Migration and Roosting Behavior of Northern Saw-whet Owls (Aegolius acadicus) During Fall Migration in Western Montana Open Access

Northern Saw-whet Owls (Aegolius acadicus) are the most-banded owl species in forested areas of North America, particularly during migration (Rasmussen et al. 2020, Nakash et al. 2023). Despite this status, we know relatively little about their movement paths and travel rates during fall migration. Their small size has limited the application of precise tracking methods used for larger raptors. Most movement information comes from recapture of banded owls at different banding stations within a migratory season. The timespan between band encounters has varied, seldom involving only one night, and the resulting data suggest that average nightly travel distances vary regionally: 13–56 km/night in mid-Atlantic states (Brinker et al. 1997), 7.7 km/night (range 1.0–44.8 km) in Wisconsin (Erdman and Brinker 1997), 28.8 km/night (±15.8 [SD] km/night) in south-central Indiana (Brittain et al. 2009), 37 km/night (± 33.0 km/night) in Alberta and Saskatchewan (Priestley et al. 2010), and 10.5 km/night in eastern North America (Beckett and Proudfoot 2011). Notably, Priestley et al. (2010) documented one owl traveling 106 km in a single night.

Although these studies provide important observations of owl movement potential and directionality, they cannot determine fine-scale routes, habitat use, or other behaviors related to migratory movement. For example, the straight-line distance between banding stations likely underrepresents the true distance an owl has traveled. Local characteristics such as topography, plant community, prey availability, roost-site potential, and anthropogenic features may influence these small-scale movements. Like many birds, migrating Northern Saw-whet Owls may also use stopover sites, a behavior generally inferred when a banded owl is captured more than once at the same place in a season. In addition, we can draw few conclusions about habitat use or preferences during migration from the conditions near banding stations, because most owl captures occur with the use of an audiolure that may attract owls from great distances.

Resident, migratory, and overwintering Northern Saw-whet Owls occur in late fall in the Bitterroot Valley of western Montana, USA, but relatively little is known about their migrations through the Intermountain West. This region lacks the network of long-term owl banding stations common elsewhere in North America. These owls migrate through the region in large numbers (Frye 2012, Nakash et al. 2023) and we suspected they traveled south through the Bitterroot Valley using the fairly continuous woody cover of the Bitterroot River floodplain. To augment knowledge of Northern Saw-whet Owl migration ecology in the Intermountain West, we used aerial and ground-based radiotelemetry to study owl movements during two fall migration periods in the Bitterroot Valley. We documented nightly travel rates, movement patterns, and roost sites of owls in the study area to investigate distances and directions traveled, stopover ecology and habitat use, and roost characteristics.

Our study took place in the Bitterroot Valley across approximately 105 km between Lolo and Conner, Montana, in an area encompassing 200 km². The Bitterroot Valley is bordered on the west by the Bitterroot Mountains and on the east by the Sapphire Mountains, and the Bitterroot River flows north through the valley bottom. The valley bottom included the Bitterroot River floodplain and adjacent low-elevation areas, extending 1.5–11 km east and west, at elevations of 1000–1200 masl. The valley bottom abruptly transitioned westward into the Bitterroot Mountains (typical elevations approximately 1200–2400 masl, with peaks exceeding 3000 masl). Foothill topography east of the valley was more complex and transitioned more gradually up into the Sapphire Mountains. Peak elevations approach 2700 masl, but most of the Sapphire Mountains featured lower and more rolling topography than the Bitterroot Mountains. Plant communities and land use varied within each topographical area, including floodplain forest, agricultural, and residential areas within the valley bottom; native grasslands, shrublands, draws, agricultural, and residential areas within the foothills; and mixed conifer forest mostly administered by the US Forest Service but interspersed with some private residential areas within the mountains.

We conducted a pilot season in fall 2013 to test glue-on transmitter attachment, transmitter performance, and air-based telemetry methods to derive the methods outlined below. Capture work took place from 23 September to 7 October in 2014 and from 14 September to 6 October in 2015. Tracking extended for a week after the last owl capture in both years. No appreciable snow accumulated within the study area during these periods.

We captured and released owls at one valley-bottom location at the northern end of the study area (46°41′N, 114°00′W). This location was within a 9-km² expanse of floodplain forest featuring primarily ponderosa pines (Pinus ponderosa) that straddled both sides of the Bitterroot River. Foothill topography constrained the width of the valley bottom at this location to <4.5 km wide, creating a bottleneck effect relative to the width of the valley farther south.

We used mist nets set between low woody vegetation and audiolures to capture owls. Our effort varied based on capture rate and typically lasted 2–4 hr per evening. We opened nets just after dark and checked them every 15–30 min depending on capture rates and weather conditions, while continuously broadcasting the owl “advertising call” (Rasmussen et al. 2020). Upon release, owls had a minimum of 6 hr of darkness to continue migratory movements, if they chose to do so. We limited nightly captures to a maximum of 10 owls to facilitate subsequent tracking efforts using radio telemetry.

We banded each owl with a standard US Geological Survey aluminum leg band, aged it based on Pyle (2008), and sexed it based on Brinker et al. (1997). We extracted DNA from feather samples and assigned sex using the CHD gene (Kahn et al. 1998) for 17 owls classified as unknown sex based on morphometrics. We assessed body condition based on muscle-mass scoring around the keel (Scott 2020).

We attached temporary VHF radio transmitters (Advanced Telemetry Systems model A2420; 1.3 g; 3-sec pulse rate) to each owl by gluing a small piece of medical gauze to back feathers between the scapulae. We then glued transmitters to this gauze. We trimmed antennas to tail length. We held owls for 5 min after transmitter application to observe behavior and ensure transmitters and antennas were sitting centered on the owl’s back. All auxiliary materials attached to the owl weighed less than 3% of body mass. We released all owls on site within 30 min of capture and prior to midnight.

We performed daily searches for transmitter signals from the air using a Cessna 182 Skylane airplane mounted with antennas and a scanning receiver. We attempted to fly set transects that gave us coverage of the entire valley. We tested transmitter signal strength, behavior, and detectability from the plane in various topographies prior to setting our route. In general, we set transects 8 km apart as airplane antennas detected transmitters at a distance of 3–6.5 km depending on topography. Flight route and scanning tactics varied by location in the valley, number of owls to track, previous days’ owl locations, and weather conditions that limited flight capabilities. Our goal was to survey the entire study area with close to equal effort.

We deviated from transects to locate owls when we detected a signal. To determine location, we isolated the frequency and flew circular patterns to identify the likely roost location, usually a clump of bushes or a tree. The pilot and the technician operating the receiver both helped determine signal strength and directionality. We mapped all documented owl locations onto georeferenced aerial imagery using an iPad, and then removed that owl’s frequency from the channel scan and resumed searching. We also deviated from transects when we approached the location of an owl’s signal mapped on a previous day. If we did not detect a signal close to a previous day’s location, we circled out to a 16-km radius. These circles often took us out of the set transect area, east into the foothills and peaks of the Sapphire Mountains. We could not make similar forays west into the Bitterroot Mountains, as the steeper, higher topography and deep canyons made flying unsafe. We also did not scan for owls more than 2.4 km north of their previous location to prioritize documenting southward migratory movements. To help ensure we did not miss a signal, we searched similar circles around last known locations for up to 3 d and included undetected owls in our channel scan for up to a week after transmitter deployment. We spaced captures and releases to minimize concurrent scanning for more than five owls in any part of the valley, but sometimes had more than 10 owls in a scan at the beginning of flights near the release location.

When possible, we attempted ground-based confirmation of owl locations. This allowed us to determine the accuracy and precision of air-based detections, collect roost information, and determine if a transmitter was no longer attached to an owl. The technician in the plane communicated with a technician on the ground to track owls within hours of detection from the air. Ground-based actions were biased toward locations close to the release site, owing to our ability to access both the private and adjacent public land in this area, the logistical ease of tracking multiple owls in one general area, and the early notification of owl locations in this area during the tracking flight.

We used a handheld telemetry receiver and 3-element Yagi antenna (Receiver R410, Advanced Telemetry Systems, Isanti, MN, USA) to track owls on the ground and spent up to 45 min attempting to visually confirm an owl’s location. In many cases, we could not accomplish this objective and classified location precision as: general area, stand of vegetation, tree/shrub, owl, or discarded transmitter.

To calculate nightly distances traveled, we used ArcGIS (Esri, Redlands, CA, USA) to measure travel distances between sequential locations. For owl locations mapped by air and ground, we used the more precise ground location to calculate distances. Tests during our pilot season and comparisons between mapped air and ground locations showed that locations mapped from the air seldom differed by more than 0.4 km from the actual ground location. To be conservative, all reported distances for this study should be considered accurate to ±0.8 km. We acknowledge that the precision of similar distance estimates may vary in topography and vegetation different from that on the relatively flat but forested Bitterroot River floodplain. Further, the derived estimates reflect minimum distances traveled per night, because the owls may not have traveled in straight lines between locations.

Given the assessed accuracy of mapped locations, we did not classify nightly distance traveled as a “movement” if the calculated distance between mapped points was 0.4 km or less. We categorized directionality of both individual movements and entire flight paths using the following broad categories: N, NE, E, SE, S, SW, W, NW.

We classified owls as migrant, resident/overwintering, or unknown status based on their movement patterns. We defined owls as migrants if, over the course of all tracked movements, they moved >2.4 km in a non-northerly direction away from the capture location. We based this threshold on the geography of the study area, as the banding site was located near the northern end of a riparian forest patch that extended 2.4 km to the south. Movements beyond this distance would require owls to move through a nonforested landscape. We included owls in this category for which we never detected a signal, assuming they had moved out of the study area in the hours after release. We defined birds as resident/overwintering if they only made local movements of <2.4 km upon release, movements of >2.4 km but in a northerly direction, or a combination of the two prior to either the end of tracking or the retrieval of a shed transmitter. We defined owls as unknown status if they shed their transmitter within 2.4 km of, or moved north from, the capture location.

We defined stopover events as when a migrant owl stayed within 2.4 km of a previous location for more than 1 d and then moved at least 2.4 km farther, or we subsequently failed to detect it altogether, suggesting movement out of the study area. Because we may have captured some owls in the middle of a stopover and some owls may have removed their transmitter during a stopover, we classified stopover events as follows: unknown beginning/unknown end; unknown beginning/known end; known beginning/unknown end; known beginning/known end. If an owl removed its transmitter within hours after capture or immediately moved out of the study area upon release, we classified their stopover status as unknown. We did not count the night of capture in the calculation of stopover length, given the possibility an owl was migrating at the time of capture. We used Fisher’s Exact Test to determine if stopover behavior was related to either year or body condition.

We used broad categories to classify the plant communities around roost locations as follows: floodplain forest (floodplain forest on valley bottom [ca. 1000 masl elevation] with continuous woody vegetation); local floodplain forest (floodplain forest on valley bottom with continuous woody vegetation within 5 km of owl capture and release location); floodplain intermittent forest (floodplain forest on valley bottom with intermittent or discontinuous woody vegetation and often interspersed residential development); low-elevation forest (continuous mixed-conifer forest at low elevations [1000–1200 masl] outside of the 100-yr floodplain of the Bitterroot River); riparian (deciduous trees and shrubs along small, low-elevation streams running primarily east-west); draw (shrubby draws at mid elevations [1200–1500 masl] in foothill topography and oriented primarily east-west); mid-elevation forest (continuous mixed-conifer forest at mid elevations within foothill topography); mid-elevation intermittent forest (discontinuous mixed-conifer forest or woodlands at mid elevations within foothill topography); and high-elevation forest (continuous mixed-conifer forest at elevations >1500 masl). We could assign broad categories for most locations, even when we could not precisely verify an owl’s location on the ground. We did not classify a used plant community when the margin of error for a location straddled two or more of the defined communities.

We documented finer-scale roost-site characteristics in the Bitterroot River floodplain around the capture and release location, where access and logistic feasibility were greatest. We assessed roost characteristics at multiple scales depending on how accurately we could pinpoint an owl’s location. For all detections, we noted dominant overstory plant cover on the floodplain, which included ponderosa pine, black cottonwood (Populus balsamifera), quaking aspen (P. tremuloides), mixed deciduous shrub, or mixed forest (i.e., no predominant tree species). We noted the substrate species if we tracked an owl to a stand of uniform species or an individual tree/shrub. For locations where we tracked an owl to an individual tree or shrub, we also recorded substrate height, the diameter at breast height (dbh) of tree roosts, and the presence/absence of pellets or whitewash. For locations where we tracked an owl to a tree and did not flush it, we described owl conspicuousness as: not seen, well hidden, partially obscured, or easily visible. For locations where we could see the owl, we calculated the perch height by taking two angle readings (to the owl and the base of the tree) using a clinometer and two distance readings (horizontal distance to the tree and line-of-sight distance to the owl) using a rangefinder. These measurements allowed us to calculate the vertical segment above and below eye level using trigonometric relationships, and the sum of these segments provided the perch height. We used binoculars to search for whitewash and did not search for pellets if the roost was low and we could have disturbed an owl. We noted the presence of other owls when relevant and, for specific roost sites used by multiple owls, we recorded roost characteristics only once.

We captured and tracked 89 owls: 36 in 2014 and 53 in 2015. Based on movement behavior, we classified 65 owls as migrant, 18 as resident/overwintering, and six as unknown status. Most migrants were hatch-year females in poor body condition with low levels of muscle around their keel at the time of capture (Table 1).

We documented 68 movements of 55 migrant owls on sequential nights. These movements ranged from 2.6–65.7 km per night, with most movements falling between 2.6–8.0 km per night (Fig. 1). Average nightly distance traveled was 13.2 ± 13.0 (SD) km. We failed to detect 10 transmittered owls within our study area, indicating either they quickly moved more than 10 km in the hours after release or their transmitters failed prematurely.

Although many owls moved out of the valley in the first few nights after release, our method allowed us to track some individuals a relatively long distance from the banding station. We tracked one owl approximately 96 km over 10 nights and eight owls >32 km from the release site. These owls all showed variability in nightly distance traveled (Fig. 2).

Owls tracked multiple nights traveled due south less often than southeast and southwest combined (Fig. 3) and used foothill topography on both the east and west sides of the valley (Fig. 2). Of the six owls that had an overall northerly movement path, all had multiple days of detection in the vicinity of the banding station prior to leaving the area entirely, thus limiting our ability to observe a broader-scale flight direction. If owls made greater northerly movements, we likely would have missed them owing to logistical constraints (e.g., study area boundary and channel scan protocol). The 10 owls that left the study area the night of release could have traveled in any direction but due south, as movements in that direction would have been detected via airplane telemetry.

Of the 55 migrant owls tracked for multiple days, 35 (64%) showed stopover behavior involving 40 individual stopover events, and five owls each made two stopovers. Stopover events occurred in all phases of the lunar cycle. The proportion of migrants exhibiting stopover behavior was 65% in 2014 and 46% in 2015 and we observed no difference between years (P = 0.15). Stopover events with a known beginning and end averaged 4.6 ± 3.2 (SD) nights and spanned 2–13 nights (Fig. 4). However, 22 stopover events had an unknown beginning, end, or both, making it difficult to determine true length. We could not determine if owls were on a stopover when captured, so stopovers were likely more frequent and longer than calculated.

Of the 202 migrant-owl day locations with a classified plant community, 33% were in local floodplain forest, 20% were in mid-elevation forest, 18% were in floodplain forest, and 29% were scattered among other communities (Table 2). Once owls moved from the vicinity of the banding station, we most often detected them in mid-elevation forest and floodplain forest.

We assessed 60 roosts used by 32 different owls in the local floodplain forest and found most (67%) in floodplain areas where ponderosa pine predominated. Of 58 roosts we could pinpoint to a substrate type, most (72%) occurred in trees and live shrubs (21%). Of 56 roosts we could identify to substrate, 73% were ponderosa pines. No owls roosted in black cottonwood and only two roosted in quaking aspen. Willows (Salix spp.) were the most common deciduous shrub used (Table 3).

Although we often pinpointed an owl to a roost tree or shrub, our ability to assess roost height was confounded by the unexpected heights owls roosted in trees. We could only see the owl and thus calculate its perch height in 20 instances. A majority (55%) of these roosts were in shrubs close to the ground, but perch heights varied from as low as 0.08 m in a clematis vine (Clematis ligusticifolia) to as high as 26 m in a ponderosa pine (Table 4). We most often located a roost tree, searched for the owl for 15–45 min both visually and with handheld telemetry equipment, and in the end had to classify an unseen owl’s perch height as somewhere “high in tree” (n = 31). Although ponderosa pines were used as roost substrates 41 times, we could identify the roost location in only six cases owing to the perch heights used. In all six cases, we observed owls nestled within needles and cones on lateral branches away from the trunk. The range of heights at which we found an owl in ponderosa pines was 3–26 m.

Owls concealed themselves well, even when not high in a tree. In the 52 instances where we recorded owl conspicuousness and did not accidentally flush it (occurred twice), we classified 63% of the owls as unseen, 29% as partially obscured, 6% as well hidden, and one owl (2%) as easily visible. We found no pellets at 39 roosts where we actively searched. We found no whitewash at 32 (84%) searched roosts and small amounts at six (16%) roosts.

We had one instance in which three owls roosted in the same ponderosa pine on the same day, although not on the same branch or at the same height. Based on overall movement behavior, we classified one of these owls as resident/overwintering and the other two as migrants. We did not observe any individual owls using the same roost for multiple days; even during stopovers, owls changed roost locations each night (Fig. 5).

Despite decades of capturing and banding Northern Saw-whet Owls during migration, our understanding of owl behavior at this time of year is limited. The nightly distances that we documented with telemetry fell within the range documented in previous studies based on band re-encounters. However, our tracking of owls on a nightly basis, rather than averaging movements across nights, provided a clearer picture of the variability in nightly travel distances, the potential for longer-distance nightly movements than revealed by averages, and the prevalence of stopovers during migration. Logistical constraints and the temporary nature of transmitter attachments prevented us from documenting longer travel distances comparable to those resulting from recapture events (e.g., 106 km; Priestley et al. 2010). Nonetheless, we tracked eight owls more than 48 km from the release site, surpassing the results of previous tracking efforts. These multi-day tracks also emphasized variability in nightly travel distances that can be obscured by averaging the distance between band re-encounters.

The only other study that specifically investigated the stopover behavior of Northern Saw-whet Owls used recapture data as an indirect indication of stopover rates on the Delmarva Peninsula in the eastern USA, and yielded a recapture rate of 31–40% in irruptive years and <13% in non-irruptive years (Whalen and Watts 2002). Our estimated stopover rate of >60% exceeds these values, and owls in our study may have stopped over more than our data revealed. Given that we documented one stopover of almost 2 wk, perhaps some owls classified as overwintering/residents moved on after our study ended or after their tags fell off. Use of longer-lasting tags and automated tracking systems, such as the Motus network (Taylor et al. 2017), may help to improve classifications and understanding of the comparative behaviors of migrants, overwintering owls, and local residents.

The sample size and logistical constraints of our study limit the potential for identification of a mechanism for high stopover rates. Whalen and Watts (2002) suggested that stopover rates among eastern owls were high when owl density was low (i.e., in nonirruptive years). We did not detect a significant difference in the proportion of migrant owls stopping over between years, even though 2015 was considered irruptive at the closest banding station in Lucky Peak, Idaho (G. Kaltenecker, Intermountain Bird Observatory, pers. comm.). Whalen and Watts (2002) also hypothesized that low prey abundance or available roosts prompts owls to move on versus stay in one area. Perching opportunities were not limited in our system, neither close to the release site where we observed the most owls nor generally throughout our study area. We did not study prey availability, but areas around the banding station and much of the rest of the study area below the highest elevations lacked snow cover that might otherwise have inhibited owls from foraging for small mammals such as voles (Microtus spp.) and deer mice (Peromyscus maniculatus). Stopovers occurred in all phases of the lunar cycle, suggesting these events were not related to moon illumination, although we acknowledge that moon phase often amounts to only a coarse representation of lunar illumination (Śmielak 2023).

Our data did not support our hypothesis that owls would use the continuous cover of the Bitterroot River floodplain to travel in a southerly direction. In general, owls did not move directly south following the river corridor and instead traveled along the adjacent foothills or left the valley entirely without detection. Although woody cover exists continuously along the Bitterroot River, the density varies. Deciduous vegetation dominates some areas and deciduous species are either actively shedding their leaves or already lack foliage during peak owl migration, minimizing their potential concealment and thermal benefits. Indeed, all but one roosting event in a deciduous substrate occurred on or before 1 October, when some deciduous trees and shrubs still had leaves. The lone exception was a roost in a clematis vine on 6 October 2014, where the owl roosted deep within the cover of multiple layers of the woody vine itself.

In addition to a potential lack of continuous conifer cover, most floodplain forest in the study area occurred on private land, with consequent human alteration, residential and commercial development, various levels of ambient light at night, and other land-use practices that might alter habitat quality or attractiveness to a migrating owl. That said, we tracked multiple owls into residential areas and some owls roosted near homes or other human structures.

In contrast, the floodplain close to the banding station offered approximately 200 km² of nearly continuous forest with a high ponderosa pine component and mixed land ownership of protected private land and state-managed forest. We suspect the topography at the northern end of the valley funneled migratory owls into this area and, once there, the extensive coniferous forest cover provided good roosting and foraging opportunities for extended periods.

The relatively small temporal and geographic scale of our study compared to the distances Northern Saw-whet Owls may travel to reach overwintering grounds limits potential inferences regarding overall migratory paths, destinations, and potential nomadic behavior. The owls we tracked moved in all southerly directions from the release location. Once out of the study area, they might have continued to show non-uniform migration directions similar to those documented in Alberta and Saskatchewan (Priestley et al. 2010) and in the Appalachian region (Beckett and Proudfoot 2011). Current tracking efforts in our region and elsewhere using the Motus network (e.g., Holroyd 2024, Stone 2024) should eventually provide more insight about the movement ecology of Northern Saw-whet Owls, including the nonsoutherly movements documented from band re-encounters that have prompted discussions of nomadism (Marks and Doremus 2000, Beckett and Proudfoot 2011).

Evidence of Northern Saw-whet Owl reliance on forested habitat is prevalent in the literature. Past studies have stressed the importance of coniferous forest as cover for migrating owls (Brinker et al. 1997, Priestley et al. 2010), that migration timing may correlate with the seasonal loss of deciduous cover (Brinker et al. 1997), and that the presence of forest cover likely dictates the predominant direction of migratory pathways (Priestley et al. 2010). The owls in our study had available forest cover in all directions, which might explain their variable travel directions across the range from southeast to southwest. However, unless they took extremely tortuous routes to follow isolated stringers of woody cover, many owls would still have had to cross expanses of treeless habitat (e.g., grasslands, shrublands) to get to foothill and upland coniferous forests, or travel generally southward.

Our study is the first to provide detailed documentation of roosts used by Northern Saw-whet Owls during migration. Our findings complete documentation across the full annual cycle of owls using conifers for roosting, with prior studies having demonstrated their importance during breeding (Boula 1982), winter (Mumford and Zusi 1958, Grove 1985, Swengel and Swengel 1992, Churchill et al. 2000), and spring (Hayward and Garton 1984). Similar to other studies (Hayward and Garton 1984, Grove 1985, Swengel and Swengel 1992, Churchill et al. 2000), we also found owls perched away from the tree trunk in the area of densest cover and highest concealment in a ponderosa pine. Owls blended in well and looked superficially similar to pinecones in the instances we could spot them. Our telemetry work allowed us to locate owls roosting high in trees, often so high we could not determine their precise location. We suspect roost sites high in ponderosa pines offer a thermal advantage. The topography east of the Bitterroot River floodplain blocked the morning sun and consequently the top third of ponderosa pines absorbed solar radiation for several hours before the bottom portions.

Many studies have used passive searching for roosts using indicators such as pellets or whitewash (Mumford and Zusi 1958, Boula 1982, Grove 1985, Swengel and Swengel 1992), but we rarely detected such clues. These observations suggest that relying on passive methods alone can bias findings toward low roost sites and potentially miss roosts high in trees. In a majority of cases, we could not see owls we knew were present, and we considered our highest detection at 26 m a field triumph.

In conclusion, our fine-scale mapping of Northern Saw-whet Owl movements and roost sites during migration in Montana advances understanding of the movement ecology of a commonly banded species. This fine-scale information will help interpret the more generic locations from multiyear nanotags used with the communal Motus receiving network (Stone 2024). Motus-related research should further improve understanding of longer-distance travel paths and rates, behaviors like partial migration and nomadism, accurate stopover durations, activity patterns at fine temporal scales, and possibly relationships between nightly activity periods and factors such as moonlight and wind direction. The two methods together will greatly clarify aspects of Northern Saw-whet Owl migration that currently are not well understood.

This work was funded by MPG Ranch and was conducted with all required state and federal permits (Federal Bird Banding Permit #23878 and Montana Scientific Collector’s Permits 2014-117 and 2015-107). We thank the following technicians for their help with data collection: W. Blake, I. Perks, E. Rasmussen, and M. Seidensticker. We thank D. Schwaderer and his team of pilots at Western Montana Aviation for their superb aviation skills. M. and J. Larson, formerly of the Owl Research Institute, contributed to field work and method development during the 2013 pilot season. M. Scofield contributed to data analysis and W. Nielson made maps for this manuscript. Independent reviews by J. Marks and L. T. Priestely also helped improve an earlier version of this report. The authors declare no conflicts of interest.