Application of Radon-Deficit Technique for Site Characterization and Machine Learning Integration: Case Studies and Emerging Insights

The radon-deficit technique has proven to be a valuable tool for environmental site characterization, particularly in detecting subsurface organic contamination. This work highlights its successful application in two contaminated sites, validated by consulting firms and supported by independent data collection campaigns. In the first case study, the technique effectively identified previously undetected DNAPL (Dense Non-Aqueous Phase Liquid) accumulations and optimized the placement of monitoring wells. Similarly, in the second case, radon-deficit data delineated areas potentially impacted by LNAPL (Light Non-Aqueous Phase Liquid) contamination, refining the sampling approach and complementing existing geochemical methods.

Building on these findings, a study is underway to integrate long-term radon data with machine learning (ML) techniques. By analysing environmental variables such as soil moisture, temperature, and atmospheric conditions, this approach aims to reduce the uncertainties inherent in radon-deficit data interpretation. Preliminary results indicate that ML models, such as Random Forest and Artificial Neural Networks, can enhance the predictive accuracy and reliability of the technique, paving the way for standardized protocols in site assessments. This integration represents a significant step toward more robust and scalable applications of radon-deficit methods in environmental monitoring.