Divergence of above-and belowground C and N pool within predominant plant species along two precipitation gradients in north China

Divergence of aboveand belowground C and N pool within predominant plant species along two precipitation gradients in north China X. H. Ye, X. Pan, W. K. Cornwell, S. Q. Gao, M. Dong, and J. H. C. Cornelissen State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, Beijing 100093, China Systems Ecology, Department of Ecological Science, VU University, Amsterdam, De Boelelaan 1085, 1081 HV Amsterdam, the Netherlands Key Laboratory of Hangzhou City for Ecosystem Protection and Restoration, College of Life and Environmental Sciences, Hangzhou Normal University, Hangzhou 310036, China School of Biological, Earth and Environmental Sciences, University of New South Wales, Sydney, Australia University of Chinese Academy of Sciences, Beijing 100049, China


Introduction
Plant carbon economy and nitrogen economy are closely coupled (Shaver and Chapin, 1991;Aerts and Chapin, 2000), and this coupling has important feedbacks to terrestrial carbon budgets and climate (IPCC, 2007;Chapin et al., 2009).To capture and utilize light, nutrients, carbon and water, plants allocate resources both to aboveground shoots and belowground roots.Nitrogen is obtained belowground via roots and partly transported to the aboveground shoots to support photosynthesis principally through the enzyme Rubisco and other parts of the photosynthetic systems, while carbon is obtained aboveground via photosynthesis and partly transported to belowground roots to support both acquisition of water and mineral nutrients.Important differences Published by Copernicus Publications on behalf of the European Geosciences Union.
in allocation strategy occur both among species and among abiotic environments within species (Lambers et al., 1998;Poorter and Nagel, 2000).When a given resource is in short supply, plants typically allocate resources differently to compensate.If the resource in short supply is belowground for example, limited water, species and populations within species typically will allocate more resources belowground to compensate (Brouwer, 1963;Tilman, 1991;Gleeson and Tilman, 1992;Reynolds and Chen, 1996;Craine et al., 2002).So comparing dry versus mesic sites, it is well documented that the allocation of biomass shifts to a higher proportion of roots versus leaves (Orians and Solbrig, 1977;Mooney et al., 1978).
There is a key shift in the use of nitrogen when water is in short supply.Increased N allocation to leaves can support greater Rubisco-N investment per unit leaf area, which in turn can drive a greater photosynthetic capacity per unit transpiring area (Field and Mooney, 1986).Faster photosynthetic production at a given stomatal conductance increases intrinsic water use efficiency (Field et al., 1983;Buckley et al., 2002;Wright et al., 2003), which may compensate for prolonged periods of stomatal closure.Going from mesic to dry sites, species commonly allocate more N per leaf area partially by increasing N per unit leaf mass and partially by increasing the leaf mass per area (Wright et al., 2002).This pattern has been documented both among species in Australia (Wright et al., 2001) and across China (Han et al., 2011), and within species in Hawaii (Cornwell et al., 2007).
Interestingly, aridity creates a condition under which biomass may be allocated in greater quantities belowground, but N is needed aboveground to facilitate greater water use efficiency.This suggests a potential decoupling of pool sizes for C and N, above-and belowground, with the pattern in pool size determined by the relative effect sizes of the trends in biomass allocation and [N].These shifts could have implications for biogeochemical functions -especially the residence time of C in the biosphere, both via leaf quality to herbivores and litter quality to decomposers.Both herbivory and decomposition are known to be at least in part a function of tissue C : N ratio (Swift et al., 1979;Austin and Vitousek, 2000;Perez-Harguindeguy et al., 2000).Here we argue that a particular consequence of increasing divergence of above-versus belowground N and C pool with increasing aridity could be accelerated by land degradation; as many semi-arid and arid lands are grazed by livestock, the proportional removal of a plant's aboveground N would increase with aridity, thereby potentially leading to faster exhaustion of whole-plant N to support productivity.Such positive feedback to land degradation by livestock would be a significant and previously unrecognized consequence of increased allocation of N to leaves to increase water use efficiency.
We investigate the above-and belowground pool sizes for C and N in the predominant plant species across two precipitation gradients in the semiarid to arid region of northern China.We then decompose the pool size patterns into the Going from mesic to dry sites, the allocation of biomass shifts to a higher proportion of roots versus leaves and higher leaf N concentrations.
respective drivers -shifts in both the allocation of biomass and in the concentration of macronutrients.Our approach is to use three predominant species, which together represent much of the biomass along these gradients.Our hypothesis is that, from mesic towards dry sites, above-and belowground C pool and N pool will diverge as a consequence of increasing biomass allocation belowground combined with greater leaf N concentration (Fig. 1).To our knowledge, this is the first large-scale field study to explicitly dissect carbon pool and nitrogen pool both between and within organs of whole plants.

Study area and species
This study was conducted along two transects, each following a major precipitation gradient, in the Inner Mongolia Autonomous Region (hereinafter, Inner Mongolia), N China (Fig. 2).The North-East China Transect (NECT) ranges from Erenhot to Chifeng city (longitude 112.12 to 117.42 • E, latitude 43 to 44 • N, altitudes 920 to 1300 m), with mean annual precipitation (MAP) ranging from approx.200 to 400 mm; The Ordos Transect (OT) stretches also from the west to east in Ordos City (longitude 107.00 to 111.31 • E, latitude 38 to 39 • N, altitudes 1200 to 1500 m), with MAP ranging from 160 to 440 mm.Along the west of NECT, temperature only shows a modest trend, with mean annual temperature (MAT) from 1 to 3 • C, mean temperature of the coldest month (CMT) from −20 to −12 • C (January), and mean temperature of the warmest month (WMT) from 19 to 24 • C (July).The precipitation from May through September accounts for almost 80 % of MAP (Zhang et al., 1997).The warmest sites have the lowest relative inner-annual variability in temperature and the coldest sites have the highest variability, while the driest sites have the highest seasonality in precipitation and the wettest sites have the lowest variability along NECT (Ni and Zhang, 2000).The Ordos Transect, about 450 km south of NECT, features a weak eastto-west temperature gradient (MAT from 6 to 9 • C, CMT from −13 to −10 • C (January), and WMT from 20 to 23 • C (July)) and a strong east-to-west precipitation gradient, with June through September accounting for 60-80 % of MAP (Niu and Li, 1992).
Artemisia ordosica Krasch (Asteraceae) is a deciduous semi-shrub with plumose, linearly lobate leaves, occurring in the Inner Mongolia Autonomous Region, Ningxia Autonomous Region and Shanxi Province of N China.Its root system is mainly distributed in the upper 30 cm of sandy soils.This species is overwhelmingly dominant in the semifixed and fixed dune landscape of the Mu Us Sandland, which is the main body of the Ordos Plateau (Zhang, 1994).Stipa grandis P. Smim is a tall bunchgrass and can grow to almost 1 m at the peak of the growing season (late August).Its aboveground parts die off in autumn and new growth is in spring.It often forms nearly pure stands in the semi-arid region of N China.Stipa bungeana Trin. is smaller than S. grandis, with height ranging from 20 to 60 cm.It is distributed in N, SW and SE China as well as in Mongolia and Japan.Both Stipa species predominate on the loamier soils in the region, as opposed to A. ordosica on the sandy substrates.

Field sampling and laboratory analysis
The field investigation was conducted from 6 August to 18 September 2006, i.e. the first 3 weeks along NECT and the latter 3 weeks along OT.Along each transect this was during the peak biomass period and well before the onset of senescence (which is in September at NECT; end of September/early October at OT).In total 46 sites were chosen along NECT and 41 sites along OT (Fig. 2), each two adjacent sites at intervals of more than 10 km.MAP at each site was extracted from the interpolated climate grids (New et al., 2002), using the Raster package in R (Hijmans and van Etten, 2011).The site values closely matched previous estimates based on regional regressions with altitude, latitude and longitude for NECT (Zhang et al., 1997) and OT (Niu and Li, 1992), respectively.
Three 1 m × 1 m quadrats for grass and/or 2 m × 2 m quadrats for semi-shrub were measured at each site.Within each quadrat, all living biomass was harvested, separated into species and oven-dried at 70 • C for ≥ 24 h to constant mass before they were weighed.For each species, cover %, height and number of individuals were also measured within each quadrat at the same time.Leaves of S. grandis in NECT and of S. bungeana and A. ordosica in OT were collected for lab analyses within each quadrat, as well as the stems of A. ordosica in OT.For each species roots were collected by taking three 10 cm diameter soil cores from 0-40 cm depths, mixed in situ into one composite sample.Roots of S. grandis in NECT and roots of S. bungeana and A. ordosica in OT were hand-sorted and washed clean for lab analyses.It is often difficult to differentiate between the roots of different species, but it was perfectly doable to differentiate between the roots of the three target species in this study.Both S. grandis and S. bungeana are bunch-grasses and could be identified visually; their roots are fibrous and shallow.Confusion with Leymus can be excluded, as this grass is rhizomatous and has its root distribution much deeper in the soil profile than Stipa.The roots of the shrub A. ordosica were easily recognized by their strong and typical aroma as well as by their morphology.Most other species in NECT and OT were nongrass herbaceous species, such as Allium, Carex, Euphorbia, Iris, Potentilla and Taraxacum.Only roots of the three target species were hand-sorted and washed clean for lab analyses, and all other roots in NECT were mixed into one composite sample.
Leaf, stem and root biomass of the three species were measured after ≥ 24 h oven-drying at 70 • C. Leaf, stem and root potassium (K) concentration was determined using atomic absorption spectrophotometry (Model-932, GBC Scientific Equipment, Melbourne, Australia), while nitrogen (N) and phosphorus (P) concentrations were analysed according to the micro-Kjeldahl method (Kjeltec 2300 Analyser Unit, www.biogeosciences.net/12/457/2015/Biogeosciences, 12, 457-465, 2015 Sweden), and total carbon (C) concentration was determined using the method of K 2 Cr 2 O 7 − H 2 SO 4 solution digestion with oil-bath heating (Dong et al., 1996;Zheng et al., 2010).Also, standard reference materials were used throughout to guarantee the measurement accuracy and repeatability.
In total, 15 of the sites provided data for plant aboveground and underground biomass and element contents for A. ordosica in OT, 22 sites for S. bungeana in OT, and 35 sites for S. grandis in NECT.Unfortunately, it was not logistically possible to measure soil nutrient availabilities simultaneously and repeatedly at so many points along the two gradients, given that nutrient availability is notoriously variable in time.However, we expect nutrient availabilities to have varied little over each of the transects for given species, as both Stipa species were generally sampled from unfertilized loamy soils and A. ordosica generally from unfertilized sandy soils, The remarkably flat pattern for root N content with MAP for all three species (see Results, Fig. 4b) strongly suggests that there was no correspondence between soil nutrient availability and precipitation regime, although we cannot exclude effects of nutrient availability on intra-specific variability in nutrient-related traits at finer spatial scale.

Data analysis
Statistical analyses were performed using SPSS13.0software (2004, SPSS Inc., USA).We analysed variation in each trait for the effect of species and precipitation using ANCOVA, with species as an independent variable and precipitation as a covariate.And we analysed relationship between each trait and precipitation for each species using linear regression.Data for leaf [P], leaf [K], root [P], root [K], and proportion of aboveground N, C, P and K content were log-transformed before analyses in order to meet the assumptions of normality.

Results
The relative contribution of A. ordosica to the plant community biomass ranged from 13 to 86 % (mean 52.6 %) and increased with higher precipitation within the sandy region in OT (r 2 = 0.319, P = 0.035) (Fig. 3a).S. bungeana made up a higher proportion of the total biomass (range 10 to 91 %, mean 36.8 %) on the loamier OT sites where A. ordosica did not dominate.S. bungeana on the OT achieved a greater dominance compared to S. grandis in NECT (range 1 to 80 %, mean 29.8 %).For neither Stipa species was the proportion correlated with precipitation (P > 0.1).

Plant biomass and its allocation along transects
The ANCOVA results showed that the proportion of aboveground biomass increased linearly with MAP in all three species (df = 1, F = 8.135, P = 0.006); the slope of proportion of aboveground biomass as a function of MAP was not heterogeneous across species.The size of the trend, the crucial element affecting pool sizes, was a mean of 7.0 % additional allocation aboveground for every 100 mm yr −1 of additional rainfall.Also, there was a significant difference in elevation of these regressions among species (df = 2, F = 6.521,P = 0.003); S. grandis in NECT had the largest proportion aboveground biomass allocation, followed by A. ordosica and S. bungeana in OT.Only the latter species by itself increased aboveground biomass with MAP (r 2 = 0.497, P < 0.001, Fig. 3b).

Effects of species and precipitation on plant element concentrations
Precipitation had a negative and significant relationship with leaf nitrogen concentration (LNC) and leaf phosphorus concentration (LPC), as well as significant overall effects on proportion of aboveground C, P and K content (Table 1).There were significant interactions for species by precipitation on most plant traits analysed (Table 1).

Proportion of aboveground plant N, C, P and K pools
Precipitation had a significant effect on the proportion of N and C pool aboveground, and this effect differed among species (significant interactions, see Table 1).S. grandis had the highest proportion of N and C pool aboveground, followed by A. ordosica and S. bungeana respectively.There was a significant positive relationship between MAP and proportion of N (slope 0.129 % per 100 mm yr −2 , r 2 = 0.315, P = 0.007) and C aboveground content (slope 0.239 % per 100 mm yr −2 , r 2 = 0.583 and P < 0.001) in S. bungeana, while A. ordosica showed only a weaker trend and S. grandis none at all (Fig. 4d).
Both species and precipitation had significant effects on the proportion of P and K pool aboveground and also the interactions were significant (Table 1).Comparable with N pool, there were significant positive relationships between MAP and proportion of P (slope 0.212 % per 100 mm yr −2 , r 2 = 0.631, P < 0.001) and K aboveground content (slope 0.239 % per 100 mm yr −2 , r 2 = 0.620, P < 0.001) in S. bungeana, but not in the other two species (Fig. S1d).

Discussion
The aboveground versus belowground pool size and concentrations of carbon and nitrogen are linked, and understanding the shifts across a precipitation gradient requires a wholeplant perspective (Van der Putten et al., 2001;Wardle et al., 2004;Bardgett et al., 2005;De Deyn et al., 2008).Because of the many fungi and animals associated with the plantsmutualists, pathogens, herbivores, decomposers -the whole plant C and N strategy is crucial for carbon storage, food webs, soil water and nitrogen retention and cycling (Hooper et al., 2000;Bardgett and Wardle, 2003).Changes in the stochiometry and balance between these pools owing to differences or changes in precipitation regime are therefore likely to impact greatly on ecosystem function and biodiversity.
Two quantities affect the proportion of N above-versus belowground: the relative allocation of biomass and the concentration of N in that biomass.A priori we expected the proportion of biomass aboveground to be greater in mesic versus dry places (Orians and Solbrig, 1977;Mooney et al., 1978).We also expected Inner Mongolia to show similar tissue N concentration patterns to China at larger scale (Han et al., 2011), Australia (Wright et al., 2001(Wright et al., , 2002)), Hawaii (Cornwell et al., 2007), Panama (Santiago et al., 2005), and western North America (Maherali and DeLucia, 2001) -that is, the concentration of N in leaves would be greater in dry versus mesic environment.For the pool size of N, because these two components were expected to change in opposing directions, we expected the result to be a function of the relative effect sizes of the two trends.Because, to the best of our knowledge, the effect sizes for biomass and N concentration have not previously been measured in one system, we started from the null hypothesis that the two trends would cancel each other out, resulting in similar proportions of N aboveground versus belowground pool size across aridity gradients.
We found that like other aridity gradients across the world (Maherali and DeLucia, 2001;Austin and Sala, 2002;Santiago et al., 2005;He et al., 2008), there was a shift from more aboveground towards more belowground biomass allocation with decreasing precipitation in S. bungeana (Fig. 3b).Also consistent with our expectation, there was a significant increase in leaf N concentration from the drier towards the more mesic sites in S. grandis and A. ordosica, and a weaker increase in leaf N concentration in S. bungeana (r 2 = 0.167, p = 0.059) (Fig. 4a).Integrating both patterns, we found that for one species, S. bungeana, there was a significant positive trend in proportion of N aboveground with precipitation (Fig. 4d).The other two species did not show relationships that differed significantly from flat.This was neither due to differing biomass allocation, the slopes of which relationships were very similar (Fig. 3b); nor to root N concentration, which also showed similar pattern for the three plant species (Fig. 4b).Instead, S. bungeana did not increase leaf N concentration in response to aridity to nearly the extent that the other two species did (Fig. 4a).This meant that unlike in the two other species where N and biomass shifts came close to equal effect sizes, in S. bungeana the biomass effect dominated and the proportion of N aboveground increased with precipitation.
Species differ in their intraspecific variation -both for biomass allocation and the extent to which they adjust their N concentration.While intraspecific variation in leaf traits (including N concentration and leaf mass per area) has been relatively widely studied (e.g.Shipley and Almeida-Cortez, 2003;McGill et al., 2006;Ackerly and Cornwell, 2007;Cornwell and Ackerly, 2009;Albert et al., 2010;Messier et al., 2010), there is still no consensus on why species differ in the magnitude of their intraspecific variation.We suggest that intraspecific variation in a few key traits, especially N and C allocation, is crucial for understanding how much of N and biomass cycling occurs belowground versus aboveground as living biomass and then on the surface of the soil as leaf and stem litter.
In our study of three predominant species in Inner Mongolia, in two cases (clearly in S. grandis, borderline in A. ordosica) intraspecific variation in biomass allocation balanced shifts in the concentration of leaf N. In the third species, S. bungeana, biomass allocation changed, but the shift in N concentration did not fully compensate.Where S. bungeana was predominant, a greater proportion of N was held aboveground in mesic versus dry places.In contrast, where S. grandis was predominant, the pool sizes of N above-versus belowground were similar in mesic versus dry places, but the pool size of C was not.There was much more C aboveground in mesic versus dry places, but that C was accompanied with a lower concentration of N.
We propose one potentially important consequence of the decreasing aboveground allocation to biomass that is richer in N as aridity gets more severe.Logically, at the level of the plant individual, a given proportion of consumption of aboveground biomass by herbivores in a more arid site will remove a relatively large proportion of the plant's total N pool, which has to be compensated for to the extent possible by N uptake from the soil.Under strong grazing regimes, uptake may not keep up with N removal and this might lead to exhaustion of the aboveground N pool and eventually lead to shoot death, with consequences for the soil-stabilizing function of the vegetation in these erosion-prone areas.Thus, increasing divergence of above-versus belowground N and C pool with increasing aridity could accelerate land degradation; as many semi-arid and arid lands are grazed by livestock, often heavily, the dominant vegetation could drive a positive feedback on land degradation by livestock at low precipitation regimes.This situation will not be unique to Inner Mongolia, as extensive regions in a similar climate zone in at least West, Central and East Asia and North America (Fig. 2, insert) host similar vegetation types and similarly heavy use by livestock.This possibility may deserve in-depth study that would have to extend to vegetation regrowth patterns and N and C pathway and pool via livestock at different animal densities.It would also be important to test whether and how the possible feedback suggested above-might be influenced by possible changes in preference of livestock for different food plants as related to changing leaf nutritional value for the different species, with consequences for their abundance hierarchies at different points along the precipitation gradient.

Conclusions
The C and N cycle are usually tightly coupled, because the proportion of N that is needed in specific tissue types is relatively constrained -N is usually low in structural tissue and high in absorbing (roots) and photosynthetic (leaves) tissue (Sterner and Elser, 2002).Therefore, decoupling of the C and N pool across different climates is a paradox.Changing water availability potentially decouples the C and the N pool because of greater allocation to high N leaves in dry sites.Here we have fully decomposed above-and belowground C and N allocation and stoichiometry.We have found examples of consistent N pool sizes above-and belowground and a shift to a greater proportion of belowground N in drier sites depending on the species.We suggest that precipitation gradients do potentially decouple the C and N pool, but the exact nature of the decoupling depends on the dominant species' capacity for intraspecific variation.Even in cases where the above-versus belowground N and C pool themselves do not change proportionally, smaller biomass allocation to shoots with higher N concentrations in drier areas might facilitate a positive feedback to land degradation by livestock.

Figure 1 .
Figure 1.The allocation of biomass under different water regimes.Going from mesic to dry sites, the allocation of biomass shifts to a higher proportion of roots versus leaves and higher leaf N concentrations.

Figure 2 .
Figure 2. The sampling sites in Inner Mongolia, China.The 46 northern sites (1) were along the North-East China Transect (NECT), and the 41 southern sites (2) were along the Ordos Transect (OT).The map in the upper left corner shows places in the world with a similar climatic regime to the study area -with similar annual and seasonal pattern of precipitation and temperature; that is, with mean annual precipitation ranging from 160 to 440 mm, and most of MAP occurring from May through September, and with mean annual temperature from 1 to 9 • , mean temperature of the coldest month from −20 to −10 • (January), and mean temperature of the warmest month from 19 to 24 • (July).

Figure 3 .
Figure 3. Relationship between percentage of vegetation biomass due to Stipa/Artemisia and mean annual precipitation (a), percentage of biomass aboveground in Stipa/Artemisia and mean annual precipitation (b), both along the two transects.

Figure 4 .
Figure 4.The relationship between mean annual precipitation and plant traits: (a) leaf nitrogen (LNC) and leaf carbon concentration (LCC), (b) root nitrogen (RNC) and root carbon concentration (RCC), (c) stem nitrogen (SNC) and stem carbon concentration (SCC), and (d) proportion of N and C aboveground.

Table 1 .
Effects of species and mean annual precipitation on each plant trait in the two transects.Bold values indicate P < 0.05.