In North America, passerine birds migrate in a broad-front fashion between breeding and nonbreeding grounds along relatively well-documented corridors, stopping for periods during these migrations to rest and replenish fuel stores (Gauthreaux 1991, Able 2004, Liechti et al. 2013). At this broad spatial scale (macroscale of 100s of km), directionality of migration movements are predictable (Mabey 2004). The site-specific timing and spatial patterns of movements of migratory birds at individual locations (microscale of 1–10 km) along these corridors are less understood, particularly because variation in weather and topography can influence these patterns (Mabey 2004, Liechti et al. 2013, Pocewicz et al. 2013). Understanding movement patterns on these smaller microscales, though, is required to identify whether anthropogenic developments, such as wind installations or other tall structures built along migration corridors, can lead to migratory disruption or collision risk.
Nocturnally migrating passerines have been recorded at a range of altitudes up to and over 5000 m (Liechti and Schaller 1999, Liechti and Schmaljohann 2007, Schmaljohann et al. 2009), but in some instances a significant number of birds have been found to migrate in the lower altitudes, below 1500 m above ground level (agl; Able 2004, Mabee et al. 2006, Longcore et al. 2008, Schmaljohann et al. 2008, Dokter et al. 2011, Kemp et al. 2013). Depending on site, this range of lower flight altitudes may occur in one season only (Schmaljohann et al. 2009) or during both spring and fall movements (Bruderer 1997). Migrants will typically climb to altitudes where they encounter favorable winds, which are used to maximize flight ranges on a given amount of fuel stores (Klaassen et al. 2012, Marques et al. 2014). Wind layers at greater heights above ground often are less subject to surface friction, and thus create more stable and predictable winds to facilitate migration (Klaassen et al. 2012, Dokter et al. 2013), however, choice of altitudes for migratory movement must also be balanced against constraints imposed by extreme temperatures and lower oxygen for respiration with increasing altitude (Able 2004, Schmaljohann et al. 2008). This results in nocturnal migrants settling their movement each night within the most profitable wind layer available that balances these factors (Kemp et al. 2013). Flight altitudes generally peak early in the evening and are typically higher than those observed later in the evening/early morning (Mabee et al. 2006); this altitudinal profile of nocturnal migrants during a night, though, is influenced by topography and wind (Bruderer 1997). Because of the lower flight altitudes relative to ground level imposed on birds when moving over mountain ranges, migrants can be exposed to terrain forced winds that may oppose the principal direction of migration (Bruderer 1997, Liechti et al. 2013). Under such conditions, the flight paths of migrants may be scattered and subject to topographic features and structures on the landscape (Mabey 2004), further adding to the unpredictable patterns of microscale movement.
Understanding microscale movement patterns, and the factors that govern them, may prove a useful tool in determining the potential impact of wind power development on passerine migration. During the spring and fall periods, nocturnally migrating passerines are the most abundant avian group encountering wind energy facilities (Marques et al. 2014), which is reflected in the high proportion of passerine carcasses that are typically found at wind projects (Johnson et al. 2002, Zimmerling et al. 2013, Erickson et al. 2014). Passerines typically compose 80% of all fatalities at such installations, most of which involve nocturnal migrants (Mabee et al. 2006, Kuvlesky et al. 2007). Although this likely constitutes a small fraction of overall population sizes (Loss et al. 2013, Zimmerling et al. 2013), effects on birds remains an integral component of environmental assessments for wind energy developments (Zimmerling et al. 2013) because of their potential to add to cumulative impacts on avian populations. The general knowledge of the interactions between birds and wind turbines has substantially increased since the infancy of the wind energy industry (Marques et al. 2014), but before-and-after development studies on how wind facilities affect the migratory behavior of passerines are still sparse (Kuvlesky et al. 2007).
Avoidance behavior by nocturnal migrants will strongly influence the mortality rates observed at wind energy projects (Chamberlain et al. 2006), but the scale of avoidance to wind turbines for nocturnally migrating birds is almost unknown (Liechti et al. 2013). Two levels of magnitude are expected for the avoidance of wind turbines: (1) macroscale avoidance where birds alter their flight path to circumnavigate an entire wind energy installation; and (2) microscale avoidance where birds alter their flight movements while they are passing within the boundaries of a wind energy installation (Marques et al. 2014). In this study, we documented the microscale patterns of nocturnal migrant movements through a 144 MW wind energy project in northeast British Columbia, Canada. Using X-band marine radar units, equipped with an electronic interface system, we recorded the movement patterns and altitudes of migrants to determine if spatial patterns of flight differ between preoperational and operational periods of the wind energy facility.
We collected radar data on nocturnal migrants at the Dokie I Wind Energy Project located in northeast British Columbia, Canada (55°41'28"N 12218'06"W) during the spring and fall migration periods from 2008 to 2012 (Fig. 1). The site is located in the eastern foothills of what is considered the Northern Rockies, which lie in a north-northwest to south-southeast orientation. The project is situated on two ridges ranging in elevation from 1200 m to 1400 m above sea level. The project underwent site construction (roads, turbine pads, etc.) from 2008–2009. At this time five widely-spaced (> 1 km apart) turbine towers were initially constructed, of which only two had blades attached and none contained any operational hydraulics, and so were inactive. Construction then halted until the late summer of 2010. Turbine erection resumed following our spring surveys, and several additional turbines were erected but nonoperational during the fall survey, and the full construction completed in the fall following our survey period. Turbines become operational and commenced energy production in early 2011 prior to the spring migration period of that year. The Dokie I Wind Energy Project is a 144 megawatts (MW) installation comprising 48 Vestas V90 3MW wind turbines that have a tower height of 80 m, a rotor diameter of 90 m and a rotor swept area of 6362 m². Fifteen turbines are placed on the smaller, southern ridge and 33 turbines on the larger, northern ridge. For analytical purposes, we considered the first three years of the study (2008–2010) the preoperational period for the wind energy facility and the final two years (2011–2012) to represent the operational period, based upon whether turbines were actively rotating during these periods.
We recorded movement patterns of nocturnal migrants around the wind project using two Furuno X-band marine radar units (model 1954C, 12kW, 9,000 MHz, 1.83 m open array antennas with beam width of 1.9° horizontal and 22° vertical - Furuno Electric Company Ltd. Miki Japan) equipped with an electronic interface system (signal digitizer [XIR3000B] and WinHorizon software [Version 184.108.40.206 - Russell Technologies Inc., North Vancouver, BC; http://www.russelltechnologies.ca/). Details on radar set up and calibration of the avian detection system used are provided as supplementary material. One of the radar units was set in the surveillance position (antenna rotating on the typical horizontal axis) to record passage rates and determine bearing of migrants. This radar’s antenna was custom angled upwards to 15° such that the main lobes of the beam span from ~ +4° to +26° above the horizon. The second unit was set in the vertical position (radar mounted at 90°, so that the antenna rotated through the vertical axis) to record heights of targets. Both radars were set to 1.5 km detection range on short-pulse length (80 ns at pulse repetition frequency [PRF] = 2100 Hz). Sea clutter and rain removal settings were turned off, and gain set at maximum level (76 on scale of 0–100) that balanced maximum resolution without introducing clutter (see supplemental material for additional information on radar set up). These settings were standardized among years and the radars were set in the same locations each season. The radar units were set in locations where minimal interference from ground clutter was present. The horizontal radar was mounted approximately 2.0 m above ground and oriented to true north. The vertical radar was mounted approximately 1.5 m above the ground and the antenna was aligned with the proposed/constructed turbine arrays allowing us to determine heights of targets as they passed above the turbine strings (Fig. 1). Radars were generally operational from 21:00 to 05:00 each night in the spring and 20:00 to 06:00 in the fall, which reflected sunset to sunrise in each season. Recording dates varied slightly among years, but surveys were timed to coincide with the previously documented peak periods of spring passerine migration (mid- to late-May) and fall migration (late-August to early September) each year (Jacques Whitford-AXYS Ltd. 2006, Pomeroy et al. 2007).
Radar imagery was analyzed using radR (Taylor et al. 2010), an open-source, R-based platform (https://radr-project.org/). This platform uses algorithms to distinguish and track moving targets from stationary objects. radR uses the first five rotations of the radar to identify and ignore stationary returns. It then searches for detections moving in predictable paths, based on user inputs for specific variables, such as target size and the number of successive rotations on which targets are detected before initiating tracking. We tested a variety of settings for each of four user-defined settings that accounted for the greatest variability in detecting/tracking targets (see Table A1.1), utilizing the combination of settings that results in the highest congruence between known manually tracked targets and those autotracked by radR (R² = 0.94; see Fig. A1.1). Output data on tracked targets from both radars were further processed in the statistical program R (R Development Core Team 2017) to compile specific data on track location, length, bearing, and speed (horizontal radar data). Height of targets as they passed over the vertical radars were determined at a detection range of 1.5 km, but to further investigate the number of birds aloft in lower altitudes closer to the turbines, we also recorded the number of targets detected in six height categories (0–150 m, 151–300 m, 301–450 m, 451–600 m, 601–750 m, and > 751 m agl; vertical radar data).
We obtained wind vector data from the National Centers for Environmental Prediction (NCEP) North American Regional Reanalysis (NARR) dataset using the three hour composite data for the zonal wind (U wind) and meridional wind (V wind). The latitude and longitude of the study site was matched with the nearest NARR 32 km by 32 km grid cell center, which was approximately 10 km from the study site, and we obtained data from the 825 mb pressure level, which corresponds to approximately 500 m agl at the study site. This altitude was chosen because it corresponded to average heights of migrants detected during preliminary analysis. We calculated the wind vector as the direction (degrees), with respect to true north, toward which the wind was blowing for each hour of surveying across both seasons and all years. For each period, we also recorded the wind speed in m/s.
Utilizing information on the average bearing and flight speed of migrants across years, coupled with the wind direction and speed for each hour of monitoring, we calculated a measure of flow-assisted flight for migrants using the Wind Profit equation of Kemp et al. (2013). This equation utilizes the seasonal migratory “goal,” the direction in which migrants are attempting to travel, and the assumed average migratory flight speed that birds are attempting to maintain. We set the migratory goal as the average bearing of migrant targets detected on the horizontal radar for each season (all years combined), which we calculated across all individual tracks using CircStats (Agostinelli 2009) package in R (vs 3.4.0, R Development Core Team 2017). We also report the relative length vector of this average bearing (rho), which ranges from 0–1 with higher values indicating a longer vector, e.g., more focused directionality to mean bearing. We assumed an average migratory flight speed of 12 m/s (Bruderer and Boldt 2001, Kemp et al. 2013), which also closely matches the average flight speed of radar targets in our study under low-wind conditions (unpublished data). Kemp et al.’s (2012, 2013) flow-assisted flight (or Wind Profit) assumes that migrants attempt to maintain this migratory goal and average flight speed, and that the wind direction/wind speed can either assist this, e.g., a strong tailwind, or hinder this, e.g., strong headwind, objective. The flow-assisted value is a function of the angular difference between wind direction and the direction of the birds’ migratory goal, taking into account the wind speed. Derived values are vector scores that range from ~ +20 (full tailwind at wind speed that provides full flight assistance to maintain migratory flight speed) to ~ -20 (a headwind that requires high energetic outlay from the migrant to maintain the target flight speed and bearing). All values between these indicate varying wind displacement for which the birds must partially compensate to maintain intended direction and speed.
We calculated the wind profit for each hour of monitoring for all nights across both seasons and all years for which radar data was gathered. We then used a General Linear Mixed Model (lme4 package for R; Bates et al. 2015) to compare the hourly flow-assistance (dependent variable) offered to migrants by both season (spring vs fall) and operational phase of the project (preoperational vs operational; fixed effects) while controlling for variation from individual nights as a random effect.
We determined flight altitudes (meters agl) of targets using the vertical radars as they passed above the installation. Radars have higher likelihood of detecting small targets closer to the radar, but simultaneously the spreading of the beam width means a larger volume of sky is sampled at greater distances. Although this provides a high likelihood, with the radar setting used, of detecting passerine-sized targets up to at least 1000–1200 m from the radar, and larger targets to the full 1500 m detection radius, the distribution of target heights was still slightly skewed. As a result, we used a General Additive Model (GAM) with a gamma function and log link using the mgcv package for R (Wood 2006) to compare the flight altitudes of migrants (dependent variable) in relation to the operational phase (preoperational vs operational) with hour of night after sunset as the smoothed function. Each season, spring and fall, was run as a separate model.
To further investigate the effect of turbine operation on migration, we determined the number of birds aloft in each 150 m height interval during the preoperational vs operational years. We compared the number of birds aloft in the 0–150 m height category—the heights overlapping the turbine area, and thus constituting the highest potential collision risk—with the total number of detected migrants in height bins > 150 m, e.g., above turbines, using a generalized linear mixed model (glmer) in the package lme4 using a binomial model with a logit link. Operational phase (preoperation vs operational) was added as a fixed effect, and variation accounted for by individual nights added as a random effect. All graphics were created using either QGIS (vs 2.8), or the ggplot2 (Wickham 2009) package for R.
Over the five-year survey period we autotracked a total of 2,293,814 aerial targets with the horizontal radar and a total of 598,834 aerial targets with the vertical radar (Tables 1 and 2) with the radR processing. Periods of poor weather were excluded from the survey effort, which partially explains the variation in annual and seasonal survey effort. Technical difficulties with the radar equipment also caused minor gaps in the collection of data and these periods were excluded from the survey effort. In general, more targets were detected by the horizontal radar because of the larger detection area being covered by the radars in their respective orientations. Finally, a higher number of targets were consistently detected in the fall migration (Tables 1 and 2).
Across the survey period the predominant wind vectors at the study site were generally to the northwest. The circular mean wind vector across all spring migration periods (2008–2012) was 355.25° (rho = 0.65) and the circular mean wind vector in for the fall migration periods was 341.27° (rho = 0.74). The circular mean target track bearing for the spring migrants (all years combined) was 301.39° (rho = 0.45), while the circular mean target track bearing for the fall was 125.95° (rho = 0.27; Fig. 2). Because wind directionality was relatively consistent during both migration periods, most transit during the northward spring migration occurred during tailwind conditions, while southern migration in the fall is predominantly under headwind conditions. Indeed, flow-assistance (wind profit) for migratory movement was higher in spring than fall (F1, 536 = 358.34, P < 0.0001; Fig.3), but did not differ within season based on operational phase of the wind installation (Operational Phase: F1, 536 = 0.04, P = 0.83; Interaction between Season x Operational Phase: F1, 536 = 2.18, P = 0.14).
The altitude of targets detected during spring migrations were significantly lower during the preoperational phase relative to the operational phase (GAM model: Estimate = -0.21 ± 0.0034 SE, t = -63.06, P < 0.0001; Fig. 4). The median altitude during the preoperational years was 360 m agl, whereas during the two operational years it was 453 m agl. The approximate estimate of the smoothing term explaining variation in altitudes across time of night was also significant (edf = 5.92 F = 197.8 P < 0.0001, R²(adj) = 0.029 Deviance explained = 2.2%); there was a slight rise in altitudes from the first hours after sunset which then stabilized and remained steady for the remainder of the night. There appeared to be a similar pattern between preoperational and operational phases of the wind installation, but shifted upward in the operational phase (Fig. 4).
Similarly, the altitude of targets detected during fall migrations were also significantly lower during the preoperational phase (median = 321 m agl) relative to the operational phase (median = 389 (GAM model: Estimate = -0.16 ± 0.0021 SE, t = - 76.58, P < 0.0001; Fig. 4), and the smoothing term explaining variation in altitudes across time of night was also significant (edf = 5.90 F = 1286.00, P < 0.0001, R²(adj) = 0.027 Deviance explained = 2.4%). As with spring migration, that altitudes of targets in the fall tended to be lowest early in the evening, increasing and then staying relatively stable across the remaining hours of tracking throughout the night (Fig. 4).
When controlling for the effects of individual nights as a random effect, there was no difference in the number of targets detected in the lowest height category (0–150 m agl) relative to the other altitudes because of either season (binomial GLM: z = 0.12, P = 0.90) or operational phase (z = 1.17, P = 0.24), nor was there an interaction between these variables (season x operational phase: z = -0.083, P = 0.93; Fig. 5). However, the number of birds in the second altitude category (151–300 m agl) relative to other height categories did differ by operational phase of the wind installation (z = 2.13, P = 0.033), but not by season (z = -1.45, P = 0.15); the proportion of birds in this altitude category decreased during the operational phase of the study relative to the preoperational phase, and the effect was similar in both spring and fall migration (Fig. 5). There was no interaction effect between season and operational phase in the number of targets detected in the 151–300 m agl altitude category (z = 1.20, P = 0.23).
The number of nocturnal migrants per hour detected around the Dokie I Wind Energy facility varied by season, with generally higher passage rates in the fall than in the spring on both the vertical and horizontal radars. This is expected because both adult and juvenile birds are moving during the postbreeding season in the fall (Harmata et al. 2000, Otter et al. 2014). Further, some fall migrants may have been bats, which are more prominent in the fall migration (Kunz et al. 2007). The fall migratory season, however, had less favorable winds that provided much lower flow-assistance to migrants than occurred in the spring. One response to the difference in flow-assistance between seasons we may have expected would be for birds to fly at different altitudes (e.g., Kemp et al. 2013) but we found that the vertical distribution of migrants was similar between seasons in the preoperational surveys, either measured in median altitudes or proportions of detected migrants in each 150 m height category. Where we did see shifts in altitude was in response to operational phase of the turbines, but the effect was largely parallel in either season despite the seasonal differences in flow-assistance.
If migrants were adjusting height to avoid collisions with operational wind turbines, we expected to see a reduction in the number of targets at the lowest height category (0–150m agl, co-occurring with the turbines) during operational phases of the wind farm. The number of targets detected in this lowest height categories (0–150m agl) was, however, not affected by operational phase. Yet, in preoperational years, the highest proportion of migrants were found not in this lowest height category, but in the airspace immediately above (151–300m agl). During the years in which turbines were operational, there was a significant reduction in the proportion of targets detected in this airspace, instead appearing to shift into higher height categories (Fig 5). This was also reflected in a higher median hourly altitude of migrants tracked during the operational period, which occurred in both seasons. As altitudes were increased by approximately 70 m (fall) to 90 m (spring) upwards, this suggests that nocturnal migrants may respond to the presence of the wind project by adjusting their altitudes.
The wind turbines in our study site were 120 m tall and during the operational phase of the study there were still migrants moving within this airspace. However, this constituted less than 20% of the detected targets using the airspace around the wind turbines even under preoperational conditions. Although there was no reduction in the proportion of targets within this airspace when turbines were operational, our vertical radar could not resolve whether those targets were making adjustments in their flight paths to move in the airspace between turbines. Regardless, our data do support for microscale adjustments in migratory behavior because birds increased flight altitude by 70–90 m in response to the presence of wind turbines. Because the median heights of nocturnal migrants were 200 m or more above turbine heights (125 m) during preoperational surveys in either season, this suggested the impact of development from this particular wind energy project was low.
One caveat to this assessment is that our radar surveys were conducted on clear nights. This is a constraint of utilizing X-band radars; because these utilize wavelengths capable of resolving passerine-sized targets, they also resolve water vapour (fog, clouds, rain) and this precludes accurately surveying during precipitation and dense fog. However, these periods have been implicated as potentially high-risk weather conditions for bird collisions with wind turbines (Marques et al. 2014). However, we feel that the data from our studies likely reflects a low collision potential at this particular installation, because our low predicted collision rates are consistent with postconstruction carcass searching on the site. These searches yielded very low estimated collision rates of < 0.01% (Stantec Consulting Ltd. 2012a,b, Otter et al. 2014). Further, those carcass search estimates were conducted almost daily from March through October in the two operational years, covering a greater period and varying weather conditions than were possible with this radar tracking study. Thus, although adjustments to altitude during the operational phase were minor, our results and those of carcass searches suggest that turbines at this site are detected and minor altitudinal adjustments are made that would serve to further reduce already low collision risk, similar to flight adjustment patterns of diurnal migrants at this same installation (Johnston et al. 2014). The low collision rates at this facility may reflect its location; the site appears to be situated on a migratory route, but does not appear to constitute a major stop-over site or wintering ground, and is no more extensively used as a breeding site than other areas in the region. Most detected nocturnal and diurnal migrants appear to be simply moving through the site.
Avoidance rates of nocturnally migrating species toward wind turbines have not been extensively studied, but conservative estimates of 98% to 99% avoidance have been used in collision-risk models (Chamberlain et al. 2006, Liechti et al. 2013). Nocturnal migrants primarily comprise passerine species that are relatively abundant and widespread so subtle changes in avoidance rates can have large implications on the accuracy of collision-risk models (Chamberlain et al. 2006). Microscale avoidance at our study site seems restricted to small adjustments to increase altitude during the operational phase of the wind installation, however, quantifying the degree of microscale avoidance remains a challenge because some migrants are flying below the heights of turbines and may be taking evasive actions at individual turbines. Preliminary results from night vision cameras set up at and between wind turbines in the study site suggested that there may have been a slight decrease in the number birds moving near turbines compared to the number of birds detected between turbines, which would influence avoidance rates (Walsh 2012).
In conclusion, at the microscale level, nocturnal migrants showed some indications of adjusting their movements around the wind energy facility during the operational period, yet their typical migratory behavior was also not placing them in collision risk situations for the most part. Subtle adjustments in altitude may be occurring, which would further reduce collision risk at the wind project level. Changes at the macroscale level may be occurring to avoid wind energy facilities, although we found little evidence suggesting fewer detected targets during migratory seasons during the operational phase of the installation.
The radar data from this study showed that there was high nocturnal migration movement through the northeast British Columbia region in both spring and fall, providing context to the number of migrants potentially exposed to collision-risk situations (Otter et al. 2014). At the Dokie I Wind Energy Project, during two years of postconstruction mortality monitoring an estimate of 35 fatalities (birds and bats combined) were found during the periods when we were conducting radar surveys (Otter et al. 2014). After correcting these numbers for searcher efficiency and scavenger impact, the estimated annual mortality rate was < 0.01% of known migrants (Otter et al. 2014), but this estimate was based solely on the total number of detected radar tracks during the operational years. It did not account for the vertical stratification of migrants, and the data in the current manuscript to suggest that migrants may adjust altitude in response to operational turbines. The detailed data on nocturnal migration could suggest that avoidance behavior in postconstruction can reduce estimates of collision risk that were based on preconstruction monitoring. These avoidance rates may, however, reflect regional variation in collision risk, so we recommend using similar methodology to detail passage rates in relation to mortality rates in postconstruction years to derive accurate, local collision risk estimates. Further, avoidance behavior may vary by avian guild, and the way in which species utilize the site. The greatest impacts of wind installations on avian populations appears to coincide with placement of turbines in important breeding sites, particularly when placed between nesting and feeding grounds (e.g., Stienen et al. 2008, Dahl et al. 2013), migratory bottlenecks, stop-over or wintering sites (Drewitt and Langston 2006). Thus, siting of wind developments in relation to avian use and behavior within sites should likely be the first consideration in management (Marques et al. 2014).
We thank the many people involved in the collection of the radar data including D. Walsh, N. Johnston, A. Pomeroy, and M. Schmidt, as well as J. Brzustowski and P. Taylor for their assistance with the program radR. H. Russell, M. Laziner, and S. Yu provided technical support on the operation of the radar digital interface system. Stefanie LaZerte was instrumental in helping with the R coding for data management and analysis. We also thank Rob Bryce for operating his Easy Star II Airplane and Greg Sanders at Tech Helicopters for use of the Robertson R22 in radar calibration. The Dokie General Partnership, EarthFirst Canada Inc., and Stantec provided logistic and financial support. Additional funding support was provided by the Natural Sciences and Engineering Research Council (NSERC) Strategic Projects Grants, Special Opportunities Grants, and Collaborative Research and Development Grants programs, Environment Canada Grant-in-Aid of Research, UNBC, and Aboriginal Affairs and Northern Development Canada’s Northern Scientific Training Program. We acknowledge and thank the West Moberly First Nation for supporting our research in their traditional territory and H. Schmaljohann, F. Liechti, and an anonymous reviewer for their comments on earlier drafts of the manuscript.
Able, K. P. 2004. Birds on the move: flight and migration. Cornell Lab of Ornithology, Ithaca, New York, in association with Princeton University Press, Princeton, New Jersey, USA.
Agostinelli, C. 2009. Circular Statistics, from “Topics in circular Statistics” (2001) S. Rao Jammalamadaka and A. SenGupta, World Scientific. R Foundation for Statistical Computing, Vienna, Austria.
Bates, D., M. Mächler, B. Bolker, and S. Walker. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67:1-48. http://dx.doi.org/10.18637/jss.v067.i01
Bruderer, B. 1997. The study of bird migration by radar - Part 2: Major achievements. Naturwissenschaften 84:45-54. http://dx.doi.org/10.1007/s001140050348
Bruderer, B., and A. Boldt. 2001. Flight characteristics of birds: I. radar measurements of speed. Ibis 143:178-204. http://dx.doi.org/10.1111/j.1474-919X.2001.tb04475.x
Chamberlain, D. E., M. R. Rehfisch, A. D. Fox, M. Desholm, and S. J. Anthony. 2006. The effect of avoidance rates on bird mortality predictions made by wind turbine collision risk models. Ibis 148:198-202. http://dx.doi.org/10.1111/j.1474-919X.2006.00507.x
Dahl, E. L., R. May, P. L. Hoel, K. Bevanger, H. C. Pedersen, E. Røskaft, and B. G. Stokke. 2013. White-tailed eagles (Haliaeetus albicilla) at the Smøla wind-power plant, Central Norway, lack behavioral flight responses to wind turbines. Wildlife Society Bulletin 37:66-74. http://dx.doi.org/10.1002/wsb.258
Dokter, A. M., F. Liechti, H. Stark, L. Delobbe, P. Tabary, and I. Holleman. 2011. Bird migration flight altitudes studied by a network of operational weather radars. Journal of the Royal Society Interface 8:30-43. http://dx.doi.org/10.1098/rsif.2010.0116
Dokter, A. M., J. Shamoun-Baranes, M. U. Kemp, S. Tijm, and I. Holleman. 2013. High altitude migration at temperate latitudes: a synoptic perspective on wind assistance. PLoS ONE 8(1):e523000. http://dx.doi.org/10.1371/journal.pone.0052300
Drewitt, A. L., and R. H. W. Langston. 2006. Assessing the impacts of wind farms on birds. Ibis 148:29-42. http://dx.doi.org/10.1111/j.1474-919X.2006.00516.x
Erickson, W. P., M. M. Wolfe, K. J. Bay, D. H. Johnson, and J. L. Gehring. 2014. A comprehensive analysis of small-passerine fatalities from collisions with turbines at wind energy facilities. PLoS ONE 9(9):e107491. http://dx.doi.org/10.1371/journal.pone.0107491
Gauthreaux, S. A. 1991. The flight behavior of migrating birds in changing wind fields: radar and visual analyses. Integrative & Comparative Biology 31:187-204. http://dx.doi.org/10.1093/icb/31.1.187
Harmata, A. R., K. M. Podruzky, J. R. Zelenak, and M. L. Morrison. 2000. Passage rates and timing of bird migration in Montana. American Midland Naturalist 143:30-40. http://dx.doi.org/10.1674/0003-0031(2000)143[0030:PRATOB]2.0.CO;2
Jacques Whitford AXYS Ltd. 2006. Wildlife monitoring report - Fall 2006 - Dokie Wind Energy Project. Jacques Whitford AXYS Ltd., Burnaby, British Columbia, Canada.
Johnson, G. D., W. P. Erickson, M. D. Strickland, M. F. Shepard, D. A. Shepard, and S. A. Sarappo. 2002. Collision mortality of local and migrant birds at a large-scale wind-power development on Buffalo Ridge, Minnesota. Wildlife Society Bulletin 30:879-887.
Johnston, N. N., J. E. Bradley, and K. A. Otter. 2014. Increased flight altitudes among migrating Golden Eagles suggest turbine avoidance at a Rocky Mountain wind installation. PLoS ONE 9(3):e93030. http://dx.doi.org/10.1371/journal.pone.0093030
Kemp, M. U., J. Shamoun-Baranes, A. M. Dokter, E. van Loon, and W. Bouten. 2013. The influence of weather on the flight altitude of nocturnal migrants in mid-latitudes. Ibis 155:734-749. http://dx.doi.org/10.1111/ibi.12064
Kemp, M. U., J. Shamoun-Baranes, E. van Loon, J. D. McLaren, A. M. Dokter, and W. Bouten. 2012. Quantifying flow-assistance and implications for movement research. Journal of Theoretical Biology 308:56-67. http://dx.doi.org/10.1016/j.jtbi.2012.05.026
Klaassen, M., B. J. Hoye, B. A. Nolet, and W. A. Buttemer. 2012. Ecophysiology of avian migration in the face of current global hazards. Philiosophical Transactions of the Royal Society B 367:1719-1732. http://dx.doi.org/10.1098/rstb.2012.0008
Kunz, T. H., E. B. Arnett, B. M. Cooper, W. P. Erickson, R. P. Larkin, T. J. Mabee, M. L. Morrison, M. D. Strickland, and J. M. Szewzak. 2007. Assessing impacts of wind-energy development on nocturnally active birds and bats: a guidance document. Journal of Wildlife Management 71:2449-2486. http://dx.doi.org/10.2193/2007-270
Kuvlesky, W. P., L. A. Brennan, M. L. Morrison, K. K. Boydston, B. M. Ballard, and F. C. Bryant. 2007. Wind energy development and wildlife conservation: challenges and opportunities. Journal of Wildlife Management 71:2487-2498. http://dx.doi.org/10.2193/2007-248
Liechti, F., J. Guélat, and S. Komenda-Zehnder. 2013. Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines. Biological Conservation 162:24-32. http://dx.doi.org/10.1016/j.biocon.2013.03.018
Liechti, F., and E. Schaller. 1999. The use of low-level jets by migrating birds. Naturwissenschaften 86:549-551. http://dx.doi.org/10.1007/s001140050673
Liechti, F., and H. Schmaljohann. 2007. Wind-governed flight altitudes of nocturnal spring migrants over the Sahara governed by wind. Ostrich 78:337-341. http://dx.doi.org/10.2989/OSTRICH.2007.78.2.35.115
Longcore, T., C. Rich, and S. A. Gauthreaux. 2008. Height, guy wires, and steady burning lights increase hazard of communication towers to nocturnal migrants: a review and meta-analysis. Auk 125:485-492. http://dx.doi.org/10.1525/auk.2008.06253
Loss, S. R., T. Will, and P. P. Marra. 2013. Estimates of bird collision mortality at wind facilities in the contiguous United States. Biological Conservation 168:201-209. http://dx.doi.org/10.1016/j.biocon.2013.10.007
Mabee, T. J., B. A. Cooper, J. H. Plissner, and D. P. Young. 2006. Nocturnal bird migration over an Appalachian ridge at a proposed wind power project. Wildlife Society Bulletin 34:682-690. http://dx.doi.org/10.2193/0091-7648(2006)34[682:NBMOAA]2.0.CO;2
Mabey, S. E. S. 2004. Migration ecology: issues of scale and behavior. Pages 66-68 in S. Savitt Schwartz, editor. Proceedings of the Wind Energy and Birds/Bats Workshop: Understanding and Resolving Bird and Bat Impacts. Washington, D.C. May 18-19. RESOLVE Inc., Washington, D.C., USA.
Marques, A. T., H. Batalha, S. Rodrigues, H. Costa, M. J. R. Pereira, C. Fonseca, M. Mascarenhas, and J. Bernardino. 2014. Understanding bird collisions at wind farms: an updated review on the causes and possible mitigation strategies. Biological Conservation 179:40-52. http://dx.doi.org/10.1016/j.biocon.2014.08.017
Otter, K. A., M. I. Hartley, M. V. d’Entremont, N. N. Johnston, D. Walsh, J. E. Bradley, M. Willie, and A. C. Pomeroy. 2014. Avian movement at the Dokie I Wind Energy Project 2008-2012: analysis of potential conflict between wind development and avian migration in the northern Rocky Mountains. Natural Resources and Environmental Studies Institute (NRESi) Technical Report #1. University of Northern British Columbia, Prince George, British Columbia, Canada. [online] URL: http://www.unbc.ca/nres-institute/technical-report-series
Pocewicz, A., W. A. Estes-Zumpf, M. D. Andersen, H. E. Copeland, D. A. Keinath, and H. R. Griscom. 2013. Modelling the distribution of migratory bird stopovers to inform landscape-scale siting of wind development. PLoS ONE 8(10):e75363. http://dx.doi.org/10.1371/journal.pone.0075363
Pomeroy, A. C., P. Thomas, C. Morcos, A. Labrosse, J. Preston, C. Bryden, M. V. d’Entremont, and K. A. Otter. 2007. Wildlife monitoring report 2007 - Dokie and Wartenbe Wind Energy Projects. Technical Data Report for EarthFirst Canada Inc. Jacques Whitford AXYS Ltd., Burnaby, British Columbia and the Center for Wind Energy and the Environment, University of Northern British Columbia, Prince George, British Columbia, Canada.
R Development Core Team. 2017. R A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org
Schmaljohann, H., F. Liechti, E. Bächler, T. Steuri, and B. Bruderer. 2008. Quantification of bird migration by radar - a detection probability problem. Ibis 150:342–355. http://dx.doi.org/10.1111/j.1474-919X.2007.00797.x
Schmaljohann, H., F. Liechti, and B. Bruderer. 2009. Trans-Sahara migrants select flight altitudes to minimize energy costs rather than water loss. Behavioral Ecology and Sociobiology 63:1609-1619. http://dx.doi.org/10.1007/s00265-009-0758-x
Stantec Consulting Ltd. 2012a. Dokie Wind Energy Project 2011 operational wildlife monitoring report. Stantec Consulting Ltd., Burnaby, British Columbia, Canada.
Stantec Consulting Ltd. 2012b. Dokie Wind Energy Project 2012 operational wildlife monitoring report. Stantec Consulting Ltd., Burnaby, British Columbia, Canada.
Stienen, E. W. M., W. Courtens, J. Everaert, and M. Van de Walle. 2008. Sex-biased mortality of Common Terns in wind farm collisions. Condor 110:154-157. http://dx.doi.org/10.1525/cond.2008.110.1.154
Taylor, P. D., J. M. Brzustowski, C. Matkovich, M. L. Peckford, and D. Wilson. 2010. radR: an open-source platform for acquiring and analysing data on biological targets observed by surveillance radar. BMC Ecology 10:22. http://dx.doi.org/10.1186/1472-6785-10-22
Walsh, D. A. 2012. A technique for avian nocturnal migration research: determining flight pattern of local nocturnal migrants around pre-operation wind turbines. Thesis. University of Northern British Columbia, Prince George, British Columbia, Canada.
Wickham, H. 2009. ggplot2: elegant graphics for data analysis. Springer, New York, New York, USA.
Wood, S. N. 2006. Generalized additive models: an introduction with R. Chapman and Hall/CRC, Boca Raton, Florida, USA.
Zimmerling, J. R., A. C. Pomeroy, M. V. d’Entremont, and C. M. Francis. 2013. Canadian estimate of bird mortality due to collisions and direct habitat loss associated with wind turbine development. Avian Conservation and Ecology 8(2):10. http://dx.doi.org/10.5751/ACE-00609-080210