Statistical requirements for noise mapping based on mobile measurements using bikes

Abstract

This research presents a modeling framework that allows checking the statistical requirements for producing noise maps based on mobile measurements. First, a sound field of reference is created based on a micro-simulation traffic modeling coupled with acoustic modeling, which outputs sound levels each second on a grid of receivers. The aggregated indicators (LAeq) calculated from this sound field serve then as reference. Mobile targets performing measurements evolve within the simulation, aiming to estimate these indicators. The difference between the reference noise map and the one generated by the moving receivers, characterized by the Root Mean Square Error (RMSE), is computed for different aggregation radius of mobile receivers, and as a function of the number of passes-by and to the distance to its nearest cross street. It is observed that the mobile sampling is actually possible and the RMSE can be reduced by setting an optimal aggregation radius and a minimum number of passes-by. With the optimal parameters, 95% of the mobile samples fall within an estimation error interval of [−3.0, 2.2] dBA from the reference. It is also shown that the distance to the nearest cross street affects the estimation error depending on the traffic flow, producing a RMSE greater than 2 dB for distances lower than 30 m.