Palms, peccaries and perturbations: widespread effects of small-scale disturbance in tropical forests

Spatial patterns of distribution and density

In initial neighbourhood analyses using a bivariate form of the O-ring statistic with a homogeneous Poisson null model we found a significant large-scale effect, indicating that large palms are located in areas of somewhat lower sapling density (data not shown). When we corrected for this effect using the heterogeneous Poisson null model, randomly displacing palms within a 30 m neighbourhood, we found, as expected, notable small-scale repulsion of saplings from mature palms at distances < 3 m (Figure 1A). Considering immature palms, the large scale effect with respect to saplings existed as a tendency but was not significant and there were no significant small-scale effects (Figure 1B). Saplings were neither repulsed from or attracted to immature palm trees. When we extended these analyses to other fruiting trees, contrary to our predictions, there was no significant repulsion or attraction of saplings from neither male nor female Myristicaceae trees and the observed O-ring values remained within the simulation envelopes (Figure 1C and 1D).

Spatial patterns of community structure

A total of 1,471 5 × 5 m quadrats contained or were adjacent to a quadrat containing a mature palm ('palm quadrats') and the remaining 8,301 quadrats were defined as uninfluenced by mature palms ('non-palm quadrats', Figure 2). Both stem density and species richness were lower in palm quadrats, by about 0.5 of a stem/species. When we accounted for stem density, rarefied species richness was equivalent between palm and non-palm quadrats, indicating that the lower observed species riches in palm quadrats was primarily caused by a fewer individuals.

These results were confirmed in ISAR neighbourhood analyses, were we found a significant repelling effect of large arborescent mature palms on surrounding species at distances < 5 m (Figure 3A). This result implies that mature palms have a limited direct negative effect on the diversity of surrounding sapling species. In contrast, immature arborescent palms had no such effect on surrounding species diversity, and the observed ISAR values for the null model generally remained within the simulation envelopes, indicating a significant but very small repelling effect at only one distance category (3 m, Figure 3B).

Dynamic demographic effects of large trees on seedlings

As expected, one-yr survival of new seedlings near (< 10 m) to mature palms was significantly lower than survival of seedlings far (> 10 m) from mature and immature palms (Figure 4). Survival of new seedlings near to immature palms was also significantly lower, but only when also near mature palms. However, unexpectedly, survival of new seedlings near to male and far from female Myristicaceae trees was significantly lower than survival of seedlings far from from sexes (Figure 4), survival of new seedlings near female trees was not significantly lower (although in these cases our analyses suffer from small sample sizes).

Dynamic demographic effects of large trees on saplings

There were some significant effects of palm and Myristicaceae trees on sapling dynamics. Contrary to our predictions, saplings showed no significant difference in the probability of survival in areas near to (< 10 m) or far from (≥ 10 m) large arborescent palms, mature or not (Figure 5A, blue circles), and the probability of negative growth of saplings near to immature palms was greater than that of saplings far from either mature or immature palms (Figure 5B, blue circles). In accordance with our predictions, resprouting rates were higher in saplings near to palm trees compared to areas far from both immature and mature palms (Figure 5C, blue circles).

When we looked at the effects of mature Myristicaceae trees on sapling dynamics, we found fewer significant effects. Sapling survival near to male but far from female Myrsticaceae trees was significantly higher compared to saplings far from both sexes (Figure 5A, red squares). However, these same saplings also had a significantly higher probability of negative growth (Figure 5B, red squares). There was no effect of Myristicaceae trees on resprouting rates (Figure 5C).

Implications for forest diversity and community structure

Given the difficulties of reliably measuring seedling growth, it was not possible to repeat all three tests of dynamic data in both seedlings and saplings. Growth and resprouting were analysed solely in saplings. However, it seems that significant mortality of seedlings occurs in the vicinity of palm trees, but that saplings can generally withstand physical damage caused by peccaries and/or palm fronds, and rather suffer reduced growth and can potentially resprout. This difference in mortality implies that different mechanisms are at work at different life history stages and that if seedlings can grow large enough, their fate is more secure [42]. The differences in spatial structure and sapling density around palms are therefore more likely to be formed at the seedling stage than the sapling stage. The small-scale repulsion of saplings (Figure 1B) may be a consequence of the immediate negative effect of large palms on seedlings (falling palm fronds cause increased mortality and stem breakage in neighboring saplings) and the large-scale effects (Figure 1A) could be a long-term cumulative population dynamics effect that results from many generations of palms having a negative impact on nearby seedlings.

The impacts of palms on the forest structure are of smaller spatial scale than that hypthesized by Peters et al. [15], who suggested that the effect of palms extended out to 7 m and that at least a quarter of lowland forest could be impacted by falling palms fronds per year. We found significant negative effects of palm trees out to only 3 m (in terms of over-dispersion of saplings [spatial pattern]) and 5 m (in terms of species repelled [ISAR]). Given the spatial pattern and density of palm trees in our plot, the area of forest predicted to be directly affected by this process is therefore reduced to 17%, although we noted above that these effects could multiply over generations leading to large areas of lower sapling density.

The negative impacts identified above appear limited to large palms and do not involve other fruiting trees such as the Myristicaceae. We would conclude therefore that this process is not a disturbance that affects large swathes of forest (or everywhere there is a tree that produces large peccary-consumed fruit). If the effect is limited to the environment of palms, this creates heterogeneity in the forest, which may lead to a wide variety of regeneration niches for species that can withstand the effects of falling leaves, fruit and rooting frugivores [43]. Once established, however, these individuals may have better access to light and soil resources, given the diffuse nature of palm canopies and the large amounts of decaying fruit and animal faeces that likely accumulate around the base of palm trees. The stilt-roots of many arborescent palms probably increase nutrient availability by trapping leaves and organic matter.

This kind of widespread small-scale disturbance creates heterogeneity in regeneration niches in both space and time. Fruiting and frugivory by peccaries is episodic, and palms generally mature one infructescence at a time. Furthermore, in the aseasonal forest of Yasuní, Iriartea fruit production is not confined to a few months of the year (Nancy Garwood, unpublished data). Confoundingly, palm fronds are also produced regularly throughout the year. Beck et al. [36] found no difference in leaf-fall rates between two widely separated forests in Peru. In seasonal sites such as Yasuní, therefore, where seedling germination takes place throughout the year, some individuals may escape damage simply by growing enough before the next frond or infructescence falls. In more seasonal forests where germination is often restricted to short periods, individuals are of similar ages and/or vulnerabilities and so may suffer greater impact. A further test of our hypotheses could take place in seasonal forest and examine if seedlings and saplings suffer greater mortality during the fruiting season (therefore more likely caused by peccaries) or during the non-fruiting season (more likely leaf-fall).

The ubiquitous nature of the palm family in Neotropical forests implies that given some effect, palm trees could be a powerful force structuring these forests [15]. Several species of tree dominate large areas of lowland forests in western Amazonia, and arborescent palms such as Iriartea are among this 'oligarch' group [24, 26, 28]. Widespread small-scale mosaics created by these palms may affect the composition of advanced regeneration for when a larger canopy gap does open. However, we note that peccaries are among the first large mammals to be hunted out of forest near to human habitation. Hunting will likely have dramatic effects on forest structure [32].

If reduced survival and growth around palms is predictable and selective, what can we hypothesize about species that occur in the neighbourhood of palms? Species that are better able to survive damage by falling fronds or rooting peccaries would be more likely to survive in areas close to palm trees. From our community-level analyses we cannot identify traits of individual species. However, species with underground storage organs or cotyledons might find an advantage in germinating near to a large palm because of their increased resource availability allowing resprouting following damage [15]. We identified a number of species that had significantly higher or lower abundance around palms - identification of key functional traits of these species implicated by this pattern would be useful. An advantage might even apply to the seedlings of arborescent palm trees themselves, if they can survive predation and falling fronds. Iriartea is one of the most widespread and abundant tree species in western Amazonia. If Iriartea seedlings can survive better in the adult zone of influence than the majority of non-palm species, this may go some way to explaining how such abundant populations can be maintained. Future work should therefore look for more evidence of selection potential of the process that Peters et al. [15], Beck et al. [36] and we have identified. What species are selected for? Does species diversity change predictably from seedlings to saplings? Are there species that appear more frequently together in proximity to palms than in other parts of the forest? Furthermore, a broader-scale comparison with other forests may be helpful. For example, the forests of SE Asia are not dominated by palms as are Neotropical ones, and yet peccaries have a significant impact on seedling and sapling communities [31].

Palms themselves are not immune to the effects of branch falls [44] and suffer negative density dependence [45], caused by falling fronds [22, 46–48], peccaries [32, 34], and pathogenic fungi [49]. Although palm trees significantly impact non-palms, the combination of falling fronds and peccary seed predation and fungal attack inhibit population growth such that palms may currently be at maximum density in the forest.

Tree and sapling data

A 25-ha permanent forest dynamics plot (FDP; http://www.ctfs.si.edu) is located inside Yasuní National Park (0°41S, 76°24W; [50]), a still largely wilderness, tropical lowland aseasonal rain forest in eastern Ecuador [51]. Mean annual rainfall is approximately 2800 mm and total monthly rainfall is almost never ≤ 100 mm. Mean monthly temperature is 25-27°C [50]. The FDP ranges from 216 to 248 m a.s.l: it includes two ridges and an intervening valley that floods for brief periods. From 1995-1999, all freestanding stems ≥ 1 cm diameter at breast height (dbh at 1.3 m), excluding lianas, were tagged, mapped and identified [50]. In the first census, 1104 morph species were recorded, comprising 151,230 individual trees [50]. The FDP was recensused in 2002-2003 [52].

There are four species of arborescent palm in the Yasuní FDP: Astrocaryum chambira Burret, Iriartea deltoidea Ruiz & Pav., Oenocarpus bataua Mart., and Socratea exorrhiza (Mart.) H. Wendl. The majority of individuals are Iriartea (Table 1). Although the relationship between stem diameter and height in palms is not as robust as that in dicotyledon trees, Iriartea generally reaches maturity around 15 cm dbh [53]. Because the height of palms is not measured on the FDP, we took this diameter as the cut-off between mature and immature large palms. In order to examine the interaction between frugivory and leaf-falls, we compared mature fruiting palms (> 15 cm dbh) to large palms without fruit, being those 10-15 cm dbh.

The size at which tropical trees mature is difficult to examine and often unknown for many species and can depend on a number of factors such as light availability, nutrient and water resources. However, the Myristicaceae at Yasuní have been studied intensively for the previous 10 years, and trees have been visited multiple times to look for signs of reproduction. The family is dioecious and all reproductive (ie. mature) male and female trees have been identified and sexed from flowers and/or fruit [54–58]). The environment below a male and female Myristicaceae adult should be qualitatively similar, allowing the separation of the effect of fruit crop and associated frugivores from the effects of light availability or habitat quality on the performance of seedlings and saplings.

Seedling data

Within the 25-ha FDP, 339 1 m² seedling plots were established along trails in 2002 [59]. They were arranged around seed traps in sets of three as part of a long-term community phenology study. Survival, growth and recruitment are monitored every year and we use data from 2003 to 2008.

Spatial patterns of distribution and density

In order to examine patterns of attraction and repulsion between two sets of spatially-explicit points, we used an extension of the O-ring statistic for bivariate patterns [61] to test, for example, the density of saplings (1-2 cm dbh) at various distances r around large palms (> 15 cm dbh). More formally, the O-ring statistic O ₁₂ (r) for a bivariate pattern composed of type 1 and type 2 points yields the average density of type 2 point within rings of radius r and width dr around the points of pattern 1. The O ₁₂ (r) is also called neighbourhood density function and is closely related to the pair correlation function g ₁₂ (r), i.e., O ₁₂ (r) = λ 2g₁₂(r)[61, 62] where λ2 is the average density of type 2 points in the plot.

To find out if the distribution of saplings in the neighbourhood of palms or Myristicaceae trees is different from a distribution expected by chance for each of these analyses we initially tested two different null models of spatial distribution. The first null model retained the spatial distribution of saplings and assumed a random distribution of palms (i.e., a homogenous Poisson null model [61]). This null model therefore compares the observed neighbourhood density of saplings around palms with sapling densities in random locations. If the neighbourhood density of saplings around focal trees is smaller (or larger) than expected for randomly relocated focal trees we can infer a negative (or positive) impact of focal trees on saplings. However, we could observe the same pattern if (larger-scale) environmental heterogeneity acted such that the focal trees were located in areas of higher or lower overall sapling density. We therefore used a second null model that approximately factors out such larger-scale environmental variability and reveals the "pure" smaller-scale effects of focal trees on saplings (e.g. [63–65]). The second test also assumed a random distribution of palms, but the palms were only displaced within a given neighborhood of 30 m. The neighborhood of 30 m is somewhat larger than the expected range of direct interactions between focal trees (i.e., palms and adult Myristicaceae trees) and saplings (< 10 m based on ePeters2004). Technically, this null model was implemented as a heterogeneous Poisson process with a non-parametric kernel estimate of the intensity function λ1(x) of the focal trees (for details see [64]).

We made these comparisons for (i): non-palm sapling distributions versus mature and immature palms and (ii) non-palm sapling distributions versus male and female Myristicaceae. All analyses were run for 199 simulations and 95% simulation envelopes were calculated. These analyses were conducted using the Programita software package [61].

Spatial patterns of community structure

While the previous analysis focused on overall sapling density we also carried out tests where we compared the species diversity of sapling communities that were influenced by palms to sapling communities that were not. For the first two tests we compared sapling (1-2 cm dbh) communities in 5 × 5 m quadrats that either contained ≥ 1 palm or were adjacent to a quadrat containing a palm to quadrats that were > 1 quadrat distant from a palm-quadrat. This accounted for the up to 10 m distance over which palm fronds fall. In the third test, we examined the individual species-area relationships between palms and saplings.

First, we tested whether sapling density, species richness and rarefied species richness per two individuals were lower in quadrats influenced by palms, using the 'vegan' R package [66]. This provided a basic description of the community spatial structure with minimal consideration of species identity.

Second, we compared the community composition of saplings in quadrats using a multiple response permutation procedure (adonis), in 'vegan' [66]. This method is analogous to a multivariate analysis of variance and provides a robust test of whether there is a significant difference between the species composition of palm and non-palm quadrats. Palm and non-palm quadrats may have equivalent species richness yet these species need not be identical. We also compared the abundance of every species in the plot in palm and non-palm quadrats using a proportion test, with the expected relative abundance set to 85% of individuals in non-palm quadrats (the proportion of all quadrats that were non-palm). Species with significantly higher or lower abundances than expected (P < 0.05) were identified.

ISAR is a statistic to analyse the spatial diversity structure in forest ecosystems and reconciles common species area relationships [67–69] and the individual perspective of point-pattern analysis [63, 70]. The ISAR allows for a subtle assessment of the effect of palms (or adult Myristicaceae) trees on the non-palm sapling diversity in their neighbourhood. If negative interactions of the focal species on saplings dominate, the target species would decrease the proportion of species in its neighborhood (i.e., a diversity repeller). Conversely, if the target species exerts positive interactions to non-palm saplings, the target species would accumulate and maintain an over-representative proportion of sapling diversity in its proximity (i.e., being a diversity accumulator). Finally, if positive and negative interactions are weak or even out, the focal species would behave neutrally.

Similar to the case of the neighbourhood density described above, environmental heterogeneity may cause repeller or accumulator effects in the absence of species interactions if the focal species is predominantly located in areas of low or high sapling diversity. We therefore contrasted the observed ISAR(r) function to that of the same null models as for the analyses of neighbourhood density. In the first null model we distributed individuals of palms randomly over the entire plot. The second test also assumed a random distribution of palms, but the palms were only displaced within a given neighborhood of 30 m to remove in the null model potential interactions with saplings and to guarantee that the displaced palms will be as well located in areas with low sapling density. The neighborhood of 30 m is somewhat larger than the expected range of direct interactions between focal trees. These analyses were conducted in Programita, for more details of ISAR, see Wiegand et al. [64].

We hypothesize that large palms are a diversity repeller because of the disturbance effects of falling fronds or rooting damage by palm fruit consumers in the vicinity of palms. Only a subset of species would have the traits necessary for resprouting and continued growth following damage from nearby palms.

Dynamic demographic effects of large trees on seedlings

To examine the effects of proximity to large palms and Myristicaceae on seedling survival, for each of the 339 seedling plots located in the FDP we calculated the total number of new recruits of all species between 2003 and 2007. We then calculated the number of each cohort surviving one year. We modelled the proportion of survivors in each plot as a function of location relative to mature and immature large trees with a generalised linear model with binomial errors as above.

Dynamic demographic effects of large trees on saplings

We examined the survival, growth and resprouting of saplings 1-2 cm dbh from the first to the second plot census as a function of whether the sapling was near to (within 10 m) or far from (> 10 m) focal trees (palm or Myristicaceae). First, we modelled the probability of survival to the second census of all saplings alive in the first census. Second, we modelled the probability of each sapling resprouting (new sprouts growing up and out from damaged or broken stems). Third, growth rate for each sapling from the first census to the second census was calculated as growth = (dbh1 dbh0)/(time1 time0) and divided into negative (1) or ≥ zero growth (0). We modelled the probability of a negative growth rate between censuses. In all three cases we used a generalised linear model with binomial errors. We modelled survival (or growth or resprouting, all binary response variables) as a function of proximity to the nearest focal tree (mature and immature palm or male and female Myristicaceae): survival ~ mature palms * immature palms. We coded each sapling as near or far (</> 10 m) from focal trees. We included the interaction term because saplings could be near to a mature palm but far from an immature palm, for example.