Vapour intrusion from the vadose zone—seven algorithms compared

Abstract

Vapours of volatile organic compounds (VOCs) emanating from contaminated soils may move through the unsaturated zone to the subsurface. VOC in the subsurface can be transported to the indoor air by convective air movement through openings in the foundation and basement. Once they have entered the building, they may cause adverse human health effects. Screening-level algorithms have been developed, which predict indoor air concentrations as a result of soil (vadose zone) contamination. The present study evaluates seven currently used screening-level algorithms, predicting vapour intrusion into buildings as a result of vadose zone contamination, regarding the accuracy of their predictions and their usefulness for screening purpose. Screening aims at identifying contaminated soils that should be further investigated as to the need of remediation and/or the presence of an intolerable human health risk. To be useful in this respect, screening-level algorithms should be sufficiently conservative so that they produce very few false-negative predictions but they should not be overly conservative because they might have insufficient discriminatory power. For this purpose, a comparison is made between observed and predicted soil air and indoor air concentrations from seven reasonably well-documented sites, where the vadose zone was contaminated with aromatic or chlorinated VOCs. The seven screening-level algorithms considered were: Vlier–Humaan (Be), Johnson and Ettinger model (USA), VolaSoil (NL), CSoil (NL), Risc (UK) and the dilution factor models from Norway and Sweden. Calculations are presented in two scatter plots (soil air and indoor air), each containing the predictions versus the observations. Differences between predicted and observed VOCs concentrations were evaluated on the basis of three statistical criteria to establish their accurateness and the usefulness for screening purposes. Results from the applied criteria are presented in a table and figures. It was found that the screening-level algorithms investigated tended to overestimate soil air concentrations more than indoor air concentrations. Differences between predictions and observations were up to three orders of magnitude. The algorithms with the highest accuracy for predicting the soil air concentration are in ascending order the Johnson and Ettinger model (JEM), Vlier–Humaan and VolaSoil algorithms. For the indoor air, it is concluded that all algorithms have a tendency to overestimate the predicted indoor air concentrations, except for the JEM and Vlier–Humaan algorithms, which produced frequent underestimations. Several earlier studies have investigated the accuracy of some of the screening-level algorithms for vapour intrusion and the results presented in the present study agree with the findings. However, the present study presents the accuracy of vapour intrusion algorithms via three statistical criteria that allow their ranking. The present study also determines the suitability of screening-level algorithms as screening tool. It is found that algorithms may rank differently as to accuracy and suitability as a screening tool. The algorithms with the highest accuracy for predicting the indoor air concentration are the JEM and Vlier–Humaan algorithms. The most suitable algorithms to serve for screening purposes are CSoil, VolaSoil and Risc, since they are sufficiently conservative, have fewer false-negative predictions and still have sufficient discriminatory power. Given the over-predictions and under-predictions of the algorithms considered, a combination of modelling and measurements will often be required to produce multiple lines of evidence for the presence of an intolerable human health risk or the need for remedial actions at a site. Integrated programmes of modelling and field observations can reduce the uncertainty of predicted soil air and indoor air concentrations, and a tiered approach is presented in this study.