Altitudinal vegetation belts in the high-Andes of central Chile ( 33 ° S ) Pisos altitudinales de vegetaci 6 n en los Andes de Chile central ( 33 ° S )

The limits of alpine vegetation belts have been established mainly based on physiognomic criteria. However, a more objective approach for fixing limits of vegetation belts are methods based on species composition and relative abundance of each species. While these methods are more time consuming, they are more detailed and permit the detection of physical factors affecting the limits of vegetation belts. In this paper we: 1) describe the altitudinal changes of vegetation above timberline, 2) compare vegetation belts defined with physiognomy and two floristic methods (a qualitative one based on altitudinal changes in species composition, and a quantitative one based on changes in dominant species); and 3) detect some environmental factors responsible for the altitudinal distribution of alpine vegetation between 2100 and 3700 masl in the Andes of Santiago, central Chile (33°S). There was a complete agreement between the different methods in delimiting the subalpine belt. However, in the lower alpine belt (the cushion's belt) floristic methods subdivided it in 23 sub-belts. In the floristic methods, elevations 3500-3700 that formed the higher alpine belt segregated in separate ways because they have no species in common. Physiognomic descriptions lose relevant information about species distribution, especially at higher elevations. Mean annual temperature and nitrogen content of soil are the main environmental factors affecting the altitudinallimits of vegetation belts in the central Chilean Andes.


INTRODUCTION
In high mountains, climate change abruptly with elevation, especially above timberline (Ferreyra et al. 1998).Temperature, depth of permafrost and growing season decrease with elevation, whereas solar radiation and wind speed increase with elevation, affecting the composition and relative importance of plant species inhabiting these zones (Bliss 1971, Billings 1974).In addition to macroclimate, other environmental variables such.as slope aspect, slope inclinqtion and substrate may affect plant distribution in high mountain systems (Squeo et al. 1993).
On an altitudinal gradient, vegetation changes are often gradual, though at certain elevations those variations are sufficiently conspicuous to allow the recognition of discrete units usually called altitudinal vegetation belts (Moravec 1989, Kitayama 1992).The definition and delimitation of vegetation belts has been mainly based on physiognomic criteria, where the relative abundance of different life-forms at different elevations has been the most used (Squeo et al. 1993).
Vegetation limits based on physiognomic methods is relatively easy to obtain in the field, and often is considered meaningful for habitat description (Goldsmith & Harrison 1983).However their results are highly subjective because they depend on the investigator criterion (op.cit.).For example, several authors have described altitudinal changes in alpine vegetation at different latitudes in the Chilean Andes (e.g., Villagrán et al. 1980, 1981, 1983, Quintanilla 1980, 1996, Arroyo et al. 1988, Squeo 1991, Squeo et al. 1993, 1994 ), defining three to four altitudinal vegetation belts depending on latitude and the physiognomic criteria used (Squeo et al. 1994 ).
In the Andes of central Chile (33°S) Quintanilla ( 1980,1996) based on altitudinal transects done in the Rio San Francisco basin, defined only two altitudinal belts: an Andean belt located between 2300 and 3000 masl characterized by a shrubby steppe dominated by spiny species, and a nivoglacial belt from 3000 up to 3600 m with many rosette and cushion plants.Arroyo et al. ( 1981 ), for this same area, defined three altitudinal belts: 1) a Sub-andean scrub belt between 2200-2600 m elevation, characterized by the dominance of postrate shrubs, 2) a lower Andean belt between 2700-3100 m, characterized by the dominance of cushion plants, and 3) an upper Andean belt between 3200-3800 m, characterized by a sharp decrease in vegetation cover, with dominance of gramineous (tussock grasses) and small rosette plants.Teillier et al. (1994 ), in the El Morado National Park, also located in the Andes of Santiago, defined the following limits for vegetation belts: Sub-Andean belt 2000-2500 m, lower Andean belt 2600-3000 m and higher Andean belt: 3000-3500 m.
A more objective approach for delimitation of vegetation belts are methods based on species composition and relative abundance of each species (Goldsmith & Harrison 1983 ).While these methods are more time consuming, because they rely on measurements or samples of vegetation, they are more detailed in the description and delimitation of vegetation.The delimitation can be done only based on changes in the species composition, only based on changes in the dominant species, or both (Goldsmith & Harrison 1983).The main difference between these methods is the sampling effort involved and the accuracy of the realized delimitation ( op.cit).However, the most relevant feature of these methods is that they permit (with the aid of correlative methods) the detection of physical factors involved in the delimitation of vegetation belts (Goldsmith & Harrison 1983 ).For example, many studies in different alpine zones have established that temperature and some characteristic of the substrate are the main determinants of the vegetation belt limits (Chabot & Billings 1972, Squeo et al. 1993, Vebizquez 1994, Fernandez-Palacios & de Nicolas 1995, Ferreyra et al. 1998).As far as we are aware, only Squeo et al. (1993) have determined physical factors that affect the limits of alpine vegetation in the Chilean Andes.
The objectives of this paper are: 1) to describe the altitudinal changes of vegetation above timberline, 2) to compare vegetation belts defined with 1 physiognomic and 2 floristic methods, and 3) to detect some environmental factors that may be responsible for the altitudinal distribution of vegetation belts in the Andes of central Chile.
The Andes of central Chile is characterized as a young mountain range.The geology of the study zone is composed by two large lithological facies: "Farellones Formation" and "La Parva-Cerro Colorado Formation" (sensu Thiele 1980).Farellones Formation (21 00 to 3000 masl) is a stratified lithologic sequence, assigned to a Miocene age ( 18-20 m.y BP), made up by la vas, tuffs and breccias of andesitic, rhyolitic and basaltic composition alt~rnating with sedimentary rocks (Thiele 1980).The sedimentary rocks are mainly of lacustrine origin corresponding to conglomerates, sandstones and finely stratified shales (Vergara et al. 1988, Thiele 1980).La Parva -Cerro Colorado Formation is a set of volcanic and subvolcanic rocks mainly made up by rhyolitic porphyry, often found in structures such as domes and necks, associated with stratified rocks compound of lava flows, tuffs and breccias of andesitic and rhyolitic composition.Thiele ( 1980) assigned a Pliocene age to this formation.however some recent data, suggests a Miocene age, and therefore including this formation within the Farellones Formation (Beccard et al. 1986).• According to the structural composition of soil, because of the complex geology and geomorphology, there is a gradient from loam-type soils at lower elevation to clay-type soils at higher elevations.Organic matter content is< 3% and, as occur in alpine soils (Retzer 1973), decreases with elevation.
Central Chile is characterized by a Mediterranean-type climate (di Castri & Hajek 1976), with high interannual variability in rainfalls (Aceituno 1990).Precipitation mainly occur in the form of snow during winter, although at higher elevations there occur summer storms with either flash hail or snow fall (Santibafiez & Uribe 1990).Snow cover remains for 3 to 5 months depending on elevation and solar exposure (Rozzi et al. 1989).Mean annual temperature at 2600 masl is close to 6.5°C decreasing to 3°C at 3150 m (Cavieres & Arroyo, 1999).Annual lapse rate is 6.1 °C/km, but showing seasonal variation, 4.0°C/km in winter and 7.0°/km in summer (op.cit.).

Vegetation sampling
Immediately above the Kageneckia angustifolia D. Don (Rosaceae) treeline, that is found at 2000 m (Le6n 1991), vegetation was sampled every 100 m between 2100 and 3700 m.Eighteen -50 m long transect, perpendicular to slope, were established and plants sampled with the point-intercept methods.We estimated species richness, absolute and relative cover of vegetation for the following life forms: Annuals, Perennial herbs, Cushions and Shrubs, according to Marticorena (in press).From 2100 to 2900 m, transects were sampled on south-facing slopes, while between 3000-3700 m transects were placed on southeastfacing slopes due to topographic difficulties.Inclination of slopes, and percent of rocks were also registered at each transect.
At each elevation, a soil sample (0.2 m depth) were taken with a metallic bore, stored in sealed plastic bags and sent it to a chemical laboratory (Agrolab Ltda.)where content of available nitrogen and phosphorous, pH, electric conductivity and texture (percent of clay) were determined.Content of available nitrogen and phosphorous were determined with the methods of Bremen and Olsen, respectively (Pearcy et al. 1989, Agrolab Ltda., comm. pers.), while electric conductivity and texture were determined with atomic absorption in ammonia acetate (pH= 7) and Bouyoco's densimeter method, respectively (op.cit.).

Vegetation belts
The relative abundance of the different life-forms at each elevation was used (sensu Squeo et al. 1993) to determine the limits vegetation belts (physiognomic approach).
In the floristic approach we used both a qualitative and a quantitative method (cf., Krebs 1989).For the qualitative analysis, we built a matrix with presence and absence of species at each elevation.We calculated the similarity index proposed by Baroni-Urbani & Buser (1976) for each pair of elevations.We chose this index because it can be easily tested for statistical significance.Afterwards with the similarity index we built a dendrogram using an UPGMA-type algorithm.To determine clusters that are statistically significant we used the method described by McCoy et al. ( 1986).First we calculated the significance of similarities using the table provided in Baroni-Urbani & Buser (1976).Then we transformed the similarity matrix into a matrix of statistical significance in which we replaced the actual values by"+","-" and "0" signs, according to whether the value of the index is significantly higher than expected at random, significantly lower, or not significantly different, respectively.Afterwards, at every fork we tested the significance of the segregation between groups with an analysis of independence that compared the distribution of pluses and minuses between the elevations that are clustered at any particular fork.In those forks that showed significant differences in the independence test, the homogeneity of each group was tested throughout the parameter DW (see McCoy et al. 1986 for details).We considered a group of altitudes as a significant cluster if the independence showed significant differences and DW was positive.
In the quantitative analysis we carried out an analysis of similarity using both species composition and relative cover of each species at each altitude.The Pears on's correlation index was used as a distance index for the analysis of similarity (Krebs 1989).Correlation coefficients are completely insensitive to proportional differences between communities (op.cit.).With the correlation matrix we built a dendrogram using an UPGMAtype algorithm with SYSTAT 6.0, which re-scales data, calculating and plotting 1-r, where r is the correlation coefficient.The statistical significance of the resulting clustering was established calculating the distance 1-rc, where re is the critical value of the correlation index at 95% of confidence.

Gradient analysis
As a general approach to relating the floristic and structural variation of the vegetation with environmental variation a principal component analysis (PCA), as included in the SYST AT 6.0 package, was applied using plant cover as the primary variable.In order to detect environmental factors involved in the altitudinal delimitation of vegetation, the first two unrotated factors extracted with the PCA were multiple-regressed against the following environmental variables: mean annual temperature, slope inclination, nitrogen content of soil, phosphorous content of soil, pH, electric conductivity of soil, percentage of clay in soil and percentage of rocks in transects (Table 1).
As an overall trend, species richness significantly decreased with elevation (r = 0.82, n = 17, P < 0.05) (Fig. 2a).However, species richness initially increased from 33 at 2100 m to 39 species at 2400 m.A decrease to 27 species was observed at 2500 m, followed with a rise to 45 species at 2700 m.From this altitude upwards, species richness decreased monotonically up to 3700 m where only one species (Bromus setifolius) was found (Fig. 2a).This indicates a loss of ea. 5 species for each 100 m increase in elevation.
Vegetation cover also significantly decreased with elevation (r = 0.95, n = 17, P < 0.05) (Fig. 2b ).Between 2100 and 2900 m vegetation cover averaged around 50% with a maximum value of 66% at 2400, decreasing abruptly toward higher elevations (Fig. 2b ).Dominance of plant life-forms varied with elevation (Fig. 3).Shrubs were the dominant lifeform from 2100 to 2500 m, decreasing in importance at higher elevations, although a secondary peak was observed at 2700 m (Fig. 3).A clear replacement of shrubs by cushion plants was observed at 2500 masl.Cushion plants reached maximum cover at 2500-2900 m, with a second peak at 3100-3400 m (Fig. 3 ).Perennial herbs showed high cover throughout all the altitudinal transect, being the dominant life-form between 2500-3300 m, and the only life form found from 3500 up to 3700 m.Annual herbs, the life form with the narrowest elevation range, were found between only 2100 and 2500 m, with low relative abundance (Fig. 3 ).

Vegetation belts
According to the distribution of life forms, physiognomic limits for the three Andean vegetation belts could be defined as follow: the Variaci6n altitudinal de la riqueza de especies (a) y cobertura vegetal (b) entre Ios 2.100 a 3.700 m de altitud en laderas de exposici6n sur en Ios Andes de Santiago (33°S).Barras error representan ± 2 EE.
subandean belt: 2100 -2500 m, characterized by the dominance of shrubs and the presence of annuals (Fig. 3).The lower Andean belt: 2600 -3400 m, characterized mainly by the abundance of cushion plants.The higher Andean belt: 3500 -3700 m with dominance of perennial herbs (Fig. 3).
Figure 4a shows the dendrogram resulting from applying the UPGMA on the Baroni-Urbani & Buser similarity index matrix.According to the McCoy et al. (1986) method there were 4 significant clusters, with 3600 and 3700 elevations segregating individually.Cluster 1 encompassed 2100 -2500 m, being characterized by the presence of the shrubs Chuquiraga oppositifolia, Ephedra chilensis, Acaena andina y A. splendens; the annual species Collomia biflora, Leucheria congesta and Chaetanthera spp.; and the perennial herbs Euphorbia collina, Montiopsis umbellata and M. andicola.Cluster 2 included 2600 -2800 m characterized by the presence of the cushion-plants Laretia acaulis and Anarthrophyllum gayanum, and the perennial herbs Nassauvia aculeata, Plantago barbata, Hypochaeris tenuifolia, Calopappus acerosus, Montiopsis cistiflora and Tristagma spp.Cluster 3 included elevations 2900-3100 m characterized by the presence of the cushionplant Azorella monantha and the perennial herbs Nassauvia pyramidalis and Senecio francisci.Cluster 4 comprised elevations 3200 -3500 m, characterized by the presence of the cushion Azorella madreporica and the perennial herbs Nastanthus caespitosus, Oxalis compacta and Viola philippii.The highest two altitudinal stations (3600 and 3700 m), with the presence of single species (Barneoudia major and Bromus setifolius, respectively) segregated early in the cluster analysis.

Gradient analysis
The ordination of each elevation with the 2 principal factors extracted with PCA is shown in Fig. 5.The 2 principal factors accounted for 39.6% of the total variance.Factor I accounted for 21.9% while factor 2 accounted for 17. 7%.According to a multiple regression analysis factor 1 was significantly correlated with mean annual temperature (Table 3 ), while factor 2 was significantly correlated with nitrogen content of soil (Table 3).Squeo (1991) reported 99 species for the alpine vegetation of Cerro Diente in the Chilean Patagonia.These numbers compare, for example, with the 332 species found in the alpine vegetation of 2 volcanoes in Mexico (Veilizquez 1994 ), or the 294 vascular plant species found for the alpine vegetation in the southern Appalachian mountains (Wiser et al. 1996).
The xeric conditions that characterize the Andes of central Chile and Argentina (Villagran et al. 1983) explain, in part, the lower species richness and vegetation cover (that never exceeded 50%) in comparison with other mountains of the world (Villagran et al. 1981, Arroyo et al. 1988).Shrubs and annuals were restricted to lower elevations (Fig. 3 ), suggesting that those lifeforms are more vulnerable to the colder climatic conditions of higher elevations.The height of plant, or how close to the soil the plant grows, is a very important factor for survival in cold habitats (Korner & Larcher 1988).In cold windy climates, small plants maintain higher temperatures than the surrounding air (Cabrera et al. 1998, Cavieres et al. 1998), enabling them to maintain a metabolic activity irrespective of a low ambient temperature.This may explain the restriction of shrubs to relatively low elevations while cushions and rosette perennial herbs, which grow close to the ground, are dominants at higher elevations.Additionally, Squeo et al. ( 1991Squeo et al. ( , 1996) ) reported that the decrease in plant heights with elevation is closely related with the capacity to tolerate low temperatures.While tall shrubs avoided freezing temperatures, mainly through super-cooling, small plant tended to tolerate such condition.Recently, Rada et al. ( 1999) suggested that limitation in growth rate due to low temperature and water availability could also be involved in the altitudinal zonation of life-forms.The decrease in shrub cover with elevation would also be related with the fact that long-lasting snow at higher elevations reduces the growing season and permafrost impedes root growth, both factors affecting secondary growth (Squeo et al. 1996).
The restriction of annuals to lower elevations has been reported for other alpine zones (Billings &Mooney 1968,Bliss 1971, Villagninetal.1981).Annuals require completing their life cycle within a single growing season.Favorable period for growth decrease with elevation (Kudo 1991, Stanton et al. 1994 ), thus annuals could not complete their life cycle at higher elevations.Furthermore, at higher elevations germination and establishment of seedling of annual plants could be inhibited strongly due to repeated freeze-thought events in soil (Callaghan & Emmanuelsson 1985).
The absolute dominance of perennial herbs at higher elevations suggests that a very limited range of life forms is viable on sites with short snow-free period and freezing temperatures.The low stature, the capacity to tolerate freezing temperatures, and the possession of underground perennial organs (such as rhizomes and bulbs) that help in maintaining the plant alive under the snow, allow them to take advantage of the short snow-free period (Billings & Mooney 1968).
According to the physiognomic criterion, limits for the subandean belt were 2100 -2500 m.In the floristic approach, both qualitative and quantitative cluster analysis joined elevations from 2100 to 2500 m (Figs. 5 and 6).These clusters were characterized by the presence and dominance of prostrate shrubs and annual species, which corresponded to the definition of the subandean belt, thus there were a complete agreement between both approaches.According to physiognomy, the lower Andean belt (the cushion's belt) extended from 2600 to 3400 m.In the qualitative floristic method, 3 clusters (including elevation 3500 m) were recognized (Fig. 5), while in the quantitative method only 2 clusters were found (Fig. 6).These differences could be related with the fact that in the qualitative floristic method each cluster was characterized by the presence of cushions (Laretia acaulis 2600-2800 m; Azorella monantha 2900-3100 m; A. madreporica: 3200-3500 m), however the dominance of other species (e.g., Phacelia secunda, Senecio spp.) between 2900-3400 m excluded elevation 3500 m in the quantitative methods irrespective of the replacement of Azorella monantha by A. madreporica.Thus, the lower Andean belt can be divided in 2-3 sub-belts depending on the floristic method used.A similar result was reported by Squeo et al. (1993), who also divided this belt in 2 sub-belt depending on the dominant cushion species.This suggests that delimitation of alpine vegetation based only on a physiognomic criterion lose relevant information about the distribution of cushion species.
According to physiognomy limits for the higher Andean belt, characterized by the dominance of perennials herbs, were 3500-3700 m.In the quantitative analysis all these elevations segregate in separate ways because they have no species in common, however in all cases the dominant species was a perennial herb.Thus, there is a complete correspondence between both types of delimitation.However, in the qualitative analysis, the presence (although with a very low cover) of Azorella madreporica at 3500 m joined this elevation with the cluster characterized by the presence of cushion plants, irrespective of the dominance of perennial herbs.
It has been reported that diversity and productivity on mountains decrease with elevation (MacArhtur 1972), although there are mountains were extreme aridity at lower elevations override this general trend (Villagnin et al. 1983, Arroyo et al. 1988).The major role of altitude and associated low temperatures on vegetation in mountains has been largely recognized (e.g., Ward & Dimitri 1966, Troll 1973, Fernandez-Palacios & Nicolas 1995, Ferreyra et al. 1998).The results of the present study seem to confirm this rule, because mean annual temperature showed a significant correlation with factor 1 in the PCA analysis.Although, there is a difference around 1 0°C in mean annual temperature between 2100-3600 m, according to the observed seasonal variation in lapse-rate (Cavieres & Arroyo 1999) the difference in temperature is higher (ea., 12 °C) during the growing season.Such a steep decrease in temperature could explain in part the abrupt decrease in species richness reported here (Fig. 2).
Many studies have shown that in addition to temperature, variables related to the substrate affect plant distribution in high mountain systems (Kitayama 1992, Squeo et al. 1993, Velazquez 1994, Wiser et al. 1996).In this study we found that the nitrogen content of soil, is the second most relevant factor that determined the altitudinal distribution of vegetation in the Andes of central Chile.Kitayama (1992) found that in addition to nitrogen, phosphorous and exchangeable magnesium and potassium were the properties of soil that determined the altitudinal distribution of vegetation in Mount Kinabalu, Borneo.Wiser et al. ( 1996) reported that pH, and exchangeable cations were determinants on distribution of vegetation in the southern Appalachian mountains.Squeo et al. (1993) found that in the Andes of the desert zone of Chile, the altitudinal zonation of vegetation produced by thermic and radiation gradients is further modified by the chemical characteristic of soil mentioning pH, manganese, cooper, zinc, iron, calcium, etc. but failing to establish what chemical element is the main determinant of vegetation changes.An additional property of soil that has been reported as relevant for explain the distribution of vegetation is soil moisture (Velazquez 1994, Wiser et al. 1996, Ferreyra et al. 1998), especially in mountains with mediterranean-type climates where xeric conditions develop during the growing season (Morefield 1992).
In summary, altitudinal delimitation of alpine vegetation based on physiognomic methods, while fast and easy to apply, it is highly subjective and loses relevant information about species distribution, especially at higher elevations.On the other hand, methods based on species composition and relative abundance of each species, while more time consuming, they are more detailed and permits the detection of physical factors involved in the altitudinal delimitation of alpine vegetation.

Fig. 1 .
Fig. 1.Map of the study zone showing the elevations where vegetation was sampled.Mapa de la zona de estudio donde se indican !as altitudes de muestreo de la vegetaci6n.

Fig. 2 .
Fig. 2. Altitudinal variation of species richness (a) and vegetation cover (b) from 2100 to 3700 m elevation in south-facing slopes in the Andes of Santiago (33°S).Error bar represent± 2 SE.

Fig. 3 .
Fig. 3. Altitudinal variation of different life-forms cover from 2100 to 3700 m elevation in the Andes of Santiago (33°S).

Fig. 5 .Fig. 4 .
Fig. 5. Ordination of elevations along the two principal components extracted with a PCA for the relative cover of each species from 2100 to 3700 m elevation in the Andes of Santiago (33°S).Ordenaci6n de !as altitudes con Ios dos componentes principales obtenidos con un PCA para la cobertura reiativa de cada especie desde Ios 2.100 a 3.700 m en Ios Andes de Santiago (33°S).

TABLE 3
Multiple-regression summary for"principal factors extracted with PCA for the relative cover of each species from 2100 to 3700 m elevation in the Andes of Santiago (33°S).*Indicate significant regression (P < 0.05)