/t,); t , depends on the grid size of the meteorological input data. It is the characteristic time it takes for a puff to reach the size of a grid element. In the current version it is set at 105s ( ~ 30 h), corresponding to a grid size of approximately 60 × 60 kin.

In the reservoir layer we set the constant c" to 0.4 m 2 s -z and in the mixed layer to 0.6 m 2 s -2, to fit the observations compiled by Gifford (1985).

2.5. Vertical fluxes between both layers

During the journey of the puff, material may be transported from one layer to the other. In the model, the only way for material to cross the boundary between both layers is through variations of the mixing height. The corresponding flux of material can be written as:

mass flux = w e e(h), (7)

where c(h) is the concentration at the height z = h in the mixed layer or the reservation layer depending on the sign of we, and we is the entrainment velocity. At z = h, there is a discontinuity in the concentration profile in the model. When wo is positive the concentration in the reservoir layer (c(z+h)) is taken; when w e is negative the concentration in the mixed layer (c(zTh)) is taken.

The entrainment velocity, w e, is defined as:

dh we - ~ - - ~ h , (8)

where ~'h is the mean vertical velocity at z = h. But in the model the vertical component of the wind is neglected.

2.6. Advection of puff~

Wind analyses or prognoses on several grids are available to advect the puffs. To calculate the traject- ory of the puff, the horizontal wind vector at the centre of mass of the puff is estimated by interpolation. Horizontally, this is done by simply taking the values from the nearest gridpoint at which wind data are available. Vertically, interpolation or extrapolation of the available wind vectors at the nearest two levels takes place, according to a power law. The power p is obtained from the wind fields that are input to the model.

The wind speed and direction at the centre of mass of a puff are obtained by vertical interpolation using a weighting factor of (Zp/Zl) p.

2.7. Puff projection and puff splittin9

Puff projection. Wind speed and wind direction are functions of space and time. Part of this variability is resolved by the wind fields that are used. The distinction between advection and turbulent diffusion is based on the resolution. Puff growth in subgrid scales is parameterized. The parameterization is described in Sections 2.3 and 2.4. Dispersion on larger scales is resolved and dealt with explicitly by the model.

When the horizontal dimensions of the puffs reach the resolved scales (the grid spacings) of the wind field, the puffs are projected on the grid. For each grid cell, a new puff is generated that will be advected separately. The large projected puff is deleted (Fig. 2). When two or more puffs overlap, their contributions to the mass in a grid box are added. The characteristics of the puffs, like layer heights, are weighted by mass for their contribution to the characteristics of the newly generated puff. This way optimal use is made of the spatial resolution of the meteorological input.

In the initial stage the number of puffs may grow rapidly, therefore puff projection is performed only

•Y

\ _ J

o r = 0.25Ax Ay

r ~

, . . ~ - -

Fig. 2. Puff projection.

AE(A) 26:17-I


once per simulated day. On the other hand, during a long simulation period, the number of puffs is largely restricted to the number of grid cells.

Puff splitting. Horizontal wind vectors will also vary with height (vertical shear of the horizontal wind). The daily cycle of the height of the turbulent boundary layer, in combination with the daily cycle of wind shear, will cause material to disperse horizontally (e.g. McNider, 1988). To account for this process, an algorithm is developed that splits a puff in two: a mixed layer puff and a reservoir layer puff. This will be the case over a land surface, at the end of the day, when stability changes sign and the nighttime boundary layer is formed. The two independent puffs will then be advected by a different wind vector (Fig. 3). In general a puff will be split once a day, depending on the meteorological situation.

2.8. Dry deposition

Deposited material may contaminate the soil and field crops, and may contribute significantly to human exposure to toxic substances or to radiation. Dry deposition and wet deposition are therefore important processes accounted for by the model. Dry deposition will reduce concentrations in the mixed layer. Material in the reservoir layer is decoupled from the surface and will therefore not be affected by dry deposition.

Dry deposition of material from the atmosphere is conceptually thought to take place through three resistances in series. The aerodynamic resistance describes the resistance due to turbulent diffusion of material from the atmosphere to a very shallow layer near the surface. The aerodynamic resistance depends on height and stability. The stability correction functions that are used are given by Dyer (1974) and Van

© ©

~ y Z ~ . ~ res.1.-

x.1.

one puff x one puff x

reffl: | [res.l

/ ' ~ " mix . l .~ - - -~ / two puffs x two puffs x

Fig. 3. Puff splitting depicted in four stages. Stage h one puff with mass in both layers; stage 2: a shallow nighttime boundary layer is developed; stage 3: the puff is split in two separate puffs, one in the mixed layer and one in the reservoir layer; stage 4: a daytime mixed layer

is developed; both puffs have mass in this layer.

Ulden and Holtslag (1985) for unstable situations (L<0); for stable situations (L>0) a formulation proposed by Holtslag and De Bruin (1988) is used.

The sublayer resistance is defined as the resistance to transport from z = Zo to z0c, where Zo is the roughness length for momentum and z0¢ is the roughness length for transport of gases. It is parameterized to conform to Wesely and Hicks (1977).

The surface resistance describes the ability of the surface to take up the airborne material and depends therefore on the chemical and/or (bio) physical properties of both receiving surface and depositing material. For most substances that may be released into the atmosphere the surface resistance can be estimated only crudely. In the model, r s is a constant represent- ing the whole model area and the dispersed material. The surface resistances needed for dispersion simulations following the Chernobyl disaster, were estimated to be ~ 500 s m- 1 for both 1-131 and Cs-137 (De Leeuw et al., 1986).

2.9. Wet deposition

Wet deposition may bring large amounts of material down to the ground surface. The model can give valuable information to estimate human exposure. However, precipitation is on many occasions a subgrid process, therefore the model can only roughly estimate the wet deposition. The below-cloud scavenging of polluted air by raindrops is parameterized as:

dc/dt = - cA. (9)

The scavenging coefficient A depends on the spectrum of raindrop sizes, the solubility of the gas, the rain intensity and chemical properties of the pollutant and raindrop. In the puff model we use the simple parameterization for the scavenging coefficient:

A = ( I/d, (10)

where I is the rain intensity and d the mixing height or the thickness of the reservoir layer. The scavenging ratio ( is defined as the equilibrium ratio of concentration in precipitation and air.

It is assumed that scavenging takes place more efficiently in the reservoir layer than in the mixed layer, due to the difference of the drop spectra. The washout coefficient ((/d) is a uniform input parameter for each of the layers of the model, that is set according to the properties of the dispersed pollutant (e.g. 0.07 s-~(m/h) -1 in the reservoir layer for 1-131 and Cs-137; De Leeuw et al., 1986).

In-cloud scavenging (or rain-out) is accounted for in a highly parameterized way by assuming a higher washout ratio in the reservoir layer.

2.10. Chemistry and radioactive decay

In the model radioactive decay and chemical re- actions are sink terms, described by linear decay:

dc/dt= -k~c, (11)

where k~ is the radioactive decay rate or chemical


removal rate. To estimate air concentrations at one 2 se÷16. position, contributions of a number of puffs are added. Chemical transformation, radioactive decay, dry and

2.0e+16 wet deposition are processes modelled in such a way that they affect each separate puff. Consequently these processes can only be described by a linear expression 1 Se*t6. like Equation (11). Chemical transformation or radioactive decay of material already deposited on the ground is not modelled. 1 0e+16.

3. A P P L I C A T I O N T O T H E C H E R N O B Y L A C C I D E N T

The operational puff model has been applied to the Chernobyl accident. Emissions from the reactor started on 26 April at approximately 1 UTC. The dispersion of 1-131 and Cs-137 was simulated for nearly 9 days. For most parts of Europe radioactivity measurements during this period were available for comparison with model values. A detailed description of the meteorological situation during the Chernobyl episode is given by Smith and Clark (1989).

3.1. Specification of the source term, parameter settings and meteorological data used

In this case study the source term and parameter settings used are the same as defined in the model intercomparison study ATMES. Figures 4A and 4B give the daily emission of Cs-137 and 1-131 aerosols, respectively. The effective emission height is also indicated. The half-life of 1-131 due to radioactive decay is set at 8.05 days. Radioactive decay of Cs-137 is neglected. Washout coefficients are chosen: 0.0111 s- l(m/h)- 1 for the mixed layer and 0.070 s- 1 (m/h)-1 for the reservoir layer. In the calculation of the deposition a surface resistance of 500 s m I is assumed. For this case study, calculations are made using precipitation and wind data from the limited area model at KNMI. The wind fields are analysed observations at the 1000 and. 850 hPa levels. The heights to which the wind vectors are assigned are 200 and 1500 m. Every 3 h a new set of wind data is used. Linear interpolation in time is performed for the intermediate hours.

The accumulated precipitation is a 3 h forecast, since the meteorological model cannot provide analysed precipitation data. The precipitation amounts are divided equally over 3 h.

The mixing heights for this study are obtained from schemes developed by various authors. For stable situations (L > 0) we used the diagnostic formulation from Nieuwstadt (1981). For unstable situations 6h/Ot is estimated using a scheme developed by Van Dop (1982), that is based on Tennekes (1973) and Drie- donks (1981). In general these schemes produce realistic mixing heights with maximum values of ~ 2500 m and minimum values of ~ 50 m (Verver and Scheele, 1989).

[3. 5e+16-

0 ( A )

20e+17'

H=1500m

26 27 28 29 30 I 2 3 4 5

<---April May--->

1.5e+17-

1.0e+17.

H=1500m

H=600m

H=300m

0.5e+17.

[3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( B ) z6 27 28 29 ~o , 2 3 4 5 6

<---April May--->

Fig. 4. Emission data used in the model intercomparison study ATMES. Depicted are the emission height H and the emitted amount of Cs-137 (A) and 1-131 (B) for each day.

3.2. Results and discussion

Nine days of the dispersion of Cs-137 were simulated (the total simulation time was 222 h). Figures 5A to 5D show the distribution of air concentrations of Cs-137 near the surface (z=4 m. Figure 5 shows the total accumulated wet and dry deposition of Cs-137, up to 5 May, 0600 UTC. On the edges of the figure the grid spacings of the model are indicated.

During the first few days material is transported towards Sweden. In a second stage a flow to the south- west moves material into central Europe. In Figs 5B and 5C it is shown how material covers large parts of Europe during the next days. Figure 6 reveals high deposition areas in the south of Germany and a part of northern Italy. On 4 May, material moved further to the northwest (Fig. 5D).

Comparison of model values with observations. Table 1 gives the total accumulated dry and wet deposition of Cs-137 up to 5 May 1986, 0600 UTC for most European countries. For comparison estimates of the total deposition based on observation in some

3186 / G.H.L. VERVER and F. A. A. M. DE LEEUW

. . . . . . . . . . . . . .

0.2 ~tm -~ :5 ~,t ~'~ 20~tm -~ 200~1m -~

I ?

0.2 ~lm "a 5 ~ m -3 2ONto -a 200Bqm -3

,cL .............. : ......... ....

. ~ . . . . ~ - ~ . ~ . . . . ~ ; : ' . . . . . . . . . . . . . . . . t 0.2_Bqm_ -3 .5.p5.~-. 3 .2.P~.m; 3 2 0 0 B q m "3

~7_.-', ~::k,~Z.CL..._/'-", I ....

. . . . . ~ . . . . ~ . . . . . . . . . . . . . . . . . . 0_:2 Bqm_-3..5..Bq.m_? .2.0..~.~. 3 200Bqm -3

Fig. 5. (A) Concentration distribution of Cs-137 over Europe for 28 April 1986, 1200 UTC. KNMI/RIVM Puff model. (B) Concentration distribution of Cs-137 over Europe for 30 April 1986, 1200 UTC. KNMI/RIVM Puff model (C) Concentration distribution of Cs-137 over Europe for 2 May 1986, 1200 UTC. KNMI/RIVM Puff model.

(D) Concentration distribution of Cs-137 over Europe for 2 May 1986, 1200 UTC. KNMI/RIVM Puff model.

European countries are also given. There is a good agreement between the observations and the model values (correlation coefficient ~0.8). However, a sub- stantial amount of Cs-137 that contributed to the observed deposition was deposited after 5 May, the day our simulations ended. Therefore, on average the total accumulated deposition values are underestimated.

Figure 7 shows the locations where measurements of air concentrations of 1-131 are available. The observational data of 1-131 presented are taken from Raes et al. (1990). Figure 8 depicts model concentrations and observed concentrations versus time for eight sites. During the first days after the initial release material was transported directly to Scandinavia. Its

arrival in Stockholm was modelled approximately 6 h too early. Concentrations are overestimated by a factor ~ 5. One and a half days later, concentrations began to rise in Budapest, which was also found by the model. The first peak concentration (30 April) found by the model overestimates the concentrations that were observed. In May, however, fair agreement is found. Arrival times and modelled concentrations in Vienna agreed very well with observations. The low concentrations observed in Paris were also found by the model, although underestimated by a factor of ~ 3. The relatively high model concentrations on 4 May were not confirmed by the observations. Model values for Bilthoven, The Netherlands, show a very good agreement with observations, both for arrival time


. . . . ',, ) / ( , . , : ; , . . . -~ . . . _ ~ - ~ , , ~:,

. . . . . i , " ~ m o i , - i • i o ,

r , " ~, t J l # i ~"~,. ~ - ~ -' 11"% I t '~ a " i ~ I • , / 1 1

• i II .

...... (,,i ." .,, ,.~ o

], t . ~.

. k,". :

0:..5...Bqm;2 2 59. m -2 500Bq -2 5000 m -2

Fig. 6. Accumulated total deposition of Cs-137 over Eu- rope up to 5 May, 0600 UTC. KNMI/RIVM Puff model.

L ' ~ P~ri, Varl**

Fig. 7. Locations where air concentrations of 1-131 are measured and compared with model output.

Table 1. Total accumulated dry and wet deposition of Cs-137, up to 5/5/1986, 6 GMT (for the model values), for some countries. Observations

refer to the whole period (OECD, 1987; CCRX, 1988)

No. Country grid points Model Obs.

Albania 6 0.0 Belgium 7 1.9 0.8 West Germany 63 4.6 4.0 Bulgaria 16 1.9 East Germany 29 0.6 Denmark 8 0.1 0.8 Finland 43 0.7 France 144 1.9 1.3 Greece 29 0.3 3.5 Great Britain 56 0.3 0.9 Hungary 22 4.2 Ireland 17 0.1 Italy 73 4.9 6.0 Yugoslavia 74 1.2 Netherlands 11 1.4 1.8 Norway 53 0.0 Austria 24 0.7 15 Poland 73 2.9 Portugal 23 0.0 Romania 58 3.4 Soviet Union 315 4.9 Spain 131 0.0 0.0 Czechoslovakia 39 6.2 4.0 Iceland 11 0.0 Sweden 86 4.7 8.0 Switzerland 12 8.8 7.1 Northern Africa 20 0.0 North Sea 222 0.2 Atlantic Ocean north 369 0.1 Atlantic Ocean south 176 0.0 Mediterranean 418 0.6 Baltic Sea 117 8.5

and concentrat ions. In Varese, Italy, arrival t imes were s imulated very well, and concent ra t ions show fair agreement. High model concent ra t ions tha t were found for Marcoule , France, on 3 and 4 May were not

confirmed by observations. Fo r Harwell, U.K., arrival t imes are, again, modelled very well, concent ra t ions were underes t imated by a factor of 4.

A scatter plot of model concent ra t ions versus ob-


Conc.cntration I- 131 (Bq/m3)

8O

6O

40

2O J I

*

0 0 24 48

S t o c k h o l m ( S )

. i ii

it

s I

i i !

e i | e t i

' I ; I t I e s w

],.i , '~

. . . . . . . . Model

Observations

72 96 120 144 168 192 216 240

Concentration I- 131 (Bq/m3)

3O

20

10

0 0

time (hours)

B u d a p e s t ( H )

t

s s s

. . . . . . . . Model

Observations

I

I |

I |

I s

I s

" " - I - - - | " - - i - - -

24 48 72 96 120 144 168 192 216 240

time (hours)

Fig. 8. (Continued).

served concentrations of 1-131 is depicted (Fig. 9). accuracy of the concentration estimates is lower, but There is an average underprediction of 0.5 Bqm -3, fair agreement is found for most sites. There are the correlation coefficient is 0.7. The ability to predict several uncertain factors influencing the observed arrival time is a very useful feature of the model. The concentration data, such as the source characteristics,


C o n c e n t r a t i o n s I- 131 (Bq /m3)

100

90

80

70

60

50

40

30

20

10

0 • • i , , , i , . , i , la l

0 24 48 72

V i e n n a

• | • •

96 t ime

| | I I I I # | # !

. . . . . . . Model

Obse rva t ions

I

L I

• , . . . , . ~ . , . . " 7 . , . - ,

120 144 168 192 216 240 (hours)

Concentrations 1-131 (Bq/m3)

2.0

1.5

1.0

0.5

o.0

Par i s

" . . . . . . . Model

Observations

0 24 48 72 96 120 144 168 192 216 240

time (hours)


the measurement accuracy and subgrid phenomena (orography, coastlines, precipitation, clouds). Large concentration gradients in the plume, occurring during the first few days, made model results very sensitive to the advection speed and direction.

4. S U M M A R Y A N D C O N C L U S I O N S

In this paper we have described an operational puff dispersion model that will be used in real-time at KNMI and RIVM in the case of accidental release of


Concentrations I- 131 (Bq/m3)

10

B i l t h o v e n (NL)

. . . . . . . . Model

Observations

4 '

2

0

1:, I : o"

! ~" ~ :

e e

24 48 72 96 120 1 ~ 168 192 216 240

time (hours)

Concentration I-131 (Bq/m3)

60

50

40

30

20

I0

Varese (I)

. . . . . . . . Model Observations

l i l |

. . . , . . . , . . . , . . . , . . . .

24 48 72 96 120 144 168 192 216 240

time (hours)


possibly hazardous material into the atmosphere. The model has a simple two-layer structure: a turbulent mixed layer that evolves in time, and a reservoir (or "residual") layer on top of that. When a puff becomes

much larger than the grid spacings of the meteorological fields, it is split into a number of new puffs. This way dispersion by the resolved scales of turbulence is accounted for explicitly. The model itself is independ-


Concentrations I- 131 (Bq/m3)

25

20,

15,

10,

0 0

Marcoule (F)

. . . . . . . Model Observations

e e a e

I *

I wm

I ee !

I

I

• , .

time (hours)

Concentration I- 131 (Bq/m3)

5 - r

Harwell (UK)

. . . . . . . Model

Observations

0 0 24 48 72 96 120 144 168 192 216 240

time (hours)

Fig. 8. Model concentration (averaged over 6 h) and observed concentrations of 1-131 aerosol for eight European sites. The time axis indicates hours after 26 April 1986, 000(3

GMT.


Observations 10 3

10 2

101

10 °

10 -1

1 0 2

10 .4 I 10 -4

t i l l m

10 3 10-2 i0-~ l0 ~, 101 10 2 10 3

Model

Fig. 9. Modelled concentrations of 1-131 vs observations (Bq m -3) for eight sites. (The minimum value is

put to 10 4Bq m-3.)

ent of the source of meteorological input data: analyses and prognostic fields may be used.

The Chernobyl accident was used as a test case for the model. It was found that the total accumulated deposition of Cs-137 is simulated well. Reasonably good estimates are given for 1-131 concentrations at eight European sites (correlation coefficient 0.7, average bias - 0 . 5 Bq m - 3). The arrival time of the cloud is

simulated very well. The model participates in the atmospheric trans-

port model evaluation study ATMES, where a thor- ough evaluation is made of the model performance in the Chernobyl test case.

REFERENCES

ApSimon H. M., Goddard A. J. H. and Wrigley J. (1985) Long-range atmospheric dispersion of radioisotopes--I The MESOS model. Atmospheric Environment 19, 99-I 11,

CCRX (1988) Rapportage aanvullend meetprogramma Tsjernobyl CCRX-rapport (in Dutch). Ministerie van Volkshuisvesting, Ruimtelijke Ordening en Milieu- beheer's Gravenhage.

Csanady G. T. (1973) Turbulent Diffusion in the Environment. D. Reidel, Dordrecht.

Driedonks A. G. M. (1981) Dynamics of the well-mixed atmospheric boundary-layer. Thesis, Royal Netherlands Meteorological Institute, The Netherlands.

Dop H. Van, De Haan B. J. and Engeldal C. (1982) The KNMI mesoscale air pollution model. Scientific report W.R. 82-2, KNMI, De Bilt, The Netherlands.

Dyer A. J. (1974) A review of flux-profile relationships Boundary-Layer Met. 7, 363-372.

Van Egmond N. D. and Kesseboom H. (1983) Mesoscale air pollution dispersion models II; Lagrangian PUFF-model, and comparison with eulerian GRID model. Atmospheric Environment 17, 265-274.

Van Egmond N. D. and E. Wirth (1986) WMO working group on assessment of radiation dose commitment in Europe due to the Chernobyl accident, Bilthoven, The Netherlands, 25-29 June 1986. WMO regional office for Europe, Copenhagen.

Van Egmond N. D., Onderdelinden D. and Kesseboom H. (1984) Correlation spectrometry as a tool for mesoscale air pollution modelling. In Proc. 13th Int. Tech. Meetin# Air Pollution Modelling and its Application, lies de Embiez, France, 14-17 September 1982. Plenum Press, New York.

Gifford F. A. (1977) Tropospheric relative diffusion observations. J. appl. Met. 16, 311-313.

Gifford F. A. (1985) Atmospheric diffusion in the range 20 to 2000 kilometers. In Proc. Air Pollution Modellin# and its Application V, St. Louis, MO, 15-19 April 1985. Plenum Press, London.

Golder D. (1972) Relations among stability parameters in the surface layer. Boundary-Layer Met. 3, 47-58.

Hanna S. R. (1986) Lateral dispersion from tall stacks. J. Clim. appl. Met. 25, 1426-1433.

Holtslag A. A. M. and De Bruin H. A. R. (1988) Applied modelling of the nighttime surface energy balance over land. J. appl. Met. 27, 689-704.

Holtslag A. A. M. and Van Ulden A. P. (1983) A simple scheme for daytime estimates of the surface fluxes from routine weather data. J. Clim. appl. Met. 22, 517-529.

Klug W. (1991) Overview of ATMES: an intercomparison study of results from log-range transport models after Chernobyl. Paper presented at the 19th ITM on Air Pollution Modelling and its Applications, Ierapetra, Greece, 29 September-4 October 1991.

De Leeuw F. A. A. M., Van Aalst R. M., Van Rheineck Leyssius H. J., Kesseboom H., Van Egmond N. D., Scheele M. P., Van Dop H. and Van Ulden A. P. (1986) The Chernobyl accident: model calculations of concentration and deposition of radioactive releases. Paper submitted to the expert meeting on the Assessment of the Radioactive Dose Commitment in Europe due to the Chernobyl Acci- dent, RIVM, Bilthoven, The Netherlands, 25-27 June.

McNider R. T. (1988) Influence of diurnal and inertial boundary-layer oscillations on long-range dispersion. At- mospheric Environment 22, 2445-2462.

Nieuwstadt F. T. M., (1981) The steady-state and resistance laws of the nocturnal boundary-layer: theory compared with observations. Boundary-Layer Met. 20, 3-17.

OECD (1987) The radiological impact of the Chernobyl accident. OECD countries OECD, Paris.

Pasquill F. (1974) Atmospheric Diffusion, 2nd edn. Wiley, New York.

Pasquill F. and Smith F. B. (1983) Atmospheric Diffusion, 3rd edn. Ellis Horwood, Chichester, U.K.

Raes F., Graziani G., Grossi L., Marciano L., Peirs D., Pedersen B., Stanners D. and Zarimpas N. (eds) (1990) Radioactivity measurements in Europe after the Cherno- byl accident, part 1: air. Commission of the European Communities. Joint Research Centre. Ispra, Italy.

Saffman P. G. (1962) The effect of wind shear on horizontal spread from an instantaneous ground source. Q. JI R. Met. Soc. 88, 382-393.

Smith F. B. (1965) The role of wind shear in horizontal diffusion of ambient particles. Q. Jl R. Met. Soc. 91, 318 329.

Smith F. B. and Clark M. J. (1989) The transport and deposition of airborne debris from the Chernobyl nuclear power plant accident with special emphasis on the conse- quences to the United Kingdom. Scientific paper 42, Meteorological Office, Bracknell, U.K.

Tennekes H. (1973) A model for the dynamics of the inversion above a convective boundary layer. J. atmos. Sci. 30, 558-567.

Turner D. B. (1969) Workbook of atmospheric dispersion estimates. Office of Air Programs, publication no. 999-AP- 26, Public Health Service, Cincinnati, OH.

Van Ulden A. P. and Holtslag A. A. M. (1985) Estimation of atmospheric boundary layer parameters for diffusion applications. J. Clim. appl. Met. 24, 1196-1207.

Venkatram A. (1988) An interpretation of Taylor's statistical


analyses of particle dispersion. Atmospheric Environment 22, 865-868.

Verver G. H. L. and Scheele M. P. (1989) Influence of non- uniform mixing heights on dispersion simulations following the Chernobyl accident. In Proc. 17th Int. Meeting Air Pollution Modelling and its Application, Cambridge, U.K., 19-22 September 1988. Plenum Press, New York.

Verver G. H. L., De Leeuw F. A. A. M. and Van Rheineck- Leyssius H. J. (1990) Description of the RIVM/KNMI puff dispersion model. KNMI publication 177/RIVM rapport 222501001, KNMI, De Bilt, The Netherlands.

Wesely M. L. and Hicks B. B. (1977) Some factors that affect the deposition rates of sulphur dioxide and similar gases on vegetation. J. Air Pollut. Control Ass. 27, 1110 1116.

Documents

An operational puff dispersion model