See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/260406415

Secondary dispersal by ants promotes forest regeneration after deforestation

Article in Journal of Ecology · February 2014

DOI: 10.1111/1365-2745.12226

CITATIONS READS

57 309

3 authors, including:

Silvia C. Gallegos Matthias Schleuning

Martin Luther University Halle-Wittenberg Senckenberg Biodiversität und Klima - Forschungszentrum
23 PUBLICATIONS 240 CITATIONS 214 PUBLICATIONS 9,159 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Silvia C. Gallegos on 17 March 2016.

The user has requested enhancement of the downloaded file.

Journal of Ecology 2014, 102, 659–666 doi: 10.1111/1365-2745.12226

Secondary dispersal by ants promotes forest

regeneration after deforestation
Silvia C. Gallegos1,2,3*, Isabell Hensen1,4 and Matthias Schleuning1,2
1
Institute of Biology/Geobotany and Botanical Garden, Martin Luther University Halle-Wittenberg, Am Kirchtor 1,
D-06108 Halle, Germany; 2Biodiversity and Climate Research Centre (BiK-F) & Senckenberg Gesellschaft fu €r
3
Naturforschung, Senckenberganlage 25, D-60325 Frankfurt (Main), Germany; Herbario Nacional de Bolivia (LPB) –
Instituto de Ecologıa – MNHN, Universidad Mayor de San Andre  s, Campus Universitario Cota Cota, La Paz, Bolivia;
4
and German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Deutscher Platz 5e, D-04103
Leipzig, Germany

Summary
1. Many remnants of tropical forests are surrounded by deforested areas dominated by successional
vegetation. Although secondary dispersal by scatter-hoarding rodents and ants may influence seed
predation and seedling recruitment, very little is known about its importance in degraded forest eco-
systems.
2. We studied the effects of secondary dispersal on seed predation, germination and seedling recruit-
ment of the primarily bird-dispersed tree Clusia trochiformis in a tropical montane forest in Bolivia.
We carried out exclosure experiments that allowed or excluded access to seeds by vertebrates in
three habitat types (forest interior, degraded habitat close to and degraded habitat far from the forest
margin) in a spatial block design at six sites. We offered a total of 1440 seeds (both with and with-
out an aril) and marked half of them with a thread to follow their fate after 48 h and after 1 month.
3. We found that secondary dispersal by ants was highest in the forest interior but was also frequent
in degraded habitats close to and far from the forest edge. Secondary dispersal significantly
increased seedling recruitment, particularly in the degraded habitats, probably because seeds were
often dispersed to sites beneath the leaf litter. Recruitment success increased significantly with dis-
persal distance. High recruitment of secondarily dispersed seeds in the degraded habitat was due to
the combined effect of reduced predation and increased germination of seeds that had been moved
by ants.
4. Synthesis. In the absence of secondary dispersal, seed germination and seedling recruitment were
very low in degraded habitats. Seed dispersal by ants substantially increased natural regeneration in
the deforested habitats. Our experiments thus demonstrate that the effects of biotic interactions on
plant demography can vary across habitat boundaries at small spatial scales and that secondary dis-
persal is a crucial and overlooked process that can aid the regeneration of deforested habitats in the
tropics.
Key-words: aril, Clusia, directed dispersal, forest regeneration, germination, human disturbance,
plant population and community dynamics, plant–animal interactions, seed predation, tropical forest

by vertebrates, especially birds, are secondarily dispersed by

Introduction
ants and scatter-hoarding rodents (Vander Wall & Longland
Most tropical tree species are dispersed by animals attracted 2004). Secondary seed dispersers rearrange seed shadows
by their fleshy diaspores (Howe & Smallwood 1982). Seed generated by primary seed dispersers (Levey & Byrne 1993;
dispersal of these trees often constitutes several consecutive Passos & Oliveira 2002) and may benefit the recruitment suc-
steps by different agents (Vander Wall & Longland 2004). cess of the dispersed plant species (Passos & Oliveira 2002;
For instance, many plant species that are primarily dispersed Christianini & Oliveira 2010; Hirsch et al. 2012). While pri-
mary dispersal mostly reduces density-dependent seed mortal-
ity near the parent plant (Schupp, Jordano & G
omez 2010),
*Correspondence author: E-mail: [email protected] secondary dispersal may reduce seed predation by moving
© 2014 The Authors. Journal of Ecology © 2014 British Ecological Society
660 S. C. Gallegos, I. Hensen & M. Schleuning

seeds to safe sites (Wenny 1999; Vander Wall & Longland dispersers, especially to birds and ants (Passos & Oliveira
2004) and may increase germination due to directed dispersal 2002). We conducted dispersal and predation experiments
to suitable microhabitats (Passos & Oliveira 2002; Briggs, with seeds of C. trochiformis in three habitat types: forest
Vander Wall & Jenkins 2009) in spatially heterogeneous envi- interior, degraded habitat close to the forest edge and
ronments (Wenny 2001). Thus far, studies of the effects of degraded habitat far from the edge in a tropical montane for-
seed dispersal on forest regeneration have rather focused on est in Bolivia. Specifically, we asked the following questions:
the effects of primary than of secondary seed dispersal (e.g. (i) How does secondary dispersal and seed predation differ
Aide & Cavelier 1994; Wunderle 1997; but see Cole 2009) between animal agents and habitat types? (ii) Is the risk of
and sometimes have imprecisely equated secondary seed seed predation related to secondary dispersal? (iii) Does sec-
removal with seed predation (Vander Wall, Kuhn & Beck ondary dispersal have habitat-specific effects on germination
2005). and seedling recruitment? To our knowledge, this is the first
Many tropical forests have been deforested (FAO 2010), study testing the importance of secondary seed dispersal and
and investigation of the processes that influence the regener- predation for forest regeneration in the highly disturbed
ation of deforested areas is an important research field in Andean forest ecosystems.
applied ecology (Lamb, Erskine & Parrotta 2005). Despite
the recognized importance of secondary seed dispersal in
Materials and methods
tropical forests, few studies have compared secondary seed
dispersal between forests and deforested habitats. Secondary
STUDY AREA
dispersal may aid in reducing recruitment barriers at spa-
tially heterogeneous deforested sites by directed dispersal to The study was conducted between February and March 2012 at the
suitable microhabitats (Briggs, Vander Wall & Jenkins 2009) eastern cordillera of the Bolivian Andes, in the vicinity of Chulumani
and by reducing the intensity of seed predation (Duncan & village in La Paz (16°240 S, 67°310 W). The study area is located in
Chapman 2002; Cole 2009). Previous studies on ants have the montane forest of the Yungas region between 1900 and
2500 m asl. Mean annual temperature is 20.8° C and mean annual
found context-specific effects of forest degradation on sec-
precipitation is 1459 mm, with a peak between January and March
ondary dispersal, and there is evidence for both heightened
(Molina-Carpio 2005). The forests around Chulumani have been
seed removal (Zelikova & Breed 2008; Christianini &
highly fragmented, mainly by uncontrolled human-induced fires
Oliveira 2013) and reduced seed removal (Dominguez- (Killeen et al. 2005) and the extension of coca (Erythroxylum coca)
Haydar & Armbrecht 2011; Zwiener, Bihn & Marques plantations. Only two relicts of mature continuous forest of ca.
2012) by ants in degraded habitats. The roles of scatter- 3000 ha each remain in the area. These relicts are surrounded by a
hoarding rodents have been shown to change from second- fire-degraded area dominated by Pteridium arachnoideum (Kaulf.)
ary dispersers in large Amazonian forest fragments to seed Maxon (bracken) and successional shrubs. The litter of bracken
predators in small fragments (Jorge & Howe 2009), while covers approximately 70% of the soil in these habitats. Details of
Cole (2009) found that large seeds were only rarely second- microclimate and soil properties of forest and degraded habitats in the
arily dispersed by scatter-hoarding rodents in pastures in study area are described elsewhere (Lippok et al. 2013).
Costa Rica. The scarcity of previous studies in tropical mon-
tane forests and the inconclusive findings of previous work STUDY SYSTEM
call for a better understanding of secondary seed dispersal
and seed predation in natural and disturbed tropical forests. Clusia trochiformis (Clusiaceae) is a dioecious tree distributed in the
tropical Andean montane forest from Venezuela to Bolivia. It is a
In particular, there is a lack of studies testing effects of
common tree in mature forests and also grows in the fire-degraded
secondary seed dispersal and seed predation on plant germi-
habitats. Many Clusia species are able to grow in open and dry con-
nation and seedling recruitment. This information is essential ditions (L€uttge 2007) and have nursing effects on other plant species
for understanding the quantitative importance of secondary (e.g. Dias & Scarano 2007). Mature fruits of C. trochiformis are pres-
seed dispersal and seed predation for tropical forest regener- ent from December to March. Fruits are globular capsules (mean 
ation (Reid & Holl 2013). SD = 12.2  1.5 mm length, 8.9  1.5 mm wide, n = 50) that open
The tropical Andes have suffered high deforestation rates in to expose four diaspores (10.2  1.5 mm length, 4.1  0.7 mm
previous decades, especially due to frequent fires, and many wide, n = 50). Each diaspore (fresh weight: 0.3  0.1 g, n = 50)
formerly forested areas are dominated by ferns and early- contains 1–5 seeds (1.97  1.18 seeds, n = 200 diaspores) covered
successional shrubs (Lippok et al. 2013). The deforested areas by a red lipid-rich aril, that is, a fleshy, nutrient-rich seed appendage.
are characterized by harsh abiotic conditions that limit seed Seeds are recalcitrant, that is, rapidly lose their viability after
dispersal.
germination and recruitment of forest species (Hartig & Beck
Seeds of Clusia trochiformis require humid conditions for germina-
2003; Lippok et al. 2013). In this experimental study, we ask
tion. Microclimatic conditions on the forest floor are suitable for seed
whether secondary dispersal contributes to forest regeneration
germination, whereas microclimatic conditions in the deforested habi-
by reducing seed predation and increasing germination and tats are harsher and seed germination may be restricted to moist
seedling recruitment of the widespread, primarily bird- microsites beneath the litter layer. Accordingly, recruitment rates of
dispersed tree Clusia trochiformis Vesque (Clusiaceae). Clusia C. trochiformis were higher with litter presence in a simultaneous
species produce seeds that are surrounded by a lipid-rich aril sowing experiment in the deforested habitat (S.C. Gallegos, I. Hensen
that is highly attractive to primary and secondary seed & M. Schleuning, unpubl. data).

© 2014 The Authors. Journal of Ecology © 2014 British Ecological Society, Journal of Ecology, 102, 659–666
 Secondary dispersal promotes reforestation 661

The fleshy arils of C. trochiformis attract many bird species, for and invertebrates in different habitat types, we carried out an exclo-
example Mionectes striaticollis, Chiroxiphia boliviana, Tangara sure experiment. At six sites, three in each of the two large forest re-
xanthocephala and Catharus ustulatus, which are the primary dispers- licts and separated by at least 1 km, we placed a 240-m-long transect
ers of C. trochiformis. Birds frequently drop intact or partially eaten perpendicular to the forest edge (Fig. 1). Each transect comprised
diaspores, containing aril remnants, and many diaspores fall spontane- three habitat types: forest interior (160 m from the forest margin),
ously from the fruiting tree to the ground. In the study area, seeds are degraded habitat close to the forest edge (5 m outside from the forest
readily visited by different species of ants. margin) and degraded habitat far from the forest edge (80 m outside
To identify the ant species attracted to the seeds, we obtained data from the forest margin). In each site and habitat type, we established
on ant attendance by dedicated point observations and by ant observa- a single block including two pairs of depots (i.e. groups of seeds)
tions during revisions of the experiments (see species list in Table S1 separated by ca. 10 m; each pair had one depot with a wire exclusion
in Supporting Information). For point observations, we offered 10 cage (25 9 25 9 15 cm, 1.2 cm mesh), staked to the ground with
Clusia seeds with an aril (one-seeded diaspores) and 10 seeds without 7 cm nails to exclude vertebrates (see Passos & Oliveira 2002), and
an aril on the ground. The observations were carried out for 1 h in one depot without an exclusion cage (hereafter, open). Each depot
each habitat, twice at each site. We recorded the attracted ant species was separated from the other by at least 5 m. A depot comprised a
and their behaviour, estimated the distance they moved the seeds and group of 20 seeds, which were placed on a 15 9 15 cm square of
recorded the final location of dropped seeds. We collected individuals fabric, that is, a polyester–cotton material that allowed rain drainage
of each ant species that interacted with the seeds for identification. and was fixed to the ground in order to relocate the seeds. In each
Voucher specimens were deposited at the Universidad Mayor de San depot, we offered seeds with or without an aril (arillate vs. cleaned),
Andr
es at the Colecci
on Boliviana de Fauna (CBF), La Paz, Bolivia. and with or without a thread (thread vs. non-thread), including five
Apart from ants, we did not observe any other invertebrate taxa fre- seeds for each factor combination. In total, 1440 seeds were deposited
quently attracted to Clusia seeds. Among vertebrates, we recorded in 72 depots at six sites. To exclude effects of human handling on
several species of rodents in the study area (for details, see Table S2). seed attendance, all offered seeds were manipulated with latex gloves.
We did not observe rodents caching Clusia seeds. Ants readily approached and handled the presented seeds and moved
unrestricted over the fabric. Ants were able to move seeds with
threads without difficulty. Sporadically, we observed threads entan-
EXPERIMENTAL DESIGN gled in the vegetation, which could have resulted in some instance in
reduced dispersal distances.
For the experiments, we collected ripe fruits from six individuals of
To estimate secondary dispersal and predation, we established the
C. trochiformis. We manually separated diaspores from the fruits and
depots the day before the experiment started and relocated seeds from
selected diaspores containing only one seed to ensure similar sizes of
each depot at 08:00 and 17:00 h during the first 48 h. An additional
all offered seeds. To assess the effect of aril presence, in half of the
observation was conducted 1 month after seed presentation. During
seeds, the aril was removed with latex gloves and water. To follow
each observation, seeds with a thread were searched for within a 2.5 m
the fate of individual seeds, half of the seeds, both arillate and
radius, and seeds without a thread were searched for within a 1 m radius
cleaned, were equipped with a thread labelled with a unique numeric
because they were more difficult to relocate. We deliberately did not
ID. The thread (20 cm nylon fishing line) was passed through the
assess the rare events of long-distance seed dispersal by ants but
endosperm of the seeds with a needle and burned at the end; the
focused on the frequent seed movements at a small spatial scale. Sec-
embryos of most seeds were left intact by this treatment. To evaluate
ondary seed dispersal was indicated if the thread was relocated with the
the relative importance of seed dispersal and predation by vertebrates

160 m 5m 80 m

Close Far
Forest Degraded

Depot
Block

Arillate Cleaned

Exclosure Open

Fig. 1. Experimental study design of secondary dispersal and seed predation experiments. Clusia trochiformis seeds were offered in three habitat
types: forest interior and degraded habitats close to and far from the forest edge. In each habitat type, we offered seeds in four depots, each con-
taining 20 seeds. Two depots were caged (excluding vertebrates), and two were open. In each depot, we offered an equal number of arillate and
cleaned seeds and of seeds with and without an attached thread. The experiment was replicated at six sites in montane forest remnants in Bolivia.

© 2014 The Authors. Journal of Ecology © 2014 British Ecological Society, Journal of Ecology, 102, 659–666
662 S. C. Gallegos, I. Hensen & M. Schleuning

intact seed still attached, and in case of seeds without a thread, if they model. All statistical analyses were conducted with the R statistical soft-
had been moved away from the fabric without damage. If a thread was ware, v.2.15.2 (R Development Core Team 2012) and with dedicated
recovered without a seed or if a seed was partially eaten, predation was packages lme4 (Bates, Maechler & Bolker 2011) and MuMIn (Barton
assumed. For each relocated seed, the direction and distance from the 2012).
seed to the fabric edge were assessed. In the observations after 1 month,
we relocated seeds again and additionally checked if they had germi-
nated in the meantime. Seeds that did not germinate and were not pre- Results
dated were recorded as dead seeds. All seeds and threads that could not During point observations of Clusia seeds, we mostly
be relocated were excluded from all analyses. We approximated seed-
observed ants of Pheidole spp. (in groups) and Acromyrmex
ling recruitment from the germination rates observed during the first
sp. (single individuals) dispersing the seeds and removing the
month after the start of the experiment, because seedling recruitment
after 1 month was the major limiting factor in a simultaneous sowing
aril after seed relocation. By contrast, Solenopsis spp.
experiment with the same species. In this study, approximately two- recruited many workers to arillate seeds and usually removed
thirds of the recruited seedlings survived the first year after germination the aril on the spot. Overall, ants carried about 20% of the
(S.C. Gallegos, I. Hensen & M. Schleuning, unpubl. data). dispersed seeds to their nests. More than a half of the other
moved seeds were left beneath the litter layer, while the
remaining seeds were left on top of litter or on patches of
DATA ANALYSIS bare soil. Ants moved the seeds over short distances (n = 172
observations: mean  SE = 17.4  2.9 cm; maximum dis-
Most events of secondary dispersal by ants were recorded during the
first 48 h after presentation. We therefore analysed differences in sec-
tance = 100 cm).
ondary dispersal among habitats and treatments for this time interval In the experiment, 58% of the offered seeds were removed
and calculated the proportion of seeds that were moved away from from the depots. Of these seeds, 42% were removed from
the fabric. In contrast, most events of seed predation were recorded exclusion cages (attributable to ants) and 58% were removed
after 1 month. Because of the higher detectability of seeds with a from open treatments (attributable to ants and vertebrates;
thread after 1 month, we restricted the analysis of seed predation to Fig. 2). We relocated 81% of the removed seeds after 48 h
seeds offered with a thread and calculated the proportion of seeds that and 49% after 1 month. At the end of the experiment, we
had been predated 1 month after presentation. Analysis of effects of relocated 89% of the removed seeds offered with a thread and
secondary dispersal on seed predation was also restricted to seeds 20% of the removed seeds offered without a thread.
with a thread and was based on the data from observations 1 month
Secondary dispersal during the first 48 h after presentation
after presentation. Seed germination was defined as the proportion of
was lower in the degraded habitats than in the forest interior
seeds that had germinated after 1 month, excluding all predated seeds
from the analysis (predated seeds could not germinate). Accordingly,
(Fig. 2a); this difference was only significant between forest
seedling recruitment was defined as the proportion of seeds that ger- and degraded habitat far from the edge (Table S3a). Second-
minated after 1 month, but including predated seeds in the analysis ary dispersal was significantly higher in open than in exclo-
(predated seeds did not recruit) and thus measures the recruitment sure treatments, and arillate seeds were more likely to be
success of all offered seeds during the first month after presentation. dispersed than cleaned seeds (Fig. 2b). Differences in second-
In all statistical models of seed germination and recruitment, thread ary dispersal between exclosure and open treatments were
presence was included as a covariate to account for potential effects weaker in arillate than in cleaned seeds (Table S3a). Mean
of thread presence on germination probabilities. secondary dispersal distances were short (mean  SE =
We compared secondary seed dispersal and seed predation among 13.6  0.6 cm; maximum distance = 75 cm).
habitat types (forest interior, degraded habitat close to and degraded Seed predation also decreased from forest to degraded
habitat far from the forest edge), exclosure (exclosure, open) and aril
habitats, but this effect was not significant (Fig. 2c and
(arillate, cleaned) treatments with generalized linear mixed-effect
Table S3b). Seed predation was higher in open than in
models (GLMMs), assuming binomial error distributions of the
response variables. To account for the spatial design of the experiment,
exclosure treatments (Fig. 2d). This effect was more pro-
we included site, block and depot identities (nested within each other) nounced for cleaned than for arillate seeds (Fig. 2d and
as random factors in all models. To assess the effect of secondary dis- Table S3b). We found higher predation rates of secondarily
persal on seed predation, we included secondary dispersal as a fixed dispersed seeds than of non-secondarily dispersed seeds in
effect additional to habitat and treatment levels. Similarly, to analyse the forest, but the opposite in degraded habitats far from the
effects of secondary dispersal on seed germination and seedling recruit- edge (Fig. 3a and Table S4a). Dispersal distance had no
ment, we included secondary dispersal as an additional fixed effect. To effect on seed predation in any habitat type (Fig. 3b and
assess effects of distances of secondary dispersal on seed predation, ger- Table S5a).
mination and recruitment, we additionally fitted statistical models only Seed germination, excluding all predated seeds, was much
including seeds that had been secondarily dispersed and included thread
higher in the forest than in the degraded habitats (Fig. 3c).
presence, and main and interaction effects of dispersal distance and hab-
Secondary dispersal positively affected seed germination in
itat type as fixed effects. Dispersal distances were log-transformed prior
to the analysis. In all analyses, we fitted models for all possible combi-
all habitats. Seeds offered with an aril germinated at a higher
nations of predictor variables (including second-order interaction terms) proportion than cleaned seeds, but only if seeds were second-
and compared model fits according to the Akaike Information Criterion, arily dispersed (Table S4b). Dispersal distance had a positive
corrected for small sample size (AICc, Burnham & Anderson 2002). In effect on seed germination in all habitats, the effect being
all cases, we selected the model with the lowest AICc as the single best largest in degraded habitats (Fig. 3d and Table S5b).

© 2014 The Authors. Journal of Ecology © 2014 British Ecological Society, Journal of Ecology, 102, 659–666
 Secondary dispersal promotes reforestation 663

60
(a) (b)

Secondary dispersal (%)

Arillate
50 Cleaned

40

30

20

10

0
Close Far Exclosure Open
Forest Degraded

60
(c) (d)
50
Fig. 2. Proportion of secondarily dispersed
Seed predation (%)

seeds (a) in the three habitat types and (b) in 40

exclosure and aril treatments. Proportion of
predated seeds (c) in the three habitat types 30
and (d) in exclosure and aril treatments.
20
Shown are means + SE. Secondary dispersal
of Clusia trochiformis was assessed during 10
the first 48 h after presentation. Seed
predation of the same species was assessed 0
Close Far Exclosure Open
1 month after presentation. Experimental
details are shown in Fig. 1. Forest Degraded Treatments

Seedling recruitment, including also predated seeds, was non-myrmecochorous plants primarily dispersed by birds in
higher in the forest than in degraded habitats (Fig. 3e). Sec- the Neotropics (e.g. Passos & Oliveira 2002; Christianini &
ondary dispersal affected seedling recruitment differently in Oliveira 2013). This highlights the general importance of ant-
the three habitat types. In the forest, recruitment was not mediated seed dispersal in tropical forests. Higher secondary
influenced by secondary dispersal, whereas recruitment of sec- dispersal of arillate than of cleaned seeds is also consistent
ondarily dispersed seeds increased in the degraded habitat with other studies on non-myrmecochorous species (e.g. Pizo
close to and far from the edge (Fig. 3e and Table S4c). & Oliveira 1998; Christianini & Oliveira 2010). In our experi-
Recruitment of arillate seeds was enhanced only in seeds that ment, ants removed the aril of a very high proportion of aril-
had been secondarily dispersed (Table S4c). Dispersal dis- late seeds (95%) within the first 48 h. This is likely to be a
tance had a positive effect on seedling recruitment in all habi- benefit for the plants because aril removal reduces the risk of
tat types (Fig. 3f and Table S5c). seed mortality due to fungal infestation (Ohkawara & Akino
2005). Arils of Clusia species have high lipid contents, for
example 83% in C. criuva from Brazil (Passos & Oliveira
Discussion
2002), which is also found in many other vertebrate-dispersed
Our study reveals the importance of secondary seed dispersal diaspores (Stiles 1993). The presentation of diaspores with
for seedling recruitment in degraded tropical forests. Second- high lipid contents may thus be an evolutionary adaptation to
ary dispersal reduced the risk of predation and increased seed- both primary and secondary seed dispersers.
ling recruitment in degraded habitats, but not in the forest. The contribution of vertebrates to secondary seed dispersal,
Dispersal distances were positively associated with recruit- especially of arillate seeds, was smaller than that of ants. In
ment success in all habitats, but distance effects were greatest other neotropical ecosystems, scatter-hoarding rodents, such
in degraded habitats. This shows that effects of biotic interac- as agoutis (Dasyprocta spp.), are important agents of second-
tions on plant demography can vary across habitat boundaries ary dispersal (Cole 2009; Hirsch et al. 2012). These species,
at small spatial scales. however, prefer to cache large seeds, whereas small seeds are
usually immediately consumed (Jansen et al. 2002). The
small size of Clusia seeds could be an explanation for the
SECONDARY SEED DISPERSAL
rather limited importance of rodents as secondary seed dis-
Ants were the main secondary dispersers of Clusia trochiformis persers in our system.
(Table S1). They were mostly attracted to arillate seeds, Overall, secondary dispersal decreased from the forest to
which is corroborated by the high proportion of arillate seeds degraded habitats. A similar pattern was found comparing
that were secondarily dispersed from exclosure treatments, forest with regenerating habitats in some previous studies
compared with the small proportion attributable to vertebrates (Dominguez-Haydar & Armbrecht 2011; Zwiener, Bihn &
in open treatments. Our result of higher secondary dispersal Marques 2012; but see Zelikova & Breed 2008; Christianini
by ants than by vertebrates is consistent with other reports of & Oliveira 2013). Habitat-specific differences may be due to
© 2014 The Authors. Journal of Ecology © 2014 British Ecological Society, Journal of Ecology, 102, 659–666
664 S. C. Gallegos, I. Hensen & M. Schleuning

(a) 100 (b) 100

Seed predation (%)

Seed predation (%)

80 80

60 60

40 40

20 20

0 0
Close Far 6 10 20 30 50 100
Forest Degraded Distance of secondary dispersal (cm)

(c) 100 (d) 100

Seed germination (%)

Seed germination (%)

80 80

60 60

40 40

20 20

0 0
Close Far 6 10 20 30 50 100

Forest Degraded Distance of secondary dispersal (cm)

(e) (f)
100 100
Seedling recruitment (%)

Seedling recruitment (%)

80 80
Fig. 3. Effect of secondary dispersal (left
60 column) on (a) seed predation, (c) seed
60
germination (excluding predated seeds) and
40 40
(e) seedling recruitment (including predated
seeds); and effect of secondary dispersal
20 20 distance (right column) on (b) seed predation,
(d) seed germination and (f) seedling
0 0 recruitment. In all cases, fitted values from
Close Far 6 10 20 30 50 100 minimal adequate models are shown for
Distance of secondary dispersal (cm) seeds of Clusia trochiformis in the respective
Forest Degraded habitat type (see model details in Tables S4
Forest and S5). Models of dispersal distance only
Undispersed seeds
Degraded close to the edge include seeds that had been secondarily
Secondarily dispersed seeds
Degraded far from the edge dispersed.

differences in ant composition between habitats and a higher & Wilson 1990). For instance, some Pheidole spp. exhibited a
diversity of seed-dispersing ants in the forest (Majer 1992, dual role as seed predators and secondary dispersers in other
see Table S1). Despite these habitat differences, secondary tropical environments (Levey & Byrne 1993). Some Pheidole
dispersal by ants was also frequently recorded in the degraded ants in our system may have acted as seed dispersers and preda-
habitat in our study system. tors, although we never observed ants consuming seeds.
We found an interesting relationship between secondary dis-
persal by ants and reduced predation risk in degraded habitats,
SEED PREDATION
particularly far from the forest edge. This effect of secondary
Seed predation was higher in open treatments than in those seed dispersal on seed predation has three possible explana-
excluding vertebrates, especially for cleaned seeds, highlighting tions. First, ants move seeds under the leaf litter, making them
the importance of vertebrates as seed predators. Small mam- less likely to be detected by predators (Clark, Clark & Jacobi
mals, especially rodents, are widely recognized as seed preda- 1991). Secondly, rapid seed dispersal by ants reduces the prob-
tors, limiting seedling recruitment and abundance (e.g. Bricker, ability of seed predation shortly after seed release (Turnbull &
Pearson & Maron 2010). Accordingly, pieces of Clusia seeds Culver 1983). Thirdly, seed dispersal by ants reduces the clus-
were found in the stomach and faecal contents of several rodent tering of seeds, decreasing their probability of being encoun-
species in the study area (O. Osco & S.C. Gallegos unpubl. tered (Hulme 1994). In contrast to the effects in the degraded
data, see also Table S2). In addition to rodents, ants are also habitat far from the forest edge, there was a tendency of higher
important seed predators in various ecosystems (H€ olldobler predation of secondarily dispersed seeds in the forest. This may
© 2014 The Authors. Journal of Ecology © 2014 British Ecological Society, Journal of Ecology, 102, 659–666
 Secondary dispersal promotes reforestation 665

be due to ants exhibiting dual roles as seed dispersers and pre- experimental findings confirm this notion and indicate that
dators in the forest (Levey & Byrne 1993). Weak effects of sec- secondary dispersal increased seedling recruitment, as a result
ondary dispersal on seed predation in the degraded habitat close of increased germination and reduced predation rates. This
to the forest edge could be explained by a transitional ant com- shows that the secondary reorganization of seed shadows by
munity comprising ‘predator’ species found in the forest as well ants, although covering only small spatial scales, can be a sub-
as ‘mutualistic’ species prevalent in open habitats. tle and pivotal trigger of seedling recruitment in the short term
and probably of forest recovery in the long term. Secondary
seed dispersal is therefore a crucial although often overlooked
SEED GERMINATION AND SEEDLING RECRUITMENT
process that aids the natural regeneration of spatially hetero-
Secondary dispersal had a positive effect on seed germination geneous deforested habitats in tropical ecosystems.
in all habitat types and increased seedling recruitment in the
degraded habitats only. Ants can increase germination and
Acknowledgements
recruitment success in three possible ways: first, by removing
the aril from the seeds; secondly, by increasing distances The study was funded by German Science Foundation (DFG) and supported by
between seeds and thus by reducing seed and seedling densi- the research funding programme ‘LOEWE – Landes-Offensive zur Entwicklung
Wissenschaftlich-€okonomischer Exzellenz’ of Hesse’s Ministry of Higher Edu-
ties; and thirdly, by moving the seeds to more suitable micro- cation, Research and the Arts. We thank K. B€ohning-Gaese for her support.
habitats (directed dispersal). Aril removal has been previously We are very grateful to S.G. Beck, F.V. Saavedra and D. Lippok for their help
found to have positive effects on seed germination (Pizo & in the logistics and field work, and to M. Peters for his helpful suggestions.
We also thank L. Marconi, R. Velasco, E. Aguirre, A. Guzman, V. Baltz and
Oliveira 1998; Christianini, Mayh
e-Nunes & Oliveira 2007; J. Rojas for field assistance, J.M. Limachi for ant identification, O. Osco for
Lima, Oliveira & Silveira 2013). This pattern was consistent rodent identification and D. Bowler for language polishing in the manuscript.
with our results, but only if arillate seeds were also secondar- G. Matlack and two anonymous referees made important comments to improve
this manuscript. The authors have no conflict of interest to declare.
ily dispersed. Many studies have shown that seed dispersal by
ants usually covers short distances (e.g. G
omez & Espadaler
1998; Christianini & Oliveira 2010), especially if the ant spe- Data accessibility
cies involved are small (G
omez & Espadaler 1998), such as
Data set available from the BiK-F Data & Metadata Repository (Gallegos et al.
those in our study (most species were <1 cm). Nevertheless, 2014).
we found that seed germination and recruitment increased
with the distance of secondary dispersal. Distance effects may
be driven by reduced density-dependent mortality related to
References
pathogens and decreased competition among seeds and seed- Aide, T. & Cavelier, J. (1994) Barriers to lowland tropical forest restoration in
the Sierra Nevada de Santa Marta, Colombia. Restoration Ecology, 2, 219–
lings (Augspurger & Kelly 1984; Hammond 1995; Gallery,
229.
Moore & Dalling 2010). Directed seed dispersal by ants to Augspurger, C.K. & Kelly, C.K. (1984) Pathogen mortality of tropical tree
more suitable microhabitats has been previously demonstrated seedlings: experimental studies of the effects of dispersal distance, seedling
density, and light conditions. Oecologia, 61, 211–217.
(e.g. Passos & Oliveira 2002, 2004) and can be explained by
Barton, K. (2012) MuMIn: Multi-Model Inference. R Package Version 1.7.2.
the transport of seeds to nest sites or to sites beneath the leaf http://CRAN.R-project.org/package=MuMIn
litter. High nutrient availability and suitable microclimatic Bates, D., Maechler, M. & Bolker, B. (2011) lme4: Linear Mixed-Effects Mod-
els Using S4 Classes. R Package Version 0.999375-39. http://CRAN.R-
conditions close to ant nests and beneath the litter layer can
project.org/package=lme4
increase seedling recruitment (e.g. Culver & Beattie 1980; Bricker, M., Pearson, D. & Maron, J. (2010) Small-mammal seed predation
Pizo & Oliveira 1998; Passos & Oliveira 2004). Directed dis- limits the recruitment and abundance of two perennial grassland forbs. Ecol-
ogy, 91, 85–92.
persal is likely to be the most important mechanism for the
Briggs, J.S., Vander Wall, S.B. & Jenkins, S.H. (2009) Forest rodents provide
positive effect of secondary dispersal on seedling recruitment directed dispersal of Jeffrey pine seeds. Ecology, 90, 675–687.
in degraded habitats. Microclimatic conditions beneath the lit- Burnham, K.P. & Anderson, D.R. (2002) Model Selection and Multimodel
Inference: A Practical Information-Theoretic Approach, 2nd edn. Springer,
ter layer increased seed germination and seedling recruitment
New York.
of C. trochiformis in a simultaneous sowing experiment (S.C. Christianini, A.V., Mayh
e-Nunes, A.J. & Oliveira, P.S. (2007) The role of ants
Gallegos, I. Hensen & M. Scheluning, unpubl. data), and in the removal of non-myrmecochorous diaspores and seed germination in a
neotropical savanna. Journal of Tropical Ecology, 23, 343–351.
most of the seeds dispersed by ants were left beneath the leaf
Christianini, A.V. & Oliveira, P.S. (2010) Birds and ants provide complemen-
litter, probably increasing germination and recruitment proba- tary seed dispersal in a neotropical savanna. Journal of Ecology, 98, 573–
bilities. 582.
Christianini, A.V. & Oliveira, P.S. (2013) Edge effects decrease ant-derived
benefits to seedlings in a neotropical savanna. Arthropod-Plant Interactions,
7, 191–199.
Conclusion
Clark, B.K., Clark, B.S. & Jacobi, E.A. (1991) Ability of prairie rodents to find
seeds in plant litter. American Midland Naturalist, 126, 385–391.
Secondary seed dispersal is an important process influencing
Cole, R.J. (2009) Postdispersal seed fate of tropical montane trees in an agricul-
plant population dynamics in spatially heterogeneous environ- tural landscape, southern Costa Rica. Biotropica, 41, 319–327.
ments (Wenny 2001). Because small-scale gradients in micro- Culver, D.C. & Beattie, A.J. (1980) The fate of Viola seeds dispersed by ants.
American Journal of Botany, 67, 710–714.
climates are particularly steep in disturbed habitats, directed
Dias, A.T.C. & Scarano, F.R. (2007) Clusia as nurse plant. Clusia: A Woody
dispersal of seeds towards suitable microhabitats has the poten- Neotropical Genus of Remarkable Plasticity and Diversity (ed. U. L€uttge),
tial to increase vegetation recovery after deforestation. Our Ecological Studies Vol. 194, pp. 55–72. Springer-Verlag, Berlin, Heidelberg.

© 2014 The Authors. Journal of Ecology © 2014 British Ecological Society, Journal of Ecology, 102, 659–666
 666 S. C. Gallegos, I. Hensen & M. Schleuning

Dominguez-Haydar, Y. & Armbrecht, I. (2011) Response of ants and their seed Passos, L. & Oliveira, P.S. (2004) Interaction between ants and fruits of Guapi-
removal in rehabilitation areas and forests at El Cerrej
on coal mine in ra opposita (Nyctaginaceae) in a Brazilian sandy plain rainforest: ant effects
Colombia. Restoration Ecology, 19, 178–184. on seeds and seedlings. Oecologia, 139, 376–382.
Duncan, R.S. & Chapman, C.A. (2002) Limitations of animal seed dispersal for Pizo, M.A. & Oliveira, P.S. (1998) Interaction between ants and seeds of a non-
enhancing forest succession on degraded lands. Seed Dispersal and Frugi- myrmecochorous neotropical tree, Cabralea canjerana (Meliaceae), in the
vory: Ecology, Evolution and Conservation (eds D.J. Levey, W.R. Silva & Atlantic forest of southeast Brazil. American Journal of Botany, 85, 669–674.
M. Galetti), pp. 437–450. CAB International, Wallingford, Oxford, UK. R Development Core Team (2012) R: A Language and Environment for Statis-
Food and Agricultural Organization of the United Nations (2010) Global Forest tical Computing. R foundation for Statistical Computing, Vienna, Austria.
Resources Assessment 2010. Food and Agricultural Organization of the Uni- Reid, J.L. & Holl, K.D. (2013) Arrival 6¼ Survival. Restoration Ecology, 21,
ted Nations, Rome. 153–155.
Gallegos, S.C., Hensen, I. & Schleuning, M. (2014) Data from: Secondary dis- Schupp, E.W., Jordano, P. & G
omez, J.M. (2010) Seed dispersal effectiveness
persal by ants promotes forest regeneration after deforestation. BiK-F Data & revisited: a conceptual review. The New Phytologist, 188, 333–353.
Metadata Repository, doi:http://dx.doi.org/10.12761/SGN.2014.1. Stiles, E.W. (1993) The influence of pulp lipids on fruit preference by birds.
Gallery, R.E., Moore, D.J.P. & Dalling, J.W. (2010) Interspecific variation in Vegetatio, 107/108, 227–235.
susceptibility to fungal pathogens in seeds of 10 tree species in the neotropi- Turnbull, C.L. & Culver, D.C. (1983) The timing of seed dispersal in Viola
cal genus Cecropia. Journal of Ecology, 98, 147–155. nuttallii: attraction of dispersers and avoidance of predators. Oecologia, 59,
G
omez, C. & Espadaler, X. (1998) Seed dispersal curve of a Mediterranean 360–365.
myrmecochore: influence of ant size and the distance to nests. Ecological Vander Wall, S.B., Kuhn, K.M. & Beck, M.J. (2005) Seed removal, seed
Research, 13, 347–354. predation, and secondary dispersal. Ecology, 86, 801–806.
Hammond, D.S. (1995) Post-dispersal seed and seedling mortality of tropical Vander Wall, S.B. & Longland, W.S. (2004) Diplochory: are two seed dispers-
dry forest trees after shifting agriculture, Chiapas, Mexico. Journal of Tropi- ers better than one? Trends in Ecology & Evolution, 19, 155–161.
cal Ecology, 11, 295–313. Wenny, D.G. (1999) Two-stage dispersal of Guarea glabra and G. kunthiana (Me-
Hartig, K. & Beck, E. (2003) The bracken fern (Pteridium arachnoideum (Kaulf.) liaceae) in Monteverde, Costa Rica. Journal of Tropical Ecology, 15, 481–496.
Maxon) dilemma in the Andes of Southern Ecuador. Ecotropica, 9, 3–13. Wenny, D.G. (2001) Advantages of seed dispersal: a re-evaluation of directed
Hirsch, B.T., Kays, R., Pereira, V.E. & Jansen, P.A. (2012) Directed seed dis- dispersal. Evolutionary Ecology Research, 3, 51–74.
persal towards areas with low conspecific tree density by a scatter-hoarding Wunderle, J.M.J. (1997) The role of animal seed dispersal in accelerating native
rodent. Ecology Letters, 15, 1423–1429. forest regeneration on degraded tropical lands. Forest Ecology and Manage-
H€olldobler, B. & Wilson, E.O. (1990) The Ants. Belknap Press, Cambridge, ment, 99, 223–235.
Massachusetts. Zelikova, T.J. & Breed, M.D. (2008) Effects of habitat disturbance on ant com-
Howe, H.F. & Smallwood, J. (1982) Ecology of seed dispersal. Annual Review munity composition and seed dispersal by ants in a tropical dry forest in
of Ecology and Systematics, 13, 201–228. Costa Rica. Journal of Tropical Ecology, 24, 309–316.
Hulme, P.E. (1994) Post-dispersal seed predation in grassland: its magnitude Zwiener, V.P., Bihn, J.H. & Marques, M. (2012) Ant-diaspore interactions dur-
and sources of variation. Journal of Ecology, 82, 645–652. ing secondary succession in the Atlantic forest of Brazil. Revista de Biologıa
Jansen, P.A., Bartholomeus, M., Bongers, F., Elzinga, J.A., den Ouden, J. & Tropical, 60, 933–942.
Van Wieren, S.E. (2002) The role of seed size in dispersal by a scatter-
hoarding rodent. Seed Dispersal and Frugivorous: Ecology, Evolution and Received 26 September 2013; accepted 13 February 2014
Conservation (eds D.J. Levey, W.R. Silva & M. Galetti), pp. 209–225. CAB Handling Editor: Glenn Matlack
International, Wallingford, Oxford, UK.
Jorge, M.L.S.P. & Howe, H.F. (2009) Can forest fragmentation disrupt a condi-
tional mutualism? A case from central Amazon. Oecologia, 161, 709–718. Supporting Information
Killeen, T.J., Siles, T.M., Soria, L. & Correa, L. (2005) Estratificaci
on de vege-
taci
on y cambio de uso de suelo en los Yungas y Alto Beni de La Paz. Eco- Additional Supporting Information may be found in the online
logıa en Bolivia, 40, 32–69. version of this article:
Lamb, D., Erskine, P.D. & Parrotta, J.A. (2005) Restoration of degraded tropi-
cal forest landscapes. Science, 310, 1628–1632.
Levey, D.J. & Byrne, M.M. (1993) Complex ant-plant interactions: rain-forest Table S1. List of ant species attracted to seeds of Clusia trochiformis
ants as secondary dispersers and post-dispersal seed predators. Ecology, 74, in different habitat types and their behaviour towards seeds.
1802–1812.
Lima, M.H.C., Oliveira, E. & Silveira, F. (2013) Interactions between ants and
Table S2. List of rodent species registered in the different habitat
non-myrmecochorous fruits in Miconia (Melastomataceae) in a neotropical
savanna. Biotropica, 45, 217–223. types in the study area.
Lippok, D., Beck, S.G., Renison, D., Gallegos, S.C., Saavedra, F.V., Hensen, I.
& Schleuning, M. (2013) Forest recovery of areas deforested by fire Table S3. Results from mixed-effects models of the effects of habitat
increases with elevation in the tropical Andes. Forest Ecology and Manage- type, exclosure and aril treatments on secondary dispersal and seed
ment, 295, 69–76.
L€
uttge, U. (2007) Physiological ecology. Clusia: A Woody Neotropical Genus
predation.
of Remarkable Plasticity and Diversity (ed. U. L€uttge), pp. 187–234. Ecolog-
ical Studies Vol. 194. Springer-Verlag, Berlin, Heidelberg. Table S4. Results from mixed-effects models of the effects of sec-
Majer, J.D. (1992) Ant recolonisation of rehabilitated bauxite mines of Pocßos ondary dispersal, habitat type, exclosure and aril treatments on seed
de Caldas, Brazil. Journal of Tropical Ecology, 8, 97–108.
predation, germination and seedling recruitment.
Molina-Carpio, J. (2005) R
egimen de precipitaci
on en la cuenca de Huarinilla-
Cotapata, La Paz-Bolivia. Ecologıa en Bolivia, 40, 43–55.
Ohkawara, K. & Akino, T. (2005) Seed cleaning behavior by tropical ants and Table S5. Results from mixed-effects models of the effects of
its anti-fungal effect. Journal of Ethology, 23, 93–98. secondary dispersal distance and habitat type on seed predation, ger-
Passos, L. & Oliveira, P.S. (2002) Ants affect the distribution and performance mination and seedling recruitment.
of seedlings of Clusia criuva, a primarily bird-dispersed rain forest tree.
Journal of Ecology, 90, 517–528.

© 2014 The Authors. Journal of Ecology © 2014 British Ecological Society, Journal of Ecology, 102, 659–666

View publication stats