Differences in tail feather growth rate in storm-petrels breeding in the Northern and Southern hemisphere: a ptilochronological approach

Introduction

Moulting and breeding are energetically costly stages of the annual cycle of birds. The costs of feather synthesis can be illustrated by the fact that metabolic rate during moult increases by more than 100% compared to pre-moulting (Lindström, Visser & Daan, 1993). Feather production costs are linked with body mass in a way that moult is relatively more demanding for smaller birds (Lindström, Visser & Daan, 1993). Additionally, moult gaps in the remiges and/or rectrices formed after losing old feathers reduce aerodynamic performance, mostly through affecting manoeuvrability (Hedenström & Sunada, 1999; Slagsvold & Dale, 1996) and less so through increased flight costs (Hedenström & Sunada, 1999).

The costs of breeding (e.g., incubation and chick provisioning) are apparent in the increased field metabolic rates (e.g., 11% from incubation to chick rearing in Australasian gannets, Morus serrator) (Green et al., 2013) and increased stress levels (e.g., higher feather corticosterone concentrations in giant petrels, Macronectes spp.) in successful compared to failed breeders (Crossin et al., 2013). Increased reproductive costs negatively affected the breeding success in the following year, and birds may even forego breeding if the costs are too high (Crossin et al., 2013; Minguez, 1998; Pratte et al., 2018).

Due to the high energetic costs of moulting and breeding, trade-offs may emerge regarding energy allocation between them, e.g., as shown by decreasing chick quality when artificially increasing parental flight costs (Mauck & Grubb, 1995; Navarro & González-Solís, 2007). Indeed, in many avian species these two life-stages are temporally separated, with complete moult following the breeding period. Failed breeders and non-breeders often take advantage of the absence of breeding duties by advancing moult (Alonso et al., 2009; Barbraud & Chastel, 1998; Crossin et al., 2013; Hemborg, Sanz & Lundberg, 2001; Mumme, 2018; Ramos et al., 2018). In contrast, individuals that breed relatively late in the season (Stutchbury et al., 2011), or that have higher foraging costs during the breeding period (Alonso et al., 2009), moult later in the season. Moreover, individuals of some species may suspend moult until they arrive at the wintering areas (Catry et al., 2013; Ramos et al., 2009), providing some flexibility in allocation of energy between moulting, breeding and migration. The extent of this flexibility partially depends on environmental circumstances (e.g., day-length linked to latitude or food availability) (Hemborg, Sanz & Lundberg, 2001; Terrill, 2018), and the trade-off between moulting and breeding may even differ strongly between closely related species (e.g., in Northern, Fulmarus glacialis, and Southern, F. glacialoides, fulmars; Barbraud & Chastel, 1998). For instance, some seabird species overlap breeding and moulting, although populations with higher foraging costs show less overlap than populations with lower costs (e.g., in Cory’s shearwaters, Calonectris diomedea borealis; (Alonso et al., 2009). Moult-breeding overlap may therefore only be possible when energetic demands can be met, e.g., when food availability is high (Alonso et al., 2009; Barbraud & Chastel, 1998). Likewise, moult-breeding overlap seems more prevalent in sedentary than migratory species (Bridge, 2006), though several migratory species adopt this strategy as well (Alonso et al., 2009; Barbraud & Chastel, 1998; Ramos et al., 2009).

Investigating the trade-off in energy allocation between moulting and breeding may be challenging in pelagic seabirds as they are only available for researchers when they come to land for breeding. As at least part of the moulting period is often completed away from the breeding colony, studying their energy management during feather growth may prove difficult. Ptilochronology may offer a way to retrospectively determine the relative amount of energy available during moulting in seabirds, and so evaluate their energy allocation towards feather production. The method is based on feather growth rate, which is determined by the mean feather growth bar width (Grubb, 1989; Grubb, 2006). Growth bars are alternating light and dark bands formed during feather growth. It is generally assumed that one growth bar is formed over a period of 24 h (Grubb, 2006; Jovani et al., 2011; White & Kennedy, 1992), making it a convenient measure for feather growth rate.

Mean growth bar width is linked with nutritional status, with birds foraging in areas with higher food availability having relatively larger growth bars (Grubb, 1989; Hill & Montgomerie, 1994). However, within species, growth bar width has also been related to other feather traits (i.e., positively to feather size (De la Hera, Pérez-Tris & Tellería, 2009; Hargitai et al., 2014; Le Tortorec et al., 2012; Pérez-Tris, Carbonell & Tellería, 2002), though not in all species (De la Hera, Pérez-Tris & Tellería, 2009; Pérez-Tris, Carbonell & Tellería, 2002), and negatively to feather quality (Marzal et al., 2013). Inter-species comparisons have shown that growth bar width is positively correlated with feather length and mass. This correlation is negatively allometric, such that species with larger feathers have relatively lower growth rates per unit of feather length (De la Hera, DeSante & Milá, 2012). A similar correlation has been found between feather growth rate and body size, with larger species having higher absolute feather growth rates, but lower relative growth rates per unit of body size (Rohwer et al., 2009).

The aim of our study was to compare relative energy availability during moult between pelagic storm-petrel species with contrasting moult-breeding strategies, i.e., moult-breeding overlap or non-breeding moult. In order to understand the inter- and intra-specific differences in energy availability during moult we compared feather growth rates with feather length. Additionally, to infer the relative energy allocation for each of the species towards moulting, we compared their observed feather growth rate with feather growth rate data for other species found in literature. This study is the first to compare differences in expected feather growth rates between similar species breeding in both hemispheres. Due to their small size and pelagic life-style the non-breeding period of storm-petrels can be hard to study but thanks to recent developments in technology specific migration routes of some species are being discovered (Pollet et al., 2014; Halpin et al., 2018; Martínez et al., 2019; Lago, Austad & Metzger, 2019). Our study adds to the understanding of storm-petrel migratory, moulting and breeding strategies by giving some, admittedly indirect, insights into their energy management.

Since larger feathers have been linked to a higher growth rate both within (De la Hera, Pérez-Tris & Tellería, 2009; Hargitai et al., 2014; Le Tortorec et al., 2012; Pérez-Tris, Carbonell & Tellería, 2002) and between species (De la Hera et al., 2011), we expected to find positive correlations between feather length and growth bar width both within and between the four storm-petrel species. Since the studied species adopt contrasting moult-breeding strategies, we expected to find differences in feather growth rate relative to feather length between the two strategies, indicating differences in relative energy allocation towards moult.

Materials and Methods

Results

Feather characteristics

We found significant differences between the species in FGR and FL (Kruskal–Wallis test, FGR: χ² = 214.35, p < 0.001; FL: χ² = 248.35, p < 0.001). Post-hoc tests (Dunn test, p < 0.001) revealed that black-bellied storm-petrels had a higher FGR than Wilson’s storm-petrels and the Southern species had a higher FGR than the Northern species (Table 1). FL differed significantly between all species pairs (Dunn-test, p < 0.001) except between black-bellied and Leach’s storm-petrels (Table 1). Only the Southern species showed a significant positive correlation between FGR and FL, with Wilson’s storm-petrels showing a weak positive correlation (Spearman correlation, r_s = 0.215, p = 0.002) and black-bellied storm-petrels a moderately positive correlation (r_s = 0.513, p = 0.04) (Fig. 1, Table 2). For the Northern species we found no correlation between FGR and FL (p > 0.05) (Table 2).

Table 1:

Results of the post-hoc Dunn test for inter-specific differences in feather length (FL) and feather growth rate (FGR) for each studied species.

Species	Var	Black-bellied storm-petrel (Z, p)		European storm-petrel (Z, p)		Leach’s storm-petrel (Z, p)
European storm-petrel	FL	12.177,	<0.001
	FGR	10.363,	<0.001
Leach’s storm-petrel	FL	0.698,	0.243	−13.763,	<0.001
	FGR	9.438,	<0.001	−1.218,	0.111
Wilson’s storm-petrel	FL	7.587,	<0.001	−8.446,	<0.001	8.850,	<0.001
	FGR	3.602,	<0.001	−10.831,	<0.001	−9.528,	<0.001

DOI: 10.7717/peerj.7807/table-1

Notes:

Var, variable, p-values ≤α∕2(α = 0.05) are bolded.

Table 2:

Spearmans Rank Correlation output for correlations between feather growth rate (FGR) and feather length (FL) for each studied species.

Species	S	rho	p-value
European storm-petrel	2,3327	0.004	0.976
Leach’s storm-petrel	27,756	−0.001	0.993
Wilson’s storm-petrel	1,046,299	0.215	0.002
black-bellied storm-petrel	1,977.7	0.513	0.004

DOI: 10.7717/peerj.7807/table-2

Notes:

S

sum of squared rank differences

rho

Spearmans rank correlation rho

P-values ≤ 0.05 are bolded.

Relative energy availability

In the full PGLS model (AIC = −497.59), FL (p < 0.001) and group (i.e., Procellariiformes vs. non-Procellariiformes) had a significant effect (p < 0.001) on FGR but the interaction FL * group did not (p = 0.729) (Fig. 2A, Table 3). The PGLS model with group as a predictor (hereafter optimised PGLS model) was better (AIC = −499.47) than the PGLS model without group (AIC = −484.99, ΔAIC = 14.48) (Table 4). The model with group and interaction did not differ from the model without the interaction (AIC = −497.59, ΔAIC = 1.88). After multiplying Procellariiformes’ FGR by two (i.e., PGLSadj), FL still had a significant effect on FGR (p < 0.001) (Fig. 2B, Table 3). Neither group (p = 0.912) nor the interaction FL * group was significant (p = 0.729) (Table 3). The PGSLadj model without the group predictor (AIC = −501.46) was not different from the PGLSadj model including group (AIC = −499.47, ΔAIC = 1.99) nor was the PGLSadj model including group different from the model including group and interaction (AIC = 497.59, ΔAIC = 1.88) (Fig. 2B, Table 4).

Table 3:

Phylogenetic generalized least squares (PGLS) models for feather growth rate (FGR) based on feather length (FL).

PGLS models, with Pagel’s λ based on the phylogenetic tree, were fitted to log10 FGR as response variable and log10 FL as predictor, for data found in literature (n = 164 species) and this study (n = 4 species). To determine whether the Procellariiform species considered (n = 6 species) differed in number of growth bars (GB) formed per 24 h, we added a group term (group 1 = non-Procellariiformes, group 2 = Procellariiformes) and its interaction to the full PGLS model (no. 1). Terms were dropped based on significance and improvement of AIC (no. 2 & 3). To test whether Langston and Rohwer’s (1996) suggestion that Procellariiformes might form two GBs per 24 h was true, Procellariiformes FGR was doubled (PGLSadj model) and an analogous set of models were tried. Pagel’s λ is the phylogenetic signal, with values between 0 and 1.

Model	No.	Predictor	AIC	Pagel’s λ	Estimate	SE	t-value	Pr(>—t—)
PGLS	1	Intercept	−497.59	0.935	−0.500	0.107	−4.669	<0.001
		Log10(FL)			0.531	0.043	12.320	<0.001
		Group			−0.156	0.402	−0.388	0.699
		Log10(FL):Group			−0.067	0.194	−0.347	0.729
	2	Intercept	−499.47	0.935	−0.493	0.105	−4.708	<0.001
		Log10(FL)			0.527	0.042	12.588	<0.001
		Group			−0.294	0.068	−4.309	<0.001
	3	Intercept	−484.99	0.962	−0.590	0.112	−5.288	<0.001
		Log10(FL)			0.554	0.045	12.374	<0.001
PGLSadj	1	Intercept	−497.59	0.935	−0.500	0.107	−4.669	<0.001
		Log10(FL)			0.531	0.043	12.320	<0.001
		Group			0.145	0.402	0.361	0.719
		Log10(FL):Group			−0.067	0.194	0.347	0.729
	2	Intercept	−499.47	0.935	−0.493	0.105	−4.708	<0.001
		Log10(FL)			0.527	0.042	12.588	<0.001
		Group			0.008	0.068	0.110	0.912
	3	Intercept	−501.46	0.935	−0.490	0.102	−4.814	<0.001
		Log10(FL)			0.527	0.041	12.760	<0.001

DOI: 10.7717/peerj.7807/table-3

Notes:

All p-values ≤ 0.05 are bolded.

The individual residuals of predicted FGR based on the optimized PGLS model differed significantly between the studied storm-petrel species (Kruskal–Wallis; χ² = 198.92, df = 3, p < 0.001)(Table 5) and hemispheres (Welch t-test, p < 0.001)(Fig. 3). The residuals for the Southern species were significantly higher than those of the Northern species (mean Southern = 0.126, mean Northern = −0.091, t_158.56 = −21.371). The residuals differed significantly from zero for all four species. Both Northern species had negative residuals, while both Southern species had positive residuals (Student’s t-test, ESP: t₅₁ = −5.847, p < 0.001; LSP: t₅₄ = −8.367, p < 0.001; WSP: t₁₉₉ = 25.310, p < 0.001; BBSP t₂₈ = 14.460, p < 0.001) (Fig. 3, Table 6).

Discussion

The Spearman’s rho correlations showed significant, positive relationships between mean growth bar width and feather length for the Southern storm-petrel species, but not for the Northern species. Additionally, using the PGLS model, we found that the Southern species had a higher feather growth rate than predicted while the Northern species had a lower feather growth rate than predicted.

The difference in residual length between the studied species, and between the hemispheres may be associated with a difference in relative energy availability during moulting between species of both hemispheres, possibly caused by their different moult-breeding strategy. The Southern species, both moulting during the non-breeding period (Beck & Brown, 1972), are free from breeding duties during moult and may use all available energy for feather synthesis, while the Northern species, showing moult-breeding overlap (Amengual et al., 1999; Arroyo et al., 2004; Bolton & Thomas, 2001), have to allocate that energy between moulting and breeding. The differences between storm-petrels from both hemispheres in trade-offs in energy allocation between moulting and breeding may affect the correlations between feather length and feather growth rate, resulting in a lack of significant correlations between feather length and mean growth bar width in the Northern species in contrast to significant relationships between feather length and mean growth bar width in the Southern species.

Moult-breeding overlap in the Northern species has so far only been shown in the Mediterranean subspecies of the European storm-petrel (Hydrobates pelagicus melitensis) (Amengual et al., 1999; Arroyo et al., 2004), in a British population of European storm-petrels (Scott 1970 in Cramp et al., 1977), in Canadian populations of the Leach’s storm-petrel (Ainley, Lewis & Morrell, 1976), but the overlap extend is not (yet) generally accepted in Northern populations. However, preliminary stable-isotope analyses show that tail feather isotopes of the Northern species are more closely matched with blood isotopes collected during the breeding season, than those of the Southern species (Ausems et al., in prep.). This seems to indicate that both the feathers and blood were synthesised under more similar foraging conditions and in similar foraging areas, strengthening our conviction that the Northern species at least partially overlap their tail moult and breeding.

Austral summer is short and primary production is highest only in favourable conditions (i.e., longer daylight hours and retreating sea ice) (Arrigo, Van Dijken & Bushinsky, 2008; Murphy et al., 2016). The peak abundance of the main prey of the Southern storm-petrels, Antarctic krill, Euphausia superba, usually lasts from December to February (Food and Agriculture Organization of the United Nations, 2019; Ross & Quetin, 2014). The relatively short period of high food abundance and possible competition over it e.g., from penguins, whales and krill fisheries (Barlow et al., 2002; Descamps et al., 2016; Ratcliffe et al., 2015), could inhibit the Southern storm-petrels from overlapping moult and breeding as there is no longer enough food available at the end of the breeding season. Additionally, compared to the North Atlantic and Arctic ocean, the highly productive oceanographic features in the Southern Ocean, such as ocean fronts and eddies, occur over larger spatial scales and are usually farther away from the breeding colonies, forcing the birds to take longer foraging trips (Bost et al., 2009). During the non-breeding period birds free from the constraint of central-place foraging may exploit these highly productive areas freely, which may explain the relatively higher than predicted daily feather growth rate of the Southern species.

In contrast, both Northern storm-petrels have been reported to show moult-breeding overlap, including the moult of tail feathers (Ainley, Lewis & Morrell, 1976; Amengual et al., 1999; Arroyo et al., 2004; Bolton & Thomas, 2001), though Leach’s storm-petrels seem to start moulting relatively earlier in the breeding season than European storm-petrels. In the North Atlantic, around the Faroe Islands, primary production peaks over a longer period (Eliasen, 2017) as it is not linked to sea ice cover. Thus, food abundance might still be sufficient for moulting at the end of the breeding season for the Northern species. However, primary production varies strongly between years, which could lead to distinct inter-annual differences in food abundance for the storm-petrels (Bonitz et al., 2018; ICES, 2005; ICES, 2008). As food availability may thus be unpredictable for the Northern species during breeding, individuals may make different choices in prioritising either moult or reproduction, leading to obscured relationships between mean growth bar width and feather length.

Langston & Rohwer (1996) suggested that Laysan albatrosses (Phoebastria immutabilis) may form two growth bars per 24 h because their main prey (i.e., various squid species) is active at night and the albatrosses forage for them at dawn and dusk. This would result in two activity-rest cycles per 24 h, which would explain the formation of two growth bars daily as growth bar formation has been linked to sleep or rest rhythms (Jovani et al., 2011). Indeed, after doubling the feather growth rate of the Procellariiformes, their correlation between feather growth rate and feather length was very similar to that of the other orders (Fig. 2B), as shown by the lack of a significant group effect in the PGLSadj model. This seems to confirm Langston & Rohwer’s (1996) suggestion that Procellariiformes form two growth bars per 24 h. The four studied storm-petrel species may also have two activity-rest cycles per 24 h, which is consistent with the main prey activity of the studied storm-petrels during the breeding season, myctophid fish and krill. These prey species have a nocturnal activity similar to the prey of the albatrosses (Hedd & Montevecchi, 2006; Siegel, 2012), and several seabirds, including storm-petrels, forage in more oceanic habitats during the non-breeding period where they seem to increase their intake of myctophid fish (Watanuki & Thiebot, 2018).

We are aware of several possible limitations of the present study. Although growth bar widths have originally been linked to relative nutritional condition (Grubb, 1989; Hill & Montgomerie, 1994), it is not a direct measurement of food availability and the results should be interpreted with caution in that regard (Murphy & King, 1991). In this study we used feather growth rate as a way to retrospectively infer energy availability during moulting, as direct examinations of diet and food availability during moulting were impossible due to the pelagic nature and small body size of our study species. In order to put the feather growth rates observed in our study species into perspective, we compared their growth rates with data found in literature. However, while the reported measuring methods where similar between the studies, sampling techniques may have differed. In some studies samples were taken from museum specimens (e.g., De la Hera, DeSante & Milá, 2012; Rohwer et al., 2009) while others collected samples from live birds (e.g., De la Hera et al., 2011). This could lead to a bias in the condition of the birds sampled, as museum specimens may come from individuals in relatively poor health, or from relatively young individuals. Sample sizes per species ranged between 1 and 54 (De la Hera, DeSante & Milá, 2012) and for some species multiple sources were found (e.g., Sitta carolinensis: De la Hera, DeSante & Milá, 2012; Dolby & Grubb, 1998; Grubb & Cimprich, 1990. Especially Passeriformes where highly represented (n = 160 observations) in the model, while orders with larger species were under-represented. The model may therefore be less appropriate for larger species, but since the four storm-petrel species studied fall in a highly represented body size category in the model we feel it is appropriate to use here. Due to differences in the studied species’ availability, large differences in sample sizes in our research occurred: the Northern species were comparatively abundant during mist-netting sessions, while the Southern species were not. Additionally, nests of Wilson’s storm-petrels were more accessible and concentrated than black-bellied storm-petrel nests, which were spread out over larger areas and more often located on inaccessible cliffs and ledges. Nevertheless, our study provides the first comparison of relative energy availability during tail-feather moult of storm-petrels differing in moult-breeding strategies and breeding in different hemispheres.

Our results suggest that for many pelagic seabirds ptilochronology may be a useful, non-invasive, and often only feasible, tool to study their relative energy allocation to feather growth during the non-breeding period when they are hardly accessible to researchers. Due to their specific life-history traits, pelagic seabirds may be especially interesting for ptilochronology studies as one may expect different patterns of feather growth compared to other species.

Conclusions

We expected to find positive correlations between feather length and feather growth rate both within and between species, and to find differences in relative energy availability during moulting between species with differing moult-breeding strategies. The results of our analyses showed distinct differences in relative energy availability between four species of storm-petrels. The Southern species had a higher feather growth rate than predicted by a model based on data from multiple species and orders, while the Northern species had a lower feather growth rate than predicted. We suggest that all these differences can be attributed to the different moult-breeding strategies the species adopt, as the Southern storm-petrel species show no moult-breeding overlap while the Northern species do overlap both stages of the annual cycle. The better relative energy availability of the Southern species during moult may be explained by the fact that they change their feathers during the non-breeding period and can thus use all free energy for feather synthesis. In contrast, the Northern species have to allocate their energy between breeding and moulting. Our study shows how different moult-breeding strategies may affect the relative energy availability or energy allocation during moult of migratory pelagic seabirds. Additionally, we showed that at least a subset of the Procellariiformes likely forms two growth bars per 24 h instead of one, probably associated with the diel migration of their main prey species.

Supplemental Information