Grazing elevates litter decomposition but slows nitrogen 1 release in an alpine meadow 2

Nutrient cycling is a key ecosystem function whereby the processes of litter decomposition and N 12 release in the soil-plant interface are vitally important but still not clear in the alpine ecosystems. We carried out 13 a 3-year study to improve our understandings in nutrient cycle and develop strategies for restoring the degraded 14 grasslands on the Qinghai-Tibetan Plateau. We established the grazing (GP) and grazing exclusion paddocks 15 (GEP), then identified litter species composition and analyzed litter and soil chemical characteristics. Litter 16 decomposition and N release were monitored by incubating litter ‘in situ’ and across paddocks over 799 days. We 17 found that grazing did not change plant species composition, but increased litter N; while grazing exclusion 18 increased litter mass of palatable species and promoted soil organic carbon. Litter decomposed faster in GP, while 19 N release was faster in GEP. Incubation site environment had more but litter source had less impact on litter 20 decomposition and N release. Therefore, grazing and grazing exclusion had different impacts on litter 21 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66 Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c © Author(s) 2018. CC BY 4.0 License.

decomposition and N release but both elevated nutrient cycle.The implications of our findings in restoring the degraded grasslands on the Qinghai-Tibetan Plateau were discussed.

Introduction
Qinghai-Tibetan Plateau represents an important eco-region in China (Wen et al., 2010), where the alpine grasslands cover more than 85% of total area and are regarded as the major natural pastures (Dong et al., 2010).
However, the ecosystems in this region have continuously suffered from severe degradation mainly driven by climate change, overgrazing, over-cultivation and poor management (Han et al., 2008;Li et al., 2009;Wu et al., 2009;Feng et al., 2010;Wu et al., 2010), with an increasing degradation rate of 1.2-7.44%per year (Ma et al., 2007).Since 1990's, the restoration of degraded grasslands has attracted highly attention (Kang et al., 2007;Han et al., 2008), and some efforts have recently focused on the grassland restoration and maintenance by increasing aboveground plant abundance (Niu et al., 2009) and biodiversity (Wu et al., 2009;Niu et al., 2010), and improving soil property (Cao et al., 2004;Wu et al., 2010;Sun et al., 2011).However, litter decomposition and N release, the key factors regulating the nutrient cycle and availability in the soil-plant interface (Carrera and Bertiller, 2013), are still not clear in the alpine ecosystems (Luo et al., 2010;Zhu et al., 2016).We carried out this study to investigate the effect of grazing on plant community, litter quality, litter decomposition and N release in an alpine meadow, aiming to improve our understandings in the nutrient cycle and to develop management strategies for restoration of ecological function in alpine pastures.
It is generally known that herbivore grazing may induce short-term ecophysiological changes in overall litter quality as well as longer-term shifts in plant community composition.At the short-term ecophysiological level, herbivore grazing may promote the plant species producing high-quality litter (Holland and Detling, 1990;Sirotnak and Huntly, 2000;Olofsson and Oksanen, 2002;Semmartin et al., 2004Semmartin et al., , 2008)).Because the loss of plant Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.tissues caused by grazing may favour the grazed species with a higher re-growth rate and greater nutrient contents in plant tissues due to the higher nutrient uptake (see Holland and Detling, 1990;Olofsson and Oksanen, 2002;Semmartin et al., 2008).At the long-term community level, selective foliar grazing of herbivores may alter the competitive interaction and recruitment patterns of plant species, which will change their abundance and life form structure (Bardgett and Wardle, 2003;Semmartin et al., 2008;Wu et al., 2009;Niu et al., 2010).For instance, herbivores usually concentrate on the most palatable plants (e.g., species with high nutrient and low fibre contents), which will favour dominance by the less unpalatable species (Garibaldi et al., 2007), resulting in the high inputs of low-quality litter to soil and thus a reduction of decomposition rate, nutrient availability and nutrient cycling (Ritchie and Knops, 1998;Moretto et al., 2001;Olofsson and Oksanen, 2002).However, empirical evidences of variances in litter quality input and soil nutrient cycle caused by grazing are still scarce and controversial (Garibaldi et al., 2007).
It is supposed that the higher nutrient content in plant tissue usually results in the faster litter decomposition, and the high nutrient mineralization and availability in soil (Olofsson and Oksanen, 2002).At the ecosystem scale, the chemical characteristics of plant litter, for example the carbon:nitrogen ratios (C:N) and/or nitrogen and lignin content, are often regarded as the indicators of litter quality (Aerts, 1997;Strickland et al., 2009).Many studies have demonstrated a positive correlation between litter decomposition rate and N content, or a negative relationship between litter decomposition rate and initial lignin content and C:N or lignin:N ratios ratios (e.g., Wardle et al., 2002;Aerts et al., 2003;Semmartin et al., 2004;Garibaldi et al., 2007;Luo et al., 2010;Vaieretti et al., 2013).
Except litter quality (its chemical composition), the climate (mainly temperature and humidity) and decomposing organisms (their abundance and activity) are the two main factors controlling litter decomposition (Coûteaux et al., 1995;Aerts, 1997;Semmartin et al., 2004;Keeler et al., 2009;Berg and McClaugherty, 2014; Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.Zhu et al., 2016).Generally, climate is the dominant factor over litter quality in areas subjected to unfavourable weather conditions (Coûteaux et al., 1995), because decomposer activity regulating litter decomposition is largely temperature and moisture dependent (De Santo et al., 1993).Usually the activity of decomposers increases with temperature (Coûteaux et al., 1995); and as soil moisture level rises, the metabolic activity of decomposers normally increases until an optimum plateau is reached (Orsborne and Macauley, 1988).Thus it is likely that the optimum response of decomposers to temperature and moisture is determined largely by the local climate (De Santo et al., 1993).However, under the favourable conditions, litter quality may largely prevail as the regulator and remain important until the late decomposition stages (Coûteaux et al., 1995).
Most studies on litter decomposition affected by the herbivores usually focus on the forest, grassland or crop ecosystems in the temperate areas, largely ignoring that in the alpine zones.Moreover, to date our knowledge on the decomposition of litter particular that of mixed-litter in the grasslands in the alpine regions is still lacking, though some efforts have been made (e.g., Wu et al., 2009;Luo et al., 2010;Zhu et al., 2016).In this study, we firstly set up the grazing (GP) and grazing exclusion paddocks (GEP) and examined the subsequent plant composition, litter quality and soil characteristics.We then subsequently investigated whether litter quality had any effects on litter decomposition, N release and soil property by collecting litter mixtures from both GP and GEP and incubating them 'in situ' and across GP and GEP.Based on the knowledge outlined above, we tested: (1) whether grazing may improve litter quality with higher nutrient content (i.e., N) and change plant community with lower biomass of palatable species detected whereas the opposite is the case in GEP, and (2) whether grazing will elevate litter decomposition and N release and thus improve soil properties whereas the opposite is the case in GEP.Results of the present study may provide insight into our understandings in nutrient cycle in the alpine regions in general, which may also further provide a scheme to assist in the development of strategies for restoring the degraded grasslands on the Qinghai-Tibetan Plateau in particular.

Experimental site
This study was conducted in a typical alpine meadow, eastern Qinghai-Tibetan Plateau, NW China (N33°59′, E102°00′, altitude 3,500 m above sea level).The mean annual temperature is 1.2°C , ranging from -10°C in January to 11.7°C in July, with approximately 270 days of frost per year.The mean annual precipitation is 620 mm over the last 35 years, occurring mainly during the short and cool summer (Niu et al., 2010).The years during this study (i.e., 2009 ~ 2012) were climatically typical (Sun et al., 2015), and the mean temperature and total precipitation of each month during the experiments were showed in Supplemental Fig. 1.
The pasture selected for experiments is larger than nine hectares (including experimental and buffer areas) and regularly used for Tibetan sheep and yak grazing during the grazing seasons (May ~ October).The slopes are less than 5%, which is the gentle topography of the area, and the soil attributes in the experimental areas were similar after a long-term grazing history with the same grazing pattern.The soil type in the experimental area is alpine meadow soil, similar to the primarily Mat-Cryic Cambisols (Wu et al., 2010).

Litter composition and quality
To measure the annual litter composition and determine whether plants could recover without grazing, three To measure the litter composition, litter of different species was identified and then separated into two groups according to the palatability by the Tibetan sheep (Niu et al., 2009(Niu et al., , 2010;;Wu et al., 2009; see Supplementary Table 1): (1) palatable species -preferred and desirable species, and (2) unpalatable speciesundesirable and toxic species.
To measure the litter quality, two treatments were tested: (1) GP-litterlitter of all species from aboveground was collected from the GP, (2) GEP-litterlitter of all species from aboveground was collected from the GEP.Litter from a quadrat was mixed after the measurement of litter composition and then oven-dried at 60°C for 48 h.The dry litter was ground and stored in a zip-lock bag with 10 g per bag.There were six replicates for each treatment.The contents of lignin, cellulose and hemicellulose were measured by following van Soest et al. (1991).Organic carbon concentration (C) was measured by the K2Cr2O7-H2SO4 oxidation method of Walkley-Black (Nelson and Sommers, 1996).Briefly, a known weight of plant sample was treated with potassium dichromate in the presence of concentrated sulfuric acid and digested at a low temperature (≈ 30°C ) for one hour.
The excess of potassium dichromate did not reduce the organic matter, as it was titrated back against a standard solution of ferrous sulfate.The C was then calculated on the basis of the quantity of ferrous sulfate consumed (Chen et al., 2016).The total Kjeldahl nitrogen (N) and total phosphorus (P) were analyzed using a FIAstar 5000 flow injection analyzer (Foss Tecator, Högnä s, Sweden) (Chen et al., 2016).We also calculated the ratios of C:N, lignin:N, cellulose:N and hemicelluloses:N. of litter composition and quality, and thus litter decomposition and N release.Therefore, we also examined the effects of grazing and grazing exclusion on soil characteristics.We randomly selected five sampling sites from each experimental paddock, and soil samples (n = 30 in total) were collected at depth 0-10 cm using a bucket auger (10 cm in diameter) in October of 2010, 2011 and 2012.The same methods used to test litter quality (i.e., Nelson and Sommers, 1996;Chen et al., 2016) were applied to estimate the soil organic carbon (SOC), total nitrogen (TN) and total phosphorus (TP).

Litter decomposition and N release
In this experiment, we carried out four treatments: (1) GP-GP, litter of all species was collected from and incubated in the GP; (2) GEP-GEP, litter of all species was collected from and incubated in the GEP; (3) GP-GEP, litter of all species was collected from the GP but incubated in the GEP; (4) GEP-GP, litter of all species were collected from the GEP but incubated in the GP.The treatments 1 and 2 were referred as "in situ" incubation treatments.
The treatments 3 and 4 were across incubation treatments which were to improve our better understanding "homefield advantage" effect on litter deposition (John et al., 2011).
For each sample soil particles attached to litter were cleaned off with a soft brush, and samples were air-dried for three days.Dry litter collected from each quadrat were cut to ≈ 5 cm length and 10 g litter was packed into a nylon litter-bag (15 cm × 20 cm with mesh size of 0.35 mm) (Cornelissen, 1996), which may prevent any loss of material and has no effect on litter decomposition (Cornelissen et al., 1999).On 20 th Oct 2010, the packed litter was incubated above the soil surface by fastening to the ground surface with four steel stakes to avoid being removed by the sheep and small animals (Vaieretti et al., 2013).The small animals were the plateau pika, Ochotona curzoniae (Hodgson) in the present study.For each treatment, 24 packed litter-bags were incubated with 20 cm apart from each other to reduce the mutual interference.Three litter bags were retrieved after a Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.incubation period of 56,141,247,391,444,582,695 or 799 days (i.e., on 15 th December 2010, 10 th March, 24 th June and 4 th September 2011, and 7 st January, 24 th May, 14 th September and 27 th December 2012, respectively).
There were a total of 144 packed litter-bags used in this experiment.Retrieved litter was brought back to the laboratory and cleaned by removing any extraneous material attached.They were weighed after being oven-dried at 60º C for 48 h.Samples were ground and stored in a zip-lock bag for further chemical analyses as mentioned above.We estimated the litter decomposition and N release as the percentage of dry weight remaining at the end of each incubation period (Cornelissen et al., 1999;Vaieretti et al., 2013).

Statistical analyses
A goodness-of-fit test (Shapiro-Wilk test, Univariate Procedure) was used to test the normality of data before mean comparison using analysis of variance (ANOVA, GLM Procedure).All those data were normally distributed.
Data on litter quality (Table 1) were analyzed using ANOVA followed by Tukey's Studentized multiple range test.While the difference between the palatable and unpalatable litter quantity in GP or GEP was compared by paired-t test (TTEST Procedure) (Fig. 2).Data on the proportion of litter mass or N remaining (Fig. 3-5), litter quality (content of organic carbon, nitrogen, phosphorous and so on) (Table 1), and soil SOC, TN and TP (Fig. 1) were analyzed using ANOVA followed by least significant difference test (LSD test) for multiple comparisons.
The decay rate (k) of litter mass during the incubation period (Table 2) was assessed using a negative exponential model according to Swift et al. (1979): y = a*e -(t * k) , where y is the dry mass of litter remaining in the litter bags at time t (days), a is the initial litter mass.The difference in decay rate between treatments was compared according to Julious (2004)  To provide quantified information of how environmental conditions (i.e., decomposition sites, the GP and GEP) and litter source (i.e., litter collected from the GP and GEP) affecting the final litter decomposition and N release, a multivariate regression model (GLM Procedure) employed by Vaieretti et al. (2013) was used (Table 3): litter decomposition or N release = site + litter source + site × litter source + ϵ, where ϵ is the model error.The proportion of site environment, litter source and their interaction contributing to variability of litter decomposition or N release was calculated as: their sum of square divided by the total sum of square.The Type I sum of square was used because of the balanced design of this experiment.All analyses were done using SAS 9.3 (SAS Institute Inc., Cary, NC, USA).Rejection level of H0 was set at α < 0.05.Mean (± SE) was presented in Fig. 1-5.

Litter composition and quality and soil property
Fifty-five plant species (mostly forbs and graminoid grasses with several legumes and sedges) were identified, and they presented in both GP and GEP, except the Gentiana macrophylla Pallas, which was only found in the GP (Supplemental Table 1).However, even though the annual litter mass of unpalatable species in both GP and GEP was similar (ANOVA: F1,38 = 3.43, P = 0.0717), litter mass of palatable species was significantly greater in the GEP than in the GP (ANOVA: F1,38 = 75.32,P < 0.0001), which contributed significantly more to the total litter mass in the GEP than in the GP (ANOVA: F1,38 = 114.66,P < 0.0001) (Fig. 1).The litter mass was not significantly different between palatable and unpalatable species in the GP (Paired-t test: t19 = 0.96, P = 0.3510); however, in the GEP, litter mass of palatable species was significantly greater than that of unpalatable species (Paired-t test: t19 = 7.17, P < 0.0001) (Fig. 1).
As shown in Table 1, litter collected from GP had significantly higher C and N but significantly lower hemicelluloses and hemicelluloses:N than that collected from GEP.No significant difference was found in other compounds or compound ratios between different litter sources though the cellulose, C:N, lignin:N and cellulose:N were lower in litter collected from GP (Table 1).
The concentrations of soil TN and TP were not significantly different between GP and GE for each year (LSD = 0.0002~0.0015and 0.0003~0.0004for TN and TP, respectively; P > 0.05) (Fig. 2a-b).Similarly, there was no significantly difference in SOC concentration between the GP and GE in 2010 and 2011 (LSD = 0.0169 and 0.0111 for 2010 and 2011, respectively; P > 0.05), while in 2012 SOC was significantly higher in the GE than in the GP (LSD = 0.0138, P = 0.0279) (Fig. 2c).

Litter decomposition
The proportion of litter mass remaining significantly decreased with incubation duration (LSD test: LSD = 3.23 ~ 4.96, P < 0.0001) (Fig. 3).It is found that the lower environmental temperature (< 0°C ) between November and March (Supplemental Fig. 1) might have significantly slowed the litter decomposition (Fig. 3); but increasing temperature since April (Supplemental Fig. 1) significantly accelerated litter decomposition (Fig. 3).The time period required to achieve 50 % decomposition of litter mass was about 19 months in GP-litter, which was faster than that in GP-GEP-litter (i.e., about 23 months) (Fig. 3).
As shown in Table 2, the decomposition rate (k) of litter incubated in the GP was significantly higher than that in the GEP (non-overlap of 83.4% CL), i.e., for the 'in situ' treatments k in GP-GP > k in GEP-GEP) and for the across treatments k in GEP-GP > k GP-GEP.The final proportion of litter mass remaining was significantly lower in GP-GP and GEP-GP than in GP-GEP (LSD test: LSD = 2.51, P < 0.0001) (Fig. 5a).

N release
Compared to litter decomposition, the dynamics of N release were more complicate.Generally, the percentage of of GEP-GP (Fig. 4).Since January in the second winter (2012), the percentage total N remaining significantly decreased until the end of experiments (Fig. 4).The final proportion of total N remaining was significantly higher in the GEP-GP and significantly lower in the GEP-GEP and GP-GEP (LSD test: LSD = 5.36, P = 0.0002) (Fig. 5b).

Contribution of site environment and litter source to litter decomposition and N release
The multivariate regression model indicates that both site environment and litter source significantly affected litter decomposition and N release with various contributions (Table 3).Site environment contributed respectively near 25% and 50% more to litter decomposition and N release than did litter source (Table 3).Furthermore, the model predicts that GP resulted in significantly greater litter decomposition (8.13%) but significantly fewer N release (9.73%) than did GEP (F1,8 = 62.48 and 57.49 for litter decomposition and N release, respectively; P < 0.0001).
Results show that litter collected from GEP decomposed significantly faster (2.5%) but released N significantly slower (9.27%) than that collected from GP (F1,8 = 34.99 and 15.80 for litter decomposition and N release, respectively; P < 0.01).
Significant interaction of site environment and litter source was found on litter decomposition (F2,8 = 7.10, P = 0.0286), i.e., litter collected from GEP but incubated in GP decomposed significantly faster (also see Fig. 5a).
However, interaction of site environment and litter source was not significant on N release (F2,8 = 2.76, P = 0.1350).
Grazing of herbivores may indirectly alter the species composition and functioning of grasslands by inducing shifts in plant competitive interactions and recruitment patterns, and thus changes in species abundances and life form structure (Bardgett and Wardle, 2003;Garibaldi et al., 2007;Semmartin et al., 2008;Wu et al., 2009;Niu et al., 2010;Chaneton, 2011).However, our results indicate that herbivore grazing did not alter plant community composition in terms of species inventory, as species found in the GEP mostly presented in the GP.According to Sun et al. (2011), herbivores caused changes in species composition is grazing-intensity dependent.Thus our results imply that a moderate stocking rate of 4 Tibetan sheep/ha with a short annual grazing period (i.e. from April to October 2010) in the GP did not change the species composition.To confirm this conclusion, a long-term grazing exclusion experiment will provide more confident proofs.However, our results show that herbivore grazing significantly altered species composition in terms of species abundance or palatability with significantly less palatable litter produced in the GP than in GEP (Fig. 1).
The low mass of palatable species may attribute to two causations.First, on the Qinghai-Tibetan Plateau herbivore grazing will can in the short plant height and small leaf area of palatable species (Sun et al., 2011).This may be an effective ecological strategy of palatable species to avoid being grazed and the continuous disturbance, through increasing resistance to grazing with lower competition for light (van der Wal et al., 2000;Falster and Westoby, 2003;Niu et al., 2009Niu et al., , 2010)).Second, most palatable species on the Qinghai-Tibetan Plateau (mostly the grasses and sedges, see Supplemental Table 1) are tall and more likely grazed by preferential grazing (Sun et al., 2011), subsequently resulting in the lower mass of palatable species in the GP.Furthermore, our results show that there was no significant difference in the litter mass of unpalatable species between the GP and GEP (Fig. 1), disagreeing with the assumption that removing the canopy of palatable species will allow the intra-and interspecific competition for light which ultimately favours the establishment of short, less-palatable species (Sternberg et al., 2000;Pavlů et al., 2008;Wu et al., 2009;Sun et al., 2011).It is well known that litter quality is usually determined by the content of different chemical compounds such as soluble C, N and P, as well as lignin or lignin:N ratio, and litter of high quality usually has higher N content but lower lignin and lignin:N ratio (e.g., Aerts, 1997;Olofsson and Oksanen, 2002;Wardle et al., 2002;Garibaldi et al., 2007;Strickland et al., 2009).In this study, grazing might have improved litter quality at least to some degrees by significantly increasing the N content and potentially lowering the hemicellulose content and C:N, lignin:N, cellulose:N and hemicellulose:N ratios (Table 1).Therefore, our results agree with previous studies that grazing may promote litter quality (e.g., Sirotnak and Huntly, 2000;Olofsson and Oksanen, 2002;Semmartin et al., 2004).

Litter decomposition
For a given climate region, the ecological processes of litter decomposition are regulated by incubation environment (i.e., grazing/grazing exclusion and soil property in this study) and litter quality.Our results indicate that herbivore grazing played a major role in litter decomposition on the Qinghai-Tibetan Plateau.Firstly many studies have demonstrated that litter quality is one of the most important factors affecting the litter decomposition, and litter with higher N content but lower lignin and lignin:N ratio will decompose faster (Aerts, 1997;Olofsson and Oksanen, 2002;Wardle et al., 2002;Garibaldi et al., 2007;Strickland et al., 2009).By following this line, it may be assumed that regardless of incubation site, litter collected from the GP will decompose faster than that collected from GEP.It is true for the "in situ" incubation treatments, i.e., decomposition rate was significantly greater in GP-GP than in GEP-GEP; however, for the across incubation treatments opposite results were found, i.e., it was significantly greater in GEP-GP than in GP-GEP (Table 2 and Fig. 5a).In fact the multivariate regression model shows that litter collected from the GEP had less mass remaining (i.e., 2.5%) compared to that collected from the GP.In the study of tissue and fertilizer N affecting decomposition of conifer litter, Perakis et Moreover, Aerts (1997) suggests that litter chemistry is not a good predictor on litter decomposability in the cold temperate region.Our experimental area locates 3,500 m above sea level with a typical alpine climate having a mean annual temperature of 1.2°C and ranging from -10°C in January to 11.7°C in July.
Alternatively, for a given litter source (i.e., collected from GP or GEP) a greater decomposition rate might be caused by the grazing herbivores when litter was incubated in GP (Table 2 and Fig. 5a), because grazing herbivores may modify site conditions for litter turnover through physical changes in the soil through herbivore activities, such as trampling and urine/dung deposition (Takar et al., 1990;Fahnestock and Knapp, 1994;Luo et al., 2010).Such argument is demonstrated by the results of multivariate regression model: (1) the significantly greater effect of incubation site environment (≈ 25%) than that of litter source or quality, and (2) the the significantly greater effect of GP (8.13%) than that of GEP (Table 3).Grazing elevates litter decomposition is widely reported (e.g., Takar et al., 1990;Fahnestock and Knapp, 1994;Garibaldi et al., 2007;Semmartin et al., 2008;Luo et al., 2010).
On the Qinghai-Tibetan Plateau, Wu et al. (2009) report that a long-term (9 years) grazing exclision favors the increase of soil total nitrogen, soil organic matter, soil organic carbon, soil microbial biomass carbon and soil carbon storage.It is intersting that in the present study, only the SOC significantly increased after three-years grazing exclusion (Fig. 2c); however, grazing exclusion did not significantly modify soil proporty in terms of TN and TP.The increase of SOC in GEP may be because grazing exclusion prevents the reduction of outflow of palatable litter by the herbivores (Fig. 1), and the organic C locked within plant tissues will be returned to the soil during litter decomposition (Bardgett and Wardle, 2003;Wu et al., 2009).Holland and Detling (1990) and Ågren et al. (1999) state that increasing carbon availability in soil will promote decomposer growth and activity even at the low nitrogen concentrations.However, the expected results, i.e. significant higher litter decomposition rate caused by the possible increasing decomposer mass and/or activity in the GEP (Wu et al., 2009), were not observed (Table 2 and Fig. 5a).Therefore, during a relatively short period of time the soil property is unlikely significantly changed through herbivore urine/dung deposition and may have less effect on litter decomposition under cool environments on the Qinghai-Tibetan Plateau.
Our study site is a typical alpine meadow with a long and cold winter and spring (Supplemental Fig. 1).
Climate is the dominant factor regulating litter decomposition and nutrient cycling in such area, and the activity of decomposers may be inhibited during the cold seasons (Coûteaux et al., 1995).These may have interpreted the slow progresses of litter decomposition (Fig. 3) during the first winter and spring seasons when the low temperature inhibited decomposer growth and activity.While the optimum response of decomposers to temperature and moisture occurs during summer and early autumn, elevating the litter decomposition rate (Fig. 3) when both temperature and precipitation simultaneously increase (Supplemental Fig. 1), as reported by Orsborne andMacauley (1988) andDe Santo et al. (1993).Because of the slow biological processes, the decomposition rates estimated in this study (k = 0.38 ~ 0.49, Table 2) is lower than global ones (k ≈ 0.75) with similar latitudes (N30 ~ 40°) (Zhang et al., 2008).

N release
N release is more complex compared to litter decomposition.N release from litter may involve any one or both procedures of N immobilization and N mineralization, where the former results in the accumulation of N in the litter and the latter causes the release of N from the litter (Manzoni et al., 2008).Swift et al. (1979) and Berg and McClaugherty (2014) report that the biological decomposition of litter is mainly carried out by microbial decomposers, including bacteria and fungi, which have higher N/C value compared with most litter types.This property of decomposers creates a high N demand for growth.Therefore, Manzoni et al. (2008)  though a considerable fraction of assimilated C is respired, the decomposers often still require some inorganic N uptake during at least the early phases of decomposition.The increase of N remaining in the GP-GP, GEP-GEP and GEP-GP at beginning of incubation season (i.e., from October to December 2010) in the present study (Fig. 4a-c) provided direct evidence supporting the above assumption of Manzoni et al. (2008).
However, our results also show an increase of N remaining from a lower content in June 2011 to a higher content in September 2011 (Fig. 4).It may be because that the fast decomposition rate of litter during summer (Fig. 3) results in utilization of C by decomposers which increases N:C ratio.Similarly, Bosatta and Balesdent (1996) and Manzoni et al. (2008) have demonstrated a positive correlation between the decomposer N:C and respiration rate (the efficient carbon use) defining the actual nutrient requirement of decomposers; thus the higher respiration rate of decomposers during summer will elevate N:C ratio.Manzoni et al. (2008) also indicate that when N:C is high, large amounts of mineral N are immobilized, which will in turn increase the litter N concentration.However, N release starts after the accumulative N level reaches a critical level (Berg and Staaf, 1981), causing a significant decrease of N remaining during April ~ June 2011 and during the end incubation period (May ~ December 2012) (Fig. 4).
Based on the above knowledge, inverse patterns of litter decomposition and N release were found, i.e., the greater litter decomposition rate was (Fig. 5a), the lower N release rate became (Fig. 5b).Many studies of litter decomposition have found an increase in relative nitrogen concentration in litter as the litter decomposition proceeds (e.g., Aber and Melillo, 1980;Fahey et al., 1991;Gallardo and Merino, 1992).Aber and Melillo (1980) state that an inverse-linear relationship between the remaining biomass and nitrogen concentration in litter is validated for a large number of litter decomposition studies.
The inverse patterns of N release compared to litter decomposition is also demonstrated by the results of multivariate regression model: (1) althought the incubation site environment had significantly greater effect (≈ Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.50%) on N release than did litter source or quality (Table 3), N release was significantly lower in GP (9.27%) than in GEP; (2) litter collected from GP had significantly lower N remaining than did GEP.The latter result contradicts the assumption of Berg and Staaf (1981) that the relatively higher initial nitrogen concentration (see Table 1 for GP litter) may contribute to the higher nitrogen immobilization in litter.The cause is unknown but this phenomenon indicates that litter chemistry may not be a good predictor on litter N release as Aerts (1997) suggests for litter decomposability in the cold temperate region.

Conclusion
Our findings provide insight into our understandings in the litter decomposition and N release affected by herbivore grazing in a typical alpine meadow.Compared to grazing exclusion, grazing with a moderate stocking rate could improve litter quality with higher N but lower hemicellulose and hemicellulose:N ratio, and herbivore activity significantly elevated litter decomposition rate regardless of litter source.While grazing exclusion was more likely to maintain significantly more palatable litter, promoted N release and increased SOC.The different effects of livestock grazing and grazing exclusion on litter decomposition and N release may have implications in the management of alpine meadows on the Qinghai-Tibetan Plateau.For example, grazing exclusion for one season following by livestock grazing will be a good option which allows plant recovery, promotes N release, elevates litter decomposition, and thus contributes to the restoration of degraded grasslands.   1. Mean (± SE) initial contents of carbon (C, mg g -1 ), nitrogen (N, mg g -1 ), phosphorus (P, mg g -1 ), lignin (mg g -1 ), cellulose (mg g -1 ) and hemicellulose, and C:N, lignin:N, cellulose:N and hemicellulose:N ratios in litter*.Table 1.Mean (± SE) initial contents of carbon (C, mg g -1 ), nitrogen (N, mg g -1 ), phosphorus (P, mg g -1 ), lignin 593 (mg g -1 ), cellulose (mg g -1 ) and hemicellulose, and C:N, lignin:N, cellulose:N and hemicellulose:N ratios in litter*.
grazing (GP, 30 m × 20 m) and three grazing exclusion paddocks (GEP, 100 m × 200 m) were established when all aboveground plants are dormant in October 2009.Grazing in GP started with an optimal moderate stocking rate of 4 Tibetan sheep/ha since April 2010.The mean body weight of sheep was about 38 kg when used for the experiment.In October 2010, 20 quadrats (0.5 m × 0.5 m) were randomly established within the GP or GEP, and three sampling methods [i.e., half alongside, half along diagonal, and two sub-quarters (0.25 m × 0.25 m) along Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.diagonal; see Supplemental Fig. S2] were designed to minimize the sample variance caused by the uneven litter distribution and to ensure the similar composition and quality of litter used for this and next experiments.The aboveground portion of all dormant plants (i.e., litter) from each quadrat was sampled for two purposes: (1) measurement of litter composition and quality, and (2) measurement of litter decomposition and N release.
The soil environment modified by land use pattern is thought to be another important driver of variations Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.
Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.total N release did not significantly change during the first winter when temperature < 0°C , except that it significantly increased during the incubation period of December 2010 and March 2011 (first winter) in treatments Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.
Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.al. (2012) revealed that high initial litter N slows decomposition rate in both early and late stages in unfertilized plots.Fertilizer N is usually not applied in the pastures on Qinghai-Tibetan Plateau (F.H personal comment).
Author contributions.YS and FH designed the experiments, YS, ZW and SH Chang performed research and collected data.XZH and YS analysed data and prepared the manuscript, and all authors contributed to the writing.Competing interests.The authors declare that they have no conflict of interest.Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-66Manuscript under review for journal Biogeosciences Discussion started: 23 February 2018 c Author(s) 2018.CC BY 4.0 License.

Figure captions Figure 1 .
Figure captions

Figure 2 .
Figure 2. Comparison of soil property between GP and GEP: (a) TN, (b) TP and (c) SOC.*Significant difference was only

Figure 3 .
Figure 3. Decrease of litter mass remaining with incubation time.For each treatment columns with the different letters are

Figure 4 .
Figure 4. Dynamics of total N remaining with incubation time.For each treatment columns with the different letters are

Figure 5 .
Figure 5. Percentage of litter mass (a) and total N remaining (b) at the end of experiments.Columns with the different letters

Figure 1 .Figure 2 .Figure 3 .Figure 4 .Figure 5 .
Figure 1.Estimated mean (± SE) annual litter mass in GP and GEP.For each category, columns with different 604 , i.e. there is no significant difference in decay rate if their 83.4% CL overlap.The decay rate and 83.4 % CI were estimated by fitting the negative exponential model to a nonlinear least square regression model (NLIN Procedure).

Table 2 .
Estimated litter decay rate (k) in different incubation environments.

Table 2 .
Estimated litter decay rate (k) in different incubation environments.

Table 3 .
Estimated contribution (%) of incubation site environment (Site, GP and GEP) and litter source (Source, 600 GP and GEP) to litter decomposition and N release by multivariable regression model.*