Geospatial modelling of road-traffic noise levels and frequency and the attributable burden of annoyance and sleep disturbance in Accra, Ghana

Background and aim: Limited city-wide data on environmental noise and sources in rapidly growing sub-Saharan African (SSA) cities constitutes a major barrier for investigating health impacts as well as environmental policy making. In a first of its kind study in SSA, we modelled and predicted noise levels and road-traffic-specific sounds in Accra, Ghana, and estimated the attributable burden of being highly annoyed and sleep disturbed in high-spatial resolution. Methods: From 2019-2020, we collected measurements of sound levels and audio recordings along the roadside in a large-scale campaign. The audio was processed with a deep learning acoustic classifier to identify the frequency of road-traffic sounds. We combined the acoustic data with geospatial predictors in land use regression models (mixed models/random forest) to predict noise levels (Lden, Lnight) and the frequency of road-traffic-specific sounds across the city. Finally, by combining population exposures to predicted Lden and Lnight with literature informed exposure-response relationships and disability weights, we estimated the attributable burden of being highly annoyed and sleep disturbed in aggregate and by census enumeration area (median size: 0.03km2). Results: Predicted road-traffic sounds were prevalent throughout the day (median: 81% of the time present) and nighttime (median: 62%) in Accra. Furthermore, 99% of the population in lived in census enumeration areas where average Lden and Lnight surpassed WHO guidelines for road-traffic noise (Lden <53; Lnight <45). Noise exposures in Accra translated into 21% and 7% of the population highly annoyed and sleep disturbed, with significant variation across areas, and a combined 10,761 Disability Adjusted Life Years lost. Conclusions: In an area of the world where noise research is severely lacking, this work can support epidemiological studies, burden of disease assessments, and the development of policies and interventions that address noise exposure within Accra. Keywords: Noise, Africa, health burden, land-use-regression, audio processing, road-traffic noise