Gyrfalcon Polulation and Reproduction in Relation to Rock Ptarmigan Numbers in Iceland
In courtesy of Dr. Olafur K Nielsen and The Peregrine Fund
Icelandic Institute of Natural History, Urriðaholtsstræti 6 –8, P.O. Box 125,
212 Garðabær, Iceland. E-mail: email@example.com
I have studied the population ecology of the Gyrfalcon (Falco rusticolus) since 1981 on a 5,300-km2 study area in northeast Iceland harbouring 83 traditional Gyrfalcon territories. The main questions addressed relate to the predator–prey relationship of the Gyrfalcon and its main prey, the Rock Ptarmigan (Lagopus muta). Specifically, how does the Gyrfalcon respond functionally and numerically to changes in ptarmigan numbers? Field work involved an annual census to determine occupancy of Gyrfalcon territories, breeding success, and food. Also, Rock Ptarmigan were censused annually on six census plots within the study area. The Rock Ptarmigan population showed multi-annual cycles, with peaks in 1986, 1998 and 2005. Cycle period was 11 or 12 years based on the 1981−2003 data. The Gyrfalcon in Iceland is a resident specialist predator and the Rock Ptarmigan is the main food in all years. The functional response curve was just slightly concave but ptarmigan densities never reached levels low enough to reveal the lower end of the trajectory. Occupancy rate of Gyrfalcon territories followed Rock Ptarmigan numbers with a 3−4 year time-lag. Gyrfalcons reproduced in all years, and all measures of Gyrfalcon breeding success—laying rate, success rate, mean brood size, and population productivity—were significantly related to March and April weather and to Rock Ptarmigan density. The Gyrfalcon data show characteristics suggesting that the falcon could be one of the forces driving the Rock Ptarmigan cycle (resident specialist predator, time-lag). The Iceland Gyrfalcon and Rock Ptarmigan data are the longest time series on the Gyrfalcon–Rock Ptarmigan relationship, and the only series that indicate a coupled predator–prey cycle for the falcon and its ptarmigan prey. Received 20 April 2011, accepted 28 July 2011.
NIELSEN, Ó. K. 2011. Gyrfalcon population and reproduction in relation to Rock Ptarmigan numbers in Iceland. Pages 21–48 in R. T. Watson, T. J. Cade, M. Fuller, G. Hunt, and E. Potapov (Eds.). Gyrfalcons and Ptarmigan in a Changing World, Volume II. The Peregrine Fund, Boise, Idaho, USA. http://dx.doi.org/10.4080/gpcw.2011.0210
Key words: Iceland, predator−prey relationship, time-lag, numerical response, functional response.
THE GYRFALCON (FALCO RUSTICOLUS) is a large diurnal raptor with a circumpolar, arctic and subarctic distribution (Cramp and Simmons 1980, Cade et al. 1998, Potapov and Sale 2005,
Booms et al. 2008). The breeding habitat is generally harsh, and winter conditions are such that in parts of its range the Gyrfalcon is migratory (Poole and Bromley 1988, Burnham 2007). In other parts, it is sedentary with just local movements of juvenile birds (Nielsen and Cade 1990a), and in between are partly migratory populations (Platt 1976). The Gyrfalcon is a specialized predator, and within most of its range the main food is one or two species of ptarmigan, the Willow Ptarmigan (Lagopus lagopus), and the Rock Ptarmigan (L. muta) (Figure 24.9. in Nielsen 2003). The Gyrfalcon is unusual among raptors that prey upon other birds; they go through the breeding season preying on the adult segment of the ptarmigan population at a time of year when ptarmigan abundance is lowest (Hagen 1952a, Cade 1960). The most common pattern among other diurnal raptors that prey upon birds is to time their breeding so that the period of peak food demand—the nestling period—coincides with fledging of their prey (Newton 1979, Newton and Marquiss 1982b) to capitalize on this annual peak in prey abundance. In contrast, it has been postulated that the timing of the Gyrfalcon’s breeding period matches the dispersal period of juvenile Gyrfalcons with the annual peak in number of the most vulnerable ptarmigan (Cade 1960, Nielsen 2003). Ptarmigan chicks enter the diet of the Gyrfalcon when the ptarmigan are c. four weeks old, and it is only when they are c. 80 days old that they have grown their primaries and acquired full flying capabilities. In Iceland, this window of annual peak abundance of newly fledged, naïve Rock Ptarmigan with impaired flying capabilities lasts from late July to early September, coinciding with the dispersal period of juvenile Gyrfalcons.
Some ptarmigan populations, sympatric with Gyrfalcons, have multi-year, cyclic fluctuations of numbers. In Fennoscandia, the Willow Ptarmigan has traditionally had 3to 7-year cycles (Myrberget 1984, Andrén et al. 1985, Angelstam et al. 1985), with 10-year cycles in the eastern part of Siberia and in North-America (Keith 1963, Bergerud 1970, Andreev 1988, Bergerud 1988). Rock Ptarmigan in Iceland (Nielsen and Pétursson 1995) and Alaska (Weeden and Theberge 1972) have c. 10-year cycles. Similar cycles are known for many other species of Arctic and boreal herbivores (Elton 1924, Keith 1963). Many believe these periodic fluctuations to be caused by trophic interactions (Berryman 2002), either between the herbivore and its food source (Bryant and Kuropat 1980, Bryant 1981b, Bryant 1981a), or the herbivore and its predators (Hanski et al. 1993, Krebs et al. 1995, Korpimäki and Krebs 1996, Turchin et al. 1999), or its pathogens (Berryman 1996, Hudson et al. 1998, Holmstad et al. 2005a, Holmstad et al. 2005b). It goes without saying that a specialist predator such as the Gyrfalcon, utilizing a cyclic prey base, is faced with greatly contrasting food regimes, alternating between years of plenty and scarcity. It has been assumed that these large scale changes in the Gyrfalcon’s prey base are reflected in the population size and fecundity of the raptor (Palmer 1988, Cade et al. 1998, Booms et al. 2008).
Studies have been done on the Gyrfalcon−ptarmigan relationship in the Yukon Territory (Mossop and Hayes 1994) and the Northwest Territories (Shank and Poole 1994) of Canada, in two areas in Norway (Tømmeraas 1993, Selås and Kålås 2007), in Sweden (Nystrom et al. 2005), and in Iceland (Nielsen 1999). In Iceland, the numerical and functional response of the Gyrfalcon to changes in Rock Ptarmigan density was studied from 1981 to 1997 (Nielsen 1999). In this paper, I use these same data series extended to 2010 to describe how changes in Rock Ptarmigan density affect food composition of the Gyrfalcon, how populations of the two species relate to each other in time, and finally how Rock Ptarmigan numbers and weather interact with respect to fecundity of the Gyrfalcons.
Study Area.—The study area covers 5,327 km2 located in northeast Iceland and centered on Lake Mývatn (N65°60′, W17°00′). This area is characterized by rolling hills rising from the coast to 600–800 m above sea level at the southern border, 100 km inland. Several glaciated valleys, isolated mountains, and larger mountain masses break this relief (highest peak 1,222 m a.s.l.). Two major glacial rivers border the study area, Skjálfandafljót in the west, and Jökulsá á Fjöllum in the east.
The climate of Iceland is maritime with cool summers and mild winters. July is the warmest month and February the coldest. The mean temperature for these months respectively from 1981−2010 on the study area was 9.5°C and −1.0°C on the coast at Mánárbakki on Tjörnes Peninsula, and 10.6°C and −3.7°C inland at Reykjahlíd, Lake Mývatn. Average annual precipitation decreases from the coast inland. It was 598 mm from 1981−2010 at Mánárbakki, and 464 mm at Reykjahlíð. At Lake Mývatn, the last frost in spring on average was on 12 June, and the first frost in autumn was on 5 September (Eythorsson and Sigtryggsson 1971, Einarsson 1979, Icelandic Met Office http://www.vedur.is).
The dominant vegetation types are heath and meadow vegetation, which cover 3,003 km2. Other important types are wetlands of various kinds, 327 km2; Downy Birch (Betula pubescens) woods and shrubs, 156 km2; and moss heath, 2 km2. Sparselyor un-vegetated land covers 1,659 km2, and lakes and rivers, 180 km2 (Vegetation map of Iceland, 1:500,000, Icelandic Institute of Natural History). Important heath plants include species belonging to the Ericaceae, the Dwarf Birch (Betula nana), the Tea-leaved Willow (Salix phylicifolia), and various species of grasses (Poaceae), sedges (Carex spp.), moss, and lichens. Three species of terrestrial mammals—Arctic Fox (Vulpes lagopus), Mink (Mustela vison) and Wood Mouse (Apodemus sylvaticus)—and 61 avian species breed in the area.
The avifauna is characterised by large populations of waders, waterfowl, and seabirds. In summer, the Rock Ptarmigan is common on heath and grassland habitats. Winter habitats include alpine areas, lava fields and birch shrubs (Nielsen 1993). Natural predators of adult Rock Ptarmigan are the Gyrfalcon, Raven (Corvus corax), Arctic Fox, and Mink. The Rock Ptarmigan is a game bird and is harvested in autumn; in 2010 the open season was 18 days from 29 October to 5 December. The Rock Ptarmigan is the only ptarmigan species breeding in Iceland.
Rock Ptarmigan Population.—Each spring, territorial Rock Ptarmigan cocks were counted on six plots within the Gyrfalcon study area. The combined size of these plots was 26.8 km2 (range 2.4–8.0 km2). Each plot was censused once during May (mean date 20 May, SD = 5.49, range 7 May—6 June). The census was usually conducted by two observers in the late afternoon (time 17:00–24:00) or the early morning hours (time 05:00–10:00). The positions of territorial cocks were plotted on a map, as were the locations of all kills. A “kill” was the remains of a Rock Ptarmigan killed and eaten after arrival on the census plot in spring. The total number of cocks in spring was measured as the sum of the number of territorial cocks censused and killed (for a detailed description of census plots and methods, see Nielsen 1996). The Rock Ptarmigan index used for this study was the annual mean density of cocks on the six plots.
Gyrfalcon Population.—Gyrfalcons have traditional nesting territories (Tømmeraas 1993, Burnham et al. 2009). The term “nesting territory” refers to the area defended by the Gyrfalcons around the nest, usually only the nesting cliff and its immediate surroundings. The nesting territory was the unit surveyed in this study. Information on the location of Gyrfalcon nesting territories in the study area already existed at the start of the study in 1981, from local knowledge and the archives of the Icelandic Institute of Natural History (IINH). A nesting territory always had several nest sites. A “nest site” referred to the actual location of the nest. Nest sites belonging to the same nesting territory could be on different cliffs, with the territorial pair alternating between them in different years. Whether adjacent nest cliffs were regarded as belonging to the same nesting territory or not was based on their history of occupancy. By definition, only one Gyrfalcon pair could use a specific nesting territory at any one time. At the start of the study in the 1980s, there were 81 traditional Gyrfalcon nesting territories recognized in the study area (Nielsen 1986). By 2010, there were 83 Gyrfalcon nesting territories recognized in the study area. Nest sites in two different nesting territories (3.4 and 4.2 km apart respectively), and used by the same banded females at the start of the study, have since been split among four pairs.
Field work was conducted between May and mid-August from 1981 to 2010. During the first visit in spring, the nesting territory was determined either as “unoccupied” if no signs of Gyrfalcons were found at any of the known nest sites, roosts, and perches within the territory, or “occupied” if definite signs of activity were observed. To be classified as occupied, the territory had to have a breeding pair, or in case of non-breeders, an active roost with some combination of bird sightings, new food remains, fresh droppings, moulted down and feathers, or pellets. Occupied territories could hold: (1) successful breeding pairs; (2) unsuccessful breeding pairs; (3) non-breeding pairs; and (4) unknown occupants. Successful pairs fledged at least one young. Unsuccessful pairs laid eggs but failed at some stage before fledging young; proof of breeding was observation of eggs or chicks, or an incubating adult. Territories with non-breeding pairs included territories where sightings were made of adult pairs or where proof was found of courtship feeding. Territories with unknown occupants included territories where clear signs of occupancy were found but no indications of courtship activities and only single birds observed. Classification into any of the four occupancy categories could be based on signs alone. Brood size was determined when the nestlings were 4–6 weeks old. Most of the nests were revisited after young had fledged. Nestlings found at that time, and dead since the last visit, where taken into account when calculating mean brood size.
I used the following variables to describe the relative size of the territorial Gyrfalcon population and its fecundity (cf. Newton and Marquiss 1982a, Steenhof et al. 1997, McIntyre and Adams 1999): (1) occupancy rate, the proportion of nesting territories surveyed that were occupied by Gyrfalcons; (2) laying rate, the proportion of occupied territories where eggs were laid; (3) success rate, the proportion of laying pairs that fledged chicks; (4) mean brood size, the average brood size at fledging for successful pairs; and (5) population productivity, mean number of fledglings per occupied nesting territory, calculated as the product of variables no. 2, 3 and 4 above. Population data for 1993 were not included in the current analysis as only 48 territories were visited, the territories visited were not chosen randomly, and they were biased towards “good” territories.
Gyrfalcon Diet.—Gyrfalcons started to bring prey to the nesting territory at the onset of courtship feeding in the 2nd half of March to the 1st half of April. During courtship, laying, and incubation, prey remains accumulated at plucking locations within the territory some distance away from the nest site itself. These plucking sites were usually on top of the nesting cliff, on the slope below it, or on either rim if the nesting territory was centered within a gorge. It was only after the young had hatched that prey remains started to accumulate in the nest. During the nestling stage, most prey remains were found in the nest or at the bottom of the cliffs below it. When the young fledged, they used many of the same plucking locations as their parents’ in spring, but also many new sites so prey remains were much more spread out than in spring.
The food study was based on collections of prey remains and pellets from successful nesting territories only. Nest sites were visited two or three times during the summer to collect remains. The final visit was always made after young had fledged. Prey collections from territories with aborted breeding attempts were excluded from this analysis, as were prey collections from successful territories where collecting was not possible within the nest itself. These exclusions were made because of predictable changes in prey selection over the course of a season (Nielsen 2003). The only way to have comparable samples of diet was to collect prey remains at all sites where they accumulated, and over the course of the entire breeding season, courtship through fledging. Complete prey collections were made at 71 nesting territories, representing 521 successful breeding attempts. A total of 45,014 individuals were identified in the prey collections.
A prey collection consisted mostly of skeletal remains, but also feathers and pellets. Identification of most prey items was done in the field. Problematic specimens were brought back to the lab and identified with the aid of a reference collection. Each prey collection was separated into species and age groups, and the minimum number of individuals belonging to each group found by counting the most frequently occurring bone representing one individual. For the Rock Ptarmigan, this was nearly always the sternum (Langvatn 1977), but for smaller species this item was more often either one of the wings or the feet. Feather pattern or bone structure was used to distinguish between young of the year and adults. Pellets were only analysed for legs and bones of small birds, mainly young waders and passerines. Masses of prey items were taken from IINH files, and Nielsen (1986, unpubl. data).
Gyrfalcon Functional Response.—The number of adult Rock Ptarmigan consumed by Gyrfalcons during the breeding season (PC) was calculated using the formula:
(FC + MC + YC) × PGD falcons = number of males × 111 days × daily food requirements (241 g day−1); YC = consumption by young = 54 days × daily food requirements (169 g day−1); PGD = proportion of adult Rock Ptarmigan biomass in the diet of Gyrfalcons; WG = mean weight of Rock Ptarmigan in spring 537 g (unpubl. data). The per capita consumption was calculated by dividing PC by the total number of falcons. It was assumed that all territories were occupied by pairs. Daily food requirements were calculated in accordance with Lindberg (1983). These figures included waste (estimated at 20%), and gross energy intake (estimated assimilation efficiency 70%). Daily energy expenditure was taken as 2.5 × basal metabolic rate (BMR) for adults and 1.7 × BMR for young. Weights used for calculations of BMR were 1,355 g for adult males, 1,831 g for adult females, 1,262 g for young males and 1,573 g for young females (Cade et al. 1998). The sex ratio of young was taken to be equal (Nielsen 1986), and consumption by young used in the calculations was the average value for the two sexes. The breeding season as represented by the food data was taken as 111 days long, including 14 days for courtship, 43 days for laying and incubation, 47 days for nestling period, and seven days post-fledging period. Both the courtship-period and the post-fledging period were longer than stated above. I used a conservative estimate of what fraction of those periods my food collections covered, namely the last two weeks of the courtship period and the first week of the post-fledging period. It was assumed that the estimated number of prey items consumed equalled the number hunted, i.e., there was no surplus killing. The data, per capita consumption, was graphed against Rock Ptarmigan densities and the Holling’s type two functional response function (Holling 1959) was fitted to the data using the solver application in Excel (Liengme 2000). Weather.—Weather data used in the analysis was collected during the years 1981−2010 at six weather stations within or bordering the study area (courtesy of the Icelandic Met
FC + MC + YC) × PGD
PC = ———————
WG × 100
￼where FC = consumption by female Gyrfalcons = number of females × 111 days (the length of breeding season) × daily food requirements (301 g day−1); MC = consumption by male Gyr-
Office http://www.vedur.is/). The stations were: Grímsstaðir, Mánárbakki, Mýri, Mývatn, Möðrudalur, and Staðarhóll. The weather variables studied were for the months January through April, or for the following combinations of months: January and February, January through March, January through April, February and March, February through April, and March and April. The variables were: (1) mean temperature (°C); (2) accumulated precipitation (mm); (3) days with wind ≥ 18 m per sec; (4) days with precipitation ≥ 1 mm; (5) days with precipitation as snow; (6) days with 100% snow cover; and (7) mean percentage snow cover.
Statistical Analysis.—All statistical tests were run using the software STATISTICA (http://www.statsoft.com/). Statistical significance level was set at P ≤ 0.05 for all tests. I used ANOVA to test for differences in Rock Ptarmigan density among years (6 observations × 30 years) and among decades (60 observations × 3 decades). The variance was = 1.206, P = 0.233), but the distribution was right-skewed and did not conform to normality. I used the natural logarithm (ln) to transform density values prior to analysis to normalize the frequency distribution. I used Chi-square to test for differences in occupancy rate, laying rate, and success rate among years. For success rate, cells (years) with fewer than five observations were added to adjacent cells prior to analysis. To test for differences in mean brood size among years I used ANOVA. To test for a relationship between occupancy rate, laying rate or success rate as dependent variables, and spring Rock Ptarmigan density and spring weather as explanatory variables, I used a logit regression with binomial distribution and logit link function (Hill and Lewicki 2006).
For model selection I used “best subset,” and the AIC criteria to select amongst models. Wald statistics were used to test the significance of the regression coefficients. Forward stepwise multiple linear regression was used to test for a relationship between mean brood size or population productivity (dependent variables), and spring Rock Ptarmigan density and spring weather (explanatory variables) (Hill and Lewicki 2006). The linear regression was evaluated with the F-test, and model parameters with the t-test. I used autocorrelation and partial autocorrelation, a time series analysis technique, to study the lag process in the Rock Ptarmigan time series (Chatfield 1989). Autocorrelation measured the degree of association between numbers in a time series. Autocorrelation coefficients plotted against their lag gave the autocorrelation function that provided an objective estimate of the dominant cycle period. Partial autocorrelation gave the correlation between current population and the population at some time in the future, but controlled for years between the points. Partial autocorrelation coefficients plotted against the lag gave the partial autocorrelation function and revealed the process order or the number of lags needed to model the population fluctuations. Similarly, I applied cross-correlations to assess the degree of temporal synchrony of the Rock Ptarmigan population and the Gyrfalcon occupancy rate, laying rate, and importance of Rock Ptarmigan in Gyrfalcon diet (Chatfield 1989). Correlation coefficients, rt, were calculated with different time-lags, t. Correlation coefficients calculated between synchronously fluctuating parameters yielded high positive values with t = 0 years, and rapidly decaying values with an increasing time-lag. Non-synchronous parameters did not correlate strongly. Fluctuating parameters in opposite phases yielded high negative correlation with t = 0 years. Prior to analysis, time series were detrended using trend subtract (x = x − (a + b × t)) in the Time Series module, and standardized to zero mean and unit variance. Data points for 1993 were interpolated using adjacent values.