radon in groundwater - Mark- och vattenteknik - KTH

Radon in groundwater - Influencing factors and prediction methodology for a Swedish environmentRadon prediction using Risk VariableModellingThe RV modelling, which was based onlinear regression, generated risk values associatedwith each group of radon concentrations.Calibration of the model was based onhalf of the wells comprising the database(2209 wells) and the result is presented inthe form of a boxplot (Fig. 4).Radon concentrations were classified intodifferent concentration ranges based on theregulatory limits applicable in Sweden. It wasobserved that risk values became more negativewith increasing radon concentrations,although the standard deviations of thegroups were considerable.A test (or validation) of the Risk VariableModel was conducted on the other half ofthe dataset, comprising 2209 wells and distributedin 12 subareas. Risk values associatedwith corresponding radon concentrationswere computed. The results arepresented in the form of a scatterplot(Fig. 5), in which the median radon concentrationprevailing in each area was plottedagainst the corresponding median risk valuein that area. As can be observed in Fig. 5, ahigh negative correlation (-0.87) existedbetween the two plotted quantities, implyingthat the higher the radon concentration, themore negative the risk value. A similar trendwas observed in the calibration stage.A prediction map in GISRaster maps of elevation, bedrock, soil, landuse and uranium concentration in bedrockwere reclassified and the different classeswere assigned their corresponding ratings.An overlay operation was performed to get afinal risk map. Each pixel represents a finalrisk value equivalent to the sum of the riskvalues. The factor distance to fracture zonewas not included in the map productionsince it was not a surface map. Differentwater bodies (e.g. lakes, rivers) were thenadded to the final map. Three categories ofrisk areas were defined: Low risk for all finalrisk values, FRV > 0; medium risk for all,–5 < FRV < 0; high risk for all, FRV < -5.The risk map in a sample study area isshown in Fig. 6.In Paper I, the prediction map based on theRV method was compared to the map ofinterpolated airborne uranium concentrationsin the bedrock. It was observed thatsome regions with low to moderate uraniumconcentrations in the bedrock were definedas high-risk areas in the corresponding RVmap. However, such observations need tobe compared with field measurements infuture work. It should also be mentionedthat geological materials (bedrock and uraniumdistribution) used in the various analysesand available on a scale of 1:50 000 donot always reflect the geological conditionsthat prevail at a specific site.15