A statistical regression framework for evaluating noise exposure through outdoor-to-indoor sound propagation models

Urban noise exposure has been shown to have significant adverse effects on human health. Studies based on outdoor noise levels, rather than indoor levels, risk introducing bias when identifying an exposure-response relationship with health outcomes. This bias is of particular concern for vulnerable groups (e.g., children) who spend most of their time indoors. This highlights the need for a comprehensive approach to modelling outdoor-to-indoor sound propagation for the estimation of indoor noise levels. For both outdoor and indoor propagation, energetic approaches are used, while the facade sound insulation is modeled via the Transfer Matrix Method, incorporating the facade element size and bridge connections. Validation of this modelling approach against sound insulation measurements has revealed good agreement. The employment of the developed models resulted in the generation of datasets, facilitating the implementation of statistical learning (random forest) models for estimating (equivalent) indoor noise levels using predictors associated with outdoor levels and factors related to the indoor and outdoor environment, and facade structure. The proposed approach has the potential to estimate indoor levels for entire cities from predicted outdoor noise levels (noise maps) and built environment characteristics.