WHO Environmental Noise Guidelines for the European Region: A Systematic Review on Environmental Noise and Annoyance

Abstract

Background: This paper describes a systematic review and meta-analyses on effects of environmental noise on annoyance. The noise sources include aircraft, road, and rail transportation noise as well as wind turbines and noise source combinations. Objectives: Update knowledge about effects of environmental noise on people living in the vicinity of noise sources. Methods: Eligible were published studies (2000-2014) providing comparable acoustical and social survey data including exposure-response functions between standard indicators of noise exposure and standard annoyance responses. The systematic literature search in 20 data bases resulted in 62 studies, of which 57 were used for quantitative meta-analyses. By means of questionnaires sent to the study authors, additional study data were obtained. Risk of bias was assessed by means of study characteristics for individual studies and by funnel plots to assess the risk of publication bias. Main Results: Tentative exposure-response relations for percent highly annoyed residents (%HA) in relation to noise levels for aircraft, road, rail, wind turbine and noise source combinations are presented as well as meta-analyses of correlations between noise levels and annoyance raw scores, and the OR for increase of %HA with increasing noise levels. Quality of evidence was assessed using the GRADE terminology. The evidence of exposure-response relations between noise levels and %HA is moderate (aircraft and railway) or low (road traffic and wind turbines). The evidence of correlations between noise levels and annoyance raw scores is high (aircraft and railway) or moderate (road traffic and wind turbines). The evidence of ORs representing the %HA increase by a certain noise level increase is moderate (aircraft noise), moderate/high (road and railway traffic), and low (wind turbines). Strengths and Limitations: The strength of the evidence is seen in the large total sample size encompassing the included studies (e.g., 18,947 participants in aircraft noise studies). Main limitations are due to the variance in the definition of noise levels and %HA. Interpretation: The increase of %HA in newer studies of aircraft, road and railway noise at comparable L_den levels of earlier studies point to the necessity of adjusting noise limit recommendations. Funding: The review was funded by WHO Europe.

Keywords: annoyance; environment; exposure-response; meta-analysis; surveys; traffic noise.

Figure 1

Flow-chart of the study selection process (following the PRISMA flow-chart, Moher et al. [14]). Selection criteria are explained in Supplementary Materials S3.

Figure 2

Scatterplot and quadratic regression of the relation between L_den and the calculated %HA for 12 aircraft noise studies, together with ERFs by Miedema and Oudshoorn ([4], red), and Janssen and Vos ([20], green). Notes: (1) The size of the data points corresponds to the number of participants in the respective study (size = SQRT(N)/10). (2) If two results from different studies fall on the same data point, the last point plotted may mask the former one. (3) The black curve is derived from aggregated secondary data, while the red and green curves are derived from individual data. In addition, the mathematical models used for establishing the three functions differ.

Figure 3

Scatterplot and regression lines of the relation between L_den and the calculated %HA for five “high-rate change” (red curve) and five “low-rate change” (black curve) airport noise studies, together with exposure-response function by Miedema and Oudshoorn ([4], green curve). Notes: (1) The size of the data points corresponds to the number of participants in the respective study (size = SQRT(N)/10). (2) If two results from different studies fall on the same data point, the last point plotted masks the former one. (3) The red and black curves are derived from aggregated secondary data, while the green curve is derived from individual data. In addition, the mathematical models used for establishing the three functions differ.

Figure 4

Meta-analysis of 15 aircraft noise studies, based on correlations between individual L_den or L_dn and annoyance raw scores, Random Effects Model. The right part of the graph contains a forest plot of the correlations and their respective 95% confidence intervals. The figures of the last row indicate the summary estimates.

Figure 5

ORs and 95% confidence intervals for the OR referring to a %HA increase by a 10 dB increase (from 50 to 60 dB L_den) aircraft noise. The right part of the graph contains a forest plot of the ORs and their respective 95% confidence intervals. The figures of the last row indicate the summary estimates.

Figure 6

Scatterplot and quadratic regression of the relation between L_den and the calculated %HA for 25 road traffic noise studies (black line), together with the exposure-response function by Miedema and Oudshoorn [4] (red line). Notes 6: (1) Black symbols refer to valley studies, red symbols refer to Asian studies, and green symbols refer to European no-valley studies. (2) The size of the data points corresponds to the number of participants in the respective study (size = SQRT(N)/10). (3) If two results from different studies fall on the same data point, the last point plotted may mask the former one. (4) The black curve is derived from aggregated secondary data, while the red one is derived from individual data.

Figure 7

Quadratic regressions of the relation between L_den and the calculated %HA for 25 road traffic noise studies, (“full WHO data set”, black) vs. 10 studies (dashed green, same data set excluding the Alpine and Asian studies). For comparison, the exposure-response function by Miedema and Oudshoorn ([4], road) is shown (red), together with the respective confidence interval. Note: The black and green curves are derived from aggregated secondary data, while the red curve is derived from individual data.

Figure 8

Meta-analysis of 21 studies using Pearson correlations between L_den or L_dn and road traffic noise annoyance raw scores. The right part of the graph contains a forest plot of the correlations and their respective 95% confidence intervals. The figures of the last row indicate the summary estimates.

Figure 9

ORs and 95% confidence intervals for the observed “highly annoyed” increase by 10 dB increase (from 50 to 60 dB or 55 to 65 dB L_den or L_dn) road traffic noise. The right part of the graph contains a forest plot of the ORs and their respective 95% confidence intervals. The figures of the last row indicate the summary estimates.

Figure 10

Scatterplot of the relation between L_den and %HA including ten railway noise studies. The quadratic regression (black line) was calculated excluding the Shinkansen data. In addition, the exposure-response function by Miedema and Oudshoorn ([4], railway, red curve) is shown together with the confidence interval. Notes: (1) The size of the data points corresponds to the number of participants in the respective study (size = SQRT(N)/10). (2) If two results from different studies fall on the same data point, the last point plotted may mask the former one. (3) The black curve is derived from aggregated data, while the red one is derived from individual data.

Figure 11

Meta-analysis of eight studies using Pearson correlations between L_Aeq,24h and railway noise annoyance raw scores. The right part of the graph contains a forest plot of the correlations and their respective 95% confidence intervals. The figures of the last row indicate the summary estimates.

Figure 12

Odds Ratios and 95% confidence intervals from seven studies (based on observed data) for the increase of the rate of “highly annoyed” persons from 50 to 60 dB L_Aeq,24h railway noise. The right part of the graph contains a forest plot of the ORs and their respective 95% confidence intervals. The figures of the last row indicate the summary estimates.