A new high-resolution indoor radon map for Germany using a machine learning based probabilistic exposure model

• DOI: 10.48550/arxiv.2310.11143
• Publication Date: 2023-01-01
• Authors: Eric Petermann; Peter Bossew; Joachim Kemski; Valeria Gruber; Nils Suhr; Bernd Hoffmann
• Keywords: Quantile

Radon is a carcinogenic, radioactive gas that can accumulate indoors. Therefore, accurate knowledge of indoor radon concentration crucial for assessing radon-related health effects or identifying radon-prone areas. Indoor at the national scale usually estimated on basis extensive measurement campaigns. However, characteristics sample often differ from population due to large number relevant factors control such as availability geogenic floor level. Furthermore, size does not allow estimation with high spatial resolution. We propose model-based approach allows more realistic distribution higher resolution than purely data-based approach. A two-stage modelling was applied: 1) quantile regression forest using environmental and building data predictors applied estimate probability function each level residential in Germany; (2) probabilistic Monte Carlo sampling technique enabled combination weighting floor-level predictions. In this way, uncertainty individual predictions effectively propagated into variability aggregated The results show an approximate lognormal arithmetic mean 63 Bq/m3, geometric 41 Bq/m3 95 %ile 180 Bq/m3. exceedance 100 300 are 12.5 % (10.5 million people) 2.2 (1.9 people), respectively.