Phytolith-occluded carbon (PhytOC) is an important long-term stable carbon fraction in grassland ecosystems, and plays a promising role in global carbon sequestration. Determination of the PhytOC traits of different plants in major grassland types is crucial for precisely assessing their phytolith carbon sequestration potential. Precipitation is the predominant factor in controlling net primary productivity (NPP) and species composition of the semiarid steppe grasslands. We selected three representative steppe communities of the desert steppe, the dry typical steppe and the wet typical steppe in Northern Grasslands of China along a precipitation gradient, to investigate their species composition, biomass production and PhytOC content for quantifying its long-term carbon sequestration potential. Our results showed that (i) the phytolith and PhytOC contents in plants differed significantly among species, with dominant grass and sedge species having relatively high contents, and the contents are significantly higher in the below- than the aboveground parts. (ii) The phytolith contents of plant communities were 16.68, 17.94 and 15.85 g kg^-1 in the above- and 86.44, 58.73 and 76.94 g kg^-1 in the belowground biomass of the desert steppe, the dry typical steppe and the wet typical steppe, respectively; and the PhytOC contents were 0.68, 0.48 and 0.59 g kg^-1 in the above- and 1.11, 0.72 and 1.02 g kg^-1 in the belowground biomass of the three steppe types. (iii) Climatic factors affected phytolith and PhytOC production fluxes of steppe communities mainly through altering plant production, whereas their effects on phytolith and PhytOC contents were relatively small. Our study provides more evidence on the importance of incorporating belowground PhytOC production for estimating phytolith carbon sequestration potential, and suggests it crucial to quantify belowground PhytOC production taking into account of plant perenniality and PhytOC deposition over multiple years.

Sampling sites

This study was conducted at three sites in the steppe region of central Inner Mongolia along a climatic gradient of increasing annual precipitation, that is, at a desert steppe site (within Sunite Right Banner, at 43°511'N, 113°42'E), a dry typical steppe site (within Maodeng farm of Xilinhot city, at 44°50'N, 116°36'E), and a wet typical steppe site (within West Ujimqin Banner, at 45°43'N, 118°30'E). The region experiences a temperate semiarid climate. The mean annual temperature (MAT) were 0.19, 2.67 and 3.32℃, and mean annual precipitation (MAP) were 182, 278 and 342 mm, respectively at the desert steppe, the dry typical steppe and the wet typical steppe sites; and 75%-85% of annual precipitation falls in the plant growing seasons from May to September (average of the 1960-2016 period). In the year for field study (2016), the annual temperature (TEMP) were 0.98, 3.49 and 4.50℃ and annual precipitation were 189, 309 and 299 mm, and the plant growing-season precipitation were 129, 215 and 237 mm, respectively in the desert steppe, the dry typical steppe and the wet typical steppe sites. The desert steppe site is on a calcic brown soil, whereas the other two steppe sites are on chestnut soil. The humus layer is 15-30 cm in the soil profile, and the calcic horizon (mostly CaCO₃) is 30-60 cm in the soil profile, both increasing from the desert steppe to the dry typical steppe and the wet typical steppe. The dominant species of the vegetation are Stipa klemenzii, Cleistogenes songorica, Allium bidentatum and Salsola collina in the desert steppe, Leymus chinensis, Stipa krylovii and Cleistogenes squarrosa in the dry typical steppe, and Leymus chinensis, Stipa grandis and Cleistogenes squarrosa in the wet typical steppe (Table 2).

Field sampling

In each of the three steppe sites along the precipitation gradient, three delineated plots of 20m × 20m located in three separate farms were selected for plant and soil sampling. These plots were on flat ground and covered with representative native steppe communities, and fenced to exclude animal grazing at the beginning of the plant growing season in 2016. Plant and soil samples were collected from these plots at the end of August. Five quadrates of 1m ×1m were set up at the center and four corners of each delineated plots, and all standing live and dead vascular plants (that was produced during the current season) in these quadrats were harvested at ground level species by species, dried to a constant weight at 65℃ and weighed. The dry mass of all plant species per quadrat averaged over five replicates was used to determine the aboveground plant biomass at peak plant biomass time, and this was also used to approximate ANPP of the grassland (Scurlock et al., 2002). The belowground biomass and its distribution profile (0-70 cm) were measured using the soil coring method, and BNPP (during the plant growing season from May to October) of the studied grassland was obtained from previous studies (Chai et al., 2014; Hou et al., 2014).

Whole plants of dominant species in each sampling plot (within a separate farm) were collected by digging up each individual to a depth of 20 cm below ground level, and then each individual was cut into two parts: aboveground part (shoots) and belowground part (noted as roots, but it includes plant roots and rhizomes as well as shoot stumps buried below the soil surface). A sample of about 300g dry matter of the shoots and the roots of each plant species was collected in each of the three sampling plots at each site. The samples were washed with deionized water, dried at 65℃ and then cut into pieces (<5mm) for phytolith analysis.

The soil bulk density and moisture content of soil profile (0-70 cm) were obtained by the cutting-ring method and the oven drying method (at 105℃). The soil samples were collected using soil cores (diameter=7cm) and air dried at ambient temperature in the laboratory.

Sample analysis

The phytoliths within plant parts were extracted using a microwave digestion process (Parr et al., 2001) followed by a Walk-Black type digestion to ensure the purity of the phytoliths (Parr & Sullivan, 2014; Walkley & Black, 1934). Two duplicates were analyzed for each plant sample. The exacted phytoliths were dried at 65℃ to a constant weight. The PhytOC was determined using the PhytOC alkalidis solution spectrophotometer method (Yang et al., 2014). In this method, sodium hydroxide solution was used to dissolve the Si compound in order to release the occluded organic carbon from the phytoliths, then potassium dichromate (K₂Cr₂O₇)-sulfuric acid (H₂SO₄) solution was used to oxidize the released organic carbon, and the concentration of Gr³⁺ produced in this oxidation was determined by spectrophotometer with its absorbance at 590 mm wavelength. The organic carbon concentration was calculated based on the amounts of potassium dichromate consumed, and the accuracy and repeatability of the method were well verified against the results obtained with acid dissolution-Elementar Vario MAX CN method (Germany) (Yang et al., 2014). The phytoliths and PhytOC contents of the two parts of each plant species were calculated as the average of the three replicate plots. For each species, the ratio of aboveground to belowground biomass (shoots/roots) was calculated based on the sampled plant individuals. The ratio was used to calculate the belowground biomass of the species in 1 m² based on the measured aboveground biomass of the species.

The air dried soil samples was separated into the 100 mesh soil samples and the 10 mesh soil samples. The soil organic carbon (SOC) was determined with the 100 mesh soil samples using the method of classical potassium dichromate (Walkley & Black, 1934), and the soil pH and bioavailable Si content were analyzed with 10 mesh soil samples using a pH meter and silicomolybdic acid method (Yang et al., 2018), respectively.

Data calculations and statistics

The formula for calculating phytolith and PhytOC contents in plant species were as follows:
    phytolith content (g kg^-1) = phytolith weight (g) / dry biomass (kg) [1]
    PhytOC content (g kg^-1) = PhytOC weight (g) / dry biomass (kg)         [2]

The community weighted mean contents of phytolith (Phytolith_CWM) and PhytOC (PhytOC_CWM) were also calculated using the contents of phytolith and PhytOC in the species of the community and the relative biomass of the species (as weight), similar to the calculation of other community-weighted mean plant traits (Ricotta and Moretti, 2011).

    phytolith_CWM (g kg^-1)=∑_i phytolith content_i (g kg^-1)×biomass_i (%)            [3]
   PhytOC_CWM (g kg^-1)=∑_i PhytOC content_i (gkg^-1)×biomass_i (%)               [4]

i enumerates each species.

    PhytOC stock (kg ha^-1)= PhytOC_CWM (g kg^-1)×biomass (kg ha^-1)×10^-3         [5]
    PhytOC production flux (kg ha^-1 yr^-1) = PhytOC_CWM (gkg^-1)×NPP (kg ha^-1 yr^-1)×10^-3           [6]

One-way ANOVA and Duncan's multiple range test were performed to examine the difference in phytolith and PhytOC contents among different parts of plant species. A principal component analysis (PCA) of PhytOC content and production parameters and environmental factors were performed to show their interrelations. R 3.3.3 was used for all the statistics and Sigma Plot 12.0 was used for figures.