The effects of seed predators on the recruitment of mangroves

Authors


Peter J. Clarke, Botany, The University of New England, Armidale 2351, Australia (e-mail pclarke1@metz.une.edu.au).

Summary

  • 1Propagule (diaspore) predation by crabs has been shown to be a major source of mortality for mangroves. We measured predation by crabs on seeds of nine tropical mangrove species in multifactorial experiments by following the fates of tethered propagules.
  • 2We tested whether planting, intertidal position and canopy gaps influenced predation of propagules and whether the predation of propagules was reduced in the presence of conspecifics. We also tested if predation influenced patterns of propagule establishment.
  • 3Mortality due to predation ranged from 22 to 100%, with Aegiceras corniculatum > Avicennia marina > Bruguiera parviflora > Aegialitis annulata > B. exaristata > Ceriops australis > C. decandra=B. gymnorrhiza > Rhizophora stylosa. Initial predator preference was correlated with the size of propagules.
  • 4Propagule planting (prone vs. implanted) and canopy type had the largest magnitude of effects across all species for treatment effects. Propagules dispersed in the prone position had more mortality while those dispersed into canopy gaps were generally less preyed upon. Three species were tested for dominance-predation by regression of stand relative density with final predation by crabs for canopy treatments. No species had significant effects that supported the hypothesis.
  • 5Predation by crabs often changed with intertidal position but showed no consistent pattern among species or gap treatments. Interactions of canopy treatment and tidal position showed that predation by crabs did not have a major influence on the zonation of mangroves in our study sites.
  • 6Analyses of covariance of predation and establishment showed that establishment success is strongly controlled by predation in six of the nine species tested. This suggests that herbivores have a greater impact on recruitment than do microhabitat effects on resources. The combined effects of predator refuge and growth preference enhance recruitment in large canopy gaps. Crab predators appear to maintain the floristic similarity between canopy gaps and surrounding forests in tropical mangrove forests of northern Australia by removing allopatric species from gaps.

Introduction

Mangroves have large non-dormant diaspores and the resulting absence of a buried seed bank increases the importance of establishment in structuring mangrove communities (Clarke et al. 2001). Pre- and post-dispersal predation of mangrove propagules is common (seeds, fruits and seedlings) (Watson 1928; Macnae 1968; Smith et al. 1989; Robertson et al. 1990; Farnsworth & Ellison 1997; Minchinton 2001), possibly because the large maternal investment in these offspring increases their attractiveness to predators. In particular, seed predation by crabs has been shown to be a major source of mortality for mangrove propagules in tropical communities (Smith 1987a,b; Smith et al. 1989; Osborne & Smith 1990; McKee 1995; Sousa & Mitchell 1999), suggesting the likelihood of predator-mediated recruitment. However, few studies of seed predation in mangrove stands directly link predation with measurements of seedling recruitment, hence measures of predation alone may not reveal its influence on plant distribution.

Patterns of both pre- and post-dispersal predation in mangroves have led to the proposition that seedling recruitment in the understorey is lower close to conspecific adults (Smith 1987a; Smith et al. 1989). This model, known as the dominance-predation hypothesis, has subsequently been tested independently by three studies (McKee 1995; McGuinness 1997a; Sousa & Mitchell 1999), all of which found little support for it. Nevertheless, differences in propagule predation by crabs have been shown among species, position in the intertidal zone, distances from creeklines and times of year (Smith 1987a,b; Smith et al. 1989; Osborne & Smith 1990; Clarke & Myerscough 1993; McKee 1995; Patterson et al. 1997; Sousa & Mitchell 1999) and these may influence establishment (McKee 1995).

Recruitment in mangroves, like other closed forest communities, is controlled by gaps in the canopy generated by disturbances such as windstorms, lightning, cyclones, pathogens and wood borers (Smith 1992; Smith et al. 1994; Feller & McKee 1999; Clarke & Kerrigan 2000). Differences in rates of seed predation between gaps and adjacent forests have been shown in a number of studies, including one in mangrove forests (Osborne & Smith 1990), and suggest that canopy gaps act as spatial refuges from predators as well as locations where light and nutrient availability are enhanced (Clarke & Kerrigan 2000). However, tests for differential predation among different sized gaps among three species of mangroves in Central America (Sousa & Mitchell 1999) did not observe the pattern expected from this generalization or from the proposal that predation is reduced in the low intertidal due to less forage time. Such ecological studies should be repeated, not only in different locations but also the same locations at different times (Sousa & Mitchell 1999).

We tested the effects of predation by crabs on the recruitment of nine mangrove species in northern Australia using multifactorial manipulative experiments. We tested the hypotheses that predation by crabs decreases (i) in canopy gaps, (ii) in the low intertidal mangrove position, (iii) where three conspecifics grow and (iv) if propagules are implanted. The first and second test whether crab foraging behaviour influences spatial patterns of predation, the third tests whether conspecifics may escape predation though predator satiation (the dominance-predation hypothesis, Smith 1987a), and the fourth the influence of timing and nature of dispersal. These tests are also related to the more general model of differential distribution in mangrove species where crab predation has been proposed as a mechanism for zonation (Louda 1989).

Methods

SPECIES AND STUDY SITE

Diaspores of water-dispersed mangroves from a range of plant families and germination types were selected for study. Nine tropical species of mangroves (Table 1) were collected by picking mature fruit from trees, except for Ceriops australis, which had to be collected from the strandline. The fresh weights of all diaspores were recorded within 3 days of collection. All diaspores were collected from mangrove stands (Lucinda and Cape Cleveland) within 100 km of Townsville, Australia (19°17′ S, 147°03′ E), and were screened for pre-dispersal insect damage. Nomenclature follows Tomlinson (1986) and Duke (1992).

Table 1.  Propagule characteristics of mangrove species used in experiments. Vivipary, germination descriptions and propagule type are from Tomlinson (1986). Other data from Clarke et al. (2001). Nomenclature for Avicennia from Duke (1992)
SpeciesFamilyViviparyGrowth fromMean fresh mass of propagules g (SE)Mean length of propagules (cm)Time until root initiation (no. of days)Initial shoot growth in relation to salinity
Aegialitis annulataPlumbaginaceaeCryptoviviparousShrub0.25 (0.01)310No growth in freshwater
Aegiceras corniculatumMyrsinaceaeCryptoviviparousShrub0.51 (0.01)48Optimal growth 5% sw
Avicennia marina var. eucalyptifoliaAvicenniaceaeCryptoviviparousTree5.16 (0.14)34No salinity effect
Bruguiera exaristataRhizophoraceaeViviparousTree4.30 (0.15)98Optimal growth 5% sw
Bruguiera gymnorrhizaRhizophoraceaeViviparousTree24.96 (0.97)1814Optimal growth 0–50% sw
Bruguiera parvifloraRhizophoraceaeViviparousTree2.53 (0.05)218Optimal growth 0–50% sw
Ceriops australisRhizophoraceaeViviparousTree1.60 (0.03)138No salinity effect
Ceriops decandraRhizophoraceaeViviparousTree2.74 (0.06)158Optimal growth 0–50% sw
Rhizophora stylosaRhizophoraceaeViviparousTree35.37 (1.09)2314Optimal growth 50–100% sw

Experiments were conducted in the mangrove forests at Bowling Green Bay on the north-east coast of Australia (19°17′ S, 147°03′ E), where 14 species occur, during the wet season of 1996. The experimental site was adjacent to the location used by Smith (1987a) in his experiment on the effect of seed predation and canopy dominance. At the seaward edge of the mangroves there was an area about 100 m wide, dominated by Rhizophora stylosa (6–8 m tall) with occasional Avicennia marina. At the upper tidal level, patches of forest about 10–20 m wide were dominated by either Ceriops australis or Bruguiera exarista and interspersed with mixed stands of Bruguiera exaristata with occasional Avicennia marina and B. gymnorrhiza. Canopy gaps were cut in stands of R. stylosa (lower intertidal), and C. australis and B. exaristata (upper intertidal), and the predator-dominance hypothesis could therefore only be tested for these three species. This was done by regressing the amount of predation, after 4 weeks of observations, with the relative density of adult conspecifics for each canopy treatment.

The most important propagule predators at this location were the large herbivorous crabs Sesarma messa, S. smithii and S. fourmanori (Grapsidae) (Giddins et al. 1986; Mitcheli 1993). Sesarma messa is generally more abundant in the high intertidal in the seaward end of these tropical estuaries (Frusher et al. 1994). Burrow densities were similar among low and high intertidal positions (c. 0.5 m−2). Other post-dispersal propagule predators include rodents, macropods and insects (Robertson et al. 1990), although none were observed in our sites.

EXPERIMENTAL DESIGN

We measured the rate of crab predation on propagules by monitoring the fate of tethered propagules as in previous studies (Smith 1987a,b; Smith et al. 1989; McKee 1995; McGuinness 1997a,b,c; Sousa & Mitchell 1999). McGuinness (1997c) explicitly tested for artifacts of tethering and location and found no evidence that it caused biases for habitat comparisons, although it was likely to underestimate the intensity of herbivory. In each replicate, five propagules of nine species were tethered on independent 50 cm lengths of twine and placed in a prone position on the sediment. Those species with an elongated propagule axis (Bruguiera exaristata, B. gymnorrhiza, B. parviflora, Ceriops australis, C. decandra and Rhizophora stylosa) had a further five propagules tethered and implanted vertically to establish them in the sediment. This provided the planted treatment with two levels (prone or implanted).

Four canopy gap treatments were applied in two replicate patches in each of two forest areas (low and high intertidal). The four canopy treatments were large cut gaps (c. 225 m2), small cut gaps (c. 50 m2), small cut gaps with shade covers, and adjacent sites with a complete canopy cover. Thus, there was a total of 16 experimental plots, each with three replicate sets of five tethered propagules (six sets if both prone and implanted) evenly spaced at least 1 m apart. This design accounts for spatial variation within plots (three replicates) and variation among them (two patches) (Underwood 1997). A propagule was considered dead following the conventions used in previous studies (Smith 1987a,b; Smith et al. 1989; McKee 1995; McGuinness 1997a,b,c; Sousa & Mitchell 1999). The fate of propagules was monitored daily for the first week, and then at weekly and monthly intervals. The density of trees adjacent to cut gaps and in plots with canopy cover was measured in 50-m2 areas and the relative density of conspecifics for each of the 16 plots was calculated (relative density = conspecific density/total density).

DATA ANALYSES

We used four- and three-factor avova to examine the effect of canopy gaps (four levels), intertidal position (two levels), planting (two levels) and patches (two nested levels) on the mean number of propagules killed by crabs per set of five propagules during the first 4 weeks of observations. The same designs were then applied to the number of propagules taking root and developing shoots (establishment) at 7 weeks, with the canopy effect reduced to two levels (gaps and canopy) due to patchy recruitment. Comparison of cumulative mortality among species was made with a single-factor anova pooled across canopy gap and intertidal position; implanted seedlings were treated separately. Analyses of covariance (ancova) were performed on numbers of propagules establishing at 7 weeks across canopy treatments with numbers of propagules eaten as a covariable for six species that had some recruitment. Cochran's test was used to test for homogeneity of variances and ln(x + 1) transformations were applied where necessary. Scheffe's test was used for post hoc comparisons of treatment means (Underwood 1997). Rather than primarily relying on tests of significance in these models (P-values), we also use estimates of variance components or ‘magnitudes of effects’ (ω2) in our anovas (Graham 2001). The methods used follow standard approaches in statistical tests and follow the recommendations of Graham (2001) with respect to negative variance estimates common in nested designs.

Correlation analyses tested the relationship between initial propagule predation (day 2) and propagule mass, length and regional abundance of species. Regression analyses were also applied to the relationship between the relative density of conspecifics and mean number of propagules killed by crabs for each of the 16 plots.

Results

PREDATION BY CRABS

Numbers of propagules removed and killed by crabs ranged from 22.5 to 100% among species (Table 2). Those species with small propagules (Aegialitis annulata, Aegiceras corniculatum and Avicennia marina) had the highest mortality and were the first to be consumed, whereas those with larger elongated propagules were slower to be consumed, especially when implanted. There was a significant correlation between propagule length and the cumulative mortality due to predation by crabs (r2 = 0.64, P < 0.01) but neither propagule mass (r2 = 0.28, P > 0.05) nor relative abundance of propagules (r2 = 0.09, P > 0.05) affected mortality.

Table 2.  Summary of the proportion of propagules killed by crabs and the proportion of propagules establishing, averaged across all treatments. Treatment effects of planting, canopy condition and intertidal level on predation of dispersed propagules and subsequent establishment are summarized
SpeciesMean percentage killed after 50 daysEffects of predation by crabsMean percentage established after 50 daysEffects of predation and other factors on establishment. Mean percentage contribution of predation to establishment failure shown
Aegialitis annulata80.8Initial gap refuge but most propagules killed after 50 days0No establishment. 80.8% fail due to predation by crabs
Aegiceras corniculatumAvicennia marina var. eucalyptifolia100.0Initial gap refuge but all propagules killed0No establishment. 100% fail due to predation by crabs
 97.1Initial gap refuge and intertidal effects but most killed after 50 days2.5Low establishment. 99.5% fail due to predation by crabs
Bruguiera exaristata77.3Implanted propagules escape predation1.7Low establishment. 75.8% fail due topredation by crabs. The remainder lacked optimal growth conditions
Bruguiera gymnorrhiza57.1Implanted propagules, low intertidal and gaps escape predators15.4Low establishment in high shore gaps. 65.4% fail due to predation by crabs. The remainder lacked optimal growth conditions
Bruguiera parviflora83.8Implanted propagules escape predators12.5Low establishment in high shore gaps. 94.4% fail due to predation by crabs. The remainder lacked optimal growth conditions
Ceriops australis70.6Implanted propagules escape predators23.8Moderate establishment patterns in gaps. 92.8% fail due to predation by crabs
Ceriops decandra57.1Low levels of predation36.0Moderate establishment patterns in gaps. 87.8% fail due to predation by crabs. The remainder lacked optimal growth conditions
Rhizophora stylosa22.5Low levels of predation, lowshore and gaps are refuges46.8Establishment patterns not strongly controlled by predation but greatest high on the shore. 22.9% fail due to predation by crabs

Within species, rates of propagule mortality by crabs differed among canopy treatment, establishment and nested patch effects (Table 3 for species which were not implanted, Table 4 for species with two planting levels). The largest variances (> 50%) were mainly associated with random spatial factors (among replicates within plots and among the two patches), although planting (prone vs. implanted) and canopy effects were significant in some species (Tables 3 and 4). Rates of mortality were significantly higher for those propagules placed in a prone position on the surface of mangrove sediments (Figs 1 and 2a). Differences in mortality between prone and implanted propagules were greater for those that had greater circumferences (R. stylosa and Bruguiera gymnorrhiza compared with B. exaristata, B. parviflora, C. australis and C. decandra, Fig. 2a).

Table 3.  Analyses of variance for number of propagules killed by crabs: (a) Aegialitis annulata day 2, (b) Aegiceras corniculatum day 2, (c) Avicennia marina ssp. eucalyptifolia day 6. Canopy and intertidal level as fixed factors. Nested plots have been pooled with main effects. Symbols for level of significance: NS = not significant; * P < 0.05; ** P < 0.01; *** P < 0.001. Magnitude of effect (ω2) at this level includes the sum of all variance below patch, note large spatial effects
Sourced.f.abc
MSω2MSω2MSω2
Canopy: C318.0***181.89*301.90 NS0
Intertidal level: I10.18 NS03.00*20.52 NS0
C × I33.52 NS121.89*03.29*11
Residual402.62700.46681.3589
Transformation None None None 
Cochran's test NS NS NS 
Table 4.  Analyses of variance for number of propagules killed by crabs: (a) all species, (b)Bruguiera exaristata, (c) Bruguiera gymnorrhiza, (d) Bruguiera parviflora, (e) Ceriops australis, (f) Ceriops decandra, (g) Rhizophora stylosa. Analyses at day 2 and day 50 for canopy, intertidal level and propagule planting as fixed factors. Nested plots have been pooled with main effects. Symbols for level of significance: NS = not significant; * P < 0.05; ** P < 0.01; *** P < 0.001
 d.f.abcdefg
MS 2MS 50MS 2MS 50MS 2MS 50MS 2MS 50MS 2MS 50MS 2MS 50MS 2MS 50
Planting: P1115.0***28.3**21.0***27.1***  0.5*84.4***  5.0 NS8.2*3.0 NS21.1**1.5 NS15.0*48.3***
Canopy: C3 38.1***81.6***17.7***15.0***  0.4*13.6** 14.6**19.1***7.3***30.4***0.5 NS16.7**4.3*
Intertidal level: I1 6.45 NS8.0 NS14.2***7.6*< 0.1 NS45.4*** 2.0 NS4.2 NS5.5*0.8 NS1.5 NS0.7 NS16.7***
C × I3 6.1 NS10.8*7.9***5.0*  0.1 NS21.3*** 6.4 NS1.3 NS3.2 NS2.4 NS0.8 NS6.6 NS5.6**
C × P3 1.2 NS2.4 NS4.0 NS0.6 NS  0.1 NS1.4 NS 7.5*0.8 NS0.5 NS1.3 NS0.2 NS2.4 NS0.4 NS
I × P1 0.1 NS1.3 NS5.5*0.3 NS  0.1 NS7.0 NS 2.0 NS0.4 NS0.1 NS1.4 NS0.4 NS1.5 NS1.5 NS
C × I × P3 1.2 NS6.0 NS1.2 NS4.7 NS  0.1 NS5.3 NS> 0.1 NS0.8 NS0.1 NS2.1 NS0.5 NS3.9 NS2.0 NS
Residual80 3.43.62.361.8  0.12.3 2.61.81.22.10.53.61.4
Transformation nonenonenonenoneln(x + 1)nonenonenonenonenoneln(x + 1)noneln(x + 1)
Cochran's test *NS*NSNSNSNSNSNS*NSNSinvariantNS
Figure 1.

Mean (± SE) number of propagules per set of five propagules killed by crabs pooled across all species and patches for propagule treatments of planting, intertidal position and canopy manipulations 2 days after the experiment commenced.

Figure 2.

Mean (± SE) number of propagules per set of five propagules killed by crabs, 6 days after the experiment commenced, pooled across: (a) canopy, intertidal position and patches for treatments of planting in prone (solid) and implanted positions (hatch); (b) planting, intertidal position and patches for effects of canopy treatments; (c) planting, canopy treatment and patches for effects of intertidal position; high shore (hatch), low shore (solid).

The effects of canopy treatments on mortality of propagules differed among species but, in general, propagules were preyed on more heavily in forests with intact canopies and in small gaps whether or not they were shaded to mimic canopy cover (Figs 1 and 2b, Table 4). Shading was used to control for disturbance effects (other than light), and differences between shaded gaps and adjacent forest with an intact canopy indicated that factors other than light sometimes reduced predation in canopy gaps (Figs 1 and 2b). Nevertheless, canopy was the most important factor and, although C*I effects were observed in some species, planting was the next most frequent effect (Tables 3 and 4, Fig. 1).

Intertidal position did not show consistent effects (Figs 1 and 2c, Tables 3 and 4) but two species (B. gymnorrhiza and R. stylosa) were eventually (after 7 weeks) more preyed upon in the high intertidal than the low intertidal, whereas B. exaristata was preyed upon more in the low shore position (Fig. 2c, Table 4). Position on the shore initially influenced predation in large canopy gaps for smaller sized propagules (A. annulata, A. corniculatum and A. marina) where more propagules were killed in the low intertidal during the initial days of the experiment.

No significant relationship was found between adult conspecific relative density and numbers of propagules eaten by crabs for the three species tested across all canopy treatments (r2 = 0.08, P > 0.05) or individual canopy treatments (natural canopy r2 = 0.22, P > 0.05; artificial shade r2 = 0.14, P > 0.05; cut gaps r2 = 0.01, P > 0.05).

ESTABLISHMENT

Numbers of propagules establishing ranged from 0 to 46.8% (Table 2, Fig. 3a), with the majority of propagules surviving being those that were artificially planted (Fig. 3a). Two species (Aegiceras corniculatum and Aegialitis annulata) did not establish, whilst four species had less than 20% establishment (A. marina, B. exaristata, B. gymnorrhiza, B. parviflora) (Table 2). Field observations suggested that poor root growth inhibited establishment of those propagules that survived predation. Canopy treatments had significant effects in five species of the remaining seven, with enhanced establishment in large canopy gaps in three (Table 5, Fig. 3b). Small canopy gaps did not appear to enhance establishment in any species (Fig. 3b). These effects are explained by reduced predation in large gaps as final numbers killed by crabs was a significant covariable for species with more than 10% establishment (ancova results in Table 5). Intertidal position had significant effects in three species (B. gymnorrhiza, B. parviflora, R. stylosa) but there was no consistent enhancement of establishment at either position on the shore (Fig. 3c).

Figure 3.

Mean (± SE) number of propagules established, 7 weeks after the experiment commenced, pooled across: (a) canopy, intertidal position and patches for propagules dispersed in the prone (solid) and implanted (hatched) positions; (b) intertidal position and patches for effects of canopy treatments; (c) planting, canopy treatment and patches for effects of intertidal position; high (hatch), low (solid).

Table 5.  Analyses of variance and covariance for number of propagules established at 7 weeks: (a) Avicennia marina ssp. eucalyptifolia, (b) Bruguieraexaristata, (c) Bruguiera gymnorrhiza, (d) Bruguiera parviflora, (e) Ceriops australis, (f) Ceriops decandra, (g) Rhizophora stylosa. Analyses for canopy, intertidal level and propagule orientation as fixed factors. Nested plots have been pooled with main effects. Symbols for level of significance: NS = not significant; * P < 0.05; ** P < 0.01; *** P < 0.001
 d.f.a MSb MSc MSd MSe MSf MSg MS
Planting: P10.67 NS40.04***4.17 NS25.22***27.09**356.5***
Canopy: C30.75 NS0.47 NS3.90*8.11**11.14***13.93**5.18**
Intertidal level: I10.75 NS0.67 NS24.00***10.67*0.52 NS0.84 NS10.01**
C × I30.75 NS0.47 NS7.41**1.16 NS1.47 NS1.07 NS0.955 NS
C × P30.47 NS1.90 NS0.33 NS2.90 NS2.32 NS0.37 NS
I × P10.67 NS13.17***0.67 NS1.50 NS1.26 NS0.5 NS
C × I × P30.47 NS4.58**0.94 NS2.42 NS5.87 NS1.33 NS
Residual#800.50.224.471.571.583.260.66
Transformation ln (x + 1)ln (x + 1)NoneNoneNoneNoneNone
Cochran's test **NSNSNSNSNS
Canopy30.37 NS12.45***1.36*0.61 NS0.90 NS5.21 NS
Predation (covariable)10.19 NS56.05***28.11***115.95***212.85***163.22***
Canopy × covariable30.24 NS7.24*0.97 NS0.53 NS0.58 NS2.27 NS
Residual880.251.390.420.690.743.45

The relationship between numbers of propagules establishing and the numbers that had been killed by predators was not always singular (Table 2), i.e. some propagules that escaped predation did not establish and died from other causes. Three species (B. exaristata, B. gymnorrhiza, R. stylosa) had more than 20% of establishment failure attributed to factors other than predation by crabs (Table 2). Nevertheless, the broad patterns of establishment appear to be determined by initial propagule mortality due to crabs, e.g. B. gymnorrhiza and R. stylosa propagules were killed by crabs less often in the low intertidal and were better able to establish there.

Discussion

We have shown, by manipulative field experiments, that predator-mediated recruitment is likely in tropical mangrove forests of northern Australia. Both predation by crabs and establishment of the remaining propagules varied among species and treatments. Planting had the greatest effect, followed by canopy condition and then position on the shore. The largest spatial effects, however, were among patches indicating the idiosyncratic nature of crab foraging (McGuinness 1997a, 1997b). Species and spatial differences in propagule predation do not explain the distribution and abundance of mangrove stands across the shore or along tidal gradients in estuaries. They may, however, partially explain the process of gap colonization in mangrove communities.

INTERTIDAL FORAGING HYPOTHESIS

We found that the rates of predation by crabs on a wide range of mangrove propagules did not vary strongly with position along the intertidal gradient. Only two species (B. gymnorrhiza and R. stylosa) showed higher overall rates of propagule predation by crabs higher on the shore, typically reported elsewhere (Smith 1987b; Smith et al. 1989; Osborne & Smith 1990; McGuinness 1997b). Similar patterns were also present in Smith's (1987a) landmark study, with three species showing less predation by crabs in the low intertidal zone. Differences in foraging by sesarmid crabs appeared to account for these patterns because they consume up to 30% of leaf litter in the low- and mid-shore Rhizophora forests, whilst in the high shore Bruguiera and Ceriops forests, crabs remove 70% of litter (Robertson & Daniel 1989). In Smith's experiments less predation was also found at low and high intertidal levels of A. marina where conspecifics were present. These effects could be a result of the absence of sesarmid crabs from these zones because the high-intertidal Avicennia forests are dominated by microphagous ocypodids rather than sesarmid crabs; hence these forests are less likely to be subject to propagule predation. Converse patterns of propagule predation have been found in Panama, again relating to differences in predator guilds across the shore (Sousa & Mitchell 1999). We conclude that rates of predation can vary with intertidal position but a general ‘time for foraging’ model does not apply.

GAP REFUGE HYPOTHESIS

We found patterns of less predation of propagules by crabs in large canopy gaps relative to areas with an intact canopy and shaded small gaps. Through time, however, these differences became less pronounced and only the Rhizophoraceae had significant differences in numbers preyed on by 7 weeks, particularly when they were implanted. These results are consistent with those found by Osborne & Smith (1990) but differ from those found in Panama, where no differences could be detected (Sousa & Mitchell 1999). The reduced effect of a gap refuge through time in our study probably reflects the very low level of background propagules when we did our study and the small gap size relative to those measured by Osborne & Smith (1990). Smith (1992) suggested that increases in soil temperature associated with canopy gaps may underlie these patterns as the crab fauna in gaps is dominated by ocypodids, which prefer warm sediments, rather than sesarmid seed predators. We conclude that large canopy gaps can provide spatial refuges from crab predation in our study system.

DOMINANCE-PREDATION HYPOTHESIS

Patterns of propagule predation by crabs in mangroves have led to the suggestion that seedling establishment is lower where conspecific adults grow (Smith 1987a; Smith et al. 1989) but this model is poorly supported (McKee 1995; McGuinness 1997a; Sousa & Mitchell 1999). Our tests similarly found no consistent pattern of crab predation associated with the dominance of the three species in the forest canopy. One species (R. stylosa) had some reduced predation where adults were abundant, corresponding to Smith's original outcome. This is not surprising given that the experiment was conducted in the same location and that R. stylosa was preyed upon in the high shore environment. These results highlight one of the problems associated with assessing dominance-predation in mangroves where dominance of species is also related to position on the shore. Thus controlling for shoreline effects, independently of dominance, may be the only method to better test this hypothesis. We conclude that a dominance–predation effect may occur in some mangrove stands but it is not a general phenomenon among the three species tested.

DOES PREDATION REGULATE RECRUITMENT?

We found that establishment was strongly controlled by predation of propagules in six of the nine species in our experiments (Table 2). Mangrove species with small-sized propagules that are high in nutrients and low in fibre are the most vulnerable to crab predators (Smith 1987a; McKee 1995) and had little or no establishment success. Hence predators may control the relative abundance of species and account for the dominance of Rhizophora stylosa in north Australian tropical estuaries. Naturally implanted propagules are also more likely to escape predators and establish, highlighting the importance of the dispersal phase in recruitment success (McKee 1995; McGuinness 1997b; Patterson et al. 1997). Not all experimental studies, however, have shown crabs to have negative effects on recruitment (Siddiqi 1995; Minchinton 2001), and crab perturbation of sediment can enhance establishment in temperate mangrove systems (Minchinton 2001). Nevertheless, in our study, herbivores have a greater impact on initial recruitment than did microhabitat resource effects. This suggests that there is recruitment limitation for propagules dispersed in a prone position. Strong seed predator effects on recruitment have also been reported in rain forests where gaps also provide refuges from predators (see review by Wright 2002).

We found that initial rates of predation and subsequent establishment were consistent among spatial factors, resulting in patchy recruitment of seedlings. For example, B. gymnorrhiza and R. stylosa had reduced numbers of seeds killed by crabs low on the shore, resulting in better overall establishment. Enhanced establishment of species in some light gaps also appears related to spatial refuges from predators for several species (B. exaristata, B. parviflora, C. australis, C. decandra), which is subsequently reinforced by better growth in gaps. Recruitment can therefore be manifested through multiple life history stages, starting with propagule predation generating patterns that are subsequently enhanced through growth. Our results show that enhanced recruitment starts with ‘dispersal refuges’, through implanting at low tide, followed by ‘spatial refuges’ from predators in large canopy gaps. This process may explain the similarity in floristic composition between canopy gap recruits and adjacent forests with intact canopies (Clarke & Kerrigan 2000). This is because allopatric species dispersing into canopy gaps are more likely to be preyed upon as they are dispersed in a prone position. In contrast, propagules from the surrounding canopy are more likely to implant in the sediment at the gap edges and escape predation, thus the gap is filled with like species. We therefore propose that seed predators provide the mechanism for a ‘direct replacement’ model for colonization of gaps in tropical mangroves.

Acknowledgements

We wish to thank the staff at the Australian Institute of Marine Science (AIMS) for their support throughout the project. In particular we are grateful to Barry Clough, Dan Alongi, Janet Ley, Lindsay Trott and Alistair Robertson, for sponsorship at AIMS and logistical support. We are indebted to Paul Dixon and to Otto Dalhaus for diaspore collection. For assistance in the field we would like to thank Kellie Mantle, John Steer, Greg Calvert, Christine Westphal, Karyn Andersen and Angus Galletly. Tom Smith and Alistair Robertson inspired the study. The referees and editors provided valuable comments and improvements to the manuscript. The Queensland Department of Primary Industries and the National and Marine Parks Services granted permission for diaspore collection and the cutting of gaps. The project was funded by an Australian Research Council grant A19530936.

Ancillary