INTRODUCTION
This study focuses on the response of plant functional types to environmental gradients. Plant functional types (hereafter referred to as PFTs) represent sets of species that demonstrate either a similar response to the environment or have similar effects on major ecosystem processes (Gitay & Noble Reference GITAY, NOBLE, Smith, Shugart and Woodward1997). Therefore, they can be regarded as a powerful proxy for ecological mechanisms.
Due to the abstract nature of the classified units, this concept enables comparisons across ecosystems, landscapes and even biomes that differ in species composition.
PFTs are grouped according to their morphological, physiological, regenerative or phenological features, qualities or traits, respectively (Lavorel & Garnier Reference LAVOREL and GARNIER2002).
Functional response traits characterise species’ responses to a given environment or disturbance regime (Lavorel & Garnier Reference LAVOREL and GARNIER2002). They are linked with soil processes, water availability, climatic conditions, disturbances and other resources or impacts (Diaz et al. Reference DIAZ, LAVOREL, McINTYRE, FALCZUK, CASANOVES, MILCHUNAS, SKARPE, RUSCH, STERNBERG, NOY-MEIR, LANDSBERG, ZAHNG, CLARK and CAMPBELL2007).
From the functional perspective, which is always reductionist to some extent, growth forms provide essential information for instance on the strategy of plants regarding the use of energy and water (Baldocchi et al. Reference BALDOCCHI, XU and KING2004), carbon storage (Diaz & Cabido Reference DIAZ and CABIDO1997) and the support of static body weight against wind (Rowe & Speck Reference ROWE and SPECK2005). In semi-arid environments, where water is a limiting factor, plants develop morphological features to reduce heat loading in the leaves through reflection and the absorption of solar radiation, and water-storage organs to buffer the effects of drought stress. In addition, morphological features such as thorns and spines serve as defence organs that deter herbivores (Cornelissen et al. Reference CORNELISSEN, LAVOREL, GARNIER, DIAZ, BUCHMANN, GURVICH, REICH, TER STEEGE, MORGAN, Van Der HEIJDEN, PAUSAS and POORTER2003, Roininen et al. Reference ROININEN, VETELI, PIIROINEN, Varmola, Valkonen and Tapaniene2007). Similarly, the life span (annual vs. perennial) of herbs and graminoids reflects the temporal limitations of resource availability, resulting in a relatively large investment of annuals into reproductive traits compared to vegetative ones (Diaz & Cabido Reference DIAZ and CABIDO1997, Silvertown Reference SILVERTOWN2004).
Only a few studies have investigated relationships between PFTs and environmental gradients in the African tropics (e.g. Skarpe Reference SKARPE1996). Most of the available studies were conducted in temperate climates (de Bello et al. Reference DE BELLO, LEPS and SEBASTIA2005, Reference DE BELLO, LEPS and SEBASTIA2006, Diaz et al. Reference DIAZ, CABIDO, ZAK, MARTINEZ CARRETERO and ARANÍBAR1999, Stevens et al. Reference STEVENS, COX, STRAUSS and WILLIG2003) and the neotropics (Condit et al. Reference CONDIT, HUBBELL and FOSTER1996, Kraft et al. Reference KRAFT, VALENCIA and ACKERLY2008). Our study aims to investigate the response of PFTs along altitudinal gradients and to examine the linkages between PFTs and measured environmental variables. In the face of climate change and the discussion on upward shifts, research on the ecological traits of altitudinal gradients is particularly required in sensitive tropical ecosystems.
Generally, (1) spatial (structural, morphological), (2) temporal (e.g. seasonal or life cycle related) and functional qualities (mechanism-oriented) can be distinguished. The latter can be subdivided into (3) traits that reflect biotic interactions (such as herbivory) and environmental stresses (such as drought and cold) and (4) traits that represent adaptations to the availability of abiotic resources (such as nutrients and water).
We hypothesize that the relative abundance of PFTs responds specifically to altitude. Altitude represents a composite gradient of several climatic variables, which are closely linked to a variety of environmental factors. Here, we want to identify those aspects in the functioning of vegetation that show the strongest responses along altitudinal gradients. (1) Structural attributes and PFTs: growth forms are mainly an adaptation to climatic constraints. We expect the relative abundance of woody species exhibits to be positively correlated with altitude because of climatic constraints at lower elevations. However, gradients and responses can be non-linear for instance as a consequence of frost, cloudiness and precipitation. Herbaceous species and graminoids are not that directly dependent on climatic conditions and should more typically display a correlation to slope inclination and/or to soil nutrient availability than to altitude. (2) Temporal attributes and PFTs: here, we expect the relative abundance of deciduous woody species to be higher in semi-arid i.e. at lower elevation zones, where seasonal rhythms are more pronounced, while the relative abundance of evergreen woody species should increase with altitude, where rainfall occurs throughout the year. Likewise, the relative abundance of short-lived herbaceous species is expected to show a correlation to slope and aspect as a consequence of seasonal patterns in moisture and nutrient availability. (3) Attributes and PFTs related to biotic interactions (herbivory) and stress: we expect the relative abundance of plants with thorns or spines and those that form dense clonal tussocks to be higher at lower altitudes in response to drought stress; whereas the relative abundance of rosettes and rhizomes should be higher at higher altitudes in response to stress from the cold. (4) Attributes and PFTs related to the availability of water and nutrients: In this category we expect a positive correlation between N-fixers, surface stone cover and slope because these site conditions are directly related to soil moisture; whereas succulent species should be more abundant at lower altitudes, where water shortage occurs.
METHODS
Study area
The study area is located at 5°42′–6° 20′N, 37°17′–37°59′E (Figure 1), about 500 km south of the Ethiopian capital, Addis Ababa. The landscape is characterized by a great physiographic diversity that is related to the tertiary volcanic and tectonic processes (Mohr Reference MOHR1971). It consists of plateaux, escarpments, a block mountain and the Rift Valley where Lakes Abaya and Lake Chamo are located (Wana & Beierkuhnlein Reference WANA and BEIERKUHNLEIN2010).
Figure 1. Map of the study area in the southern part of the Ethiopian rift valley with locations of plots for data collection.
Rainfall in the study area shows a pronounced bimodal seasonal distribution. The mean annual rainfall recorded for Arbaminch, located at 1200 m, was 888 mm in the period from 1987 to 2005, whereas it was 1235 mm for Chencha (2700 m) for the period from 1972 to 1980 and 1990 to 2004 (data is missing between 1981 and 1989) (Wana & Beierkuhnlein Reference WANA and BEIERKUHNLEIN2010). The main rainy season, which accounts for about 40% of the mean annual rainfall, occurs from April to June, while the ‘small’ rainy season is in September and October.
According to the atlas of the potential vegetation of Ethiopia and Eritrea, the vegetation of the Rift Valley part of the study area belongs to the Acacia–Commiphora woodland and bushland proper (ACB) whereas the highlands of Gughe and Amaro mountains belong to the dry evergreen Afromontane forest (DAF) (Friis et al. Reference FRIIS, SEBSEBE and BREUGEL2010).
Sampling scheme
The sampling design was based on the relative distribution of the area along altitudinal gradients. We consulted Shuttle Radar Topography Mission (SRTM) digital elevation data sources to establish a digital elevation model for the study area. The digital elevation data were made available by the International Centre for Tropical Agriculture (http://srtm.csi.cgiar.org). The spatial classification of elevation at the interval of 200 m was extracted from the DEM. The number of plots was then selected depending on the relative distribution of elevation for representative sampling following a similar pattern of elevation distribution along the Gughe-Amaro Mountains (detailed description of the sampling scheme was provided in Wana & Beierkuhnlein Reference WANA and BEIERKUHNLEIN2010).
Vegetation and environmental data
Fieldwork was carried out in two phases, November 2006 to January 2007, and in December 2007, respectively. Both phases of fieldwork were conducted after the ‘small’ rainy season. In total 74 plots, each with a size of 20 × 20 m, were established (Figure 1), where environmental data, plant species and presence-absence information was recorded. We encountered a gamma diversity of 475 plant species on our plots (Appendix 1). For those species that could not be identified precisely in the field, specimens were taken to the National Herbarium at the Addis Ababa University for identification or confirmation. Functional traits were recorded for every species, whenever possible directly in the field for the following traits: thorns/spines, rosettes, legumes (N-fixing), and succulence. Additionally, plant functional types regarding the life span for herbs and graminoids (annuals vs. perennials), leaf phenology (deciduous vs. evergreen) for woody species, the presence of rhizomes/stolons and potential N-fixation (legumes) were checked in the published flora volumes of Ethiopia and Eritrea (Edwards et al. Reference EDWARDS, MESFIN and HEDBERG1995, Reference EDWARDS, SEBSEBE and HEDBERG1997, Reference EDWARDS, MESFIN, SEBSEBE and HEDBERG2000; Hedberg & Edwards Reference HEDBERG and EDWARDS1989, Reference HEDBERG and EDWARDS1995; Hedberg et al. Reference HEDBERG, EDWARDS and SILESHI2003, Reference HEDBERG, FRIIS and EDWARDS2004).
Altitude and aspect were recorded using a Garmin GPS 3.1. The slope was recorded using a clinometer. Aspects were coded prior to analysis following Zerihun et al. (Reference ZERIHUN, FEOLI and LISANEWORK1989). For stone cover, plots were divided into four subquadrats and it was visually estimated and averaged to yield the percentage of stone cover in a plot. Five soil samples were collected at 0–30 cm depth from each corner and from the centre of the plot and mixed for a composite soil sample.
Classification of plant functional types
Plant functional types were classified according to four categories representing structural traits (1), temporal traits (2), adaptations to biotic interactions and stress (3), and adaptations to abiotic resource availability (4). (1) The first category was based on traits of the whole plant. Here, we found mainly adaptation to climatic constraints and competition, resistance to strong winds and support of static weight such as self-support or climbing (Rowe & Speck Reference ROWE and SPECK2005). We differentiated between the growth forms of woody plants, herbs, graminoids, climbers and ferns. Epiphytes were excluded from the analysis because of rarity. Ferns are primarily a taxonomic category. However, we regarded ferns as a separate functional group due to the specific restrictions in water regulation. Ferns cannot respond to water shortage by stomatal regulation. (2) The second category was related to the fluxes of resources and how species respond to temporal changes in environmental conditions (e.g. seasonality). In this category woody species were subdivided into deciduous vs. evergreens while herbs and graminoids were subdivided into annuals vs. perennials. (3) The third group was based mainly on the specific response to physical and biotic stresses. The physical stresses that were expected in our system included the cold (low temperature in high-altitude areas) and drought in the semi-arid zones along the lower-elevation zones while biotic stresses included herbivory and the predation of seeds and fruits, which reduces reproductive success and has an impact on population demography. (4) Finally, traits related to the access and storage of resources such as water and nutrients were applied and species were attributed to the groups of hemi-parasites, succulents and nitrogen fixers (Table 1).
Table 1. Classification of plant functional types (PFTs) and number of species per group.
1 Three species were counted double (one for thorns/spines and two species for rosette and rhizomes) thus the total number of species for this category yields 478 (the total number of species encountered in the field were 475).
Before analysing the data, for each PFT the presence-absence data of species were pooled in order to generate the total number of species for the corresponding PFT within each plot and then the proportion (%) of each PFT was calculated from the total number of species in a plot. Thus, abundance is defined here as the proportion (percentage) of a given plant functional type from the total number of species belonging to all PFTs in a plot.
Soil data
The soil samples were analysed for soil organic matter and then converted to soil organic carbon by dividing the percentage soil organic matter by 1.72 (SOC) (Schumacher Reference SCHUMACHER2002). Texture, total nitrogen (TN), and available phosphorus (AP) were analysed in the analytical service laboratory of the International Livestock Research Institute (ILRI) in Addis Ababa, and the second batch of soil samples were analysed in the Debre Zeit Plant and Soil Laboratory of the Ethiopian Agricultural Research Organization. Nitrogen was analysed using the Kjeldahl method, phosphorus by the Bray method, texture using the hydrometer method, and organic matter by following the Walkey and Black wet-oxidation method.
Statistics
The PFTs and most environmental data were transformed into logarithmic (log(x+1)) scale for normality. However, soil sand and silt fraction were normally distributed and therefore not transformed. A constrained linear ordination technique (RDA) was used to investigate the relationship of PFTs with environmental factors using the software CANOCO 4.5 (ter Braak & Smilauer Reference TER BRAAK and SMILAUER2002). Forward automatic selection and partial Monte Carlo permutation was used to test the significance of the relationship between PFTs and environmental variables.
RESULTS
PFT-environment linkages
The measured environmental variables explained relatively large proportions of variance for temporal attributes (44.7%) and growth forms (37.4%) from the total inertia in the PFT data (Table 2). However, the total variance explained for PFTs related to biotic interactions and resource availability (water/nutrient) was relatively small (Table 2). The amounts of explained variance shared by the first two axes for biotic interactions (herbivory) and stress and abiotic site conditions (water/nutrients) were 24% and 26%, respectively (Table 2).
Table 2. Constrained linear ordination analysis (canonical RDA) of plant functional types and the amount of variance explained (species data were log(x+1)-transformed, scaling with inter-species correlations and standardized by species centring), PFTs refers to plant functional types.
The first axes for all categories of PFT data were strongly correlated with altitude and soil organic carbon. These two environmental variables were intercorrelated (r = 0.58, results not shown). On the other hand, soil sand fraction was strongly correlated with the second axes for all categories of PFTs excluding growth forms. Slope inclination was strongly correlated with the second axis for the growth form based classification.
Forward selection of environmental variables showed that altitude was found to have a significant effect (P = 0.002) on the relative abundance of plant functional types for all categories (Table 3). Slope inclination was significantly related to growth form (P = 0.01) and temporal attributes (P = 0.01). Similarly, SOC was found to be significantly related to the relative abundance of growth forms (P = 0.002) and temporal attributes (P = 0.01) (Table 3). However, soil sand fraction only had a significant effect on the temporal attributes (P = 0.006) while surface stone cover was significantly (P = 0.01) related to the relative abundance of PFTs attributed to biotic interactions and stress (Table 3).
Table 3. The relationships among environmental variables and plant functional types based on Partial Monte Carlo permutation tests (** = significant at P ≤ 0.01, * = significant at P < 0.05); SOC stands for soil organic carbon, Alt = altitude, Stone = surface stone cover, C:N = carbon:nitrogen ratio, AP = available phosphorus).
Graminoids (grasses and sedges) were highly negatively correlated with SOC but were positively related to available phosphorus (Figure 2a). Herbs were strongly related to slope inclination (Figure 2a) rather than to altitude in the study area. Tussock grasses and plants with thorns and spines were highly correlated to surface stone cover and available phosphorus and negatively related to altitude (Figure 2b). Rhizomatous plants and rosette-forming herbs were found to be more abundant in areas of higher altitude, with high soil sand content and a low amount of soil available phosphorus (Figure 2b).
Figure 2. A redundancy analysis of plant functional types and environmental variables; growth forms (a), biotic interactions/stress (b), temporal attributes (c) and nutrient- and water-related plant functional types (d); SOC = soil organic carbon, AP = available phosphorus, Alt = Altitude, Stone = surface stone cover.
With regard to the abundance of PFTs related to temporal attributes (leaf seasonality and life span), the abundance of woody species was clearly differentiated between higher and lower altitudes (Figure 2c). Deciduous woody species were more abundant at lower altitudes while evergreen woody species were abundant in the higher elevation zones (Figure 2c). Deciduous woody species showed a strong negative relationship with soil organic carbon and were positively correlated with surface stone cover, whereas evergreen woody species were positively correlated with soil sand fraction and SOC (Figure 2c).
Annuals and perennials (graminoids and herbs) were strongly correlated with aspect, slope inclination, and soil silt fraction. However, the pattern of abundance of annuals and perennials was not clearly differentiated probably due to the aggregation of comparatively large groups such as graminoids and herbs. With respect to PFTs related to nutrients and water acquisitions: nitrogen fixers were strongly correlated with higher surface stone cover and C/N-ratio whereas succulents showed a strong positive relationship with available phosphorus in the soil (Figure 2d). Hemi-parasites, however, did not appear to have any relationship with the measured environmental variables
DISCUSSION
In this study we find that topographic factors such as altitude, inclination and aspect play an important role in differentiating the relative abundance of PFTs across the landscape. Furthermore, it is the content of soil organic carbon that explains patterns in functional types. These parameters do not have a direct influence as such. The relief influences the temperature, insolation and water regimes, whereas organic carbon reflects nutrient availability and nutrient exchange capacity but also the legacy of a site in terms of its age and maturity.
The influence of the topographic relief is apparent in water and solar energy distributions across spatial gradients, which in turn control photosynthetic processes (productivity). Understandably, the diversity of woody species is closely related to productivity (O'Brien Reference O'BRIEN1998, O'Brien et al. Reference O'BRIEN, FIELD and WHITTAKER2000). Altitude can be considered to be a good proxy variable for the distribution of water and energy in high-mountain regions.
The topographic heterogeneity of slopes and exposure (aspect) also affects the (re)distribution of vital resources such as water and solar energy. This is supported by the strong explanatory effect of slope inclination for growth forms and temporal attributes.
In mountain regions and in semi-arid ecosystems slope inclination and slope exposure (aspect) commonly play an important role in structuring the relative abundance of PFTs since the redistribution of water and nutrients is affected by the relative slope position (Breckle Reference BRECKLE2002). In the face of climatic changes, it would be rather naïve only to consider the altitudinal gradients in montane ecosystems. The small-scale heterogeneity of the microclimate that is a consequence of slope and aspect has to be considered as well and can provide options for habitat shift at the same elevational zone. In the investigated mountain region, slope inclination and aspect are important factors that explain patterns of species distribution and diversity (Wana & Beierkuhnlein Reference WANA and BEIERKUHNLEIN2010). Their role in the semi-arid environment is very critical in forming patchy vegetation and barren lands which in turn affects the run-on/run-off processes, soil depth, water retention and nutrient status of the soil (Pueyo & Alados Reference PUEYO and ALADOS2007, Wilkinson & Humphreys Reference WILKINSON and HUMPHREYS2006). Slope aspect affects productivity and plant species composition, respectively (Gong et al. Reference GONG, BRUECK, GIESE, ZHANG, SATTELMACHER and LIN2008, Wana & Beierkuhnlein Reference WANA and BEIERKUHNLEIN2010). Teshome et al. (Reference TESHOME, DEMEL and SEBSEBE2004) reported from the Gamo Gofa region, south of the current study area, a low cover of herbaceous vegetation in the steep-slope area dominated by a Commiphora cyclophylla plant community. Our study, however, showed a strongly positive correlation of herb abundance with slope inclination.
Depending on seasonal availability, plants modify rooting depth and resource segregation in order to access water and nutrients in the soil (Silvertown Reference SILVERTOWN2004). In our study graminoids (grasses and sedges) show strong negative correlations with SOC. Apparently, this is not a direct correlation because the largest source for SOC is litter-fall from trees, while soil organic matter (and hence SOC) replenishment in grass-dominated ecosystems mainly depends on the die-back of grass roots as a source of organic matter (Troeh & Thompson Reference TROEH and THOMPSON2005). Graminoids show a strong positive correlation to AP in the soil at the lower elevation range.
PFTs related to biotic interactions (herbivory) and stresses (drought and cold) are clearly separated along altitudinal gradients. Tussock grasses and species with thorns/spines were abundant at lower elevation where drought is expected to be a limiting factor while rhizomes and rosettes were found to be abundant in the higher elevation zones. The combination of drought stress with grazing tends to favour tussock plants and plants with thorns or spines while the combination of the cold and grazing appears to favour plants with rhizomes or rosettes.
In clonal grasses, tussock formation is an adaptation strategy to environmental stress and disturbance such as fire and grazing. In semi-arid environments tussocks enhance survival by forming organic debris, fine soils and through retaining soil moisture (Pugnaire & Haase Reference PUGNAIRE and HAASE1996). In addition, they buffer soil erosion along slopes. Thorns and spines help to reduce heat or drought stress (Cornelissen et al. Reference CORNELISSEN, LAVOREL, GARNIER, DIAZ, BUCHMANN, GURVICH, REICH, TER STEEGE, MORGAN, Van Der HEIJDEN, PAUSAS and POORTER2003) by dissipating heat loading on the surface of the leaves and stems or absorbing solar radiation in addition to their role as a mechanical deterrent against vertebrate herbivory. Rhizomes and stoloniferous traits are features that are presumed to be related to reproductive success. They may help organisms to buffer the effect from an unexpected occurrence of drought or frost (Diaz & Cabido Reference DIAZ and CABIDO1997). If not destroyed, rhizomes deliver the competitive advantage to re-occupy space after disturbance events (e.g. grazing or mowing) because of their ability to store metabolic products and therefore quickly resprout (Klimesova et al. Reference KLIMESOVA, LATZEL, DE BELLO and VAN GROENENDAEL2008).
The abundance of deciduous woody species in the lower altitudes appears to reflect the response to drought. Leaf shedding is a strategy to reduce moisture loss in semi-arid environments. Herbaceous species, both annuals and perennials, were strongly related to soil nutrients (C:N, AP). The availability of phosphorus and organic sources of nitrogen in the soil varies seasonally with soil organic matter turnover, temperature, pH and water availability in the soil (Troeh & Thompson Reference TROEH and THOMPSON2005).
The abundance of PFTs related to nutrient and water acquisition was significantly related to altitude, which was highly correlated with the first RDA axis. Succulents tended to be associated with a sandy soil texture and with high values of AP in the soil. Due to their capacity to store water in the tissue, they are able to buffer the effect of seasonal water shortage in the lower elevation zones. Nitrogen fixers are less dependent on the organic turnover of nitrogen and are even negatively correlated to SOC. The strong positive relationship of nitrogen fixers to surface stone cover may be due to the availability of moisture in the matrix, which reduces moisture loss through evaporation.
Topographic attributes (altitude and slope) and SOC play an important role in differentiating the relative abundance of PFTs in the investigation area. The categorization of PFTs reveals the differential response of plants to the environmental gradients that structure the vegetation in a landscape.
Further attempts to characterize PFTs and their response to environmental gradients should include traits related to fire, water harvesting (e.g. rooting depth), degree of ramification, and drought-avoidance strategies. The occurrence and the relative abundances of PFTs are directly linked to ecosystem processes, with the datasets at the same time being coarse enough to capture the patterns and processes at a large scale (Beierkuhnlein & Jentsch Reference BEIERKUHNLEIN, JENTSCH and Henry2005). Thus, PFTs aggregate a variety of species responses to the environment and are powerful input data for global land-cover modelling.
ACKNOWLEDGEMENTS
We are very grateful to the three anonymous reviewers for critical comments on an earlier version of the manuscript. We would like to thank the National Herbarium (AAU) and its staff for providing access to the Herbarium and for their kind assistance. We thank Florian Fritzsche and Volker Audorff for comments on an earlier draft. This research was supported by DAAD and the School of Graduate Studies at the University of Addis Ababa. Axel Bedouin has partly supported the soil laboratory analysis through the NUFU project.
Appendix 1. Species and plant functional types. Abrevations: Cl = Climber, Rh = Rhizomes/stolons, Nf = Nitrogen fixers, An = Annuals, He = Herbs, Ro = Rosettes, Pa = Parasites, De = Deciduous, Wo = Woody, Th = Thorns/spines, Su = Succulents, Ev = Evergreens, Fe = Ferns, Tu = Tussocks, Pe = Perennials, Gr = Graminoids, Na = No adaptation/does not fit into the specified strategies.
