Can grazing response of herbaceous plants be predicted from simple vegetative traits?

Authors

  • Sandra Díaz,

    Corresponding author
    1. Instituto Multidisciplinario de Biología Vegetal (CONICET–UNC) and FCEFyN, Universidad Nacional de Córdoba, Casilla de Correo 495, Vélez Sársfield 299, 5000 Córdoba, Argentina; and *Department of Agricultural Botany, Faculty of Agricultural, Food and Environmental Quality Sciences, Hebrew University of Jerusalem, PO Box 12, Rehovot, 76100 Israel
    Search for more papers by this author
  • Imanuel Noy-Meir,

    1. Instituto Multidisciplinario de Biología Vegetal (CONICET–UNC) and FCEFyN, Universidad Nacional de Córdoba, Casilla de Correo 495, Vélez Sársfield 299, 5000 Córdoba, Argentina; and *Department of Agricultural Botany, Faculty of Agricultural, Food and Environmental Quality Sciences, Hebrew University of Jerusalem, PO Box 12, Rehovot, 76100 Israel
    Search for more papers by this author
  • and * Marcelo Cabido

    1. Instituto Multidisciplinario de Biología Vegetal (CONICET–UNC) and FCEFyN, Universidad Nacional de Córdoba, Casilla de Correo 495, Vélez Sársfield 299, 5000 Córdoba, Argentina; and *Department of Agricultural Botany, Faculty of Agricultural, Food and Environmental Quality Sciences, Hebrew University of Jerusalem, PO Box 12, Rehovot, 76100 Israel
    Search for more papers by this author

Sandra Díaz, Instituto Multidisciplinario de Biología Vegetal (CONICET–UNC) and FCEFyN, Universidad Nacional de Córdoba, Casilla de Correo 495, Vélez Sársfield 299, 5000 Córdoba, Argentina (fax 00 54 3514332104; e-mailsdiaz@com.uncor.edu).

Summary

  • 1 Range management is based on the response of plant species and communities to grazing intensity. The identification of easily measured plant functional traits that consistently predict grazing response in a wide spectrum of rangelands would be a major advance.
  • 2 Sets of species from temperate subhumid upland grasslands of Argentina and Israel, grazed by cattle, were analysed to find out whether: (i) plants with contrasting grazing responses differed in terms of easily measured vegetative and life-history traits; (ii) their grazing response could be predicted from those traits; (iii) these patterns differed between the two countries. Leaf mass, area, specific area (SLA) and toughness were measured on 83 Argentine and 19 Israeli species. Species were classified by grazing response (grazing-susceptible or grazing-resistant) and plant height (< or > 40 cm) as well as by life history (annual or perennial) and taxonomy (monocotyledon or dicotyledon).
  • 3 Similar plant traits were associated with a specific response to grazing in both Argentina and Israel. Grazing-resistant species were shorter in height, and had smaller, more tender, leaves, with higher SLA than grazing-susceptible species. Grazing resistance was associated with both avoidance traits (small height and leaf size) and tolerance traits (high SLA). Leaf toughness did not contribute to grazing resistance and may be related to selection for canopy dominance.
  • 4 Plant height was the best single predictor of grazing response, followed by leaf mass. The best prediction of species grazing response was achieved by combining plant height, life history and leaf mass. SLA was a comparatively poor predictor of grazing response.
  • 5 The ranges of plant traits, and some correlation patterns between them, differed markedly between species sets from Argentina and Israel. However, the significant relationships between plant traits and grazing response were maintained.
  • 6 The results of this exploratory study suggest that prediction of grazing responses on the basis of easily measured plant traits is feasible and consistent between similar grazing systems in different regions. The results challenge the precept that intense cattle grazing necessarily favours species with tough, unpalatable, leaves.

Introduction

Land use and management actions on plant communities induce responses and changes in the community. These vegetation responses, in turn, modify the economic and other benefits that can be obtained from the managed system. The explanation and prediction of plant community responses to land use and management factors is therefore a major objective in applied as well as theoretical ecology. The basic unit for which the response is usually observed is at the population or species level. However, results from studies of particular communities, if they consist simply of species lists with their associated responses, cannot be generalized or compared beyond the limits of the local situation. Comparison and integration of community responses observed in floristically distinct regions, and extrapolation to predict plant responses in new situations, require a transformation to a level more general than the species. In regions with a rich and imperfectly known flora, the species level may be problematic even for primary observations of responses. Therefore, there is a growing need to understand and predict plant responses to different land management factors in terms of plant traits that are easily measured and at the same time ecologically meaningful (Díaz & Cabido 1997; Lavorel et al. 1997; Westoby 1998; Hodgson et al. 1999; Weiher et al. 1999).

Grazing by domestic herbivores has both extensive and profound impacts on plant communities. Vegetation changes in response to grazing management decisions often modify primary plant and animal production and economic returns from the system, as well as other benefits like conservation value. The identification of plant types and traits that explain and predict the response of species and communities to grazing intensity is one of the main tools in management of grazing systems. Almost a century of empirical research in rangelands in different parts of the world has been dedicated to identifying the response of individual plant species to varying grazing intensities. For some ecosystems, in particular North American grasslands, the results from many studies have led to the formulation of generalizations on the plant types or plant traits associated with a negative or a positive response to grazing intensity (grazing decreasers or grazing increasers and invaders, respectively; Dyksterhuis 1949; Ellison 1960).

Generalizations across communities and continents on plant traits that can explain, or predict, responses of plant species to grazing intensity can be derived by theoretical considerations (Milchunas, Sala & Lauenroth 1988; Westoby 1999) and evaluated by comparing or integrating information from communities in different regions. The comparative approach requires measurement and recording of a common set of traits in different species, together with independent information on species responses to grazing. Recently, attention has focused on identifying a minimum set of the most important plant traits that must be considered in relation to grazing response (Landsberg, Lavorel & Stol 1999; Weiher et al. 1999; Westoby 1999).

Decreases in plant height and leaf size in response to grazing by ungulates have been repeatedly documented in the literature (Sala et al. 1986; McNaughton & Sabuni 1988; Noy-Meir, Gutman & Kaplan 1989; Díaz, Acosta & Cabido 1992; Landsberg, Lavorel & Stol 1999). Leaf toughness or resistance to tearing has been associated with low palatability and grazing avoidance (Coley, Bryant & Chapin 1985; Herms & Matson 1992; Grime et al. 1996; Cornelissen et al. 1999). In the last few years, the importance of specific leaf area as a trait that integrates plant investment into growth vs. defence and storage, and its value as an easily measured indicator of relative growth rate, have been highlighted by several authors (Garnier 1992; Lambers & Poorter 1992; Reich, Walters & Ellsworth 1992; Westoby 1998; Hodgson et al. 1999). Recently, Westoby (1999) has proposed a scheme in which specific leaf area and plant height play a major role in response to disturbance, in particular to grazing. According to his model, plants with high specific leaf area should be favoured under conditions of heavy, non-selective, grazing, whereas plants with low specific leaf area should predominate when stocking rate is moderate or low, thus allowing grazers to feed selectively.

Within this context, we analysed species sets from temperate, subhumid, upland grasslands of central Argentina and northern Israel in order to answer the following questions.

  • 1Do plants with contrasting responses to grazing differ in terms of easily measured traits, such as plant height, life history, leaf area, leaf mass, specific leaf area or leaf tensile strength?
  • 2What is the trait or combination of traits among those measured that can better explain and predict the grazing response of species in this set?
  • 3Are the answers to the first two questions similar in sets of species from two different floras, and in the combined species set?

Methods

Study sites

The Argentine study area was the upland grasslands of Sierras Grandes (Córdoba Mountains, central Argentina), developed mostly on granitic substrate at altitudes ranging from 1400 to > 2200 m a.s.l., with an annual rainfall ranging from c. 840 to c. 912 mm and an annual mean temperature ranging from c. 12 to c. 8 °C. The climate is temperate subhumid with a monsoonal rainfall distribution (Acosta et al. 1989). According to Díaz, Acosta & Cabido (1994) and Díaz et al. (1999), the region has a relatively long history of grazing by native (mostly Camelidae) and domestic ungulates, with moderate to high stocking rates of livestock of European origin during the last 400 years. At present the area is grazed by cattle and to a lesser degree by horses and sheep. Plant community composition and species responses to grazing have been studied extensively by Acosta et al. (1989), Cabido, Acosta & Díaz (1989), Díaz, Acosta & Cabido (1992, 1994), Pucheta, Díaz & Cabido (1992) and Pucheta et al. (1998).

The Israeli study site was Mediterranean subhumid grassland on stony basaltic soil near Almagor on the Korazim Plateau in Galilee, at an altitude of 100 m a.s.l. (300 m above the Sea of Galilee) The mean annual rainfall is c. 500 mm (Almagor rainfall station) and the mean annual temperature is 21 °C (Anonymous 1995). The region has a very long history of intense grazing by domestic herbivores (Noy-Meir & Seligman 1979; Perevolotsky & Seligman 1998). Domestic goats and sheep have been documented in this region from about 8500 bp and domestic cattle before 7000 bp (Smith 1995). At present, most of the area is grazed by cattle at moderate to high intensities. The responses to grazing of the plant community and of many individual species is fairly well known from previous studies (Noy-Meir, Gutman & Kaplan 1989; Noy-Meir 1995; Noy-Meir & Sternberg 1999).

Species set

The Argentine species set included 83 native species for which information was available on both response to grazing and on basic plant and leaf traits. This set is a fairly representative sample of the entire herbaceous flora of the area and includes most of the common species. Exotic species in the area are mostly associated with agricultural plots and road verges. In grazed sites, they represent an extremely low proportion of total vegetation cover and species richness, and do not change significantly under different grazing intensities (Díaz, Acosta & Cabido 1994). Therefore, they were not considered in this study. The Israeli species set consisted of 19 common species of known grazing response, from which leaf trait measurements could be obtained in April 1999. The Israeli sample represented the main functional types and grazing responses encountered in the flora, although not necessarily in exact proportion to their relative frequencies. The combined Argentine–Israeli species set thus consisted of 102 species.

Trait measurements

Measurements of leaf traits for each species were taken from a sample of at least 24 fully expanded, but not senescing, healthy leaves from at least six sexually mature individuals. Mean area per leaf (leaf area in mm2) was measured on the fresh leaves by scanning five to eight samples of three to five leaves at a time and estimating their area in the image with a scanner and the software Optimetrics® (Bioscan, Edmonds, USA). Mean dry mass per leaf (leaf mass in mg) was estimated by oven-drying until a constant weight (60 °C) and then weighing the same leaf samples. From these two measurements, mean specific leaf area (SLA) was calculated (mm2 mg−1). Leaf tensile strength, or resistance to tearing, was measured with a tensile-strength meter constructed on the basis of Hendry & Grime (1993), and expressed as force needed per unit of width of a leaf sample (N mm−1). Plant height was expressed as a categorical variable, with species classified into low (< 40 cm maximum foliage elevation) or tall (> 40 cm). Life history, i.e. annual vs. perennial, was determined on the basis of the regional floras. Taxonomic affiliation of the species was defined at the highest level, as monocotyledons vs. dicotyledons.

Grazing response

On the basis of previous studies (Acosta et al. 1989; Cabido, Acosta & Díaz 1989; Noy-Meir, Gutman & Kaplan 1989; Díaz, Acosta & Cabido 1992, 1994; Pucheta, Díaz & Cabido 1992; Pucheta et al. 1998; Noy-Meir & Sternberg 1999), species were initially divided into three categories. Grazing decreasers or grazing-susceptible species (G−) were consistently more abundant in ungrazed or lightly grazed sites than in moderately to heavily grazed sites (> 25% consumption of above-ground net primary productivity). Grazing increasers (G+) were species that consistently showed the opposite trend. The third group (G0) included species that did not respond consistently to grazing intensity and those that had maximum abundance at some intermediate grazing intensity. The G0 group on its own was too small for statistical inference and was therefore combined with G+ into one grazing-resistant group (G0+). It should be noted that grazing response as evaluated here includes both direct and indirect (community-mediated) effects of grazing on a species. For example, a G+ species may benefit from the reduced competition and increased light availability in grazed sites, rather than from growth stimulation by grazing. The proportion of G− and G0+ plants was not significantly different between the Argentine and the Israeli species sets (P = 0·30 by Fisher’s exact test). A complete list of species and grazing responses is given in the Appendix.

Statistical analysis

Three types of statistical analysis were carried out, in order to answer different questions. First, to examine trait differences between species with different responses to grazing, grazing response (categorical) was defined as the independent variable, while each of the other species traits in turn was the dependent variable. For continuous traits (leaf mass, leaf tensile strength and SLA), the differences between grazing response groups in the values of the trait, transformed to natural logarithms (ln) to normalize distribution and to equalize variances, were evaluated by the t-test and simple analysis of variance. Significance of the difference is presented as P, the probability of a value of t or F, respectively, being greater than the observed value. The proportion of variance explained (r2) is evaluated as the ratio of sum of squares in the model (grazing response) to total sum of squares. Means for continuous traits are presented without the transformation, to facilitate biological interpretation. For categorical traits (life history, height, taxonomic affiliation and country), the significance of difference in traits frequencies between grazing response groups was evaluated by Fisher’s exact test for 2 × 2 contingency tables. A measure of the proportion of variance explained (R2) in this case is the ratio of negative log-likelihoods (U) in the logistic regression of the categorical trait on grazing response (Sall, Lehman & Saul 1996). The analyses were carried out for Argentine (n = 83) and Israeli (n = 19) species sets separately and for the combined (n = 102) species set.

A second type of statistical analysis was designed to examine relationships between different traits that may affect the interpretation of the relationships of each of those traits with grazing response. In particular, the relationships of SLA (as dependent variable) with other traits were examined in detail, because SLA has been proposed as an integrated or diagnostic trait in the definition of plant functional types. Relationships between pairs of continuous traits were evaluated by linear correlation and regression of ln-transformed values, for the separate country sets and the combined data set. Relationships between a continuous (in particular SLA) and a categorical variable were analysed by t-test and by analysis of variance of ln-transformed values. In some cases, the joint effects of a categorical and a continuous variable (covariate) on another continuous variable (SLA) were analysed by general linear model anova.

In the third type of statistical analysis, grazing response (G−/G0+) was defined as the dependent variable to be predicted, and both continuous and categorical species traits were defined as the independent variables or potential predictors. Because grazing response is a binary categorical variable, only the probability of a species being G− or G0+ can be predicted, and the appropriate statistical model is logistic regression. Analyses with single species traits and with combinations of two or more traits were carried out in search of the ‘best’ model for prediction of grazing response, i.e. the model with the highest R2 (the ratio of negative log-likelihoods U). These analyses were in general carried out only with the combined data set, but in some cases the separate country sets were analysed to clarify specific points.

Results

Trait differences between species with different responses to grazing

Continuous traits

In the combined data set, all four continuous traits were significantly different between grazing response groups, i.e. between G− and G0+ species (Table 1). In decreasing order of variance of the trait accounted for by grazing response, G− species had larger leaves (by mass, then by area) and stronger leaves than G0+ species. G0+ species had on average higher SLA, but this difference accounted for only 4% of the variance in SLA and was not as highly significant as for the other traits. The trends described for the combined data set for the four traits in relation to grazing response were qualitatively maintained in the separate data sets for both Argentina and Israel (Table 1).

Table 1.  Mean values of continuous traits for grazing-resistant (G0+) and grazing-susceptible (G−) species response groups in the combined Argentine–Israeli data set (n = 102 species) and in the separate Argentine (n = 83) and Israeli (n = 19) sets. Means are of untransformed data. P is the significance of the difference in ln-transformed trait values between grazing response groups, by anova. r2 is the adjusted proportion of variance accounted for by grazing response in the same analysis. The traits are ranked by r2 in the combined data set. Significantly (P < 0·05) higher values for each trait are presented in bold
TraitMean G0+Mean G−r2P
Combined data set
Leaf mass (mg)  26·9108·40·200< 0·001
Leaf area (mm2) 41113630·125< 0·001
Leaf tensile strength (N mm−1)   2·49   8·670·098< 0·001
Specific leaf area (mm2 mg−1)  16·1  12·60·044  0·012
Argentina
Leaf mass (mg)  18·6  74·00·191< 0·001
Leaf area (mm2) 1885870·129< 0·001
Leaf tensile strength (N mm−1)   2·78  10·710·158< 0·001
Specific leaf area (mm2 mg−1)  13·9  10·90·063  0·012
Israel
Leaf mass (mg)  87·5217·40·244  0·018
Leaf area (mm2)150638180·167  0·047
Leaf tensile strength (N mm−1)   1·05   2·200·013  0·282
Specific leaf area (mm2 mg−1)  27·0  18·10·350  0·004

For all continuous traits, there were highly (P < 0·001) significant differences between the species sets from the two countries (Table 2, top). Plants in the Israeli set had on average higher leaf mass, higher leaf area and higher SLA, but lower leaf tensile strength, compared with Argentine plants. Despite these large differences, the ‘grazing effect’, defined as the mean difference in traits (measured on a ln-scale) between grazing response groups, was remarkably similar in the two countries (Table 2, bottom). In the cases of leaf mass and leaf area, the grazing effect in the combined data set was as great or greater than in both separate country data sets (Table 2, bottom). The mixing of two data sets representing different floras with different ranges of these traits did not mask the relationship between trait and grazing response but rather strengthened it. For SLA, the grazing effect in the combined data set was slightly smaller than for both separate data sets, that is some masking occurred by the combination for this trait. For leaf tensile strength, the grazing effect on the combined data set was intermediate between the effects observed in separate country sets.

Table 2.  Means of leaf traits, and the differences between means of these traits in grazing-resistant (G0+) and grazing-susceptible (G−) species response groups (both in ln-transformed values). Both means and grazing response differences are presented for the combined species set and the separate country sets. The last column shows the differences between Argentine and Israeli species sets in the ln-transformed means and grazing response differences. Bold type indicates differences that are significantly different from zero (P < 0·05, t-test)
TraitsCombinedArgentinaIsraelDifference Argentina–Israel
Means of ln (variable)
Leaf mass (mg) 2·49 2·19 3·83−1·64
Leaf area (mm2) 5·05 4·61 6·98−2·37
Leaf tensile strength (N mm−1) 0·71 0·90−0·14+1·04
Specific leaf area (mm2 mg−1) 2·56 2·43 3·14−0·71
Differences ln (variable) between G0+ and G−
Leaf mass (mg)−1·72−1·61−1·65+0·04
Leaf area (mm2)−1·40−1·24−1·26+0·02
Leaf tensile strength (N mm−1)−0·86−1·06−0·54−0·52
Specific leaf area (mm2 mg−1)+0·32+0·37+0·39−0·02

Categorical traits

The categorical trait that showed the strongest association with grazing response in the combined data set was plant height: 76% of G− species were tall (> 40 cm) vs. only 13% tall species among G0+ species (P < 0·001 by Fisher’s exact test for 2 × 2 contingency tables). Also significant was the difference in the representation of monocotyledons and dicotyledons in the grazing response groups: 72% of G− plants were monocotyledons, compared with only 35% monocotyledons in G0+ plants (P = 0·001). The proportion of perennials (vs. annuals) was somewhat greater in the G− group than in the G0+ group, but this difference was not significant in the combined species set (G− 80% perennials, G0+ 69%, P = 0·21) or in the Argentine set alone (G0− 95%, G+ 83%, P = 0·28). Perennials represented 86% of the Argentine species set and only 11% of the Israeli set.

Relationships between sla and other variables

There were significant relationships between traits that must be taken into account in the interpretation of the relationships between traits and response to grazing. Most obviously, leaf area and leaf mass were strongly and significantly associated, considering both the whole database and individual countries (Fig. 1). The regression lines for the two countries had almost equal slopes, but the intercept was significantly higher for Israeli plants (3·64 ± 0·33, 95% confidence interval) than for Argentine plants (2·74 ± 0·20). Israeli species showed consistently larger leaf area per unit leaf mass (mean SLA = 24·2) than Argentine species (mean SLA = 12·7, difference significant at P < 0·001).

Figure 1.

The relationship between leaf area and leaf dry mass, both transformed to natural logarithms, over all species in the data set. Separate regression lines for 83 Argentine species (solid symbols) and 19 Israeli species (empty symbols). Triangles = grazing-susceptible species (G−); circles = grazing-resistant species (G0+).

In the combined data set, annuals had significantly higher SLA (mean = 21·8) than perennials (mean = 12·7, R2 = 0·24, P < 0·001). A similar but weaker trend was also found in the Argentine set alone (annuals 17·1 vs. perennials 12·6, R2 = 0·062, P = 0·013). In the Israeli data set, all but two species were annuals, which explains in part the higher mean SLA in this country. However, Israeli annuals also had significantly greater SLA than Argentine annuals (R2 = 0·33, n = 27, P < 0·001). Thus, the combination of country and life-history factors explained a somewhat larger proportion of variance in SLA (R2 = 0·27) than either life history (R2 = 0·24) or country (R2 = 0·21) alone. Dicotyledons had significantly higher SLA (17·2) than monocotyledons (12·8), but this explained only 8% of the variance in SLA in the combined data set (R2 = 0·08, P = 0·002). In the Argentine data set, low plants had on average somewhat higher SLA (14·0) than tall plants (9·7; P = 0·001), but in the combined data set there was no significant relationship between SLA and plant height.

SLA was significantly and strongly negatively correlated with leaf tensile strength in the Argentine and in the combined sample of species (Table 3). In other words, leaves with high SLA tended to be weaker, as might be expected from mechanical considerations. This factor alone accounted for 33% of the variance of SLA in the combined data set, considerably more than life history (24%) or country (21%). An additive bifactorial model including highly significant (P < 0·001) effects of both life history (annuals > perennial) and leaf tensile strength (negative) predicted as much as 43% of variation in SLA among species.

Table 3.  Univariate relationships between SLA (as dependent variable) and each one of the other leaf traits in the combined species set and in the separate country sets, by regression with ln-transformations of all variables. Parameters presented are the slope parameter (b) of regression, the significance (P) of the slope being different from zero, and the proportion of variance accounted for by regression (R2). Traits are ranked by R2 in the combined data set. Significant (P < 0·05) regressions are presented in bold
 CombinedArgentinaIsrael
Independent variableSlopeR2PSlopeR2PSlopeR2P
Leaf tensile strength−0·3050·332< 0·001−0·2930·305< 0·001−0·0010·0000·916
Leaf area+0·0800·039  0·026−0·0030·000  0·939−0·1130·1950·033
Leaf mass−0·0530·011  0·147−0·1460·138< 0·001−0·1300·3700·003

Some unexpected effects were observed in the statistical correlations between SLA and the two variables of which it is the ratio, leaf area and leaf mass. In the sample of Israeli species, SLA was negatively correlated not only with leaf mass (which might be expected if only from a pure algebraic consideration) but also with leaf area (P = 0·033), i.e. larger leaves also tended to have more mass per unit area. In the Argentine sample there was no correlation between leaf area and SLA, with rather large and independent variations of both variables. All combinations, large and high SLA, large and low SLA, small and high SLA, and small and low SLA leaves, were common (Fig. 2). When samples from both countries were combined, a new pattern appeared that was different from those observed in the separate samples (Table 3). In the combined data set, SLA showed a weak but significant positive correlation with leaf area, while the negative correlation between SLA and leaf mass disappeared. Inspection of the scatter of species in the SLA vs. leaf area graph showed that the Israeli species extended the circular cloud of Argentine species into a region of larger leaves with higher SLA (Fig. 2), hence the positive correlation in the combined set.

Figure 2.

The relationship between specific leaf area and leaf area, both transformed to natural logarithms, over all species in the data set. Separate regression lines for 83 Argentine species (solid symbols) and 19 Israeli species (empty symbols). Triangles = grazing-susceptible species (G−); circles =grazing-resistant species (G0+).

Best predictors of plant response to grazing

When species response to grazing (− vs. 0+) was selected as the dependent (predicted) variable, the best single predictor of this response, in the combined set of species, was plant height as categorical variable (Table 4). Only 8% of low or medium height (< 40 cm) species were G−, compared with 66% of species taller than 40 cm.

Table 4.  Best logistic regression models for prediction of the probability of grazing response of a species (− or 0+) from other traits, in the combined data set (n = 102), allowing both continuous and categorical predictors to be included in the models. All one-variable models are presented, ranked by the proportion of variance explained by the model R2. Only significant two-variable models are presented, ranked by R2. The ‘effect’ for categorical variables is a trait state and the associated grazing response; for continuous variables, + (or −) means that higher values of the variable predict a higher probability of + (or −) grazing responses. Significance (P) levels of the effects are denoted as follows: NS (P > 0·05), * (P < 0·05), ** (P < 0·01), *** (P < 0·001)
 One-variable models
 Variable 1
R2TraitEffectP
0·306HeightLow +***
0·203Leaf mass***
0·124Leaf area***
0·097Leaf strength**
0·093Dicot/monocotDicots +**
0·048SLA+*
0·011Life history NS
0·005Country NS
 Two-variable models
 Variable 1Variable 2
R2TraitEffectPTraitEffectP
0·401HeightLow +***Life historyAnnuals +**
0·369HeightLow +***Dicot/monocotDicots +*
0·361HeightLow +***Leaf strength*
0·349Leaf mass***Leaf strength***
0·347Leaf area***Leaf strength***

The second best single predictor of grazing response was leaf mass: larger leaves tended to be associated with a negative response to grazing. A similar association with leaf area was significant but with a lower determination coefficient. Additional significant single predictors of decreasing effectiveness were leaf tensile strength (stronger leaves, G− species) and taxonomic affiliation (dicotyledons were more G0+). SLA, although marginally significantly associated with grazing response (higher SLA with G0+ species), was the continuous variable with the smallest predictive value (Table 4).

Life history alone was not a significant predictor of grazing response in the combined data set, although a somewhat greater proportion of annuals (83%) than of perennials (73%) was classified as having a positive or neutral response to grazing (G0+; Tables 4 and 5). However, when life history was added as a second factor to height, it increased considerably the predictive power of the model, resulting in the best two-factor model for predicting response to grazing (Table 4). This was because, among tall plants, the proportion of G0+ plants was significantly greater in annuals than in perennials (69% vs. 6%; Table 5). Among low plants there was no such significant difference between life histories, because almost all the low plants are G0+. Because there are more low than tall plants in the data set, the life-history effect found among tall plants gets swamped by the low plants and the life-history effect alone is not significant in the whole data set (Table 5). When the height effect is already in the model, the life-history effect (expressed among tall plants) significantly improves the prediction of grazing response (Table 4). Although the data indicated an interactive effect of height and life history on grazing response, the additional interaction term was not quite significant (P = 0·056).

Table 5.  Association between grazing response and life history, in different plant height categories. Each cell contains the number of species with each grazing response within the given life history, as absolute number and as percentage of the total number of species with this life history. The significance of association (lack of independence) is given by Fisher’s exact test for 2 × 2 contingency tables
  Grazing response class 
Height classLife historyG−G0+P (Fisher)
Tall plants (n = 29)Annuals 4 (30%) 9 (69%) 
 Perennials15 (94%) 1 (6%)0·001
Low plants (n = 73)Annuals 1 (6%)15 (94%) 
 Perennials 5 (9%)52 (91%)0·607
All plants (n = 102)Annuals 5 (17%)24 (83%) 
 Perennials20 (27%)53 (73%)0·320

When height was excluded from the model to predict grazing response, a good prediction could also be obtained by a two-factor model combining leaf tensile strength with leaf mass or leaf area (Table 4). SLA did not significantly contribute to any two-factor model. The best prediction of whether a species in our database responded negatively to grazing or not, was achieved by combining plant height, life history and leaf mass. In this three-factor model all effects were significant and it slightly improved the determination coefficient (0·44) compared with the two-factor model with height and life history only (0·40).

Discussion

Similar plant trait responses were associated with grazing by livestock in Argentina and Israel. In general, species that showed a positive or neutral response to grazing tended to have shorter height and smaller, more tender, leaves, with higher SLA compared with species that showed a negative response to grazing. Short plant height and small leaves are typical mechanisms of grazing resistance by grazing avoidance (sensuBriske 1996, 1999) that have been documented in many other studies (Sala et al. 1986; McNaughton & Sabuni 1988; Noy-Meir, Gutman & Kaplan 1989; Díaz, Acosta & Cabido 1992; Landsberg, Lavorel & Stol 1999). However, another common mechanism of grazing avoidance, leaf toughness (and low SLA), that is usually associated with low palatability (Coughenour 1985; Grime et al. 1996; Cornelissen et al. 1999) did not contribute to grazing resistance in this species set. On the contrary, grazing resistance was associated with tender leaves and (weakly) with high SLA, as predicted by Westoby (1999) for heavily grazed areas, suggesting higher growth rate as a mechanism of grazing tolerance (sensuBriske 1996, 1999).

The trends reported here for Argentine and Israeli grasslands match the predictions of the models of McNaughton (1984) and Milchunas, Sala & Lauenroth (1988) for subhumid grasslands with a long evolutionary history of grazing: intense grazing favours short plants with high regrowth rates, rather than tough, unpalatable plants. High growth rates tend to be negatively correlated with quantitative defences, such as leaf toughness (Herms & Mattson 1992). These results contradict a classical precept of range management, that palatability in grassland communities responds inversely to grazing intensity (Ellison 1960). Although the latter phenomenon has often been observed in some regions of the world (North America, Ellison 1960; South Africa, Morris, Tainton & Hardy 1992), the present results support the view that response to grazing is more diverse and depends on the evolutionary and climatic context of grazing in different regions (Perevolotsky & Seligman 1998). For example, grasslands in Argentina, presumably with a similar evolutionary history of grazing, and under relatively similar management regimes, differ in their responses to grazing. Responses similar to the ones described in this article have been found for some moderately to heavily grazed subhumid grasslands (Posse, Anchorena & Collantes 1996, 2000). On the other hand, opposite trends have been reported for semi-arid grasslands, in which continuous grazing is often associated with tough and unpalatable plant species (Distel & Boó 1996). This might be related to lower resource availability or lower grazing intensity. According to general models, grazing avoidance should be favoured against grazing tolerance in situations of water or nutrient scarcity (Herms & Mattson 1992; Hobbie 1992). According to Westoby (1999), lower grazing intensity in a short growing season may allow more effective avoidance of tough leaves by grazers.

G− species in productive Argentine and Israeli grassland grazed by cattle tended to be tall, with large, tough, leaves. This suggests that leaf toughness, rather than serving to deter grazers, is related to the advantages of having stiff, erect, leaves to search for light in the closed canopy that develops in the absence of grazing. In Argentina, large leaves with low SLA are mostly fibrous leaves of tall perennial tussock grasses (‘chocolate-box’ leaves; sensuGrubb 1986; Cunningham, Summerhayes & Westoby 1999), whereas in Israel species with large leaves and low SLA are broad-leafed annual dicotyledons, with tender lamina and strong thick veins (‘kite’ structure; sensuGrubb 1986; Cunningham, Summerhayes & Westoby 1999).

In general, Argentine species showed lower SLA and higher leaf tensile strength than Israeli species. This higher degree of sclerophylly does not seem to be associated with nutritional or water-balance deficiencies. Although Córdoba montane grasslands receive considerably more precipitation than Galilee grasslands, seasonality (rainfall strongly concentrated to the warm season) determines a much higher evapotranspiration rate, and therefore the moisture regime can be considered roughly similar. Above-ground net primary productivity is similar (around 300 g m−2 year−1 in both cases; Seligman & Gutman 1979; Pucheta et al. 1998), and N and P availability is not as low as to be commonly limiting in the soil of either system. It may be speculated that the higher occurrence of sclerophylly in the flora of these Argentine mesic grasslands is linked with their historically less intense grazing regime, and therefore stronger selection for canopy dominance, compared with the eastern Mediterranean (Milchunas, Sala & Lauenroth 1988; Perevolotsky & Seligman 1998; Díaz et al. 1999).

From a practical point of view, the best single predictor of response to grazing in our joint data set was plant height, followed by leaf mass. The best prediction of whether a species in our database responded negatively to grazing or not, was achieved by combining plant height, life history and leaf mass. Despite the suggestion of its key role in understanding plant trait responses to grazing (Díaz & Cabido 1997; Westoby 1999), SLA showed a poor predictive value of grazing response. Within the context of our data set, plant height and leaf mass appeared as better predictors of grazing response than SLA. They are also considerably easier to measure, as SLA has some operational complications (Weiher et al. 1999; Wilson, Thompson & Hodgson 1999).

The ranges of plant height, leaf size, leaf strength and SLA differed markedly between the species samples from the two countries. The patterns of correlation between some of these structural traits were also different between countries. However, the significant relationships between these plant traits and grazing response were maintained and in some cases strengthened by combining the Argentine and Israeli species sets. This exploratory study allows cautious optimism regarding the prospects of generalizing relationships between simple plant traits and grazing responses across continents, in communities and grazing systems that share similarities in productivity, grazing history and present grazing management. In particular, the results indicate that intense cattle grazing in productive natural grasslands with a long history of grazing will result mainly in an increase of short species with small tender leaves, at the expense of tall species with large tough leaves. Animal intake and productivity in the intensely grazed rangeland will therefore be limited by small bite size, but not necessarily by low palatability or low primary productivity.

Acknowledgements

We are grateful to F. Vendramini, S. Basconcelo and C. Ribbert for their invaluable contribution to trait measurements, and to our research team for constant support. Members of IMBIV assisted in plant identifications, and D. Abal-Solis drew the graphs. Research leading to this paper was supported in Argentina by Universidad Nacional de Córdoba, CONICET, CONICOR, SECyT–UNC, IAI ISP I and III, Fundación Antorchas, the European Union (CI1*-CT94-0028) and the Darwin Initiative (DETR-UK); and in Israel by the Israel Science Foundation of the Academy of Sciences and Humanities. This is a contribution to IGBP-GCTE Task 4.3.2.

Received 3 October 2000; revision received 1 February 2001

Appendix

Table 6. Species included in the data set and their response to grazing by domestic ungulates. G−= grazing-susceptible; G0+= grazing-resistant (see text for definition and further details). Nomenclature follows Zuloaga et al. (1994) and Zuloaga & Morrone (1996, 1999) for Argentine species, and Zohary & Feinbrun-Dothan (1966–86) for Israeli species
SpeciesFamilyGrazing response
Argentina  
Acicarpha tribuloidesCalyceraceaeG0+
Adesmia bicolorFabaceaeG0+
Agrostis montevidensisPoaceaeG0+
Alternanthera pumilaAmaranthaceaeG0+
Astragalus parodiiFabaceaeG0+
Bidens andicola var. decompositaAsteraceaeG0+
Bothriochloa laguroidesPoaceaeG0+
Briza subaristataPoaceaeG0+
Bromus auleticusPoaceaeG−
Bromus catharticusPoaceaeG−
Bulbostylis juncoidesCyperaceaeG0+
Cardionema ramosissimaCaryophyllaceaeG0+
Carex fusculaCyperaceaeG0+
Cerastium arvenseCaryophyllaceaeG0+
Chaptalia integerrimaAsteraceaeG0+
Chevreulia sarmentosaAsteraceaeG0+
Chloris retusaPoaceaeG0+
Cologania ovalifoliaFabaceaeG−
Cotula mexicanaAsteraceaeG0+
Cuphea glutinosaLythraceaeG0+
Cyperus reflexusCyperaceaeG−
Deyeuxia hieronymiPoaceaeG−
Dichondra repens var. holosericeaConvolvulaceaeG0+
Eleocharis albibracteataCyperaceaeG0+
Eragrostis lugensPoaceaeG0+
Eryngium agavifoliumApiaceaeG0+
Eryngium nudicauleApiaceaeG0+
Euphorbia serpensEuphorbiaceaeG0+
Festuca hieronymiPoaceaeG−
Festuca tucumanicaPoaceaeG−
Galactia marginalisFabaceaeG0+
Gamochaeta filagineaAsteraceaeG0+
Gentianella parvifloraGentianaceaeG0+
Glandularia dissectaVerbenaceaeG0+
Gnaphalium gaudichaudianumAsteraceaeG0+
Hieracium giganteum var. setulosumAsteraceaeG−
Hypochaeris argentinaAsteraceaeG−
Hypoxis humilisHypoxidaceaaeG−
Juncus achalensisJuncaceaeG0+
Juncus uruguensisJuncaceaeG0+
Lachemilla pinnataRosaceaeG0+
Lepidium bonarienseBrassicaceaeG0+
Melica macraPoaceaeG−
Mitracarpus cuspidatumRubiaceaeG0+
Muhlenbergia peruvianaPoaceaeG0+
Nothoscordum inodorumLiliaceaeG0+
Noticastrum marginatumAsteraceaeG0+
Oenothera indecoraOnagraceaeG0+
Oxalis sexenataOxalidaceaeG0+
Paspalum notatumPoaceaeG0+
Paspalum quadrifariumPoaceaeG−
Pfaffia gnaphaloidesAmaranthaceaeG0+
Piptochaetium montevidensePoaceaeG0+
Plantago australisPlantaginaceaeG0+
Plantago brasiliensisPlantaginaceaeG0+
Poa resinulosaPoaceaeG0+
Poa stuckertiiPoaceaeG−
Pratia hederaceaCampanulaceaeG0+
Relbunium richardianumRubiaceaeG0+
Rhynchosia sennaFabaceaeG0+
Rumex acetosellaPolygonaceaeG0+
Schizachyrium microstachyumPoaceaeG0+
Schizachyrium spicatumPoaceaeG0+
Schkuhria pinnataAsteraceaeG0+
Setaria parvifloraPoaceaeG0+
Sisyrinchium unguiculatumIridaceaeG0+
Sorghastrum pellitumPoaceaeG−
Spergula ramosaCaryophyllaceaeG0+
Sporobolus indicusPoaceaeG0+
Stenandrium dulceAcanthaceaeG0+
Stipa eriostachyaPoaceaeG−
Stipa filiculmisPoaceaeG−
Stipa neesianaPoaceaeG0+
Stipa tenuissimaPoaceaeG−
Stipa trichotomaPoaceaeG−
Stylosanthes gracilisFabaceaeG0+
Tagetes argentinaAsteraceaeG0+
Taraxacum officinaleAsteraceaeG0+
Trifolium repensFabaceaeG0+
Tripogon spicatusPoaceaeG0+
Vicia gramineaFabaceaeG−
Vulpia myurosPoaceaeG0+
Zephyranthes longistylaAmaryllidaceaeG0+
Israel  
Avena sterilisPoaceaeG0+
Psoralea bituminosaFabaceaeG−
Brassica nigraBrassicaceaeG0+
Bromus alopecurusPoaceaeG0+
Cephalaria joppensisDipsacaceaeG−
Hordeum bulbosumPoaceaeG−
Hordeum spontaneumPoaceaeG−
Linum pubescensLinaceaeG0+
Medicago granadensisFabaceaeG0+
Phalaris paradoxaPoaceaeG0+
Pimpinella creticaApiaceaeG0+
Raphanus rostratusBrassicaceaeG0+
Rapistrum rugosumBrassicaceaeG0+
Scabiosa proliferaDipsacaceaeG0+
Synelcosciadium carmeliApiaceaeG−
Trifolium nigrescensFabaceaeG0+
Trifolium pilulareFabaceaeG0+
Trifolium purpureumFabaceaeG0+
Triticum dicoccoidesPoaceaeG−

Ancillary