Introduction

Nitrogen (N) availability is an important factor limiting productivity in most terrestrial ecosystems (Vitousek 1982; Aerts and Chapin 2000) and the understanding of N dynamics is crucial for sustainable management of grazed natural grasslands. Consequently, many studies focused on N cycling in grazed ecosystems. However, the processes in which plant available N is affected by grazing seem to be highly variable. In semi-arid grassland ecosystems, soil biogeochemical processes are controlled by seasonal and inter-annual rainfall variation and temperature (Schimel and Parton 1986; Varnamkhasti et al. 1995; Austin et al. 2004; Augustine and McNaughton 2006) and these effects may override the long-term impact of grazers on N availability and N dynamics (Biondini et al. 1998).

Herbivores can stimulate N cycling by accelerating nutrient turnover rates via digestion and trampling effects, and the increased plant available N might stimulate plant productivity (McNaughton et al. 1997; Frank et al. 2000). On the other hand, grazing was found to be associated with N losses from grazed sites (Lavado et al. 1996; Ritchie et al. 1998; van Wijnen et al. 1999; Wang et al. 2006), mainly caused by processes such as NH3 volatilization from urine patches (Schimel et al. 1986; Frank and Evans 1997), leaching and biomass export. Gradually depleting N pools can reduce plant available N and eventually decrease grassland productivity.

During the limited time of preferential growth conditions, temperature and moisture patterns usually synchronize plant productivity and mineralization rates (Burke et al. 1997; Frank and Groffman 1998). However, our understanding of the impact of variable soil water content on mineralization processes and inorganic N availability is limited in semi-arid ecosystems. Furthermore, information about the effects of rainfall variability or grazing on NO −3 -N and NH +4 -N availability is rare for natural semi-arid grassland ecosystems, even though N-forms might be essential in terms of pathways of N losses and preferential plant N uptake and, thus, also for ecosystem performance under drought stress and climate variability. Kahmen et al. (2006) found that plant species and functional groups contrasted each other in the relative contribution of inorganic N form uptake, suggesting that plants occupy distinct niches with regard to their N-uptake, which is potentially affecting ecosystem functions. Furthermore, different N form availability to plants will likely alter carbon turnover rates (Austin et al. 2006) and carbon costs related to carbon uptake and assimilation (Bijlsma et al. 2000). How these naturally driven N-form availability dynamics interact with grazing is largely unknown.

Seasonal mineral N dynamics can be described by sequential collection and analysis of soil samples, the collection of soil water by suction caps, or the installation of ion-exchange resins. Several studies illustrated that the latter method can be successfully used for time-integrated monitoring of plant available N in semi-arid ecosystems (Dodd et al. 2000; Hook and Burke 2000; Augustine and McNaughton 2006), while soil sampling is a widely used method to analyse instantaneous N availability.

In this study, we compared the impact of grazing and rainfall variability on seasonal N availability. We combined time integrated monitoring (ion exchange) with “snapshot” measurements of conventional soil mineral N extraction and linked the N availability dynamics to aboveground net productivity and plant N uptake. Comparing both methods’ ability to reflect soil N availability, Johnsen et al. (2005) found contradicting results under manipulated soil moisture and temperature. While high soil Nmin values are not necessarily reflecting plant available-N during dry periods, the resin N measurements are highly dependent on sufficient soil water content to enable ion exchange between soil and resin surfaces, which might lead to inaccurate measurements under dry conditions. We expected high fluctuations of N availability during the growing season caused by pronounced wet-dry cycles that are typical for the semi-arid climate of Inner Mongolia grassland ecosystems. Therefore, we decided to apply both methods to obtain a more comprehensive overview of the system’s N fluxes. We hypothesized that the synchronized process dynamics of N availability and the ecosystem’s net productivity are driven by intra- and interannual and seasonal rainfall variability, as well as, by a pronounced grazing effect on N availability. Understanding the relevance of grazing and rainfall pattern for N availability might be helpful to understand the ecosystem’s stress response mechanisms and to assess resilience potentials to land use changes and climate variability.

Material and methods

Study area

The study area is located in the Xilin River Basin of Inner Mongolia autonomous region of China, nearby the Inner Mongolia Ecosystem Research Station (IMGERS, 43°38´N, 116°42´E). We analysed the seasonal and inter-annual N availability on three grassland sites characterized by varying grazing intensities. Stocking rates (average number of animals per hectare and year) of Inner Mongolia grasslands are not constant throughout the year. The bulk of a herd consists of mother sheep, each usually raising one lamb from early spring. We defined the mother sheep and lamb as one sheep unit (SU). Study sites were a 28 ha long-term grazing exclosure (hereafter: site UG, position 43°33′10″N; 116°40′33″E) fenced in 1979, a 42 ha moderate winter grazing site (WG, 43°32′57″N; 116°40′04″E) with an average annual stocking rate of 1 SU ha−1, and a 100 ha long-term heavily grazed site (HG, 43°34′54″N; 116°40′37″E) located about 3 km NE of the other two sites, with an annual average stocking rate of 2 SU ha−1 for more than 30 years. Due to the fact that a long-term factorial grazing experiment was not available in this region, a pseudo-replicated design was unavoidable and, therefore, site differences cannot be stated as pure grazing effects. However, the comparison of sites subjected to long-term extremes of grazing intensities (UG versus HG) provides information reflecting, to a certain extent, the upper and lower limits of N response to grazing. Site selection was done in close co-operation with the Institute of Botany, Chinese Academy of Sciences.

By thermic and hygric criteria, the semi-arid continental climate defines an annual growing season of about 5 months between early May and late September (Fig. 1). In this period, average monthly temperatures are above 5°C reaching a maximum of 19°C in July. More than 85% of the mean annual precipitation (MAP: 348 mm) occurs between May and September, with a clear maximum in July. In terms of MAP, we experienced a normal year in 2004 with 325 mm, a very dry year in 2005 with only 166 mm and a year with nearly normal rainfall amount in 2006 (304 mm). However, 2006 showed unfavourable distribution with no pronounced summer rainfall peak, but instead considerable rainfall at the end of the growing season in September. Soils have a sandy loam texture (sand 44–69%, silt 21–38% and clay 11–19%) and are classified as calcic Chernozems (IUSS Working Group WRB 2006). The predominant perennial vegetation is classified as a Stipa grandis P. Smirn. and Leymus chinensis (Trin.) Tzvel. dominated steppe community.

Fig. 1
figure 1

Climate diagram depicting meteorological data from the Inner Mongolia Grassland Ecosystem Research Station (IMGERS, 43°38′N, 114°42′E). The striped bar at the bottom indicates the frost period, the grey bar the growing season defined by mean monthly temperature > 5°C and precipitation > 2× temperature

Full size image

Measurement of N dynamics

Plant available NH +4 -N and NO −3 -N was measured with PST-1 ion exchange resin capsules (UNIBEST, Montana, U.S.) filled with mixed anion-cation resins. Capsules were installed with five replications per site into the main rooting zone at 15 cm depth. Installation and collection of capsules were synchronized with aboveground biomass sampling (see below). Data from 2004 concluded that temporal dynamics of aboveground biomass could be monitored accurately by less frequent harvesting. Sequential sampling was, therefore, performed with higher time resolution in 2004 (bi weekly) and lower resolutions in 2005 (3 weeks intervals) and 2006 (monthly).

At each sampling time a new set of resin capsules were installed and capsules from the proceeding period were collected. To position a new capsule, a grass sod, approximately 15 × 15 × 15 × 15 cm in volume, was lifted and a small cavity was formed in order to assure good contact between capsule and surrounding soil, the new capsule was positioned and the grass sod replaced.

Prior to chemical extraction of harvested capsules, soil particles adhering to the capsules’ surface were removed by gently moving capsules in de-ionized water. Capsules were stored at 2°C until extraction. Ions were removed by sequential shaking in three batches of 20 ml of 2 M HCl. Extracts were analysed with a rapid flow auto-analyzer (Autoanalyzer II, Bran & Lübbe, Norderstedt, Germany) after neutralization with 2 M NaOH. Data of N availability over sampling periods are presented as average daily N flux through the resin capsule surface (mg N m−2 d−1). To obtain time-integrated seasonal values of N availability, data of resin N availability of the sampling periods were cumulated and divided by the total number of days.

On each sampling date in 2005 and 2006, soil cores of 4 cm diameter were taken from 0–20 cm depth with 4 replications per grazing site, each soil sample consisting of three subsamples. Replications were at least 20 m away from each other and close to the resin sample area. Soil was cooled in ice-boxes and frozen (−18°C) until analyses. Water content of fresh samples was determined gravimetrically and data were corrected for soil bulk density, which differed between sites (Steffens et al. 2008). NH +4 and NO3 − from 12 g soil (fresh weight) were extracted with 100 ml 0.01 M CaCl2 and concentrations analysed with the rapid flow auto analyser mentioned above. Soil mineral N (Nmin) was presented as kg N ha−1. Compared to soil Nmin extractions, resin capsules are dependent on mass flow to adsorb ions at their exchanging surface. Therefore, methodological differences most likely appear during drought periods (Johnsen et al. 2005). Similarly, freezing and thawing events during spring time, affecting the top-soil N dynamics, may not be captured by resin capsules placed at 15 cm depth, while soil Nmin is potentially more sensitive by integrating N availability over the entire 0–20 cm depth.

Seasonal dynamics of soil water content (SWC) were measured with three replications on each site at 30-min intervals with calibrated Theta-probes (Type ML2x, Delta-T Devices, Cambridge, UK) (for more information: see Zhao et al. 2007).

Aboveground biomass and plant N uptake

At each of the three sites, aboveground biomass was collected on seven dates from 1 m2 plots in 2004 and 2005 (2004: June 11, June 27, July 12 and 27, August 11 and 27, and September 16; 2005: May 11, June 2 and 23, July 15, August 7 and 28, and September 22) and on five dates in 2006 (May 15, June 11, July 12, August 12, and September 12) with five (2004), seven (2005) and eight (2006) replicates (Gao et al. 2008). Biomass was separated into green parts and dead material and dried at 75°C for 48 h. Peak live biomass was considered to represent ANPP of sites UG and WG. For each sampling date, ANPP of site HG was calculated as the difference between biomass sampled from 1 m2 inside exclosure cages of 1.5 m × 1.5 m size at time(i+1) and the biomass measured outside exclosure cages at time(i). N content of shoot biomass samples were analysed by dry combustion with CN elemental analyzer (ThermoFinnigan, Flash EA Egelsbach, Germany). Plant N uptake was calculated by multiplying aboveground green dry matter with N concentration.

Data analysis

Statistics for all data were performed with SAS procedures from software version 9.0 (SAS Institute Inc., Cary, NC, USA). Data were tested for normal distribution of residuals with a Shapiro-Wilk test using PROC UNIVARIATE. Statistical significance of the site, years and interaction was tested with PROC GLM. Resin-N availability and soil mineral N data were analysed with a two-factorial ANOVA, with grazing sites and years as fixed factors. Seasonal dynamics of N availability was analysed with a two-factorial ANOVA, with grazing sites and sampling periods as fixed factors. Comparison of means was based on Tukey’s HSD test. Regression analyses between resin N availability and both plant N uptake and ANPP (Fig. 3a and b) were illegitimate due to unpaired sampling design of the replicates from biomass, resin and soil mineral N collection. To describe the relationship between available NO −3 -N fraction and SWC (Fig. 6), a regression analysis was performed with Sigma plot 9.0, fitting a sigmoid function (three parameter) to 235 observations of resin N availability from 16 sampling periods during 2004-2006 of three different grazing sites and the corresponding SWC.

Results

N availability

Resin-N availability varied among the three growing seasons with the highest resin-N availability in 2004 and the second highest in 2006 with mean daily N-fluxes (mg N m−2 d−1) during the growing seasons of 4.96 in 2004, 2.93 in 2006 and 1.65 in 2005 (MSDTukey = 0.93). Both, resin capsules and soil mineral-N extraction indicated higher N availability in 2006 compared to 2005. Mean values of soil Nmin (kg/ha) were 12.3 in 2005 and 28.2 in 2006 (MSDTukey = 3.11). Considering intra-annual N dynamics, both methods indicated generally low values of N availability in 2005 throughout the growing season compared to 2006.

In 2004, mean daily resin-N availability (mg N m−2 d−1) at site UG was higher (7.52) compared to that at sites WG (4.03) and HG (4.59). Higher N availability at site UG was measured in three of four sampling periods (Fig. 2). In 2005 and 2006 resin-N availability showed no differences among the three grazing sites over the growing season. However, soil Nmin indicated site differences at the last sampling period of 2005 and 2006 with higher N availability at site HG compared to site UG.

Fig. 2
figure 2

Seasonal dynamics of mean daily resin-N availability (mg N m−2 d−1) and soil mineral-N extractions (kg N ha−1) at different grazing sites (heavy grazing, winter grazing and no grazing) as well as precipitation amounts during sample periods for the growing seasons 2004–2006 (no soil mineral-N data available for 2004). Error bars represent the standard error and letters indicate differences by ANOVA analysis (P < 0.05)

Full size image

We compared N availability of different grazing sites as measured by resin-N with ANPP and plant N uptake (Fig. 3 a, b). Resin N availability appears to be a reliable proxy of plant N availability and uptake. Highest resin-N availability, ANPP and plant N uptake were measured in 2004. However, significant differences between grazing sites in ANPP in 2005 were not reflected in resin N availability. Similarly, apparently higher resin N availability in 2006 compared to 2005 was not reflected in correspondingly higher ANPP. Resin N availability is supposed to indicate plant N uptake but not necessarily ANPP; aboveground plant N-uptake was better related to N availability than ANPP.

Fig. 3
figure 3

a, b: Resin-N availability in relation to aboveground net primary productivity (ANPP) and plant N uptake measured in the growing seasons 2004–2006 at sites of different grazing intensities (UG = grazing exclosure, WG = winter grazing, HG = heavy grazing) in a semi-arid grassland of Inner Mongolia. Average daily resin-N uptake (mg m−2 d−1) was calculated for the period from onset of growing season until peak biomass time (data are plotted as mean values with standard errors)

Full size image

NH +4 and NO3 − availability

Resin NH +4 -N availability differed between years but not between grazing sites (Table 1, Fig. 4). During the growing seasons of 2004 (1.86 mg N m−2 d−1) and 2005 (1.52 mg N m−2 d−1), resin NH +4 -N availability remained at a constant and similar level, while it was significantly lower in 2006 (1.07 mg N m−2 d−1, MSDTukey = 0.36). The strong site x year interaction for resin NO −3 -N originates from higher availability at site UG compared to sites HG and WG in 2004, while there were no site differences in 2005 and 2006. Resin NO −3 -N availability in 2005 was very low compared to 2004 and 2006, while differences in resin NH +4 -N availability were less pronounced (Fig. 4).

Table 1 ANOVA results of total N and N-form availability measured by ion exchange resin capsules (2004–2006) and soil mineral N extraction (2005–2006) at three grazing sites in semi arid grassland of Inner Mongolia.
Full size table
Fig. 4
figure 4

Mean annual NO −3 -N and NH +4 -N availability at different grazing sites (UG = grazing exclosure, WG = winter grazing, HG = heavy grazing) measured by resin capsules (2004–2006) and soil Nmin (2005–2006). Only significantsignificant differences between grazing sites (P < 0.05) are indicated by different letters, data are plotted as mean values with standard errors

Full size image

With regard to soil mineral N extraction, mean NH +4 -N availability in 2006 (20.2 kg N ha−1) was higher compared to 2005 (10.6 kg N ha−1, MSDTukey = 3.07). Soil NO −3 -N availability showed site × year interaction. NO −3 -N availability was higher at site HG compared to the other sites in 2006. Comparing both methods, resin NH +4 -N availability was lower compared to NO −3 -N availability in 2006; while soil mineral N extraction indicated the opposite trend of N-form availability (Fig. 4).

Differences in resin NO −3 -N and NH +4 -N availability were analysed in terms of NO −3 -N fraction of N availability. In Fig. 5, the NO −3 -N fraction, average SWC and the precipitation are plotted for all sampling periods during 2004–2006. Dynamics of the NO −3 -N fraction were generally reflected by seasonal changes of SWC and precipitation pattern. In 2004 and 2006, the NO −3 -N fraction (between 0.5 and 0.8) increased with increasing SWC and decreased when SWC became drier. Only in 2005, was there no reflected variability in SWC to corresponding changes in NO −3 -N fractions. During that year, the proportion of NO −3 -N measured by resin capsules decreased strongly, indicating that NH +4 -N was the dominant available inorganic N-form. Summarizing, in wetter years the NO −3 -N fraction was the dominant N-form available to plants, following the dynamics of SWC. Resin NO −3 -N fraction correlated positively with SWC (Fig. 6). When SWC decreased below 8 Vol. %, NH +4 -N dominated the resin available N, while above 13 Vol. % of SWC the resin available N-form was dominated by NO −3 -N.

Fig. 5
figure 5

Seasonal dynamics of NO −3 -N fraction of total resin available N, soil water content and precipitation during 19 sample periods from 2004–2006 (mean values pooled over three grazing sites). Soil water content data were not available for sampling periods with frost

Full size image
Fig. 6
figure 6

NO −3 -N fraction of the total resin available inorganic N in relation to the average soil water content. Data of 16 sampling periods during the growing seasons 2004 (6), 2005 (6) and 2006 (4) at three different grazing sites

Full size image

Discussion

Impact of grazing and rainfall on N availability

Grazing had less impact on the dynamics of resin N availability compared to seasonal and annual rainfall distribution. However, in a year of favourable rainfall conditions (2004), we found a grazing effect with higher resin-N availability at site UG. This finding confirms the negative grazing effect on N availability reported by (Lavado et al. 1996; Ritchie et al. 1998; van Wijnen et al. 1999; Wang et al. 2006) and is likely related to a long-term N export and lower soil organic matter content (Steffens et al. 2008). N is limiting ANPP of this grassland system under conditions of high water availability (Brueck et al. 2010). It can be assumed that negative effects of grazing on N availability become more evident under wet conditions and constrain ANPP to a greater extent. Grazing effects on resin-N availability were absent in the dry year (2005) and in a year of unfavourable rainfall distribution (2006), indicating that drought suppressed processes within the N-cycle. In twelve (resin-N) and six of eight (soil Nmin) sampling periods we did not observe increased N availability under grazing as reported in other studies (McNaughton et al. 1997; Frank et al. 2000; Garibaldi et al. 2007) performed in more humid grasslands. However, two Nmin sample periods, both at the end of the growing season in 2005 and 2006, indicated a higher N availability at site HG compared to site UG (Fig. 2). Explanations would be speculative. The last sampling period in 2006 was characterized by comparably high SWC, while in 2005 the grazing effect was observed under dry conditions. Results obtained from resins did not confirm soil Nmin data in contrasting sample periods, indicating that both methods should be carefully considered if N availability in ecosystems is assessed.

We conducted the study during three growing seasons of very low to average precipitation amounts during the vegetation growth period (May–September). The extended drought period from May to September 2005 with very low rainfall (120 mm) and low average soil water content (see Fig. 5) resulted in the lowest inorganic N availability compared to both other years (2004: 235 mm, 2006: 251 mm; see Fig. 2). Thus, rainfall amount and distribution and the corresponding changes in soil water content apparently had a higher impact on inorganic N availability than grazing. This effect of water availability on inorganic N availability was confirmed by two different methods (resin-N and soil Nmin). The number of studies analysing the effects of climate variation and grazing on N availability of semi-arid grasslands is limited and results are controversial. Climatic factors had a larger impact on N dynamics than grazing in semi-arid grasslands of North-America (Biondini et al. 1998) and in a South African savanna (Stock et al. 2010), while other studies provided clear evidence that despite pronounced annual variability in rainfall, large herbivores strongly influenced N availability and productivity in semi-arid grasslands (Augustine and McNaughton 2006).

The high plant available N in spring 2006 might reflect a N accumulation during the previous years’ drought when plant N uptake was limited due to water shortage. However, decomposition and mineralization processes still continued, at least for aboveground litter decomposition (Giese et al. 2009). This high N availability, as indicated by soil Nmin, was, however, not confirmed by resin-N data. This discrepancy between both methods may be related to the resin capsule position in 15 cm depth, where N turn-over processes are delayed due to continuous freezing or low temperatures, while Nmin may have captured topsoil N dynamics during freezing-thawing periods (see material and methods).

N availability and plant productivity

Does the inorganic N availability measured by ion exchange resins and soil mineral-N extractions reflect ecosystem productivity? In temperate semi-arid ecosystems the low winter temperatures and drought periods during summer are main drivers of plant and microbial activities. In the limited time of preferential growth conditions, temperature and moisture patterns usually synchronize plant productivity and mineralization rates (Frank and Groffman 1998). Basically, annual ecosystem net primary productivity and plant N uptake dynamics are related to soil inorganic N availability (Fig. 3). Plant N uptake was better related to N availability compared to ANPP. Two aspects can be considered in this context: (i) the higher investment of assimilates to the rooting system in 2006 compared to 2005 (Gao et al. 2008) and (ii) the higher N concentration in plant tissue and higher total plant N uptake in 2006 (Gao 2007). Since productivity in semi-arid grassland ecosystems occurs for the major part belowground, root production and shoot N concentration need to be additionally considered when the correlation between inorganic N availability and net primary production is analysed.

Additionally, to abiotic factors controlling seasonal dynamics of nutrient availability, internal N pools are most likely relevant in order to interpret the relationship between productivity and N availability. Depending on the species that are abundant under different grassland management systems, a considerable amount of N can be resident in belowground organs supporting growth in dry periods, while continuous nutrient availability would be required if grassland sites are dominated by annuals and species with low root N storage capacity. Plant internal nutrient cycling is consequentially not reflected by any standard soil N availability index. This aspect deserves further attention to consolidate our understanding of seasonal soil N dynamics and productivity.

Plant available N-form

Plant available N-form dynamics and consequences for natural grassland ecosystem processes are increasingly noticed and exploration will be intensified in the near future (Kahmen et al. 2006; Austin et al. 2006). Information about the soil inorganic N-form, as influenced by rainfall variability, in semi-arid ecosystems is limited. Yahdjian et al. (2006) found a trend of higher NO −3 -N concentrations in Nmin samples with decreasing rainfall, while NH +4 -N concentrations were unaffected. High NO −3 -N accumulations during drought was explained by reduced ion mobility in thin water films of dry soils simultaneously inhibiting plant N-uptake. However, both methods we used to quantify plant inorganic N availability did not indicate higher NO −3 -N availability in the dry year 2005 (Fig. 4). Stark and Firestone (1995) found inhibition of nitrification below a soil water potential of -0.6 MPa. We speculate that low soil water potentials explain low NO −3 -N availability during drought, which points out the significance of further investigations on plant and microbial N assimilation as related to soil water potential in this grassland ecosystem.

Availability of N forms has consequences for plant performance due to the concomitant carbon costs for uptake and assimilation of inorganic N (Bijlsma et al. 2000). In an ecological context, seasonal and inter-annual differences in plant available N-form might explain competitive dominance of grass species in semi-arid grasslands. Studies in Patagonian steppe indicated that the majority of grassland species preferentially take up NO −3 -N (Austin et al. 2006). As indicated by predominant resin NH +4 -N availability in 2005, plants that preferentially take up NH +4 -N might have a competitive advantage under dry conditions in the semi-arid grasslands of Inner Mongolia. Changes of N form availability, thus, may have considerable consequences for ecosystem species composition dynamics with possible feedback processes altering the N cycling, including the carbon turnover.

Conclusions

Rainfall variability triggers processes of N availability and N-form supply in this semi-arid ecosystem. During our studies, grazing had a comparably minor impact on N availability. In general, we found a positive relationship between N availability and plant productivity. The positive relationship between N availability and both ANPP and plant N-uptake indicates ecosystem N limitation if water is sufficiently available in this ecosystem. Our study emphasizes that the effect of rainfall on N-form availability deserves further attention in studies considering interactive effects of grazing and N on plant species composition. Additionally, our results indicate that methods to quantify N availability and dynamics (resin versus soil Nmin) need to be carefully examined in terms of time integrated and instantaneous measurements.