Tómas G. Gunnarsson, School of Biological Sciences, University of East Anglia, Norwich, NR4 7TJ, UK. E-mail: firstname.lastname@example.org
1Buffer effects occur when increases in population size result in an increasing proportion of a population inhabiting poor quality sites. When there are fitness costs to inhabiting poor sites, buffer effects can potentially regulate population sizes. In migratory populations, the regulatory capacity of buffer effects will clearly be influenced by their role in both the breeding and nonbreeding seasons, but previous studies have been restricted to one season only.
2Icelandic black-tailed godwits Limosa limosa islandica are currently increasing in number and previous studies have revealed a large-scale buffer effect operating on the wintering grounds.
3Here, we reconstruct the pattern of population expansion and colonization of new breeding grounds, in relation to breeding habitat quality, to investigate whether a buffer effect is also operating during the breeding season.
4Godwit breeding success is higher in marsh habitats than in dwarf-birch bogs. Prey densities are also higher in marsh habitats, and breeding success increases with the density of shallow pools, which are more common on marsh sites. Large lowland basins with higher marsh coverage were colonized earlier than small ones with low marsh coverage. Recent colonizations have been into basins that are closer to occupied sites and have higher cover of dwarf-birch bog.
5Thus godwits appear to be expanding into poorer quality breeding habitat as well as poorer quality winter habitat. The large spatial scale of these analyses and the fitness costs of occupying poor quality sites suggest that this double buffer effect is likely to play a key role in regulating this expanding population. In most migratory populations, some level of density dependence is likely to operate at both ends of the range. Double buffer effects may therefore be a common phenomenon and an important mechanism regulating migratory populations.
Population regulation can act both through density dependence within sites, due for example to resource depletion, interference competition, territoriality or dominance rankings (Sutherland 1996a), and through larger-scale changes in distribution resulting in poorer sites being used disproportionately at higher population densities (the buffer effect, Kluyver & Tinbergen 1953; Brown 1969). Although there are many studies showing within-site density dependence (e.g. Brooke 1979; Goss-Custard, Clarke & Durrel 1984; Newton, Rothery & Dale 1998), including those resulting from local buffer effects (Rodenhouse et al. 2003; Kokko, Harris & Wanless 2004), there are very few studies of large-scale buffer effects. This is largely due to the practical difficulties and extensive data requirements of studying these processes at large spatial and temporal scales. Species known to be changing in abundance and distribution provide an opportunity to explore the role of buffer effects in population regulation. In a buffer effect system, at low population sizes higher quality habitats are preferentially occupied but, as population size increases, an increasing proportion of the population occupies poorer quality habitats. If demographic processes vary with habitat quality, then expansion into poorer quality sites can lead to a reduction in average survival and/or fecundity in the population, which could eventually result in population regulation. The reverse pattern should result in local extinctions from poorest quality habitats first and highest quality habitats last. A range of studies have shown buffer effect patterns (expansion into or contraction out of particular habitat types with fluctuations in population size, e.g. Suter 1995; Summers et al. 1996; Chamberlain & Crick 1999) but the demographic implications of these patterns have rarely been established (Ferrer & Donazar 1996; Gill et al. 2001).
The size of migratory populations will be limited by the interactions between density-dependent processes operating at either end of the migratory range (Sutherland 1996b; Newton 2004). In populations where some level of density dependence exists at both ends of the range, separate buffer effects may operate on both the wintering and the breeding grounds (and potentially also on migratory stopover sites). A double buffer effect would potentially have greater implications for population regulation than buffer effects in one season only, as both survival and reproductive success would decline with increasing population size.
The Icelandic black-tailed godwit Limosa limosa islandica breeds almost exclusively in Iceland and winters on the coasts of western Europe (Gunnarsson et al. 2005), and has undergone an extensive population increase during recent decades (Prater 1975; Musgrove et al. 2001). At the start of the twentieth century, black-tailed godwits bred only in a limited area in southern Iceland (Slater 1901). By about 1930, the species had spread to new breeding areas in western Iceland (Gudmundsson 1951). Since then, the godwit population has increased greatly (current population size =c. 47 000 individuals, Gunnarsson et al. 2005) and the current breeding distribution covers most lowland areas in Iceland (Fig. 1). The reasons for this population increase are unknown.
Our previous studies have shown that a buffer effect is operating in the godwit population during winter; between 1970 and 1996, the population expanded into estuaries where prey intake rates and adult survival were significantly lower than on traditionally occupied winter sites (Gill et al. 2001). The reduction in mean survival, as a consequence of an increasing proportion of the population inhabiting poor quality winter sites, could clearly be compounded by a similar reduction in fecundity, if a buffer effect were also operating in the breeding season. Our objectives in this study were therefore to assess whether the changes in breeding distribution of this population also follow a large-scale buffer effect, with expansion into poorer quality breeding sites.
reconstructing the pattern of population expansion
Standardized surveys of biodiversity in Iceland are rare. However, the Icelandic Institute of Natural History in Reykjavík encourages and collects submissions of natural history records from local residents. Additions of new breeding species to the local fauna are frequently reported. We collated these records, together with records from published sources, in order to date the major patterns of colonization of breeding sites by black-tailed godwits. In total, 118 records of first or recent breeding were obtained from these sources. Recent breeding refers to cases where godwits colonized a site within the previous 10 years, but the exact year of first breeding is not known; these constitute approximately half of the records. These sources allowed us to identify dates of first or recent breeding for 65 individual breeding sites. However, in order to avoid pseudoreplication, analyses were based on 39 individual basins (geographically distinct areas below an elevation of 200 m.a.s.l. and generally isolated by surrounding higher ground) throughout Iceland. The earliest breeding date of these sites was used as the colonization year for each basin.
geographical correlates of population expansion
The probability of new sites being colonized is likely to be influenced by both their size and proximity to occupied sites. We constructed a General Linear Model (GLM) to explore the effects of basin size and distance from the nearest occupied basin on the timing of colonization of the 39 basins. Both basin size and distance between basins were obtained digitally using mapping software and 1 : 50 000 maps of Iceland. Distances between newly colonized basins and the nearest other breeding location were measured as a straight line, in km, between the nearest edges of both basins. The size of the basin was measured as the area of continuous habitat in km2, excluding open water. The basins ranged in area from 1 to 845 km2 (mean ± SD 109 km2 ± 190) and the distance of each basin from the nearest occupied basin ranged from 5 to 200 km (mean ± SD 69 km ± 74·4).
habitat correlates of population expansion
In order to explore the influence of habitat structure on the pattern of expansion, the timing of colonization was related to the proportional occurrence of different habitat types within each basin. A large-scale survey of the habitat associations of birds in lowland areas was carried out in Iceland in late May 2001, 2002 and 2003, with different parts of Iceland being surveyed in each year. This survey recorded information about the distribution of habitat types in 20 of the lowland basins for which godwit colonization data are available. The survey recorded eight different habitat types at 2-km intervals along roads throughout each lowland basin (Table 1). The mean number of survey points per basin was 36 ± 78·1 SD. Godwits are most commonly found in two of these habitat types (Table 1): (1) dwarf-birch bogs are homogenous bogs characterized by dwarf-birch, Betula nana, often with Carex rostrata or Eriophorum spp. present, and (2) marsh is an aggregation of habitat types dominated by sedges, Carex spp., indicating a high water level on an annual basis. Marshes often have shallow pools and are usually rather flat and intersected by tracts of mesic grassland. The records of godwits on grasslands, agricultural fields (principally hayfields) and river plains (Table 1) refer largely to foraging, rather than breeding birds. Elevation (m.a.s.l.) at each survey point was also recorded by GPS and used to calculate mean elevation within each lowland basin.
Table 1. Habitat types recorded in a large-scale survey of breeding waders throughout lowland Iceland between 2001 and 2003. The number of survey points containing each habitat type and the number and percentage of points at which black-tailed godwits were recorded are shown
No. survey points
Points with godwits
% with godwits
The proportional occurrence of marshes and dwarf-birch bogs were incorporated with basin area and distance to the nearest occupied basin in a second GLM, for the 20 basins for which both habitat survey data and first breeding records were available.
In order to assess the differences in the quality of marshes and dwarf-birch bog habitats for breeding godwits, we examined reproductive success and resource abundance in these two habitat types on 14 sites in southern Iceland (six dwarf-birch bogs and eight marshes) in 2002 and 2003. The cryptic nature of godwit nests and chicks makes it difficult to accurately measure fledging success over the large number of sites needed for a cross-habitat comparison of breeding success. An index of productivity was therefore calculated for each site, using a similar approach to Grant et al. (2000). On each of the 14 sites, surveys of territorial birds were carried out every 3–4 days from mid-April until all birds had left in late summer. The number of pairs breeding on the site was calculated as the mean of the three maximum counts during the first part of incubation (late May to mid June). During the latter part of the fledging period (second to third weeks of July), the mean number of pairs with chicks was calculated from the mean of the two maximum counts during this period. This allowed calculation of the proportion of pairs on each site that had chicks that were close to fledging. During the period just prior to fledging, chick presence is clearly indicated by the agonistic behaviour of the adults. To validate this, we observed 27 godwit pairs (across 10 sites) whose behaviour suggested the presence of chicks. In 25 cases, chicks were observed within 40 min, suggesting that adult behaviour is a strong indicator of chick presence. In addition, observations of 11 colour-marked broods showed that adults with chicks stay on the breeding site after hatching, suggesting that movements of broods into and out of sites is unlikely to confound the results. Eight study sites (four each on dwarf-birch bog and marshes) were surveyed in both 2002 and 2003, and a further six were only surveyed in one year. On the eight sites surveyed in both years, breeding success was highly correlated and consistent between years (slope = 1·01, r = 0·77, P < 0·05). Thus all 14 sites were included in the analyses and averages of the 2 years were used for the eight relevant sites.
The abundance of invertebrate prey was quantified with both pitfall traps, to survey surface invertebrates, and hand-netting of invertebrates within vegetation. On 11 of the study sites (four on dwarf-birch bog and seven on marsh), five to 13 pitfall traps of 12 cm diameter were randomly placed throughout each site during the period between chick hatching and fledging in 2002. These traps were monitored for c. 30 days and all animals with body lengths exceeding 3 mm were counted. As the number of traps varied between sites, the unit of comparison was the mean number of animals/trap-day at each site. In addition, at eight of the study sites (four each on dwarf-birch bog and marsh), 10 nonoverlapping sweeps through vegetation with a 40-cm diameter hand-net were carried out at five to six random points. These surveys took place during the second week of July in 2003, when the number of chicks was at a peak. Again, all animals with body lengths exceeding 3 mm were counted.
Both marshes and dwarf-birch bogs contain shallow pools in which adult godwits feed extensively. These pools are a source of invertebrates, of which Tipulidae and Chironomidae larvae are most abundant (Gardarsson 1998). Across the 14 sites in this study, the number of permanent pools and the area of each site was recorded digitally using mapping software and scanned aerial photographs.
breeding habitat quality
The mean index of breeding success on marsh habitat was significantly higher than on dwarf-birch bog habitat (proportion of pairs with chicks on marsh = 0·54 ± 0·06 SE, n = 8 sites; dwarf-birch bog = 0·29 ± 0·06 SE, n = 6; t12 = 2·54, P = 0·026). Prey densities were also higher in marsh habitats (hand-net samples: marsh = 188 ± 22·3 SE prey/sample; dwarf-birch bog 65 ± 10·76; t36 = 2·2, P = 0·034, pitfall samples: marsh = 3·7 ± 0·39 prey/trap-day, dwarf-birch bog = 2·5 ± 0·31; t25 = 2·37, P = 0·026).
The average number of pools in the marsh habitat was significantly higher than in the dwarf-birch bog habitat (10·5 km−2 and 1·6 km−2, respectively, t12 = 2·8, P = 0·014). There was also a strong positive relationship between the density of pools and the index of breeding success (Fig. 2).
major patterns in distribution changes
Figure 3 shows the number of records of new breeding locations each decade since 1911. Most new colonizations appear to have occurred between 1960 and 1990, at which time the population was rapidly expanding (Gill et al. 2001).
Early records of black-tailed godwits between 1900 and 1907 only report godwits from the southern lowlands of Iceland (Slater 1901; Hantjsch 1905; Hörring 1908), and particularly from the more southern part. The first breeding records outside the southern lowlands were from western Iceland in 1913 (Blöndal 1944) (Fig. 4). Around 1930, godwits started breeding in the upper part of the southern lowlands and between 1939 and 1943 they colonized three major basins in northern Iceland (Jonasson 1944), and extended their range in the south (Einarsson 1955). Between 1950 and 1960 they colonized smaller basins in the west, after which they started breeding in north-east Iceland. Godwits are still being found in new breeding locations every year (e.g. Gunnarsson & Thorisson 2004).
factors affecting the pattern of expansion
The pattern of expansion of breeding godwits across the 39 lowland basins for which colonization dates were known was strongly influenced by basin size, with larger basins being colonized first (GLM: F3,38 = 21·9, P < 0·001, Fig. 5), but not by distance to the nearest occupied basin or by the interaction of basin size and distance to nearest occupied basin.
Basin size was also a significant predictor of the pattern of expansion across the 20 lowland basins for which both colonization year and habitat structure were known, but the area of marsh habitat and the distance to the nearest occupied site were also significant predictors (Table 2). Recently occupied sites are smaller, have a lower proportion of marsh habitat (Fig. 6a) and are closer to occupied sites than older sites. By contrast, godwits have been progressively occupying areas where dwarf-birch bog is more common (Fig. 6b). The frequency of marsh and dwarf-birch bog habitats did not change significantly with altitude (marsh: R2 = 0·17, P = 0·059, n = 20; dwarf-birch bog: R2 = 0·03, P = 0·449, n = 20), there was no significant correlation between the frequency of the two habitats within individual basins (r = −0·38, P = 0·072, n = 20) and colonization year was not related to altitude (R2 = 0·01, P = 0·658, n = 20). Year of colonization was also not related to the frequency of occurrence of any of the other habitat types in which godwits were recorded (Table 1, agriculture: R2 = 0·11, P = 0·165, n = 20; grasslands: R2 = 0·02, P = 0·609, n = 20; river plains: R2 = 0·003, P > 0·705, n = 20). Frequency of marsh habitat occurrence did not correlate significantly with either basin size or distance to nearest occupied basin (r = 0·45, P = 0·052, n = 20 and r = 0·04, P = 0·869, n = 20, respectively). Frequency of dwarf-birch bog did show a significant negative correlation with size of basin (r = −0·526, P = 0·025, n = 20) (which may be why dwarf-birch bog occurrence was not retained in the GLM), but not with distance to nearest occupied basin (r = 0·065, P = 0·780, n = 20).
Table 2. Results of a general linear model of the factors predicting the colonization pattern of black-tailed godwits across 20 lowland basins around Iceland. See text for details. Overall model fit: F3,19 = 83·05, R2 = 0·95, P < 0·001
Marsh in area (%)
Distance to nearest breeding area
Basin size (km2)
The buffer effect describes the large-scale, density-dependent variation in patterns of habitat use resulting from changes in population size. In buffer effect systems, the use of poorer quality habitats is expected to increase disproportionately as populations expand, and decrease disproportionately as population size declines. In this study, the two breeding habitat types used by black-tailed godwits differ greatly in both prey abundance and reproductive output. Godwits have been expanding into areas where the poorer quality breeding habitat type is more common, whereas traditionally occupied sites have significantly more high-quality breeding habitat. This strongly suggests that a large-scale buffer effect has been operating on the breeding grounds during the expansion of the godwit population. The presence of a buffer effect has already been established in this particular population on the wintering grounds, where birds have been expanding into areas where survival is lower (Gill et al. 2001). Counts of black-tailed godwits on the wintering grounds in Britain and Ireland suggest that the rate of population increase has slowed in the last few years (Pollitt et al. 2003). This may well be a consequence of this double buffer effect, as it suggests that per capita survival and fecundity should be declining as a result of the extensive expansion into poor quality breeding and wintering sites.
The Icelandic black-tailed godwit population provides a rare opportunity to explore a century of distribution change of virtually the entire breeding population of a migratory bird. During the last 100 years, godwits have spread from breeding in a restricted area in the southern lowlands of Iceland, to currently occupying lowland basins all around the country. The early occupation of the small number of large, lowland basins around the country has been followed by expansion into smaller basins, closer to each of these large basins. Patch size and isolation are common predictors of distribution changes in many studies, both empirical and theoretical (e.g. Verboom et al. 1991; Selmi, Boulinier & Faivre 2003; Tischendorf, Bender & Fahrig 2003), although the relative importance of each varies between study systems. In this study, large basins also tended to have lower proportions of poor quality habitat. Thus larger sites are both likely to be easier to locate and they are likely to be selected because of the higher probability of finding good quality breeding habitat.
It is possible that the observed pattern of changing habitat use could be influenced by changes in the proportional availability of the habitats through time. However, while marshes and dwarf birch bogs have been extensively altered by drainage and conversion to agricultural land, their relative loss does not appear to differ greatly (Oskarsson 1998; Thorhallsdottir et al. 1998). Comparisons of land maps from 1930 and 1996 do not suggest that drainage has had a disproportionate impact on either marsh or dwarf-birch bog.
The availability of data describing the changes in summer and winter distribution of the increasing population of black-tailed godwits, together with data quantifying the demographic consequences of variation in habitat quality, provided an ideal opportunity to explore the potential for population regulation through the buffer effect. The double buffer effect that we have described operates through variation in habitat quality at both ends of the migratory range, although buffer effects could also operate on the stopover sites that are used on migration. In most migratory species, it is highly likely that some form of density dependence operates in both seasons (e.g. Goss-Custard et al. 1995; Sutherland 1996a; Pettifor et al. 2000), and thus double buffer effects may be a common and widespread form of population regulation.
This work has been possible because of the numerous observers who have reported godwit occurrence and colonization patterns in Iceland. For helpful discussions about godwit distribution and expansion we thank Arnthor Gardarsson and Kristinn H. Skarphedinsson. We thank Hersir Gíslason for help with Fig. 4. For financial support we thank the British Chevening Scholarships, the Icelandic Research Fund for Graduate Students, Tyndall Centre for Climate Change Research and NERC.