The effects of individual tree species on species diversity in a tropical dry forest change throughout ontogeny

Abstract

Understanding how diversity is maintained in species‐rich communities, such as tropical forests, remains a challenge in ecology. Recent work suggests that the controversy between competing theories could be better resolved by considering the spatial scale at which different processes rule community assembly.

Here we use individual species–area relationships (ISAR) to evaluate the spatial organization of tree diversity around individuals of different species in a completely‐mapped tropical dry forest in south Ecuador. We test two hypotheses. First, stressful environmental conditions promote facilitative interactions that will generate spatial signals of accumulation of diversity around individual trees – contrary to what has been reported in humid tropical forests. Second, spatial signals will shift through ontogeny. As, as larger, older trees generate new microsite conditions that affect the recruitment of younger, smaller trees.

We compute ISAR functions for adult trees, for young trees and a new crossed‐ISAR function measuring the accumulation of diversity of young trees around the old trees. We compare observed ISARs to the expectations of inhomogeneous Poisson (i.e. null) models controlling for the effects of environmental variation and habitat association on tree distribution.

Although the prevalent response among adult trees was not different from null expectations, which means that the organization of diversity in this size class could be explained by environmental heterogeneity alone, most species accumulated more diversity than expected over short spatial scales in the small stem size class. Only two species accumulated significant diversity in the crossed‐ISARs. Our study indicates the role of facilitation in the organization of plant diversity in this dry forest, but that facilitation is limited to some key species acting on early life stages and accumulating diversity around them.

Our results demonstrate the benefit of considering different life‐stages and crossed analyses to disentangle the processes affecting community assembly in tropical dry forests.

Understanding how diversity is maintained in tropical ecosystems remains a challenge in ecology (Chesson 2000, Wright 2002) and has fostered the elaboration of multiple hypotheses (Janzen 1970, Connell 1971, Hubbell et al. 1999, 2001, Chave et al. 2002, Giles et al. 2004, Volkov et al. 2005). For instance, the neutral theory of biodiversity (Hubbell 2001, and recently revisited by Rosindell et al. 2011), predicted that stochastic immigration and mortality determine species coexistence at fine spatial scales, with the level of species richness determined by the size of the regional species pool. The theory therefore minimizes the influence of deterministic processes such as those related to ecological assembly rules. Some works suggest the importance of ecological drift (i.e. random changes in species relative abundance) for structuring local communities (Rosindell et al. 2011), evidence shows that clumped multi‐species distributions are the norm in mega‐diverse tropical forests (Wiegand et al. 2007a). However, these clumped patterns could also be compatible with differential responses to small‐scale environmental heterogeneity (Ashton 1969, Grubb 1977, John et al. 2007), or with the existence of negative density dependences (Janzen 1970, Connell 1971).

Recently, Chase (2014) pointed out the necessity of explicitly considering different spatial scales to reconcile the predictions of alternative, competing theories. In fact, spatial point pattern analysis, when applied in conjunction with a priori hypotheses based on ecological theory, has the power to disentangle the ecological processes responsible for the structure of plant populations and communities at a complete range of spatial scales (Wiegand and Moloney 2004, McIntire and Fajardo 2009). Within the wide panoply of point pattern techniques, the framework of individual species area relationships (ISAR, Wiegand et al. 2007b) allows researchers to address community assembly from completely‐mapped plant data by estimating how species richness is spatially arranged around the individuals of particular species. Using some appropriate null models to control for the effects of environmental filtering, ISAR allows a subtle assessment of the effects of particular species on, and their responses to, local diversity (i.e. to local species richness) at multiple spatial scales (Wiegand et al. 2007b). More precisely, ‘diversity accumulators’ and ‘diversity repellers’, i.e. species whose individuals are surrounded by more and fewer species than expected at a particular spatial scale, respectively, compared to ‘neutral’ species that experience the average species richness in their neighborhoods. The proportion of these plant types within a community could shed light on the mechanisms ruling coexistence in mega‐diverse plant communities. For instance, the prevalence of diversity accumulators has been related to the dominance of positive plant‐to‐plant interactions, whereas the existence of diversity repellers and neutral species has been respectively related to the prevalence of competition and stochastic assorting (Wiegand et al. 2007b).

Wiegand et al. introduced the ISAR approach in their 2007 study of mega‐diverse, moist tropical forests. They found a weak prevalence of species‐specific effects on local diversity and explained it as the consequence of balanced multi‐specific interactions in a mild environment. Here, we aim to test the generality of this finding and advance the knowledge of tropical forests assembly mechanisms by testing the prevalence of diversity‐structuring species in a dry tropical forest. We hypothesize that under the harsher environmental conditions imposed by seasonal drought, individual species should play a greater role in structuring diversity. This idea is based on the well‐established increase in the importance of positive interactions between plants as the environment becomes less productive and more stressful (Callaway 1997, Bowker et al. 2010). Under such conditions, some species (e.g. ‘nurse species’) may ameliorate the environmental conditions for other species by reducing the frequency and intensity of some physical constraints and thus allowing for increased local species diversity (Hacker and Gaines 1997).

In drylands and tropical seasonal forest, in particular, the recruitment, growth and survival of trees are profoundly limited by seasonal water availability (Blain and Kellman 1991, Murphy and Lugo 1995, Sampaio 1995, Tewksbury and Lloyd 2001). Under such conditions, positive plant–plant interactions are expected to be more prevalent than negative, competitive interactions (Bertness and Callaway 1994, Maestre et al. 2009, Espinosa et al. 2013). Our hypothesis is that large trees of some key species would act as ‘nurses’ by increasing local humidity and/or reducing direct sunlight. The benign microclimate created by the nurse individuals would allow habitat expansion for other species and significantly increase species richness in the vicinity of the facilitators (in comparison to other areas of the forest or to the neighborhoods of other trees that do not exert this facilitative effect).

A weakness when analyzing the organization of local diversity around individuals is the assumption that all individuals of a particular species structure diversity in a similar way. This does not seem realistic since species‐specific effects on diversity are necessarily asymmetric. We hypothesize that ontogeny affects the relationship between individual trees and community diversity: while large and mature individuals of a species can modify, by softening or exacerbating, the environmental limitations under their canopies and thereby affect local diversity, small trees (i.e. juveniles) can only respond to the conditions generated by large individuals. Therefore, we further hypothesize that the role of species in relation to the organization of diversity shifts along ontogeny, with dominant adult individuals generating conditions, and small ones only perceiving the environmental conditions generated by adults.

We worked in an Ecuadorian tropical dry forest ecosystem where environmental stress is linked to water shortage during extended dry periods (Blain and Kellman 1991, Murphy and Lugo 1995, Sampaio 1995, Espinosa et al. 2012). We evaluated the relationship of each tree species with taxonomic diversity (i.e. species richness) by computing individual species–area relationships (ISAR) and more specifically we tested the following two hypotheses: 1) the proportion of species acting as accumulators is higher in comparison with tropical rain forests as expected from the more‐stressful conditions occurring in dry tropical forests, and 2) diversity patterns around large and small individuals are markedly different because of their different role as generators or perceivers of environmental conditions.

Methods

Study site

The study plot was located in the Ecological Reserve Arenillas (hereafter referred to as REA from the Spanish‐language acronym). REA is located at the farthest southwestern tip of Ecuador in El Oro province, between the towns of Arenillas and Huaquillas, and covers approximately 17 ha with altitude ranging from 0–300 m a.s.l. (Fig. 1a). The climate is characterized by a distinct rainy season (January–April) with average precipitation of 515 mm, and an eight‐month dry season with only 152 mm (averages for 45 yr of records for Huaquillas weather station). The mean daily maximum temperature is 25.2°C, with variation of 3.4°C between the coldest and warmest months. The lowest temperatures occur during the dry season (Fig. 1b). According to Sierra (1999), REA contains three different types of vegetation: the dry scrub of lowlands (0–50 m), the deciduous forest of lowlands (50–200 m) and the semi‐deciduous forest of lowlands (100–300 m). These dry forests are considered the most endangered ecosystems in Ecuador (Gentry 1977, Sierra 1999) and constitute part of the Tumbesian biogeographical region. REA shelters one of the last relict dry forests in the Ecuadorian Pacific Coast, where dry forest remnants are heavily fragmented and severely degraded (Linares‐Palomino et al. 2010).

Figure 1

(a) Location of Ecological Reserve Arenillas (REA) in south Ecuador. Grey circles show the study area near ‘Pintag Nuevo’. (b) Climatic diagram in the REA from the Huaquillas weather station, with 45 yr record period (50 m a.s.l.). The black vertical pattern depicts the humid months, while the dotted one shows when aridity prevails. (c) Structure of the forest during the dry season.

We established a permanent plot in the center of REA in 2009, within a very well‐conserved area known as ‘Pintag Nuevo’, which is covered by a transitional formation between dry deciduous forests and dry scrubs of lowlands. The most conspicuous tree species in the area are Ceiba trichistandra (Bombacaceae), Tabebuia chrysantha and Tabebuia bilbergii (Bignoniaceae), together with other species like Colicodendron scabridum (Capparaceae) and Croton spp. (Euphorbiaceae) which become more important in the dry scrub formation. From January to May 2010 and 2011, we delimited a square 9 ha plot with 225 20 × 20 m (400 m²) subplots. All the trees and scrubs with DBH > 5 cm were tagged, measured and identified. During the dry season (July to September 2010 and 2011), all of the marked individuals were mapped using a Leica total station, model TS02‐5power with a precision of < 5 cm.

Spatial pattern analysis

ISAR functions estimate the expected number of species within circular areas with radius r around an average individual of a target species t. Following Wiegand et al. (2007b), we first calculated the bivariate emptiness probability P_tj (0, r) that species j was not present in the circles with radius r around the trees of the target species t (without counting the focal stem if t = j) and then summed up 1 − P_tj(0, r) for all N species present in the plot. ISAR is then estimated as:

for each species t.

In order to evaluate whether or not the ISAR varies with ontogeny as postulated, we computed the ISAR for three complementary sets of data. First, we computed ISAR for ‘adult’ individuals with DBH ≥ 10 cm, which in our ecosystem is the dominant life stage (Linares‐Palomino and Ponce‐Alvarez 2009); these functions summarize how community richness is organized around individual species once they reach maturity and allowed for comparisons with the results from other forests were this DBH limit has been employed (Wiegand et al. 2007b). Second, we computed ISAR functions for ‘juvenile’ individuals (i.e. stems between 5 and 10 cm DBH), which allowed the tendency of young individuals to be located in taxonomically richer or poorer neighborhoods to be assessed. Finally, we computed a ‘crossed’ ISAR, evaluating the richness of small individuals (< 10 cm DBH) of species around the larger individuals (≥ 10 cm DBH) of the target species. This allowed the role and the importance of individual species at their mature stage as promoters or inhibitors of regeneration niches and opportunities for juveniles to be assessed. In order to minimize edge effects (Wiegand and Moloney 2004), we computed ISAR(r) only for trees that were located at a distance ≥ r from the border of the plot.

For each target species, we computed Monte Carlo envelopes from simulated ISAR patterns of the same species. Simulations were built from inhomogeneous Poisson models (hereafter HPm, or ‘null model’) fitted to the spatial pattern of each individual species and life stage (Wiegand et al. 2007a, 2007b). The HPm was based on an intensity surface of the target species estimated for a Gaussian kernel with σ = 25 m. This null model accounts for ‘first‐order effects’ in the spatial distribution of individuals (i.e. spatial variation of intensity related to unmeasured large scale environmental heterogeneity) and controls for the effects of habitat association, i.e. the increased or decreased likelihood that an individual will occur at a given location (Wiegand et al. 2007a, 2009, Rayburn and Wiegand 2012).

To assess effects at the different r scales, we computed Monte Carlo simulation envelopes for each species based on the 99 simulations of the fitted inhomogeneous null models. If the empirical ISAR(r) was at larger a given scale r than the second highest ISAR(r) of all 99 simulations of the null model, the species was regarded at scale r as a diversity accumulator with an approximate α level of 0.05 (Wiegand et al. 2007b). Conversely, if the empirical ISAR(r) was smaller at a given scale r than the second smallest ISAR(r) of all 99 simulations, the species was regarded at scale r as a diversity repeller. If the empirical ISAR(r) was within the range of the null model, the species was considered neutral at scale r (see Fig. 2 for examples of accumulator or repeller species). As the simulation envelopes cannot be interpreted as confidence intervals because of simultaneous inference (i.e. one test for each scale r), we additionally used a goodness‐of‐fit (GoF) test to assess departures of the empirical ISAR curves from the null model (Loosmore and Ford 2006, Perry et al. 2006). The GoF was calculated in three different ranges: from 0 to 10 m, 11 to 20 m and 21 to 30 m radii. Each target species was classified as an accumulator, a repeller or neutral based on the GoF results in each range of radii. Thus, when the species statistically departed above or below the null model within a given range of radii, it was classified as an accumulator or repeller species, respectively. If the ISAR was not significantly different, the species were classified as neutral.

Figure 2

Individual species–area relationships and species–area relationships at REA. (a) Upper panel: ISAR curves for 15 adult individuals (grey solid lines) and for two example species (Caesalpinia glabrata, black dotted line and Colicodendron scabridum, red solid line). The black solid line shows the common SAR for individuals ≥ 10 cm of DAP. Middle and lower panels: observed ISAR (black line) and simulated ISAR functions (grey lines) for the two example species. (b) Upper panel: crossed ISAR for 15 adult individuals (grey solid lines) and for two example species (Caesalpinia glabrata, black dotted line, and Chloroleucon mangense, red solid line). Black solid lines show the common SAR for individuals ≥ 10 cm of DAP. Middle and lower panels: observed ISAR (black line) and simulated ISAR functions (grey lines) for the two example species. (c) Upper panel: ISAR curves for 21 juvenile individuals (grey solid lines) and for one example species (Jacquinia sprucei, black dotted line). Black solid lines show the common SAR for individuals < 10 cm of DAP. Lower panel: observed ISAR (black line) and simulated ISAR functions (grey lines) for the example species.

In addition, we compared the individual ISAR functions with the common SAR (species area relationship) and computed for each species the difference ISAR(r) – SAR(r). This difference measures how much (and at which scales) the empirical ISAR of each species deviates from the common SAR, which is its expected value if there were not effects of habitat association or dispersal limitation. To complement the information provided by the test of ISAR against the heterogeneous null models (HPm), we tested the difference ISAR–SAR against a null model of complete spatial randomness (CSR). This basic test determines whether the focal species is located in areas with relatively‐low or high species richness (Wiegand and Moloney 2014: 410), without accounting for any possible generating mechanism. Thus, a species with significant and non‐significant results for the CSR and HPm tests, respectively, would probably be responding to habitat association (i.e. it appears in a habitat that host more species).

The SAR (r) curve was estimated by setting a regular 5 × 5 m grid of points in the REA and computing the average richness in circles of variable radius r around the points. To assess the significance of the difference ISAR–SAR, we computed Monte‐Carlo envelopes by simulating 99 CSR patterns for each species and re‐computing the ISAR function and the difference ISAR–SAR for each simulated pattern. These envelopes were employed to conduct a GoF test on the whole range of distances considered (i.e. from r = 1 to 75 m).

ISAR analyses were computed in the R environment (R Core Team) using the packages spatstat (Baddeley and Turner 2005) and idar (de la Cruz et al. unpubl.).

Results

We mapped 4194 individual trees with DBH greater than 5 cm; of these, 2896 belonged to 36 tree species, 909 to 15 treelet species and 389 to one succulent. The average abundance per hectare was 466 individuals; trees were more abundant than treelets, with 321 and 101 individuals per ha respectively. The abundance was more evenly distributed among trees since 90% of individuals belonged to 11 species, being Tabebuia billbergii, whereas only two species, Cynophalla mollis and Croton rivinifolius, accounted for 78% of individuals in the treelets class. The mean and standard deviation of taxonomic richness in REA was 2.8 ± 1.7 within a 5 m radii, 7.7 ± 2.5 at 10 m, 19.8 ± 3.19 at 30 m, and 25.7 ± 2.8 at 50 m.

We restricted the computation of ISAR to species having at least 15 individuals in the corresponding life stage category. We computed the ISAR within the juvenile life stage for 21 species, whereas for adults and the crossed ISAR were computed for 15 species (Table 1).

Table 1. p‐values of GoF tests for three spatial ranges: 0–10, 11–20 and 21–30 m. Grey background: species that behaved as an accumulator in the evaluated range. Black background: species that behaved as a repeller in the evaluated range. Neutral behavior is showed as white cell, and n.c. indicates that ISAR was not calculated. R10: range between 0 and 10 m of radii. R20: range between 11 and 20 m of radii. R30: range between 21 and 30 m of radii

Taken together, the most prevalent behavior of species at all spatial scales was neutral in the three ISAR analyses (Fig. 3). In the case of adult trees, only a 20% (3 tree species) and 13% (2 species) of species behaved as accumulators and repellers, respectively, over small spatial scales (Fig. 3). These percentages were even lower in the crossed (i.e. adult–juvenile) analyses. In the analyses of juveniles alone, the percentage of species behaving as accumulators peaked at approximately 50% between 7 and 8 m and, at these scales, it was larger than the percentage of neutral species. These deviations from neutrality were not caused by just a few species but, instead, most species (16 out of 23) behaved as accumulators or repellers at some spatial scales (Table 1).

Figure 3

Proportion of significant richness accumulators and repellers at REA. (a) Crossed ISAR; proportion of species with large trees behaving as significant accumulator or repellers of small individuals diversity, (b) adult ISAR (≥ 10 cm DBH); (c) juvenile ISAR (< 10 cm DBH).

When considering the adult size class, four species behaved as accumulators – two of them, Armatocereus cartwrightianus and Eriotheca ruizii at all spatial scales – and another two behaved as repellers (Chloroleucon mangense and Geoffroea spinosa). In the small stem size class, no species behaved as repellers and 11 out of 21 behaved as accumulators. It is noteworthy that for r = 7–8 m, the proportion of young accumulator species surpassed the neutral ones. In the crossed ISAR, three species behaved as accumulators and two as repellers (Table 1).

Two species, Caesalpinia glabrata and Eriotheca ruizii, were accumulators both in the adult and the crossed ISAR analyses (C. glabrata also was accumulator in the juvenile ISAR). None of the species that were repellers in the adult analyses were repellers in the crossed ISAR and vice versa.

In all three analyses, most species had ISAR values larger than the plot SAR at all spatial scales (Fig. 4; Supplementary material Appendix 1, Table A1), especially in the crossed and juvenile analyses.

Figure 4

Difference between ISAR and common SAR for each species in radii between 1 and 50 m. Grey, black and red lines show respectively neutral, accumulator and repeller species.

Discussion

One of the most pervasive debates in ecology is whether communities are stochastically assembled or are responding to environmental and/or biotic filters (Weiher et al. 2011). Based on Niche theory, we hypothesized that in the seasonally‐dry forests of southern Ecuador, the harshness the environmental conditions would favour dominance of facilitative interactions (Callaway 1997) and that these would affect the spatial structure of plant diversity at fine spatial scales. A first inspection of our results show that these positive interactions (i.e. prevalence of accumulators) are not as common as initially expected. At all spatial scales considered, most species behave as ‘neutral’, i.e. they are surrounded by the expected richness under a null model that only accounts for environmental filtering for the target species. This is similar to what has been reported in other tropical humid forests, where between 65 and 75% of species behaved as neutrals (Wiegand et al. 2007b). Thus, our first hypothesis was not supported; there is no evidence of increased prevalence of accumulator species in this dry tropical forest compared with data reported for moist tropical forest. However, this finding does not completely discard the possible influence of niche‐related processes since ISAR values for most species are larger than the values of the plot SAR (Fig. 4). Higher species richness around individuals of some species than around a regular grid of points might involve different mechanisms; for instance, some species sharing microhabitat preferences within the plot, the presence of some key species attracting others into their proximity or both mechanisms. The predominance of neutral species in the ISAR analyses, where the null models explicitly accounted for spatial heterogeneity, and the absence of a clear relationship between dispersal limitation and ISAR behavior (Supplementary material Appendix 2, Table A2), discards the prevalence of key species and instead points to the existence of environmental filters. In fact, Jara‐Guerrero et al. (2015) have documented that the spatial distribution of 75% of REMA species respond to some kind of environmental heterogeneity.

The results of the ISAR analyses for the juvenile individuals (DBH < 10 cm) also points to the existence of some environmental factor driving community assembly. The differences between the juvenile ISARs and the SAR values computed using only the young individuals are much larger than in the case of adults. Additionally, a large number of species appear as accumulators at short spatial scales (especially at very fine scales). Due to their small size, it is not expected that juvenile individuals of any species act as nurse plants for other individuals. This suggests that the accumulation of species richness would instead be the consequence of microsite variability, with higher numbers of species occupying favorable microsites. Habitat association related to variation in germination and early survival with environmental conditions (Kobe 1996, Caspersen and Kobe 2001, Schurman and Baltzer 2012) could explain these responses. As we accounted for coarse scale variability with the heterogeneous Poisson null models, these fine scale deviations in the accumulated richness around juveniles reinforce this idea. Obviously, these beneficial effects could be enhanced by other fine‐scale processes such as species‐herd protection (i.e. enhanced survival linked to a reduced risk of transmission of species‐specific pests and pathogens because of the reduced intra‐specific density in more‐diverse assemblages; Peters 2003, Comita et al. 2010).

Some adult trees may be responsible for these favorable microhabitats through a facilitative effect. In fact, three abundant species (Caesalpinia glabrata, Eriotheca ruizii and Cynophalla mollis) showed an accumulator behavior in the crossed‐ISAR analyses. The adult individuals of these species significantly accumulated more richness of young individuals than expected according to the null models. Two of these trees, Caesalpinia and Eriotheca, also retain the accumulator behavior in the adult size class. These species could be acting as community engineers, favoring the recruitment of multiple species through the enhancement of environmental conditions (Soliveres et al. 2011, Espinosa et al. 2013) or by promoting dispersal (i.e. by a perch effect; Pausas et al. 2006; a relatively high 39% of tree species in REA are dispersed by animals including birds; Jara‐Guerrero et al. 2015). In fact, both Caesalpinia and another four accumulator species are Fabaceae and could be enhancing soil nitrogen content in their neighbourhood. In the case of Cynophalla mollis, during field work we constated that this evergreen tree was preferred by birds as a resting site during the night, which could increase dispersal under its canopy. An alternative explanation is also possible: these supposed engineer species are simply the winners of the competition for space among all species that detect, and recruit into, intrinsically favorable areas. However, the fact that these two species are also accumulators in the adult life stage suggest that these species might not be especially good competitors.

One possible explanation for the reduced prevalence of accumulator behavior in most species through ontogeny could be the displacement of their individuals within the rich microsites. This would affect the power of the statistical tests and would make them appear mostly as neutrals in the adult class (Baldeck et al. 2013). However, some studies have documented changes in the outcome of plant–plant interactions through ontogeny (le Roux et al. 2013). This results from large individuals requiring more resources and being more competitive which could shift habitat preferences during development (Webb and Peart 2000, Comita et al. 2007).

Another possible explanation for the scarcity of accumulators among the adult class could be that the recruitment of adults and young trees had occurred under different environmental conditions. For example, adults could have recruited after some perturbation whereas juveniles would have recruited in more stable conditions (Chacón‐Labella et al. 2014). We have no information about recent disturbances in the REA, but the incidence of extreme climatic events such as ENSO fluctuations could not be ruled out. In general, these dry tropical forests regenerate quickly after disturbance (Josse and Balslev 1994).

Also, it is worth noting the scarcity of repeller species in the REA: there were none among the juvenile class, only two in the adult class and another two in the crossed analyses. This repeller behavior could be the consequence of a clumped spatial structure, where the excess of conspecifics trees in the neighborhood of each individual would not let enough room for individuals of other species to accumulate. However, the spatial structure of these repellers is not different from other neutral or accumulator species (Supplementary material Appendix 2, Table A2).

In conclusion, contrary to our expectations, in this tropical dry forest most species were neutral with respect to the structuring of species richness. However, this does not preclude the influence of facilitative processes in the organization of diversity. The adults of some species accumulated significantly more diversity than expected, which is probably linked to the existence of positive plant–plant interactions. Discrepancies along ontogeny are the norm for most species, which confirms the importance of analyzing ISAR and crossed‐ISAR functions for different life stages.

Acknowledgements

This work has been partially supported by projects A/024796/09 and A/030244/10 financed by Agencia Española de Cooperación Internacional y para el Desarrollo (AECID), projects Islas‐Espacio CGL2009‐13190‐C03‐02 and Mountains CGL2012‐38427 financed by the Spanish Ministerio de Ciencia, project REMEDINAL3 financed by Comunidad de Madrid, project PIC 08 138 – Ecuador Secretaria Nacional de Educación Superior Ciencia, Tecnología e Innovación (SENESCYT, < www.senescyt.gob.ec>), SENESCYT scholarship 2008‐2 and project IBERO‐REDD+ financed by CYTED. A special acknowledgement to Ministerio de Ambiente del Ecuador and the Ministerio de Defensa del Ecuador for the facilities and operational support while developing field work. We thank also Mercedes Martinez and David Duncan for language revision.

Supplementary material (Appendix ECOG‐01328 at < www.ecography.org/appendix/ecog‐01328 >). Appendix 1–2.