It's not just noise: The consequences of inequitable noise for urban wildlife

Spatial Analysis of Urban Noise Pollution: We conducted a spatial analysis of the distribution of noise pollution across HOLC grades for 83 cities in the United States (data in HOLC_Noise_City_results.csv). To be included in the study, the city needed to be included in both datasets used in the analysis: 1) the Mapping Inequality Project dataset on the distribution of HOLC grades across cities, and 2) the U.S. Department of Transportation, National Transportation Noise Map 2018. Any cities in which the distribution of HOLC grades did not include all four grades (A-D) were excluded from the analysis, which largely excluded cities with population sizes below 100,000 people. To evaluate noise exposure across HOLC grades for each city in our study, we acquired spatial data on the distribution of HOLC grades across U.S. cities from the Mapping Inequality Project. We also acquired data on road, rail, and aircraft noise (hereafter transportation noise models), from the U.S. Department of Transportation, National Transportation Noise Map (2018). The transportation noise models represent potential exposure to transportation noise reported on a decibel scale in a 30m x 30m pixel resolution. Here noise represents the average noise energy produced by road, rail, and aviation networks over a 24-hour period, measured in A-weighted decibels (dBA) (LAeq, 24h) at sampling locations deployed across a uniform grid in each city at an elevation of 1.5 m above ground level. Noise levels below 35 dBA are assumed to have minimal negative impacts to humans and the environment and thus are represented with null values in the transportation noise models. For each HOLC grade and each city, we used zonal statistics in ArcGIS Desktop v. 10.7 to summarize the median noise levels and area covered by excess noise (i.e., values > 35 dBA). We used the resulting zonal statistics estimates and the formula from Collins et al. (2019) to calculate an area-corrected measure of excess noise: N = (r * Md)/a where N is excess noise in each HOLC grade (with units of dBA/900m2); r is the area covered by the 30m x 30m pixels with noise values >35 dBA across all polygons of the same HOLC grade in each city; Md is the median transportation noise value (in dBA) for those same pixels; and a is the total area of all polygons of the same HOLC grade in each city. Thus, N represents a measure of both the level of noise and the area covered by excess noise in a given HOLC grade for each city. Literature Review on the Impacts of Noise to Urban Wildlife: To assess the effects of noise on wildlife in urban environments, we conducted a literature review using Thompson’s ISI Web of Science and adapting the methods of Shannon et al. (2016). We adjusted of Shannon et al.’s search criteria to include urban phrases, resulting in the following search terms (TS=(WILDLIFE OR ANIMAL OR MAMMAL OR REPTILE OR AMPHIBIAN OR BIRD OR FISH OR INVERTEBRATE) AND TS=(NOISE OR SONAR) AND TS=(CITY OR *URBAN OR METROPOLITAN)). We only selected papers published between 1990 and 23 June 2021 (i.e., the date we conducted our search) within the ISI Web of Science categories of ‘Acoustics’, ‘Zoology’, ‘Ecology’, ‘Environmental Sciences’, ‘Ornithology’, ‘Biodiversity Conservation’, ‘Evolutionary Biology’, and ‘Marine Freshwater Biology’. This returned 691 peer-reviewed papers, which we filtered so only empirical studies focused on documenting the effects of anthropogenic noise on wildlife in urban or suburban ecosystems or the effects of urban noise on wildlife in rural environments were included in the final data set (n = 207). We excluded reviews, meta-analyses, methods papers, and research that took place outside of urban or suburban areas where the noise was not explicitly denoted as urban (e.g., omitted studies that measured traffic noise by parks and reserves in rural areas). For the 241 articles previously analyzed in of Shannon et al. (2016), one of our authors reviewed each paper to determine which studies were focused on urban noise (n = 46). We then verified whether there were significant biological responses to a particular noise level threshold, noting each noise level if multiple biological responses were recorded. We recorded responses to noise into one of eight possible biological response categories, many of which were taken or modified from the biological response categories utilized in Shannon et al. (2016). The following were the biological response categorical values: movement behavior, vocal behavior, physiological, population, mating behavior, foraging behavior, vigilance behavior, life history / reproduction, and ecosystem. For any new articles published since the Shannon et al. (2016) dataset (n = 354) or those published between 1990 and 2013 but not reviewed by Shannon et al. (n = 96), two of our authors reviewed each paper to first determine which studies met our criteria (n = 161) and then compiled data on a number of variables of interest, including the noise levels and their resulting biological responses that were statistically significant. For this subset of papers, one author was randomly assigned a list of papers and then a second author was randomly assigned to assess the accuracy of the data collected by the first author. Any discrepancies were discussed as a group until an agreement was reached. Noise categories (environmental, transportation, industrial, multiple, other) were chosen for each paper by noting the explicitly stated source or description of urban noise described in the methodology. Noise levels and their units were reported for each paper, with only noise levels reported in decibels (dB) being used in data analysis. All terrestrial papers used a reference pressure of 20 microPascals (μPa). Due to the low sample size of aquatic studies (n = 4), differences in reference pressures, and varying sound intensities amongst aquatic studies, we only included terrestrial studies in statistical analyses and figures. We recorded the sound metric used (i.e., SPL, SPL Max, Leq) for each paper, but were unable to convert the various sound metrics given to a single sound metric for standardization during analysis. Thus, there were various sound metrics used in the analysis of the data extracted from the literature search, in particular for the cumulative weight-of-evidence curve, which poses a limitation in the comparison of noise levels amongst papers. Additionally, we recorded the weightings for each noise level, with many of the papers being A-weighted (dBA; n = 100) and Z-weighted (dBZ; n = 4). These weightings relate to typical characteristics of sounds as observed by humans. Many papers, however, did not record the weighting and/or the exact sound metric used.