Dynamic noise mapping: A map-based interpolation between noise measurements with high temporal resolution

• Published in: Applied acoustics 2016-01, Vol.101, p.127-140
• Main Authors: Wei, Weigang, Van Renterghem, Timothy, De Coensel, Bert, Botteldooren, Dick
• Format: Article
• Language: English
• Subjects: Dynamic noise map; Environmental noise monitoring; Sound propagation outdoors
• ISSN: 0003-682X; 1872-910X
• DOI: 10.1016/j.apacoust.2015.08.005

Since the introduction of the Environmental Noise Directive, strategic noise mapping has been used as a tool for noise policy in many European countries. Although these strategic noise maps have their merits, they also have some shortcomings: accuracy in predicted noise levels in shielded or quiet areas is not very high, the maps fail to capture sounds that are less easy to predict, and above all the dynamics of the sound environment are not included. However, these dynamics might be important to evaluate sleep disturbance and noise annoyance. In this paper, a model to dynamically (every 15min) update a noise map based on measurements is proposed. This model relies on reasonable good source and propagation models and a not-very-dense measurement network. The least mean squares method (LMS) is used for tuning model parameters. To avoid an under-determined system, the number of degrees of freedom is reduced by grouping the sources and propagation paths into different categories. Source strengths and propagation path attenuations in the same category are corrected by offsetting the same small values from their base levels. The map-based interpolation is performed jointly on LAeq,L10 and L90, and takes into account 1/3-octave band spectra. The efficiency of the proposed method was validated in a case study in the Katendrecht district of Rotterdam, the Netherlands. The results showed that more than 75% of the LAeq predictions are closer to the measurement than the ab initio calculations based on traffic data. Values for L10 and L90 are closer to measurements for 55% and 90% of the observations, respectively.