THE CONDOR

A JOURNAL OF AVIAN BIOLOGY

Volume 99 Number 3

The Condor 9915677584 0 The Cooper Omithologmd Society 1997

THE INFLUENCE OF , PREY ABUNDANCE, SEX, AND BREEDING SUCCESS ON THE RANGING BEHAVIOR OF ’

JOHN M. MARZLUFF~, BRYAN A. KIMSEY, LINDA S. SCHUECK, MARY E. MCFADZEN, MARK S. VEKASY AND JAMES C. BEDNARZ~ Greenfalk, 8300 Gantz Avenue, Boise, ID 83709

Abstract. We studied the ranging behavior and habitat selection of radio-tagged Prairie Falcons (F&o mexicanus) during the breeding seasonin southwesternIdaho. The distri- bution and numbersof Townsend’s ground squirrels(Spermophilus townsendii), the primary prey of Prairie Falcons in our study area, varied in response to drought during the study period. Prairie Falcons ranged over large areas (ca. 300 km*) and increasedtheir foraging ranges in responseto declining ground squirrels.Reptiles and were preyed upon most frequently when squirrelswere rare. Males and females differed little in their use of space. Successfulpairs ranged over smaller areasthan non-nestersand unsuccessfulpairs. Falcons nesting near habitat most suitable for ground squirrelsranged over smaller areasthan those nesting farther from such habitat. Home ranges containedsignificantly more winterfat (Cer- atoides lanata) and native perennial grasses(especially secunda), and significantly less salt desert shrubsand exotic annual grassesthan expected based on availability. Salt desert shrubswere found less than expected,based on availability in core areaswithin home ranges. Selection for winterfat and bluegrass in core areas was contingent upon selection at the larger scale of the home range; falcons with-home ranges containing more winterfat and bluegrassthan expected based on availability were less selective in their placement of core areaswith respectto thesehabitats. We believe salient featuresof Prairie home ranges result largely from patchy distribution of landscape features associatedwith different den- sities and availabilities of Townsend’s ground squirrels. Key words: Prairie Falcon, Townsend ground squirrel, home range, habitat selection, radio telemetry, conservation.

INTRODUCTION requirements (Hunt 1993, Squires et al. 1993), Prairie Falcons (F&o mexicanus) are one of the wintering behavior (Beauvais et al. 1992), nest- most common raptors of montane desert and ing behavior (Kaiser 1986, Holthuijzen 1990), shrub of western . Nu- and productivity (Ogden and Hornocker 1977, merous studies have documented their nest site Allen et al. 1986). Despite this research, our characteristics (Runde and Anderson 1986, Al- knowledge of Prairie Falcon ranging habits and len 1987), abundance (Platt 1974, 1981), forag- habitat use remains limited because these birds ing behavior (Phipps 1979, Haak 1982), habitat range over large, remote areas where observing them is difficult. Some of this difficulty can be overcome by the use of radio-telemetry, but pre- ’ Received 29 August 1996. Accepted 4 April 1997. vious work involved relatively small sample * Current address:College of Forest Resources,Uni- sizes of radio-tagged Prairie Falcons (5 14; versity of , Seattle, WA 98195-2100, Dunstan et al. 1978, Hunt 1993). e-mail: [email protected] 3 Current address: Department of Biological Sci- The most intensive studies of Prairie Falcons ences, ArkansasState University, State University, AR have been conducted in the Snake River Birds 12467. of Prey National Conservation Area (NCA) of

[5671 568 JOHN M. MARZLUFF ET AL. southwestern , where abundance and pro- ful” if 21 chick reached an estimated age of 30 ductivity have been measured continuously for days (80% of actual age to fledging, Steenhof 21 years. The NCA has the densest concentra- 1987). We rappelled to nests when chicks were tion of nesting Prairie Falcons in the world; ap- at least 30 days old to obtain brood counts, proximately 200 pairs nest on basalt cliffs that which reliably discriminate successful from border 100 km of the Snake River. Falcons spe- failed breeders (Marzluff and McFadzen 1996). cialize on Townsend’s ground squirrels (Sper- We did not always know if falcons using a nest- mophilus townsendii) to such an extent that their ing area successfully laid eggs. We therefore es- arrival on, and departure from, nesting areas is tablished three categoriesfor analyses: (1) “Un- closely correlated with ground squirrel emer- successful pairs” were pairs found throughout gence and estivation, and their productivity is the breeding seasonin a previously documented closely correlated with indices of squirrel abun- nesting area that did not produce 30-day-old dance (U.S. Dept. Interior 1979, Steenhof et al., young, (2) “Non-nesters” were nonbreeding in- unpubl. data). dividuals not associated with a previously doc- In this paper we investigate behavioral link- umented nesting area, and (3) “Successful ages between Prairie Falcons and Townsend’s pairs” produced 30-day-old young. For some ground squirrels that help explain why ground analyses, unsuccessful pairs and nonbreeders squirrel population dynamics influence Prairie were lumped for comparison with successful Falcons. We measured the behavior of falcons pairs. In all analyses of travel distance, breeding during four years of varying squirrel abundance pairs were classified as successfuluntil the time to determine: (1) the degree to which falcons they lost their clutch or brood. relied on ground squirrels for prey, (2) whether the degree of reliance on ground squirrels influ- DIETS enced ranging behavior and habitat use, (3) how We observed a sample of nests each year to doc- sex and breeding success influenced the rela- ument falcon reliance on Townsend’s ground tionships between prey abundance and ranging squirrels. From 1991 through 1994, we observed behavior, and to (4) provide managers with a prey delivered to breeding females and nestlings prescription for optimal falcon habitat. at 62 nests during 235 days between late April and mid-June. Nests were constantly observed METHODS from 20 min before sunrise to 15 min after sun- STUDY AREA set by two observers, one in the morning and We studied Prairie Falcons in a 198,616 ha por- another in the afternoon. Observations were tion of the NCA selected for an investigation of made with 10 X 50 binoculars and 15-45X the effects of military training on falcons. It in- spotting scopes from blinds or vehicles at dis- cluded relatively homogenous areas east and tances between 70 and 300 m (see Holthuijzen west of a military training area and extended 1990). Observers were trained before making north enough to encompass foraging ranges of observations and were rotated among nest sites falcons (U.S. Dept. Interior 1996). The terrain to minimize observer bias. We classified prey and vegetation were dominated by shrub (Arte- delivered by species or class (mammal, , misia tridentata, Chrysothamnus nauseosus, reptile, insect) when possible. Only fresh items Ceratoides lanata, and Atriplex confertifolia) (blood usually visible, prey supple and not de- and grassland (Poa secunda and Bromus tecto- hydrated) were included in counts of deliveries rum) flats punctuated by buttes and rolling hills to reduce multiple counting of cached items. surrounding the Snake River Canyon. Falcons nested in the Snake River Canyon and hunted RADIOTELEMETRY surrounding plains to the north and east. Selection and capture of falcons. The Prairie Falcons radio-tagged for this study were ran- TERMINOLOGY domly selected from approximately 200 pairs “Nesting area” is a stretch of cliff where nests breeding in the 1 lo-km stretch of the NCA from are found year after year, but where no more Walters Ferry to Bruneau. This area was strati- than one pair has ever bred in one year Falcons fied based on position of nesting areas relative nest in potholes, cracks, and ledges on cliffs in to a military training area into a west, central, nesting areas. A nesting attempt was “success- and east region (Fig. 1). We attempted to ra- PRAIRIE FALCON RANGING BEHAVIOR 569

Study Area Radio Tracking Zones 20 - m Contour Intervals Receiver Sites

FIGURE 1. Topographyand locationof studyarea. Receiver sites where radio trackingwas conductedand generalizedtracking zones are indicated.Bold lines from river to northeasternborder of studyarea delineate threesampling strata (west, central, east). dio-tag equal numbers of falcons in the west and in 1992,36 in 1993, and 31 in 1994. These sam- central strata, and approximately equal numbers ples are independent as no individuals were ra- of males and females each year. Falcons from dio-tagged in more than one year. the east stratum were radio-tagged only in 1992 Two 2-person teams trapped each year from (1 falcon) and 1993 (5 falcons). March to May. They observed each nesting area We investigated 68 randomly selectednesting until a falcon exhibited signs of occupying the areas for the possibility of trapping Prairie Fal- nesting area (perching, courting, copulating, or cons in 1991, 52 in 1992, 86 in 1993, and 76 in defending the area), and then placed traps as 1994. We rejected nesting areas if: (1) adults close as possible to the cliff containing the nest. were not present or not exhibiting territorial be- To avoid trap placements dangerousto personnel havior (n = 142), (2) sites were heavily dis- and possibly disruptive to falcons (Clugston turbed by human recreational activities (n = l), 1990), we captured falcons away from the nest (3) sensitive raptors, primarily Ferruginous using noose harnesses or dho-gazas (Bloom Hawks (Buteo regalis), would be disturbed by 1987). We lured falcons to the traps with live trapping (n = lo), (4) trap placement sites were Rock Doves (Columba livia), European Star- unavailable (n = 4), or (5) falcons were not re- lings (Stumus vulgaris), House Sparrows (Pass- sponsive to the trapping methods (n = 30). We er domesticus), or Great Homed Owls (Bubo trapped during courtship, egg-laying, incubation, virginianus). and brooding, and captured 34 adult Prairie Fal- When an individual was captured, it was cons in 1991, 37 in 1992,40 in 1993, and 41 in hooded immediately (Bloom 1987). We placed 1994. We radio-tagged 28 of these in 1991, 34 a radio transmitter (12-16 g with Teflon@ straps 570 JOHN M. MARZLUFF ET AL. and leather sternum patch) on each captured fal- falcons were tracked relatively evenly through- con that we believed resided in the preselected out the season and through the daylight hours. nesting area where we were trapping, unless we Tracking sessions were suspended when light- felt eggs in the abdomen (1992-1994), or the ning or heavy rain threatened personnel or re- individual had received a transmitter in a pre- ceiving equipment. vious year (1992-1994). The general configu- Radio-tracking teams used 4-element, Yagi ration of the harness followed Buehler et al. antennas and programmable scanning receivers (1995), with the modification of a leather ster- (Advanced Telemetry Systems, Isanti, MN) to num patch added to distribute the pull of the sample sequentially for radio frequencies of in- backpack evenly acrossthe breast (Vekasy et al. strumented falcons. Antennas were hand-held or 1996). We also weighed, measured, and banded placed on 2-7-m towers to increase the recep- each bird. Sex was determined by wing chord tion range. Bearings were obtained by sighting length (U.S. Dept. Interior 1977). hand-held compassestoward signals in 1991 and We monitored each falcon for at least 1 hr 1992. In 1993 and 1994, compassrosettes bolted after release to determine if the transmitter pack- to towers enabled us to eliminate the use of age and/or handling adversely affected the bird’s hand-held compassesand increased the consis- flight or behavior. Each bird was monitored tency of orientation from each site. Null-peak again for 2 hr (in most cases on the next day) antenna arrays were not used because they did to ensure that the transmitter was still on and not increase bearing accuracy (Clugston 1990). that the bird’s behavior appeared normal. Com- Trackers used 2-way radios to alert other team parisons of radio-tagged birds to untagged, con- members of the presence of a falcon in a zone. trol birds indicated no significant effects of tag- When 2 3 trackers had a bird’s signal, they took ging on productivity or behavior (Vekasy et al. simultaneous bearings to the signal. All bearings 1996). to falcons were entered into laptop computersto Tracking protocol. We radio-tracked falcons determine the error associated with a given tri- from fixed sites in seven zones ranging in size angulation and to identify systematic errors re- from 50 to 80 km* that covered foraging areas sulting from misaligned compasses,electrical in- on benchlands adjacent to canyon nesting areas. terference, etc. We repeated triangulation at- After conducting a series of tests on transmitters tempts on each bird until either a 95% confi- placed throughout the study area (beacons) dur- dence ellipse around the location was < 1,000 ing winter 1992, we further divided the study ha, the bird was deemed out of range, or the area into eight zones (70-120 km2) which were bird’s signal was detected by < 3 trackers. We small enough to allow detection from 4-6 fixed allowed at least 30 min between successivelo- receiver sites of most radio-tagged falcons using cation estimates with ellipses < 1,000 ha on the the zone (Fig. 1). Receiver sites were located same bird to reduce dependency among esti- with a global positioning system (GPS) accurate mates. Although multiple estimates of the same to within 5 m. Most receiver sites were on prom- bird’s locations are not truly independent, loca- inent buttes, ridges, and outcroppings, but some tion estimates 30 min apart should be represen- were on flat terrain (Fig. 1). Receiver sites were tative of a falcon’s use of the study area (White typically > 6 km from the falcons being tracked and Garrott 1990). Furthermore, these locations and were positioned in a zone to allow the tak- should not include diurnal or seasonal bias in ing of simultaneous bearings that minimized tri- home range estimation becausethey were evenly angulation error (White and Garrott 1990). distributed throughout the study period (Ander- Teams of four to six personsradio-tracked fal- sen and Rongstad 1989). cons in each zone from approximately 15 The design of our zones may bias home range April-15 July each year from 1991 to 1994. estimation because falcons could not be tracked Each zone was sampled in random order with while they were at their nests below the canyon the constraint that each be sampled once in the rim (Marzluff et al. 1994). However, our goal morning (30 min before sunrise until 13:30) and was to define ranges and habitat use while for- once in the afternoon (13:30 until approximately aging, therefore activities at the nest were less sunset) every 14-16 days. Two zones usually important. Radio-tracking in each zone twice ev- were sampled each day (one in the morning and ery two weeks (once in the morning and once one in the afternoon). Each year all radio-tagged in the evening) minimized potential biases as- PRAIRIE FALCON RANGING BEHAVIOR 571 sociated with falcons foraging outside of a zone nests. Our definitions of adequately sampled on a particular day. males and females are further justified because Selecting location estimates and falcons for correlations between home range size and the analysis. Subsamples of location estimates for number of location estimates used to calculate individual falcons were selectedfor use in home home range size were very weak (rs’ < 0.20, range analyses based upon their accuracy. Be- Ps’ > 0.40; Marzluff et al. 1992). This implies cause confidence ellipses associated with loca- that less frequently (but “adequately”) sampled tion estimates (Lenth 1981) are weakly correlat- ranges were not unusually small, as one might ed with actual accuracy of an estimate (Marzluff expect if the intensity of sampling influenced de- and Kimsey, unpubl. data), we developed a re- termination of home range size. gression model to estimate accuracy based on beacon tests. The model was verified by pre- HOME RANGE AND RANGING BEHAVIOR dicting accuracy of remote estimates when fal- Home range size, use of area within ranges, and cons were simultaneously observed by mobile travel distance from nests and other centers of ground crews independent of crews at fixed lo- activity were computed with Ranges V (Ken- cations. Using Andrew’s estimator to produce a ward and Hodder 1995). We estimated home point estimate relatively insensitive to signal ranges with harmonic mean methods, convex bounce (Lenth 1981), we accounted for 55% of polygons, and hierarchical incremental cluster variation associatedwith distance between esti- analysis (Dixon and Chapman 1980, Kenward mated and actual locations (linear error) by us- 1987). Nest sites were not used in the delinea- ing the distance from the estimated location to tion of home ranges. For harmonic mean iso- the center of a “tracking region.” The tracking lines, we reduced plotting error by using a 40 X region is a dynamic polygon with vertices 40 grid and location estimates in the center of formed by positions of trackers involved in a the grid squares(Spencer and Barrett 1984). Te- particular location estimate. We used location lemetry location point estimateswere used with- estimates with estimated linear errors smaller out accounting for telemetry error in all calcu- than the upper 75% quartile of the error distri- lations of area. This may lead to slight under- bution in our calculations of home range. This estimates of home ranges and core areas, es- allowed us to objectively censor estimates with pecially using convex polygons, but relative large linear errors while keeping accurate and differences in range size among individuals are precise estimates. not affected (Senchak 1991). The adequacy of radiotelemetry sampling of We used the minimum convex polygon to rep- individual falcons was assessedby relating in- resent the maximum home range size and the crease in home range size to successivelylarger 95% harmonic mean range as a general repre- samples of locations for each falcon (incremen- sentation of the area typically used by falcons. tal analysis; Kenward 1987). Individual males The 95% harmonic mean range was probably with 2 40 location estimates were considered to the best estimator because it excluded infre- be adequately sampled because incremental quently used outlying points, and matched, with analysis indicated that 85% of the area used by only slight distortion, the area in which we ob- males was sampled by these first 40 fixes (Mar- tained location estimates (Squires et al. 1993). zluff et al. 1992). Location estimatesfor females We used minimum convex polygons for esti- were more difficult to obtain because they spent mation of habitat available to falcons within most of their time out of radio contact below the their home range because it described the max- canyon rim until their nestlings were ready to imum area used, minimized inclusion of areas fledge. Incremental analysis indicated that 25 es- where we never located birds, and did not rely timates of a female’s locations defined 75% of on statistical distributions of locations. Cluster the total area used by the female (Marzluff et al. techniques (Kenward 1992) produced the best 1992). We therefore considered females to be representationsof “core areas” (areas of dispro- adequately sampled if they had 2 25 location portionately high use within ranges). estimates with at least 65% taken from the time We investigated effects of year, sex, nest lo- of fledging to post-fledging independence (1 cation, and breeding successon ranging habits month after fledging; McFadzen and Marzluff with multivariate analyses of variance (MAN- 1996) when females ventured farthest from their OVAs). Measures of spatial use were partitioned 572 JOHN M. MARZLUFF ET AL. into three groups for use as dependent variables the number of telemetry locations per km2 with in separate MANOVAs as follows: (1) average the percent cover of shrubs and grassesin each travel distance from the nest to telemetry loca- km2 of the study area. These large grid cells tions and 95% harmonic mean home range as were used because of the error associatedwith indices of average area use, (2) maximum travel telemetry locations. Second, to document selec- distance from the nest to telemetry locations, tivity in the location of home ranges by individ- minimum convex polygon home range, and ual falcons, we compared the proportion of hab- maximum width of home range for maximum itats used in convex polygon home ranges to the area use, and (3) 90% and 95% cluster-based proportion of habitats available within the usual home ranges for core area use. Sample sizes flight range of falcons (the area from 21 km were insufficient to investigate all four factors north of the Snake River Canyon [the average (year, sex, nest location, breeding success) in maximum travel distances observed for radio- any single analysis because there were few un- tagged birds] to 7 km south of the canyon [the successfulpairs in 1992 and few successfulpairs full extent of vegetation sampling south of the in 1993. Therefore, we initially subdivided the canyon]). Third, to document selectivity within sample into successful pairs and unsuccessful the home range by individual falcons, we com- pairs plus non-nesters. Sample size was suffi- pared the proportion of habitats used within core cient for the use of a two-factor (year and sex) areas (defined by cluster analysis) to the pro- MANOVA to analyze successful pairs, but the portion of habitats available within each individ- smaller sample of unsuccessful pairs was ade- ual’s convex home range. We examined habitat quate only for two, single-factor (1 for year, 1 in clusters that included 95% of locations be- for sex) MANOVAs. The influence of breeding cause ranges showed little change in the rate of success was investigated separately for males area increase for cluster polygons that included and females. For all analyses, the nest site uni- from 20% to 95% of the locations, but typically versal transverse mercator (UTM) coordinates increased sharply thereafter both in area within (or trap site coordinates, if no nest was estab- ranges and size variation between ranges, which lished) were included as covariates in the MAN- indicated that the remaining 5% of locations OVAs. were outliers. We combined each adequately sampled, suc- We quantified habitat (vegetation) in two cessfully breeding individual’s travel distances ways. First, we calculated percent cover of shrub to test the influence of sex, year, and stage of and grass sampled at 684 sites from 1990 nesting cycle on travel in a three-factor ANOVA through 1994. Percent cover at each site was es- (nest location was used as a covariate). Four timated by point frame interception (Floyd and stages of the nesting cycle were determined by Anderson 1982) at 252 points randomly distrib- back-dating from estimated age at banding or uted along seven, 50-m transects.Percent cover from the last date of observation before nest fail- of 16 species (Artemisia spinescens,Arte- ure as follows: (1) territory establishment/incu- misia tridentata, Atriplex canescens, Atriplex bation, (2) early brood rearing (nestlings < 21 confertifolia, Atriplex nuttallii, Bromus tecto- days old), (3) late brood rearing (nestlings 2 21 rum, Ceratoides lanata, Chrysothamnusnauseo- days old), and (4) postfledging (the first month sus, Chrysothamnus viscidiflorus, Grayia spi- after fledging). Because the number of locations nosa, Poa secunda, Salsola iberica, Sarcobatus per individual for each stage of the nesting cycle vermiculatus, Sitanion hystrix, Tetradymia gla- was not equal, each location estimate in the brata, Vulpia spp.) per 1 km2 of the study area ANOVA was assumed to be independent. Any was derived by interpolating between sites using pseudoreplication (Hurlbert 1984) in the analy- a GIS-based kriging procedure (technique that sis may increase Type I error, but assessments assignsplant cover for areas not sampled based of the relative importance of each factor should on cover of nearby sampled areas). This mea- be valid. surement of habitat was used to test for selection at all three scales discussedabove. Second, we HABITAT SELECTION determined the types and areas of habitats from We investigated habitat selection at three scales. Landsat thematic mapper satellite imagery. Hab- First, to document general associationsbetween itats were classified on the basis of dominant falcons and underlying vegetation, we correlated vegetation and included sagebrush/rabbitbrush. PRAIRIE FALCON RANGING BEHAVIOR 573 winterfat, salt desert shrub, grassland, cliff, and mental units and Bonferonni-adjusted confi- water habitats (Knick et al. 1997). In addition, dence intervals). Values presented are K 5 SE. we delineated all areas used for agriculture since 1979 (including fallow fields) from a composite RESULTS of the 1979 Snake River Birds of Prey vegeta- DIETS tion map (U.S. Dept. Interior 1979), 1993 Bu- Townsend’s ground squirrels comprised the ma- reau of Reclamation agriculture maps, and the jority of prey utilized by Prairie Falcons each classified satellite imagery. Resolution of the year, but the proportion of prey types delivered habitat map was 50 m X 50 m (resampled from fresh to the territory differed significantly 30-m pixels in the satellite image). We used > among years (G, = 88.6, P < 0.001). Parents 5% ground cover of shrubsto separateshrub and delivered more Townsend’s ground squirrels in grassland classes.Accuracy of the classification 1992 than in 1993 or 1994 (Fig. 2). Conversely, in separating shrub and grassland areas was more reptiles and birds were delivered in 1993 80%; accuracy in separating individual habitat and 1994 than in 1992 (Fig. 2). classes was 64% (Knick et al. 1997). Because falcon locations were least closely correlated RADIO TELEMETRY with this quantification of habitat, we only used As a result of large areas covered by Prairie Fal- it to investigate selection at the first scale dis- cons, we were between 141 m and 14.6 km (f cussed above. = 7.7 km +- 38 m, n = 7,213 locations) from We determined significance of selectivity in them when we obtained acceptable location es- the location of home ranges and location of core timates. Bearings obtained by tracking moving areas within home ranges by calculating selec- falcons across such large distances were rela- tion ratios (proportion of habitat class used/pro- tively imprecise (SD of bearing errors = 16.4”), portion of habitat type available) for each habitat but unbiased (a deviation of estimated bearings type (Manly et al. 1993). We normalized selec- from true bearings = 1.03 O, n = 584 estimates tion ratios by using their natural logarithm. The of beacon locations; not significantly different sampling unit was the individual falcon and av- from 0, paired tss3= 1.52, P = 0.13). Estimated erage selection ratios were calculated for our linear error between location estimates and fal- sample of falcons. We calculated a 95% confi- con positions ranged from O-6.7 km (ff = 3.0 dence interval around each ratio average after a km ? 20 m, n = 7,213). Although the errors are Bonferroni adjustment for multiple comparisons. large (approximately 10% of maximum distance Selection ratios that did not include 0 in their between points in a falcon’s range), their unbi- confidence interval indicated significant (o. = ased nature means that comparisons of relative 0.05) selection for (ratio > 0) or against (ratio differences between groups of birds are mean- < 0) particular habitats. ingful. The error associatedwith telemetry locations may bias tests of habitat selection, especially if HOME RANGE AND RANGING BEHAVIOR error is large relative to habitat patch size (White Average home range size was ca. 300 km*, de- and Garrott 1986, Nams 1989). We reduced this pending upon calculation method (Table 1). Fal- potential bias by basing our assessmentsof hab- cons traveled 7 km on average, and an average itat use on entire home ranges and core areas maximum of 21.7 km from their nests (Table 1). rather than basing them on individual location The greatest distance a falcon was located from estimates. Although sizes of such areas are in- its nest was 38.3 km. Falcon use of these large fluenced by telemetry error, this influence is less ranges was not distributed in an even manner. than the bias created by using individual loca- Rather, 90% of locations were confined to l-7 tions to determine habitat use (Senchak 1991). core use areas that included only 38% of the Areas estimated without accounting for teleme- total range. try error may underestimate actual area (Sen- Average Prairie Falcon home range size was chak 1991), which might lead to overestimates nearly twice as large in 1993 and 1994 com- of the significance of habitat selection. However, pared to 1991 and 1992 (Z 95% harmonic mean this is unlikely in our study because we used range ? SE for: 1991 = 227 km* Z?22; 1992 = conservative methods to assign significance to 204 km2 ? 26; 1993 = 341 km* -+ 36; 1994 = habitat selection (individual falcons as experi- 400 km* ? 38). Year-to-year changes in use of 514 JOHN M. MARZLUFF ET AL.

I Unidentified m Reptile IIUII Bird EZZI Mammal I TGS

8 11 2-- 35 6 39 80 18 9 1 12 19 60 36 41 40 29

CENTRAL WEST CENTRAL WEST CENTRAL WEST

1992 1993 1994 FIGURE 2. Percentageof fresh prey items delivered by adult Prairie Falcons to their territories in the central and western sampling areas (see Fig. 1 for area locations) in 1992-1994. Unidentified prey items presentedhere were excluded from the analysis of annual variation in prey items. Numbers of prey of each type are adjacent to histograms.TGS = Townsend’s ground squirrel.

TABLE 1. Travel distancesand home range characteristicsof radio-tagged Prairie Falcons. Measures are av- eraged for the entire sample and a variety of subsamples.Table entries are mean, sample size, and standard error.

Average Maximum Size of home range (ha) Size of core use distance d,stance Maximum ______area (ha) traveled traveled distance from nest from nest across home convex 95% harmomc Sample (m) (m) range (m) polygon Ill.%" 95% cluster 90% cluster All falcons 7,031 21,748 27,561 31,472 29,757 16,203 11,351 77 77 93 95 95 95 95 290 738 909 1,570 1,772 893 718 Successfulpairs 7,021 21,253 26,773 29,113 25,959 14,865 10,597 58 58 58 58 58 58 58 352 815 1,175 1,882 1,887 1,115 860 Unsuccessfulpairs 7,063 23,259 28,675 34,536 36,753 16,488 11,985 19 19 19 19 19 19 19 484 1,651 2,020 4,322 5,298 2,065 1,631 Non-nesters N/A N/A 29,094 35,838 34,613 20,233 13,146 18 18 18 18 18 2,056 3,164 3,824 1,942 1,933 Males 7,044 23,384 29,075 34,187 31,935 16,999 12,338 45 45 53 53 53 53 53 350 909 1,221 2,016 2,436 1,099 907 Females 7,013 19,447 25,553 28,045 27,009 15,219 10,128 32 32 40 42 42 42 42 501 1,127 1,310 2,400 2,542 1,464 1,132 PRAIRIE FALCON RANGING BEHAVIOR 575 area were significant for falcons that successful- inant vegetation (compare R* for models in Table ly reared 30-day-old young, and these differ- 3). ences were apparent in average, maximum, and The distribution of falcon location estimates core use of area (Table 2). Unsuccessful pairs per km* was heavily skewed to the northwest and non-nesters did not exhibit significant an- portion of the study area where Sandberg’s blue- nual differences in ranging habits regardless of grass (Poa secunda) and big sagebrush (Arte- the type of travel or range considered (Table 2). misia tridentata) were most common (Figs. 4 Males and females did not differ greatly in and 5). Greasewood (Surcobatus vermiculatus), their use of area (Table 1). Average, core, and four-winged saltbush (Atriplex canescens), and maximum use of area varied little between males Russian thistle (Salsola iberica) also were pos- and females that successfully reared 30-day-old itively correlated with the occurrence of falcon young (Table 2). Travel during the nesting cycle locations (Table 3). In contrast, falcon location by males and females differed significantly estimates were rarest in habitats with abundant among years (3-way interaction between stage rabbitbrush (Chrysothamnus nauseosus and C. of nesting cycle, sex and year in ANOVA com- viscidijlorus), cheatgrass (), paring travel distances of adequately sampled and bottlebrush squirreltail (Sitanion hystrix; Ta- birds: F9,4247= 4.0, P < 0.001). Males traveled ble 3). The general association of falcons with farther than females during territory establish- habitats dominated by sagebrush,winterfat (Cer- ment/incubation except in 1992 (Fig. 3). Fe- atoides lanata), and grassland was confirmed in males typically traveled farther than males dur- the analysis of habitat classes defined from sat- ing late brood rearing, except in 1992 when they ellite imagery (Table 3). traveled extensively during incubation/territory Areas used by falcons (2 1 location ktn*) establishment (Fig. 3). The greatest change in averaged 5% cover of big sagebrush,8% cover travel distance during the nesting cycle occurred of Sandberg’s bluegrass, and 2% cover of win- in 1992 when both sexes greatly reduced travel terfat (Fig. 6). Bluegrass was the dominant veg- during brood rearing and post-fledging. etation in areas used by falcons. In contrast, Non-nesters and breeders after nest failure cheatgrassdominated areas where falcons were ranged over larger areas than successfulbreeders not located. Areas containing many falcon lo- (Table l), but this difference was significant only cations (2 10 locations ktn2) had 12% cover of for average area use by males (Table 2). Female bluegrass,5% cover of big sagebrush,2.5% cov- use of core area was similarly related to breed- er of winterfat, 5.5% cover of cheatgrass, and ing success,but the trend only was marginally 6.7% cover of Russian thistle (Fig. 6). significant (Table 2). Maximum area used did Amount of land in agriculture was weakly and not differ with respect to breeding successfor positively associatedwith falcon abundance (Ta- males or females. ble 3). Agricultural lands are primarily inter- Falcons that nested in the northwest portion spersed throughout sage, winterfat, and blue- of the study area ranged over smaller areas than grass habitats in the northwest portions of the those that nested in the southeastportion. This study area. trend was significant for average travel of suc- Placement of home range. Prairie Falcons’ cessful and unsuccessful pairs and for use of home ranges contained a high percent cover of core area by unsuccessfulpairs (Table 2). Max- plant speciesassociated with Townsend’s ground imum area used did not vary significantly with squirrels. Home ranges contained over twice the nest location, regardlessof breeding success(Ta- percent cover of winterfat and bluegrass than ble 2). was available within the usual flight range of falcons (see Methods). Likewise, ranges con- HABITAT SELECTION tained significantly more of the annual grass Si- General association between falcons and vege- tanion hystrix than expected based on availabil- tation. Habitat features explained a significant, ity (Table 4). In contrast, falcon home ranges but moderate (< 4/3)portion of the variation in had fewer salt-desertshrubs (Atriplex canescens, the abundance of falcon locations (Table 3). The A. conferttfolia, Grayia spinosa,Sarcobatus ver- number of telemetry locations per km* was more miculatus, Tetradymia glabrata) and grasses sensitive to variation in percent cover of indi- (Bromus tectorum, Agropyron spicatum) than vidual plant speciesthan to variation in the dom- expected based on availability.

PRAIRIE FALCON RANGING BEHAVIOR 517

m Females ~ 0 Males

8000

77i 8000 z g 4000

L 0 8 c 12000 27 fi?iJ 8000

55 4000

c 0

12000

8000

Incubation / Early Late First Month Territory Brood Brood Afler Establishment Rearing Rearing Fledging FIGURE 3. Averagedistance traveled from their nestsby malesand femaleseach year throughoutthe nesting cycle. Nestingcycle periodsare definedin the Methods.Error barsare SE. Samplesizes are given aboveerror

Falcons nesting in the northwestern portion of in the placement of the home range (Table 4). the study area were more selective for sagebrush Most salt desert shrubs(Atriplex confertifolia, A. and bluegrass than those nesting in the south- nuttalli, and Sarcobatus vermiculatus) occurred east. Correlations between nest location (UTM in core areas less than expected. However, Atri- coordinates) and each individual falcon’s selec- plex canescensoccurred in core areas more fre- tion coefficient for vegetation were strongestfor quently than expected based on availability. sagebrush(n = 98; North UTM: r = 0.52, P < The degree of selectivity within home ranges 0.001; East UTM: r = -0.58, P < O.OOl), and was related to the location of the home range in also significant for bluegrass (n = 98; North the study area. Individual falcons that had more UTM: r = 0.22, P = 0.03; East UTM: r = bluegrass than expected based on availability in -0.26, P = 0.01). Selection for winterfat de- their home range were less selective for blue- creased significantly from north to south (n = grass in their placement of core areas (r = 98; North UTM: r = -0.25, P = 0.02). -0.52, n = 98, P < 0.001). Selection within Selection within the home range. Falcons were ranges for winterfat also was less in relation to less selective within home ranges than they were the extent of selection for winterfat in the place- 578 JOHN M. MARZLUFF ET AI..

TABLE 3. Importance of plant species coverage (determined by kriging between data obtained from vegetation transects) versus habitat categories (defined by dominant plant species classified from satellite imagery) in explaining variation in number of Prairie Falcon radiotelemetry locations per square kilometer. Table entries are significant (P < 0.05) t-statistics testing the importance of each variable in a stepwise multiple regression of vegetation on the number of falcon radiotelemetry locations for 199 l-1994 combined. Variables not remaining in model had nonsig- nificant (ns) t-values. Summary statistics for regression models are listed at bottom of each column.

Plant 5peci.z coverage Dominant habitat category

Species t-value Habitat i-value

Artemisia tridentata 6.3 SagebruMrabbitbrush 9.7 Atriplex canescens 5.6 Salt desert shrubs Atripkx confertifolia 1.9 Winterfat 9n; Atrip1e.x nuttallii -2.3 Grassland 4.9 Bromus tectorum -7.5 Agriculture 2.7 Ceratoides lanata 13.4 Water Chrysothamnus nauseosus -5.0 Cliff/rock outcrop 8n; Chrysothamnus viscidijorus -9.8 Grayia spinosa -4.8 Poa secunda 23.3 Salsola iberica 5.4 Sarcobatus vermiculatus 5.7 Sitanion hystrix -6.4 Tetradymia glabrata -4.2 Vulpia spp.’ ns Agropyron spicatum ns F 110.2 54.3 df 14,346s 5,3005 P

’ Vulpra octafrora and V. michrosfachys were not reliable d&nutted m the field.

ment of the home range (r = -0.39, n = 98, P prey abundance declined (Dunstan et al. 1978, < 0.001). Selection for bluegrass and winterfat Harmata et al. 1978, Squires et al. 1993), (3) within ranges with less than expected percent long travel distances and large home ranges for cover of these increased the overall per- males and females during incubation and terri- cent cover of bluegrassand winterfat within core tory establishment (Dunstan et al. 1978, Haak areas. Percent cover within core areas of blue- 1982), (4) increased length of travel distances grass and winterfat increased significantly with for females as the nesting cycle progressed the degree of selectivity within home ranges for (Dunstan et al. 1978), (5) use of many distinct bluegrass and winterfat (all Ps < 0.001). core areas within the foraging range (Hunt 1993), and (6) the importance of native grass- DISCUSSION land habitats (Hunt 1993). During four years of study, in which weather We believe the size and use of area within and prey abundance varied greatly (U.S. Dept. Prairie Falcon home ranges in the NCA results Interior 1996, Van Horne et al., in press), we largely from the patchy distribution of landscape documented the ranging behavior and habitat features associated with different densities and use of 98 radio-tagged Prairie Falcons. Previous availabilities of Townsend’s ground squirrels. In studies of Prairie Falcon behavior and habitat years of normal rainfall, ground squirrels reach use investigated few birds and usually only last- their highest densities in habitats with abundant ed two years. Despite these limitations, many cover of the native Sandberg’s bluegrass (Pm conclusions previously reported were confirmed secunda; Van Horne et al., in press). Following in our study, specifically: (1) greater reliance on a drought, Townsend’s ground squirrels survive alternate prey when ground squirrel populations better in native sagebrushhabitats, and congre- declined (Phipps 1979, Steenhof and Kochert gate around agricultural fields (Van Horne et al., 1988), (2) use of slightly larger areas by males in press). We located Prairie Falcons most fre- than females and increased use of area when quently in areas with high percent cover of PRAIRIE FALCON RANGING BEHAVIOR 579

Prairie Falcon Contours ii

a 1-5 m 6-10 5-0 Kilometers 11-15 m 16-20 m 20+ FIGURE 4. Association of Prairie Falcons and Sandberg’s bluegrass throughout the study area. Percentage coverage of bluegrassis indicated by shading as determined by kriging between vegetation transects.Falcon abundance for 1991-1994 is portrayed by contours determined by splining (a spatial interpolation method) between counts of radiotelemetry locations obtained per km*.

Sandberg’s bluegrass, winterfat, and big sage- Effects of the 1992 spring drought supportthe brush, probably because of the greater abun- importance of ground squirrel distribution as a dance of ground squirrels there. Moreover, in- determinant of Prairie Falcon foraging behavior. dividual falcons in the northwest portion of the The drought strongly influenced Townsend’s study area, where bluegrasslsagelwinterfat hab- ground squirrel densities and distributions (Van itats are close to nesting areas, brought more Horne et al., in press), and we saw these changes ground squirrels to their nestlings and ranged reflected in falcon ranging habits and habitat over significantly smaller areas than falcons use. Ground squirrel populations increased from nesting in the southeast where bluegrass/sage/ moderately high in 1991 to extremely high in winterfat is less common and far from the nest- 1992, then crashedto the lowest levels recorded ing cliffs (Figs. 2, 4, and 5). in the NCA in 1993 and 1994 (Van Home et al., 580 JOHN M. MARZLUFF ET AL.

_ Prairie Falcon Contours Arfemesia fridenfafa Coverage (%) I-6 5-O Kilometers 6-10 11-15 I 16-20 m 20+

FIGURE 5. Association of Prairie Falcons and big sagebrush throughout the study area. See Figure 4 for details.

in press). As expected, falcons used relatively nests apparently allowed falcons to dramatically small areas in 1992 and large areas in 1993 and reduce travel during brood rearing and post- 1994. However, the extreme reduction in area fledging periods in 1992, and rely on winterfat used and reduced use of bluegrass habitat in habitats more than in previous years (Table 3). 1992 was likely accentuated because a drought Similar weather conditions in 1994 causedblue- caused senescence of bluegrass by early May, grass to again senesce early (Van Horne et al., and ground squirrels, especially juveniles, in press), but squirrels were rare, and falcons moved as bluegrass senesced(Van Horne et al., continued to range far from their nests through- in press). Ground squirrels also congregated out the nesting cycle, concentrating their activ- near alfalfa fields close to falcon nesting areas, ities in habitats likely to contain ground squirrels where they could forage, although the fields (e.g., bluegrass and sagebrush). themselves did not provide burrows for refuge. Prairie Falcons used habitats likely to contain The abundance of available prey close to falcon ground squirrels more frequently during years of PRAIRIE FALCON RANGING BEHAVIOR 581

m Big sagebrush TABLE 4. Habitatselection for homerange placement EEI Cheatgrass anduse of spacewithin homeranges by PrairieFalcons. ED VWeffat Selectioncoefficients equal In (habitatused/habitat avail- icrm Sandbergs’ bluegrass able),and entries in body of table are averagedfor n = ä EI Russian thistle 1 98 falcons. Habitats used in the convex polygon home range are comparedto habitatsavailable in the typical flight range of falcons within the area sampledfor veg- etation (from 21 km north of the Snake River Canyon to 7 km southof the canyon) to test selectionfor place- ment of the home range. Selectionfor placementof core areas is tested by comparing habitats used within the core to habitats available within the convex polygon home range. Asterisks indicate significant (* = P < 0.05, ** = P < 0.01) selection for (ratio > 0.0) or against(ratio < 0.0) habitats.

Selection by mdividuals for placement of:

Home 95% core Plant species range area

0 >o >9 Artemisia tridentata -0.06 -0.12 Atriplex canescens -0.63** 0.31” Number of Falcon Locations I km* Atriplex confertifolia -1.52** -0.46** FIGURE 6. Coverageof dominantplant speciesper Atriplex nuttallii 0.11** -0.47** km2where no falcon radiotelemetry locationswere ob- Bromus tectorum -0.58** 0.04 tained, where at least one location was obtained, and Ceratoides lanata 0.71** -0.10 where 10 or more locations were obtained. Coverage Chrysothamnus viscidijorus -1.74** -0.13 of cheatgrassdecreases while bluegrassand winterfat Grayia spinosa -0.89** -0.06 increases in areas with progressively more falcon lo- Poa secunda 0.51** -0.09 cations rr = number of km’ cells with appropriate Salsola iberica -0.09 -0.04 number of falcon locations. Sarcobatus vermiculatus -1.01** -2.92** Sitanion hystrix 0.20** -0.06 Tetradymia glabrata -3.82** -0.15 low squirrel abundance than during years of Vulpia spp.’ -0.10 -0.07 high squirrel abundance. Apparently bluegrass/ Agropyron spicatum -0.86** -0.15 sage/winterfat habitats may support the highest 1 Vulpia octafiora and V. michrostachys were not reliably delineated m abundances of squirrels when populations are the field. low. Thus, maintenance of bluegrass/sage/win- terfat habitat is probably important to the man- continuously when away from their canyon nests agement of Prairie Falcon populations. Even (Phipps 1979), we suspect that observations of though falcons routinely consumed other prey actual foraging would show an even strongeruse types (birds and reptiles) in years of low squirrel of bluegrass/sage/winterfatmosaics by falcons. abundance, Townsend’s ground squirrels make Prairie Falcons in the NCA had larger home up the majority of falcon diets, especially in ranges than Prairie Falcons in other areas (Dun- years when falcon productivity is high (Fig. 2). stan et al. 1978, Harmata et al. 1978, Hunt 1993, Variation in habitat accounted for only l/3 of Squires et al. 1993). Only Haak (1982) reported the variation in the observed locations of ra- ranges (228 kn?) close to the size we docu- dio-tagged falcons in the study area. However, mented (298 km*). Some of the differences may the reliance of falcons on native shrub and grass- be methodological: (1) we did not record loca- land mosaics may be greater than reported here tion estimates at nests because of topography because our estimated locations were associated (such locations were, however, also excluded by with large error. This should have little effect on Squires et al. [1993] and Hunt [1993]), (2) our our assessmentsof habitat use in home ranges study included years of low prey abundance, (3) and core areas because these area measurements we tracked birds throughout the nesting cycle rely, not a single location, but on a collection of from territory establishment to dispersal, and (4) locations which are unbiased. Obtaining loca- we used fixed tracking sites at varying distances tions remotely also prevented us from knowing from the nest sites that allowed us to wait for what the falcons were doing at the time a loca- birds to visit all parts of their range. tion was determined. Although falcons forage Beyond the methodological factors, it also is 582 JOHN M. MARZLUFF ET AL. possible that home ranges are larger in the NCA may be the only productive breeders in the pop- because of the way in which high-quality-prey ulation (Smith and Johnson 1985). The openness habitat is distributed. Schoener (1968) reported of agricultural habitats, especially pastures and that Prairie Falcons had much larger home rang- hayfields, may be conducive to successfulfalcon es than expected for birds their size. He specu- foraging (Haak 1982) and may provide a variety lated, and Harmata et al. (1978) concurred, that of prey types, such as voles and passerines, to this was possibly due to dispersion of habitats supplement falcon diets when ground squirrel containing prey and constraints of the falcon’s populations crash. These possible advantages hunting style. Our results support this idea. Na- may explain the slight positive association of tive grass habitats most likely to contain Town- falcons with agriculture during our study, which send’s ground squirrels were 5-20 km from nest occurred during years of mild to severe drought. sites (Fig. 4), home range size increased with However, the advantage of agriculture in our declining prey, and home range size increased study area may be unique because agricultural from northwest to southeast within the study fields are sparsely dispersed among native hab- area in parallel with reduced percent cover of itats thereby increasing habitat diversity. Con- perennial grass. version of large tracts of native vegetation to Because of the mosaic structure‘of prey hab- agriculture would adversely impact falcons be- itat in the NCA, increased foraging areas re- cause overall prey abundance is much lower in quired during years of low prey abundance may agricultural lands than in native shrubland (U.S. be too large for effective brood rearing. Falcons Dept. Interior 1979). in the NCA apparently cannot range over more Controlling is perhaps the most im- than 300 km2 and still provide the required food portant management practice beneficial for Prai- and vigilance to their nestlings. Falcons traveled rie Falcons. Shrubsteppehabitats are vulnerable over areas of 350-400 km2 in years when their to fire, which converts shrub and perennial grass reproduction was poor (1993, 1994) but only communities into exotic annual communities covered 200-280 km2 in years when their repro- which are eventually dominated by cheatgrass duction was good (1991, 1992; Table 1). More- (Yensen 1982). Cheatgrass communities can over, when prey declined those falcons nesting support ground squirrel populations, but popu- in the southeast,who ranged over areas 2 300 lations in such areas are more susceptible to km2 even when squirrels were abundant, were drought and therefore provide less stable prey the first to show poor reproduction and usually populations than those in shrub/perennial grass had poorer reproduction than pairs nesting in the mosaics (Yensen et al. 1992, Van Horne et al., northwest where squirrel habitat was abundant in press). close to nests (Steenhof et al., unpubl. data). Ideal habitat for foraging falcons probably in- ACKNoWLEDGMENTS eludes a mosaic of shrubs and grasses, in which This study was funded primarily by the Idaho Army shrub patches 210 ha provide cover and forage National Guard (IDARNG) under U.S. Army contract for squirrels, particularly during drought periods DAAD05-90-0135 and numerous agreements admin- istered by W. S. Seegar. The U.S. Bureau of Land (Van Home et al., in press). Our observation that Management (BLM) and the U.S. National Biological sagebrt=h and winterfat covers approximately Service provided additional funding and support. This 8% of the ground used most bv falcons should studywas part of the cooperativeBLM/IDARNG pro- not be interpreted to mean a continuous shrub ject.. J. McKinley, R. Townsend and the armies of Greenfalk trackers were invaluable with data collec- coverage. Rather, areas of perennial grasses tion. S. Knick, S. Watts, and B. Van Horne kindly pro- should be interspersed with stands of winterfat vided unpublisheddata on vegetation and Townsend’s and big sagebrush. Open areas are apparently ground squirrel use of the study area for our analysis. needed by falcons to catch squirrels (Haak D. Dyer and T Zariello provided GIS support. K. 1982), but shrub cover may reduce squirrels’ Steenhof, M. Kochert, R. Kenward, B. Van Home, and J. Rotenberry made valuable comments on earlier abilities to detect raptors making it potentially is drafts of the manuscript. easier for falcons to surprise their prey (Sharpe et al. 1994). LITERATURE CITED Agricultural borders may be especially im- ALLEN, G. T 1987. Prairie Falcon aerie site charac- portant habitats for Prairie Falcons in drought teristics and aerie use in North Dakota. Condor years because ground squirrels in these areas 89:187-190. PRAIRIE FALCON RANGING BEHAVIOR 583

ALLEN,G. T., R. K. MURPHY,K. STEENHOF,AND S. W. KENWARD,R. E. 1992. Quantity versus quality: pro- PLATT. 1986. Late fledging dates, renesting, and grammed collection and analysisof radio-tracking large clutchesof Prairie Falcons. Wilson Bull. 98: data. D. 231-246. In I. G. Priede and S. M. Swift 463-465. [eds:]:Wildlife telemetry-remote monitoring and ANDERSEN,D. E., AND 0. J. RONGSTAD. 1989. tracking of . Ellis Horwood, Chichester, Home-range estimatesof Red-tailed Hawks based UK. on random and systematic relocations. J. Wildl. KENWARD,R. E., AND K. H. HODDER.1995. Ranges Manage. 53:802-807. V. Software for analyzing location data. BEAUVAIS,G., J. H. ENDERSON,AND A. J. MAGRO. Inst. Terrestrial , Wareham, UK. 1992. Home range, habitat use and behavior of KNICK, S. ‘I, J. T ROTENBERRY,AND T. J. ZARRIELLO. Prairie Falcons wintering in east-centralColorado. 1997. Supervised classification of Landsat the- J. Raptor Res. 26:13-18. matic mapper imagery in a semi-air rangelandby BLOOM,F! H. 1987. Capturing and handling raptors. nonparametric discriminant analysis. Photogram- p. 99-123. In B. A.-GironPendleton, B-A. Mill- metric Engineering & Remote Sensing 63:79-86. sau. K. W. Cline. and D. M. Bird leds.1. Rantor LEN~H,R. V. 1981. On finding the sourceof a signal. managementtechniques manual. Nati. Wildl. Fed., Technometrics23:149-154. Washington, DC. MANLY, B. E J., L. L. MCDONALD,AND D. L. THOMAS. BUEHLER,D. A., J. D. FRASER,L. S. MCALLISTER,M. 1993. Resource selection by animals. Chapman R. FULLER,AND J. K. D. SEEGAR.1995. Captive and Hall, London. and field tested radio attachment techniques for MARZLUFF,J. M., AND M. E. MCFADZEN. 1996. Do Bald Eagles. J. Field Omithol. 66:173-180. brood counts accurately measure productivity? CLUGSTON,D. 1990. Preliminary evaluation of pro- Wilson Bull. 108:151-153. posed techniquesto monitor Prairie Falcon move- MARZLUFF,J. M., M. S. VEKASY,AND C. COODY.1994. ments, habitat occupancy,and food delivery rates Comparative accuracyof aerial and ground telem- to the nest in responseto military training in the etry locationsof foraging raptors.Condor 96:447- Orchard Training Area (OTA), p. 41-58. In K. 454. Steenhof [ed.], Snake River Birds of Prey research MARZLUFF,J. M., M. VEKASY, L. S. SCHUECK,M. project annual report. U.S. Dept. Int., BLM., Boi- MCFADZEN,R. R. TOWNSEND,B. A. K~MSEY,J. 0. se, ID. MCKINLEY,AND C. J. COODY. 1992. Influence of DIXON, K. R., AND J. A. CHAPMAN.1980. Harmonic military training on the behavior of raptors in the mean measure of animal activity areas. Ecology Snake River Birds of Prey Area, 1992, p. 49-143. 61:1040-1044. In K. Steenhof ted.], Snake River Birds of Prey DUNSTAN,T. C., J. H., HARPER,AND K. B. PHIPPS. Area 1992 annu. rep. U.S. Dept. Int., BLM, Boise, 1978. Habitat use and hunting strategiesof Prairie ID. Falcons, Red-tailed Hawks, and Golden Eagles. MCFADZEN,M. E., AND J. M. MARZLUFF.1996. Be- U.S. Dept. Int., BLM, Denver Federal Center, havior of Prairie Falcons (F&o mexicanus) dur- Denver, CO. ing the nesting and fledging-dependenceperiods FLOYD,D. E., AND *J.E. ANDERSON.1982. A new point under fluctuatingprey conditions.Bird Behav. 11: frame for estimating cover of vegetation. Vegeta- 81-89. tio 50: 185-186. NAMS, V. 0. 1989. Effects of radiotelemetry error on HAAK, B. 1982. Foraging ecology of Prairie Falcons sample size and bias when testing for habitat se- in northern . M.Sc. thesis, State lection. Can. J. Zool. 67:1631-1636. Univ., Corvallis, OR. OGDEN, V. T., AND M. G. HORNOCKER.1977. Nesting HARMATA,A. R., J. E. DURR,AND H. GEDULDIG.1978. density and successof Prairie Falcons in south- Home range, activity patterns and habitat use of western Idaho. J. Wildl. Manage. 41:1-11. Prairie Falcons nesting in the Mojave Desert. U.S. PHIPPS,K. B. 1979. Hunting methods,habitat use and Dept. Int., BLM, Denver Federal Center, Denver, activity patterns of Prairie Falcons in the Snake co. River -Birds of Prey Natural Area, Idaho. M.Sc. HOLTHUIJZEN,A. M. A. 1990. Prey delivery, caching, thesis. Western Illinois Univ.. Macomb. IL. and retrieval rates in nesting Prairie Falcons. Con- PLAY, S. ‘W. 1974. Breeding status and distribution dor 92~475-484. of the prairie falcon in northern New . HUNT, L. E. 1993. Diet and habitat use of nesting MSc. thesis, Oklahoma State Univ., Stillwater, Prairie Falcons (F&o mexicanus) in an agricul- OK. tural landscapein southernAlberta. M.Sc. thesis, PLAIN, S. W. 1981. Prairie Falcon: aspectsof popu- Univ. , Edmonton, Alberta, . lation dynamics, individual vocal identification, HURLBERT,S. H. 1984. Pseudoreplicationand the de- marking and sexualmaturity. Ph.D. diss., Brigham sign of ecological field experiments. Ecol. Mon- Young Univ., Provo, UT ogr. 54:187-211. RUNDE,D. E., AND S. H. ANDERSON.1986. Character- KAISER,T J. 1986. Behavior and energeticsof Prairie istics of cliffs and nest sites usedby breeding Prai- Falcons (F&o mexicanus) breeding in the west- rie Falcons. J. Raptor Res. 20:21-28. em Mojave desert. Ph.D. diss., Univ. California, SCHOENER,T W. 1968. Sizes of feeding territories Los Angeles. among birds. Ecology 49:123-141. KENWARD,R. E. 1987. Wildlife radio tagging. Aca- SENCHAK,S. S. 1991. Home ranges and habitat selec- demic Press, London. tion of Red-shouldered Hawks in central Mary- 584 JOHN M. MARZLUFF ET AL.

land: evaluating telemetry triangulation errors. techniques,Part 6, Sp. no. 355.0. U.S. Fish Wildl. M.Sc. thesis, Virginia Polytechnic Inst., Blacks- Serv. and Can. Wildl. Serv. Washington,DC. burg, VA. U.S. DEPARTMENTOF THE INTERIOR. 1979. SnakeRiver SHARPE, F? B., R. L. SCHOOLEY, AND B. VAN HORNE. Birds of Prey specialresearch report. BLM, Boise, 1994. Behavior of Townsend’s ground squirrelsin ID. different habitat types, p. 188-139. In K.-Steenhof U.S. DEPARTMENTOF THE INTERIOR. 1996. BLMI led.1. Snake River Birds of Prev Area 1994 annu. IDARNG researchproject final report. Natl. Biol. ;ep.‘k.S. Dept. Int., BLM, Boise, ID. Serv., Boise, ID. SMITH,G. W., AND D. R. JOHNSON.1985. Demography VAN HORNE,B., OLSON, G. S., SCHOOLEY,R. L., CORN, of a Townsend ground squirrel population- in J. G., AND K. P BURNHAM.In press. The effects southwesternIdaho. Ecologv 66:171-178. of drought and prolonged winter on demography SPENCER,W. D., ANDR. H. BA&ETT. 1984. An eval- of Townsend’s ground squirrelsin different shrub- uation of the harmonic mean measurefor defining steppe habitats-Ecology. carnivore activity areas. Acta Zoologica Fennica VEKASY,M. S.. J. M. MARZLUFF.K. STEENHOF.M. N. 171:255-259. KOCHERT,AND R. L. LEHMAN. 1996. Influence of SQUIRES,J. R., S. H. ANDERSON,AND R. OAKLEAF. radio-transmitterson Prairie Falcons. J. Field Or- 1993. Home range size and habitat-use patterns nithol. 67:680-690. of nesting Prairie Falcons near oil developments WHITE, G. C., ANDR. A. GARROTT.1986. Effects of in northeasternWyoming. J. Field Omithol. 64: biotelemetry triangulation error on detecting hab- itat selection. J. Wildl. Manage. 50:509-513. l-10. WHITE,G. C., AND R. A. GARROTT.1990. Analysis of STEENHOF,K. 1987. Assessing raptor reproductive wildlife radio-tracking data. Academic Press, San successand productivity, p. 157-170. In B. A. Gi- Diego, CA. ron Pendleton, B. A. Milsap, K. W. Cline, and D. YENSEN.D. L. 1982. A grazing historv of southwest M. Bird [eds.], Raptor management techniques Idaho with emphasis& thg Snakekiver Birds of manual. Nat. Wildl. Fed.. Washington.DC. Prey Area. Prog. Rep. U.S. Dept. Int., BLM, Boi- STEENHOF,K., AND M. N. KOCHERT.1588, ‘Dietary re- se, ID. sponsesof three raptor species to changing prey YENSEN,E., D. L. QUINNEY,K. JOHNSON,K. TIMMER- densitiesin a natural environment. J. Anim. Ecol. MAN, AND K. STEENHOF.1992. Fire, vegetation 57:37-48. changes, and population fluctuations of Town- U. S. DEPARTMENTOF INTERIOR.1977. North Ameri- send’s ground squirrels. Am. Midl. Nat. 128: can bird banding manual, Vol. 2, bird banding 299.-3 12.