Isotopic values of plants in relation to water availability in the Eastern Mediterranean region

Abstract

Plant C and N isotope values often correlate with rainfall on global and regional scales. This study examines the relationship between plant isotopic values and rainfall in the Eastern Mediterranean region. The results indicate significant correlations between both C and N isotope values and rainfall in C₃ plant communities. This significant relationship is maintained when plant communities are divided by plant life forms. Furthermore, a seasonal increase in C isotope values is observed during the dry season while N isotope values remain stable across the wet and dry seasons. Finally, the isotopic pattern in plants originating from desert environments differs from those from Mediterranean environments because some desert plants obtain most of their water from secondary sources, namely water channeled by local topographic features rather than direct rainfall. From these results it can be concluded that water availability is the primary factor controlling C and N isotope variability in plant communities in the Eastern Mediterranean.

Introduction

Water availability, measured as rainfall, is argued to be the most influential factor affecting the C (δ¹³C) and N isotopic composition (δ¹⁵N) of plants in semi-arid and arid environments (Stewart et al. 1995; Handley et al. 1999; Amundson et al. 2003; Swap et al. 2004). The relationship between rainfall and the isotopic composition of plants has been demonstrated in many regions, but has not been investigated in the Eastern Mediterranean. This paper tests whether a similar relationship exists in this region. Although a dominant factor, rainfall is not the only variable that may affect water availability (for example see Schulze et al. 1996). In particular, topography and seasonality may also affect the relationship between rainfall and plant isotopic composition (Ehleringer and Cooper 1988; Smedley et al. 1991). The Eastern Mediterranean climate is characterized by a strongly seasonal pattern of rainfall, with rain falling nearly exclusively in the cool winter season and steep rainfall gradients that range from over 1,000 mm/year to below 100 mm/year within a small geographic area (Goldreich 2003). Likewise, topography influences water availability by impacting the water drainage and retention capacities of local soils. The effect of topography on isotopic values can be particularly pronounced in arid regions (Ehleringer and Cooper 1988; Farrell et al. 1996).

This study documents the relationship between water availability and the C and N isotopic values of plants on a community level in the Eastern Mediterranean and evaluates the effect of seasonality and topography on this relationship. The relationship between rainfall and plant isotopes is established through intensive sampling of plants across north-south and east-west rainfall gradients in the Eastern Mediterranean. The effect of seasonality on water availability is investigated by collecting data on both wet (October–April) and dry (May–September) season plants, while the impact of topographic setting is investigated by comparing the isotopic signatures of plants growing in elevated (exposed ridges) versus depressed (dry washes) settings in arid regions.

Theoretical background

C isotope discrimination and water use efficiency

The following equation estimates the C isotope discrimination (¹³Δ) in C₃ plants (Farquhar et al. 1989):

$$ {}^{13}\Updelta = a\,{\frac{{P_{\text{a}} - P_{\text{i}} }}{{P_{\text{a}} }}} + b\,{\frac{{P_{\text{i}} }}{{P_{\text{a}} }}} = a + (b - a){\frac{{P_{\text{i}} }}{{P_{\text{a}} }}} $$

(1)

Where the constant a (=4.4‰) is fractionation occurring during the diffusion of CO₂ through the stomata (Craig 1953), and the constant b (=27‰) is the fractionation associated with CO₂ fixation by ribulose bisphosphate carboxylase/oxygenase (RuBisCO; Farquhar and Richards 1984). P _a refers to air and P _i to intracellular partial CO₂ pressure. Reduced stomatal conductivity caused by drought stress leads to a decrease in the P _i/P _a ratio and results in a decrease in ¹³Δ (Farquhar et al. 1982). The δ¹³C values of plants (δ¹³C_p) can be directly related to ¹³Δ though this requires an a priori assumption about the δ¹³C values of atmospheric CO₂ (δ¹³C_a):

$$ \delta^{ 1 3} {\text{C}}_{\text{p}} \, \left( \permille \right) \, = {\frac{{\delta^{13} {\text{C}}_{\text{a}} - {}^{13}\Updelta }}{{1 + {}^{13}\Updelta }}} $$

(2)

The present-day δ¹³C_a value is ~−8‰, yet this value may vary seasonally on local and global scales, most importantly as a consequence of fossil fuel combustion (Keeling et al. 2005).

Given the above, C₃ plants growing under conditions of water stress are expected to have elevated δ¹³C values compared to plants growing under optimal water conditions (Farquhar and Sharkey 1982). A number of studies have found a negative correlation between mean annual rainfall and the δ¹³C of C₃ plants (Stewart et al. 1995; Swap et al. 2004). Studies undertaken in seasonal, semi-arid environments also report higher δ¹³C values in grassland plant species and pine needles collected in the dry versus wet season (Smedley et al. 1991; Klein et al. 2005). Despite the strong association between rainfall and the δ¹³C values of C₃ plants, other factors including genetic, structural, and functional plant adaptations and soil characteristics can also change the water use efficiency (WUE; the ratio of water loss to biomass gained) and thus may influence the relationship between rainfall and the δ¹³C values of C₃ plants (Ehleringer and Cooper 1988; Comstock and Ehleringer 1992; Schulze et al. 1996, 2006; Dawson et al. 2002 and references therein). Given the potential confounding effect of these variables, the relationship between rainfall and δ¹³C values of C₃ plants must first be established. Here, it is hypothesized that the δ¹³C values of C₃ plant communities will be significantly negatively correlated with rainfall in the Eastern Mediterranean region. Additionally, the δ¹³C values of C₃ plants growing during the hot and dry season are expected to be significantly higher than those of plants growing during the cool and wet season. In contrast to C₃ plants, the isotopic composition of C₄ plants is not expected to be affected by water stress (Farquhar et al. 1982; Farquhar 1983).

N isotope distribution in plants and its relation to water availability

A negative correlation has also been reported between the N isotopic values of plants and bulk soils and rainfall (Heaton 1987; Austin and Vitousek 1998; Handley et al. 1999; Amundson et al. 2003; Swap et al. 2004). The association between rainfall and plant δ¹⁵N values is not straightforward—while plants fix C directly from the atmosphere, they obtain N from the soil or through a symbiotic relationship with N-fixing microorganisms. Many climate-linked factors such as the activity of soil microorganisms, variability in soil N sources (i.e., NO₃ ⁻, NH₄ ⁺), and plant metabolism also influence plant δ¹⁵N values (Handley and Raven 1992; Högberg 1997). Handley et al. (1999) suggested that the observed negative correlation between plant δ¹⁵N values and rainfall is a product of water availability and soil N sources during plant growth. An increase in the δ¹⁵N values of both soils and plants under conditions of decreasing rainfall has been attributed to more open N cycling in ecosystems with higher N loss relative to turnover (Austin and Vitousek 1998). Plant physiology has also been shown to be at least partly responsible for higher δ¹⁵N values in water-stressed grasses (Lopes and Araus 2006). Despite gaps in current understanding of the mechanism, it is hypothesized that there will be a negative relationship between the δ¹⁵N values of plant communities and rainfall in the Eastern Mediterranean. It is also hypothesized that the δ¹⁵N values of C₃ plants that grow during the dry season will be significantly higher than those that grow during the wet season.

An independent measure of water availability during plant growth, the degree of concentration of N in C₃ plant leaves (measured as the C/N ratio), is also used here. Plants growing in regions with low annual rainfall improve light use efficiency by increasing leaf N concentration relative to plants growing in wetter regions (Field et al. 1983; Cunningham et al. 1999).

Materials and methods

Plant collection

Plant samples (n = 1,400; for a complete list of plant species, season of collection, life form, and other attributes for each site see Appendix A in ESM) were collected from 27 sites along the major north-south and west-east rainfall gradients in Israel and the Palestinian Authority ranging from 1,000 to below 100 mm rain/year (Table 1; Fig. 1). Plants were collected in undisturbed and protected areas under a permit granted by the Israel Nature and National Parks Protection Authority to minimize the biasing effect of anthropogenic activity. For the same reason, plant collection was avoided in river beds except arid environments (see below). The exact location of each collection site was recorded using a GPS device; the coordinates were later used to extract rainfall data from the Hebrew University GIS center databases. Three plants from each of ca. 15 dominant species located within a 10-m radius of a randomly assigned sampling point were collected at every site. Sun-exposed leaves of shrubs and trees were collected from the outer surface of the crown at ca. 1.5 m above the ground surface.

Table 1 Description of plant sampling sites, including location, rainfall, elevation data, and the site means of C (δ¹³C; ±1 SD) and N isotopic composition (δ¹⁵N; ±1 SD) values in C₃ plants

Fig. 1

Relief map indicating plant collection sites in the Eastern Mediterranean region. Plant collection sites are numbered in black (see Table 1 for site description). Rainfall data are depicted by isohyets (values in gray; mean annual rainfall; mm/year). Gray arrows represent the major rainfall gradients: north-south (NS) and west-east (WE)

The sampling locations along the north-south and west-east rainfall gradients were chosen independently in consultation with rainfall maps at roughly 100-mm per year increments. Higher isotopic variability was expected in the transitional areas between Mediterranean and desert environments (200–75 mm/year) and thus sampling stations were added at frequencies <100 mm/year. The plants were collected in two episodes representing the peak of the wet and dry season. The dry season collection took place in July–August 2006 and the wet season collection took place in March 2007. Recently expanded leaves were collected during the wet season, and also in the dry season when available. While most of the lignified species collected in this study are evergreen, winter deciduous tree species (n = 1) and drought deciduous shrub and tree species (n = 6) were also collected. Many of the dwarf shrub species replace short-lived leaves typical of the wet season with long-lived leaves in the dry season (see Appendix A in ESM for the longevity of leaf and photosynthetic stems of the lignified species sampled here).

To test whether topographic differences affect water availability and the δ¹³C and δ¹⁵N values of plants in arid environments, five wet season desert sites were sampled. Plants were collected in each of these sites from dry wash beds (wadis) and adjacent exposed ridges. Data collected from exposed ridges or dry washes at other arid sites were also included in the sample. Since desert plants that grow in fissures on smooth dolomite rock outcrops (DWD) also trap water, plants collected from these locations were also included with the dry wash sites (see Table 1 for DWD sites; Danin 1983). The eastern dry wash desert sites in this study are fed by locally drained rainwater rather than spring water or major washes that might be contaminated by N from fertilizers and sewage. The southern dry wash desert sites drain water from the higher Negev hills which do not support agriculture or major human settlements (Bedard-Haughn et al. 2003; Segal-Rozenhaimer et al. 2004; for site locations see Fig. 1).

Plant processing

The plants were identified to the species level in the field and were then dry-pressed and frozen and shipped to Harvard University for analysis. In the laboratory plant leaves and other non-lignified parts were ground in liquid nitrogen using a mortar and pestle. The plant species (n = 241) were then collapsed into the following major Raunkiær life forms: annuals (therophytes n = 104), forbs (hemicryptophytes n = 36), geophytes (n = 21), dwarf shrubs (chamaephytes n = 56), shrubs and trees (phanerophytes n = 24; Danin and Orshan 1990). Trees and shrubs were collapsed into a single group since semi-arid Mediterranean trees often branch close to the ground making the two groups functionally and in many cases taxonomically similar. Geophytes were collected only in the wet season since in most cases their above-ground growth is limited. Dry and green plant specimens collected in the dry season of 2006 were separated during processing and analysis. This strategy was designed to illuminate the isotopic differences between wet and dry season plants—specifically, what part of the dry season isotopic pattern is derived from growth that reached completion during the previous rainy season (represented by dry plants), and what part also included growth under dry season conditions (for green and dry plant list see Appendix A in ESM). Other than a few dwarf shrubs (n = 3), all of the dry plant species collected in the dry season (n = 70) were annuals or forbs. An equal contribution of the different species that comprise each functional group was ensured by weighing standardized amounts (20 mg) of dried plant material from each species into a reaction tube prior to homogenization. The plant mixtures were homogenized using a micro bead-beater and the product was weighed (2 mg) in duplicate into tin boats and shipped to Boston University Stable Isotope Laboratory for analysis. (For a complete list of collected plant species see Appendix A in ESM.)

Isotopic and C/N ratio analyses

Plant samples were analyzed using an automated continuous-flow isotope ratio mass spectrometer (Preston and Owens 1983). The samples were combusted in a EuroVector Euro elemental analyzer. The combustion gases (N₂ and CO₂) were separated on a GC column, passed through a GVI diluter (GV Instruments) and reference gas box, and introduced into the GVI IsoPrime isotope ratio mass spectrometer. Water was removed using a magnesium perchlorate water trap. Ratios of ¹³C/¹²C and ¹⁵N/¹⁴N were expressed as the relative per mil (‰) difference between the samples and international standards [Vienna Pee Dee belemnite (V-PDB) carbonate and N₂ in air] where:

$$ \delta X = \, \left( {R_{\text{sample}} / \, R_{\text{standard}} - 1} \right) \, \times 1,000\left( \permille \right) $$

where X = ¹³C or ¹⁵N and R = ¹³C/¹²C or ¹⁵N/¹⁴N

The sample isotope ratio is compared to a secondary gas standard with an isotope ratio that was calibrated to international standards. For ¹³C_V-PDB the gas was calibrated against NBS 20 (Solnhofen limestone, −1.05 ± 0.02‰). For ¹⁵N_air the gas was calibrated against atmospheric N₂ and IAEA standards N-1, N-2 [(NH₄)₂SO₄, 0.4 ± 0.2‰ and 20.3 ± 0.2‰ respectively], and NO ^–₃ (KNO₃, 4.7 ± 0.2‰). All international standards were obtained from the National Bureau of Standards in Gaithersburg, Maryland. Continuous flow also reported mass% C and mass% N data from which the C/N ratio was calculated. The %C and %N values were calibrated against known quantities of the peptone and glycine standards.

Rainfall and climatic data acquisition

Mean rainfall data from weather stations located within 10 km of plant collection sites were obtained from the Israeli Meteorological Service for the years 1961–1990. Additional information from weather stations was obtained from the meteorological service of the Israeli Ministry of Agriculture (http://www.moag.gov.il/csv/index.html). Rainfall data for plant collection sites that were not located in the immediate proximity of weather stations were taken from the Hebrew University of Jerusalem GIS center database for the same years. The rainfall data were extrapolated from weather stations using the inverse distance weight algorithm method. Rainfall amounts can vary greatly from year to year in arid regions. Since the dry season plant data reflect growth in the 2005/2006 wet season while the wet season data derive from 2006/2007, the total amount of rainfall for the 2 years was compared for 13 sites. Data for this analysis were taken from the Israel Central Bureau of Statistics (ICBS) (http://www.cbs.gov.il/reader/shnaton/templ_shnaton.html?num_tab=st01_04&CYear=2007; http://www.cbs.gov.il/reader/shnaton/templ_shnaton.html?num_tab=st01_04&CYear=2008). The difference in rainfall for the 2005/2006 and 2006/2007 wet seasons was insignificant (paired t test, P = 0.509), and can therefore be treated as representative of a single year.

The high resolution rainfall data available for this study from the GIS database cover a range of 30 years (1961–1990). Rainfall values averaging the 4 years leading up to and during the study period (2003–2006) would have been preferred, since they best represent water availability during the growth of the sampled plants, but these data are not currently available in GIS format. It was thus important to test whether the long- and short-term rainfall data sets are comparable. The same ICBS dataset of 13 sites used in the previous comparison was used to test the question. The amount of rainfall measured between 2003/2004 and 2006/2007 was significantly lower than the mean annual rainfall measured between 1961 and 1990 (paired t-test, P = 0.014), but rainfall from the two groups was also highly correlated (y = 0.94x − 9.5, R ² = 0.95, P < 0.0001). The intercept of the regression line formed between the recent years’ and long-term rainfall data sets indicate a small yet consistent difference of approximately 10 mm/year. While the mean annual rainfall in recent years was smaller than the mean value for the years 1961–1990, the distribution of rainfall appears to be uniform across climatic regions. Next, the effect of a 10-mm difference in rainfall on the regression equations presented in the “Results” section (see below) was investigated. The effect of a 10-mm decrease in the mean annual rainfall on the predicted plant isotope values is approximately 0.01–0.1‰ for δ¹³C and 0.04–0.36‰ for δ¹⁵N. The low end of the range represents the wet regions (1,000 mm/year) while the high end of the range represents more arid regions (100 mm/year). These differences are very small and fall within the 95% confidence intervals for the expected isotopic values presented in the “Results” section. The use of the 30-year-averaged data is thus not expected to have had any significant effect on the results presented here, and they are preferred for their substantially more robust coverage of the study area.

Statistical analysis

The relationship between rainfall and the δ¹³C and δ¹⁵N values of plants was examined using linear regression analysis. The logistic trend-line formed between mean annual rainfall and plant isotopic composition was corrected using the natural log (ln) of the rainfall data; 95% confidence intervals were calculated for the resulting trend-lines. Paired t-tests were used to detect seasonal and topographic differences in plant δ¹³C and δ¹⁵N data originating from the same site. Seasonal and topographic differences among sites were investigated using a t test. Non-normally distributed data were examined using a Mann–Whitney rank sum test. Seasonal differences between regression slopes and intercepts (elevations) were analyzed using a t test (Zar 1999). The regression lines of multiple plant life forms were compared following the analysis of covariance procedure detailed in Zar (1999).

Results

Correlation between rainfall and the isotopic composition of plants

Both the mean δ¹³C and δ¹⁵N values of C₃ plants were negatively correlated with rainfall (Table 2). The omission of arid dry wash sites from the δ¹³C analysis and exposed ridge sites from the δ¹⁵N analysis (see below) resulted in higher correlations (Table 2).

Table 2 The linear relationship between mean C (δ¹³C) and N isotopic composition (δ¹⁵N) values and the C/N ratio from C₃ plant sites and mean annual rainfall

There are significant correlations between the δ¹⁵N values of each of the four life forms and rainfall (annuals n = 32, R ² = 0.38, P < 0.001; forbs n = 28, R ² = 0.34, P < 0.001; dwarf shrubs n = 32, R ² = 0.54, P < 0.001; shrubs and trees n = 19, R ² = 0.77, P < 0.001). The correlation between rainfall and the δ¹³C values of each of the plant life forms, however, was significant only for annuals and dwarf shrubs; and only when desert dry wash sites were excluded from the analysis (annuals n = 27, R ² = 0.40, P < 0.001; forbs n = 24, R ² = 0.13, P = 0.090; dwarf shrubs n = 24, R ² = 0.33, P = 0.004; shrubs and trees n = 19, R ² = 0.30, P = 0.055; but see seasonal data below).

No relationship was found between the δ¹³C values of C₄ plants and rainfall (n = 24, R ² = 0.0007, P = 0.905; Appendix B in ESM). The δ¹⁵N values for C₄ plants were consistently higher than the values measured for C₃ plants from the same site (+2.0‰ difference; paired t test, P < 0.001; Appendix C in ESM). Since most of the C₄ plant species in this study are dwarf shrubs (chenopods) they were also compared to C₃ dwarf shrubs from the same sites with similar results (+2.1‰ difference; paired t test, P < 0.001). The regression slopes of the δ¹⁵N of C₃ plants on rainfall and the δ¹⁵N of C₄ plants on rainfall were statistically similar (P = 0.936).

Seasonal variability in isotopic composition and C/N ratio of C₃ plants

The mean δ¹³C values of C₃ plants collected in both the wet and dry seasons were negatively correlated with rainfall (Table 2). The regression slopes of rainfall and the δ¹³C values of C₃ plants collected in the wet versus dry season were significantly different (P < 0.05), with the dry season site means exceeding wet season site means (paired t test, P < 0.001; Fig. 2a). When green and dry plants collected in the dry season were separated, the slope and intercept of the regression line formed between the δ¹³C values of dry plants and rainfall were similar to those for the wet season sites (slope P = 0.99; intercept P = 0.22; Fig. 3a). The slope and intercept of dry season green and dry plants were significantly different (slope P < 0.05; intercept P < 0.01; Fig. 3b), with the δ¹³C values of green plants being consistently higher than those of dry plants. The green to dry plant ratio in plants collected in the dry season sites shows a positive correlation with rainfall. Green plants thus comprise a larger portion of the total plants collected in wetter regions compared to drier regions (Fig. 3c). The latter observation demonstrates that the dry season regression line slope and intercept that include all plants are driven by green plants in wetter environments while dry plants are more influential in drier environments.

Fig. 2

Correlation between mean annual rainfall, mean site isotopic composition, and C/N ratio of C₃ plants in the wet and dry seasons. Plants collected in the 2006 dry season indicated by triangles, dashed regression lines; plants collected in the 2006/2007 wet season indicated by squares, solid regression lines; 95% confidence intervals of the regression lines indicated by dotted lines. a Mean site C isotopic composition (δ¹³C) values versus rainfall; closed symbols sites with >150 mm rainfall/year, gray symbols exposed ridge sites, open symbols dry wash sites (not included in the regression analysis). b Mean site N isotopic composition (δ¹⁵N) versus rainfall; closed symbols sites with >150 mm rainfall/year, open symbols dry wash sites, gray symbols exposed ridge sites (not included in regression analysis). Note that the dry and wet season regression lines overlap exactly. c C/N ratio versus rainfall for all sites. For statistics see Table 2

Fig. 3

Contribution of dry and green plants to the relationship between δ¹³C values of dry season plants and rainfall. a Dry season, C isotope site means of dried plants indicated by triangles, gray regression line; wet season site means indicated by squares, solid regression line with 95% confidence intervals indicate by dotted lines (dry wash sites not included in regression analysis). b Dry season site mean regression line indicated by bold solid line is compared to the regression of green (open triangles, hatched regression line) and dry plants (closed triangles, dotted regression line) with 95% confidence intervals (dry wash sites not included in regression analysis). c Correlation between green/dry plants ratio per site (number of species from each group) during dry season and mean annual rainfall

Higher δ¹³C values were also found in dry than in wet season C₃ plants when they were separated into life forms (paired t test, annuals P < 0.05; forbs P < 0.01; dwarf shrubs P < 0.01; shrubs and trees P < 0.001; Fig. 4). The intercepts of the regression lines of rainfall on the δ¹³C values of wet versus dry season plants were also significantly different when each of the four life forms was analyzed independently (Table 3). The regression slopes, however, do not show any seasonal difference (Table 3). In summary, the seasonal δ¹³C data indicate two primary trends. First, plant δ¹³C values increase as rainfall decreases in both the wet and dry seasons and second, δ¹³C values are significantly higher in the dry than the wet season.

Fig. 4

Relationship between mean δ¹³C and δ¹⁵N values of plants according to life form and mean annual rainfall. Plant δ¹³C regression lines exclude desert dry wash sites and δ¹⁵N regression lines exclude desert exposed ridge sites. Wet season plant sites are indicated by squares, solid regression lines; dry season plant sites are indicated by triangles, dashed regression lines. Open symbols depict dry wash sites and gray symbols depict exposed ridge sites. a annuals, b forbs, c geophytes (wet season data only), d dwarf shrubs, and e shrubs and trees. See Table 3 for statistics

Table 3 Linear regressions of seasonal isotopic composition of plant life forms and mean annual rainfall

A comparison of the regression slopes and intercepts of rainfall on the δ¹³C values of herbaceous life forms (annuals and forbs) collected in the wet versus dry season suggests that seasonal differences in δ¹³C values are more pronounced in sites with high annual rainfall (Table 3; Fig. 4). This pattern does not occur in the lignified plant groups (e.g., dwarf shrubs, shrubs and trees; Fig. 4). When the four plant life forms were compared, there was no significant difference between the regression slopes of plant δ¹³C values on rainfall in the wet and dry seasons, but there were significant differences in the intercepts of the wet season plants from the different life forms (wet season, slopes P = 0.223, intercepts P < 0.05; dry season, slopes P = 0.492, intercepts P = 0.223). The plant life forms cluster into two major groups based on their dry season regression slopes. The first group includes dwarf shrubs, and shrubs and trees which have steep regression slopes, while the second group includes annuals and forbs that have mild regression slopes (Table 3). These trends fully correspond with the above division of dry season plants into green (dwarf shrubs; shrubs and trees) and dry (annuals and forbs) groups.

The mean δ¹⁵N values of C₃ plants collected in both the wet and dry seasons were negatively correlated with rainfall (Table 2). No significant seasonal differences, however, were detected in the δ¹⁵N values of C₃ plants (paired t test P = 0.166). Furthermore, the regression slopes and intercepts for rainfall on the δ¹⁵N values of wet versus dry season C₃ plants were indistinguishable (slope P = 0.980; intercept P = 0.900; Fig. 2b). A comparison of the slopes of the wet and dry season regression lines for each of the life forms showed no difference (Table 3). In contrast, there is a borderline significant difference among the intercepts of dry season life forms (wet season, slopes P = 0.234, intercepts P = 0.945; dry season, slopes P = 0.353, intercepts P = 0.058). This difference is the outcome of a significant increase in δ¹⁵N values in dry versus wet season plants in the annuals group (paired t test, P < 0.01; Fig. 4a). Finally, there were no seasonal differences in the δ¹⁵N of C₄ plants (paired t test, P = 0.624; regression slopes P = 0.987, intercepts P = 0.963; Appendix C in ESM).

The mean C/N ratio of C₃ plants increased with increasing rainfall in both seasons (Table 2; Fig. 2c), and also correlated with δ¹⁵N values (wet season R ² = 0.43, P < 0.01; dry season R ² = 0.45, P < 0.05). Pronounced seasonal differences were found in the intercepts of the regression lines of C/N ratios on rainfall (intercept: P < 0.001), though the slopes were similar (slope P = 0.463). Seasonal variation in the C/N ratio is mostly related to large changes in %N (81%) and only minor changes in %C (8%; Fig. 5). During the wet season, the relative N content in non-lignified plant parts was much higher than in the dry season (paired t test, P < 0.001; Fig. 5). These differences were not limited to annual plants, but were evident in all other life forms including shrubs and trees which are metabolically active during the dry season (paired t test, annuals P < 0.01; forbs P < 0.001; dwarf shrubs P < 0.001; shrubs and trees P < 0.05).

Fig. 5

Seasonal trends and correlation between site means of %N (a) and %C (b) and rainfall. Dry season sites indicated by gray triangles, dashed regression line; wet season sites indicated by empty squares, solid regression line. The difference between seasons is significant (t test, %N P < 0.0001; %C P < 0.01)

Dry wash versus exposed ridge isotopic distribution in arid environments

The mean δ¹³C values of C₃ plants growing in exposed ridge microhabitats were consistently higher than those of plants growing in dry washes from the same sites (paired t test, P < 0.01; Fig. 6a). The same pattern was observed when comparing the mean δ¹³C values of exposed ridge versus dry wash C₃ plants from the same life form (annuals paired t test, P < 0.01; dwarf shrubs paired t test, P < 0.01; the absence of shrubs and trees from exposed ridges and the near absence of geophytes during the dry season prevented a similar analysis for these life forms). In one case (Metzokei Dragot) a significant difference was measured between annual grasses growing on an exposed ridge and a dry wash located merely 10 m away (e.g., Stipa capensis dry wash = −28.2‰; exposed ridge = −24.6‰). A significant topographic difference also exists when the means of all exposed ridge and all dry wash sites are compared (t test, P < 0.01).

Fig. 6

Mean δ¹³C and δ¹⁵N values of C₃ plants collected from exposed ridge (closed squares) versus dry wash (open squares) desert sites. Dry wash and exposed ridge wet season samples collected from the same location are linked by lines. a δ¹³C values, b δ¹⁵N values. Legumes (diamonds) in dry wash and exposed ridge sites are also plotted (mean = −1.05 ± 0.45‰, n = 5)

The effect of topography on δ¹⁵N values was much harder to isolate given the wide range of factors contributing to the isotopic pattern of desert soils. Nevertheless, the mean δ¹⁵N value of non-N-fixing C₃ plants from all dry wash sites was higher than for exposed ridge sites (paired t test, P < 0.05; Fig. 6b). When divided into life forms similar differences were visible between exposed ridge and dry wash sites (paired t test, annuals P < 0.01; dwarf shrubs P < 0.01; but not in forbs, P = 0.204). The mean δ¹⁵N (±1 SE) value of plants growing on exposed ridges (1.7 ± 3.2‰) is closer to that of N-fixing legumes (−1.05 ± 0.5‰) than to that of dry wash plants (5.23 ± 2.4‰; Fig. 6b).

In summary, a strong correlation between rainfall and plant δ¹⁵N values is found for both seasonal and annual data when site values are averaged or divided by plant life forms. These correlations are maintained despite a significant decrease in the N content of leaves during the transition from the wet to dry season. A strong correlation is also found between rainfall and plant δ¹³C values when the data are divided into wet and dry seasons. Likewise, the regression formed between the δ¹³C values of dried dry season plants and rainfall is similar to that of wet season plants. The slope of these regression lines is also similar to that of the regression line of green dry season plants and rainfall, but the intercept for the latter is significantly higher. In arid environments, plants growing in dry washes have significantly higher δ¹⁵N values and lower δ¹³C values than exposed ridge plants. Finally the slope of the regression line formed between the δ¹⁵N values of C₄ plants and rainfall is similar to that of C₃ plants, but the intercept is significantly higher.

Discussion

The annual and seasonal relationship between rainfall and δ¹³C values of plants

The results of this study show a significant correlation between annual rainfall and the δ¹³C values of C₃ plants collected during the wet season in the Eastern Mediterranean. The same results are obtained for the dry season, but only when green plants are removed from the analysis. The regression line formed between rainfall and dry season green plants is different yet still highly significant. The δ¹³C values of dry season green plants are consistently higher than the values of dry plants, or wet season plants (Fig. 3b). An exception is the desert microhabitat where water availability is influenced more by topography and edaphic factors than by rainfall (Fig. 6a). This pattern was repeated for all plant life forms (Table 3; Fig. 4). The correlation between rainfall and δ¹³C values is evident in both the north-south and east-west rainfall gradients measured here (Fig. 1). This result supports previous studies reporting that drought stress is a major determinant of C isotope fractionation in C₃ plants on a community level (Stewart et al. 1995; Swap et al. 2004). Regression lines formed between ln rainfall and the δ¹³C site values for C₃ plants in this study are comparable to the trends published for plant communities from other regions (Stewart et al. 1995; Austin and Vitousek 1998; Swap et al. 2004; Zheng and Shangguan 2007; Song et al. 2008; Fig. 7). Unfortunately, C data have not been compiled to describe a global trend and thus how the data presented here fit a global trend cannot be shown.

Fig. 7

Regional and local trends in rainfall and C₃ plant δ¹³C. Linear regression line showing the relationship between wet season δ¹³C values of C₃ plant sites (thick black line), and dry season green plants (thick black dashed line), with 95% confidence intervals indicated (thin black dashed line), both from the current study. These are compared to data collected at other locations in other studies including: Hawaii, y = −2.07x − 12.51, R ² = 0.83 (1 Austin and Vitousek 1998); Western Australia, y = −2.91x − 8.70, R ² = 0.69 (2 Stewart et al. 1995); y = −2.62x − 9.95, R ² = 0.65 (3 Swap et al. 2004); China loess plateau, y = −4.73x + 2.64, R ² = 0.82 (4 Zheng and Shangguan 2007); and Tibetan Plateau, y = −1.28x − 18.55, R ² = 0.76 (5 Song et al. 2008)

In contrast, the δ¹³C values of C₄ plants analyzed here are not correlated with rainfall (Appendix B in ESM; for theoretical background see Farquhar 1983). The relative contribution of C₄ plants to plant communities, however, is correlated with rainfall in the studied region. C₄ plants first become noticeable in areas receiving less than 350 mm of mean annual rainfall. The number of C₄ species increases as annual rainfall decreases (Vogel et al. 1986). The presence of C₄ plants in the Eastern Mediterranean is an indicator of arid environments. Further work is needed to establish measures to separate the δ¹³C values of drought-stressed C₃ plants from C₄ plants in arid environments.

One of the most important results of this study is the strong dichotomy in the δ¹³C values of plants collected in the wet versus dry seasons (Figs. 2a, 4). This pattern was first identified in North American annual grasses and forbs (Smedley et al. 1991; Ehleringer et al. 1992). In this study, this seasonal pattern is recorded in all plant life forms across both north-south and east-west rainfall gradients (Table 3; Fig. 4). It is possible that it derives from the striking seasonal rainfall partitioning in the region of study. Although the gradients are characterized by dramatic differences in rainfall during the wet season, the dry season is hot, dry and prolonged.

The gentle regression slope and weaker correlation between rainfall and the δ¹³C values of dry season plants is a product of combining dry and green plants that co-occur in sample collection sites (Fig. 3). Dry plants better reflect the conditions of the previous rainy season when these plants grew (Fig. 3a). Green plants during the dry season, however, form a separate regression line marked by significantly elevated δ¹³C values compared with those from the wet season (Fig. 3b). This trend suggests that green plants (dwarf shrubs, shrubs and trees) maintain a certain level of metabolic activity under increasing drought stress conditions. Elevated δ¹³C values do not seem to be associated with senescence in dry season mature leaves (Kao et al. 2002). The strong correlation between dry season green plants and rainfall is probably also the combined product of ground water availability, evaporative demand, and improved WUE in desert shrubs and trees. Escudero et al. (2008) demonstrated that long-lived leaves from lignified species have higher δ¹³C values than short-lived leaves, due to reduced transpiration. They argued that this adaptation was designed to combat seasonal and annual shortages of water by minimizing the risk of leaf desiccation. This argument can also be extended to dwarf shrubs that shift from short- to long-lived leaves between the wet and dry seasons (Appendix A).

The relationship between rainfall, C/N ratio and plant δ¹⁵N

The mean δ¹⁵N value of non-N-fixing C₃ plants is highly correlated with mean annual rainfall when arid exposed ridge sites are excluded from the analysis (see discussion on desert N below; Table 2; Fig. 2b). The regression lines formed between rainfall and the δ¹⁵N values of wet and dry season plants were identical, demonstrating year-round stability in the δ¹⁵N isotopic values of C₃ plant communities. The correlation between rainfall and plant δ¹⁵N values agrees with studies that demonstrated local and global plant ¹⁵N enrichment along negative rainfall gradients (Handley et al. 1999; Swap et al. 2004; Fig. 8). Nevertheless, there is a difference in the regression slope of the Eastern Mediterranean and the regional and global trends that were reported in Handley et al. (1999) and Swap et al. (2004; see also references therein). The wetter sites sampled in this study on average had low δ¹⁵N values compared with the global trend, but these values fall within the range of variation reported in the global trend.

Fig. 8

Global and local trends in the relationship between rainfall and the N isotopes of C₃ plants. Local relationship between rainfall and the annual δ¹⁵N values of C₃ plant sites (open squares), represented by linear regression line (thick black line) with 95% confidence intervals (dashed lines). Exposed ridge desert plant sites are indicated by gray squares. Regression line (thin black line) showing the global relationship between rainfall and the δ¹⁵N values of C₃ plant sites (open diamonds) n = 68, y = −2.22x + 16.07, R ² = 0.36, P < 0.0001 (data taken from Handley et al. 1999; Swap et al. 2004)

The absence of seasonal differences in the δ¹⁵N values of all of the C₃ plant life forms with the exception of the annual grasses indicates that the physiological response of plants to seasonal drought stress does not cause ¹⁵N enrichment in non-lignified plant stems or leaves. Plant physiology has been argued to be at least partially responsible for increased δ¹⁵N values in drought-stressed annual plants in controlled studies (Lopes et al. 2004; Lopes and Araus 2006, for the opposite trend see Robinson et al. 2000). The absence of enrichment in other plant life forms may be related to two processes. First, while annual grasses are fully dependent on external N sources, other plant life forms also use internally stored N (Chapin et al. 1990). It is possible that the use of internal N sources late in the growing season dampens the increase in δ¹⁵N values. Second, the pattern may be related to the senescence of plant parts. The increase in the C/N ratio of dry season plants shown here is primarily caused by a decrease in leaf N content and is likely associated with senescence of the analyzed plant portions. Perennial plants conserve valuable N by resorbing it from senescent portions. This process is not accompanied by ¹⁵N enrichment of the senescent parts (Garten 1993; Kao et al. 2002; Kolb and Evans 2002). However, given that N conservation is not essential in annual plants, N loss due to senescence may be more pronounced in annuals than in other life forms, leading to higher δ¹⁵N values in the dry season. Since the dry season plants were collected late in the season, it cannot be determined whether the lack of internal N sources or plant senescence caused the observed pattern in annual plants.

Although C₄ chenopods had consistently higher δ¹⁵N values than C₃ plants, the regression slopes of rainfall on the δ¹⁵N values of C₃ plants were similar to those of C₄ plants. To date, only data from Wooller et al. (2005) have shown higher δ¹⁵N values in C₄ than C₃ plants in a given microenvironment (but see Aranibar et al. 2008). Although the explanatory mechanism for this phenomenon remains unclear, the elevation of δ¹⁵N values in C₄ chenopods must be taken into account when reconstructing the relative contribution of C₄ plants to terrestrial food webs.

High δ¹⁵N values in arid region plants were accompanied by low C/N ratios primarily because of higher N content in non-lignified portions (Fig. 2). This result is not unexpected since the leaves of plants growing in areas with low annual rainfall are known to have higher N content per leaf mass than plants growing in wetter conditions (Wright et al. 2004). The low C/N ratios and high δ¹⁵N values of plants from arid regions and the correlation between C/N ratio and plant δ¹⁵N values, further strengthen the argument that high plant δ¹⁵N values are a product of aridity rather than anthropogenic N input into the environment.

The isotopic pattern in desert dry washes versus exposed ridges

As expected, the δ¹³C values of plants growing on exposed ridges were higher than those growing in dry washes at the same site, while the opposite was true of the δ¹⁵N values (Fig. 6). Variation in water availability caused by topographic differences in arid regions is likely interfering with an otherwise clear relationship between rainfall and plant isotopic composition (Table 2). While significant differences are apparent between dry wash and exposed ridge plants, this study was not designed to investigate the mechanisms that govern these differences. Instead, we propose an explanation for these results based on published data. Dry washes collect water from surrounding areas and are often located closer to aquifers than exposed ridges and thus water availability is much greater than that provided by annual rainfall alone. In addition, in most cases the dry washes are better sheltered from desiccating winds and less exposed to direct solar radiation. Because water availability is heightened in the growing season, dry wash plants can afford to relax water conservation strategies unlike plants growing on exposed ridges that receive water only from rainfall (Ehleringer and Cooper 1988). This results in more negative δ¹³C values in dry wash than exposed ridge plants. The isotopic differences between exposed ridge and dry wash plants were upheld when plants were separated by life form despite diverse water utilization and growth strategies (Fig. 4).

The dichotomy in the δ¹⁵N values of desert plants from dry wash versus exposed ridge sites might be related to previously observed differences in N cycling in semi-arid and arid regions (Hook and Burke 2000; Aranibar et al. 2004; Zaady 2005). The combination of relatively high temperatures and high water availability in dry washes during rain events stimulates the rapid proliferation of soil microorganisms that deplete below-ground oxygen sources (Farrell et al. 1996). Under these temporarily anaerobic conditions, denitrifying bacteria scavenge O from soil NO₃₋ causing the rapid volatilization of ¹⁵N-depleted N₂, thus increasing the δ¹⁵N value of the residual soil N (McKeon et al. 2005). Moreover, the runoff water that concentrates in dry washes also transports more N further stimulating N volatilization (Peterjohn and Schlesinger 1991; Zaady and Shachak 1994). The outcome of these processes is the profound ¹⁵N enrichment of the soil in dry washes. The N from these soils is assimilated by plants growing in dry washes resulting in the patterns of ¹⁵N enrichment shown here. Unlike dry washes, the N in exposed ridge locations is predominantly supplied by N-fixing microorganisms that inhabit the soil crust as well and by eolian deposition (West 1990; Evans and Ehleringer 1993; Böhlke et al. 1997). The lack of water probably limits further denitrification of soil N and consequently plants obtain N depleted in ¹⁵N and more similar to atmospheric N values.

Conclusion

The negative relationship between mean annual rainfall and plant N and C isotope composition shown in the current study is in agreement with other regional studies. A secondary trend indicating an increase in mean δ¹³C values of dry season C₃ plants is attributed to drought stress, which is associated with a decrease in leaf N content. Mean δ¹⁵N values, however, remain stable regardless of seasonal drought stress and are likely attributable to long-term processes in soil N cycling. Variability in the δ¹³C and δ¹⁵N values of C₃ plants in arid regions is caused by topographic differences in water availability independent of mean annual rainfall. These results indicate that water availability is the primary factor controlling the δ¹³C and δ¹⁵N values of plant communities in the Eastern Mediterranean.