Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Knowledge of the spatial scale of the dispersal service provided by important seed dispersers (i.e. common and/or keystone species) is essential to our understanding of their role on plant ecology, ecosystem functioning and, ultimately, biodiversity conservation.
Carnivores are the main mammalian frugivores and seed dispersers in temperate climate regions. However, information on the seed dispersal distances they generate is still very limited. We focused on two common temperate carnivores differing in body size and spatial ecology – red fox (Vulpes vulpes) and European pine marten (Martes martes) – for evaluating possible functional diversity in their seed dispersal kernels.
We measured dispersal distances using colour-coded seed mimics embedded in experimental fruits that were offered to the carnivores in feeding stations (simulating source trees). The exclusive colour code of each simulated tree allowed us to assign the exact origin of seed mimics found later in carnivore faeces. We further designed an explicit sampling strategy aiming to detect the longest dispersal events; as far we know, the most robust sampling scheme followed for tracking carnivore-dispersed seeds.
We found a marked functional heterogeneity among both species in their seed dispersal kernels according to their home range size: multimodality and long-distance dispersal in the case of the fox and unimodality and short-distance dispersal in the case of the marten (maximum distances = 2846 and 1233 m, respectively). As a consequence, emergent kernels at the guild level (overall and in two different years) were highly dependent on the relative contribution of each carnivore species.
Our results provide the first empirical evidence of functional diversity among seed dispersal kernels generated by carnivorous mammals. Moreover, they illustrate for the first time how seed dispersal kernels strongly depend on the relative contribution of different disperser species, thus on the composition of local disperser assemblages. These findings provide a key starting point for understanding and modelling plant population processes that include mammal-mediated seed dispersal, such as connectivity, range expansion and colonization.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The spatial scale at which seed dispersal occurs determines the distribution, dynamic and genetic structure of plant populations and, therefore, of plant species assemblages (see reviews by Cain, Milligan & Strand 2000; Nathan & Muller-Landau 2000; Levin et al. 2003; Levine & Murrell 2003; Nathan et al. 2008). In the case of endozoochorous plants, seeds are usually dispersed by a diverse array of vertebrate frugivores (e.g. Herrera 1984; Wheelwright 1985). Differences in traits such as body size, diet and spatial behaviour are responsibly for variations between frugivore species in the seed dispersal kernels they generate (e.g. Dennis, Wescott et al. 2007; Jordano et al. 2007; Spiegel & Nathan 2007), understanding as kernel the probability density function of dispersed seeds relative to distance from maternal plants (Levin et al. 2003). In fact, some disperser species typically deposit most seeds at short distances, while others do it at long distances (e.g. Jordano et al. 2007; Spiegel & Nathan 2007). Such differences have important consequences on the functional role of each animal-vector on gene (genotypes) flow patterns and spread potential of plant populations. In this line, short-distance seed dispersers would be promoting gene flow and recruitment within populations, while long-distance dispersers would be promoting gene flow among populations and tracking vacant areas for colonization (Cain, Milligan & Strand 2000; Nathan et al. 2008). Thus, a detailed knowledge of the seed dispersal kernel generated by important dispersers (i.e. common and/or keystone species; Gaston 2010) is essential for our understanding of their role on plant ecology, ecosystem functioning and, ultimately, biodiversity conservation (Lundberg & Moberg 2003; Trakhtenbrot et al. 2005; Montoya et al. 2008).
Carnivorous mammals (Order Carnivora; hereafter carnivores) are the main mammalian ‘frugivores-seed dispersers' in temperate climate regions (Debussche & Isenmann 1989; Herrera 1989; Willson 1993; López-Bao & González-Varo 2011), where they can feed on a large fraction of fleshy-fruited plant assemblages (e.g. 40% in south-eastern Spain; Herrera 1989). They are considered long-distance seed dispersers as compared with distances mediated by resident (non-migrating) passerine birds (Hickey et al. 1999; Jordano et al. 2007; Koike et al. 2011). Furthermore, many medium-sized carnivores (e.g. foxes, martens and badgers) are common species that can inhabit both undisturbed and anthropogenic landscapes (Pereira & Rodríguez 2010; López-Bao & González-Varo 2011). However, despite the key services they likely provide across a broad array of ecosystems, we still know very little about the seed dispersal kernels mediated by such species (but see Hickey et al. 1999; Jordano et al. 2007) and, to our knowledge, no study has yet evaluated differences between kernels generated by different carnivore species.
To date, the most relevant study on this issue was carried out by Jordano et al. (2007) who estimated dispersal distances for a guild composed of red fox, stone marten and Eurasian badger – but the three species were pooled for analyses. Using molecular markers and maternal assignments, Jordano et al. (2007) estimated that carnivores dispersed most seeds between 650 and 700 m and suggested longer dispersal events (>1500 m) because they were unable to identify the maternal origin for a subset of genotyped seeds (see Jordano 2007 for a likelihood-based analysis of this sample). In fact, molecular studies have several limitations in their application to long-distance seed dispersers owing to the exponential decrease of maternal assignments when increasing the distance from genotyped plants (Ashley 2010). In addition, they generally lack from an explicit sampling strategy aiming to detect long-distance dispersal (LDD) events. For example, we are unable to discern if those dispersal events above 1500 m estimated by Jordano et al. (2007) represent the tail of an unimodal curve or, alternatively, a portion of a more complex multimodal curve (e.g. Russo, Portnoy & Augspurger 2006; Lenz et al. 2011).
Certainly, measuring animal-mediated seed dispersal distances is particularly challenging. Methods such as direct observations of animal movements (Gómez 2003; Heymann et al. 2012) and molecular markers are unfeasible if we are to measure seed dispersal kernels generated by carnivores, which show an elusive behaviour and large spatial requirements (Gittleman et al. 2001). Models combining animal movement (inferred by telemetry studies) and gut retention times (trials in captive animals) have been used to estimate seed dispersal distances in birds (Holbrook & Smith 2000; Spiegel & Nathan 2007; Lenz et al. 2011), mammals (Hickey et al. 1999; Campos-Arceiz et al. 2008; Koike et al. 2011) and lizards (Santamaría et al. 2007). However, the alteration of the animal's activity patterns imposed by space limitation offers a good reason to be sceptical about gut retention times obtained in captive carnivores. More important is the fact that carnivores use their faeces for territorial marking (Macdonald 1985; Gorman 1990), which must also contribute to weaken the correlations between gut retention times measured in captivity and real seed dispersal distances generated in the wild, probably highly influenced by this carnivore behaviour.
Here, we present for the first time real measures of the whole range of seed dispersal distances mediated by carnivorous mammals based on a 3-year field experiment. We used colour-coded seed mimics embedded in common figs (Ficus carica) that were offered to carnivores in feeding stations simulating source trees. The exclusive colour code of each simulated tree allowed us to assign the exact origin of seed mimics found later in carnivore faeces. Furthermore, we designed a sampling strategy aimed to detect the longest dispersal events. The two main objectives of this study were (i) to measure the overall seed dispersal kernels generated by carnivores and (ii) to evaluate possible differences among carnivore species (i.e. functional heterogeneity) in their seed dispersal kernels. We focused on two common temperate carnivores: red fox (Vulpes vulpes, Canidae; 3–14 kg; home range: 200–600 ha) and European pine marten (Martes martes, Mustelidae; 0·8–1·8 kg; home range: 50–150 ha, Cavallini 1996; Dekker, Stein & Heitkönig 2001; Zalewski, Jedrzejewski & Jedrzejewska 2004; Wilson & Muttermeier 2009). Given that both species differ in body size and home range, we expected that foxes would generate a more extensive seed dispersal kernel than martens.
Materials and methods
The study was carried out in the Devesa da Rogueira woodland (800–1400 a.s.l.), located in Serra do Courel (42°37′N–7°05′W), a mountainous area (c. 250 km2) of the northwest Iberian Peninsula (Fig. S2a). The natural vegetation of the study site consists of mixed woodlands (Betula alba, Quercus robur, Q. petraea, Fagus sylvatica, Ilex aquifolium, Sorbus aucuparia and Taxus baccata) and heathlands (Erica spp.). Four carnivore species – red fox, pine marten, Eurasian badger (Meles meles, Mustelidae) and common genet (Genetta genetta, Viverridae) – consume fruits regularly in this locality (López-Bao & González-Varo 2011J.P. González-Varo, J.V. López-Bao & J. Guiti\xE1n, unpublished data). Fleshy-fruits comprise an important part of their diet in this region from summer to winter (Guitián & Munilla 2010; López-Bao & González-Varo 2011), as has been observed in other areas of Europe (reviewed in Rosalino & Santos-Reis 2009). In the Serra do Courel alone, carnivores consume fruits and disperse seeds of at least 11 wild and six cultivated fleshy-fruited species (López-Bao & González-Varo 2011; J.P. Gonz\xE1lez-Varo, J.V. L\xF3pez-Bao & J. Guiti\xE1n, unpublished data); all of them are woody species, mostly trees.
The Feeding Experiment
We offered experimental fruits containing coloured seed mimics to the carnivores in three different years (2008, 2009 and 2010). We used figs of the common fig tree (Ficus carica) as experimental fruits (c. 50 mm diameter; Fig. S1). We selected this fruit because it is frequently consumed by carnivores in the study area (see López-Bao & González-Varo 2011) and also because recruitment of F. carica seedlings is null in the wild (J.P. González-Varo, J.V. López-Bao & J. Guitián, pers. observ.), which ensures that the experiment did not spread a cultivated species towards natural habitats. Seeds mimics were coloured plastic beads (see Fig. S1) of (mean ± SD) 2·6 ± 0·1 mm diameter and 27·1 ± 2·8 mg weight (n = 20, see similar mimics in Varela & Bucher 2006; Koike et al. 2011). We did not detect a loss of original colours after immersing them in concentrated solutions of HCl in the laboratory denoting that (i) such colour is correctly identified after gut passage and (ii) the colorant is not released in the animal's guts (see Fig. S1). Furthermore, none seed mimic was subsequently found broken in pieces in carnivore faeces. The size of such mimics is within the modal range of wild seeds dispersed by carnivores in the study area (whole range, from c. 0·2 mm diameter in bilberry Vaccinium myrtillus to c. 10 mm length in blackthorn Prunus spinosa; López-Bao & González-Varo 2011). Varela & Bucher (2006) and Koike et al. (2011) have reported in foxes and bears, respectively, that gut passage time did not vary significantly either among different seed species or different plastic marks, so we considered that dispersal distances at which carnivores deposited our seed mimics were similar to those distances at which they deposit seeds naturally. We prepared the experimental figs embedding seed mimics in their pulp (10 seeds per fig) with the help of tweezers. The offerings were performed between September and October of every year, coinciding with the peak of the fruiting season of fleshy-fruits in the study area.
Every September, we set up ‘feeding stations’ where we offered experimental figs simulating fruits on the ground below the canopy of individual fruiting trees (Fig. S2). Each feeding station consisted of an area of c. 60 m2 where a total of six feeders (aluminium trays of 18 × 12 × 3 cm) were placed at both sides of a path (see Figs S1 and S2). We geo-referenced each feeding station using the centroid of the polygon generated by the spatial position of the six feeders. Every year, an exclusive colour code was assigned to the seed mimics offered in each feeding station. In this manner, we were able to assign directly the source (feeding station) of any seed mimics dispersed and subsequently found in carnivore faeces (see analogous method in Mack 1995). In 2008, we set up three feeding stations (I, II and III) that were also used in 2009 and 2010 (Table S1; see Fig. S2). We set up three additional feeding stations in 2009 (IV, V and VI) plus other two in 2010 (VII and VIII) making a total of eight feeding stations (n = 3, 6 and 6 feeding stations in 2008, 2009 and 2010, respectively; see details in Table S1). Feeding stations were placed at different locations along the main paths of the study site, with distances between them ranging between 280 m and 1760 m (Fig. S2).
In 2008, which was the first year and served as pilot study, we offered experimental fruits twice (2 weeks between offerings), placing six experimental fruits (figs) per feeder in each offering (n = 36 fruits per feeding station × 2 offerings). In 2009 and 2010 (main study years), we initially performed the same offering (n = 36 fruits per feeding station), but during the next 2 weeks, we increased fruit offerings in those feeders in which carnivores removed fruits previously to enhance fruit consumption and thus the probability of seed recapture. In each study year, we placed a few sardines in olive oil near the trays at the beginning of offerings to facilitate their location by carnivores. Given that (i) we typically offered less than 100 figs per feeding station and year (mean = 88, SD = 30) and (ii) our offerings overlapped with the fruiting peak of many fleshy-fruited plants consumed by carnivores in the region (with crops of hundreds or even thousands of fruits), we considered negligible any effect of the experiment on the spatial behaviour of animals involved due to a possible habituation to food supply, a process that requires long-time periods in carnivore species (López-Bao, Rodríguez & Palomares 2008).
Sampling Strategy: Searching for the Longest Dispersal Distances
Once fruits were offered, we monitored fruit removal from the feeding stations and searched for carnivore faeces twice a week from September to November. Faeces were intensively searched along the extant network of available mountain paths surrounding the feeding stations (see maps in Fig. S3). Previous studies in the study area suggest that both species deposit most of their faeces along paths (López-Bao & González-Varo 2011; J.P. González-Varo, J.V. López-Bao & J. Guitián, unpublished data).
Every carnivore faeces was broken and analysed in situ and its position was geo-referenced using a GPS. Faeces were identified by trained people at the carnivore species level according to shape, size and smell in combination, a procedure that is commonly used for the identification of carnivore faeces (e.g. Guitián & Munilla 2010; López-Bao & González-Varo 2011). Carnivore identification was hugely facilitated by the fact that all the faeces that contained seed mimics were defecated only a few days before. Faeces that could not be properly assigned to any carnivore species were classified as ‘non-identified’. We calculated the seed dispersal distances from UTM coordinates of both the faeces where seed mimics were found and their corresponding feeding stations. Given the pronounced slopes in our mountain landscape, we calculated real distances (using x, y and z coordinates) instead of Euclidean distances (only x and y), which were on average 2·0% lower.
We used two criteria for searching the longest seed dispersal events. First, we searched for faeces along all available paths within an area generated by merging 2-km buffers centred on all the feeding stations used each year. Second, once we found carnivore faeces with seed mimics, we enlarged the minimum distance sampled around each feeding station 1-km from the outermost scat with seed mimics belonging to that station. We continued applying the second criterion until we found no faeces with seed mimics within the 1 km ring around the distance from the outermost faeces with seed mimics (see scheme and details in Fig. S4). The size of the path network sampled was of c. 70 km length over an area of c. 40 km2, and the overall walking distance after all surveys was c. 900 km.
To evaluate biases in our sampling efforts in relation to the distance to the fruit sources (feeding stations), we obtained the number of km sampled within 20 non-overlapping concentric rings at intervals of 200 m up to 4000 m. To do this, we generated a map with the path network used to search for carnivore faeces using ArcGIS 9 (Esri Inc., Redlands, CA, USA). We also calculated the number of km sampled per ring area (km km−2) as a relative estimate of our sampling effort at each distance class (see Fig. S4). We estimated the sampling effort around each feeding station and the overall sampling effort by averaging all feeding stations. Sampling effort estimates were calculated separately for the main study years (2009 and 2010).
We obtained the frequency distribution and descriptive statistics of seed dispersal distances generated by both the whole guild of carnivores and the two target species (red fox and pine marten). We built 200-m distance classes histograms, a spatial scale that combined both high resolution in frequency distributions and the presence of seeds in almost all distance classes within the dispersal range. All analyses were performed at the seed level as we found non-significant differences in dispersal distances between faeces containing seed mimics and individual seed mimics (P > 0·1 in all cases; see details in Fig. S5). We tested for statistical differences in both average seed dispersal distances (using Mann–Whitney U-tests) and distance distributions (using Kolmogorov–Smirnov tests) between both carnivore species. We also tested for differences between the 2 years for which we obtained most of our data set (2009 and 2010).
We defined a priori an absolute definition (sensu Nathan et al. 2008) of long-distance dispersal (LDD) as those dispersal events longer than 1 km. We set this threshold in accordance with (i) the spatial scale of many fleshy-fruited plant (meta) populations, often separated from each other by many hundreds of metres or few kilometres (e.g. Jordano 2007; Spiegel & Nathan 2007) and (ii) bird-mediated seed dispersal distances, typically below 1 km from source trees (e.g. Jordano et al. 2007). We calculated the proportion (%) of LDD events generated by both the whole carnivore guild and the two target species.
We fitted probability density functions to describe mathematically the seed dispersal distributions generated by the carnivore guild, the red fox and the pine marten. We fitted the following density functions to the data: exponential, normal, lognormal, Student's t, gamma and Weibull. We estimated the parameters of each distribution based on maximum-likelihood estimation. Because some of our empirical distance distributions showed multiple peaks (up to four peaks; see Results), we also fitted mixture distributions. In such cases, we first partitioned the data into unimodal distributions based on visual inspection (see also Russo, Portnoy & Augspurger 2006; Lenz et al. 2011). Second, we fitted the different probability density functions to each of these data partitions separately (see also Higgins, Nathan & Cain 2003). Third, we fitted mixture distributions using the parameters estimated of best-fitting unimodal (partitions) distributions as starting points. The Akaike Information Criterion (AIC) was used for model selection (Burnham & Anderson 2002). In the case of multimodal distance distributions, we selected the best model by comparing AIC-values of unimodal and multimodal (with two, three and four peaks) density functions. All kernels were fitted with R 2·13·1 (R Development Core Team) using the packages ‘fitdistrplus’ (ver. 0·1–2, Delignette-Muller et al. 2011), ‘bblme’ (ver. 1·0·3, Bolker 2011a) and ‘emdbook’ (ver. 1·3·1, Bolker 2011b).
Finally, we tested for associations between seed recapture and sampling effort (absolute and relative) in relation to the distance with Spearman's correlations. Correlations were performed using those 200-m distance classes included within the whole range where we found seed mimics (i.e. all < 3000 m; n = 15 rings of 20).
Fruit Removal and Recapture of Seed Mimics
Each year, most (>90%) figs offered were removed from the trays at the end of the experiment. The norm was that whole figs disappeared from the trays, indicating that large animals such as carnivores had consumed them. However, we found signs of removal by wild boars (Sus scrofa) in a feeding station (VIII in 2010; Table S1), which caused important damages to the trays. Exceptionally, we found some figs partially consumed by birds (peaked and with bill marks) and small rodents (gnawed and with teeth marks).
We found and analysed 2027 carnivore faeces during the whole study period (641 in 2008, 888 in 2009 and 498 in 2010). We recovered a total of 665 seed mimics found in 98 carnivore faeces (see Table S1). These seed mimics accounted for c. 5% of seeds offered (665 of 13 220 seeds). We found 53 seeds in 10 faeces in 2008 (8% of all seeds), 447 seeds in 63 faeces in 2009 (67%) and 165 seeds in 25 faeces in 2010 (25%). By carnivore species, 63 faeces belonged to red fox (451 seeds), 32 to pine marten (206 seeds), 2 to badger (six seeds) and 1 (2 seeds) was classified as ‘non-identified’ but also considered for the whole assemblage of carnivores. Notably, red fox (68%) and pine marten (31%) accounted for c. 99% of all seeds recovered. A chi-squared contingency test revealed that the frequency of seeds dispersed by these two species differed significantly between 2009 and 2010 (χ2 = 114, P < 0·001); while most seeds recovered in 2009 were dispersed by red fox (84%), most seeds recovered in 2010 were dispersed by pine marten (59%). Remarkably, 10·2% (10 of 98) of faeces with seed mimics had seeds belonging to two different stations, being distances between such stations of 410–1220 m (mean = 575 m); seven were from red fox, two from pine marten and one from Eurasian badger.
There were nonsignificant differences between foxes and martens in the number of seed mimics found in their faeces (MW U-test: Z = 0·22, P = 0·825). It ranged from 1 to 30 seeds in faeces of foxes (mean = 6·4, median = 3, Q25–Q75 = 2–9) and from 1 to 25 seeds in those of martens (mean = 6·1, median = 4, Q25–Q75 = 2–8).
In 68% of faeces containing seed mimics, we only found seeds and fruit skins of common figs used for the experiment. In 24%, we found seeds of at least six fleshy-fruited species (five wild species and one cultivated species). In 12%, we found prey remains, mainly of small mammals, passerine birds and insects. In 60% of faeces analysed without seed mimics, we found the same seeds species (including common fig F. carica) plus other nine species (five wild species and four cultivated species).
Seed Dispersal Kernels
Seed dispersal distances generated by the whole carnivore guild ranged from 7 to 2846 m, with mean and median distances of 1035 and 822 m, respectively (see Table 1 and Fig. 1a). Forty per cent of seeds belonged to LDD events as they were recovered at distances longer than 1 km. Furthermore, 45% of LDD events (18% of all seeds) were recovered at distances above 2 km. The kernel was characterized by the presence of several peaks of decreasing size. In fact, multimodal functions fitted better to empirical data (based on AIC) than unimodal ones (see details in Fig. S6a). The best-fitting density function had four peaks (tetramodal) described by Weibull, log-normal, normal and Student's t distributions, respectively (Table S2, Fig. 1a). As reported above, 68% and 31% of seeds of this kernel were dispersed by red foxes and pine martens, respectively. However, their contribution was not similar along the kernel but it was spatially structured, with martens being responsible of most of the shortest dispersal events and foxes responsible of the longest ones (Fig. 1a).
Table 1. Summary statistics of dispersal distances (m) generated by the studied carnivore guild and the two target species
Statistics are given for all study years pooled (2008, 2009 and 2010) and separately for 2009 and 2010 (main years). ‘Max.’ is the maximum dispersal distance recorded. ‘LDD (%)’ is the percentage of long-distance dispersal events defined a priori as distances longer than 1 km.
More importantly, we found strong differences between foxes and martens in the seed dispersal kernels they generated (see Fig. 1), both in terms of average values (MW U-test: Z = 13·8, P < 0·001) and distance distributions (KS-test: P < 0·001). Dispersal distances generated by foxes ranged from 56 to 2846 m, with mean and median distances of 1321 and 1101 m, respectively (see Table 1 and Fig. 1b). In addition, more than a half (52·5%) of fox-dispersed seeds belonged to LDD events, half of which (50·2%) were recorded at distances above 2 km (26·4% of fox-mediated seed dispersal). The kernel showed several peaks of similar size and, again, multimodal functions fitted better (based on AIC) than unimodal ones (see Fig. S6b). The best-fitting density function had four peaks described by log-normal, Weibull, normal and Student's t distributions, respectively (Table S2 and Fig. 1b). By contrast, dispersal distances generated by martens ranged from 7 to 1233 m, with mean and median distances of 408 and 260 m, respectively, and only 11·2% of LDD events (Fig. 1c). The shape of the kernel generated by pine marten was described by a gamma distribution and was characterized by a tail that drops at short (<1 km) distances (Table S2 and Fig. 1c). Finally, it is worth mentioning that six seeds mimics belonging to three different feeding stations and recovered in two faeces of Eurasian badger were found at 1113, 1473 and 2209 m (Fig. 1a).
Interannual Variations in Seed Dispersal Kernels
As pointed above, the contribution of each species to the kernel generated by the whole carnivore guild was significantly different in 2009 and 2010 (see above), which could be considered fox- and marten-dominated years, respectively (see Fig. 2). As a result, there were significant between-year differences in both distance distributions (KS test: P < 0·001) and average dispersal distances (MW U-test: Z = 7·8, P < 0·001). Particularly, dispersal distances and LDD events were on average approximately twofold higher in 2009 compared with 2010 (see summary statistics in Table 1). Moreover, the kernels in 2009 and 2010 looked like ‘fox-mediated’ and ‘marten-mediated’ kernels, respectively (see Fig. 2). In 2009, the best-fitting density function had four peaks described by log-normal, log-normal, normal and Student's t distributions, respectively (Table S2; see Fig. S6c). In 2010, the best-fitting density function had two peaks (one large- and one small-peak) described by exponential and Student's t distributions, respectively (Table S2; see Fig. S6d).
Regarding the red fox, we found significant differences in distance distributions (KS test: P < 0·05) but not in average dispersal distances (MW U-test: Z = 0·2, P = 0·85) between 2009 and 2010 (see Table 1). In the case of pine marten, we found significant between-year differences in both distance distributions (KS test: P < 0·001) and average dispersal distances (MW U-test: Z = 7·6, P < 0·001), being seeds dispersed longer in 2009 than in 2010 (see Table 1).
Several lines of evidence indicated that measured seed dispersal distances were not a consequence of sampling effort biases. First, the absolute (km) and relative sampling effort (km km−2) per distance class (mean ± SD) were, respectively, of 2·80 ± 1·15 km and 1·93 ± 1·59 km km−2 in 2009, and of 2·75 ± 1·02 km and 1·87 ± 1·65 km km−2 in 2010 (non-significant between-year differences: MW U-test: Z < 0·24, respectively, P > 0·80; see Fig. 3). Second, in both years, absolute effort reached its maximum values between 1200 and 3000 m and dropped at longer distances, in accordance with maximum seed dispersal distances recorded and our sampling strategy (Fig. 3). As a consequence of ‘radius/area’ (km km−2) relationships, the relative effort reached its maximum value at the shortest radii and then dropped very slowly from 400 to 4000 m (Fig. 3).
Finally, in 2009, the frequency of seed mimics found in each ring (n = 15 rings) was not correlated either with absolute or relative sampling effort (Spearman's rs = −0·19 and 0·40, P = 0·50 and 0·14, respectively). However, in 2010, the correlation with absolute effort was negative and nearly significant (rs = −0·51, P = 0·052), and positive and significant with relative effort (rs = 0·85, P < 0·001). These correlations were associated to the shape of the kernel in 2010, which was supported by the lack of significant correlations between the frequency of seed mimics dispersed by the red fox (i.e. the long-distance disperser) and both sampling effort metrics in 2009 and 2010 (all rs < |0·28|, P > 0·30). Clearly, our sampling strategy achieved a substantial sampling effort well over (>1 km) the longest dispersal event recorded (Fig. 3).
By tracking dispersed seed mimics that were offered in a feeding experiment, we measured for the first time the (virtually) whole range of real seed dispersal distances generated by two widespread carnivores that act in many ecosystems as frugivores and seed dispersers. Following an explicit sampling strategy, we substantially searched for dispersed seeds well over (more than 1 km) the longest dispersal event recorded (Fig. 3), which provide a strong reliability to the kernels we report. As far we know, this is the most serious, rigorous and robust sampling scheme followed for tracking carnivore-dispersed seeds.
We found that the kernel generated by the carnivore guild was characterized by multimodality and also by a huge fraction of LDD events (Fig. 1). More interestingly, we found a marked functional heterogeneity in the seed dispersal kernels generated by the two studied carnivore species: multimodality and LDD in the case of the red fox and unimodality and short-distance dispersal in the case of the pine marten (Fig. 1). As a consequence, emergent kernels at the guild level were highly dependent on the relative contribution of each carnivore species (Figs 1 and 2). Our study thus demonstrates that seed dispersal patterns strongly depend on the composition of local disperser assemblages.
The number of seed mimics recovered accounted for c. 5% of those offered. Notice however that this is a gross estimate of recapture rate (i.e. offered/recovered), thus, lower than the net percentage (i.e. ingested by carnivores/recovered), as we are not fully confident that all offered fruits were consumed by the target carnivores. Despite the fact that a large proportion of seed mimics seems to have been deposited in places outside the path network sampled (for example in badger or genet latrines), this gross estimate is within the range of recapture rates of other marking methods used to measure animal movement – such as bird ringing (typically < 1%; Frías, Serradilla & Escudero 2009) and butterfly tagging (4–41%; Ricketts 2001). On the other hand, the fact that gut passage times of seed mimics are similar to those of real seeds regardless of their size (see Varela & Bucher 2006; Koike et al. 2011) makes reliable extrapolating dispersal distances to the assemblage of fleshy-fruited plants dispersed by carnivores. Finally, it is worthy to mention that our method is much cheaper than methods that combine gut retention times and animal movement (which require from animal capture, space for performing the experiments in captivity, and the logistical costs of radio-tracking) or genetic approaches (see also caveats of both methods in Introduction). However, it requires from a deep knowledge on the study system for maximizing fruit removal and recapture rates; specifically, on the sites where target disperser species deposit their faeces. Finally, it is important to point out that our sampling approach could not succeed in the same way in systems without an extensive path network, such as the one studied here.
Features and Functional Heterogeneity of Carnivore-mediated Seed Dispersal Kernels
The studied carnivore guild dispersed seeds up to nearly 3 km (2846 m) from seed sources, corroborating the role of this guild as long-distance seed dispersers (Jordano et al. 2007). Remarkably, 40% of carnivore-dispersed seeds were found farther than 1 km (LDD events) and nearly 20% farther than 2 km (Fig. 1a). Such distances and frequencies are well over the range of those generated by most passerine birds which, together with carnivores, serve as seed dispersers to many temperate fleshy-fruited plant species (see Herrera 1989; Jordano et al. 2007; Herrera, Morales & García 2011; López-Bao & González-Varo 2011). Our results thus confirm and expand the functional differences reported by Jordano et al. (2007) between carnivores and birds in terms of seed dispersal distances and highlight the key role of carnivores for gene flow among plant populations as well as for colonization and establishment of new populations.
However, considering our target carnivore guild, most of the above-mentioned features can be attributable to the red fox given that it accounted for nearly 70% of carnivore-dispersed seeds. More than a half of fox-dispersed seeds (52%) were found farther than 1 km (LDD events) and more than a quarter (26%) farther than 2 km. By contrast, only 11% of marten-dispersed seeds were found farther than 1 km and no seed was found at distances above 2 km. As predicted, maximum seed dispersal distances generated by each frugivore species (i.e. the spatial scale of its seed dispersal service) depended on its home range size. Home ranges of red foxes typically range between 200 and 600 ha (Cavallini 1996; Dekker, Stein & Heitkönig 2001). If we (conservatively) assume circular-shaped home ranges, such areas would imply maximum seed dispersal distances (i.e. their diameters) of c. 1600 and 2800 m. Exceptionally, home ranges of c. 1000 ha have been reported as well (Cavallini 1996), which would imply maximum dispersal distances of c. 3500 m. Regarding pine martens, Zalewski, Jedrzejewski & Jedrzejewska (2004) reported mean and maximum home ranges of nearly 50 and 150 ha, respectively, which would imply maximum dispersal distances of c. 800 and 1400 m, respectively. Surprisingly, such humble estimates of potential maximum dispersal distances fit well to our empirical data. Indeed, the maximum dispersal distances mediated by each carnivore species were nearly identical in two different years (Fig. 3). Our results thus strongly support the idea that the home range size of frugivore species – generally positively correlated with their body size (Makarieva, Gorshkov & Li 2005) – will shape the range of seed dispersal distances they generate (Spiegel & Nathan 2007).
The distance distribution of fox-dispersed seeds fitted to a tetramodal kernel in which the probability of seed deposition does not decay gradually with distance from maternal trees, but that reaches high values at short-, intermediate- and even at long-distance classes (see Fig. 1b). By contrast, the distance distribution of marten-dispersed seeds fitted to a unimodal kernel in which the probability of seed deposition decays sharply with distance from source trees (see Fig. 1c). Reasonably, such differences should depend on particular behaviours of each disperser species (e.g. Russo, Portnoy & Augspurger 2006), including spatial patterns of foraging activity and scent territorial marking in relation to the local landscape structure. For example, the multimodality found in the kernel mediated by foxes is congruent with their movements, which are characterized by short and rapid trips between foci of interest within a defined home range and also by the use of routine travel routes (Adkins & Stott 1998). Another factor likely influencing the observed distances would be the distribution of carnivore's home ranges in relation to the location of feeding stations, which may explain the differences between 2009 and 2010 in distance distributions of marten-dispersed seeds (Table 1).
Taken together, both kernels highlight that the role for gene flow among plant populations and for colonization of distant vacant areas is not evenly distributed among different carnivore species. In our study case, these roles are clearly carried out by the red fox but not by the pine marten.
Clues about the Spatial Variation from the Temporal One
There were strong differences between 2009 and 2010 in the relative contribution of foxes and martens to the total number of carnivore-dispersed seed mimics recovered. We found 376 fox-dispersed seed mimics in 2009 (84%) and only 66 in 2010 (40%), and we found 71 marten-dispersed seed mimics in 2009 (16%) and 97 in 2010 (59%). Thus, the shift in 2010 seems to reflect lower rates of fig consumption by foxes, which in turn might be explained by multiple non-mutually exclusive causes. A possible explanation would be changes in the number and/or identity of individuals sampled each year due to home range shifts or to mortality events (Meek & Saunders 2000; Dekker, Stein & Heitkönig 2001). Alternatively, and even sampling the same individuals that used the same home ranges, between-year differences in fig consumption could have mirrored temporal variation in local food availability.
More importantly, such strong interannual variation served as a natural and spontaneous experiment that allowed us to test how different disperser assemblages (here represented by different relative contributions of each carnivore species) generate different seed dispersal kernels. Clearly, the kernel generated by the carnivore guild in 2009 (fox-dominated year) resembled a fox-type kernel while in 2010 (marten-dominated year) it resembled a marten-type kernel (Fig. 2). Our results illustrate how a huge proportion of LDD events will be lost not only in landscapes without foxes (scenario of Fig. 1c), but also in those landscapes where they are very scant (scenario of Fig. 2b). As far as we know, this is the first experimental demonstration that seed dispersal kernels strongly depend on the relative contribution of different disperser species. Thus, seed dispersal kernels will not only vary in space and time mirroring spatiotemporal variation in disperser assemblages, but will also vary between fleshy-fruited plant species due to differences in the strength of their interactions with each particular disperser (see Bascompte, Jordano & Olesen 2006).
This study provides the first empirical evidence of functional heterogeneity among seed dispersal kernels generated by carnivorous mammals. In addition, interannual variation in relative fruit removal by different species gave us the opportunity of empirically demonstrating that seed dispersal kernels strongly depend on the relative contributions of each disperser species, thus on the composition of local disperser assemblages. Our results highlight that the key role of long-distance seed dispersal mediated by carnivores for gene flow among distant plant populations and for the establishment of new populations is not evenly distributed among carnivore species. Our findings provide starting points for modelling plant population processes that include animal-mediated seed dispersal, such as connectivity and colonization. Whenever carnivores were involved as seed dispersers, these findings should also contribute to a better understanding of the susceptibility/resilience of natural ecosystems to threats such as climate change, disperser hunting, species invasions and habitat fragmentation.
We thank Alicia Criado, Ana Lombardero, Tania Veiga, Lourdes Sotelo, Ignacio Munilla, Luna Puentes, Alejandro Rodríguez, Martiño Cabana and Anxos Romeo for essential help at different parts of this study. We especially thank Juan M. Morales for help and advice on kernel fitting. Alfredo Valido, Juan M. Morales, Pedro Jordano and two anonymous referees provided us with insightful comments that helped us to improve the final version of this paper. This study was funded by the Galician Regional Government (project PGIDIT 05RFO 20001 PR).