Effects of variation in breeding habitat on Ring Ouzel Turdus torquatus productivity and chick condition

Abstract

Capsule Territory habitat influenced clutch size and within-brood variability of chick condition in Ring Ouzel Turdus torquatus.

Aims To assess the relationship of Ring Ouzel productivity and chick condition with territory habitat, in order to improve understanding of the importance of breeding habitat for population change.

Methods Productivity and chick condition were recorded for Ring Ouzel pairs in a small area of upland Scotland and these were related to vegetation and physical characteristics for all breeding territories using General Linear Models.

Results We found that clutch size and variability of chick condition varied with Ring Ouzel breeding habitat.

Clutch size was related to variation in territory habitat, broadly increasing with territory grass, sedge and rush cover. The best-supported models for within-brood variability of chick condition included fledgling number and territory habitat, with chick condition broadly becoming more variable with territory bracken cover. Relationships between breeding habitat and brood size, fledgling number or mean chick condition were not supported by the data, although statistical power was low.

Conclusion Breeding habitat may be a driver of demographic rates of Ring Ouzel, with the effects of poor habitat being passed on indirectly through chick condition. The population-level importance of these effects is unclear, but this study highlights a possible link between recent observed habitat changes and first-year survival, the demographic parameter contributing most to variation in population growth rate in UK Ring Ouzels.

Spatial and temporal heterogeneity in bird breeding habitat can cause variation in individual productivity and survival, and thus variation in demographic parameters (Donovan et al. 1995, Paredes et al. 2012, Aubry et al. 2013, Catry et al. 2013) and population size (Holmes & Sherry 2001, Amar et al. 2011). Anthropogenic changes in habitat extent and quality have been the most important cause of declines in terrestrial biodiversity, particularly birds, in the past century (Sala et al. 2000, Green et al. 2005). In order to understand the causes of population declines, it is therefore particularly important to understand species' relationships with their breeding habitat.

The Ring Ouzel Turdus torquatus is a migratory passerine that breeds in upland areas of Europe and southwest Asia. Ring Ouzels were widespread throughout the UK before the start of the 20th century (Holloway 1996), but have since declined dramatically in both abundance and distribution (Baxter & Rintoul 1953, Gibbons et al. 1993, Wotton et al. 2002, Sim et al. 2010). The magnitude of the decline led to the species' inclusion on the UK Red List of Species of Conservation Concern (Eaton et al. 2009). Although monitoring outside the UK may be too limited to detect trends in breeding populations, such a rapid decline appears to be largely confined to the UK, with other European populations apparently remaining stable or slowly increasing (Burfield 2002, Burfield & Brooke 2005, Sim et al. 2010), or slowly declining (von dem Bussche et al. 2008). Globally there has been a small decline in abundance, but the species remains Least Concern (Birdlife International 2012). It is therefore possible that the cause of the recent decline in the UK is due to the particular ecology of the UK Ring Ouzel population. A recent proliferation of research into Ring Ouzels in the UK has found no single cause of the decline, however, a number of hypotheses have some support, including changes in breeding habitat (Burfield 2002, Buchanan et al. 2003, Sim et al. 2007), changes in wintering habitat (Ryall & Briggs 2006), changes in hunting pressure on migration (Burfield & Brooke 2005) and climate change (Beale et al. 2006).

In the UK, Ring Ouzels have two salient habitat requirements during the breeding season: heather-dominated steep slopes for nesting, and grass swards or grass-heather mosaics for foraging (Arthur & White 2001, Burfield 2002). Grass-dominated areas are especially important in providing earthworms Lumbricidae, the most important component of chick diet (Burfield 2002). Since the middle of the 20th century, dramatic anthropogenic changes have taken place in the vegetation of the UK uplands. Increasing sheep numbers caused grazing pressure to reach a peak in the 1990s, reducing the extent and quality of heather moorland (Thompson & Macdonald 1995, Fuller & Gough 1999); in recent years sheep numbers have declined. Large areas of upland were planted with non-native conifers, leading to direct loss and fragmentation of habitat and increasing the area of habitat for mesopredators. These changes have had positive or negative impacts on different upland bird species, depending on their autecology (Evans et al. 2006, Pearce-Higgins & Grant 2006, Amar et al. 2011). Due to these recent vegetation changes acting on both of the key breeding habitat types for Ring Ouzels, land-use change has attracted interest as a potential cause of the Ring Ouzel decline, but our understanding of this possible relationship is incomplete.

Burfield (2002) found Ring Ouzel breeding sites were more likely to be occupied if territory grass moor cover was high, while Sim et al. (2007) found that breeding sites were more likely to remain occupied if territory heather cover remained high. Buchanan et al. (2003) found occupancy was more likely to decline in areas which were initially heather-grass mosaics, which is where heather cover declines caused by grazing are most rapid (Clarke et al. 1995). The same study also found occupancy declines in areas close to conifer plantations, suggesting a negative effect of predation, reduced grazing or delayed population-level effects of direct loss of breeding habitat (Buchanan et al. 2003). However, although the statistical relationships between Ring Ouzel breeding habitat and territory occupancy are well known, the mechanisms for these relationships are unstudied.

Reproduction is one of the most costly aspects of bird species' life history, with high energetic demands on adults in terms of egg development, providing food and caring for young while avoiding predation (Drent & Daan 1980). Within a breeding season, the costs of reproduction can vary (spatially or temporally) between pairs due to heterogeneous environmental variables, such as habitat and/or food availability. Pairs can pass on high costs of reproduction to the number of chicks fledged (Martinez 2012), to juvenile survival (Piper et al. 2012), to adult survival (Low et al. 2010) or to a combination of these traits (Lambrechts et al. 2004, Kerbiriou & Julliard 2007, Santangeli et al. 2012, Seward et al. 2013). Variation in habitat can therefore potentially influence any demographic parameter.

In UK Ring Ouzels, productivity is high and of low variability relative to other demographic parameters (Sim et al. 2011), and so perhaps variation in reproductive cost is absorbed by other demographic parameters. Interestingly Beale et al. (2006) found that UK weather in the breeding season had no effect on contemporary productivity or occupancy in Ring Ouzels, but influenced occupancy in the following year. This might indicate that effects of environmental variation during the breeding season are passed on to juvenile or adult survival in Ring Ouzels. In a study of demographic parameters in a UK Ring Ouzel population, Sim et al. (2011) found that first-year survival contributes the most to variation in population growth rate. However, although adult survival contributes less to variation in population growth rate, adult survival in Ring Ouzels was low relative to other similar species (Sim et al. 2011). A more detailed understanding of the relationships between breeding habitat and demographic parameters in Ring Ouzel is, therefore, required.

In our study site, nests are spatially distributed fairly evenly (Arthur & White 2001), perhaps as a result of high territoriality around the nest. However, vegetation types are seemingly less evenly distributed than Ring Ouzel nests (pers. obs.), and so adults vary widely in the distance they have to fly to acquire food for themselves and for their chicks (Burfield 2002, pers. Obs.). Assuming energy costs increase with flying distance, nests are therefore apparently distributed along a gradient in habitat quality. In this study we explored whether such variation in habitat quality influences productivity or chick condition in Ring Ouzels. To do this we examined potential relationships between habitat and clutch size, brood size or fledgling number, and between habitat and mean chick condition or within-brood variability of chick condition.

METHODS

Study area and timing

Ring Ouzels were studied in an area of approximately 11 km² around Loch Lee (56.90°N, 2.95°W) in north-east Scotland. The site is a steep-sided upland valley, characterized by heather moorland on the hills and rough grassland on the lower slopes and around the loch, with much of the site a mosaic of both habitat types. The site is part of a private estate managed for Red Grouse Lagopus lagopus, Red Deer Cervus elaphus and Sheep Ovis aries. The area was selected for its high, stable density of Ring Ouzel pairs.

Productivity

Productivity and chick condition were measured from April to July 2012. Nests were located by watching for adult birds visiting the nest in areas of historical breeding activity or other suitable breeding habitat. Three nests (15% of those located) in the area were not safely accessible. Once a nest had been located, visits were made every 3–7 days to count eggs and chicks. Once the chicks were of a suitable size (> 5 days after hatching), each chick was ringed with a metal British Trust for Ornithology ring and a unique combination of three colour rings. Ringing took place as part of a different long-term project to colour-ring Ring Ouzels in the UK, but benefited this study in clarifying whether fledging had been successful. To minimize disturbance, no nest was visited twice on the same day, and only on one occasion was a nest was visited on consecutive days. Nests were not visited during precipitation. Nest visits were limited to 15 minutes duration for ringing, but otherwise were less than 1 minute. Nest outcome was inferred using the criteria of the British Trust for Ornithology's Nest Record Scheme (Ferguson-Lees et al. 2011). By these criteria, no nests were predated at the egg stage, and three nests (17.6% of the nests studied) were fully predated at the chick stage. These instances were not incorporated into analyses because this study focusses on the influence of breeding habitat through food supply, rather than through predation. We note that predation can be related to the effects of food supply, and vice versa, but the number of predated nests in this study was too low to analyse such an effect. We also assumed any partial loss of brood between hatching and fledging was due to starvation and/or disease: partial predation has been recorded in Ring Ouzel (Arthur 1994) but is rare. Three response variables were constructed from the data: ‘clutch’, being the maximum clutch size; ‘brood’, being the maximum brood size and ‘fledglings’, being the brood size at fledging.

Chick condition

During ringing, two measurements of each chick were taken. Minimum tarsometatarsus length (to nearest 0.1 mm) – from notch of knee to front edge of bent foot (Svensson 1992) – and mass (to nearest 0.1 g) were measured. Tarsometatarsus length increases linearly with age over days 5–10 in Ring Ouzels (Arthur & White 2003). The residuals from a linear regression of mass on tarsometatarsus were then used as an estimate of chick ‘condition’ – the nutrient reserves available to a chick, standardized by age (Jacobs et al. 2012). It was important (Green 2001) for this index that the following assumptions were upheld: the strongest relationship between mass and tarsometatarsus was linear (F _{1, 44} = 140.6, P < 0.001); and ‘condition’ was independent of tarsometatarsus length (F _{1, 44} < 0.001, P = 1). Condition was normally distributed and homoscedastic with tarsometatarsus. Two response variables were constructed from the data: ‘mean condition’, the mean condition for each brood; and ‘within-brood variability of condition’, the standard deviation of condition for each brood.

Habitat

Habitat characteristics were measured from July to August 2012. To describe habitat, vegetation was sampled on a grid of points within 450 m of any nest. The 450 m radius represents the zone of availability (hereafter ‘territory’) within which most (96%) foraging trips take place (Burfield 2002). Ring Ouzel often feed in small habitat patches (pers. obs.), and so a relatively high resolution of points (50 m × 50 m spacing) was used to capture this important fine-scale habitat variation. At each point (identified by GPS), the two most abundant (by proportion of ground cover) vegetation types within a 2500 cm² quadrat were recorded, and their respective ground cover (to the nearest 10%) and maximum live height (to the nearest 5 cm) were estimated. To bring the fieldwork within the limits of the study, grasses, sedges and rushes (except Soft Rush Juncus effusus, being much taller than the other members) were each amalgamated into one category, as were mosses; otherwise plants were identified to the species- or genus-level. Other non-vegetation ground cover categories recorded were: rock; water; ‘bare ground’ (no vegetation cover on soil); ‘dead vegetation’ (dead vegetation unattached to live plants) and ‘burnt vegetation’ (dead vegetation with black charring). Points overhung by trees were omitted from analyses, because Ring Ouzels in the UK rarely feed in woodland (Burfield 2002). The mean number of points per territory was 213 (range 143–255). Finally, the altitude of each nest was estimated using a GPS device.

Due to the large number of vegetation and ground cover variables, we determined the main axes of variation in the habitat data by principal component analysis (PCA). PCA rotates a data set to summarize most of the variation in a small number of linearly uncorrelated axes. Habitat variables included in the PCA consisted of the number of points for which a certain ground cover type was the most abundant, per territory. Only ground cover types which were the most abundant at more than 5% of the total surveyed points (across all territories) were included in the PCA. Also, to characterize other habitat aspects which are important to Ring Ouzel breeding habitat (Burfield 2002, von dem Bussche et al. 2008, Ciach & Mrowiec 2012), the following variables were included in the PCA: number of points where the maximum height of the first and second dominant species are both ≤ 20 cm (‘short vegetation’); number of points where the first dominant species covered only 40% or less of the quadrat (‘diverse vegetation’); number of points at which recently burnt material was found (‘burnt vegetation’); and the altitude of the nest. PCA performs better when variables have the same scaling: altitude was of different units (m) to the ground cover data (number of points per territory), and so all habitat variables were standardized to mean 0 and variance 1. PCA was then carried out on the correlation matrix of these habitat variables. The ‘broken-stick’ method (Borcard et al. 2011) was used to select axes for further use by their importance according to their eigenvalues. PCA scores from axes with eigenvalues higher than the broken-stick values were then retained for use as explanatory variables.

Analysis

Preliminary analysis showed relationships between some Ring Ouzel variables: these relationships were, therefore, also included in candidate models. Clutch size was included in models for brood size and fledgling number, while fledgling number (as a proxy for brood size at measurement: there was no loss of chicks between chick measurement and fledging from any nest) was included in models for chick condition. Ring Ouzels make multiple breeding attempts in a season, and preliminary analysis showed that data for some response variables were influenced by whether they were from the first or second brood. Too few (two) second broods were documented to properly incorporate this effect into models, and so the analysis was only conducted on first broods. Spatial autocorrelation was not found (Moran's I test) in any response variables, so a non-spatial framework was used for statistical analyses.

Given that almost all of the habitat variables used in the PCA have been individually highlighted in previous studies as being important to Ring Ouzels, we had no a priori theory about how the individual habitat variables would influence Ring Ouzel variables, and so no reason to ignore particular variables or combinations of variables. Therefore, we took an exploratory approach: all combinations of the candidate explanatory variables were fitted as linear models (using least squares) for each response variable (Sanderson et al. 2009). Within these, we ranked models in descending order of AIC_c (Akaike's Information Criterion adjusted for small sample size) (Burnham & Anderson 2002, Johnson & Omland 2004). The best-supported models were considered to be those for which ΔAIC_c < 2 (Bulluck & Buehler 2008). Spatial data were prepared in QGIS 1.8.0 (Quantum GIS Development Team 2012). Statistical analyses were carried out in R version 3.0.2 (R Core Team 2013).

RESULTS

Grass/sedge/rush, heather (Calluna vulgaris, Erica cinerea and Erica tetralix), bracken Pteridium and moss were the most abundant ground cover types (out of 32 ground cover types, Table 1), each being dominant at more than 5% of the surveyed points, covering 82.1% of the surveyed points between them. These ground cover types, along with short vegetation, diverse vegetation, burnt vegetation and nest altitude were included in the PCA. The PCA found three uncorrelated axes (PC1, PC2 and PC3) whose eigenvalues were higher than the corresponding random broken-stick components. These axes respectively explained 40.2%, 23.0% and 16.3% of the variance in habitat, characterizing 79.5% of the relevant information from the habitat data between them (Table 2). PC1, the axis explaining the greatest variation in the habitat of the study site, represented a positive gradient from bracken-dominated territories to territories dominated by all other habitat types. PC2 represented a gradient from low-altitude, grass-, sedge- and rush-dominated territories to high-altitude bracken-dominated territories. PC3 represented a gradient from heather-dominated territories to territories dominated by short vegetation.

Table 1. Recorded vegetation and ground cover categories.

English name	Scientific name
Bare ground
Bell Heather	E. cinerea
Blaeberry	Vaccinium myrtillus
Birch	Betula sp.
Bog Asphodel	Narthecium ossifragum
Bog Myrtle	Myrica gale
Bracken	Pteridium
Buttercup	Ranunculus sp.
Clover	Trifolium sp.
Cowberry	Vaccinium vitis-idaea
Cross-leaved Heath	E. tetralix
Crowberry	Empetrum nigrum
Dead vegetation
Docks and Sorrel	Rumex sp.
Dog's Mercury	Mercurialis perennis
Fern (except Bracken)	Pteridophyta
Foxglove	Digitalis purpurea
Grass/sedge/rush (except Soft Rush)	Poaceae, Cyperaceae, Juncaceae
Horsetail	Equisetum sp.
Ling	C. vulgaris
Mosses	Various
Other angiosperm
Plantain	Plantago sp.
Rock
Self-heal	Prunella vulgaris
Soft rush	J. effusus
Stinging Nettle	Urtica dioica
Thistle	Various
Water
Willow	Salix sp.
Wood sage	Teucrium scorodonia
Woodrush	Luzula sp.

Table 2. Correlation coefficients for the relationship between habitat variables and the three principal components used in statistical analyses.

Habitat variables	PC1	PC2	PC3
Altitude of nest	0.248	0.533	−0.334
Grass/sedge/rush – dominated points	0.470	−0.359	−0.006
Heather-dominated points	0.260	0.330	−0.509
Bracken-dominated points	−0.322	0.530	0.042
Moss-dominated points	0.466	−0.141	−0.311
Points with short vegetation	0.221	0.319	0.544
High-diversity points	0.385	0.272	0.340
Points with recently burnt vegetation	0.363	0.033	0.346

Descriptive statistics for Ring Ouzel productivity and chick condition variables are presented in Table 3. The sample size in our study is low, leading to high levels of uncertainty in model selection: for clutch and variability of chick condition there were multiple plausible models, including the intercept-only model. The best-supported model for clutch size included solely the habitat variable PC2, with clutch size decreasing with PC2 (Tables 4 & 5). The intercept-only model was also a plausible model for clutch size, but had about a quarter less support (w = 0.33) than the habitat-only model (w = 0.42). The best-supported model (w = 0.75) for brood size described a positive relationship with clutch size alone. The best-supported model (w = 0.49) for fledgling number was also a positive relationship with clutch size alone.

Table 3. Summary statistics, Ring Ouzel productivity and chick condition variables.

Response variable	Mean (±se)	n
Clutch	3.86 ± 0.14	14
Brood	3.73 ± 0.15	15
Fledglings	3.50 ± 0.20	14
Mean condition	0.19 ± 0.89	13
Variability of chick condition	2.06 ± 0.24	13

Table 4. Model statistics, best-supported (ΔAIC_c < 2) models for Ring Ouzel productivity and chick condition variables w = Akaike weight.

Response variable	Model	ΔAICc	w	R^2
Clutch	PC2	0	0.42	0.17
	Intercept only	0.48	0.33
Brood	Clutch	0	0.75	0.84
Fledglings	Clutch	0	0.49	0.23
Mean condition	Intercept only	0	0.60
Variability of condition	Fledglings	0	0.21	0.25
	PC1 + Fledglings	0.07	0.21	0.40
	PC2	1.17	0.12	0.18
	Intercept only	1.34	0.11
	PC1	1.51	0.10	0.15
	PC1 + PC2	1.80	0.09	0.32

Table 5. Parameter estimates (±se) for the best-supported (ΔAIC_c < 2) models for Ring Ouzel productivity and chick condition variables.

Response variable	Model	Intercept	PC1	PC2	Clutch	Fledglings
Clutch	PC2	3.87 ± 0.13		−0.18 ± 0.09
	Intercept only	3.86 ± 0.14
Brood	Clutch	−1.33 ± 0.65			1.33 ± 0.17
Fledglings	Clutch	−0.21 ± 1.81			0.96 ± 0.48
Mean condition	Intercept only	0.19 ± 0.89
Variability of condition	Fledglings	−0.10 ± 0.99				0.61 ± 0.27
	PC1 + Fledglings	−0.10 ± 0.88	−0.26 ± 0.13			0.58 ± 0.25
	PC2	2.13 ± 0.22		−0.28 ± 0.15
	Intercept only	2.06 ± 0.24
	PC1	1.94 ± 0.23	−0.28 ± 0.16
	PC1 + PC2	2.01 ± 0.21	−0.26 ± 0.14	−0.26 ± 0.13

For mean chick condition, no combination of habitat variables or fledged brood size constituted an improvement over the intercept-only model (w = 0.60). The best-supported model (w = 0.21) for within-brood variability of chick condition was a positive relationship with fledgling number. However, the next-best-supported model (w = 0.21), a negative relationship with PC1 and a positive relationship with fledgling number, had an AICc just 0.07 higher, and explained nearly two-thirds more (0.40) of the variation in within-brood variability of chick condition. Four other plausible models had about half the level of support of the two best-supported models: including univariate models for PC1 and PC2, respectively, a model for both PC1 and PC2 together, and the intercept-only model. In these other models, within-brood variability of chick condition varied negatively with both PC1 and PC2.

DISCUSSION

Ring Ouzel clutch size varied with breeding habitat quality, broadly increasing with grass, sedge or rush cover, and decreasing with bracken cover and altitude. Although pair-level productivity has not previously been linked with breeding habitat in this species, grass is an important component of Ring Ouzel foraging habitat during the breeding season (Burfield 2002). Our analysis suggested that breeding habitat had no effect on the transition from clutch size to brood size or from clutch size to fledgling number. Although clutch size was related to habitat, this relationship was quite weak, explaining only a small proportion of the variation in clutch size, and only being related to a minor axis of habitat variation (PC2). Therefore, other (unstudied here) factors may play a more important role in determining clutch size. Furthermore, clutch size only explained a small proportion of the variation in fledgling number, and so change in breeding habitat is unlikely to contribute importantly to population change through productivity alone.

We found no strong support for a relationship between breeding habitat and mean chick condition. Janiga (1992) found mean Ring Ouzel chick condition to be slightly lower in large broods than smaller broods, which we did not replicate. Our study has a much smaller sample size than Janiga (1992), and so perhaps any such minor relationship was undetectable.

Ring Ouzel hatch asynchronously, with the later-hatching chicks remaining underweight for their age (Janiga 1992). Similarly to Arthur & White (2003), we found that chick condition became more variable with fledged brood size, which perhaps simply describes the existing asymmetry of chick condition set up by hatching asynchrony, which would be greater in larger broods. We found evidence that within-brood variability of chick condition increased along a gradient from grass-, sedge- and rush-dominated territories to bracken-dominated territories. This relationship is perhaps more noteworthy than the relationship between clutch size and territory habitat, because it describes a relationship of Ring Ouzel vital rates with the most important axis of variation in territory habitat (PC1). Together with fledgling number, this habitat variable explained 40.3% of the variation in within-brood variability in chick condition.

Piper et al. (2012) similarly found a negative relationship between local breeding habitat quality and within-brood variability of chick condition in Great Northern Diver Gavia immer. In other species, chicks fledging later than their siblings often have lower post-fledging survival (Cam et al. 2003). In poor foraging habitat, life history theory would lead us to expect adults to feed younger chicks disproportionately less, in order to increase the probability of fledging/survival of the other chicks above a critical level. This would increase the within-brood variability of chick condition: however, in poor foraging habitat, we would also expect a negative relationship between habitat quality and mean chick condition, which (unlike Piper et al. 2012), we did not replicate. Therefore, unless our sample size was too small to detect an existing relationship between habitat quality and mean Ring Ouzel chick condition, the ecological mechanism of a relationship between territory habitat and variability of chick condition alone is unclear. Egg size can mediate the relationship between breeding habitat and chick condition (Williams 1994, Bańbura et al. 2010): future studies may benefit from measuring this variable.

Ordination methods like PCA reduce the interpretability of relationships involving the rotated variables, and thus it is difficult to interpret the individual effects of ground cover types on Ring Ouzel variables. However, PC1 broadly represents a gradient from high bracken cover to all other ground cover types, and so there is an indication that chick condition becomes more variable with territory bracken cover. Bracken cover in the UK has increased over recent centuries, associated with changes in herbivory regimes (Marrs & Watt 2006). There has been a decline in cover over the last few decades (Pakeman et al. 2000, Marrs & Watt 2006, Carey et al. 2008), but this recent overall decline masks complex dynamics in bracken cover, with bracken cover increasing in Scotland, the core area for Ring Ouzel in the UK.

In nests where there is a large difference in condition between the best- and poorest-condition chicks, the poorest-condition chicks may be below a critical level of condition for survival in the juvenile period. Therefore, a population-wide increase in variability of chick condition may result in a greater proportion of chicks being below a critical level of condition, and thus an overall decline in juvenile survival. This potentially links variation in breeding habitat with the demographic variable most important in population change in UK Ring Ouzels, first-year survival. If upland vegetation has moved along a gradient from grass, sedge and rush to bracken in recent years, then such habitat change may have played a role in the population decline in UK Ring Ouzel. However, Sim et al. (2013) found that although chick condition positively influences first-year survival in Ring Ouzel, chick condition is only a minor contributor to this demographic parameter. Furthermore, territory habitat (itself summarized by PCA) only explained a small proportion of variation in within-brood variability in chick condition. Therefore, unless high in magnitude, habitat-related declines in minimum chick condition might make little contribution to population change. Further research could elucidate whether breeding habitat has varied in a direction which could influence variability of chick condition, and whether the magnitude of such a change is great enough to contribute importantly to population change. However in many species, effects of natal condition are not confined to the juvenile stage (Metcalfe & Monaghan 2001, Lindström & Kokko 2002, Reid et al. 2006), and can have a greater influence on survival at the adult stage than at the juvenile stage (van de Pol et al. 2006).

Previous studies (Sim et al. 2007) have found the availability of heather to be an important determinant of territory occupancy, yet we did not find heather to importantly influence productivity or chick condition. However, although heather cover varied greatly spatially throughout the study site, variation in territory heather cover between pairs was low, potentially leaving little variation to influence productivity or chick condition. It may be that heather cover is an important cue in site selection in Ring Ouzels, but other habitat dimensions determine productivity and chick condition after site selection. However, our use of PCA to summarize territory habitat makes it difficult to rule out an individual effect of heather after territory occupation.

Parents may vary their own survival to absorb costs of environmental variation during the breeding season. For example, Low et al. (2010) found that adult workload is higher in poorer quality habitat for Northern Wheatear Oenanthe oenanthe: the rate of provisioning visits to the nest was the same, but the distances flown per visit were much higher. This variation in adult energy expenditure of up to 30% resulted in a decline in adult survival with decreasing habitat quality. It was beyond the remit of our study to examine effects of variation in breeding habitat on parent survival, but as adult survival is low in UK Ring Ouzels (Sim et al. 2011), the relationship between breeding habitat and adult survival in this species is worthy of further research.

Proving a direct relationship between habitat and productivity or chick condition in an observational study is problematic, because adult quality often covaries with habitat (Coulson 1968). Variation in adult quality can then be passed on in terms of variation in productivity, chick condition, juvenile survival and adult survival (Galbraith 1988, Sydeman & Emslie 1992, Hasselquist et al. 1996). Such a spurious relationship between habitat and productivity or chick condition would mean that changes in habitat have no effect on demographic parameters. In order to tease apart the causal relationships between habitat and clutch/egg size, and between habitat and variability of chick condition, experimental studies would be required.

ACKNOWLEDGEMENTS

We are grateful to Invermark Estate for site access and to Abby Saunders for help with fieldwork. We thank Will Cresswell and two anonymous reviewers for valuable feedback.