*Present address and correspondence: Dr Allan M. Strong, School of Natural Resources, Aiken Center, University of Vermont, Burlington, Vermont 05405, USA. (802) 656–3007. E-mail: email@example.com
1. Food availability has been considered one of the most important factors limiting bird populations, yet the effects of food abundance on non-breeding insectivores has rarely been investigated. We studied the effects of food abundance on the body condition of ovenbirds (Seiurus aurocapillus L.), a sexually monomorphic, ground-foraging, Neotropical–Nearctic migrant warbler during two winters in three habitats on their wintering grounds in Jamaica.
2. Prey biomass decreased from early to late winter in all habitats. Concurrently, ovenbird body mass, corrected for differences in structural size, decreased significantly in five of six habitat–year combinations. Only in second growth scrub habitat in the 1995–96 winter did ovenbirds show no decrease in corrected body mass, and during this period there was no significant decrease in ant biomass, the dominant prey item of ovenbirds in all habitats.
3. Significant positive correlations were found between ovenbirds' rate of feather regrowth and the biomass of ants on their home ranges in early winter, and between overwinter change in ovenbird body mass and the biomass of ants on their home ranges in late winter.
4. The results of both the habitat- and home-range-based analyses suggest a similar threshold of ant biomass (2·5–3·0 mg m−2), below which ovenbirds did not maintain their body mass.
5. The results suggest that late winter rainfall mediates the biomass of prey abundance, which in turn affects the ovenbirds' overwinter body condition.
Although food limitation is often investigated at the population level, numerous sublethal effects (e.g. loss of body mass, delay of moult or migration) may be manifested only at the level of the individual and thus cannot be quantified through traditional census techniques. Body condition indices provide an assessment of an individual's physiological state and their ability to deal with stress (Brown 1996; Marra & Holberton 1998). Few studies, however, have documented a direct link between food availability and body condition (Newton 1998 and references therein), especially for insectivorous species (but see Carrascal et al. 1998). This is due in part to the spatial and temporal variability in arthropod abundances and the difficulty in differentiating between prey availability and prey abundance for insectivores (Cooper & Whitmore 1990; Wolda 1990; Strong 2000).
Different body condition indices provide different information about the physiological state of an individual (Brown 1996); thus, multiple condition indices may be preferable in determining overall body condition. Body mass corrected for structural size provides an indication of an individual's gross nutrient stores (Piersma & Davidson 1991). However, body mass does not necessarily provide an accurate picture of the composition of nutrient stores (Lindström & Piersma 1993). Fat scores (Helms & Drury 1960) are an accurate predictor of lipid composition (Krementz & Pendleton 1990; Rogers 1991), but the relationship between fat storage and food availability may not be straightforward. Individuals that have access to adequate food resources may not store excess fat because increased wing loading may decrease flight performance and increase susceptibility to predation (Lima 1986). In contrast, individuals that are more likely to encounter food shortages may store fat in anticipation of energy deficits (Blem 1990). Additionally, ptilochronolgy has been used to assess the long-term nutritional status of non-breeding birds (Grubb 1989; Grubb & Cimprich 1990; Grubb 1991; but see Murphy & King 1991).
Other measurements, although not indicative of the condition of an individual, can provide additional information about food limitation. For example, if adult birds prevent young birds from gaining access to sites with greater food availability, then age ratios may be an indicator of food resources. Additionally, the proportion of birds that remain at a site throughout the winter (overwinter persistence) has been used as a minimum estimate of survival rates (Holmes, Sherry & Reitsma 1989; Sherry & Holmes 1996).
Despite this array of condition indices, the effects of food availability on the condition of wintering Neotropical–Nearctic migrants remains poorly understood. For example, non-breeding American redstarts (Setophaga ruticilla) exhibit non-random habitat use patterns, with sex ratios (Marra, Sherry & Holmes 1993), body mass loss, corticosterone levels (Marra & Holberton 1998), density, and overwinter persistence (Sherry & Holmes 1996) varying among habitats. Yet despite investigations into seasonal (Lovette & Holmes 1995) and habitat-specific (Parrish & Sherry 1994) resource levels, the link between demographic parameters and food availability is inferential. Although circumstantial (Price 1981) and direct (Katti & Price 1996) evidence of the effect of food availability on the body condition of non-breeding migrant Palearctic warblers (Sylviidae) has been documented, data for Neotropical–Nearctic migrants remain limited (but see Greenberg 1992). Understanding the relationship between body condition and food availability is critical to elucidate the factors that affect habitat quality, information necessary to develop habitat management strategies for migrants on their wintering grounds.
The ovenbird (Seiurus aurocapillus, Parulidae) is a ground-foraging Neotropical migrant warbler with life-history traits that make it amenable to studies of body condition and food availability. In winter, ovenbirds forage nearly exclusively by gleaning prey from the leaf litter surface (Zach & Falls 1979), and their prey, > 60% ants, can be readily quantified (Strong 2000). Additionally, non-breeding ovenbirds occur in a variety of structurally distinct habitat types (Wunderle & Waide 1993) that span a gradient of prey abundances. Further, ovenbirds are readily captured in mist nets, so large sample sizes can be procured for examining overwinter body condition indices.
We tested two hypotheses regarding the relationship between food availability and body condition of non-breeding ovenbirds. First, across habitats, changes in body condition indices (corrected body mass and subcutaneous furcular fat) and indices of habitat quality (age ratios and overwinter persistence), both within and between winters, will correspond to prey biomass. Second, an individual's body condition index (change in body mass and rate of feather regrowth) will be correlated with prey biomass on their home range.
Ovenbirds were studied in plots of approximately 5 ha in each of three structurally and ecologically distinct habitat types in Jamaica: shade coffee (n = 3 sites), second growth scrub (n = 2), and undisturbed dry limestone forest (n = 2). The shade coffee plots, James Hill (18°10′N, 77°20′W, 630 m elevation), Baronhall Estates (18°12′N, 77°22′W, 555 m elevation), and Coleyville farm (18°11′N, 77°30′W, 880 m elevation) had well-developed overstories consisting primarily of Pseudalbizzia berteroana at James Hill and Inga vera at Baronhall and Coleyville Farm. Because most coffee bushes at the James Hill study plot were cut to < 1 m following the 1995–96 field season, we used the Coleyville Farm site in 1996–97. The second growth scrub plots were located 10 km west of Black River (18°02′N, 77°57′W, 5 m elevation). The plots, approximately 1 km apart, were dominated by logwood (Haematoxylum campechianum), an introduced tree which formed a roughly monospecific overstorey across the majority of both plots. The dry limestone forest plots were located at Portland Ridge, 12 km south-east of Lionel Town (17°44′N, 77°09′W, 100–120 m elevation). These plots, < 2 km apart, were characterized by Metopium brownii, Bursera simaruba, and to a lesser extent Thrinax parviflora. For additional details on study sites see Strong (2000).
Body condition indices
During the winters of both 1995–96 and 1996–97, we visited each plot twice for 5–7 days, first during the early winter (November–December) and approximately 12 weeks later in late winter (February–March). These periods are well after ovenbirds arrive in Jamaica (late September to mid-October) and well before they depart for the breeding grounds (late April to early May; R. and A. Sutton, unpublished data), thus minimizing the effects of changes in body condition due to migratory demands. We attempted to capture all ovenbirds on each plot by saturating the plots with mist nets and by luring ovenbirds to mist nets with taped recordings of their songs and calls. Eight to 15 mist nets per day were used, resulting in 415–580 net hours per plot per visit. Each captured ovenbird was marked with a unique combination of two coloured leg rings and a numbered aluminium US Fish & Wildlife Service ring. Unflattened wing chord and tail length (to the nearest 0·5 mm), bill length (measured from the tip of the bill to the distal edge of the nares to the nearest 0·1 mm), tarsus length (to the nearest 0·01 mm; Pyle et al. 1987), and body mass (to the nearest 0·1 g using an electronic balance) were measured on each captured bird. Age was classified as adult (after hatch year during November–December or after second year during January–March) or immature (hatch year during November–December or second year during January–March) using the criteria of Pyle et al. (1987) and Donovan & Stanley (1995). Because of the sexually monomorphic plumage of ovenbirds, only 12% could be sexed using length of the unflattened wing chord. Additionally, we collected the right third rectrix (tail feather) on each ovenbird captured during the early winter to induce feather regrowth for ptilochronological analyses (Grubb 1989).
Six indices of body condition and habitat quality (overwinter change in body mass, body mass corrected for structural size [hereafter corrected body mass], furcular fat score, rate of rectrix regrowth, age ratio, and overwinter persistence) were used in two types of analyses (habitat-based and home-range-based). For habitat-based analyses, sampling units were all birds captured within each habitat–season combination, allowing broad comparisons between condition indices and prey biomass. However, this obscures potential within-habitat variation, which we addressed in the home-range-based analyses. In the home-range-based analyses, sampling units were all individuals that were captured in both the early and late winter.
To calculate corrected body mass in habitat-based analyses, we regressed body mass of ovenbirds that had no visible furcular fat against tarsus length, wing length, bill length, and tail length. The result of this analysis was the equation:
bill and tail length provided no additional predictive value. Corrected body mass in subsequent analyses was calculated as actual body mass minus predicted body mass. For home-range-based analyses of birds captured in both early and late winter periods, we used uncorrected body mass, because structural measurements of individuals were assumed to be constant over the course of the winter. Passerines can show significant diurnal changes in body mass and fat stores (Webster 1989; Graedel & Loveland 1995). However, we found no differences in capture times among habitats (three-way anova, F = 0·02, d.f. = 2 and 380, P > 0·10, range for habitat means = 0838–0849), years (F = 2·69, d.f. = 1 and 380, P > 0·10, range for year means = 0829–0855), or seasons (F = 2·21, d.f. = 1 and 380, P > 0·10, range for season means = 0832–0835), and for the home-range-based analyses of recaptured birds, capture times did not differ between early and late winter (t = 1·16, d.f. = 72, P > 0·25, mean for early winter = 0848, mean for late winter = 0906). Thus, we did not attempt to correct for diurnal variation in body mass.
Body lipid stores were quantified using visible furcular fat categories: 0 (none), 1 (trace) 2 (fat forming a solid sheet across the bottom of the furculum), and 3 (fat filling furculum; Holmes et al. 1989). For analyses, we pooled fat classes 2 and 3 because < 2% of ovenbirds had fat scores of 3.
The width of daily growth bars on regrown feathers provides an index of the availability of nutrients during the period of feather growth, such that birds in food-poor sites should show narrower growth bars than birds in sites with greater food availability (Grubb & Cimprich 1990; Grubb 1991). To assess the rate of induced feather regrowth, we used ptilochronology, following the protocol of Grubb (1989), with the exception that we included feathers that had 4 or 5 visible growth bars (n = 10), as opposed to the minimum of 6 that Grubb used in his analyses. Inclusion or omission of these feathers had no affect on the outcome of the analyses. Only feathers of birds that were captured in early winter and recaptured in late winter were included in this analysis.
Overwinter persistence was used to estimate (minimum) overwinter survival. Overwinter persistence was calculated as the proportion of colour-ringed ovenbirds on the plot in early winter that were detected (either by recapture or resighting) in the late winter. Because colour-ringed ovenbirds were difficult to observe, the majority of the detections were through recapture.
The prey available to each ovenbird captured in the early winter was estimated using 20 (1995–96) or 30 (1996–97) visual invertebrate counts (Holmes 1966; Strong 2000) per home range. Each count consisted of a 5-min scan of a 0·25-m2 quadrat of the forest floor. During each 5-min scan, all invertebrates observed within the quadrat were recorded to order in 1 mm size categories. Half of the counts were conducted in early winter and half in late winter, resulting in 10 counts per season per home range in 1995–96 and 15 counts per season per home range in 1996–97. Visual invertebrate count sites were located using random distances (0–25 m) and bearings (0–359°), which were paced out from each capture location. Most visual invertebrate counts were conducted from 11.00 to 16.00 EST. Repeated sampling (07.00–17.00) at fixed locations showed no diurnal variation in prey numbers (Strong 2000).
For habitat-based analyses, all counts within each habitat–season were pooled, resulting in 220–390 counts per habitat per season. For the home-range-based analyses, we included only the visual counts surrounding that individual's capture location. During the non-breeding season, ovenbirds are site faithful; however, they do not actively defend a territory boundary unless an intruding conspecific is detected (Strong 1999). Rather, they defend only the area around their current location (i.e. spatio-temporal territories; sensuWilson 1980). This method of territory defence results in some overlap of territory boundaries. Thus, ‘home range’ better describes the ovenbird's non-breeding pattern of spatial use. Based on data from radio-tagged ovenbirds on the same study plots, a 50-m diameter approximates the core area of a non-breeding ovenbird's home range (Strong 1999). However, if birds were not captured in the centre of their home range, our invertebrate sampling protocol may have led to a portion of the counts being conducted in an adjacent home range. Because colour-ringed ovenbirds were difficult to observe we cannot evaluate the magnitude of this potential error. However, most of the variation in invertebrate biomass was among habitats rather than among home ranges within habitats, so adjacent home ranges probably contained comparable invertebrate biomass.
To assess ovenbirds' response to food resource levels, we calculated two measures of prey abundance: ant abundance and the abundance of ‘total edible prey’. Ants alone were used as one measure because they made up 62% of all prey items, and were present in 100% of ovenbird regurgitation samples at these study sites (Strong 2000). ‘Total edible prey’ included invertebrate prey types that were present in at least 25% of ovenbird regurgitation samples: ants, beetles (96%), holometabolous larvae (primarily Lepidoptera and Coleoptera, 51%), snails (< 20 mm in length, 38%), spiders (38%), Hymenoptera other than ants (30%), and Orthoptera (including roaches, 25%; Strong 2000).
To convert numerical abundance to estimates of dry mass, we derived length–weight regressions (pooled across invertebrate orders) from voucher specimens of potential prey items (Table 1). Prey items were measured with an optical micrometer on a dissecting microscope to the nearest 0·1 mm, dried at 100 °C. for 24 h, and weighed to the nearest 0·1 mg. Dry mass was then estimated for each prey item in the visual counts by substituting the mid-point of each size category (e.g. 1·5 mm for 1·00–1·99 mm) into the resulting regression equations.
Table 1. Statistics for length–weight regression equations of all Jamaican leaf litter invertebrates represented in at least 25% of ovenbird regurgitation samples
We found no differences in body mass or corrected body mass between sites within habitats (all t < 1·64, all P > 0·10). Further, differences between sites within habitats existed for only one of the three habitats in analyses of age ratio (coffee; G = 4·82, P < 0·05), and fat scores (second growth scrub; G = 18·38, P < 0·01). Therefore, for the habitat-based analyses, we pooled sites within habitats based on their proximity and structural similarity.
For home-range-based analyses, we included only birds that were captured in both the early and late winter seasons. The relationship between body condition indices (rate of rectrix regrowth and change in body mass) and prey biomass on individual home ranges was examined with linear regression. We eliminated one bird that we determined to be an outlier because its corrected body mass was nearly twice as great and ant biomass on its home range was three times as great as any other individual. Deletion of this point substantially reduced the F-statistic. Estimates of early winter prey biomass were used in analyses of rectrix regrowth because early winter conditions were more appropriate indicators of food availability for the period of actual feather regrowth since rectrices were > 90% regrown after 30 days (A.M. Strong, unpublished data; Grubb 1989). For the analyses of changes in body mass, we used estimates of late winter prey biomass on each bird's home range, based on the assumption that change in body mass over the course of the winter was related to food availability at the end of that winter. We used BMDP version 7·01 (Dixon 1992) for all analyses.
Ovenbird body condition
To examine variation in condition indices (corrected body mass, predicted body mass) and prey biomass (ants and total edible prey), we used three-way anovas with habitat (coffee, second growth scrub, and dry limestone forest), year (1995–96 and 1996–97) and season (early winter and late winter) as grouping variables. Ovenbird corrected body mass declined significantly from early to late winter in all habitats in both years (Table 2), with the exception of second growth scrub habitat in 1995–96, in which ovenbirds exhibited a non-significant increase in corrected body mass (t = 1·45, d.f. = 72, P = 0·15, Fig. 1a). Corrected body mass also varied significantly among habitats, with early winter corrected body masses greater in coffee and dry limestone habitats than in second growth scrub (Table 2, Fig. 1a). Corrected body mass did not vary between years, but there was a significant habitat–season interaction (Table 2) due to greater seasonal declines in coffee and dry limestone relative to second growth scrub. We found no differences in body mass predicted from wing and tarsus lengths between seasons or years and no significant interactions (Table 2), suggesting that seasonal and annual changes in corrected body mass were not due to changes in the structural size of captured individuals. However, there were significant differences in predicted body mass among habitats (mean + SD = 19·3 + 0·5 g, 18·9 + 0·6 g, and 19·0 + 0·7 g, for shade coffee, second growth scrub, and dry limestone forest, respectively)
Table 2. anova results of the effect of habitat (shade coffee, second growth scrub, and dry limestone forest), year (1995–96 and 1996–97), and season (early winter = Nov–Dec and late winter = Feb–Mar) on corrected body mass (= actual body mass based on wing and tarsus length – predicted body mass), and predicted body mass (based on wing and tarsus length) of a sample of non-breeding ovenbirds in Jamaica, and ant biomass, and total edible prey biomass (Hymenoptera, Coleoptera, holometabolous larvae, snails, spiders, and Orthoptera, including Dictyoptera) on plots used to study the body condition of ovenbirds. Data for corrected body mass and prey biomass are shown in Fig. 1
Corrected body mass
Predicted body mass
Total edible prey
Error d.f. for predicted and corrected body mass = 380, error d.f. for ant biomass and total edible prey biomass = 3708.
To examine variation in the distribution of categorical variables (age ratios, fat scores, and overwinter persistence) among habitats, seasons, and years, we used G-tests. In both years and in both seasons, ovenbirds exhibited significant differences in fat scores among habitats (all G > 12·8, all d.f. = 4, all P < 0·05, Fig. 2). Fat scores were highest in dry limestone habitats, with 51·1% of all birds having fat scores > 2, compared to 15·4% in coffee and 27·3% in second growth scrub habitats. We found no significant differences in fat score distributions between the early and late winter (all G < 4·0, all d.f. = 2, all P > 0·05) with the exception of second growth scrub in 1995–96, in which the proportion of fat scores > 2 increased from early to late winter (G = 15·7, d.f. = 2, P < 0·01).
Pooled across seasons, the proportion of immature birds did not vary among habitats in 1995–96 (G = 3·10, d.f. = 2, P > 0·10, Table 3) or in 1996–97 (G = 2·65, d.f. = 2, P > 0·10). Further, ovenbird age ratios did not vary between seasons in any habitat–year combination (all G < 2·0, all d.f. = 1, all P > 0·10). The proportion of immature birds tended to remain constant or decrease slightly from early to late winter.
Table 3. Overwinter persistence (calculated from the proportion of ovenbirds colour-ringed in early winter that were detected in late winter) and seasonal changes in the percentage of immature ovenbirds wintering in three habitats in Jamaica
Early winter = November – December, late winter = February – March
Persistence is calculated from the early winter sample.
In 1995–96, overwinter persistence varied significantly among habitats (G = 11·15, d.f. = 2, P < 0·01; Table 3) ranging from 80% in shade coffee to < 40% in dry limestone forests. In 1996–97 overwinter persistence was similar among habitats (G = 0·43, d.f. = 2, P > 0·25). Shade coffee was the only habitat in which overwinter persistence varied between years (G = 8·71, d.f. = 1, P < 0·01).
Ptilochronological analyses could be conducted only on recaptured birds therefore we pooled data across years because of small sample sizes. Rectix regrowth rates varied among habitats (F = 7·81, d.f. = 2 and 66, P = 0·001) and were significantly lower in dry limestone forest than in the other two habitats (Bonferroni post hoc comparison, both P < 0·003).
In both years and in both seasons, ant biomass varied significantly among habitats (Table 2) and was greatest in second growth scrub, followed by coffee, and dry limestone forests (Fig. 1b). In all habitats and in both years, ant biomass decreased significantly from early to late winter (Fig. 1b, Table 2); this seasonal decline was significant in all habitats and years (all t > 2·44, all P < 0·02), except second growth scrub and dry limestone forest in 1995–96 (both t < 1·6, both P > 0·10). Ant biomass was greater in the winter of 1995–96 than in 1996–97. There was one significant interaction (habitat × year), which was caused by a greater decrease in ant biomass in second growth scrub in 1996–97 relative to the other habitats.
The pattern for biomass of total edible prey was similar to that of ants. In both years and in all habitats, dry mass of total edible prey decreased significantly from early to late winter (Fig. 1b, Table 2); however, the proportional decrease from early to late winter was greater than for ants. There was a significant habitat effect, with second growth scrub having the greatest biomass of total edible prey in both years and seasons, followed by coffee and dry limestone forest (Table 2). Additionally, the biomass of total edible prey decreased in all habitats from 1995–96 to 1996–97, similar to the pattern shown for ants. There were no significant interactions. On average, ants made up a greater percentage of the total edible prey in late winter (55·7%) than in early winter (39·3%).
The rate of ovenbird rectrix regrowth was significantly positively correlated with the early winter biomass of ants on their home ranges (F = 7·67, d.f. = 1 and 67, P < 0·01, r = 0·32, Fig. 3, but showed no statistically significant correlation with the early winter biomass of total edible prey (F = 1·98, d.f. = 1 and 67, P > 0·15, r = 0·171). The change in ovenbird body mass from early to late winter was positively correlated with the biomass of ants on their home range in late winter (F = 4·85, d.f. = 1 and 69, r = 0·26, P = 0·031, Fig. 4). Further, the relationship between change in ovenbird body mass and the late winter biomass of total edible prey was not statistically significant (F = 1·66, d.f. = 1 and 69, P = 0·20, r = 0·15). Figures 3 and 4 both show that although there was considerable variation in prey biomass among habitats, individuals with low rectrix regrowth rates and greater decreases in body mass were clustered in dry limestone forest habitat.
These results support the hypothesis that ovenbird winter body condition is directly related to prey biomass, especially the biomass of ants. This hypothesis is supported most strongly by the home-range-based analyses, which showed significant correlations between feather regrowth rate and early winter ant biomass and between overwinter change in body mass and late winter ant biomass. Additionally, in the habitat-based analyses, Ovenbirds showed significant early to late winter decreases in corrected body mass in conjunction with seasonal declines in prey biomass. The exceptional habitat–year combination helps prove the rule. In second growth scrub habitat in 1995–96, ant biomass did not decrease from early to late winter and ovenbird corrected body mass correspondingly did not decrease.
Fat and protein stores
Dry limestone forests supported the lowest biomass of ants in all sampling periods, the greatest seasonal declines in biomass of total edible prey, the greatest seasonal decreases in corrected body mass of ovenbirds, and the lowest rectrix regrowth rates. However, ovenbirds in dry limestone habitats had the greatest proportion of fat scores in categories > 2 (with the exception of second growth scrub in late winter 1995–96), suggesting that these individuals maintained high fat stores in anticipation of deteriorating environmental conditions. Similar strategies of increased fat storage in poor-quality habitats have been shown for species wintering in the temperate zone (Nolan & Ketterson 1983; Blem & Shelor 1986; Dawson & Marsh 1986), but to date, data for wintering Neotropical migrants have been equivocal (Winker, Rappole & Ramos 1990; Greenberg 1992).
Although ovenbirds in dry limestone forests showed the most extreme seasonal decreases in corrected body mass and the greatest fat scores, ovenbirds in all habitats maintained relatively constant fat scores during the winter. Assuming both that fat is deposited in a consistent manner among fat depots (Blem 1976) and that furcular fat is a good predictor of total body lipid composition (Krementz & Pendleton 1990; Conway, Eddleman & Simpson 1994; Rogers 1991), then the seasonal declines in corrected body mass do not appear to have resulted from decreases in fat stores. Rather, our results suggest that decreases in corrected body mass during the winter were due to catabolism of muscle tissue (sensuMarra & Holberton 1998) and that in most years and in most habitats, decreased late winter food availability results in protein catabolism.
The degree to which this decrease in body condition affects fitness, however, is unclear (King & Murphy 1985). Decreased muscle mass could be advantageous, simply by decreasing total body mass and decreasing metabolic demands (Freed 1981). In contrast to ovenbirds on the breeding grounds, non-breeding ovenbirds do not use elevated song perches to defend home ranges. Thus, they spend most of their time on the ground, perhaps decreasing the need for increased flight muscle mass. Historically, predation pressure was probably low throughout Jamaica, which may also have decreased the need for increased pectoral muscle mass. Mammalian predators were absent and avian predation was limited to the American kestrel (Falco sparverius) and the merlin (Falco columbarius). Neither of these falcons is likely to exert measurable predation pressure on ovenbird populations because American kestrels are primarily found in open habitats (unlike ovenbirds) and merlins are relatively uncommon. However, if decreased pectoral muscle mass and subsequent decreases in flight performance do occur, recent introductions of feral cats (Felis domesticus) and mongooses (Herpestes auropunctatus) may affect the probability of predation. This possibility cannot be evaluated from our data. Although muscle catabolism may not affect winter survival, it may potentially delay spring departure, and decrease both physiological condition upon arrival at breeding locations and nesting success as shown for American redstarts (Marra & Holberton 1998; Marra, Hobson & Holmes 1998).
Assessment of winter food resources
Ant biomass appeared to be a better index of late winter habitat quality than the biomass of total edible prey. This conclusion was supported most strongly by the results of the home-range-based analyses in which we found significant correlations between body condition indices and ant biomass, but no significant relationship with total edible prey. These results support the conclusion that ants are a critical food resource for ovenbirds in Jamaica.
Ovenbirds, however, do not appear to use ant biomass as the primary cue in initial habitat selection. Assuming ovenbirds use despotic mechanisms when selecting winter habitats (sensuSherry & Holmes 1996), age ratios should correspond with early winter ant biomass. However, we found no differences in age ratios among habitats, suggesting that either (i) ovenbirds do not use despotic mechanisms in initial habitat selection; or (ii) ovenbirds settle randomly in relation to habitat quality (based on food availability). We captured ovenbirds by using taped recordings of their songs and calls in all habitats, suggesting that at least some individuals defend home ranges from conspecifics (see also Rappole & Warner 1980). Thus, despotic interactions occurred during the non-breeding period. Instead, we hypothesize that ovenbirds settle randomly with respect to late winter habitat quality by using the early winter biomass of total edible prey as the primary cue in initial habitat selection, a cue that provides a misleading index of late winter habitat quality.
The early winter biomass of total edible prey items was less variable across habitats (2·3–4·9 mg per 0·25 m2) than ant biomass in either season (early winter: 0·3–2·9 mg per 0·25 m2, late winter: 0·1–2·6 mg per 0·25 m2). Therefore, the lack of variation in age ratios among habitats may be a result of ovenbirds' use of early winter biomass of total edible prey as the primary cue in initial habitat selection. However, over the course of the winter, the biomass of total edible prey decreased by an average of 55% overall and 82·5% in dry limestone habitats, whereas the average biomass of ants decreased by only 35%. Thus, by using total edible prey as a cue in initial habitat selection, ovenbirds may settle randomly with respect to late winter habitat quality as measured by ant biomass.
The mechanisms behind non-breeding season habitat selection have important conservation implications for understanding habitat-specific carrying capacities for migrant birds. For example, dry limestone forests may appear to be suitable habitat upon initial settlement because of the relatively high biomass of total edible prey items. However, as the dry season progresses, biomass of total edible prey drops to < 1 mg per 0·25 m2 and relative to other habitats, these forests may act as an ecological trap for ovenbirds. This result highlights the importance of sampling prey biomass during the late winter to provide the most accurate assessment of habitat quality for migrants. However, it is also possible that total edible prey may be a better index of late winter habitat quality in other parts of the ovenbird's winter range.
At least two alternatives to the above hypothesis could be operating. First, despotic interactions may affect some factor other than the ones we measured (e.g. sex ratios). That some degree of sexual habitat segregation occurred, was suggested by significant differences in predicted body mass among habitats (Table 2). However, smaller individuals (probably disproportionately females; Van Horn & Donovan 1994) were more prevalent in second growth scrub habitat, which supported greater prey abundances. Thus, sexual habitat segregation via dominance mechanisms is unlikely. Another alternative hypothesis is that ovenbirds use vegetative characteristics to assess winter food availability indirectly. Correlational data suggest that this is a cue by which ovenbirds select home ranges for reproduction (Smith & Shugart 1987). However, given the diversity of habitats that ovenbirds use during the winter (Wunderle & Waide 1993), this hypothesis appears unlikely and may be difficult to test.
Integration of resident and migratory species in the neotropics
In Panama, a study of the diets of non-breeding Neotropical–Nearctic migrant passerines showed that relative to resident species, migrant diets were dominated by small, hard-bodied prey, or larger prey that were obviously noxious or distasteful (Poulin & Lefebvre 1996). They concluded that small, hard-bodied prey items were a ‘just better than nothing’ food resource. However, our data demonstrate that although ants are poor nutritionally (Zach & Fall 1978; Bell 1990), if present at sufficiently high densities, ovenbirds can maintain or increase body mass. In fact, both the habitat-based and the home-range-based analyses suggest a similar threshold of ant abundance (2·5–3·0 mg dry mass ants per 0·25 m2), below which ovenbirds were not able to maintain their body mass for the duration of the winter. In the habitat-based analyses, significant seasonal declines in ovenbird corrected body mass occurred when late winter ant biomass was less than 2·6 mg per 0·25 m2. In the regression of change in body mass vs. late winter ant biomass, the x-intercept (point at which body mass change = 0) was 2·7 mg ants per 0·25 m2. Our data also support the hypothesis that ovenbirds are integrated into tropical avifauna by consuming prey that are underexploited by resident species (Rappole & Warner 1980; Greenberg 1995; Poulin & Lefebvre 1996; Strong 2000). However, given adequate quantities of these ‘suboptimal’ resources, ovenbirds can maintain their body mass and increase fat stores.
Temporal variation in food resources
Seasonal and annual variation in rainfall affect arthropod populations markedly, with decreased populations in xeric sites and population declines during the dry season (Wolda 1978; Levings 1983; Levings & Windsor 1984; Pearson & Derr 1986; Frith & Frith 1990). Our data show a similar result, with the biomass of both ants and total edible prey decreasing from November to April, the Neotropical dry season. This reinforces the hypothesis that rainfall indirectly affects the condition of non-breeding Neotropical migrant insectivores through its effect on arthropod populations (Sherry & Holmes 1996; Katti & Price 1996).
The results from second growth scrub imply such a scenario for ovenbirds wintering in Jamaica. Overwinter (November– March) rainfall in 1995–96 in second growth scrub sites was 392·7 mm as compared to 161·6 mm in 1996–97 (Petroleum Company of Jamaica, unpublished data). During 1995–96, ant biomass decreased by only 8·7% from early to late winter, Ovenbirds showed no significant change in corrected body mass, and median fat scores increased from 1 to 2. In the following year, ant biomass decreased from early to late winter by 43%, corrected body mass showed a significant decline, and fat scores remained constant. Both ant biomass and the biomass of total edible prey items were significantly lower in 1996–97 than the previous winter and overwinter persistence decreased (albeit non-significantly) from 50% in 1995–96 to 34% in 1996–97. Other studies have also shown ant abundance and diversity are affected negatively by desiccation (Levings 1983). Taken together, these results suggest that by affecting arthropod numbers, annual variation in rainfall can influence body condition, habitat quality, and perhaps wintering carrying capacity for ovenbirds.
Because the effects of rainfall on arthropod populations vary across taxa (Levings & Windsor 1982; Levings & Windsor 1984), modelling the effects of rainfall on populations of Neotropical migrants may be impossible without additional information on the diets of particular species (Faaborg & Arendt 1992). However, because ovenbirds' diet (in Jamaica) is dominated by a single taxon, looking for correlations between Caribbean rainfall data and ovenbird populations returning to breeding areas may be productive.
Whether the observed changes in body condition were within the physiological limits that ovenbirds can tolerate without impairing vital physiological functions (sensuKing & Murphy 1985) is unknown. Although the overwinter persistence data suggest some variation in mortality and/or emigration rates among habitats, documenting a linkage between seasonal declines in body condition and mortality rates is impossible with these data. Given that prey biomass affected absolute body mass, corrected body mass, rate of rectrix regrowth, and the potential for emigration, increased rates of predation and starvation may result, especially in sites with low food availability (dry limestone) and in years with low rainfall.
This study would not have been possible without field assistance from K. Convery, K. Hannah, M. Johnson, P. Marra, W. Merkle and R. Papish. Financial support was provided by NSF grants to T.W. Sherry (Tulane University) and R.T. Holmes (Dartmouth College), Sigma Xi Grants-in-Aid of Research, the Chicago Zoological Society, and a fellowship from the Louisiana Educational Quality Support Fund. Miss Yvette Strong and Miss Andrea Donaldson of The Jamaican Natural Resources Conservation Authority gave me permission to conduct research in Jamaica, and we thank them for their assistance. Mr E. Ziadie and members of the PWD Gun Club granted us permission to work at Portland Ridge and Newton provided assistance and Hospitality at those sites. A. Palmer, Mr Williams, and S.C. Jane of the James Hill coffee plantation, D. Reed, J. Minot, and Mr Clark of the Baronhall plantation and C. Gentiles and L. Barrett at Coleyville Farms kindly gave permission to work on their land and showed warm Hospitality during our visits. The Petroleum Company of Jamaica granted me permission to work at the second growth scrub sites. Finally, thanks to R. and A. Sutton, who were instrumental in helping me find study sites, and providing lodging and support throughout the course of the study. This manuscript benefited from the constructive comments of M. Johnson, R. Holmes, P. Marra, and P. Stouffer, and two anonymous referees.
Received 20 October 1999;revisionreceived 13 April 2000