The Appendix 1 provides information from all papers compiled in the SLR.
Databases And Journals
Upon completing the fourth stage of the systematic review, 63 papers were selected, the majority (75%) from the Scopus database (n = 576) and 25% from Science Direct (n = 191). Among the selected papers, it was observed that 17% of the papers were published in the journals Applied Acoustics (n = 11), 11% (n = 7) in Noise Mapping, 6% (n = 4) in Sustainability (Switzerland) and Transportation Research Part D: Transport and Environment, 5% (n = 3) in the IOP Conference Series: Earth and Environmental Science and Science of The Total Environment, 3% (n = 2) in Civil and Environmental Engineering, Environment International, Environmental Research, Environmental Science and Pollution Research and ISPRS Archives. The remaining, 2% in each journal (n = 1), were from the journals: Building and Environment, Cartographic Journal, Case Studies on Transport Policy, Complexity, Energies, Environment, Development and Sustainability, Environmental Challenges, Environmental Pollution, Global Health Journal, IEEE Access, International Journal of Environmental Research and Public Health, International Journal of Sustainable Development and Planning, Journal of Environmental Health Science and Engineering, Journal of Environmental Management, Journal of Environmental Planning and Management, Journal of Urban and Environmental Engineering, Land Use Policy, PLoS ONE, Proceedings of Meetings on Acoustics and Sustainable Cities and Society.
The results indicate the wide application of noise maps in the most diverse fields of science. There were publications in journals (n = 57) and also conference papers (n = 6). Thus, even without a local noise prediction model, research and elaboration of maps were encouraged in publications.
One of the exclusion criteria was to remove studies in countries that already had a local TNM. Most of the articles excluded in the extraction phase (n = 85) were from studies that were carried out in places that already had an TNM, e.g., European countries, Japan and the USA. After this exclusion phase, 63 papers remained.
Among the papers collected (n = 63), it was identified that the countries that performed the most studies with noise maps were China (n = 12), India (n = 9), Brazil (n = 7), Turkey (n = 5), Ecuador (n = 3), Malaysia (n = 3), Singapore (n = 3), Canada (n = 2), Colombia (n = 2), Iran (n = 2), South Korea (n = 2) Taiwan (n = 2), Australia, Bangladesh, Chile, Czech Republic, Ghana, Indonesia, Japan, Jordan, Korea, Nigeria, Slovakia and Thailand, (each with 1 study) (see Fig. 2). Also, there was one study that was performed at a generic place so it could be replicated in any country, including those that do not have a local TNM (Forssén et al. 2022).
One of the reasons why there have been few studies that include traffic noise maps is the difficulty of finding a noise prediction model that represents the local reality (Pozzer et al. 2018; Bakowski et al. 2019). In addition, there is the difficulty of making noise maps in places where there is no normative requirement, such as the European Noise Directive (END) (The European Parliament and the Council of the European Union 2002). It is also difficult to obtain data in underdeveloped countries where there is no continuous noise monitoring data, vehicle count data or even local topography (Bocher et al. 2019; Alam et al. 2020; Raess et al. 2021; Adulaimi et al. 2021).
Prediction Model Used As References In Studies
Most of papers compiled in the first stage of the selection were from research in places that already had a local TNM because these countries encourage and, even, require the elaboration of noise maps. It can be seen that without a local model, it is necessary to choose the model that best fits the region. Furthermore, the TNM may often require a considerable amount of input data in the model and, depending on the location, such data are scarce because the country is underdeveloped or there is no or little available geoprocessing data, few noise monitoring stations or a lack of vehicle counting technology, which can make the elaboration of noise maps difficult or even impossible.
The most used noise prediction models, which can be adapted to the local reality, were: RLS-90 (n = 9), NMPB (n = 9), FHWA (n = 7), ISO 9613 (n = 6), CoRTN (n = 4), CNOSSOS-EU (n = 3), Nort Test Method (n = 2), ISO 1996 (n = 2) and STL-86 (n = 1). It is noteworthy that the Chinese studies (n = 3) that used the emission model proposed in The Technical Guidelines for Environmental Impact Assessment (HJ 2.4–2009) based on Specifications for Environmental Impact Assessment of Highways (JTG B03-2006), which has a series of correction parameters based on the FHWA model (Zou et al. 2019), were accounted for in studies that used the FHWA model.
Several of selected papers used the interpolation method, such as the Empirical Bayesian Kriging (EBK), used for Inversion of the distribution of traffic noise, or the inverse distance weighting (IDW) technique (n = 14) that is based on GIS modelling. Land use regression (LUR) modelling was also used (n = 3). In order to develop a traffic noise prediction model that represents the local reality, 8 studies proposed their own model.
Thus, Fig. 3 presents the noise prediction models used as references in the selected studies. Papers in red are from standard TNM.
In most studies (n = 36) the authors adapted a prediction model to the local reality, while 25 papers used some interpolation method, IDW technique or even proposed a model itself. Two studies did not mention the TNM used in the noise map (Chan et al. 2019; Raess et al. 2022).
Among the noise prediction models that were most used in the studies (RLS-90, NMPB, FHWA, respectively), it is important to observe the input data in the programs. For the NMPB, speed and traffic flow type of each vehicle category, road platform surface category, road gradient are entered (Besnard and LeDuc 2009; Garg and Maji 2014). For the RLS-90 the following are required: data on type of traffic, speed, topography information, number of light and heavy vehicles, vehicle flow, road surface data and, also, parking information when available (der BUNDESMINISTER FÜR VERKEHR 1990; Garg and Maji 2014). For the FHWA, data on the type of traffic, vehicle flow, speed, road surface data are entered (Barry and Reagan 1978; Garg and Maji 2014). An important fact is that, although the RLS-90 was updated in 2019 to the RLS-19, no studies published from that year, used the current model (DER BUNDESMINISTER FÜR VERKEHR 2019).
One reason why these TNM are the most used may be due to the smaller amount of input data required in the models. An example of a TNM requiring a large amount of specific data is the CNOSSOS-EU model (Kephalopoulos, Stylianos Paviotti and Anfosso-Lédée 2012), which is more complex. In this model there are more categories of vehicles to be accounted for, which can be a hindrance when collecting data in countries that do not require the obligatory elaboration of noise maps. Furthermore, maps that use interpolation methods or the IDW technique exhibited a lower level of graphic detail in the noise maps.
Among the papers selected, there were studies that analyzed more than one TNM or in more than one place (Sonaviya and Tandel 2019, 2020; Bravo-Moncayo et al. 2019a; Wosniacki and Zannin 2021; Liu et al. 2021; Gozalo and Escobar 2021). Appendix 2 presents the TNM used in each country.
Most of the papers compiled were from China (n = 13), consequently, this was the place where the most different noise prediction models were used. Among them, FHWA was the most used TNM and there were also the same number of papers that used noise prediction models proposed by the authors themselves. India was the country that most used different TNMs: CoRTN (n = 4), ISO 9613, RSL-90, EBK interpolation method (n = 2, each) and ISO 1996-1/2 (n = 1).
It was observed that each country used different TNMs or proposed their own model. From the SLR, it was found that the models can be adapted or created for the local reality. However, for the purposes of monitoring sound quality and also for noise prediction, it is important to have a country standard to standardize the modeling of noise maps.
Software Used In Studies For The Elaboration Of Noise Maps
In order to identify which software was most used in studies that did not have a local TNM, an analysis was performed on the compiled works. Murphy and King (2010) emphasize the importance of the competent authorities to specify the spatial interpolation method to be used in the states for the compilation of strategic noise maps. The results of papers compiled showed that the most used software was SoundPLAN (n = 14), followed by ArcGIS (n = 13), CadnaA (n = 9), GIS (n = 7), Predictor LimA (n = 5), q-GIS (n = 3), E-map (n = 2), NoiseModelling (n = 2), DISIA and SurferTM (together), ERDAS Imagine, MATLAB, NoiseExplorer app and NoiseSystem (n = 1, each). Two studies did not mention the software used in the research. The software used to create the noise maps can be seen in Fig. 4.
Most studies used specific software for the elaboration of noise maps such as SoundPLAN, Cadna-A, Predictor Lima, NoiseModelling, DISIA, NoiseExplorer or NoiseSystem. In addition to the interface aimed at making the maps, some software allows a choice of the TNM to be used and also allow the mapping of noise in 3 dimensions, which provides the level of noise exposure also in the Z direction (e.g., facades). Used in residential areas, this option can assess the number of people affected by noise in high buildings (Zhao et al. 2017).
Several studies used the geographic information system (GIS), either in commercial programs such as ArcGIS or EDAS Imagine, or in free software with open source code such as q-GIS. The advantage of using this for the elaboration of noise maps is in the customization of the source code and in the choice of the way to interpolate the data.
Most of the papers that were based on GIS modeling did not mention sound maps in 3D. In some models that use commercial software, it is possible to create 3D maps. This is the case in the study by Alam et al. (2020) which also identified that most researchers used GIS as a tool for the development of a 2D noise map. Thus, although there is software that also performs an analysis in the Z direction, 2D maps have been used extensively in environmental impact studies such as air pollution, soil pollution and noise in the existing environment (Alam et al. 2020).
Measurement Time
The duration and the measurement times were analyzed in this research. Given that there is no local regulation in these countries, it is important to analyze the period of the day and the duration of the measurement for the vehicular traffic noise data collected for the realization of maps.
Thus, most studies performed measurements from 7:00–17:00 h for daytime, 18:00–20:00 h for evening and 23:00–5:00 h for night time. The highest incidence of times cited by papers that performed measurements at peak times was from 7:00–9:00 h; 10:00–12:00 h; 16:00–18:00 h, for daytime. 18:00–19:00 h for evening and; 23:00–00:00 h for night time.
When analyzing the measurement time, most studies performed measurements during 15 min (n = 11) and 10 min (n = 10) (Fig. 5). Among measurements that took less than 1 hour, the average measurement time was 17.82 min.
Various parameters can be adopted when creating a noise map. For purposes of monitoring noise over time, it is important that the parameters adopted in the modeling are the same. Among them is grid spacing (or mesh size) and the size of the rectangles will determine the accuracy of the map. As the size of the grid is reduced, in order to reduce the interpolation errors, the computational processing time increases. Thus, optimizing mesh spacing involves trying to balance accuracy in sound mapping and the processing time (The European Parliament and the Council of the European Union 2002). Figure 6 (a) presents the different grid sizes adopted in the papers.
Parameters (Heights And Grid Sizes) Used In Noise Maps
Among the papers that cited the grid spacing adopted (n = 27), most adopted the size of 10 x 10 m (n = 11), followed by 5 x 5 m (n = 4), 20 x 20 m (n = 3), 30 x 30 m (n = 2) and 200 x 200 m (n = 3). The grid sizes of 1 x 1 m, 3 x 3 m, 8 x 8 m, 125 x 125 m and 140 x 140 m were used once in each paper.
Another parameter used is the height of the noise map, which is often prepared at the same height in relation to the SPL meter for calibration purposes. The European Parliament and the Council of the European Union (2002) suggests for strategic noise mapping purposes, assessment points should be 4.0 ± 0.2 m (3.8 to 4.2 m) above ground. Regarding noise exposure in and around buildings, other heights can be chosen, but they must never be less than 1.5 m above the ground, and the results must be corrected for an equivalent height of 4.0 m (The European Parliament and the Council of the European Union 2002).
For countries that do not have a TNM standard, the papers that mentioned the height from the ground level (n = 39 of 63), most used 1.5 m (n = 13), 1.2 m (n = 9), 4.0 m (n = 7), 2.5–3.0 m (n = 3), 4.5 m (n = 2) and 2.0 m (n = 1) (see Fig. 6 (b)). It is important to note that heights of 1.5 m and 1.2 m are commonly used because of the height of the SPL meter mounting tripod.
In addition, there were 2 papers that performed measurements at a height of 1.2 m from the ground level and adjusted the grid to a height of 4.0 m (Cai et al. 2019; Paiva et al. 2019). There were 2 papers that only reported the grid height of 4.0 m (Bravo-Moncayo et al. 2019b; Beran et al. 2021) and the work of Beran et al. (2021) tested the heights of 4, 8, 16 and 20 m above the terrain.
In summary, for the purposes of modeling the noise maps, the grid size of 10 x 10 m and the height of the SPL meter of 1.5 m were the most adopted parameters.
Future Perspectives In This Field
As mentioned, noise maps are a diagnostic tool used to pursue control strategies, inform the public and make stakeholders aware of noise related issues. Therefore, the importance of creating noise maps has been growing over time (Yang et al. 2020a; Khomenko et al. 2022). In addition, with the growth of cities and consequently the population and the number of vehicles, traffic noise has become a pollutant that must be constantly monitored.
It is important to highlight that from the year 2020 onwards, there was the impact of the pandemic (COVID-19), which became one of the biggest global health crises in recent decades and influenced the behavior of many people. During this period, many inhabitants changed their habits, especially regarding use of public transport, preferring to travel in private vehicles (Arimura et al. 2020; Benita 2021). Thus, there has been an increase in the number of studies regarding changes in vehicular traffic and how this is impacting noise even after the pandemic (Arimura et al. 2020; Benita 2021; Xu et al. 2022).
Figure 7 shows the number of papers compiled by year of publication. It is important to note that this SLR was carried out until mid-2022 (May 30th) and, despite this, the number of compiled papers has already surpassed the publications of 2018 and, practically, equaled the number of papers of 2020.