Development of a multi-level predictive soundscape model to assess the soundscapes of public spaces during the COVID-19 lockdowns Free

The recent developments in the standardization of soundscape as both a research and engineering field have highlighted the need for models which can predict likely soundscape assessment from objective measurements. Such a need was highlighted during the COVID-19 lockdowns. The unprecedented restrictions in human activity presented a unique opportunity to investigate the urban noise impacts of drastic reductions in traffic noise and human sounds, but simultaneously made it impossible to carry out standard methods of soundscape assessment (i.e., in-person surveys or soundwalks). To address this, a multi-level linear regression model was developed based on an existing database of soundscape surveys and binaural recordings to predict how the soundscapes of 13 locations in London and Venice would have likely been perceived during the lockdowns based on objective measurements. To build this model, a feature selection process was applied to an extended suite of psychoacoustic metrics and a variable characterising the context of each location to identify a minimum set of input features and model structure. This presentation will demonstrate the development of this model, its application in the COVID case study and corresponding results, and will discuss the potential for future applications of a similar predictive soundscape modelling framework.