Interactiveffects of prey and weather on golden eagle. reproduction. KAREN STEENHOF, MICHAEL N. KOCHERT and TRENT L. MCDONALD* 66,

Transcription

1 Interactiveffects of prey and weather on golden eagle 66, Journalof Animal Ecology 1997, KAREN STEENHOF, MICHAEL N. KOCHERT and TRENT L. MCDONALD* US National Biological Service, Raptor Research and Technical Assistance Center, 97 Lusk Street, Boise, ID 8376, USA; and * Western EcoSystems Technology, 23 Central Avenue, Cheyenne, Wyoming 821 USA Summary 1. The of the golden eagle Aquila chrysaetos was studied in southwestern Idaho for 23 years, and the relationship between eagle and jackrabbit Lepus californicus abundance, weather factors, and their interactions, was modelled using general linear models. Backward elimination procedures were used to arrive at parsimonious models. 2. The number of golden eagle pairs occupying nesting territories each year showed a significant decline through time that was unrelated to either annual rabbit abundance or winter severity. However, eagle hatching dates were significantly related to both winter severity and jackrabbit abundance. Eagles hatched earlier when jackrabbits were abundant, and they hatched later after severe winters. 3. Jackrabbit abundance influenced the proportion of pairs that laid eggs, the proportion of pairs that were successful, mean brood size at fledging, and the number of young fledged per pair. Weather interacted with prey to influence eagle reproductive rates. 4. Both jackrabbit abundance and winter severity were important in predicting the percentage of eagle pairs that laid eggs. Percentage laying was related positively to jackrabbit abundance and inversely related to winter severity. 5. The variables most useful in predicting percentage of laying pairs successful were rabbit abundance and the number of extremely hot days during brood-rearing. The number of hot days and rabbit abundance were also significant in a model predicting eagle brood size at fledging. Both success and brood size were positively related to jackrabbit abundance and inversely related to the frequency of hot days in spring. 6. Eagle was limited by rabbit abundance during approximately twothirds of the years studied. Weather influenced how severely eagle declined in those years. 7. This study demonstrates that prey and weather can interact to limit a large raptor population's productivity. Smaller raptors could be affected more strongly, especially in colder or wetter climates. Key-words: Aquila chrysaetos, black-tailed jackrabbits, Idaho, Lepus californicus, modelling. Ecology (1997) 66, Introduction breeding areas (Galushin 1974; Hamerstrom 1979; Smith et al. 1981; Korpimaki & Hongell 1986; Kor- Numerous studies of raptors have emphasized the pimaki, Lagerstr6m & Saurola 1987; Korpimaki & close relationship between raptor and Norrdahl 1991). In other species, most individuals the densities of principal prey species (e.g. Woffinden remain on nesting territories but do not lay eggs dur- & Murphy 1977; Adamcik, Todd & Keith 1978; Smith ing periods of low prey abundance (Southern 197; & Murphy 1979; Smith, Murphy & Woffinden 1981; Adamcik et al. 1978; Village 1981). Still other raptors Korpimaki 1984, 1992). In some species, most indi- lay fewer eggs and/or experience increased egg or nes- viduals respond to prey declines by moving to other tling mortality when prey populations are low (Cave 35

2 351 K. Steenhof, M.N. Kochert & T.L. McDonald Ecology, 66, ; Smith & Murphy 1979; Village 1981; Korpimaki 1984, 1987, 1992). Few studies have documented weather as a limiting factor for raptor. Weather can affect raptors through its effect on prey or on hunting behaviour, and adverse weather also can directly affect the survival of both adults and young. Rainfall may influence foraging success and thereby affect either a female's body reserves prior to egg-laying or a male's ability to deliver prey to an incubating female (Hirons 1982; Village 1986). European kestrel Falco tinnunculus L. laying dates are influenced by winter and spring precipitation and spring temperatures (Cave 1968), and nesting failures correlate with the amount of spring rainfall (Village 1986). Annual nesting success rates of African black eagles Aquila verreauxi Lesson in Rhodesia (Gargett 1977) and honey buzzards Pernis apivorus L. in Europe (Kostrzewa 1989) are also related inversely to rainfall. Red-tailed hawk Buteo jamaicensis Gmelin nestling mortality in Alberta is related to the biomass of food delivered to the nest as well as the frequency of rain (Adamcik, Todd & Keith 1979). Cold and rain are associated with mortality of small nestlings of both northern goshawks Accipiter gentilis L. and common buzzards Buteo buteo L. in Europe (Kostrzewa & Kostrzewa 199). In Wyoming, prairie falcon Falco mexicanus Schlegel pairs abandoned their nesting attempts during a spring snowstorm (Squires, Anderson & Oakleaf 1991). Bald eagle Haliaeetus leucocephalus L. reproductive success is negatively correlated with the sever- ity of spring weather (Swenson, Alt & Eng 1986). Severe winter weather depleted nesting populations of common barn-owls (Tyto alba Scopoli) and reduced their productivity the following spring (Marti & Wagner 1985). In Germany, the density of kestrel pairs and the percentage of pairs laying eggs are closely related to winter temperatures and snow cover as well as vole abundance (Kostrzewa & Kostrzewa 1991). Kestrels are more likely to starve during severe winters, and surviving kestrels are less likely to lay eggs because of their reduced mass. Winter severity, however, does not affect density or laying rates of the larger common buzzards and northern goshawks (Kostrzewa & Kostrzewa 1991), and Tengmalm's owl Aegolius funereus L. laying dates are unaffected by winter temperatures and snow depth (Korpimaki & Hakkarainen 1991). Golden eagle Aquila chrysaetos L. reproductive rates are known to fluctuate with prey densities (Smith & Murphy 1979; Tjernberg 1983; Watson & Langslow 1989; Bates & Moretti 1994). Failure of eagles to lay eggs during periods of low prey abundance has been documented in both Sweden (Tjernberg 1983) and the western United States (Smith & Murphy 1979). However, some evidence suggests that this response may be mitigated by mild winter weather (Tjernberg 1983). Spring weather also may affect survival of golden eagle nestlings. Mosher & White (1976) con- cluded that golden eagle young are susceptible to thermal stress during the first 6 weeks after hatching, particularly between 3 and 6 weeks of age. Beecham & Kochert (1975) found that heat stress was a significant mortality factor for nestling eagles in Idaho. In this paper we examine 23 years of data on golden eagle reproductive rates in relation to both weather conditions and prey abundance. Golden eagles that nest in southwestern Idaho are year-round residents (Steenhof, Kochert & Moritsch 1984), and blacktailed jackrabbits Lepus californicus Gray are their principal prey (Steenhof & Kochert 1988). We predicted that rabbit abundance would influence whether female eagles laid eggs and whether pairs would successfully raise young. We also predicted that these effects would be modified by weather. Winter severity should influence the proportion of females that lay eggs, particularly when increased energy demands occur during periods of decreased prey availability. Spring weather should influence nesting success of pairs that lay eggs: cold, wet weather during incubation and early brood-rearing could result in chilling of eggs or small young; extremely hot weather could cause increased nestling mortality. Study area The study area was 243 ha of public and private lands within the Snake River Birds of Prey National Conservation Area (NCA) in southwestern Idaho, USA (42?', 115?'W). The principal physiographic feature of the study area is the Snake River Canyon, with basalt cliffs ranging from 2 to 125m in height. Topography above the canyon is generally flat or slightly rolling with a few isolated buttes. Elevation ranges from 77 m in the canyon bottom to 1 m at the rim. Native vegetation in the area is characteristic of a shrub steppe community, with big sagebrush Artemisia tridentata Nutt., shadscale Atriplex confertifolia [Torr. and Frem.] Wats. and winterfat Ceratoides lanata [Pursh] J.T. Howell vegetation associations (US Department of the Interior 1979). Wildfires destroyed approximately 5% of the shrub cover between 198 and 1985 (Kochert & Pellant 1986), resulting in large stands of exotic annual grasses (primarily Bromus tectorum L.) and forbs. Annual precipitation averages 2cm and occurs mainly in winter; summers are hot and dry. Methods EAGLE NESTING DENSITY AND REPRODUCTION From 1971 to 1994, the entire NCA was searched for nesting golden eagles. Nearly all golden eagle nesting territories had been identified by previous research when our surveys began (Beecham 197; Kochert 1972). We located occupied territories by observing

3 352 Golden eagle Ecology, 66, territorial activity, courtship, brood-rearing activity, eggs, young, or any other conspicuous field sign (e.g. whitewash at a roost). Any area of cliff or power line where nests were found in successive years and where no more than one eagle pair bred at one time was considered a nesting territory (Newton & Marquiss 1982). In each year, we considered the number of nesting pairs to be equivalent to the number of occupied territories. From 1971 to 1994 we monitored reproductive activity at all known eagle nesting territories. We surveyed each territory at least twice each year from foot, boat, helicopter or fixed-wing aircraft. Most surveys prior to 1984 were from the ground; most surveys from 1984 to 1994 were by helicopter. We used fixed- wing aircraft only in Between 1971 and 1981, we entered some nests as often as every 4 days to assess nestling diet (Steenhof & Kochert 1988). To minimize stress to eggs or young, we did not enter nests during inclement weather. From 1985 to 1994, we conducted at least two helicopter surveys each year. The first survey in March determined if territories were occupied and if adults were incubating. We rechecked territories with incubating birds during a second flight in late May or early June to assess productivity. All territories for which the breeding outcome could not be determined from the helicopter were surveyed from the ground. We categorized eagle pairs that showed no evidence of egg laying after repeated observations or after climbing into and examining potential nests as 'nonlaying.' We confirmed 'laying' if an occupied territory contained an incubating adult, eggs, young, or any other indication that eggs were laid (e.g. fresh eggshell fragments in fresh nesting material). A pair was 'successful' if it produced > 1 young that reached 8% of normal fledging age [i.e. 51 days (Steenhof 1987)]. A small number of eagle nests (< 1) inspected after young had fledged were considered successful if: (i) a nest platform decorated the same year was worn flat and contained fresh prey remains; (ii) fresh faecal matter covered the back and extended over the nest's edge; and (iii) no dead young birds were found within a 5-m radius of the nest (Steenhof & Kochert 1982). Our analysis of nesting success and productivity excluded nesting attempts where research personnel had treated diseased young, installed shade devices, radioed adults, or transferred young in or out of nests. These activities, which may have influenced the outcome of a nesting attempt (M. Kochert, unpublished data), occurred at fewer than 15% of the nesting attempts over all years. We also excluded nests where investigators were known to have caused egg or chick mortality (n = 6). We assigned median hatching dates to all nests where we observed young prior to fledging (n = 873). We used a photographic ageing key (Hoechlin 1976) to age young. We were unable to determine hatching dates at nests where young died or fledged before we could age them, and we were unable to assign laying dates to pairs that failed during incubation. We considered the following reproductive parameters in our analyses: percentage of territorial pairs laying eggs; percentage of laying pairs that were successful; brood size at fledging; and number of young fledged per territorial pair. Brood size at fledging was based on counts of young in successful nests at or after 8% of fledging age (51 days). It includes information on clutch sizes and egg and nestling survival rates at successful nests, which we did not measure directly. Number of young fledged per territorial pair was a derived variable that described overall productivity of the population. It was calculated as the product of percentage of pairs laying, percentage of laying pairs successful, and brood size at fledging. PREY ABUNDANCE Each year from 1977 to 1991, black-tailed jackrabbit abundance was assessed by spotlighting from a vehicle along line transects in May and June (Smith & Nydegger 1985). Jackrabbit densities for the NCA were estimated using the program DISTANCE (Buckland et al. 1993). In 1971, 1973 and the average number of jackrabbits seen per day from late March to July by a 2-person raptor survey crew was tabulated from field notes. Jackrabbits per crew-day and jackrabbit densities from spotlighting were correlated (r = -86) from 1977 to 1983, and the relationship (y = x, where y = predicted density andx = jackrabbits seen per crew-day) was used to estimate pre-1977 jackrabbit densities. We used the mean of previous and current spring densities ('midpoint' or estimated winter density) in our analyses. This measure incorporates information about rabbits from both years that could influence eagle. We assumed that the midpoint density would best approximate rabbit density during egg-laying in late winter and incubation in early spring. WEATHER We obtained temperature and precipitation data from US Department of Commerce climatological summaries for three weather stations (Bruneau, Grand View and Swan Falls) in or near the NCA. We obtained data on snow and heating degree days from summaries for the Boise Airport. We analysed five variables that described the severity of winter immedi- ately preceding the breeding season: number of days when the temperature never rose above 2?F (-7?C); number of days when the temperature fell below?f (-18 C); number of days with snow cover; the last date of continuous snow cover on the ground; and heating degree days. One heating degree day is accumulated for each whole Fahrenheit degree that the mean daily temperature was below 65?F (18 3?C). We tallied heating degree days for the entire winter

4 353 K. Steenhof, M.N. Kochert & T.L. McDonald Ecology, 66, (November to February) and for the two coldest winter months (December and January). Other variables were summarized for the entire winter as well as for the period just prior to egg-laying (15 January to 15 February). Extreme heat was measured from 15 May to 15 June each year by counting the number of days when temperatures exceeded 32?C and 35?C. We also summarized the amount of precipitation during incubation (24 February to 1 April), brood-rearing (11 April to 15 June), and the entire nesting period (24 February to 15 June). ANALYSIS We undertook a model building process to investigate how prey and weather interact to affect golden eagle abundance, breeding chronology and. We first examined a SYSTAT (Wilkinson 1989) Pearson correlation matrix of data from all years to reduce the number of variables by eliminating non-independent or highly correlated variables. This reduced the pool of 19 weather and jackrabbit variables to 7. The pool of explanatory variables available in the model building process for each dependent variable was further reduced for two reasons. First, we did not use a potential explanatory variable if the connection between it and a certain dependent variable did not make sense biologically. For example, it did not make sense to use May/June temperatures to explain variation in hatching dates because hatching occurred before May. Secondly, we kept the number of terms low in the models for which only data points were available because we did not want to overfit the data for certain dependent variables. As a result, the pool of variables available as explanatory variables differed in each model building and are listed separately for each dependent variable in the subsections below. Distributional assumptions also differed for each variable due to the different nature of dependent variables. All of our modelling fits under the framework of 'general linear modeling.' We did not build a model for one variable, number fledged per pair, because our best estimator for this variable was a function of three dependent variables that we modelled, and consequently its distributional properties were unknown. Nesting abundance and breeding chronology To assess factors influencing the number of nesting pairs and breeding chronology, we modelled number of nesting pairs and hatching date using regular normal theory multiple regression and performed backward elimination of explanatory variables to reduce the number of terms. Terms in the full model were estimated winter jackrabbit densities per square kilometre (JRW), winter heating degree days (HDD), year, and the two-way interactions (JRW:YEAR, JRW:HDD and HDD:YEAR). Terms were eliminated based on the magnitude of the individual term's F statistic, which for normal theory models is roughly equivalent to the Akaike's information criterion (AIC) used below. Backward elimination stopped when elimination of additional terms drastically reduced the predictive power of the model, as measured by the residuals and the F statistic. Reproduction We modelled the relationships between three dependent variables that characterize : percentage of pairs laying eggs, percentage of laying pairs successful, and brood size at fledging. The explanatory variables used in the modelling process were estimated winter jackrabbit density (JRW) and the following weather variables: total precipitation (PRECIP), precipitation during brood rearing (PRECIPB), number of days in January and February below-18?c (LT18), number of days in May and June above 32?C (GT32), number of days with snow cover (SNOW), and winter heating degree days (HDD). A generalized linear response surface model was fitted by backwards elimination with a forward look at each step (Weisberg 1985, chapter 8). The backward elimination was carried out using the S + (? 1995; MathSoft Inc., version 3-3 for Windows) routine step.glm. At each step in the elimination process, single terms were deleted from the model, and the AIC statistic (Akaike 1973, 1985) of the reduced model was calculated. If deleting a term improved the AIC statistic (i.e. lowered AIC), the term that lowered AIC the most upon deletion was eliminated. The forward look at each step involved adding single terms eliminated in previous steps. If the AIC statistic was reduced upon addition of a term, the term that reduced AIC the most was added back to the model (Statistical Sciences 1995). Models similar to the final model selected via backwards elimination were investigated to see if any further improvement in the model could be achieved. Statistical significance of terms in the final model was assigned by dropping each in turn and performing a likelihood ratio test using the 'full' and 'reduced' likelihoods. Percentage of pairs laying eggs. The initial logistic response surface model in the backwards elimination process for percentage of laying pairs included JRW, HDD, SNOW, JRW2, HDD2, SNOW2, and the twoway interactions JRW:HDD, JRW:SNOW and HDD:SNOW. The logistic response surface model assumed the number of pairs laying eggs each year to be a binomial random variable with index mk (mk = total pairs in year k) and probability Pk. Under the binomial assumption, the logistic model related logit(pk) = lg(pk/(l-pk)) to a linear function of the explanatory variables and obtained coefficient estimates by maximizing the binomial likelihood function. All hypotheses about model coefficients were tested using likelihood ratio tests. The predictive power of other explanatory variables, relative to the

5 354 Golden eagle Ecology, 66, final model selected in the backwards elimination process, was investigated. Additional variables considered after backwards elimination were LT18, PRE- CIP, current year jackrabbit densities, previous year jackrabbit densities, and the interaction of previous year and current year jackrabbit densities. The weather variables of PRECIP and LT18 were investigated by first dropping any weather variables from the final model and then adding PRECIP and LT18 one at a time to assess their predictive power. Current year jackrabbit density, previous year jackrabbit density, and the interaction were investigated by first dropping all jackrabbit variables from the model and then adding all three to assess their predictive power. Percentage of laying pairs successful. The initial logistic response surface model in the backward elimination procedure for percentage of laying pairs that successfully raised young included JRW, GT32, HDD JRW2, GT322, HDD2 and the interactions JRW:GT32, JRW:HDD and GT32:HDD. The logistic response surface model assumed the number of successful pairs each year to be a binomial random variable with index mk (mk = total pairs in year k) and probability Pk The logistic model related logit(pk) = log(p/(l-pk)) to a linear function of the explanatory variables and obtained coefficient estimates by max- imizing the binomial likelihood function. All hypotheses were tested using likelihood ratio tests. The additional predictive power of other explanatory variables was investigated, relative to the final model selected in the backwards elimination process. Additional variables considered after backwards elimination were PRECIP, LT18, SNOW, current year jackrabbit densities, previous year jackrabbit densities, and the interaction of previous year and current year jackrabbit densities. The weather variables of PRECIP, LT18 and SNOW were investigated by first dropping any weather variables from the final model and then adding each in turn to assess their predictive power over current year jackrabbit density, previous year jackrabbit density, and the interaction were investigated by first dropping all jackrabbit variables from the model and then adding all three to assess their predictive power. Brood size at fledging. The initial model in the backwards elimination procedure for brood size at fledging included JRW, HDD, GT32, PRECIPB, JRW2, HDD2, GT322, PRECIPB2, and all two-factor interactions among linear terms. The response surface model used for brood size assumed that the average brood size in each year was a normal random variable with variance proportional to the inverse of sample size in each year. Coefficients in the linear model were estimated through weighted least squares techniques with weights equal to the sample size in each year (Table 1). Results GOLDEN EAGLE NESTING POPULATIONS An average of 31 golden eagle pairs occupied nesting territories in the NCA each year from 1971 to 1994 (range = 28-35, SD = 2-1; Table 1). The percentage of pairs that laid eggs each year (range: 38-1%, mean = 7%, SD = 15 5; Table 1) was relatively more variable than the number of pairs occupying territories (cv = 19-6 vs. 6-6; t = 4-444, P < -1). The percentage of laying pairs that successfully raised young also showed considerable variation among years (range: 32-8%, mean = 6%, SD = 13-9, cv = 23-2). Brood size at fledging, in contrast, remained fairly stable across the years (range = 1-2-, mean = 1-56, SD = -22, cv = 14-). Estimated number of young produced per pair reflected the variation in percentage of pairs laying and percentage of laying pairs successful (range = , mean = 79, SD = 35, cv = 44 3), as did total number of young produced by the population (range = 5-41,mean = 25, SD = 1-3, cv = 41-2). Hatching dates ranged from 17 March to 18 May (mean = 13 April, SD = 1 days). The earliest mean annual hatching date was 5 April, and the latest was 21 April. JACKRABBIT POPULATIONS Estimated jackrabbit densities ranged from 1 3 to 54 8 per km2 between 1971 and Jackrabbit densities were highest in 1971, but they decreased sharply in 1973 and remained low through to 1978 (Fig. 1). Numbers were high from 1979 to 1981, but the peak reached in 1979 was lower than that attained in Estimated densities declined again between 1981 and 1982, and the estimated population size between 1984 and 1986 was even lower than in the mid-197s. In 1987, populations began rebuilding to a third peak in 1992, decreased sharply in 1993, and rebounded slightly in Jackrabbit populations had peaked at successively lower densities from 1971, to , and Knick & Dyer (1997) distinguished two jackrabbit population phases (high and low) in the NCA. During our 23-year study, we identified 8 years when jackrabbit populations were at high phases and 15 years when they were in low phases of their population cycle (Fig. 1). Jackrabbit abundance was not correlated with any of the weather variables measured (r's < -4; P's > -1). EAGLE-JACKRABBIT-WEATHER RELATIONSHIPS Nesting abundance and breeding chronology Regression analyses showed that the number of eagle pairs occupying nesting territories each year was unrelated to rabbit density and winter severity. Year, however, was the single significant variable in the final model for the number of nesting pairs. The number of

6 i 355 K. Steenhof, M.N. Kochert & T.L. McDonald Table 1. Nesting success and productivity of golden eagles in the NCA, Sample sizes are in parentheses Year Mean 31-5 SD 2-1 CV 6-6 Number of occupied % of pairs nesting territories laying 1% (31) 65% (34) 67% (33) 75% (32) 68% (34) 82% (34) 75% (32) 97% (3) 87% (31) 1% (3) 87% (3) 96% (27) 38% (32) 55% (29) 78% (32) 74% (31) 8% (3) 87% (3) 9% (29) 86% (29) 9% (29) 63% (3) 79% % of laying pairs Brood size at Number fledged Total successful fledging per pair* fledged 62% (21) 44% (18) 53% (17) 56% (18) 5% (16) 59% (17) 77% (13) 63% (24) 76% (21) 73% (22) 79% (24) 67% (18) 59% (17) 42% (12) 4% (14) 32% (25) 52% (23) 63% (19) 8% (25) 74% (23) 56% (16) 8% (15) 5% (12) 6% (19) 1-17 (11) - (8).4 (1).53 (11) -61 (1).54 (1).73 (1).98 (15) -86 (18) 1-18 (19) 1-27 (18) 1-38 (14) 1.1 (9) -64t (4) -16 (5) -26 (8).34 (13) -59 (14) -69 (2) 1-15 (17) 1-6 (11) -83 (14) 1-23 (7) *Based on percentage of pairs laying, percentage of laying pairs successful, and brood size at fledging. tbased on percentage of pairs successful and brood size at fledging E High 3 -.o2- L-rl z 1 Low [ I, i I i i I I I I I I I I I I I I I I I i i i i i i i I i I I I I i, i, i i,. i i Fig. 1. Black-tailed jackrabbit densities (n km-2) in the Snake River Birds of Prey National Conservation Area, The shaded line separates 'high' and 'low' phases of the population cycle. eagle pairs in the NCA showed a slight but significant primarily to estimated winter jackrabbit densities decrease during the 24-year study. Year explained (JRW) and also to winter severity, as measured by 57% of the variation in number of pairs (P < -1) HDD. Eagles hatched earlier in years when jack- according to the linear regression: number of pairs rabbits were abundant, and they hatched later after = 5'4--23 x YEAR. severe winters. The final model relating hatching dates Golden eagle hatching dates, in contrast, showed no to explanatory variables was: hatching date = Ecology, 66, significant trends over time, and instead were related -236 x JRW + -2 x HDD. The three-dimen-

7 356 Golden eagle %J%J '" v 2 '-7. ;_, JRW 5 36 Fig. 2. Hatching dates of golden eagles in the Snake River Birds of Prey National Conservation Area in relation to estimated winter black-tailed jackrabbit abundance and winter severity (JRW), as measured by heating degree days (HDD), sional point cloud and estimated surface appear in Fig. 2. however, revealed that weather interacted with prey to influence eagle reproductive rates. Reproduction Preliminary scatter plots (Fig. 3) and correlation analyses showed that all four measures of eagle were strongly related to estimated winter jackrabbit densities but not to the weather variables we measured (r's < -5, P's > -5). Modelling exercises, Percentage of pairs laying eggs. Both estimated winter jackrabbit density (JRW) and winter severity, as measured by heating degree days (HDD), were important in predicting the percentage of eagle pairs that laid eggs. The final estimated logistic model relating percentage of pairs laying to explanatory variables was: )., 7-6? 5 a lo ioo 3s 9 8 o 7 o cm r 3 t _ -_ - 5 a I II I I5 3 I I Ecology, 66, c ) la m * 'o L ? a s~-. - * U) vc 1 lm. 1- o E z 4 _ a I I I I I I I Jackrabbits km-1 Fig. 3. Golden eagle in relationon to estimated winter black-tailed jackrabbit densities in the Snake River Birds of Prey National Conservation Area, *?

8 357 K. Steenhof, M.N. Kochert & T.L. McDonald logit(pk) = x JRW x HDD x JRW:HDD. All three terms in the model were significant (P = 95 for JRW, P = -6 for HDD, P = -19 for JRW:HDD). Three-dimensional and contour plots of the estimated response surface (Fig. 4) show that winter severity was most important in predicting laying rates when jackrabbit densities were low. To illustrate the interaction between jackrabbit numbers and heating days, consider the following calculations: if the severity of winter increased from 45 degree days to 5 degree days and jackrabbit density was low (5 km-2), the estimated percentage of pairs laying eggs fell from 72% to 63%; if the severity of winter increased from 45 degree days to 5 degree days and jackrabbit density was high at a value of 3 per square kilometer, the estimated percentage of pairs laying eggs was approximately constant, ranging from 93% to 97%. Percentage of laying pairs successful. The variables most useful in predicting percentage success were estimated winter jackrabbit densities (JRW) and number of days with temperatures above 32?C (GT32). The final logistic model relating percentage of laying pairs successful to explanatory variables was: logit(pk) = x JRW--251 x JRW x GT32. All terms in the final model were significant (P= 16 for JRW, P= -223 for JRW2, P = -276 for GT32). Three-dimensional and contour plots of the estimated surface (Fig. 5) show how the number of extremely hot days during brood-rearing affects success rates. The quadratic response surface model predicts that to maintain estimated percentage success at 5% when the number of days above 32 C increases from 5 to 1, the density of jackrabbits must increase by three rabbits km-2; to maintain a higher estimated percentage success of.9.8 CD n. "^^ ? a ?~~~~~~~~~~~* ~~ u.b/6 / *./ Ecology, 66, ~~~8 t ~~~ ~~~~~~~~~~~~~~~~~~~O' HDD Fig. 4. Three-dimensional and contour plots of the response surface depicting the model describing percentage of golden eagle pairs laying eggs in the Snake River Birds of Prey National Conservation Area,

9 358 Golden eagle C) GT32 Fig. 5. Three-dimensional and contour plots of the response surface depicting the golden model eagle pairs successful in the Snake describing percentage of River Birds of laying Prey National Conservation Area, % when the number of days above 32?C increases from 5 to 1, the density of jackrabbits must increase by 8-5 rabbits km-2. Brood size at fledging. Both winter rabbit densities (JRW) and the number of hot days during broodrearing (GT32) also were significant (P = -24 for JRW, P = -29 for GT32) in the final model relating brood size at fledging to explanatory variables: (mean brood size) = x JRW x GT32. Three-dimensional and contour plots of the esti- mated surface are shown in Fig. 6. The magnitude of the estimated coefficients was relatively small, and the practical importance of the model for predicting Ecology, 66, future changes in brood size should be considered. Under the least favourable conditions within the range of observed data, jackrabbit density was 1-5 km2, GT32 was 18, and the estimated average brood size was 1-25 chicks per nest. In the most favourable situation within the range of observed data, jackrabbit density was 37 km2, GT32 was 2, and estimated average brood size was 1-87 chicks per nest. All other estimated average brood sizes were between 1-25 and Discussion Long-term productivity of golden eagles in southwestern Idaho (-79 young per pair) was similar to that of golden eagle populations in other parts of the world. Watson (1957) reported that golden eagles in

10 359 K. Steenhof, M.N. Kochert & T.L. McDonald CD cu lal 4 (I) la 2~ I GT32 Fig. 6. Three-dimensional and contour plots of the response surfae depicting the model describing brood size at fledging for golden eagles in the Snake River Birds of Prey National Conservation Area, ? Ecology, 66, Scotland produced an average of '8 young per pair; Phillips et al. (199) observed a productivity rate of -78 in Montana and Wyoming; and Bates & Moretti (1994) reported -82 young per pair in Utah. As in other studies of golden eagles (Smith & Mur- phy 1979; Tjernberg 1983; Watson & Langslow 1989; Bates & Moretti 1994), prey abundance was an important factor influencing long-term in southwestern Idaho. Nesting eagles in southwestern Idaho were tied closely to populations of black-tailed jackrabbits. Although eagles used alternate prey, they showed strong preferences for jackrabbits even when rabbit populations were low (Steenhof & Kochert 1988). Numbers of territorial pairs were unrelated to annual rabbit numbers or the weather variables mea- sured. These findings are consistent with those of Brown & Watson (1964) and Watson & Lansglow (1989), who found stable numbers of golden eagle pairs in Scotland where territorial behaviour apparently limits the number of nesting pairs. Our results contrast with Galushin's (1974) prediction that raptors with a specialized and unstable diet should fluctuate in synchrony with prey numbers, but they confirm Galushin's (1974) observation that such nomadic behaviour is less common in resident, non-migratory populations. Because eagles are relatively long-lived, it is advantageous for them to remain on territory even when prey conditions for breeding are unfavourable (Newton 1979). In the population we studied, the overall annual reproductive output was influenced most strongly by

11 36 Golden eagle Ecology, 66, the proportion of pairs that laid eggs. As we predicted, golden eagle laying rates were related to food and weather prior to the nesting season. Both hatching dates and the percentage of pairs laying eggs were related to estimated winter jackrabbit densities and the severity of the winter preceding the nesting season. In Scotland, prey abundance was the most important factor influencing barn owl laying dates (Taylor 1994; p. 128), but owl laying dates also correlated closely with temperatures immediately before breeding. When Idaho jackrabbits were in the low phase of their population cycle and after cold winters, eagles either laid eggs later or early clutches failed before hatching. Fewer golden eagles laid eggs in Idaho when jackrabbits were scarce, and fewer laid eggs after severe winters. The severity of winter influenced the percentage of pairs laying eggs primarily at low phases of the jackrabbit population cycle. During peaks in the jackrabbit cycle, the percentage of pairs laying eggs was relatively high regardless of how cold the winter was. Well-fed eagles can probably withstand cold better than food-stressed birds (Stalmaster & Gessaman 1984). Our results are consistent with Tjernberg's (1983) finding that eagle is related to prey and weather prior to egg laying. Weather during the nesting season also influenced eagle in our study. The number of extremely hot days during brood-rearing was important in explaining the variability in both percentage of laying pairs successful and mean brood size at fledging. Both variables were related positively to jackrabbit abundance and inversely to the number of hot spring days. Tomback & Murphy (1981) found that underfed ferruginous hawk Buteo regalis Gray nestlings were especially vulnerable to heat stress and that the combined effects of inadequate food and high temperatures caused increased nestling mortality in years when prey populations were low. The number of golden eagle young fledged per pair was the product of all three independent measures of (percentage of pairs laying, percentage of laying pairs successful, and brood size at fledging), and it reflected the patterns described above. Jackrabbit abundance influenced the number of young fledged per pair in all years. In peak rabbit years, number of young fledged per pair was related inversely to the number of hot days in May and June. During low phases of the jackrabbit population cycle, number fledged per pair was related primarily (and inversely) to HDD and secondarily (and inversely) to the number of days > 32?C in May and June. Weather can affect raptors directly, and it also can influence them indirectly by affecting prey densities or prey behaviour. Reproduction of Aquila and other raptor populations in the southern hemisphere decreased following drought that reduced prey densities (Ridpath & Brooker 1986; Hustler & Howells 199). However, in our study, both cold and heat affected eagles independently of their effects on the abundance of the eagle's principal prey. Weather conditions were similar during low and high phases of the jackrabbit population cycle. There was no evidence that rabbit populations were influenced by winter severity or spring heat. Instead, rabbit abundance in our study area was unrelated to weather conditions and fluctuated cyclically, as has been documented in other parts of western North America (Gross, Stoddart & Wagner 1974; Smith 1987). Weather, however, may have affected rabbit behaviour and their vulnerability to eagles. Weather also may have affected the availability of alternative prey due to either behaviour or population changes that we did not measure. Our study demonstrated how prey populations and weather can interact to limit a raptor population's productivity. Other non-migratory populations of raptors that depend on cyclic prey in arctic and temperate environments are likely to show similar patterns [e.g. kestrels (Kostrzewa & Kostrzewa 1991); snowy owls (Wiklund & Stigh 1986); and barn owls (Taylor 1994)]. The shrub steppe environment we studied may be unique in that both extreme cold and extreme heat affected nesting populations. Rainfall, however, may be a more important factor for raptors in higher precipitation zones (Cave 1968; Adamcik et al. 1979; Village 1986). Precipitation was not a significant factor in our southwestern Idaho study area, where rainfall rarely exceeded 5 cm during the nesting season. Variations in prey and weather should have more serious consequences for smaller, shortlived raptors, because the relative costs of egg formation are higher for smaller species (Newton 1979). Kostrzewa & Kostrzewa (1991) found that, unlike kestrels, northern goshawks and common buzzards are able to withstand severe winters in Germany because of their large masses, their relatively low energy demands, and their consequent ability to thermoregulate effectively. Our results, however, show that even larger raptors can be influenced by cold weather at more southern latitudes. Our results re-emphasize the importance of longterm studies. Shorter investigations would have been unable to identify the effects of weather on eagles because weather effects were overshadowed by prey effects and were only apparent in some years. Modelling can be an effective approach to understand how weather and prey factors interact. Previous studies had documented short-term responses of raptors to severe weather conditions, but our study is one of the few that has examined how more subtle differences in weather and prey interact over the long term to influence populations. Our results support Wiens' (1977) theory of variable environments in which resources are limiting in some, but not all, years. With golden eagles in southwestern Idaho, years when prey limit were approximately twice as frequent as years when prey were superabundant. Weather influenced how severe the effect of decreased prey would be.

12 361 K. Steenhof, M.N. Kochert & T.L. McDonald Ecology, 66, Acknowledgements This paper is a contribution of the National Biological Service's Raptor Research and Technical Assistance Center (formerly the US Bureau of Land Management [BLM] Snake River Birds of Prey Research Project). The BLM provided funds from 1972 to From 1984 to 1989 Pacific Power and Light Company provided funds for aerial surveys. The Idaho Army National Guard (IDARNG) assisted with funding from 199 to 1994 as part of the Cooperative BLM/ IDARNG Research Project. The Idaho Cooperative Wildlife Research Unit provided the 1971 golden eagle data. J. H. Doremus and A. R. Bammann supervised the collection of nesting data between 1976 and 1983, and J. H. Doremus assisted with data collection from 1983 to 199. R. N. Lehman, L. B. Carpenter, and their crews helped to collect data from 199 to The study could not have been completed without the assistance of numerous field technicians and dedicated volunteers, especially J.R. Oakley, T.E. Hamer, O.R. Fremming and M. W. Collopy. Data on jackrabbit abundance were provided by M. L. Wolfe and L. O. Oftedahl, Department of Wildlife Science, Utah State University (BLM contract 525- CT5-13), G. W. Smith, D. J. Johnson, N. C. Nydegger, Department of Biology, University of Idaho (BLM contract 525-CT5-12), M. L. McHenry (Idaho Army National Guard), J.H. Doremus (BLM Bruneau Resource Area), S. T. Knick and S. E. Watts (National Biological Service). M. W. Collopy, E. Korpimaki, J. M. Marzluff, J. T. Rotenberry, and anonymous referees critically reviewed drafts of the manuscript. L. McDonald provided valuable input on statistical modelling. D. M. Parrish tabulated and automated most of the data, and A. Teutsch prepared the graphics. References Adamcik, R.S., Todd, A.W. & Keith, L.B. (1978) Demographic and dietary responses of great horned owls during a snowshoe hare decline. Canadian Field-Naturalist, 92, Adamcik, R.S., Todd, A.W. & Keith, L.B. (1979) Demographic and dietary responses of red-tailed hawks during a snowshoe hare decline. Canadian Field-Naturalist, 93, Akaike, H. (1973) Information theory as an extension of the maximum likelihood principle. Second International Symposium on Information Theory (eds B. Petrov & F. Csaki), pp Akademiai Kiado, Budapest, Hungary. Akaike, H. (1985) Prediction and entropy. A Celebration of Statistics (eds A. Atkinson & S. Fienberg), pp Springer, New York. Bates, J.W. & Moretti, M.O. (1994) Golden eagle (Aquila chrysaetos) population ecology in eastern Utah. Great Basin Naturalist, 54, Beecham, J.J. Jr (197) Nesting ecology of the golden eagle in southwestern Idaho. MSc thesis, University of Idaho, Moscow. Beecham, J.J. Jr & Kochert, M.N. (1975) Breeding biology of the golden eagle in southwestern Idaho. Wilson Bulletin, 87, Brown, L.H. & Watson, A. (1964) The golden eagle in relation to its food supply. Ibis, 16, Buckland, S.T., Anderson, D.R., Burnham, K.P. & Laake, J.L. (1993) Distance Sampling: Estimating Abundance of Biological Populations. Chapman and Hall, New York. Cave, A.J. (1968) The breeding of the kestrel, Falco tinnunculus, L., in the reclaimed area Oostelijk Flevoland. Netherlands Journal of Zoology, 18, Galushin, V.M. (1974) Synchronous fluctuations in populations of some raptors and their prey. Ibis, 116, Gargett, V. (1977) A 13-year population study of the black eagles in the Matopos, Rhodesia, Ostrich, 48, Gross, J.E., Stoddart, L.C. & Wagner, F.H. (1974) Demographic analysis of a northern Utah jackrabbit population. Wildlife Monographs, 4, 71 pp. Hamerstrom, F. (1979) Effect of prey on predator: voles and harriers. Auk, 96, Hirons, G.J.M. (1982) The effects of fluctuations in rodent numbers on breeding success in the tawny owl Strix aluco. Mammal Review, 14, Hoechlin, D.R. (1976) Development of golden eaglets in southern California. Western Birds, 7, Hustler, K. & Howells, W.W. (199) The influence of primary production on a raptor community in Hwange National Park, Zimbabwe. Journal of Tropical Ecology, 6, Knick, S.T. & Dyer, D.L. (1997) Distribution of black-tailed jackrabbit habitat determined by GIS in southwestern Idaho. Journal of Wildlife Management, 61, Kochert, M.N. (1972) Population status and chemical contamination of golden eagles in southwestern Idaho. MSc thesis, University of Idaho, Moscow. Kochert, M.N. & Pellant, M. (1986) Multiple use in the Snake River Birds of Prey Area. Rangelands, 8, Korpimaki, E. (1984) Population dynamics of birds of prey in relation to fluctuations in small mammal populations in western Finland. Annales Zooligici Fennici, 21, Korpimiki, E. (1987) Clutch size, breeding success and brood size experiments in Tengmalm's owl Aegolius funereus: a test of hypotheses. Ornis Scandinavica, 18, Korpimaki, E. (1992) Diet composition, prey choice, and breeding success of long-eared owls: effects of multiannual fluctuations in food abundance. Canadian Journal of Zoology, 7, Korpimaki, E. & Hakkarainen, H. (1991) Fluctuating food supply affects the clutch size of Tengmalm's owl independent of laying date. Oecologia, 85, Korpimiki, E. & Hongell, H. (1986) Partial migration as an adaptation to nest-site scarcity and vole cycles in Tengmalm's Owl Aegoliusfunereus. Vdr Figelvdrld, Supplement, 11, Korpimaki, E. & Norrdahl, K. (1991) Numerical and functional responses of kestrels, short-eared owls, and longeared owls to vole densities. Ecology, 72, Korpimaki, E., Lagerstr6m, M. & Saurola, P. (1987) Field evidence for nomadism in Tegmalm's Owl Aegolius funereus. Ornis Scandinavica, 18, 1-4. Kostrzewa, A. (1989) The effect of weather on density and success in honey buzzards Pernis apivorus. Raptors in the Modern World (eds B. -U. Meyburg & R. D. Chancellor), pp World Working Group on Birds of Prey, Berlin, London & Paris. Kostrzewa, R. & Kostrzewa, A. (199) The relationship of spring and summer weather with density and breeding performance of the buzzard (Buteo buteo), goshawk

13 362 Golden eagle (Accipiter gentilis) and kestrel (Falco tinnunculus). Ibis, 132, Kostrzewa, R. & Kostrzewa, A. (1991) Winter weather, spring and summer density, and subsequent breeding success of Eurasian kestrels, common buzzards, and northern goshawks. Auk, 18, Marti, C.D. & Wagner, P.W. (1985) Winter mortality in common barn-owls and its effect on population density and. Condor, 87, Mosher, J.A. & White, C.M. (1976) Directional exposure of golden eagle nests. Canadian Field-Naturalist, 9, Newton, I. (1979) Population Ecology of Raptors. Buteo Books, Vermilion, South Dakota. Newton, I. & Marquiss, M. (1982) Fidelity to breeding area and mate in sparrowhawks Accipiter nisus. Ecology, 54, Phillips, R.L., Wheeler, A.H., Lockhart, J.M., McEneaney, T.P. & Forrester, N.C. (199) Nesting ecology of golden eagles and other raptors in southeastern Montan and northern Wyoming. US Fish and Wildlife Service Technical Report 26, Washington, DC. Ridpath, M.G. & Brooker, M.G. (1986) The breeding of the wedge-tailed eagle (Aquila audax) in relation to its food supply in arid western Australia. Ibis, 128, Smith, D.G. & Murphy, J.R. (1979) Breeding responses of raptors to jackrabbit density in the eastern Great Basin desert of Utah. Raptor Research, 13, Smith, D.G., Murphy, J.R. & Woffinden, N.D. (1981) Relationships between jackrabbit abundance and ferruginous hawk. Condor, 83, Smith, G.W. (1987) Mortality and movement within a blacktailed jackrabbit population. PhD thesis, Utah State University, Logan. Smith, G.W. & Nydegger, N.C. (1985) A spotlight, linetransect method for surveying jack rabbits. Journal of Wildlife Management, 49, Southern, H.N. (197) The natural control of a population of tawny owls (Strix aluco). Journal of Zoology, London, 162, Squires, J.R., Anderson, S.H. & Oakleaf, R. (1991) Prairie falcons quit nesting in response to spring snowstorm. Journal of Field Ornithology, 62, Stalmaster, M.V. & Gessaman, J.A. (1984) Ecological energetics and foraging behavior of overwintering bald eagles. Ecological Monographs, 54, Statistical Sciences (1995) S-PLUS Guide to Statistical & Mathematical Analysis, Version 3.3, StatSci, a division of MathSoft, Inc., Seattle, WA. Steenhof, K. (1987) Assessing raptor reproductive success and productivity. Raptor Management Techniques Manual (eds B. A. Giron Pendleton, B. A. Millsap, K. W. Cline & D. M. Bird), pp National Wildlife Federation, Washington, DC. Steenhof, K. & Kochert, M.N. (1982) An evaluation of methods used to estimate raptor nesting success. Journal of Wildlife Management, 46, Steenhof, K. & Kochert, M.N. (1988) Dietary responses of three raptor species to changing prey densities in a natural environment. Ecology, 57, Steenhof, K., Kochert, M.N. & Moritsch, M.Q. (1984) Dispersal and migration of southwestern Idaho raptors. Journal of Field Ornithology, 55, Swenson, J.E., Alt, K.L. & Eng, R.L. (1986) Ecology of bald eagles in the greater Yellowstone ecosystem. Wildlife Monographs, 95, 46 pp. Taylor, I. (1994) Barn owls: Predator-Prey Relationships and Conservation. Cambridge University Press, Cambridge. Tjernberg, M. (1983) Prey abundance and reproductive success of the golden eagle Aquila chrysaetos in Sweden. Holarctic Ecology, 6, Tomback, D.F. & Murphy, J.R. (1981) Food deprivation and temperature regulation in nestling ferruginous hawks. Wilson Bulletin, 93, US Department of the Interior (1979) Snake River Birds of Prey Special Research Report. Bureau of Land Management, Boise, Idaho. Village, A. (1981) The diet and breeding of long-eared owls in relation to vole numbers. Bird Study, 28, Village, A. (1986) Breeding performance of kestrels at Eskdalemuir, south Scotland. Journal of Zoology, London (A), 28, Watson, A. (1957) The breeding success of golden eagles in the north-east highlands. Scottish Naturalist, 69, Watson, J. & Langslow, D.R. (1989) Can food supply explain variation in nesting density and breeding success amongst golden eagles Aquila chrysaetos? Raptors in the Modern World (eds B.-U. Meyburg & R. D. Chancellor), pp World Working Group on Birds of Prey, Berlin, London & Paris. Weisberg, S. (1985) Applied Linear Regression, 2nd edn. John Wiley & Sons, New York. Wiens, J.A. (1977) On competition and variable environments. American Scientist, 65, Wiklund, C.G. & Stigh, J. (1986) Breeding density of snowy owls Nyctea scandiaca in relation to food, nest sites, and weather. Ornis Scandinavica, 17, Wilkinson, L. (1989) Systat: the Systemfor Statistics. Systat, Inc. Evanston, IL. Woffinden, N.D. & Murphy, J.R. (1977) Population dynamics of the ferruginous hawk during a prey decline. Great Basin Naturalist, 37, Received 11 December 1995; revision received 18 July 1996 Ecology, 66,