Rebuilding soil organic C stocks in degraded grassland by grazing exclusion: a linked decline in soil inorganic C

Study sites

The studied site was located in the Yunwu Mountain National Nature Reserve (36°10′–36°17′N, 106°21′–106°27′E), 45 km northeast of Guyuan City, Ningxia Autonomous Region, China in the hinterland of the Loess Plateau at an altitude of 1,700–2,148 m. It belongs to the temperate semi-arid climate zone. The mean annual precipitation for the period 1,980–2,014 was 425 mm, with 60%–75% of total precipitation occurring between July and September. The main soil types are locally known as ‘mountain gray cinnamon’ soil and ‘black loess’, which are Alfisols and Mollisols, respectively, according to the USDA classification system. Atmospheric precipitation mainly replenishes water resources. The region’s dominant plants are Stipa bungeana, Stipa grandis, Thymus mongolicus, Artemisias acrorum, and Potentilla acaulis.

Since 1980, the regional government has implemented several mountain closure and grazing prohibition measures, closing plots to livestock grazing and making them available for study. We selected sites in the Yunwu Mountain Nature Reserve that had similiar soil and vegetation types, were distributed over a distance of up to 45 km with a rolling topography, and had been fenced for 15 or 30 years (F15 and F30, respectively) together with similar sites still being grazed (G). There were three replicate sites in each case. Site information appears in Table 1.

Experiment design and sampling

We designed to investigate the differences among the G, F15, and F30. Three 10 m ×10 m plots were set for each level, with nine sites in total. At each of the nine sites, we selected a homogeneous sampling area deemed visually representative of the site with an area of approximately 900 m². Three 1 ×1 m square quadrats were placed at 10 m intervals along a transect line for sampling. A six cm diameter auger (S1 Canada) was used to extract soil samples within each quadrat (three samples per quadrat) to a 500 cm depth. Soil samples were packed into Ziploc bags on site for processing in the laboratory. The 0–40 cm soil depth samples were collected in 10 cm depth increments, and samples from 20 cm depth increments were collected from 40–500 cm, making a total of 27 samples per core. To find the bulk density (BD), we used a 51-mm soil sampling drill (S1 Canada) to obtain an undisturbed soil core from 0 to 500 cm, and a soil bulk sampler method to measure the BD (g cm⁻³) of the different soil layers (0–10, 10–20, 20–30, 30–40 cm; from 40 to 500 cm, samples were taken every 20 cm). In the undisturbed soil cores from the FG and G plots, the sampler was 100 cm³ and the sampler’s inner diameter was 50 mm (Li et al., 2019). In the laboratory, we determined the subsamples’ soil moisture content, then sample remainders were air dried and passed through a two mm sieve to remove mixed litter and roots. Nine cores per sampling site were then bulked together for each soil layer. The analyses for SOC, SIC, and pH were performed in triplicate on subsamples from the bulked site samples. SOC and SIC were measured using an elemental analyzer (Vario EL/micro cube, Germany) and soil pH value was measured using an acidity agent (PHS-3C pH acidometer, China).

Soil carbon calculations

The SOC and SIC stocks (Mg hm⁻²) were calculated as follows (Li et al., 2019):

(2.1)${\displaystyle SOCstorage=\sum _{i=1}^n{D}_i\times B_i\times SO{C}_i}$ (2.2)${\displaystyle SICstorage=\sum _{i=1}^n{D}_i\times B_i\times SI{C}_i}$

Where n is the number of soil layers, D_i is the soil depth (cm), B_i is the soil bulk density (g cm⁻³), SOC_i and SIC_i are the SOC content and SIC content (%) of soil layer_i.

The soil total carbon (STC) stocks were calculated as: (2.3)${\displaystyle STCstorage=SOCstorage+SICstorage}$

And the changes in soil C stocks (C change, Mg hm⁻²) were: (2.4)${\displaystyle Cchange=C_t-C_0}$

Where C_t represents soil stocks in the fenced grassland (Mg hm⁻²), and C₀ is the soil stocks in the grazed grassland (Mg hm⁻²).

Lastly, the rate of soil carbon stock change (RSS, Mg hm⁻² yr⁻¹) was estimated by the linear regression of change in soil C stocks with time since fencing: (2.5)${\displaystyle Cchange=f\left(\Delta time\right)=y_0+K\times \Delta time}$

Where y ₀ is a constant, K is the rate of soil C stock change (Mg hm⁻² yr⁻¹), and △ time is the fenced time interval (i.e., years from grazing cessation to the present).

Statistical analyses

Various statistical analyses were performed using the software packages SAS 9.4, Origin 2017, and Microsoft Excel. Since soil depth data taken from the same cores can suggest correlation across soil depths, all data were first analyzed using One-Way ANOVA comparing the time spent fenced off with soil depths as a statement of “ mixed effect model” in Proc GLM in SAS (Li et al., 2019). For presentation and statistical purposes, the bulked samples from the nine sites (each consisting of 27 soil layers) were averaged across each of the six soil layers (0–40, 40–100, 100–200, 200–300, 300–400, and 400–500 cm soil depths). The soil depth effect had a degrees of freedom (df) of 5; the interaction effect between time fenced and soil depth had a df of 10, with an error df of 30 and a total of 53 (9 × 6–1) df in each repeated measures ANOVA. SAS provides two options for adjusting the p-value of the main-factor ×repeat-factor interaction across soil depths, and we used the adjusted p-value based on the Greenhouse-Geiser Epsilon. We then processed the SAS repeat measure output using the other packages. For example, we used SPSS to obtain the standard errors of individual means across soil depths, and Minitab Version 10.51 to perform Pearson correlation analyses.

Soil bulk density, soil moisture (SWC), and pH

The observed soil bulk density values ranged from 0.91 g cm⁻³ in the 0–40 cm soil layer of the F15 plots to 1.67 g cm⁻³ in the 100–200 cm layer of the F30 plots. For subsoil with low SOC, typical values were 1.2–1.28 g cm⁻³. Values for the 0–40 cm layer were lower, reflecting the accumulation of SOC in the upper soil. For the grazed plots, the mean bulk density in the 0–40 cm soil depth was 1.09 g cm⁻³, and was reduced to less than 1.00 g cm⁻³ in the F15 and F30 plots (Table 2). The results from repeated measures ANOVA showed that time fenced and soil depth both significantly affected soil bulk density (F = 23.10, p < 0.0001; Table 3). The high bulk density of 1.67 g cm⁻³ in the 100–200 cm layer of the F30 plots was atypical, since all the others were observed at less than 1.30 g cm⁻³, so we noted a correlation of –0.802 (p = 0.008) between soil bulk density and total carbon (TC) (Table 4). The pattern of variation in soil water content (SWC) generally followed the pattern of variation in SOC, with a correlation of r = 0.849 (Table 4; p < 0.001). Thus, SWC was higher in the surface horizon than in deeper horizons, and lower in grazed than in F15 and F30 grasslands (Table 2). Although SWC was (not unexpectedly) higher in the 100–300 cm soil horizon than in the 300–500 cm soil horizon (9.4% and 10.3% on average, respectively; p = 0.399), SOC values were lower at the same depths (0.47% and 0.29%, respectively; p = 0.362).

All soil pH values were alkaline, ranging from the lowest observed value of 7.76 in the 0–40 cm soil layer of F30 plots to 9.20 and 9.18 in the 200–300 and 300–400 cm soil layer of grazed plots. Notably, across soil depth and fencing duration data, pH values were strongly negatively correlated with SOC and strongly positively correlated with SIC (r =  − 0.890, p < 0.001 and r = 0.713, p = 0.001, respectively; Table 4). pH was consistently lower in F15 and F30 plots than G plots in soil up to 400 cm deep.

Soil organic and inorganic carbon

In grazed grassland, SOC g kg⁻¹ ranged from 14.3 g kg⁻¹ in the 0–40 cm soil horizon to 3.1 g kg⁻¹ in the 400–500 cm soil horizon (p < 0.05). In F30 grassland, the corresponding values were 22.9 g kg⁻¹ (a 60% increase compared to G, p < 0.05) and 3.1 g kg⁻¹ (unchanged). For the soil depth of 200–300 cm, F15 had lower SOC levels than grazed grasslands but F30 had the highest, and for the 300–500 cm soil depth, the SOC levels of F15 were not significantly different than grazed grasslands but F30 had higher levels (Table 5). By contrast, SIC values changed much less with soil depth, but were higher in the deep soil layers compared to surface levels in grazed sites (p < 0.05). The opposite was true for F30 and F15 grassland when compared to grazed grassland (grazed 0–40 cm soil depth, 17.1 g kg⁻¹ SIC; grazed 400–500 cm soil depth, 18.9 g kg⁻¹ SIC; F30 0–40 cm soil depth, 10.8 g kg⁻¹ SIC; F30 400–500 cm soil depth, 16.9 g kg⁻¹ SIC) (Table 6; p = 0.0046 for the GG adjusted, time fenced ×soil depth interaction). Repeated measures ANOVA showed that time fenced and soil depth both significantly affected SOC content (F = 502.4, p < 0.0001; Table 3).

The SOC and SIC data are presented as total carbon stocks (Mg hm⁻²) in Figs. 1–3 show the changes relative to grazed sites. These data illustrate that the SOC accumulation at this set of sites extended below 100 cm soil depth and continued 15 and 30 years after fencing, as evidenced by a F15–F30 separation and positive movement relative to grazed plots at the 0–200 cm soil depth (Figs. 2A, 3A). The rate of change for years 0–30 was different from the rate of change for years 0–15, particularly in the 0–200 cm soil depths where SOC accumulation occurs. In all three of the upper soil layers (0–40, 40–100, and 100–200 cm soil depths), SOC accumulation at the 40–100 cm soil depth occurred faster in the first 15 years after fencing, and occurred mainly in years 15–30 at the 100–200 cm soil depth. A reciprocal trend was seen in SIC data, but because of greater variability in this data, the trend could not be confirmed as statistically significant.

After comparing the SOC and SIC data, the effects of fencing on SIC stock were considered non-significant below 200 cm soil depths. However, in the 0–200 cm soil depth, the SIC stock decrease was larger than the SOC stock increase, to the extent that total soil C stocks (SOC stock + SIC stock; TC) across the 0–500 cm sampled soil profile were lower at fenced sites compared to grazed sites (Grazed 1345 Mg hm⁻²; F30 1310 Mg hm⁻²; p = 0.0479).

Total carbon stock percentages of the SOC and SIC pools

The average SOC stock was 270 Mg hm⁻² in grazed grassland, 304.7 Mg hm⁻² in F15, and 461.6 Mg hm⁻² in F30 grasslands. SOC stock significantly increased at the F30 sites, compared to the F15 sites, while the SIC stock showed an opposite trend. SIC stocks decreased with time fenced, suggesting that STC did not significantly differ with years fenced (Fig. 4).

SOC stocks for the 0–40 cm soil depth significantly increased at F30 but not F15 sites (Fig. 5A), while at 40–100 cm and 100–200 cm soil depths, SOC was significantly greater (p < 0.05) in F15 than in grazed sites, and significantly greater (p < 0.05) again in F30 than in F15 sites (Figs. 5B, 5C). In contrast, SIC stocks were lower in F15 and F30 sites compared to grazed sites at all soil depths, and the reduction was greater for F30 than F15 sites at deeper soil depths of 40–100 and 100–200 cm (p < 0.05; Figs. 5D–5F). When observing the combined effect of SOC and SIC changes, the STC for the 0–40 cm soil depth was reduced at F15 sites compared to grazed sites but had recovered at F30 sites; at 40–100 cm soil depths, the STC was reduced at F15 and F30 sites compared to Grazed sites; and at 100–200 cm soil depths, the STC did not differ among F15, F30, and grazed sites (Figs. 5G–5I).