Variation in carbon isotope discrimination in Cleistogenes squarrosa (Trin.) Keng: patterns and drivers at tiller, local, catchment, and regional scales.

Understanding the patterns and drivers of carbon isotope discrimination in C(4) species is critical for predicting the effects of global change on C(3)/C(4) ratio of plant community and consequently on ecosystem functioning and services. Cleistogenes squarrosa (Trin.) Keng is a dominant C(4) perennial bunchgrass of arid and semi-arid ecosystems across the Mongolian plateau of the Eurasian steppe. Its carbon isotope discrimination (((13))Δ) during photosynthesis is relatively large among C(4) species and it is variable. Here the ((13))Δ of C. squarrosa and its potential drivers at a nested set of scales were examined. Within cohorts of tillers, ((13))Δ of leaves increased from 5.1‰ to 8.1‰ from old to young leaves. At the local scale, ((13))Δ of mature leaves varied from 5.8‰ to 8.4‰, increasing with decreasing grazing intensity. At the catchment scale, ((13))Δ of mature leaves varied from 6.2‰ to 8.5‰ and increased with topsoil silt content. At the regional scale, ((13))Δ of mature leaves varied from 5.5‰ to 8.9‰, increasing with growing-season precipitation. At all scales, ((13))Δ decreased with increasing leaf nitrogen content (N(leaf)). N(leaf) was positively correlated with grazing intensity and leaf position along tillers, but negatively correlated with precipitation. The presence of the correlations across a range of different environmental contexts strongly implicates N(leaf) as a major driver of ((13))Δ in C. squarrosa and, possibly, other C(4) species.

Introduction

Grasslands have been experiencing rapid shifts in botanical composition, structure, and functioning driven primarily by global climate change and human disturbances (Alward ; Grime ; Sage and Kubien, 2003). Such changes may also shift the dominance of the two distinct photosynthetic pathways, C3 and C4, which differ in the mode of initial CO2 fixation and have profound effects on ecosystem processes (Farquhar ; Hunt ; Sage and Kubien, 2003; Kuzyakov ). Most of the world's C4 species are grasses (Teeri and Stowe, 1976; Ehleringer ). The ratio of C3 to C4 photosynthetic types is an important component in understanding ecosystem processes because it can affect biomass production, carbon sequestration, and water cycling (Tieszen ; Pyankov ; Sage and Kubien, 2003). It can also influence the terrestrial carbon isotope signal, which is used for the land–ocean partitioning of CO2 fluxes (Still ; Suits ).

The actual C3/C4 ratio is often obtained from the stable carbon isotope compositions of C3 and C4 plants and bulk samples in a two-member mixing model (Murphy and Bowman, 2007), where bulk samples may comprise soil organic matter (Wittmer ), n-alkanes (Zhang ), wool (Auerswald ), or other materials. This approach is based on the difference in discrimination against the heavy carbon stable isotope (Δ) during photosynthesis between C3 and C4 plants (Farquhar ). A single mean C4 end-member value is usually assumed in the two-member mixing model (Murphy and Bowman, 2007; Wittmer ), whereas variable C3 end-member values are estimated, based on the well known correlations between the Δ of C3 plants and environmental conditions (Mole ; Männel ; Wittmer ). The reason for this is that Δ of C4 plants is reported to be less responsive to environmental variables than that of C3 plants (Van de Water ; Murphy and Bowman, 2009). However, this view has been challenged by a number of studies that have examined variations in Δ of C4 plants, and which suggest that the Δ of C4 species can vary with environmental variables (Henderson ; Buchmann ; Wang ) and that this variability is related to ecophysiological responses and resource efficiencies (Henderson ; Ghannoum ; Wang ). For instance, the 13Δ of Amaranthus cruentus was ∼6‰ higher in low-light leaves than in high-light leaves (Tazoe ). The 13Δ of Bothriochloa ischaemum increased 2‰ with an increase in annual precipitation from 350 mm year−1 to 700 mm year−1 (Liu ). In Saccharum spp., Δ declined by 2‰ with increasing leaf nitrogen content (Nleaf) in a nitrogen-stress experiment (Meinzer and Zhu, 1998).

The Δ of a plant can be obtained from the carbon isotope composition of a plant (δ13Cplant) and that of the air (δ13Cair) in which it grows:

13Δ of C4 plant biomass is influenced by many factors, including isotope effects during diffusion of CO2 through the boundary layer, stomatal pore, cell walls, membranes, and in the liquid phase, dissolution, hydration by carbonic anhydrase, fixation of bicarbonate by phosphenolpyruvate carboxylase (PEPc) in mesophyll cells, fixation of CO2 by Rubisco in bundle sheath cells, leakage of CO2 from bundle sheath cells to mesophyll cells, and carbon allocation and partitioning, including photorespiration and dark respiration (Farquhar, 1983; von Caemmerer and Furbank, 2003). This complexity has not been analysed in full detail, mainly due to lack of instrumentation and experimental protocols for the measurements of all the necessary variables. To solve the problem, Farquhar (1983) proposed a simplified, but testable version of the model, which was subsequently used widely to investigate the influence of the supposedly strongest drivers of variation in Δ. This model states that Δ of C4 plants depends on bundle sheath leakiness (Ф, the proportion of C fixed by PEP carboxylation, which subsequently leaks out of the bundle sheath) and the ratio of internal and ambient CO2 concentrations (Ci/Ca):

where a is the discrimination of C during diffusion of CO2 through stomata (4.4‰), b3 is the discrimination during fixation by Rubisco [27‰ for C4 plants; von Caemmerer ], and b4 is the discrimination during hydration of CO2 to HCO3– and fixation by PEPc.

Eqn (2), predicts that Δ variation during photosynthesis of C4 plants depends on b4, Ф, and on Ci/Ca. These variables in turn depend on a host of environmental conditions. Temperature is the biggest influence on b4, inducing variation from –4.8‰ at 33.8 °C to –6.2‰ at 21.0 °C (Henderson ). Ф is an important ecophysiological trait that is directly associated with photosynthetic quantum yield and nitrogen use efficiency of C4 plants (Henderson ; Ghannoum ; Wang ). Any CO2 that leaks from bundle sheath cells decreases the energetic efficiency of C4 photosynthesis, through futile costs of ATP for the regeneration of PEP (Furbank ). Ф depends on the balance of the biochemical capacities of the C4 cycle in the mesophyll cells and the C3 cycle in bundle sheath cells, and the CO2 conductance of the bundle sheath exterior wall (von Caemmerer and Furbank, 2003). Factors that could affect Ф include the activity ratio of Rubisco to PEPc (Farquhar, 1983), such as Nleaf (Sage ), temperature (Pittermann and Sage, 2001), and light (Tazoe ). Ci/Ca depends on the ratio of stomatal conductance to photosynthetic activity and, hence, influences intrinsic water use efficiency (Ghannoum ). Because there is an interaction between the influences of Ф and Ci/Ca on Δ, Δ can increase or decrease with Ci/Ca depending on the term (b4+b3Ф–a) (Farquhar, 1983). For Ф>0.37, this term is greater than zero with Δ>4.4‰ and it increases with Ci/Ca. For Ф<0.37, this term is smaller than zero with Δ<4.4‰ and it decreases with increasing Ci/Ca. For Δ differing from 4.4‰, Δ will thus vary with Ci/Ca if Ф is constant.

Studies of the Δ of C4 plants and its response to environmental variables have been conducted primarily in grasslands in North America, Europe, and Australia (Schulze ; Murphy and Bowman, 2009; Ghannoum ) or in experiments under controlled conditions (Ranjith ; Tazoe ). Few studies of this type have been reported from the mixed C3/C4 grassland of the Mongolian plateau, where the C4 community of the semi-arid steppe is dominated by Cleistogenes squarrosa (Trin.) Keng, a C4 (NAD-ME) perennial bunchgrass. C. squarrosa has a wide range occurring at meadow steppe, typical steppe, desert steppe, and sand dune ecosystems across the Mongolian plateau. The Δ in C. squarrosa is well above 4.4‰ (Pyankov ; Chen ; Gong ) and it is known to vary with soil moisture (Chen ).

In this study the Δ variability in C. squarrosa and its controlling factors were investigated across a range of ecological scales on the Mongolian plateau. It was hypothesized that Δ variability in C. squarrosa is driven by factors controlling Nleaf, which influence Ci/Ca and Ф. Factors influencing Nleaf may include leaf age, grazing pressure, soil properties, or weather conditions, all of which vary at different scales. Hence the Δ of C. squarrosa was studied with a nested set of scales (i.e. tiller, local, catchment, and regional) to identify and quantify controls of Nleaf and, in turn, Δ.

Material and methods

Study area

The study area was located between 106.15 °E and 117.13 °E longitude and 40.62 °N and 45.57 °N latitude in the Inner Mongolia Autonomous Region, China (Fig. 1). This area of ∼200 000 km2 constitutes the regional scale. Samples were taken at altitudes ranging from 1010 to 1635 m elevation. The temperate, semi-arid climate is characterized by warm summers and cold, dry winters (Chen, 1988) with mean annual temperature (MAT) between 0.3 °C and 5.6 °C and mean July temperature (MJulT) between 18.7 °C and 22.7 °C. Mean annual precipitation (MAP) increases from 160 mm year−1 in the western part to 350 mm year−1 in the eastern part, and 60–80% of the annual total occurs in the May–August growing season (Bai ). Regional sampling took place in five different years (2005–09) to take account of the large inter-annual variation in precipitation in the region.

Fig. 1.

Nested hierarchy of sampling sites in the Inner Mongolia grassland. Leaf position along a tiller constitutes the tiller scale, plots in a grazing experiment constitute the local scale, the Xilin river basin constitutes the catchment scale, and the steppe area of Inner Mongolia constitutes the regional scale. Numbers on the tiller scale are phytomer units from base.

The Xilin river basin covers an area of 10 786 km2 in the northeast of the region. Within this catchment, MAP increases from 275 to 346 mm year−1 along a northwest–southeast precipitation gradient, with elevation ranging from 1019 to 1482 m. MAT varies from 0.3 °C in the southeast to 2.5 °C in the northwest of the basin, with MJulT ranging from 18.7 °C to 20.9 °C. The soils in the Xilin river basin include phaeozems, shallow and stony soils at rocky outcrops, and sandy soils of low water-holding capacity in sand dune areas (Wiesmeier ).

A grazing experiment (116.67 °E, 43.56 °N) near the Inner Mongolia Grassland Ecosystem Research Station (IMGERS) of the Chinese Academy of Sciences at the middle reach of the Xilin river represents the local scale. MAT, MJulT, and MAP at IMGERS are 0.4 °C, 19.0 °C, and 337 mm year−1, respectively (Zheng ).

Meteorological data and δ13C of atmospheric CO2

Growing season precipitation was geostatistically determined according to Wittmer for each sampling location and time of sampling. This was done by interpolating between meteorological stations and taking into account 1.5×1.5 km2 resolution maps of MAP, mean monthly precipitation, MJulT, and MAT (Climate Source, Inc., Corvallis, OR, USA). The mean daily precipitation (mm d−1) between 1 May, which is the date that usually corresponds with the onset of growth of C. squarrosa (Liang ), and the day of sampling was calculated. This measure of precipitation takes account of the fact that samples were collected on different days and in different years.

δ13Cair was obtained from a long-term monitoring station in Ulan Uul (Climate data online; NOAA NNDC, 2009), Mongolia, ∼460 km northwest of IMGERS.

Soil analyses

On the catchment scale sampling locations were selected to cover the full range of soils as known from soil mapping campaigns. Soil pits were dug to 1 m at locations representing the mapping units and the soils were sampled horizon-wise. A comprehensive set of soil properties was measured (for details see Wiesmeier et al., 2011), including bulk density, texture, pH, and nitrogen content. Either topsoil properties (e.g. silt content) because water uptake by C. squarrosa is restricted to the topsoil (Yang ), or whole-soil properties (e.g. N stocks) were used.

Grazing experiment

The grazing experiment had 28 annually grazed plots (Fig. 1, grazing experiment) within two experimental blocks differing in topography (flat and gently sloping). The plots had been grazed between June and September each year since 2005 at seven stocking rates: 0, 37.5, 75.0, 112.5, 150.0, 187.5, and 225.0 sheep km−2 with three times higher stocking densities during the 4 months of grazing. Further information about the grazing experiment can be found elsewhere (Schönbach ).

C. squarrosa sampling

The tiller scale was sampled in an ungrazed plot, where no leaves were lost by grazing. Leaves from several individuals were sampled and pooled according to their phytomer position (Fig. 1). Leaves were collected in mid-July, end of July, and mid-August 2008. Resampling over time allowed us to distinguish between new leaves and old leaves grown between sampling dates. Old leaves at low phytomer positions were then compared between consecutive sampling dates to identify changes over time.

At the local, catchment, and regional scales, mature, fully expanded, and sun-exposed leaves from several non-flowering individuals were sampled. At the local scale, leaves were collected from the plots of the grazing experiment in June, July, and August of 2007, and in May, June, and July of 2008. These two years differed in precipitation: 2007 was dry with 240 mm year−1, and 2008 was an average year with 362 mm year−1. At the catchment and regional scales, samples were collected in each of the five years 2005–09. The precise locations and elevations of all sampling positions were recorded with a mobile GPS. In total, 51, 144, 64, and 47 samples at the tiller, local, catchment, and regional scales, respectively, were collected.

Isotope and Nleaf analysis

All samples were dried at 60 °C for 48 h in a forced-draught oven and ground to homogeneity with a ball mill. Carbon isotope composition and nitrogen content were then measured using an elemental analyser (NA 1110; Carlo Erba, Milan, Italy) interfaced (ConFlo III; Finnigan MAT, Bremen, Germany) to an isotope ratio mass spectrometer (Delta Plus; Finnigan MAT) following the protocol by Wittmer . Carbon isotope data were specified as δ13C relative to the Vienna Pee Dee Belemnite standard:

where Rsample and Rstandard are the ratios of C/12C in the sample and standard.

The precision for sample repeats was better than 0.15‰ for δ13C and 0.04% for nitrogen content in dry matter.

Statistical analysis

Linear regressions were used to evaluate relationships between Δ, Nleaf, and the environmental variables of interest at a certain scale: leaf position, grazing intensity, soil properties, and growing season precipitation. The coefficient of determination was tested with a two-sided test for significance of the regression. Confidence intervals (95%) were calculated for the relationship between Δ and Nleaf and for the population. All procedures were carried out in SPSS Version 16.0 (SPSS, Inc., Chicago, IL, USA).

Results

Relationship between Δ and leaf position at the tiller scale

The Δ of C. squarrosa decreased and Nleaf increased from the lowermost leaf to the uppermost leaf (Fig. 2). Thus, young leaves had a low Δ but high Nleaf. The Δ mainly changed from leaf 5 to younger leaves, while Nleaf mainly changed among leaves older than leaf 7. Hence in a bulk sample of undefoliated plants, the variation in Δ can mainly be attributed to younger leaves whereas the variation in Nleaf can mainly be attributed to older leaves.

Fig. 2.

Influence of leaf position (phytomer 1 is the first leaf from base) on Δ and Nleaf of C. squarrosa. Means of three sampling dates in an ungrazed plot; error bars denote the standard error. Note that more than seven phytomers were found in August and only one sample was available for phytomer 9.

The Nleaf of a leaf produced at a given time remained unchanged between mid-July (average=1.6%, SD=0.2%) and August (average=1.6%, SD=0.3%), which suggested no nitrogen redistribution over time. Nleaf of the three uppermost (newly grown) leaves (average=2.4%, SD=0.1%) was significantly higher (P<0.01) in August than those grown in July (average=1.7%, SD=0.2%).

13Δ, stocking rate, and topography at local scale

At the local scale,Δ varied between 5.8‰ and 8.4‰ with an SD of 0.5‰. Nleaf varied between 1.8% and 3.8% with an SD of 0.4%. Δ decreased (R2=0.04, N=144, P<0.05) while the Nleaf increased (R2=0.11, N=144, P<<0.001) with stocking rate (Fig. 3). Δ and Nleaf also varied with year (P<0.001) and month (P<0.001), but not with topographic position (P>0.05).

Fig. 3.

Influence of stocking rate on Δ and Nleaf of C. squarrosa. Each data point shows the mean and standard error of 24 samples taken from four plots with different topographic positions and management systems.

13Δ and soil variables at catchment scale

Across the Xilin river catchment, Δ of C. squarrosa varied between 6.2‰ and 8.5‰ (SD=0.5‰), Nleaf varied between 1.5% and 3.7% (SD=0.5%), and topsoil silt content varied from 2% to 40% (SD=10%). Δ increased with topsoil silt content (R2=0.42, N=24, P<0.001) despite there being differences in slope between the two years (P<0.05), but Nleaf was not correlated with silt content (P>0.05) (Fig. 4). Similar patterns were found with soil nitrogen stocks. The Δ increased significantly with soil nitrogen stocks (R2=0.40, N=25, P<0.01), but there was no relationship between soil nitrogen stocks and Nleaf (P>0.05). The influence of soil nitrogen stocks and silt content on Δ could not be separated due to their correlation with each other; this also applied to other soil properties.

Fig. 4.

Influence of topsoil silt content on Δ and Nleaf of C. squarrosa at the catchment scale in 2007 (filled symbols), a dry year, and in 2008 (open symbols), a normal year.

13Δ and precipitation at regional scale

As expected, the range of 13Δ of C. squarrosa (3.4‰) was considerably larger at the regional scale than at the local (2.6‰) and catchment scales (2.3‰). Precipitation during the growing season had a significant, positive effect on 13Δ (R2=0.31, N=47, P<0.001) and a negative effect on Nleaf (R2=0.56, N=47, P<<0.001) (Fig. 5). The 13Δ of C. squarrosa increased (P<0.05) whereas Nleaf decreased (P<0.01) from the desert steppe to the typical steppe, and to the meadow steppe.

Fig. 5.

Influence of growing season precipitation on 13Δ and Nleaf of C. squarrosa at the regional scale. Growing season precipitation was calculated based on year and time of sampling (see Materials and methods).

13Δ and Nleaf

At tiller, local, and regional scales, 13Δ was negatively correlated with Nleaf (Fig. 6) although the range of values and factors that affected the variation in Nleaf differed between scales (Figs 2, 3, 5). In spite of no significant relationship at the catchment scale (P>0.05), all except four of the total of 47 data points were included in the 95% confidence intervals for all pooled samples. Nleaf explained 23% of the variation in Δ when all samples were pooled. When regressed separately, the slope of the regression was significantly steeper at the tiller scale than other scales (P<0.01), but remained within the 95% confidence interval of the common regression slope for the other scales (Fig. 6).

Fig. 6.

Correlations between Δ and Nleaf of mature leaves of C. squarrosa at the regional (steppe area of Inner Mongolia), catchment (Xilin river catchment), local (grazing experiment), and tiller scales (an ungrazed plot). The central line in each panel denotes the regression for the pooled samples, the inner envelopes (solid lines) are the 95% confidence intervals for the regression, and the outer envelopes (dashed lines) are the 95% confidence intervals for individual samples.

Discussion

This study represents the first comprehensive examination of the Δ variability of a C4 (NAD-ME) perennial bunchgrass, C. squarrosa, and its controlling factors across a range of scales (i.e. tiller, local, catchment, and regional scale) on the Mongolian plateau. The results reported here have shown that Δ of C. squarrosa was large (on average: 7.2‰) and, at the region scale, it varied over a 4‰ range (5–9‰) across the Inner Mongolian steppe. This range is larger than that for other C4 species in the study area. For instance, Salsola collina, a dicot sampled together with C. squarrosa in the grazing experiment, varied by 3‰ under identical conditions (data not shown). A previous study has reported that B. ischaemum, a dominant C4 perennial bunchgrass on the Chinese loess plateau, varied by only 2‰ across its range (Liu ). This suggested that C. squarrosa could be a model species for studying the effect of environmental conditions on the Δ of C4 plants, and that Δ of the C4 community is likely to vary, largely due to the change in relative abundance of C. squarrosa.

Many studies have demonstrated that foliar δ13C of plants is negatively correlated with water availability (e.g. soil moisture and annual precipitation) across an environmental gradient (Farquhar ; Stewart ; Liu ). The mechanisms underpinning these patterns, however, have yet to be explored. The results reported here indicate that Nleaf has a strong and consistent effect on Δ of C. squarrosa at all scales and explained 23% of the variation in Δ, even though the factors influencing Nleaf varied at different scales. Therefore confidence can be placed in this relationship, and it may be expected that other environmental variables that affect Nleaf would also affect Δ in a similar way.

Factors influencing Nleaf

Factors that affect Nleaf include factors that control N availability, such as fertilization (Bai ), factors that influence N uptake, such as root density and root activity (Farooq ), and factors that relate to ontogenesis and organogenesis, such as the availability of water and nutrients, which determine the velocity of leaf expansion and leaf dimensions where grazing could contribute (Anten ).

The influence of leaf position on Nleaf may have been influenced by soil temperature, which was ∼12.6 °C in late May and increased to 21.1 °C in mid-August (average of 10 d at 10 cm depth, data from IMGERS). N uptake of C4 is especially sensitive to low temperature (Farooq ). Remarkably, neither Nleaf nor Δ of newly formed leaves changed between mid-July and August, indicating that the influence of leaf position was not caused by redistribution of N from old to young leaves. The increase in leaf nitrogen with grazing intensity is well known and relates to the faster cycling of N and decrease in mean leaf age (Mattson, 1980). Decreasing Nleaf with increasing precipitation is likely to be related to the dilution effect caused by additional plant growth. Temperature-related N uptake may also add to the effect of precipitation, as regional temperatures and precipitation are closely negatively correlated (R2=0.93, Bai ) and hence their effects cannot be separated. Finally, the Birch effect of enhanced mineralization after a pronounced dry spell (Birch, 1964; Austin ) could also contribute to improved N availability in drier areas where rainfall occurs infrequently.

At the catchment scale, the influences on Nleaf are less clear. Δ increased with silt content whereas Nleaf did not, although the relationship between Δ and Nleaf was similar for this scale as for the other scales. At this scale the variation of soil is expected to have a strong influence but, in contrast to leaf position, grazing intensity, or precipitation acting on the other scales, the soil cannot be arranged along only one single variable like silt content. Soil is an aggregation of several soil properties, which may exert influence on Nleaf and in turn on Δ.

The effect of Nleaf on Δ

In principle, the variability in Δ could be derived from: (i) discrimination associated with photosynthesis; (ii) post-photosynthetic fractionation, which includes CO2 release in dark respiration, or (iii) any other simplifications included in Eqn (2). Although the latter processes have not been studied in detail in C4 plants (but see Kromdijk ; Pengelly ), evidence for the contribution of post-photosynthetic fractionation comes from the fact that Δ obtained by online measurement of photosynthetic CO2/12CO2 exchange is often less than Δ estimated from the carbon isotope composition of leaf dry mass (Evans ; Kubásek ). For instance, Δ obtained from online measurement was ∼0.75–1.5‰ lower than that from leaf dry mass (Kubásek ). Respiration and carbon allocation are responsive to environmental conditions including temperature and water stress (Ryan, 1991; Flexas ) and thus can contribute to the observed variability. However, changing chemical composition seems unlikely to explain the correlation of Δ with Nleaf. Proteins differ isotopically from cellulose only by 4‰ (Benner ). A difference in Nleaf of 2%, corresponding to a difference in protein content of 12%, is insufficient to explain a difference in Δ of 3‰. Also, secondary compounds, which increase with leaf age and may thus relate to Nleaf, cannot explain the variability of Δ at the tiller scale, because the young leaves present in mid-July did not change in isotopic composition until mid-August. Hence variation in leaf Δ was most likely caused by the variation in discrimination during photosynthesis.

In Eqn (2), Δ mainly depends on Ф and Ci/Ca as b4 changes little within the 4 °C range in air temperature in the study area. Ci/Ca should increase with stomatal openness (Turner ), but decrease with the activity of carboxylating enzymes (Sage ; Sage and Pearcy, 1987), which should increase with Nleaf (Ranjith ; Sage ). Ф depends on the activity ratio of Rubisco and PEPc, which is influenced by the allotment of nitrogen to them (Sage ). At the local scale, Ci/Ca was measured for young mature leaves close to the top of tillers by Zheng simultaneously with our sampling. Ci/Ca varied between 0.41 and 0.66 (Table 1), similar to the range of 0.34–0.59 found by Chen et al. (2005) (Table 2). Using these Ci/Ca data and the Δ data reported here, Ф was estimated using Eqn (2). This indicated that Ф ranged between 0.56 and 0.58 (Table 1). Nleaf from this study correlated with both Ci/Ca and Ф although variation of Ф was small. Therefore, Δ seemed to be influenced mostly by Ci/Ca, which is related to water use efficiency.

Conclusions

The findings reported herein demonstrate that Nleaf is a key factor driving leaf Δ variation of C. squarrosa. Increasing Nleaf was associated with decreased Δ, which seemed to be related to improved water use efficiency. The leaf-level response of Δ to Nleaf was found at all scales from tiller to local, catchment, and regional although the reasons for Nleaf variation differed. Within cohorts of tillers, Δ decreased with leaf position. Within the field experiment site, Δ decreased with grazing intensity. Within the Xilin river catchment, Δ decreased with decreasing topsoil silt content. Across the regional study area, Δ decreased with decreasing growing season precipitation. Accordingly, it is proposed that water use efficiency increased for leaves produced later in the growing season, and that it also increased with grazing intensity. But it declined with increasing silt content and water-holding capacity of the soil and precipitation during the growing period.

Table 1.

Variation in the ratio of internal to ambient CO2 concentration (Ci/Ca) and leakiness (Ф) for the local scale and correlations with Nleaf

Variable	Minimum–maximum	Correlation with Nleaf
Ci/Ca	0.41–0.66	–0.866
Ф	0.56–0.58	+0.832

Gas exchange measurements were taken from Zheng .

Table 2.

Leakiness (Ф) and ratio of internal to ambient CO2 concentrations (Ci/Ca) of NAD-ME plants from published sources

Species	Ф	Ci/Ca	References
Amaranthus caudatus	0.37	0.43–0.46	Kubásek et al. (2007)
Amaranthus edulis	0.32	0.43	Henderson et al. (1992)
Amaranthus retroflexus	0.40	0.40–0.90	Sage and Pearcy (1987)
Atriplex lentiformis	0.54–0.62	–	Zhu et al. (1999)
Atriplex rosea	0.45	0.47	Henderson et al. (1992)
C. squarrosa	–	0.34–0.59	Chen et al. (2005)
Cynodon dactylon	–	0.12–0.28	Mantlana et al. (2008)
Eleusine coracana	0.40	0.38	Henderson et al. (1992)
Eleusine indica	0.37, 0.27	0.48, 0.58	Farquhar (1983)
Eragrostis lehmanniana	0.45–0.57	0.08–0.70	Fravolini et al. (2002), Mantlana et al. (2008)
Panicum schinzii	0.32	0.38	Henderson et al. (1992)
Total range	0.27–0.62	0.08–0.90
Mean	0.42	0.44