Nitrogen fertilizer regulates soil respiration by altering the organic carbon storage in root and topsoil in alpine meadow of the north-eastern Qinghai-Tibet Plateau

Soil respiration (Rs) plays a critical role in the global carbon (C) balance, especially in the context of globally increasing nitrogen (N) deposition. However, how N-addition influences C cycle remains unclear. Here, we applied seven levels of N application (0 (N0), 54 (N1), 90 (N2), 126 (N3), 144 (N4), 180 (N5) and 216 kg N ha−1 yr−1 (N6)) to quantify their impacts on Rs and its components (autotrophic respiration (Ra) and heterotrophic respiration (Rh)) and C and N storage in vegetation and soil in alpine meadow on the northeast margin of the Qinghai-Tibetan Plateau. We used a structural equation model (SEM) to explore the relative contributions of C and N storage, soil temperature and soil moisture and their direct and indirect pathways in regulating soil respiration. Our results revealed that the Rs, Ra and Rh, C and N storage in plant, root and soil (0–10 cm and 10–20 cm) all showed initial increases and then tended to decrease at the threshold level of 180 kg N ha−1 yr−1. The SEM results indicated that soil temperature had a greater impact on Rs than did volumetric soil moisture. Moreover, SEM also showed that C storage (in root, 0–10 and 10–20 cm soil layers) was the most important factor driving Rs. Furthermore, multiple linear regression model showed that the combined root C storage, 0–10 cm and 10–20 cm soil layer C storage explained 97.4–97.6% variations in Rs; explained 94.5–96% variations in Ra; and explained 96.3–98.1% in Rh. Therefore, the growing season soil respiration and its components can be well predicted by the organic C storage in root and topsoil in alpine meadow of the north-eastern Qinghai-Tibetan Plateau. Our study reveals the importance of topsoil and root C storage in driving growing season Rs in alpine meadow on the northeast margin of Qinghai-Tibetan Plateau.

Human activities such as agricultural expansion, industrial development and deforestation have enormously altered the rate of N deposition 1,2 . Terrestrial ecosystems have been experiencing approximately twice the input of N worldwide compared to the level of the pre-industrial era 2 . According to the results of Lamarque et al. 3 , N deposition in the global terrestrial ecosystems will range from 60 to 100 Tg (N) per year by 2100. Moreover, the rate of N deposition will increase 2.5-fold in the next century 4 . Excess N has caused a range of ecological problems such as a reduction of biological diversity, affecting net primary productivity, community structure and alterations to the C cycle 5,6 . One of the main scientific problems to be addressed is to determine how N deposition influences C cycle 7,8 . Grasslands serve as a very important C pool. They store approximately 34% of the terrestrial global C stock 9 . Soil respiration (Rs) is an important source of C efflux from soil surface to atmosphere 10 . A small change in Rs can alter the concentration of CO 2 in the atmosphere 11,12 . Therefore, understanding the magnitude of CO 2 fluxes emitted from grassland and grassland C pool are crucial to estimate the contribution of grassland ecosystems to the global C budget 13 . Although there have been a lot of studies about the effects of N deposition on Rs, the results remain controversial. For example, Wang et al. 14

Results
Biotic and abiotic variables. During the two experimental years, the above-and below-ground biomass were higher in the N-addition treatments than N0 treatment. Meanwhile, the above-and below-ground biomass showed an increasing trend with increasing amounts of N addition up to the N5 treatment, and then decreased in the N6 treatment (Table 1). N application had no significant impact on litter biomass and soil bulk density in 0-10, 10-20 and 20-30 cm soil layer during the two experimental years. In 2014 and 2015, the soil temperature at 5 cm depth (ST5), soil moisture content at 10 cm depth and (SM10) soil microbial biomass carbon (SMBC), soil organic carbon (SOC) and soil total nitrogen (TN) contents were significantly increased in the N3, N4, N5 and N6 treatments relative to N0 treatment.
Soil respiration and its components. During the two experimental years, the Ra, Rh and Rs showed similar seasonal variation among the various N-addition treatments, with the higher values found in August and lower values found in October (Fig. 1). Meanwhile, during the growing seasons (May -October), the Ra, Rh and Rs showed an increasing trend with increasing amounts of N-addition up to the N5 treatment, and then decreased in the N6 treatment. The growing seasons mean Ra, Rh and Rs were significantly increased in the N3, N4, N5 and N6 treatments in both experimental years compared to the N0 treatment (Fig. 1).
carbon and nitrogen storage. The C storage showed an increasing trend in plant and root with increasing amounts of N addition up to the N5 treatment, and then decreased in the N6 treatment. However, N addition had no impact on litter C or N storage. The vegetation C storage was significantly increased in the N application treatments compared to the N0 treatment in 2014 and 2015. N storage in plant, litter and root showed similar trends to those of C storage. At 0-10 and 10-20 cm soil layer, SOC storage and total N storage differed significantly in the different N-addition treatments, whereas in deeper soil, N application had no significant impact on SOC or N storage (Figs 2 and 3). The SOC storage in the 0-30 cm soil layer was significantly increased in the N3, N4, N5 and N6 treatments in both experimental years (Fig. 2). While the soil total N storage in the 0-30 cm soil layer was significantly increased in the N2, N3, N4, N5 and N6 treatments in both experimental years. The ecosystem C and N storage all showed an increasing trend with increasing amounts of N-addition up to the N5 treatment, and then decreased in the N6 treatment (Figs 2 and 3). Ecosystem C storage and N storage were significantly increased in all of the N-addition treatments except the N1 treatment in both experimental years.
Soil respiration and its components in relation to carbon and nitrogen storage. A principal component analysis (PCA) was used to determine the correlations among the C and N storage in plant, litter and soil (Fig. 4). The results showed that the first, second, and third axes explained 64.17%, 12.24% and 9.57% of the standardized variance, respectively. The highly weighted parameters (loading value > 0.900) on PC1 were soil C storage in the 0-10 cm layer, soil N storage in the 0-10 and 10-20 cm layers, and root C and N storage. The highly loaded parameter for PC3 was soil C storage in the 10-20 cm layer.
We chose the parameters with loading value greater than 0.900 in the PCA results, ST5 and SM10 as predictors to establish structural equation modelling (SEM) to explore the impact of C and N storage, soil temperature and soil moisture content on growing season Rs (Fig. 5). The results explained the changes in Rs well (goodness of fit index was 0.752; χ 2 = 7.99 and P = 0.18). The results showed that C storage had significant effect on growing season Rs (P < 0.05). While N storage, soil temperature and soil moisture had no significant effects on growing season Rs (P > 0.05). Simultaneously, the path coefficient of ST5 on Rs (0.42) was higher than the path coefficient of SM10 on Rs (0.36). Therefore, we concluded that C storage (in root, the 0-10 and 10-20 cm soil depths) is the most predominant variable controlling growing season soil respiration. Meanwhile, ST5 has a stronger effect than SM10 on soil respiration. Rs, Ra and Rh had positive linear relationships with root C storage, 0-10 cm soil layer C storage, 10-20 cm soil layer C storage ( Fig. 6), soil temperature and soil moisture (Fig. 7). Multiple linear regression model showed that root carbon storage, soil organic carbon storage in 0-10 cm layer and soil organic carbon storage in 10-20 cm layer explained more than 94.5% of Rs, Ra and Rh (Table 2).

Discussion
Soil respiration and its components. During the two experimental years, the growing season Rs was significantly increased in the N-addition treatments, which suggests the stimulate effects of N fertilizer on Rs in alpine meadow ecosystem. Similar results were found in alpine steppe 26 and semiarid grassland ecosystem 25,30 . Meanwhile, N-addition significnatly stimulated the growing season Ra and Rh during the two experimental years. Therefore, we concluded that the stimulated effect of N-addition on Rs was mainly due to the increase of Rh and Ra in alpine meadow on the northeastern Qinghai-Tibetan Plateau. The alpine meadow is an N limited grassland ecosystem and has strong capacity to immobilize external N 31 . Therefore, the above-and below-ground biomass was significantly increased after N-addition. This can increase the supply of C substrates to support root and microbial internal stoichiometric requirements and thereby increase Rh 32 . Generally, root activity and biomass are major factors driving Ra 33,34 . Soil C concentration is an important regulator of Rh 21 . The N-addition increases the soil N availability and thus promotes the plants growth in grassland 31 . The increased soil N content may meet the demand of soil microorganisms for N source, thus promoting microbial biomass and activity. Our study showed that the above-and below-ground biomass was significantly increased in the N2-N6 treatments relative to the N0 treatment (Table 1). Larger root biomass represents more root surface area and associated with higher Ra 35 . Therefore, the increased root biomass may improve the Ra. Previous studies also indicated that N-addition increased the root biomass and Ra 25,36 . However, a privious study conducted in a artificial forest showed that 8 years N-addition had no impact on Rs 37 . They interpreted this due to a decrease in microbial biomass after N-addition. Generally, N fertilizer could increase the N source for soil microorganism, and thus increased the microbial biomass and activity. In the current study, the microbial biomass carbon was significantly increased in N3-N5 treatments during the two experimental years ( Table 1). The increased microbial biomass may improve the Rh and thereby increase Rs. Compared to the N5 treatment, the mean Rs in the N6 treatment was decreased by 4.0% and 6.9% in 2014 and 2015, respectively. In the current study, root biomass showed a similar hump-shape pattern response to Rs with increasing level of nitrogen addition (Table 1). Before the N saturation point (180 kg N ha −1 yr −1 ), root biomass increased, whereas after the saturation point, root biomass decreased. Meanwhile, the high-level N addition (>180 kg N ha −1 yr −1 ) can lead to soil acidification 38 , this in turn will accumulate poisonous substances and inhibit the activities of soil microorganisms 39 . Our research showed that the topsoil microbial biomass carbon in the N6 treatment was significnatly decreased by 9.7% and 5.4% in 2014 and 2015, compared to the N5 treatment. Thus, the decreased root biomass and microbial biomass carbon in the N6 site may decrease the root respiration and microbial respiration, and ultimately decrease the soil respiration. www.nature.com/scientificreports www.nature.com/scientificreports/ carbon and nitrogen storage. N-addition increased the content of soil available nitrogen, and thus stimulated the plant growth, and the plant C and N storage increased as a result. The experimental site was grazed in the cold season (November to May) by Tibetan sheep at a high livestock density (6.67-7.22 head per hectare). Almost all the above-ground biomass was removed by the herbivores until May of the next year. There was only little litter (33-38 g/m 2 ) in the experimental site. Meanwhile, there was no significant difference in litter biomass or litter C and N contents among the N-addition treatments. Therefore, N application had no significant impact on litter C or N storage in current study. Previous researches about the effect of N enrichment on soil C storage were still highly controversial. For example, N addition significantly increased the 0-10 cm layer SOC storage in a Europe forest ecosystem 40 . However, some studies reported that N application significantly decreased the SOC storage 41,42 . In these studies, N additions decreased the root biomass which led to little C allocation to belowground, and thus the SOC pool was decreased. In our study, the C storage in the 0-30 cm soil layer increased by 0.6-11.8% in N-addition treatments. Wang et al. 43 reported that the 0-40 cm layer SOC content increased by 34.7% after two years of N addition (200 kg N ha −1 yr −1 ) in an alpine meadow on the Qinghai-Tibetan Plateau. Dong et al. 44 found that the SOC storage in 0-60 cm soil layer increased by 21.7% after 15 years of N addition (200 kg N ha −1 yr −1 ) in a farmland ecosystem of the North China Plain. Generally, SOC storage comes from photosynthetic carbon fixation. Plant root is crucial in controlling SOC pool through root decomposition and exudation. In our study, N addition caused a surge of dominant Elymus nutans, which have relatively greater abundance of leaf area. During the growing seasons, the plentiful sunlight makes the photosynthesis stronger in the N application treatments 45 . Therefore, N addition should stimulate root activity and increase root biomass 46 , thereby increasing soil C and N storage. In addition, the soil temperature difference between day and night is large in alpine meadow on the Qinghai-Tibetan Plateau 47 . The low temperature in the night may restrict the microbial and root activity that limited Ra and Rh 48 . Therefore, the high photosynthetic carbon fixation and low decomposition of organic matter made the higher SOC storage in the N application treatments in alpine meadow ecosystem on the Qinghai-Tibetan Plateau.
N addition influences the composition and structure of the vegetation community (Table 1) and thus influences ecosystem C and N storage 49 . Generally, C and N storage increased with increasing N-addition rate in the current study. However, relative to the N5 level, the highest level of N addition (216 kg N ha −1 yr −1 ) decreased the vegetation, soil and ecosystem C and N storage. These results indicate that C and N storage appears to saturate at a high rate of N addition (216 kg N ha −1 yr −1 ) in excess of the N saturation point (180 kg N ha −1 yr −1 ). In cold region, temperature is the most important factor limiting plant growth 36 . Plants may have a threshold of nitrogen uptake in alpine meadow due to the low temperature, above which plant may not absorb more nitrogen even in high N-addition treatment. Furthermore, the high rate of N addition can lead to soil acidification, which in turn www.nature.com/scientificreports www.nature.com/scientificreports/ will limit plant and root growth, and thus decrease C and N storage in plant and root 36,50,51 . Our research also indecated that the above-and below-ground biomass in the N6 treatment were signifcantly decreased by 15.9% and 9.3%, respectively, compared to the N5 treatment. The decomposition of dead root and litter is the main source of soil organic matter 24 . The decreased plant and root biomass in the N6 site may decrease the vegetation and soil C and N storage. Accordingly, high N addition beyond the N saturation point (180 kg N ha −1 yr −1 ) decreased vegetation, soil and ecosystem C and N storage.

correlations between soil respiration and its components and each of c and n storage.
Previous researches about the factors driving Rs have mainly focused on ST, SM, plant, litter and below-ground biomass 52,53 ; and soil C and N concentration 18 . Generally, ST and SM are two key abiotic factors that control daily and seasonal variation in Rs across different vegetation types 54 . Previous study in a Tibetan Kobresia  www.nature.com/scientificreports www.nature.com/scientificreports/ pastures found that ST5 had a stronger impact than did SM10 on growing season Rs 22 . However, studies conducted in temperate grassland and steppe grassland reported that SM10 had a stronger impact than did ST5 on growing season Rs 24,55,56 . In the present study, the SEM analysis revealed that soil temperature had a stronger effect than did soil moisture on Rs during the growing seasons. Soil temperature and moisture affect respiration by affecting respiratory enzymatic activity and substrate supply 57 . Generally, low temperature and soil water content limit the enzymatic activity and diffusion of soluble substrates; while high temperature and soil moisture also limit the enzymatic activity and the diffusion of oxygen, thus, both of which can suppress Rs 57 . Therefore, appropriate soil temperature and soil moisture are essential for soil respiration. Yan et al. 58 reported that there was a threshold (14%) at which the effects of volumetric soil water content on Rs changed, above which soil moisture is not the limiting factor of Rs 58 . Meanwhile, a previous study indicated that there www.nature.com/scientificreports www.nature.com/scientificreports/ was a threshold at which the effects of volumetric soil water content on soil respiration changed; that is, the two were positively correlated at soil water contents of 30% or lower, and negatively correlated at soil water contents of >30% 59 . In our study, the volumetric soil moisture in growing season was 25.3-27.2%, which is no longer the factor that influences the diffusion of soluble substrates. Furthermore, the alpine meadow on the Qinghai-Tibetan Plateau is a temperature limited ecosystem 60 . The mean ST5 of the growing season was between 10.1°C and 10.9°C, which may be below the threshold of soil temperature, and thus limit the respiratory enzymatic activity.
Although soil temperature and soil moisture had important effect on Rs, some previous researches indicated that some biotic parameters might have stronger effects on the Rs rate than do either SM10 or ST5 23,61 . For example, previous studies showed that root and soil carbon storage significantly control the Rs through manipulating Ra and Rh in forest ecosystem 23,62 . Li et al. 24 also revealed that the root and topsoil C pools were two most important factor that controlling the Rs in an alpine meadow on the Qinghai-Tibetan Plateau. In our study, the SEM results showed that C storage (in root, 0-10 and 10-20 cm soil depths) have greater influences on Rs than do ST5 or SM10 (Fig. 5). Indicating that root C storage and C storage in topsoil (in 0-10 and 10-20 cm soil depth) are two important factors driving Rs. This mainly attribute to the impacts of belowground C storage on the production of detritus and exudates for soil microbial activities 23 . C storage is a complex parameter representing the vegetation biomass and C concentration. The increase in belowground C storage was related to an increase in root biomass and thus increased Rs. We found that N-addition significantly increased the root and soil carbon storage (Fig. 2). General linear regression analysis indicated that growing season Rs, Ra and Rh had significant positive correlation with root carbon storage and soil carbon storage (Fig. 6). The reason for this was the increased soil and root C storage that can meet the microbial demand for C source, and thereby increased Rh. Meanwhile, the combined root carbon storage, 0-10 cm and 10-20 cm soil layer carbon storage explained 97.4-97.6% variations in Rs; explained 94.5-96% variations in Ra; and explained 96.3-98.1% in Rh, which was higher than the combined soil temperature and soil moisture did (Table 2). Therefore, the growing season soil respiration and its components can be well predicted by the root and topsoil organic C storage in alpine meadow of the north-eastern Qinghai-Tibetan Plateau.

conclusion
Short-term N-addition can significantly increase the soil respiration and its components, C and N storage in plant, root and topsoil (0-10 cm and 10-20 cm soil layer). Meanwhile, our results revealed that the stimulated effect of N-addition on Rs was mainly due to the increase of Rh and Ra. In addition, the ecosystem C and N storage appears to saturate at a high rate of N addition (216 kg N ha −1 yr −1 ) in excess of the N saturation point (180 kg N ha −1 yr −1 ). However, the current study was a short term experiment, and how Rs and C storage respond to the long-term N-addition remains unclear. Therefore, further studies should consider these for better understanding of the impacts of increasing N enrichment on C cycling in grassland ecosystem. The SEM results showed that C storage in root, 0-10 and 10-20 cm soil depths have greater influences on soil respiration than do soil temperature or soil moisture. Meanwhile, multiple linear regression model showed the combined root carbon storage, topsoil carbon storage explained 97.4-97.6% variations in Rs; explained 94.5-96% variations in Ra; and explained 96.3-98.1% in Rh, which was higher than the combined soil temperature and soil moisture did. Therefore, the growing season soil respiration and its components can be well predicted by the root and topsoil organic C storage in alpine meadow of the north-eastern Qinghai-Tibetan Plateau. The SEM results also indicated that soil temperature had a greater impact on growing season soil respiration than did volumetric soil moisture. Considering productive functions and ecosystem C storage, the optimum level of N fertilizer is 180 kg N ha −1 yr −1 in alpine meadow of the northeast margin of Qinghai-Tibetan Plateau.

Materials and Methods
Study area. The research was performed in an alpine meadow region of the northeast margin of Qinghai-Tibetan Plateau located in Zhuaxixiulong Township, Tianzhu Tibetan Autonomous County, Gansu Province, PR China (37°11′N, 102°46′E; 2960 m a.s.l.). The area is approximately 180 km northwest of Lanzhou, PR China. The mean annual temperature in the study area was 0.13 °C (varying from −11.4 °C in January to 11.2 °C in July) and mean annual precipitation was 414.98 mm, approximately 63.3-88.2% of the rainfall falling between June and September (climatological data from 1951 to 2016) 63 . In this region, the vegetation growing season is from May to October. The soil in the study area belongs to alpine chernozem. The soil organic carbon and total nitrogen content of the 0-10 cm soil layer at the beginning of the study ranged from 47.6 to 49. www.nature.com/scientificreports www.nature.com/scientificreports/ field measurements and sampling. Soil respiration and its components. In late May of 2013, four intact, untrenched subplots (1 m × 1 m) were randomly established in each plot. Meanwhile, we excavated a trench (60 cm diameter × 70 cm deep, no root deeper than 70 cm) on the outside edges of each untrenched subplot. For a total of twelve untrenched and trenched subplots of each N addition treatment, respectively. The location of each untrenched and trenched subplots were at least 1 m away from the margin to avoid edge effects. The trenches were lined with a nylon mesh bag (0.038 mm mesh, 60 cm diameter × 75 cm deep) to prevent root growth into the plots (but allow the movement of water, microorganism, organic matter and minerals). We removed the root using a 2-mm sieve and then refilled the trenches with the root-free soil layer by layer 25,66 . Therefore, the CO 2 efflux surveyed in the treated plots was Rh, and CO 2 efflux surveyed in the untrenched subplots was Rs. The difference between treated subplot and untrenched subplot was Ra. This root exclusion method has been widely used in the field to partition Ra and Rh 25,66,67 .
Soil CO 2 flux was measured from May to October in 2014 and 2015 with a portable automated soil CO 2 flux system (LI-8100 A, LI-COR Inc., Lincoln, USA). A PVC collars (20 cm diameter and 15 cm height) were inserted into each untrenched and trenched subplots to a depth of 3-5 cm below the surface for soil CO 2 flux measurement one day before each survey day. The soil was tamped outside of the collar to prevent air leakage. All soil CO 2 flux measurements were taken between 09:00 and 14:00 on sunny days. Before the measurements, all plants in each PVC collar were clipped. The measurements were conducted twice a month (early and late in the month) for four days each time. ST5 and SM10 were measured simultaneously near each PVC collar along with the Rs measurement by a probe connected to the LI-8100 A.
Vegetation structure and soil properties. Vegetation and soil sampling were conducted in early September of 2014 and 2015. In each plot, eight 1 m × 1 m quadrats were randomly established 1 m from the edge. In each quadrat, the height and density of each plant and the aboveground, belowground and litter biomass were investigated. Aboveground biomass was cut at ground level for species and all litter in each quadrat was collected by hand. We used a 10 cm inner diameter auger to measure the root biomass in 0-10, 10-20 and 20-30 cm soil layer in each harvested quadrat. We collected eight soil cores at each plot, and twenty-four soil cores per N-addition treatment. We separated out the root by washing the soil samples within a 0.5-mm mesh bag. The plant, litter and root samples were first dried at 105 °C for 30 min, and then dried at 65 °C to constant weight. We used a cutting ring (100 cm 3 volume) to measure the soil bulk density at 0-30 cm soil depth in each of the harvested quadrat (i.e., the soil was divided into 3 layers of 10 cm each).
We used a 3.5 cm inner diameter auger to take soil samples in each harvested quadrat (3 layers of 10 cm each) to measure SOC, TN contents. Eight cores were collected in each plot, and twenty-four soil cores per N-addition treatment. Soil samples were air-dried and sieved to pass a 2 mm mesh sieve for testing the SOC and TN. Meanwhile, the surface layer (0-10 cm) soil was collected with a 3.5 cm inner diameter auger to measure the SMBC content in each harvested quadrat. Eight cores were collected in each plot, and twenty-four soil cores per N-addition treatment. The soil samples were immediately sieved through a 2-mm screen in the field and kept refrigerated. chemical analysis. The plant, litter, root and soil organic carbon contents were measured using a total organic carbon analyzer (Multi N/C 2100 s, Analytik Jena Germany). The plant, litter, root and soil total N contents were measured via the Kjeldahl method 68 . The SMBC content was measured via the chloroform fumigation extraction method. The carbon in the extract was measured through dichromate oxidation method, and the conversion coefficient was 0.38 69 . Statistical analysis. The plant C storage and N storage were calculated by the methods of Fang et al. 70 ; the soil organic C storage and N storage were calculated using the methods of Ma et al. 71 . The C storage and N storage in the alpine meadow ecosystem were considered the sum of the plant, root, litter and soil C storage and N storage, respectively.
Repeated-measures ANOVA was applied to compare the significant differences in Rs, Ra and Rh, biotic and abiotic variables among the N-addition treatments. Considering the strong correlations among C and N storage in plant, litter, root, 0-10 cm soil layer, 10-20 cm soil layer, 20-30 cm soil layer, we used the PCA to determine the primary axes of covariation among the variables. Based on the PCA results, we chose the parameters with loading value greater than 0.900 and soil temperature and volumetric soil moisture as predictors to establish the SEM to identify the direct and indirect effect pathways on soil respiration. SEM was conducted with the Amos software program ver. 21.0.
Meanwhile, linear regression was conducted with the Origin software program ver. 8.5 to examine the relationships between soil respiration components (Rs, Ra and Rh) and the following: root C storage, 0-10 cm soil layer C storage, 10-20 cm soil layer C storage, soil temperature and soil moisture. The equation was as follow: R a bx (1) = + where R is Rs, Ra, or Rh; x is root C storage, 0-10 cm soil layer C storage, 10-20 cm soil layer C storage, soil temperature or soil moisture, and a and b are fitted parameters.
We used the following multiple linear regression to test the combined effect of root carbon storage, 0-10 cm layer carbon storage and 10-20 cm layer carbon storage on soil respiration rate (conducted with the Origin software program ver. 8.5): 10 20 where R is Rs, Ra, or Rh, RCS is root carbon storage, SOCS 10 is 0-10 cm layer carbon storage, SOCS 20 is 10-20 cm layer carbon storage, and c, d, e and f are fitted parameters. (2019) 9:13735 | https://doi.org/10.1038/s41598-019-50142-y www.nature.com/scientificreports www.nature.com/scientificreports/ We used the following multiple linear regression to test the combined effect of soil temperature and moisture on soil respiration rate (conducted with the Origin software program ver. 8.5): where R is Rs, Ra, or Rh, ST5 is soil temperature at 5 cm depth, SM10 is volumetric soil moisture at 10 cm depth, and g, h, and i are fitted parameters. All statistical analyses were conducted using the SPSS software program ver. 19.0 (SPSS, Chicago, IL).