Measurement network for urban noise assessment: Comparison of mobile measurements and spatial interpolation approaches A. Can a,⇑ , L. Dekoninck b , D. Botteldooren b a LUNAM Université, IFSTTAR, AME-EASE, F-44341 Bouguenais, France b Acoustics Group, Department of Information Technology, Ghent University, St. Pietersnieuwstraat 41, 9000 GENT, Belgium article info Article history: Received 31 May 2013 Received in revised form 19 December 2013 Accepted 17 March 2014 Keywords: Noise mapping Mobile measurements Spatial interpolation Noise indicators Measurement network optimization abstract This paper investigates the relevance of different interpolation techniques to improve the spatial resolu- tion of urban noise maps, in complement to measurements achieved at fixed stations. Interpolation tech- niques based on mobile measurements are compared to usual spatial interpolations techniques, namely Inverse Distance Weighting and Kriging. The analyses rely on a measurement campaign, which consisted of nearly 8 h of geo-referenced mobile noise measurements performed at random moments of the day, conducted simultaneously with continuous measurements collected at five fixed stations located on the inner city of Gent, Belgium. Firstly, a procedure is proposed to build a noise map with a high spatial resolution (one point every 5 m). The procedure relies on both mobile and fixed measurements: the mobile measurements are used to capture spatial variations on the network, and the measurements at fixed stations are used to capture the temporal variations. The map produced is then used as reference to compare the interpolation tech- niques based on a significantly more sparse measurement set. The spatial interpolation techniques tested fail in predicting accurately the noise level variations within streets. The explanation given is that they do not offer a sufficient covering of the network, and assume spatial variations which are not coherent with traffic dynamics or street configurations. Inversely, mobile measurements cover the entire network. As a result, they allow a more accurate prediction of noise levels even if very short samples are used, provided that the procedure used to estimate noise levels includes a spatial aggregation, which aims at smoothing the high spatial variations inevitable with short samples. Moreover, mobile measurements can advantageously be used to optimize, through a Genetic Algorithm, the locations where to install fixed stations, promising an efficient noise monitoring at reduced opera- tional costs. Ó 2014 Elsevier Ltd. All rights reserved. 1. Introduction The strong recent urbanization and the increasing demand of city dwellers for a better quality of life have led to the development of policies towards sustainable cities. A consequence in regards to noise is the enactment of the European Noise Directive, which made mandatory strategic noise maps for cities with more than 100,000 habitants [1]. Those maps play an important informative role, pointing black points and quiet zones; this information has for example a significant impact on the property market [2]. They can also be a powerful tool for comparing the impact of different noise reduction strategies [3,4], provided that a special care is gi- ven during the traffic modelling step [5–7]. Although the directive leaves some liberty concerning the methods to produce noise maps, modelling based on traffic data collection and sound propagation calculation is the most wide- spread technique [8]. Noise contour maps with grid spacing of less than 10 m are then recommended in urban areas [8]. Spatial inter- polation techniques can be used to determine noise levels with a greater resolution than the initial grid of results [9,10]. It has been shown however that this step can yield uncertainties, especially if maps are built based on wide initial grids [11]. Beyond their known advantages, modelling based on traffic data collection and sound propagation calculation has the disadvantage of needing some prior data collection and network acquisition steps that are long and costly. Moreover the estimation of noise levels within shielded streets requires time consuming sound propagation calculations [12,13]. Finally, some discrepancies can be observed between modelled maps, which mainly focus on road http://dx.doi.org/10.1016/j.apacoust.2014.03.012 0003-682X/Ó 2014 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. Tel.: +33 2 40 84 58 53. E-mail address: arnaud.can@ifsttar.fr (A. Can). Applied Acoustics 83 (2014) 32–39 Contents lists available at ScienceDirect Applied Acoustics journal homepage: www.elsevier.com/locate/apacoust