Landscape patterns of avian habitat use and nest success are affected by chronic gas well compressor noise
Anthropogenic noise is becoming a dominant component of soundscapes across the world and these altered acoustic conditions may have severe consequences for natural communities. We modeled noise amplitudes from gas well compressors across a 16 km2 study area to estimate the influence of noise on avian habitat use and nest success. Using species with noise responses representative of other avian community members, across the study area we estimated gray flycatcher (Empidonax wrightii) and western scrub-jay (Aphelocoma californica) occupancy, and gray flycatcher nest success, which is highly dependent on predation by western scrub-jays. We also explore how alternative noise management and mitigation scenarios may reduce area impacted by noise. Compressor noise affected 84.5% of our study area and occupancy of each species was approximately 5% lower than would be expected without compressor noise. In contrast, flycatcher nest success was 7% higher, reflecting a decreased rate of predation in noisy areas. Not all alternative management and mitigation scenarios reduced the proportion of area affected by noise; however, use of sound barrier walls around compressors could reduce the area affected by noise by 70% and maintain occupancy and nest success rates at levels close to those expected in a landscape without compressor noise. These results suggest that noise from compressors could be effectively managed and, because habitat use and nest success are only two of many ecological processes that may change with noise exposure, minimizing the anthropogenic component of soundscapes should be a conservation priority.
KeywordsAnthropogenic noise Occupancy patterns Natural gas well compressor Nest success Soundscape
All landscapes include soundscapes, which are the dynamic acoustic environments that characterize different locations (Schafer 1977). These sounds may include biological sounds, sounds from moving abiotic features (e.g., wind, rain, water), as well as anthropogenic sounds. A recent surge of studies investigating the influence of anthropogenic noise (henceforth “noise”) on wildlife has shown that noise may have severe negative consequences for a diverse array of taxa (Barber et al. 2010). Though evidence for impacts of noise on natural communities is growing, and the scale to which noise penetrates natural habitats is extensive (Barber et al. 2010), to date, studies have not examined ecological changes in response to noise at broad-scales.
There have been two main research directions involving consequences of noise on wildlife, but neither has focused on broad-scale patterns (i.e. >100 s ha). First, a number of studies have focused on how noise disrupts acoustic communication by masking acoustic signals and, in turn, how animals adjust signals to mitigate the masking effects of noise (e.g., Slabbekoorn and Peet 2003; Gross et al. 2010). These studies have primarily focused on birds, but amphibians and terrestrial mammals have also received attention (e.g., Rabin et al. 2006; Egnor et al. 2009; Parris et al. 2009). A second direction has involved the influence of noise on species abundances and densities. A long line of studies suggests that noise from roadways may exclude species from otherwise suitable areas (e.g., Reijnen et al. 1995; Forman et al. 2002; Rheindt 2003), but most findings have been confounded by uncontrolled variables associated with human-generated noise. For example, noise caused by traffic co-varies with changes in vegetation, edge effects, moving vehicles, pollution intensity, and mortalities from animal–vehicle collisions. In urban areas, many of these same features may change as acoustics vary across the landscape, but species common to urban areas have broader environmental tolerances than species that avoid urban areas (Bonier et al. 2007); therefore, responses by these species may not represent typical responses within a taxon. More recent studies that controlled for confounding factors associated with urban and roadways, and without focusing on species common to urban areas, have shown that noise may not only reduce bird habitat use (Bayne et al. 2008; Francis et al. 2011) and pairing success (Habib et al. 2007), but may also change avian communities and predator–prey interactions (Francis et al. 2009).
A critical step to understanding the full impacts of the anthropogenic component of soundscapes requires that local scale studies are properly scaled to characterize the magnitude of noise disturbances on a landscape-level. Only then can we fully appreciate the overall impacts of noise on ecological processes and be able to determine best management practices that can realistically be implemented to mitigate the negative effects of noise.
Here, we aim to (i) quantify the anthropogenic component of a soundscape resulting from gas well compressor noise (see below for details), (ii) pair landscape-level acoustic patterns with avian habitat use and nest success, and (iii) evaluate how acoustic and ecological patterns change under alternative energy extraction management practices and mitigation scenarios. First, we evaluate several candidate models that predict compressor noise amplitude with respect to distance from the source and use the best model to map noise amplitudes across a 16 km2 study area representative of our study region. Our goal is to provide amplitude values that are representative of the region as a whole. Second, to illustrate how compressor noise can alter ecological processes across a large area, we use predictions from our noise amplitude model to calculate broad-scale occupancy rates for two species, the gray flycatcher (Empidonax wrightii) and western scrub-jay (Aphelocoma californica). We also evaluate nest success for gray flycatchers, which is highly dependent on predation by western scrub-jays (Francis et al. 2009). Finally, we examine how several alternative mitigation and management practices may restore a more natural soundscape by decreasing the proportion of the landscape impacted by compressor noise.
Study area and noise measurements
For previous studies, we had used the arrangement of well pads with noise-generating compressors (treatment sites) and well pads without compressors (control sites) to determine the influence of compressor noise on natural avian populations and communities. Unlike studies along roadways (Reijnen et al. 1995; Forman et al. 2002; Rheindt 2003) or in urban areas (Nemeth and Brumm 2009), human activity and vegetation does not differ on and around well pads with and without compressors (Francis et al. 2009). Therefore, effects of noise are separated from other confounding variables that complicate numerous studies that have investigated the influence of noise on wildlife along roadways or along urban gradients. Additionally, because noise can severely bias an observer’s ability to locate birds (Pacifici et al. 2008), compressors were turned off for 2 h during surveys and nest searches so that our ability to detect birds and find nests would be similar between sites with and without compressors.
We chose our study area extent because it is representative of well (≈1.78 ± 0.34 SD/km2) and compressor densities (0.81 ± 0.25 SD/km2) throughout RCHMA (Francis, unpublished data) and may serve as initial values to estimate the full extent of noise exposure in the gas-producing region. Our study area included 28 active wells (1.74 per km2) and 12 active compressors (0.75 per km2) in 2005 (Fig. 1). Woodlands surrounding only three of these wells were used for our nesting study (Francis et al. 2009), and roughly half were used for related studies (Francis et al. 2011; Francis, unpublished data). In woodlands surrounding these study sites, plus that of over 70 wells outside the area under consideration here, we measured compressor noise amplitudes and at a subset of these sites, we also recorded background noise using a Marantz PMD 660 Digital recorder using a directional shotgun microphone (Audio-technica AT-815). For all measurements, the distance from the nearest compressor (on treatment sites) or wellhead (on control sites) was recorded. Noise amplitude measurements were taken with NIST certified sound pressure meters (Casella® model CEL 320 and CEL 1002 converter) for approximately 2 min and, for most locations, on three separate occasions (different days and times) to control for the effects of atmospheric variability on amplitude values. Measurements were discarded and retaken when aircraft noise was audible, when birds were vocalizing within ≈30 m, which could bias measurements, and when wind conditions reached category three (≈13–18 km/h) on the Beaufort Wind Scale. At each location, we measured mean amplitude (equivalent continuous noise level (Leq), fast response time) with A- and C-weighting. Here, we used C-weighted decibels (dB(C)) values in all analyses because A-weighting filters acoustic energy below 1.0 kHz, which includes frequencies used by the western scrub-jay for communication (Curry et al. 2002), but also by other members of the avian community, including, but not limited to, the mourning dove (Zenaida macroura; Francis et al. 2009) and brown-headed cowbird (Molothrus ater; Lowther 1993).
Noise amplitude models
To estimate noise amplitudes across the study area, we used mean noise amplitude measurements from 1140 individual locations in piñon-juniper woodlands ranging from 2 to 517 m from gas wells (on control sites) and compressors (on treatment sites) near 86 different wells. Woodland terrain surrounding each well varied from relatively flat to terrain with small hills, canyons and rocky outcroppings. The typical ambient noise amplitude uninfluenced by compressor noise was assigned as 55.00 ± 0.14 SE dB(C), calculated as the mean value from measurements at 541 locations surrounding quiet control sites. Hereafter we refer to this value as baseline amplitude. To characterize those areas where noise amplitudes have been increased due to compressor noise, we evaluated the strength of support for several general models predicting noise as various functions of distance from the source. Our candidate models considered noise amplitude as (i) a linear function of distance (linear), (ii) a power function of distance (power, which was distance raised to an attenuation coefficient), (iii) a natural logarithm of distance (loge), (iv) an exponential decay function of distance (edistance), (v) a sound propagation model based on spherical spreading and attenuation (SA), and (vi) a null intercept only model in which noise was unrelated to distance from the source (null). For all models considered (except null), we estimated not only the distance coefficient, but also the amplitude at the source because we lacked noise measurements at a distance of zero meters.
For our model selection procedure, we used an information-theoretic approach to evaluate support for competing candidate models (Burnham and Anderson 2002) with Akaike’s Information Criterion corrected for small sample sizes (AICc). We ranked models based on differences in AICc scores (∆AICc). Candidate models with ∆AICc scores within two of the best model were considered to have strong support, and those within four ∆AICc were considered to have some support. All analyses were performed in R (R Development Core Team 2009).
Noise, occupancy, and nest survival spatial analyses
Utilizing the best supported noise model we computed the area affected by compressor noise within the study area. Compressors that could have sound waves that extend into the study area but were not located within its boundaries were excluded from the analysis. Noise amplitudes were mapped as multiple circular concentric buffers from the compressor location using 2 dB(C) intervals and the area affected by each amplitude bin was calculated. Subsequently, we computed the cumulative proportion of area affected by each of these amplitude intervals. However, for visual clarity we used 5 dB(C) intervals for generating the maps. Because baseline ambient noise amplitude in the study area is 55 dB(C), we used this value as the sound matrix in which noise from the compressors were embedded.
Our noise measurements were taken from over 1,000 locations surrounding 86 wells situated in areas that varied in terrain, thus the data from which the models were derived implicitly accounted for variation in topography that influences sound propagation. For obvious reasons, this modeling approach is more general than an approach that explicitly accounts for topography and likely results in noise exposure predictions that differ from a topographically explicit model. Yet the general modeling approach reflects our goal to approximate the extent of noise exposure across RCHMA without explicitly accounting for variation in terrain. Additionally, there were some occasions when our models predicted that small areas were exposed to noise from two or more compressors. In the majority of these cases (>90%), the increase in amplitude was smaller than the 2 dB(C) steps that we used as our noise exposure intervals; therefore, we did not combine noise amplitudes from multiple sources and the values we report here are conservatively low. All spatial analyses were conducted using ArcGis 9.3 (ESRI 2008).
Alternative management and mitigation scenarios
The alternative noise management and mitigation scenarios we considered involved (i) sound barrier walls around compressors located on existing wells, (ii) a central compression station resulting in a single point source of compressor noise (multiple tightly-grouped compressors) rather than many across the landscape (see Fig. 1), and (iii) a central compression station surrounded by sound barrier walls.
Noise mitigation with sound barrier walls exists in a few areas of RCHMA where wells are adjacent to residential property (C. D. Francis, pers. obs.). Though compressors with sound barrier walls are often encased on all four sides, in some cases barriers encase compressors on only three sides. We measured amplitude at 30 m behind walls on three sides and on one side lacking the barrier and found that amplitudes on open sides (81.11 ± 0.61 SE dB(C), n = 4 wells) were similar to those measured on compressors lacking barriers (81.94 ± 0.55 SE dB(C), n = 9 wells; two-tailed t = 0.92, df = 11, P = 0.38) and that amplitudes on sides with walls were roughly 10 dB(C) lower than open sides (71.23 ± 0.56 SE dB(C), n = 4 wells; two-tailed t = 17.76, df = 6, P < 0.001). Based on these measurements, we mapped noise amplitudes assuming that noise barrier walls reduce noise by 10 dB(C) at a distance of 30 m from the compressor on all sides (assuming four-sided noise barriers). Here we used the top-model describing noise amplitudes but adjusted the noise at the source to achieve a 10 dB(C) reduction at 30 m.
Finally, use of noise barrier walls on compressor stations is also realistic. In a third scenario, we assume noise levels can also be reduced by 10 dB(C) at 30 m from the central compressor station and map the results with the top-model.
Noise amplitude models
Results from model selection procedure describing noise amplitude (dB(C)) as a function of distance from the source using Akaike’s information criterion adjusted for small sample size
Of the three management scenarios, the alternatives with noise-reducing walls around individual wells with compressors and a central compression system with walls resulted in the largest areas with baseline noise amplitudes (86.3 and 80.2%, respectively; Figs. 4 and 5). These scenarios also had the smallest proportions of the landscape where the listening area was reduced by 50% or more (6.3% for individual compressors with walls and 9.2% for central compression with walls). The scenario with central compression without walls resulted in 100% of the spatial extent exposed to amplitudes above baseline values with over 58% of the area exposed to amplitudes at least double the sound pressure of the baseline level. Additionally, over 95% of the landscape had a reduced listening area of at least 50%.
Anthropogenic noise now reaches an unprecedented proportion of terrestrial landscapes, emanating from urban centers, aircraft, transportation networks, motorized recreation, and natural resource extraction (Barber et al. 2010). To our knowledge, this is the first study to investigate the spatial extent of chronic noise exposure from natural resource extraction and to link this exposure to ecological patterns that change in response to noise. The distances at which noise from a single compressor may impact birds in our study system are comparable to those reported by previous studies. For example, Bayne et al. (2008) reported that noise could affect bird communities up to 700 m from compressor stations in boreal forests of Alberta, Canada. Our results suggest that the impact from a single compressor without noise-reducing walls may be at least 700 m based on C-weighted values and nearly this far for A-weighted values. Regardless of the decibel weighting used, noise exposure from a compressor station lacking noise-reducing walls will be much farther than from a single compressor. Under 2005 conditions, compressor noise reached over 80% of the study area, but the use of noise-reducing walls around existing compressors could lessen the spatial extent impacted by over 70%.
Despite the predicted effectiveness of noise-reducing walls at decreasing the area exposed to compressor noise, our results for broad-scale estimates for gray flycatcher and western scrub-jay occupancy would increase by only 5% and flycatcher nest success would be reduced by 8%. These apparently small changes occur because amplitudes within 5 dB(C) of baseline levels dominate the landscape under current management practices. Taken individually, these small improvements in occupancy may appear trivial compared to the relatively high costs associated with reducing noise amplitudes (see Bayne et al. 2008 for estimated costs of reducing compressor noise by 4 dB). However, it is critical to stress that reduced habitat use in response to noise is not restricted to the two species under consideration here, but that over one dozen species appear to be noise sensitive in our study area (Francis et al. 2009). Even though the analysis of flycatcher nest success was conducted independently of scrub-jay occupancy, it must be noted that jays depredate nests of many avian species; therefore, the increased nest predation risk associated with higher jay occupancies will be spread over dozens of bird species, many of which will also experience increased occupancy with decreased noise. Moreover, occupancy and nest success are only two ecological patterns that may be influenced by noise exposure and many additional processes may change in response to noise (see below).
Ecological changes from noise exposure
Increases in nest success with increased noise amplitudes is not unique to flycatchers in our study area; similar patterns were observed for spotted towhees (Pipilo maculatus), chipping sparrows (Spizella passerina), and the pooled breeding bird community (Francis et al. 2009). Though we show how flycatchers may have increased rates of nest success throughout RCHMA due to the negative effect of noise on the scrub-jay, a major nest predator in the area, this pattern may not exist in other landscapes exposed to noise where different nest predators may fail to respond to noise, or even respond positively, which may be the case for those nest predators that rely primarily on olfactory and visual cues to locate prey.
Though birds nesting in areas exposed to compressor noise may benefit from decreased predation risk, which results in higher nest success, this single measure of reproductive success does not fully capture the full range of trade-offs that may occur for nesting birds exposed to noise. For example, great tit (Parus major) clutch size and the subsequent number of hatchlings and fledglings declines with increased noise exposure (Halfwerk et al. 2011). Additionally, males may experience decreased pairing success relative to those in quiet areas. In Alberta, Canada, territory-holding male ovenbirds (Seiurus aurocapilla) were 17% less likely to successfully pair with a female in areas adjacent to compressor stations relative to males holding territories in areas adjacent to quiet well pads (Habib et al. 2007). A similar pattern was observed for reed buntings (Emberiza schoeniclus) in Zurich, Switzerland (Gross et al. 2010). The exact mechanisms responsible for the reduced pairing success are unknown but may be due to females’ decreased ability to detect and discriminate signals masked by noise, or, alternatively, the pattern may be an epiphenomenon of intraspecific competition. For instance, a greater proportion of territory-holding male ovenbirds in noisy areas were young and inexperienced relative to those in quiet areas (Habib et al. 2007). Because older, experienced males often acquire better territories (Holmes et al. 1996) and have higher pairing rates (Saether 1990; Bayne and Hobson 2001), younger territory-holding males in noisy areas may be perceived as low quality mates relative to those in quiet areas. Studies determining whether reduced pairing success in noisy areas is typical across species and identifying the precise mechanisms responsible for this reduction are urgently needed.
That breeding birds may potentially benefit from noise in one sense, such as through reduced predation risk or a competitive release due to decreased densities of species less tolerant of noise (Francis et al. 2009; Slabbekoorn and Halfwerk 2009), but also suffer from detrimental effects, such as reduced pairing success or clutch size, limits our ability to quantify the net impact of noise on populations and communities. In terms of avian reproductive success, the way in which each breeding variable (e.g., breeding occupancy, pairing success, clutch size, nest predation) responds to changes in noise amplitude and frequency will probably differ. For example, territory occupancy may have a near linear response to amplitude, pairing success may be described by a threshold response, and other stages may be described by different linear and nonlinear functions. Unfortunately, our ability to understand how incremental changes to noise conditions influence these responses is restricted by the common use of study designs representing two extreme points on what is actually a continuous gradient of noise exposure (i.e., quiet vs. noisy areas) (Habib et al. 2007; Nemeth and Brumm 2009; Gross et al. 2010). Studies evaluating how processes respond to incremental changes to noise conditions are needed to determine how effects interact to influence population recruitment and long-term stability in landscapes exposed to noise. Finding locations that permit the isolation of noise from other confounding stimuli to examine noise along amplitude and/or frequency gradients will be challenging, but critical to efforts scaling up population and community responses to noise to a broad-scale.
Another gap in our current understanding is how noise may influence animal movements. On fine spatial scales, con- and heterospecific acoustic signals are critical for spatial perception that may guide individuals to mating partners and prey items, or repel them from competitors and predators (Slabbekoorn and Bouton 2008). On broad spatial scales, the use of acoustic cues for movements within a landscape (“soundscape orientation”) may guide movements at great distances. For example, tropical coral larvae are attracted to acoustic cues to locate reef substrate for settlement and anthropogenic noise may disrupt larvae ability to use sound for orientation (Vermeij et al. 2010). Similar problems are likely to exist in terrestrial environments because relatively small increases in noise amplitude may severely reduce an individual’s listening area. Where soundscapes are dominated by anthropogenic noise, it is likely that noise may present similar problems in terms of connectivity as do physically altered vegetation structure and the spatial arrangement of the soundscape becomes important. Among the management scenarios we considered, adding walls around individual compressors or around central compression stations would greatly increase natural soundscape connectivity.
Extent of noise in RCHMA and the San Juan basin and noise management
RCHMA has 400 km2 with federally owned mineral rights (BLM 2003). Extrapolating from our analyses, 338 km2 of RCHMA is exposed to elevated noise amplitudes due to compressors, and 244.4 km2 have a listening area reduced by one-half or more. These estimates are most certainly low because they do not include noise amplitudes from traffic along the dendritic network of access roads, existing compressor stations along major pipelines, or drilling activities from the 600 new wells drilled each year in BLM managed land in NW New Mexico (Engler et al. 2001; BLM 2003).
In 2003, BLM anticipated an increase of nearly 10,000 new wells throughout the 5,666 km2 of federally administered lands within the New Mexico portion of the San Juan Basin (BLM 2003), and half of the new wells are expected to have wellhead compression (Engler et al. 2001). This expansion in energy development is well underway and corresponds to considerable habitat loss and fragmentation for new well pads, pipelines and access roads. Compounding these changes to natural habitat with increased noise exposure from more compressors will likely only intensify the problems that arise due to these other anthropogenic forces. It is also important to recognize that this basin is not unique in this expected increase in energy extraction activities. Other landscapes throughout the United States and the world are slated for increases in gas extraction, such as the Marcellus Shale in New York, Pennsylvania and other eastern states (Kargbo et al. 2010) and the Barnett Shale in Texas (Bowker 2007). The anthropogenic component of soundscapes in these regions will almost certainly grow without management efforts.
For heavily developed landscapes with wellhead compression, use of noise-reducing walls may provide a good option to restore the natural soundscape and bring those ecological processes that had changed in response to noise back towards baseline levels. In landscapes at early stages of energy extraction development, central compression with noise-reducing walls may be a better option, especially if noise amplitudes are reduced by more than 10 dB(C) and if central compression may service many more wells than what we used in our model (12 compressors and 28 wells total). Given the uncertainties regarding the number of compressors needed within a compressor station and the potential effectiveness of noise-reducing walls, it is unclear how well this approach will reduce the extent of noise exposure, but it may be possible to reduce noise exposure to a much smaller proportion of the landscape. In all cases, it is clear that without noise-reducing walls, the industry’s soundscape footprint will be much larger and companies and land managers should work together to minimize the spatial distribution of this industry’s impact on natural communities and maintain some semblance of the natural soundscape.
We thank our research assistants for field and lab support, especially R. Kennedy, N. Kleist, and P. Nylander, and BLM Farmington Field Office Biologist J. Hansen. We also thank two anonymous reviewers for useful suggestions and comments on an earlier version of this manuscript. This study was primarily supported by NSF DDIG (# IOS 0910092), BLM, through the Colorado Plateau Cooperative Ecosystems Studies Unit, ConocoPhillips, Williams Energy, and University of Colorado Department of Ecology and Evolutionary Biology, plus the Animal Behavior Society, Southwestern Association of Naturalists, Colorado Field Ornithologists, Edna Baily Sussman Foundation, Mountain Studies Institute, and an Ariel Appleton Research Fellowship with support from the Wells Fargo Foundation. C.D.F. was also supported by the National Evolutionary Synthesis Center (NESCent, NSF EF-0905606).
- Bureau of Land Management [BLM] (2003) Farmington resource management plan with record of decision. Bureau of Land Management, US Department of Interior, Farmington Field Office, FarmingtonGoogle Scholar
- Burnham KP, Anderson DR (2002) Model selection and inference: a practical information-theoretic approach. Springer-Verlag, New YorkGoogle Scholar
- Curry RL, Peterson AT, Langen TA (2002) Western Scrub-Jay (Aphelocoma californica). In: Poole A (ed) The birds of North America online. Cornell Lab of Ornithology, Ithaca, NY. Available from http://bna.birds.cornell.edu.bnaproxy.birds.cornell.edu/bna/species/712. Accessed 22 January 2011
- Engler TW, Brister BS, Chen H-Y, Teufel LW (2001) Oil and gas resource development for San Juan basin. Report to Bureau of Land Management, Albuquerque Field Office, AlbuquerqueGoogle Scholar
- ESRI (2008) ArcGis 9.3. ESRI, RedlandGoogle Scholar
- Francis CD, Ortega CP, Cruz A (2011) Vocal frequency change reflects different responses to anthropogenic noise in two suboscine tyrant flycatchers. Proc Royal Soc B. doi:10.1098/rspb.2010.1847
- Gross K, Pasinelli G, Kunc HP (2010) Behavioral plasticity allows short-term adjustment to a novel environment. Am Nat 176. doi:10.1086/655428
- Lowther PE (1993) Brown-headed Cowbird (Molothrus ater). In: Poole A (ed) The birds of North America online. Cornell Lab of Ornithology, Ithaca, NY. Available from http://bna.birds.cornell.edu.bnaproxy.birds.cornell.edu/bna/species/047. Accessed January 2011
- Occupational Safety and Health Administration (OSHA) (2009). Available from http://www.osha.gov/SLTC/noisehearingconservation/index.html
- R Development Core Team (2009) R: a language and environment for statistical computing. R Foundation for Statistical Computing, ViennaGoogle Scholar
- Rheindt FE (2003) The impact of roads on birds: does song frequency play a role in determining susceptibility to noise pollution? J Ornithol 144:295–306Google Scholar
- Saether B-E (1990) Age-specific variation in reproductive performance of birds. Curr Ornithol 7:251–283Google Scholar
- Schafer RM (1977) The soundscape: our sonic environment and the tuning of the world. Destiny Books, Rochester, VermontGoogle Scholar
- Sterling JC (1999) Gray Flycatcher (Empidonax wrightii). In: Poole A, (ed) The birds of North America online. Cornell Lab of Ornithology, Ithaca, NY. Available from http://bna.birds.cornell.edu.bnaproxy.birds.cornell.edu/bna/species/458. Accessed September 2010