Data Collection and Filtering. PTTs were programmed to transmit for 10 hr and turn off for a 40-hr recharging period. Standard and auxiliary location classes (LC) were derived by the Argos System (www.argos-system.org) and disseminated by CLS America (www.clsamerica.com). We used a systematic hybrid filtering algorithm to remove implausible Argos locations (Douglas et al. 2012). Filtering criteria considered location quality class, distance moved, movement rate, and turning angle. Locations were retained if their LC was 1, 2, or 3. Auxiliary locations (LCs 0, A, B, and Z) within 5 km of a preceding or subsequent location were retained by virtue of spatial redundancy, and remaining auxiliary locations were included only if resultant movement rates were <80 km/hr and the internal angles (α, in degrees) formed by preceding and subsequent vectors (of lengths d1 and d2 km) were not suspiciously acute (α > −25 + β × ln[minimum (d1,d2)], where β = 15). We assigned β = 15 because it performed well for our specific tracking data during both stationary and migratory periods. PTTs reported a total of 19,709 standard and auxiliary locations; 3757 (19%) of the auxiliary class locations were removed by filtering. We further filtered locations by retaining only the best quality location (highest LC) per duty cycle. This resulted in 1503 total locations (92% standard quality) and 8–191 locations per owl (mean = 57.8 ± 44.0 [SD] locations). Root-mean-square location errors, as estimated by Douglas et al. (2012), were 1.0, 2.5, and 4.3 km for unfiltered LCs 3, 2, and 1, respectively, and 6.3, 4.2, 6.2, and 5.2 km for filtered LCs 0, A, B, and Z, respectively.

PTTs were equipped with sensors that monitored motion, temperature, and battery voltage. We estimated causes of tracking termination following Klaassen et al. (2014). We considered PTT failure likely when the loss of contact was preceded by an extended period of poor transmission performance or low battery voltage. Probable mortality occurred when the PTT transmitted continuously from the same location, the motion sensor stopped incrementing, and the temperature sensor emulated ambient conditions. In any of these circumstances, we retained an owl's last active location. We attempted to recover motionless transmitters to assess an owl's fate and potential cause of mortality as soon as was logistically feasible.

Movement Patterns. We defined four seasonal periods to describe Short-eared Owl movements. Autumn and spring movement periods began when new locations showed that owls were consistently moving away from summer and winter areas, and ended when the owls started making only localized multidirectional movements. Owls tended to be relatively sedentary during summer and winter in between the spring and autumn movement periods. We classified migratory stopovers as periods during spring or autumn when owls temporarily occupied a localized area, which we defined as moving <35 km within 48 hr. Because transmitters were not active every day, we could not always determine the exact dates of departure and arrival. We estimated departure and arrival dates as the midpoint of transmission gaps ≤7 d, but did not estimate departure and arrival dates based on gaps >7 d.

We investigated ground speed (km/hr) by constructing all pairwise vectors with LC 3, 2, or 1 endpoints that spanned 2–10 hr. We examined diurnal movement patterns by assigning the more probable movement vectors (those >5 km/hr) to daytime (0600–1800 H) or nighttime (1800–0600 H) based on the vector's midpoint local time. We report cumulative distance moved based on the subset of filtered locations with the highest quality location for each transmitter duty cycle. We calculated all distances as great-circle-routes (i.e., orthodromes). We calculated local time for a given location by offsetting its Greenwich Mean Time (GMT) by −1 hr for each 15° of west longitude.

We evaluated wind conditions when owls undertook substantive movements and during stopovers. We extracted wind speed and direction at 10 m aboveground from the National Centers for Environmental Prediction (NCEP) North American Regional Reanalysis (NARR) daily datasets (Mesinger et al. 2006) with 3-hr periodicity as described in Rutledge et al. (2006). For substantive owl movements, we considered tracking vectors that exceeded 500 km during 50–80 hr (movement between two sequential transmitter duty cycles). We averaged the NARR wind data over the duration of the movement vector, within an area along the flight path buffered laterally to each side by one-quarter of the flight-path length. At stopover sites, we averaged the NARR wind data for the duration of the stopover. Because our diurnal analysis indicated that owls moved primarily at night, we restricted the wind data to the local nocturnal period (i.e., 1800–0600 H). Following Kemp et al. (2012), we calculated a tailwind index (TWI) using the averaged u (east/west) and v (north/south) wind-vector components (m/s), and the compass direction of migration (ω) as indicated by the tracking data:

.

The tailwind index expresses the amount of the wind vector that is parallel to the direction of the birds' movement, where positive values represent a favorable tailwind and negative values a headwind.

We evaluated general patterns of habitat use during migratory stopover and seasonal stationary periods by first creating a buffer around each location, with radius equal to the root- mean-square error of the estimated location accuracy of the relevant LC. We extracted general habitat characteristics by overlaying each buffer on the North American Land Change Monitoring System (NALCMS. 2010). We standardized the extracted habitat information using an inverse-area scaling, in which smaller buffers with higher location accuracy were given a higher proportional weight than larger buffers with lower location accuracy. We performed all spatial mapping and analyses in ArcMap 10 (ESRI 2011). Unless otherwise noted, descriptions of movements are based on all birds combined across years and capture locations. Summary statistics are presented as the range followed parenthetically by the mean ± SD.

Autumn Migration. Fourteen owls departed summer ranges on Seward Peninsula during a 47-d period from 8 August to 25 September (mean departure date 3 September ± 12.7 d). Seven interior owls began autumn migration during a 15-d period from 31 July to 15 August (3 August ± 5.9 d). Owls that migrated from interior sites departed an average of 30 d earlier than owls that migrated from Seward Peninsula.

Seven owls dispersed from Seward Peninsula in a predominantly southeasterly direction, and their movements converged within approximately an 85-km-wide zone in interior Alaska that encompassed the Tanana River valley and the primary 2010 interior capture location (Fig. 1A). Four other owls deviated from this pattern (Fig. 1A). One owl flew northeast from Seward Peninsula across the northern Yukon Territory before orienting to the southeast. Three owls veered to the south and crossed the Alaska Range into coastal southcentral Alaska.

Figure 1. 

(A) Autumn movements of 14 Short-eared Owls captured on Seward Peninsula (orange lines) and 12 owls captured at two sites in interior (black lines) Alaska. Yellow dots denote location and duration of stopovers. Triangles denote last known location of partial migratory tracks, and dots denote locations where owls completed migration. (B) Seasonal movements of four Short-eared Owls following first documented winter period. Black and orange lines delineate spring and autumn migration routes, respectively. Other symbols are as for (A). (C) Movements of one owl that remained in southern Idaho for a full year after migrating south from Alaska. For all maps, orthodromes are depicted as straight lines in this gnomonic map projection.

Movements of most Seward Peninsula and interior owls (n = 14) converged in a 200–400-km-wide zone east of the Rocky Mountains, extending through the Yukon Territory, northeastern British Columbia, Alberta, and southwestern Saskatchewan (Fig. 1A). One interior owl deviated from this corridor and migrated through the Northwest Territories and southern Manitoba. Four Seward Peninsula owls flew mostly west of the Rocky Mountains, including one owl that crossed 850 km of open water over the Gulf of Alaska. An interior owl was relocated in southern British Columbia after a 48-d data gap, which also suggested a southward trajectory west of the Rocky Mountains.

Among 11 owls that crossed the Canadian border into the conterminous United States, nine migrated through Montana and western North Dakota and two migrated through central Washington (Fig. 1A). Seven owls completed their autumn migrations in 61–147 d (85.2 ± 31.8 d) and settled on winter ranges between 22 October and 27 December. The duration of autumn migration averaged more than 1.5 times longer for three interior owls (104 ± 36.9 d) than it did for four Seward Peninsula owls (63 ± 15.1 d). Winter destinations for these seven owls spanned 21° of latitude from central Montana to southern Texas, and 24° of longitude from central California to western Kansas (Fig. 1A).

Completed autumn migrations spanned 3205–6886 km (4722 ± 1156 km) and were on average longer for owls originating on Seward Peninsula (5280 ± 1185.8 km) than for owls originating at interior sites (3979 ± 672.2 km). Other migration tracks terminated prematurely because of owl mortality or transmitter failure. Notably, the transmitters of one owl captured on Seward Peninsula and one owl captured at an interior location both failed when the owls were within 35 km of each other in Zacatecas, Mexico, after the owls had travelled 6008 and 7057 km, respectively (Fig. 1A). A third owl traveled 6553 km from Seward Peninsula to southern Texas before its transmitter failed (Fig. 1A).

Peak movements occurred almost entirely at night between 1800 and 0600 H local time (Fig. 2). Owls migrated at rates of 44.5–340.7 km/d (146.5 ± 101.2 km/d). Estimates of in-flight ground speed reached as high as 40 km/hr (Fig. 2). Substantial movements (≥500 km in <3 d) during autumn migration occurred when the wind speed and direction along the flight corridor were slightly favorable (i.e., a light tailwind) or neutral (Fig. 3), but long-distance movements did not necessarily appear dependent on tailwinds. Notably, however, the owl that crossed the Gulf of Alaska followed a trough of low pressure that generally provided favorable tailwinds throughout most of the >800-km overwater flight (Fig. 4).

Figure 2. 

Tracking speeds of Short-eared Owls during local daytime and nighttime hours. More and faster movements occurred at night. To represent movements of flying owls, samples (in parentheses) were limited to vectors of 2–10 hr duration that were bounded by standard quality Argos locations (LC 3, 2, or 1) and represented ground speeds >5 km/hr. The upper range of tracking speeds (35–45 km/hr) represents conservative (straight line) maximum estimates of sustained (2–10 hr) movement rates over the landscape. Boxes span the interquartile range (IQR), whiskers extend to 1.5 times the IQR, means are denoted with diamonds, and medians with horizontal lines.

Figure 3. 

Long-distance movements (50–80 hr and >500 km) of Short-eared Owls in relation to wind conditions. Positive values represent tailwinds and negative values represent headwinds.

Figure 4. 

Near-surface (1000 millibar) wind conditions (National Oceanographic and Atmospheric Administration 2016) during the flight of a Short-eared Owl over the Gulf of Alaska. Black dots are Argos satellite tracking locations and white dots are estimated locations for the beginning and end of the open-water crossing. (A) Beginning wind conditions when the owl encountered a low pressure system in the northern Gulf of Alaska and appeared to follow tailwinds around the southern side of the low. (B) Ending wind conditions showing a trough of easterly tailwinds that prevailed as the owl completed the second half of the crossing.

All seven owls that completed autumn migration made one or more stops along their southbound routes (4.1 ± 1.3 stops per owl; Fig. 1A). Length of stay ranged from 2–48 d (11.7 ± 14.8 d) and the combined duration of autumn stopovers ranged from 18–75 d (47.3 ± 19.1 d). Stopovers accounted for 29–58% (46.9 ± 10.9%) of the total migration periods for individual owls. The total duration of stopovers was >1.5 times longer for interior owls (61 ± 17.8 d, n = 3) than for Seward Peninsula owls (37.0 ± 13.9 d, n = 4). Excluding stopovers, interior owls took slightly longer to complete their autumn migrations (43.3 d ± 28 d) than Seward Peninsula owls (34.2 ± 14.3 d). Shorter stopovers often coincided with the occurrence of headwinds, whereas the average tailwind index was neutral during long stopovers (>14 d; Fig. 5).

Figure 5. 

Migratory stopover duration of Short-eared Owls in relation to wind conditions. Positive values represent tailwinds and negative values represent headwinds.

Owls occupied primarily open habitats during stopovers, including grasslands and shrublands (64 ± 19.8% of the area occupied by individual owls), and croplands (21 ± 21.3%), but they also occupied stopover areas that contained forest habitat (15 ± 11.5%).

Winter. We tracked four owls during a complete winter period, ranging in length from 92–162 d (134 ± 31.3 d). Three owls that originated on Seward Peninsula wintered in central California, southern Idaho, and northern Colorado, and one interior owl wintered in western Utah (Fig. 1B). We tracked a fifth owl, which travelled from interior Alaska to central Montana, for only a portion of the winter season. Four owls were relatively sedentary during winter, remaining within a <20-km radius, whereas one owl moved among several sites within a 100-km radius.

Owls wintered in a variety of open habitats. Three owls in southern Idaho, central Montana, and northern Colorado occupied a mixture of natural and introduced grasslands (52 ± 16.3% of occupied area), croplands (43 ± 17.7%), and natural shrublands (5 ± 1.3%). Two other owls occupied almost exclusively croplands in the Central Valley of California (99% of occupied area) and shrublands in southern Utah (98% of occupied area).

Spring Migration. Three owls initiated spring migration between 30 March and 30 April (12 April ± 15.9 d), traveled 115–1375 km (743 ± 630.1 km) from their wintering areas, and settled on summer ranges in southern Idaho, eastern Montana, and southern Alberta (Fig. 1B). The fourth owl departed on 20 March and traveled 2461 km from central California to northern British Columbia paralleling its southward migration route, at which point its transmitter failed (Fig. 1B).

The owl that traveled between western Utah and southern Alberta made one 15-d stop in southern Idaho, during which time it occupied croplands (54% of occupied area), natural shrublands (23%), and natural grasslands (23%).

Summer. We tracked three owls throughout a second summer. A female that wintered in western Utah arrived in eastern Montana on 10 April and remained there for 121 d. A female that wintered in southern Idaho moved to a nearby summering area in southeastern Idaho on 1 May and remained there for 161 d (Fig. 1C). The timing, length of stay, and movement patterns of these two females suggested that they may have nested. In contrast, a male that wintered in western Utah arrived in southeastern Alberta on 30 April, and spent the summer moving among several locations in southeastern Alberta and southwestern Saskatchewan (Fig. 1B).

The two females occupied predominantly natural habitats in summer: natural grasslands (83%) and croplands (17%) in eastern Montana, and a mosaic of natural shrublands (86%), natural grasslands (10%), and cultivated croplands (4%) in southern Idaho. In contrast, the male in Alberta occupied primarily croplands (86%) and less often natural grasslands (14%).

Second Autumn Migration. On 9 October, the female in southern Idaho began a series of short movements in the vicinity of her summer range, and she was last located on 14 December in her previous wintering area (Fig. 1C). The female in eastern Montana departed her summer range on 9 August, spent 83 d migrating south to her previous winter location in northern Colorado, and then moved 175 km north into southeastern Wyoming, where she remained until her transmitter failed on 15 February. She made one stop for 68 d in grassland (92% of occupied area) and shrubland (8%) habitats in northeastern Wyoming. The southern Alberta/Saskatchewan male was still on his summer range when transmitter failure occurred on 12 October (Fig. 1B).

Fate of PTTs and Owls. Sensor data and PTT recoveries indicated that 20 of 26 owls probably (13 owls) or definitely (seven owls) died during the study. Two owls died near capture sites before migrating south, 13 died during the first autumn migration, three during winter, one in spring, and one during the second autumn migration. The total travel distance and duration from deployment locations to the locations of probable or confirmed mortalities averaged 2591 ± 1729 km and 125 ± 50 d.

We found owl remains at six of seven sites where we recovered PTTs. At the seventh site, we found only a chewed harness. Evidence at recovery locations, such as plucked feathers, bones with attached tendons stripped of meat, and proximity of perches, suggested that raptor predation was the likely cause of death for four owls; however, we could not rule out post-mortality scavenging. Sensor data suggested that six PTTs were on live owls when the batteries failed, four during the first autumn migration period and two during the subsequent winter.

Potential Mortality/Transmitter Effects. Placing a transmitter on a bird affects it in some, usually unknown, manner ranging from short-term behavioral modifications to potentially long-term physical effects, including death (Reynolds et al. 2004, Steenhof et al. 2006, Barron et al. 2010). Attributes of Short-eared Owls that may make them sensitive to transmitters include light wing-loading and an aspect ratio maximized for slow buoyant flight, forward thrust just above stalling speed, and aerial agility (D. Holt pers. comm.). We observed no obvious direct negative effects of transmitters on owls encountered during a 7–23 d post-deployment period on Seward Peninsula. Three of the owls were nesting females that continued to incubate eggs or brood young. Furthermore, the chronology and movement patterns of two females suggested that they successfully nested during the second summer tracking period.

Most (92%) owls survived the initial summer period, but then the mortality rate increased substantially during autumn migration. Our results are consistent with other studies that have documented higher mortality rates for raptors during migration than during stationary periods. The mortality rate for adult Ospreys (Pandion haliaetus), Marsh Harriers (Circus aeruginosus), and Montagu's Harriers (Circus pygargus) was six times higher during migratory periods than during stationary periods (Klaassen et al. 2014). Reduced apparent survival of Burrowing Owls (Athene cunicularia) was related to an increased number of storms during autumn and spring migration (Wellicome et al. 2014).

We acknowledge that PTTs could have contributed to higher mortality by reducing the owls' maneuverability and increasing their susceptibility to predation or accidents. Nevertheless, raptor predation was suspected as the cause of death for most recovered owls and, although the natural predation rates are unknown, several raptor species are known to prey on Short-eared Owls in North America, including Rough-legged Hawks (Buteo lagopus; Bechard and Swem 2002), Gyrfalcons (Falco rusticolis; Booms et al. 2008), Peregrine Falcons (Falco peregrinus; Sooter 1942), Snowy Owls (Murie 1929, Levin et al. 1977), and Great Horned Owls (Bubo virginianus; Killpack 1951). Short-eared Owl remains are regularly observed in raptor nests in western Alaska, and with increasing frequency in years of high owl abundance (T. Booms unpubl. data). On one occasion, remains of eight Short-eared Owls were found in a single Peregrine Falcon nest in northern Alaska (T. Swem pers. comm.). The morphologically and behaviorally similar Long-eared Owl (Asio otus) is the most frequent victim of raptor predation among European owls (Mikkola 1976, Marks et al. 1994).

Booms et al. (2014) listed raptor predation as a likely secondary threat facing Short-eared Owls in North America. Populations of some open-landscape or edge-associated species of potential predators (e.g., Peregrine Falcon and Red-tailed Hawk [Buteo jamaicensis]) have increased concurrent with a decline in Short-eared Owls (Sauer et al. 2012). Thus, although we cannot rule out PTTs having exacerbated the risk, the evidence gathered during this study and previously by others is consistent with raptor predation being an important source of mortality for Short-eared Owls that probably contributed to the high mortality observed during this study.

Management Implications. Multiple lines of evidence indicate a long-term, range-wide decline in the abundance of Short-eared Owls in North America since at least 1966 (Holt 1986, Wiggins et al. 2006, Sauer et al. 2012). Loss of native grasslands and shrublands is believed to be one of the primary causes of the decline (Wiggins et al. 2006, Booms et al. 2014). This and other studies demonstrate that Short-eared Owls are capable of exploiting a wide range of open habitats, including croplands, grasslands, and shrublands (Wiggins et al. 2006). Assessing the comparative value of different habitat types for Short-eared Owls during different parts of their annual cycle is an important area for future research. Additionally, although brief tracking periods and few multiyear samples limited relevant insight, the owls we tracked demonstrated little interannual range fidelity. Hence, the available information suggests that Short-eared Owls occupy areas across larger spatiotemporal scales than we typically consider when assessing a species' dependence on a particular parcel of land. The Short-eared Owls we tracked used a patchwork of diverse open habitats across a large area of western North America, which demonstrates that effective conservation of this species requires a continental perspective (Booms et al. 2014). Furthermore, our study highlights that areas used intermittently among years by Short-eared Owls should not necessarily be classified by managers as of low importance.

Although the causes of mortality in this study were not readily apparent, the high number of owl deaths during autumn migration suggests that the stresses of long-distance migration may play a major role in determining Short-eared Owl demography. A better understanding of seasonal and overall survival rates, and the factors influencing them, is an important missing component to conserving Short-eared Owl populations. Further, we interpret the timing and location of mortalities to suggest that conservation measures may be most needed at latitudes south of Alaska where the majority of mortalities occurred.