Recruitment dynamics of a fleshy-fruited plant (Olea europaea): connecting patterns of seed dispersal to seedling establishment

Authors


Pedro J. Rey (tel. +34 953212145; fax +34 953212141; e-mail prey@ujaen.es).

Summary

1  Little is known about the consequences of seed-disperser activity for plant demography. We compared the spatial patterns of seed dispersal generated by frugivorous birds with those of seedling survival for the shrub Olea europaea. We examined the relative importance of dispersal in determining plant recruitment and tested whether the initial dispersal pattern persisted throughout recruitment.

2  We quantified the processes affecting each stage of regeneration (seed within a ripe fruit, dispersed seed, seedling and sapling) in different microhabitats, and evaluated transition probabilities between stages. We could then determine the overall probability of a seed in a ripe fruit becoming a sapling, and compare the probability of such an event occurring in different microhabitats.

3  Only 9.3% of the emerged seedlings reached the sapling stage, whereas 35.3% of the seeds were dispersed; 27.0% of dispersed seeds produced seedlings and 62.9% of saplings survived for 2 years. Seedling survival was therefore the critical link in regeneration. Water stress was responsible for more than 70% of seedling losses, which suggests that abiotic factors (mainly rainfall) may account for most of the fluctuation in recruitment in this species.

4  Neither post-dispersal seed predation nor germination caused changes in the initial spatial distribution of seeds, but differences in the requirements of seeds and seedlings then caused spatial uncoupling. The most favourable places for seeds were the worst for seedlings, and consequently frugivore-generated dispersal patterns differed from the final spatial pattern of recruitment.

5  Recruitment under conspecifics was nearly zero and dispersers are therefore crucial if recruitment is to occur. Their effect on the amount of recruitment was, however, overwhelmed by processes acting on the seedling stage.

6  For Olea europaea, the pattern generated by short-term recruitment dynamics persists in the long-term spatial distribution of saplings.

Introduction

Seed dispersal is a key process in plant population dynamics (Harper 1977). Although it is implicitly assumed that animal dispersers can influence plant population dynamics, their role has seldom been evaluated (Estrada & Fleming 1986; Jordano 1992; Willson 1992; Fleming & Estrada 1993; Herrera 1995). Over the last two decades, most studies of seed dispersal have focused on just one of the several stages leading to recruitment (but see Herrera et al. 1994; Houle 1998; Clark et al. 1999). Conclusions about the evolutionary and ecological importance of dispersers may therefore be limited by an absence of information about the remaining stages (Schupp 1995; Schupp & Fuentes 1995). If however, we view dispersal as one event within a sequence, it is possible to determine the relative importance of dispersers in driving recruitment dynamics by contrasting their effect with subsequent interactions.

The result of recruitment in heterogeneous habitats can be expressed in terms of the number and spatial distribution of new individuals incorporated into the population, both of which can be influenced by seed dispersers. Dispersers can limit recruitment if the probability of a seed being dispersed is lower than the probability of a dispersed seed becoming an established plant. However, fleshy-fruited trees usually produce large crops that are dispersed by a coterie of animal species (Jordano 1992; Herrera 1995), and we therefore hypothesize that dispersal is unlikely to be the limiting process in the recruitment dynamics of such species.

Nevertheless, dispersers may disseminate different numbers of seeds in a variety of microhabitats, and thus either shape the spatial distribution of the seedlings or limit recruitment by depositing large numbers of seeds in places that are unsuitable for recruitment. Processes acting at different recruitment stages are likely to be independent and evidence of such ‘uncoupling’ (Jordano & Herrera 1995) would indicate the complexity of selective forces in play during the life cycle of the plant, including seed–seedling conflicts (Schupp 1995). Different outcomes for each stage in different microhabitats (spatial discordance, Jordano & Herrera 1995) are partly a consequence of uncoupling because the recruitment-driving processes, which may have a strong effect on spatial pattern, can vary between microhabitats (Shibata & Nakashizuka 1995; Houle 1998). We hypothesize that the spatial pattern of recruitment will initially be established by seed dispersers but may then be altered by uncoupling and spatial discordance (Schupp 1995; Schupp & Fuentes 1995).

We use a demographic approach to explore our hypotheses linking seed dispersal to recruitment in Olea europaea var. sylvestris. Brot. (wild olive, hereafter Olea), a Mediterranean bird-dispersed tree. We use a simple model of the plant life cycle, composed of life stages connected by transition probabilities (Harper 1977; Jordano & Herrera 1995) that are determined by a number of processes. A process is any event that affects the probability of a propagule moving from one stage to the next: pre-dispersal seed loss (whereby the seed fails to reach the dispersed seed stage) is a process and may be caused by biotic factors, like granivorous insect larvae, or by abiotic factors, such as frost. Our approach is exploratory rather than predictive. We analyse the influence of biotic and abiotic factors on each recruitment stage (pre-dispersed seed, dispersed seed, seedling and sapling), and then estimate the transition probabilities (hereafter TP) between stages and the overall probability of a seed in a ripe fruit becoming an established juvenile plant (i.e. overall probability of recruitment, hereafter OPR). Comparison of TPs and OPRs between microhabitats allows us to explore the extent and influence of spatial discordance, and, by analysing each process, to determine which one influences the final spatial pattern and dynamics of recruitment.

Our design allows us to address some of the most relevant and current questions concerning the link between seed dispersal and recruitment (see, e.g. Jordano & Herrera 1995; Schupp & Fuentes 1995; Schupp 1995; Houle 1998; Clark et al. 1999); specifically: (i) which are the critical life stages and processes in recruitment, (ii) how do processes operating at different stages interact to affect final recruitment patterns, (iii) are the best sites for seeds also the best for seedlings, i.e. is seed–seedling conflict in operation, and (iv) does the activity of seed dispersers determine the spatial distribution of recruitment, or is it counteracted by the existence of strong spatial discordance?

Methods

Study species and study area

Olea grows along the northern Mediterranean basin and bears hermaphrodite flowers (4–6 mm in diameter), from late April to mid June, in axillary racemes on the previous year's wood. Pollination is anemophilous with partial self-compatibility (Fernández 1979). Functional dioecy, through some individuals expressing only male function, has been suggested (Jordano 1987). The fruits are drupes that ripen in autumn and winter, and contain a single seed dispersal unit (embryo plus endosperm) wrapped in a hard endocarp (the whole unit, hereafter referred to as the seed). Fruit production follows a supra-annual cycle of 2 or 3 years, depending on climatic conditions. Insect pests, especially Dacus oleae (Diptera), a pulp predator, and Prays oleae (Lepidoptera), a seed predator, can prevent seed dispersal. A number of small bird species also act as pre-dispersal predators by consuming only the pulp of the fruits (mainly Fringilla coelebs and Parus major) or the seed (Carduelis chloris and Coccothraustes coccothraustes). Small-to medium-sized frugivorous birds, mainly species of the genera Turdus and Sylvia, are the dispersal vectors. Seeds are dormant for around 20 months during which time they may be consumed by rodents (almost exclusively Apodemus sylvaticus) and more rarely by ants (Messor capitatus, M. barbarus and Aphaenogaster senilis). Epigeal germination occurs during the second winter after dispersal. The emerged seedling can be affected by water stress, physical damage (e.g. by trampling and fallen branches), generalist vertebrate and invertebrate herbivores, sap-feeding insects (mainly Euphyllura olivina, Homoptera), and fungal attack. Approximately 1 year after germination, the seedling acquires adult vegetative traits (woody stems and wax-covered leaves) and becomes a sapling. Reproductive age is achieved after 7 or 8 years under garden conditions, but much later in the field (P. Rey, personal observation). The adult plants can live for hundreds of years.

The study was conducted in dense scrubland at Sierra Sur de Jaén (37°40′ N, 3°45′ W; Jaén province, southern Spain). Between the lowlands and the mountains of the region, the landscape is typically made up of a mosaic of large patches of the scrubland (of a few hectares in area) scattered among olive orchards, pine forests and old fields. The total scrub cover in the study site is 66.8% and, of this cover, Pistacia terebinthus accounts for 30.2%, Olea europaea 13.9%, Quercus coccifera 11%, Phillyrea latifolia 5.8% and Rhamnus lycioides 5.2%. Other species, such as Pistacia lentiscus, Crataegus monogyna and Rosa canina, each account for less than 1%.

The climate is of the Mediterranean type (Elías & Ruiz 1977), with a mean temperature of 16 °C and a mean annual rainfall of 539 mm. During the seasonal drought (June to October, with rainfall dropping to about 10 mm per month in July and August), the mean monthly temperature ranges from 23.0 °C in June to 27.4 °C in July (SinambA Difusión 1996). The area suffered a severe supra-annual drought from 1991 to late 1995, followed by one rainy year (745 mm) and two average years.

Olea trees fruited copiously during the 1992–93 and 1993–94 fruiting seasons, having respectively 9999 ± 5988 (range 599–20638) and 3237 ± 5179 (range 0–28183) fruits per tree. In 1994–95 and 1995–96 the trees hardly produced any seeds, but again set large crops in 1996–97.

Sampling design

Fig. 1 summarizes the stages, processes and factors analysed in this study: pre-dispersed seeds are those contained in a full-sized fruit; dispersed seeds are the seeds deposited on the ground after consumption by dispersers; seedlings are classed as newly emerged plants still bearing the cotyledons (we consider a seedling to have emerged when its cotyledons have expanded completely); saplings are classed as plants that have acquired vegetative adult characters, generally about 1 year after germination.

Figure 1.

Diagram of recruitment to show stages (rectangles), processes (ovals) and factors influencing processes (left hand side). Values for process-specific transition probabilities (TPs) and cumulative probabilities up until that stage (shown in italics) are given. Stage-specific TPs can be obtained as the product of the process-specific TPs involved in each stage.

The analyses also give information about the spatial dynamics of recruitment. We consider microsites occupied by different shrub species to represent different microhabitats, and the transition from pre-dispersed to dispersed seed (representing the probability of arrival) in each leads to the first spatial array in the recruitment diagram (Fig. 2). Similarly, the transitions from dispersed seed to seedling and from seedling to sapling give rise to new spatial arrays characterized by the differing probability of emergence (here defined as the ratio of density of emerged seedlings : density of seeds deposited by birds) and survival, respectively.

Figure 2.

Diagram of the spatial dynamics of recruitment. Each column depicts recruitment dynamics in a microhabitat defined by the shrub species present. Values shown are process-specific transition probabilities (TPs), with overall probabilities of recruitment given at the base of each column (shown in italics). The width of the box line reflects the relative suitability of each microhabitat for overall recruitment within each stage. A, seed arrival; E, seedling emergence; Ss, seedling survival; Sps, sapling survival.

Assuming that the processes are independent, the product of the process-specific TPs occurring within each stage will determine the stage-specific TP, and the product of all TPs will determine the OPR. Table 1 summarizes the analyses undertaken.

Table 1.  Summary of the probabilities and sampling procedures used in this study
ProbabilityAcronymCohort studiedSample unitWeighting variable1Assessed
  • 1  

    The weighting variable is the sampling stratum with which each probability was sampled. The relative abundance of each stratum was used to calculate population TPs and OPR. Weighting variables were not necessary to calculate TPs in the analyses by microhabitat (i.e. the spatial dynamics of recruitment).

  • 2

    Probabilities used in the exploration of the overall probability of recruitment.

  • 3

    Probabilities used in the exploration of the spatial dynamics of recruitment.

Escape from pre-dispersal losses2PreD1994/95TreeTree/tree crop sizeNumber of healthy fruits
Seed dispersal2D1994/95TreeTree/tree crop sizeNumber of healthy fruit removed by birds
Seed arrival3A1994/950.5 × 0.5 m seedfall quadratsSeed density
Escape from post-dispersal predators2
Germination2
PostD
G
1996/97
1994/95
Protected seedfall traps and trays open to predators
Experimental sowing sites
Relative cover of microhabitats
Relative cover of microhabitats
Number of dispersed seed predated
Number of seeds germinated
Seedling emergence3E1994/952 × 2 m sampling plotsNumber of seedlings
Seedling survival2,3Ss1994/952 × 2 m sampling plotsRelative cover of microhabitatsNumber of plants (seedlings) surviving the first year
Survival of 1-year-old saplings2,3Sps11994/952 × 2 m sampling plotsRelative cover of microhabitatsNumber of plants (saplings) surviving the second year
Survival of 2-year-old saplings2,3Sps21994/952 × 2 m sampling plotsRelative cover of microhabitatsNumber of plants (saplings) surviving the third year
Survival of old-saplings in first year of study2,3OSps1Before 19944 × 4 m sampling plotsRelative cover of microhabitatsNumber of old-saplings surviving after 1 year
Survival of old-saplings in third year of study2,3OSps3Before 19944 × 4 m sampling plotsRelative cover of microhabitatsNumber of old-saplings surviving after 3 years
Survival of old-saplings in fourth year of study2,3OSps4Before 19944 × 4 m sampling plotsRelative cover of microhabitatsNumber of old-saplings surviving after 4 years

Pre-dispersed seed

Dispersal probability and the proportion of damage by biotic and abiotic factors were estimated as described in detail in Alcántara et al. (1997a, b). Briefly, during the 1993–94 fruiting season, we marked four branches on all the fruiting trees at the study site (n = 39). We monitored the fruits on each branch fortnightly and counted the number affected by each of the factors shown in Fig. 1. From these data, we estimated the number of seeds dispersed as the number of healthy ripe fruits that had disappeared from each branch between censuses. This number was corrected for fruits abscissed or dropped by birds, based on fruit counts made in quadrats placed beneath the branches.

This stage can also be influenced by certain characteristics of the habitat and individual trees. For each fruiting tree we measured: crop size, mean fresh weight of fruit, seed and pulp, mean pulp : seed ratio, percentage of scrub cover in an area 6 m around the trunk, and distance to the nearest conspecific reproductive adult.

Dispersed seed

The spatial distribution of seeds after dispersal is analysed as detailed in Alcántara et al. (2000a). Briefly, in April 1994 (after the fruit removal period) we sampled seed density around each tree along four transects placed at right angles to each other. Within each transect, we counted seed density in 0.5 × 0.5 m quadrats located 0, 1, 3, 6, 9 and 12 m away from the trunk. In each quadrat we noted the microhabitat (O. europaea, P. terebinthus, P. lentiscus, Q. coccifera, R. lycioides, P. latifolia, C. monogyna, R. canina or open interspaces).

Full details of the monitoring of post-dispersal seed predation are given in Alcántara et al. (2000b). Briefly, in October 1996 we placed 10 sampling stations randomly in each of the five most abundant microhabitats throughout the study site (O. europaea, P. latifolia, P. terebinthus, Q. coccifera and open interspaces). Each sampling station consisted of a seed-fall trap (a 26 × 33 × 5-cm aluminium pan protected by a wire mesh to prevent seed removal by rodents) and an adjacent plastic-net tray (15 × 15 × 3 cm) that was open to predators. We collected the seeds from each trap fortnightly until October 1997 and moved them into the tray to expose them to predators. Litter that accumulated in the trays was not removed so that seeds became naturally hidden during the study period. This method yields the number of fallen seeds at each sampling point (i.e. collected by the seedfall traps) as well as the number of seeds predated (i.e. those that had disappeared from the trays), and thus estimated the probability of escape from post-dispersal predation.

In December 1994, 2400 seeds collected during the analysis of the spatial distribution of seeds were distributed within 66 sowing sites, half under scrub and half in open sites. Each sowing site was protected with a cage made of 30 × 30 × 10-cm wire mesh to prevent rodent predation. Before sowing, we removed all the seeds that were naturally present in these sites. In subsequent fruiting seasons, we prevented seed arrival by attaching a translucent piece of cloth to the cage ceiling. The sowing sites were monitored for germination fortnightly throughout the period of natural seedling emergence (December to May). We considered that germination had occurred whenever the endocarp was split and the radicle had emerged.

Seedlings and saplings

In December 1994, we chose fifty-three 2 × 2-m permanent seedling sampling plots; 28 of these plots were located on four randomly positioned transects (seven plots per transect, 5 m between plots). To compensate for the low frequency of some microhabitat types in these transects, we randomly chose 10 additional plots under Olea, 10 under Phillyrea and 5 under Quercus. Thus, our sampling design was closer to a stratified random than a strict random design.

From January 1996, we monitored the number of seedlings within the plots at monthly intervals. Seedlings emerging in that year were labelled and checked monthly for survival until December 1998. The development of mature characteristics (at which point they were classified as saplings) and the cause of mortality if they died were recorded.

To determine the survival probability of older saplings, we also marked all non-reproductive individuals already present in 1994 (hereafter called ‘old-saplings’) within and around the seedling sampling plots. A 4 × 4 m plot was centred on each seedling plot to obtain a larger sampling unit for plants in this less abundant stage. We monitored the survival of old-saplings in December 1995, 1997 and 1998.

Transition probabilities

Transition probabilities (TPs) were calculated as the ratio of number of individuals completing a stage : number of individuals entering that stage. This ratio is used directly when computing TPs within microhabitats (i.e. in the exploration of spatial dynamics). To extrapolate microhabitat information to the population OPR, it is necessary to weight TP values by the relative abundance of the sampling stratum (trees or microhabitats, see Table 1). Interpretation of some TPs requires consideration of the limitations in our sampling procedures. Our estimate of escape from post-dispersal predation (PostD) has two shortcomings. First, this estimate was calculated for the 1996–97 season, and consequently does not relate to the cohort monitored for all the other processes. The values could lead to under- or over-estimates of the level for the 1994 cohort (see Results) because inter-annual variation in seed predation by rodents is common (Schupp 1988, 1990; Ostfeld et al. 1997). Secondly, the period for which seeds were exposed to predation (11 months) was shorter than that necessary for germination (around 20 months) and this could lead to an over-estimated PostD. Seeds are, however, naturally buried by litter under field conditions and detection by rodents is reduced (Hulme 1994). For a given microhabitat, A (the probability of seed arrival) was defined as the ratio of mean seed density in this microhabitat : the sum of mean densities of seeds in all the microhabitats, and therefore provides a measure of the relative likelihood of seeds being deposited in each microhabitat. In the absence of germination data from each microhabitat, our values for E (the probability of seedling emergence in a specific microhabitat) were obtained assuming that the seedlings that emerged in 1996 and 1997 came from those seeds dispersed during the 1993–94 fruiting season that escaped post-dispersal predation (note that the trees did not produce seeds in 1994–95 and 1995–96). Thus, we calculated E as the ratio of the mean number of seedlings emerged : the mean number of seeds arriving in each microhabitat.

Statistical analyses

Parametric statistical analyses (correlations and anovas) were used. When necessary to meet assumptions of the tests, data were angular transformed in the case of percentages, or Freeman-Tukey transformed in the case of density values (Zar 1984). Non-parametric tests (Spearman rank-correlation, Kruskal–Wallis analysis of variance and Kendall concordance test) were used when the assumptions for parametric tests were not met even after transformation. Repeated measures anovas were used to compare the mean percentages of sapling and old-sapling surviving between microhabitats and years. In these cases, the variable ‘year’ was considered as the within-subject effect, and the variable ‘microhabitat’ as the between-subject effect. Data are reported as means ± 1 SE.

Results

Stagewise analysis and overall probability of recruitment

Pre-dispersal stage

On a per tree basis, pre-dispersal seed damage was high and variable (45.4 ± 4.7% of the crop; range 5.6–100%). Abiotic damage to seeds was higher than biotic damage, but the effect was not significant (27.7 ± 4.4%, range 1.6–98.7%, and 17.6 ± 2.2%, range 0–61%, respectively; paired t-test: t = 1.71, d.f. = 38, P = 0.09). Escape from pre-dispersal losses was highly correlated with the overall abiotic damage, but not with biotic damage, or with any characteristics of the tree and the habitat around it (Table 2). Although seed dispersal was relatively low as a percentage of the initial crop size (32.9 ± 4.6%; range 0–87.5%), 60.59 ± 6.58% (range 0–100%) of the fruits that escaped abiotic and biotic damage were dispersed. Seed dispersal was negatively correlated with abiotic damage and positively correlated with both fruit and pulp fresh weight (Table 2).

Table 2.  Product-moment correlations between the TPs of processes occurring during the pre-dispersal stage (escape from pre-dispersal losses and dispersal) and factor types, tree characteristics, and habitat features around the trees (n = 38 trees). **P < 0.01, *P < 0.05, NS = not significant
VariableEscape from pre-dispersal lossesDispersal
Proportion of loss by biotic factors −  0.27 NS −  0.10 NS
Proportion of loss by abiotic factors −  0.90 ** −  0.57 **
Crop size0.18 NS0.14 NS
Fruit fresh weight0.13 NS0.35 *
Seed fresh weight0.12 NS0.26 NS
Pulp fresh weight0.14 NS0.38 *
Pulp : seed weight ratio0.04 NS0.27 NS
Percentage scrub cover around the tree0.03 NS0.21 NS
Distance to the nearest conspecific reproductive tree −  0.13 NS −  0.19 NS

Dispersed seeds

Overall, 63.4 ± 15.8% (range 0–100%) of the dispersed seeds escaped subsequent predation. This percentage was not correlated with seed density (rs = − 0.24, P > 0.05, n = 28). The probability of seedling emergence (E) in the 1994 cohort (weighted mean density of emerged seedlings : weighted mean seed density in the seed rain) was 0.3333, giving a value of PostD for this cohort (E1994/G1994) of 0.8655. This is higher than the value obtained directly (0.7008) which may be an underestimate, although it does not change the general pattern of the results. This value is used in Fig. 1. Average germination was relatively low (28.7 ± 2.7%), and differed significantly between open interspaces and areas under scrub (means 16.6 ± 2.4% and 40.2 ± 3.8%, respectively; F1,64 = 28.25, P < 0.01).

Seedlings

In 1996, 1564 seedlings emerged in the sampling plots, resulting in a mean density of 7.4 ± 1.4 seedlings m−2 (range 0–44). Overall, 10.8 ± 2.1% of the seedlings survived to reach the sapling stage (range 0–50%). We identified the cause of mortality of 90.7% of the seedlings; these were water stress (70%), herbivory (10%), fungal attack (5.8%), other causes (trampling, physical damage, non-fungal infections) less than 1% each. Seedling survival did not depend on seedling density (r = 0.13, P > 0.05, n = 43 plots) or on the summed proportion of loss by biotic factors (r = − 0.13, P > 0.05, n = 43), but was negatively related to the summed proportion of loss by abiotic factors (r = − 0.41, P < 0.01, n = 43). Neither abiotic nor biotic factors were correlated with seedling density (r = − 0.13 and 0.11, respectively, P > 0.05, n = 43 in both cases).

Saplings

Sapling survival was 77.8 ± 6.56% (range 0–100%) in the first year and 77.8 ± 7.27% (range 0–100%) in the second. The initial sapling density was not correlated with survival (first year: r = 0.25, P > 0.05; second year: r = 0.005, P > 0.05; n = 20 in both cases). The starting density of older saplings was 0.17 ± 0.05 saplings m−2 (range 0–1.87). For the 129 individuals present, survival was 77.2 ± 7.19% (range 0–100%) after the first year, 86.5 ± 5.1% (range 0–100%) after the third, and 92.3 ± 5.12% (range 0–100%) after the fourth.

Overall probability of recruitment

Process-specific and cumulative probabilities of recruitment are summarized in Fig. 1. A large decrease in the number of potential recruits occurred prior to the seedling stage (cumulative TP indicates that more than 90% of the potential recruits in the cohort did not survive to reach this stage). However, the most critical individual stage was the seedling stage, and seedling survival in the first year of life was the most critical process during recruitment, as seen from it having the lowest TP value (0.09). Cumulative TP declined sharply (by an order of magnitude) during this stage but, after reaching the sapling stage, the survival probability increased by almost one order of magnitude (TP = 0.74 and 0.85 for first and second year saplings, respectively). This high probability tended to continue to increase slowly as the saplings became older, as shown by values of 0.84, 0.95 and 0.96 for older saplings in the first, third and fourth year of the study, respectively.

Spatial dynamics of recruitment

Recruitment in each microhabitat is summarized in Fig. 2. The density of seed arrival differed significantly between microhabitats (Kruskal–Wallis: H(4, n = 512) = 195.1, P < 0.001; see mean densities in Table 3). The probability of seed arrival (A) was maximum under Olea, whereas it was very low in open interspaces. Intermediate values were recorded in Quercus, Phillyrea and Pistacia microhabitats. In contrast, the probability of escape from post-dispersal seed predation (PostD) did not differ significantly between microhabitats (F3,24 = 1.00, P > 0.05; sites receiving no seeds were excluded from the analysis). Mean density of emerged seedlings differed significantly between microhabitats (F4,46 = 6.64, P < 0.001; see Table 3 for mean values). The probability of seedling emergence (E) was more than 0.5 under Phillyrea, Olea and Quercus (Fig. 2), whereas seeds in open interspaces and under Pistacia had little probability of producing seedlings. Ss, the probability of seedling survival, was generally low (Fig. 2), but significantly different between microhabitats (F4,36 = 2.64, P < 0.05). No seedlings survived in open interspaces, and Ss was extremely low under conspecifics (TP = 0.04). The highest Ss occurred under Pistacia and Phillyrea (more than four times higher than under Olea). Mortality of seedlings caused by abiotic or biotic factors did not differ significantly between microhabitats (F4,36 = 1.44 and 1.81, respectively, P > 0.05 in both cases), although the incidence of water stress did (F4,36 = 3.42, P < 0.05).

Table 3.  Effect of microhabitat on seed density, seed predation, density of emerged seedlings, and starting density of old-saplings (mean across microhabitats ± 1 SE). Microhabitats with the same superscript are not significantly different in post hoc comparisons (Least Significant Difference post hoc test)
  OleaPhillyreaQuercusPistaciaOpen interspaces
  1. 1   This figure is the summed density of seedlings that emerged during 1996 and 1997, which were assumed to come from seeds dispersed in 1994 (see Methods).

Seed density (m−2)37.2 ± 2.8a14.4 ± 3.4a15.6 ± 3.4a9.3 ± 2.0b3.4 ± 0.8c
Seed predation (%)38.5 ± 12.9a14.3 ± 8.0a52.3 ± 15.4a25.0 ± 20.9a
Emerged seedlings1 (m−2)20.7 ± 5.8a9.4 ± 2.6a8.4 ± 2.6a0.9 ± 0.6b0.2 ± 0.1b
Old-saplings (m−2)0.06 ± 0.03a0.30 ± 0.11b0.61 ± 0.33b0.05 ± 0.02a0.01 ± 0.01a

The probability of sapling survival (Sps) was much higher and more homogeneous between microhabitats, but was lowest under Olea (Fig. 2). A repeated measures anova (Table 4a) indicated no significant effects of year, microhabitat or their interaction on Sps. Initial density of old-saplings differed significantly between microhabitats (F1,41 = 4.92; P < 0.01; see mean densities in Table 3), and their survival probability (Osps) followed the same trend as for saplings (Tables 4b & 5). The lack of a significant microhabitat × year interaction indicates that the inter-annual trend in survival was similar between microhabitats.

Table 4.  Results of repeated measures anovas on the percentage survival of (a) the 1996 sapling cohort and (b) the old-saplings. Microhabitat was used as the between-groups factor and year as the within-group factors. Open interspaces were not included in the analyses as no seedlings reach the sapling stage P > 0.05 in all cases
  Fd.f.
(a) 1996 saplings
Microhabitat1.073, 17
Year0.581, 17
Interaction0.223, 17
(b) Old-saplings
Microhabitat2.673, 15
Year0.582, 30
Interaction1.456, 30
Table 5.  Survival probability for old-saplings in each microhabitat in the first, third and fourth year of study. Data are averaged across sampling units
  First yearThird yearFourth year
Olea0.6667 ± 0.18260.5833 ± 0.14430.7500 ± 0.2500
Phillyrea0.8598 ± 0.12320.9000 ± 0.07240.9514 ± 0.0374
Quercus0.8875 ± 0.05150.9167 ± 0.08330.9500 ± 0.0500
Pistacia0.7500 ± 0.14431.0000 ± 0.00001.0000 ± 0.0000

The overall probability of recruitment (OPR, Fig. 2) indicates that seeds under Quercus had the highest probability of reaching the sapling stage, followed by those under Phillyrea, whereas there was no recruitment in open interspaces and little under Olea or Pistacia. There was a clear lack of concordance across microhabitats in the stage-specific transition probabilities (Kendall Coefficient of Concordance = 0.19, P > 0.05, n = 8, d.f. = 3; open interspaces excluded). Interestingly, the long-term sapling distribution mirrored the net results of the short-term processes, i.e. the density of the old-saplings was positively correlated across sampling plots with the density of 2-year-old saplings from the study cohort (r = 0.63, n = 48, P < 0.001).

Discussion

Our approach to the analysis of population recruitment dynamics allowed us to determine: (i) the relative role of biotic and abiotic factors during recruitment, (ii) the probability of a seed in a ripe fruit becoming an established plant, (iii) the processes and stages involved in the recruitment dynamics that acted as the most critical limitations upon recruitment, and (iv) the spatial dynamics of recruitment.

Relative importance of biotic and abiotic factors during recruitment dynamics

Abiotic rather than biotic agents are the most important factors determining recruitment dynamics in wild olive. The only other study that has analysed the effect of both biotic and abiotic factors simultaneously on the probability of escape from pre-dispersal damage or mortality in Mediterranean fleshy-fruited plants reported low and similar incidence for each group of factors (Krüsi & Debussche 1988). Studies on bird-dispersed plants have focused on the incidence of biotic factors, especially insects (Manzur & Courtney 1984; Jordano 1987, 1989; Kirschik et al. 1989; Sallabanks & Courtney 1992; Valburg 1992a,b; Traveset 1993). We found the incidence of insect damage to be in the same range as for other populations of Olea in southern Spain (6 and 27% in consecutive years, Jordano 1987; 21%, Alcántara et al. 1997b) as well as for other fleshy-fruited plants in the Mediterranean, such as Juniperus communis (3 − 33% in different populations; García 1998) or Cornus sanguinea (15%, Krüsi & Debussche 1988).

Abiotic rather than biotic factors accounted for the variation between trees in the probability of seeds avoiding pre-dispersal damage. We also found a negative relationship between abiotic damage and seed dispersal, but this was not strong enough to outweigh the effect of dispersers on the number of potential recruits entering the dispersed-seed stage. In fact, much undamaged fruit remained on the trees at the end of the season (21.7 ± 4.7%), suggesting that the large crops sated both dispersers and pests (Alcántara et al. 1997c; see also Jordano 1987 and Herrera et al. 1994 for other Mediterranean species).

Among the factors leading to post-dispersal damage, only rodents caused substantial seed losses. Information on Mediterranean fleshy-fruited plants (García 1998; Herrera 1984; Herrera et al. 1994; Hulme 1992, 1997 Verdú & García-Fayos 1996) shows variable levels of post-dispersal seed predation, ranging from 0.4% for Crataegus monogyna to 62.9% for Pistacia lentiscus; our estimate for Olea (36.6%) falls in the middle of this range.

Almost 40% of the unpredated seeds germinated in the field. This value is well below the potential germination capability of Olea europaea under laboratory conditions (almost 100%; Mitrakos & Diamantoglou 1984; Voyiatzis 1995). The laboratory studies indicated a strong effect of temperature on the release of dormancy, and the clear differences in germination success found between scrub-covered sites and open interspaces suggest that abiotic microenvironmental factors may have played a major role.

Abiotic factors, especially drought, were also the major cause of low seedling survival. Moreover, their effect may be even more important than estimated here as in other years three cohorts monitored at the same study site showed much lower recruitment (more than 95% of seedlings of 1995, 1997 and 1998 cohorts died before completing their first year of life; P. J. Rey et al., unpublished data). Such sensitivity, which has been observed in other plant species (Andersen 1989; Houle 1994), is common under the harsh summer conditions of the Mediterranean climate (Herrera et al. 1994; Bustamante et al. 1996; García 1998).

Sapling survival was, however, very high. This suggests that the gradual acquisition of the adult traits typical of a drought-tolerant tree (Angelopoulos et al. 1996; Schwabe & Lionakis 1996) helps the plant to cope with potentially damaging abiotic and biotic factors. Similar trends of high sapling survival have been found in other species (De Steven 1991; Herrera et al. 1994; Houle 1994; but see García 1998).

Critical stages and processes in the recruitment dynamics of olea

The OPR value (0.00557) indicated that less than 0.6% of Olea seeds became 2-year saplings and, if we further assume that the old-saplings data actually represent subsequent survival for our main study cohort, only 0.4% of seeds would survive to become 7-year-old plants. Such a low OPR value can result either from a succession of processes with relatively low TP or from a single process with critically low TP. Dispersal probability (0.5261) was much higher than the product of the subsequent transition probabilities (0.0158). Together with the fact that dispersal did not have the lowest transitional value, this finding suggests that dispersal was not the most critical process and (in accordance with our prediction) does not limit recruitment of Olea.

The critical process for recruitment in this species was seedling survival (Fig. 1). Factors affecting this process are therefore particularly likely to influence the species' recruitment dynamics. Our results indicate that water stress was the main determinant of seedling survival, and results for other cohorts in this population (P. J. Rey et al., unpublished data) indicate that it may cause nearly 100% mortality in some years. Recruitment dynamics of Olea therefore appear to depend on the occurrence at the appropriate point in each recruitment period of climatic conditions (particularly rainfall levels) that permit early survival of the seedlings. Biotic factors, in particular avian dispersers, may well affect the amount of recruitment, but their contribution is overwhelmed by the effects of abiotic factors, such as climate cycles, on the seedling stage.

Spatial dynamics of recruitment

Our field design tested the hypothesis that the spatial pattern of frugivorous seed dissemination persists throughout subsequent recruitment. We therefore explored the demographic consequences of seed dispersal, which are of paramount importance (Howe 1989; Schupp et al. 1989; Schupp 1993; Venable & Brown 1993; Herrera et al. 1994) and yet are poorly understood (Schupp & Fuentes 1995).

The distribution of Olea seeds across microhabitats is concordant with the spatial pattern of seedling emergence. The high positive correlation across microhabitats between seed rain and initial seedling density (rs = 0.90, P < 0.05, n = 5) indicates that variability of post-dispersal seed predation and failure in germination between microhabitats are not large enough to counteract the pattern generated by dispersers. Once the seedling stage is reached, however, differential survival of seedlings between microhabitats causes a clear difference in the relative suitability of the microhabitats (see also Gill & Marks 1991; Houle 1992; Herrera et al. 1994). The most favourable places for seeds became the worst for seedlings (see Fig. 2), indicating that a seed–seedling conflict (Schupp 1995) does occur in our population. For example, the microhabitat under Olea itself showed the highest probability of seed arrival and also scored high in seedling emergence, but it was the worst (apart from open interspaces) for seedling and sapling survival. The opposite occurred under Pistacia, where the probabilities of seed arrival and seedling emergence were lowest and the probabilities of seedling and sapling survival were the highest. Such uncoupling between seed and seedling stages is common in recruitment dynamics (De Steven 1991; Houle 1992, 1994, 1998; Herrera et al. 1994; Horvitz & Schemske 1994; Schupp 1995).

The spatial discordance was mainly due to the marked contrast between the high probability of seed arrival under Olea (47% of the dispersed seeds reached this microhabitat) and the low probability of survival of post-germination stages in those sites. In contrast, in other microhabitats the spatial pattern generated by dispersers persisted throughout all the recruitment stages. Probabilities ran in parallel throughout the recruitment stages under Quercus, Phillyrea and, apart from a very low probability of seedling emergence, under Pistacia.

In order to achieve successful recruitment, Olea seeds must escape the inhibitory influence of established conspecifics and reach sites where other scrub species give cover. It is at these sites that facilitation occurs. This is their regeneration niche (Grubb 1977), or their habitat choice (Bazzaz 1991). Although widespread seed dispersal seems to be a fundamental characteristic for this species, many seeds were nevertheless not dispersed and of those that were a high percentage were placed under conspecifics. This apparent inefficiency of the dispersers (Schupp 1993), together with the absence of a match between the spatial patterns of dispersed seed and either the 1-year-old seedlings or saplings, suggests that later acting factors rather than dispersal agents act to shape the spatial dynamics of recruitment in this species.

Acknowledgements

The authors benefited from the involvement of Alfonso M. Sánchez-Lafuente and Francisco Valera throughout the course of this work. We also wish to thank Andrés Aguilera, Yayo Fernández, Yiyo García, José L. Garrido, Antonio Manzaneda and José M. Ramírez for their inestimable help during fieldwork. Two anonymous referees helped us to considerably improve this manuscript.

Received 28 September 1999revisionaccepted 17 January 2000

Ancillary