HAL Id: hal-03233773
https://hal.archives-ouvertes.fr/hal-03233773
Submitted on 26 May 2021HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Achievements of Dynamap project
Giovanni Zambon, Chiara Confalonieri, H. Eduardo Roman, Fabio Angelini,
Roberto Benocci
To cite this version:
Giovanni Zambon, Chiara Confalonieri, H. Eduardo Roman, Fabio Angelini, Roberto Benocci. Achievements of Dynamap project. Forum Acusticum, Dec 2020, Lyon, France. pp.1685-1690, �10.48465/fa.2020.1073�. �hal-03233773�
ACHIEVEMENTS OF DYNAMAP PROJECT
Giovanni Zambon
1Chiara Confalonieri
1H. Eduardo Roman
2Fabio Angelini
1Roberto Benocci
11 Department of Earth and Environmental Science (DISAT), University of Milano Bicocca, Milano, Italy
2 Department of Physics “G. Occhialini”, University of Milano Bicocca, Milano, Italy
Correspondence : giovanni.zambon@unimib.it
ABSTRACT
DYNAMAP is a European Life project realized with the aim of developing a dynamic acoustic mapping system to detect and represent the noise impact generated by road infrastructures in real time. To this end, the project involved the development and installation of an automatic monitoring system, based on a limited number of low-cost permanent noise monitoring stations. The Zone 9 in the city of Milan is one of the two pilot area chosen for the implementation of the system. In particular, it is representative of a typical urban scenario. The “real time” noise map is built starting from the data recorded by the 24 sensors homogeneously distributed in the pilot area, each one representative of a number of roads sharing similar characteristics (e.g. daily traffic flow). To evaluate the accuracy and reliability of the system a testing campaign was prepared. The collected field measurements have been so compared with the predicted noise level of Dynamap system. Finally, a statistical analysis has been implemented to try to reduce the error associated with such prediction below 3 dB and then try to optimize the system.
1. INTRODUCTION
Dynamap, a co-financed project by the European Commission through the Life+ 2013 program, aims at developing a dynamic approach to noise mapping, owing to its capability to update environmental noise levels through a direct link with a limited number of noise monitoring sensors. Dynamap has been developed in the city of Milan and the motorway surrounding Rome [1-2]. As for Milan, all network stretches have been grouped in clusters roads sharing similarities among the 24-h continuous acoustic monitoring of the hourly equivalent
LAeq1h levels [3-5]. From this analysis, we developed a
model for predicting the traffic noise of an arbitrary road stretch by means of a non-acoustic parameter [6-8]. We divided the whole set of about 2000 road stretches, the pilot area of the city of Milan is made of, into six groups, each one represented by a noise map. The updating of the noise map is obtained with the information from the traffic noise taken from 24 monitoring stations appropriately distributed over the urban pilot area of Milan [9-10]. The initial tests of calculation reliability and of correction of systematic errors, inherent the
calculation scheme are reported in [11-14]. In this paper, we provide the latest results about the uncertainty of noise mapping prediction related to Dynamap system in the pilot area of Milan named District 9.
2. DYNAMIC NOISE MAPPING: OPERATIONS OF DYNAMAP IN MILAN
An acoustic map has been assigned with each group of roads g (six base maps), the pilot area of Milan has been divided into. Operatively, each noise sensor i (24 for Milan) records a signal after filtering the anomalous events [15-17]. The signal is integrated to obtain an
equivalent level Leqτ,i over a predefined temporal interval
τ (τ = 5, 15, 60 min). Thus, we get 24 Leqτ,i values every τ
min, each one corresponding to a recording station i. To
update the acoustic maps, we deal with variations , ( ),
where the time t is discretized as t = nτ and n is an integer, defined according to
δ , ( ) = , ( , )( )( .)− , ( , )( )( .)
(1)
where Leqref,M(g,j)(Calc.) is a reference value calculated from
the acoustic map of group g (CADNA model) at the time interval Tref = (08:00–09:00) at the point corresponding to the position of the M(g, j)-th station. CADNA model provides mean hourly Leq values over the entire city of Milan with a resolution of 10 m given a set of input traffic flow data, thus representing a reference static acoustic map. Here, we have chosen the reference time Tref = (08:00–09:00) for convenience, since it displays rush-hour type of behaviour. The temporal ranges within the day are conventionally chosen as
τ = 5 min for (07:00–21:00); τ = 15 min for (21:00–01:00); τ = 60 min for (01:00–07:00).
Once all the , ( ) values have been obtained, the six
acoustic maps can be updated corresponding to each group g by averaging the variations in Eqn. (1) over the four values j in each group, according to
δ ( ) = ∑ δ , ( ) (2)
The first quantity we need to know is the value of Leq(s)ref(g) at the point s at the reference time (8:00-9:00)
is provided by the calculated (CADNA) acoustic base
map. The absolute level ( , ) at location s
at time t = nτ can then be obtained by properly adding the contribution of each base map with its variation
calculated according to Eqn. (2).
τ( , ) = 10 ∙ ∑ 10
( ) ( ) δτ( )
(3)
(3) We obtain the so called “scaled map” (dynamic map).
3. MESUREMENT CAMPAIGN
A preliminary validation of the described procedure has been done by comparing the measurements performed in 21 sites within the Zone 9 (purple stars in Fig. 1), randomly selected and equally distributed in the six road groups, with the predictions obtained from the recordings of the 24 monitoring stations for the same sites. The analysis needs the recorded data to be previously cleaned up from eventual anomalous events (ANEs) extraneous to the actual vehicle noise. A dedicated algorithm integrated in Dynamap sensors have been developed to this purpose [18-20].
Figure 1. Area 9 of the city of Milan. Colors correspond to the 6 different groups of streets. Black triangles and purple stars represent the sites where the monitoring stations are installed and the position of test measurements, respectively.
4. TEST OF DYNAMAP PREDICTIONS In the following, we report the comparison between traffic noise measurements with the corresponding
Dynamap predictions of Leqt for t=(5,15,60) min. In
particular for this purpose, we calculate the Mean
deviation, <εLeq>:
< ℇ >= 1∑=1 , ( )− , ( )
(4) where the summation index i extends over four time
zones (24h - NTot =190; day 7-21 - N5min =168; evening
21-01 - N15min =16; night 01-07 - N1h =6). In Table 1 we
report the total daily mean deviation (24h) and the mean deviations within each time zone t.
Site <ε(24 h) Leq> [dB] <εLeq> + (07-21 h) [dB] <εLeq> (21-01 h) [dB] <εLeq> (01-07h) [dB] 1 4.9±1.3 5.0±1.3 4.9±0.8 3.0±0.9 3 2.8±1.7 2.8±1.6 2.5±1.9 2.7±2.5 4 3.0±2.7 5.1±1.6 2.2±0.7 2.3±1.2 5 2.0±1.3 2.1±1.3 0.8±0.7 2.6±1.0 6 3.2±1.9 2.9±1.7 5.3±2.0 6.9±2.0 7 1.7±1.5 1.4±1.1 4.4±2.3 3.4±2.2 8 2.0±1.3 2.0±1.3 1.4±1.2 3.1±2.2 9 1.3±0.9 1.4±0.8 0.5±0.4 2.8±2.1 10 6.4±2.5 7.0±2.1 2.0±1.1 2.3±1.0 11 3.2±2.3 2.9±1.9 5.2±2.1 8.4±3.6 12 7.5±2.3 7.1±2.1 11.6±3.3 7.1±1.7 13 4.0±1.3 4.1±1.4 3.7±0.9 5.0±1.1 14 3.4±1.6 3.6±1.6 1.9±1.3 2.2±1.0 15 3.0±1.2 2.9±1.2 3.1±1.0 4.2±1.4 16 8.4±1.5 8.3±1.5 9.2±1.0 8.6±1.7 17 1.4±0.8 1.4±0.8 1.2±0.6 1.0±1.0 18 6.5±1.5 6.5±1.6 6.7±0.8 6.3±0.7 19 4.0±1.2 4.2±1.1 3.7±0.6 1.2±0.6 20 4.7±2.3 4.7±2.3 4.8±2.3 6.0±1.6 21 1.7±1.1 1.6±1.1 2.1±1.0 1.7±0.7
Table 1. Summary of the results.
Data analysis shows a constant deviation between measurement and prediction; in relation to different sites this occurs over the entire measurement period or only in single time zone.
5. PREDICTION CORRECTION
In Fig. 2, we compare the measurement in Site 16 with the corresponding prediction as obtained by applying Eqn. (3). The error band represents the propagation error applied to Eqn. (3) and associated with the variability of
, within each group g. As we can observe, we have a
discrepancy between predictions and measurements. The mean error over the 24 hours is 8.4 dB. This may suggest the presence of a systematic error inherent the DYNAMAP calculation method which is based on the traffic flow used as input in CADNA software for the calculation of each base map.
Figure 2. Comparison of traffic noise measurements and Dynamap prediction at Site 16. In the figure, the 1σ confidence level is displayed.
To quantify this discrepancy and try to correct it we
calculate, for each site, the relative mean deviation (εL)
between hourly traffic noise measurement level, Leq1h(Meas.), and the corresponding hourly Dynamap
prediction level, Leq1h(DYN.),over the day and night period,
defined as
ℇ = ∑ ( , ( .) , ( .))
, ( .) (5)
where the summation index i extends over two time zones
(day 7-21 N1h =14; evening-night 21-07 N1h =10).
The relative error is then averaged for all roads belonging to the same group, in order to represent the average hourly values of the road group (ε ). During the measuring campaign, we simultaneously record the traffic flows. The comparison between the measured and calculated traffic flow shows that the model provides more reliable results for roads belonging to groups g2-g5 and fails for g1 [see ref. 13]. So we also evaluate the
relative deviation (εF) between measurement and model
for the logarithm of the traffic flow at the reference time, Log F(8:00-9:00),
ℇ = ( : : ) . ( ( : : ) )
( : : ) . (6)
where Log(F(8:00-9:00)Model) is the logarithm of the flows
from 8:00 to 9:00 of the 2012 traffic model. Than we calculate the mean deviation of all sites belonging to the same group, .
These values for and are plotted in Fig. 3,
illustrating that, in general, higher traffic flow deviations are correlated with higher noise level errors. In particular, group g1 presents both higher values of flow and noise level deviation. Therefore, we associate the deviation of traffic flow calculations from the corresponding measurements with an additional systematic error to be taken into account within the Dynamap scheme.
Figure 3. Relative mean hourly deviation between traffic noise measurements and the corresponding Dynamap
predictions vs the relative deviation between the
logarithm of traffic flow measurements and the corresponding model calculations at the reference hour (8:00-9:00) for each group separately. The results refer to: Day time (07:00-21:00) (Left panel), and Evening-Night time (21:00-07:00) (Right panel) periods.
So, we apply a correction to the predicted noise levels in Site 16. The corrected prediction for Site 16 is illustrated in Fig. 4. In this case, the uncertainty band includes both the statistical and systematic errors.
Figure 4. Comparison of traffic noise measurements and Dynamap prediction in Site 16. In the figure, the total error is displayed.
In Table 2, the mean prediction error before, <εLeq>N, and
after, <εLeq>C, the systematic error correction for all the
available sites is shown. All values are in dB. In Table 2, we also report the prediction error after a correction based on the use of median-averaged values in each group g, <εLeq>M. This quantity is less sensitive to outliers and,
consequently, it improves, in almost all case, the corrections. In table 3, the error averaged over the roads belonging to each group is displayed.
Site Group <εLeq>N <εLeq>C <εLeq>M
10 g1 5.0 5.2 5.2 11 g1 4.5 4.0 4.1 18 g1 6.4 6.1 6.1 20 g1 5.3 5.5 5.6 7 g2 1.9 4.0 2.9 12 g2 7.8 2.5 3.8 14 g2 2.8 3.1 1.6
3 g3 1.8 2.3 2.5 6 g3 4.2 4.5 5.9 8 g3 2.1 1.5 2.0 21 g3 1.8 1.5 0.7 13 g4 4.1 2.0 0.8 15 g4 3.3 2.6 1.3 16 g4 8.4 2.6 4.2 1 g5 4.5 3.0 4.4 5 g5 1.9 2.4 1.2 9 g5 1.4 1.9 1.0 19 g6 3.4 1.3 1.3
Table 2. Mean prediction error before, <εLeq>N, and after
systematic error correction for all the considered sites
based on the mean, <εLeq>C, and median average,
<εLeq>M. All values are in dB.
Group <εLeq(g)>N <εLeq(g)>C <εLeq(g)>M
g1 5.3 5.1 5.2 g2 4.2 3.2 2.8 g3 2.5 2.5 2.8 g4 5.3 2.4 2.1 g5 2.6 2.4 2.2 g6 3.4 1.3 1.3
Table 3. Mean prediction error before, <εLeq(g)>N, and
after the systematic error correction, averaged over each
group g, based on the mean, <εLeq(g)>C, and median
average, <εLeq(g)>M. All data are in dB.
Therefore, excluding group g1, for which a specific analysis will be developed, the prediction error of roads belonging to other groups, upon systematic error
correction <εLeq>C, is less than 3 dB for each site with the
exception of sites 6 (g3) and 7 (g2). The latter must be treated differently if we require that the 3 dB constrains must apply to all sites belonging to a group. As an example, consider site 6 (g3). Correcting the predicted noise level using its own relative traffic flow deviation (not the group mean) expressed in percentage terms [1+ εL(g)], we obtain the results reported in Fig. 5, that
correspond to <εLeq>C = 1.1 dB. 7: 0 0 8: 0 0 9: 0 0 1 0: 0 0 1 1: 0 0 1 2: 0 0 1 3: 0 0 1 4: 0 0 1 5: 0 0 1 6: 0 0 1 7: 0 0 1 8: 0 0 1 9: 0 0 2 0: 0 0 2 1: 0 0 2 2: 0 0 2 3: 0 0 0: 0 0 1: 0 0 2: 0 0 3: 0 0 4: 0 0 5: 0 0 6: 0 0 45 50 55 60 65 70 75 L e q1h [d B ] Time [h] Site 6 Measurement
Site 6 Locally Corrected Dynamap
Site 6 Statistical Error
Figure 5. Comparison of traffic noise measurements and Dynamap (corrected) prediction in Site 6.
This result suggests that in order to get an effective correction, the relative error between the measured and the model traffic flow (8-9) in a given road stretch has to be bound within an interval that depends on the group it belongs to. In Fig. 6, for example, is reported the relative
mean hourly deviation between traffic noise
measurements and the corresponding Dynamap
predictions, εL, versus the relative deviation between the
logarithm of traffic flow measurements and the corresponding model calculations at the reference hour
(8:00-9:00), εF, for each site of group 3. The graphs in
Fig. 7 have been obtained assuming for simplicity that the
relation between εL and εF is linear within group 3 (see
Fig. 6). In this case in order to get a prediction error < 3dB for each site, the relative error on the traffic flow can vary of about ±10% with respect to the value centered on the error of the single site as it can be observed in Fig. 7 for Sites 3,6 and 21.
In Fig. 7, the minimum prediction error is obtained “near” the corresponding site-specific flow error. It does not match exactly the value reported in Fig. 6 because we are
using a linear dependence between εL and εF. In other
words, the mean value of the relative error on the traffic flow of a given group g, , (the one that has been used in the correction procedure of Dynamap prediction) must be bound within an interval that can be determined as follows: if we take centered on the minimum of the
relative error of the site-specific traffic flow, εF,S6m for the
case of Site 6, it means that = εF,S6m can have a
maximum standard deviation σ= ±(0.10) to satisfy the
condition about the mean prediction error , <εLeq> < 3dB.
Therefore, must belong to an interval (εF,S6m - 0.10,
εF,S6m + 0.10). This procedure has to be repeated for each
sites of the group. If these conditions are met, all sites
will have <εLeq>, < 3dB. This means that the traffic
model must provide flow values for the streets belonging to each group with comparable accuracy.
Figure 6. Relative mean hourly deviation between traffic noise measurements and the corresponding Dynamap
predictions, εL, versus the relative deviation between the
logarithm of traffic flow measurements and the corresponding model calculations at the reference hour
Figure 7. Mean prediction error <εLeq> as function of the
relative traffic flow error (8-9) εF for Sites 6,9 and 21.
The graphs have been obtained assuming for simplicity
that the relation between εL and εF is linear within group 3
(see Fig. 6).
As for roads characterized by low traffic flows, such as those belonging to group g1, the application of the correction based only on the relative deviation of the local traffic flow is not effective, because in these cases, the noise level is not mainly determined by the local traffic but by that of busier nearby roads. In these cases, we may think to reassign them to other groups, applying the correction of the group whose contribution in the prediction of the noise level is predominant.
6. CONCLUSIONS
The proposed method of noise mapping calculation in Milan is essentially a statistically based approach whose predictive capability over the entire road network is based on the assignment to each road stretch of an available non-acoustic parameter. This is an essential requirement that must be used to generalize Dynamap project to other cities. The non-acoustic parameter must show a high degree of correlation with the traffic noise level, that in our case was in the range of (60-65)%. In terms of accuracy, the predictive capability of the Dynamap scheme is mainly associated with the related accuracy of the chosen non-acoustic parameter. For this reason, a poor accuracy of the non-acoustic parameter is directly reflected on the noise prediction error. For example, in the case in which the non-acoustic parameter is provided by a traffic flow model, as here for the total daily traffic flow, eventual errors in the traffic noise predictions can be traced to a discrepancy between the measured and the model predicted hourly traffic flows.
We have estimated prediction errors of Dynamap for the District 9 of the city of Milan. Dynamap predicts real-time traffic noise at any site within the pilot zone, by relying on the noise measurements of 24 monitoring stations properly located within the zone. For practical purposes, the whole set of roads has been divided into six groups, within which the constituting roads display
similar traffic noise dynamics. Therefore, the process generates six dynamic acoustic maps. The predicted noise level at a generic location is given by a combination of the six dynamically updated maps. In order to investigate the error associated with such prediction, we performed new traffic noise and traffic flow measurements. Besides the expected statistical errors due to the traffic noise fluctuations within each group of roads, our preliminary analysis shows a systematic error inherent to the adopted traffic model. By correlating the Dynamap prediction error to the traffic flow error for each group of roads, the overall error can be reduced. The large error bars present in Fig. 3 are also the results of the poor statistics, which needs to be significantly improved. However, the illustrated method is able to correct the predicted noise levels in a generic location and, therefore, limiting the overall mean error within 3 dB for all groups of roads. A significant improvement would be obtained by implementing a more realistic traffic flow model. This would reduce the systematic error and, therefore, enhance the overall reliability of Dynamap prediction.
In terms of quality of the obtained estimations, the description of the group behavior is rather satisfactory, being the prediction error bounded within about 3 dB. However, some criticalities still remain in terms of the noise prediction at a specific site due to the large deviations around the mean value (see e.g. Fig. 3. In the latter, we get a maximum deviation of 5.5 dB using the mean group correction (Site 20 group 1) and 5.9 dB for the median-average correction (Site 6 group 3) (see Table 2). In order to reduce this deviation one should use updated traffic flow models and eventually, the possibility to increase the number of groups to better tune the range of variability of the non-acoustic parameter. This fact affects prominently group 1 whose non-acoustic parameter has a broad range over three orders of magnitude (from 1 to 1000 vehicles per day).
7. REFERENCES
[1] G. Zambon, R. Benocci, A. Bisceglie, H. E. Roman, P. Bellucci: “The life dynamap project: Towards a procedure for dynamic noise mapping in urban areas”, Applied Acoustics 124 (2017) pp. 52–60. [2] R. Benocci, P. Bellucci, L. Peruzzi, A. Bisceglie, F.
Angelini, C. Confalonieri, G. Zambon: “Dynamic noise mapping in the suburban area of Rome (Italy), Environments – MDPI, 6, 7, (2019), (79).
[3] G. Zambon, R. Benocci, G. Brambilla. “Statistical Road Classification Applied to Stratified Spatial Sampling of Road Traffic Noise in Urban Areas”, Int. J. Environ. Res. 2016, 10(3):411-420.
[4] M. Smiraglia, R. Benocci, G. Zambon, H. Roman, “Predicting hourly traffic noise from traffic flow rate model: Underlying concepts for the dynamap project”, Noise Mapping 3 (2016).
[5] G. Zambon, R. Benocci, G. Brambilla, “Cluster categorization of urban roads to optimize their noise
monitoring”, Environmental monitoring and
assessment (2016) 26.
[6] G. Zambon, R. Benocci, A. Bisceglie, H. E. Roman. “Milan dynamic noise mapping from few monitoring stations: Statistical analysis on road network”. In: The 45th INTERNOISE, Hamburg, Germany, 2016.
[7] X. Sevillano, J. Claudi Socoró, F. Alías, P. Bellucci, L. Peruzzi, S. Radaelli, P. Coppi, L. Nencini, A. Cerniglia, A. Bisceglie, R. Benocci, G. Zambon. “DYNAMAP – Development of low cost sensors networks for real time noise mapping”, Noise Mapping, 3, 1, 2016.
[8] G. Zambon, F. Angelini, R. Benocci, A. Bisceglie, S. Radaelli, P. Coppi, P. Bellucci, A. Giovannetti, R. Grecco, “DYNAMAP: a new approach to real-time noise mapping”, Proc. EuroNoise 2015, Maastricht, May 31 - June 3 2015
[9] G. Zambon, R. Benocci, A. Bisceglie, H.E. Roman, M. Smiraglia; “DYNAMAP project: procedure for noise mapping updating in urban area”; INTER-NOISE and INTER-NOISE-CON Congress and Conference Proceedings, 2017, 5490-5496.
[10] G. Zambon, H.E. Roman, M. Smiraglia, R. Benocci. “Monitoring and prediction of traffic noise in large
urban areas” (Review), Applied Sciences
(Switzerland) 8 (2), 2018.
[11] G. Zambon, F. Angelini, M. Cambiaghi, H. E.
Roman, R. Benocci, “Initial verification
measurements of Dynamap predictive model ”, Proc. EuroNoise 2018, Crete, May 27 - 31 2018 [12] R. Benocci, F. Angelini, M. Cambiaghi, A.
Bisceglie, H. Roman, G. Zambon, R. M. Alsina-Pages, J. C. Socoro, F. Alas, F. Orga, “Preliminary results of Dynamap noise mapping operations”, in: INTER-NOISE and NOISE-CON Congress and Conference Proceedings, 26-29 August 2018, Chicago, Illinois, USA, Institute of Noise Control Engineering.
[13] R. Benocci, A. Molteni, M. Cambiaghi, F. Angelini, H. E. Roman, G. Zambon, “Reliability of DYNAMAP traffic noise prediction”, Applied Acoustics, (2019), 156, 142-150.
[14] R. Benocci, C. Confalonieri, H.E. Roman, F. Angelini, G. Zambon: “Accuracy of the Dynamic Acoustic Map in a Large City Generated by Fixed Monitoring Units”, Sensors 20, 2, 2020.
[15] F. Orga, J. C. Socoro´, F. Alıas, R. M. Alsina-Pages, G. Zambon, R. Benocci, A. Bisceglie, “Anomalous noise events considerations for the computation of road traffic noise levels: The dynamap’s milan case study”, in: Proceedings of the 24th International
Congress on Sound and Vibration, 23-27 July 2017, London, UK, pp. 23–27.
[16] G. Zambon, R. Benocci, F. Orga, R. M. Alsina Pages, F. Alıas, J. C. Socoro´, “Real-time urban traffic noise maps: The influence of anomalous noise events in milan pilot area of dynamap”, in: INTER-NOISE and INTER-NOISE-CON Congress and Conference Proceedings, 27 - 30 August, Hong Kong, China, Vol. 255, Institute of Noise Control Engineering, pp. 3647–3656.
[17] J. C. Socorò, F. Alıas, R. M. Alsina-Pages, “An anomalous noise events detector for dynamic road traffic noise mapping in real-life urban and suburban environments”, Sensors 17 (2017) 2323.
[18] J. C. Socoro´, X. Sevillano, F. Alıas, “Analysis and automatic detection 388 of anomalous noise events in real recordings of road traffic noise for the life dynamap project”, in: INTER-NOISE and NOISE-CON Congress and Conference Proceedings, 21-24 August 2016, Hamburg, Germany, Vol. 253. [19] F. Alıas, J. C. Socoro´, “Description of anomalous
noise events for reliable dynamic traffic noise mapping in real-life urban and suburban soundscapes”, Applied Sciences 7 (2017) 146. [20] R. M. Alsina, F. Orga, F. Alías, G. Zambon, J. C.
Socoró, R. Benocci “Removing Anomalous Events for Reliable Road Traffic Noise Maps Generation: from Manual to Automatic Approaches” Applied Acoustics, Vol. 151, August 2019, pp. 183-192
