Verification of railways noise mapping using CNOSSOS-EU: case study on freight trains Álvaro Grilo 1 INERCO Acústica S.L. Tomas Alba Edison, 2 (Isla de la Cartuja), Seville, Spain. Javier E. Mitjavila 2 INERCO Acústica S.L. Tomas Alba Edison, 2 (Isla de la Cartuja), Seville, Spain.

ABSTRACT Freight trains are a relevant source of environmental noise and require an acoustical assessment to quantify the exposure of the population using noise mapping. Once the results of the noise calculations have been developed, it might be necessary to adopt mitigation strategies that can be validated using these noise models. It is becoming increasingly common to undertake some form of validation exercise to cross-reference the calculated levels with measurements to reduce the uncertainty in the action planning. Based upon the results of the monitoring, the measured data may then be stratified regarding the source data and meteorological data and introduced into the calculation model, to enable the model to replicate the situation during the measurement windows (meta-models). The results from the calculations of each of these meta-models may then be compared with the noise measurement results as the basis for the validation study. This paper presents an overview of a verification project where such a process was followed using the CNOSSOS-EU as the basis of comparison.

1. INTRODUCTION

In the past decades, the adverse impacts on human health related to noise have become a major worldwide concern, to both administration and the general public. The European Environment Agency (EEA) estimated [1] that railways are the second most dominant source of environmental noise in Europe, with nearly 22 million people affected and, a greater proportion of people exposed to noise levels higher or equal than 65 dB during the night-time. The European Commission also pointed out that the rolling noise of freight wagons is the main source of railway noise [2] [3]. In addition, rail freight traffic is expected to increase by more than 50 % by 2030, compared with 2010 levels [4] with the highest growth among the inland freight transport modes and increases its modal share by 3% in 2050 [5].

In terms of noise assessment and control regulations in Europe, the Environmental Noise Directive 2002/49/EC [6] requires all EU countries to develop strategic noise maps and action plans for areas where major roads, railways, airports, and agglomerations are present. Action plan implementation implies an investment to assure that the noise levels comply with national and international noise

1 agrilo@inerco.com 2 jemitjavila@inerco.com

guidelines/regulations. Thus, it is a common practice to check the efficacy of the proposed action plan through the implementation of the selected mitigation measures into the noise modelling, developed according to the Common NOise aSSessment methOdS (CNOSSOS‐EU) [7]. The verification process of these noise models through noise monitoring at select sites shows is becoming a wide solution for boosting the public confidence in these maps and contributing to further evaluation of the efficacy of the action planning [8].

This paper aims to describe the verification methodology of a CNOSSOS-EU noise model through noise monitoring for a freight network in Latin America. The verified noise model process will serve later as the main supporting point for a further proposal of feasible and cost-effective noise mitigation measures.

2. DESCRIPTION AND ASSESSMENT OF RAILWAYS NOISE

In Harmonoise WP1.2 [9], an individualization of the noise sources that compose the term “railway noise”, alongside the behaviour of said components, is performed. The report states that railway noise could be divided in traction noise (motor noise and auxiliary noise), rolling noise (interaction between wheel and rail), and aerodynamic noise.

Figure 1. Sound pressure level as a function of train speed. Extracted from [9]

As presented, in Figure 1, the total railway noise is dominated by traction noise at low speeds (below 50 km/h). On the other hand, the rolling noise dominates total noise level between 50 km/h, and 300 km/h and aerodynamic noise becomes predominant around 300 km/h. The average speed of freight trains goes between 50 and 80 km/h and thus, traction noise and rolling noise are the two major noise components to take into account when assessing freight train noise. However, it also states the existence of other components of noise, due to specific operating conditions such as during bridge passing, curve passing, rail joint passing, and braking [9]. A brief description of braking noise and acceleration from standstill noise is presented:

• The braking noise is especially annoying for both passengers and residents if squeal noise and the banging of wagons are present. The squeal noise spectrum is most of the time composed of one or several pure frequencies [10]. • The acceleration from standstill noise is characterized by an increase in the noise coming from the motor (due to increasing revolutions to make the train start moving) but, unlike during pass- by, the train moves slower (and so the noise event last longer).

As an example, the evolution of sound pressure level reported in a monitoring station near the studied railway track is shown in Figure 2.

‘Sound pressure level dB(A) ‘Train speed [knv/h)

Figure 2. Sound pressure level evolution of freight train pass-by during the braking procedure (left)

and acceleration from standstill procedure (right).

Railway noise is a complex problem where different noise sources, different physical generation mechanisms and behaviour interactions. In accordance with the latter, the key to developing an effective action plan is to follow an assessment methodology that considers all noise sources (and their particularities) presented earlier. The common practice, to assess railway noise impact and validate action plan results, is to perform a noise model using specific calculation methods for railway noise. In Europe, the Common NOise aSSessment methOdS were developed by the European Commission, to be used by the EU Member States (as calculation methods) for strategic noise mapping following adoption as specified in the Environmental Noise Directive 2002/49/EC [6].

The descriptors, in CNOSSOS-EU model, for railway vehicles include the vehicle type, number of axles per vehicle, brake type, and wheel measure. On the other hand, descriptors for track type include the track base, railhead roughness, rail pad type, additional measures, rail joints, and curvature. With this information, the equivalent sound sources related to railway noise are calculated: • Rolling noise, as equivalent noise line source allocated to h=0.5 m. • Traction noise, as equivalent noise line source allocated between h=0.5 m and 4.0 m, depending on the vehicle type. • Aerodynamic noise effects, as two equivalent noise line sources allocated at h=0.5 m and 4.0 m. • Impact noise is associated with the source at 0.5 m. • Squeal noise is associated with the sources at 0.5 m. • Bridge noise is associated with the source at 0.5 m.

& 3B i 3 g (wap) janay aanssaug punos.

The calculation method is only valid for determining noise in the frequency range from 63 Hz to 8 kHz (railway noise). In general, there shall be no reliance on default input values or assumptions unless the cost of real data collection is highly disproportionated. As consequence, in the application of the method, the input data shall reflect the reality, leading to the development of custom values using other methods that accurately represent the situation at hand [11] [12] [13]. In fact, within the quality framework of the Directive 2015/996/EC, the input values affecting the emission level of railways shall be determined with at least the accuracy corresponding to an uncertainty of ± 2dB(A) in the emission level of the source (leaving all other parameters unchanged). In addition, the guidance for CNOSSOS‐EU implementation [14] considers “essential”, the validation of said strategic noise maps using measurement data. 3. METHODOLOGY OF VERIFICATION PROCESS

The strategic noise map verification process, proposed in this paper, considers a combined methodology to validate the noise model (CNOSSOS-EU method) using specific measurement methods for railway noise. The methodology framework is based on the fact that the results of the acoustic tests can be used to obtain an idea of the degree of trust (confidence level) of the noise levels

(yap) jana7 ainssaug punos tues Tues wists Tr9e'6 wiseis Testis wives Teves wets Trees twee Tres ties Ties woe'e Toes war's Tests weve Travis

calculated through acoustic models when the comparison between noise level measurements and calculated strategic noise mapping results is assessed in terms of uncertainty.

The methodology used a noise measurement campaign based on real-time sound level monitoring to capture sound levels adjacent to railways sources, and noise calculations for each of the measured situations as a means of undertaking verification of noise models previously developed. The outputs of the measurement campaign and noise modelling were compared within the verification procedure using statistical analysis alongside an assessment of uncertainty. The workflow of the overall verification process is summarised in Figure 4 below.

Figure 3. Verification process schematic for case study

The case study will be developed on a single-track line with different passing sidings (where braking, acceleration from standstill, and acceleration pass-by operation are present). Regarding equipment, the freight train configuration is comprised by two (2) diesel-electric locomotive (AC- traction) and 150 wagons that operate on day-time and night-time. The structure of the railways is a ballast platform with wooden sleepers (boron treated) of conventional width and a maximum curvature radius of 7000 m. The trains pass-by occur at 25 miles/hour (during a train meet at siding or at low-speed sections of railway) and 40 miles/hour, and alternate between complete cargo and empty cargo. As mentioned before, during a train meet, operations such as full brake, acceleration from standstill, acceleration pass-by. The studied scenarios are described in Table 1. Table 1: Description of scenarios and scope of noise campaigns.

Scenario Measurement

Total observation interval (approx.)

Number of

Positions

samples

Full brake. 10 15 days 23

Acceleration 10 15 days 18

Constant speed at 40

miles/hour. 30 21 days 340

Constant speed at 25 miles/hour 5 5 days 54

Constant speed at 40 miles/hour. (ISO 1996-2) 5 3 days 195

=] 7

3.1. Noise campaign

Noise levels were measured using type 1 sound level meters with an outdoor microphone kit, placed at a height of 4m above the underlying terrain, following: • ISO 1996-2 Acoustics - Description, measurement, and assessment of environmental noise - Part 2: Determination of environmental noise levels: The main objective of this campaign was to obtain the real contribution of the specific noise sources under investigation (railways) whilst attempting to minimize the influence of extraneous ambient noise. The measurement campaign undertaken included a simultaneous collection of noise level data, noise source operational data (audio-video recordings and speed gun), and meteorological conditions. The noise level measurements and an uncertainty assessment were undertaken following ISO 1996-2 at each of the selected locations (sensitive uses), in the surroundings of the platform (30-125 meters). • ISO 3095 ¨Acoustics — Railway applications - Measurement of noise emitted by rail-bound vehicles¨: ISO 3095 specifies measurement methods and conditions to obtain reproducible and comparable exterior noise emission levels and spectra for all kinds of vehicles operating on rails or other types of fixed tracks. This methodology was used to characterize the pass-by under different operating conditions (previously described scenarios).

Figure 4. Test disposition: ISO 3095 Braking operation (up-left), ISO 3095 Acceleration from standstill (up-right), ISO 3095 constant speed in curve and straight track (down-left), and ISO 1996-

2 long term monitoring (down-right). * Symmetrical sound level meter position on the other side of

the track.

The data processing activities for long-term monitoring include the correlation of noise events with operational procedures (pass-by, brake, start-up), as well as the removal of intervals with extraneous results, due to residual sound or meteorological conditions (high-speed wind or rain). After the selection of the valid data (under favourable meteorological conditions according to ISO1996-2), the partial contribution of railway noise to the measured environmental noise was calculated from individual event identification and calculation of Sound Exposure level, SEL, and L Aeqpass-by of each operation, according to ISO 3095. The extrapolation to day (L d ) and night periods (L n ) was developed considering the real traffic flow existing during the measurement interval and partial contribution in line with ISO1996-2 for long-term monitoring.

3.2. Noise mapping

Regarding noise modelling, one meta-model per scenario was developed using the software CadnaA. These meta-models shall include all the necessary details that allow evaluating the noise emission (number of trains) and propagation present during the measurement period (favourable conditions, incident noise, and distance to obstacles). Thus, to guarantee the similarity between the models and reality, the following specifications were followed:

• The three-dimensional model, where the source and receiver are located, should be developed considering reliable (accurate) geo/topographic information. For the case study, an extensive LiDAR campaign was developed as part of the project, reaching high-degree accuracy results (0.5 m steps vertically with height uncertainty of ±0.1 m). • In addition to the latter, the location of each receiver within the noise model was confirmed to be at the same location as the sound level meter, including the distance to reflectors and the main noise sources. • The operator of the railways provided accurate information about the network, including:

• Railways lines 3D geometry data; • Locomotive and freight wagon types • Rail vehicles per train; • Trains per day and night period • Daily movements • Type of platform, ballast, and sleepers • Special track elements such as bridges, joints, and overpasses • Curvature Radius (for squeal noise); and • Number of joints or switches / 100m • Due to operational limitations, the rugosity of the track surface was not assessed but the implemented track roughness control maintenance program assured low values. The parameters for the case study noise modelling are summarized in the table below. Table 2: Description of scenarios and scope of noise campaigns.

Parameter Value Comments

Maximum Error (dB): 0dB Maximum search radius (m): 2000 m Minimum source-receiver distance (m): 1000 m Max. Order reflection= 2 (Receivers calculation)

General configuration

Ground absorption 0.7 (Other areas with semi-soft ground) 1 (Platform with ballast and dense vegetation areas)

Meteorological conditions (Temperature and humidity) Those existing during measurement campaign

Propagation conditions 100% favourable

1. Vehicle type d (loco) / a (freight wagon) Table 2.3a

2. Axles per vehicle 6 / 4 Table 2.3a

3. Brake type c. cast-iron block Table 2.3a

4. Wheel measure n. no measure Table 2.3a

The dynamic parameters of each noise source were adapted in each meta-model, to match the values captured during each defined emission window. The latter will allow to establish each of the comparison scenarios. Any known variations between the monitored situation and the modelled scenario were noted and reported as part of the discussion on the outcome of the verification process.

Figure 5. Noise monitoring station (left) and noise model developed in CadnaA (right).

3.3. Confidence level assessment

Set out below is a summary of the results of the verification process, specifying the level differences between the measured and calculated noise levels for each of the analysed scenarios. The level differences are presented in terms of average difference for all the studied receptors for each scenario, according to the following expression:

∑ {𝐿 𝐴𝑒𝑞,𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑,𝑖 −𝐿 𝐴𝑒𝑞,𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑,𝑖 } 𝑛 𝑖=1

𝑛

𝐿𝑒𝑣𝑒𝑙 𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 [𝐷 𝑗 ] =

Where:

• D j is the average difference between measured and calculated values for scenario j . • 𝐿 𝐴𝑒𝑞,𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑,𝑖 is the calculated noise level for the measurement position i of scenario j . • 𝐿 𝐴𝑒𝑞,𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑,𝑖 is the measured noise level for the measurement position i of scenario j . • n is the total number of evaluated scenarios for scenario j .

All average level differences were calculated considering, equally, the noise indicators for day and night periods, independently of the distance to the source or measurement height (arithmetic average of these values for each scenario). Positive values of parameter D represent an overestimation, i.e., the calculated level is higher than the measured level, whilst negative values represent an underestimation, i.e., the calculated level is lower than the measured level. 4. RESULTS

The following figures and table show the correlation coefficient (R 2 , linear regression) and deviation (D) found between noise modelling results and measurement data, for the different studied scenarios. Table 3: Difference level (D) and linear correlation coefficient (R 2 ) for the studied scenarios.

Scenario D (dBA) Correlation coefficient R 2 (linear regression)

Full Brake 3.4 0.58

Acceleration 1.5 0.70

25 mi/h 0.8 0.96

40 mi/h 0.9 0.94

40 mi/h (ISO 1996-2) 1 0.96

Figure 6. Correlation graph for studied scenarios: Full brake (up-left), Acceleration (up-right), 25

mi/h (middle-left) and 40mi/h (middle-right), and for long-term monitoring (40 mi/h-down)

Noise calculations of low-speed scenarios (brake and acceleration) show the highest deviation in comparison with the measured values, with an overestimation within the interval 2-3 dBA and a lower correlation of the linear regression (<70%, the noise emission of the train during these scenarios is highly variable and dependent on the driver's expertise, and the presence of squeal noise). Regarding

constant speed scenarios, it can be verified how the model is validated from a conservative point of view, with a deviation of 1 dBA for the near field receivers (ISO 3095), as well as other receivers in the surroundings of the platform (≈95% of linear correlation between calculated and measured noise levels). The results of the noise prediction model slightly overestimated the measured sound levels and were within the range of uncertainty of the measurements (2-9 dBA, 95% of coverage probability). 5. DISCUSSION

The discussion of the results was undertaken in line with the approach within IMAGINE WP1 final report [15], including an assessment of the measured levels to compare with calculated levels for the monitored scenarios. Any comparison between noise level measurements and calculated strategic noise mapping results can only provide a general indication of the level of confidence for the specific measurement locations, and even then, only if reviewed alongside the uncertainty assessment.

Based upon the results presented in Section 4, it is possible to average the results for each scenario across all the measurement locations to provide a summary of the average level difference and global lineal correlation, as presented in Figure 7 below. The verification of the noise model for the studied freight train concludes an overestimation of the measured values (1,5 dBA) considering all evaluated scenarios, presenting a conservative estimation (1 dBA) for tracks with constant speed (40 mi/h) at mid-distance receivers (30-125 metros). The global correlation between calculated and monitored noise levels has been higher than 96% for all measured locations. The obtained deviations are similar to the results of other projects where verification of CNOSSOS-EU methodology was undertaken (underestimation of 0,6 dBA considering constant speed and the occurrence of different train classes) [16].

To provide a general indication level of confidence for the Noise Maps (to be considered during action planning stage), it is relevant to discuss some factors that may influence the uncertainties associated with the analysis undertaken, within the verification process.

Figure 7. Overall level difference of verification process and correlation graph between

measurement data and noise model for all scenarios

4.1 Measurement Uncertainty

Any measured value (noise indicator in this case) should be expressed along with the associated uncertainty of the applied methodology of measurement, with a chosen coverage probability. Figure 8 shows the results expressed as L± U (L is the measured value; U is the combined standard

uncertainty considering a coverage probability of 95% by convention). This means that the real value would be within the range [L-U,L+U] with a 95% confidence interval (CI).

Figure 8. Overall level difference of Verification Process considering 95% of coverage probability

If the 95% CI of the measured levels, for each type of measurement situation, is applied to the average level difference determined for the same situation, it may be concluded that the calculated noise levels are all within the 95% CI of the measured noise levels for each situation. This graph illustrates that the average noise level differences between calculated and measured levels are all within the interval [0,8; 3,4] dBA for all evaluated scenarios.

One of the main challenges of the measurement process according to ISO 3095 is related to the hard conditions required for the characterization of braking noise and acceleration from standstill noise, which were hardly achievable in regular operation when different train drivers are operating the vehicles. The reported standard deviations of the measured value near the platform during brake and acceleration scenarios (4 and 2,6 dBA respectively) were significantly higher than during stable speed tests (between 1 and 2,1 dBA). The experience of the driver conditioned the presence of squeal and impact noise, which was critical for the variability of the results. 4.2 Input data uncertainty

The uncertainty of the input data used for the noise calculation might have a significant influence on the results of the noise map verification, in the form of noise model uncertainty. Some aspects are discussed below, as examples of the potential influence of input data uncertainty. • The speed of the trains could not be estimated from direct measurement (speed gun) for all considered train pass-by. An unusual distribution of vehicles velocity might introduce a higher deviation of the prediction model that uses as input an average speed by section, less representative of the real speed of the train at some locations (due to driving behaviour and pedestrian level crossing). Moreover, the level of acceleration and braking depended on the driver's expertise for each scenario evaluated. It is strongly recommended to increase the number of samples in these scenarios to reduce the uncertainty of the measured data ( n < 25 samples). • The lack of onsite data about rail roughness and the transfer function to the platform may be a source of uncertainty in the noise modelling, especially when no differences were found during measurements between full cargo and empty vehicles. This situation should be studied in detail in future research. • Uncertainty related to the digital terrain model (DTM) was discarded because an extensive LiDAR campaign was developed during the project and the accuracy degree was very high. The simplification of the DTM guaranteed that no influences were found within the buffered area, where the noise monitoring stations were located.

• Finally, the noise emission might be overestimated due to the mixed operation of vehicles with iron-cast and LL composite brakes. • 4.2 Calculation method uncertainty

CNOSSOS-EU is based on defining individual rail vehicles and trains are built from multiple vehicles. However, the database only holds five vehicles. For this case study, diesel locomotive type vehicles and freight wagon type vehicles were considered applicable to the study case. The general deviation of the railway noise obtained during the verification process may be due to the misalignment between the selected categories of vehicles within the model and the real studied vehicles (and their maintenance status) and the use of railhead roughness was set as the default value (well maintenance according to reported maintenance plan). In addition, some guidelines suggest that CNOSSOS-EU overestimated the noise level when the speed is less than 50 km/h [17] contributing to the overestimation in braking and acceleration scenarios.

Figure 9 shows an analysis of the spectrum analysis of noise level at 7,5 meters from the track, comparing calculated and measured values. The given differences (especially at low frequencies) might be explained by the previously discussed factor about the calculation method.

Figure 9. Spectral comparison of calculated and measured values at 7,5 meters and constant speed

(25 mi/h-left, 40 mi/h-right).

Finally, the implementation of the corrigenda of Directive 996/15/EC [18] and new guidelines and recommendations by different working groups and national institutions might influence the results and conclusions of this study and shall be considered in future research [17]. 6. CONCLUSIONS

The verification of the noise model for the studied freight train concludes an overestimation of the measured values (1,5 dBA) considering all evaluated scenarios, presenting a conservative estimation (1 dBA) for tracks with constant speed (40 mi/h) at mid-distance receivers (30-125 meters). This represents a good level of agreement between the noise mapping and the measured noise levels, and in all cases, the average level differences are within the coverage of the 95% CI of the measured noise levels, and may therefore be considered equivalent results. In general, based on the above conclusions, it can be stated that the use of verified CNOSSOS-EU noise models for freight trains allows the effective use of software simulations to manage and verify the effectiveness of required action plans. 7. REFERENCES 1. European Environment Agency. “Environmental noise in Europe”. Publications Office, 2020. 2. Policy Department Structural and Cohesion Policies, European Parliament. "Reducing Railway

Noise Pollution." 2012.

3. Grupo de trabajo «Emisiones sonoras del ferrocarril» de la Comisión Europea. "Documento de

debate sobre las estrategias y prioridades europeas para la reducción del ruido ferroviario. http://ec.europa.eu/transport/rail/ws/doc/position-pap." 2003. 4. European Environment Agency. EEA Report No 10/2014 Noise in Europe. http://www.eea.europa.eu/publications/noise-in-europe-2014 , 2014. 5. European Commission, Directorate-General for Climate Action, Directorate-General for Energy,

Directorate-General for Mobility and Transport, Zampara, M., Obersteiner, M., Evangelopoulou, S., et al., “EU reference scenario energy, transport and GHG emissions: trends to 2050”. Publications Office, 2016. 6. Directive 2002/49/EC of the European Parliament and of the Council of 25 June 2002. Relating

to the assessment and management of environmental noise. n.d. 7. European Directive 996/2015 and further modifications. n.d. 8. European Commission Working Group Assessment of Exposure to Noise (WG-AEN). "Position

Paper, Good Practice Guide for Strategic Noise Mapping and the Production of Associated Data on Noise Exposure, Version 2" 2006. 9. Research and technology department physics of the railway system and comfort. Harmonoise

WP1.2 Technical report "HAR12TR-020118-SNCF10". HAR12TR-020118-SNCF10, n.d. 10. Małgorzata Szwarc, Bożena Kostek, Józef Kotus, Maciej Szczodrak & Andrzej. “Problems of

Railway Noise—A Case Study”. International Journal of Occupational Safety and Ergonomics, 2011. 11. M. Paviotti, S. J. Shilton, R. Jones and N. Jones. "“Conversion of existing railway source data to

use CNOSSOS-EU”. Euronoise, 2015. 12. Extrium. "P053 - Process Applied to Establish CNOSSOS-EU/National Method Equivalence for

Rail Source Data" 2012. 13. Tohmmy Bustad, Trafikverket / Siddharth Venkataraman, KTH. "“Influence of replacing Cast

Iron brakes with Disc brakes for freight wagons on typical Swedish railway line – comparative evaluation with CNOSSOS-EU and TWINS”." n.d. 14. Paviotti, and Anfosso-Lédée. Common noise assessment methods in Europe (CNOSSOS-EU).

Publications office of the European Union, 180p,. 10.2788/31776. hal-00985998, 2012. 15. IMAGINE WP1 Final Report, “Guidelines and good practice on strategic noise mapping

Deliverable 8 of the IMAGINE project IMA01-TR22112006-ARPAT12”. n.d. 16. Grilo, A. et als. (2019) “Verification of Noise Mapping in Serbia Using CNOSSOS-EU:2015 and

the EU Interim Method. INTERNOISE 2019”, Madrid (oral presentation). n.d. 17. Centro de Estudios y Experimetnración de Obras Públicas. Ministerior de Fomento. Ministerio

para la Transición Ecolócia y el Reto Demográfico. "Guía para la aplicación del método CNOSSOS-EU en la modelización del ruido producido por las circulaciones ferroviarias en las infraestructuras de ADIF y ADIF AV." Madrid, 2022. 18. Commission Delegated Directive (EU) 2021/1226 of 21 December 2020 amending, for the

purposes of adapting to scientific and technical progress. "Annex II to Directive 2002/49/EC of the European Parliament and of the Council as regards common noise assessment method." n.d. 19. European Environment Agency. “Good practice guide on noise exposure and potential health

effects EEA Technical report. No 11/2010. EEA Technical report No 11/2010”, Copenhagen: Office for Official Publications of the European Union, 2010. 20. UIC. “Environmental Noise Directive Development of Action Plans for Railways”. UIC Noise

Expert Network, 2008. 21. World Health Organization. "Environmental Noise Guidelines for the European Region." 2018. 22. UIC. “Railway noise Technical Measures Catalogue”. UIC Noise Expert Network, 2013. 23. Directive 2008/57/EC of the European Parliament and of the Council of 17 June 2008 on the

interoperability of the rail system within the community (recast) (oj l 191, 18.7.2008, p.1) commission regulation (EU) no 1304/2014 of 26 November 2014 on the technical specification for interoperability relating to the subsystem ‘." 2014. 24. European Commission. "Rail freight noise reduction." 2015.