SCOPE 50 - Radioecology after Chernobyl, Chapter3, Atmospheric Pathways

The atmosphere can provide rapid dispersal and deposition of radionuclides out to very long distances from a release. Some of the processes involved in such transport and deposition are described in succeeding sections. Measurement and numerical modelling methods are outlined subsequently.

3.2 ATMOSPHERIC PROCESSES

Radionuclides released into the atmosphere are subjected to a variety of physical processes that determine their eventual fate (Figure 3.1). Some of the most important processes, from a dose assessment point of view, are associated with atmospheric dispersion and subsequent removal of the radioactivity from the atmosphere. Dispersion of radionuclides injected into the planetary boundary layer (the lowest few kilometres of the atmosphere) occurs as a result of both the transport (advective) and mixing (diffusive) properties of the atmospheric circulations that occur on local to global scales. The radionuclides are initially dispersed by local scale circulations that occur over time scales of seconds to minutes and over horizontal dimensions of up to a few kilometres. Dilution and mixing is induced by turbulent eddies and wind shear. In the vertical these eddies are limited in size to the mixing layer, whereas in the horizontal they extend up to the synoptic scale of large weather systems.

The advective component (the mean transport of radionuclides) is determined from measurements of winds at the surface and aloft collected by various wind measurement systems operated as part of local, regional or global scale monitoring networks. Meteorological forecast models use such data to produce analysed windfields over a region and forecast future development. The winds observed at a particular location may be thought of as consisting of a mean wind component and a superimposed fluctuating component (commonly referred to as eddies) that can be related to the diffusive (turbulent) properties of the flows. Wind fluctuations larger than the cloud tend to transport the entire cloud (advection) downwind rather than diffuse it, while eddies that are about the same size as the cloud will cause the cloud to grow rapidly and dilute the radioactivity. Thus, atmospheric turbulence consists of a full spectrum of eddies that range from distances of thousands of kilometres down to molecular dimensions. There is a continuous transfer of kinetic energy from the largest eddies to the smallest eddies where the energy is dissipated. The turbulent intensity of the atmosphere is primarily dependent upon the complexity of the underlying terrain, the wind shears as a function of height, and the vertical temperature distribution in the atmosphere. Thus, one expects greater turbulent intensities over complex terrain than over flat terrain, and during periods when strong solar heating causes vertical temperature instabilities.

Figure 3.1 Schematic diagram of processes influencing radionuclides in the atmosphere.

Much research effort has been focused on relating the diffusion of atmospheric pollutants to measurements of various meteorological variables. In particular, boundary-layer turbulence is often related to vertical temperature gradients in the atmosphere, and to the variability of the horizontal wind directions and speeds. Since the theoretical foundations of these relationships are generally poorly understood, it is necessary to rely on empirical relationships based on data acquired from meteorological field experiments. An example of these relationships is the Pasquill

Gifford curves which describe the rate of diffusion of a pollutant released into the boundary layer. These curves may be considered to give a measure of the mechanical component of turbulence due to frictional drag near the surface and the thermal component due to the vertical heat fluxes.

The radioactivity is removed from the atmosphere by two principal mechanisms: precipitation scavenging and dry deposition. Precipitation scavenging is the removal of particulate matter and gases from the atmosphere by various types of precipitation. The process involves incorporation of the radioactivity into the rain water and the subsequent deposition of the material onto the surface of the Earth. The deposition rates, which determine the residence time of the material in the atmosphere, can markedly affect the downwind surface and airborne concentration patterns. The removal of radioactive particles and gases from the atmosphere by precipitation scavenging depends on complicated microphysical and microchemical processes that are functions of the conditions both within and outside natural cloud bearing layers. Some of these are nucleation scavenging, gas and particle diffusion to cloud and rain droplets, aerodynamic and electrostatic capture, and thermophoresis and diffusiophoresis. Removal from the atmosphere by dry deposition is important within the surface layer where the airborne radioactivity may come in contact with the surface by a variety of mechanisms that include diffusion, gravitational settling, impaction, interception, electrostatic effects, diffusiophoresis and thermophoresis. Since these processes are extremely complicated and poorly understood, it is usual to simulate them by means of a ,deposition velocity, which is defined as the deposition flux divided by the air concentration (see Section 3.2.3).

Material deposited on the surface may undergo resuspension due to wind stresses and by mechanical disturbances as a result of human activities. This resuspension process may continue to occur over a contaminated area for long periods of time. The physical processes involved in the resuspension process are also poorly understood and difficult to quantify. Resuspension studies are often based on the concept of a resuspension factor, which is the ratio of the airborne concentration to the surface concentration. These factors, however, may range over many orders of magnitude, and therefore are of limited use for predicting the inhalation exposure due to a resuspended contaminant. For numerical modelling purposes, the resuspension rate is often preferred. It is a function of various physical processes such as particles moving in saltation and surface creep. Since the rates are believed to depend on the soil type, particle size, surface wetness, surface winds, and atmospheric stability, they also vary over several orders of magnitude. Thus, our ability to predict resuspension rates is limited due to our lack of understanding of the physical processes involved and also the great variability of observed rates. The concepts above are expanded in the succeeding sections (see Sections 3.2.5 and 3.2.6).

Radioisotopes may be used as tracers of atmospheric processes and much has been learned from observations of their behaviour. The movement of radioisotopic tracers is controlled by the meteorological processes in the atmosphere; a distinction can be made between stratospheric and tropospheric processes. Radioactive tracers confined within the troposphere (e.g. activity of low-yield nuclear tests, reactor accidents, etc.) are subjected to tropospheric wind systems only. Tracers reaching the stratosphere are initially controlled by the stratospheric processes and later, after transport to the lower part of the atmosphere, by tropospheric meteorology. Thus, these tracers move with the wind systems of both the regions.

Wind systems, both in the stratosphere and troposphere, are predominantly zonal, i.e. west to east or east to west. The tropospheric winds are predominantly west to east at higher latitudes with the wind speeds increasing with height up to the level of the jet stream. Moreover, intensity of the jet stream varies with location and height. At lower latitudes the winds are often easterly. However, the zonal winds are modified by the presence of cyclones and anticyclones resulting in the wind system `meandering' and having north

south components. In addition, at lower levels, regional wind systems are observed at certain locations, e.g. the monsoons of the Indian Ocean region, were due to the uneven heating of land and sea; we observe wind systems changing directions with the seasons. Meridional transport can also take place by circulations of the Hadley cell type in the troposphere at low latitudes. In this circulation pattern, winds are raised near the equator, move towards the poles and descend to the Earth's surface in the sub-tropical regions. Movement towards the equator compensates for the poleward shift at higher altitudes. Radioactive debris from nuclear tests of France and China have been used to trace the Hadley cell circulations.

Radioactive tracers are subject to the regional wind systems when they reach the lower levels of the atmosphere. They can be studied to give information on the meteorological phenomena described above. Some of the applications in which radioactive tracers have been employed are:

Residence times of stratospheric particulates have been estimated using measurements of radioactive tracers carried high into the atmosphere by megaton yield nuclear tests. The residence time is calculated assuming exponential removal of the material by meteorological processes. Since removal at lower levels of the atmosphere is much more rapid than in the stratosphere, the decrease in stratospheric levels is estimated by collecting samples of long-lived tracers ¹³⁷Cs and ⁹⁰Sr at ground level and following their year-to-year decrease in activity at similar periods of the annual cycle (to even out seasonal variations in stratospheric

tropospheric exchange). Confirmation is obtained by high-altitude sampling.

Radioisotope tracers from low-yield tests are excellent tracers for wind systems in the troposphere. A study of the detection dates of radioactive debris indicated that for the middle and high latitude tests, the movement has been predominantly by the westerly winds, possibly under the influence of jet streams. For example, Chinese test debris reached Japan in 1

3 days, the USA in 3

6 days and Central Europe in 7

8 days; the global circulation time was 10

14 days. The fast zonal movement due to the jet streams results in the activity distribution being a maximum in the latitude of testing and decreasing at higher and lower latitude; the geographical distance is not relevant.

Regional air mass circulations, like the monsoon wind systems of the Indian subcontinent, have been studied using isotopic tracers. The source of the Indian summer monsoon has been traced using debris from the low-yield French tests in Polynesia in the southern hemisphere. Many of these tests were in the summer period of the northern hemisphere. A study of the travel times of debris from the French test series of 1968 to various locations in the globe showed detection early on the west coast of India (in about 16

22 days). This is similar to the travel time of the Chinese test debris in spite of its being in the same hemisphere. A comparison of the levels of activity also showed that the west coast region of India receives more activity, up to an order of magnitude higher compared to other stations of the northern hemisphere. This study confirmed that a significant southern hemispheric component exists in the Arabian Sea branch of the monsoon. This type of study would have been very difficult by conventional meteorological means as few stations operate over the oceans of the equatorial regions to collect the required data continuously.

The levels of radon activity can also be used as an indicator of air mass mixing. This is based on the fact that air masses which travel a long distance over the oceans have low radon activity as the emanations over the oceans are much less as compared to that over the continents. Hence an abrupt increase in radon over oceanic areas indicates the presence of continental air.

The tropospheric aerosol residence time, i.e. the mean time spent by aerosols in the atmosphere before deposition on the surface, can also be measured using the ratios of radon daughters. This is based on the principle that in a parent

daughter series of radioisotopes, the growth of the daughter is affected by the non-radioactive removal rate constant. The required condition is that the radioactive half-life should be of comparable magnitude to or longer than the `removal rate' for the relative concentrations of parent and daughter to be significantly affected. ²¹⁰Bi/²¹⁰Pb ratio studies can provide this information, there being no source of ²¹⁰Bi in the troposphere other than the parent ²¹⁰Pb. Residence times of about a week for the lower troposphere have been obtained, the value increasing slightly at higher levels.

The values of tropospheric residence times have also been estimated from the decrease of tropospheric fission product debris but this may be in error as it involves meridional diffusion and other uncertainties.

The physical and chemical form of a radionuclide affects its behaviour in the atmosphere and may itself be subject to transformation during atmospheric transport. As indicated in Chapter 1, the radionuclides may be emitted in various forms, including:

Thus radionuclides may enter the atmosphere in a wide range of physico-chemical forms and may be modified during transfer, particularly by interaction with the ambient aerosol. The range of travel, and depositional properties, are crucially dependent upon the physico-chemical form. These matters may be illuminated by considering the typical size distribution of the atmospheric aerosol

an idealized aerosol size distribution appears in Figure 3.2; it comprises three modes, as follows:

Figure 3.2 Schematic diagram of the size distribution (expressed as mass per increment in log particle diameter) and formation mechanisms for atmospheric aerosols (adapted from Harrison, 1990).

Aerosol coagulation processes affect particles of all sizes, although they are most rapid for small particles at high number densities. A plume of radioactive particles mixing with ambient air will lead to coagulation of radionuclides with the ambient aerosol. This will both alter the size distribution of the aerosol and affect its physicochemical behaviour. For instance, when a hydrophobic nuclei mode radioactive particle coagulates with a larger hygroscopic particle in the ambient aerosol, its properties will be modified by incorporation into the accumulation mode, and by the presence of hygroscopic material which will influence its ability to act as a condensation nucleus.

Measurements of the size distribution of airborne radionuclides reflect the physical processes described above. Aerosol generated in nuclei mode is rapidly lost by deposition or transformation to larger sizes. It is thus normally observed only close to a source. Giant mode particles are also usually rather locally generated (including by resuspension), although atmospheric turbulence processes can keep them airborne for appreciable distances (Pye, 1987); tens, hundreds or exceptionally thousands of kilometres. Knowledge of the occurrence of very large particles (i.e. >8 µm diameter) is restricted due the difficulties of collecting them from the atmosphere with conventional air samplers. Their range of travel is likely to be very short, although there is evidence from Chernobyl of very large `hot' particles depositing tens or hundreds of kilometres from source.

Thus in the Chernobyl emission, whilst approximately 27 per cent of the volatile elements Cs and I was deposited within 80 km, 60 per cent of the more refractory Zr and Ce, expected to be associated with large particles, was deposited within this range (USSR State Committee, 1986). The major part of the activity reaching Western Europe was associated with particles of <2 µm (Jost et al., 1986; Winkelmann et al., 1987); larger particles containing a higher proportion of refractory elements reached Scandinavia (Devell et al., 1986; Persson et al., 1987).

Accumulation mode aerosol is the vector of efficient long-range transport of aerosol. It has a very low deposition velocity (see Section 3.2.3), typically 0.01

0.l cm s^-1, leading to a lifetime with respect to dry deposition of 12

120 days in a boundary layer of 1 km depth. In practice, lifetimes of 7

30 days are typical.

Aerosol particles which experience high relative humidities are liable to hygroscopic growth, a process crucial to cloud formation and precipitation scavenging. The particles which grow most rapidly by absorption of water comprise soluble salts. Very small particles (< ca. 0.1 µm), even when soluble, do not take up water except at appreciable supersaturation owing to the high energies associated with their surface curvature (the Kelvin effect). Thus even water-soluble accumulation mode particles require a small supersaturation to grow into cloud droplet sizes (ca. 10 µm) at which their dissolved salt concentration is very low and hence the suppression of vapour pressure at their surface is negligible. Supersaturations of up to ca 3 per cent occur in the updraughts associated with convective cloud formation and under these conditions many water-soluble particles act as cloud condensation nuclei. Small water-insoluble particles will not condense water under these conditions, but the presence of a water-soluble component to the particle, perhaps as a result of coagulation with the background aerosol after emission, may lead to incorporation of both the hydrophilic and hydrophobic sections of the particle being incorporated in a cloud droplet. There has been some speculation that the presence of water vapour in the Chernobyl emissions deriving from the reactor coolant circuits may have contributed to an enlargement of particles in the plume upon cooling, with consequent tendency to more rapid deposition.

Aerosol particles can sometimes acquire electrical charges which profoundly influence their behaviour. Normally, atmospheric aerosol has little associated charge (it is in a Boltzmann charge equilibrium), but there has been speculation that aerosol emitted from Chernobyl may have been charged. The presence of charge can lead to considerably altered properties, for example either enhanced or decreased coagulation rates dependent upon whether charging is bipolar or unipolar. It has been suggested that the frequency of electrical storms may have been increased due to charged radioactive aerosols arising from Chernobyl emissions.

It is convenient to distinguish `wet' deposition mechanisms, which involve the intervention of rain or some other form of precipitation, from those that result from the direct interaction of airborne material with the surface. The latter collectively constitute `dry' deposition. Thus, dry deposition is continuous, whereas wet deposition is intrinsically episodic. Deposition cleans the atmosphere but places radioactivity on the surface. Atmospheric radionuclides result in doses to humans and other organisms via external exposure and by inhalation; deposited radioactivity may cause external radiation exposure and may enter food chains over extended periods. Often, deposited radioactivity contributes a larger total dose than direct exposure from the air, and the rate of the deposition processes determines the distribution and intensity of the contamination of terrestrial and aquatic ecosystems and of the radiation exposures that result.

Studies of the transport from the atmosphere to the surface have been reviewed by Chamberlain (1991), Garland (1980), Garland et al. (1987) and Nicholson (1988b). Both gases and particles are transported to the immediate vicinity of the surface by turbulent mixing (see Section 3.2.1 above). Close to the surface is an air layer where turbulence is suppressed and transport is dominated by molecular diffusion (for gases and vapours) or a combination of Brownian diffusion, inertial effects and sedimentation for particles. At the surface specific mechanisms for removal depend on physical, chemical and physiological properties of the surface as well as the depositing material. Thus, deposition involves a sequence of steps, and the overall rate of deposition cannot exceed that of any step.

where F is the flux density of the depositing material to a unit area of land including any vegetation or other structure it supports, and C(z₁) is the concentration at a height z₁, where C may be measured or specified.

Here r_a(z₁) is the resistance for transport by turbulence to the vicinity of the surface; it can be estimated using the experience of agricultural micrometeorologists in studying the exchange of water vapour, heat and momentum at extensive crop surfaces. An additional resistance, r_b, is found to be necessary in such studies to account for observed behaviour; it can be loosely identified with transport very close to the surface. Additional resistance due to the interaction at the surface itself is represented by r_s.

r_a varies with wind speed, surface roughness and the atmospheric stability (e.g. see Garland 1980). Typical values at moderate wind speeds are 30 s) m^-1 for forest and 70 s) m^-1 for grassland, but values are often several times higher in the low winds common at night. r_b is comparable in magnitude with r_a. r_b varies as (diffusion coefficient)^-2/3, and is roughly 100 times larger for typical small particles (0.3 µm diameter) than for typical gas. Although r_s is known to be negligible for particles below about 5 µm diameter, it is the controlling resistance for most gases, and generally field measurements are necessary to establish its value.

The noble gases, krypton, xenon and radon, are not very soluble and do not react with common surfaces. There may be some loss at water surfaces, but this is unlikely to be significant, and for practical purposes r_s is infinite and the deposition velocity zero. Radioisotopes of these gases are thus lost predominantly as a result of radioactive decay.

Iodine has a number of isotopes of radiological significance of which the chief, ¹³¹I,contributed an important fraction of the dose following the Windscale and Chernobyl accidents (see also Section 3.5.2 and 3.5.3). The dose due to this short-lived isotope arises through its deposition to vegetation, and its prompt and efficient transfer via cows to dairy products. Several airborne forms have been postulated including the element (I₂ or possibly I), methyl iodide (CH₃I), hypoiodous acid (HOI), IONO₂, I₂O₂ and iodine attached to particles (Jenkin et al., 1985). Field measurements indicate a mixture with similar proportions of a readily absorbed gas (I₂, I or HI) a less absorbable gas (CH₃I or possible HOI) and particulate I on particles of 0.1 to 2 µm diameter. Studies of deposition show that I₂ is absorbed via stomata and on the leaf cuticle; in field conditions r_s is small and for grass v_g ~1 cm s^-1. Forest might have a v_g several times higher, but this would be of little practical importance. CH₃I has a deposition velocity about four orders of magnitude smaller, and particulate iodine about two orders smaller than I₂. Thus, the deposition rate is largely determined by the fraction of the reactive form. In practice, the net deposition velocity for the mixture appears to be about 0.3 cm s^-1.

The long-lived ¹⁴C is released from nuclear fuel during reprocessing as ¹⁴CO₂, and has also been enhanced as a result of nuclear weapon tests. It is incorporated into vegetation during photosynthesis and permeates all living organisms and food chains. In the long term, all biotic carbon is formed with practically the same isotopic composition as the atmospheric CO₂. Following a short release, a deposition velocity might be useful in estimating contamination, but v_g would depend on the rate of photosynthesis, and might be of order 0.05 cm s^-1 during daylight, but zero at night.

The deposition of airborne particles is strongly influenced by particle size (see Nicholson, 1988b, for a review of available information). Table 3.1 shows the Brownian diffusion coefficient and sedimentation velocity for particles spanning the range of interest. In normal wind speeds turbulent velocities of order 0.3 m s^-1 are common in the boundary layer: particles bigger than 100 µm diameter clearly could not be supported by these turbulent motions, and would fall out more or less along trajectories determined by their sedimentation velocities and the mean wind. (In stronger winds, or in the organized motions in convective storms, a higher threshold would be appropriate). Particles of around 30 µm and smaller would be readily mixed throughout the boundary layer, and particles of intermediate size would be transported to varying degrees by the turbulent flow, while subject to rapid deposition by sedimentation and inertial impaction.

Particles smaller than about 30 µm would be conveyed to the vicinity of the surface by turbulent motions assisted by gravity according to particle size. Within the air layers close to surface elements where turbulence is damped, mechanisms which cause particles to deviate from the flow become important. Brownian motion becomes significant for particles of ~0.1 µm and smaller (Table 3.1), and leads to an increase of deposition velocity for smaller sizes (Figure 3.3). Particles of 2 or 3 µm and larger have stopping distances comparable in magnitude with the smaller surface elements in vegetation canopies (Table 3.1). In consequence, they have sufficient inertia to carry them to grass blades and similar obstacles. Inertial impaction explains the rise in deposition velocity up to about 10 µm. The further rise for larger particles is chiefly due to the increase in sedimentation. Particles larger than ~5 µm in diameter may bounce at surfaces, if both particle and surface are dry. This may sometimes offset the contribution of impaction to deposition. However, typical deposition velocities of 10 µm particles to grass are above 2 cm s^-1, and deposition to forest may be several times higher.


Particle diameter	Brownian diffusion	Sedimentation	Stopping
	coefficient	velocity	distance^b
(µm)	(mm²S^-1)	(mm s^-1)	(mm)

0.01	0.052
0.03	5.6 x 10^-3
0.1	6.8x10^-4	8.7x10^-4	2.6x10^-5
0.3	1.3 x 10^-4	4.2 x 10^-3	1.3 x 10^-4
1	1.3x10^-5	0.035	1.1x10^-3
3	8.0 x 10^-6	0.28	8.4 x 10^-3
10		3.0	0.093
30		27	0.84
100		300	9.3

^a For particles of density 1000 kg m^-3.
^bThe distance travelled by a particle projected into still air with an initial velocity of 0.3 m s^-1).

Figure 3.3 Experimental data from several sources, on deposition of small particles to grass. The symbols and hatched area represent the distribution of field and wind-tunnel data. (Copyright AEA.)

None of these mechanisms is effective in the size range 0.1 to 2.0 µm. Here deposition is due to the interception of particles following the flow of air past surface elements. Particles which touch the surface adhere and are deposited. The microscopic roughnesses constituted by leaf hairs and other surface structures may increase the surface area of vegetation and may contribute to this effect, but the deposition rate in this region is expected to be of order 0.03 cm s^-1. Some field measurements indicate much larger deposition velocities, but there are a number of possible sources of difficulty in the interpretation of these data, and the general behaviour of submicron particles in the atmosphere is consistent with the value indicated (Nicholson, 1988b). Measured deposition velocities of several nuclides determined in Sweden after Chernobyl appear in Figure 3.4. The more volatile elements, Cs and I, are in the accumulation mode and show the lowest values.

Water covers large areas of the Earth's surface, and deposition to water is important as a cause of contamination of aquatic systems as well as decontamination of the atmosphere. Water surfaces are smooth aerodynamically, even when waves are generated by strong winds. Generally, deposition velocities for particles of 0.1 µm. to 10 µm diameter would be predicted to be smaller than those for vegetated surfaces, and would not much exceed the sedimentation velocity through much of the range. However, most atmospheric particles are hygroscopic, and they are expected to grow by accretion of water vapour, in the shallow layer of air close to the sea surface where the air is near saturated with vapour. If the particles grow quickly enough to reach their equilibrium size at saturation within this layer, the deposition velocity for most particles of practical interest should increase to ~ 1 cm s^-1 (Slinn and Slinn, 1980).

Figure 3.4 Dry deposition on various grass surfaces at Studsvik for 28

30 May. Iodine is corrected for the presence of 70 per cent gaseous iodine (from Devell 1991; reproduced by permission of the Swedish Radiation Protection Institute).

It was recognized well before Chernobyl, that in the event of a major nuclear accident, precipitation processes could act as a concentration mechanism leading to areas of enhanced deposition (e.g. ApSimon and Simms, 1986). Not surprisingly therefore the major problems after Chernobyl coincided with areas of high rainfall, especially over hilly and mountainous regions of Europe. Many countries found a correlation between deposition, D_dep^,and amounts of precipitation scavenging the radioactive cloud:

where W is the dimensionless 'wash-out ratio', X the concentration in air near ground level and J the precipitation rate. Observations of bomb fall-out and radioactive aerosols from Chernobyl indicate that typical values of W for small aerosols (~1 pm diameter) commonly lie in the range between 100 and 1000 Bq per litre per Bq m^-3 mean concentration in the air. Alternatively a 'wash-out model' is envisaged, whereby the rain falls through the radioactive material from above, uniformly removing radioactive material from a vertical column of air with an efficiency dependent on the rainfall rate at the ground, and depositing it instantaneously at the surface. This leads to the definition of a wash-out coefficient expressed as an integral over the raindrop spectrum

where N(a) is the number density of raindrops of radius a, E(a) is their collection efficiency for the pollutant, and V(a) is their terminal velocity. The difference between the two approaches is that the latter depends on the vertically integrated air concentration, whereas the former depends on what is likely to be measured, namely the time-integrated air concentrations near ground level.

These simple techniques provide a straightforward approach to mapping wet deposition. For example in the UK, Smith (Clark and Smith, 1988) used wash-out ratios and observed rainfall amounts to map deposition in considerable detail and pick out hot-spots of ¹³⁷Cs/¹³⁴Cs from Chernobyl. In an emergency situation such methods are dependent on detailed rainfall data being readily available. ApSimon et al. (1989) illustrated the potential of weather radar observations of rainfall in conjunction with a wash-out model to give a very rapid assessment within a matter of hours in any future emergency. The use of such remote sensing techniques is expanding steadily, especially in Europe.

However, not surprisingly, quite large variations are observed in the scavenging efficiency of precipitation. For example near Gavle in Sweden, which was one of the areas subject to the highest levels of deposition of ¹³⁷Cs/¹³⁴Cs from Chernobyl outside the USSR (~100 kBq m^-2), the hot-spot areas not only coincided with the highest rainfall, but the efficiency of removal in the precipitation was an order of magnitude higher than in some other less contaminated areas of the country (Persson et al., 1987). Similarly in the UK, for example, the problem areas for sheep farming coincided with upland areas, again subject to much enhanced deposition.

To appreciate the efficiency of removal in precipitation systems we first have to consider the chemical and physical form of the radionuclides. In a major accident a whole range of radionuclides may be released; and their characteristics may depend on the exact nature of the accident

for example on core temperatures and volatility. Thus after Chernobyl there were two distinct components of the release; the first similar in nature to particles of fuel (with some depletion of the volatiles possibly); and the second due to volatilization and condensation of nuclides. The latter is a form of gas-to-particle conversion, and often results in rather small aerosols (<<0.1 µm), which quickly attach to ambient aerosols, particularly in the accumulation range (~0.1 up to 1 or 2 µm). A special case is ¹³¹I which is so volatile that it is still partly in the gaseous state at ambient temperatures and pressures (for example about two-thirds of the total ¹³¹I was observed as gaseous iodine, and one-third attached to ambient aerosols after Chernobyl). This also introduces complexities into the rain chemistry.

The character of the radioactive aerosols determines the probability that they will readily nucleate water droplets when cloud first forms. As air becomes supersaturated there is a competitive process between the atmospheric particles, with the largest soluble particles being the most successful at capturing the available moisture. The unnucleated dry aerosols remain in the interstitial air. This is very important because the probability of capture by a falling raindrop is generally very much higher for a nucleated cloud droplet than for a bare dry aerosol.

A simple example of this is provided by the feeder

seeder mechanism over hills and upland areas. The situation occurs when there is some general rain from overlying `seeder' cloud; air forced to rise over the hill cools and becomes saturated forming the `feeder' cloud. Rain falling from the seeder cloud picks up additional moisture from the feeder cloud, leading to heavier rainfall over and downwind of the summit region. However the order of magnitude difference in capture efficiency of cloud droplets relative to bare aerosols can lead to far more dramatic enhancement, and hot spots of deposition over a hill-top for radionuclides attached to efficient cloud condensation nuclei. The exact location of hot spots is very complex, depending on meteorological and topographical factors. The feeder

seeder mechanism has been extensively studied by Choularton and co-workers (Hill et al., 1986).

Even in the absence of a seeder cloud and precipitation, hill-cloud can lead to very much enhanced deposition. The effective deposition velocities for cloud droplets can easily be in excess of 10 cm s^-1, especially over rougher surfaces such as forests, and with high wind speeds. This is more than two orders of magnitude higher than typical dry deposition velocities for bare sub-micron aerosols over, say, grassland. Studies of acid deposition have indicated that direct deposition from cloud can provide a significant proportion of the total annual sulphur deposition in sensitive areas, even at moderate altitudes of 500 m or less above sea level; also fog droplets can be very efficiently captured by obstacles such as bushes, hedges or isolated trees (ApSimon et al., 1991b). Sea-fog or morning mist may also enhance deposition over low-lying ground.

Much of the wet deposition of ¹³⁷Cs/¹³⁴Cs across Europe from Chernobyl occurred in convective shower systems. For simplicity we shall consider a shallow shortlived shower. In such a shower material is continuously drawn into the updraught from the boundary layer of the atmosphere

see Figure 3.5. As the air rises it condenses, forming cloud droplets which coalesce, eventually forming rain droplets large enough to fall down despite the upward current of the updraughts. The whole system acts rather like a giant vacuum cleaner filtering a large volume of boundary layer air, and depositing a proportion of the moisture and pollutant material in a small area below the updraught with the precipitation. The first rain which falls has scavenged a full column of undepleted material in the updraught and is often substantially more contaminated than later precipitation, which has often reached a higher altitude and colder temperatures (with ice-phase processes) where a greater proportion of the vapour has condensed out. If incorporated in cloud droplets the radioactivity will be more efficiently deposited than interstitial aerosol, when factors such as the aerosol size will be significant. As much as half the radioactive aerosols entering the storms will generally be vented from the top of the storms. This still reduces concentrations at ground level but is an important mechanism for transferring contamination to higher levels in the free troposphere.

In a shallow shower system with rain falling through the updraught, evaporation of the rain and hence cooling of the air eventually quenches the updraught, and precipitation generally ceases within about half an hour. More intense and persistent convective systems tend to be deeper; also wind-shear leads to displacement of the falling precipitation relative to the updraughts, so that evaporative cooling occurs in air outside the column (forming a downdraught of colder air) and does not quench the storm. The complex intertwined updraught and downdraughts in a thunder-cloud lead to large local wind changes and the characteristic anvil shape of the cloud. Ice particles are formed in different ways. As air falls further below freezing, say

10 °C, direct nucleation can take place on small ice nucleating aerosols; the latter are quite different from the cloud condensation nuclei, being insoluble and often minerals or clays. These small ice nuclei can then nucleate super-cooled droplets and raindrops forming larger ice nuclei, and resulting in conversion of water to ice phases. Further growth of ice particles can occur by such processes as diffusion of vapour and riming (collection and freezing of droplets, etc.). When diffusion occurs the vapour may be supplied by evaporation of water droplets which can lead to resuspension of the bare aerosols, reducing scavenging efficiency. Riming, on the other hand, transfers the radioactivity to the large ice meteors which are likely to precipitate leading to deposition.

The vacuum cleaner effect of convective storms drawing in large volumes of air and depositing radioactivity in precipitation over a relatively narrow area was evident in many locations across Europe after Chernobyl, but was particularly acute in the vicinity of Gomel some 140 km from the accident. In this region of Byelorussia hot spots occurred at levels in excess of 1500 kBq m^-2where precipitation had fallen. ApSimon et al. (1991 a) show that the concentrated deposition from a simple summer storm, generating about 5 mm of precipitation (amounting to a deposition of 1.2 kBq m^-2 for air contaminated with 1 Bq m^-3 of ¹³⁷Cs aerosol), is sufficient to account for the observed levels, given the estimated release rates of Sedunov et al. (1990).

The full complexity of warm and cold cloud processes is also encountered in frontal systems. These are very likely to be important in an accident situation, passing over northern mid-latitudes on average every two to three days. In such situations warm air from the south is drawn northwards, over-riding colder air, and giving rise to a conveyor belt of ascending air in the warm sector ahead of the cold front. Such systems are however much more complex to model. In particular it is extremely difficult to deduce the trajectories of material during passage close to the warm and cold frontal surfaces from analysed windfields produced by forecasting models. Such windfields are often only specified every three to six hours (or even 12-hourly), during which time the frontal surfaces can move 100 km or more, and the discontinuities in winds at the frontal surfaces are poorly resolved by the spatial resolution of the grid. This emphasizes the importance of increasing spatial and temporal resolution in forecasting models to obtain accurate simulations of transport within low-pressure systems, a topic of interest in tropospheric chemistry as well as in the context of radionuclides.

Resuspension caused major problems in the vicinity of Chernobyl requiring repeated decontamination. It is the general term used to describe the process in which previously deposited material becomes entrained in the atmosphere. Strictly, the material ought to have been previously deposited via an atmospheric process, although the term is often used to describe the entrainment of material originating from accidental spills, etc. The potential effects of resuspension may be considered as two-fold: the spread of contamination to adjoining clean areas and the occurrence of an inhalation hazard. A brief summary of resuspension is given in this section. For further information, the reader is referred to more detailed reviews (Linsley, 1978; Sehmel, 1980a, 1984; Nicholson, 1988a).

In order to assess resuspension, the resuspension factor (atmospheric concentration/ground contamination per unit area) and resuspension rate (fraction resuspended per unit time) have both been used. The resuspension factor suffers the disadvantage that resuspension from upwind sources cannot be easily taken into consideration and the resuspension rate, although preferable for use in many model calculations, is not easily measured directly. In practice, resuspension factors are usually used since they are more readily determined.

Resuspension can occur due to a shearing action by the wind and by mechanical agitation of the deposited species. Such mechanical agitations might include agricultural activity, vehicle or pedestrian movements and cleaning actions. In practice, wind resuspension from surfaces such as vegetation includes the effects of a movement of individual elements (e.g. leaves) and hence, resuspension can include the effects of momentum imparted on to surface material as well as the shear of the airstream. Wind tunnel measurements have indicated resuspension rate and factors which typically increase in proportion to a given power of wind speed. This power is usually greater than or around 3 (Nicholson, 1988a). Traffic resuspension measurements have also indicated an increase in resuspension with vehicle speed, and this reflects the importance of enhanced turbulence and higher tyre shear at greater speeds (see Nicholson and Branson, 1990).

Resuspension induced by falling hydrometeors can occur (e.g. Garland, 1979) and results from the air movement created just before impaction on a surface or by a movement of the surface due to momentum being imparted on contact. Such resuspension is sometimes termed `tap and puff' and has been well known by biologists as being instrumental in the dispersion of spores. However, after a period of rainfall surfaces are likely to become wetted and surface tension effects will reduce the resuspension rate. In addition, it is possible that precipitation may result in a downward translocation of deposited material from the vegetation to the underlying soil surface or to parts of the plants from which resuspension is unlikely.

Due to the variety of surface types and conditions found in the environment, as well as the variations likely in environmental conditions, resuspension is likely to vary significantly according to time and location.

Resuspension of particles by the wind can occur because of the shear forces exerted by the airstream. Because larger particles present a greater area on which the shear forces can act, and because they may protrude beyond the viscous sub-layer associated with the underlying surface, they are more easily resuspended than small particles. Similarly, because the momentum imparted on particles due to surface movement increases greatly with particle size then the effects of an accompanying increase in adhesion force are overcome and the largest particles are most easily resuspended.

Because of gravitational effects, there is a particle size (~500 µm diameter) at which particle movement is likely to be along a surface (termed surface creep) rather than via the atmosphere. It has been noted that particles with sedimentation velocities sufficiently smaller than the vertical velocities present in the atmosphere may remain in suspension. Gillette et al., (1974) equated a threshold for resuspension in which the sedimentation velocity was 12 to 68 per cent of the friction velocity. This, in practice, means that particles greater than a few tens of micrometres are unlikely to be fully resuspended in even very windy conditions. Particles larger than this rapidly fall back to earth and are said to be saltating. It is a matter of definition whether these are considered to be resuspended. They can certainly lead to a spread of contamination but are unlikely to represent any kind of respiration problem.

It is important to note that resuspended trace species are usually associated with soil particles or other host particles. Thus, the resuspended material will not have the same size distribution as the original depositing species and, consequently, an inhalation risk cannot be evaluated from such a knowledge of the size of the original material.

Resuspension factors have been summarised by Sehmel (1980a) and were found to lie in the range 10^-10

10^-4 m^-1 when the effects of wind were considered alone. When the effects of mechanical resuspension stresses were included, this range was extended to 10^-10

10^-2 m^-1. Such a wide range of values presents problems in adequately assessing the effects of resuspension in dose predictions. The resuspension factor is very dependent upon the area contaminated since horizontal transport plays a major role in the measured value.

A time dependence of resuspension has been widely noted (e.g. Nicholson, 1988a). The exact nature of the time dependence relationship has been described as either inverse power (e.g. Garland, 1979) or as a negative exponential (e.g. Linsley, 1978). Both resuspension rates (and factors) have been found to rapidly fall off after an initial contamination event and this can be explained by a number of effects. Firstly, in the early stages after deposition the most easily resuspended particles are removed from the surface. Thus, there is an initial high rate of resuspension. This will rapidly fall but will remain above zero since there will remain the statistical chance that wind gusts or eddies will arrive, in time, with sufficient velocity to remove the less readily removed material. Secondly, after the initial resuspension there will be a translocation of deposited material to shielded parts of vegetation (e.g. leaf nodes) and to the soil. Finally, in the much longer term, mixing of the deposited material with surface soil will limit resuspension from the soil itself and resuspension from vegetation might be controlled by soil splash onto plants and subsequent resuspension by wind or mechanical stresses.

Considering the time dependence of resuspension, it is perhaps not surprising that such a wide range of resuspension factors have been reported. Consequently, the adoption of a resuspension factor (or rate) for modelling purposes will depend strongly on the time period over which effects are to be integrated.

Low-level radioactive effluent is routinely discharged, under governmental authorizations, from some industrial establishments into coastal waters. In the sea the radionuclides are distributed between dissolved and particulate-associated phases, the fractions being different for different elements.

In some coastal areas, a small proportion of the radionuclides are blown back onto land in seaspray; an effect most pronounced for the particle-reactive elements such as plutonium and americium. Whilst this phenomenon is likely to be widespread, it is most readily measurable close to the points of radionuclide discharge.

The highest doses to man via this pathway (McKay and Pattenden, 1990) are believed to arise near the British Nuclear Fuels Sellafield works on the Cumbrian coast of the Irish Sea (summarized by Peirson, 1988). However the process has also been observed close to the AEA-Technology Establishment, Dounreay, Caithness, Scotland (Pattenden et al., 1989a,b) and the reprocessing works at Cap de la Hague, Normandy, France (Martin et al., 1981).

Uptake by man is mainly through direct inhalation of aerosol or indirectly via inhalation of resuspended soil and consumption of foodstuffs. Whilst the doses received are currently small relative to those from contaminated sea food or external exposure, they arise mainly from the long-lived particle-reactive actinides of plutonium-239, plutonium-240 and americium-241. The transfer is thus sustained by long-lived radionuclides which may be retained for long periods by fine sediment on the seabed and it is therefore likely to persist long after discharges cease. Consequently it is important to be aware of the pathway and to be able to estimate its likely magnitude now, in the past and particularly in the future when its relative importance might increase.

Sea to land transfer is observed most readily through analysis of soils. Levels generally increase with increasing proximity to the sea and the isotopic ratios become increasingly consistent with those in seawater.

In Cumbria a large number of surface (0

15 cm) soil samples were collected within a 50 km radius of Sellafield between 1977 and 1979 (Eakins et al., 1981). These showed that within a coastal strip, 13 km north to 27 km south of Sellafield and stretching 10 km inland, the accumulated deposition of Pu and Am was greater than could be attributed to other processes (see Figure 3.6). This was the first clear evidence of the sea to land effect. Later atmospheric measurements (Pattenden et al., 1980) showed further that (i) an aerosol with a substantial fraction of large particles (indicated by higher coast-to-land ratio in deposition compared with air) was involved, consistent with it being seaspray and (ii) the plutonium and americium in this seaspray aerosol were enriched relative to their concentrations in seawater (by up to about two orders of magnitude). The studies around Cap de la Hague, though less extensive, suggested the operation of seaspray processes similar to that seen in Cumbria. The Cumbria measurements indicate that the ratio of soil plutonium derived from sea-to-land transfer to that from weapons fallout is about 7 near Sellafield, falling to unity in other parts of west Cumbria. The total excess Pu in soils (40

70 GBq) is ~0.01 per cent of that discharged from Sellafield into the sea (440 TBq up to 1978). Autopsy measurements indicate that excess plutonium in tissues of people living near Sellafield, but not employed there, derives from aerial discharges in the early history of the works, and not from sea-to-land transfer.

Around Dounreay, the transfer appeared to be dominated by a different mechanism. Coastal soil surveys revealed a number of anomalously high accumulations of actinides within a few hundred metres of the coast. In some of these areas, the values were comparable to those found in Cumbria, despite far lower seawater levels (Pattenden et al., 1989a,b). The suggested transfer medium here was spume, a stable foam commonly occurring along the northern Scottish coastline. Spume has a very high fine particulate loading and the actinides are very particle reactive, consequently the process of spume being blown onshore and washed into the soil could explain the high accumulations observed. Spume is not so readily transported inland as seaspray or as uniformly generated, and thus the sea-to-land effect observed in Caithness was more patchy and fell off more rapidly with distance from the coast.

Figure 3.6 Distribution of radionuclide deposition (Bq m^-2) in soil in West Cumbria.

In Cumbria, it has been observed that the airborne and deposited material associated with sea-to-land transfer contains radionuclides which can be related mainly to the particulate phase in nearby seawater. However, the actinide concentrations, as measured by the ratios to stable Na, are 1

2 orders of magnitude greater than those in seawater. Thus an enrichment mechanism is involved, due to scavenging of particulates by bubbles rising through the water column to the surface. At the surface the bubbles burst, generating droplets enriched in particulates, some of which are incorporated in the aerosol over the sea. The strong tendency of actinides to associate with particles means that the mechanism is particularly,important for these radionuclides (Walker et al., 1986). It is worth noting that the mechanism is a natural one and the actinides act as tracers for particulate material in seawater.

It has been speculated that organic coatings, arising as a result of surfactants released by plankton, may enhance scavenging through their attachment to the bubbles and particles. Such coatings, by reducing surface tension, reduce the dissolution rate of bubbles and thus increase the number of smaller bubbles of relatively high surface area reaching the surface. On inorganic particles these coatings render particle surfaces more hydrophobic and thus more attractive to rising bubbles. Recent work reported by Cloke et al. (1991) supports this hypothesis and suggests a linear relationship between aerosol particulate content and seawater surfactant concentration. Thus seasonal and localized increases in surfactant seawater concentration could be, in part, the cause of variations in aerosol actinide content.

Radiation doses to the population due to sea-to-land transfer have been estimated for Cumbria where a large proportion of the population is exposed through inhalation and food chain contamination. These estimates have been made using a largely empirical model (Howorth and Eggleton, 1988), and its validity outside Cumbria and its predictive capabilities thus are not yet established. However the radiological significance of sea-to-land transfer in Cumbria (and by implication, elsewhere) is certainly small, and on the basis of the empirical model probably ~2 per cent of the ICRP limit of 1 mSv per year for 1987.

3.3 ENVIRONMENTAL MEASUREMENTS

The Chernobyl accident clearly illustrates the need for measurements of radionuclide concentration in the air and on the ground to go hand-in-hand with modelling studies. Measurements on their own are rarely sufficient to answer all the questions that are inevitably raised, and this was particularly true following Chernobyl. Models complement measurements and measurements correct and strengthen models. All models require appropriate input data; in this case details of the locality, height and time profiles of the emissions. They also require suitable meteorological data. Principal of these are wind and precipitation fields in space and time, but other parameters may also be necessary. Actual requirements will change with the scale and character of the emission.

The remainder of this chapter contains three parts: the first describes in outline the various techniques and the uncertainties associated with radiological sampling methods. The second considers how meteorological information is collected, and in so doing implies the accuracy likely to be achieved in the interpolated fields. The third part of the chapter discusses the wide range of atmospheric dispersion and deposition models, and gives an example of how these models have been ^,validated' against the results of the Chernobyl accident.

A variety of different methods are used for monitoring the concentrations of radionuclides in air. Most commonly, a sampling apparatus is used to collect intermittent whole air samples or to filter radionuclides of interest from the air. The collected sample is taken back to the laboratory for concentration analyses. Some instruments are available which allow in situ measurements to be made, either directly after sampling or even while sampling is in progress. Generally, different techniques are used for sampling gases and particles.

For the measurement of particles, the most common sampling methods are based on filtration of large volumes of air. The large volumes required provide a major constraint upon the time resolution of measurements. For applications where filters are to be counted directly for alpha particle emissions, a membrane type filter material which allows minimal penetration is desired. However, such filters tend also to restrict flow and thus limit the volume of air which can be sampled. If the radioactive material is to be chemically removed from the filter matrix before analysis, a filter medium which is easily dissolved and has low ash content and whose ash does not interfere with the chemical separation is required.

Filters cannot generally be used for collecting gaseous radionuclides although some chemically impregnated materials are available. Gaseous species are generally sampled by either collecting whole air samples or preferably by removing the gaseous nuclides from an air stream via some type of absorption process. The most widely used absorbent is activated charcoal, which is almost a universal absorbent and effectively removes nuclides such as radon, krypton, xenon, and gaseous iodine from an air stream with high efficiency. The efficiency of removal depends on flow rate and can be increased by cooling the charcoal to dry ice or even liquid nitrogen temperatures. Cold traps are frequently used to condense tritiated water vapour from the atmosphere. Similarly, cryogenic separations are used to remove noble gases such as krypton-85 from whole air streams. ¹⁴CO₂ is often sampled by bubbling air through a caustic solution of NaOH where the carbon is chemically trapped.¹⁴C can also be sampled by drawing high volumes of air through a dry molecular sieve.

Once the nuclide of interest has been collected, a number of different methods are used to infer its concentration in air. The choice of method is usually governed by the nuclide half-life, the type of radiation emitted (penetrating power), the sampling matrix and the speed with which the information is required. In many cases particulate collected on filters can be analysed directly in the field either immediately after collection or even during collection. This is generally true when the radionuclide of interest emits fairly energetic gamma rays (e.g. ¹³⁷Cs, ¹³¹I). Most commonly, filters are removed from the pump after sampling and brought to the laboratory where they are either counted directly in a fixed geometry and/or the collected particles are chemically separated from the filter for subsequent counting. In general gamma-emitting radionuclides such as ¹³¹I or ¹³⁷Cs, and radon and thoron progeny can be measured by counting the filter directly.

For most low-level alpha and beta emitter measurements the filter is generally dissolved and the radionuclide of interest chemically separated and then electroplated onto a metal planchet or evaporated onto a thin plastic film. The resultant purified and concentrated low mass sample is then counted in a vacuum chamber using a high-quality solid-state alpha spectrometer. Further information upon radioanalytical procedures is given in Section 8.3.

All of the various analysis methods, whether based on directly counting the sampling matrix itself (filter, charcoal, etc.) or a prepared extracted low mass sample, usually rely on comparison with radioactive standards of the same geometry traceable to one of the major national or international standards laboratories (NIST, IAEA, etc.). Most reputable laboratories routinely participate in cross-calibration exercises sponsored frequently by organizations such as the IAEA.

Routine high-volume air sampling for fission products is carried out in the environs of almost all of the world's nuclear facilities. Many nations also maintain monitoring networks to detect fallout from weapons tests or potential reactor accidents (or inadvertent releases). Many programmes in place during the 1950s and 1960s for monitoring fallout from nuclear weapons tests were terminated or scaled back after atmospheric testing was halted in 1962. In the immediate intervals following the Chernobyl reactor accident, a number of nations, particularly in Europe, increased the frequency of their routine monitoring programmes or activated previously inactive stations.

Nuclear facilities also often routinely monitor the external gamma ray exposure rate (ionization) in air. This quantity is a measure of both the concentration of radionuclides in the atmosphere and the deposition on the surrounding ground. Radon, a naturally occurring noble gas, is monitored routinely at selected sites around the world as a tracer of atmospheric transport of other pollutants. A number of studies have monitored ⁸⁵Kr, which is released during reprocessing of reactor fuel rods, as an atmospheric tracer in order to test local and intermediate range atmospheric diffusion. The US government, originally through its Naval Research Laboratory and since 1963 by the Department of Energy's Environmental Measurements Laboratory, has since 1957 collected monthly air samples throughout the world at a network of over two dozen sampling sites.

3.3.1.4 Sampling methods and instrumentation for the measurement of surface deposition on soil and vegetation

The deposition of radionuclides onto the Earth's surface is also monitored at many sites throughout the world. Radionuclide deposition density or concentration in soil or vegetation is monitored in situ as well as via routine sample collection. Generally, in situ monitors are designed to collect precipitation or to intercept depositing radionuclides. Samples are also collected by directly funnelling precipitation into an ion exchange column which removes most radionuclides of interest from the liquid. The ion exchange bed is removed for analysis at the end of the sampling period. Dry deposition is much more difficult to assess; it is usually collected in buckets or on other surfaces such as gummed film (sticky paper), trays of granular material, or fibrous material which simulates grass. Most such collectors do not discriminate between deposition occurring during precipitation as opposed to dry deposition. A combination wet/dry collector is now used at many monitoring sites to collect separate samples of wet and dry deposition.

Once the radionuclides have been deposited onto the soil, vegetation or other surface it is generally difficult to monitor routinely the deposition density. Techniques have been developed, however, which allow rapid qualitative and often even quantitative estimates of the deposition of gamma-emitting radionuclides in situ (NCRP, 1976). These in situ gamma spectrometric techniques utilize large-volume germanium diodes to obtain a spectrum of the incident gamma radiation above a soil half-space. This spectrum can then be unfolded using the known detector characteristics and certain assumptions regarding the radionuclide depth and surface distribution to estimate the concentration of many radionuclides on or near the soil surface. Rapid surveys of deposition of gamma-emitting radionuclides are also frequently carried out from aircraft utilizing large arrays of gamma scintillation detectors. Although these aerial measurements generally provide only estimates of total activity or gamma exposure rate, they are useful for rapidly surveying large areas and providing guidance for ground sampling.

The major problem in most monitoring programmes is obtaining samples which are representative of the average deposition over a large area. Frequently, the deposition via atmospheric transport processes is non-uniform and large numbers of samples are required to characterize adequately the average deposition over even a small area or the average concentration in any particular matrix. Sophisticated statistical sampling protocols have been devised to minimize the number of samples needed, but generally such sampling programmes are time-consuming and expensive.

A number of global and national sampling programmes of radionuclide deposition density were carried out during the 1950s and 1960s to monitor fallout from nuclear weapons tests. Many of these programmes were terminated after nuclear testing in the atmosphere was halted in 1962. Deposition is still usually monitored routinely in the vicinity of nuclear facilities by many governments and most such facilities maintain their own regularly scheduled routine sample collection programmes designed to demonstrate regulatory compliance with emission limits. Many of these facilities are also surveyed periodically by airborne gamma ray spectrometry to detect previous undocumented releases and establish baseline levels in case of an accident.

Radioactive material (like other pollutants) injected into the atmosphere can be carried long distances before being deposited to the underlying surface. Although a typical distance-scale is 1000 km, the scale can vary considerably according to the physical nature of the material, the current state of the atmosphere and the effective height of release. For large releases it is necessary to follow the material over such long scales and further, but much smaller releases may only be of significance out to one or two kilometres, or at most a few tens of kilometres. Transport over such relatively short distances can be assessed using one of several programs run on desk-top computers, using surface-based meteorological data as input collected either from instrumentation on-site or from a standard meteorological observing station which is ideally within about 30 km distance.

The better programs require a data-base of surface elevation expressing local topography which can result in a significantly variable windfield across the area of concern, especially in rather stable conditions. In addition the meteorological data should include, as a minimum, wind speed and direction, net incoming radiation, air temperature, relative humidity, medium and low-level cloud amounts, the precipitation rate (if any), the surface roughness length, and some knowledge of the state of the ground (principally how damp it is, as may be inferred from the 'present weather code' in a standard meteorological observation). If radiosonde or other upper-air data are available, then information on boundary layer depth, buoyancy frequency in the air above the boundary layer and the temperature jump across the capping inversion should also be used.

Important though small releases may be, in the rest of this section emphasis will be placed on larger releases where material has to be followed through the atmosphere over hundreds or thousands of kilometres. Generally models which attempt to simulate the transport and deposition over large ranges rely on the output from numerical weather prediction (NWP) models, either simply using analysed fields resting heavily on observed data (but may use predicted values in areas of sparse data input), or on totally predicted fields when looking ahead to future movements of the plume. The transport models take from these NWP models as a minimum wind and temperature profiles, boundary layer depths and estimates of precipitation type and intensity in both space and time. These NWP models require large fast computers and depend on the vast input of a wide variety of observational data collected throughout the hemisphere or globe. The nature of these data and the means by which they are collected will now be briefly summarized.

These NWP models require, as input, data describing the quantitative state of the atmosphere. This demands regular, reliable and accurate measurements of pressure, temperature, wind, cloud and precipitation. No single observing method is capable of providing all the data required and a series of monitoring networks of different types, using different techniques, has been established.

Many countries have a mix of stations taking surface observations; some are `key' stations with a spacing of about 150 km used to define the broad weather pattern over the country. These are often supplemented by other stations manned by meteorologists to give more local small-scale details, but all professionally manned stations are relatively costly to maintain. Other data are collected by synoptic automatic weather stations (SAWSs) and by auxiliary observers such as lighthouse keepers, coastguards and private individuals.

Out at sea observations are made by personnel on board dedicated weather ships, and on voluntary observing ships and from gas and oil platforms and rigs. Many ships are now equipped with automatic transmission equipment, sending data via meteorological satellites more reliably and at substantially lower cost than by other means. Other ships send their data in coded form by radio and telex to reach national meteorological centres via coastguard stations. Worldwide there are some 7500 ships from 49 countries making such observations when they are at sea. In spite of the number of observing ships, the typical density of ship reports received from some areas is still woefully inadequate. Over a one-week sampling period in a recent winter only 15 surface observations were received on average for each midday observing time from merchant ships in the entire Atlantic west of the British Isles (i.e. from 50° to 60 °N, 10° to 60 °W), since the major shipping lanes lie further to the south. It would require about 200 such observations together with satellite imagery to achieve an adequate description of the broad features of weather disturbances at the surface. Even fewer ship observations are received at night due to a lack of a duty radio operator on many vessels during the `silent' hours. Whilst at first sight this may not seem too serious in the present context, forecasts of the movement of airborne radioactivity may be seriously degraded by such sparsity of observations in relatively distant upstream areas.

Upper-air measurements of geopotential, temperature, humidity and wind are made routinely from suitably equipped stations around the world, using balloon-borne radiosondes, up to heights of around 30 km at 00.00 and 12.00 GMT. These stations require reception, tracking and analysis systems and these as well as the sondes are periodically upgraded with the latest technologies. The measurements give information that are clearly short-time samples and that are averaged to some extent along the vertical track of the sonde as it ascends. The upper winds are obtained by tracking the sonde by radar or by using a Navaid radiosonde which receives and retransmits low-frequency radio signals of the type used for aircraft navigation.

An increasing number of boundary-layer radiosonde systems are being established which measure geopotential, temperature and humidity up to about 5000 m. These give much more detailed profiles in the region of particular interest to pollution transport modelling. Other systems, like doppler acoustic sounders, are also capable of inferring low-level winds and turbulence intensities. These however are still largely research tools and are thus not widely available.

In recent years an avionics system known as ASDAR (Aircraft to Satellite Data Relay) has been developed to collect meteorological data on board commercial aircraft, especially those which overfly data-sparse areas, and to transmit the data automatically via satellites to the telecommunications networks. Valuable though these are (particularly for defining the jetstream), reporting is inevitably spasmodic. Aircraft reports, which do not originate from ASDAR, are however sometimes inconsistent, and subject to communication delays.

Weather radar is a very valuable source of information not only to the NWP models but also to the transport and deposition models tracking pollution. Radars provide a high-resolution picture of almost real-time precipitation. Traditionally rainfall has been measured using funnel rain-gauges but, because of the large spatial and temporal variability of precipitation, a large number of such gauges are required for accurate estimates of the total rainfall over any particular area such as the cloud of radioactive debris released from a damaged reactor. Radar, however, measures rainfall and snowfall instantaneously over a large area, thus simplifying the collection and transmission of data. This is not to say that conventional gauges are redundant. For example, a number of rain-gauges which can telemeter their data to the radar base are required to calibrate the radar returns and thereby yield optimum estimates of surface rainfall. This is particularly important in mountainous terrain where the radar beam cannot always `see' down to ground level.

An example of the high resolution of the radar data is provided by the situation in the UK where individual radars operate on a cycle which repeats every 5 minutes and data are processed and reduced to measurements of precipitation in both 2 km x 2 km and 5 km x 5 km areas aligned with the National Grid. Although the precipitation estimates have greatest accuracy within about 75 km radius of the radar, daily totals of rainfall are achieved for each 5 km square within a 210 km radius. In addition, the composite radar images produced from the radar network every 15 minutes are also archived.

Many countries in Western Europe have come together in the last few years to exchange and harmonize their radar data to provide composite maps of rainfall. This process is bound to continue, bringing in an increasing number of countries.

Observations from polar-orbiting (typically at about 1000 km altitude) and geostationary satellites (at about 36 000 km above the equator), able to sense radiation from the atmosphere, provide opportunities for meteorologists to make a wide variety of very relevant measurements. By detecting the variable solar radiation reflected by clouds and the Earth's surface during daylight hours, a conventional image can be formed and transmitted back to Earth. Furthermore, by monitoring certain infra-red wavelengths, images of cloud can be obtained at all times of the day. Microwave radiometers offer the prospect of measurements of the rate of rainfall over the sea that are about as accurate as those now obtained from surface-based radar. Also by measuring infra-red or microwave radiation at a range of wavelengths, some of which are strongly absorbed and some of which are not, it is possible to discern the temperature of the atmosphere as a function of height. Although satellite soundings provide a denser network of observations in both time and space than is available from the conventional radiosonde network, there are significant problems to overcome if the desired accuracy is to be achieved. An Along Track Scanning Radiometer (ATSR) has been fitted to the European Space Agency Satellite ERS-1 and provides measurements of sea-surface temperature to an accuracy better than 0.5 °C.

As already seen, no single system on its own can provide a totally adequate precipitation field. Considerable research effort is therefore in progress to combine information from various sources to provide the best composite pictures. One such system operated in the UK is the FRONTIERS computer-based system that receives radar and satellite data in image format and manipulates them under forecaster control to generate improved images of the current rainfall pattern. By `improved' is meant enhanced, extended and quality-controlled, and interpreted by the forecasters' understanding of the meteorological situation.

3.4 MODELLING

Modelling of atmospheric dispersion and deposition of radionuclides is undertaken for a variety of reasons. For example detailed modelling of some of the processes described above, such as the fate of radioactive aerosols entering a convective storm, aid in understanding the role and interaction of specific microphysical and meteorological factors. More general models, incorporating simpler parameterizations of advection, turbulent mixing, radioactive decay and removal by dry and wet deposition, are directly applicable to estimating the environmental consequences of radionuclides released to the atmosphere. These estimates are particularly important in the immediate aftermath of an accidental release in alerting and guiding monitoring teams to areas most likely to have been significantly contaminated. The type of model appropriate will depend on the spatial scale of importance, differentiated below as local within 10 to 20 km of a single source, regional out to about 100 or 200 km, continental involving a range of the order of 2000 km or more, and hemispherical or global. These different scales are principally determined on the one hand by the magnitude and duration of the release, and on the other by the differing assumptions implicit to the models. For example on the local scale, and on this scale only, the meteorological parameters may be assumed stationary in time whilst any part of the release traverses the `scale', and the material if released in the turbulent mixing layer continues to spread vertically.

The type of model to be selected is highly dependent upon the specific purpose for which it is to be used. The selection process must consider a variety of factors such as: (1) the spatial and temporal scales to be modelled, (2) response time requirements, (3) source term characteristics, (4) the complexity of the terrain and meteorology over the domain of interest, (5) meteorological and radiological data availability, (6) model complexity relative to operator training, (7) desired modelling accuracy, and (8) computational requirements. Atmospheric dispersion models are used for emergency preparedness planning as well as for dose assessments during and after an accidental release of radioactivity. In many cases a simple model coupled with climatological data may be sufficient for basic emergency preparedness planning purposes. However, accident analysis, both during and after the accident, should be based on results obtained from a model that faithfully simulates the physical processes responsible for pollutant dispersion and utilizes the available meteorological data.

By integrating the radiological measurements with the model predicted concentration patterns, it is possible to estimate the source term. This, however, requires a careful optimization of the agreement between the measurements and model calculations in order to acquire realistic results. Models, in conjunction with weather forecasts, can be used to increase the effectiveness of the measurement teams by directing them to areas where the concentration patterns are expected.

The calculated patterns of time-integrated atmospheric concentration and deposition of the important nuclides provide a starting point for estimating the doses to the populations exposed; either used directly in conjunction with the appropriate dose-factors to calculate inhalation doses, or doses due to external irradiation from the cloud; or in conjunction with further models of transfer through the surface environment, and into food chains, to estimate doses from external irradiation and ingestion. This can either be for individuals exposed to more acute doses leading to direct radiation effects or requiring ameliorative action, or to population or individual doses to indicate the stochastic effects. Such analysis is extremely useful in the context of emergency situations; both in treating hypothetical releases of radioactivity for planning emergency responses and risk assessment; and in the course of a real emergency as a tool for interpreting the available information and aiding decisions on the appropriate actions and control measures. Also, if as in the case of Chernobyl, there are large uncertainties about exactly what is being released at the source, comparison of the agreement between model estimates for different assumed releases and measurements downwind can be used to estimate the source terms for the important nuclides, and their variation over time. Once the immediate emergency is over, the release brought under control, more detailed modelling reconstructions can assist in the post-accident analysis.

The level of complexity of atmospheric dispersion models varies greatly. In treating hypothetical releases for risk studies detailed accuracy in the spatial and temporal pattern is not necessary, nor is it appropriate to use a very sophisticated model requiring substantial computer time when a rapid response is required close-in to an accident; too complex a model, with a large number of variable parameters, can just be confusing in such a situation. However it is also important to recognize the simplifications introduced by a model, and the limitations induced by the indeterminant nature of the atmosphere. These are described below for the different types of model applicable over different distance scales. In essence they all represent a solution of the equation for the concentration Q:

A variety of models are available for evaluating the consequences of an atmospheric release of radioactivity. The simplest of these models is the Gaussian plume model, which requires only the wind speed and direction at the release location along with estimates of atmospheric stability and source term. At the other end of the spectrum are complex three-dimensional models that are capable of including the effects of terrain and spatially varying meteorology. This spectrum of model capability can be divided into the following three generic categories:

These models can provide real-time dose assessments for distances out to 5

10 km from the source point, depending on the roughness of the terrain and the complexity of the meteorology at the time of the accident.

Beyond 10

20 km or so topographical features and changing meteorological conditions complicate the dispersion. This requires sophisticated models for the windfields taking account of ground contours and surface characteristics. Alternatively techniques have been developed to interpolate between available wind measurements over a region, ensuring that the windfields deduced are mass consistent (Lange, 1978). It is important in such models to be able to represent such features as sea breezes, and the behaviour of flows over hills and in valleys. With these windfields, concentrations can either be estimated by integrating equation (3.5) over a three-dimensional grid of cells, or using an assembly of particles (either using statistical Monte-Carlo simulation or particle in cell).

Over long distances, possibly involving transport across international frontiers, the first question following an emission is where will material be transported to; or alternatively if radioactivity is unexpectedly detected, where has it come from? This requires trajectory modelling, either following material forwards from a source, or backwards from the point of observation, and depends on the availability of windfield data over the region of interest. These may be taken directly from weather forecasting models, or derived indirectly from pressure fields using a quasigeostrophic approximation. After Chernobyl a wide variety of trajectories were produced, reflecting a plethora of different techniques and windfield data. Some of these trajectories referred to different altitudes; some were two-dimensional horizontal trajectories at a constant level, and others three-dimensional using vertical winds directly, or isentropic trajectories adjusting altitude to maintain a constant potential temperature. The diversity of results indicates the considerable uncertainty in estimating continental scale transport, which may have been somewhat obscured in the Chernobyl situation by the prolonged nature of the release. It stresses the point that the modelling of atmospheric dispersion over such distances can be no better than the windfields used, and the importance of current improvements in forecasting models with much finer spatial grids, providing data at frequent intervals (ideally hourly). The accuracy of trajectories decreases markedly with increasing travel time, and is normally considered very unreliable beyond about 72 hours, although in the case of Chernobyl the material took 8 days to cross Europe.

Some models may not adequately reflect a very important characteristic of plume behaviour at continental scale which was very evident in the Chernobyl cloud, namely that beyond about 24 hours travel the plume becomes increasingly fragmented and contorted, often with filaments and blobs. This complex behaviour results from both synoptic-scale variations in the winds associated with mobile weather fronts, depressions and anticyclones, and the variable interaction of vertical motions and marked wind-shear with height.

The simplest long-range dispersion models merely involve assuming a plume along the estimated trajectory in a similar fashion to an extended version of the Gaussian plume model used at shorter distances, and often simplified to give uniform vertical mixing over a constant mixing layer. Alternatively for uniform continuous routine releases, or risk studies, statistical models are sometimes used which ignore bending of the trajectories and changing wind-speeds, and use merely windrose data at the source into different directional sectors. However for more sophisticated quantitative analysis of downwind concentrations and deposition, slightly more complex models are required. These can be differentiated into two types

the Lagrangian models which essentially follow the histories of component elements of the release across the region, and the Eulerian models which simulate the dispersion of material through a three-dimensional grid of cells spanning the region of interest.

Lagrangian puff models have been widely used in continental scale studies, not only of radionuclides but of acidic and other atmospheric pollutants. They use horizontal windfields, and treat pollutant emissions as a series of puffs which are advected as columns of polluted air along the calculated trajectories. They vary in complexity according to the detail in which they model vertical and horizontal dispersion and other pollutant processes according to the meteorological conditions encountered. One of the earliest studies of the European scale transport from Chernobyl was undertaken with such a model, MESOS (by ApSimon et al., 1989), which had already been applied to the Windscale accident in 1957 in the UK

see Section 3.5.2. As was apparent from the results, difficulties arise when there are marked changes in wind height, or the motion is very three-dimensional in nature as in the upward motion within the frontal system over Scandinavia. Another difficulty is the venting of material aloft in convective clouds. The advantage is that such Lagrangian models give a very direct relationship between emissions and the contamination they give rise to. They are well-suited to estimating source terms and their variation in time, as they differentiate between different parts of the release; also to estimating transfrontier fluxes. It is also easier to revise predictions in the light of early measurements and updated forecasting of windfields as the situation evolves following an accident release. Their computer requirements are still relatively modest, and hence they are still appropriate for use in risk studies where large numbers of potential accident situations need to be investigated (e.g. ApSimon et al., 1985).

The Eulerian grid models allow a full three-dimensional treatment in integrating the transport equation over a three-dimensional array of grid cells, but are far more demanding on computer time. They have been particularly useful in treating situations with complex atmospheric chemistry, but this is less critical when treating radioactive aerosols. Nevertheless Eulerian models were successfully applied to the Chernobyl situation; for example the GRID model of RIVM in The Netherlands (van Egmond and Kesseboom, 1983). The Eulerian models still cannot resolve subgrid scale processes such as cloud venting and convective storms; and also suffer from some numerical difficulties such as numerical diffusion, which is particularly important close to a point source. Nor do they overcome the intrinsic errors in the windfields and advection; and it is crucial to have mass-consistent windfields which satisfy the continuity to avoid violation of conservation of the mass of pollutant material. Their main draw-back for application in an emergency situation is their inability to distinguish contributions of different elements of the release, except by repetition of the calculations.

An approach which overcomes some of the difficulties of the Lagrangian models so far discussed whilst retaining most of their advantages involves Monte-Carlo particle simulations. The release is treated as a sequence of particles which are advected according to the evolving windfields in space and time, with random perturbations in each time-step to represent the effects of turbulent displacements with respect to the mean windfield. They are thus good for treating the complex three-dimensional nature of the windfields, but are heavy on computer resources since large numbers of particles have to be tracked. This difficulty can be reduced by the use of parallel processing techniques, and by making use of the whole trajectory of each particle through a set of surface grid-cells, instead of just the instantaneous distributions every few hours. There are no problems of numerical diffusion as in the Eulerian models, but care has to be taken to avoid artificial accumulation of particles in regions of lower turbulence by inappropriate simplification of the technique.

Quantities of the key radionuclides can be associated with each emitted particle. The airborne fractions can be reduced according to the probability of decay and of particle deposition by dry and wet processes. Wet deposition, and cloud venting to the free troposphere which can be so important with convective activity (and may well have accounted for substantial hemispherical scale transport after Chernobyl), can be treated statistically with a resolution determined by the available data on cloud distributions and precipitation. This is an area where further work on modelling is required.

Since Chernobyl there has been an increased interest in development of modelling capabilities to simulate continental scale dispersion. Accordingly a long-range model evaluation study, ATMES, was established jointly by IAEA, WMO and the CEC using observations after Chernobyl. Joint estimates of the emissions were used, and the REM (Radioactivity Environmental Monitoring) data bank established at the Joint Research Centre, Ispra (Girardi et al, 1987) provided carefully compiled and screened measurements relating to daily patterns of air concentrations of ¹³¹I and ¹³⁷Cs and their deposition across Europe. Modellers were asked to provide corresponding estimates, many of them making use of the same meteorological data set provided by ECMWF (European Centre for Medium Range Weather Forecasting), for the Chernobyl period. Various statistical analyses have been applied to compare modelled results and observations and test spatial and temporal agreement, drawing on experience with earlier model evaluation studies in North America based on perfluorocarbon tracer experiments.

Twenty-one models from 14 countries participated in the study, which varied unduly in sophistication from very simple trajectory models to three-dimensional particle models and Eulerian grid models. It was evident that different models appeared to give better agreement according to different tests, but there was no clear indication that the most sophisticated models performed appreciably better than the simplest. However, there are major diferences between the model predictions, indicating that the results are sensitive to the manner in which the atmospheric structure and processes are parameterized and represented

for example the wide variation in treatments of mixing layer depth.

Comparing the participating models was difficult in view of uncertainties in the given source terms and observations, and also because some models may already have been adjusted or tuned to this particular accident. The results are being analysed further, and will be reported in full. Proposals have been put forward to carry the model comparisons further by conducting a tracer experiment in Europe, similar to those already undertaken in North America (ANATEX, CAPTEX) to test models in a real-time application

a very difficult but relevant test for their practical use in an emergency situation. This will emulate a release of inert gases and non-depositing pollutants.

3.5 APPLICATIONS OF MODELS TO IMPORTANT ACCIDENTS

The three major accidents at Three Mile Island, Windscale and Chernobyl have provided unique opportunities to evaluate models and test their usefulness. The latter two both released radionuclides that deposited to the surface by a variety of processes, in contrast to the inert non-depositing tracers available for experimental releases. For all three accidents models were applied to simulate the atmospheric transport, and compared with the radiological observations.

Several modelling approaches were taken to estimate the radiation dose received by the affected population due to the release of noble gases (mainly ¹³³Xe) from the damaged reactor. This included several independent studies based on atmospheric dispersion modelling in conjunction with the environmental radiation measurements, as well as spatial interpolation of the radiation measurements made at numerous locations surrounding the site. The various dispersion modelling efforts used a variety of models that ranged from Gaussian to complex three-dimensional models.

Meteorological data were available from the on-site tower and from several local sources. The meteorological conditions during the first five days of the accident, when the highest release rates occurred, consisted mainly of up- and down-river wind flows. From the morning on 28 March, when the release was initiated, until mid-afternoon on 29 March, the winds were primarily from the southeast

southwest direction at approximately 3 m s^-1. Subsequently the winds rotated to a northeast

northwest direction with an average speed of 1

2 m s^-1. Calm and highly variable conditions were observed during the night of 29

30 March. These calm and variable conditions continued until the evening of 30 March when strong and steady southerly winds of about 3 m s^-1 returned. On 1 April the winds rotated into the westerly to northwesterly directions with speeds generally ranging between 1 and 3m s^-1.

Atmospheric dispersion modelling, based on these meteorological conditions, permitted elucidation of the temporal evolution of the time-integrated dose pattern. This is illustrated in Figure 3.7 which shows the evolving integrated dose pattern (Knox et al., 1981). These results were generated by means of a three-dimensional mass-consistent windfield model coupled with a particle-in-cell transport and diffusion model using a normalized 1 million curie release that varied in time according to that reported by the General Public Utilities (GPU), the plant operator (Woodard, 1979). The figure shows how the dose pattern was quickly and predominantly established by the generally south-to-north flow during the initial release period. Thereafter, the low levels of release resulted in only relatively minor but discernible changes in the initial dose pattern. Note particularly the southward extension of the pattern from 29 to 30 March, the east and southeast spread from 30 to 31 March, and finally the 'diffusion-like' effect of nine days of synoptic and diurnal meteorological variations with a small source term.

Integration of atmospheric dispersion patterns with the environmental radiation measurements permitted estimation of the total integrated dose to the exposed populations. Using the US Department of Energy (DoE) aerial radiation measurements made within the plume on a regular basis over a two week period, resulted in the DOE integrated dose pattern shown in Figure 3.8 (Hull, 1980). Note that the highest doses occurred in the areas immediately north of the plant with secondary nodes extending in the southeast and easterly directions. A similar pattern was generated by GPU after integrating the TLD measurements with their finite plume dispersion modelling (Woodard, 1979). These studies indicate that a total dose of 100 mrem was exceeded over an area extending several kilometres in a northeast to northwest direction from the plant, although the highest measured off-site dose by the TLD network was 75 mrem.

Figure 3.7 Normalized calculated integrated dose patterns in units of mSv due to TMI-2 noble gas release on (a) 29 March (24-h integration), (b) 30 March (48-h integration), (c) 31 March (72-h integration), and (d) 7 April 1979 (240-h integration). The patterns are based on the normalized release of 37 PBq of ¹³³Xe. (Reproduced by permission of the Lawrence Livermore National Laboratory. All rights reside with the US Government.)

Integration of the dose pattern with the population distribution throughout the affected region led to an assessment of the collective dose equivalents. The most credible estimates are given in Table 3.2. The estimates resulting from analysis of the DOE aerial radiation measurements was 2000 person-rem (Hull, 1980) in contrast to 3300

3400 person-rem obtained by the US Government Ad Hoc Interagency Study Group on the basis of spatial interpolation of the TLD measurements by either atmospheric dispersion modelling or by inverse distance scaling (Ad Hoc, 1979; Pasciak et al., 1981). The GPU obtained 3300 person-rem by combining dispersion modelling with a 10 million curie source term and the TLD measurements. The lowest collective dose estimate, 500 person-rem, was derived on the basis of atmospheric dispersion modelling using a 2.4 million curie source term (Auxier et al., 1979). Thus, the central estimate is about 3300 person-rem with a range of several orders of magnitude.

Figure 3.8 Estimated dose pattern derived from the DOE aerial measurements from 28 March-3 April 1979. The units are mSv. (Reproduced by permission of the Lawrence Livermore National Laboratory. All rights reside with the US Government.)

Table 3.2 Estimates of collective dose equivalent due to noble gas releases from TMI-2 accident


		Person-Sv
Group	Technique	Best value	Range

DOE	Aerial measurements	20
GPU	TLD measurements and	33	16.566
	dispersion modelling
Ad Hoc Group	TLD measurements and	34
	dispersion modelling
Ad Hoc Group	Spatial interpolation	33	1653
	of measurements
President's	Source term and	5	0.5050
Commission	dispersion modelling

In the case of the much earlier accident at Windscale in 1957 relatively simple models were used at the time to estimate source terms from atmospheric concentrations measured over the United Kingdom (Crabtree, 1959). Subsequently the situation has been used to evaluate models developed for nuclear accident risk studies (e.g. ApSimon et al., 1989).

The release took place mainly in two peak periods, and the situation was complicated by the passage of a weak frontal system at about the time of the first peak. During the second peak, after water was poured on the pile early on the morning of 11 October 1957, the meteorological conditions were more straightforward with a steady well-defined plume blowing towards the south. This was well represented within a few tens of kilometres by a simple Gaussian plume model (Clarke, 1974).

Over the longer distance scales the trajectory model MESOS described the general character of the dispersion fairly satisfactorily despite the complications. The initial material was transported eastwards across England, subsequently turning southwards to reach London and the continent where it was detected during the following day.`

Concentrations in the southward plume from the second peak in the release were also in good agreement with observations; although as travel distances increased towards southern England the model overestimated spreading westward over Wales

possibly due to effects of the Welsh mountains in blocking the spread in that direction. Over the next two to three days the material stagnated in the anticyclone before moving off eastwards; again this was consistent with the observations. A model prediction, which could not be confirmed because of lack of measurements, was the fumigation action of the early part of the release to the surface over northeast England as the mixing layer dispersed on the morning of 11 October. Such situations can lead to temporary higher concentrations as material is mixed down to the surface.

Estimates of deposition of the principal nuclide released, ¹³¹I,were consistent with model estimates using a dry deposition velocity of 3 x 10^-3 m s^-1. Since no significant rainfall scavenged the material there could be no studies of wet deposition (although some cloud over hills may explain some observations of enhanced areas of deposition

another orographic effect not represented in current models).

Radionuclides released after the accident at the Chernobyl nuclear power plant were extensively monitored and were detected throughout the northern hemisphere. Theoretical models simulating the processes that affect radioactivity transport and deposition have been used in conjunction with the observations to deduce how the material dispersed on local to global scales. This dispersion is discussed here along with selected model results. The accident at Unit No. 4 of the Chernobyl plant occurred at 1.26 a.m. local time on 26 April, 1986. It is described in Section 2.7.

Gravitational settling of the coarser material (particles with a radius greater than 10 µm) released over the first five days dominated the deposition pattern close to the site. This pattern emphasizes the potential importance in nuclear accidents of hot fuel particles and the need for further studies of their behaviour in the environment. During this time winds at Chernobyl blew from all directions. Material from the initial release was carried westward; subsequently, over 26 and 27 April the plume swung to the southwest and northwest with fairly light and variable winds. During 28 April material went mainly north, turning toward the east on 29 April and then to the southeast and south to complete the circle on 30 April. The calculated deposition agrees very well with the observed distribution of radiation levels for the zone within 30 km of the plant, where evacuation took place (see Figure 2.5).

In the night following the explosion the initial plume travelled westward at a height of a few hundred metres in stably stratified air. During the day of 26 April, convective motions mixed the cloud vertically up to 2

3 km. Significant amounts were also transported to higher levels in cumulus clouds. The plume turned northward, reaching Scandinavia on 27 April. The following day radioactivity was incorporated in a developing frontal precipitation system, and localized peaks of wet deposition (in excess of 100 kBq m^-2) occurred in parts of central Scandinavia. A detailed analysis of this phase has been undertaken in Sweden.

Subsequently, radionuclides spread from Poland southwestward across Europe, south of a ridge of high pressure pushing across northern Europe from the Atlantic Ocean. Some of this radioactivity then turned northward to the United Kingdom behind the anticyclone that developed over Denmark. Other material travelled eastward across the Soviet Union and southward to Turkey and Greece. Much of the radioactivity eventually left Europe by moving across the North Sea and Scandinavia. Deposition of the most important long-lived nuclide, caesium 137, did not decrease smoothly with travel distance but was enhanced in regions where rain or snow intercepted the plume (see Figure 3.9).

What such a large-scale map fails to reveal is the extremely patchy nature of the deposition, which was often concentrated in a few small areas within a country and was highly correlated with local variations in rainfall amounts during passage of the radioactivity. This has been demonstrated in Scandinavia, the United Kingdom, and many other European countries. Nevertheless, except in a few particular areas, contamination from the Chernobyl accident outside the USSR was generally low compared with normal overall annual exposures from natural sources of radiation. An interesting observation is the effect of ionization on conductivity of the air over the areas of higher deposition, which in Sweden have been suggested to be associated with increased frequencies of lightning.

Figure 3.9 Total deposition across Europe of caesium-137 from the Chernobyl accident as estimated using the MESOS model of Imperial College, London (ApSimon and Wilson, 1987).

On the scale of the northern hemisphere, simulations such as those undertaken by the Lawrence Livermore National Laboratory and the Japan Meteorological Institute provide good visual demonstrations of the dispersal of the radionuclides, represented by an assembly of particles. These indicate how the radioactive material became segmented during the first day, with one part heading toward Scandinavia and hence over central Europe as already indicated, while the remainder was transported in an easterly direction across Asia to Japan, the North Pacific, and the North American continent (see Figure 3.10). These simulations imply that a significant fraction of the radioactivity was transported at elevated levels in the atmosphere

up to 4 km in the Japan estimates and 10 km for the Lawrence Livermore model.

Figure 3.10 Calculated spatial distribution of radioactivity over the northern hemisphere 10 days after the Chernobyl accident as illustrated by the Lawrence Livermore National Laboratory and the US Department of Energy (Lange et al. 1992).

Scientists have used modelling results in conjuction with observations to deduce the quantities of various nuclides released from the Chernobyl reactor and their variation in time between 26 April and 6 May. This task is complicated by the different behaviour of different parts of the release over different distance scales. Thus, emissions of large particles could not be directly evaluated from observations at longer distances, and there are uncertainties about the height attained by the initial release on the first day and its complex transport within the frontal system. Nevertheless, estimates from European and global modelling assessments are in agreement within a factor of about 2 (see Table 3.3).


Source of estimate		Iodine-131	Caesium-137
		(10¹⁷ Bq)	(10¹⁶ Bq)

USSR		2.8	4.7
MESOS model		1.7	3.9
	(European fallout)
Lawrence Livermore		6.0	8.9
	(northern hemisphere scale)

Note: All estimates are based on values of decay to 6 May 1986 as in the Soviet report to the International Atomic Energy Agency. Refers only to material travelling beyond the local zone where particles were deposited. Allowance for vertical dilution by convective mixing in clouds could increase this estimate by up to 30 per cent.

Analysis of the time-dependent variation from the MESOS model at the Imperial College in London accords with two peaks in the release and indicates a greater ratio of the less volatile ¹⁰³Rurelative to ¹³⁷Cs as core temperatures rose during the second peak. This is fully consistent with the Soviet accounts of the accident.

3.6 RECOMMENDATIONS AND RESEARCH NEEDS

Historically, much of the development of a scientific understanding of atmospheric transport of pollutants came from studies of radioactive tracers, either derived from bomb fallout, or in planned laboratory and field experiments. This knowledge was rapidly transferred to stable element species, as processes affecting pollutants such as heavy metals and sulphur oxides became a focus of interest. Nowadays, major atmospheric releases of man-made radioactivity are very infrequent and, as exemplified by Chernobyl, highly unpredictable. Because of the relatively short lifetimes of radioactive species in the atmosphere, experimental observations of atmospheric processes following such an accident must be made very rapidly and with a minimum of preparation. When this is considered, it is remarkable how much was learned from measurements after Chernobyl.

Many of the real research needs in this field can only be met fully by studies of nuclear accidents. However, research must nonetheless proceed at other times, and will depend upon laboratory and field experiments, theoretical studies, and in particular studies of processes using non-radioactive elements present in the atmosphere, with extrapolation to closely related radioactive elements and species. Thus, for example, research on terrestrial resuspension processes can readily progress through use of stable elements, or particles labelled in other ways, for instance, using fluorescent dyes.

3	Atmospheric Pathways
	Co-ordinator: R. M. Harrison
	Contributors: H. M. ApSimon, H. Beck, L. Devell, M. Dickerson, J. A. Garland, G. Graziani,
		P. Gudiksen, W. A. McKay, U. C. Mishra, K. W. Nicholson, F. B. Smith and
		C. S. Shapiro

	3.1 Introduction
	3.2 Atmospheric Processes
		3.2.1 Atmospheric Transport Processes and Tracers
		3.2.2 Transformation Processes and Aersosol Behaviour
		3.2.3 Dry Deposition
		3.2.4 Wet Deposition Processes
		3.2.5 Resuspension from Land
		3.2.6 Sea to Land Transfer
	3.3 Environmental Measurements
		3.3.1 Concentration Monitoring
		3.3.2 Meteorological Measurements
	3.4 Modelling
		3.4.1 Introduction
		3.4.2 Types of Atmospheric Dispersion Model
		3.4.3 Long-range Transport Models
		3.4.4 Model Evaluation Studies after Chernobyl
	3.5 Applications of Models to Important Accidents
		3.5.1 Three Mile Island
		3.5.2 Windscale
		3.5.3 Chernobyl
	3.6 Recommendations and Research Needs