† Current address: Bamfield Marine Station, Bamfield, British Columbia, Canada V0R 1BO. E-mail: email@example.com
Avian seed dispersal mutualisms are characterized frequently by stochastic interactions between birds and fruits; however, many studies report coarse-scale correlations in annual abundances of birds and fruits at particular locales (i.e. ‘phenological synchrony’). This study tested the geographical consistency of phenological synchrony in a meta-analysis of data from 14 biogeographic locations. Data from a single site in British Columbia, Canada, were then used to test the dispersal facilitation hypothesis, which postulates that synchronous bird–fruit abundance patterns result from deterministic seed dispersal processes (i.e. avian fruit consumption). Results showed that phenological synchrony is a geographically consistent pattern. However, fruit production occurred after peak periods of bird abundances in British Columbia. Although phenological patterns were asynchronous at this site, observational and experimental fruit removal patterns supported the dispersal facilitation hypothesis. Avian fruit consumption covaried with bird abundances, suggesting selection may favour earlier fruit production and increased phenological synchrony. Environmental data suggest that earlier fruit production is constrained by cold spring temperatures, which inhibit the activity of pollinators and earlier dates of fruit maturation. Overall, the results show that phenological synchrony is a geographically consistent pattern in seed dispersal mutualisms. However, decoupled bird–fruit abundance patterns may occur despite deterministic processes favouring phenological synchrony.
Avian seed dispersal mutualisms were once thought to be characterized by stable ecological relationships between birds and plants. It was argued that consistent bird–fruit interactions favoured the evolution of fruit characteristics to correspond with the foraging behaviour of fruit-eating birds (e.g. McKey, 1975; Thompson & Willson, 1979). However, recent studies have documented stochastic bird–fruit interactions, resulting from extensive spatial and temporal variability in the identity of bird species interacting with particular plant species (Guitián et al., 1992; Jordano, 1993, 1995; Sallabanks, 1993; Herrera, 1995, 1998). The unpredictability of bird–fruit interactions is likely to preclude pairwise co-evolutionary processes among particular plant and bird species. Consequently, patterns in fruit characteristics, which reflect efficient seed dispersal processes, are now believed to be rare (see Herrera, 1995, 1998; Levey & Benkman, 1999). However, one pattern in bird–fruit interactions has been observed consistently in avian seed dispersal systems. Positive correlations between annual fruit production and the abundance of fruit-eating birds have been documented in a diverse array of geographical locations (e.g. Herrera, 1984; Noma & Yumoto, 1997). The apparent ubiquity of bird–fruit ‘phenological synchrony’ suggests, therefore, some underlying structure to seed dispersal interactions. However, the geographical consistency of this relationship has yet to be globally synthesized.
The ecological processes responsible for phenological synchrony are also poorly understood. Consistent interactions among individual species within seed dispersal mutualisms are undoubtedly rare (see Herrera, 1995). However, stable relationships may exist among groups of birds and plants, as individual species may be functionally similar with respect to the mutualism (Jordano, 1993). Accordingly, the ‘dispersal facilitation’ hypothesis predicts that phenological synchrony results from co-evolutionary interactions among groups of plants and birds (Thompson & Willson, 1979). More specifically, selection favours plants producing fruit when birds are maximally abundant, as they probably receive increased rates of fruit consumption and subsequent seed dispersal (Thompson & Willson, 1979; Smith-Ramírez et al., 1998). Although it makes testable predictions for the ecological consequences of phenological synchrony, the dispersal facilitation hypothesis has received little empirical investigation.
In this study, the global consistency of phenological synchrony was evaluated in a meta-analysis of bird–fruit abundance data from 14 biogeographic regions. The dispersal facilitation hypothesis was then evaluated with field data from a previously uninvestigated site in British Columbia, Canada. At this site, temporal relationships between fruit production and fruit-eating bird abundances were evaluated to test for fine-scale phenological synchrony. Observational and experimental data on fruit removal rates were then used to critically evaluate the dispersal facilitation hypothesis, by testing whether avian fruit consumption covaries with avian abundances.
The geographical consistency of phenological synchrony was evaluated with data from 19 studies in 14 biogeographic regions (Table 1). Meta-analyses were conducted on monthly abundances of fruit-eating birds and the number of plant species producing ripe fruits. These variables were chosen because they were common to most studies suitable for analysis. Several relevant investigations were omitted because they investigated small numbers of species, reported only seasonal results or were conducted for less than 1 year. The month of peak fruit production and bird abundances (i.e. modes of annual distributions) were compared using non-parametric angular correlation (Zar, 1999). Three geographical regions (i.e. North-eastern United States, Spain and Costa Rica) were represented by several studies that were conducted in different regional locales (Table 1). To maintain spatial independence of the dataset, separate tests were made with all combinations of single sites from these three regions.
Table 1. Geographical location, collection methods and source of data used in meta-analysis. Abbreviations in the data column describe bird censusing methods followed by methods used to quantify fruiting phenologies
A: total abundance of fruit-eating birds, At: total number of birds, B-M: biomass, FA: total fruit abundance, m-n: mist net captures/h., S: total number of plant species bearing ripe fruits, s-t: seed trap, u-s: forest understorey, v: visual counts, SE: south-east, NE: north-east, W: west, years: number of years data were collected (average used in analysis), % proportion of total, ?: longitude and latitude not reported.
All field data were collected in the Pacific Rim National Park (PRNP), Vancouver Island, British Columbia, Canada (48°80′N, 125°20′W). Environmental conditions in the area are highly seasonal; summer and autumn experience mild temperatures and moderate precipitation, while almost continuous rainfall and cooler temperatures prevail in winter and spring. PRNP lies within the coastal western hemlock biogeoclimatic zone and is dominated by undisturbed conifer forest. Although the forest canopy is dominated by wind-dispersed conifers, bird-dispersed shrubs are common below the forest canopy (Table 2). Eight bird species are the primary fruit consumers in PRNP (Table 2). Three are migrants, which travel to the area in the summer to breed. Two species reside in PRNP throughout the year, but are present in greater abundances in summer. The remaining three species are residents that maintain fairly constant annual populations. All but the band-tailed pigeon (Columba fasicata) consume fruits opportunistically to supplement diets consisting mainly of invertebrate prey. More detailed descriptions of the avifauna, flora and environmental conditions of PRNP are reported in Campbell et al. (1997), Pojar & Mackinnon (1994) and Harding & McCullum (1994), respectively.
Bird–fruit abundance patterns were assessed with field observations of fruiting phenologies and a previously published dataset on bird abundances. The fruiting phenologies of the 10 most common bird-dispersed shrubs were quantified from January to December 1999. The fruiting stalks of six fruits from 12 plants of each species were marked with green tape and their status was evaluated at 10-day census intervals. Monthly bird abundances were obtained from a study conducted between 1972 and 1973 at several locations in PRNP (Hatler et al., 1978). To test whether bird–fruit abundance patterns were synchronized annually, monthly numbers of shrub species with ripe fruits present on marked plants were compared to monthly bird abundances with non-parametric angular correlation.
To test the dispersal facilitation hypothesis, patterns in the dates on which marked fruits were removed by birds were evaluated. Dates of fruit removal were established by inspecting the health of ripe fruits and the ground below marked plants. Preliminary observations indicated that: (1) characteristic markings were left behind on fruiting stalks after removal by birds; (2) fruits rarely fell from parent plants without showing signs of either desiccation, stem damage or pathogenic attack prior to fruit fall; and (3) following fruit damage, fruits were avoided by birds and usually fell from parent plants while attached to stems. Consequently, dates of fruit removal were quantified through observations of fruiting stalks and the prior status of fruits, which were then crossed-checked by inspecting the ground below marked plants. To examine whether removal rates differed across the fruiting season, two-factor analysis of variance (ANOVA) was used to test for temporal differences in the proportion of fruits removed from each plant. Each marked plant was assigned to one of two ripening categories. The six plants with the earliest average ripening date were categorized as ‘early’ while all later fruiting plants were categorized as ‘late’. To account for the confounding effect of species, the species identity of each plant was also included as a factor in the ANOVA. Therefore, both factors (species and ripening category) were fully crossed and not considered fixed.
A fruit removal experiment, using fruits collected from other locales in southern British Columbia, was also conducted to test the dispersal facilitation hypothesis. The region has a diverse topography and climate; therefore it was possible to collect ripe fruits from other locations both before and after natural periods of fruit production in PRNP. Fruits of both Rubus spectabilis and Sambucus racemosa ripened prior to PRNP at a rain-shadow site, 100 km east of the study site (48°80′N, 123°°60′W). They ripened after their respective fruiting intervals at a high elevation site 50 km south-east (48°50′N, 124°50′W), and at the same time 10 km east of PRNP. Three trials were conducted with each species, using fruits collected from these locales. The first trial was conducted approximately 20 days prior to the natural fruiting interval of each species at PRNP (‘early’). Importantly, the early trial in Rubus spectabilis was conducted prior to fruit maturation of any species in PRNP. The ‘middle’ fruiting trial was conducted during their natural, approximate peaks in fruit production, and the ‘late’ trial was conducted 20 days after each species’ fruiting interval concluded. During each trial, three fruits of R. spectabilis or 50 fruits of S. racemosa were fastened to a single L-bracket attached to 12 distantly spaced conifer trees. Trials were left in the field for 2 days and the proportion of fruits removed from each L-bracket was compared between trials of each species separately with Mann–Whitney U-tests with Bonferroni corrections.
Results from the meta-analysis showed a strong positive association between peak fruit production and fruit-eating bird abundances (Fig. 1). Positive relationships were found when all data points were included (n = 19 (raa)s = 0.438, P < 0.01) and never dropped below (n = 14 (raa)s = 0.290, P < 0.05) when individual sites in NE United States, Costa Rica and Spain were analysed separately. However, four regions had peak bird abundances that differed from peak fruit production by more than 2 months (Israel, 2; Malaysia, 3; British Columbia, 7 and Costa Rica, 11). Fine-scale patterns in bird–fruit abundances in British Columbia were also asynchronous (Fig. 2). Monthly bird abundances were unrelated to the number of plant species producing fruit each month ((raa)s = –0.165, P > 0.2).
Despite asynchronous bird–fruit abundances, fruit removal rates covaried with temporal differences in avian abundances. The proportion of fruits removed from each plant marked in phenological observations differed among species (F = 3.2, P = 0.002) and phenological categories (F = 6.1, P = 0.015). Both factors did not interact (F = 1.0, P = 0.422), indicating that the effect of phenological categories was similar among species. Therefore, plants producing fruit earlier in the season had higher removal rates (Fig. 3).
Similar results were observed in the fruit translocation experiment (Fig. 4). The proportion of fruits removed in R. spectabilis was higher in the early trial relative to both the middle (U′ = 135, P < 0.001) and late trial (U′ = 131, P < 0.001). However, the middle and late trials did not differ (U′ = 96, P = 0.184). Similarly, S. racemosa fruits were removed in higher proportions in the early (U′ = 118, P = 0.002) and middle (U′ = 124, P < 0.001) trials relative to the late trial, but the early and middle trials did not differ (U′ = 81, P = 0.594).
Results from the meta-analysis showed that phenological synchrony is a geographically consistent, coarse-scale pattern in avian seed dispersal mutualisms. Corresponding peaks in fruit production and bird abundances indicate that bird–fruit interactions are related consistently over large scales of space and time. However, correlated peaks of fruit production and bird abundances do not necessarily imply that annual distributions of bird–fruit abundances are tightly correlated. Fine-scale patterns in fruit production may be quite different from patterns in avian abundances, despite overlapping peaks in their yearly distributions. It should therefore be reiterated that meta-analyses only provide support for coarse-scale patterns in phenological synchrony.
Although most biogeographical regions supported phenological synchrony, four sites deviated from the overall pattern. At two of these sites (Israel and Costa Rica), fruiting peaks coincided with secondary annual peaks in bird abundances (Izhaki & Sarfriel, 1985; Loiselle & Blake, 1991). Therefore, fruit production occurred during elevated periods of avian abundances, albeit during smaller peaks within annually bimodal bird distributions. In contrast, fruit production and bird abundances showed very weak monthly fluctuations in Malaysian dipterocarp forests (Wong, 1986). Consequently, dates obtained for this site do not appear to represent true annual differences in either fruit production or bird abundances.
Field data from British Columbia were also asynchronous. However, both fruit production and bird abundances showed a single, well-pronounced annual peak. Therefore, the processes responsible for asynchrony in British Columbia appear to be different from the three sites discussed above. Observed phenological patterns may be influenced by the collection of bird and fruit data in different years. Several studies have demonstrated that fruiting phenologies are sensitive to environmental fluctuations (Gorchov, 1987; Fuentes, 1992; Guitián, 1998). Therefore, annual climatic differences may have biased observed phenological patterns. However, a short-term dataset, documenting seasonal bird–fruit abundances in PRNP from the same year, corroborates results reported here (K.C. Burns, unpublished). These data were not reported because they were collected for less than a year. They illustrate a clear, but somewhat less pronounced, pattern in phenological asynchrony. It is possible that the use of data from different years exaggerates the magnitude of phenological asynchrony, but this does not explain the results.
Patterns in fruit removal supported the dispersal facilitation hypothesis. Both observational and experimental rates of fruit removal declined through time concurrently with avian abundances. Moreover, experimental fruit consumption rates were highest in May, prior to fruit production by any species at PRNP. This indicates that birds would consume fruits rapidly if they ripened in spring, prior to observed ripening times. It also suggests that increased avian fruit removal early in the fruiting season may select for earlier dates of fruit maturation and increased phenological synchrony.
Two ecological processes could provide an explanation for phenological asynchrony. First, coupled bird–fruit abundance patterns may be attributed to the movement patterns of fruit-eating birds relative to fruit resources (i.e. ‘resource tracking’; see Blake & Loiselle, 1991; Levey & Stiles, 1992; Rey, 1995; Kannan & James, 1999). Consequently, constraints to the capacity of birds to track temporal fluctuations in fruit resources in PRNP may account for phenological asynchrony. Most migratory birds arrive in PRNP in May, and apparently persist on a diet of insects until fruits become available (except the band-tailed pigeon, Columba fasicata, which feeds on young shoots and buds). A protein-rich diet of insects is probably necessary for successful avian reproduction (Thompson & Willson, 1979; Stapanian, 1982). The arrival of birds prior to fruit production appears to be adaptive, and phenological asynchrony cannot be explained by the movement patterns of fruit-eating birds alone.
Alternatively, several factors external to bird–fruit interactions may constrain earlier fruit production. Phylogenetic inertia is frequently invoked to explain patterns in fruit characteristics (see Herrera, 1995; Eriksson & Ehrlén, 1998; Smith-Ramírez et al., 1998). However, phylogenetic relatedness does not appear to constrain fruit ripening times at PRNP, as several plant families are represented by species with very different fruit ripening times. For example, R. spectabilis, R. parviflorus and Pyrus fusca are all in the family Rosaceae, but are the first, fourth and tenth species in the fruiting sequence, respectively (K.C. Burns, unpublished). Therefore, fruiting patterns do not appear to be overly influenced by phylogenetic constraints.
On the other hand, earlier fruit production may be constrained by the strongly seasonal climate of PRNP. All shrub species are pollinated by either insects (mainly Hymenoptera) or hummingbirds (i.e. Selasphorus rufus), which migrate to PRNP in early May (Hatler et al., 1978). Both taxa are temperature sensitive and their activity is undoubtedly constrained by the cool, wet weather of early spring (see Pojar, 1974; Hatler et al., 1978). Most shrub species flower in mid-May, after pollinators become seasonally active or migrate to PRNP, and the earliest ripening shrub species takes more than a month to mature (K.C. Burns, unpublished). Therefore, environmental constraints on the pollination process provide a probable explanation for constraints to earlier fruit production and synchronous bird–fruit abundance patterns.
Although this study focused on geographical variation in phenological synchrony, stochastic phenological relationships have been documented over longer time-scales at individual locales. In particular, work by Herrera (1998) has been cited as evidence against the generality of phenological synchrony (Levey & Benkman, 1999). Herrera’s study showed that interannual variation in fruit-eating bird abundances was related more strongly to environmental conditions than to the abundances of fleshy fruits. However, these data document among-year variation in bird–fruit abundance patterns during annual peaks in abundances, not within-year variation. In other words, they evaluate relationships in the amplitude of annual cycles, not relationships among their wavelengths. As a result, Herrera’s study does not test the generality of phenological synchrony, as defined above.
More than 20 years ago, Thompson & Willson (1979) documented fruit consumption patterns that covaried with bird abundances in eastern North America. From these results, they predicted that fruits produced during periods of elevated avian abundances should have higher probabilities of consumption and subsequent seed dispersal. Observational and experimental results from this study support these predictions, even though precise patterns in phenological synchrony were not observed. Phenological asynchrony in British Columbia appears to be anomalous, not only because de-coupled abundance patterns are uncommon, but also because fruit consumption patterns appear to favour the evolution of phenological synchrony. Results from this study therefore support predictions that deterministic seed dispersal processes can occur among assemblages of fruit-eating birds and fleshy-fruited plants, over large spatial and temporal scales.
Funding for this study was provided by grants from the Bamfield Marine Station, the Canadian Government, the Fulbright Foundation and The University of California, Los Angeles. The manuscript was greatly improved by helpful advice and comments from J.L. Dalen.
K.C. Burns is a graduate student at UCLA conducting research on terrestrial plant ecology on Vancouver Island, Canada. His research focuses on niche relationships among bird-dispersed plants and how avian seed dispersal influences the distribution of bird-dispersed plants on islands off the coast of British Columbia.