Skip to main content
Journal of Urban Health : Bulletin of the New York Academy of Medicine logoLink to Journal of Urban Health : Bulletin of the New York Academy of Medicine
. 2011 Jan 12;88(1):1–13. doi: 10.1007/s11524-010-9501-1

A Study of Riders' Noise Exposure on Bay Area Rapid Transit Trains

Alexis Dinno 1,, Cynthia Powell 2, Margaret Mary King 1
PMCID: PMC3042080  PMID: 21225356

Abstract

Excessive noise exposure may present a hazard to hearing, cardiovascular, and psychosomatic health. Mass transit systems, such as the Bay Area Rapid Transit (BART) system, are potential sources of excessive noise. The purpose of this study was to characterize transit noise and riders’ exposure to noise on the BART system using three dosimetry metrics. We made 268 dosimetry measurements on a convenience sample of 51 line segments. Dosimetry measures were modeled using linear and nonlinear multiple regression as functions of average velocity, tunnel enclosure, flooring, and wet weather conditions and presented visually on a map of the BART system. This study provides evidence of levels of hazardous levels of noise exposure in all three dosimetry metrics. Leq and Lmax measures indicate exposures well above ranges associated with increased cardiovascular and psychosomatic health risks in the published literature. Lpeak indicate acute exposures hazardous to adult hearing on about 1% of line segment rides and acute exposures hazardous to child hearing on about 2% of such rides. The noise to which passengers are exposed may be due to train-specific conditions (velocity and flooring), but also to rail conditions (velocity and tunnels). These findings may point at possible remediation (revised speed limits on longer segments and those segments enclosed by tunnels). The findings also suggest that specific rail segments could be improved for noise.

Electronic supplementary material

The online version of this article (doi:10.1007/s11524-010-9501-1) contains supplementary material, which is available to authorized users.

Keywords: Trains, Noise exposure, Auditory health, Cardiovascular health, Hypertension, Psychosomatic stress

Introduction

Bay Area Rapid Transit (BART) is a regional rapid transit rail system connecting portions of San Francisco, Alameda, San Mateo, and Contra Costa counties. During a typical weekday, passengers take about 360,000 rides on the system.1 While customer satisfaction with the transit system is generally high, noise on BART trains is consistently reported as one of the greatest factors leading to dissatisfaction,2 as borne out by complaints reported to local news media.35 Indeed, we have personally witnessed child and adult passengers covering their ears while riding BART, wearing ear protection such as earplugs or earmuffs, and pairs of individuals leaning close together and shouting at high volume in order to carry on a conversation. Excessive noise levels associated with other above- and belowground mass transit systems have been documented.6,7 Most recently, studies of the New York City subway system indicate that noise levels are sufficiently high to be injurious to the hearing health of some portion of the ridership.8,9

Noise exposure is a concern on BART for several reasons. As a nuisance, noise inhibits conversation and can be an unpleasant sensory experience. More concerning, however, are the known and suspected physiological effects of noise on humans. Chronic exposure to high levels of noise is well established as contributing to hearing loss.1012 Mounting recent evidence suggests that chronic noise provokes the hypothalamic–pituitary–adrenal axis, activating negative endocrine and vascular outcomes, as seen in the association between chronic noise exposure and increased risk of hypertension,1316 increased risk of myocardial infarction,10,13 and psychosomatic stress.10,17 Findings have also shown that children living amid chronic noise have elevated levels of stress-induced hormones as well as elevated blood pressure.18,19 There is also evidence of negative behavioral outcomes of noise exposure as well,10,20 including negative behavioral effects of noise on children’s cognition, concentration and memory,17,1921 and on school performance.22

The level of noise on trains results from many interacting factors,23 including wheel roundness (i.e., deviation from perfectly circular) and wheel trueness (side-to-side wobble in shape), rail condition (e.g., “corrugation,” banking, points, etc.), the speed of a train, whether a particular line segment is in open air or in a tunnel, and the points and curves of a specific line segment among other factors.

This study attempts to quantify BART passengers’ potential exposures to hazardous levels of noise using a convenience sample of line segments central to the transit system. We also ask whether different measures of noise exposure are explained by average velocity, tunnel enclosure, flooring, and wet weather conditions.

Methods

During January, February, and March of 2009, line segments (i.e., the portions of a BART line in a specific direction between one station and the next) were sampled by convenience. The number of samples for the reported segments ranges from 1 to 10.

Measures

Noise dosimetry measurements were made using a Quest Q 300 logging noise dosimeter clipped to the belt, with its microphone clipped to the top of the left shoulder approximately 10 cm from the left ear. Measurements were made separately for each direction in any particular line segment; measurements began with the closing of train doors and ceased with their opening. The dosimeter’s calibration was checked using a QC-10 Calibrator (114 dB at 1,000 Hz) a few minutes before the first measurement on each day that measurements were made, and the dosimeter was recalibrated 2 months before measurements commenced. The dosimeter was set to integrate sound levels over 1-minute intervals with a 3-dB exchange rate, an 80-dB threshold, an 85-dB criterion, and a 115-dB upper limit. We collected average (Leq), maximum (Lmax), and peak (Lpeak) sound levels. Leq and Lmax were A-weighted (dBA) and Lpeak was unweighted (dB).

In addition, the presence of newer hard composite flooring (versus older carpeting), rainwater on the ground, and full or partial enclosure of a line segment by a tunnel were recorded and coded as indicator variables. Average velocity (km/h) was constructed using line segment lengths from BART GIS data and duration as measured by recorded start and stop times. For two records made when boarding at the Millbrae station, there was a wait of several minutes from the time the doors shut to the time the train commenced moving, and both records were excluded in the analyses of average velocity. In all but three cases, measurements were made from the bicycle/wheelchair seat, and in those other three cases were made from the bicycle/handicap/elderly seating.

Descriptive Statistics

Unpresented histograms (see Electronic Supplementary Material [ESM]) characterized the overall distributions of our three dosimetry measures. Leq was massed at about 82 dBA and was skewed somewhat below the mean, with low measures likely resulting from the threshold of 80 dB, which thus are likely to be slightly understating the true Leq level. Of the 268 noise dosimetry measurements of Leq, 60 (22%) were above 85 dBA. Measures of Lpeak massed around 112 dB, but skewed several standard deviations above the mean. Six recorded Lpeak levels exceeded 120 dB, the World Health Organization’s (WHO) guideline threshold for hearing impairment in children,24 and three reported Lpeak levels exceeded 140 dB, the threshold for hearing impairment in adults used by both the National Institute of Occupational Health and Safety and the WHO;24,25; the maximum recorded Lpeak level was above 147 dB (this level was slightly over the upper end of the dosimeter’s listed range, but a sensitivity analysis limiting this value of 144 gave substantively the same results in Table 4, with difference in the restricted model estimate appearing at the 5th significant figure). Lmax was massed just under 90 dBA, but is slightly skewed to the right. One hundred forty-one measurements of Lmax were >90 dBA, four measurements of Lmax were >100 dBA, and the maximum was >105 dBA. These are very high levels, well exceeding the levels cited in the US Environmental Protection Agency’s26 examination of maximum allowable exposures. We considered the effects of time of day as a proxy for passenger noise on all three noise measures using multiple nonparametric smoothing regressions (presented in ESM), but found no relationship in each case. Figures 1a, b, and c illustrate the average and maximum recorded values for all three metrics along each line segment for which there were two or more measures in order to visually characterize the noise exposure of particular rides (color maps annotated with measurements are available in ESM). These maps were generated using the numbers in Table 1, which associates average and maximum recorded for each dosimetry measure.

Table 4.

Parameter estimates for full and restricted and nonlinear least squares models (Leq) and ordinary least squares models (Lpeak and Lmax)

  Full model Restricted model
Parameter estimate (SD) p valuea Parameter estimate (SD) p valuea
y: Leq (dBA)
Inline graphic 82.9 (1.05) <0.001 83.0 (1.04) <0.001
Inline graphic 0.489 (0.070) <0.001 0.516 (0.077) <0.001
Inline graphic (break at 53 km/h) −0.430 (0.124) 0.001 −0.462 (0.122) <0.001
Inline graphic 6.78 (2.00) 0.001 5.06 (0.680) <0.001
Inline graphic 0.260 (0.145) 0.094
Inline graphic −0.278 (0.193) 0.169
Inline graphic 1.98 (0.658) 0.004 1.81 (0.856) 0.003
Inline graphic −0.706 (1.06) 0.503
RMSE 5.030 5.095
R2 0.479 0.457
y: Lpeak (dB)
Inline graphic 118 (5.98) <0.001 113 (0.507) <0.001
Inline graphic −0.352 (0.297) 0.351
Inline graphic −5.92 (5.96) 0.387
Inline graphic 0.425 (0.298) 0.305
Inline graphic 1.07 (0.517) 0.118
Inline graphic −0.121 (0.889) 0.889
RMSE 4.632 4.766
R2 0.073 0.000
y: Lmax (dBA)
Inline graphic 88.2 (2.07) <0.001 88.2 (2.07) <0.001
Inline graphic −0.115 (0.102) 0.310 −0.114 (0.107) 0.258
Inline graphic 2.51 (2.12) 0.310 2.50 (2.11) 0.258
Inline graphic 0.309 (0.104) 0.006 0.309 (0.104) 0.005
Inline graphic 1.56 (0.104) <0.001 1.52 (0.412) 0.001
Inline graphic −0.375 (0.858) 0.660
RMSE 3.505 3.499
R2 0.4708 0.462

All models accounted for clustering in line segments, thereby estimating robust standard errors. N = 266 for all models. Both the full and restricted nonlinear least squares models of Leq converged in 11 iterations

aAll p values were corrected for multiple comparisons using the false discovery rate41, using p.adjust() in R version 2.9

Figure 1.

Figure 1.

Figure 1.

Figure 1.

a Map of mean and maximum Leq (dBA). b Map of mean and maximum Lpeak (dB). c Map of mean and maximum Lmax (dBA). The number of observations is overlaid on each line segment, except those with a single measurement. The shape and position of line segments have been distorted to facilitate visual discrimination and should be interpreted as schematic.

Table 1.

Dosimetry measures for 51 BART line segments with two or more measures

Segment N Duration Leq (dBA) Lpeak (dB) Lmax (dBA)
12th St–19th St 6 64 (3.4) 88 (4.1) 93 77 (8) 82 109 (4.2) 118
12th St–West Oakland 5 264 (51.3) 84 (2.9) 89 69 (7.4) 80 111 (0.7) 112
16th St–24th St 6 101 (2.1) 92 (0.7) 93 82 (1.2) 83 114 (0.7) 115
16th St–Civic Center 8 119 (7.8) 94 (1.5) 96 87 (1.7) 90 119 (10.6) 140
19th St–12th St 5 65 (8.3) 87 (3) 90 79 (3.3) 82 110 (1.7) 111
19th St–MacArthur 7 222 (104.1) 91 (2.1) 94 79 (2.7) 82 116 (0.9) 117
24th St–16th St 10 103 (7.5) 92 (1.7) 95 86 (2) 90 115 (2.6) 120
24th St–Glen Park 4 143 (4) 93 (0.7) 93 84 (0.6) 85 111 (0.7) 112
Ashby–MacArthur 5 167 (5.9) 95 (1.4) 97 82 (1.4) 84 113 (0.9) 114
Ashby–Berkeley 6 127 (8.3) 98 (1.9) 101 90 (1.9) 93 113 (1.7) 115
Balboa Park–Daly City 4 216 (54.8) 92 (1) 94 80 (2.5) 83 115 (1.3) 116
Balboa Park–Glen Park 4 117 (3.3) 97 (1.1) 98 91 (1.2) 92 115 (1.3) 116
Bay Fair–Hayward 2 218 (2.8) 87 (2.5) 89 77 (3) 79 113 (1.5) 114
Bay Fair–San Leandro 2 272 (81.3) 91 (9.6) 97 73 (3) 75 129 (25.9) 147
Civic Center–16th St 8 117 (2) 92 (1.9) 94 86 (1.6) 88 111 (1.7) 114
Civic Center–Powell 10 70 (3.9) 89 (3) 94 78 (2.8) 82 111 (1.5) 114
Coliseum–Fruitvale 2 173 (5.7) 84 (1.5) 85 68 (5.2) 72 110 (3) 112
Coliseum–San Leandro 3 221 (4) 90 (2.4) 93 79 (1.8) 80 111 (1.6) 113
Colma–Daly City 4 187 (16.1) 93 (1) 94 81 (1.7) 83 112 (1.2) 114
Colma–South San Francisco 4 156 (4.8) 89 (1.2) 90 78 (2) 80 111 (1.4) 113
Daly City–Balboa Park 4 183 (5.4) 91 (2.8) 93 80 (3.8) 83 114 (1.4) 115
Daly City–Colma 4 216 (5.4) 84 (1.6) 86 73 (3.7) 77 112 (1.1) 113
Berkeley–Ashby 3 118 (3.1) 94 (1.6) 95 86 (1.7) 87 112 (2.7) 115
Berkeley–North Berkeley 5 124 (0.7) 88 (2.3) 91 81 (2.3) 85 113 (1.7) 115
Embarcadero–Montgomery 8 59 (1.4) 89 (1.5) 91 82 (1.1) 84 114 (1.4) 116
Embarcadero–West Oakland 10 394 (9.5) 97 (1.8) 100 87 (1.1) 89 114 (1.3) 117
Fruitvale–Coliseum 3 172 (8.1) 92 (1.2) 93 82 (0.7) 83 114 (1.3) 114
Fruitvale–Lake Merritt 2 234 (2.1) 94 (3.7) 96 82 (1.7) 83 112 (1.2) 113
Glen Park–24th St 4 139 (3.5) 96 (0.7) 96 86 (0.8) 87 115 (1) 116
Glen Park–Balboa Park 4 120 (7.8) 95 (1.1) 96 87 (0.8) 88 113 (1.1) 114
Hayward–Bay Fair 2 219 (9.2) 85 (2.5) 87 74 (3.4) 76 112 (2.6) 114
Lake Merritt–12th St 2 154 (2.1) 81 (0.1) 81 62 (0.7) 63 111 (2.4) 113
Lake Merritt–Fruitvale 3 207 (3.8) 95 (1.4) 97 83 (0.8) 84 116 (3.7) 120
Lake Merritt–West Oakland 2 317 (39.6) 85 (0.7) 85 68 (5.7) 72 111 (0.6) 112
MacArthur–19th St 6 179 (28.3) 90 (1.6) 93 80 (1.9) 84 116 (1.2) 118
MacArthur–Ashby 5 180 (16.3) 87 (2.3) 89 75 (3.4) 79 109 (1.7) 111
MacArthur–Rockridge 2 116 (5.7) 83 (3) 85 66 (3.5) 69 117 (7.2) 122
Millbrae–San Bruno 4 490 (300.7) 91 (1.1) 92 79 (2.1) 81 113 (0.4) 113
Montgomery–Embarcadero 8 65 (7.4) 89 (2.2) 93 81 (3) 84 113 (1.4) 115
Montgomery–Powell 8 66 (3.9) 84 (2.3) 88 73 (4.5) 80 106 (1.4) 108
North Berkeley–Berkeley 4 134 (20.1) 90 (0.5) 90 82 (1.4) 84 120 (15.4) 143
Powell–Civic Center 8 75 (10) 87 (3.1) 93 78 (2.7) 83 109 (1.3) 112
Powell–Montgomery 10 73 (13.3) 86 (3.7) 94 73 (5.6) 78 107 (1.9) 111
San Bruno–Millbrae 4 351 (111.5) 88 (3) 92 76 (5) 81 111 (1.4) 113
San Bruno–South San Francisco 4 179 (8.3) 96 (1.3) 97 86 (1.2) 87 114 (1.2) 115
San Leandro–Bay Fair 2 193 (2.1) 89 (1) 90 81 (1.7) 82 116 (3.3) 118
South San Francisco–Colma 4 172 (16.3) 92 (0.9) 93 80 (1.8) 81 112 (0.7) 113
South San Francisco–San Bruno 4 179 (9.9) 95 (0.6) 96 87 (1.4) 88 112 (0.8) 113
West Oakland–12th Street 4 211 (24.7) 86 (2) 88 75 (2.5) 77 111 (0.9) 112
West Oakland–Embarcadero 7 380 (15.6) 101 (3.1) 105 92 (1.7) 94 118 (6.8) 133
West Oakland–Lake Merritt 5 315 (33.9) 89 (2.1) 91 74 (2) 76 112 (2.4) 115

Numbers given are mean (SD) maximum. N is the number of observations made on each line

These data can be used to describe exposure to transit noise on BART cars under different plausible transit scenarios. We present three such commute scenarios corresponding to actual weekday round-trip commutes of two of the authors during the study period. The average time of these commute scenarios ranged from 58 to 73 minutes, and each scenario comprised 24 line segments round-trip. The minimum mean Leq exposure at different levels by mean time in transit for these three commute scenarios is detailed in Table 2. On average, riders experience a minimum exposure between 54 and 61 min/day at Leq ≥ 70 dBA and between 19 and 23 min/day at Leq ≥ 85 dBA due to noise while BART cars are in motion. Riders on the MacArthur to Daily City commute experience a minimum exposure of 7 min/day at Leq ≥ 95 dBA.

Table 2.

Minimum mean Leq exposure times for three round-trip commute scenarios (h:mm:ss)

  24th street and North Berkeley 24th street and Hayward MacArthur and Daly City
Round-trip mean transit duration 0:57:57 1:13:06 1:02:25
Mean Leq
 ≥70 dBA 0:53:33 1:01:13 0:53:33
 ≥75 dBA 0:51:14 0:45:28 0:51:14
 ≥80 dBA 0:36:27 0:35:44 0:36:27
 ≥85 dBA 0:22:38 0:18:33 0:22:38
 ≥90 dBA 0:08:27 0:06:20 0:08:27
 ≥95 dBA 0:00:00 0:00:00 0:06:34

The number of line segments where mean Lmax exceeds specific levels for these three commutes is detailed in Table 3. Riders in the commute scenarios experience a minimum of between 20 and 22 exposures per day at mean Lmax ≥ 85 dBA, between 10 and 14 exposures per day at mean Lmax ≥ 90 dBA, and between 2 and 5 exposures per day at mean Lmax ≥ 95 dBA. Riders on the two commute scenarios including MacArthur station experienced at least one exposure per day to mean Lmax ≥ 100 dBA.

Table 3.

Line segments with mean Lmax at different levels for three round-trip commute scenarios

  24th street and North Berkeley 24th street and Hayward MacArthur and Daly City
Total number of segments round-trip 24 24 24
No. of segments with mean Lmax
 ≥70 dBA 24 22 24
 ≥75 dBA 24 22 24
 ≥80 dBA 24 22 24
 ≥85 dBA 22 20 21
 ≥90 dBA 12 10 14
 ≥95 dBA 4 2 5
 ≥100 dBA 1 0 1

Data Analysis

We explained our three noise dosimetry measures by velocity, tunnel enclosure, flooring type, and weather conditions using multiple regression frameworks. We began with a full model, removing predictors stepwise according to the highest p value. Preliminary multivariate nonparametric smoothing regression27,28 of all three dosimetry measures (Eqs. 1a1c) suggested that only the relationship between average velocity and Leq was nonlinear (Figure 2), with the linear main effect of velocity virtually saturating at about 53 km/h. Our preliminary nonparametric additive running-line smoothing regressions were fit using the running package 2.0.0 in Stata (these models used centered average velocity). Such nonlinearities are both substantively interesting and violate the assumption of linearity, which may bias linear regression estimates. Accordingly, we modeled the effect of average velocity on Leq using nonlinear least squares regression to model a break point in the linear relationship (Eqs. 2a and 2b). Lpeak and Lmax were modeled using multiple ordinary least squares regression. All regression analyses were conducted in Stata,29 and average velocity was centered. Full regression models included all predictor variables, plus multiplicative interaction terms for centered average velocity and degree of tunnel enclosure. We estimated full models which included all predictors and interactions (Eqs. 3a and 4a) and restricted models which retained only predictors at the α = 0.05 level (Eqs. 3b and 4b).

graphic file with name M1.gif 1a
graphic file with name M2.gif 1b
graphic file with name M3.gif 1c
graphic file with name M4.gif 2a
graphic file with name M5.gif 2b
graphic file with name M6.gif 3a
graphic file with name M7.gif 3b
graphic file with name M8.gif 4a
graphic file with name M9.gif 4b

where vc is centered average velocity (in km/h); vb is the change in slope of average velocity at the break, modeled by max(average velocity—θv, 0); θv is the estimated breakpoint at which centered average velocity changes slope, T is the presence of a tunnel longer than three cars on the line segment (1 = tunnel present); vcT is the multiplicative interaction between vc and T; f is the presence of newer hard floor instead of older carpet (1 = with hard floor); w is the presence of rainwater on the ground during the ride (1 = water on ground); and Inline graphic are the model error terms, adjusted for clustering by line segment, assumed normal.

Figure 2.

Figure 2.

Effect of average velocity on Leq modeled with a nonlinear breakpoint in the effect of average velocity at 52 km/h (thick black line) overlaid on top of a restricted nonparametric smoothing model of Inline graphic—(thick black line) with 95% point-wise confidence intervals (thin gray lines) and the raw data.

Despite the fact that our sample was by convenience, our study obtains very high statistical power (>0.95) with respect to Cohen’s30 post hoc analysis of power to detect a change in the sample correlation coefficient (R2) due to the inclusion of the independent variables. We also have very high power (>0.90) using Kelley and Maxwell’s method31 in our restricted models of both Leq and Lmax, but are underpowered for the unrestricted models and for Lpeak generally (see ESM).

Results

Clustered regression results are presented in Table 4. We found that average velocity had different effects on our three dosimetry measures. Leq increased linearly with average velocity by 0.52 (95% CI = 0.36–0.67) dBA km−1 h−1, with that effect almost completely saturating to 0.05 dBA km−1 h−1 (95% CI = −0.34–0.45) above approximately 53 km/h as illustrated in Figure 2. Lpeak was not found to be significantly related to average velocity. Lmax was found to decrease linearly by −0.11 dBA (95% CI = −0.32–0.09) in cars running on line segments without tunnels, but to increase linearly by 0.19 dBA (95% CI = 0.15–0.24, calculated as described in Figueiras et al.,32, p. 2100) in cars running on segments with tunnels.

Leq increased by 5.1 dBA (95% CI = 3.7–6.4) on line segments enclosed by tunnels. Lmax increased by 2.5 dBA (95% CI = −1.7–6.7), with the above described significant interaction with average velocity.

Presence of the newer composite flooring was associated with an increase of 1.8 dBA (95% CI = 0.58–3.1) in Leq and was associated with an increase of 1.5 dBA (95% CI = 0.69–2.3) in Lmax. Flooring was not associated with Lpeak.

The presence of water on the ground was not associated with any of our three noise dosimetry measures.

Discussion

This small study provides evidence of potential noise exposures that may be deleterious to the health of BART passengers. The Leq and Lpeak levels reported here are comparable, although somewhat louder, to in-car noise levels recently reported in the New York Metro subway system.8,9 The reported Leq exposure durations ≥85 dBA in the three commute scenarios translate to 40–48% of the maximum daily noise exposure levels set by the EPA to broadly protect population hearing.24,26 This implies compounded noise-related risks for those riders who reside or work in very noisy environments. Our Leq and Lmax measures indicate exposure to very loud noise for periods somewhat comparable to the daily ranges associated with increased cardiovascular and psychosomatic health risks.10,12,15,17,18,20,3337 However, we warn that these studies generally treated periodic exposures throughout the day, such as those due to proximity to rail systems or traffic. Most BART trips are likely to extend beyond one line segment; for round-trip commuters, such exposure will double in the course of a day. This implies chronic exposure to persistent levels of noise during the workday and may present a threat of hypertension and other health problems associated with chronically heightened psychosomatic stress. Lpeak levels indicate acute exposures potentially damaging to adult hearing on about 1% of rides from one station to the very next station and acute exposures potentially damaging to children’s hearing on about 2% of such rides.24 Hearing may also be threatened by BART noise indirectly as many people employ headphones while riding BART (e.g., for digital musical players), and BART noise may drive riders to raise headphone volume to damaging levels.

While recognizing that passenger exposures to loud noises on BART are unlikely to exceed an hour or two per day and thus likely to present only a small health risk to individuals, we also consider this from a population perspective; small increases in individual risk for health problems caused by chronic exposure, when multiplied across large populations—such as the hundreds of thousands of riders each weekday—may amount to large public health concerns.38,39 Moreover, from a vulnerabilities perspective,40 populations already under stress suffer greater extremes and greater uncertainty in health outcomes as a result of stresses; because BART serves the elderly, school-age children, and socioeconomically marginalized communities, we find vulnerability to noise especially concerning and a needed avenue for further research.

We have provided evidence that the noise to which passengers are exposed may be due not only to car-specific conditions (velocity and flooring) but also to rail conditions (speed limit and tunnels). These findings may point at possible remediation (revised speed limits on longer segments and those enclosed by tunnels). The findings also suggest the possibility that specific line segments could be physically improved for noise. Factors not considered here—such as wheel and brake conditions or rail conditions—may also contribute to noise levels.

This study has several limitations. First, the small sample size does not permit an estimation of the distribution of dosimetry responses for each line segment. A thorough sampling of every line segment in the BART system would also give a better picture of passenger exposure. Likewise, we did not account for clustering of variance by car that a larger study would (for example, using cross-classified multilevel models). Better dosimeters could provide more finely spaced measurements permitting a more nuanced visual characterization of gradients of noise dosimetry along single line segments and assessment of the relationships between more instantaneous measures of velocity and dosimetry measures. Such finely spaced measurements could also permit total counts of Lpeak and Lmax events as recommended by WHO,24 rather than the “at least one event per line segment” given by these measures in this study. Our use of average (rather than instantaneous) velocity biases regression results toward finding smaller effects since if the effects of instantaneous velocity on noise are positive, “average velocity” will be slightly lower than instantaneous velocity when dosimetry measures are high and conversely will be higher than instantaneous velocity when dosimetry measures are low.

We conclude by noting that BART’s operation appears to produce several kinds of noise-related health hazard. While news reports indicate that BART took steps to improve rail condition in 2009,35 it remains to be seen if and how passenger noise exposure will be affected. BART, being a public institution, should serve its passengers at a minimum by communicating the health hazard imposed by the noisy conditions under which it operates, perhaps even suggesting ways for passengers to protect themselves from hazardous noise to, most fully, by making trains quieter. BART could also establish ongoing noise dosimetry measures for the protection of riders’ health. Such a surveillance system could also provide better understanding of velocity/noise measures since instantaneous train speed is available to BART operators.

Electronic Supplementary Material

ESM 1 (4.8MB, docx)

(DOCX 3856 kb)

Acknowledgments

We thank Craig Ishida and the Environmental Health & Safety Department of California State University East Bay for the loan of the dosimeters used in this study. We thank Professor Tom Dolan of Portland State University’s Department of Speech and Hearing Sciences manuscript feedback before submission. This study was unfunded.

References

  • 1.2008 BART customer satisfaction survey. Commissioned study report. San Francisco, CA: Corey, Canapary & Galanis Research; 2009. [Google Scholar]
  • 2.2008 BART station profile study. Commissioned study report. San Francisco, CA: Corey, Canapary & Galanis Research; 2009. [Google Scholar]
  • 3.Curiel JSF. BART noise needs to be silenced. San Francisco Chronicle. December 11, 2008; B-2.
  • 4.John from cyberspace noise in the tube, kids in the car and kudos for the queen. The Oakland Tribune. Available at: http://www.insidebayarea.com/oaklandtribune. April 25, 2007.
  • 5.Rocha A. BART goal: stop SCREEECH! The Examiner. October 25, 2007.
  • 6.Bhattacharya SK, Bandyopadhyay P, Kashyap SK. Calcutta metro: is it safe from noise pollution hazards? Ind Health. 1996;34(1):45–50. doi: 10.2486/indhealth.34.45. [DOI] [PubMed] [Google Scholar]
  • 7.Gershon RRM, Qureshi KA, Barrera MA, Erwin MJ, Goldsmith F. Health and safety hazards associated with subways: a review. J Urban Health. 2005;82(1):10–20. doi: 10.1093/jurban/jti004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Gershon RRM, Neitzel R, Barrera MA, Akram M. Pilot survey of subway and bus stop noise levels. J Urban Health. 2006;83(5):802–812. doi: 10.1007/s11524-006-9080-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Neitzel R, Gershon RRM, Zeltser M, Canton A, Akram M. Noise levels associated with New York City’s mass transit systems. Am J Publ Health. 2009;99(8):1393–1399. doi: 10.2105/AJPH.2008.138297. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Vernez Moudon A. Real noise from the urban environment how ambient community noise affects health and what can be done about it. Am J Prev Med. 2009;37(2):167–171. doi: 10.1016/j.amepre.2009.03.019. [DOI] [PubMed] [Google Scholar]
  • 11.Ising H, Kruppa B. Health effects caused by noise: evidence in the literature from the past 25 years. Noise Health. 2004;6(22):5–13. [PubMed] [Google Scholar]
  • 12.Lusk S. Noise exposures: effects on hearing and prevention of noise induced hearing loss. AAOHN J. 1997;45(8):397–405. [PubMed] [Google Scholar]
  • 13.Babisch W, Kamp I. Exposure–response relationship of the association between aircraft noise and the risk of hypertension. Noise Health. 2009;11(44):161–168. doi: 10.4103/1463-1741.53363. [DOI] [PubMed] [Google Scholar]
  • 14.Haralabidis AS, Dimakopoulou K, Vigna-Taglianti F, et al. Acute effects of night-time noise exposure on blood pressure in populations living near airports. Eur Heart J. 2008;29(5):658–664. doi: 10.1093/eurheartj/ehn013. [DOI] [PubMed] [Google Scholar]
  • 15.Jarup L, Babisch W, Houthuijs D, et al. Hypertension and exposure to noise near airports: the HYENA Study. Environ Health Perspect. 2008;116(3):329–333. doi: 10.1289/ehp.10775. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Stansfeld S, Haines M, Brown B. Noise and health in the urban environment. Rev Environ Health. 2000;15(1–2):43–82. doi: 10.1515/reveh.2000.15.1-2.43. [DOI] [PubMed] [Google Scholar]
  • 17.Passchier-Vermeer W, Passchier WF. Noise exposure and public health. Environ Health Perspect. 2000;108(Suppl 1):123–131. doi: 10.2307/3454637. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Babisch W, Beule B, Schust M, Kersten N, Ising H. Traffic noise and risk of myocardial infarction. Epidemiology. 2005;16(1):33–40. doi: 10.1097/01.ede.0000147104.84424.24. [DOI] [PubMed] [Google Scholar]
  • 19.Evans GW, Lercher P, Meis M, Ising H, Kofler WW. Community noise exposure and stress in children. J Acoust Soc Am. 2001;109:1023–1027. doi: 10.1121/1.1340642. [DOI] [PubMed] [Google Scholar]
  • 20.Stansfeld SA, Matheson MP. Noise pollution: non-auditory effects on health. Br Med Bull. 2003;68(1):243–257. doi: 10.1093/bmb/ldg033. [DOI] [PubMed] [Google Scholar]
  • 21.Matheson M, Stansfeld SA, Haines M. The effects of chronic aircraft noise exposure on children’s cognition and health: 3 field studies. Noise Health. 2003;5(19):31–40. [PubMed] [Google Scholar]
  • 22.Sanz SA, García AM, García A. Road traffic noise around schools: a risk for pupil’s performance? Int Arch Occup Environ Health. 1993;65(3):205–207. doi: 10.1007/BF00381157. [DOI] [PubMed] [Google Scholar]
  • 23.Harris CM, Aitken BH. Noise in subway cars. Sound Vibration Mag. 1971; February, pp. 12–14.
  • 24.Guidelines for community noise. WHO expert task force report. Geneva: World Health Organization; 1999. [Google Scholar]
  • 25.Criteria for a recommended standard: occupational noise exposure revised criteria 1998. DHHS (NIOSH) Publication. Cincinnati, OH: Department of Health and Human Services (DHHS), Public Health Service, Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health; 1998. [Google Scholar]
  • 26.Protective noise levels: condensed version of EPA levels document. EPA report. Washington, DC: U.S. Environmental Protection Agency; 1979. [Google Scholar]
  • 27.Friedman JH, Stuetzle W. Projection pursuit regression. J Am Stat Assoc. 1981;76(376):817–823. doi: 10.2307/2287576. [DOI] [Google Scholar]
  • 28.Royston P, Cox NJ. A multivariate scatterplot smoother. Stata J. 2005;5(3):405–412. [Google Scholar]
  • 29.Stata statistical software: release 11. College Station, TX: StataCorp LP; 2009. [Google Scholar]
  • 30.Cohen MP. Sample size considerations for multilevel surveys. Int Stat Rev. 2005;73(3):279–287. doi: 10.1111/j.1751-5823.2005.tb00149.x. [DOI] [Google Scholar]
  • 31.Kelley K, Maxwell S. Sample size planning for multiple regression: power and accuracy for omnibus and targeted effects, chapter 11. In: Alasuutari P, Bickman L, Brannen J, editors. Sage Handbook of Social Research Methods. Thousand Oaks: Sage; 2008. pp. 166–192. [Google Scholar]
  • 32.Figueiras A, Domenech-Massons JM, Cadarso C. Regression models: calculating the confidence interval of effects in the presence of interactions. Stat Med. 1998;17(18):2099–2105. doi: 10.1002/(SICI)1097-0258(19980930)17:18&#x0003c;2099::AID-SIM905&#x0003e;3.0.CO;2-6. [DOI] [PubMed] [Google Scholar]
  • 33.Babisch W. Traffic noise and cardiovascular disease: epidemiological review and synthesis. Noise Health. 2000;2(8):9–32. [PubMed] [Google Scholar]
  • 34.Babisch W. Transportation noise and cardiovascular risk: updated review and synthesis of epidemiological studies indicate that the evidence has increased. Noise Health. 2006;8(30):1–29. doi: 10.4103/1463-1741.32464. [DOI] [PubMed] [Google Scholar]
  • 35.Barregard L, Bonde E, Ohrstrom E. Risk of hypertension from exposure to road traffic noise in a population-based sample. Occup Environ Med. 2009;66(6):410–415. doi: 10.1136/oem.2008.042804. [DOI] [PubMed] [Google Scholar]
  • 36.Bluhm GL, Berglind N, Nordling E, Rosenlund M. Road traffic noise and hypertension. Occup Environ Med. 2007;64(2):122–126. doi: 10.1136/oem.2005.025866. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Kempen EEMM, Kruize H, Boshuizen HC, Ameling CB, Staatsen BAM, Hollander AEM. The association between noise exposure and blood pressure and ischemic heart disease: a meta-analysis. Environ Health Perspect. 2002;110(3):307–317. doi: 10.1289/ehp.02110307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Gordis L. Epidemiology. 4. Philadelphia: W.B. Saunders; 2008. [Google Scholar]
  • 39.Pearce N. Traditional epidemiology, modern epidemiology, and public health. Am J Publ Health. 1996;86(5):678–683. doi: 10.2105/AJPH.86.5.678. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Levins R, Lopez C. Toward an ecosocial view of health. Int J Health Serv. 1999;29(2):261–293. doi: 10.2190/WLVK-D0RR-KVBV-A1DH. [DOI] [PubMed] [Google Scholar]
  • 41.Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B Methodol. 1995;57(1):289–300. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

ESM 1 (4.8MB, docx)

(DOCX 3856 kb)


Articles from Journal of Urban Health : Bulletin of the New York Academy of Medicine are provided here courtesy of New York Academy of Medicine

RESOURCES