Annual assessment of highway traffic noise using two channel short-term measurements

AbstractResearch has shown that for residential area nearby (up to 50 m) exurbia roads, there are cases where the evaluation of noise impact needs more than using the calculation methods alone. Thus, in order to confirm (or not) the implementation of designed measures for noise abatement procedures in these areas, some direct acoustical measurements of traffic flow are required. The present paper deals with the development of new approach to the methodology that allows truly estimate the annual Lden value using the one short-term (till 1 h) time interval measurement. The method is based on the original extrapolation procedure that allows to adapt results obtained in representative shortterm time interval to the annual average rating time periods of day (12 h), evening (4 h) and night (8 h), for which the Lden value is assessed. The procedure is based on the statistical processing of Sound Exposure Levels (SEL) distribution histogram of all vehicles passing by and registered on the reference microphone. Th...


Introduction
Noise in the areas near the exurbia roads and highways is mainly produced by road traffic flow. The EC noise directive (Directive 2002/49/EC 2002 introduces the mandatory noise descriptor L den for noise pollution evaluation. This annual average descriptor depends on the annual average day L d , evening L e and night L n equivalent sound pressure levels and for the Lithuanian case (Lietuvos Respublikos triukšmo… 2004), it may be expressed as follows: . .
So, the contribution of night level is most significant, but the contributions from day and evening levels are the same. Thus, noise for pollution assessment purposes must evaluate the annual values of L d , L e and L n . Generally, this is done by calculation methods used in noise mapping for the areas near the roads. In these commercial (Bendtsen 1999;Steele 2001;Pamanikabud, Tansatcha 2003) and new developed (Cho et al. 2004;King, Rice 2009;Hamet et al. 2010) methods the possible realizations of noise mapping calculations are described. This may be presented as follows: where: L eq,T is the evaluated noise level for time interval T (day, evening or night); L ref is the calculated level at reference (initial for further calculations) point (normally from 7.5 m to 15 m from road centre line at 4 m height); C rec is the correction due to receiver point position or environmental sound propagation correction (i.e. receiver distance from noise source, ground attenuation, screening, reflections are taken into account); C road is the correction taking into account parameters that characterise a road itself (gradient, surface type and quality). In addition, the correction due to weather conditions may be considered as it is described in Lithuanian Standard (LST ISO 1996-2:2008. The initial data for L ref calculation are the annually averaged one whole-day (by hours) traffic flow data (every vehicle category amount and speed) and Sound Exposure Levels (SEL) database, for all considered transport categories depending on the speed values.
However, there are cases for residential areas nearby (till 50 m) the roads where additionally the direct acoustical measurements of traffic flow are required to confirm or not the implementation of designed measures (noise barriers construction, building façade sound insulation improvement, speed limiting, etc.) for noise abatement in these areas.
Generally annual L den assessment as shown in (Gaja et al. 2003) may be accomplished using a sampling technique that results in few randomly chosen whole-days noise monitoring measurements. In (Ng, Tang 2008) was shown that one day L den assessment may be done using arbitrarily chosen few representative short-term measurement periods. The possibility to reduce common measurement time for assessment the annual level L den based on the partially deterministic model for choosing the number of successive representative measurement days is considered in (Jagniatinskis, Fiks 2014). The best results obtained were for representative one-week continuous measurements giving an uncertainty for annual L den of about 2 dB.
Near exurbia roads, the overall noise is not stable and time-history comprises individual vehicle passing by events as shown in Fig. 1.
In this case, there is a possibility to perform the annual L den assessment by permanent measurements in representative short-term period and not exceeding 1 h. Such assumption is based on the following: -annual average (hourly for one whole-day) traffic flow data (amount of considered vehicle categories and speed) to be available; -measurement time period may be chosen, so that will include enough amount of considered vehicle categories passing by to meet the statistically representative time interval sample required by the measurement accuracy (LST ISO 1996-2:2008; also, the conformity to average traffic parameters on the considered road in chosen measurement time interval is required; -for exurbia roads applying the traffic registration methods (video recording; traffic flow measuring equipment, e.g. 'TRAX Flex' or method presented by Mato-Méndez and Sobreira-Seoane (2011), etc.) it is possible to identify the most significant vehicle passing by (that is valid for Lithuanian case, where traffic flows are mostly not steady). That allows to recognize the sound level time history of these passing by (see Fig. 1 as example) and determine the corresponding SEL; -when these SEL's values are determined in the immediate vicinity of the road (e.g. 7.5÷15 m from road centre line -reference measurement point), they may be considered as the basic data that characterize the initial sound energy pro- -the 'receiver' correction C rec may be evaluated using the results of simultaneous direct equivalent level measurements on two microphones placed in the reference and receiver points; -the 'road' correction C road is included in the sound level time history of every vehicle passing by. Thus, all parameters presented in formulas (1) and (2) may be determined.

Acquisition of Initial Data
The basic data were obtained using a two-channel sound level meter 'Symphonie 01 dB-Stell' with a two free field microphones G.R.A.S. to accomplish the produced sound by traffic flow sound time history registrations on two microphones simultaneously (see Fig. 1 as example).
To collect noise emission data the reference point microphone is located in the immediate vicinity of the road at the distance from the centre-line of a road from 7.5 m to 15 m, while the receiver point microphone is located in the considered investigated site where the annual L den value has to be evaluated. The height of both microphones above the ground surface level may be of 4 m or 1.5 m, depending on the aim of the measurements. The sound time history of traffic flow in the chosen representative time interval is registered on both microphones simultaneously, as the two discreet sets of instantaneous A-weighted sound pressure levels L pA,100ms . Therefore, the time-domain 100 ms sampling is applied. Correspondingly, a non-sound traffic flow data were registered at the same time: category of vehicle (in Lithuania the only light and heavy categories of vehicle are considered), as well as their speed. The weather conditions are also controlled by the weather station 'Kestrel 4500' , to correspond to normal conditions, as it is stated in standard (LST ISO 1996-2:2008. Therefore, for the assessment of annual L den value following parameters must be directly measured (determined), obtained and calculated: -directly measured equivalent continuous Aweighted sound pressure levels at reference point (L Aeq,T,ref ) and at receiver point (L Aeq,T,rec ) in chosen measurement representative time interval T = T repr ; -directly determined traffic flow data in time interval T: the number and average velocity of passing by heavy transport/vehicles, N T, heavy and V T, heavy respectively; the number and average velocity of passing by light transport/vehicles, N T,light and V T, light respectively; -declared by road department the hourly one whole-day annual average number of heavy N h, heavy and light N h, light transport and also annual average speed V Y, heavy and V Y, light for heavy and light transport respectively; -the set of A-weighted SEL, L AE, i (i = 1, ..., N, where N is the number of all passing by vehicles) of every vehicle passing by, determined by calculations from the recorded time history of instantaneous A-weighted sound pressure levels L pA,100ms acquired at reference microphone point. From the above presented parameters, the only SEL data are obtained, by using specially developed software. This software allows to recognize every vehicle passing by event (see Fig. 2 as example of three events) and calculate the individual event L AE (SEL) value applying the common formula: where: L A (t) is a separated individual passing by time history of L pA,100ms values (dBA); t E is a corresponding separated time interval (s) and t ref = 1 s. After the SEL value for every vehicle passing by event are determined using formula (3), they are grouped with a step of 1 dB to build a corresponding histogram of SEL distribution densities like analogously was done (Jagniatinskis et al. 2011).
So, it helps to get an initial measured histogram (h k = n k /N, k = 1, …, J), where: n k is the amount of L AE, i (SEL) values falling into k-th interval (or the number of vehicle passing by, whose exposure level L AE, i lies in interval L AE, k ± 0.5 dB) and N = N light + N heavy . In this histogram H* % (H* = N heavy /N) of all 100 % SEL values belong to the heavy vehicle. This paper proposes that SEL values of heavy transport are greater than SEL values of light transport. Thus, exist one of SEL values  Fig. 3. Index of k that corresponds to this SEL* value is denoted as k*.

Algorithm for Annual L den Determination
The below presented procedure describes the order of operation for annual composite descriptor L den determination at the receiver point in the investigated residential site (the appropriate block-diagram is presented in Fig. 4): 1) Simultaneous direct using of two microphones acoustic measurements at the reference and receiver points and non acoustic traffic flow data registration (measurements) in time interval T = T repr seem to be too large to be representative. The time history of instantaneous L pA,100ms in the reference and receiver points data storage.
where: t i is time duration in seconds of the most significant part (usually area of event until the sound pressure level has failed at least 10 dB below the maximum level of event) of i-th event (while t ref = 1 s); L A,i,j is the time history of instantaneous A-weighted sound pressure levels L pA,100ms of i-th vehicle passing by event (so, t i = 100 ⋅ (n(i)-1)[ms] and n(i) is a number of samples of i-th event) acquired at the reference microphone point. Thus, the appropriate histogram (h k , k = 1, …, J) for SEL (L AE,i ) values may be built as described in Section 1.
6) Determination of SEL* value and an appropriate index k* that separates SEL's of light and heavy vehicle passing by from cumulative function of obtained representative histogram and measured proportion of heavy vehicle H* as described in Section 1 (see Fig. 3 where: N T is a declared number of light and heavy vehicle passing by in rating time interval T (day, evening or night). This formula is derived from the common formula presented by Jagniatinskis and Fiks (2012), which describes the relationship between the values of SEL's histogram obtained in representative (measurement) time interval T repr and corresponding equivalent level: 10) Determination of desired composite annual level L den using formula (1).

Determination of Coefficients for Extrapolation
Separately for every rating time interval T, two coefficients R T, light and R T, heavy need to be determined. They are applied to recalculate the representative SEL's histogram, which is obtained from measurements data in representative time interval into another rating time interval with corresponding proportion of light and heavy vehicle. If proportion of light and heavy vehicle in any rating time interval is the same as in representative time interval, the representative histogram is applied to this rating time interval and coefficients R T, light and R T, heavy are equal to 1. Otherwise, these coefficients must be determined in the following manner: formally, as follows from Fig. 3 index k* for which corresponds the SEL* value that separates SEL's of light and heavy vehicle passing by events may be determined from cumulative function of histogram using formula: where: H* is measured proportion of heavy vehicle for representative time interval. So k* is the highest of k values when the condition in the figure brackets is fulfilled. In addition, this is enough to deal with heavy vehicle statistics. So, if for any rating time interval the declared proportion of heavy vehicle is H d ≠ H* and the part of histogram belonging to heavy transport is denoted as The finally extrapolated histogram components are found by multiplying the measured representative histogram components h k by the coefficients R T, heavy and R T, light .

Experimental Results
The related measurements were carried out on the 6 September 2012 on the 57 km arterial highway No A5 Kaunas-Marijampolė. The measurement site scheme is presented in Fig. 5.
The reference microphone (point 1) was placed at 15 m distance from centre line of arterial road (two-line in both directions) at height of 4 m above ground level. The receiver microphone (point 2) was placed at 36 m distance from road centre line at height of 4 m above ground level. Both microphones were placed in the free acoustic sound field.
The permissible vehicle speed on this road is 90 km/h. Measured from two microphones sound indicator data were simultaneously collected in one hour interval from 13:45 to 14:45. During this time, the interval three separate short-term measurements were restarted newly: from 13:45 to 14:00, from 14:00 to 14:25 and from 14:25 to 14:45. In order to estimate the accuracy of the measurement, these three measurements were performed under repeatability conditions when the measurement time interval was changing as 15 min,  Tables 1 and 2. The presented data in Table 2 show that they correspond to the average data of the year 2011 and may be considered as a representative sample of the annual data. The number of heavy transport passing by in all considered measurement time periods is much larger than 30, which satisfies the requirements of (LST ISO 1996-2:2008), as described in Section 1. Additionally, as shown in Tables 1 and 2, the percentage values of heavy transport in measurement time intervals fully correspond to the declared annual average percentage for considerated periods. Thus, the collected measurement data may be considered as representative and the algorithm for these data extrapolation to the rating time periods may be applied.
Firstly, the assessment of the environmental sound propagation correction using formula (4) was performed. The results are presented in Table 3.
Next step represents the post processing of all recorded levels using the algorithm presented in Section 2 (step 5) to determinate the SEL for all detected passing by events in three chosen measurement time intervals. So, the appropriate three distribution histograms for every time interval may be obtained as described in Section 1. Determination of SEL* value and an appropriate index k* from cumulative function for each of three representative histograms are presented in Fig. 6, using the measured proportion of heavy vehicle H* (see Table 2).
For every considered three time intervals, three desired indexes of k*, that separate in SEL's histograms SEL values of light transport from SEL values of heavy transport, may be found from Fig. 6. So k* values for histograms in the first and third time intervals are equal to 21 and for the second one -is equal to 20 (the histograms for time intervals 2 and 3 start from 62 dBA, while histogram for time interval 1 starts from 61 dBA).
Consequently, the extrapolation procedure as described in Section 3 may be carried out. From formula (10), the percentage increment D % that presents the heavy transport percentage changing in rating time periods T in comparison with the heavy transport percentage in measurement representative time intervals, as well the coefficients R T,heavy and R T,light in formulas (11) and (12), may be calculated. The results are presented in Table 4.  As presented in step 8 of algorithm in Section 2, the annual average levels L d,ref , L e,ref , L n,ref for day, evening and night rating time intervals at reference point now may be calculated using formula (6). And finally, the annual average levels L d , L e , L n for day, evening and night rating time intervals at receiver point of investigated living environment using formula (8) as well the desired composite annual level L den are being estimated, by applying the formula (1). All results were obtained separately for 3 measurement time intervals and are presented in Table 5.
In Fig. 7, as an example, the measured histogram of first time interval and corresponding extrapolated histogram for evening rating time period are presented. From the final results presented in Table 5, it may be concluded that the desired composite annual level L den assessed for considered point in living environment (bolded values) has a standard deviation of 0.27 dB, that shows a good repeatability of results and allows to hope that the expanded uncertainty of suggested method with 95 % probability of tolerances is not exceeding 3 dB limit. Also, it is important to note that calculated data in reference point 1 annual L den value for considered road using the known NordTest method (Bendtsen 1999) show a 77.7 dB result that is in tolerance limits of measured results.

Conclusions
1) The research of this paper shows a possibility to perform the annual L den assessment by permanent measurements in representative short-term time interval (till 1 h) in the living areas near (till 50 m) suburban arterial roads for Lithuanian traffic flow conditions. The appropriate algorithm for L den assessment is suggested. Firstly, it is based on the checking for initial measured data to be a representative ones, which proposes that the number of heavy and light vehicle passing by must fulfil the measurement accuracy requirement of standard (LST ISO 1996-2:2008; the proportion of heavy and light vehicle must be similar to declared annual average proportion of these vehicle categories for the day time, when measurements are carried out; meteorological conditions must be stable and normal and road surface must be dry. 2) The measurement procedure allows to estimate the sound propagation correction by simultaneous sound levels acquisition on reference and receiver microphones. The correction due to road pavement characteristics (profile, type, exploitation time) is included automatically in the acquired sound levels.
3) The proposed algorithm contains the original extrapolation procedure that allows to adapt results obtained in representative time interval to the annual average rating time periods of day (12 h), evening (4 h) and night (8 h) for those to whom the L den value is assessed. Procedure is based on the statistical processing of SEL distribution histogram of all vehicles passing by events, being registered by the reference microphone. 4) The accomplished experimental measurements in considered living site in three separate measurement time intervals show that the standard deviation of repeatability of estimated L den values (69.9 dB, 69.4 dB and 69.5 dB) is 0.27 dB. At the reference point obtained L den results (78.4 dB, 78.1 dB and 77.8 dB) are in limit of tolerances with the result of the wellknown NordTest calculation method of 77.7 dB.