Identifying Predators and Fates of Grassland Passerine Nests Using Miniature Video Cameras

Introduction

Populations of many grassland passerines have declined in North America in recent decades (Peterjohn and Sauer 1993, Knopf 1994, Herkert 1995, Igl and Johnson 1997). Nest predation may contribute to these declines (Basore et al. 1986) or limit population recovery. Predation rates may vary widely among areas with different predator communities (Miller and Knight 1993). Loss and fragmentation of nesting habitat may increase levels of predation on grassland bird nests (Johnson and Temple 1990), but the effect of habitat fragmentation may depend on the composition of the local predator community (Nour et al. 1993). To understand why predation rates vary and how habitat features interact with predation, we must be able to accurately assess causes of nest failure and identify nest predators of grassland passerines.

Direct observations of predators at passerine nests are rare (Pettingill 1976, Sealy 1994) and may be biased toward diurnal and slower-moving predator species. Indirect evidence at the nest is of limited value because predators of passerine nests often leave little or no sign (Custer 1973, Hussell 1974, Major and Gowing 1994, Hernandez et al. 1997a). Limited sign increases the difficulty of determining nest fates, causes of nest failure, and identity of predators. Nevertheless, sign has been used by many researchers to attribute nest failures to various types of predators (Best 1978, Wray et al. 1982, Moors 1983, Vickery et al. 1992, Christman and Dhondt 1997). This procedure may be adequate when researchers have considerable knowledge of the predator community, but extrapolating to areas where predator communities differ can lead to errors (Brown et al. 1998). Furthermore, direct evidence linking sign left at nests with specific predators is largely lacking for grassland passerines.

A variety of camera systems have been used to document predation and other activities at nests in woody habitats (Innes et al. 1994, Major and Gowing 1994, Franzreb and Hanula 1995, Sykes et al. 1995, Dearborn 1996, Kristan et al. 1996). Most of these systems are unsuitable for use in grasslands where cameras must be very close to nests to avoid having the view obscured by vegetation. These systems generally have large cameras and mounting systems that would be readily visible near grassland nests and could affect rates of abandonment, predation, and brood parasitism. In addition, systems with still-frame cameras generally use tripping devices and artificial flash that could disrupt parental behavior at such short distances.

Many researchers have avoided the difficulties associated with studying natural nests by using artificial nests to study predation (Major and Kendal 1996). Artificial nests connected to still cameras (Picman 1987, Savidge and Seibert 1988, Reitsma et al. 1990, Danielson et al. 1996) and/or containing plasticine eggs (Mřller 1989, Nour et al. 1993, Bayne et al. 1997, Rogers et al. 1997, Hannon and Cotterill 1998) have been used to identify egg predators in a variety of habitats. However, factors that differ between artificial and natural nests, such as egg size or type (Haskell 1995, DeGraaf and Maier 1996, Bayne et al. 1997), nest appearance (Martin 1987), scent, and presence of adult birds (MacIvor et al. 1990, Sloan et al. 1998, Wilson et al. 1998) likely affect the species composition of their nest predators. Given the potential biases in predation studies that use artificial nests, Martin (1987:928) expressed the need for "direct determination of nest predators and their relative importance in nest predation."

In 1995-96, we worked with electronics specialists to develop a camera system specifically designed to monitor predation and other events at grassland passerine nests. We conducted field evaluations of prototype camera systems in 1996 and 1997, deploying cameras at nests of 10 species of grassland passerines in North Dakota. In this paper we present (1) a description of the camera system; (2) results of our camera evaluations; (3) egg, nestling, and nest fates and causes of nest failure at camera-monitored nests; (4) identities of predators and timing of predation events videotaped at nests; (5) documentation of sign left at nests by known predators; (6) a comparison of predation rates by nest stage, type, and height; and (7) an assessment of the accuracy of assigning nest fates, causes of nest failure, and predator types using conventional methods.

Study Areas

Cameras were used to monitor passerine nests in a variety of grassland habitats in Stutsman and Barnes counties, North Dakota. We monitored nests on 8 grassland sites 10-30 km from Jamestown (46°54'N, 98°42'W) in 1996, and on 14 sites 3-20 km from Woodworth (47°09'N, 99°23'W) in 1997. Most sites were public lands managed by the U.S. Fish and Wildlife Service; 6 were privately owned pastureland or retired cropland that had been planted to perennial grass. Vegetation included native and introduced species that had been subjected to various management practices (idle, hayed, grazed, burned); all fields were idle during our study. Most sites used in 1996 were dominated by grasses 30-80 cm tall; 1 site was dominated by forbs (Melilotus sp., Artemisia sp.) >1 m high. Sites used in 1997 were dominated by short to medium grasses (5-45 cm tall in late summer) with 0-20% forb cover, 0-25% shrub cover (mostly Symphoricarpos spp.), and 0-25 cm of litter. In 1997, habitat blocks were larger and more isolated from human activity.

Methods

Camera System Design

We chose video recording because (1) sensors used as tripping devices for still-frame cameras are likely to be activated by vegetation, parent birds, and other non-target objects; (2) shutter sounds and camera flashes can affect predation events; and (3) single images may fail to distinguish between predators and non-predatory visitors of nests. We recorded continuously because using a tripping mechanism to start the VCR results in a delay long enough to completely miss some events at the nest (Franzreb and Hanula 1995). We used 24-hr time-lapse recording because it provided about 4-5 images/sec, which we believed would capture even the fastest predation events.

Prototype camera systems were built by J. Christensen (Christensen Designs, Manteca, California), R. Fuhrman (Fuhrman Diversified, Seabrook, Texas), and D. Garcelon (Institute for Wildlife Studies, Arcata, California). Use of company names does not imply endorsement by the U.S. Geological Survey. Each system included a miniature black-and-white electronic-board camera in a waterproof housing with infrared (940-950 nm) light-emitting diodes around the lens to provide cryptic illumination at night. Camera lenses had fixed focal lengths of 3-4 mm with 51-88° horizontal field of view. Camera housings were cuboid or cylindrical, about 4 cm across, with volumes of 43-93 cm³. Most cameras were mounted on a wooden dowel that was pushed into the ground; an adjustable bracket allowed the camera to be positioned quickly. Cameras were typically 10-30 cm from nest contents. The camera was connected by 40-50 m of cable to a DC-powered time-lapse VCR (Panasonic model AG-1070DC) and a deep-cycle marine battery (120 amp-hr). An interface box (8 × 5 × 3 cm) with a connector for a portable video monitor was located 0.5-1.75 m from the nest to aid camera placement. The VCR was housed in a waterproof case with external connectors for the camera, battery, and portable monitor.

Data Collection

A crew of 3 (1996) to 6 (1997) people searched for nests from sunrise to early afternoon by systematically walking through fields swinging 2-m poles or dragging a 20-m rope to flush birds. Species, status, and location were recorded for each active passerine nest that was found; eggs were candled (Lokemoen and Koford 1996) and nestling ages were estimated (0 = day of hatch) to predict hatching and fledging dates. Nests were classified by height (on or above ground level) and type (open or covered by vegetation). A small flag was placed 4 m north of each nest. Nests without cameras were checked on estimated hatching and fledging dates, or at least once per week.

Cameras were deployed at a sample of nests. We tried to select representative numbers of nests (1) at egg and nestling stages, (2) on and above the ground, (3) on each area searched, and (4) for each nesting species. Daily visits were made to each VCR to check nest status (using a portable monitor), change tapes, and check battery power. The battery was changed about every 3 days. Each camera was left in place until the nest was destroyed, abandoned, or fledged young.

When a nest with or without a camera was no longer attended by adult birds, we searched for evidence of predation (damage to the nest; eggshell fragments, prey remains, disturbed vegetation, scats, etc. ≤1 m from the nest) and fledging (feather sheaths in the nest, fledglings or alarm-calling adults nearby). Nests were considered hatched if ≤1 egg hatched, and fledged if ≤1 nestling left the nest. Nests that failed were classified as abandoned (≤1 host egg or nestling remaining in unattended nest) or destroyed. Causes of failure were classified as predation, brood parasitism, observer interference, nest tipovers, starvation, or unknown. For the sake of simplicity, animals responsible for egg or nestling losses were referred to as predators (and the losses predation) even if they did not consume what they removed or destroyed.

Data Analysis

We assessed passerine response to cameras near their nests by comparing the proportion of camera nests that were abandoned <1 day after camera deployment to the proportion of nests without cameras that were abandoned <1 day after they were located and marked. For non-abandoned camera nests, we also calculated the length of time elapsed until an adult perched or sat on the nest following camera deployment (hereafter, return time). We used a 2-sample Kolmogorov-Smirnov test to compare distributions of return times between clay-colored sparrows (the only species with an adequate sample to treat separately) and all other species combined (see Table 1 for scientific names of nesting species).

We calculated Mayfield daily survival and predation rates (Mayfield 1961, 1975; Johnson 1979) for nests with and without cameras. Egg-laying days and the day a nest was found or a camera deployed were not included as exposure days. To test the effect of cameras, we assessed exposure days and fates of camera nests following the same protocol used for nests without cameras. For camera nests, information that would have been recorded during a scheduled nest visit was obtained by viewing a sample of videotape from the appropriate day (hereafter, simulated nest visit). When a mortality occurred between nest visits, the midpoint of the interval was used as the day of mortality. We used daily survival and predation rates with Sauer and Williams (1989) χ² procedure to test non-camera nests for effects of nest stage (incubation or nestling), height, and type, and to compare nests with and without cameras for effects of 1 treatment (camera) and 2 classification (stage, height) variables.

To evaluate the accuracy of assigning nest fates using information from nest visits, we compared actual fates of camera nests to fates assigned using simulated nest visits and observed sign. We also compared actual and simulated fates of eggs and nestlings, and actual and simulated daily survival rates.

We used data from camera nests to examine effects of nest characteristics on predator risk during day and night. In most calculations of predation rate, a nest is not considered depredated until the last viable egg or nestling is removed. As a measure of predation risk, we were interested in when the first egg or nestling was removed. We therefore calculated initial-predation rate based on nests that had not been depredated previously. Initial-predation rate was calculated by dividing the number of nests with predation by the number of exposure days accrued at successful nests and at depredated nests before the first egg or nestling loss occurred. We classified initial predation events as diurnal (after sunrise, before sunset) or nocturnal, and used χ² tests to examine initial-predation rate by nest stage, height, type, and day-night category.

We used Spearman's rank correlation to assess the association between nestling age (using oldest nestling) and initial-predation rate and survival rate calculated for each day (Pollock et al. 1989). We only used nestling ages 0-8 days because most young fledged by 9 days and sample sizes beyond 8 days were limited.

Results

During 1996 and 1997, we monitored fates of 364 nests of 13 passerine species (Table 1); 279 nests contained eggs and 85 contained nestlings when found. Cameras were deployed at 69 nests of 10 passerine species. At the time of camera deployment, 48 nests contained eggs, 6 were hatching, and 15 contained nestlings.

Response of Nesting Passerines to Cameras

Sixteen (23%) of 69 nests were abandoned <1 day after camera deployment. Abandonment rate was lower for 363 nests without cameras (χ²₁ = 51.99, P = 0.001), only 7 (2%) of which appeared to be abandoned <1 day after they were found. The actual abandonment rate for non-camera nests may be higher if some abandoned nests were scavenged and misclassified as depredated. For camera nests, abandonment was related to nest stage (χ²₁ = 5.76, P = 0.02), with only 1 abandonment occurring after hatch. Abandonments at camera nests were not restricted to any species, nest type, nest height, study site, camera type, camera distance from nest, or date or hour of camera deployment. In 1996, 3 (18%) of 17 camera nests were abandoned; in 1997, 13 (25%) of 52 were abandoned.

At 51% of 49 non-abandoned nests for which return times were measured, an adult returned to the nest in ≤30 min of camera deployment; at 76% of nests, an adult returned in ≤ 60 min. Return times to 11 of 38 nests in 1997 were longer than the longest return time in 1996 (61 min). The cumulative distribution function of return times differed (D = 0.385, P = 0.035) between clay-colored sparrows and other species. Clay-colored sparrows accounted for all 8 nests where adults were absent for >90 min after camera deployment. Absences of 26 clay-colored sparrows ranged from 1 min to 11.4 hr. Those absent for several hours returned just before dark; thus, the earlier in the day the camera was deployed, the longer the nest was unattended.

Egg, Nestling, and Nest Fates

Fates of eggs and nestlings were determined for 47 of 53 camera nests that were not abandoned (Tables 2 and 3); at the other 6 nests, equipment failed or (in 2 cases) was removed before nest completion. Of the 45 nests monitored during incubation that had known fates, 51% hatched (Table 2). Of the 41 nests monitored after hatching that had known fates, 56-68% fledged young (Table 3). At 6 nests, we documented ≥1 young departing the nest during or immediately after depredation of its nest mate(s) (hereafter, forced fledging). The percentage of nests classified as successful depends on whether forced fledgings are accepted as fledged. These young were within a day of minimum ages considered normal for fledging. In 1 case, however, videotape showed that the fledgling was caught and eaten by the predator just outside the nest.

We documented many events that could lead to errors when nest fates and causes of failure are assigned using information from periodic nest visits. Examples include (1) fledgings induced by predators at 6 nests; (2) 2 aboveground nests that gradually tipped over, dumping nestlings on the ground; (3) a parent that knocked 2 2-day-old nestlings out of the nest when it flushed at night; (4) a red fox (see Table 6 for scientific names of predators) scavenging eggs that had been punctured by a brown-headed cowbird; (5) removal of 3 eggs from 1 nest by a cowbird(s) that laid no parasitic eggs; and (6) a thirteen-lined ground squirrel killing and removing a parent from the nest.

Based on information from simulated nest visits, we correctly determined fates of 40 of 47 nests (22 fledged, 17 depredated, 1 apparent starvation). For 2 of the depredated nests, however, the stage at which predation occurred was incorrectly assigned to incubation rather than nestling stage.

Fates of ≤7 (15%) of 47 nests could have been misclassified using nest visits. For 5 nests in which all fledgings were forced by predators, 3 were classified as fledged and 2 were classified as depredated. For 1 of the nests classified as fledged, videotape indicated that no young survived. Of the 2 nests that tipped over, 1 fledged young but was classified as depredated; the other would have been classified as depredated or as destroyed due to unknown causes, depending on whether or not the young were scavenged before a nest inspection occurred.

Despite these errors, estimates of daily survival rates were similar when calculated using actual nest fates and using nest fates assigned from simulated nest visits (Table 4). For the incubation stage, the simulated sample included 2 more mortalities than the actual sample but a similar number of exposure days. For the nestling stage, the simulated and actual samples had a similar number of mortalities, but the actual sample included more exposure days because predation events often occurred after the mid-point between nest visits.

Fates of 138 nestlings from 40 non-abandoned nests were recorded on videotape. Using information from simulated nest visits, 67 of these were correctly classified as fledged and 49 were correctly classified as depredated. Fates of ≤22 nestlings (16%) could have been misclassified using information from nest visits. Four nestlings that were force fledged were classified as depredated and 4 were classified as fledged. Eight depredated nestlings were classified as fledged. One nestling that fledged from a tipped nest was classified as depredated; 5 others that died after falling out of tipped nests would have been classified as depredated or as destroyed due to unknown causes.

Daily Survival and Predation Rates

For nests without cameras, Mayfield survival rates were higher (χ²₁ = 6.42, P = 0.011) and predation rates lower (χ²₁ = 3.68, P = 0.055) for ground nests than for aboveground nests. Nest losses other than predation (i.e., mortality rates excluding predation) also were lower for ground nests (χ²₁ = 9.25, P = 0.002). Survival rates were not related to nest stage (χ²₁ = 0.44, P = 0.51), but predation rates tended to be lower on eggs than nestlings (χ²₁ = 2.45, P = 0.12). We detected no difference in survival (χ²₁ = 0.11, P = 0.74) or predation rates (χ²₁ = 0.00, P = 0.98) between open and covered nests. We found no evidence of interactions of nest stage, height, and type (χ²₁ = 0.33-3.22, P = 0.20-0.85).

When testing for camera effects on daily survival and predation rates, we found evidence of a 3-way interaction among nest stage, height, and camera presence for predation rates (χ²₁ = 4.49, P = 0.034) and survival rates (χ²₁ = 2.82, P = 0.093). The interaction resulted from the high predation rate estimated for aboveground camera nests during incubation (Fig. 1). For this group, the number of exposure days (36) was too small to provide reliable estimates. After eliminating this group, we found no evidence of other interactions (χ²₁ = 0.06-3.03, P = 0.22-0.97), and detected no camera effect on either survival (χ²₁ = 0.30, P = 0.58) or predation rates (χ²₁ = 0.95, P = 0.33). However, there appeared to be a tendency for predation rates to be lower on camera nests than non-camera nests (Fig. 1).

For camera nests, effects of nest type on initial-predation rate differed between night and day (χ²₁ = 4.88, P = 0.027). Open nests were more vulnerable than covered nests during the day (χ²₁ = 5.93, P = 0.015), but there was no difference at night (χ²₁ = 0.51, P = 0.47; Table 5). We found no evidence of an interaction between time of initial predation and nest height (χ²₁ = 2.14, P = 0.14), or between time of initial predation and nest stage (χ²₁ = 0.10, P = 0.75). Sample sizes were too small to reliably test for interactions among nest characteristics within day and night.

Initial-predation rate increased with nestling age for days 0-8 (r_s = 0.63, P = 0.070). No relation was found, however, between daily nest survival rate (measured when the last nestling was lost) and nestling age (r_s = -0.32, P = 0.41).

Predator Identification and Behavior

We videotaped ≥11 species removing or destroying eggs or nestlings at 29 nests of 10 passerine species. Predators were identified to species at 21 nests and genus at 3 nests (Table 6; sample images in Fig. 2). At 2 nests, we narrowed predator identity to 2 species or genera. Predators at 3 nests never came into camera view and thus were classified as unknown. Large predators (deer, canids) were difficult to identify because only parts of their heads were in view. Wider-angle lenses used in some cameras were meant to reduce this problem, but image quality decreased as field of view increased. Images produced during the day were sharper than those produced at night with infrared light, although shadows produced by bright sunlight sometimes made daytime images difficult to interpret.

All 4 avian predations of camera nests occurred during daylight; the 22 mammalian predations were scattered throughout the 24-hr period. Only mice were documented depredating nests both day and night. At 6 nests, eggs or nestlings were depredated during >1 day or night (22-116 hr). In all these cases the predators were small: mouse (1 nest), thirteen-lined ground squirrel (3), Franklin's ground squirrel (1), and brown-headed cowbird (1). At 17 nests, all nest contents were destroyed or force fledged within a single day or night. At 15 of these nests, all contents were gone in < 15 min; at 2 nests (thirteen-lined ground squirrel, mouse), all contents were gone in 2-3 hr.

All predators removed 1 egg or nestling from the nest at a time, thus allowing some nestlings to escape (i.e., 8 forced fledgings at 6 nests). Forced fledgings were induced by white-tailed deer (2 nests), Franklin's ground squirrels (2), and thirteen-lined ground squirrels (2) at nests of 5 different species.

Evidence at Nests

No sign was evident at 12 of 26 depredated nests for which the final fate was documented on videotape (Table 6); 9 of the 12 nests with no sign were on the ground. Nest bowls were damaged at 6 of 12 depredated aboveground nests and only 2 of 14 depredated ground nests. Counting 2 nests that tipped over, aboveground nests accounted for 8 of 10 nest bowls that were disturbed, destroyed, or left with holes. Large mammals (badger, canid, deer) left no sign at nests they depredated. Ground squirrels varied from leaving no sign to destroying the nest bowl, and left 2-4-cm holes in 3 nests. Mice sometimes opened eggs in nests, leaving small eggshell fragments, but sometimes removed eggs from nests before opening them. Two avian predators left no sign, whereas a third completely destroyed a nest bowl and a fourth left punctured eggs.

Parent birds and scavengers modified evidence left by predators at some nests. A mouse killed 3 of 4 5-day-old clay-colored sparrow nestlings and left them in the nest; the parents removed the dead nestlings over the next 17 hr. Western meadowlark eggs punctured by a cowbird were scavenged by ants and beetles within 24 hr, and by a red fox within 54 hr. An adult clay-colored sparrow ate the shell and remaining contents of an egg 1 min after the mouse that had opened the egg was chased from the nest.

The length of time parent birds continued to attend depredated nests was measured for 23 camera nests. The duration of parental attendance following the last predation event was highly variable (range 0-19.6 hr) and averaged 6.8 ± 1.2 (SE) hr. The number of parental visits following the last predation also was highly variable (range 0-41) and averaged 9.7 ± 2.3. Of 15 nests last depredated during daylight, 5 were still being attended the next day; however, all 5 were deserted by 1035 hr. Of 8 nests last depredated during the night, 6 were still attended the following morning; the last of these was deserted by 1217 hr. Seven hours after predation, no nests were visited >3 times/hr.

Discussion

Predator Identification and Behavior

Nest predators identified in this study span a broad range of taxonomic groups and feeding habits. Most predators we videotaped were suspected to be predators of grassland passerine nests, although few have been documented. White-tailed deer are known to eat passerines caught in mist nests and have been suspected of nest predation (Sealy 1994), but our 2 videotaped events may be the first documentation of white-tailed deer eating nestlings. Deer mice (Peromyscus maniculatus) have been implicated as predators of grassland nests (e.g., Lane 1968, Rogers et al. 1997), but jumping mice (Zapus spp.) have not. Clutch destruction by brown-headed cowbirds (without brood parasitism) has been surmised from indirect evidence (e.g., Arcese et al. 1996) but not documented.

Although cameras proved useful for identifying predators, our predator list undoubtedly is incomplete. Striped skunks (Mephitis mephitis) and raccoons (Procyon lotor) are common in North Dakota grasslands (Sargeant et al. 1993) and known nest predators, yet we did not videotape either species depredating a nest. Although snakes were not documented on tape, our nest searchers encountered a plains garter snake (Thamnophis radix) depredating Savannah sparrow nestlings. Snakes probably are easier to observe depredating nests than other species because many are diurnal and snakes swallowing prey may not leave nests quickly when humans approach.

The dynamics of predator populations over time and space increase the difficulty of completely documenting a predator community. During our study, for example, skunk and canid populations were reduced in some areas by rabies (Greenwood et al. 1997) and mange (S. H. Allen, North Dakota Game and Fish Department, personal communication), respectively. A predator list obtained with cameras also might be incomplete or biased if cameras affect the likelihood of predation by certain species. Some predators (e.g., corvids) may be attracted to novel things in their environment, and thus more likely to find camera nests than non-camera nests. In contrast, predators such as red fox and coyote are believed to be wary of novel things in their environment (Hernandez et al. 1997b) and of human scent (MacIvor et al. 1990), and therefore may avoid camera nests. Although we documented 2 canids removing eggs from camera nests during our study, it is possible that canid predation was underrepresented in our data. In any case, populations of a given species may differ in their reaction to novel items and human scent, depending on the extent to which those populations have been persecuted by humans (Birkhead 1991:221, Götmark 1992:80) or habituated to human presence.

Sign and Nest Fates

Major disturbances at failed passerine nests (e.g., destroyed nest bowl, trampled vegetation) frequently have been attributed to predation by large mammals; empty nests with no disturbance have been attributed to predation by birds, snakes, or small mammals (Best and Stauffer 1980, Wray et al. 1982, Hoover et al. 1995, Patterson and Best 1996, Christman and Dhondt 1997). Minor disturbance and eggshell fragments in nests have been attributed to predation by small mammals, and circular holes in the bottom of nests to predation by snakes (Best 1978). These criteria were generally wrong for classifying predators of grassland passerine nests on our study areas. Variability of sign within species and overlap among species observed in our study and others (Major 1991, Brown et al. 1998, Marini and Melo 1998) indicate that evidence at nests often is unreliable for identifying predators of passerine nests.

Sign at nests also may lead to misinterpretations of nest fates. We documented cases in which successful nests appeared to have been depredated and depredated nests appeared successful. Because mistakes in assigning nest fates with sign were made in both directions, they were not reflected in overall estimates of survival rates. However, the lack of large differences in our actual and simulated Mayfield survival rates does not preclude the potential for significant levels of error from nest-visit data in other studies. Studies that involve a single nest type may be more likely to exhibit a bias in the direction of errors. Our results, for example, suggest that predation is more likely to be overestimated for aboveground nests and underestimated for ground nests.

The level of error deemed acceptable depends in part on research objectives. The maximum error rate in our sample would not likely have changed the relative importance of predation as an agent of nest failure. Most studies have implicated predation as the main cause of nest failure, often by large margins (Martin 1992). Our error rate may have been large enough, however, to change results of studies that compare predation rates to assess the effects of nest and habitat characteristics (e.g., edge, block size).

Attendance by adult birds is another type of evidence researchers use to assess nest fate when visiting nests. We found that parent birds may attend nests up to 20 hr after the last predation event. Thus, it may be misleading to assume a nest is still active based solely on the presence of an attending adult.

Predation Rates and Risk Factors

Although we did not detect a statistical difference in predation rates between nests with and without cameras, the possibility that predation was slightly lower on camera nests (as suggested by Fig. 1) could not be eliminated. If true, it may indicate that some types of predators avoided camera nests. Alternatively, it may indicate that nests surviving long enough to be in the camera treatment (i.e., surviving >1 day after being found and >1 day after camera setup) had a lower likelihood of predation than other nests.

Knowing the time of predation events at camera nests allowed us to take a new approach in exploring predation risks associated with different nest characteristics (eggs-nestlings, ground-aboveground, open-covered bowl). Time of predation events rarely has been documented in the past. Picman and Schriml (1994) documented times that predators were photographed at artificial nests and Nolan (1978:413) separated night and day predation by inspecting prairie warbler (Dendroica discolor) nests at least twice a day. They speculated that the relative importance of diurnal and nocturnal predation depended on the composition of the predator community. This was supported by our study, in which the preponderance of ground squirrel predation resulted in a higher proportion of daytime predations.

Some researchers have detected no difference in predation rates between incubation and nestling stages of passerines (Zimmerman 1984, Suárez and Manrique 1992, Cresswell 1997, Roper and Goldstein 1997); others have reported higher predation rates during incubation (Dixon 1978, Best and Stauffer 1980, Martin 1992) or during the nestling stage (Young 1963, Schaub et al. 1992, Suárez and Manrique 1992, Morton et al. 1993, O'Grady et al. 1996). Higher predation rates on nestlings might result from increased parental activity or begging calls of young (Cresswell 1997, Roper and Goldstein 1997). Because most activity at nests ceases at night, any increased vulnerability to predation resulting from activity levels should occur during the day (Roper and Goldstein 1997). Although small sample size limited our ability to test this conclusively, our comparisons of diurnal and nocturnal initial-predation rates illustrate a potentially important investigative approach made possible by camera data.

Beginning with Mayfield (1961), many researchers have calculated daily nest survival separately for incubation and nestling stages, acknowledging the potential for these to have different predation rates. Most researchers are unable to test the assumption that predation rates are constant within each of these stages. If this assumption is incorrect, nest survival estimates based on exposure days may be biased (Klett and Johnson 1982). Given that nest contents and parental activity change little through incubation (Mayfield 1961, Roper and Goldstein 1997), one might expect predation rates to be relatively constant during this period (but see Willis 1973, Klett and Johnson 1982). In contrast, because of changes in nestling biomass and activity, and in parents' nest visitation rate, one might expect predation rates to increase through the nestling period (Young 1963). Data from our camera nests support this expectation; initial-predation rate increased with nestling age.

Initial-predation rate provides a better measure of predation risk relative to nest stage or nestling age than does daily survival rate. Daily survival rate includes sources of nest loss other than predation and is associated with the age of nestlings at the time the last nestling is lost rather than when the first nestling is lost. The loss of the first nestling (initial predation) more likely coincides with a predator's discovery of the nest.

The relation between predation rate and nest height has long interested researchers (e.g., Martin 1993), but few grassland nest studies have examined it. Knapton (1978) found that clay-colored sparrow nests with bases ≤10 cm from the ground were more successful than those built higher in vegetation. Aboveground nests may be more vulnerable to predation if they are more conspicuous than ground nests. If so, the risk should be greatest during the day. This prediction was not supported by our sample of camera nests, but more data are needed to test this hypothesis conclusively.

Overall, we found that aboveground nests suffered higher mortality than ground nests, even when predation mortalities were excluded. Aboveground nests can fall apart or tip over, are more vulnerable to wind damage, and may be more visible to cowbirds. The relative importance of these and other mortality factors can be difficult to assess using only information from nest visits.

Many researchers have examined the relation between nest concealment and predation rate (Nolan 1978, Martin and Roper 1988, Martin 1992, Cresswell 1997); however, few have assessed whether differences in nest bowl structure affect concealment from predators. Studies by Mřller (1989) and Wray et al. (1982) suggest that the importance of nest type and concealment depends on the composition of the predator community. Conflicting results among studies about the importance of nest concealment (Martin 1992, Götmark et al. 1995, Howlett and Stutchbury 1996) may reflect differences in predator species, activity patterns, and hunting methods. Increased vulnerability to predation resulting from differences in nest visibility should be most pronounced in daylight. Data from our camera nests support this prediction; we found that open nests had higher initial-predation rates during the day. The fact that we detected no difference in predation rates between open and covered non-camera nests suggests that the importance of nest cover may only be apparent if diurnal and nocturnal predation can be separated.

Nest Abandonment and Parental Return Rates

The risk of abandonment for camera nests may be influenced by several factors. We found that abandonment was more likely if cameras were deployed at nests during incubation than during brood rearing. However, we have deployed cameras at 4 nests during egg laying (P. J. Pietz, unpublished data) and none of these was abandoned. Many species of birds are considered more vulnerable to disturbance during earlier stages of nesting (Livezey 1980, Lanyon 1994, Martin and Gavin 1995, Hill and Gould 1997). Tolerance for disturbance early in nesting may vary considerably among species (Thompson et al. 1999), individuals, or environmental conditions.

Based on abandonment rate and return times, birds seemed to be more disturbed by cameras in 1997 than 1996. Abandonment rate was considerably lower in a subsequent study in 1998 (2-3 of 29 nests; P. J. Pietz, unpublished data). Conditions in 1997 differed from the other 2 years in that cameras were deployed earlier in the day, study areas were more remote, and weather conditions were more stressful for nesting birds.

Management Implications

The potential for bias in predation rates based on camera data may depend on the predator community. If important predator species are attracted or deterred by cameras, predation rates could be overestimated or underestimated. Even if no camera bias exists, the relative importance assigned to various predator species may be affected if sample sizes are small or predator communities are dynamic. Further, both the predators documented and their relative importance may be affected by how cameras are spatially deployed. If cameras are clustered, they will be exposed to fewer individuals of large predator species, the same individual may depredate multiple nests, and individuals that encounter >1 camera nest might learn to associate cameras with nests.

The benefits of information obtained with cameras should be weighed against the risk of increasing abandonment rates. Risk of abandonment can be reduced by delaying camera deployment until after hatch, but at the price of missing predation events that occur during the egg stage. Some evidence suggests that abandonment risk may be reduced by deploying cameras after noon.

Despite these caveats, cameras can provide information on predator and nesting ecology that is impossible to obtain by other means. For example, knowing the identity of predators, their behavior at nests, and timing of predation events can help investigators identify nest and habitat characteristics that increase nest vulnerability.

Biologists should be discouraged from attempting to identify predators using only sign at failed nests. This common practice is probably unreliable even for coarse levels of identification (e.g., large mammals). Sign is least likely to be useful in studies of ground-nesting passerines because most ground nests are not damaged during predation.

Errors interpreting sign that result from alteration in sign (e.g., by parent birds, scavengers) may be reduced by visiting nests more frequently, unless visits themselves affect nest outcomes (e.g., influencing predation, inducing abandonment or premature fledging). This risk is increased by the need to approach most grassland nests closely to check contents. Because adults attended nests after predation events, we recommend that empty, attended nests be checked again on the subsequent day to assess nest fate.

Cameras provide more specific and accurate information on fate of nest contents and causes of nest failure than can be obtained from periodic visits. However, detailed information from cameras presents some challenges to conventional nest-fate terminology and classification. As this type of information accumulates, researchers may need to set standards for handling the more common events, such as forced fledging.

As illustrated above, cameras proved effective for evaluating field methods currently used to assess nest success, causes of nest failure, and predator types. Such evaluations can help researchers choose the most appropriate and cost-effective methods for obtaining the data they need. Cameras also may prove useful for evaluating effects of management and research activities on grassland nesting birds.

Literature Cited