Publication
Since the 80s, indoor radon, a radioactive noble gas, became a public health concern, with the WHO declaring it a carcinogen in 1987. Accurate radon measurement is crucial for estimating exposure and shaping policies to reduce lung cancer. Nowadays, the reliability of radon measurement is challenged by different trends: 1) the advent of IoT and environmental monitoring in buildings. 2) The increased awareness of indoor air quality, especially following the SARS-CoV-2 pandemic, increasing interest in monitoring indoor air quality and radon. 3) Renovating buildings to reduce energy use often worsens indoor air quality. This thesis addresses these key challenges in assessing indoor radon. Following introduction and literature review, Chapter 3 provides new insights into the performance of passive dosimeters and real-time radon devices. Current methods for measuring radon are inconsistent, making it difficult to benchmark concentrations and accurately assess exposure. The study compared passive dosimeters and three price-based grades of real-time monitors at low (300 Bq/m3) and high (2'000-3'000 Bq/m3) concentrations in lab conditions. Consumer-grade sensors underperformed compared to research-grade models, while medium-grade sensors matched consumer-grade at low levels but performed like research-grade at high levels. Performance disparities were larger at low radon concentrations. Finally, passive dosimeters improved with longer exposure and higher levels. Results show real-time sensor performance correlates with price, emphasizing the need to choose devices based on measurement goals to reduce errors. Lab experiments also revealed no significant impact of common aerosol sources on real-time sensors, confirming their robustness against environmental conditions. Chapter 4 evaluates the performance of real-time radon measurement devices in real-world conditions to ensure a smooth transfer of knowledge from the lab to practical use. The results show that real-time sensors performed similarly in real-life conditions as they did in the lab. Additionally, passive measurement protocols used in Switzerland were compared and indicated long-term passive measurements were only slightly affected by duration, affirming the reliability of 3- and 6-month measurements, even for periods shorter than one year. Chapter 5 explores alternatives to direct radon measurement by comparing statistical methods (e.g. multiple linear regression, logistic regression, random forest regression/classification) to predict indoor radon levels based on environmental and buildings factors. After providing a list of method, through a PRISMA systematic literature review, results showed that predictive accuracy varies depending on the dataset, with random forest classification achieving up to 85% accuracy, although this remains insufficient. No method can predict indoor radon levels without error, highlighting the need for further research. Researchers must carefully choose prediction me