Spatial patterns of seed dispersal and seedling recruitment in Corema album (Empetraceae): the importance of unspecialized dispersers for regeneration

Authors

  • María Calviño-Cancela

    Corresponding author
    1. Departamento de Ecoloxía e Bioloxía Animal, Universidade de Vigo, E.U.E.T. Forestal, Campus Universitario, 36005 Pontevedra, Spain
    • María Calviño-Cancela, Departamento de Ecoloxía e Bioloxía Animal, Universidade de Vigo, E.U.E.T. Forestal, Campus Universitario, 36005 Pontevedra, Spain (e-mail maria@uvigo.es).

    Search for more papers by this author

Summary

  • 1Spatial patterns of seed dispersal and seedling recruitment of Corema album were examined among and within habitats to determine the quantitative importance of different dispersers in each type of habitat, and their effectiveness in carrying seeds to suitable habitats for seedling recruitment.
  • 2Gulls, blackbirds and rabbits were, respectively, the main dispersers (45%, 40% and 15% of Corema album seeds). Within habitats, blackbirds disperse seeds mainly to female Corema album shrubs, while gulls and rabbits disperse seeds mainly to open ground.
  • 3The quantitative role of dispersers varies among habitats because of their habitat preferences, causing the spatial pattern of seed rain to differ.
  • 4Open ground has the highest density of seedlings and the highest seedling-to-seed ratios. Regeneration is more active in the pioneer scrub than in the mature scrub and the herbaceous vegetation.
  • 5Gulls, rather than specialist frugivores, are the most effective dispersers in carrying seeds to suitable sites for recruitment.

Introduction

Seed dispersal is a keystone demographic bridge between adult and seedling stages (Harper 1977) and the natural regeneration of plant populations depends to a large extent on seed dispersal (Jordano 1993; Gavin & Peart 1997; Wunderle 1997; Ganzhorn et al. 1999). The seeds of fleshy fruits are dispersed via defecation or regurgitation and arrive at sites that may differ in their suitability. Environmental conditions are more critical in the early stages of the life cycle, when mortality is high (Harper 1977), and success in the steps involved in recruitment is therefore a good indicator of the suitability of a particular site. The spatial pattern of seedling recruitment integrates the critical post-dispersal processes of seed survival, germination and early seedling survival. An animal will be a good disperser for a species if its dispersal pattern matches that of seedling recruitment, i.e. it carries seeds to sites where the probability of seedling recruitment is high.

Seed dispersal patterns have been analysed both indirectly, by observing the movements of fruit-eating animals (Herrera & Jordano 1981; Izhaki et al. 1991), and directly, by collecting either seeds visible on the ground (Salomonson 1978; Herrera 1984; Santos et al. 1999) or from traps (Debussche & Isenmann 1994; Herrera et al. 1994; Kollmann & Goetze 1998). Seed traps may not record seed rain accurately in open areas (Kollmann 1995) as, when visible, they may attrack dispersers (e.g. if birds use them as perches, Kollmann & Goetze 1998) or be avoided. Alternatively faeces can be collected directly from the ground, particularly in snow-covered (Herrera 1984) or sandy areas (this study) where they are easily visible (as well as traps being hard to conceal).

I investigated the natural dispersal of seeds of Corema album (L.) D. Don (Empetraceae), an endangered shrub endemic to the west coast of the Iberian Peninsula. I concentrate on: (i) the patterns of seed deposition at the among- and within-habitat levels; (ii) the quantitative role of individual disperser species and their spatial pattern of seed deposition; (iii) the patterns of seedling recruitment at the among- and within-habitat levels; and (iv) the overlap between the patterns of the species-specific seed deposition and seedling recruitment. At the study site, fruits of Corema album are consumed by gulls (Larus cachinnans Pallas) and blackbirds (Turdus merula L.), as well as some small passerines, and rabbits (Calviño-Cancela 2000). Gulls have not previously been considered as seed dispersers of a fleshy-fruited plant and the role of rabbits has been little studied (but see Muñoz Reinoso 1993; Nogales et al. 1995). Blackbirds, however, are typically frugivorous and known to be involved in dispersal of many fleshy-fruited plants (e.g. Herrera 1984; Jordano 1987; Théry 1989). This unusual assemblage of frugivores allows comparison of the spatial patterns of seed dispersal generated by very different species, and assessment of whether specialist frugivores (e.g. T. merula) are necessarily the best dispersers.

Materials and methods

STUDY PLANT SPECIES

Corema album (L.) D. Don (Empetraceae) is found in the Iberian Peninsula (ssp. album) and the Azores Islands (ssp. azoricum Pinto da Silva). It grows in coastal habitats, mainly on sand dunes but also at rocky sites, and in the Azores Islands, on volcanic lava and ash fields, as a low, trailing shrub that rarely exceeds 1 m in height. Corema album is wind-pollinated and dioecious (see Guitián et al. 1997 for reproductive biology), producing quasi-spherical drupes, 9.0 ± 0.1 mm (n = 30) in diameter, which are white when ripe and have a sugary and water-rich pulp (Herrera 1987). Each fruit has three (exceptionally four or two) elongated seeds (0.48 ± 0.01 mm in length, n= 30) with a thick woody endocarp. In the study site, fruiting occurs in summer and early fall and peaks in August and early September at 595 ± 26 ripe fruits m−2 (Calviño-Cancela 2000). Seeds are dispersed mainly in summer but lie dormant for at least 1 or 2 years (personal observation). Seeds germinating and seedlings emerging during a particular winter will reflect production 2 or 3 years (or even more) previously.

STUDY SITE

The study was carried out during 1999 and 2000 in the Figueiras and Muxieiro sand dunes in the Cíes Islands Natural Park (42°15′ N, 8°53′ W, NW Spain). The climate is Mediterranean (Guitián & Guitián 1990). The annual average temperature for 1945–74 was 13.8 °C (January, 8.5 °C; July, 19.8 °C; 877 mm annual precipitation), with a notable drought period in summer, mainly in July and August

An important population of Corema album, with c. 1500 adult individuals, is found on partially fixed dunes in both pioneer (PS) and mature (MS) stages of the scrub succession (Guitián & Guitián 1990). In PS, vegetation cover (Table 1) is low (28.1%), with C. album forming 82.6% of the total vegetation cover. In MS, vegetation cover is higher (64.6%), of which a significant amount is herbaceous, and Cistus salvifolius L. is now the dominant shrub (C. album represents only 13.6% of the vegetation cover, Table 1). Although other fleshy-fruited plants (Daphne gnidium L., Solanum nigrum L. and Osyris alba L.) are present, their contribution to fruit availability is virtually insignificant compared with C. album. A third habitat type was also studied; C. album is absent from areas of herbaceous vegetation (HV), where the low vegetation cover (40%) consists of species such as Artemisia crithmifolia L., Helichrysum picardii Boiss & Reuter, Iberis procumbens Lange and Pancratium maritimum L.

Table 1. Number of pellets of Larus cachinnans, Turdus merula and Oryctolagus cuniculus sampled and percentage cover of the five microhabitat categories in each of three types of habitat, using pooled data for 1999 and 2000
Microhabitat categoryMicrohabitatsHerbaceous vegetationPioneer scrubMature scrub
Cover (%) Larus cachinnans Turdus merula Oryctolagus cuniculusCover (%)Larus cachinnans Turdus merula Oryctolagus cuniculus Cover (%) Larus cachinnans Turdus merula Oryctolagus cuniculus
FemaleFemale            
Corema album Corema album  000  014.0  089  0 4.5 0247  0
Other shrubsMale            
  Corema album  000  0 9.2  2 8  6 4.3 0 95  0
  Daphne gnidium  000  0 0.1  0 0  0 3.6 0 37  0
  Ulex europaeus  000  0 0  0 0  0 6.1 0  0  4
  Cistus salvifolius  000  0 0  0 0  014.7 0  2 10
  Scrophularia frutescens  000  0 0  0 0  0 3.6 0 11 10
 Other shrubs 000  0 0  0 0  0 0.5 0  0  2
Herbs Helichrysum picardii  6.300 16 0.6  0 0  0 4.1 0  0 10
  Artemisia crithmifolia 15.700 23 0.9  0 0  3 4.6 0  4 13
 Other herbs15.000 39 3.3  0 0 24 6.1 2  8 12
Open groundOpen ground60.02420064.71404148732.216 31215
OthersMosses 3.000 16 0  0 0  012.5 0 28455
 Pine stumps 000  0 0  0 0  0 0.1 0  4  0
 Rocks 000  0 7.2  732  1 0.5 0  6  2
 Others 000  0 0  0 0  0 2.6 0  0 90

Microhabitats within each vegetation type (habitat) were categorized as follows: (a) female Corema album; (b) other shrubs (i.e. male Corema album, and all other species); (c) herbs, where herbaceous species covered more than 10% of the surface; (d) open ground, i.e. bare sand or sand with herbaceous plant cover lower than 10%; and (e) others (see Table 1).

Nomenclature of plant species follows Castroviejo et al. 1986–97). See Guitián & Guitián (1990) for more information about climate and vegetation in the area.

POST-DISPERSAL SEED DISTRIBUTION

Post-dispersal seed distribution was studied during the fruiting seasons of 1999 and 2000 along linear transects arranged in May 1999, according to a stratified design. The numbers of faeces and regurgitations within 50 cm on one side of the transect were counted. For each of the three habitat types (MS, PS and HV) two parallel series of transects were arranged in two areas c. 150 m apart, with a total of 230 m in 16 transects (10 or 15 m long) per habitat. In 1999, weekly counts were carried out from 18 July to 29 August, with a seventh visit being made 16 days later on 8 September. In 2000, weekly recordings were carried out from 24 July to 26 August, with a sixth visit on 11 September. The microhabitat of each excrement or regurgitation of Larus cachinnans was noted, as well as for faeces of Turdus merula, Oryctolagus cuniculus and small passerines. The spatial pattern of small passerines’ faeces was not analysed due to their low abundance. Faeces of Oryctolagus cuniculus and Turdus merula and faeces and regurgitations of Larus cachinnans can be distinguished by their shape and size.

I collected samples from Turdus merula (n = 150 faeces), Larus cachinnans (n = 50 faeces and 30 regurgitations) and Oryctolagus cuniculus (n = 1470 faeces) during the fruiting season and estimated the number of dispersed seeds per species as the product of the number of recorded faeces and the average seed contents per faecal item.

SEED PREDATION

Once faeces are deposited on the ground, seed predators can modify the original seed distribution patterns until the next sampling date. Counting faeces directly therefore includes the effects of predation as well as post-dispersal seed distribution. Seed predation was surveyed during the period of fruit availability in 1999 and 2000, when post-dispersal seed distribution was studied using groups of 10 seeds contained in trays. The trays (8 cm × 8 cm × 3 cm in height) were made out of 1 mm mesh white plastic trellis to provide a good drainage and nailed to the ground. Trays (16 per treatment) were placed in PS and MS: (a) underneath female C. album; (b) c. 50 cm from a female C. album; (c) underneath male C. album; (d) underneath Daphne gnidium (Ms only due to the low cover in PS); and (e) in open ground, and in HV in open ground (40 trays). In both years, the trays were placed 15 days before the first sampling of post-dispersal seed distribution and checked for seed predation after 3 days, after 1 week and then every 15 days.

SEED VIABILITY

Seeds were collected from pellets and from branches of 30 plants (10 fruits per plant), and tested with 2,3,5 triphenyl tetrazolium (TTZ) (ISTA 1985). For each treatment (seeds regurgitated by gulls, defecated by gulls, by blackbirds and by rabbits, and control seeds), 20 replicates of 10 seeds were tested.

SEEDLING DISTRIBUTION

The spatial patterns of the newly germinated seedlings (1-year-old seedlings) were surveyed in May 1999 and May 2000, just before the high mortality that usually occurs in summer. All the seedlings within 50 cm on one side of the same transects used to estimate post-dispersal seed distribution were recorded and marked, and those 1999 seedlings that survived till 2000 (2-year-old seedlings) were noted, together with the microhabitat where seedlings appeared.

The seedling-to-seed ratio was computed as the number of 1-year-old seedlings recorded in a given year divided by the average number of viable seeds deposited in that habitat or microhabitat in the 2 years of study (see Debussche & Isenmann 1994; Houle 1998). Because of the variable dormancy period of the seeds, it is not possible to know which seedlings correspond to dispersal in a given year. Nevertheless, I assume that the variation between years in the patterns of seed distribution in habitats and microhabitats is low (see Results), and therefore that the seedlings arose from seed patterns similar to those observed in this study. In addition, I assume that the variation of seed distribution between years is significantly lower than that due to differences in habitat and microhabitat suitability for seedling recruitment (see also Debussche & Isenmann 1994; Houle 1998), so that the seedling-to-seed ratio could be used to compare their relative suitabilities (Nakashizuka et al. 1995). A seedling survival index was also computed as the proportion of the seedlings recorded in 1999 that survived till 2000.

STATISTICAL ANALYSIS

Because of their non-normal distributions, the density of seeds and seedlings was analysed using Kruskal–Wallis tests, with multiple comparisons of pairs by the nonparametric Nemenyi test (equal replicates per group) and the Dunn test (unequal replicates per group), with critical values corrected for multiple comparisons (Zar 1996). Seed viability was analysed using one-way anova with multiple comparisons of pairs by the Bonferroni test. Levels of significance throughout this study are as follows: *P < 0.05; **P < 0.01; ***P < 0.001.

Results

SEED PREDATION

No sign of either predation or removal was detected for seeds in trays or in the faeces of seed dispersers. Thus, I assume that the distribution of faeces reflects the true seed deposition pattern.

POST-DISPERSAL SEED DISTRIBUTION

Quantitative importance of seed dispersers

Larus cachinnans was quantitatively the most important seed disperser in both years, dispersing c. 45% of seeds (n = 7457 in 1999, and n= 8297 in 2000) compared with c. 40% for Turdus merula. These two species together therefore dispersed the great majority of seeds, with Oryctolagus cuniculus and small passerines contributing only 15% and less than 0.3%, respectively. The low abundance of small passerine faeces made further analyses of their spatial pattern impossible.

Among-habitat patterns

The density of seeds dispersed by each species differed among habitats (Table 2). For Larus cachinnans it was significantly higher in PS than in the other two habitats, with HV receiving fewer seeds. For Turdus merula HV was again lower, but MS received the highest density. The pattern by Oryctolagus cuniculus was less pronounced; although HV received the highest density of seeds and PS the lowest, the differences were only significant in 1999.

Table 2. Density of Corema album seeds (seeds m−2) dispersed by each of three species and total in the three types of habitat in the 1999 and 2000 study seasons (mean ± 1 SE). Values of Kruskal–Wallis test (H) are indicated. Superscript letters indicate the results of multiple comparisons of pairs with the Nemenyi test; means with different letters in each row are significantly different (P < 0.05)
Habitat Larus cachinnans Turdus merula Oryctolagus cuniculus Total
19992000199920001999200019992000
HV 0.1a ± 0.1 0.1a ± 0.10.1b ± 0.10.05b ± 0.05  2.2b ± 0.42.0 ± 0.4   2.5b ± 0.4   2.1b ± 0.4
PS11.3b ± 2.313.1b ± 2.1 3.5a ± 1.1 3.5a ± 0.8  1.0a ± 0.41.2 ± 0.215.9a ± 3.117.7a ± 2.3
MS 3.2a ± 1.5 2.7a ± 1.4  9.0a ± 4.6 9.9a ± 4.71.9a,b± 0.82.1 ± 0.313.8a ± 4.914.7a ± 4.6
H30.5***32.9***19***23.2***6.7*5.5ns21.7***30.9***

Within-habitat patterns

Within-habitat patterns of seed dispersal varied widely between seed dispersers. Larus cachinnans showed a clear preference for open ground (Table 3), which received 94% of gull faeces and regurgitations (Fig. 1 and Table 1).

Table 3. Density of Corema album seeds (seeds m−2) dispersed by Larus cachinnans, Turdus merula, Oryctolagus cuniculus and total to the microhabitats of the three types of habitats in the 1999 and 2000 study seasons (mean ± 1 SE). Values of Kruskal–Wallis test are indicated (H) and superscript letters indicate the results of multiple comparisons of pairs with the Dunn test. Means with different letters in each row for each disperser are significantly different (P < 0.05)
DispersersMicrohabitatsHerbaceous vegetationPioneer scrubMature scrub
199920001999200019992000
Larus Female      
cachinnans Corema album 0a0a0a0a
 Other shrubs  1.4a ± 1.4   1.4a ± 1.30a0a
 Herbs0000a     2.5a ± 2.5  0.9a ± 0.9
 Open ground  0.3 ± 0.3 0.6 ± 0.6 48.6b ± 18.0  52.9b ± 10.7  26.9b ± 13.9 34.9b ± 20.5
 Others00 30.1a ± 23.1  10.3a ± 5.00a 36.9b ± 10.7
 H1.438ns1.438ns42.532***43.943***17.741**18.529***
Turdus Female      
merula Corema album  36.7a ± 11.2  41.6a ± 8.9   348.2a ± 243.9450.3a ± 263.8
 Other shrubs  3.1b ± 1.7  11.3b ± 9.9  14.8b ± 10.4   12.2b ± 9.0
 Herbs000b0b0b  5.8b ± 2.3
 Open ground  0.2 ± 0.1 0.1 ± 0.1  2.1b ± 1.0   2.5b ± 1.4    6.5b ± 1.5  4.3b ± 1.3
 Others0010.8a,b ± 5.8   6.8b ± 1.0    6.8b ± 2.4 20.0b ± 8.0
 H4.543ns1.438ns25.25***26.789***33.133**26.069***
Oryctolagus Female      
cuniculus Corema album 0a0a0a0a
 Other shrubs  0.2a ± 0.2   1.4a ± 1.2  0.3a,b ± 0.1  0.4a ± 0.2
 Herbs 6.2a ± 1.03.95 ± 1.0 6.4a,b ± 2.6  7.4a,b ± 2.7  1.0a,b ± 0.4  1.5a ± 0.6
 Open ground 3.6b ± 0.8 3.8 ± 0.9  3.2b ± 1.2   4.0b ± 0.8   3.8b ± 1.3  3.7b ± 0.9
 Others 2.3b ± 1.3 3.4 ± 1.30a     3a ± 5.6  23.8b ± 15.5 36.9b ± 10.7
 H21.425***0.467ns20.959***21.982***15.754**50.338***
TotalFemale      
  Corema album 36.7a,b ± 11.2 41.6a,c ± 8.9  613a,c ± 317  450a ± 263
 Other shrubs  5.9a ± 2.0  17.1a ± 10.8  15.8a ± 3.9 12.6b ± 9.0
 Herbs 6.2a ± 1. 0 4.0 ± 1.0ns  6.4a ± 2.6  7. 4a,b ± 2.7  3.4a,b ± 2.6  8.2b ± 2.8
 Open ground4.0a,b ± 0.8 4.5 ± 1.0ns 53.9b ± 18.7  59.4c ± 11.3  46.1c ± 16.142.9a,b ± 20.9
 Others 2.3b ± 1.3 3.4 ± 1.3ns40.9a,b ± 24.920.1a,b,c± 7.229.5a,b,c ± 14.7 56.9a ± 15.5
 H6.0*1.1ns24.0***25.2***31.0***37.1***
Figure 1.

Microhabitat selection by dispersers in the three types of habitats. Bars represent the percentage of the surface covered by each microhabitat category and percentage of excrements of each disperser recorded. If percentage of excrements of a species is higher than percentage cover in a given microhabitat, positive selection (attraction) for that microhabitat by that species is indicated and vice versa.

Turdus merula dispersed the seeds mainly to female Corema album (Fig. 1), both in MS and PS, and was responsible for the arrival of most of the seeds reaching this microhabitat (99.7% in MS and 99.6% in PS, Fig. 2). T. merula faeces, however, were found in all microhabitat categories (Fig. 1, Table 3), with 12% found in open ground (55% within 50 cm from female Corema album).

Figure 2.

Number of seeds in each microhabitat due to each disperser in the three types of habitat.

Oryctolagus cuniculus showed a more variable pattern, with the highest density of seeds in herbs in PS and HV, but in others followed by open ground and herbs in MS (Table 3). Open ground received most of the pellets (55% vs. 9% for herbs, see Fig. 1 and Table 1), but since it covered a much higher proportion of the surface than herbs, seed density was lower. Few rabbit pellets were found under shrubs.

Contributions to habitats and microhabitats

Overall, the density of seeds dispersed in HV was significantly lower than in the other two habitats (Table 2). The density in PS was higher than in MS in both years but differences were not significant.

In PS, Larus cachinnans was the main disperser in both years (Fig. 2), responsible for 69% of seeds vs. 25% and 7% for T. merula and O. cuniculus (n = 8091). This species therefore, determined seed arrival among microhabitats, with most seeds falling on open ground (73% in 1999 and 78% in 2000, 84% of which were dispersed by L. cachinnans).

In MS, Turdus merula was the main disperser (Fig. 2), dispersing 64% of seeds in 1999 (n = 3184) and 67% in 2000 (n = 3454), compared with 25% and 15% for L. cachinnans and O. cuniculus. Female Corema album received the highest proportion of seeds (41% in 1999 and 44% in 2000, of which 99.7% were dispersed by T. merula, with open ground and Others receiving c. 30% and 15%).

In HV, Oryctolagus cuniculus was the main disperser (Fig. 2, 91% vs. c. 5% for both L. cachinnans and T. merula). The densities of seeds were quite low in all microhabitats (Table 3), but herbs, with the highest density, were significantly different from others, with the lowest. However, most seeds fell in open ground (65%), followed by herbs (32%), while others received only 3%.

Number of seeds per faecal item

The number of seeds per faecal item differed widely among dispersers: from c. one seed in O. cuniculus pellets (0.97 ± 0.04, n= 1470) to 11 ± 0.5 for T. merula (n = 150) and 27 ± 2.6 for Larus cachinnans (n = 30). Gulls also dispersed seeds in their regurgitations, which each contained 252 ± 36.5 seeds (n = 50).

Variability in seed density

Patchiness was higher for seeds dispersed by O. cuniculus than those dispersed by L. cachinnans and T. merula (CVs of 54%, 17% and 16%, respectively). Among habitats, it was higher in HV than in PS and MS (CVs 108% vs. 63% vs. 21%). CV in the microhabitats varied between MS and PS. Patchiness in female Corema album and other shrubs was higher in PS than in MS (CV 122% vs. 55% and 92% vs. 35%). However, patchiness in open ground and others was higher in MS than in PS (CV 115% vs. 89% and 75% vs. 54%).

SEED VIABILITY

Nearly 80% of the seeds were viable but there were differences among sources (Table 4), such that seeds recovered from faeces were 10% less viable than those from regurgitations or collected directly from the plants. Seed processing in the gut appears to decrease viability similarly for all three dispersers (Table 4).

Table 4. Viability (%) of Corema album seeds (mean ± 1 SE) sampled from the plants (Control), and dispersed by Larus cachinnans (faeces and regurgitations), Turdus merula and Oryctolagus cuniculus. The F-test and multiple comparisons of pairs with the Bonferroni test are included: means with different superscript letters are significantly different
GroupsSeed viability (%)
Control86.5 ± 2.2a
Larus cachinnans regurgitations87.0 ± 1.5a
Larus cachinnans faeces77.0 ± 2.7b
Turdus merula faeces76.0 ± 2.8b
Oryctolagus cuniculus faeces77.0 ± 2.2b
F4, 95 d.f.5.721***

PATTERNS OF SEEDLING RECRUITMENT

Densities of seedlings were very different among habitat types (Table 5), with most of the seedlings being found in PS (95%, n= 390). Seedling density in MS was quite low (lower than 0.1 in both years) and even lower in HV (lower than 0.05). Within habitats, 98.9% of seedlings were recorded in open ground, where density was therefore higher (Table 6). Seedling-to-seed ratios were very low (Table 6), the highest values being in open ground in PS.

Table 5. Seedling recruitment of Corema album in the three habitats (mean ± 1 SE). The Kruskal–Wallis test and multiple comparisons of pairs with the Nemenyi test (P < 0.005) are included: means with different letters in each row are significantly different
Habitat typeSeedlings m−2 (1999)Seedlings m−2 (2000)
Herbaceous vegetation0a0.05 ± 0.02a
Pioneer scrub0.95 ± 0.10b2.24 ± 3.87b
Mature scrub0.02 ± 0.02a 0.1 ± 0.06a
H39.2***22.3***
Table 6. Density of seedlings and seedling-to-seed ratio in the microhabitats of the three habitats in 1999 and 2000 (mean ± 1 SE). Microhabitats where no seedlings were recorded are not included
HabitatMicrohabitatSeedlings m−2Seedling : seed ratio
1999200019992000
HVOpen ground00.098 ± 0.0450 0.014 ± 0.008
PSFemale Corema album0.096 ± 0.0930.121 ± 0.1180.0012 ± 0.00110.0058 ± 0.0056
 Open ground1.927 ± 0.4908.830 ± 1.753 0.064 ± 0.013 0.136 ± 0.023
MSOpen ground0.028 ± 0.0270.397 ± 0.203 0.011 ± 0.011 0.075 ± 0.044

The spatial pattern of seedlings did not resemble that of seeds at either among-habitat (chi-square = 327, P < 0.001, d.f. = 2) or within-habitat (chi-square = 101, P < 0.001, d.f. = 4, in PS) levels. In PS, the only habitat with sufficient seedlings, seedling distribution was not related to percentage cover of microhabitats (chi-square = 193, P < 0.001, d.f. = 4).

Only six seedlings recorded in 1999 survived till 2000, all in open ground in PS, with a density of 0.13 ± 0.05 seedlings m−2. Survival ratio of seedlings in the first year was only 7%.

Discussion

THE ROLE OF UNSPECIALIZED DISPERSERS

One of the more remarkable findings of this study is the role of Larus cachinnans and Oryctolagus cuniculus, with L. cachinnans being quantitatively the most important disperser of Corema album. Fruits are not a common component in gull diets, although they can be abundant in certain sites and seasons (see Cramp & Simmons 1998). Gulls are unlikely to be significant dispersers for most plant communities, but Corema album is a coastal plant and Larus cachinnans is very abundant in this area (22 000 pairs in the island, Munilla 1997). All the gulls observed feeding on fruits were juveniles that were just leaving the breeding colony. In spite of their low energy value (compared with fish or crustaceans), fruits are very easy to obtain and this can increase their profitability for young individuals with less-developed feeding skills, according to the optimal-foraging theory (Krebs & McCleery 1984). There are several studies on the role of O. cuniculus as seed disperser, although mainly in herbaceous communities (Welch 1985; Malo & Suárez 1995), rather than for fleshy-fruited plants. Its quantitative role here is low, probably because of its low population density, despite the frequent fruit consumption evidenced by seeds in the faeces.

SPATIAL PATTERNS OF SEED DISPERSAL

The patterns of seed dispersal reflect the habitat and microhabitat preferences of seed dispersers (Blake & Hoppes 1986; Levey 1988; Kollmann & Pirl 1995).

The preferences of Larus cachinnans for open ground and for PS may be explained by considering morphological constraints on its movements, and by fruit availability. Gulls, unlike passerines, are not adapted to perching on shrubs and prefer open ground where it is easier for them to move. In PS, female individuals of Corema album are isolated and their fruits are therefore more accessible to gulls than in the denser vegetation (particularly the higher cover of shrubs) in MS. Although gulls are not specialized in fruit consumption, they tended to avoid areas like HV, where Corema album is not present.

Turdus merula prefers scrub (MS and PS) to herbaceous habitats (HV), as do most frugivorous passerines, which usually generate a scrub-centred pattern of seed dispersal (Gorchov et al. 1993; Debussche & Isenmann 1994; Kollmann & Pirl 1995). Denser and higher perching sites in the scrub may provide more fruits as well as protection against predators (Howe 1979; Snow & Snow 1988). In both MS and PS T. merula disperses the seeds mainly to female Corema album, rather than under other shrubs, including male C. album. Frugivorous passerines usually disperse seeds underneath fruit-bearing plants, where they spend a long time foraging and perching (Herrera 1984; Izhaki et al. 1991; Debussche & Isenmann 1994; Herrera et al. 1994). Rocks, mosses and open ground receive an appreciable number of faeces of T. merula, which uses rocks as perching sites (personal observation, García et al. 1996 for T. torquatus) and may search for arthropods in the mosses. Passerines do not usually carry seeds to clearings (Izhaki et al. 1991; Debussche & Isenmann 1994; Herrera et al. 1994); Turdus merula, however, often forages on the ground, from where it can reach C. album fruits: 55% of faeces found in open ground were within 50 cm of female C. album, and a significant number were close to rocks (16%) or pine stumps (8%), which are used as perches from which faeces may be secondarily dispersed.

Oryctolagus cuniculus seems to avoid shrubs but its preferences between other microhabitats are not clear.

DISPERSERSCONTRIBUTION TO SEED ARRIVAL AT HABITATS AND MICROHABITATS

The quantitative importance of dispersers combines with their preferences among habitats and microhabitats to determine the patterns of seed deposition. Distribution among microhabitats will depend on the preferences of the most prevalent dispersers in a habitat. Thus, most seeds in PS appeared in open ground (due to L. cachinnans) but underneath female Corema album in MS (due to T. merula), whereas in HV O. cuniculus does not show a clear preference among non-shrub microhabitats.

The difficulty in exporting seeds to HV is noteworthy (see also Gorchov et al. 1993; Debussche & Isenmann 1994; Kollmann & Pirl 1995) and seems to be due to most dispersers preferring habitats where fruits are abundant (Herrera 1985; Kollmann & Pirl 1995). Seed dispersal to open habitats depends on unspecialized feeders, whose movements are not related to fruit availability (see also Santos et al. 1999), as here where rabbits may have an important role in the colonization of new areas.

SEEDLING RECRUITMENT

One-year-old seedlings are scarce and are concentrated in open ground in PS. Seedling emergence seems to be a true bottleneck in the recruitment of Corema album, imposing an important limitation on both quantity and spatial distribution. Regeneration in PS is more active than in MS (see also Debussche & Isenmann 1994; Kollmann & Pirl 1995), and open ground seems to be the most suitable microhabitat. Seed dispersal patterns were very consistent in the 2 years of study, and coincide also with the patterns observed in a preliminary study in 1998 (Calviño-Cancela 2000), so that the calculated seedling-to-seed ratios should reflect the suitability of sites for establishment. Seedling recruitment appears to be strongly dependent on the place where seed arrives: environmental conditions only seem to be favourable in open ground.

SEED DISPERSAL EFFECTIVENESS: SPECIALIZED VS. UNSPECIALIZED SEED DISPERSERS

Spatial patterns of seedling recruitment do not resemble those of seed dispersal. The efficiency of seed dispersal will therefore depend to a large extent on the dispersers’ ability to carry seeds to open ground. Frugivorous passerines are bad dispersers for C. album because, as a consequence of their specialization in fruit consumption, they tend to disperse the seeds underneath the shrubs and fail to carry them to clearings (Izhaki et al. 1991; Debussche & Isenmann 1994; Herrera et al. 1994). This can be an advantage for those species whose germination and seedling survival are favoured by the environmental conditions underneath shrubs (Debussche & Isenmann 1994; Verdú & García-Fayos 1996) but not for species that need to reach clearings for recruitment (e.g. Corema album). Although frugivorous passerines are the main dispersers of fleshy-fruited plants in the temperate areas (Snow & Snow 1988), other dispersers, such as gulls or rabbits (this study), or carnivorous mammals (Bustamante et al. 1992), may play an important role in regeneration even when their quantitative importance is low. Therefore, although T. merula disperses many more seeds than O. cuniculus (40.1 vs. 14.8%), their contribution to open ground in PS, where recruitment is concentrated, is almost the same (7.5% vs. 8.0%). Thus, the role of non-passerine dispersers should receive more attention for plants that need seed arrival at clearings for regeneration. Therefore, a higher degree of specialization in fruit consumption does not necessarily imply a higher effectiveness in seed dispersal, and can even be detrimental in certain circumstances.

Acknowledgements

I would like to thank the staff of the Natural Park of Cíes Islands for their logistical support, as well as the Servicio de Medio Ambiente Natural (Pontevedra) for permitting my field work in the natural park. I also thank José Guitián for stimulating my work on this topic, Carlos Herrera and Pedro Jordano for their advice during the initial planning of this study, Adolfo Cordero for his comments on earlier drafts of the manuscript and his continuous support, Johannes Kollmann for constructive comments on the manuscript, and Jacquie Deverell for checking the English. This study was supported by fellowships from Secretaría Xeral de I + D (Xunta de Galicia) and a project from the Consellería de Medio Ambiente (Xunta de Galicia).

Ancillary