The timing, duration and pattern of moult and its relationship to breeding in a population of the European Greenfinch Carduelis chloris

Transcription

1 Ibis (2005), 147, Blackwell Publishing, Ltd. The timing, duration and pattern of moult and its relationship to breeding in a population of the European Greenfinch Carduelis chloris I. NEWTON* & P. ROTHERY Centre for Ecology and Hydrology, Monks Wood Research Station, Abbots Ripton, Huntingdon, Cambs PE28 2LS, UK Moult was studied in 1 year among Greenfinches trapped in a garden in east-central England. Over the period June December 2003, 333 captures of 179 individual adults provided information on breeding condition, moult, body weight, sex and age (yearling or older adult, equivalent to birds in their second or later calendar years, respectively). About 95% of all birds (sex and age groups combined) started primary feather moult from 2 July to 14 August, and finished from 10 October to 22 November. The mean date of moult onset in the population as a whole was 24 July. On average, males began 8 days before females, and yearlings began 6 days before older birds. The mean duration of moult was 100 days, whether the figure was calculated for the population as a whole or just for the 36 individual birds that were caught more than once during moult. However, moult rate was slightly slower, and moult duration slightly longer, in yearlings than in older adults of both sexes. No evidence was found for any systematic relationship between moult onset date and rate (duration). Breeding and moult overlapped by up to 5 weeks or more in individual birds, and some birds probably started to moult as early as the incubation stage of their last clutch of the season. The cloacal protuberance (taken as indicative of breeding condition) had regressed in all males by the time the fifth primary was shed, and the brood patch had regressed and re-feathered in all females by the time the fourth primary was shed. The bulk of feather replacement in the secondary, tail and body tracts occurred in the second half of primary moult, and after cloacal protuberances and brood patches were completely regressed. In all birds examined near the end of primary moult the secondaries were still growing, and would have continued growth for up to another 19 days or more, extending the end of the moulting season into December. Body mass during moult was affected significantly by sex and age, as well as by time of day, amount of food in gullet, reproductive condition and date. No firm evidence emerged that body mass was affected by moult stage, after allowing for effects of date and other variables (although there was a non-significant negative relationship between moult stage and body mass in males). In the population as a whole, the breeding season (from first egg-laying to independence of last young) was spread over 21 weeks and moult over 24 weeks. With an overlap between the two events at the population level of up to 9 weeks, the two processes together took up to 36 weeks, some 69% of the year. Much less attention has been devoted to the study of moult in birds than to reproduction and migration, the other major events in the annual cycles of many species. In particular, little is known of the factors that underlie variation in the timing and duration of *Correspondence. ine@ceh.ac.uk moult among adult individuals from the same population, or of the factors that underlie any overlap between breeding and moult (Newton 1966, Middleton 1978, Morton & Morton 1990, Jenni & Winkler 1994). Our aim in this paper is to examine the timing, duration and pattern of moult in adult Greenfinches Carduelis chloris in relation to reproduction. The paper is based on 333 captures of British Ornithologists Union

2 668 I. Newton & P. Rothery different adults, all caught for examination in the same garden at Pickworth in south Lincolnshire, UK (52 54 N, 0 25 W) during the 7 months June December 2003 inclusive. As in the rest of Britain, Greenfinches are resident in this area year-round, and at least the majority of individuals do not migrate. Previous studies of moult among adult Greenfinches were conducted by Newton (1968), based on analyses of British Trust for Ornithology Moult Cards, and by Thorne (1976) based on birds caught at a site in Cambridgeshire, about 80 km to the south-southeast of Pickworth. Although useful for comparative purposes, these earlier studies were based on smaller samples than the present one, enabled less detailed investigation and were made before statistical methodology designed especially for the analysis of moult data became available (Underhill & Zucchini 1988, Rothery & Newton 2002). In addition, no relevant data were collected in the earlier studies that enabled the moult to be placed within the context of breeding or other events in the annual cycle. METHODS The birds for the present study were caught in a cage trap or in mist-nets placed near hanging feeders, all baited with sunflower seeds. For each bird at each capture, note was made of sex and age (yearling [hatched in the previous year] or older adult [hatched before the previous year], based on criteria in Svensson 1992), body mass, wing-length (maximum chord, as an index of body size, Gosler et al. 1998), date and time of capture, amount of food in gullet (on a 0 3 scale), reproductive and moult condition. In males, reproductive condition was assessed from the cloacal protuberance (recorded as regressed as in the non-breeding season, or swollen), and in females on the presence or absence of a brood patch (recorded as feathered or unfeathered). Moult in the flight and tail feathers was recorded in the conventional way (Newton 1966) by allocating a score to each feather, according to its state of growth: 0, old; 1, pin-brush stage; 2, brush to one-quarter grown; 3, one-quarter to one-half grown; 4, one-half to three-quarters grown; and 5, three-quarters to fully grown. For purposes of analysis, feathers that were missing were allocated a score of 0.1 and feathers that were fully grown but still sheathed at the base were allocated 4.9. The primaries were moulted in sequence, inner (1) to outer (9) and, except at the start and end of moult, most birds had two or three primary feathers in growth at once at different stages. Total moult score was taken as the sum of the scores of all nine primaries of one wing (0 45), which gave an index of stage of moult in each bird at the time of capture. In birds in which the left and right wings were slightly out of phase, a mean score was taken for each feather (equivalent feathers of each wing) and for total moult score. A similar scoring system was used to record moult in the six secondary flight feathers of each wing and the six large feathers of each half of the tail. The moult in different tracts of body feathers was also recorded, but for present purposes these data are reduced to a single variable: body moult not started, in progress or finished. In the Greenfinches examined, all the tail and body feathers were moulted within the period of primary feather moult, but moult of the secondary feathers continued for a time beyond the end of primary moult. Nevertheless, total primary score could be taken as a useful index of the state of moult as a whole (see later). For some purposes, the total primary feather scores (PS) were converted to mass of new feather material grown. This was conducted by taking the weights of freshly grown primary feathers from 10 individuals (Table 1), calculating the mean weights of each of the nine feathers, and allocating a mass score (PMS) to each new or growing feather in every moulting bird caught, according to the stage of growth reached (equivalent to PFMG, percentage feather mass grown, of Summers et al and others). This procedure gave a measure of moult stage that was influenced less than the numerical score by the variable lengths and weights of the different primary feathers (see Summers et al. 1983). In consequence, PMS gave a more linear relationship with date than did PS, and hence more reliable Table 1. Lengths and masses of each primary feather (1 9, inner to outer) in the Greenfinch, based on measurements from 10 individuals, five males and five females; standard errors are shown in parenthesis. Primary feather Length (mm) Mass (g) (0.58) 87.8 (2.16) (0.61) 95.3 (2.65) (0.70) (2.84) (0.70) (2.90) (0.76) (3.68) (0.83) (4.58) (0.88) (4.76) (0.81) (4.90) (0.73) (5.17)

3 Timing, duration and pattern of moult in European Greenfinch 669 estimates of moult start date and duration in individuals (Dawson & Newton 2004). Estimates of the mean timing and duration of moult in different age and sex categories were made using the methods of Underhill and Zucchini (1988). (For appraisal of these methods in another finch species, see Newton & Rothery 2000.) For most birds allocation to the two age classes, one year or older, presented no problem, on the basis of wing coverts as described by Svensson (1992). However, near the end of moult first-year birds had adopted full adult plumage, so could no longer be distinguished from older birds unless they had been ringed previously. The small numbers of unringed birds involved were classed as of unknown age, and excluded from analyses that involved comparison of age groups. Dates are taken as day 1 (1 June) to day 214 (31 December). RESULTS Estimation of moult parameters In this initial analysis we ignore the fact that some birds were trapped more than once during moult, and on each occasion regard these birds as independent components of a cross-sectional random sample from the population. This gave a total of 333 records, 112 for yearling males, 69 for older males and eight for unknown age males, together with 72 for yearling females, 69 for older females and three for unknown age females. From these records, the mean moult onset date and mean moult duration were estimated using the Underhill Zucchini model for Type 2 data [i.e. using all three categories of birds: moult not started, in moult (actual scores), and moult finished]. For birds in moult, we used both the moult score (PS) and the new feather mass score (PMS). However, the PMS was preferred because its increase through moult was more linear than that of PS (Dawson & Newton 2004). On applying the model to all birds, ignoring age and sex differences, the estimated mean onset date was earlier using PS [day 49 ± 1.0 (mean ± 1 se), 19 July] than using PMS (day 54 ± 1.2, 24 July). The estimated standard deviation was also larger for PS (12.0 days ± 0.67) than for PMS (10.8 days ± 0.61), possibly reflecting the slightly curvilinear increase of PS with time. Estimated moult duration was almost the same on both scoring methods, at around 100 days (on PS, duration = ± 2.5 days; on PMS, duration = 99.8 ± 2.3 days). The fit of the model (using PMS) is shown in Figure 1, based on the mean Figure 1. Fit of model showing primary feather mass score (PMS) in relation to date of capture for all Greenfinches (sex and age groups combined). Lines show the mean rate of increase for a bird starting on the mean onset date, with lower and upper 97.5% percentiles. Some 95% of birds started during 2 July 14 August, and finished during 10 October 22 November. Dates are given as days 1 214, 1 June 31 December. Figure 2. Estimated moult onset dates of all captured Greenfinches (sex and age groups combined), calculated from the fitted model in Figure 1. The solid line shows the Normal curve with a mean onset date of 54 and standard deviation of 10.4 days. onset date and the lower and upper 97.5 percentiles. Most values of PMS fall within the expected range of variation. The moult onset dates of all birds are shown in Figure 2, estimated from the moult rate calculated from the model, i.e. estimated onset = t T PMS, where t = examination date (days from 1 June) and T = mean moult duration (days). There was no obvious departure from the normal distribution of onset dates assumed in the Underhill Zucchini model. The standard deviation of the estimated values was 10.4 days, which was close to the value in the fitted model, mentioned above. An estimated 95% of birds started moult within the period 2 July 14 August.

4 670 I. Newton & P. Rothery Table 2. Estimates of (a) mean onset date (b) standard deviation of onset date and (c) moult duration (in days) for females (F) and males (M) in different age groups using feather mass score (PMS) with the Underhill Zucchini model, Type 2 data (see text). Day 1 = 1 June, age 1 yearling; age 2 older adult. Age F (se) M (se) F M (se) Z (F M) Mean (se) (F + M)/2 (a) (2.3) 46.4 (1.7) 8.79 (2.84) 3.09** 50.8 (1.4) 25 July 16 July 20 July (1.9) 53.2 (1.6) 7.69 (2.53) 3.04** 57.0 (1.3) 30 July 23 July 27 July Age (2.98) 6.78 (2.37) 1.10 (3.81) (1.91) Z (age 2 1) ** ** Mean (age 1 + 2)/ (1.5) 49.8 (1.2) 8.24 (1.90) 4.34*** 53.9 (1.0) 28 July 19 July 23 July (b) (1.40) 9.30 (0.92) 2.35 (1.68) (0.84) (1.27) 7.63 (0.92) 2.93 (1.57) (0.78) Age (0.95) 1.67 (1.60) 0.58 (2.30) (1.15) Z (age 2 1) Mean (age 1 + 2)/ (0.95) 8.47 (0.65) 2.63 (1.15) 2.29* 9.79 (0.57) (c) (5.7) (3.7) 5.27 (6.84) (3.4) (4.4) 99.1 (4.3) 5.61 (6.10) (3.1) Age (7.20) 3.61 (5.66) (9.16) (4.56) Z (2 1) 2.01* * Mean (age 1 + 2)/ (3.6) (2.83) 0.17 (4.58) (2.3) Test for differences uses Z = difference/se[difference]. P < 0.10, *P < 0.05, **P < 0.01, ***P < An important question was whether the rate of moult varied systematically with the onset date: in particular, did birds that started late in the season moult most rapidly? If this was the case, there should have been more spread in the range of start dates than in the range of finish dates. This possibility was investigated by using Type 1 data (i.e. frequencies of birds not started moult, in moult and finished), and fitting a model that allowed for different variance in start and finish dates. The fit of this model was then compared by the maximum likelihood test to the fit of a similar model with no difference in variance between start and finish dates (Newton & Rothery 2000). There was no significant difference in the fit of the two models for any sex or age group (yearling 2 2 males, χ 1 = 0.60, n = 112, P = 0.44; older males χ 1 = , n = 69, P = 0.76; yearling females χ 1 = 1.27, 2 n = 72, P = 0.26; older females χ 1 = 1.93, n = 69, P = 0.16). This test thus gave no evidence of any systematic variation in moult duration that was associated with onset date. Age and sex differences in onset dates were examined by fitting a model separately to each of the four categories of birds: yearling males, yearling females, older males and older females, ignoring the few birds of unknown age. The different age and sex categories showed consistent differences in mean onset dates, males beginning on average about 8 days earlier than females, and yearlings about 6 days earlier than older birds (Table 2), with no evidence of a sex age interaction. There was a suggestion of less spread in the onset dates of males than females, and of a slightly faster moult in older birds than younger ones. Similar conclusions emerged from analyses based on PS as on PMS, with differences in the actual values from PS (not shown) similar to those for PMS in Table 2. Repeat observations on the same individuals Repeat observations on birds caught more than once during moult were used to provide a direct check on the assumption of a linear increase in PS or PMS. They were also used to provide estimates of the mean date of onset and duration of moult via regression analyses of PS and PMS on date, for comparison with the above estimates obtained by using the Underhill Zucchini models based on all captures. In total, 77

5 Timing, duration and pattern of moult in European Greenfinch 671 Figure 3. Rates of increase in primary mass scores (PMS) of 27 individual Greenfinches examined more than once during moult, with first and last capture dates more than 7 days apart. birds were caught only once during moult (so were excluded from the present analyses), but 26 birds were caught twice, five birds were caught three times and five birds were caught four times. These individual records generally supported the notion of a linear increase in primary scores (PMS) with date (Fig. 3), but in many birds the span of days between captures was relatively small, which limited the scope for detecting non-linearity. Formal tests for non-linearity were performed by fitting linear, quadratic and cubic regressions, allowing for different onset dates (intercepts) in different birds. The fit of each model was measured by the percentage of variation within birds accounted for, rather than by the percentage of total variation accounted for. The latter included a component of variation between birds, which was accommodated by allowing a different intercept for each bird. For both PS and PMS, a linear model accounted for a large percentage of the variation within birds (r 2 > 96.4%). Adding quadratic and cubic terms made little difference (r 2 < 97.5% in all cases), although statistically significant effects emerged in the cubic term for PS (P < 0.01), and in both the quadratic (P < 0.01) and cubic (P < 0.001) terms for PMS. Differences in slopes (moult rates) among individuals were detected using both PS (F 35,15 = 6.54, P < 0.001) and PMS (F 35,15 = 3.49, P = 0.006), and the effects remained significant after allowing for quadratic and cubic terms in the slopes. Inspection of the data showed that the biggest deviations from the mean slope occurred in relatively few birds, mainly those examined at short intervals in the early stages of moult. To estimate the mean onset date and duration of moult in birds caught more than once during moult, we fitted a linear regression model with a separate intercept for each bird, and a common slope. With the average of the fitted lines given by y = a + bt, the mean onset date and duration were estimated as a/b and 1/b, respectively, with the score scaled to lie between 0 and 1. The estimates are shown in Table 3, alongside those based on the Underhill Zucchini model. For PMS, the estimates of mean onset dates were very similar for both methods. For PS, larger differences between the results from different methods were apparent, although the same patterns held across groups. There was reasonable agreement in estimates of duration for yearling males and for adult females, but less agreement for adult males and yearling females. These regression estimates were, of course, based on smaller samples than were those derived from the Underhill Zucchini model, but this analysis supported the view that the mass-based PMS provides a more satisfactory scoring system than the proportionate length-based PS. Any relationship between onset date and duration could not be examined satisfactorily from recaptured Table 3. Linear regression estimates of onset dates and durations of moult for Greenfinches examined more than once during moult, compared with estimates obtained by the Underhill Zucchini model. Mean onset Duration Sex Age Recaptures in moult U Z model Type 2 data Recaptures in moult U Z model Type 2 data PMS Male Female PS Male Female

6 672 I. Newton & P. Rothery Table 4. Percentage of birds with tracts in growth against number of primaries shed. Note last category (10) corresponds to birds having completed primary moult. No. of primaries shed n Secondaries Tail Body birds, because estimates of individual onset dates were themselves dependent on estimates of individual moult rates. This could give rise to a spurious relationship between moult rate and onset date, because any bird that moulted quickly by chance between successive capture dates would have both a shorter estimated duration and a later estimated onset date. Moult pattern So far, we have been concerned only with the moult of the primary feathers, and their use in recording the progress of moult. In this section, we examine the moult of the body, secondary and tail feathers in relation to primary moult (Table 4). Body moult in different individuals began between the shedding of the second and fifth primary feathers, moult of the secondaries began between the shedding of the fifth and seventh primaries, and moult of the tail feathers between the shedding of the third and sixth primaries. In other words, considerable individual variation was apparent in these respects. However, the bulk of feather replacement in all these other tracts occurred in the latter half of primary moult, and after the cloacal protuberance or brood patch had regressed completely. The last tract to finish moulting was the secondaries, and in all individuals trapped at an appropriate stage the secondaries were still in moult after the last (outermost) primary feather was fully grown. In terms of timing, while primary moult began on average around day 54 (24 July), secondary moult began on average around day 100 (8 September), some 46 days later, and nearly halfway through the total primary moult period (Fig. 4a, Table 4). Tail Figure 4. Secondary (a) and tail (b) scores in relation to primary scores in individual Greenfinches. moult began slightly earlier, around day 90 (29 August), some 36 days later than the primary moult (Fig. 4b, Table 4). While the average duration of primary moult was around 100 days, secondary moult is likely to have continued in some birds for up to 19 days or more beyond this date. This is because, in the most extreme individuals, the last (innermost) secondary was shed as the last (outermost) primary reached full length, and to judge from captive birds (Newton 1967) the last secondary takes about 19 days to grow. Body mass during moult The body weights of Greenfinches varied significantly with date, time of day, amount of food in the gullet (visible through the skin when the dorsal neck feathers were blown aside), wing-length (as an index of body size), reproductive condition (as indicated by states of the cloaca in males and the brood patch in females) and primary moult score (Table 5). They also varied between the sexes, with male weights

7 Timing, duration and pattern of moult in European Greenfinch 673 Table 5. Relationships (Pearson correlations) between body mass and various factors likely to affect mass for male (above diagonal) and female Greenfinches. Body mass PS Day Time of day Food in crop Wing length Body mass 0.29*** 0.34*** 0.22** 0.38*** 0.29*** PS 0.20* 0.95*** *** 0.23** Date 0.25** 0.91*** *** 0.22** Time of day 0.23** Food in gullet 0.31** 0.30** 0.28** Wing-length 0.26** 0.19* 0.17* * *P < 0.05, **P < 0.01, ***P < Figure 5. Body masses of male (a, c) and female (b, d) Greenfinches in relation to stage of moult (primary score, PMS) and date. Masses are shown as measured, uncorrected for time of day or other influences. Lowess lines show the underlying trend. averaging slightly higher than females for much of the year, but females weighing more than males in the breeding season (Fig. 5). This latter difference accounted for some different patterns during moult: whereas males showed a steady increase in body weight throughout the moult period, June December, females showed an initial decrease to about midprimary moult (from their high reproductive weights), followed by an increase paralleling that in males (Fig. 5). Our aim here was to explore whether body mass was influenced by stage of moult after allowing for the potentially confounding effects of the other variables. No differences in mean body mass were observed between yearling and older adults so the two age groups were combined, but sexes were treated separately throughout.

8 674 I. Newton & P. Rothery Table 6. Comparison of wing-lengths in the same group of individual Greenfinches examined just before and just after moulting primaries 7 8 (which, being the longest primaries, are the only ones that influence wing-length). The initial group of birds were moulting primary feathers but had not yet reached P7, whereas the second group had finished growth of the 8th primary. Birds caught up to 31 December contributed to the latter wing-length measure. Sex Age Difference (se) after before t (df) P-value Males (0.32) 6.70 (26) < (0.67) 1.50 (10) 0.17 Overall 1.98 (0.24) 6.93 (39) < Females (0.50) 3.33 (12) (0.52) 4.99 (9) Overall 2.06 (0.37) 5.65 (22) < Difference between young and adults: Males: 1 2 = 1.15 (se = 0.740, t 36 = 1.55, P = 0.26). Females: 1 2 = 0.90 (se = 0.718, t 21 = 1.25, P = 0.23). We first considered the correlations between primary moult score and the other variables. Primary score was correlated highly with date (as expected), amount of food in the gullet and wing-length (Table 5). The relationship with gullet food arose because, as days shortened throughout the season, an increasing proportion of the birds examined had food in the gullet. The relationship with wing-length arose because, while in the first part of the season measurements were made on old primary feathers, towards the end of the season they were made on newly grown feathers. Comparison of wing-lengths between premoult and post-moult birds revealed a mean increase of around 2 mm (approximately 2%), associated with the replacement of old feathers by new ones (Table 6). This reduced the usefulness of wing-length over the moult period as a measure of body size, and in the statistical analyses below wing-length was excluded (but accounted for in another way; see later). In a multiple regression model, in which variables were added in predetermined sequence, body mass showed significant relationships in both sexes with time of day, amount of gullet food, primary score and date. The full model accounted for 27% of the total weight variation in males and for 39% of that in females. However, primary score and date were correlated very closely (r = 0.90), and after allowing for date and other variables no significant relationship between mass and primary score was apparent, either in linear (males, F 1,129 = 1.12, P = 0.29; females, F 1,95 = 0.16, P = 0.69) or quadratic models (males, F 1,129 = 0.25, P = 0.62, females, F 1,95 = 2.42, P = 0.12). Repeating the analysis, but entering date and primary score in reversed order in the model, the effect of primary score emerged as significant for linear (males, F 1,129 = 5.83, P = 0.017; females, F 1,95 = 6.45, P = 0.013) and quadratic effects for females only (males, F 1,129 = 0.01, P = 0.97; females, F 1,95 = 11.05, P < 0.001). After allowing for score, the linear effect for date was statistically significant in males (F 1,95 = 4.17, P = 0.043) and the quadratic effect for females (F 1,95 = 13.4, P < 0.001). Excluding date from the model, primary score, time of day and gullet food together accounted for 24% of the weight variation in males and for 30% in females. This analysis implied that the weight trends with score could be accounted for by effects of date, but not vice versa. This supported the view that date rather than primary score was the more important of the two variables in affecting body mass. Effects of time of day and gullet food in a linear model showed no significant differences between the sexes. Unfortunately, gullet food was not quantified in the initial weeks of the study, so many birds had missing values. When this variable was excluded from the analysis, the eligible sample sizes increased from 136 to 189 for males and from 102 to 144 for females. Nevertheless, the findings were essentially similar to those mentioned above. All the above analyses ignore the fact that many birds were represented more than once in the samples. Because such pseudo-replication could lead to spurious statistical significance we restricted further analysis to birds caught more than once, and partitioned the variance into within-bird and betweenbird components. The between-bird component emerged as statistically significant, and accounted for about 50% of the total weight variation in each sex. Adding wing-length did not account for a significant amount of the residual variance. Trends within birds (caught more than once) were then examined by allowing a separate term for each individual. The aim was to take out any intrinsic variation between individuals (including that in wing-length). As before, the effects of primary score were not significant after removing the effect of date, while the effect of date emerged as strongly significant after allowing for the effect of primary score (males, F 1,58 = 9.49, P = 0.003, females, F 1,36 = 11.66, P < 0.001). After allowing for other variables, there was a suggestion of a negative effect on weight of increasing moult score in males (F 1,58 = 3.43, P = 0.069) but not in females (F 1,36 = 0.79, P = 0.38). In general, all coefficients were similar in models with

9 Timing, duration and pattern of moult in European Greenfinch 675 Table 7. Estimated coefficients in linear regression model relating weight to time of day, food in gullet, date (day 1 = 1 June) and primary score (PS). For males, quadratic effects of day and PS not statistically significant. Estimated coefficient (se) Sex Model term Population level Individual level Males Constant 25.2 (0.44) 25.2 (0.42) Time of day 0.11 (0.028) (0.027) Food in crop 0.35 (0.11) 0.22 (0.11) Date (0.0066) (0.0068) PS (0.020) (0.021) Females Constant 29.5 (1.1) 30.6 (1.1) Time of day 0.15 (0.042) 0.11 (0.037) Food in crop 0.42 (0.12) 0.53 (0.13) Date (0.019) (0.024) Date (quadratic) ( ) ( ) PS (0.047) (0.086) PS (quadratic) (0.0011) (0.0019) and without individual bird effects (Table 7). This helped to confirm that trends at the population level across birds could be accounted for by trends within birds. The general conclusion was that, although several variables influenced body weight, we had no firm evidence that moult stage influenced weight independently of date. If moult affected body mass in any way, the influence was negative in males, but this effect was insignificant and in any case was more than compensated for by the positive effect of advancing date. We think it unlikely that date affected body mass directly, but is best regarded as a surrogate measure for such seasonal factors as daylength and temperature. Moult in relation to breeding Many of the males in the early stages of moult had a swollen cloacal protuberance while many of the females had a de-feathered brood patch, both indicative of reproductive condition. It was assumed that females whose brood patch was oedematous and vascularized, as well as unfeathered, were incubating and brooding small nestlings whereas those whose brood patch was simply bare were feeding well-grown young, or had finished. To quantify reproductive condition in relation to stage of moult, the proportions of birds with a swollen cloaca (males) or a de-feathered brood patch (females) were examined in relation to stage of primary moult, as indicated by the latest primary feather in the series that had been shed when the bird was captured (Fig. 6a). Figure 6. Proportion of Greenfinches examined that had a swollen cloaca (males, crosses) or brood patch (triangles, solid line) in relation to stage of primary moult (a) and date (b).

10 676 I. Newton & P. Rothery Almost all males entered moult with a fully developed cloacal protuberance, and most females entered moult with a de-feathered brood patch. There was a clear downward trend in these proportions as moult progressed, with females ahead of males. For both sexes, the slope of the relationship (on a logistic scale) was similar. The cloacal protuberance in all males had regressed by the time the fifth primary was shed, and the brood patch in all females had regressed and re-feathered by the time the fourth primary was shed. None the less, these findings suggested considerable overlap between breeding and moult in this Greenfinch population. An alternative way to examine the seasonal regression in breeding condition was to estimate the proportion of birds with a regressed cloaca or brood patch in relation to date (Fig. 6b). The median date, when 50% of the males had a large cloacal protuberance, was estimated at day 70 (9 August), which is later than the mean date of moult onset (day 50). The median date, when 50% of females had a defeathered brood patch, was estimated at day 58 (28 July), which is almost equal to the estimated mean onset of moult in the female population. The rate of decline in the percentage of males with a large cloacal protuberance was similar to the rate of decline in the percentage of females with a de-feathered brood patch (Fig. 6b). Despite the general linear increase in PMS and PS with date, in many birds the first two primaries were shed in quick succession and were well grown (fully grown in two birds) before the third and other primaries were shed. As the second primary takes 21 days to grow (in captive birds, Newton 1967), this implied a gap of up to 3 weeks between the shedding of the second and third primaries. Between other feathers, the shedding interval is likely to have been in the region of 9 14 days, on average (Newton 1967). All birds with an interrupted primary moult at P2 showed no moult in any other tract and were in obvious reproductive condition, which may have delayed the continuation of moult. Not all birds showed this interruption and, especially in birds moulting later in the season, primary moult progressed smoothly through the series. A previous study of Greenfinches in southern England indicated a surplus of males, with many more non-breeding males than females in the local population (Eley 1991, cited in Cramp & Perrins 1994). Non-breeding birds were likely to begin moulting earlier than breeders, the difference in proportions between sexes affecting the difference in their mean start dates. For most of the breeding season sex ratios were difficult to measure accurately, because females spent much more time at the nest than males and were therefore less available for capture at feeding sites. Restricting analysis to dates after 10 September, when the last observation of a dependent fledgling signalled the end of the breeding season, we trapped 28 different males and 20 females (excluding juveniles). This ratio of 1.4 males to one female did not differ significantly from unity (Z = 1.15, P = 0.12, one-tailed test). With such a small sample, we could draw no firm conclusion on the sex ratio in our population (95% confidence interval = ), which left open the possibility that a surplus of non-breeding (early moulting) males contributed to the difference in mean moult onset dates found between the sexes. DISCUSSION Feeding and catching Of the 214-day study period, bait was provided and birds were caught on a total of 81 days (38% of the total), usually several days at a time, separated by intervals of up to 3 weeks when no food was provided. No adults were caught more than once on the same day (in contrast to juveniles not considered here), and intervals between successive captures of adults that were re-trapped on different days ranged between one day and 207 days (n = 172 intervals). The mean interval between successive captures of the same individuals was 33.8 days (sd = ± 40.0 days), with no differences in this respect between sex and age groups. The total food provided over the whole 214-day period was about 30 kg, around 140 g per day over the whole period, and 370 g per day for the trapping days alone. At least 179 different adults were caught over the study period, and several hundred juveniles were caught, many more than once. With such relatively meagre and irregular food provision between so many birds, at a time when natural food was plentiful, our supplementary food is unlikely to have had any significant effects on the seasonal pattern of breeding and moult in the local Greenfinches. We therefore assume that the patterns recorded were unaffected by our sporadic local food provision. On a different issue, we have no way of knowing whether our trapping methods gave a random or non-random cross-section of the adult population.

11 Timing, duration and pattern of moult in European Greenfinch 677 Moult timing and duration Our preliminary analysis of methodology revealed that a score based on the mass of new feather growth (PMS) gave a more linear increase with date than did the conventional numerical primary score (PS), and therefore gave more reliable estimates of moult onset date and duration (confirming Summers et al and Dawson & Newton 2004 for other species). The estimate of mean moult onset date was some days later on PMS than on PS, and the spread in onset dates a few days less. On PMS, which was judged as giving the most reliable estimates, 95% of adults started moulting between 2 July and 14 August, with a mean start date of 24 July. On average, males started moulting about 8 days earlier than females and yearlings about 6 days earlier than older birds. The mean duration of primary moult was 100 days (se = ± 2.3 days) and on average was probably slightly longer in yearlings than in older birds. All body and tail feathers were moulted within the period of primary moult, but moult of the secondary flight feathers continued for up to about another 19 days beyond the end of primary moult. Our estimate of mean primary moult duration (at 100 days 95% confidence limits days) was not dissimilar to two previous estimates of 97 days, based on 89 captures at Wicken Fen in Cambridgeshire (Thorne 1976) and of 115 (range ) days based on six captive Greenfinches examined daily near Oxford (Newton 1967). Our current estimate was longer than the 85 days estimated from British Trust for Ornithology moult cards by Newton (1968), but this estimate was based on birds drawn from many different localities and years, and pre-dated the development of appropriate statistical methodology. Mean onset date was estimated at 22 July for Wicken Fen birds and at 21 July for the six captive birds, compared with 24 July for birds in the present analysis. From Finland, where summers are shorter than in Britain, the moult duration of Greenfinches was estimated at days (Lehikoinen & Niemelä 1977). Moult and body mass In male Greenfinches, average body mass increased progressively through moult and female body mass declined initially (reflecting regression from reproductive condition), and then increased in the same way as males (although remaining lower throughout this phase). In both sexes, body mass was found to vary significantly with time of day, amount of food in gullet, wing-length (presumed to reflect body size), reproductive state (especially in females), moult score and date. Some of these variables, especially moult score and date, were intercorrelated, and after allowing for such correlations in multiple regression analyses no firm evidence was obtained that the stage of moult influenced body mass independently of the other variables. If moult had any influence, it would have had a slight negative effect on the body mass of males, but this was more than offset by an increase related to the other factors examined. The general upward trend in body mass during moult was associated with the progressively shortening days and cooling temperatures through summer and autumn. Similar trends in body mass during moult were found in Bullfinches Pyrrhula pyrrhula and Great Tits Parus major (Newton 1966, Gosler 1994, respectively); and like many other small birds, Greenfinches (apart from reproductive females) are at their heaviest in midwinter, when large body reserves are carried. Moult in relation to breeding One of the most striking findings from our current study was the substantial overlap between the breeding and moult of many individuals, using the state of the cloaca or brood patch as measures of reproductive condition. Males with a swollen cloaca were found until about halfway through primary moult, and females with a de-feathered brood patch until about a third of the way through. Brood patches seemed to regress earlier in females than did cloacal protuberances in males, but the rate of decline in the proportion of reproductive individuals in the trapped samples was the same for both sexes. This could have been a genuine difference between the sexes that would hold even between members of a pair, or it might have been due to more non-breeders (which might be expected to start moult earlier on average than breeders) among the males. Assuming that females with swollen brood patches were still brooding eggs or young, some must have started moult during the incubation stage of their last clutches of the season, and the same presumably held for breeding males. This implies an overlap between breeding and moult in some individuals of more than 5 weeks. Greenfinches feeding flying young were seen up to 10 September, by which time all adults in our sample had long since started moulting. The substantial overlap between reproduction and moult in individuals was of interest in relation to the

12 678 I. Newton & P. Rothery pattern of moult in Greenfinches. Most birds began body, tail and secondary moult after the cloaca and brood patch had fully regressed, and almost all the feathers in these other tracts were replaced in the latter half of primary moult. Hence, despite the overlap between breeding and moult in this species, the peak periods of nutritional demand for the two processes were largely separated. Greenfinches thus achieved through overlap a long period for primary moult, as well as a long breeding season, but without a clash in periods of peak food demand between the two processes. In addition, no other finch species breeding in Britain is known to continue secondary moult so long beyond the end of primary moult (Evans 1966, Newton 1966, 1967, 1968). Greenfinches are resident year-round within Britain, without the need for a long migration. This may enable them to continue moulting into November or December, much later than the moult periods of migratory finches (Evans 1966, Newton 1968, 1972). A similar extension into winter holds for the Bullfinch, which is also resident (Newton 1966). It is, of course, possible that extensive overlap between breeding and moult, or a late moult per se, might incur thermoregulatory costs or lower feather quality, leading in turn to reduced overwinter survival (Dawson et al. 2000), but these aspects were not examined. Breeding and moulting seasons of the whole population Because different Greenfinches finished breeding and began moulting at different dates from one another, the amount of overlap between the breeding and moulting seasons of the population as a whole was much greater than the overlap within individual birds. With an approximately 6 7-week spread in the moult onset dates of the population, the latest birds could have fitted in an extra brood in the period since the earliest started. Based on British Trust for Ornithology nest record cards, Monk (1954) estimated the start of egg-laying among Greenfinches in southern Britain at about 15 April, and our own observations revealed adults feeding flying fledglings until at least 10 September, which gives a breeding season for the population from the start of egg-laying of at least 21 weeks, and a population overlap between breeding and moult of more than 10 weeks. The first birds to start primary moult began before 2 July and the last finished after 22 November, giving a total primary moult period of about 21 weeks. To this must be added another 3 weeks or so for the completion of secondary moult in the latest birds. Hence, in this Greenfinch population the total moulting season (24 weeks), counted from the start of feather loss, lasted somewhat longer than the total breeding season (21 weeks) counted from the start of egglaying. The two processes together occupied about 36 weeks, some 69% of the annual cycle. One of the most difficult estimates to obtain in some multibrooded bird species is the dates at which individual birds finish breeding. This is largely because, with cover at its thickest in late summer, late nests are difficult to find in sufficient numbers, thus precluding a reasonable estimate of finishing dates for the population as a whole. Also, many casual observers are likely to give up before the end of the season, as nest finding becomes more difficult and birds gradually stop breeding. The measures used here from trapped birds, namely cloacal protuberance and brood patch development, provide an alternative and perhaps more representative measure of end of breeding dates, which could be obtained with much less effort than nest details, and used in annual or regional comparisons. Previous studies have used moult start dates (Newton 1999) or migration dates (Browne & Aebischer 2003) as measures of the timing of the end of breeding, but these measures are more difficult to record than the state of the cloaca or brood patch. REFERENCES Browne, S.J. & Aebischer, N.J Temporal changes in the migration phenology of Turtle Doves Streptopelia turtur in Britain, based on sightings from coastal bird observatories. J. Avian Biol. 34: Cramp, S. & Perrins, C.M The Birds of the Western Palearctic, Vol. 8. Oxford: Oxford University Press. Dawson, A., Hinsley, S.A., Ferns, P.N., Bonner, R.H.C. & Eccleston, L Rate of moult affects feather quality: a mechanism linking current reproductive effort to future survival. Proc. Roy. Soc. Lond. B 267: Dawson, A. & Newton, I Use and validation of a molt score index corrected for primary-feather mass. Auk 121: 1 7. Evans, P.R Autumn movements, moult and measurements of the Lesser Redpoll Carduelis flammea cabaret. Ibis 108: Gosler, A.G Mass-change during moult in the Great Tit Parus major. Bird Study 41: Gosler, A., Greenwood, J.J.D., Baker, J.K. & Davidson, N.C The field determination of body size and condition in passerines: a report to the British Ringing Committee. Bird Study 45: Jenni, L. & Winkler, R Moult and Ageing of European Passerines. London: Academic Press. Lehikoinen, E. & Niemelä, P Varpuslintujen sulkasadon tutkimus. Lintumies 12:

13 Timing, duration and pattern of moult in European Greenfinch 679 Middleton, A.L.A The annual cycle of the American Goldfinch. Condor 80: Monk, J.F The breeding biology of the Greenfinch. Bird Study 1: Morton, G.A. & Morton, M.L Dynamics of postnuptial moult in free-living Mountain White-crowned Sparrows. Condor 92: Newton, I The moult of the Bullfinch Pyrrhula pyrrhula. Ibis 108: Newton, I Feather growth and moult in some captive finches. Bird Study 14: Newton, I The moulting seasons of some finches and buntings. Bird Study 15: Newton, I Finches. London: Collins. Newton, I An alternative approach to the measurement of seasonal trends in bird breeding success: a case study of the Bullfinch Pyrrhula pyrrhula. J. Anim. Ecol. 68: Newton, I. & Rothery, P Timing and duration of moult in the Bullfinch: an appraisal of different analytical procedures. Ibis 142: Rothery, P. & Newton, I A simple method of estimating timing and duration of avian primary moult using field data. Ibis 144: Summers, R.W., Swann, R.L. & Nicoll, M The effect of methods on estimates of primary moult duration in the Redshank Tringa totanus. Bird Study 30: Svensson, L Identification Guide to European Passerines. Stockholm: privately published. Thorne, C.J.R Primary moult of finches in Cambridgeshire. Wicken Fen Group Rep. 8: Underhill, L.G. & Zucchini, W A model for avian primary moult. Ibis 130: Received 8 July 2004; revision accepted 18 March 2005.