J. Field Ornithol., 72(4):556–572

CORVID SURVEY TECHNIQUES AND THE RELATIONSHIP BETWEEN CORVID RELATIVE ABUNDANCE AND NEST PREDATION

JOHN M. LUGINBUHL,JOHN M. MARZLUFF AND JEFFREY E. BRADLEY College of Forest Resources, University of Washington, Seattle, Washington 98195 USA

MARTIN G. RAPHAEL U.S. Forest Service, Pacific Northwest Research Station, Olympia, Washington 98512 USA

DANIEL E. VARLAND Rayonier, Hoquiam, Washington 98550 USA

Abstract.—We conducted a four-year study on the Olympic Peninsula of Washington to assess the relationship between corvid (Gray [ canadensis], Steller’s Jay [ stelleri], [ brachyrhynchos] and [Corvus corax]) abun- dance and the risk of nest predation. We assessed risk of predation through the use of artificial mid-canopy nests and assessed corvid abundance using a variety of techniques in- cluding point-count surveys, transect surveys, and the broadcast of corvid territorial and predator attraction calls. Point counts of corvid abundance had the strongest correlation with predation on artificial nests containing eggs. The relationship between nest predation rate and corvid abundance was strongest when study plots were used as replicated measures of landscape conditions rather than as independent samples. We suggest using the maximum value for each corvid attained from several temporally replicated point-count surveys in each study plot. Corvid point-counts should be conducted on days with light winds (Ͻ20 kph) and no more than light precipitation. Use of attraction calls is important for gaining a meaningful measure of corvid abundance. Their use may overrepresent corvids at the local plot scale but is important in assessing the landscape scale presence of wide-ranging (Amer- ican Crows) and often non-vocal (Gray Jays) corvids.

TE´ CNICAS DE ENCUESTAS DE CO´ RBIDOS Y LA RELACIO´ N ENTRE ABUNDANCIA RELATIVA DE ESTOS Y DEPREDACIO´ N DE NIDOS Sinopsis.—Durante cuatro an˜os llevamos a cabo un estudio en la Penı´nsula Olı´mpica de Wash- ington para determinar la relacio´n entre la abundancia de cuervos (Perisoreus canadensis), (Cya- nocitta stelleri), (Corvus brachyrhynchos), (Corvus corax) y el riesgo de la depredacio´n de nidos. Evaluamos el riesgo de depredacio´n a trave´s del uso de nidos artificiales colocados en la estrata media de arboladas. Adema´s, de terminamos la abundancia de co´rbidos utilizando te´cnicas va- riadas que incluyeron conteos de puntos, transectos y el uso de grabaciones de la llamada terri- torial de co´rbidos y las llamadas de atraccio´n de depredadores. Los conteos de puntos de abun- dancia de co´rbidos tuvieron la mayor correlacio´n de depredacio´n de nidos artificiales con huevos. La relacio´n entre la tasa de depredacio´n de nidos y la abundacia de co´rbidos tuvo mayor fortaleza cuando las a´reas de estudios fueron utilizadas como re´plicas de medidas de las condiciones del paisaje en vez de muestras independientes. Sugerimos el uso del valor ma´ximo de cada especie de co´rbido, obtenido de varias re´plicas de puntos de conteo en cada estacio´n de estudio. Los puntos de conteo de co´rbidos se deben conducir en dı´as con poco viento (menor a 20 kph y no ma´s de una precipitacio´n liviana. El uso de grabaciones de llamadas es de gran significado para determinar la abundancia de los co´rbidos. Su uso pudiera sobre representar los co´rbidos en la escala de un punto particular, pero es importante para fijar, en la escala de paisaje, la presencia de especies de amplia distribucio´n o de poca vocalizacio´n.

Nest predation is pervasive and arguably the most important factor limiting avian productivity (Nice 1957; Ricklefs 1969; Wilcove 1985; Mar-

556 Vol. 72, No. 4 Counting Corvids [557 tin 1993a,b). Corvids are efficient nest predators (Angelstam 1986; Marz- luff and Balda 1992) whose populations have increased dramatically in the western United States (Marzluff et al. 1994) and in urban areas world- wide (Fraissinet 1989; Konstantinov 1996) over the last century. This in- crease may have important implications for the conservation of open- nesting , because many studies have shown a positive correlation between various indices of corvid abundance and predation rates on near- by nests (Angelstam 1986; Johnson et al. 1989; Andren 1992). The poten- tial for corvids to limit other populations poses a need for a consis- tent method to assess corvid numbers and index the risk of nest preda- tion. Such a method would allow managers to assess the risk of corvid predation simply by surveying corvids. Development of this method requires a thorough understanding of the biotic and abiotic factors that affect our ability to detect corvids, whether measures of corvid abundance correlate with nest predation, and the spatial scale at which any correlation exists. To this end, we conducted a four-year study (1995–1998) to assess the relationship between corvid (Gray Jay [Perisoreus canadensis], Steller’s Jay [Cyanocitta stelleri], Ameri- can Crow [Corvus brachyrhynchos] and Common Raven [Corvus corax]) abundance and corvid nest predation. We examined a variety of survey protocols to determine what factors affect corvid surveys and how well corvid abundance predicts the rate of nest predation at simulated Mar- bled Murrelet (Brachyramphus marmoratus) nests. Marbled Murrelets are federally threatened birds that nest in the middle to upper canopy of mature coastal forests of the Pacific Northwest (Ralph et al. 1995b). Mar- bled Murrelets suffer high nest predation rates, and corvids (Common Ravens and Steller’s Jays) are suspected to have caused the majority of known nest failures (Nelson and Hamer 1995a; Miller et al. 1997).

METHODS Study area.—The study area was located on the western side of the Olympic Peninsula of Washington State, north and south of Forks, Wash- ington (47Њ56ЈN, 124Њ23ЈW) in the Hoh, Soleduck, Quinault, and Queets River drainages. Study plots consisted of mixed-conifer forest ranging in age from 80–250 yr and in size from 37–106 ha. Plot elevations varied between 45 and 600 m. The study area is adjacent to the major concen- tration of murrelets in Washington (Varoujean and Williams 1995) and is in a landscape used substantially by nesting murrelets (Horton and Harrison 1997). We conducted experiments and surveys in a total of 49 plots in 12 landscape categories. Landscape categories consisted of all possible com- binations of two classes of forest fragmentation, three classes of forest structure, and two classes of proximity to human-use areas. Fragmentation classes were fragmented (plot Ͼ75% surrounded by clear-cuts) and con- tinuous (plot Ͼ75% surrounded by mature forest). Levels of plot frag- mentation (and plot size) varied somewhat due to the limited number of suitable forest patches occuring ‘‘naturally’’ throughout the study area. 558] J. M. Luginbuhl et al. J. Field Ornithol. Autumn 2001

TABLE 1. Summary of experimental design. Numbers in the table refer to study plots in each treatment.

Forest structure Very complex old Simple mature Complex mature growth (80–120 yr) (80–120 yr) (Ͼ200 yr) Distance from human use areas Ͻ1km Ͼ5km Ͻ1km Ͼ5km Ͻ1km Ͼ5km Surrounding forest landscape Fragmented 5 4 5 4 3 5 Continuous 3 4 3 3 5 5

Fragmented plots were delineated using existing forest edges, while con- tinuous plots were more arbitrarily delineated from within larger, more continuous forest landscapes. Structure classes were simple mature (80– 120 yr, canopy single storied with few gaps), complex mature (80–120 yr, canopy 1–2 storied with many gaps), and very complex old growth (Ͼ200 yr, canopy multi-storied with many gaps). The two proximity to human- use area classes were near (Ͻ1 km from plot edge) and far (Ͼ5 km from plot edge). We defined ‘‘human-use’’ as any human development that could potentially attract corvids (i.e., towns, farms, campgrounds, high- ways, and dumps; Table 1). Each plot (spatial replicate) was examined in 2–4 subsequent years (temporal replicates). Artificial nest experiments.—We climbed trees and placed nests using techniques that minimize disturbances that might cue predators to the nest’s location, such as damage to the bark from spurs or human scent trails left from touching the bole or limbs. We followed Perry (1978) to place fishing line, then 4-mm cord, and finally an 11-mm static climbing rope, into a likely nest tree. We avoided contact with the tree by climbing with ascenders and rappelling to the ground. We wore latex or vinyl gloves while taking measurements and preparing the nest. We marked nest trees on the ground with white plastic flagging hung in a random direction approximately 3 m from the tree. The entire process of climbing, placing nests, and rappelling took approximately 90 min. Murrelet nests are easy to mimic because eggs are laid in simple de- pressions on moss-covered branches, eggs are sometimes left unattended for several hours during incubation, and nestlings are left alone for much of the day after they reach3dofage(Nelson and Hamer 1995b; Manley 1999). We simulated nests at typical heights and locations for murrelets on moss-covered branches with diameter Ͼ11 cm, within the live crown, Ͼ15 m above the ground, well covered from above (x¯% overhead cover ϭ 83.0%, SE ϭ 0.46, n ϭ 671) and close to the trunk (x¯ distance to bole ϭ 40.2 cm, SE ϭ 0.93, n ϭ 671; Marshall 1988; Singer et al. 1991; Hamer and Nelson 1995). We placed a total of 905 artificial murrelet nests (Table 2). Nests were placed in separate trees at least 50 m apart, and only six nests were placed Vol. 72, No. 4 Counting Corvids [559

TABLE 2. Summary of nest distribution within experimental design. Numbers in table refer to total nests and, in parentheses, nests with eggs in each treatment.

Forest structure Simple mature Complex mature Very complex old growth (80–120 yr) (80–120 yr) (Ͼ200 yr) Distance from human use areas Ͻ1km Ͼ5km Ͻ1km Ͼ5km Ͻ1km Ͼ5km Surrounding forest landscape Fragmented 90 68 88 69 64 109 (45) (35) (42) (35) (31) (55) Continuous 36 98 47 36 100 100 (18) (48) (24) (18) (50) (50) in a given plot at any one time. The resulting low densities of our nests (0.6–0.17 nests/ha) reduced the effects of area-restricted searching prac- ticed by Common Ravens and American Crows (Marzluff and Balda 1992) and decreased the possibility that high nest densities may cause predators to associate our activities with food rewards (Sieving 1992; Major et al. 1994). Two artificial nests, one with an egg and one with a nestling, were placed at each of three distances from the forest edge (Ͻ50 m, approxi- mately 100 m, and Ͼ200 m from the edge). We simulated nest contents with plastic eggs and taxidermy mounts of nestlings. We painted eggs to resemble Marbled Murrelet eggs and coated them with wax (household paraffin) to aid predator identification. The plastic eggs were slightly larger than actual marbled murrelet eggs (64 ϫ 44 mm vs. 59.5 ϫ 37.4 mm; Nelson and Hamer 1995b). Nestling models were made from domestic chicken (Gallus sp.) chicks preserved with bo- rax. Nestling models were dark-colored (mostly black), approximately 10 cm long, and were placed in a posture that imitated a crouched or sleep- ing nestling. Both egg and nestling models were placed in cedar chips for a minimum of 12 hours prior to placement to limit human scent. Although decay of mounted nestlings was limited (they appeared visually unchanged after 30 days), they emitted odor perceptible to humans. Mar- bled Murrelet nests with a well-developed fecal ring also give off odor perceptible to humans from 2 m away (T. Hamer, pers. comm.). We used eggs and nestlings because they represent both life-history stages of mu- rrelets vulnerable to nest predation, artificial eggs are highly attractive to corvids (Heinrich et al. 1995) and provide visual cues to potential pred- ators, and nestlings offer olfactory cues as well as visual ones, which may better mimic a real nest and attract scent-oriented predators (Ratti and Reese 1988; Major 1991; Whelan et al. 1994; Darveau et al. 1997). Nests were monitored every other day for 30 d, the approximate in- cubation and brooding period of murrelets (DeSanto and Nelson 1995). All models included a motion-sensitive radio transmitter (Advanced Te- lemetry Systems models 5903 and 7PN) that allowed for frequent, remote nest checks from outside of the study plot (Willebrand and Marcstrom 560] J. M. Luginbuhl et al. J. Field Ornithol. Autumn 2001 1988). Remote monitoring allowed us to determine date of predation while limiting human presence at nest sites. Remote monitoring also al- lowed us to reclimb simulated nests immediately after predation, before other predators or heat from sunlight obscured clues to predator identity left in wax. Motion-sensitive radio transmitters in models allowed us to accurately monitor the fate of artificial canopy nests. A total of 705 of 905 nests (81%) were depredated during the 30 d trials. Twenty-one percent of 737 depredated eggs and nestlings were moved by predators and were relo- cated with the aid of telemetry. There were 106 cases (12% of 905 nests) where telemetry indicated predation and we found no signs of distur- bance upon reclimbing. We simply continued to monitor these ‘‘false alarms.’’ Rarely (19 cases or 2.1% of 905 nests), nests were disturbed without being recorded through telemetry. These experiments were re- done in a new location to determine accurately the timing of predation. We used a variety of techniques to identify potential predators. All eggs and transmitters inside nestlings were coated with household paraffin to record marks left by predators (Møller 1987; Haskell 1995). We moni- tored an additional 82 artificial nests using 35-mm cameras attached to an active infrared motion detection system (Trailmaster௡ Model TM 1500, Goodson and Assoc., Inc., Lenexa, Kansas, as described by Hernandez et al. (1997). These cameras imprint photographs with date and time, and are equipped with auto-advance, allowing photography of subsequent predators without researchers revisiting the nest. They provided an im- portant way to calibrate the predator identification we based on marks in wax. Nests observed with cameras were not included in the determination of predation rate due to the potential biases associated with the presence of camera equipment. We refined our experimental nest methods after the 1995 season, re- quiring that some data be screened from analysis. In this first season, two experiments where telemetry did not indicate predation did in fact show signs of predation upon removal of the experiment. These were classified as failed experiments and were dropped from the analysis (in subsequent seasons, data from failed experiments was replaced by redoing the ex- periment in a new tree). We also excluded nest predation data from 31 nest placements with cameras in 1995 (in subsequent years, cameras were used only at placements in addition to the six used for rate determina- tion). Corvid surveys.—Our overall survey design was loosely based on the standard point-count techniques suggested by Ralph et al. (1993, 1995a). We surveyed corvids from 7–17 points spaced evenly on a 250-m grid within each plot (plot size dictating the number of survey points). Each count lasted 10 min, during which time all vocal and visual detections were recorded as either being within a 50-m radius of the observer, or beyond a 50-m radius (but still within the habitat of interest). We did not attempt the similar variable radius point count suggested by Ralph due to the difficulty in determining distances to birds detected greater than Vol. 72, No. 4 Counting Corvids [561

50 m from point center in the dense forests in our study area. Corvids were also counted as we walked between survey points and tallied sepa- rately from corvids detected at point-count stations. All corvid detections were sketched onto a map of the survey area to minimize double-count- ing. At 1–3 selected point-count centers we used corvid territorial calls (Knight & Hale crow hunting call) and ‘‘predator-attraction’’ calls (Low- man ‘‘Circe’’ predator call) to maximize our chances of seeing and hear- ing corvids. Call-points were placed at least 500 m from each other and at least 125 m from plot edge. Calls were used during a second 10-min point after a standard passive survey at that point. Corvid detections dur- ing the second 10-min call-point were tallied separately from detections during the initial passive point. The pattern of call use during the 10-min call-point was: 30 s crow call, 1 min silence, 30 s predator call, 1 min silence, 30 s crow call, 1 min silence, 30 s predator call, 5 min silence. To examine the effects of attractant calls, we conducted a paired control count at the point location immediately before point locations where calls were used. At these controls, the observer conducted a second passive 10- min count (without the use of attractant calls) after the standard passive survey at that point. This design allowed us to control for the potential bias of increased observer presence (20 min instead of 10; Verner 1985). Counts from different points within a plot were averaged to give a mean number of each corvid species per point-count. Averages were used be- cause of varying plot size. Each plot was surveyed two times during the period corresponding to the Marbled Murrelet nesting season, the first series between 6 April and 22 June, and the second between 1 July and 26 August. Precipitation intensity (none, intermittent light, steady light, intermittent moderate, steady moderate; U.S. National Weather Service 1995) and approximate wind speed (Beaufort scale) were also recorded during surveys in order to evaluate how these varying weather conditions impacted our ability to detect corvids. Analysis.—Statistical analyses were performed using version 8.0 of SPSS (1997). We used analysis of variance to test for the effects of wind speed and precipitation levels and a paired t-test to examine the effect of at- tractant calls on corvid detection rates. During corvid surveys, we found that some birds (especially ravens, crows and Steller’s Jays) could be heard at great distance. Since we minimized double-counting of individual birds, we tended to detect these louder, more vocal corvids on points early in surveys of individual plots (a relatively large proportion of each plot could be surveyed from each point for these ‘‘louder’’ vocalizing species) and were not recorded if detected again at later points. Due to this lack of independence between survey time periods, we did not rigorously analyze the number of corvid detections by time of day. We investigated relationships between measures of corvid abundance and nest predation using standard parametric correlation and regression techniques. We examined 30 measures of corvid abundance representing all possible combinations of five species classes (crows, ravens, Gray jays 562] J. M. Luginbuhl et al. J. Field Ornithol. Autumn 2001

TABLE 3. Numbers of leaving signs at 905 artificial nests containing eggs and nes- tlings. Numbers in parentheses are percentages of column total.

Nest contents Nestling (%) Egg (%) Corvids jay 12 (2.6) 171 (37.9) crow/raven 9 (2.0) 28 (6.2) unknown corvid 2 (0.4) 17 (3.8) Small mammals mouse 50 (11.0) 20 (4.4) squirrel 33 (7.3) 14 (3.1) mustelid sp. 1 (0.2) 1 (0.2) unknown small mammal 76 (16.7) 17 (3.8) Other unknown small bird 1 (0.2) 10 (2.2) insect — — 2 (0.4) unknown predator 200 (44.1) 72 (16.0) No predation 70 (15.4) 99 (22.0) Total 454 (100.0) 451 (100.0)

and Steller’s jays separately and all corvid species together), three survey techniques (passive point-count surveys, transect surveys, and point-count surveys with attractant calls) and two data summary techniques (grand mean of per point [or 10 min of transect] detections in a season and maximum per point [or 10 min of transect] average for each survey in a season). We examined six measures of nest predation: average days to predation for all nests, per cent predation for all nests, average days to predation for only nests preyed upon, average days to predation for nests containing eggs, and average days to predation for only egg nests preyed upon. All tests of relationships between corvid abundance and nest pre- dation were one-tailed.

RESULTS Artificial nest experiments.—A variety of predators were identified using tooth and bill imprints on wax and later confirmed by photography. Pred- ators of eggs were more often identified than predators of nestlings (80% vs. 48%) because wax was on the outside of eggs, but beneath the skin of chicks. Predators were typically identified to genus using imprints (Ta- ble 3) but to species using photography (Table 4). Predation on simulated eggs differed significantly from predation on simulated nestlings. Eggs and nestlings were depredated at different rates (mean time until predation of eggs: 15.1 d, SE ϭ 0.51, n ϭ 451; mean time until predation of chicks: 12.7 d, SE ϭ 0.47, n ϭ 454; t ϭ 3.5, P Ͻ 0.01) and by different predators (eggs: 81% avian vs. 19% mammalian Vol. 72, No. 4 Counting Corvids [563

TABLE 4. Numbers of animals photographed at 48 artificial nests containing eggs and nes- tlings. Numbers in parentheses are percentages of column total.

Nest contents Nest- Animal ling (%) Egg (%) Corvids Steller’s Jay 4 (14.3) 5 (25.0) Gray Jay 1 (3.6) 3 (15.0) American Crow — — 1 (5.0) Common Raven — — — — Small mammals Forest deer mouse (Peromyscus keeni) 4 (14.3) 1 (5.0) Northern flying squirrel (Glaucomys sabrinus) 10 (35.7) 4 (20.0) Douglas squirrel (Tamiasciurus douglasii) 2 (7.1) 1 (5.0) Townsends chipmunk (Eutamias townsendii) 1 (3.6) 1 (5.0) Other Chestnut-backed Chickadee (Parus rufescens) 6 (21.4) 2 (10.0) (Picoides villosus) — — 1 (5.0) Carrion beetle — — 1 (5.0) Total 28 (100.0) 20 (100.0) predators, n ϭ 278; nestlings: 13% avian vs. 87% mammalian predators, 2 n ϭ 184; ␹ 1 ϭ 207.7, P Ͻ 0.01). The frequent occurrence of predation at artificial nests is slightly high- er than observed at real murrelet nests. In the most comprehensive study to date, the fates of 21 murrelet nests were determined in British Colum- bia from 1994–1997 (Manley 1999). Of these, seven fledged young and 14 (66.7%) failed due to predation (12 eggs were depredated, one adult was depredated, and one nestling was found dead below the nest). This predation rate is not significantly different from the 81% we documented at artificial nests (Fisher’s exact test, P ϭ 0.11). The fates of 71 real murre- let nests are known in total for all of North America. Only one third of these nests fledged young, and it is estimated that between 38% (n ϭ 27) and 59% (n ϭ 42) of nests were depredated (I. Manley and S. K. Nelson, pers. comm.). Factors that affect corvid surveys.—We tested the effects of three factors on corvid surveys: wind, precipitation, and the use of attractant calls. Corvid detections decreased slightly with increasing wind, but this de- crease was not significant during winds under 19 kph (Fig. 1; F2,3308 ϭ 3.00, P ϭ 0.11). We did not survey on days with heavy rain because it was obvious that we detected fewer corvids on such days. However, point- counts conducted during even steady, light precipitation had significantly fewer detections than point-counts with no rain or, at most, intermittent light rain (Fig. 2; F4,3307 ϭ 5.68, P Ͻ 0.01). Though constrained to qual- itative examination due to a lack of independence between time periods, the time of day that surveys were conducted appeared to have little effect 564] J. M. Luginbuhl et al. J. Field Ornithol. Autumn 2001

FIGURE 1. Association of wind and corvid detections (mean number per 10-min point- count). Lines represent SE for total corvids (sample sizes shown above bars).

FIGURE 2. Association of precipitation and corvid detections (mean number per 10-min point-count). Lines represent SE for total corvids (sample sizes shown above bars). Vol. 72, No. 4 Counting Corvids [565

FIGURE 3. Variation in corvid detections (mean number per 10-min point-count) by time of day. on the total number of birds detected. In considering the bias toward greater detections early in surveys (06 00–07 59, the interval during which 67.4% of surveys started; Fig. 3), no strong pattern is evident in either the total number of corvids detected per hour, or in the relative propor- tion by species (08 00–13 59 h). Overall corvid detection rates were low (x¯ ϭ 1.37 Ϯ 0.12 corvids per 10 min, n ϭ 3308 point counts). We assessed the effect of attractant calls by comparing the average number of birds counted at points where calls were used to paired control-points where calls were not used. The use of attractant calls increased our detections for American Crows (average in- crease of 0.24 birds per point; t57 ϭ 2.04, P ϭ 0.05) and Gray Jays (average increase of 0.43 birds per point; t57 ϭ 2.11, P ϭ 0.04) but not for Steller’s Jays (t57 ϭ 0.58, P ϭ 0.34) or Common Ravens (t57 ϭ 0.16, P ϭ 0.60). Predicting nest predation rate from corvid abundance.—Corvid numbers were poorly correlated with the rate of predation within each forested plot. We investigated 210 correlative relationships between measures of corvid abundance and nest predation. The strengths of these correlations were not influenced by survey methods or data summary methods. Cor- relation strength did not differ when we used data from point-counts (90 relationships, r from Ϫ0.28 to 0.23, five with P Ͻ 0.05), transects (60 relationships, r from Ϫ0.28 to 0.14, one with P Ͻ 0.05), or point counts and transects (60 relationships, r from Ϫ0.29 to 0.19, four with P Ͻ 0.05) to survey corvids. Likewise, correlations remained weak whether we com- 566] J. M. Luginbuhl et al. J. Field Ornithol. Autumn 2001 bined multiple surveys per plot by averaging (90 relationships, r from Ϫ0.18 to 0.14, one with P Ͻ 0.05), or by selecting the maximum count per species per survey (120 relationships, r from Ϫ0.28 to 0.24, nine with P Ͻ 0.05). Analyzing each corvid species individually or grouping them also did not affect correlations (210 relationships, r from Ϫ0.28 to 0.24, 10 with P Ͻ 0.05). Including birds detected using attractants (180 rela- tionships, r from Ϫ0.28 to 0.19, five with P Ͻ 0.05) versus deleting them (30 relationships, r from Ϫ0.14 to 0.07, five with P Ͻ 0.05) did not affect the strength of the correlations. Taking these results in total, we found little evidence for a linear relationship between corvid abundance and predation within each plot because only 10 of the 210 correlations were significant at ␣ϭ0.05, which is as expected by chance alone. To examine the relationship between the number of corvid detections and nest predation rate at the landscape-level, we chose the maximum per point average for all corvid species together (of per plot averages calculated from each survey in a given plot in a given year) to characterize the abundance of corvids (because corvids resident in a plot may not be present every day due to their large home ranges) and the days to pre- dation for nests with egg mimics as the measure of predation rate (be- cause corvid predation was focused on nests with eggs; see Table 2 for sampling distribution of nests with eggs). This approach also made screen- ing our data for counts during high winds (over 19 kph) and heavy pre- cipitation (greater than light, intermittent rain) unnecessary since the greater average corvid counts per species were never detected during surveys in these weather conditions. The weak correlation between these two measures was typical of all plot-level comparisons (Fig. 4; r ϭϪ0.18, n ϭ 112, P ϭ 0.17). In contrast to the weak correlations between corvid abundance and predation at the plot-level, the average corvid abundance and average rate of predation for various types of landscapes were closely correlated. We attained average values for landscapes based on the apriori classifi- cation of plots into 12 landscape categories in our study design. Survival of eggs in nests declined significantly (R 2 ϭ 0.69, n ϭ 12, P Ͻ 0.01) as total corvid abundance increased within these 12 types of landscapes (Fig. 5). This relationship held only for corvid abundance measures attained using attractant calls. When corvids detected with attractants were re- moved from analysis, the relationship between abundance and predation was weak (R 2 ϭ 0.23, n ϭ 12, P ϭ 0.12).

DISCUSSION Corvid abundance and rate of predation.—Angelstam (1986) and Andren (1992) both found a clear positive correlation between corvid abundance and corvid predation on artificial ground nests in forested landscapes interspersed with varying amounts of agricultural development in central Sweden. In contrast, Gooch et al. (1991) found no relationship between Black-billed ( pica) densities and overall nest success for 15 songbird species throughout England. Johnson et al. (1989) found a pos- Vol. 72, No. 4 Counting Corvids [567

FIGURE 4. Plot-level relationship between corvid abundance and risk of nest predation. Each point represents the maximum per point average of all corvids detected versus average days to predation for nests containing eggs for each study plot. Line represents least squares regression. Symbols are coded based on the apriori design classification of study plots by stand structure (circle ϭ simple, triangle ϭ complex, square ϭ very complex), proximity to human activity (near ϭ closed symbol, far ϭ open symbol) and level of fragmentation (ϩ indicates fragmented, otherwise contiguous). itive correlation between abundance of American Crows and nest pre- dation on ground-nesting ducks in Canada, but they found no similar relationship for Black-billed . Comparing the results of these studies is difficult because of the lack of consistency in methodology in how both corvid abundance and nest predation rate were determined. Corvid abundance was determined by line transect (Andren 1992), point-counts and line transects ( Johnson et al. 1989), spot-mapping (Gooch et al. 1991), and through incidental en- counters during ‘‘general field work’’ (Angelstam 1986). Although all 568] J. M. Luginbuhl et al. J. Field Ornithol. Autumn 2001

FIGURE 5. Landscape-level relationship between corvid abundance and risk of nest preda- tion. Each point represents the maximum per point average of all corvids detected versus average days to predation for nests containing eggs for plots grouped by apriori landscape category. Line represents least squares regression. Symbols are coded by stand structure (circle ϭ simple, triangle ϭ complex, square ϭ very complex), proximity to human activity (near ϭ closed symbol, far ϭ open symbol) and level of fragmentation (ϩ indicates fragmented, otherwise contiguous; see Methods for detailed description of stand and landscape classifications). Vol. 72, No. 4 Counting Corvids [569 studies focused on ground nests, methods for examining predation varied from monitoring real nests ( Johnson et al. 1989; Gooch et al. 1991) to the use of artificial nests (Angelstam 1986; Andren 1992). A potentially more important problem with comparisons between these studies is the variation in spatial scale at which the relationship between predators and predation rates was investigated. Areas over which corvid abundance and predation rate were compared ranged in size from 0.65 km2 ( Johnson et al. 1989) to over 40,000 km2 (Gooch et al. 1991). In our study, high variability in both our measures of nest predation rate and corvid abundance obscured the relationship between these metrics when examined at the scale of a single forested plot. Averaging many repeated measures of predation rate and corvid abundance in plots of similar struc- ture within similar landscapes (comparable levels of forest fragmentation and human use) reduced this variability and exposed a strong correlation. Assessments of corvid abundance are therefore most likely indicative of egg survivorship when plots are used as replicated measures of landscape conditions rather than as independent samples. Our results indicate that using measurements of corvid abundance to assess nest predation risk is not possible at the typical scale of homogenous plots (0.5–1.0 km2 in our study). Rather, this approach should be considered useful only at a broad- er, landscape scale on the order of 5–50 km2 (based on the scale of our fragmentation and human-use measures). We suggest that the relationship between corvid abundance and nest predation is scale-sensitive. The spatial scale at which multi-species inter- actions are examined must be relevant to the scale at which these species use, and are affected by, their environment (May 1994; Bissonette 1997). Corvids are habitat generalists with relatively large home range sizes vary- ing from an average of 0.1 km2 for Black-billed Magpies (Møller 1982) to 14–18 km2 for Common Ravens and American Crows ( J. Marzluff et al. unpubl. data). Recommended corvid survey protocol.—Of the techniques we examined, our modified point-counts produced measures of corvid abundance with the strongest correlation to predation in the dense, mature forests of the Pacific Northwest. Our experience surveying corvids suggests that delin- eating and mapping the survey area is a critical first step to accurately assess relative abundance. Plots, not survey stations, are the sampling units. While our point-count stations were placed using a 250 m grid, with no points placed closer than 50 m from plot edge (to maximize coverage while minimizing detections from outside the plot), other, perhaps wider spacings may also be suitable. To maximize corvid detections and thereby increase chances of documenting all relevant predators, surveys should be conducted on days with light winds (Ͻ19 kph) and no more than light precipitation. Timing of surveys between sunrise and 1400 appears to have little effect on corvid detections. Use of attractant calls further aids in detecting corvids that can be wide ranging (American Crows) and often non-vocal (Gray Jays). Although the use of attractants may overre- present corvids at the plot level (by attracting birds into the plot from 570] J. M. Luginbuhl et al. J. Field Ornithol. Autumn 2001 the surrounding area), they are important in providing predator counts that reflect predation risk at the landscape scale. To calculate the relative risk of nest predation at the landscape scale, plots must be categorized based on similarities in plot and landscape characteristics. We suspected that forest structure (age and uniformity of the overstory) was important at the plot level and that degree of forest fragmentation and human activity that provides supplemental food for corvids (settlements, recreation areas) at distances up to 5 km from the plot were important at the landscape-level. This categorization did in- crease our ability to relate corvid abundance to rate of predation (Fig. 5). However, only forest structure was consistently associated with corvid abundance and rate of predation (simple structured forests had fewer corvids and lower predation rate; Marzluff et al. 2000; Raphael et al. in press). Fragmentation and proximity to human use were related to corvid abundance and predation rate in a complex fashion explored elsewhere (Marzluff et al. 2000; Raphael et al. in press; Luginbuhl et al., unpubl. data). Similar metrics that might influence corvid populations should be anticipated so that counts of corvids on plots can be averaged within landscapes to produce useful indices of relative risk of nests to predation by corvids.

ACKNOWLEDGMENTS This project was a cooperative venture funded by the Washington State Department of Natural Resources, the U.S. Forest Service, the U.S. Fish and Wildlife Service, Rayonier, the Olympic Natural Resources Center, the National Council for Air and Stream Improvement, the National Park Service, Boise Cascade Corporation, Willamette Industries, and the Oregon Department of Forestry. S. P. Courtney, L. Young, J. Engbring, and S. Horton were instru- mental in obtaining funding and guiding the initial development of the study design. S. Youngs, M. Vekasy, E. Neatherlin, L. Everette, N. MacRury, T. Hirzel, J. Carlson, M. Philippart, A. Agness, R. Jaffe, D. Rossman, K. Bork, C. Christy, L. Williams, and C. Apodaca assisted with field experiments and observations. D. Evans and P. Brewster conducted GIS analyses at the USFS PNW Olympia Laboratory. S. Hall, E. Seaman, and B. Rohde facilitated our research in Olympic National Park. D. Collins provided the canopy cover base map and harvest detection data. R. Perry supplied digital ortho quads. B. Galleher prepared habitat use maps and aided with GIS analyses. K. Tarvin, R. Chandler, and three anonymous review- ers provided helpful comments on the manuscript.

LITERATURE CITED

ANDREN, H. 1992. Corvid density and nest predation in relation to forest fragmentation: a landscape perspective. Ecology 73:794–804. ANGELSTAM, P. 1986. Predation on ground-nesting birds’ nests in relation to predator den- sities and habitat edge. Oikos 47:365–373. BISSONETTE, J. A. 1997. Scale sensitive ecological properties: historical context, current mean- ing. Pp. 3–31, in J. A. Bissonette, ed. Wildlife and landscape ecology: effects of pattern and scale. Springer-Verlag, New York. DARVEAU, M., L. BELANGER,J.HUOT,E.MELANCON, AND S. DEBELLEFEUILLE. 1997. Forestry practices and the risk of predation in a boreal coniferous forest. Ecol. Appl. 7:572–580. DE SANTO, T. L., AND S. K. NELSON. 1995. Comparative reproductive ecology of the auks (family Alcidae) with emphasis on the Marbled Murrelet. Pp. 33–48, in C. J. Ralph, G. L. Hunt, Jr., M. G. Raphael, and J. F. Piatt, eds. The ecology and conservation of the Vol. 72, No. 4 Counting Corvids [571

Marbled Murrelet in North America: an interagency scientific evaluation. United States Department Agriculture Forest Service General Technical Report PSW-GTR-152. FRAISSINET, M. 1989. Espansione della taccola, Corvus monedula, nei capoluoghi Italiani. Riv. Ital. Orn. Milano 59:33–42. GOOCH, S., S. R., BAILLIE, AND T. R. BIRKHEAD. 1991. Magpie, Pica pica, and songbird pop- ulations. Retrospective investigation of trends in population density and breeding suc- cess. J. Appl. Ecol. 28:1068–1086. HAMER,T.,AND S. K. NELSON. 1995. Characteristics of Marbled Murrelet nest trees and nesting stands. Pp. 49–56, in C. J. Ralph, G. L. Hunt, Jr., M. G. Raphael, and J. F. Piatt, eds. The ecology and conservation of the Marbled Murrelet in North America: an in- teragency scientific evaluation. United States Department of Agriculture Forest Service General Technical Report PSW-GTR-152. HASKELL, D. G. 1995. A reevaluation of the effects of forest fragmentation on rates of bird- nest predation. Cons. Biol. 9:1316–1318. HEINRICH, B., J. M. MARZLUFF, AND B. ADAMS. 1995. Fear and food recognition in Common Ravens. Auk 112:499–503. HERNANDEZ, F., D. ROLLINS, AND R. CANTU. 1997. An evaluation of Trailmaster(r) camera systems for identifying ground-nest predators. Wildl. Soc. Bull. 25:848–853. HORTON, S., AND P. HARRISON. 1997. Olympic experimental stare forest Marbled Murrelet surveys: 1997 annual report. Washington Department of Natural Resources, Forks, Wash- ington. JOHNSON, D. H., A. B. SARGEANT, AND R. J. GREENWOOD. 1989. Importance of individual species of predators on nesting success of ducks in the Canadian Prairie Potholes Region. Can. J. Zool. 67:291–297. KONSTANTINOV, V. M. 1996. Anthropogenic transformations of bird communities in the forest zone of the Russian Plain. Acta Ornithologica 31:53–65. MAJOR, R. E. 1991. Identification of nest predators by photography, dummy eggs, and ad- hesive tape. Auk 108:190–195. ,G.H.PYKE,M.T.CHRISTY,G.GOWING, AND R. S. HILL. 1994. Can nest predation explain the timing of the breeding season and the pattern of nest dispersion of New Holland honeyeaters? Oikos 69:364–372. MANLEY, I. A. 1999. Behaviour and habitat selection of Marbled Murrelets nesting on the Sunshine Coast. M.Sc. thesis. Simon Fraser University, Burnaby, British Columbia. MARSHALL, D. B. 1988. Status of the Marbled Murrelet in North America, with special em- phasis on populations in California, Oregon and Washington. United States Fish and Wildlife Service, Biological Report 88(30). MARTIN, T. E. 1993a. Nest predation among vegetation layers and habitat types: revising the dogmas. Am. Nat. 141:897–913. . 1993b. Nest predation and nest sites: new perspectives on old patterns. BioScience 43:523–532. MARZLUFF,J.M.AND R. P. BALDA. 1992. The . T & A D Poyser, London. ,R.B.BOONE, AND G. W. COX. 1994. Native pest bird species in the West: why have they succeeded where so many have failed? Stud. Avian Biol. 15:202–220. ,M.G.RAPHAEL, AND R. SALLABANKS. 2000. Understanding the effects of forest man- agement on avian species Wildl. Soc. Bull. 28:1132–1143. MAY, R. M. 1994. The effects of spatial scale on ecological questions and answers. Pp. 1–7, in P. J. Edwards, R. M. May and N. R. Webb, eds. Large scale ecology and conservation biology. Blackwell, London. MILLER, G. S., S. R. BESSINGER,H.R.CARTER,B.CSUTI,T.E.HAMER, AND D. A. PERRY. 1997. Recovery plan for the Marbled Murrelet in Washington, Oregon, and California. United States Department of the Interior, Fish and Wildlife Service, Portland, Oregon. . 1982. Characteristics of Magpie Pica pica territories of varying duration. Ornis Scand. 13:94–100. MøLLER, A. P. 1987. Egg predation as a selective factor for nest design: an experiment. Oikos 50:91–94. NELSON,S.K.,AND T. HAMER. 1995a. Nest success and the effects of predation on Marbled Murrelets. Pp. 89–98, in C. J. Ralph, G. L. Hunt, Jr., M. G. Raphael, and J. F. Piatt, eds. 572] J. M. Luginbuhl et al. J. Field Ornithol. Autumn 2001

The ecology and conservation of the Marbled Murrelet in North America: an inter- agency scientific evaluation. United States Department of Agriculture Forest Service General Technical Report PSW-GTR-152. , AND . 1995b. Nesting biology and behavior of the Marbled Murrelet. Pp. 57– 68, in C. J. Ralph, G. L. Hunt, Jr., M. G. Raphael, and J. F. Piatt, eds. The ecology and conservation of the Marbled Murrelet in North America: an interagency scientific eval- uation. United States Department of Agriculture Forest Service General Technical Re- port PSW-GTR-152. NICE, M. M. 1957. Nesting success in altricial birds. Auk 74:305–321. PERRY, D. R. 1978. A method to access into the crowns of emergent and canopy trees. Bio- tropica 10:155–157. RALPH, C. J., S. DROEGE, AND J. R. SAUER. 1995a. Monitoring bird populations by point counts. United States Department of Agriculture Forest Service General Technical Report PSW- GTR-149. ,G.R.GUEPEL,P.PYLE,T.E.MARTIN, AND D. F. DESANTE. 1993. Handbook of field methods for monitoring landbirds. United States Department of Agriculture Forest Ser- vice General Technical Report PSW-GTR-144. ,G.L.HUNT,JR., M. G. RAPHAEL, AND J. F. PIATT. 1995b. Ecology and conservation of the Marbled Murrelet in North America: an overview. Pp. 3–22, in C. J. Ralph, G. L. Hunt, Jr., M. G. Raphael, and J. F. Piatt, eds. The ecology and conservation of the Marbled Murrelet in North America: an interagency scientific evaluation. United States Department of Agriculture Forest Service General Technical Report PSW-GTR-152. RAPHAEL, M. G., D. M. EVANS,J.M.MARZLUFF, AND J. M. LUGINBUHL. In press. The potential effects of forest fragmentation on populations of the Marbled Murrelet. Stud. Avian Biol. RATTI,J.T.,AND K. P. REESE. 1988. Preliminary test of the ecological trap hypothesis. J. Wildl. Manage. 52:484–491. RICKLEFS, R. E. 1969. An analysis of nesting mortality in birds. Smithsonian Contr. Zool. 9: 1–48. SIEVING, K. E. 1992. Nest predation and differential insular extinction among selected forest birds of central Panama. Ecology 73:2310–2328. SINGER, S. W., N. L. NASLUND,S.A.SINGER, AND C. J. RALPH. 1991. Discovery and observations of two tree nests of the Marbled Murrelet. Condor 93:330–339. SPSS. 1997. Statistics. Version 8.0. SPSS Inc., Chicago. U.S. NATIONAL WEATHER SERVICE. 1995. Federal meteorological handbook: surface weather observations and reports. FCM-H1-1995. VAROUJEAN, D. H., II, AND W. A. WILLIAMS. 1995. Abundance and distribution of Marbled Murrelets in Oregon and Washington based on aerial surveys. Pp. 339–352, in C. J. Ralph, G. L. Hunt, Jr., M. G. Raphael, and J. F. Piatt, eds. The ecology and conservation of the Marbled Murrelet in North America: an interagency scientific evaluation. United States Department of Agriculture Forest Service General Technical Report PSW-GTR- 152. VERNER, J. 1985. Assessment of counting techniques. Curr. Orinithol. 2:247–302. WHELAN, C. J., M. L. DILGER,D.ROBSON,N.HALLYN, AND S. DILGER. 1994. Effects of olfactory cues on artificial-nest experiments. Auk 111:945–952. WILCOVE, D. S. 1985. Nest predation in forest tracts and the decline of migratory songbirds. Ecology 66:1211–1214. WILLEBRAND,T.,AND V. MARCSTROM. 1988. On the danger of using dummy nests to study predation. Auk 105:378–379. Received 7 July 1999; accepted 19 January 2001.