Environmental noise exposure has shown to have significant negative effects on people’s lives. In noise exposure studies, outdoor noise levels are usually preferred over indoor levels for investigating exposure-response relationships, introducing a systematic risk of bias. Hence, an outdoor-to-indoor propagation modelling framework is defined for estimating indoor noise levels based on outdoor levels. Particularly, by expressing outdoor and indoor sound propagation via energetic models and façade (multi-component and multi-layered) structures via computational models, outdoor-based indoor impulse responses can be generated. To validate the framework, a case study was conducted, showing that the measured and simulated sound insulation were in good agreement. Finally, this framework was applied to generate datasets of outdoor and indoor noise levels (noise indicators) based on scenarios of outdoor-indoor environments and façade structures. This allows the training of statistical learning approaches for estimating indoor noise levels and identifying important predictors. Results show that a random-forest approach outperforms the a stepwise and a neural network approach across all the employed noise indicators (RMSE<2dB). These models enable the assessment of indoor exposure and the exploration of exposure-response relationships in locations with known built environment characteristics.
Terzakis et al. (2025) studied this question.