Optimisation of 85 Kr environmental monitoring to provide a validation of atmospheric dispersion models

Transcription

1 Optimisation of 85 Kr environmental monitoring to provide a validation of atmospheric dispersion models T.G. Parker, R. A. Hill 2, I. Lowles 2, N. Chambers 2 and I. Coleman 2 BNFL, Sellafield, Seascale, Cumbria, CA20 PG, United Kingdom. 2 Westlakes Scientific Consulting Ltd., The Princess Royal Building, Westlakes Science and Technology Park, Moor Row, Cumbria, CA24 3LN, United Kingdom. INTRODUCTION At the planning Inquiry, into the construction and operation of THORP () the discharge of 85 Kr and associated doses to members of the public was a significant issue. This was partly due to the predicted dominant contribution of 85 Kr to the critical group dose (45 µsv). Changes in the dosimetry of 85 Kr (principally ICRP2 to ICRP30) have led to a reduction in the predicted dose which, for the current authorisation, is 6.5 µsv (2). Despite this reduction 85 Kr has remained an important isotope because of the scale of the annual discharge (98,900 TBq total from Magnox and THORP in 998) and it was considered important that BNFL demonstrated the validity of the dose predictions. In the early 990s, prior to the operation of THORP, BNFL had been investigating means of validating the atmospheric dispersion code (3) used in the dose prediction model. This was considered necessary, principally because the validation sets were based on flat terrain, isolated stacks; whereas Sellafield has multiple large buildings, moderate topography and is coastal. The field scale validation was to augment work already carried out with wind tunnel modelling (including buildings and terrain) of all the major stacks (4). The validation experiments were to use the classic tracer, sulphur hexafluoride (SF 6 ), during short duration releases. There were several major problems with this approach: difficulties of injection into an operating radioactive ventilation system, logistics of sampling/analysis and global warming potential of the SF 6. The SF 6 global warming was of major concern; to obtain the same data spread of wind directions and weather categories, as the experiment described in this paper, would have required approximately.6e+04 kg of SF 6. Whilst the detailed resolution of plume profiles would have been much finer, it is doubtful whether that would have justified the release of 4.0E+05 te CO 2 equivalent. The installation of a fully automated meteorological station in 995, with 0 minute data resolution and computerised data capture, opened the way for a site dispersion validation using the 85 Kr discharged from the reprocessing plants. The discharges of 85 Kr are measured to a precision of +/-0 % and have been validated with isotope production codes (5). The combination of a well characterised source with the isotope`s 0.7 year half life and chemically inert properties make 85 Kr an ideal tracer for the Sellafield Site. The environmental measurements required for the dispersion validation would then also validate the dose predictions. This paper presents field measurements of the air concentrations of 85 Kr, and the comparison between two atmospheric dispersion models; the NRPB R-9 model (3) currently in use at Sellafield and its possible successor the next generation UK-ADMS model (6). 85 Kr DOSIMETRY The biological half-life of 85 Kr is estimated at 30s (7) and therefore the dose to members of the public is dominated by external radiation. The proportions of dose, due to the β-particles (mean energy 0.25 MeV) and the associated gamma ray (54 kev), depend principally on the extent of the plume (infinite vs semi-infinite) and whether people are outside or inside. For dose assessments at Sellafield no account is taken of any beta shielding by clothing and buildings are assumed to have a shielding factor of 0.2 (8). There are two critical group locations (residences) for Sellafield, with similar levels of dose but which are affected differently depending on the source of the discharge (Magnox or THORP). At these locations the proportion of the total predicted dose, due to the 85 Kr β particles is approximately 60 %. Direct measurement of the 85 Kr air concentration at ground level (and thereby the beta component) is capable of validating up to 60 % of the predicted dose and providing confidence in the overall dose assessment. METHODS Monitoring of 85Kr air concentrations Samples of air were collected in PTFE sampling bags, with a capacity of m 3, over sample durations of either two hours or approximately 24 hours. The longer duration samples were collected at seven monitoring sites, including the BNFL Sellafield critical group locations, east of the plant and west of the plant. All the sites are shown in Figure. The short duration samples were collected at various locations, all of which were downwind of the discharging stack, at a typical distance of.0 km.

2 Figure. Map of the BNFL Sellafield site and surrounding area showing the 24 hour 85 Kr monitoring sites (numbered) and the locations of the stacks emitting 85 Kr (THORP and Magnox). Crown Copyright. Air samples were returned to the laboratory for the crude separation of the total krypton fraction including both stable and radioactive isotopes following a method similar to that described in Janssens et al (9). Air from the sample bag was passed through a series of traps maintained at a reduced pressure. H 2 O was initially removed by condensation onto the walls of a glass dewar at liquid nitrogen temperature. Subsequently, CO 2 was removed by adsorption onto molecular sieve granules at ambient temperature. Krypton, some nitrogen and other trace atmospheric constituents were then trapped in an activated charcoal column at liquid nitrogen temperature. Once the sampling bag had been emptied, the activated charcoal column was allowed to warm to room temperature and finally was heated to 80 o C. The krypton liberated from the charcoal column was preferentially retained on a conditioned molecular sieve coil (at reduced pressure and liquid nitrogen temperature). Fine separation of the total krypton fraction and beta counting to determine the 85 Kr content were performed by the Department of Subatomic and Radiation Physics at the University of Gent in Belgium. Ambient 85 Kr air concentrations were calculated from the measured ratio of 85 Kr to stable Kr assuming a constant mixing ratio of.4 ppm (by volume) of stable Kr in the atmosphere. The overall uncertainty in the determination of ambient air concentrations of 85 Kr has been estimated to be +/- 5% (0). Stack monitoring 85 Kr is emitted from two stacks on the Sellafield site: Magnox and THORP, shown in Figure. Continuous stack monitoring of 85 Kr at both stacks is conducted by BNFL in order to demonstrate compliance with the relevant discharge authorisations. The pattern of 85 Kr discharge, shown in Figure 2, differs between the two plants, with the Magnox plant producing a more continuous emission, due to fuel dissolution, and the THORP plant producing a more temporally variable emission, due to fuel shearing. Meteorological data Meteorological data were obtained from the 48 m high Sellafield meteorological mast, this site is shown as location three in Figure. The mast is equipped with anemometers, wind vanes and thermistors at heights of 2 m, 4 m, 0 m, 6 m, 24 m, 32 m, 40 m and 48 m. The datalogger attached to the system stores data, from each of the meteorological sensors, at a 0 minute time resolution. 2

3 8 Emission (GBq s - ) THORP MAGNOX 0 6:00 2:00 8:00 0:00 6:00 Time (GMT) Figure 2. Typical patterns of emissions of 85 Kr from the THORP and Magnox plants on the BNFL Sellafield site. Wind speed (u) measured at 0 m height and wind direction measured at the highest functioning wind vane were used as input to the dispersion model. Data from the higher vanes were used as they better represent the wind direction at the mean height of the 85 Kr plume. During periods when wind speed data were unavailable from anemometers at 0 m, wind speeds at this height were interpolated from measurements at other heights using the classical logarithmic wind profile equations (). Atmospheric stability, defined by the Monin-Obukhov stability length, was estimated from temperature and wind speed measurements using the micrometeorological Flux-profile relationship, assuming a constant flux layer (2). Pasquill-Gifford stability class was estimated, from the Monin-Obukhov stability length, using the relationship derived by Golder (3) for a roughness length of 0.3 m. This roughness length is typical of agricultural areas and is consistent with the roughness length used when conducting modelling of the atmospheric dispersion of radionuclides released from the Sellafield site. Hourly averaged meteorological and stack emissions data were used as input to the atmospheric dispersion models when comparing model predictions with the field measurements. The annual averaged dispersion and predicted doses at the critical group sites were determined using statistical input data on stability class distributions, wind speed distributions and wind direction distributions. These statistical distributions were determined from an analysis of the hourly data recorded between and Data were subdivided into 2 thirty degree wind sectors with the distribution of stability classes and the mean wind speed for each stability class being determined for each sector. Modelling the atmospheric dispersion of 85Kr Two Gaussian plume atmospheric dispersion models, the NRPB R-9 model (3) and the UK-ADMS model (6) were used to predict the air concentration of 85 Kr in the field. These models used input data of: measurement site location, meteorological conditions and stack emissions corresponding to each of the 85 Kr field measurement periods, ensuring that the model predictions were directly comparable with the field measurements. The main difference between the atmospheric dispersion models is that UK-ADMS uses a more modern method of boundary layer scaling, dependent on the Monin-Obukhov length, which allows for vertically inhomogeneous atmospheric turbulence to be modelled. This contrasts with the simple turbulence-typing scheme used in R-9 which assumes that atmospheric turbulence is vertically homogeneous. The UK-ADMS model can include complex effects such as coastlines, buildings and terrain, however the simple model configuration used in this study did not make use of these modules. The influence of plume rise and building induced downwash on the centreline height of the plume was included through the determination of effective stack heights, as discussed in Fulker et al (4). These heights were determined from wind tunnel studies using a scale model of the BNFL Sellafield site, the site model is shown in Figure 3. Downwind dispersion factors were determined at the centreline of a tracer plume released from the relevant stack on the site model. Effective stack heights were calculated by comparing the measured downwind dispersion factors with the predictions of both the R-9 and UK-ADMS atmospheric dispersion models. An example of such a comparison is shown in Figure 4. Corrections were applied to account for differences in the treatment of lateral dispersion between the numerical models and the wind tunnel. 3

4 .E m Figure 3: Wind tunnel model of the BNFL Sellafield site.e m DF (s m -3 ).E m 80 m 00 m.e Distance (m) Figure 4: Example of the determination of an effective stack height from wind tunnel measurements of dispersion factors (DF). Wind tunnel data uncorrected for lateral spread is shown as diamonds ( ), whilst wind tunnel data corrected for lateral spread are shown as squares ( ). The atmospheric dispersion models were run with identical hourly chronologies of meteorological data, as discussed overleaf. Hourly ground level dispersion factors (in s m -3 ) were determined using the models, treating each stack separately and using a unit emission (.0 Bq s - ). These dispersion factors were rescaled using the hourly emission rate for each stack (in Bq s - ) to give hourly estimates of stack specific air concentration (in Bq m -3 ). The hourly air concentration from both stacks was calculated as the sum of the contribution from each source. These hourly concentrations were averaged over the duration of each of the field measurement periods to provide a direct comparison with the time-averaged air concentration measured in the field. RESULTS Field measurements of 85Kr air concentrations 85 Kr air concentration measurements are summarised by site in Table. In total 02 air concentration samples were collected split between 78 long term measurements and 24 short term measurements. Ensemble means of 08.5 and 75.3 Bq m -3 were measured for long and short term sampling respectively with, as expected, higher air concentrations being measured for the short-term samples, collected immediately downwind of the stacks. Of the 24 short-term samples six were collected in neutral conditions and 8 were collected in convective conditions. Background 85 Kr concentrations were in the region of -2 Bq m -3 whilst the peak concentration measured within the 85 Kr plumes was 550 Bq m

5 Site Number of Air concentration (Bq m -3 ) measurements Mean Maximum Minimum Long term (all sites) 2 hourly (all sites) Table. Summary of air concentration measurements by site. Comparison between model predictions and short term (2 hour) field measurements The NRPB R-9 and UK-ADMS models were used to simulate the dispersion of 85 Kr, following the methods discussed in the previous section. These model simulations were conducted to reproduce the conditions encountered in the field and provide a direct comparison between predicted and measured air concentrations. A scatterplot showing the comparison between the NRPB R-9 model predictions and the short term field measurements is shown in Figure 5, whilst the comparison between the UK-ADMS model and the short term field measurements is shown in Figure 6. Both models showed a fairly wide variability between their predictions and the measurements. The R-9 model tended to both over-predict and under-predict the measured air concentrations whilst the UK-ADMS model showed more of a tendency towards over-prediction. The comparison between the model predictions and the field measurements was also evaluated statistically. The fractions of model predictions within factors of two, five and 0 of the field measurements (F-2, F-5 and F-0) were evaluated to describe the accuracy of the model. The correlation statistic (R) and the mean bias (MB, defined as the ensemble mean of the model predictions/ the ensemble mean of the measurements) were used to describe the precision of model predictions. Results of the statistical analysis, shown in Table 2, demonstrate that the UK-ADMS model was superior to the older R-9 atmospheric dispersion model in all the tests, producing less biased and more accurate dispersion predictions. Overall both models over-predict concentrations and for dose assessment purposes, therefore, there is no reason to change from the existing NRPB R-9model. F-2 F-5 F-0 R MB R ADMS Table 2: Statistical evaluation of the performance of the NRPB R-9 and UK-ADMS models in comparison with the short-term (2 hour) 85 Kr air concentration measurements. F-2, F-5 and F-0 are the fractions of model predictions within factors of two, five and 0 of the field measurements, R is the correlation statistic and MB is the mean bias. 5

6 00000 Modelled (Bq m -3 ) Measured (Bq m -3 ) Figure 5: Comparison between the short-term field measurements and the predictions of the NRPB R-9 atmospheric dispersion model. The solid line shows a : fit whilst the dashed line shows fit within a factor of two Modelled (Bq m -3 ) Measured (Bq m -3 ) Figure 6: Comparison between the short-term field measurements and the predictions of the UK-ADMS atmospheric dispersion model. The solid line shows a : fit whilst the dashed line shows fit within a factor of two. Comparison between model predictions and long- term (24 hour) field measurements Scatterplots comparing the NRPB R-9 and UK-ADMS model predictions with long term field measurements of 85 Kr are shown in Figures 7 and 8. As observed in the previous section, a considerable scatter was found between the predictions of both models and the field measurements. Both models showed a tendency towards over-prediction, particularly in the range of measured concentrations between Bq m -3. The R-9 model also showed an increased tendency to under-predict concentrations in the range Bq m -3. Both models were found to produce closer dispersion predictions in the measured concentration range between 00 0,000 Bq m -3, with the R-9 model predictions being slightly closer to the field measurements. A statistical analysis was also conducted to make a quantitative assessment of the relative performance of each model. The results, shown in Table 3, demonstrated little difference between the predictions of each model and the field measurements. F-2 F-5 F-0 R MB R ADMS Table 3: Statistical evaluation of the performance of the NRPB R-9 and UK-ADMS models in comparison with the long-term (24 hourly) 85 Kr air concentration measurements. F-2, F-5 and F-0 are the fractions of model predictions within factors of two, five and 0 of the field measurements, R is the correlation statistic and MB is the mean bias. 6

7 0000 Modelled (Bq m -3 ) Measured (Bq m -3 ) Figure 7: Comparison between the long-term field measurements and the predictions of the NRPB R-9 atmospheric dispersion model. The solid line shows a : fit whilst the dashed line shows fit within a factor of two Modelled (Bq m -3 ) Measured (Bq m -3 ) Figure 8: Comparison between the long-term field measurements and the predictions of the UK-ADMS atmospheric dispersion model. The solid line shows a : fit whilst the dashed line shows fit within a factor of two. Estimated annual offsite dose Annual doses received by members of the offsite critical groups, east and west of Sellafield, were calculated for 998 using statistically analysed meteorological data. This data was analysed as 30 o wind sectors, calculating the sector correlated stability class distributions and the average wind speeds corresponding to each stability class. Seven stability classes were used corresponding to the A-G turbulence typing scheme, as presented in (3). This method of calculation, whilst being relatively simple to implement, assumed that the release of 85 Kr was continuous throughout the year. Future work will investigate the effect of conjunction between discharge and meteorology. A comparison was also made between the 85 Kr dose calculated for the 998 meteorological conditions and the dose calculated for generic meteorological data (60 % stability class D, and a wind rose biased along the south-east to north-west axis). External gamma doses were calculated, for the R-9 model, from the ground level air concentrations assuming a semi-infinite plume. A gamma dose coefficient of 4.73E-3 Sv hr - per Bq m -3 was determined by subtracting the β dose coefficient of 4.44E-3 Sv hr - per Bq m -3 quoted in Simmonds et al (8) from the cumulative dose coefficient (including γ and β) quoted in ICRP 72 (4). The assumption of a semi-infinite plume was justified for downwind distances of km or more from a comparison of calculated gamma doses with the results of the NRPB ESCLOUD model. At distances of between 500 m and 000 m from the source external gamma doses were no longer linearly related to the surface level air concentration, hence a correction factor of 2.0 was applied in line with the ESCLOUD model predictions. External gamma radiation doses were calculated directly from the UK- ADMS model using a finite cloud approach. The dose conversion factor for beta radiation was assumed to be independent of the location of the individual, such that an identical exposure was predicted for individuals exposed within their homes and those outside. For gamma radiation from the plume a building shielding factor of 0.2 was applied for individuals resident within buildings (8). A generic building occupancy of 66 % was assumed and a further conservative assumption of 00 % site occupancy was made. The results of the dose calculations are presented in Table 4. The offsite doses for the 998 statistical meteorological data, presented in Table 4, show that the 7

8 estimates of external gamma dose were consistent between models, at both critical group locations. More substantial differences were found between the estimates of beta radiation dose and the air concentration with R- 9 predicting the higher dose at Critical Group East and UK-ADMS predicting the higher dose at Critical Group West. Such differences may be explained by the increased rate of vertical dispersion predicted by UK-ADMS to occur close to the source. This acts to enhance ground level air concentrations at short distances from the stack, whilst air concentrations further from the stack tend to be reduced by the increased atmospheric dilution. Despite these apparent differences the total doses (γ + β) predicted by the UK-ADMS and R-9 models were reasonably consistent (within 30 %) for Critical Groups East and West. Data in Table 4 also show that the choice of meteorological data was more influential on the predictions of off-site dose than the choice of model. The R-9 dose increased by 7 % at Critical Group West and 43 % at Critical Group East when the generic meteorological data were used as model input. The difference between the model estimates for these conditions was likely to be due to the higher mean wind speeds measured during 998 than estimated from the generic meteorological dataset. Model Critical group Air (Bq m -3 ) Gamma dose (µsv a - ) Beta dose (µsv a - ) Total dose (µsv a - ) ADMS (98-STAT) East West R-9 (98-STAT) East West R-9 (GEN) East West Table 4: Doses at the BNFL Sellafield aerial critical group sites due to exposure to airborne 85 Kr. Calculations were made using two atmospheric dispersion models (ADMS and R-9) and two meteorological datasets, the 998 statistical meteorological data (termed 98-STAT) and generic assumptions of long term stability class and wind speed distributions (termed GEN). A comparison was also made between the modelled and measured β dose corresponding to the field measurements. The ensemble mean modelled and measured concentrations, from both long-term and short-term measurements, were calculated and the β doses determined using the aforementioned dose coefficient, data are summarised in Table 5. The data in Table 5 provide a further example of the dispersion model over-predicting the beta dose, with UK-ADMS predicting the highest doses for the long-term sampling and R-9 predicting the highest doses for the short-term sampling. The ensemble mean β dose from all the long-term measurements corresponded to an annual dose of 0.42 µsv, which compared with values of.06 µsv and.4 µsv, from R-9 and UK-ADMS respectively. Number of Mean modelled β dose psv hr - Measured measurements NRPB R-9 UK-ADMS β dose psv hr - Long-term Short-term Table 5: Summary of modelled and measured β dose estimates for long-term and short-term 85 Kr sampling. CONCLUSIONS Using 85 Kr discharges as a dispersion tracer was successful and removed the need to release large quantities of the traditional tracer gas, sulphur hexafluoride (with a large global warming potential). This programme, of field measurement of 85 Kr concentrations, has completed the data set required to demonstrate that the dispersion behaviour of aerial discharges, from Sellafield, is adequately represented in the dose assessment models. The combination of detailed wind tunnel simulations and classical mathematical models is particularly powerful. The field measurements of 85 Kr have demonstrated that the current aerial dispersion model (NRPB- R9), used for critical group dose prediction, is conservative for both short term and long term discharges. They have also demonstrated that, in the Sellafield context, the new generation model (UK-ADMS) is also conservative but is more accurate for short duration discharges in convective conditions. The current dose assessment model overestimates the beta component of critical group dose by approximately a factor of three. It is reasonable to conclude therefore that total dose (including gamma) is also overestimated. REFERENCES. Hon Mr Justice Parker, The Windscale Inquiry, volume report and annexes 3-5 (London: HMSO) (978). 2. BNFL Application for a Variation to the Certificate of Authorisation for the Disposal of Waste Gases, Mists and Dusts from the BNFL Premises on the Sellafield Site, BNFL (996). 8

9 3. R.H. Clarke, A model for short and medium range dispersion of radionuclides released to the atmosphere. First report of a working group on atmospheric dispersion, National Radiological Protection Board, NRPB R-9 (979). 4. M.J. Fulker & S. Singh, The role of wind tunnel dispersion measurements in critical group dose assessments for new plant. In proceedings of Air pollution 94 held in Barcelona September 994, p (994). 5. D. Jackson, C.H. Zimmerman & J. Gray, Discharges of krypton from Sellafield, , and the resultant doses to members of the public. Journal of Radiological Protection, Vol. 8, No. 2, -8 (998). 6. D.J. Carruthers, R.J. Holroyd, J.C.R. Hunt, W.S. Weng, A.G. Robins, D.D. Apsley, D.J. Thomson & F.B. Smith, UK-ADMS: A new approach to modelling dispersion in the Earth s atmospheric boundary layer. Journal of wind engineering and industrial aerodynamics, Vol. 52, p (994). 7. A.D. Turkin, Dizimetriya Radioaktivnkh Gazov (Dosimetry of radioactive gases) (Moscow: Atomizdat) in Russian (973). 8. J.R. Simmonds, G. Lawson and A. Mayall, Methodology for Assessing the Radiological Consequences of Routine Releases of Radionuclides to the Environment. EUR-5760 (Luxembourg: European Commission, DG-XIII) (995). 9. A. Janssens, J. Buysse, F. Raes & H. Vanmarcke, An improved method for the sampling of atmospheric 85 Kr. Nuclear Instruments and Methods in Physics Research, B7, p (986). 0. A. Janssens, J. Buysse & E. Cottens, The measurement of low-level atmospheric krypton-85. Nuclear Instruments and Methods in Physics Research A234, p (985).. J. L. Montieth & M. H. Unsworth Principles of environmental physics, 2 nd edition. Edward Arnold, London (990). 2. A.J. Dyer & B.B Hicks, Flux-gradient relationships in the constant flux layer. Quarterly Journal of the Royal Meteorology Society, Vol. 96, p (970). 3. D.G. Golder, Relationships between stability class in the surface layer. Boundary Layer Meteorology, Vol. 3, p (972). 4. ICRP, Age dependent doses to members of the public from intakes of radionuclides. ICRP publication 72 (996) 9