Soil uranium, soil gas radon and indoor radon empirical relationships in the UK and other European countries

Transcription

1 Soil uranium, soil gas radon and indoor radon empirical relationships in the UK and other European countries Don Appleton (BGS) and Jon Miles (HPA)

2 EUROPEAN GEOGENIC RADON POTENTIAL MAP Wide range of numerical (quantitative) data types: indoor radon, soil gas radon, U in soil (measured and estimated from airborne or ground gamma spectrometry), U in rock; U in sediment. European radon potential classification based on combining layers of available input variables requires an understanding of the empirical relationships between indoor radon (main hazard to human health) and other numerical data types. European radon potential classification will also have to deal with categorical (qualitative) data (e.g. rock type and structural features, such as faults) but this is outside scope of this presentation.

3 DATA Estimated U in <2mm topsoil derived from estimated U (eu) from airborne gamma spectrometry [Estimated <2mm soil U = * HiRES eu mg/kg)] U in topsoil (5-20 cm depth) soil geochemical samples (<2mm fraction); potentially better correlation with airborne gamma spectrometry data than subsurface (20-50 cm soil U data) Radon concentrations in soil gas Radon concentrations in homes Data grouped by geological unit and sub-area (usually 1-km or 5-km grid square)

4 REGRESSION ANALYSIS Spatially and geologically grouped data should be approximately normally distributed to give most robust central estimate for least squares linear regression analysis Anderson-Darling normality tests indicate regression modelling should be based on: 1. Arithmetic means of airborne eu and soil U data 2. Geometric means of lognormally distributed soil gas and indoor radon data LS regression analysis results have limitations related to uncertainties of x and y axis data AND outliers

5 MODELS BASED ON AIRBORNE GAMMA SPECTROMETRY DATA Estimated U in <2mm topsoil derived from estimated U (eu) from HiRES airborne gamma spectrometry [Estimated <2mm soil U = * HiRES eu mg/kg)]

6 Figure 1: Least squares regression models between estimated soil U (derived from HiRES eu) and GM indoor radon (data grouped by 1-km grid square and geology; thin lines have intercepts set to 5 Bq m -3 )) 400 DINLM GM indoor radon Bq m DINMDMIX TRIMD Linear (DINLM) Linear (DINLM) Linear (DINMDMIX) Linear (DINMDMIX) Linear (TRIMD) Linear (TRIMD) estimated soil U mg/kg

7 INTERCEPTS High positive intercepts on y axis of least squares (LS) linear regression models probably due to uncertainties of indoor radon, soil gas radon and soil U data UNSCEAR (1993) estimated world mean outdoor radon (10 Bq m -3 ) and indoor contribution from building materials (6 Bq m -3 ) UNSCEAR intercept of 16 Bq m -3 inappropriate for UK where a value of 5 Bq m -3 is the best estimate for combined indoor contribution from outdoor radon and UK building materials LS regression models for UK data therefore forced to intercept y-axis at 5 Bq m -3 (thin regression lines in previous slide)

8 Figure 2: Least squares regression models between estimated soil U (derived from HiRES eu) and GM indoor radon (data grouped by 1-km grid square and geology) compared with data for Oslofjord region and N. Ireland 300 GM indoor radon Bq m DINLM DINMDMIX TRIMD N. Ireland Oslofjord Linear (DINLM) Linear (DINMDMIX) Linear (TRIMD) Linear (N. Ireland) Linear (Oslofjord) estimated U mg/kg in topsoil

9 MODELS BASED ON SOIL CHEMISTRY DATA U (mg/kg) in <2mm fraction of topsoil determined by XRF

10 250 Figure 3: Least squares regression models between U in <2mm fraction of topsoils and GM indoor radon (data grouped by 5-km grid square and geology; dashed lines have intercepts set to 5 Bq m -3 ) GM indoor radon (Bq m -3 ) DINLM INOLMST INONS WESMDMIX Linear (DINLM) Linear (DINLM) Linear (INOLMST) Linear (INONS) Linear (INONS) Linear (WESMDMIX) U mg/kg

11 MODELS BASED ON SOIL GAS RADON DATA Geometric mean of soil gas radon concentrations (Bq/L) grouped by geology and grid square

12 Figure 4: GM soil gas radon vs. GM indoor radon (data grouped by 1-km or 5-km grid square and geology) GM Indoor radon (Bq/m3) BGS Derby-Notts Carb-Perm 1km grouping BGS Derbyshire Carb Lmst 1km grouping ( data) BGS Derby-Notts Carb-Perm 5km grouping BGS INONS 1km grouping BGS INONS 5km grouping Linear (BGS Derby-Notts Carb-Perm 1km grouping) Linear (BGS Derbyshire Carb Lmst 1km grouping ( data)) Linear (BGS Derby-Notts Carb-Perm 5km grouping) Linear (BGS INONS 1km grouping) GM Soil gas radon (kbq/m3) Linear (BGS INONS 5km grouping)

13 Figure 5: GM soil gas radon vs. GM indoor radon (data grouped by 1-km or 5-km grid square and geology; intercepts set at 5 Bq m -3 ) 400 BGS Derby-Notts Carb-Perm 1km grouping GM Indoor radon (Bq/m3) BGS Derbyshire Carb Lmst 1km grouping ( data) BGS Derby-Notts Carb-Perm 5km grouping BGS INONS 1km grouping BGS INONS 5km grouping Linear (BGS Derby-Notts Carb-Perm 1km grouping) Linear (BGS Derbyshire Carb Lmst 1km grouping ( data)) Linear (BGS Derby-Notts Carb-Perm 5km grouping) Linear (BGS INONS 1km grouping) Linear (BGS INONS 5km grouping) GM Soil gas radon (kbq/m3)

14 Figure 5a: Least squares regression models between GM and AM indoor radon (data grouped by 5-km grid square and geology)

15 Figure 6: Soil gas radon vs. estimated arithmetic mean indoor radon linear regression models for England, Czech Republic and Germany Arithmetic average groundfloor indoor radon (Bq/m3) y = 5.878x y = x y = x y = x y = x y = x y = x y = x Czech; Barnet et al 2008 Tab data Czech Republic Barnet et al., 2002 Czech Republic: Barnet et al., 2004 BGS Derby-Notts Carb-Perm 1km grouping BGS Derbyshire Carb Lmst German NUTS5 units;barnet et al 2006 Kemski et al., 2009 N. Bohemian Massif; Barnet et al., 2010 Linear (Czech; Barnet et al 2008 Tab data) Linear (Czech Republic Barnet et al., 2002) Linear (Czech Republic: Barnet et al., 2004) Linear (BGS Derby-Notts Carb-Perm 1km grouping) Linear (BGS Derbyshire Carb Lmst ) Linear (German NUTS5 units;barnet et al 2006 ) Linear (Kemski et al., 2009) Soil gas radon (kbq/m3) Linear (N. Bohemian Massif; Barnet et al., 2010)

16 Indoor radon: soil gas radon ratios UK (GM) models similar to arithmetic mean SG Czech data Both significantly higher than German models based Maximum SG data Different house national and regional characteristics will also impact on IR/SG ratios Older buildings with leaky floors are likely to have higher IR/SG ratios Příbram (Czech Rep) has higher IR/SG ratio than average; lower ratios for magmatic rocks, medium ratios for metamorphic, highest ratios for sedimentary rocks.

17 Figure 7: Soil gas radon vs. estimated arithmetic mean indoor radon linear regression models for England, Italy and Croatia 800 y = x BGS Derby-Notts Carb-Perm 1km grouping Arithmetic average groundfloor indoor radon (Bq/m3) y = x y = 6.559x y = 2.02x y = x y = x BGS Derbyshire Carb Lmst BGS Derby-Notts Carb-Perm 5km grouping Italy - lower Italy - higher Croatia Linear (BGS Derby-Notts Carb-Perm 1km grouping) Linear (BGS Derbyshire Carb Lmst ) Linear (BGS Derby-Notts Carb-Perm 5km grouping) Linear (Italy - lower) Linear (Italy - higher) Soil gas radon (kbq/m3) Linear (Croatia)

18 CONCLUSIONS ASSUMPTIONS: linear relationship between surrogates for source of radon (e.g. eu or U) and radon in homes DATA DISTRIBUTIONS : eu and soil U normally distributed (therefore use AM for modelling); soil gas and indoor radon log-normally distributed (use GM) MEASUREMENT UNCERTAINTIES: impact on slopes and intercepts of LS regression models; slopes generally steeper for fixed intercepts (5 Bq m -3 in UK) PERMEABILITY OF THE GROUND: Radon concentrations in houses generated from specific U or soil gas radon concentrations generally increase with the permeability of the ground GEOGENIC RADON MAP OF EUROPE: further research required to generate robust empirical relationships for the wide range of numerical input data types (and also for categorical data types).