UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)
Flora, vegetation and ecology in the Venezuelan Andes: a case study of Ramal
de Guaramacal
Cuello Alvarado, N.L.
Publication date
2010
Link to publication
Citation for published version (APA):
Cuello Alvarado, N. L. (2010). Flora, vegetation and ecology in the Venezuelan Andes: a case
study of Ramal de Guaramacal. Universiteit van Amsterdam, Institute for Biodiversity and
Ecosystem Dynamics (IBED).
General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.
Venezuela endemic species of the Espeletiinae, found in Páramo de Guaramacal: (a-c) Ruilopezia
jabonensis; (d, e) Ruilopezia lopez-palacii; (f-h) Ruilopezia paltonioides; (i-l) Ruilopezia viridis.Chapter 6
Functional diversity of Andean forests in Venezuela changes
with altitude
6.1 INTRODUCTION
Tropical Andean forests are one of the world's biodiversity hot-spots (Myers et al. 2000). These forests have a rich biodiversity (Gentry 1995) and are highly threatened because of increasing deforestation (Etter & Wijngaarden 2000; Armenteras et al. 2003). The rising temperature in the past decades and associated upslope shifts in species distribution may impose extra threats to Andean forests (Colwell et al. 2008; Svenning & Condit 2008). Potential threats due to losses in forest cover and biotic attrition might be exacerbated by degradation in functional diversity, i.e. the variety of life-history traits presented by an assemblage of organisms (Mayfield et al. 2005; Girao et al. 2007). Decreasing functional diversity is generally seen as indication of degradation and a hazard for ecosystem resilience (Tilman et al. 1997). For example, the lower diversity in reproductive traits in forest fragments in lowland Amazonia of Brazil may have detrimental consequences for the population size of pollinators and the trophic structure (Girao et al. 2007). The principal aim of our study is to examine if functional diversity changes with altitude in undisturbed Andean forests, to contribute as reference information for studies of degraded Andean systems.
Along mountain slopes temperature change strongly defines the rate of photosynthesis (Rada et al. 1992; Cabrera et al. 1998), physiological and metabolic processes (Lambers et al. 2008), growth (Grubb 1977; Medina & Klinge 1983; Ashton 2003; Leuschner & Moser 2008), nutrient uptake (Bruijnzeel 1991; Gerold 2008; Leuschner & Moser 2008) and decomposition (Illig et al. 2008), and is therefore the principal driver of ecosystem functioning (Chapin & Körner 1998; Colwell et al. 2008; Svenning & Condit 2008). In general nutrient availability and decomposition rates decrease at higher elevations in tropical wet montane forests (Cavelier 1996). Above 1500 m, chances on occasional frost increase at higher elevations. Yet, because of the strong insolation, the maximum daily temperature remains quite similar to lowland values, resulting in a larger diurnal temperature range upslope (Hansen et al. 2002). In upper montane and subalpine rain forest (SARF), canopy trees receive a large proportion of ultraviolet light, which potentially affects growth (Flenley 1992). Lastly, terrain conditions (more summits) and the proximity of the upper forest line dictate that less space becomes available for continuous forests at higher elevations, which makes fragmentation by natural causes, in principal, more frequent. Most of these factors contribute to stronger upslope levels of ecological filtering (Keddy 1992; Weiher & Keddy 1995; Ackerly 2003) acting upon montane forest plants, reducing the number of traits relative to species (underdispersion). Alternatively, increased competition for more limiting resources at higher altitudes (for example due to the lower decomposition rates) might invoke ecological differentiation leading to higher trait diversity relative to species diversity (overdispersion) (Weiher & Keddy 1995; Mayfield et al. 2005).
Temperature-constrained processes likely become manifest in plants through variation in response traits (Gitay & Noble 1997; Naeem & Wright 2003; Violle et al. 2007) related to the energy balance (growth form, leaf shape and leaf size) (Cornelissen et al. 2003). Fragmentation hampers dispersal and cross-pollination, affecting the distribution of regenerative response traits like dispersal mode and
_______________________________________________________
pollination mode, fruit type, and flower and seed size (Cornelissen et al. 2003; Girao et al. 2007). We studied the vascular plant composition of 44 small plots located between 1330 m and 3060 m altitude in a well-protected forest reserve in the Andes of Venezuela (Cuello & Cleef 2009a). We linked each species to the above-mentioned functional traits by means of literature and herbarium studies. Randomizing the species assemblages in our relevés (Legendre et al. 1997; Dray & Legendre 2008) we tested if the composition and diversity of energy balance related traits and fragmentation related traits changed with elevation.6.2 METHODS
Study area
Ramal de Guaramacal is an outlier of the Venezuelan Andes, which lies to the southeast of Boconó, Trujillo State, approximately 120 km northeast of Mérida, in the centre of the Sierra Nevada de Mérida (9° 05–21' N and 70° 00–20' W). This mountain range reaches up to about 3100 m, and most pertains to a National Park, which includes an approximate surface area of 21,466 ha. The average yearly rainfall measured over 2002-2008 at a climate station in the study area (Laguna de los Cedros at 1980 m; 9° 15' 55'' N; 70° 13' 13'' S) was 2106 mm, and showed a unimodal pattern with February as driest month and June as wettest. Temperature average around 18 to 20° C between 1000 and 1500 m, and 9 to 12° C above 2500 m (Cuello & Barbera 1999). Above 2500 m seasonal frost may occur (Urriola 1999). The high precipitation in the area favors intense lixiviation and acidification, and acid soils predominate (Marvez & Schargel 1999).
The vegetation of Guaramacal Park area is predominantly represented by montane rain forests with height and density decreasing with altitude. These forests have been described (Cuello & Cleef 2009a) into discrete zones corresponding to lower montane, upper montane, and SARF, following Grubb (1977). The montane forests can be found from 1350 m to about 2800 m. Between 2800 m and 3130 m SARF is found intermingled in a mosaic with subpáramo formations (Cuello & Cleef 2009b).
Ramal de Guaramacal has received the status of National Park since 1988, keeping most human activities and impacts outside the park borders. Fires have occurred in the past, especially in páramo areas. Some selective timber extraction is known to occur at low intensity and generally takes place in close proximity to the park limits. In the surroundings of Ramal de Guaramacal there has been a long history of agricultural activity mainly for coffee plantation, slash and burn cultivation and extensive cattle ranging, among other land uses (Barbera 1999). However, the high ridges and steep slopes of Guaramacal have kept most of the montane forest areas with minimum disturbance.
Field methods
The fieldwork was carried out in 1995, 1996, 1999, 2003, 2005 and 2006. Montane forests were studied along the altitudinal gradient on both sides of the range with Flora, vegetation and ecology in the Venezuelan Andes
different slope expositions (Cuello & Cleef 2009a, b). Thirty five 0.1 ha (20 x 50 m) plots were surveyed, positioned a distances of 30 to 150 m between 1350 m and 2890 m altitude, and nine plots of variable size (50 m2 to 400 m2) were surveyed in SARF between 2800-3050 m. Within each plot, all rooted individuals – trees, shrubs, lianas, tall and thick-stem or climbing terrestrial herbs and hemi-epiphytes ≥ 2.5 cm DBH (diameter at breast height, taken at 1.3 m from the base of the trunk, or lower for shrubs and thick-stemmed herbs) were recorded, labeled with numbered aluminum tags and their DBH and height recorded.
A total of 2082 numbers botanical specimens of vascular plants were collected. Botanical material was processed and identified at Herbario Universitario PORT of UNELLEZ. Other herbaria, such as MO and US, were also consulted. Some specimens were sent to specialists at other institutions to confirm identification. All specimens collected have been deposited at PORT, some duplicates have been sent to VEN, MER, MERF, MO and US.
From a total of 388 morphospecies recorded from all surveys, 357 were identified to species or genus level. These we used to compiled trait state data on energy balance and fragmentation related traits, on the basis of literature, floras and botanical monographs, web searches, and herbarium voucher information (Table 6.1).
Table 6.1 Plant response traits and their respective categories or trait states considered in this study.
Trait Trait states or categories
Energy balance related traits
Growth form bamboo (39), climbing herb (41), erect herb (45), hemi-epiphytic tree (46), liana (47), palm (48), stem rosette (49), tree (52), tree fern (53), upright shrub (55)
Leaf type simple (56), dissected (57), compound (58)
Leaf size leptophyll (<0.25 cm2)(59), nanophyll (0.25-2.25 cm2)(60),
microphyll (2.25-20.25 cm2)(61), notophyll (20.25-45 cm2)(62),
mesophyll (45-182.25 cm2)(63), macrophyll (182-1640.25
cm2)(64), megaphyll (>1640.25 cm2)(65)
Fragmentation related traits
Dispersal autochory (1), anemochory (2), hydrochory (3), zoochory (4) Pollination insect (5), bat (6), bird (7), self (8), water (9), wind (10), none
(11)
Sexual system dichogamy (12), dioecious (13), monoecious (14), polygamous (15), hermaphrodite (16), none (17)
Fruit type achene (18), berry (19), capsule (20), drupe (21), fleshy capsule/pome (22), follicle (23), legume (24), naked seed (25), syncarp (26), none (27)
Fruit size tiny (<2 mm2) (28), small (2–5 mm2)(29), medium (6–15 mm2
long)(30), large (16–25 mm long)(31), ex-large (36–100 mm long)(32), huge (>100 mm long)(33)
Flower size Inconspicuous (<4 mm)(34), small (4-10 mm)(35), medium (10-20 mm)(36), large ((10-20-30 mm)(37), very large (>30 mm)(38)
_______________________________________________________
Data analysis
Detrended correspondence analysis (DCA) was done on the basis of the species-to-relevee matrix, in which the species abundances were log-transformed, and on the basis of the species-to-trait state matrix, in which all trait states were entered as dummy variables. Variance partitioning was done using Canonical Correspon-dence Analysis (CCA; scaling focus of inter-species-distances and biplot scaling type) after Borcard et al. (1992). In this, the environmental information was represented by the elevation of the relevés in m above sea level. The spatial variables were selected by means of a forward selection procedure (using a probability level of 0.05) in CCA of the log species abundances against all terms of the third-degree polynomial of the centered UTM X and Y coordinates (Legendre & Legendre, 1998). The pure effects of the explanatory variables on the species patterns were tested by means of Monte-Carlo permutation tests under reduced model, applying 499 permutations. All DCA and CCA analyses were done in CANOCO for Windows 4.5.
The relevé scores to visually show the variation in trait composition against elevation were calculated as the weighted average of the species scores of the DCA of the species-to-trait state matrix, with for each relevee the number of plants per species as weight. Fourth-corner analysis was done applying 999 permutations under models 1 and 3 (Legendre et al. 1997; Dray & Legendre 2008) with the fourthcorner function implemented by S. Dray in the ade4 package (Dray & Dufour 2007) in r 2.10.
Trait state diversity was quantified by the Shannon and the Simpson (1-D) indices. Because of few aberrant relevee sizes, also Fisher's alpha was used to reduce possible effects of variable sampling sizes. All indices were calculated (applying Vegan 1.17-2 in r 2.10) on the basis of the number of species or the number of individuals per trait state. Thus, in analogy to the calculations of species diversity, trait states were used as equivalent of species and the numbers of species or the numbers of individuals per trait state were used as equivalent of the number of individuals (Girao et al. 2007). The Pearson coefficient of the correlation of the diversity indices with elevation was tested by means of 999 permutations of the species-to-relevee matrix according to the same permutation models and the same number of permutations applied in the fourthcorner tests, after adding the reference value (Pearson correlation coefficient from the unpermuted matrix) to the distribution of the null model (Legendre et al. 1997).
6.3 RESULTS
Trait composition against elevation
In the 44 relevés a total of 357 species were recorded (see Appendix 6). Most species (85%) were fully identified. The relevee scores of the first DCA axis of the species-to-relevee matrix were highly correlated with altitude (Fig. 6.1A). The gradient length of this axis was 9.1, indicating a substantial degree of species turnover between the relevés (Hill & Gauch 1980). In space, the relevés were clustered in about five groups, the largest of which consisted of the relevés made Flora, vegetation and ecology in the Venezuelan Andes
most upslope (Fig. 6.1B). The forward selection procedure of the CCA of the log-transformed species information against the nine spatial variables produced five significant terms (X, Y, X2, Y2, and X*Y). Of these X*Y was skipped from the final analysis because of its high correlation with both X2 and Y2 terms (the variable inflation factors of this final CCA were below 5). In the variance partitioning, the pure elevation effect explained 6.2% of the species variation, and the pure spatial effect 12.9%. Both these pure effects were significant (Monte-Carlo permutation tests p=0.002). Elevation and space combined explained 2.8% and the fraction of unexplained variation was 78%.
Figure 6.1 (left). Sources of variation in vascular plant species composition in the forests of Ramal de Guaramacal in the Venezuelan Andes. A: Altitude: the association between the first DCA axis of the species-to-relevé matrix and the altitude of the relevés. B: Space: the spatial configuration of the relevés. Wider circles were made at higher elevations.
Figure 6.2 (right). The principal variation in energy balance related traits (A) and fragmentation related traits (B) extracted by means of the DCA of the species-to-trait state matrix. Most trait states are simply abbreviated; leaf1-7, fr1-6, and fl1-5 means leaf size, fruit size, and flower size in increasing order (compare Table 6.1).
_______________________________________________________
In the DCA analysis of the species-to trait state matrix, the main variation (13.7%) in energy balance related traits (Table 6.1) was clearly related to leaf size (Fig. 6.2A). Species with large leaves scored low scores along the first DCA axis and fine-leaved species high scores. The variation between palms and tree ferns (showing compound leaves), and herbs (either climbing or erect with dissected leaves) mainly determined the variation along the second DCA axis. The ordination of the fragmentation related traits was mainly driven by fruit and flower size (associated with fleshy fruits, legumes, bat pollinated flowers, autochorous dispersal and self-pollinated) with highest scores along DCA axes 1 and 2, whereas small sized flowers and fruits (associated with dispersal and pollination by wind, achenous fruits and fern reproduction) were situated towards the lower left side of the ordination diagram (Fig. 6.2B).Table 6.2 Association between elevation and the DCA axes of energy balance related traits and fragmentation related traits, as given by squared Pearson correlation coefficients (r) and their probabilities (p) obtained by means of the fourth-corner analysis applying two permutation models.
DCA axis 1 DCA axis 2
model 1 model 3 model
1 model 3 r p p r P p Energy balance related traits 0.35 0.001 0.001 -0.11 0.064 0.028 Fragmentation related traits -0.16 0.001 0.007 -0.25 0.001 0.001
Table 6.3. Association between elevation and three diversity indices of energy balance related traits and fragmentation related traits, as given by Pearson correlation coefficients (r) and their probabilities (p) obtained by applying two permutation models.
Shannon Simpson (1-D) Fisher's alpha
model 1 model 3 model 1 mode l 3 model 1 model 3 r p p r p p r p p Species-based trait diversity Energy balance related traits 0.52 0.001 0.001 0.41 0.001 0.001 0.74 0.001 0.001 Fragmentation related traits -0.58 0.001 0.022 -0.10 >0.2 >0.2 0.31 0.016 >0.2 Individual-based trait diversity Energy balance related traits -0.24 0.044 >0.2 -0.34 0.008 >0.2 0.34 0.012 0.001 Fragmentation related traits -0.66 0.001 0.002 -0.62 0.001 0.003 -0.56 0.001 >0.2 Flora, vegetation and ecology in the Venezuelan Andes
Visually, elevation correlated well to the first DCA axis of energy balance related traits (Fig. 6.3A) but less convincingly so to the second DCA axis. Both first and second DCA axes of the fragmentation related traits seemed associated to elevation (Fig. 6.3B). Fourth-corner analysis allows the selection of the most appropriate null model to test for association between functional trait composition of species in relevés and environmental properties of these relevés (Legendre et al., 1997). Because of the significant effect of altitude on the species composition in the relevés and the continuous species turnover along with altitude (Fig. 6.1A), it seemed likely the individual ecological and physiological species responses to elevation ruled the co-occurrences of species in the relevés. For this reason we selected permutation model 1 (for each species separately the abundance values are randomly distributed over the relevés; Legendre et al. 1997). However, the variance partitioning showed that the spatial effect on species composition was about twice the altitudinal effect. Spatial distance potentially hinders dispersal and limits recruitment, both important in situations when regeneration through colonization drives species composition. Because regeneration and colonization after disturbances potentially depends on largely unpredictable processes related to mass movements and tree mortality the lottery model of species assembly (Sale 1978; van der Maarel & Sykes 1993) seemed appropriate as well. Therefore, we also applied a model 3 randomization (for each relevee separately the abundance values are randomly distributed over the species; Legendre et al., 1997). The fourthcorner results (Table 6.2) were in line with our visual interpretations of the scatter plots (Fig. 6.3AB), evidencing that the distribution of energy balance and fragmentation related traits was significantly correlated with altitude.
Functional diversity against elevation
All three species-based diversity indices in energy balance related traits showed a convincing positive correlation with elevation (Fig.6.4A, Table 6.3). The altitudinal association of the individual-based diversity of these traits was weaker and less consistent (Fig. 6.4B, Table 6.3). The Shannon and Simpson indices seemed negatively correlated but these patterns depended strongly on three SARF relevés and lacked significance when tested with the lottery model of permutation. The individual-based Fisher's alpha index of energy balance related traits was positively related to elevation.
Regarding the diversity in fragmentation related traits, the overall tendency was that of a negative association with elevation (Fig. 6.4A and B; Table 6.3). However, compared to the energy balance related traits, the altitudinal correlations were weaker and more strongly influenced by outlying diversities of SARF plots. Positive SARF outliers (going against the trend of neutral or negative trends of diversity with elevation) were clearly visible in the scatters of species-based Simpson and Fisher's alpha indices (Fig.6.4A). For that reason the positive altitudinal correlation of the species-based Fisher's alpha was not convincing, even though it was significant in both permutation models. Negative outliers of SARF against elevation appeared in the scatters of the individual-based Shannon and Simpson against elevation (Fig. 6.4B).
_______________________________________________________
Figure 6.3. The association between elevation and the principal variation in trait composition of the vascular plant species in the forests of Ramal de Guaramacal in the Venezuelan Andes. Relevés scores for the trait composition were the weighted average of the species scores of the DCA axes derived from the species-to-trait state matrix, with for each relevé the number of individuals as weight. A: energy balance related traits. B: fragmentation related traits. Symbols of SARF relevés have been filled
6.4 DISCUSSION
Energy balance related traits
The forests in the Ramal de Guaramacal area varied altitudinally in the selected energy balance related traits. Also they became more diverse in these traits at higher elevations, pointing at more prominent levels of overdispersion higher up the slopes. Community assemblage rule theory (Weiher & Keddy 1995; Díaz et al. 1998) predicts that increasing levels of overdispersion might occur when better adapted species outcompete functionally related species from the local community. Leaf size contributed substantially to the altitudinal variation in energy balance related traits (Fig. 6.2A). The lower leaf size at higher altitudes in wet tropical forests has been recorded repeatedly (Grubb et al. 1963; Vareschi 1966; Sugden 1985). In the absence of pronounced dryness (as in the situation along the slopes of the Ramal de Guaramacal area) this can be explained by a lower upslope temperature, more limited hydraulic conductance of stems and associated lower mineral supplies, lower nutrient availability, and increased frost frequencies (Cavelier 1996). Overall, at higher altitudes in montane wet forests, nitrogen (Grubb 1977; Cavelier 1996) becomes more limiting. Therefore, our results Flora, vegetation and ecology in the Venezuelan Andes
suggested that competition for resources, mostly related to the capture of radiation (heat) and the uptake of minerals and nutrients, is an important driver of species composition in this part of the Venezuelan Andes.
Figure 6.4A. Scatter plots of trait state diversity against elevation, calculated on the basis of traits per species (left), or traits per individuals (right) for energy balance related traits.
_______________________________________________________
Figure 6.4B: Scatter plots of trait state diversity against elevation, calculated on the basis of traits per species (left), or traits per individuals (right) for fragmentation related traits. Symbols of SARF relevés have been filled.
The association between functional diversity of energy related traits with elevation further suggested that a temperature rise as a consequence Global Change might affect the forest functionality of Andean forests. Projected higher temperatures in Flora, vegetation and ecology in the Venezuelan Andes
the coming decades (Solomon et al. 2007) might reduce the functional diversity of energy balance related along the slopes of the Andes. This projection has only general implications, however, because the detailed mechanisms, by which species migrate as function of changing temperatures and ecological filtering influences species interactions, are still poorly understood (Svenning & Condit 2008). For example, vascular plants may migrate upslope at different speeds compared to decomposer communities, which largely define the nutrient availability. Moreover, our results suggests that even in undisturbed Andean forests functional diversity varies significantly, which implies that some lower degree of functional diversity at a certain elevation in the nearby future does not necessarily endangers the ecosystem well-being.
Fragmentation related traits
Just as with the energy balance related traits, the forest species varied in our selection of fragmentation related traits as function of altitude. The negative altitudinal association of the main variation in fragmentation trait state composition was mostly due to wind-dispersed and wind pollinated species from SARF forests (e.g. species from Alsophila, Cyathea, Dicksonia, Diplazium,
Baccharis, Diplostephium, Pentacalia, Mikania). Because of the low human
influence in the Guaramacal area, the transition between SARF forests and páramo vegetation is not sharp (for example compared to forest-páramo boundaries caused by burning; Moscol & Cleef 2009). Instead, SARF and páramo vegetation occur in a spatially well-mixed mosaic (Cuello & Cleef 2009a, b). Therefore, the predominance of the trait states related to wind transportation in our highest samples could be explained by the flow of plant propagules along forest-páramo edges (Ries et al. 2004). The comparison of second-best variation in fragmentation state composition with altitude (Fig. 6.3B) was due to the tendency that species with larger fruits and flowers occurred at relatively low elevations (e.g.
Symbolanthus vasculosus (Griseb.) Gilg., Zygia bisingula L. Rico, Drymonia crassa C.V. Morton, Tabebuia guayacan (Seem.) Hemsl., Macrocarpaea bracteata Ewan, Inga edulis Mart.). We speculate that this pattern is related to a
more important role of birds in pollination and seed dispersal at elevations above 2100 m, versus a more pronounced role of mammals (including large bats) at lower elevations.
Several SARF plots showed an outlying functional diversity compared to the altitudinal trends in the lower forest relevés. This suggests that, in contrast to the lower lying forests, the plants in the SARF relevés contained markedly more trait states relative to the number of species, and/or more plant individuals relative to the variety in trait states. Both phenomena may be caused by the increased wind flow in SARF forests enhancing the number of plants with traits related to wind transport.
Setting aside the outlying SARF patterns, and in contrast to the energy balance related traits, the diversity of fragmentation related traits tended towards a negative association with elevation, visible in both species-based and individual-based indices of the montane forests. Hence, our results indicate that along a natural altitudinal gradient in Andean rain forests, undisturbed by human influence,
_______________________________________________________
ecological differentiation regarding disturbance gets lower in upslope direction. Possibly, the increased level of upslope underdispersion is due to a higher fragmentation of the forest matrix because of stronger terrain dissections. In comparison, Girao et al. (2007) reported that lowland forest fragments in Brazil showed a lower functional diversity in reproductive traits compared to control forests. Their findings pointed towards higher frequencies of self-incompatible systems due to habitat loss. Altitudinal information on forest dynamics and forest disturbance related to mass movements and slope instabilities in the Guaramacal area is needed to further develop hypotheses about causal mechanisms explaining the upward decrease in functional diversity of disturbance related traits.Conclusions
According to our expectations, we found that functional diversity of undisturbed Andean forests in the Guaramacal area changed with altitude. This implies that temperature rise due to Global Change might affect the forest functionality of Andean forests in the near future, but not necessarily in a harmful way. Functional diversity related to energy balance traits increased in upslope direction, pointing at increased levels of ecological differentiation. We explained this by assuming more upslope competition in the Andean forests regarding capture of radiation and the uptake of minerals and nutrients. Diversity in fragmentation related traits showed an opposite pattern (more underdispersion upslope), which might relate to discontinuities in the forest matrix due to the geomorphology of mountains. SARF forests diverged from the altitudinal trends in fragmentation related traits, probably as a consequence of edge effects in the SARF-páramo mosaic, created by wind. Flora, vegetation and ecology in the Venezuelan Andes
