Discussion

Analysis of variance results indicate that the sample groups based upon the classification map of Duval (2005) represent different populations at confidence levels greater than 90 percent. Discriminant analysis results support the classification of the samples although initial classification accuracies are below 50 percent for classes Qb and Ql (refer to Table 15. Reclassification of a limited number (14) of samples that were identified as misclassified and were located near the boundary between classes improved the overall classification accuracy (refer to Table 16). All of these results support a conclusion that the classification map of Duval (2005) can be used to predict soil properties such as radon emanation and soil gas radon concentrations. The graphs of Figure 6 allow one to predict the following:

1. Structures built on soils classified as Tg (Goliad Formation) and Ql (Lissie Formation including some areas mapped as sand dunes) will have similar distributions of indoor radon with less than 5 percent greater than 4 pCi/L.

2. Structures built on soils classified as Qb will have higher average indoor radon levels than areas of Tg or Ql and might have as much as 50 percent greater than 4 pCi/L. The maximum indoor radon level is predicted to be less than about 6 pCi/L.

3. Structures built on soils classified as Qal will have higher average indoor radon levels than Qb, Tg, or Ql and might have as much as 50 percent greater than 4 pCi/L with a maximum greater than 10 pCi/L.

The results of this study clearly show that geologic studies using the methods discussed here can identify geologic units that pose higher risks of indoor radon. Because geology does not conform to county boundaries, the practice of characterizing the radon risk on a county basis should be changed. Identification of areas of higher risk would allow state and local authorities to focus their resources and educational efforts.

In order to place the above predictions in a context of human health, population data from the 2000 census were downloaded from the ESRI web site. Refer to  the U.S. Census Bureau website for details about the census. The census block data were converted to population densities and the geologic unit polygons of Page, VanSistine, and Turner (2005) were used to estimate the numbers of people living within the boundaries of those polygons. Table 17 lists the results of that calculation. According to the predictions given here, as many as 295,000 residents of the lower Rio Grande valley could be exposed to indoor radon levels greater than 4 pCi/L. According to the U.S. Environmental Protection Agency 7 out of 1,000 people who never smoked and were exposed to lifetime exposure at a level of 4 pCi/L could get lung cancer. For smokers this number increases to 62 out of 1,000. If the predictions presented here are valid, from 2,500 to 22,000 residents of the study area could get lung cancer during their lifetime.

Because these predictions are based upon calculated estimates of the radon concentrations in the soil gas and upon the assumption that the amount of radon gas entering a slab-on-grade structure in the lower Rio Grand valley will be similar to amounts observed in California, they should be viewed with caution and follow-up studies should be conducted to test them. Any such follow-up study should use a stratified random sampling protocol. The stratification areas should conform to the interpreted classes of Duval (2005) and/or the geologic units of Page, VanSistine, and Turner (2005). Regardless of the predictions for absolute indoor radon levels, relative values of average indoor radon concentrations should be consistent with the above predictions.

Table 17. Estimated U.S. populations (U.S. Census Bureau) living within areas underlain by the geologic units. Unit Qal includes units Qas and Qam of Page, VanSistine, and Turner (2005). Unit Qb includes units Qb, Qbf, and Qbm of Page, VanSistine, and Turner (2005). Unit Ql includes units Ql and Qds of Page, VanSistine, and Turner (2005). Unit Tg is that of Page, VanSistine, and Turner (2005).