Many animals living in seasonal environments move between geographically distinct breeding and nonbreeding areas, tracking optimal weather or food conditions or avoiding high predation risks. It may be expected that changes in the environment will affect the timing of such events, due to altered availability of resources or simply due to changes in timing of preceding life-cycle stages (Carey 2009, Barshep et al. 2011). During the last few decades, the earth's climate has changed significantly (Møller et al. 2010). At the same time, major changes have occurred in agricultural landscapes in western Europe through the intensification of farming practices. In many cases this has come at the demise of natural values and biodiversity (Fuller et al. 1995; Robinson & Sutherland 2002; Wretenberg et al. 2006). Because the timing of life-history events is influenced by environmental factors, such as weather, food availability and predation pressure, many changes may have occurred in the phenology of migratory species during the last few decades, especially those species heavily relying on agricultural areas.

We analysed timing of moult of primary feathers in Eurasian Golden Plovers Pluvialis apricaria, hereafter referred to as Golden Plover, systematically collected in The Netherlands during a period of 34 years. Golden Plovers depend on pastures and arable fields during migration and wintering (Jukema et al. 2001; Lindström 2010). They breed from south-eastern Greenland to the Khatanga River in western Siberia and those staging and wintering in The Netherlands originate mainly from Fennoscandia and NW Russia, although a significant proportion is made up of birds from further east, up to western Siberia (Speek & Speek 1984, Jukema et al. 2001, Piersma et al. 2005). The species is monotypic and no geographically distinct populations are recognised (Byrkjedal & Thompson 1998).

The birds arrive in staging areas in north-western Europe in early August. Depending on the severity of winter, large numbers winter in those areas or move further south, mainly to northern France. In The Netherlands, numbers peak in October–November (Kleefstra et al. 2009) and again in March (Waarneming.nl 2012). Occasional counts, organised between 1976 and 2008, showed that numbers ranged roughly between 200,000 and 400,000 birds (Bijlsma et al. 2001, Kleefstra et al. 2014). Large numbers are present on pastures and, to a lesser extent, on arable fields (Gillings et al. 2007, Kleefstra et al. 2014), where they feed on various subterranean prey, in particular earthworms (Lumbricidae; Cramp & Simmons 1983, Thompson & Barnard 1984, Gillings & Sutherland 2007), but in the 1980s a shift seems to have occurred from inland, agricultural areas to marine coastal areas (Kleefstra et al. 2014). Potentially, these birds still feed inland, but only aggregate to roost on the Wadden Sea coast. Alternatively, growing numbers may feed on intertidal mudflats, but unfortunately relevant data is anecdotal (Gillings et al. 2006). From May until July the birds are virtually absent from The Netherlands.

Golden Plovers are exceptional, although not unique, among migratory birds in that they start wing moult during breeding to complete it in autumn in staging or wintering areas (Byrkjedal & Thompson 1998, Jukema et al. 2001). Birds usually do not moult flight feathers during breeding, which may either be due to the predation risks it could incur or due to the high energetic or nutritious costs of both moult and reproduction precluding their simultaneous occurrence (Hoye & Buttemer 2011). Apparently, Golden Plovers can afford to start wing moult during breeding, which may be due to low predation risks or sufficient energy reserves. Timing of breeding of Golden Plovers depends partly on spring temperature (Pearce-Higgins et al. 2005), and during the last few decades spring and summer temperatures have increased significantly in the northern hemisphere (Møller et al. 2010) while evidence is mounting that many bird species have advanced their laying dates in response to this temperature change (Dunn & Winkler 2010). We predict that the increase in temperature has advanced the timing of breeding of Golden Plovers and thereby their moult schedule. In the meantime, agricultural intensification has led to decreasing numbers of breeding meadow birds (Donald et al. 2001, Donald et al. 2006); partly due to effects on food availability (Newton 2004, Kleijn et al. 2011). Many of the same factors affecting breeding meadow birds may impair foraging conditions of staging and wintering Golden Plovers. We hypothesize that this may have resulted in slower moulting of primary feathers.

METHODS

Golden Plovers were caught in pastures on the mainland of the northwest of the province of Fryslân, in the north of The Netherlands, as well as on the island of Terschelling, 50–80 km north of the mainland catching sites. Birds were caught using wilsternets - a centuries old catching technique that relies on a catcher using a flute and decoys to draw birds to a large clap net that is pulled over when birds are landing (Jukema et al. 2001). We analysed data from 1051 birds that were caught between 1978 and 2011 on 65 successful catching days between 2 August and 2 October. Per successful day between 2 and 81 adult birds were caught and measured. The extent of primary moult was assessed by measuring the length of each of the 10 primary feathers on an ordinal scale ranging from 0 to 5, where 0 indicates the presence of an old feather, 1 indicates the absence of a feather where an old feather has been shed to make room for a new one, and 5 the presence of a fully grown new feather (Ginn & Melville 1983). The moult score is the summed scores of the 10 primaries of one wing, therefore ranging between 0 and 50. Left and right wing moult symmetrically. Sexes cannot be distinguished based on biometrics (Jukema & Piersma 1992, Byrkjedal & Thompson 1998). Onset of primary moult can be earlier in males as Yalden & Pearce-Higgins (2002) showed in a population in the UK, but this does not seem to be a general pattern, as Byrkjedal (1978) did not detect sex differences in Norway. Birds were ringed with metal rings and immediately released after measuring. Primary moult score increases faster in the early stages of moult (Henriksen 1985), but in the range of moult scores contained in our data, scores proceeded linearly with time. All analyses were performed on adult birds of one year old or older.

Data on snow cover in the breeding areas is not available for this time span. We used the North Atlantic Oscillation (NAO) index as a covariate in our analyses because the index is correlated with weather parameters over a wide area in the northern hemisphere (Hurrell 1995, Sanz 2003). The mean index of September to November (NAOSON) is correlated with the onset and period of continuous snow cover in northern Norway, while the mean index of December to February (NAODJF) correlates with snow depth (Theakstone 2013). The index for May (NAOM) was used because it coincides with the start of the breeding season.

Statistics

Data were analysed using Generalised Linear Models (GLMs) in SPSS (v.22, IBM Corp.). Because normality tests tend to be overly conservative with large data sets, the distributions of residual values were visually inspected and found to follow a distribution close to normality but negatively skewed. Because of this nonparametric distribution of the data, we used bootstrapping to estimate model parameters and robust estimates of the covariance matrix (also called Huber/ White/sandwich estimates) which provide consistent estimates of the covariance, correcting the covariance matrix for model misspecification and sampling design. Models were compared using the Akaike Information Criterion (AIC). All statistics are shown ±1 SE.

RESULTS

Primary moult was scored for 1051 birds and scores ranged between 9 and 50 with a mean of 36.5 ± 0.2 and a median value of 37.0. We tested models with arrangements of the variables year, NAO and day number as predicting variables (Table 1). The best model included year, day and year-squared. Adding the interaction of year and day did not significantly improve the fit with moult score. Adding NAO in the presence of year also did not improve estimates of moult scores and NAO without year made worse predictions of moult score. Moult scores increased by an average of 0.35 ± 0.01 per day and also increased with year: from 1977, each year moult scores were on average 0.08 ± 0.01 higher. The yearly increase was not linear but accelerating (Figure 1), as shown by the significant, positive year-squared variable (Table 1). This resulted in no discernible differences in moult scores from 1977 to 2000, but an increase with, on average, 0.25 per year during the last decade (Figure 2). Assuming that moult score increased linearly through time, start of moult - when moult score equals 0 - would be around 26 May in the first two decades and 18 May in 2011. Completion of moult - when moult score equals 50 - is estimated to be around 15 October in the first two decades and 7 October in 2011. Note that in reality moult scores do not change linearly at the start and end of moult, which introduces a bias in these estimates. Moult duration estimated in this way was 142 days and would not have changed through the years.

Figure 1.

Average moult scores (±SE) per day in the course of a year, based on catches between 1977 and 2011.

Table 1.

Results of GLM analysis relating moult score with Year, North Atlantic Oscillation indices and Day of the year, showing parameter estimates, standard errors and P-values of all significant variables after backward deletion of non-significant variables and their interaction terms. Akaike's Information Criterion (AIC) is a measure of model fit and is used to compare models.

Figure 2.

Predicted moult scores per year adjusted to the average catching date, i.e. 24 August, following model 2 shown in Table 1.

DISCUSSION

Primary moult in Golden Plovers caught in The Netherlands has advanced by 8 days from 1990 to 2011, while there was no change in timing of moult during the preceding 12 years (Figures 1, 2). There has not been a shift in the speed of moult and therefore we assume that the start of moult gradually shifted to an earlier date in later years. Primary moult starts in the breeding areas, even during incubation (Byrkjedal & Thompson 1998), and we hypothesize that earlier moulting is an effect of earlier breeding. Because moult scores increase faster early during moult and because moult can be interrupted, we cannot predict the actual starting date of moult from our data. Indeed, over a wide geographical range, moulting started late June to early July (Byrkjedal 1978), which is at least a month later than the date estimated by extrapolating from our sample. This is in accordance with our estimate of moult duration, which is at least two weeks longer than what is indicated by more comprehensive analyses (Henriksen 1985; P. Machin unpubl. data).

No direct evidence is available of advanced breeding in Golden Plovers from Fennoscandia or Russia, but several findings indicate that it is likely to have occurred. During the last decades, spring snow cover has diminished in the northern hemisphere (Møller et al. 2010), affecting availability of nesting and feeding sites (Høye et al. 2007, Rakhimberdiev et al. 2007), and many Arctic birds arrive earlier in their breeding areas (Høye et al. 2007, Huntley et al. 2007). In Iceland, Golden Plovers arrived 0.43 d earlier each year from 1988 to 2009, and this pattern was correlated with mean local temperature (Gunnarsson & Tómasson 2011). In the period 1978–2000, spring arrival in northern Norway has advanced by almost 8 days (Barrett 2002), which suggests that the start of the breeding season has shifted likewise. In The Netherlands, south of the Golden Plovers' breeding range, another member of the Charadridae, the Northern Lapwing Vanellus vanellus, advanced laying by 10 days from 1950 to 2003 in response to increasing spring temperatures (Both et al. 2005). Although we used NAO as an integrated measure of weather conditions in the breeding range in our analyses, correlations with weather conditions vary spatially across northern Europe (Sæther et al. 2003). This variation may have weakened correlations with NAO. As a case in point, in the mountains of the Scandinavian Peninsula, winter precipitation has increased leading to later snow melt and probably later breeding (Byrkjedal pers. com.).

Apart from the availability of snow-free areas for breeding and feeding, advanced breeding may also result from earlier emergence of invertebrate prey. Pearce-Higgins et al. (2005) showed that Golden Plovers breeding in the northern UK advanced hatching by 6 days from the 1980s to the 1990s in response to the earlier emergence of crane fly (Tipulidae) larvae. Although British birds do not migrate to The Netherlands, similar processes are likely to occur in northern Scandinavia and Russia. On the Taymir Peninsula, in the Siberian Arctic, arthropod abundance is predicted to have advanced by 7 days from 1973 to 2003 (Tulp & Schekkerman 2008), which would lead to an earlier optimal hatching date of Arctic breeding waders. Advanced insect abundance in the breeding areas could also lead to an earlier start of moult in the absence of advanced breeding, if food availability triggers the start of moult. A functional connection between moult and breeding is therefore difficult to discern.

During the recent decades, post-breeding, staging Golden Plovers have redistributed to some degree over Europe, with higher numbers lingering on closer to the breeding areas (e.g. Sweden, Poland and Denmark; Jukema et al. 2001, Kleefstra et al. 2014). Apart from changing temperatures, this may also have been an effect of decreasing hunting pressure (Jukema et al. 2001). This would only have had an effect on the average moult stages we find in The Netherlands if birds with different timing of moult redistributed differently, and there is no indication that this occurs.

Based on the examples above, in combination with our finding of advanced wing moult, we conclude that Golden Plovers probably started breeding earlier during the last few decades in response to increasing spring temperatures, and that this has led to advanced moulting. But may the simultaneously occurring agricultural intensification in staging areas also have affected moult? During the last decades, grassland habitat has shrunk and farmers have created extensive monocultures of rye grasses Lolium spec., while water tables have been lowered to enable more frequent fertilisation and mowing (Donald et al. 2001). While this has had strong negative effects on most bird species breeding in farmland (Newton 2004, Donald et al. 2006, Kleijn et al. 2010), effects on birds that are staging or wintering in pastures or on arable fields might be different (Lindström et al. 2010). Fertilisation may actually lead to increased availability of earthworms in grassland (Standen 1984, Curry et al. 2008), thereby positively affecting staging or wintering birds, such as Golden Plovers (Atkinson et al. 2005). Moult speed, however, has not changed in the staging birds, implying that changes in agricultural practises have had no effect on moult.