Numbers and distribution of the Great Cormorant in Iceland: Limitation at the regional and metapopulation level

Abstract We studied a metapopulation of great cormorant (Phalacrocorax carbo) in Iceland, using complete aerial censuses of nests in 25 years during 1975–2015. Age composition was estimated in 1998–2014 by ground surveys in September and February. Brood size was estimated from aerial photographs in 2007–2015. Weather, food, breeding habitat, and density were considered as explanatory variables when examining numerical and distributional changes in the cormorant metapopulation. In 1975–1990 total nest numbers changed little, very low numbers about 1992 were followed by an annual increase of 3.5% in 1994–2015. Total nest numbers were positively correlated with estimates of spawning stocks of cod and saithe and inversely related to the subpolar gyre index (SPG‐I). During the increase in 1994–2015, average colony size at first increased and then declined. Habitat use also changed: the proportion of nests on small rocky islets (skerries) at first declined, from 69% to 44% in 1995–2003 and then increased again to about 58% in 2012–2014. Habitat changes were probably a response to changed patterns of human disturbance. Breeding density, as nests per km2 sea <20 m deep, was rather uniform among five defined regions in 1975–1996. Thereafter, densities became much higher in two sheltered regions with kelp forests and in one mostly exposed region. A second exposed region remained low and in the third nest numbers declined markedly. Thus, carrying capacity was higher in sheltered regions where cormorant breeding had historically been depressed by human disturbance. Brood size varied little among regions but declined with the years from about 2.5 to 1.8. The proportion of juveniles in September (fecundity) declined in 1998–2015 from over 0.4 to 0.3 and was inversely correlated with year and nest numbers, if outlier years were excluded, suggesting resource limitation. Survival of juvenile cormorants in September–February was estimated at 0.471 ± 0.066 SE. Commercial fish stocks and climate indices were not correlated with the proportion of juveniles. Annual survival of adults (breeding and nonbreeding) was estimated from nest counts and age composition 1999–2014, as 0.850 ± 0.026 SE and showed no trend in 1998–2014. We conclude that the metapopulation of cormorants in Iceland was resource‐limited at two levels: fecundity at the regional and winter survival at the total level.


| INTRODUC TI ON
Birds are generally more visible than other vertebrates and so can be counted with some accuracy, at least during the breeding season (McKellar, Marra, Boag, & Ratcliffe, 2014). However, most birds are relatively mobile and spend much time outside a study area, during migration or because of central place foraging. These features complicate the interpretation of complex processes such as population regulation or limitation. Another challenge with local population studies is the question of demarcation of the study population in relation to a population or metapopulation over a wider range (Hanski, 1999). Ashmole (1963) initially hypothesized that food availability within a foraging radius of breeding colonies limited seabird populations.
Ashmole's hypothesis was at first restricted to tropical pelagic seabirds but has since been extended to Arctic seabirds and has found support in both empirical work and modelling (Elliott et al., 2009;Hemerik, Van Opheusden, & Ydenberg, 2014). Thus, resource availability often determines breeding distribution and nest location often is a compromize between safety and access to food resources.
Colonial seabirds generally are central place foragers, with the breeding colony serving as the central placement to optimize access to foraging grounds (Burke & Montevecchi, 2009;Christensen-Dalsgaard, May, & Lorentsen, 2018;Elliott et al., 2009;Sandvik et al., 2016;Shoji et al., 2015;Weimerskirch, 2007). Upon population growth, colonies fill up with more nesters and reach their capacity as all nest sites become occupied, which requires recruits to seek out new territories or occupy suboptimal nest sites at the original colony (Pyk, Weston, Bunce, & Norman, 2013). Ideally, population regulation should be studied for a number of consecutive years, with regard to both local breeding colonies or populations and the total or flyway population. Ashmole's (1963) halo may well be valid for population limitation on the restricted scale of foraging radius from a seabird colony.
Local resources may indeed apply as a limiting factor to any sum of colonies but on a larger scale, the population in question may be limited by conditions away from the breeding colonies in space or time.
More recently, workers on bird and mammal populations have begun to examine population limitation in open systems where breeding, staging or wintering sites of migratory populations may be important (Gardarsson & Einarsson, 1994, 1997Gill et al., 2001;Sherry & Holmes, 1996). This naturally leads to questions of spatial as well as temporal scale.
Bird populations all over the world have responded to climate change, either by changes in numbers or altered migration or nest initiation dates (Saether, Sutherland, & Engen, 2004;Stephens et al., 2016). Climate change has been implicated in seabird studies but climate indicators have varying relationships with indices for seabird species, for instance the North-Atlantic Oscillation index (NAO) and similar indices are either positively or negatively related to timing of breeding, or have seemingly no effect (Moe et al., 2009;Wanless, Frederiksen, Walton, & Harris, 2009). Food often limits breeding birds (Martin, 1987;Newton, 1980) but few studies have considered food and climate variables simultaneously. In Norway, researchers found that fish abundance was relatively more important for European shags (Phalacrocorax aristotelis) than climate variables (Bustnes, Anker-Nilssen, Erikstad, Lorentsen, & Systad, 2013;Lorentsen, Anker-Nilssen, Erikstad, & Røv, 2015). In Eurasian wigeon at Mývatn, Iceland, the production of young was positively related to food abundance and negatively to snaps of cold and wet weather (Gardarsson & Einarsson, 1997).
In addition to climate change, that is, warming trends, there have been changes in the oceanic currents within the Northern Atlantic Ocean in recent decades which have affected flow of nutrients (Hátún et al., 2016). The subpolar gyre index (SPG-I) was colinear to the NAO until 1995 but the two became de-coupled in 1995 and the SPG-I has shown consistently negative values after that event (Berx & Payne, 2017). The subpolar gyre affects flow of nutrients within the ocean, including phosphate, nitrate and silicates (Hátún et al., 2017(Hátún et al., , 2016Johnson, Inall, & Häkkinen, 2013). SPG-I is highly correlated to body size and body mass in Icelandic arctic foxes (Vulpes lagopus), presumably because the SPG-regulated ocean forces affect food availability to the foxes, particularly in coastal habitats in west Iceland where seabirds are an important part of the fox's diet (Yom-Tov, Hersteinsson, Yom-Tov, & Geffen, 2017).
Seabirds often occur as groups of colonies that form metapopulations over large areas but only exceptionally have these been subjected to coordinated demographic studies (Oro & Ruxton, 2001).
It would appear that an ideal study population would be a whole metapopulation where the responses of many local populations can be examined and interpreted in relation to the whole.
We present a long-term study of an isolated, colonially nesting seabird population, great cormorants of the Atlantic subspecies (Phalacrocorax carbo carbo) breeding in Iceland. We chose to study 8. Annual survival of adults (breeding and nonbreeding) was estimated from nest counts and age composition 1999, as 0.850 ± 0.026 SE and showed no trend in 1998. We conclude that the metapopulation of cormorants in Iceland was resource-limited at two levels: fecundity at the regional and winter survival at the total level.

K E Y W O R D S
climate, disturbance, fecundity, food, regional and total population limitation, seabird, survival this metapopulation because (a) it is relatively small and isolated from other metapopulations of the same species, the nearest of which are found in Scotland and Greenland, 800 and 1,200 km away, respectively, (b) nests are in small but conspicuous colonies and the entire breeding population can be censused accurately from the air, (c) it can be surveyed in coastal waters throughout the year, (d) age categories can be distinguished in the field (Figure 1), making it possible to observe some demographic features with relatively small effort.
By estimating breeding numbers, fecundity, and distribution on a regional and total scale we hope to gain insight into features, such as climate, food and disturbance, likely to influence the demography of local breeding populations and how these conform to the metapopulation. Like many other seabird populations, our study population has obviously been affected by a long history of human exploitation and disturbance, an influence that became more evident during the course of this study.

| Study population
The Atlantic subspecies of great cormorant (hereafter cormorant) is a coastal seabird ranging across northern North-Atlantic shores from northwestern Europe to west Greenland and eastern North America (Cramp & Simmons, 1977;Hatch, Brown, Hogan, & Morris, 2000). In Iceland, cormorants mostly occur in shallow coastal waters <20 m deep (Gardarsson, 2008). This habitat covers about 6,900 km 2 unequally distributed along the coast of Iceland, with about half in the two west coast bays where almost all cormorants bred during the study period (Gardarsson & Jónsson, 2013), that is, Faxaflói and Breiðafjörður ( Figure 2). The two bays are separated by the 90 km long and mountainous Snaefellsnes peninsula (20-25 km wide).
Cormorants are large birds that seem to be of limited interest to avian predators but human exploitation of large colonial birds is widespread. In the low islets of Icelandic coasts, young cormorants were heavily exploited for meat. Early records suggest that, from the late 18th century until sometime in the early 20th century, the breeding distribution of cormorants was quite different from today, that is, colonies occurred mainly on the north and east coasts of Iceland (Faber, 1822;Hantzsch, 1905;Mohr, 1786). While low islets are numerous on the west coast, such islets are scarce elsewhere and cormorants breeding on north and east coasts (see Coastal wildlife and associated natural resources, historically, were an important source of livelihood but their importance declined during the 20th century. Collection of eider down and eggs, along with sheep farming (livestock had to be tended and moved among islets), remain widespread activities but can cause disturbance to the wary cormorants which respond by nesting on islets shared by few other species and thus less attractive to humans. Boat traffic, often associated with fishing near the cormorant colonies, is also a potential source of human disturbance but we did not assess it. Physical catastrophes occur on rare occasions, most often attributable to coinciding high tides and high winds.
In July-September, many cormorants disperse from the west coast breeding colonies and through the winter they are found along the whole coastline of Iceland (6,000 km long). By March, the adults return to the breeding colonies and begin to build nests that last until July. The main laying period is in late April through May. Cormorants

| Islets and colonies
Colony was our basic unit used for counting and recording breeding cormorants, defined topographically as a group of nests occupying an islet or a closely aggregated group of islets (within 100 m of each other) emerging from a common subtidal or intertidal shelf. Islets averaged 0.66 ± 0.11 (SE) ha in area (range 0.04-6.49 ha, n = 84), varied in soil cover but were devoid of woody vegetation. While each colony is topographically well defined and recorded as such, nests often moved between nearby islets from year to year within colonies (and probably among colonies also); this was rather obvious when the distance was small (e.g., <2 km) but probably could not be definitely detected at longer distances.
We grouped breeding islets according to soil cover and nesting seabirds into (a) rocky islets (or skerries) mean area 0.34 ± 0.05 ha (SE, n = 47) with almost no soil (<10% soil cover) and very few other colonially breeding seabirds except cormorants; and (b) grassy islets, mean area 1.07 ± 0.22 ha (SE, n = 37) with extensive soil (mean 38 ± 5% cover) and various breeding colonial seabirds, including the commercially valuable common eider (Somateria mollissima). We assume that human disturbance is likely to be minimal on the rocky islets (which are slippery and difficult to walk on) and more frequent on the grassy islets.
Nests in one or more groups of nests on a single islet, or sometimes a tightly packed group of islets were recorded as separate colonies.

| Breeding colonies and regions
The west coast shallow sea and islets are conveniently divided into five regions based on topography and benthic communities, one in Faxaflói (region FAX) and four in Breiðafjörður ( Figure 2). The FAX colonies were on a narrow (mostly 1-4 km) offshore belt of islets stretching along the northeastern shore of the bay. FAX (886 km 2 <20 m deep) was mostly exposed sandy shallows with rocky ridges Breiðafjörður is a complex bay, subdivided into two branches (Gilsfjörður and Hvammsfjörður), with large shallow areas (2,915 km 2 <20 m deep) and numerous islets and islands. The exposed parts are more open to the ocean swell and mostly ice-free in winter. Large areas are covered with coarse shell-sand and turf-or crust-forming algae and kelp forests are of limited extent (Gunnarsson, 1991). The inner sheltered regions have extensive kelp forests (which support relatively high densities of small fish, including young cod (Gadus morhua) and bullrout (shorthorn sculpin, Myoxocephalus scorpius), see for instance Keats, Steele, & South, 1987, Stål, Pihl, & Wennhage, 2007 and can become ice-covered in winter. These shallow areas of Breiðafjörður were divided topographically into four regions ( Figure 2): 1. BSW (462 km 2 ), a moderately exposed southwestern part with numerous small islets and a rather variable bottom. Ten colony sites were used, of which eight were skerries. BSW was mostly comprised of one group of colonies and there were small isolated colonies to the west and northeast.
2. BSE (378 km 2 ), the southeastern inlet of Hvammsfjörður, separated from the outer bay by an archipelago of densely packed islands; much of this sheltered fjord has a deep (20-40 m) soft bottom. Rock-lined channels with heavy tidal currents between the islands at its mouth support kelp stands. Fourteen colony sites were used, seven of which were skerries.
3. BNW (1,381 km 2 ), an exposed western part, bounded on the east side by a line between Skarð harbor and Skálmarnes. BNW has mixed bottom, including extensive tracts of shell-sand toward the northwest and some kelp stands in the south and east. Twenty-seven colony sites were used, 22 of which were skerries.
4. BNE (694 km 2 <20 m deep), a generally sheltered northeastern part (Gilsfjörður), north of Skarð and east of Skálmarnes, with extensive kelp forests and a mixed bottom. Fourteen colony sites were used, nine of which were skerries.

| Other relevant vertebrate species of the islands
The main marine mammals are harbor seal (Phoca vitulina) and gray seal (Halichoerus grypus), which potentially compete with cormorants for food. Animals that cause disturbance or depredate on seabird colonies, apart from man, include great black-backed gull (Larus marinus) and glaucous gull (L. hyperboreus), Arctic fox (Vulpes lagopus), white-tailed eagle (Haliaeetus albicilla), common raven (Corvus corax), and the introduced American mink (Neovison vison).

| Census of breeding colonies
Aerial photographic censuses of all known cormorant colonies in Iceland were carried out in 1975, 1983-1984, 1989-1990 and annually 1994-2015, usually in mid-May (Gardarsson, 2008;Gardarsson & Jónsson, 2013). The nest (usually, but not always, occupied) was the primary counting unit. Cormorant colonies were conspicuous from the air as white patches of bird excrement and were located and photographed using fixed-wing aircraft flying at airspeeds of about 100 knots (180 km/hr). The exact location of each colony was recorded, using maps, aerial photographs, satellite images and differential global positioning system (dGPS). Aerial observations were supplemented by ground truth obtained from a variety of written records, as well as observations supplied by ornithologists and local inhabitants.
Flight altitude (usually about 300-900 feet) and angle of view varied. Telephoto lenses (up to 300 mm) and a low angle of view were used for close views (Figure 4), for instance to distinguish cormorants from shags in mixed colonies and to estimate brood sizes.
A high or vertical angle yielded better pictures for accurate nest counts. Medium format (55 × 55 mm picture frame) cameras with diapositive color film (slides) were used in 1975-2005 but were replaced by digital cameras in 2006. Films were placed under transparent acetate and counted in a stereoscope, marking each nest with a fine needle; digital images were counted in a computer using the program SigmaScan ® .

| Food of cormorants
The cormorant is a generalist feeder. In Iceland, the main food (about half the diet) in all seasons and places 1996-2000 was the bullrout, a noncommerical species abundant in the kelp forests. Other important F I G U R E 4 Nesting cormorants and broods at Akureyjarsker (BSE) in Breiðafjörður, West Iceland. Aerial photograph taken during a brood survey 20 June 2013. Note how the cormorants are conspicuous against the white backdrop. The white color comes from the bird's droppings and both allows easy detection of colonies from the air as well as providing a convenient background for nest counts from aerial photographs cormorant foods were butterfish Pholis gunnellus, cod, saithe Pollachius virens, plaice Pleuronectes platessa, wolf-fish Anarhichas lupus, lumpsucker Cyclopterus lumpus, and the spider crab Hyas araneus (Lilliendahl & Sólmundsson, 2006). No abundance indices are available for the bullrout but the relative importance of each food fish may vary annually and prey choice probably depends more on fish size than species (Cech, Cech, Kubecka, Prchalova, & Drastík, 2008;Dias, Morais, Leopold, Campos, & Antunes, 2012;Gustavsen, 2017;Magath, Abraham, Helbing, & Thiel, 2016). We are not aware of any potential diet changes in cormorants during our study period 1975-2015.
The stocks of the commercial species cod and saithe have been  (Barrett, Røv, Loen, & Montevecchi, 1990;Gustavsen, 2017;Lilliendahl & Sólmundsson, 2006;Lorentsen, Grémillet, & Nymoen, 2004) whereas haddock and common ling were not listed among species eaten by cormorants in Iceland (Lilliendahl & Sólmundsson, 2006); and (c) the haddock and common ling indices added little informative variation to the PCA relative to those of cod and saithe, that is, common ling was highly correlated with cod (correlation matrix coefficient = 0.98) and haddock was highly correlated with saithe (correlation matrix coefficient = 0.77).

| Cormorant brood size in late June
Breeding success is a potential indicator of local habitat quality as well as a convenient measure of fecundity. The great cormorant is notoriously shy when breeding, making it difficult to study breeding success at close range by visiting the colonies. Breeding success in large samples of nests was studied, in late June in seven years, 2007-2009 and 2012-2015, using low-level aerial photography.
When disturbed by the approaching aircraft, some cormorants left their nests but most incubating birds and those with small chicks stayed, completely covering the nests. When the young were halfgrown (or more) the parent could no longer cover them and it became possible to count the number of large young (Figure 4). Nest contents were classified (eggs, young of various size classes). The relative size of the young was estimated from that of the nearby attending parents. Small young, up to about one-third size, could not be reliably counted; and thus, those were at least half-grown and up to fully grown were used to estimate brood size. When the young reached full size, they began to wander away from the nest and to form crèches and thus became less countable again. During the brood surveys, an average of 66% (range 51%-81%, n = 10,852) of the nests contained countable broods.

| Age composition in autumn and winter
In autumn (September) and in late winter (February), starting in 1998, age composition was surveyed (using spotting scopes) in the field in (a) southwest Iceland (several localities between Stokkseyri and Akranes), (b) at Snaefellsnes in west Iceland, and (c) in Húnaflói, north Iceland, mainly from Hólmavik to Vatnsnes. Study sites were selected on the basis of accessibility, distance and road connections. Most survey sites were situated outside breeding areas. About 3%-5% of the estimated total population was assigned to age class in each survey. In September, the proportion of juvenile cormorants was usually higher in the relatively accessible southwest than elsewhere in Iceland, leading to possible bias in the age composition.
For the purpose of estimating age composition of the population in September and February, we used the geometrical means of two regions, southwest (survey area 1) and northwest (survey areas 2 and 3). In September, juveniles (pied brown with a variable amount of white below) were distinguished from adults (all black). In February, three categories were distinguished: juveniles (as before), adult nonbreeders (all black, without filoplumes), and full-plumaged adult breeders (white thigh patches, white filoplumes on head and neck, nuchal crest) (Cramp & Simmons, 1977;Hatch et al., 2000; Figure 1).
We assume that the number of adult breeders equals approximately two times the numbers of nests counted in May of the same year.  (2016) Center for Atmospheric Research Staff, 2016) often is used to explain changes in species abundances (Hátún et al., 2009); and (c) oceanic changes such as strength of the Subpolar Gyre reflect changes in the relevant ecosystems (Berx & Payne, 2017;Hátún et al., 2016). In addition to these climate indices, we used averaged monthly temperatures for January and February in Stykkishólmur (Icelandic Meteorological Office, 2018) as our local winter temperature index.

| Statistical analyses
We evaluated annual trends in the data using linear regression. We compared correlations among regional nest numbers to test for spatial synchrony (using natural log (ln) of regional nest numbers) among our five regions (ten possible pairings) and applied false discovery rate (FDR) significance thresholds (which lowers p-value thresholds below α = 0.05, scaled with number of comparisons made) to consider spatial correlations simultaneously (Pike, 2011).
We used a generalized linear mixed model (PROC GLIMMIX, SAS Institute Inc., Cary, NC, USA) to evaluate relationships of weather and fish stocks to total nest numbers and proportion of juveniles in September, following an approach outlined in Jónsson, Lúðvíksson, and Kaller, (2017). This method includes variation due to year (autocorrelation) as a random effect in the model. Explanatory variables were climate indices SPG-I, NAO, AMO, average winter temperatures for the months of January and February, and spawning stocks of cod and saithe. We used backwards model selection to identify the variables which were related to our dependent variables (total nest numbers and proportion of juveniles in September). Total nest numbers were analyzed at year lags 0-5 to estimate effects on recruitment into the breeding population because cormorants generally begin breeding 3-5 years old (Bregnballe, 2006;Cramp & Simmons, 1977;Frederiksen & Bregnballe, 2001;Janiszewski, Minias, Lesner, & Kaczmarek, 2017). We saw no biological reason to lag proportion of juveniles in September. As with the spatial correlations, we applied FDR a posteriori to the outcomes of the 36 tests (six explanatory variables and six time lags).
To compare effects of year and region on productivity, brood size data (2007, 2008, 2009 and 2012, 2013, 2014, 2015) were analyzed with a multivariate analysis of variance (MANOVA), with frequencies of brood sizes (1, 2, 3, 4 and 5) within a year (scaled by sample sizes) as response variable and year and region as explanatory variables.
There were not enough degrees of freedom to test the year*region interaction. We report average brood sizes for these years; findings were the same between statistical tests on effects of year and region on average brood size and frequencies of brood sizes.

| Total nest numbers
Nearly all breeding colonies were in two large bays on the west coast, the northern half of Faxaflói, with 17% of the cormorant nests in 1975 and 29% in 2015, and Breiðafjörður with 83% of the total in 1975 and 69% in 2015. Only three complete counts were available in 1975-1990, among which total numbers were approximately stable, around 3,000 nests (Figure 6a). When annual counts began in 1994, nest numbers had recently declined sharply in all five breeding regions. In 1995, there was an all-time recorded low of 2,376 nests; after that numbers increased, reaching a high of 5,752 nests in 2014 (Figure 6a). During 1994-2015 the rate of increase was loglinear and averaged 0.035 ± 0.003 SE per year (r 2 = 0.88, p < 0.001).
In the first few years, the rate of increase was higher, for example in 1994-2001 r was 0.076 ± 0.065 (r 2 = 0.90, p < 0.001).

| Total nest numbers in relation to environmental variables
Backwards stepwise model selection indicated that cod and saithe indices were positively correlated to total nest numbers for 4 and 3 of 6 time lags, respectively (Table 1). SPG-I was inversely correlated to total nest numbers for 3 of 6 time lags (Table 1). Effects of cod and SPG-I were relatively immediate (lags 0-3 and 0-2) compared to effects of saithe which were more delayed (lags 3-5). NAO and local winter temperature were inversely correlated to total nests numbers for 1 lag each (Table 1). with any of the four regions in Breiðafjörður. Regional trends in total nest numbers did not directly reflect change in the whole metapopulation and were apparently influenced by regional factors.
In BSE, the inner sheltered part of southern Breiðafjörður

| Brood size surveys in late June
Brood size (i.e., the number of half-full grown chicks in success-

| Proportion of juveniles in September 1999-2014
In September, juvenile cormorants are distinguishable from older (1 year+) birds but breeding and nonbreeding adults look alike; and thus, the proportion of juveniles out of the total number of cormorants is therefore an underestimate of fecundity. The proportion of juveniles averaged about 0.31 and was generally higher on the SW coast (mean 0.45) than on the N coasts (mean 0.23). The geometrical mean proportion of juveniles declined during the study pe- Proportion nests on skerries (=rocky islets with <10% soil cover), see text for details the years, and with density, suggesting weak density dependence, however the relationship of (Y) proportion of juveniles with (X) year was stronger than with (X) density. Also, there were four outliers (2002, and 2005-2007), which suggest that density per se was not limiting the production of juveniles and could be a result of more than one limiting resource, such as two or more demersal fishes for which stock models were not available. A suggestion of density de-

| Age composition in February 1999-2014
Three age groups could be distinguished in February (late winter).

| Estimates of annual survival of age classes
The proportional age composition in February was combined with the absolute number of breeding adults estimated from the nest numbers in May to yield a crude estimate of numbers of the three age classes and the annual survival of adult cormorants in 1999-2014, 0.850 ± 0.026 (Table 2). The calculated number of juveniles each September, compared with the estimated number of juveniles in the following February, yielded an estimate of average juvenile winter (September-February) survival of 0.471 ± 0.066 (SE) ( Table 3).
Crude estimates of the total metapopulation, based on the three age groups distinguishable in the field in February combined with nest numbers, were used to explore the survival pattern of these groups through the period 1998-2015 ( Figure 11). This comparison shows that the adult breeding population was steadily increasing through the period (as shown also by the unabated annual increase of 3.5% in overall nest numbers; Figure 11a).  Cormorants, like many seabirds, were heavily exploited by man in the past and the population responded by altering its breeding habitat and geographical locations. A similar change in nesting habitat during the 20th century occurred with the white-tailed sea-eagle which switched from high cliffs to low coastal islets or hillocks, clearly in response to reduced persecution (Skarphéðinsson, 2003).

| D ISCUSS I ON
This reduced exploitation of the coastal resources is likely to have affected the breeding cormorant population in recent decades although mostly before the beginning of the present study, which saw the end of an apparent shift in breeding distribution from the north and east coast to the west. This shift toward west coast shallow seas and islets to cormorant colonies probably corresponded to an ideal-free distribution (Fretwell & Lucas, 1970;Røv, 1994). During such a response phase, first-time breeders probably will seek out new colonies to avoid competition with experienced adults; once such recently undisturbed colony sites become occupied, dispersal is more likely to occur in an ideal-free manner (Hénaux, Bregnballe, & Lebreton, 2007;Péron, Lebreton, & Crochet, 2010).
During our study, density of breeding regions varied strongly, apparently in response to local resources rather than changes in the total metapopulation. Thus, some form of local density dependence seems to limit breeding numbers at the regional scale (Frederiksen & Bregnballe, 2000;Røv, 1994), but the total breeding population is so far increasing steadily at the annual rate of 3.5%. This rate of increase is not shown by the young and nonbreeding sectors of the total metapopulation as estimated at the end of winter, suggesting that there may be another limitation to the numbers which may become more visible in the future, if the increase in nest numbers continues.

| Total nest numbers and changes in colony size and habitat distribution
The first census in 1975 and following estimates in 1983-1984 and 1989-1990

| The role of climate change and its possible interaction with food
The shift in SPG-I values from negative to positive in 1995 also coincided with the increased trend in total nest numbers after a low in 1991-1994. The negative trend in SPG-I during this study period follows from an event in 1995 when the NAO and SPG-I became uncoupled (Hátún et al., 2016), so it is conceivable that the generally favorable conditions for cormorants in Iceland stem from mild climate and improved feeding stocks, which in turn are associated with favorable oceanic conditions. A strong subpolar gyre (

| Historical changes related to human activity prior to 1975
Almost all cormorant nests were found within Faxaflói and  (Faber, 1822;Hantzsch, 1905;Mohr, 1786), which were then colonized mainly in the 20th century. Before this study, there may have been a period of relatively high nest numbers in the 1950s followed by a low in the 1960s. Historically, many cormorant colonies were found on the north, east and south coasts of Iceland ( Figure 2). However, these breeding sites had already been abandoned by the onset of this study in 1975. At present, the cormorant population probably is in a transitional stage, that is, from a period of heavy exploitation and disturbance of breeding sites to a period of decreased human disturbance and markedly TA B L E 2 Estimated age composition of the Iceland cormorant population in 1999-2014  In 1975In , 1983In -1984In and 1989In -1990, nest densities in all study regions were relatively low, mostly about 0.5-0.8/km 2 shallow sea <20 m. The annual increase in total nest numbers of 3.5% in 1994-2015 was not equally distributed through the breeding range in western Iceland; each region maintained partly independent changes in nest densities. In Faxaflói (FAX) changes in nest density

| Regional variation in nest densities
were not correlated with those in Breiðafjörður. In region BSW, nest densities declined from 1983 to 2008 when a slight increase (to 0.5) was noted. In the outer, most exposed Breiðafjörður region (BNW) most colonies increased during 1998 to about 2002 but after that densities levelled off or declined. Simultaneously, nest densities were increasing in the inner sheltered parts of Breiðafjörður (regions BNE + BSE). Asymptotic nest densities were generally higher in the sheltered (about 2 nests per km 2 ) than the exposed regions (about 0.8 nests per km 2 ). There are two possible explanations for this difference between outer and inner parts of Breiðafjörður: 1. The outer, more exposed regions are most likely subject to more variation in wave action which could lead to benthic up-  period 1975-2006, peaking in 1985-1993 and then decreasing (Petersen, 2001;A. Petersen, personal communication).

2.
Differences in wave stress in the regions could also be expressed through historical changes in human exploitation of cormorants.
We have very little direct information on the 19th -early 20th century recolonization of the west coast by cormorants, during a time when the breeding islets were heavily exploited by the human population, who largely abandoned these islands in the 20th century. Presumably the outer exposed islets were less accessible to cormorant harvesters and the inner islets more accessible to humans using small open boats. The outermost islets were very important as fishing stations before the mid-20th century and thus, were only colonized by cormorants shortly before the beginning of nest counts. Thus, it could be argued that the outer islets (BNW and BSW) were colonized earlier and by the time of this series of censuses, resources had become limiting in the exposed parts of Breiðafjörður but not yet in the sheltered parts (BNE and BSE).
The situation in the Faxaflói, which showed a continuing trend of increase during the study period, seems to lend support to both the explanations put forward above. The Faxaflói bay is more exposed to the ocean than the Breiðafjörður and thus probably more sensitive to benthic upheavals. There are previous records of large fluctuations in the numbers of cormorants nesting in Faxaflói (Gardarsson, 1996) and these could have been caused by fluctuating food resources.

| CON CLUS ION
At present, the total metapopulation increase of cormorant in west Iceland seems to be reaching carrying capacity, leading to colonization of other parts of the Icelandic coast, with no reduction in the rate of increase in the total breeding metapopulation. Simultaneously, the composition of the total metapopulation is likely to shift toward fewer floaters and perhaps juveniles and nonbreeding adults. We expect that further increases in this population will gradually become limited by available nest sites. However, historical colony sites around Iceland remain unoccupied and the reduction of cormorant harvest will allow many of those to become safe. New colony establishment occurs slowly but it is clearly the only way in which the present breeding population can increase further, because nest numbers at the regional level are limited by carrying capacity (nest density in relation to food resources). However, during the present study which has lasted 41 years, only two new colonies have been documented outside the continuous west coast range; presumably the expansion is restricted by behavioral constraints (especially philopatry) to remain near the natal or previous breeding colonies, or simply by dispersal distances; hence, most new colonies become established within the present range (see also Hénaux et al., 2007).
Finally, it seems likely that the age group structure of the Icelandic cormorant population will change as the total carrying capacity becomes more limiting, and that change in the level of small-fish availability will continue to have a key role in limiting the population level.
For guidance in environmental and biological matters we are especially indebted to Karl Gunnarsson, Sarah Wanless and Thomas Bregnballe. We thank Sigmundur Helgi Brink for preparing the map in Figure 2 and Erling Ólafsson for photos of cormorant in Figure 1.
We thank the editors and 2 anonymous reviewers for comments that improved an earlier version of this manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N S
Conceived and designed the study: AG. Analyzed the data: AG, JEJ.
Drafted and wrote the paper: AG JEJ.