Introduction

Noise pollution is an invisible danger, which cannot be seen but present everywhere. Noise pollution refers to unwanted or disturbing sound in the environment, caused by humans and that threaten the health or well-being of humans or animal inhabitants. Continuous exposure to unwanted sounds affects the human health both psychologically and physiologically; some of the affects to mention are hearing impairment, heart diseases, bowel movement, annoyance, tinnitus, hypertension, anti-social behaviour, sleep disturbance, stress, cardiovascular effects, and many more (Sørensen et al. 2011; Kumar 2019; Banerjee 2012; Sahu et al. 2021; Tsaloglidou et al. 2015).

Ambient noise is included as environmental quality parameter in section 5.2.8(IV) of National Environment Policy 2006 (http://www.indiaenvironmentportal.org.in/content/438249/status-of-ambient-noise-level-in-india-2015); therefore, proper monitoring and assessment of ambient noise levels in urban areas are required regularly. Road traffic noises have been reported as the most important source of noise pollution by many researchers (Mocuta 2012; Hamad et al. 2017; Koushki et al. 1993). Around 1.1 billion people between 12 and 35 years of age group are in danger of deafness (https://www.who.int/news-room/fact-sheets/detail/deafness-and-hearingloss). As per survey conducted by the founder of the digital hearing app “Mimi Hearing Technologies GmbH” and analyses of results of hearing test of 200,000 of their users and WHO noise pollution data, “the average city dweller has a hearing loss equivalent to 10–20 years older than their actual age” (https://www.weforum.org/agenda/2017/03/these-are-the-cities-with-the-worst-noise-pollution/) This shows the importance and requirements of traffic noise studies so that mitigation measures can be adopted suitably. Traffic noise measurement is time-consuming, complicated, and unfeasible at the planning and design stage. Ambient traffic noise levels can be determined by measurements or by software simulation. Simulation needs mathematical modelling of environment and traffic conditions. Prediction of traffic noise using mathematical models started somewhere 50 years back, and these models are developed considering variables like traffic speed, traffic variation, road dimensions, and environmental conditions. Urban planners use road traffic noise models to predict the noise emitted in the environment based upon traffic and road characteristics. This is more useful when a new infrastructure has to be developed, that is in a planning stage or for a road already in operation. Prediction models can be used to monitor the noise impact on the surrounding environment by giving input of few traffic and road parameters (An et al. 2013). In design of highways and for assessment of existing roads, traffic noise models are needed as an aid (Golmohammadi et al. 2007). These models are used to forecast noise levels in terms of \({L}_{eq},{L}_{10,}{L}_{90}\) etc. and can be used to plan proper mitigation measure to reduce traffic noise (Ramírez and Domínguez 2013). Many developed countries like USA, UK, and Germany have developed good models (CoRTN, FHWA TNM, RLS-90 etc.) to predict noise levels for homogenous traffic conditions. Now other developing countries are also giving trials to develop suitable noise prediction models for their countries (Lekshmi et al. 2018). Traffic noise prediction modelling trend is varying from basic regression models to genetic algorithm, artificial neural network, convolutional neural network, fuzzy system, graph theory approach etc.

From the literature survey, it is determined that a number of studies have been conducted in the field of traffic noise monitoring and various models to predict traffic noise have been developed. Therefore, in this paper, an attempt has been made to present the quantitative review of the models developed by various researchers for different cities, unfolding the main features and peculiarities of each model.

Methodology

The good research papers were identified by literature search of all databases like SCOPUS, Springer, Web of Science, Academia, Elsevier, and Taylor & Francis using terms “noise monitoring, traffic noise modelling, road traffic noise, transport noise, traffic noise index, noise pollution level, traffic noise monitoring, and traffic noise mapping.” An attempt was made to screen the identified research paper’s titles, abstracts, figures and tables, results, and then full texts, against eligibility criteria. Identified and pertinent papers were deeply analysed to extract information, and database was prepared with different details like author’s details, publication year, study location, variables considered, sampling procedure used, data analysis, specific variables found, modelling equation developed, and observed R2. Where a publication was not open access, it was requested from authors. Reference lists of these papers were also searched for literatures. Only those papers were included for review in which location and duration of sampling were well defined, and which applied developed models for noise prediction.

In total, 37 research papers were studied and summarized in the review and around 21 traffic noise models from different cities were compared.

Traffic noise monitoring

Noise is not only tough on our nerves (Faisal et al. 2008); it is bad for our physical and mental health also (Anees et al. 2017). Exposure to continuous noise levels beyond 85 dB for 8 h or more may be hazardous (WHO 2005). Growth of cities, industries, and infrastructure around the urban environment poses a health risk among urban populations (Debnath et al. 2022). “Traffic noise is the only biggest source of noise pollution and is directly proportional to the volume of vehicles” (Vijay et al. 2015). Prolonged exposure to noise develops into diseases and leads to early death, but it is not easy to identify. Research on traffic noise monitoring conducted in various cities all over the world are summarized in this section to assess traffic noise status of various countries (see Table 1).

Table 1 List of studies on road traffic noise monitoring
Full size table

Mavrin et al. (2018) assessed the impact of the noise level of road traffic on the state of the environment. The results of the investigation indicate that the measured noise levels are exceeding the maximum acceptable level (55DBA). Nury et al. (2012) performed a study to obtain traffic noise index (TNI), equivalent noise level, noise climate (NC), and hazard due to noise pollution at Sylhets’s eight major intersections. From the analysis, it was found that average noise level was approximately 74 dBA exceeding the acceptable 45 dBA limit set by Department of Environment. Dulal (2008) assessed highway traffic noise pollution and its effect in and around Agartala, India. The noise level in various locations is much higher than the standard limit prescribed in Indian Standard codes for residential area. Noise perception study indicates that 62.5% is affected by the highway noise both physiologically and psychologically. Masum et al. (2021) conducted spatiotemporal monitoring of noise levels in Chattogram city in Bangladesh. Based on land use, 123 data monitoring points selected in 41 wards of Chattogram city corporation. It was found that population experienced high noise level surpassing the values set by DOE, Bangladesh, for different land uses pattern. Results indicate that out of the 41 wards, only 3 were within the acceptable condition. McAlexander et al. (2015) measured street level noise at 99 sites located in New York City and revealed the variation of 55.8 to 95 dBA. Mishra et al. (2019) performed traffic noise analysis at 10 locations in Delhi based on land use pattern and found that the noise level at all 10 locations were above the acceptable limits set by central pollution control board.

A noise monitoring study was conducted by Swain et al. (2012) at Bhubaneshwar city; results show that the minimum value of Leq 70.4 dBA is also above the permissible limit of 70 dBA. Alam et al. (2020) analysed and evaluated traffic noise levels 07 residential areas, 03 commercial areas, 04 Industrial areas, and 02 silent zones of Delhi and revealed that Delhi is exposed to high noise levels of 60–80 dBA. Spatial and temporal variabilities of noise levels of Toronto were explored by Zuo et al. (2014), and it was concluded that 80% of sites were having noise levels higher that the permissible limit of 55 dBA. Chebil et al. (2019) carried out case study of traffic noise levels at four main roads of Monastir-Tunisia and concluded that the noise levels observed are greater than the limits of Tunisian environmental standards and the WHO standards. Goussous et al. (2014) monitored noise levels in 18 selected sites of Amman, Jordon, and revealed that the average noise level was 70 dBA which is far more than the environmental standard limit of 55 dBA. Chandio et al. (2010) revealed that traffic noise levels in Larkana city exceeds the limit of 85 dBA given in National Environmental Quality standards of Pakistan. Kupolati et al. (2010) carried out traffic noise measurement at 10 locations in Ibadan and fount traffic noise levels between 53.8 to 65.2 dBA which is more than the permissible WHO standards of 50–55 dBA. Gholami et al. (2012) analysed spatial traffic noise characteristics at 41 stations in Tehran City, Iran, in residential, medical, educational, commercial-residential, and commercial use areas. Authors concluded that average noise levels were higher than the Department of Environment standards for different land uses. The amount of violation was 14.14 dB in residential and 11.11 dB in educational areas. Chowdhury et al. (2010) conducted noise monitoring in Dhaka City, and results indicate Leq noise level of 82 dBA.

Laxmi et al. (2019) used cycle mounted sound level meter (an innovative method) to monitor noise levels in Nagpur city, India. In total, 700 monitoring stations were used and found that the Lmin values at all stations are exceeding the WHO guidelines for community noise.

Review of literature indicates that road traffic noise levels were found beyond the acceptable limits in almost all the studies; therefore, road traffic noise is a matter of concern and requires an urgent action to control the alarming levels of road traffic noise.

Traffic noise prediction modelling in developing countries

Development of new roads, investment in major highway projects, and construction of tunnels are essential for developing countries and communities. But this development leads to increase in flow of traffic and causes traffic noise that have negative impact on buildings and peoples. The impacts of road traffic on local environment must be taken into consideration by urban planning and road design. For controlling traffic noise pollution in urban areas, traffic noise prediction is required. In literature survey, it was found that many works are carried for development of a predictive traffic noise model. Review of recently developed models for various cities has been presented in this section.

Delany et al. (1976) developed CoRTN model for the department of environmental engineering, UK. This model predicts noise levels in terms of L10(A) which can be converted to Leq(A). Barry and Reagon (1978) introduced Federal Highway Administration (FHWA) method to predict traffic noise. This model is based on Leq, and an adjustment for conversion to L10 is provided in the model. Tandel and Macwan (2013) carried out the study to generate a traffic noise model for main Arterial roads of Surat, India, and to analyse various parameters affecting road traffic noise. Total 03 arterials roads were selected for study based on mix traffic flow and different land use pattern. In total, 96 data points/sampling sites were selected, 32 on each corridor (16 on each side). Measurements were carried out during peak hours (5:00 to 8:00 p.m.). Multiple linear regression analysis was performed on the combined effect of PCU, open spaces, and building height and model indicated good relation of the three parameters on noise. Kamineni et al. (2019) developed a comprehensive noise prediction model for eight important highways of Andhra Pradesh and Telangana, India. Measurements were done on each highway from 10:00 a.m. to 5:00 p.m. at an interval of 15 min using far field methodology. Scattered plots for Leq, L10 v/s traffic volume, spot speed, and carriageway width were plotted for 08 highways. The 15-min time frame models resulted in a negative correlation compared to the hourly time frame model. Konbattulwar et al. (2016) designed in vehicle noise prediction models for Mumbai Metropolitan Region, India. Data was collected by covering total road length of 403.80 km by total 22 trips conducted on 06 different routes using different types of vehicles (AC car, non-AC car, Auto, Bus). Separate model for each type of vehicle and for each type of road was developed. Awwal et al. (2021) assessed the road side noise levels on asphalt pavements and concrete pavements. For this, Skudai-Pontian Highway having road stretches with different pavement types was selected. Noise levels were measured for three weekdays in the peak hours (5:00 to 6:00 p.m.) and off-peak hours (10:00 to 11:00 a.m.) using statistical pass by method. Separate models were developed for concrete and asphalt pavement for peak hours and off-peak hours. Suthanaya (2015) modelled traffic noise for collector roads of Denpasar City, Indonesia. Tumku Umar Road was selected for measurement from 6:00 a.m. to 6:00 p.m. (12 h); in total, 48 data sets were collected at 15-min interval. Traffic volume was classified into MC, LV, and HV. It was observed; if all other factors are kept constant, then an increase of 100 motor cycle increases traffic noise LAeq by about 0.3 dB, and increases in the values of LA10, LA50, and LA90 are 0.4, 0.4, and 06, respectively. Gharibi et al. (2016) evaluated and modelled noise from traffic on the Asian Highway in Golestan National Park, Iran. For measurement of noise and independent variables, 76 sampling stations were selected at various distances (0 to 250 m) from the road using systematic random method. Sampling for 1 week from 8:00 a.m. to 8:00 p.m. at 15-min time interval is done at each sampling station. Leq-based modelling as dependent variants and 19 independent variables were performed using SPSS software.

Ranpise et al. (2021a, b) carried out research work to develop traffic noise model for main urban roads of residential and commercial areas of Surat, India. After proper execution of pilot survey, 03 roads were chosen out of which 02 were of rigid pavement and 01 of flexible pavement. Measurements of sound level were done for 16 h from 6:00 a.m. to 10:00 p.m. on each road. Three different models for all three roads were built, and subsequently, the last model was developed using data of all three roads. Ranpise et al. (2021a, b) measured ambient noise levels at major arterial roads of Surat, India, and compared them with prescribed standards, and developed a traffic noise model for arterial roads of Surat using an artificial neural network. Three arterial roads selected for the detailed survey were Athwa-Dumas Road, Adajan-Rander Road, and Udhna-Sachin Road. Continuous monitoring for 24 h from 9:00 a.m. morning to 9:00 a.m. next morning was done on all three roads.

Monazzam et al. (2014) designed a traffic noise forecasting model for highways of Ahvaz city, Iran. A total of 1344 observations were recorded at 112 stations selected on 07 roads of the city. Observations were made for 4 weekdays, three times a day. Out of 15 independent variables considered, only 9 variables were used in development of model. Golmohammadi et al. (2007) developed road traffic model for Iranian Cities; in total, 282 data sets were considered, and measurements were carried out between 7 a.m to 10 p.m. and 10 p.m. to 7 a.m. Four explanatory factors involving twelve variables were used for regression analysis, which indicated high R2 = 0913. Shalini and Kumar (2018) measured road traffic noise at 7 different locations in Varanasi, and total 14 sets of data were collected. Linear regression analysis using SPSS was performed, and model equation was developed considering traffic volume, noise climate, noise range, weightage of traffic volume, and % of heavy vehicles as independent variables. Garg et al. (2014) conducted traffic noise survey at different sites in Delhi, and four different models were developed for Leq, L10, TNI (traffic noise index), and NPL (noise pollution levels) using equivalent vehicle speed and equivalent traffic flow as independent variables. Ramakrishna et al. (2021) developed MLR and ANN models for predicting traffic noise levels in residential, commercial, industrial, and silent zones of Vijayawada, Andhra Pradesh. Four sampling locations one in each zone were selected, and data was collected for 3 days at each site, four times a day. Sooriyaarachchi and Sonnadara (2006) developed traffic noise prediction model for 08 different classes of vehicles (motorcycles, three-wheeler, car, van, double cab, Jeep, bus, and Lorry) in Srilanka considering distance from centre line (2.5 m, 5.0 m, 7.5 m, 10.0 m, 12.5 m, 15.0 m). A total 650 data sets were collected for 8 different classes of vehicles.

Kumar (2015) used genetic algorithm and regression approach to predict noise levels for Patiala city, India, using vehicle volume and percentage of heavy vehicle as variables. Mean square error of GA models is in the range of 0.5558–0.6123, while regression model shows error from 0.7575 to 0.7623. The author concluded that GA model performs much better than regression model. Cirianni and Leonardi (2012) measured noise levels at 14 sites (total 154 records) in city of villa s Giovann, Italy, and recalibrated the three regression models (Burgess (1977), CoRTN model (2011), and García and Bernal (1985)) with genetic algorithm. It was observed that GRNN (general regression neural network) is well suited for simulation of phenomenon and can be used for more complex areas and greater traffic variability. Gilani and Mir (2021) used graph theory approach for predicting traffic noise using five parameters (traffic volume, volume of heavy vehicles, traffic speed, honking, and pavement width). Data for selected variables was collected for 3 months, and noise parameters Leq, L10, and L90 were included in the study. Variables considered were assigned weightage from 1 to 5 and were incorporated into a matrix, weightage for variable interaction also decided based on human knowledge, and permanent function matrix was formed to calculate permanent noise index. Model was developed using PNI and noise parameters. Patthanaissaranukool et al. (2019) predicted noise levels for Phuket province, Thailand, using NMTHAI1.2 model, and study revealed that model is overestimating the traffic noise contribution. Lekshmi et al. (2018) developed an artificial neural network model and regression model to predict traffic noise on NH66, Kochi, India. Six sites with 500-m interval were selected, and measurements on each site for 6 h/day for 6 days were carried out. Traffic flow, speed, and percentage heavy vehicles were used as input variables in both the models. Comparison of both models indicates ANN model is more reliable for traffic noise prediction.

A comparison of different features of models developed for various cities is shown in Table 2.

Table 2 Comparison of developed models
Full size table

Figure 1 displays a timeline plot that indicates specific years when each disruptive model was introduced based on available data of R2 and MSE observed.

Fig. 1
figure 1

Timeline plot indicating specific years when each disruptive model was introduced and R2/MSE observed

Full size image

Conclusion

In this work, literature related to traffic noise monitoring and predictive modelling of traffic noise was studied. Based on the review of the literature, it is concluded that around 82% of noise monitoring studies are focused on traffic noise near roadways, and only 18% were related to traffic noise in different zones (residential, industrial, commercial, and silent zones). Noise monitoring studies are reported to have been carried out on different days and times, but effect of different seasons has not been considered. Mostly, the acoustic energy descriptor used is equivalent sound level Leq; only in some cases, the percentile levels L10 or L50 are used.

Most of the models have been developed considering average speed, percentage of vehicle, traffic volume, and road dimensions. Undoubtedly traffic noise also depends on pavement type, vegetation along roads, barriers, road surface (roughness), gradient effect, wind speed, honking of horns, reflective surface etc.; therefore, considering these factors can give a more comprehensive model. Most of the noise prediction models worldwide have been built using regression modelling; therefore, an attempt to develop traffic noise prediction model using evolutionary computing methods like genetic algorithm, fuzzy systems, and neural networks and comparison of their results with traditional regression models can bring forth certain interesting results. Further models can be developed based on studies conducted in all four seasons, all days of week, for different conditions (like dry and wet surface) of road, and these studies can be further extended to measuring in vehicle noise levels on same roads to compare the noise levels tolerated by residents, road users, and the commuters; this will help in formulating traffic noise regulations.

From the study, it can be concluded that for algorithm-based modelling, large datasets are required to get the benefit of generalization and nonlinear mapping, whereas linear regression models need least data points, more than the number of variables, which can be small set. Algorithm-based models predict better but do not quantify the effects of various factors contributing to noise, whereas basic regression models have lesser prediction accuracy but are able to quantify the effects of a factor. Therefore, algorithm-based models are more suitable for application like estimation of cost related to noise pollution and regression models can be used in planning stage where effects need to be studied.

Noise mitigation measures suggested by researchers in different studies can be broadly categories into traffic control and management, technological solutions, and road design measures. Use of intelligent transport system (ITS) in transportation planning can help to control mobility, traffic volume, vehicle speed, composition etc. Technological solution includes innovative studies like utilization of sonic crystals in construction of noise barriers, poroelastic road surfaces (PERS), and application of active noise control (ANC). Introduction of roundabouts, chicanes, dense vegetation, green area etc. are the road design measures.