INTRODUCTION
In tropical forests, tree-falls, landslides and subsistence agriculture, among others, regularly produce small gaps that offer plants new habitats for colonization and create gradients in light (Chazdon Reference CHAZDON1986, Denslow Reference DENSLOW1987, Grubb Reference GRUBB1977). Regeneration studies in forest clearings, fallows or tree-fall gaps usually consider trees (Schnitzer et al. Reference SCHNITZER, MASCARO, CARSON, Carson and Schnitzer2008) and ignore the contribution of vegetative recruitment (Lasso et al. Reference LASSO, ENGELBRECHT and DALLING2009). Plants may colonize such open habitats through germinating seeds, which are present in a seed bank or arrive by dispersal (Dalling et al. Reference DALLING, SWAINE and GARWOOD1998). Colonization may also take place through vegetative recruitment by sprouting roots or stem fragments that survive gap formation (Schnitzer & Carson Reference SCHNITZER and CARSON2001) or creeping plants that laterally invade from the adjacent forest (Schnitzer et al. Reference SCHNITZER, MASCARO, CARSON, Carson and Schnitzer2008). Creeping recruits are often the first to colonize tree-fall gaps, leading to canopy closure within a few months (Clark et al. Reference CLARK, PALMER and CLARK1999, Schnitzer et al. Reference SCHNITZER, DALLING and CARSON2000). Many such creeping plants are hemiepiphytes i.e. ground-rooted plants that spend a part of their life cycle as ground-detached epiphytes. In the neotropics, these hemiepiphytes often belong to Araceae (aroids), one of the world's most species-rich and abundant epiphytic families (Bazzaz & Pickett Reference BAZZAZ and PICKETT1980, Benavides et al. Reference BENAVIDES, WOLF and DUIVENVOORDEN2006, Reference BENAVIDES, VASCO, DUQUE and DUIVENVOORDEN2011; Croat Reference CROAT1992, Gentry & Dodson Reference GENTRY and DODSON1987, Nieder et al. Reference NIEDER, ENGWALD, KLAWUN and BARTHLOTT2000).
Young successional forests in fallows (Benavides et al. Reference BENAVIDES, WOLF and DUIVENVOORDEN2006) or floodplains (Leimbeck & Balslev Reference LEIMBECK and BALSLEV2001) in the Amazon basin may already contain a large number of species and individuals of hemiepiphytic aroids, especially in comparison to holoepiphytes. The rapid colonization by hemiepiphytic aroids might be related to particular functional traits. In many aroids, creeping stems and descending shoots exhibit abrupt morphological changes by becoming flagellar (Ray Reference RAY1992). Flagellar shoots (showing internodes that are c. 35 times as long as wide) may have rapid growth and are seen not only as a mechanism to forage for essential resources, but also to multiply and disperse meristems clonally into newly available habitats (Andrade & Mayo Reference ANDRADE and MAYO2000, Balcázar-Vargas et al. Reference BALCÁZAR-VARGAS, PEÑUELA-MORA, VAN ANDEL and ZUIDEMA2012, Ray Reference RAY1992).
Our study addressed the hypothesis that hemiepiphytic aroids are well adapted to quickly occupy small, open habitats in forests. We hypothesized that this quick aroid colonization (Benavides et al. Reference BENAVIDES, WOLF and DUIVENVOORDEN2006, Leimbeck & Balslev Reference LEIMBECK and BALSLEV2001) is mainly by seeds, vegetative reproduction through sprouting fragments of fallen aroids, or invasion through lateral growth of flagellar aroids from the closed forest nearby. Our specific hypotheses were as follows: (1) Abundant light in open habitats enhances aroid seed germination. (2) Aroids reproduce vegetatively by sprouting fragments. (3) In open habitats flagellar aroids are more abundant than non-flagellar aroids whereas in closed forests this difference is less pronounced. (4) Aroids, and especially flagellar aroids, grow faster in open habitats than in closed forests. (5) They also show faster growth in forest edges when occurring closer to open habitats. (6) Finally, they show directional growth towards the open habitat in such conditions.
METHODS
Study area
We undertook fieldwork from May 2008 to September 2009 in Amacayacu National Park and adjacent Ticuna territory, located in Colombian Amazonia (3°S, 69°–70°W). The area has a bimodal annual precipitation of c. 3200 mm, with a relatively dry period between June and September. The average annual temperature is c. 26 °C and the daily average for relative humidity is c. 86% (Rudas & Prieto Reference RUDAS and PRIETO2005). The area is covered with tropical rain forest (Rudas & Prieto Reference RUDAS and PRIETO2005). In total, 58 epiphytic aroid species were reported in the area (Benavides et al. Reference BENAVIDES, WOLF and DUIVENVOORDEN2006). Local communities use shifting cultivation for subsistence agriculture. Man-made clearings in the forest to cultivate crops are up to 2 ha in size, generally square or rectangular in shape, and are used for polyculture plantations of crops such as cassava and banana. After approximately 6 y, the clearings are abandoned and the forest regenerates (fallows). In this paper, we employ the term ‘open habitat’ as a generic term that includes recent clearings, fallows and natural tree-fall gaps, with clarification where necessary.
Fieldwork
Seed germination was studied for three aroid species that were known to fruit abundantly in the region (Philodendron ernestii, P. wittianum and P. fragantissimum). We sowed seeds out of a pooled batch that were collected from at least ten mature fruits of eight individual plants per species. P. ernestii and P. wittianum were sown in November 2008 and P. fragrantissimum was sown 2 mo later. Ten seeds were sown in every cell of a 32-cell tray. Half of the cells were filled with a mix of soil samples taken at 0–10 cm depth (after removal of the litter layer) in four mature-forest sites, and the other 16 cells with a similar mix of soil samples from three fallows. Two trays were placed on a table in a clearing under each of three light-exposure regimes (for Philodendron ernestii only one tray per light regime was used). Using shade cloths that blocked 25%, 55% and 85% of sunlight we varied light exposure. While maintaining the light exposure regime, the position of the trays on the table was changed randomly each week. The germination of all three species ceased after 2 mo. The experiment was conducted in the village of Mocagua (03°48′S, 70°15′W).
The sprouting capability was studied by placing stem cuttings c. 30 cm long (middle portion of stems with three nodes) from at least 10 adult plants of five aroid species (80 stem cuttings for each species) in four recent (< 1 y old) clearings under controlled conditions. After placement, sprouting of each cutting was recorded after 6 and 14 wk.
We monitored the recruitment and apical growth of secondary hemiepiphytic aroids during a period of 14 mo (between July 2008 and September 2009). This monitoring was done along 2-m-wide transects of variable length that covered three habitats: open habitats, comprising recent forest clearings and fallows less than 2 y old; forest edges at either side of these open habitats; and mature forests at least 200 m away from any clearing or fallow. A total of 14 transects were located through the centre of randomly picked clearings or fallows, and continued 15 to 35 m into the forest edge at either side of these. Eight additional transects of 25 m length were laid out in mature forests. In total the initial area for monitoring comprised 1760 m2 in open habitats, 1150 m2 in forest edges and 400 m2 in mature forests. Fire destroyed one transect (covering open habitat and adjacent forest edge) after 2 mo. At each transect, we mapped all aroids that were known to have a hemiepiphytic life form, and marked these with plastic tags or aluminium wire, regardless of whether or not they were attached to trees. Since hemiepiphytic aroids do not attain great height on the tree, binocular observations sufficed and no tree climbing was required. For each aroid, we recorded species name and whether the shoots were flagellar (Ray Reference RAY1992). After initial mapping, we returned every 2 mo to record recruits and their recruitment mode (vegetative or from seed) and to measure the apical growth of a randomly selected subset of approximately 25 aroid individuals, irrespective of species. This growth was measured by carefully indicating the position of the apical meristems using small cuts on the tree bark or plastic tags. We estimated the percentage of diffuse light (diffuse site factor; Anderson Reference ANDERSON1964) for each of these aroids by taking hemispherical photographs above the centre of the plant (Roxburgh & Kelly Reference ROXBURGH and KELLY1995). We took the photographs early in the morning or late in the afternoon on overcast days; these were underexposed by one stop (Hale & Edwards Reference HALE and EDWARDS2002), processed using Adobe® Photoshop® CS2 software (version 9) and analysed with the Gap Light Analyzer software program (Frazer et al. Reference FRAZER, CANHAM and LERTZMAN1999).
For the directional growth study we laid out 21 transects (others than used for the recruitment and growth study) in forest adjacent to recent clearings (11 transects) and gaps caused by the recent fall of a large tree (10 transects). Tree-fall gaps were located along a 5-km path in the mature forest, and were typically about 25 × 10 m in size. Because the falling tree had destroyed most trees and shrubs, the gap-forest transition was clear and generally occurred over a distance of 2 m or less. All transects extended 20 m into the forest, in a direction perpendicular to the transition of the forest with the clearing or the tree-fall gap. Along each transect, we randomly established at least five points. At each point, we selected the nearest hemiepiphytic aroid in each of the four quadrants, following the point-centred-quarter method (Cottam & Curtis Reference COTTAM and CURTIS1956, Engeman et al. Reference ENGEMAN, SUGIHARA, PANK and DUSENBERRY1994). For vegetatively spreading aroids, each stem was regarded as a separate individual (ramet). The line between the apex and the last two internodes defined the directional growth of the youngest internodes (Sampaio et al. Reference SAMPAIO, ARAUJO, SCARANO and STUEFER2004).
Numerical analyses
Following protocols of Zuur et al. (Reference ZUUR, IENO, WALKER, SAVALIEV and SMITH2009) to simplify the models regarding random effects, we tested our hypotheses 1 and 3 to 6 applying generalized least squares (gls) and (generalized) mixed-effects models (lme, lmer).
Hypothesis 1: for each species, we modelled the arcsine-transformed (Zar Reference ZAR1999) proportions of seeds that had germinated after 2 mo against shade (three shade classes) and soil (two classes: mature forest soil and fallow soil) using the following model:
model1 <– gls(shade+soil+shade:soil, weights = varComb(varIdent(form = ~1|soil), varIdent(form = ~1|shade)))
where weights describes the heteroscedasticity structure within shade and soil classes (Pinheiro & Bates Reference PINHEIRO and BATES2000).
Hypothesis 3: we regressed the binary response flagel (flagellar shoots present or not) against habitat (in three classes: open, forest edge and mature forest), with diffuse light levels as random slope and species nested in transect as random intercept, as follows:
model3 <– lmer(flagel ~ habitat + (diffuselight | transect/species), family = binomial)
We tested for the significance of the habitat effect by comparing this model with one from which habitat was removed, using maximum likelihood and a chi-squared test.
Hypothesis 4: the apical growth rate (r) was calculated as:
r = (L2–L1)/(T2–T1)
in which L1 and L2 were the stem length at the start (T1) and the end (T2) of the measurement period (Chao et al. Reference CHAO, PHILLIPS, GLOOR, MONTEAGUDO, TORRES-LEZAMA and VÁSQUEZ2008, Hoffman & Poorter Reference HOFFMAN and POORTER2002, Rüger et al. Reference RÜGER, BERGER, HUBBELL, VIEILLEDENT and CONDIT2011). A logarithmic transformation of r was applied to increase the symmetry of its distribution. Logarithmic growth rate was regressed against the binary factor flagel (flagellar shoots present or not), habitat (in the three classes as indicated above) and the interaction between these two variables. Initially, diffuse light levels were included as random slope and species (nested in transect) as random intercept, while habitat was added as variance covariate. During model selection, diffuse light levels were removed, resulting into:
model4 <– lme(log(r + 0.1) ~ flagel + habitat + flagel:habitat, random = (~1 | transect/species), weights = varIdent(form = ~1|habitat))
where weights describes the heteroscedasticity structure within habitat classes (Pinheiro & Bates Reference PINHEIRO and BATES2000).
Hypothesis 5: here, the logarithmic growth rate was regressed against the binary factor flagel (flagellar shoots present or not), distance of the initial position of the aroid's apex towards the border with the open habitat, and the interaction between these two variables. Initially, diffuse light levels were included as random slope and species nested in transect as random intercept. During model selection, diffuse light levels were removed, yielding the final model as follows:
model5 <– lme(log(r + 0.1) ~ flagel + distance + flagel:distance, random = (~1 | transect/species))
Hypothesis 6: we first calculated for each aroid the cosine of the angle between the line of directional growth and the transect direction. These cosine values were regressed against the aroid's distance (with precision of 0.5 m) to the border with the clearing or tree-fall gap, the binary factor flagel (flagellar shoots present or not) and the interaction between these two variables. The aroid's distance was also included as random slope and the transect was inserted as random intercept. Initial models, which included species as random effect (nested in transect), broke down because most species occurred in few transects. Therefore the final model only included data from Philodendron ernestii. It was as follows:
model6 <– lme(cosine ~ flagel + distance + flagel:distance, random = (~1 | distance/transect))
Regarding models 3 to 6, spatial autocorrelation of the residuals within transects was examined using semi-variograms and spline correlograms (Bjørnstad & Falck Reference BJØRNSTAD and FALCK2001). In none of the models was it detected. All analyses were conducted using R 2.14.1, applying nlme (version 3.1–103), lme4 (version 0.999375–42), ncf (version 1.1–4) and lattice (version 0.20–10; Sarkar Reference SARKAR2008) packages.
RESULTS
The germination experiment showed that all three aroid species germinated most successfully under conditions of low light (Figure 1, Table 1). The shading effect was highly significant for all three species. Seeds of P. ernestii germinated slightly better on mature-forest soils. No significant soil-shading interaction effects were found.
Table 1. F-values of the fixed effects in the generalized least squares model to test if light (three shade levels) or soil (two soil classes) affected the seed germination of three Philodendron species in Mocagua, Colombian Amazonia. df = degrees of freedom; the df of the denominator was 90 for P. ernestii, and 186 for the other two species * = 0.01 ≤ P < 0.05; ** = 0.001 ≤ P < 0.01; *** = P < 0.001.
Figure 1. Box-and-whisker plots to show the proportion of seeds of three Philodendron species that germinated after 2 mo in response to three levels of shading and two classes of soil substrate, in an experiment carried out in the village of Mocagua, Colombian Amazonia. The whiskers extend from the quartiles up to 1.5 times the interquartile range.
The cuttings from all aroid species sprouted within 6 wk. On average 7.5% of the cuttings sprouted after 6 wk, and 11.8% after 14 wk (3.8%, 8.8%, 8.8%, 7.5% and 8.8% after 6 wk, and 3.8%, 13.8%, 18.8%, 10% and 12.5% for Philodendron elaphoglossoides, P. ernestii, P. fragrantissimum, P. heleniae and P. wittianum, respectively).
Along the monitoring transects, we recorded a total of 2133 aroid plants, represented by 41 species and six genera (see Appendix 1). Philodendron ernestii dominated the aroid community with 565 records, which was more than twice the total abundance of the second most dominant species (Anthurium digitatum with 252 records). In total 25 species were found in clearings and fallows, 35 in forest edges and 32 in mature forests. Just over half (24) of the 41 species were capable of producing flagellar shoots. On average, the initial density of plants with non-flagellar shoots was five to six times higher than that of flagellar aroids (Table 2). The comparison of the mixed models with and without habitat as fixed effect on the presence-absence of flagellar aroids did not yield a significant difference (χ2 = 2.2; P = 0.33). Hence, evidence that the probability to record a flagellar aroid depended on habitat could not be detected.
Table 2. Hemiepiphytic aroid density with and without flagellar shoots (number of plants per 10 m2, mean ± SD) in three habitats in the forests near Amacayacu, Colombia. N = number of transects.
In the 14-mo monitoring period, aroid recruitment into clearings and fallows was surprisingly low (mean of 0.3 plants per 10 m2; Table 3), especially in comparison to the initial density of aroid plants (mean of 3.1 plants per 10 m2; Table 2). A similar number of open-habitat recruits originated by seed dispersal or through vegetative reproduction (Table 4). Most of these recruits belonged to Philodendron ernestii.
Table 3. Recruitment of hemiepiphytic aroids after 14 mo (number of plants per 10 m2, mean ± SD) in three habitats in the forests near Amacayacu, Colombia. N = number of transects.
Table 4. Number of hemiepiphytic aroids that recruited during a period of 14 mo in a total area of 1760 m2 in open habitats (recent clearings and fallows < 2 y old) near Amacayacu, Colombia. AMB codes refer to collection numbers of Ana María Benavides.
We used 632 aroids to monitor apical growth (see Appendix 1). These plants belonged to 33 species; Philodendron ernestii was most abundant (161 stems). After 14 mo, the maximum stem growth was 888 cm. Over a 2-mo period, the fastest aroid growth was 106 cm mo−1 in open habitats, 97 cm mo−1 in forest edges, and 98 cm mo−1 in mature forests. A total of 53 stems (8.4% of the stems, of which 27 were flagellar) grew at a rate > 2 m per 14 mo, and 27 stems (of which 12 were flagellar) grew at a rate > 3 m per 14 mo. At the start of the growth measurements, the stem length of the aroids was 117 ± 178 cm in open habitats, 182 ± 305 cm in forest edges and 141 ± 341 cm in mature forests (mean ± SD). The maximum stem length was 32.7 m.
Growth rate was fastest in open habitats and slowest in mature forests (Tables 5, 6). Also, flagellar aroids grew faster than non-flagellar aroids (Tables 5, 6), but the interaction effect between habitat and the binary factor flagel was not significant. Restricting the analyses to edges, again the apical growth of flagellar aroids was faster than that of non-flagellar aroids but it did not depend on distance to the open habitat (Table 6).
Table 5. Apical growth of hemiepiphytic aroids (cm mo−1, mean ± SD (N)), measured after 14 mo in three habitats in the forests near Amacayacu, Colombia. N = number of aroid plants.
Table 6. F-values of the fixed effects in the (generalized) mixed-effects models to test three hypotheses of aroid growth in open habitats and adjacent forests near Amacayacu, Colombian Amazonia, as follows: if aroids, and especially flagellar aroids, grow faster in open habitats than in closed forests (model4); if they show faster growth in forest edges when occurring closer to open habitats (model5); and if they show directional growth towards the open habitat in such conditions (model6); df = degrees of freedom; * = 0.01 ≤ P < 0.05; ** = 0.001 ≤ P < 0.01; *** = P < 0.001.
The mixed-effects modelling of directional growth was conducted on the basis of 119 plants of Philodendron ernestii, which occurred in 21 transects. In the forest edges, the distance towards the open habitat or the presence-absence of flagellar shoots did not show a detectable influence on the direction of apical growth (Table 6).
DISCUSSION
The clearings and fallows along the transects that were established to monitor aroid recruitment contained almost half of all aroid species in the area (Benavides et al. Reference BENAVIDES, WOLF and DUIVENVOORDEN2006). Nevertheless, aroid recruitment into these open habitats was low (only 10% of the initial density after 14 mo). Also, a substantial number of aroids were already present in the clearing or fallows before we started our monitoring. Whether these aroids originated from seed or from surviving adults, or fragments thereof, remains elusive; we were unable to document the recruitment mode at the initial creation of the clearings in situ. However, as all three aroid species that we tested exhibited a poor germination under high-light conditions, we speculate that a substantial part of the early aroid gap occupants originated from aroid stem fragments. Indeed, the sprouting experiment showed that, on average, more than 10% of aroid stem cuttings sprouted after 14 wk. Despite the low recruitment (in 14 mo), our results further indicated that a similar number of aroids recruited open habitats vegetatively (sprouting) or through seed dispersal. Other studies showed that fragment regeneration might even prevail over seed recruitment (Lasso et al. Reference LASSO, ENGELBRECHT and DALLING2009). Fragments may have a similar or higher survival than seedlings and higher growth rates than saplings (Dietze & Clark Reference DIETZE and CLARK2008, Khan et al. Reference KHAN, RAI and TRIPATHI1986). Our recruitment monitoring took place in man-made clearings and fallows where slash-and-burn practices and weeding may have removed previously established aroid seedlings and adult aroid stem fragments. In natural tree-fall gaps, recruitment and survival of aroid seedlings and sprouted fragments might be substantially higher.
For aroids, analogous to the lateral spread of lianas (Peñalosa Reference PEÑALOSA1984, Putz Reference PUTZ1984), flagellar shoots might facilitate the lateral invasion of open space, such as man-made forest clearings or tree-fall gaps (Andrade & Mayo Reference ANDRADE and MAYO1998, Reference ANDRADE and MAYO2000). Indeed, the apical growth of flagellar aroids in forest edges (mean of 98 cm per 14 mo) would permit such plants to reach a clearing or tree-fall gap within a few years. We found 4.2% of the aroid stems showing a 14-mo growth of ≥ 3 m. Up to a growth of 12 m y−1 seems achievable if the growth rates recorded over 2 mo were to be sustained for 1 y. However, our transect studies failed to supply other evidence to support the lateral-invasion hypothesis. Generally, flagellar aroids were less common than non-flagellar aroids, and in open habitats the chance to encounter an aroid with flagellar shoots was just as low as in forest habitats. No proof was found that the growth rate of flagellar aroids relative to that of non-flagellar aroids depended on habitat, nor did we find any effect of distance towards the open habitat on growth or growth direction of aroids in forest edges. Flagellar aroids grew faster than non-flagellar aroids and growth of aroids in general was faster in open habitats (presumably stimulated by higher light availability). Both these phenomena may well contribute to the fast aroid colonization in forest clearings, but their detection does not allow conclusions about the arrival mechanism.
In conclusion, the colonization of open habitats such as clearings and tree-fall gaps in Amazonian rain forest by hemiepiphytic aroids appears to be driven by a high post-disturbance survival of adult plants, or fragments thereof. New recruitment, especially from flagellar aroids may also occur. Further research is necessary to assess what triggers the formation of flagellar shoots in aroids, and to assess the survival and sprouting capacity of adult aroids directly after tree-fall events and clearings of the forest.
ACKNOWLEDGEMENTS
We thank Amacayacu National Park and the communities of Mocagua and Palmeras. We also thank A. C. Kuchta, Y. Campiño, L. V. Londoño and C. J. Benavides for help with fieldwork, and J. van Arkel and F. van Dunné for the hemispherical photograph processing. A. M. Benavides received subsidies from the Schure-Beijerinck-Popping fund, the Alberta Mennega fund, the AUV fund and Programme Alban. J. Wolf was supported by The Stichting Het Kronendak.
Appendix 1. Hemiepiphytic aroid species found in open habitats (recent forest clearings and fallows <2 y old), forest edges adjacent to these open habitats, and mature upland forests at least 200 m away from these open habitats, in Amacayacu National Park, Colombian Amazonia. The values represent numbers of individuals. AMB codes refer to collection numbers of Ana María Benavides.
