Are noise events from surface transport predictable? Insights from a wide measurement campaign

The negative effect of road traffic noise events on annoyance is now established. However, the assessment and monitoring of road traffic noise remain mainly based on energetic indicators, which are easy to handle but mask noise dynamic structure. Recent developments in dynamic road traffic modelling, and in urban sensor networks, suggest that introducing noise events in urban noise management is possible. This however raises statistical questions: although their inherent random origin (very noisy cars, sirens, etc.) make them hardly predictable, noise events are probably site dependent. In this paper, we rely on a measurement campaign carried out in Toulouse (France), made of 20 1h-measurement periods covering both day and night time slots, to question some statistical matters relative to road traffic noise events. Firstly, some general reflections concerning candidate indicators for describing noise events are given, in line with road traffic noise dynamics. Then, a statistical method is proposed, which selects the frequency bands of interest, and then defines a set of indicators relevant to describe the urban soundscape of the site, in terms of noise events. Finally, some insights about the predictability of noise events are deduced from the spatial distributions of the selected set of indicators.