Radon Dispersion From A South Africa Gold Mine-Tailings Dam – Measurement And Modelling

Abstract

Radon has been recognised as the main contributor to the natural radiation dose exposure to human. In addition to natural radon, human activities like mining have the potential to enhance environmental radon levels. Gold-mine tailings dams contain traces of 238U and 226Ra, leading to generation of 222Rn gas in the tailings material due to radioactive decay. Current methods used to monitor radon from the tailings dams are only able to provide a close-up of emissions in space and time. These methods cannot distinguish between tailings radon and background radon as well as the extent each contributes towards the radon content in the atmosphere. The only way to determine the increment is through dispersion modelling. This study develops a technique to accurately validate radon dispersion modelling that assesses the radon contribution from Freddies 9 (sometimes referred to as Steyn 9) tailings dam situated in Odendaalsrus, Free State Province, South Africa. This study was structured into four parts. The first part dealt with determination of the radon exhalation rate from the tailings dam. The second part dealt with measuring the ambient radon concentration in the vicinity of the dam using Radon Gas Monitors (RGMs). The third part was to measure radon gas, individual radon daughters, and the F factor at different receptor points downwind by following the direction of the wind at hourly intervals. The fourth part involved using the ISCST3 dispersion modelling code to evaluate radon transport and the effects of local variations around the tailings dam. Field data were collected during winter months of June to August in 2016 and 2017. Measurements of the radon exhalation rate, which is the source term for dispersion modelling from the tailings dam, were performed with the passive diffusion tube method. Twenty (20) tailings samples were collected for analysis. The exhalation rate (E) was found to vary from 0.045 Bq/(m2s) to 0.443 Bq/(m2s) with an average value of 0.102 Bq/m2s and a mean standard deviation of 0.087 Bq/(m2s). The ambient radon concentration values measured with the Radon Gas Monitors (RGMs), averaged over 116 locations around the tailings dam distributed over a radius of 2 km from the tailings dam, range from 38 Bq/m3 to 94 Bq/m3, with a mean concentration of 64 ± 11 Bq/m3. These values are below the 100 Bq/m3 action level stipulated by the National Nuclear Regulator (NNR) with a slight average increase compared to background levels of 60 Bq/m3. Airborne radon concentration at any given location is influenced by locally exhaled radon and dispersed radon from other locations. As a means of trying to discriminate between different radon contributors, radon gas, individual radon daughters, and the F factor were measured at different receptor points downwind and upwind by following the direction of the wind at hourly intervals. The AlphaGUARD was used to measure radon concentration and the Eberline SPA-1A alpha scintillation detector coupled to Eberline Smart Portable (ESP-2) counter were used to measure the radon daughters. The Busigin and Phillips three count method was used to calculate radon daughter concentrations and hence the F factor. The minimum value of F factor was 0.016 ± 0.012 measured upwind whereas the maximum value of F factor was 0.502 ± 0.044 measured downwind. Calculations revealed strong influence by external meteorological effects on the distribution of radon and radon daughters some distance from the tailings and background. The F factor, which indicates the “age” of the gas, and radon gas, increased to their highest values when the wind was blowing from north-northern-east (NNE). The highest radon daughter concentrations at various locations were recorded in the mornings. However, fluctuating and conflicting effects due to different meteorological conditions on the resultant atmospheric radon concentration, radon daughters’ concentration and F factor as functions of distance from the tailings downwind, made the interpretation of results difficult. To further quantitatively explain these results, an air dispersion model was applied. The USA Environmental Protection Agency's (EPA) Industrial Source Complex Short Term 3 (ISCST3) dispersion modelling code using a Gaussian plume model was used to evaluate radon transport and the effects of local variations around the tailings dam. The tailings was modelled as point, total emitting surface area (true geometry) and volume source. The true area geometry was considered as the baseline source geometry. To improve the accuracy of the model predictions as compared to traditional approaches, the area source term was corrected to account for cracks and fissures on the tailings and the geometry of tailings dam was modelled by taking into account all emitting surfaces as sources. Compared to the baseline, the model over predicted the flat ground area source by up to 274 % and under predicted the top level area source by up to 50%. The volume emission source was over predicted by up to 300% in 60% of the modelling runs and under predicted by 55% in 40% of the volume model runs. While the top level area source term produced lower concentrations at near-field ground-level receptors, accounting for the wakes effect increased the radon concentrations from the top-level area source of the tailings dam by up to 239%. From modeling results, the highest concentration predicted by the model from the true geometry source was found to be 0.843 Bq/m3, which correspond to the dose of 0.012 mSv/y to the public of due to radon from the tailings. This value is less than the 1 mSv/y dose constraint stipulated by the NNR. Model validation from statistical analysis showed a constant trend for all the scenarios, with minimum variability in the Index of Agreement (IOA), Normalized Mean Square Error (NMSE) and Fraction of predictions within a factor of two (FAC2) values. The analysis were based on the model results over five days of measurements covering both morning and afternoon. There is an under prediction in the Fractional Bias (FB) and Geometric Mean bias (MG) on day 1 afternoon. In addition, the model performed poorly on day 3 afternoon. Further validation of the model was carried out by isolating radon from different contributors using the “age” of the gas approach and applying back calculations to identify origin of the radon measured at each point downwind. As predicted by the model, the origin of the radon source was traced back to the tailings.