Increased precipitation enhances soil respiration in a semi-arid grassland on the Loess Plateau, China

Study sites

The study was conducted at the Agriculture Experimental Station of Ningxia University , Yinchuan Province of China, in Yunwu Mountain of Guyuan, Ningxia (106°21′E−106°27′E, 36°10′N−36°17′N). The climate of the area is temperate, continental monsoon. The annual average precipitation is about 439 mm and varied from 282 mm in 1982 to 706 mm in 2013. More than 50% of the annual precipitation occurs in the summer months (June to August). The average annual temperature is 7.2 °C, and varied from 5.3 °C in 1984 to 8.7 °C in 2013. The average monthly minimum temperature for the coldest month (January) was −7.2 °C, and the average monthly maximum temperature for the warmest month (July) was 19.6 °C. The annual evaporation is 1,300–1,640 mm, and the annual duration for sunshine can reach 2500 h, with a frost-free period of 112–140 days. The annual potential evapotranspiration is 1,625 mm (meteorological data from 1981 to 2017 is from the National Meteorological Administration of China). The soils are grey-cinnamon and dark loessial, as classified by the Chinese soil classification system (National Soil Survey Office, 1993). The vegetation is typical steppe and the main plant species are Stipa bungeana, Artemisia gmelinii, Stipa grandis, Artemisia frigida, Potentilla acaulis and Agropyron michnoi (Wang et al., 2020).

Experimental design

Our research site was located in a semi-arid natural grassland that was left ungrazed for 19 years. The study area was at 2,077 m, with a 7−10° slope, and a south-facing, sunny aspect. Annual precipitation in 2019 was 592 mm, which was 20% higher than average. We set three blocks with three 6 × 6 m plots in each block, to make a total of 9 plots. Each trio of plots down the gradient was considered to be a block (Fig. 1). According to local multi-year meteorological data from the study area, the mean maximum and minimum precipitation levels were about 50% and 150% of the average annual precipitation. A Rain shelter was set up on each block. Each rain-shelter was fixed to the ground by steel pillars, and transparent polyethylene plates were fixed in “V” shapes to intercept precipitation and channel it off the plot using the natural slope of the mountain, while forming a stable and well-ventilated structure (Afreen & Singh, 2019). Rain shelters intercepted half of the natural precipitation to form a reduced precipitation treatment (R50). The intercepted water was piped to an adjacent plot to form an increased precipitation area (R150). The remaining three plots were the controls (R100). Snow was collected from the rain shelters after each snowfall (R50) and was sprinkled evenly into the R150 plots. Plastic barriers were used to prevent surface runoff or leakage of soil moisture between plots. Each barrier was buried at a depth of 1.1 m and projected 10 cm above ground. Our study ran from May 2017 to May 2019.

Environmental factors

We collected field data in July 2019 which corresponded to the annual period of peak biomass. Three soil sample replicates were collected at depths of 0–9.9 cm, 10–19.9 cm, and 20–30 cm in each plot, after the litter was discarded. The soil samples were separated into two parts: one part was kept moist to determine the microbial diversity of the soil, and the other was air-dried for measurement of soil properties.

Soil organic carbon (SOC) was measured by potassium dichromate-sulfuric acid digestion, with ammonium ferrous sulfate titration. Total nitrogen (TN) was determined using an Elementar analyzer (Elementar, Vario EL III, Germany). Total phosphorus (TP) was measured using Olsen’s method (Bremner & Mulvaney, 1982). Soil pH was measured using a PHS-3C pH Meter in a 1:5 ratio of fresh soil: water slurry (Huakeyi, Beijing, China).

A 1 m² quadrat was randomly selected in each subplot to determine the plant biomass. The litter was raked and bagged and the shoots of the plants were cut at ground level. Root biomass (RB) was sampled to 30 cm in three intervals of equal depth and the soil carefully brushed off the roots. All plant samples were dried at 65 °C in an oven for 75 h and then weighed.

Soil microbial diversity was determined based on the Illumina HiSeq sequencing platform of the Majorbio Cloud Platform (http://www.majorbio.com). The bacterial primer was 338F_806R and the fungal primer was ITS1F_ITS2R. Sobs’ and Shannon’s indices were used to indicate the alpha diversity of bacteria and fungi. Coverage index, as defined by Good (Good, 1953) indicates the percentage of operational taxonomic units (OTUs) sampled in a microbial community (i.e., recovered per sample) as a percent of all OTUs found on the site (Chao, 1984; Lemos et al., 2011).

Measurements of soil respiration

Soil respiration (release of CO₂) was measured every 14–16 days by the LI-8100A portable gas exchange system from April 2019 to October 2019 (LiCor, Lincoln NE, USA; chamber 8100-103, diameter of 20 cm). Polyvinyl chloride (PVC) collars were set permanently in place in the soil one week before the first measurement to minimize soil disturbance. The height of each collar was 12 cm and the above-ground height was 3 cm, with the soil surface area and volume within the collar being 317.8 cm² and 953.4 cm³, respectively. Five collars were placed randomly in each plot, giving a total of 45 collars. The above-ground parts of the plants inside the collar were removed before taking each flux reading and the roots were left in place (Afreen & Singh, 2019). Soil fluxes were measured about every 16 days, between 9 am and 1 pm, based on weather conditions. The flux from each collar was measured for 100 s. Soil moisture was measured at a depth of 5 cm with a GS-1 Licor sensor, and the temperature was measured concurrently at a depth of 10 cm using the Licor sensor 6000-09 TC. To avoid pseudo-replication, the five values per plot were averaged for each variable to get a single datum for each measurement timepoint.

Statistical analysis

Statistical analysis was conducted using IBM SPSS (IBM, Chicago, USA). One-way ANOVA was used to process the aboveground and underground biomass of plants under different precipitation treatments. A two-way ANOVA was used to process root biomass, soil nutrient content, soil pH, and the microbial diversity index under different precipitation treatments and different soil depths. Microbial diversity was calculated by Mothur (Version v.1.30). Origin (Origin Lab 2017; Microcal, MA, USA) was used for figures. The different dates formed temporal pseudoreplication, so the significance was explored using nlme (https://svn.r-project.org/R-packages/trunk/nlme) in R (R Core Team, 2013) for soil temperature, soil moisture, and soil CO₂ flux (using the format lme: flux ∼Precipitations * Date, random = ∼1 — plot, weights = varIdent (form = ∼1 — Date)). The varIdent function in package Predictmeans v1.0.2 (https://www.rdocumentation.org/packages/predictmeans) was used to allow each timepoint to have a different variance. Package ggplot2 (https://www.rdocumentation.org/packages/ggplot2) was used for correlations of soil CO₂ flux and all other factors to filter out some variables. For example, if the correlation between Soil SOC and Soil TN was high, we only selected one of these variables, and then performed sPLS analysis (Partial Least Squares regression) (http://mixomics.org/methods/spls/). By contrast, if the correlation between the two was poor, we performed sPLS on both variables. Finally, we considered including soil moisture, RB 0-19.9, soil pH, soil fungi in a sPLS analysis to select variables influencing the predictive model of soil CO₂ flux. Stepwise regression was used to model of the main factors affecting soil respiration.

Precipitation and temperature during the trial

The annual precipitation was 420 mm, 550 mm, and 592 mm in 2017–2019. In 2018 and 2019 precipitation was 30% higher than the average level of precipitation for the last 40 years. Precipitation in 2019 from April to October (the growing season) was 562 mm, which was about 95% of the average annual precipitation (Fig. 2). The average annual air temperature was 7.9 °C in 2019, which was 9% higher than the average temperatures from 1980 to 2019. The highest temperature was 19.7 °C in July, and the lowest was −6.6 °C in January in 2019 (Fig. 2).

Soil properties

Two-way ANOVA results showed that, except for TP, soil properties had significant differences under precipitation treatments and at different soil depths (P < 0.05). Precipitation and soil depth gave the only significant interaction in SOC. The rest of the soil properties had their highest value under R100 in soil 0–30 cm deep (Table 1), with the exception of the soil TP. SOC was 23% greater under R50 as the depth of the soil layer was greater, followed by R100 (18%) and R150 (14%). SOC at R50 increased just 1% in the topsoil (0–9.9 cm) and decreased about 6% in soil depths of 20–30 cm, compared with R100, while R150 had same value with R100 in 20–30 cm soil depths but decreased about 3% in the 0–9.9 cm soil band. As the depth of the soil layer increased, soil pH gradually increased. Soil TN and SOC were highest in the topsoil (0–9.9 cm), and decreased with soil depth.

Biomass of shoot, litter and root

There were no significant differences in shoot, litter, and total biomass (TB= sum of shoots, litter, and RB 0-30) under different precipitation treatments according to one-way ANOVA’s results. Nevertheless, shoot biomass was greatest in R150, and was lower at lower precipitation. Litter biomass was greatest in R50. By contrast, root biomass (RB) 0–30 cm showed significant differences between precipitation treatments according to one-way ANOVA , with R100 having the most root biomass, and R150 and R50 being significantly lower by 52% and 65% when compared with R100, respectively (Fig. 3A). The total biomass under R50 was lower than for treatments R100 and R150. The shoot/root ratio (Aboveground biomass/RB), was largest in R150, and was significantly lower by 64% in R100 (P < 0.05) (Fig. 3B).

Different precipitation treatments and soil depths caused significant differences in root biomasses according to two-way ANOVA. All values were lower at greater soil depths (Fig. 4). The highest value of root biomass appeared at 0–9.9 cm in R100; values in R50 and R150 at 0–9.9 cm were significantly lower by 68% and 57%, respectively. Both increased precipitation and reduced precipitation significantly reduced the root biomass in the topsoil (0–9.9 cm) (Fig. 4).

Microbial richness and diversity

More than 95% of operational taxonomic units (OTUs) found in the microbial community at the study site were present in each soil sample examined. There was no significant difference in soil microbial richness via Sobs’ index and diversity via Shannon’s index under different precipitation treatments and for the precipitation*soil depth interaction. However, there was a significant difference in soil microbial richness and diversity with soil depth, and the value was greatest at 0–9.9 cm (Table 2). The Sobs’ index and Shannon’s index of bacteria and fungi gradually decreased at greater soil depth.

Soil moisture, soil temperature and soil CO₂ flux

Soil moisture, soil temperature and soil CO₂ flux showed significant differences between precipitation treatments over the entire experimental period (April–October) (Table 3). Measurements over the growing season also differed significantly for all variates and were significantly affected by precipitation levels (P < 0.05) (Table 3).

Soil moisture and CO₂ flux were greater at higher levels of precipitation while soil temperature was lower (Table 4). The response of soil moisture to the decreased precipitation treatment (−37.2% of R100) was greater than to increased precipitation (3.9% of R100). Mean soil CO₂ flux under R100 was higher by 38.9% than that for R50, while for R150 the increase relative to R100 was only 8.3% (Table 4).

All variables (soil moisture, soil temperature and soil CO₂ flux) were strongly influenced by seasonality (Table 3). Soil moisture was highest in April and May, then dropped to rise again in October, giving a “W”-shaped relationship with time. Soil temperatures peaked in summer (June-August). The soil CO₂ flux showed an upward trend from April to July, and peaked before decreasing to its lowest levels in October (Table 5).

Soil moisture was typically greatest in the R150 treatment, with the exception of April and September (Fig. 5A). The highest soil temperature values were typically seen in R50 (Table 4), but this situation was reversed in June and July (summer; Fig. 5B), when precipitation was high (Fig. 2). Overall, the soil temperature first rose and then decreased during the growing season.

Soil CO₂ flux showed the same trends as soil moisture for most months. The soil CO₂ flux was lower at the lower end of the precipitation gradient and was highest at R150 and lowest at R50. However, across the summer months, early May and early July, with enhanced precipitation (R150), the soil CO₂ flux was slight lower about 1%–10% , compared to R100 (Fig. 5C). By contrast, precipitation had little effect on soil CO₂ flux in the autumnal month of October.

Modeling soil CO₂ flux

Based on the results of the correlation analysis (Fig. S1), sPLS analysis was performed and showed that the factors most closely correlated with soil respiration were soil moisture, root biomass (RB 0-9.9, RB 10-19.9), soil pH, and fungal diversity (Fig. S2). Stepwise regression showed that soil moisture was the only input factor that significantly affected soil CO₂ flux (P < 0.001, R = 0.94, adjusted R² = 0.870), and all other variables were excluded (Table 6).