Road Traffic Noise Predictions by means of L10 Modelling with a Multilinear Regression Calibrated on Simulated Data

Estimation of road traffic noise is fundamental for the health of people living in urban areas, and it is usually assessed based on field-measured data. Real data may not always be available, anyway, and for this reason, predictive models play an important role in the evaluation and controlling of the noise impact. In this contribution, the authors present a multilinear regressive model calibrated on simulated noise levels instead that on real measured ones, correlating percentile noise levels to independent traffic variables. The model efficiency is then evaluated on two field measurement datasets by analyzing data statistics and error metrics. Results show that the model provides good results in terms of mean error (less than 1 dBA on average) even if slight underestimations and overestimations are present. The presented model, then, can be used to assess the impact of road traffic noise anytime field measurements are not available, or even predict it when designing new road infrastructures.