Radon Exposure in Canadian Residences: Predictive Models, Low-Cost Sensor Evaluation, and Multi-Family Building Assessments

Abstract

Radon gas is a naturally occurring carcinogen that infiltrates homes and buildings, where it can accumulate to dangerous levels. Although it is the second leading cause of lung cancer in Canada, public testing and mitigation remain limited. Modern changes to residential construction, particularly the move toward airtight designs, have inadvertently contributed to increased radon levels in Canada. Alongside geological variations and the emergence of new radon detection tools, the interplay of all these factors needs to be further explored. This thesis examines the residential radon exposure across Canada through three key research objectives: 1) the development of predictive models for radon risk; 2) the evaluation of low-cost electronic radon monitors (ERMs); and 3) the characterization of radon patterns in Canadian high-rise multi-family buildings. To address the first objective, random forest regression models were developed using national and regional datasets from Health Canada (HC) and the British Columbia Radon Data Repository (BCRDR), respectively. These models integrated spatial, seasonal, and building-specific variables to estimate residential radon levels. The BCRDR model explained up to 47% of the variance in radon concentrations, whereas the HC model was lower at 27%, with forward sortation area ranking consistently as the most influential predictor across both models. However, limited cross-dataset performance using the BCRDR model on the HC model highlights the ongoing challenge in the transferability of models from one region to another. The second objective involved short- and long-term colocation tests of several low-cost ERMs with reference-grade instruments in residential settings. Results show that while a few ERMs achieved an acceptable agreement with reference values (within 10% relative percent error), most demonstrated substantial variability, even among sensors of the same model. Some issues included delayed response times and wide limits of agreement. These findings emphasize the limitations of deploying ERMs for consumer or research use without regulatory supervision. For the third objective, radon levels were monitored in a 12-storey, multi-family student residence building at the University of Waterloo. Average radon concentrations were relatively low (3.7 – 8.9 Bq/m3); however, concentrations exceeding 50 Bq/m3 were recorded during periods of reduced ventilation on all floors. Notably, the conventional assumption that radon concentrations decrease with building height was challenged by the observation of the highest average concentration on the 12th floor, albeit with low absolute values. Collectively, these results provide insights into the limitations of predictive modelling for substitution in real-world testing, the performance of low-cost ERMs and reliability for use for testing, and the complex nature of indoor radon distribution, particularly in multi-family buildings. A coordinated strategy integrating predictive risk mapping, validated ERMs, and targeted testing can inform evidence-based policies and ensure that radon risk is addressed across all housing types and Canadian regions.