Research Paper

INTRODUCTION

Habitat edges have been viewed as beneficial for some wildlife because of the increased availability of different resources caused by higher vegetative diversity (Leopold 1933, Johnston 1947, Willson 1974, Wang et al. 2002). However, some species of birds seem to avoid edges because they have lower breeding success there because of increased predation and/or nest parasitism (Suarez et al. 1997, Flaspohler et al. 2001). Thus, declines in forest songbird populations caused by forest loss may be exacerbated by associated increases in edge (Suarez et al 1997, Flaspohler et al. 2001). In general, edge-related effects on forest birds seem to be most negative in landscapes where the matrix is dominated by agriculture (Parker et al. 2005), which is a rare or absent disturbance in northern ecosystems. However, even small disturbances that create internal edges such as roads and trails can cause negative effects for some species (Miller et al. 1998, Ortega and Capen 1999), and these disturbances may be present even where agriculture is absent.

In the boreal forest of western Canada, linear disturbances such as roads, pipelines, and seismic lines created primarily by the energy sector are the most widespread cause of anthropogenic edge with linear feature densities of up to 40 km/km² (Stern et al. 2018). The most common linear disturbances are seismic lines: narrow linear features (< 3–8 m wide) that extend for 10’s to 100’s of kilometers as part of exploration for oil and gas reserves (Lee and Boutin 2006). Seismic line technology has changed over time with wider, higher impact 6–8 m wide lines used exclusively until 1988 and phased out by 2001 (Government of Alberta 2021). Lower impact, narrower lines (< 4 m wide) came into use after 1988 and fully replaced wider lines in the early 2000s in several other provinces and territories (Dabros et al. 2018).

The energy sector historically left seismic lines to recover naturally with the expectation they did not create significant forest loss or edge effects. However, regeneration of vegetation along some lines can be slow or non-existent (Revel et al. 1984, Lee and Boutin 2006), varying not only with line age but with soil compaction and simplification of microtopography during construction, altered hydrology, and continued use of seismic lines by humans and off-highway vehicles after seismic line exploration is done (Dabros et al. 2018). Understory plant communities may differ significantly between seismic lines and adjacent forests decades after seismic line construction (MacFarlane 2003, Finnegan et al. 2018). Regeneration may be slower in wetter lowland seismic lines, and early-successional vegetation may be more abundant along seismic lines within open forests (Finnegan et al. 2019). Microclimate changes in forests adjacent to seismic lines are small relative to differences between forests and large openings, though differences in wind and light conditions increase with seismic line width (MacFarlane 2003, Stern et al. 2018, Franklin et al. 2021).

Although individual seismic lines cause little forest loss (< 1% in the most developed areas), the cumulative length and amount of edge associated with seismic lines is high compared to other footprints such as forest harvesting (Dabros et al. 2018). Thus, seismic lines could have a measurable effect on forest songbirds if birds show edge avoidance (Bayne et al. 2005a, 2005b, Machtans 2006). There is a growing body of research showing that seismic lines impact the habitat and behavior of a wide range of organisms: songbirds (Lankau et al. 2013, Carpenter et al. 2022, Gregoire et al. 2022); insects (Riva et al. 2018); and mammals (Tigner et al. 2014, 2015, Dickie et al. 2017). Such studies led regulators to suggest limits on new seismic line development until old lines recover naturally or are restored (DLUPC 2006, Kennett 2006, Nitschke 2008, Dabros et al. 2017). Although there have been changes in seismic line policy and newer lines are narrower (~3 m) and remove less forest cover, many of the lines cut as early as the 1950s are still visible on the landscape because of slow and variable vegetation regrowth, and because human use of the lines as travel corridors prevents them from regenerating (Lee and Boutin 2006, Bayne et al. 2011, van Rensen et al. 2015). Because of this, increasing our understanding of the impacts of seismic lines and vegetation recovery on them remains important.

Because seismic lines are narrow but extensive, the sampling area within which birds are counted along seismic lines can influence the patterns of abundance observed (Bayne et al. 2016, Crosby et al. 2023). For example, if birds are counted within a smaller point count radius along a seismic line, then more of the sampled area is impacted by the seismic line but fewer birds will be detected. In contrast, if birds are counted within a larger point count radius or using unlimited-distance point counts along a seismic line, then less of the sampled area is within or influenced by the seismic line. However, the uncertainty associated with the effect size of seismic lines may also decrease (Bayne et al. 2016). At the same time, both the total area of seismic lines and the number of birds with territories overlapping with seismic lines increase with increasing sampling area.

In general, the magnitude and direction of avian response to linear feature density seems to depend on the species-habitat associations: for example, birds associated with younger forests and shrublands are more abundant in areas with more linear features (Bayne et al. 2016) whereas species associated with older coniferous forests seem less abundant (Leston et al. 2023). Although these papers discuss a general response to seismic lines, they do not include seismic line vegetation cover as a predictor. Studies of Ovenbirds (Seiurus aurocapilla; Lankau et al. 2013), Canada Warblers (Cardellina canadensis; Gregoire et al. 2022), Dark-eyed Juncos (Junco hyemalis), and Palm Warblers (Setophaga palmarum; Carpenter et al. 2022) show that some species do respond to the vegetation structure on seismic lines and that, as vegetation structure becomes more similar to the adjacent forest, initial impacts of younger seismic line vegetation on settlement decisions by forest birds may decrease.

Our goal was to evaluate how birds respond at the species and community level to seismic line recovery. We surveyed boreal songbirds on seismic lines with varying amounts of vegetation regeneration to understand how seismic lines affect the relative abundance of birds, and community composition. We made predictions based on three hypotheses: (1) no difference in the abundance of birds near seismic lines versus the forest interior whereby vegetation in the adjacent forest is the sole driver of variation in bird abundance and richness (Forest Structure Hypothesis); (2) birds show altered abundance near seismic lines relative to the forest interior but no differences between seismic lines with varying amounts of vegetation, because the vegetation on the seismic line is always different from the adjacent forest (Permanent Edge Hypothesis); and (3) bird abundance becomes more similar to the forest interior as vegetation growth on the seismic line increases (Line Recovery Hypothesis). We predicted that the hypothesis that best predicted the response of specific species would be related to life history characteristics such as nesting and foraging guild (Norton and Hannon 1997, Machtans 2006), and perception or not of seismic lines as a landscape discontinuity (Rail et al. 1997). For example, American Redstarts (Setophaga ruticilla) are known to prefer dense deciduous shrubs and may be attracted to the seismic line within the early stages of regeneration but not to newly cut bare lines (Sherry et al. 2020), whereas Ovenbirds, which forage and nest on the ground and sing in the canopy, would not find seismic lines usable until both the leaf litter, understory, and canopy on the line are regenerated (Lankau et al. 2013).

METHODS

Study area

Our study area was in northeastern British Columbia and the southern Northwest Territories (Fig. 1). Despite its location in northern Canada, parts of the study area, like Liard Valley, were fairly warm with a coastal influence and tall forests. Forests ranged from dry uplands to wet lowlands. Uplands included white spruce (Picea glauca), trembling aspen (Populus tremuloides), paper birch (Betula papyrifera) and mixed-wood stands. Lowlands included black spruce (Picea mariana) and tamarack (Larix laricina) bog-fen complexes. Because of the large area covered by bog-fen complexes and lower population density, there are few roads in the area and both human use and off-road vehicle travel in the summer is minimal, relative to areas of energy development further south in Alberta. Seismic line density in the study area varied from < 0.1 km/km² to > 26 km/km², with much lower seismic line densities and fewer roads near study sites in the Northwest Territories than in northeastern British Columbia (Tigner et al. 2015). Energy sector footprint including seismic lines was the primary human disturbance in the study area and was much more extensive at or near our sites in British Columbia than in the Northwest Territories (Fig. 1 in Tigner et al. 2015); some forestry occurred as well, more so in British Columbia, but there was no forest harvest near our study sites (Tigner et al. 2015). Trappers used the study area but trapping generally occurred in areas with little seismic line development (Tigner et al. 2015). Seismic line vegetation varied in recovery state from bare lines without woody growth to lines covered in shrubs and small trees. Bare lines were generally used as winter roads or local all-terrain vehicle (for example, snowmobile) access routes, but four of the more revegetated lines had active human trails that were 2–3 m wide down the center of much taller online vegetation. Canopy gaps gave all lines a functional width of 5–12 m (mean width = 8 m) at forest canopy height (Lankau 2014).

Seismic line selection and study design

Sampling locations were selected in the field to get a range of variation in vegetation regrowth on seismic lines. Forest height varied between and within upland and lowland forest types from 6 to 30 m. On seismic lines, as vegetation cover increased from open to closed, bare ground cover decreased while leaf litter cover, litter depth, shrub and tree stem density, canopy height and cover, and overall vegetation density increased, though most of these variables’ values except bare ground cover remained smaller along even closed seismic lines than in adjacent forests (Lankau et al. 2013). Thus, our indices of recovery were developed using the values of vegetation variables on the seismic line relative to the adjacent forest to describe recovery state. Vegetation cover on lines ranged from completely open with bare earth to dense low shrubs/ grasses (< 1 m tall) to tall shrubs (2–3 m tall) and saplings greater than 3 m tall (Fig. 2). The oldest lines were cut 50 years before the surveys were done (upland line age: mean = 28, range = 0–48 years old; lowland line age: mean = 28, range = 0–50 years old; Lankau 2014, Tigner et al. 2015). We did not use line age to preselect lines to survey because of the lack of correlation between line age and vegetation recovery (van Rensen et al. 2015).

Bird survey methods

We conducted seismic line point counts (n = 146 upland, 224 lowland) in 2008, 2009, and 2010. We also conducted forest interior point counts (n = 163 upland, 118 lowland) that were located at least 350 m from a forest edge in the same forest types as the seismic line locations. All point counts were located a minimum of 350 m apart and from any road or other anthropogenic edge.

We conducted a single 10-minute point count at each location. We estimated the distance from the observer to each bird in three distance bins: 0–50 m, 50–100 m, and unlimited distance. Because of how far boreal songbirds can be heard, unlimited distance point counts are in effect ~150 m radius. This is why the 350 m spacing between point counts was sufficient to prevent counting birds twice. Our analysis excluded birds like ravens and hawks that have large home ranges and/or can be heard much further away. Birds were identified from songs used by males for territorial demarcation. Observers were trained as a group to identify birds and estimate distance. Observers alternated surveys between point counts along seismic lines and within forest interiors to reduce observer bias. All counts were conducted between 27 May and 30 June and within four hours of sunrise. We scheduled counts so that both the forest interior and different types of seismic lines were sampled randomly with respect to time of day and day of the year to prevent seasonal or within-day biases.

Vegetation measurements

We collected data on vegetation at both the forest interior and seismic line point count locations. For the seismic line point count locations, we sampled the seismic line itself as well as the vegetation in the forest adjacent. Vegetation on seismic lines was variable, especially on older lines. To account for this, three linear subplots (each 22.6 m²) were placed 30 m apart along a 100-m transect to capture the variation in line vegetation. Each linear subplot contained four 1-m² quadrats used to measure percent ground covered by different vegetation types. Canopy cover and litter depth were measured at the ends of each linear subplot (six measurements each for canopy cover and litter depth and 12 quadrats per seismic line point count).

Values from the three subplots’ quadrats and ends were averaged to get the mean online values for each vegetation variable, similar to the sampling design used by McFarlane (2003). There was also a single forest vegetation plot (11.3 m radius, 0.04 ha) that was located 30 m from the forest edge to avoid vegetation changes associated with the seismic line edge, but still close enough to be compared to line vegetation (Fig. 3). Each forest interior point count had a single forest vegetation plot (11.3 m radius, 0.04 ha). Figure 3 illustrates the placement of point counts and vegetation plots relative to each other and the seismic line.

In each forest plot, we measured tree density, tree height, shrub stem density, and canopy cover. Trees were identified to species. Canopy height in the forest was recorded as the modal height of the trees to avoid outliers that were particularly tall or short. Canopy cover measurements were obtained at three locations (0, 135, and 225° from the center of the forest plot). We used fewer densiometer measurements in the forest plot than on the line plot because we initially found greater variability in vegetation canopy cover between the center and sides of the line than within the forest plot. There was also a single 22.6 m² linear subplot with four 1-m² ground cover quadrats and three litter depth measurement locations per forest plot.

We measured the same variables on seismic lines with the exception that we recorded mean vegetation height instead of tree height. Additional variables were measured at each seismic line vegetation subplot and at the forest plot adjacent to the seismic line: ground cover, litter depth, and horizontal vegetation density from 0 m to 3 m in height.

The density of trees (woody plants greater than 8 cm diameter at breast height [DBH]) were measured as stems per hectare, and the density of shrub stems (defined as woody plants less than 8 cm DBH and greater than 50 cm tall) as stems per square meter. Ground covered by leaf litter, moss, grass, forbs, water, and bare ground was visually estimated within multiple 1 m² subplots per vegetation plot. Litter depth was measured to the nearest centimeter. Canopy height and seismic line vegetation height were measured using a clinometer or a graduated 8 m pole depending on the height of the vegetation. Horizontal vegetation density was measured using a 0.5 m wide cover-board: one observer held the board while the second stood 10 m away and estimated the percentage of the board that was obscured by green vegetation in four height increments (0–0.5 m; 0.5–1.0 m; 1.0–2.0 m; and 2.0–3.0 m). Angular canopy cover was measured using methods described in Nuttle (1998). In brief, the densiometer contained a spherical-shaped reflector mirror engraved with a cross-shaped grid of twenty-four 1/4″ squares. Each technician held the densiometer 12″ to 18″ in front of their body at elbow height. Assuming four equally spaced dots in each grid square, the technician systematically counted the number of quarter-square openings (dots laying over canopy openings in the mirror), then multiplied the total count by 1.04 to obtain percent of overhead area not occupied by canopy and subtracted this value from 100 to obtain an estimation of overstory density in percent. The variables common to both the seismic line vegetation subplots and the adjacent forest plot were subsequently used to create recovery indices measuring the state of seismic line vegetation recovery. Further details of the vegetation surveys are available in Bayne et al. (2011) and Lankau (2014).

Vegetation variables

We standardized all vegetation variables in the forest vegetation plots to zero mean and unit variance and then used a principal component analysis (PCA) to summarize forest structure into a series of uncorrelated PCA components that included the following variables: canopy height, canopy cover, tree density, percent deciduous trees, mean tree DBH, and tree species richness. We retained all PCA components with an eigenvalue equal to or greater than 1 (which indicates PCA components that account for more variance than accounted for by one of the original variables in standardized data), and which had loadings equal to or greater than 0.5 for at least one vegetation variable (Jackson 1993). Squaring the loadings gives the proportion of variance in each original variable explained by a particular PCA component, with loadings ~ 0.45 rated as “fair” and loadings ~ 0.55 rated as “good.” These PCA components were used as predictor covariates to describe forest conditions in our analysis of bird abundance (Jackson 1993).

Seismic line recovery index

To measure seismic line recovery, we calculated the state of vegetation on each seismic line segment and compared it in a pairwise fashion to the state of the vegetation in each seismic line’s forest vegetation plot (adjacent to the seismic line) using a relative difference equation:

(1)

where f is the forest vegetation value and l is the line vegetation value and -1 is used to change the sign so that the index is scaled from -1 (line vegetation less than forest vegetation) and +1 (line vegetation greater than forest vegetation). The point of reference is 0, which occurs when line and forest values are equal.

From the relative differences calculated at each seismic line, we created four recovery indices to assess which vegetation layers best predicted bird abundance. Specifically, we created indices for (1) ground cover recovery (leaf litter depth and cover of bare ground, moss, graminoids, water, and down woody material); (2) shrub recovery (shrub stem density and percent deciduous shrubs); and (3) tree recovery (tree density, sapling density, average height, and total canopy cover). These indices relate to the nesting and foraging guild categories for the 24 species we analyzed. The final recovery index used all 12 of the vegetation variables to create an overall recovery index. To compute the recovery indices, we first converted the relative difference scores for each vegetation variable into absolute values using the formula:

(2)

where RD is the relative difference value and (-1) scales the value from -1 to 0 so that more negative numbers indicate a greater difference between the seismic line and the adjacent forest and 0 indicates the seismic line and forest were the same.

To create the indices, we averaged the RD_abs values for each vegetation variable belonging to a structural class for each of the four separate recovery indices (RI) using the formula:

(3)

where RI_(x) is a particular recovery index (i.e., ground, shrub, tree, and all). RD_abs(1) to RD_abs(i) are the individual relative difference values for each vegetation variable and (i) is the total number of RD_abs values included in a particular index.

Bird abundance

To assess how scale influenced results, we did all our analysis three times using all birds detected within (1) 50 m radius point counts; (2) 100 m radius point counts; and (3) unlimited distance point counts. The same survey locations and visits were used in all three analyses but in the 50 m and 100 m point counts, counts of each species were based only on individual birds detected within 50 m and 100 m, respectively.

A potential complication of using point counts to survey birds is that the distance sound travels may differ between forest interior and seismic line locations. Birds singing at the edges of the more open seismic lines may be heard from further away and this could result in higher counts on seismic lines than the forest interior even if bird density were the same. We tested for this by calculating the effective detection radius (EDR) for the seismic line versus the forest interior using the binomial distance estimator in program DISTANCE 6.0 (Thomas et al. 2010) with distance classes of 50 m versus > 50 meters (Buckland 1987, Matsuoka et al. 2012). We assumed observers could estimate distance equally well in both the forest interior and on seismic lines.

Detection of bird species within point counts can also vary because of timing when point counts are conducted (within the season, during the day) and observer (Sólymos et al. 2013). We included observer as a random effect in the models, and we evaluated time of day and day of year as fixed effects to account for variation caused by survey timing.

Individual species models

We assigned each bird species in the study area into upland, lowland, or a generalist “guild” depending on how many survey locations they were detected at in each forest type. Species were assigned as upland or lowland specialists if 85% or more of the survey locations where they were detected were in one forest type. Generalist species were found at less than 85% of locations in any one forest type. We used all point counts (n = 651) when modeling species that we classified as generalists, i.e., species that use both upland and lowland habitats. For species exclusively or primarily found in upland vs lowland forest, we only used point counts from uplands (n = 309) and lowlands (n = 342) to generate individual species models.

We created six statistical models for each species at the three different sampling radii (18 models per species). For the Forest Structure Hypothesis model, we included PC1 and PC2 to describe the forest conditions at each point count location as well as day of year (# days since 1 January in the survey year) and time of day (# hours since midnight). To test the Permanent Edge Hypothesis, we used the same variables and added whether a point count was done on a seismic line versus the forest interior (seismic line = 1, forest interior = 0). Finally, to test the Line Recovery Hypothesis we ran four models with the ground, shrub, tree, and all recovery indices as linear functions. Within the models created to test the Line Recovery hypothesis we also assessed how different ways of defining vegetation state on seismic lines relative to the forest interior influenced the response by birds. In the Line Recovery models, we included an interaction between a model’s recovery index and treatment (seismic line = 1, forest interior = 0). This approach allowed us to compare bird abundance on seismic lines along a recovery gradient to the forest interior.

Analyses were done using the “lme4” package (Bates et al. 2015) in R (R Core Team 2022) to run all models using mixed effects regression. We used a Poisson error distribution in these models except when variance in a species’ count was > 2 times the mean count, in which case models used a negative binomial error distribution. Akaike’s Information Criteria (AIC) was used to select the best model (Anderson 2008). Models were ranked based on AIC weights and evidence ratios. A model was considered the best model if the AIC weight was greater than 0.6. If the AIC weight was less than 0.6 and ΔAIC less than 2, we used the principle of parsimony (model with fewest parameters) to decide which model was most predictive. We estimated marginal and conditional R² at each spatial scale with the “AICmodavg” package (Mazerolle 2023).

Species richness models

Mixed effects regression was used to evaluate alpha diversity (here, the number of bird species detected at each station) of all territorial passerines separately for upland (n = 309) and lowland point counts (n = 342). We created 10 statistical models for predicting species richness at the three different sampling radii (30 models). As in the models for abundance, we compared the Forest Structure, Permanent Edge, and Line Recovery models. We ran the interaction between a model’s recovery index and treatment (seismic line = 1, forest interior = 0) to compare predicted species richness along seismic lines in different recovery stages with species richness in forest interiors. We evaluated both linear and quadratic fit (unimodal) of different recovery indices (ground, shrub, tree, and all) in the Line Recovery models.

Community analysis

To explore differences in species composition (beta diversity) among upland and lowland seismic line and forest interior point counts, while including species that were too rare to be included within the individual species models, we analyzed community metrics. First, we performed a non-metric multidimensional scaling (NMDS) analysis on the point counts, using the “metaMDS” function in the “vegan” package and a Bray-Curtis distance matrix (Baselga 2013). After removing point counts without any detections of the 48 species detected within 50 m of point counts (n = 147), we used counts of individual species detected within 50 m for measuring community composition within point counts (n = 545). To visually explore the associations of species with each other and overlap between different point count site types (upland vs. lowland seismic line vs. forest interior), we used the “orditorp” and “ordellipse” functions to respectively plot the NDMS scores of the 48 species in the analysis and the 70% confidence ellipses of the four-point count site types. We used the “envfit” function in “vegan” to assess the relationship between point count variables and community composition, measured as the squared Pearson correlation coefficients between the NMDS axis scores of individual sites and environmental variables associated with those sites (upland vs. lowland; seismic line vs. forest interior). Additionally, we used permutational multivariate analysis of variance (PERMANOVA; permutations = 2000) to test for a significant effect of site type on community composition (“adonis2” function in the “vegan” package), and for the percentage of variance in distance scores explained by point count site type. We also tested for significant differences in community composition among site types based on pairwise comparisons (“pairwise.adonis” function in the “pairwiseAdonis” package: Martinez 2017).

We performed an indicator species analysis, using the “multipatt” function (permutations = 99) in the “indicspecies” package to measure the strength of association of individual species with point count site types, and identify potential indicator species for those point counts (De Cáceres and Legendre 2009). This analysis enabled us to measure the strength of association of rarer species with different site types.

RESULTS

We conducted 651 point counts at independent locations across three years: 370 point counts on seismic lines (224 in lowland and 146 in upland areas) and 281 forest interior points (118 in lowland and 163 in upland areas). We detected a total of 52 species of passerine birds. Of these, 24 had enough data to have abundance models that converged to a solution. Six species were found primarily in uplands, eight primarily in lowlands and 10 in both land cover classes. Species codes, names, and habitat associations are summarized in Appendix 1.

Effective detection radius

We did not find a consistent difference in EDR between seismic lines and forest interiors for the 10 most common species as confidence intervals (hereafter C. I.) overlapped. Confidence intervals for both forest interior and seismic line detections of Hermit Thrushes (Catharus guttatus) and Ruby-crowned Kinglets (Regulus satrapa) were large, indicating large observer variation in estimating how far away these species were. We determined it was valid to use unadjusted count data to assess relative differences between seismic lines and the forest interior (Fig. 4, Appendix 2).

Forest vegetation

PC 1 was positively correlated with canopy height, percent deciduous trees, and tree DBH. PC 2 was positively correlated with tree density and canopy cover (Appendix 3).

Individual species models

Within the 50-m-radius point counts, there was limited evidence of seismic lines explaining variation in the abundance of most species. However, Tennessee Warbler (Leiothlypis peregrina) and American Redstart were 52% and 800% more abundant along or within 50 m of seismic lines according to the Permanent Edge model. Hermit Thrush was better predicted by the Line Recovery model (Ground) and increased with recovery of ground vegetation along seismic lines. For all other species, the Forest Structure, Permanent Edge, and Line Recovery model all were within 2 AIC units so based on the principle of parsimony we selected the Forest Structure model for reporting (Fig. 5, Appendix 4).

Within the 100-m-radius point counts, the Forest Structure model was the top model for six generalist species (60%), three upland species (50%), and four lowland species (50%). The Permanent Edge model was the top model for White-throated Sparrow (Zonotrichia albicollis), which was 44% more abundant on average along or within 100 m of seismic lines. It was also the top model for Palm Warbler, which was 42% less abundant on seismic lines. Vegetation recovery along lines best explained abundance for three generalist species (30%), three upland species (50%), and three lowland species (50%). Fox Sparrow (Passerella iliaca), Red-eyed Vireo (Vireo olivaceus), Tennessee Warbler, and Yellow-rumped Warbler (Setophaga coronata) were best predicted by recovery of ground vegetation. Alder Flycatcher (Empidonax alnorum), American Redstart, Least Flycatcher (Empidonax minimus), and Lincoln’s Sparrow (Melospiza lincolnii) were best predicted by recovery of shrubs. Swainson’s Thrush (Catharus ustulatus) abundance was best predicted by recovery of trees. The generalists (Swainson’s Thrush, Tennessee Warbler, Yellow-rumped Warbler) were respectively 79% less, 49% less, and 60% more abundant along more recovered lines. For upland species the American Redstart and Red-eyed Vireo increased with vegetation recovery while the Least Flycatcher was virtually absent (decreased by > 99%) along the most tree-recovered lines relative to lines with the least tree cover. For lowland species the Lincoln’s Sparrow increased by 6x, while the Alder Flycatcher and Fox Sparrow decreased respectively by 83% and 98% along more recovered lines (Fig. 5, Appendix 4, Appendix 6).

Within the unlimited-distance point counts, the Forest Structure model was the top model for four generalist species (40%), four upland species (67%), and three lowland species (38%). The Permanent Edge model was the most parsimonious model for Dark-eyed Junco and Palm Warbler, which were, respectively, 24% and 43% less abundant on average along or near seismic lines. One or more Linear Recovery models best explained abundance for six generalist species (60%), two upland species (33%), and three lowland species (50%). Swainson’s Thrush and Tennessee Warbler were best explained by recovery of overall vegetation, while other species were best explained specifically by recovery of ground vegetation (Alder Flycatcher, Fox Sparrow, Yellow-rumped Warbler), shrubs (American Redstart, Least Flycatcher, Lincoln’s Sparrow), or trees (Dark-eyed Junco, White-throated Sparrow). Species that responded to seismic line recovery showed similar responses in both the 100-m and unlimited distance analyses (except Red-eyed Vireo, which was best explained by forest structure in the unlimited-distance model). Generalists like Chipping Sparrow (Spizella passerina), and White-throated Sparrow were, respectively, 48% and 62% less abundant along more recovered lines in unlimited-distance point counts, although Black-and-white Warbler (Mniotilta varia) increased by 17x along lines with fully recovered ground vegetation (Fig. 5, Appendix 4, Appendix 6).

We assessed how species responded to a specific recovery index (ground, shrub, or tree) versus the overall index. Within 50-m point counts, a specific recovery model based on ground vegetation, shrubs, or trees explained abundance better than recovery of overall vegetation in 67% of species. Within this 67% the overall recovery model explained the abundance of four species better than the Forest Structure model. Within 100-m point counts, a specific recovery model explained abundance better than the overall recovery model in 79% of species. Within this 79%, the overall recovery model explained the abundance of 9 species better than the Forest Structure model. In the unlimited-distance point counts, a specific recovery model explained abundance better than the overall recovery model in 88% of species. Within this 88%, the overall recovery model explained the abundance of 11 species better than the Forest Structure model (Fig. 5, Appendix 4, Appendix 6).

Species richness models

Within the 50-m point counts, upland species richness was 48% lower along seismic lines than forest interior (AIC weight_{RI_tree_linear} = 0.56; β_{Seismic_50} = -0.64, 95% C. I. = -1.20–0.09) using the Permanent Edge model. Using the Line Recovery model based on the tree index we found richness declined by a further 67% with more trees on the line (β_{RI_tree_50} = -1.12, 95% C. I. = -1.88 – -0.36). In contrast, lowland species richness was 6% higher along seismic lines (AIC weight_{RI_ground_linear} = 0.27; β_{Seismic_50} = 0.06, 95% C. I. = -0.22–0.34) using the Permanent Edge model. Using the Line Recovery Model, lowland richness was 36% lower when ground cover was considered recovered (β_{RI_ground} = -0.45, 95% C. I. = -0.98–0.08). However, the Permanent Edge model (AIC weight_Line = 0.18; β_Seismic = 0.25, 95% C. I. = 0.10–0.42) had similar explanatory power for predicting lowland species richness, which increased by an average of one species per station along seismic lines relative to forests (Fig. 6, Appendix 5).

Within the 100-m point counts, upland species richness was 29% lower along seismic lines before accounting for recovery (AIC weight_{RI_tree_linear} = 0.53; β_{Seismic_100} = -0.33, 95% C. I. = -0.70–0.03) and further declined by 52% with increasing recovery of trees (β_{RI_tree_100} = -0.73, 95% C. I. = -1.23 – -0.22). In contrast, lowland species richness did not significantly differ between seismic lines and forests before accounting for recovery (AIC weight_{RI_ground_linear} = 0.33; β_{Seismic_Unlim} = -0.08, 95% C. I. = -0.26–0.11) and declined by 30% with increasing recovery of ground vegetation (β_{RI_ground_100} = -0.36, 95% C. I. = -0.70 – -0.01; Fig. 6, Appendix 5).

In the unlimited-distance point counts, depending on the model, upland species richness was 27–71% lower along seismic lines (AIC weight_{RI_all_quadratic} = 0.29, β_{Seismic_Unlim} = -1.25, 95% C. I. = -2.37 – -0.12; AIC weight_{RI_tree_linear} = 0.29, β_{Seismic_Unlim} = -0.32, 95% C. I. = -0.63 – -0.00) and further declined by 67% with recovery of overall vegetation (β_{RI_all_Unlim} = -3.18, 95% C. I. = -5.89 – -0.48; β_{RI_all_Unlim}² = -4.16, 95% C. I. = -9.01–0.68) or by 47% with recovery of trees (β_{RI_tree_Unlim} = -0.32, 95% C. I. = -1.05 – -0.18). In contrast, lowland species richness did not significantly differ between seismic lines and forests before accounting for recovery (AIC weight_{RI_ground_linear} = 0.43; β_{Seismic_Unlim}= -0.12, 95% C. I. = -0.27–0.03) and declined by 31% with increasing recovery of ground vegetation (β_{RI_ground_Unlim} = -0.37, 95% C. I. = -0.64 – -0.09; Fig. 6, Appendix 5).

Community analysis

When the species and site scores from the NMDS were plotted, lowland interior sites were more negatively correlated with axis 1 while upland interior sites were more positively correlated with axis 1. Confidence ellipses for points belonging to a particular site type suggested that there was more variability in community composition among both upland and lowland seismic line point counts than among upland and lowland forest interior point counts. There was more overlap in upland and lowland seismic point count communities than upland and lowland forest interior communities. Nevertheless, species that we classified as lowland specialists were generally negatively associated with NMDS axes 1 and 2, while species that we classified as upland specialists were generally positively associated with NMDS axes 1 and 2 (Fig. A7.1). There was a stronger, significant relationship between site NMDS scores and land type (upland vs. lowland: r² = 0.016, p = 0.001) than with treatment (seismic line vs. forest interior: r² = 0.001, p = 0.540; Fig. A7.1). In the PERMANOVA, site type explained 6% of variance in community composition distances (R² = 0.06, F = 10.797, p = 0.0005). There were larger pairwise differences in community composition between upland forest interior and seismic line point counts (R² = 0.012, F = 3.434, p = 0.012) than lowland forest interior and seismic line point counts (R² = 0.010, F = 2.377, p = 0.084), and larger pairwise differences than these when comparing any combination of upland vs. lowland seismic line vs. forest interior point counts.

In the indicator species analysis, species that were significantly and most strongly associated with lowland seismic lines were Alder Flycatcher, Dark-eyed Junco, Lincoln’s Sparrow, Ruby-crowned Kinglet, Swamp Sparrow (Melospiza georgiana), Tennessee Warbler, Wilson’s Warbler (Cardellina pusilla), and Yellow-bellied Flycatcher (Empidonax flaviventris). Hermit Thrush and Palm Warbler were significantly and most strongly associated with lowland forest interior point counts. American Redstart, Black-and-white Warbler, Red-eyed Vireo, Warbling Vireo (Vireo gilvus), and White-winged Crossbill (Loxia leucoptera) were significantly and most strongly associated with upland seismic lines. American Robin (Turdus migratorius), Bay-breasted Warbler (Setophaga castanea), Canada Warbler, Cape May Warbler (Setophaga tigrina), Golden-crowned Kinglet (Regulus satrapa), Hammond’s Flycatcher (Empidonax hammondii), Magnolia Warbler, Ovenbird, Red-breasted Nuthatch (Sitta canadensis), Western Tanager (Piranga ludoviciana), White-throated Sparrow, and Winter Wren (Troglodytes hiemalis) were significantly and most strongly associated with upland forest interior point counts (Table A7.1).

DISCUSSION

Our results show that the revegetation status of the seismic lines matters when assessing the impacts of seismic lines on songbirds. This is supported by earlier papers looking at individual species responses to seismic line revegetation (Lankau et al. 2013, Carpenter et al. 2022, Gregoire et al. 2022). Recovery metrics of specific cover types (ground, shrub, and tree) were better predictors of the abundance of species than the overall metric of seismic line recovery, although the overall and specific metrics were all strongly positively correlated with each other (0.68 < r < 0.99). This suggests that there are species-specific responses to vegetation attributes on the seismic line that lead to changes in bird abundance relative to the forest interior. These results also show that treating all seismic lines as equal, as has been done in some large-scale analyses (Bayne et al. 2016), does not capture the whole picture.

Importantly, we found limited evidence that canopy birds or species associated with mature forests were negatively impacted by open seismic lines. The amount of forest habitat removed by seismic lines construction may have not been enough to significantly reduce habitat for these species (Drolet et al. 1999). Several of these species (e.g., Ovenbird, Canada Warbler) were strongly associated with upland forest interior locations in the indicator species analysis. These are the species of greatest concern because they may decline near multiple oil-gas footprint types (Leston et al. 2023) and not readily recolonize other kinds of disturbed lands (e.g., harvested lands) as considerable time must pass for tree conditions to match the adjacent forest (Leston et al. 2018). The exception was Red-eyed Vireos, which increased in abundance as seismic lines recovered. Red-eyed Vireos sing and forage in the canopy but nest in small saplings or shrubs (Cimprich et al. 2020). Thus, recovery of shrubs may provide nesting sites. Secondly, Red-eyed Vireos select deciduous forest. Our personal observation is that the gap caused by seismic lines in deciduous forest often is reduced from the sides at the canopy level as branches grow into the open space. Thus Red-eyed Vireos may respond more to changes in nesting and foraging habitat than canopy gaps. Machtans (2006) made the argument that canopy nesting and foraging birds in mature forests are not likely to respond to seismic lines because the gap created by the seismic line is often not wider than natural canopy gaps. This seems to be true for Western Tanagers, Bay-breasted Warblers, and Ruby-crowned Kinglets in our study area which did not show any response to seismic lines or changes in seismic line vegetation. Although community analyses suggested that some habitat specialists (e.g., Canada Warbler) were significantly associated with either upland or lowland interior forests, few of these species showed declines with respect to seismic lines in individual species models. However, our study sites, particularly those in the Northwest Territories, had a smaller energy sector footprint than regions further south (e.g., Alberta) and little other human footprint (e.g., harvest; Tigner et al. 2015). Perhaps the species in our study sites would show stronger responses to seismic lines or seismic line recovery in study areas with more seismic lines and/or other footprints (Gregoire et al. 2022).

For now, our study suggests that the effects of seismic lines may be comparable to effects of smaller (0.01 to 2.0 ha) forest gaps from both natural and man-made disturbances on forest songbird communities elsewhere in North America (Robinson and Robinson 1999, Greenberg and Lanham 2001, Forsman et al. 2010). The correlation between gap age, shrub regrowth, and “gap-dependent” species found by Robinson and Robinson (1999) is like the trends we found with American Redstarts: abundance increased a few years after cutting and returned to pre-disturbance numbers as the gaps closed in. On the scale of a 150 m radius point count, an 8 m wide seismic line is equivalent to a 0.02 ha gap, which may explain the similarity of our results; the width of a single seismic line may have little impact via habitat amount or configuration for most species (Drolet et al. 1999).

The species that seem to be attracted to seismic lines were not unexpected. White-throated Sparrows are more likely to cross forest gaps (Rail et al. 1997) and are often found in harvested areas and very old stands where the canopy has started to break up leading to a dense shrub or grass understory (Falls and Kopachena 2020), and seismic lines provide similar forest structures. The greatest increase in abundance for any species was for American Redstarts, which were detected most frequently near lines with heavier shrub cover and were substantially more abundant than in the forest interior. A similar pattern was observed for Lincoln’s Sparrow, and we found small increases in Alder Flycatcher, Least Flycatcher, Swainson’s Thrush, Tennessee Warbler, and Yellow-rumped Warbler abundance on open or shrubby lines. The denser shrubs on seismic lines may provide high quality foraging habitat, especially during the nesting season when insects are in high demand. This would also explain why the same species are less abundant as trees recovered because shrub density often decreases below the tree canopy.

Alpha richness showed more consistent patterns across scales than individual species. Overall, richness near seismic lines was similar to the forest interior. However, for surveys on seismic lines there was a pattern of higher richness when lines were less recovered. At the individual species level, the most common pattern (e.g., 40% of species within unlimited-distance point counts) we observed were slight increases in abundance at early stages of line recovery (where shrub density could be much greater than the adjacent forest). This supports the idea that richness of birds increases near seismic lines because they are attracted to lines with shrub cover rather than some form of sampling error. We did not detect any introduced bird species in the study area meaning that any changes in species richness were due to shifts in native boreal species. The lack of non-boreal species is due to how narrow the lines are compared to linear features like pipelines and transmission line corridors where grassland and shrubland specialists might move in (Chasko and Gates 1982, King and Byers 2002, Kalukapuge et al. 2024).

We argue that the relative abundance metric we used was robust because there was no evidence from our EDR analysis that the increase in richness we observed was caused by being able to hear birds further on seismic lines than the forest interior. Unless birds preferred to sing at the edge of the seismic line (which we did not personally observe) sound was travelling through the forest vegetation at both interior and line locations in a similar way. Another factor is that total vegetation height and density varied across the study area. Especially in lowland areas, the forest was often quite open, and one would not expect a difference in sound transmission between seismic lines and the forest interior. Finally, as the seismic lines became more revegetated, the birds’ songs were travelling through dense vegetation both on and off the lines.

Although the number of studies on how birds respond to narrow linear features in the boreal forest is limited, there are enough done at different organizational and spatial scales to start looking for consistency across studies. For example, Ovenbirds were observed treating seismic lines as territorial boundaries in multiple studies (Bayne et al. 2005a, Machtans 2006, Lankau et al. 2013), which theoretically suggests that seismic lines reduce Ovenbird habitat and abundance. However, Lankau et al. (2013) found that, when shrubs and trees reach a certain height, Ovenbird begin to include seismic lines in their territories. Gregoire et al. (2022) found a similar behavior in Canada Warbler. Detection of seismic line effects on Ovenbirds also depends on the spatial scale of analysis, with studies done at small scales (replicate of analysis = individual territory) suggesting a negative impact of seismic lines on Ovenbird abundance (Machtans et al. 2006, Lankau et al. 2013). In contrast, Bayne et al. (2005b) found no effects of seismic line density at larger scales (replicate of analysis = landscapes with multiple territories), e.g., 4 km² landscapes in areas where seismic lines were the only human disturbance. More recent studies found evidence of interactive effects between various human disturbances including seismic lines within landscapes ~3ha–9km² (Crosby et al. 2023, Mahon et al. 2019), making it difficult to accurately estimate the absolute numerical impact of seismic lines on the Ovenbird and, presumably, other species as well.

More recently, a study looked at the Palm Warbler as an indicator species of bogs and fens using the same territory mapping approach as the Ovenbird (Carpenter 2020, Lankau et al. 2013, Carpenter et al. 2022). Palm Warblers will not place their territories in areas with wells, roads, and pipelines but do include seismic lines as part of their territory, but only in areas with lower contrast in vegetation density between the lines and surrounding off-line vegetation (Carpenter 2020). In our study, Palm Warblers had the strongest negative numerical response to seismic lines but this was not correlated with vegetation on the seismic line. Conversely, seismic lines had no effect on Palm Warblers at larger spatial scales (e.g., 3–5 ha [Carpenter et al. 2022], 4 km² [Mahon et al. 2019]). A possible explanation for these differences is the legacy seismic lines we studied were wider (8 m) than the other studies of Palm Warbler abundance, which included areas with a mix of low-impact (3 m) and legacy seismic lines (8 m).

An important feature of our study is that both human use of lines and line density in our study area were much lower than in other study areas. This means that we could focus on the impact of seismic lines without cumulative impacts of well pads, roads, railways, forestry, and high-density seismic exploration (Mahon et al. 2019). The low human use also meant that these seismic lines could regenerate undisturbed and the amount of vegetation regrowth is what can be expected on 40- to 50-year-old legacy lines. It would be interesting to compare the rate and variability of vegetation regrowth and the percentage of highly recovered lines between our study area and the southern boreal forest areas with more all-terrain vehicle use. Collecting vegetation data on the ground, as we did, is very time intensive and expensive, and limits the number of locations that can be surveyed in a season. Advances in remote imaging, specifically LiDAR, will make larger-scale surveys of seismic line vegetation recovery easier and allow us to create better models of seismic line impacts because we could possibly exclude “recovered” lines from estimates of linear feature density.

When comparing seismic line impacts on birds it is important to recognize that seismic lines are different widths, but current habitat suitability models used by governments for land-use planning (Dabros et al. 2018) do not account for differences among seismic lines. We do not know what the threshold of line width is for different species, and the term “low impact” includes widths ranging from 5 m down to 1 m (Government of Alberta 2021). Territory mapping shows that several species include low-impact seismic lines in their territories (Bayne et al. 2005a, Carpenter 2020, Gregoire et al. 2022). Kalukapuge et al. (2024) also found that species had different thresholds in their responses to increasing linear feature width.

We were more likely to find increases in species abundance in response to seismic lines as the spatial scale of the analysis increased. That few differences in abundance within the 50-m point counts existed suggests that edge effects are weak at best. However, the higher numbers of zeros in the 50-m radius point count dataset also meant that it had a lower power to detect changes in bird abundance (Bayne et al. 2016). The stronger support for the Permanent Edge and Line Recovery models in the 100 m and unlimited distance point counts also suggests that these sampling areas have greater power to detect change. The reason for this is that as the point count radius increases at the same locations, the number of locations with zero counts of a species decreases, resulting in less variance in the data (Bayne et al. 2016). Evaluating how sampling area influences results is essential for comparing studies on energy sector impacts and edge effects generally. It may seem counterintuitive to use a larger sampling area when the 6-m to 8-m wide seismic line is a relatively small area within the point count circle. However, songbird territories for species like Ovenbirds, Palm Warblers, American Redstarts, and Red-eyed Vireos range in size (anywhere from < 0.5 ha to > 2.0 ha depending on habitat, population density, and food supply; Billerman et al. 2022) and are roughly circular. It is unlikely that even birds like American Redstarts that are attracted to shrubby lines would fit their entire territory onto a seismic line (e.g., 0.5 ha territory would be an approximately 80-m diameter circle but would result in an 8-m by 625-m rectangle if forced onto a conventional seismic line). Using the larger unlimited distance point counts increased the number of territories that were adjacent to or bisected by the 300-m section of line within the point count circle.

Our results and those of a growing number of studies demonstrate that seismic lines do change bird behavior and locally influence species abundance, which in turn influences species richness (but see Lankau et al. 2013 and Gregoire et al. 2022). However, the magnitude of these changes in terms of regional population size or community change is much harder to estimate because of the variable results in different areas and the fact that different patterns are revealed at different scales using different methods. For example, we caution that in other forest types we did not study (i.e., mature white spruce or pine forests) bird response to seismic line recovery may differ as Leston et al. (2023) found larger negative effects of seismic lines in those forest types. Another limitation is that point counts are a coarse measuring tool and cannot capture the detail that territory mapping, telemetry, and behavioral observations can. Actual use of seismic lines by individual birds cannot be determined within point counts unless individual birds are seen on the lines and without tracking individual birds’ activities (e.g., percentage of time budget on vs off the line, percentage of food items foraged on the line, nestling growth rates relative to proportion of parents’ time spent on vs off the line). There have also been very few studies of whether the demography of birds is different near seismic lines than the forest interior and is a key knowledge gap.

Whether the changes in bird abundance and species richness are good or bad depends on conservation goals. Depending on the analysis, species richness declined by 30 to 67% with recovery of seismic line vegetation, but this decline was probably driven by decreases by more common species as vegetation recovered. A few species increased by 40 to 300% with vegetation recovery or were up to 45% more abundant along seismic lines. Either way, it is encouraging that seismic lines regenerate, albeit slowly, and that changes in bird abundance disappear when seismic line vegetation becomes structurally more similar to the surrounding forest. Current seismic line remediation efforts generally focus on reducing impacts on woodland caribou (Rangifer tarandus caribou; Dickie et al. 2017, Filicetti et al. 2019). Our work indicates that the needs of boreal songbirds should also be included in seismic line remediation plans. Although different forest disturbances may produce differences in bird communities as forests regenerate over time (Hobson and Schieck 1999), our results and those of Tigner et al. (2015) for American marten (Martes americana) demonstrate that efforts to encourage faster shrub and tree regrowth would be beneficial to a variety of taxa. Shrubs attract birds including species at risk like Canada Warbler (Gregoire et al. 2022), allow marten to cross seismic lines, and can reduce the movement of predators like bears and wolves (Tigner et al. 2015, Dickie et al. 2017). By ensuring that vegetation recovery along seismic lines utilizes native species and considers natural variation in forest composition and structure, the energy sector could reduce impacts of seismic lines on birds and presumably other species as well.

LITERATURE CITED

Anderson, D. R. 2008. Model based inference in the life sciences: a primer on evidence. Springer, New York, New York, USA. https://doi.org/10.1007/978-0-387-74075-1

Baselga, A. 2013. Separating the two components of abundance-based dissimilarity: balanced changes in abundance vs. abundance gradients. Methods in Ecology and Evolution 4:552-557. https://doi.org/10.1111/2041-210X.12029

Bates, D., M. Mächler, B. Bolker, and S. Walker. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67(1):1-48. https://doi.org/10.18637/jss.v067.i01

Bayne, E. M., S. Boutin, B. Tracz, and K. Charest. 2005a. Functional and numerical responses of Ovenbirds (Seiurus aurocapillus) to changing seismic exploration practices in Alberta's boreal forest. Écoscience 12:216-222. https://doi.org/10.2980/i1195-6860-12-2-216.1

Bayne, E., H. Lankau, and J. Tigner. 2011. Ecologically-based criteria to assess the impact and recovery of seismic lines: the importance of width, regeneration, and seismic density. Report No. 192. Edmonton, Alberta, Canada. https://publications.gc.ca/collections/collection_2016/one-neb/NE22-4-192-eng.pdf

Bayne, E., L. Leston, C. L. Mahon, P. Sólymos, C. Machtans, H. Lankau, J. R. Ball, S. L. Van Wilgenburg, S. G. Cumming, T. Fontaine, F. K. A. Schmiegelow, and S. J. Song. 2016. Boreal bird abundance estimates within different energy sector disturbances vary with point count radius. Condor 118:376-390. https://doi.org/10.1650/CONDOR-15-126.1

Bayne, E. M., S. L. Van Wilgenburg, S. Boutin, and K. A. Hobson. 2005b. Modeling and field-testing of Ovenbird (Seiurus aurocapillus) responses to boreal forest dissection by energy sector development at multiple spatial scales. Landscape Ecology 20:203-216. https://doi.org/10.1007/s10980-004-2265-9

Billerman, S. M., B. K. Keeney, P. G. Rodewald, and T. S. Schulenberg, editors. 2022. Birds of the world. Cornell Laboratory of Ornithology, Ithaca, New York, USA. https://doi.org/10.2173/bow

Buckland, S. T. 1987. On the variable circular plot method of estimating animal density. Biometrics 43:363-384. https://doi.org/10.2307/2531819

Carpenter, T. M. 2020. Effects of energy development on habitat use of an avian peatland specialist and generalist at multiple spatial scales. Thesis. University of Alberta, Edmonton, Alberta, Canada. https://era.library.ualberta.ca/items/00950833-d823-4064-af1e-38a2fd30da2c

Carpenter, T. M., C. L. Mahon, E. Bayne, J. L. Keim, and S. E. Nielsen. 2022. Influence of in situ oilsands development on occurrence of an avian peatland generalist and specialist. Avian Conservation and Ecology 17(2):23. https://doi.org/10.5751/ACE-02206-170223

Chasko, G. G., and J. E. Gates. 1982. Avian habitat suitability along a transmission-line corridor in an oak-hickory forest region. Wildlife Monographs 82:1-41.

Cimprich, D. A., F. R. Moore, and M. P. Guilfoyle. 2020. Red-eyed Vireo (Vireo olivaceus), version 1.0. in P. G. Rodewald, editor. Birds of the world. Cornell Lab of Ornithology, Ithaca, New York, USA. https://doi.org/10.2173/bow.reevir1.01

Crosby, A. D., L. Leston, E. M. Bayne, P. Sólymos, C. L. Mahon, J. D. Toms, T. D. S. Docherty, and S. J. Song. 2023. Domains of scale in cumulative effects of energy sector development on boreal birds. Landscape Ecology 38(12):3173-3188. https://doi.org/10.1007/s10980-023-01779-8

Dabros, A., H. E. J. Hammond, J. Pinzon, B. Pinno, and D. Langor. 2017. Edge influence of low-impact seismic lines for oil exploration on upland forest vegetation in northern Alberta (Canada). Forest Ecology and Management 400:278-288. https://doi.org/10.1016/j.foreco.2017.06.030

Dabros, A., M. Pyper, and G. Castilla. 2018. Seismic lines in the boreal and arctic ecosystems of North America: environmental impacts, challenges, and opportunities. Environmental Reviews 26(2):214-229. https://doi.org/10.1139/er-2017-0080

De Cáceres, M., and P. Legendre. 2009. Associations between species and groups of sites: indices and statistical inference. Ecology 90(12):3566-3574. https://doi.org/10.1890/08-1823.1

Dehcho Land Use Planning Committee (DLUPC). 2006. Respect for the land: Dehcho Land Use Plan. Final Draft Plan-May 2006. DLUPC, Fort Providence, Northwest Territories, Canada. http://www.dehcholands.org

Dickie, M., R. Serrouya, C. DeMars, J. Cranston, and S. Boutin. 2017. Evaluating functional recovery of habitat for threatened woodland caribou. Ecosphere 8:e01936. https://doi.org/10.1002/ecs2.1936

Drolet, B., A. Desrochers, and M.-J. Fortin. 1999. Effects of landscape structure on nesting songbird distribution in a harvested boreal forest. Condor 101:699-704. https://doi.org/10.2307/1370205

Falls, J. B., and J. G. Kopachena. 2020. White-throated Sparrow (Zonotrichia albicollis), version 1.0 in A. F. Poole, editor. Birds of the world. Cornell Lab of Ornithology, Ithaca, New York, USA. https://doi.org/10.2173/bow.whtspa.01

Filicetti, A. T., M. Cody, and S. E. Nielsen. 2019. Caribou conservation: restoring trees on seismic lines in Alberta, Canada. Forests 10(2):185. https://doi.org/10.3390/f10020185

Finnegan, L., D. MacNearney, and K. E. Pigeon. 2018. Divergent patterns of understory forage growth after seismic line exploration: implications for caribou habitat restoration. Forest Ecology and Management 409:634-652. https://doi.org/10.1016/j.foreco.2017.12.010

Finnegan, L., K. E. Pigeon, and D. MacNearney. 2019. Predicting patterns of vegetation recovery on seismic lines: informing restoration based on understory species composition and growth. Forest Ecology and Management 446:175-192. https://doi.org/10.1016/j.foreco.2019.05.026

Flaspohler, D. J., S. A. Temple, and R. N. Rosenfield. 2001. Species-specific edge effects on nest success and breeding bird density in a forested landscape. Ecological Applications 11:32-46. https://doi.org/10.1890/1051-0761(2001)011[0032:SSEEON]2.0.CO;2

Forsman, J. T., P. Reunanen, J. Jokimäki, and M. Mönkkönen. 2010. The effects of small-scale disturbance on forest birds: a meta-analysis. Canadian Journal of Forest Research 40:1833-1842. https://doi.org/10.1139/X10-126

Franklin, C. M. A., A. T. Filicetti, and S. E. Nielsen. 2021. Seismic line width and orientation influence microclimatic forest edge gradients and tree regeneration. Forest Ecology and Management 492:119216. https://doi.org/10.1016/j.foreco.2021.119216

Government of Alberta. 2021. The Alberta Human Footprint Monitoring Program: a historical overview of geophysical exploration in Alberta. Alberta Environment and Parks. Policy Systems Branch, Policy Division, Geographic Science Team, Edmonton, Alberta, Canada. https://open.alberta.ca/publications/alberta-human-footprint-monitoring-program-historical-overview-geophysical-exploration

Greenberg, C. H., and J. D. Lanham. 2001. Breeding bird assemblages of hurricane-created gaps and adjacent closed canopy forest in the southern Appalachians. Forest Ecology and Management 154:251-260. https://doi.org/10.1016/S0378-1127(00)00631-9

Gregoire, J. M., R. W. Hedley, and E. M. Bayne. 2022. Canada Warbler response to vegetation structure on regenerating seismic lines. Avian Conservation and Ecology 17(2):26. https://doi.org/10.5751/ACE-02262-170226

Hobson, K. A., and J. Schiek. 1999. Changes in bird communities in boreal mixedwood forest: harvest and wildfire effects over 30 years. Ecological Applications 9(3):849-863. https://doi.org/10.2307/2641334

Jackson, D. A. 1993. Stopping rules in principal components analysis: a comparison of heuristical and statistical approaches. Ecology 74:2204-2214. https://doi.org/10.2307/1939574

Johnston, V. R. 1947. Breeding birds of the forest edge in Illinois. Condor 49:45-53. https://doi.org/10.2307/1364118

Kalukapuge, T., L. F. V. Leston, J. A. Martínez-Lanfranco, and E. Bayne. 2024. Response of boreal songbird communities to the width of linear features created by the energy sector in Alberta, Canada. Avian Conservation and Ecology 19(2):14. https://doi.org/10.5751/ACE-02730-190214

Kennett, S. 2006. From science-based thresholds to regulatory limits: implementation issues for cumulative effects management. Canadian Institute of Resources Law for Environment Canada, Northern Division. University of Calgary, Calgary, Alberta, Canada. https://doczz.net/doc/9005710/from-science-based-thresholds-to-regulatory-limits

King D. I., and B. E. Byers. 2002. An evaluation of powerline rights-of-way as habitat for early-successional shrubland birds. Wildlife Society Bulletin 30:868-874.

Lankau, H. E. G. 2014. Songbird responses to regenerating seismic lines in the boreal forest. Thesis. University of Alberta, Edmonton, Canada. https://doi.org/10.7939/R3ZT2K

Lankau, H., E. M. Bayne, and C. S. Machtans. 2013. Ovenbird (Seiurus aurocapilla) territory placement near seismic lines is influenced by forest regeneration and conspecific density. Avian Conservation and Ecology 8(1):5. https://doi.org/10.5751/ACE-00596-080105

Lee, P., and S. Boutin. 2006. Persistent and developmental transition of wide seismic lines in the western Boreal Plains of Canada. Journal of Environmental Management 78:240-250. https://doi.org/10.1016/j.jenvman.2005.03.016

Leopold, A. 1933. Game management. Charles Scribner’s Sons, New York, New York, USA.

Leston, L., E. Bayne, and F. K. A. Schmiegelow. 2018. Long-term changes in boreal forest occupancy within regenerating harvest units. Forest Ecology and Management 421:40-53. https://doi.org/10.1016/j.foreco.2018.02.029

Leston, L., E. Bayne, J. D. Toms, C. L. Mahon, A. Crosby, P. Sólymos, J. Ball, S. J. Song, F. K. A. Schmiegelow, D. Stralberg, and T. D. S. Docherty. 2023. Comparing alternative methods of modelling cumulative effects of oil and gas footprint on boreal bird abundance. Landscape Ecology 38:147-168. https://doi.org/10.1007/s10980-022-01531-8

MacFarlane, A. K. 2003. Vegetation response to seismic lines: edge effects and on-line succession. Thesis. University of Alberta, Edmonton, Alberta, Canada. https://central.bac-lac.gc.ca/.item?id=MQ87819&op=pdf&app=Library&is_thesis=1&oclc_number=57281770

Machtans, C. S. 2006. Songbird response to seismic lines in the western boreal forest: a manipulative experiment. Canadian Journal of Zoology 84:1421-1430. https://doi.org/10.1139/z06-134

Mahon, C. L., G. L. Holloway, E. M. Bayne, and J. D. Toms. 2019. Additive and interactive cumulative effects on boreal landbirds: winners and losers in a multi‐stressor landscape. Ecological Applications 29:e01895. https://doi.org/10.1002/eap.1895

Martinez, A. P. 2017. Pairwise multilevel comparison using Adonis. Version 0.4.1. GitHUB. https://github.com/pmartinezarbizu/pairwiseAdonis

Matsuoka, S. M., E. M. Bayne, P. Sólymos, P. C. Fontaine, S. G. Cumming, F. K. A. Schmiegelow, and S. J. Song. 2012. Using binomial distance-sampling models to estimate the effective detection radius of point-count surveys across boreal Canada. Auk 129:268-282. https://doi.org/10.1525/auk.2012.11190

Mazerolle, M. J. 2023. AICcmodavg: model selection and multimodel inference based on (Q)AIC(c). R package version 2.3.2. https://doi.org/10.32614/CRAN.package.AICcmodavg

Miller, S. G., R. L. Knight, and C. K. Miller. 1998. Influence of recreational trails on breeding bird communities. Ecological Applications 8:162-169. https://doi.org/10.1890/1051-0761(1998)008[0162:IORTOB]2.0.CO;2

Nitschke, C. R. 2008. The cumulative effects of resource development on biodiversity and ecological integrity in the Peace-Moberly region of Northeast British Columbia, Canada. Biodiversity Conservation 17:1715-1740. https://doi.org/10.1007/s10531-008-9376-6

Norton, M. R., and S. J. Hannon. 1997. Songbird response to partial-cut logging in the boreal mixedwood forest of Alberta. Canadian Journal of Forest Research 27:44-53. https://doi.org/10.1139/x96-149

Nuttle, T. 1998. Densiometer bias? Are we measuring the forest or the trees? Wildlife Society Bulletin 25:610-611.

Ortega, Y. K., and D. E. Capen. 1999. Effects of forest roads on habitat quality for Ovenbirds in a forested landscape. Auk 116:937-946. https://doi.org/10.2307/4089673

Parker, T. H., B. M. Stansberry, C. D. Becker, and P. S. Gipson. 2005. Edge and area effects on the occurrence of migrant forest songbirds. Conservation Biology 19:1157-1167. https://doi.org/10.1111/j.1523-1739.2005.00107.x

R Core Team. 2022. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/

Rail, J.-F., M. Darveau, A. Desrochers, and J. Huot. 1997. Territorial responses of boreal forest birds to habitat gaps. Condor 99:976-980. https://doi.org/10.2307/1370150

Revel, R. D., T. D. Dougherty, and D. J. Downing. 1984. Forest growth and regeneration along seismic lines. University of Calgary Press, Calgary, Alberta, Canada.

Riva, F., J. H. Acorn, and S. E. Nielsen. 2018. Localized disturbances from oil sands developments increase butterfly diversity and abundance in Alberta’s boreal forests. Biological Conservation 217:173-180. https://doi.org/10.1016/j.biocon.2017.10.022

Robinson, W. D., and S. K. Robinson. 1999. Effects of selective logging on forest bird populations in a fragmented landscape. Conservation Biology 13:58-66. https://doi.org/10.1046/j.1523-1739.1999.97226.x

Sherry, T. W., R. T. Holmes, P. Pyle, and M. A. Patten. 2020. American Redstart (Setophaga ruticilla), version 1.0. in P. G. Rodewald, editor. Birds of the world. Cornell Lab of Ornithology, Ithaca, New York, USA. https://doi.org/10.2173/bow.amered.01

Sólymos, P., S. M. Matsuoka, E. M. Bayne, S. E. Lele, P. Fontaine, S. G. Cumming, D. Stralberg, F. K. A. Schmiegelow, and S. J. Song. 2013. Calibrating indices of avian density from non-standardized survey data: making the most of a messy situation. Methods in Ecology and Evolution 4:1047-1058. https://doi.org/10.1111/2041-210X.12106

Stern, E. R., F. Riva, and S. E. Nielsen. 2018. Effects of narrow linear disturbances on light and wind patterns in fragmented boreal forests in northeastern Alberta. Forests 9(8):486. https://doi.org/10.3390/f9080486

Suarez, A. V., K. S. Pfennig, and S. K. Robinson. 1997. Nesting success of a disturbance-dependent songbird on different kinds of edges. Conservation Biology 11:928-935. https://doi.org/10.1046/j.1523-1739.1997.96063.x

Thomas, L., S. T. Buckland, E. A. Rexstad, J. L. Laake, S. Strindberg, S. L. Hedley, J. R. B. Bishop, T. A. Marques, and K. P. Burnham. 2010. Distance software: design and analysis of distance sampling surveys for estimating population size. Journal of Applied Ecology 47:5-14. https://doi.org/10.1111/j.1365-2664.2009.01737.x

Tigner, J., E. M. Bayne, and S. Boutin. 2014. Black bear use of seismic lines in Northern Canada. Journal of Wildlife Management 78(2):282-292. https://doi.org/10.1002/jwmg.664

Tigner, J., E. M. Bayne, and S. Boutin. 2015. American marten respond to seismic lines in northern Canada at two spatial scales. PLoS ONE 10(3):e0118720. https://doi.org/10.1371/journal.pone.0118720

van Rensen, C. K., S. E. Neilsen, B. White, T. Vinge, and V. J. Lieffers. 2015. Natural regeneration of forest vegetation on legacy seismic lines in boreal habitats in Alberta’s oil sands region. Biological Conservation 184:127-135. https://doi.org/10.1016/j.biocon.2015.01.020

Wang, W., N. Maruyama, L. Bo-wen, H. Morimoto, and G. Zhong-xin. 2002. Relationships between bird communities and vegetation structure in Honghua’erji, northern Inner Mongolia. Journal of Forestry Research 13:294-298. https://doi.org/10.1007/BF02860095

Willson, M. F. 1974. Avian community organization and habitat structure. Ecology 55:1017-1029. https://doi.org/10.2307/1940352