The following is the established format for referencing this article:Shonfield, J., and E.M. Bayne. 2023. Weak support for cumulative effects of industrial disturbance on three owl species in Alberta’s boreal forest. Avian Conservation and Ecology 18(1):9.
Human-caused disturbances are encroaching on natural areas and quantifying the relative effects of different types of disturbance, and whether they interact on a landscape to create additive or synergistic cumulative effects, will provide a better understanding of how wildlife are affected. We evaluated potential cumulative effects of industrial disturbance on habitat use of Barred Owls (Strix varia), Great Horned Owls (Bubo virginianus), and Boreal Owls (Aegolius funereus) in Alberta’s boreal forest using acoustic survey presence/absence data and boosted regression tree analysis to quantify the relative importance and interactions of different types of industrial disturbance, as well as forest age and composition. Barred Owls were more likely to be found in older mixedwood and deciduous forest, and we found evidence suggestive of cumulative effects from a negative effect of total human footprint on habitat use and an additional negative effect of roads. Great Horned Owls were found in older forest but were relatively tolerant of disturbance, and soft linear features (seismic lines, pipelines, transmission lines) had a positive effect on habitat use, possibly due to these types of disturbances creating suitable hunting habitat. Boreal Owls were more likely to be found in older coniferous forest, and the effects of disturbance did not show clear evidence of sensitivity or tolerance to human disturbance. Our results indicate the importance of forest age and composition on habitat use for these owls. Cumulative effects varied among owl species and were potentially more significant for Barred Owl; for the other two species the effects of industrial disturbance were relatively small and there was no evidence of cumulative effects. Our study also demonstrates that assessing cumulative effects of human disturbance on wildlife using boosted-regression trees can effectively help focus conservation efforts and can be used, for example, for evaluating the environment effects of new projects prior to their implementation.
Les perturbations d’origine humaine empiètent sur les milieux naturels et la quantification des effets relatifs des différents types de perturbations, ainsi que de leurs interactions dans le paysage engendrant des effets cumulatifs additifs ou synergiques, permettent de mieux comprendre comment les espèces sauvages sont affectées. Nous avons évalué les effets cumulatifs potentiels de perturbations industrielles sur l’utilisation de l’habitat par la Chouette rayée (Strix varia), le Grand-duc d’Amérique (Bubo virginianus) et la Nyctale de Tengmalm (Aegolius funereus) dans la forêt boréale de l’Alberta, à l’aide de données de présence/absence tirées de relevés acoustiques et d’une analyse par boosting d’arbres de régression afin de quantifier l’importance relative et les interactions des différents types de perturbations industrielles, ainsi que l’âge et la composition de la forêt. Les Chouettes rayées étaient plus susceptibles d’être trouvées dans les forêts mixtes et les forêts de feuillus âgées, et nous avons décelé des indices signalant des effets cumulatifs d’un effet négatif de l’empreinte humaine sur l’utilisation de l’habitat et d’un effet négatif supplémentaire des routes. Les Grands-ducs d’Amérique ont été détectés dans des forêts âgées, mais étaient relativement tolérants aux perturbations, et les éléments linéaires mineurs (lignes sismiques, pipelines, lignes de transmission) ont eu un effet positif sur l’utilisation de l’habitat, peut-être parce que ces types de perturbations créent un habitat de chasse approprié. Les Nyctales de Tengmalm étaient plus susceptibles d’être repérées dans des forêts de conifères âgées, et nous n’avons pas trouvé de signes clairs des effets des perturbations sur la sensibilité ou la tolérance de l’espèce aux perturbations humaines. Selon nos résultats, l’âge et la composition des forêts sont intrinsèquement liés à l’utilisation de l’habitat par ces hiboux. Les effets cumulatifs ont varié selon les espèces de hiboux et ont été potentiellement plus importants pour la Chouette rayée; pour les deux autres espèces, les effets des perturbations industrielles ont été relativement faibles et nous n’avons pas observé d’indices d’effets cumulatifs. Notre étude a également démontré que l’évaluation des effets cumulatifs des perturbations humaines sur la faune à l’aide d’arbres de régression boostés peut contribuer efficacement à cibler les efforts de conservation et peut être utilisée, par exemple, dans l’évaluation des effets environnementaux de nouveaux projets avant leur mise en oeuvre.
A variety of different sources of human-caused (anthropogenic) disturbance are encroaching on natural areas, and understanding the effects on wildlife is important for mitigating and minimizing the effects of disturbance. A new disturbance can introduce new stressors on wildlife, and it is important to understand if some sources of disturbance have larger effects than others and if they have additive or synergistic effects. Anthropogenic disturbance that results in a decrease or alteration of the structure of native vegetation is frequently measured as habitat loss and fragmentation. Alteration of habitat is an important factor influencing species occurrence and loss of habitat is commonly implicated in observed species declines (e.g., Quesnelle et al. 2013). In Canada, the number of species at risk in an ecozone has been found to increase with the extent of anthropogenic disturbance on the landscape (Kerr and Deguise 2004), and in Canada’s boreal forest, growing oil, gas, and forest industries are associated with increasing physical disturbance and habitat loss. Physical disturbances (e.g., roads with traffic) can also produce sensory disturbances (e.g., light, noise, and other forms of pollution) that can potentially increase the functional amount of habitat loss and reduce species occurrence (Longcore and Rich 2004, Barber et al. 2010).
Studies on the effects of anthropogenic disturbance frequently employ a reductionist approach, assessing the impact of a single type of disturbance on wildlife. For example, research on the effects of anthropogenic noise has been steadily increasing in the last two decades (Shannon et al. 2016), and some of these studies used controlled playback experiments, where noise is introduced in isolation of other forms of anthropogenic disturbance (McClure et al. 2013, Rosa et al. 2015, Ware et al. 2015, Cinto Mejia et al. 2019). This experimental approach is useful as it allows for the study of the effect of a single type of disturbance with other confounding variables removed. However, to understand how wildlife respond to multiple disturbance types present on landscapes and how they may interact additively or synergistically requires a different approach.
The concept of cumulative effects is that multiple disturbance types may have combined effects greater than the sum of separate effects. Quantifying cumulative effects is often touted as an important part of managing ecosystems and mitigating effects of disturbance, but there are a variety of definitions and possible approaches to assessing cumulative effects. Here, we use the definition that cumulative effects are the effects of stressors that overlap in space and/or time, caused by a single repeated stressor or multiple stressors (Davies et al. 2020). The concept of cumulative effects is that while each disturbance on its own may result in a small or negligible effect on wildlife, the accumulation of disturbances, or changes over time/within a region can result in a major impact on wildlife (Theobald et al. 1997). Evaluating cumulative effects of disturbance is key to resource management and wildlife conservation (Burton et al. 2014) and can help us to understand what is driving declines in species that are already at risk and potentially provide insight into which species may be at risk in the future.
Studies on the effects of anthropogenic disturbance on owls have been limited in scope, often focusing on one type of disturbance only. For example, several studies have focused on the effects of roads. Roads are a well-studied type of disturbance and have been found to have negative effects on a wide variety of species (Fahrig and Rytwinski 2009), and owls are no exception. Increased proximity to roads and increased traffic volume have been found to decrease owl density and occupancy of sites (Hindmarch et al. 2012, Silva et al. 2012), and traffic volume has been attributed to decreased fledging rates in Northern Spotted Owls (Strix occidentalis caurina) (Hayward et al. 2011). Noise from roads decreased the probability of occurrence of Tawny Owls (Strix aluco) in an urban region, but this effect was small compared to a stronger positive effect of the size of wooded areas on owl occurrence (Fröhlich and Ciac 2018). Conversely, some forms of anthropogenic disturbance have the potential to provide good habitat for owls. Some species of owls are known to inhabit and nest in urban areas (White et al. 2018, Clement et al. 2019), where prey can be abundant, and there can be numerous nesting opportunities (e.g., corvid nests and nest boxes). Some owl species have become closely associated with human-built structures, for example, the Barn Owl (Tyto alba) commonly nests in barns and other old buildings (Johnsgard 2002). Different types of anthropogenic disturbance have the potential to be both beneficial and detrimental for owls, depending on the species; however, the consequences of human activities on owl occurrence still needs to be evaluated to improve our understanding.
In the boreal forest of northeastern Alberta, the presence of several industries, including the oil and gas industry and forestry industry, has resulted in a landscape where multiple types of disturbance exist (e.g., roads, cutblocks from forestry operations, and oil processing facilities from the energy sector). Owls may be sensitive to industrial disturbance because they are highly mobile with large territories and, therefore, more likely to encounter disturbance; in addition, previous studies have found negative effects of roads and habitat loss on owls. It is important to assess the potential for cumulative effects on owls and whether certain types of disturbance are more influential on where owls are found on the landscape. Evaluating cumulative effects of different types of industrial disturbance on owl habitat use, and how species may respond differently, is important for conservation, environmental assessments, and overall landscape management. Our objective for this study is to evaluate potential cumulative effects and assess the relative importance of different types of industrial disturbance on habitat use of three common owl species in boreal Alberta: the Barred Owl (Strix varia), the Boreal Owl (Aegolius funereus), and the Great Horned Owl (Bubo virginianus).
We studied owls in the boreal forest of northeastern Alberta, within the Lower Athabasca Planning Region (LAPR). The LAPR has seen increased development in the oil and gas industry in recent years, and there are forestry operations in the region (Alberta Biodiversity Monitoring Institute 2017). The amount of human footprint varies across the LAPR; some areas have large patches of contiguous habitat and in other areas different types of disturbance are clustered closely together (Fig. 1). We selected survey locations to sample across a gradient of total disturbance. We surveyed for owls at locations in upland forested areas in the LAPR, specifically within an area south of Fort McMurray, north of Lac la Biche, and northwest of Cold Lake (Fig. 1). Forests in the study area were composed primarily of trembling aspen (Populus tremuloides), white spruce (Picea glauca), and black spruce (Picea mariana) trees. Elevation at the sampling locations ranged from 271 to 803 m, and the average elevation was 601 m.
Using autonomous recording units (ARUs), we conducted passive acoustic surveys for owls; specifically, we used SM2+ and SM4 Song Meters (Wildlife Acoustics, Inc., Maynard, Massachusetts, USA). We conducted surveys across four years from 2013–2016 in late winter–early spring when owls are actively calling in this northern boreal region. We deployed a single ARU at each survey location, spaced a minimum of 1.2 km apart and attached at a height of approximately 1.5 m on trees with a smaller diameter than the width of the ARU (18 cm for SM2+ units and 12 cm for SM4 units). We programmed units to record in stereo format at 44.1 kHz with a 16-bit resolution with gain settings of 48 dB for both the left and right channel microphones. A total of 389 ARUs were deployed specifically to detect owls in 2013, 2014, and 2015 and were set to record for 10 minutes at the start of every hour on a 24-hr basis; these owl specific ARUs recorded between the beginning of March and the end of May and were deployed for a minimum of 9 days at each survey location. Additionally, 12 ARUs used in this study in 2013 and 51 ARUs in 2016 were left out for a longer period (minimum 23 days in 2013 and 35 days in 2016) with the intent of surveying a wider variety of vocalizing species for multiple research objectives and had a reduced recording schedule to extend the battery life; these ARUs recorded for a minimum of 30 minutes between sunset and sunrise. From the owl specific ARUs deployed in 2013, 2014, and 2015 we have shown that 90% of recordings with Barred Owl calls, 97.5% of recordings with Great Horned Owl calls, and 99.3% of recordings with Boreal Owl calls occurred between sunset and sunrise (Shonfield and Bayne 2021). We tested each ARU and both microphones prior to deployment in the field each year to identify any units with non-responsive channels or degraded microphones. We tested microphone sensitivity yearly with a calibrator that produces a 1 kHz tone, any microphones that were below 10 dB than the standard value at the time of purchase, were physically damaged, or completely unresponsive were not used.
Processing acoustic data
We used automated species recognition (hereafter “recognizers”) to process acoustic recordings to detect territorial vocalizations of Barred Owls, Great Horned Owls, and Boreal Owls. Recognizers are an efficient method to obtain detections of owl calls from large volumes of audio recordings (Shonfield et al. 2018). We scanned all recordings collected from 2013 to 2016 using recognizers built in Song Scope (Wildlife Acoustics, Inc., Maynard, Massachusetts, USA) for each of these three owl species (Shonfield et al. 2018). Observers with knowledge of owl calls verified all hits generated by the program to filter out false positives. Once verified, we compiled the owl detection data to determine locations that were used and unused by each owl species.
We extracted data on human disturbance from Alberta Biodiversity Monitoring Institute’s (ABMI) Human Footprint layer 2014 version 2 (http://www.abmi.ca/home/data-analytics) using Geographic Information System (GIS) tools in ArcGIS 10.3.1 (Environmental Systems Research Institute, Inc., Redlands, California, USA). This polygon vector layer includes all types of anthropogenic disturbance across the province of Alberta, categorized into 115 feature types. These feature types include linear features (roads, seismic lines, pipelines, transmission lines, and railways), industrial and resource extraction features (well pads, compressor stations, processing plants, mines, and other facilities), and forest cutblocks. We grouped and recategorized all 115 feature types into the following six disturbance categories: industrial facilities, cutblocks, low activity clearings, high activity clearings, hard linear features, and soft linear features (Table 1). We also included an additional category for total human footprint, which was all feature types combined (Table 1). We based our decisions on assigning feature types to the new disturbance categories on the descriptions provided in the metadata file for the ABMI Human Footprint (Shonfield 2018, for more details on feature types and assigned categories). Since our dataset is from 2013–2016 and includes years before and after the 2014 ABMI Human Footprint layer, we verified for all sites in all years that the disturbances mapped matched what we had observed on the ground while setting up the ARUs.
We extracted variables on disturbance types, forest composition, and forest age in ArcGIS 10.3.1, using an 800-m radius buffer around each ARU station (an area of 201 ha), approximating the maximum detection radius of an ARU to detect owls calling (Yip et al. 2017). For each of the six categories of human disturbance (Table 1), we calculated the proportion disturbed in the buffer area from the ABMI Human Footprint layer. For forest composition, we calculated the mean proportion of coniferous trees present within forested areas from the Alberta Vegetation Inventory (AVI) for each 800-m buffer, this variable is naturally related to the proportion of deciduous and mixedwood forest. We also calculated mean forest age within forested areas in each 800-m buffer from the AVI layer.
We used boosted regression trees (BRTs) to assess the relative importance of human disturbance and other landcover variables in explaining habitat use by owls. BRTs are a machine-learning technique well-suited to modeling complex ecological data because they can handle different types of predictor variables, accommodate missing data, fit complex nonlinear relationships, and automatically handle interaction effects between predictors (Elith et al. 2008). We used BRTs to explore nonlinear effects and interactions between disturbance types as part of a cumulative effects analysis. We compiled detection data of each owl species for each location surveyed with an ARU from the recognizer data. We assumed that the detection of a species at a location indicated that the species used the area surrounding the ARU.
We ran BRTs in R version 3.4.3 (R Core Team 2019) using RStudio version 1.1.383 (RStudio Team 2018), implemented in the “gbm” package (Ridgeway 2017) and “dismo” package (Hijmans et al. 2017). A ten-fold cross-validation method was used to identify meta-parameter settings and to build models. To determine the optimal settings for the BRTs, we examined all possible combinations of a range of learning rates (0.005, 0.001, 0.0005, 0.0001) and tree complexity values (2, 3, 4, and 5) following Elith et al. (2008). The learning rate determines how much each successive tree contributes to the overall growing model, while tree complexity controls the depth of interaction between the explanatory variables. For each species, we selected the model with the combination of learning rate and tree complexity with the lowest residual deviance and at least 1000 trees and used a bag fraction of 0.75 for all models, following recommendations by Elith et al. (2008). A bag fraction of 0.75 means that at each iteration, 75% of the data are drawn at random, without replacement, from the full training set. The number of trees was determined by the lowest residual deviance across ten cross-validation folds. The output generated from the final model was used to examine the relative importance of the predictor variables for each owl species. We plotted the six predictor variables with the highest relative influence for each owl species, to examine the direction of the effects of the predictor variables with the strongest influence. We examined all pairwise interactions between predictor variables and reported those with an interaction strength greater than one.
For all three owl species, we included all six disturbance variables, forest composition, and forest age in the BRT model. For Barred Owls, we also included a variable for the presence/absence of Great Horned Owls as a potential competitor and predator that could influence the distribution of Barred Owls. For Boreal Owls, we included variables for the presence/absence of Barred Owls and Great Horned Owls, as both these larger owls are potential predators. We did not include variables for the presence/absence of the other two owl species for the Great Horned Owl model because they are typically dominant over other owl species.
We compiled recordings from ARUs at 452 unique locations surveyed between 2013 and 2016 (Fig. 1). Barred Owls were detected at 83 locations (18%), Great Horned Owls were detected at 227 locations (50%), and Boreal Owls were detected at 94 locations (21%). Survey locations varied in the proportion of the different types of human disturbance, proportion coniferous forest, and mean forest age (Table 2). Industrial facilities were the least common type of disturbance and were only present at 23% of locations surveyed, whereas soft linear features were the most common (99.8% of locations surveyed, Table 2). The types of disturbance with strong influence differed between owl species, but the proportion of coniferous forest and mean forest age were within the top four most influential variables in the BRT models for all three species (Figs. 2–4).
For each owl species, the final BRT model used a learning rate and tree complexity that resulted in the lowest residual deviance and at least 1000 trees. For Barred Owls, we used a learning rate of 0.0001 and a tree complexity of 3 (Table 3). The final model had 8800 trees and explained 13% of the variation in the data (Table 3). The strongest variable influencing use by Barred Owls was the proportion of coniferous forest (relative influence 34.4%). Partial dependency plots show a positive marginal effect on Barred Owl use in forests with 0% to 70% coniferous trees (i.e. deciduous and mixedwood forests) and a negative marginal effect on Barred Owl use once the forest composition exceeded 70% coniferous trees (Fig. 2). The second strongest variable was the proportion of hard linear features (relative influence 17.2%), followed closely by the proportion of total human footprint (relative influence 16.1%). Partial dependency plots indicate that Barred Owl use decreased with increasing proportion of hard linear features and with increasing total human footprint (Fig. 2). Interactions between variables were weak; only 4 two-way interactions had a multiplicative strength greater than one, and the strongest interaction was between the proportion of coniferous forest and total human footprint (Table 4). This interaction had a positive effect on Barred Owl use in more deciduous and mixedwood forests with low proportions of total human footprint. Other disturbance types (soft linear features, cutblocks, low and high activity clearings) had low relative influence (between 2.8% and 7%). The proportion of industrial facilities had the weakest influence of all the disturbance types (relative influence 0.8%), and the presence of Great Horned Owls had the weakest influence of all the variables tested on Barred Owl use (relative influence 0.06%).
For Great Horned Owls, we used a learning rate of 0.0005 and a tree complexity of 2 (Table 3). The final model had 3300 trees and explained 9% of the variation in the data (Table 3). The strongest variable influencing Great Horned Owl use was the proportion of soft linear features (relative influence 25.8%). Partial dependency plots showed that Great Horned Owl use increased as the proportions of soft linear features increased (Fig. 3). The second strongest variable was the proportion of industrial facilities (relative influence 17.8%), followed closely by mean forest age (relative influence 16.8%), and proportion coniferous forest (relative influence 15.7%). Partial dependency plots indicate a negative effect of increasing proportion of industrial facilities on Great Horned Owl use (Fig. 3). Forest age showed a positive marginal effect of forests 75-years-old and older, and the proportion of coniferous forest showed a negative marginal effect of forests greater than 80% coniferous on Great Horned Owl use (Fig. 3). Interactions between predictor variables were weak; the only two-way interaction with a multiplicative strength greater than one was between mean forest age and the proportion of soft linear features (Table 4). This interaction had a positive effect on Great Horned Owl use in older forests with moderate proportions of soft linear features.
For Boreal Owls, we used a learning rate of 0.001 and a tree complexity of 2 (Table 3). The final model had 5500 trees and explained 30% of the variation in the data (Table 3). Forest composition had the strongest influence on Boreal Owl use (relative influence 17.5%). Partial dependency plots indicate that Boreal Owls are more likely to be present as the proportion of coniferous forest increased (Fig. 4). Mean forest age was the second strongest variable (relative influence 16%), followed closely by the proportion of low activity clearings (relative influence 15%) and the proportion of high activity clearings (relative influence 14.6%). The effect of forest age shows a fluctuating pattern between 60 and 120 years old (Fig. 4). There were few locations surveyed in very young or very old forests, 87% of locations surveyed had a mean forest age between 60- and 120-years-old. We found positive effects of low and high activity clearings at low proportions, but as the proportion of these disturbances increased the effect became negative (Fig. 4). The majority (97%) of survey locations had less than 10% disturbed by low activity clearings, and similarly, 95% of locations had less than 10% disturbed by high activity clearings. The relative influence of Great Horned Owls was 12.6%. There was a negative marginal effect on Boreal Owl use when Great Horned Owls were absent and a positive marginal effect when they were present (Fig. 4). Great Horned Owls were present at 70% of the locations where Boreal Owls were present. Barred Owls had the weakest influence of all the variables on Boreal Owl use (relative influence 0.06%). Interactions between variables were weak; only 2 two-way interactions had a multiplicative strength greater than one, and the strongest interaction was between mean forest age and proportion of soft linear features (Table 4). This interaction had a positive effect on Boreal Owl use in more coniferous forests with low proportions of soft linear features.
We evaluated cumulative effects of anthropogenic disturbance on Barred Owls, Great Horned Owls, and Boreal Owls by using presence/absence data from acoustic surveys to assess the relative importance of disturbance types on owl habitat use, while also considering forest age and composition. Forest age and composition were important factors influencing habitat use by all three owl species and the results aligned with previous studies in this regard. Barred Owls were more likely to use deciduous and mixedwood forests, a habitat preference that is well documented in the boreal region (Mazur et al. 1998, Olsen 1999, Russell 2008). Boreal Owls were more likely to use areas with higher proportions of coniferous forest, similar to findings in other studies on Boreal Owl habitat use (Hayward et al. 1993, Lane et al. 2001). Forest age had a moderate influence on habitat use of all three owl species. We found a positive effect of forest age on Barred Owl and Great Horned Owl presence. Studies have shown that Barred Owls and Boreal Owls generally prefer older forests (Hayward et al. 1993, Mazur et al. 1998, Olsen 1999, Russell 2008). There is little evidence of Great Horned Owls showing any preference for forest age, but they sometimes use larger trees for nesting. For Boreal Owls, we found a positive effect of very old forests; similarly, a study in Finland found their survival increased with greater cover of old forest (> 80 years) (Hakkarainen et al. 2008). This highlights the importance of older forest to Boreal Owls and could be due to more potential nesting cavities in older coniferous trees than in younger coniferous trees.
The evidence for cumulative effects of disturbance for these three owl species was limited. Barred Owls showed a negative effect of total human footprint on their habitat use, as well as an additional negative effect of roads suggestive of cumulative effects albeit with a relatively limited effect size. Great Horned Owls were quite tolerant of disturbance, and we found that soft linear features (seismic lines, pipelines, transmission lines) had a positive effect on their habitat use but there was no evidence of cumulative effects of industrial disturbance. Boreal Owls did not show clear evidence that they are either sensitive or tolerant to human disturbance, and their habitat use was primarily influenced by forest age and composition.
Linear features are a common type of anthropogenic disturbance found in northern Alberta. Soft linear features, such as seismic lines, are narrow and long and thus disturb a small percentage of the landscape. However, they are ubiquitous in northern Alberta and their high densities contribute to forest fragmentation and increase the amount of edge habitat for some species (Pattison et al. 2016). Soft linear features had a weak influence on Barred and Boreal Owl use, and both showed a similar pattern of not using areas with a high proportion of soft linear features. Previous work, using some of the same owl acoustic survey data and a boosted regression tree analysis to predict spatial distribution of Boreal Owls across all of northern Alberta, found a similar result: soft linear features were included in 0–19% of territories and amounts over 4–5% had a negative effect on Boreal Owl presence (Domahidi et al. 2019). For Great Horned Owls, soft linear features had a strong positive influence on their habitat use. The vegetation on soft linear features is similar to meadow habitat, and small mammals favoring this type of habitat, such as meadow voles (Microtus pennsylvanicus), are abundant on linear features in the boreal forest (Darling et al. 2019). Great Horned Owls employ a perch-and-wait hunting style, and the edges of soft linear features likely provide ample perch locations. It is plausible that Great Horned Owls are able to benefit from seismic lines, pipelines, and transmission lines as favorable hunting areas. Hard linear features had a strong negative influence on Barred Owl use but low influence on Great Horned and Boreal Owl use. We included both roads and railways as hard linear features in our analysis, but there was only a single rail line in our study area, so we attribute effects of hard linear features primarily to roads. Multiple owl species are frequently killed on roads (Bishop and Brogan 2013) and road mortality of Barred Owls specifically has been found to increase on roads with higher speed limits and on roads in areas of higher habitat suitability (Gagné et al. 2015). However, why Barred Owl's habitat use was negatively affected by roads, but there was little effect on the other two owl species remains unknown.
Industrial facilities that produce chronic noise were present at less than a quarter of the locations we surveyed for owls. This disturbance type had low influence on Barred and Boreal Owl use, but it was the second most influential variable on Great Horned Owl use. Contrary to the present study, our previous work using occupancy analyses found that Barred Owl, Great Horned Owl, and Boreal Owl occupancy was unaffected by noise sources in our study area of northern Alberta (Shonfield and Bayne 2017). This previous study did not account for the size of industrial facilities (Shonfield and Bayne 2017) and the results here indicate the effect of facilities on Great Horned Owl use is negative only when the proportion of the area occupied by industrial facilities is ~15% of the area. Other studies have shown that the effect of noise may not be as important compared to other types of anthropogenic disturbance (Bernath-Plaisted and Koper 2016, Fröhlich and Ciac 2018, Nenninger and Koper 2018), and our results provide further evidence that this is likely the case for these three owl species. Other types of disturbance included forestry cutblocks, low and high activity clearings, and total human footprint. Forestry cutblocks were some of the largest disturbances, occupying up to 70% of the area at some survey locations. Cutblocks had a low influence on habitat use by the three owl species, but areas with high proportions of total human footprint usually included cutblocks. Total human footprint negatively influenced Barred Owl use, similar to previous published results (Shonfield and Bayne 2017), indicating that Barred Owls are sensitive to disturbances that result in the removal of trees from the landscape.
We included the presence of other owl species to account for possible spatial avoidance of larger owls as predators and competitors of the smaller owls, but we did not find any evidence of spatial avoidance between the three owl species studied at this scale. The presence of Barred Owls had little influence on Boreal Owl use, and there was a similar lack of effect of Great Horned Owl presence on Barred Owl use. Barred Owls and Great Horned Owls have been found to have overlapping home ranges where both species occur, but there may be avoidance through temporal partitioning (Laidig and Dobkin 1995). The presence of Great Horned Owls had some influence on Boreal Owl use, but it was the opposite of what we had predicted: The presence of Great Horned Owls had a positive effect on Boreal Owl use. This does not necessarily imply Boreal Owls are attracted to areas with Great Horned Owls, but it suggests Boreal Owls are not spatially avoiding Great Horned Owls, though they may be avoiding them temporally. Great Horned Owls are habitat generalists (Johnsgard 2002) and were detected at a majority of the locations used by Boreal Owls, so Great Horned Owls may be using habitats with adequate prey and suitable nesting and roosting sites that can also be suitable habitat for Boreal Owls. It is possible that Boreal Owls can reduce predation risk by short-term behavioral avoidance of Great Horned Owls in areas used by both species.
The variance explained by the model for each species was never higher than 30%, which may indicate that the overall importance of anthropogenic disturbance is low in its effect on where these owl species are found on the landscape. The fact that the models did not explain more than 30% of the variance in the data is problematic if we were using these models to predict where owls will be found on the landscape, but that was not the intention of this study. If it had been, we would have considered including additional variables describing vegetation characteristics and climatic conditions. Owl vocalization behavior related to movement patterns, diurnal, and seasonal patterns of vocalization, and intra- and inter-specific interactions all might influence whether or not an owl was observed at a location when ARUs were in place for only part of the season. The trade-off between sample size (locations) vs. time recording at one location is a challenge, and future studies should consider the potential value of leaving ARUs in place for longer periods. As more data from ARUs become available it will be important to reevaluate model accuracy. Future research could consider modeling the response variable as the total number of detections of a species, rather than presence/absence, to see if a more refined estimate is possible. We also looked only at one scale in terms of assessing cumulative effects on owls, so future research could also consider looking at different scales to see if different patterns of habitat use are evident.
We found minimal evidence for additive effects of disturbance for the variables selected in this study on the habitat use of the three species studied. The two-way interactions between predictor variables were weak, and the strongest interactions were not between disturbance types, and instead were between forest composition/age and a disturbance type. Owls are highly mobile with large territories, so it is possible that these traits allow them to tolerate disturbance by being able to move to less disturbed areas within their territories. It is also possible that some types of disturbances provide resources for owls, for example, small mammal abundance can be high on some types of soft linear features (Darling et al. 2019).
The results of this study contribute to our understanding of how anthropogenic disturbances influence where these three owl species are present in the landscape. The three owl species studied differed in their sensitivity to human disturbance. We found further evidence that Barred Owls are sensitive to human disturbance, with roads and total human footprint having negative effects on their habitat use. In contrast, Great Horned Owls were relatively tolerant of human disturbance and our results indicated that soft linear features have a positive effect on their habitat use. For Boreal Owls, our results did not provide clear evidence that they are either sensitive or tolerant to these forms of human disturbance. We found some evidence for cumulative effects of disturbance on Barred Owls from the result that total human footprint had a negative effect on their habitat use. However, the evidence for cumulative effects was minimal. Total human footprint was not a strong predictor of Great Horned Owl or Boreal Owl use, and we did not find strong interacting effects between disturbance variables. Evaluating cumulative effects can be difficult, but a boosted regression tree analysis enabled us to assess the relative importance of different types of disturbance on habitat use and explore interactions and nonlinear relationships between disturbance types as part of a cumulative effects analysis. Our study demonstrates an approach to assessing cumulative effects of disturbance in other regions and on other wildlife species. This approach could be useful for environmental assessments, where the goal is generally to assess environmental effects of new projects prior to the start of development. In addition, assessing the relative impacts of multiple types of disturbance, and how the accumulation of disturbances can affect wildlife habitat use, can help focus conservation efforts by prioritizing mitigation of disturbance types with the greatest impacts on species.
RESPONSES TO THIS ARTICLE
Responses to this article are invited. If accepted for publication, your response will be hyperlinked to the article. To submit a response, follow this link. To read responses already accepted, follow this link.
We thank N. Annich, M. Knaggs, A. MacPhail, L. McLeod, S. Wilson, and D. Yip for their assistance in the field. We thank C. Charchuk, M. Foisy, and S. Tkaczyk for their assistance developing the recognizers. We thank N. Boucher and the many students and volunteers who listened to recordings and checked the output of the recognizers. We thank H. Lankau for organizing the field recordings and maintaining the database. We thank E. Knight and two anonymous reviewers for providing helpful feedback on earlier versions of this manuscript. Funding was supported by the Natural Sciences and Engineering Research Council of Canada, the Northern Scientific Training Program, the University of Alberta North Program, the Alberta Conservation Association, the Environmental Monitoring Committee of the Lower Athabasca, Nexen Energy, and the Oil Sands Monitoring program operated jointly by Alberta Environment and Parks, and Environment and Climate Change Canada. This work was funded under the Oil Sands Monitoring Program and is a contribution to the Program but does not necessarily reflect the position of the Program.
Alberta Biodiversity Monitoring Institute. 2017. The status of human footprint in Alberta: preliminary report. https://ftp-public.abmi.ca/home/publications/documents/501_ABMI_2017_HumanFootprintReport_ABMI.pdf
Barber, J. R., K. R. Crooks, and K. M. Fristrup. 2010. The costs of chronic noise exposure for terrestrial organisms. Trends in Ecology and Evolution 25:180-189. https://doi.org/10.1016/j.tree.2009.08.002
Bernath-Plaisted, J., and N. Koper. 2016. Physical footprint of oil and gas infrastructure, not anthropogenic noise, reduces nesting success of some grassland songbirds. Biological Conservation 204:434-441. https://doi.org/10.1016/j.biocon.2016.11.002
Bishop, C. A., and J. M. Brogan. 2013. Estimates of avian mortality attributed to vehicle collisions in Canada. Avian Conservation and Ecology 8(2):2. https://doi.org/10.5751/ACE-00604-080202
Burton, A. C., D. Huggard, E. Bayne, J. Schieck, P. Sólymos, T. Muhly, D. Farr, and S. Boutin. 2014. A framework for adaptive monitoring of the cumulative effects of human footprint on biodiversity. Environmental Monitoring and Assessment 186:3605-3617. https://doi.org/10.1007/s10661-014-3643-7
Cinto Mejia, E., C. J. W. McClure, and J. R. Barber. 2019. Large-scale manipulation of the acoustic environment can alter the abundance of breeding birds: Evidence from a phantom natural gas field. Journal of Applied Ecology 56:2091-2101. https://doi.org/10.1111/1365-2664.13449
Clement, M. A., K. Barrett, and R. F. Baldwin. 2019. Key habitat features facilitate the presence of Barred Owls in developed landscapes. Avian Conservation and Ecology 14(2):12. https://doi.org/10.5751/ACE-01427-140212
Darling, A. F., L. Leston, and E. M. Bayne. 2019. Small-mammal abundance differs between pipelines, edges, and interior boreal forest habitat. Canadian Journal of Zoology 97:880-894. https://doi.org/10.1139/cjz-2018-0314
Davies, K. K., K. T. Fisher, G. Couzens, A. Allison, E. I. van Putten, J. M. Dambacher, M. Foley, and C. J. Lundquist. 2020. Trans-Tasman Cumulative Effects Management: A Comparative Study. Frontiers in Marine Science 7:1-19. https://doi.org/10.3389/fmars.2020.00001
Domahidi, Z., J. Shonfield, S. E. Nielsen, J. R. Spence, and E. M. Bayne. 2019. Spatial distribution of the Boreal Owl and Northern Saw-whet Owl in the Boreal region of Alberta, Canada. Avian Conservation and Ecology 14(2):14. https://doi.org/10.5751/ACE-01445-140214
Elith, J., J. R. Leathwick, and T. Hastie. 2008. A working guide to boosted regression trees. Journal of Animal Ecology 77:802-813. https://doi.org/10.1111/j.1365-2656.2008.01390.x
Fahrig, L., and T. Rytwinski. 2009. Effects of roads on animal abundance: an empirical review and synthesis. Ecology and Society 14(1):21. https://doi.org/10.5751/ES-02815-140121
Fröhlich, A., and M. Ciac. 2018. Noise pollution and decreased size of wooded areas reduces the probability of occurrence of tawny owl (Strix aluco). Ibis 160:634-646. https://doi.org/10.1111/ibi.12554
Gagné, S. A., J. L. Bates, and R. O. Bierregaard. 2015. The effects of road and landscape characteristics on the likelihood of a barred owl (Strix varia)-vehicle collision. Urban Ecosystems 18:1007-1020. https://doi.org/10.1007/s11252-015-0465-5
Hakkarainen, H., E. Korpimäki, T. Laaksonen, A. Nikula, and P. Suorsa. 2008. Survival of male Tengmalm’s owls increases with cover of old forest in their territory. Oecologia 155:479-486. https://doi.org/10.1007/s00442-007-0929-2
Hayward, G. D., P. H. Hayward, and E. O. Garton. 1993. Ecology of boreal owls in northern Rocky Mountains, U.S.A. Wildlife Monographs 124:3-59.
Hayward, L. S., A. E. Bowles, J. C. Ha, and S. K. Wasser. 2011. Impacts of acute and long-term vehicle exposure on physiology and reproductive success of the northern spotted owl. Ecosphere 2:65. https://doi.org/10.1890/ES10-00199.1
Hijmans, R. J., S. Phillips, J. R. Leathwick, and J. Elith. 2017. dismo: species distribution modeling. R package version 1.1-4.
Hindmarch, S., E. A. Krebs, J. E. Elliott, and D. J. Green. 2012. Do landscape features predict the presence of barn owls in a changing agricultural landscape? Landscape and Urban Planning 107:255-262. https://doi.org/10.1016/j.landurbplan.2012.06.010
Johnsgard, P. A. 2002. North American owls: biology and natural history, 2nd edition. Smithsonian Institution Press, Washington DC, USA.
Kerr, J. T., and I. Deguise. 2004. Habitat loss and the limits to endangered species recovery. Ecology Letters 7:1163-1169. https://doi.org/10.1111/j.1461-0248.2004.00676.x
Laidig, K. J., and D. S. Dobkin. 1995. Spatial overlap and habitat associations of barred owls and great horned owls in southern New Jersey. Journal of Raptor Research 29:151-157.
Lane, W. H., D. E. Andersen, and T. H. Nicholls. 2001. Distribution, abundance and habitat use of singing male boreal owls in northeast Minnesota. Journal of Raptor Research 35:130-140.
Longcore, T., and C. Rich. 2004. Ecological light pollution. Frontiers in Ecology and the Environment 2:191-198. https://doi.org/10.1890/1540-9295(2004)002[0191:ELP]2.0.CO;2
Mazur, K. M., S. D. Frith, and P. C. James. 1998. Barred owl home range and habitat selection in the boreal forest of central Saskatchewan. The Auk 115:746-754. https://doi.org/10.2307/4089422
McClure, C. J. W., H. E. Ware, J. D. Carlisle, G. Kaltenecker, and J. R. Barber. 2013. An experimental investigation into the effects of traffic noise on distributions of birds: avoiding the phantom road. Proceedings of the Royal Society B 280:20132290. https://doi.org/10.1098/rspb.2013.2290
Nenninger, H. R., and N. Koper. 2018. Effects of conventional oil wells on grassland songbird abundance are caused by presence of infrastructure, not noise. Biological Conservation 218:124-133. https://doi.org/10.1016/j.biocon.2017.11.014
Olsen, B. T. 1999. Breeding habitat ecology of the barred owl (Strix varia) at three spatial scales in the boreal mixedwood forest of north-central Alberta. M.S. thesis, University of Alberta, Edmonton, Alberta, Canada.
Pattison, C. A., M. S. Quinn, P. Dale, and C. P. Catterall. 2016. The landscape impact of linear seismic clearings for oil and gas development in boreal forest. Northwest Science 90:340-354. https://doi.org/10.3955/046.090.0312
Quesnelle, P. E., L. Fahrig, and K. E. Lindsay. 2013. Effects of habitat loss, habitat configuration and matrix composition on declining wetland species. Biological Conservation 160:200-208. https://doi.org/10.1016/j.biocon.2013.01.020
R Core Team. 2019. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
Ridgeway, G. 2017. gbm: generalized boosted regression models. R package version 2.1.3.
Rosa, P., C. R. Swider, L. Leston, and N. Koper. 2015. Disentangling effects of noise from presence of anthropogenic infrastructure: design and testing of system for large-scale playback experiments. Wildlife Society Bulletin 39:364-372. https://doi.org/10.1002/wsb.546
RStudio Team. 2018. RStudio: integrated development for R. RStudio, Inc., Boston, MA, USA.
Russell, M. S. 2008. Habitat selection of barred owls (Strix varia) across multiple scales in a boreal agricultural landscape in north-central Alberta. M.S. thesis, University of Alberta, Edmonton, Alberta, Canada.
Shannon, G., M. F. McKenna, L. M. Angeloni, K. R. Crooks, K. M. Fristrup, E. Brown, K. A. Warner, M. D. Nelson, C. White, J. Briggs, S. McFarland, and G. Wittemyer. 2016. A synthesis of two decades of research documenting the effects of noise on wildlife. Biological Reviews 91:982-1005. https://doi.org/10.1111/brv.12207
Shonfield, J. 2018. Using bioacoustics to examine the effects of industrial disturbance on owls and their prey. Ph.D. thesis, University of Alberta, Edmonton, Alberta, Canada.
Shonfield, J., and E. M. Bayne. 2017. The effect of industrial noise on owl occupancy in the boreal forest at multiple spatial scales. Avian Conservation and Ecology 12(2):13. https://doi.org/10.5751/ACE-01042-120213
Shonfield, J., and E. M. Bayne. 2021. Using bioacoustics to study vocal behaviour and habitat use of Barred Owls, Boreal Owls and Great Horned Owls. Airo 29:416-431.
Shonfield, J., S. Heemskerk, and E. M. Bayne. 2018. Utility of automated species recognition for acoustic monitoring of owls. Journal of Raptor Research 52:42-55. https://doi.org/10.3356/JRR-17-52.1
Silva, C. C., R. Lourenço, S. Godinho, E. Gomes, H. Sabino-Marques, D. Medinas, V. Neves, C. C. Silva, J. E. Rabaça, and A. Mira. 2012. Major roads have a negative impact on the tawny owl (Strix aluco) and the little owl (Athene noctua) populations. Acta Ornithologica 47:47-54. https://doi.org/10.3161/000164512X653917
Theobald, D. M., J. R. Miller, and N. T. Hobbs. 1997. Estimating the cumulative effects of development on wildlife habitat. Landscape and Urban Planning 39:25-36. https://doi.org/10.1016/S0169-2046(97)00041-8
Ware, H. E., C. J. W. McClure, J. D. Carlisle, and J. R. Barber. 2015. A phantom road experiment reveals traffic noise is an invisible source of habitat degradation. Proceedings of the National Academy of Sciences 112:12105-12109. https://doi.org/10.1073/pnas.1504710112
White, J. H., J. M. Smith, S. D. Bassett, J. L. Brown, and Z. E. Ormsby. 2018. Raptor nesting locations along an urban density gradient in the Great Basin, USA. Urban Ecosystems 21:51-60. https://doi.org/10.1007/s11252-017-0705-y
Yip, D. A., L. Leston, E. M. Bayne, P. Sólymos, and A. Grover. 2017. Experimentally derived detection distances from audio recordings and human observers enable integrated analysis of point count data. Avian Conservation and Ecology 12(1):11. https://doi.org/10.5751/ACE-00997-120111
Table 1. Descriptions of the types of human disturbance included in the analysis, and the features included in each type.
|Disturbance category||Description||Features included|
|Industrial facilities||Noise-producing industrial facilities for energy extraction||compressor stations, oil/gas plants|
|Cutblocks||Areas with trees removed by forestry operations||clearcuts, selective harvest, salvage logging|
|High activity clearings||Clearings with high levels of human activity||active well pads (oil or gas), industrial camps, other clearings with industrial infrastructure|
|Low activity clearings||Clearings with low levels of human activity||inactive or abandoned well pads (no infrastructure)|
|Hard linear features||Linear features with an impermeable surface||paved and gravel roads, railways|
|Soft linear features||Linear features with a vegetated surface||seismic lines, transmission lines, pipelines|
Table 2. Predictor variables included in the boosted regression tree models. Means and ranges of each variable are based on values of the variable within an 800-m buffer around each of the 452 locations included in the analysis. Locations present is the percent of locations surveyed with each disturbance type present (not applicable for forest age and forest composition). Locations were surveyed for owls between 2013 and 2016 using autonomous recording units.
|Industrial facilities (proportion)||0.01||0.00 - 0.45||22.6%|
|Cutblocks (proportion)||0.08||0.00 - 0.74||51.5%|
|High activity clearings (proportion)||0.02||0.00 - 0.45||62.2%|
|Low activity clearings (proportion)||0.01||0.00 - 0.31||63.9%|
|Hard linear features (proportion)||0.02||0.00 - 0.16||57.7%|
|Soft linear features (proportion)||0.03||0.00 - 0.14||99.8%|
|Total human footprint (proportion)||0.18||0.00 - 0.78||99.8%|
|Forest age (mean of forested area in years)||91.8||21.0 - 153.5||NA|
|Forest composition (proportion coniferous)||0.52||0.00 - 1.00||NA|
Table 3. Details of the final boosted regression tree model (BRT) for each owl species. For each owl species, the final BRT model used a learning rate and tree complexity that resulted in the lowest residual deviance and at least 1000 trees. The percent of variation in the data explained by each model is calculated as the residual mean deviance divided by the total mean deviance.
|Species||Tree complexity||Learning rate||Bag fraction||Number of trees||Total mean deviance||Residual mean deviance||Percent of variation in data explained by model|
|Great Horned Owl||2||0.0005||0.75||3300||1.386||1.262||9%|
Table 4. Pairwise interactions between predictor variables in boosted regression tree models for Barred Owls (Strix varia), Great Horned Owls (Bubo virginianus), and Boreal Owls (Aegolius funereus). Interaction size is reported as multiplicative strength, and only those interactions greater than one are reported here.
|Species model||Variable 1||Variable 2||Interaction size|
|Barred Owl||Forest composition||Total human footprint||4.72|
|Forest composition||Hard linear features||3.25|
|Forest composition||Soft linear features||1.51|
|Forest age||Hard linear features||1.30|
|Great Horned Owl||Forest age||Soft linear features||1.26|
|Boreal Owl||Forest composition||Soft linear features||4.18|
|Forest age||Forest composition||2.54|