noise impact map

a Institute of Transport Economics, Grensesvingen 7, N-0602 Oslo, Norway effort. The methodology should work well for mapping Europe’s ‘‘black’’ and ‘‘grey’’ areas. The EU-directive to harmonise noise exposure and noise reaction measures [1] has initiated a chain of events that will result in noise maps being produced for all major * Corresponding author. Tel.: +47 22 573 800; fax: +47 22 570 290. E-mail addresses: rk@toi.no (R. Klæboe), eer@ssb.no (E. Engelien), ste@ssb.no (M. Steinnes). Applied Acoustics 67 (2006) 620–642 www.elsevier.com/locate/apacoust 0003-682X/$ - see front matter � 2006 Elsevier Ltd. All rights reserved. � 2006 Elsevier Ltd. All rights reserved. Keywords: Noise mapping; Road traffic noise; Annoyance; Soundscape; NALden 1. Noise mapping in Europe 1.1. Pan-European noise mapping initiatives b Statistics Norway, Oterveien 23, N-2225 Kongsvinger, Norway Received 13 September 2004; received in revised form 17 August 2005; accepted 1 November 2005 Available online 3 February 2006 Abstract Apartments that are exposed to the same level of road traffic noise on front of the most exposed fac¸ade often have very different neighbourhood soundscapes. In the first part of this paper, a neigh- bourhood soundscape adjusted exposure indicator, NALden, is derived. NALden-values are designed to be used as input to traditional exposure–effect relationships to improve annoyance impact esti- mates. In the second part, generic spatial procedures are developed and implemented. These produce map presentations in the form of contiguous neighbourhood quality areas. The quality of each neighbourhood is determined from the predicted annoyance impacts for residents. Noise impact maps provide experts, politicians, and the public with high-level impact visualizations of condensed status, ‘‘what-if’’ and scenario information. Results and illustrations are based on data from the Norwegian socio-environmental survey database, and a comprehensive national noise mapping Context sensitive noise impact mapping R. Klæboe a,*, E. Engelien b, M. Steinnes b doi:10.1016/j.apacoust.2005.12.002 transport sources and urban agglomerations on a pan-European scale. Noise exposures from each noise source are to be mapped separately, and a combined exposure map pro- duced. To facilitate the progress towards common noise mapping procedures, the Euro- R. Klæboe et al. / Applied Acoustics 67 (2006) 620–642 621 pean commission has established a working group, and a good practice noise-mapping guide is available [2]. Global and local noise abatement procedures have in common that their impacts are local, and amenable to map presentations. Uniform exposure indicators, noise calculation methods, common noise mapping procedures, and pulling together the impacts of all major noise sources, constitute significant breakthroughs. These efforts will result in noise maps that are comparable with respect to methodology and content across nations and European regions. As the availability of traffic flows, traffic composition, digital maps and updated traffic counts improves, the quality of the noise maps will improve over time and cover not only the ‘‘black’’ areas where noise levels exceed 65 dB, but the larger segment of the dwellings in the ‘‘grey’’ areas of Europe exposed to between 55 and 65 dB. However, traditional noise maps have important limitations. They are notoriously dif- ficult to interpret by non-experts, politicians and the public who are neither familiar with road traffic noise exposure measures nor their associated impacts. Exposure maps often show exposures without regard to where people live, work, or otherwise are affected, whereas it is the conflict areas that are of main interest when developing action plans. Information overlay techniques can be used with noise maps to provide information on areas of conflict, but multi-layer maps quickly become overloaded with information and are doubly unsuitable for the non-experts. 1.2. Noise impact mapping as an alternative to exposure mapping The main purpose of the noise mapping efforts of Europe is to provide relevant infor- mation for global and local action plans. The maps have thus to serve as tools for obtain- ing popular support for funding noise abatement projects in competition with other projects making claims for scarce municipal, regional and national resources and to dis- seminate results in a form that the public can understand. As it is impacts and changes in impacts, not exposures, which ultimately provide ‘‘meaning’’ and the policy relevance of altered traffic flows, noise abatement policies, or noise reduction at the source, an alter- native paradigm to exposure mapping is to map impacts. Noise impact maps supplement the noise exposure maps that are needed by acousticians and planners. A map of predicted current or future impacts provides as a visualization of the expected annoyance generating potential of a given residential location. In its simplest form, impact mapping can be achieved by applying exposure–effect1 rela- tionships to noise exposure indicators for each individual dwelling. For a predicted impact, such as the degree of road traffic noise annoyance, the relevant %HA (percentage highly annoyed), %A (percentage annoyed) or mean annoyance indicator can be calcu- lated. Mapping impacts constitute a paradigmatic shift that facilitates the visualization of composite or higher level indicators. Such indicators may e.g. weigh together the 1 The term exposure–effect is used instead of dose–response as the time spent at home or in other noisy environments is not used to calculate a dose. 622 R. Klæboe et al. / Applied Acoustics 67 (2006) 620–642 impacts of annoyance and sleep disturbances, take the sensitivity of the exposed popula- tion into account, constitute a mapping of noise costs, etc. 1.3. The neighbourhood context is important Environmental impact assessments usually focus on the noise level, Lden,fac¸ade, on the most exposed fac¸ade of each dwelling. However, the immediate neighbourhood of a dwell- ing is also an important subject for analysis [3–5]. There is also an increased interest in obtaining more detailed analyses of the acoustical environment of building complexes and neighbourhoods [6]. In Sweden a large research programme [7] has been launched to improve noise calculations in shielded urban areas [7,8], explore the benefits of having access to quietness [9–12] and derive more meaningful descriptions of urban sound envi- ronments [13]. Using a more conventional approach, Klæboe et al. [14] investigated whether residents became more annoyed with road traffic noise when the neighbourhood soundscape of their apartment was adverse. The authors concluded that a noisy neigh- bourhood meant an increased level of noise annoyance at home. Traditional exposure– effect relationships neglecting neighbourhood soundscape information resulted in inferior road traffic noise annoyance predictions. They also showed that changes in noise impacts due to changes in traffic, depended on the relative quality of the neighbourhood sound- scape. However, for society to profit from the added insights of soundscape research, it is not enough to show the importance of the acoustical or perceptual soundscape, but also develop procedures that allow the improvements in methodology to be part of main stream noise calculation and noise mapping practices. 1.4. A two-tiered approach is developed This paper presents a two-tiered methodology for presenting context sensitive noise impact maps. The first tier of procedures adapt the results of Klæboe et al. [14] to improve the precision of impact estimates obtained by applying exposure–effect relationships such as those from Miedema and Oudshoorn [15]. The second tier consists of a set of generic spatial procedures that visualizes neighbourhood impacts. These are implemented using geographical information systems (GIS). The spatial routines produce noise impact maps where the colouring reveals in which neighbourhood areas the annoyance impacts are most severe. 2. Method Detailed presentations of the studies, the noise calculation methods, the calculations of the neighbourhood soundscape quality and the questionnaires have previously been pro- vided [14,16,17]. A brief description is given below. Three studies in Oslo East in 1987, 1994 and 1996 [18] and a study in Drammen con- sisting of two surveys undertaken in 1998 and 1999, respectively [19] were used for devel- oping the procedures undertaken. After quality assurance, there were 3913 respondents available for analyses of exposure–effect relationships between road traffic noise and road traffic outdoor annoyance, and 3940 for analyses of indoor annoyance. The annoyance questions distinguish between the situation outside the apartment and inside the apartment: People where first asked: ‘‘Can you hear noise from road traffic when you are right outside the apartment?’’ People were thereafter asked ‘‘Is the noise highly, 2 R. Klæboe et al. / Applied Acoustics 67 (2006) 620–642 623 somewhat or not annoying for you?’’. With respect to indoor annoyance, the questions were: ‘‘Do you hear road traffic noise (when) inside your dwelling?’’ (The answers were ‘‘Yes’’, ‘‘No’’ and ‘‘Not applicable’’, and ‘‘Do not know’’). If yes: ‘‘Is this noise highly, somewhat or not annoying for you?’’ For the statistical analyses, the response categories ‘‘Do not hear’’ and ‘‘Not annoyed’’ were merged to be compatible with the data sets employed by Miedema and Oudshoorn [15]. The 24 h equivalent noise levels at the apartments most exposed side, LAeq,24h, were cal- culated using the Nordic calculation method. To make it easier to compare the results with those produced internationally, the values have summarily been converted to A-weighted Lden,fac¸ade values – See also [14,16]. Statistics Norway has implemented a national noise annoyance mapping system [20] to serve as a monitoring device for progress with respect to the ambitious national noise reduction target. The exposure data for Oslo is used as input to illustrate the implemen- tation of the noise impact mapping when road traffic noise exposure information is avail- able for all dwellings within an urban area. 2.1. Definition of the neighbourhood maximum difference The highest equivalent road traffic noise levels that are encountered near a dwelling are used initially by Klæboe et al. [14] to indicate the quality of the neighbourhood sound- scape. The focus is on the adversity of the acoustic soundscape in the neighbourhood of a dwelling and not silent sides or supportive areas. The operational definition of the neigh- bourhood soundscape maximum noise level (Lneigh,max) is thus the highest equivalent noise exposure value encountered at dwellings or along pavement areas within a fixed distance (75 m) of an apartment. A more detailed presentation is provided in [14]. Analyses that compare the respective impacts of localized noisy and silent areas in the area are provided in [17]. For use in statistical analyses, Klæboe et al. [14] introduced the neighbourhood maxi- mum difference: Ldiff,max = Lneigh,max � Lden,fac¸ade. The neighbourhood maximum differ- ence is simply the number of decibels that the equivalent noise level in the immediate neighbourhood of an apartment exceeds the noise level at the most exposed fac¸ade of the residence. It describes the adversity of the immediate neighbourhood relative to the noise level encountered in front of the most exposed fac¸ade of the apartment itself (Lden,fac¸ade). 2.2. Illustration of the neighbourhood maximum difference To introduce these terms in a less formal manner Klæboe et al. [14] provided an illus- tration of the road traffic noise exposure of three different apartments that are located near a main road – See Fig. 1. Shielding provided by intervening buildings, or a location further away from the main road entails a much lower exposure at the most exposed fac¸ade for apartments B and C (60 dB), than for A (72 dB). 2 The questions in the 1987 were slightly different – See [16] for a detailed description. 624 R. Klæboe et al. / Applied Acoustics 67 (2006) 620–642 Due to the reduced traffic volume, the exposure level in front of the most exposed fac¸- ade of apartment D is also 60 dB. It is now possible to focus on the quality of the neigh- bourhood soundscape of apartments B, C and D that all are exposed to the same road traffic noise level at the most exposed fac¸ade. The road traffic noise levels that the residents of apartments B and C encounter when they walk along the main road exceeds the noise level in front of their apartments with as much as 15 dB. (Their neighbourhood sound- scape noise level, Lneigh,max, is 75 dB and the neighbourhood maximum difference Ldiff,max = 15.) For apartment D, the noise levels in front of the apartment and in the neighbourhood are more in line, and the road traffic noise exposure value of 60 dB pro- vides an adequate description of the road traffic noise level both the apartment and of the neighbourhood soundscape (Ldiff,max = 3). With respect to residential noise annoyance impacts, Klæboe et al. [14] showed that Fig. 1. Front row and second row apartments. Noise exposure in front of the most exposed facades Lden,fac¸ade and the relative adversity of the neighbourhood soundscape Ldiff,max. locations B and C are preferable to location A, but that location D is to be preferred over locations B and C. 2.3. Harvesting neighbourhood soundscape information In the Drammen studies, the neighbourhood soundscape noise levels were determined from noise exposure calculations. In the three Oslo studies and for the national noise map- ping efforts, spatial GIS routines where applied to the collection of geographically located residential noise levels (Lden,fac¸ade-values). The spatial routines were utilized to extract the highest equivalent road traffic noise level experienced by other residents within a 75 m radius of an apartment and this value was assigned as the apartments Lneigh,max-value – See Fig. 2. 2.4. Statistical procedures Exposure–effect relationships for road traffic noise–noise annoyance, such as those derived by Miedema and Oudshoorn [15], are obtained without statistical control for the quality of the neighbourhood soundscape. This means that the reported degrees of R. Klæboe et al. / Applied Acoustics 67 (2006) 620–642 625 annoyance were obtained from residents of different neighbourhoods, some with relatively quiet and some with relatively noisy acoustical soundscapes. It follows that exposure–effect relationships that were developed ‘‘ignoring’’ neighbour- hood soundscape information, already capture the average of the modifying effect of the soundscapes associated with the apartment and dwelling locations of the sample of Fig. 2. Illustration of the GIS-based method utilized to capture the highest equivalent noise level within 75 m (Lneigh,max). A circle of 75 m is drawn around each apartment in an urban area – left panel and the highest of the Lden,fac¸ade-values for all dwellings or pavement areas within the circular disk assigned as the Lneigh,max value. The result from applying this procedure is shown in the right panel. respondents. It is therefore necessary to correct the impact estimates only for the number of decibels the neighbourhood soundscape is quieter or noisier than ‘‘normal’’ or average for the given noise exposure level (Lden,fac¸ade). 3 To determine the effect size of a deviance from such an average value, it is necessary to elaborate on the results of Klæboe et al. [14]. The set of procedures to achieve the neighbourhood soundscape adjustment to the Lden,fac¸ade-indicator contains the following steps: 1. Statistical analyses for determining how many decibels the noise level in the neighbour- hood usually exceeds the noise level on the facade of an apartment, as a function of the road traffic noise level on the fac¸ade of a dwelling. 2. Statistical analyses of what impact a relatively more or less noisy neighbourhood soundscape than is usual has on residential road traffic noise annoyance. 3. Derivation of adjustment factors for modifying the traditional noise exposure indicator Lden,fac¸ade upward or downward to compensate for a neighbourhood soundscape that is more or less noisy than ‘‘usual’’. 4. Predicting annoyance impact by inputting the neighbourhood adjusted Lden,fac¸ade-val- ues (NALden-values) into dose–response relationships such as those derived by Mie- dema and Oudshoorn [15]. 3 We assume that the average of the impacts from the different neighbourhoods is about the same as the impact of the ‘‘average’’ neighbourhood soundscapes. 2.5. Estimation of average relative neighbourhood soundscape noise levels 626 R. Klæboe et al. / Applied Acoustics 67 (2006) 620–642 To find the ‘‘normal’’ or usual number of decibels that the highest of the equivalent noise levels in the neighbourhood exceeds the residential noise level, it is necessary to esti- mate the mean Ldiff,max-value for each given noise exposure level Lden,fac¸ade-value, that is ð dLdiff;max jLden;fac‚adeÞ.4 Due to our focus on adverse neighbourhood soundscapes and using the highest equivalent noise level within a radius of 75 m to characterize this neighbourhood soundscape, the highest Ldiff,max-values are found at lower Lden,fac¸ade-values. At very high residential noise levels, the neighbourhood noise level is usually about the same as on the fac¸ade. Exploratory studies were therefore undertaken to find out whether a simple linear regression model would be reasonable for describing the relationship between the mean Ldiff,max value ðLdiff ;maxÞ and Lden,fac¸ade. This exploration took form of charting the rela- tionship and applying curve-fitting software. Subsequent to this exploration, a linear regression model was estimated. The estimated mean neighbourhood maximum differ- ence values ð dLdiff ;max jLden;fac‚adeÞ was thereafter tabulated for Lden,fac¸ade-values between 50 and 72 dB. 2.6. Derivation of the neighbourhood maximum deviance Ldev,max After having obtained an estimate of the mean neighbourhood maximum difference the neighbourhood deviation is defined as follows: Ldev;max ¼ Ldiff ;max � ð dLdiff ;max jLden;fac‚adeÞ. The neighbourhood deviation provides information on whether the neighbourhood soundscape is noisier or quieter than ‘‘normal’’ for a given noise exposure level in front of the most exposed fac¸ade. The distributions of Ldiff,max and Ldev,max-values depend on the type of urban area that is sampled. In rural or sparsely populated areas, there may be more apartments in front row situations, while central urban areas may often have more dwellings in a second row situation. The mix can be different in different cities. Two regression lines, one based on the five Norwegian socio-acoustic studies, and one based on the national noise mapping data for Oslo are estimated and the results illustrated by means of a line graph. 2.7. Ordinal logit models for determining the impact of Ldev,max Two sets of regression models were estimated. The first set of models provides the base- lines. They are the usual exposure–effect relationships for road traffic noise annoyance right outside the apartment and indoors as a function of the noise level in front of the most exposed fac¸ade Lden,fac¸ade and ‘‘ignoring’’ neighbourhood soundscape information. The estimation results for the two baseline models have previously been documented [14]. 4 The vertical bar is read as given (the Lden-value). R. Klæboe et al. / Applied Acoustics 67 (2006) 620–642 627 The next set introduces the neighbourhood maximum deviance Ldev,max as an explanatory factor in addition to Lden,fac¸ade. The stepwise procedure makes it easy to test whether sim- pler models only featuring Lden,fac¸ade as exposure indicator are sufficient. Log-likelihood ratio tests were thus applied for determining whether the introduction of the additional vari