Analysing the spatial variation of soil respiration during the early growing season of different grasslands in China

Sampling sites

Site-level R_s during the early growing season of three types of grasslands (i.e., alpine meadow, desert steppe and typical steppe) in China were obtained in this study (Fig. 1A). The sampling sites of desert steppe and typical steppe were located in the Xilingol League, Inner Mongolia, which is in the central part of the Inner Mongolia Autonomous Region with latitudes 41°35′N–46°46′N and longitudes 111°09′E–119°58′E. This region has a temperate continental climate with cold winters and hot summers with an annual mean temperature of 1–4 °C and an annual mean precipitation of 150–400 mm. Grassland types primarily include desert steppe, typical steppe, and meadow steppe.

The alpine meadow site (37°36′N, 101°20′E) was located near the National Field Scientific Observatory for Alpine Grassland Ecosystems in Haibei, Qinghai (Fig. 1A). This site has an annual mean temperature of −1.9 °C and a mean yearly precipitation of 618 mm. In addition, this site has perennial snow and seasonal permafrost distribution. The major vegetation types are alpine meadow, alpine scrub and swampy meadow.

Measurement of R_s

A portable automatic soil carbon flux measurement system (Li-8100, Li-Cor Biosciences, Nebraska, USA) was used for measuring R_s. (Fig. 1B). In total, 18 sites were measured from May 3 to May 6, 2021, in the Xilingol League, Inner Mongolia. Eight were selected from the desert steppe, and ten were from the typical steppe. Three PVC rings with an inner diameter of 20 cm were placed for each plot. These rings were buried in the soil to a depth of three cm, keeping the aboveground height of each ring basically the same. Before each measurement, green plants in the rings were cut off to eliminate the impact of plant autotrophic respiration. The rings were placed 2 h before R_s measurement to minimize the effect of soil disturbance caused by the placement of the rings. Each measurement was made between 9:30 a.m. and 11:00 a.m. local time to ensure that plot-level measurements were spatially comparable.

At the alpine meadow site, a six-channel soil respiration measurement system (LI-8150, Li-Cor) was used to automatically and continuously measure R_s every hour (Fig. 1C). The R_s of six chambers were averaged to represent the R_sfor the site of alpine meadow site. Because there is only one site for the alpine meadow, the daily averaged R_s from the same period was selected as that of the R_s experiment in the Xilingol League (May 1–May 10, 2021) for comparative study.

Environmental factors and soil properties from field measurements

R_s was measured simultaneously at each site with environmental factors (i.e., ST and SWC) and soil properties (i.e., SOC, TN and total carbon (TC)). The ST was measured with the temperature probe connected with the LI-8100. SWC was measured using a soil moisture meter time-domain reflectometer. The soil samples were collected within the three rings when R_s measurements were completed for desert and typical steppes. However, the soil samples were collected outside the six rings for alpine meadow. Three 0–10 cm soil samples were collected. Then they were composited into one sample for soil property measurements (i.e., SOC, TN and TC). Specifically, SOC was measured using the potassium dichromate-endothermic method, and TN and TC were measured using the elemental analyser temperature combustion method.

Data from other sources

Because long-term biotic conditions may also affect the spatiotemporal variation of grassland R_s, multiyear (2000–2020) averaged normalised difference vegetation index (MNDVI) (Myneni et al., 1995) and land surface water index (MLSWI) (Jurgens, 1997) were calculated for all sampling sites. These calculations used remotely sensed surface reflectance data from Landsat with a spatial resolution of 30 m and a temporal resolution of 16 days based on an online geospatial data analysis cloud platform (Google Earth Engine, GEE) provided by Google.

Multiyear (2000–2020) averaged climate factors (i.e., mean air temperature and total precipitation) from ERA5 (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land-monthly-means) and the green vegetation index (i.e., normalised difference vegetation index (NDVI)) from Landsat were first calculated for each month in all sampling sites to characterize long-term monthly variations in hydrothermal and vegetation cover conditions of the three types of grasslands. Next, their monthly averages were calculated for all sampling sites in each type of grassland.

Remote sensed land surface temperature (LST) from Landsat and the soil moisture volume fraction from a Soil Moisture Active Passive L-Band radiometer on GEE were used for the representative analysis of sampling times and sites. Temperature-related factors were only considered to evaluate the representative sampling time. Then, the mean LST of all sampling sites was compared during the sampling time with that during the entire early growing season from 2000 to 2020. In contrast, temperature- and moisture-related factors were used to analyse the spatial representative of sampling sites by comparing LST and soil moisture in all sampling sites with that around the sample sites in each type of grassland.

Statistical analyses

The statistical analyses of R_s and related impact factors in three types of grasslands were performed using SPSS25.0 software (IBM, Armonk, NY, USA). One-way analysis of variance (ANOVA) was used to analyse the differences in R_s and impact factors among different grasslands. Because the alpine meadow contained only one sample site, the ANOVA analysis was conducted only for desert and typical steppes. To analyse the SV of R_s and related impact factors, their coefficients of variation (CV) were determined for the desert and typical steppes. Pearson’s correlation coefficients were also calculated to analyse the relationships between R_s and related impact factors. In addition, a regression model was constructed using stepwise regression to determine the crucial impact factors affecting the SV of R_s in desert and typical steppes. Origin2022 (OriginLab, Northampton, MA, USA) was used for graphing.

Representative analysis of sampling times and sites

According to comparative analyses, the mean LST of all sampling sties in each type of grasslands during the sampling time was close to that during the entire early growing season from 2000 to 2020 (Fig. 2). The LST and soil moisture volume fraction of the sampling sites had near uniform distribution within these surrounding sites (Fig. 3) for three types of grasslands. The mean LST and soil moisture volume fraction of the sampling sites were close to their corresponding value from the surrounding sites for desert and typical steppes. These findings indicated that the sampling times and sites for each type of grasslands represented the entire early growing season and a large spatial region, respectively.

Analysis of R_s and impact factors in different grasslands

Based on ANOVA results, significant (P < 0.05) differences were observed in R_s during the early growing season among the three types of grasslands (Fig. 4A). The alpine meadow had the highest R_s of 2.44 µmol m⁻² s⁻¹, followed by the typical steppe with mean R_s of 1.30 µmol m⁻² s⁻¹ and the desert steppe with mean R_s of 0.52 µmol m⁻² s⁻¹ (Fig. 4A). A minor difference in ST (P > 0.05 (i.e., not statistical significant), Fig. 4B) was observed among the three types of grasslands. In contrast, a significant difference was presented in SWC (P < 0.05, Fig. 4C). The SWC of the alpine meadow (40.32%) was higher than that of the desert and typical steppes. The mean values of SWC of the desert and typical steppes were 6.02% and 11.11%, respectively. Among the three types of grasslands, significant differences (P < 0.05) were observed in soil properties (i.e., SOC, TC, TN and soil carbon-to-nitrogen ration (C:N)) (Figs. 4D–4G) and MNDVI (Fig. 4H), but no difference was found for MLSWI (Fig. 4I).

SV of R_s during the early growing season of desert and typical steppes

R_s varied substantially during the early growing season across all sampling sites in desert or typical steppes based on the CV analysis of R_s (Fig. 5). Moreover, the CV of R_s during the early growing season of the typical steppe was larger than that of the desert steppe. Among all impact factors, the desert steppe demonstrated the largest CV of TC and the smallest CV of ST, whereas the typical steppe showed the largest CV of MLSWI and the smallest CV of C:N (Fig. 5). Desert steppe had a large CV in SOC and C:N, and the typical steppe exhibited large SV in other impact factors.

Among all impact factors, R_s during the early growing season of desert steppe revealed the highest positive correlation with MNDVI and the highest negative correlation with SWC (Fig. 6A). R_s during the early growing season of the typical steppe only depicted a strong positive correlation with SWC (Fig. 6B). Stepwise regression results (Table 1) revealed that SWC was an essential predictor of SV of R_s during the early growing season of the typical steppe (R² = 0.88, P < 0.05). However, ST was important in explaining the variables for SV of R_s during the early growing season of the desert steppe.

Differences between R_s and related impact factors during the early growing season in different grasslands

R_s during the early growing season varied significantly among the three types of grasslands. This result might be attributed to the complicated interactions of various impact factors, such as vegetation type, soil properties and site-specific climatic conditions. Desert and typical steppes had lower R_s than alpine meadows, which were consistent with previous studies (Feng et al., 2018; Wang et al., 2020). The long-term lower air temperature, higher precipitation and growing-season NDVI at the alpine meadow (Fig. 7) caused more carbon fixed by photosynthesis to be retained in the soil and thus led to the accumulation of SOC (Deng et al., 2023; Jia et al., 2019). The abundant SOC provided a sufficient source substrate for R_s. Furthermore, alpine meadows exhibited the highest vegetation cover during the early growing season, which may contribute to the highest R_s during the early growing season of the three types of grasslands (Soong et al., 2020).

As the transition zone between the typical steppe and the desert, desert steppe generally has higher air temperature, lower precipitation and vegetation cover (i.e., NDVI) (Fig. 7). It was the driest among the three types of grasslands based on the long-term seasonal variation of monthly total precipitation (Fig. 7). Its dominant vegetation are shrubs and herbs, with sparse vegetation cover and low aboveground biomass (Angerer et al., 2008; Gao et al., 2022). Furthermore, long-term human grazing activities have led to severe soil sanding and low SOC in the desert steppe (Song et al., 2017; Song et al., 2021a). During the early growing season, the low SOC may restrict the diversity and activity of soil microorganisms and thus generate low R_s in the desert steppe (Lee et al., 2021; Zhou et al., 2013). The annual precipitation and growing-season NDVI of the typical steppe were higher than those of the desert steppe (Fig. 7). This activity may cause aboveground biomass and vegetation cover of typical steppe two times higher than those of the desert steppe during the same period (Jägermeyr et al., 2014; Tang et al., 2013), which may contribute to a richer SOC. Moreover, a higher R_s was exhibited in typical steppe than in desert steppe (Fig. 4).

Factors affecting the SV of R_s during the early growing season of the desert and typical steppes

During the early growing season of the desert steppe, ST was found to be the primary factor affecting the SV of R_s based on step regression analysis (Table 1). This result was inconsistent with recent studies demonstrating that soil moisture considerably affected the R_s of desert steppe (Gao et al., 2022; Suseela et al., 2012; Zhang et al., 2021). This inconsistency might be attributed to the correlations among various variables in the desert steppe (Fig. 6A). MNDVI and MLSWI had a strong positive correlation with R_s, and ST showed a strong positive correlation with both vegetation indices. Previous studies had clearly confirmed a significant positive correlation between NDVI and aboveground biomass of grasslands (Spehn et al., 2000; Eisfelder, Kuenzer & Dech, 2012). In addition, SOC of grasslands primarily comes from the carbon fixed by vegetation photosynthesis. Therefore, higher MNDVI indicates that the desert steppe had relatively higher SOC promoting R_s. The negative correlation between MNDVI and SWC could partly explain the negative correlation between SWC and R_s during the early growing season of the desert steppe.

In contrast, SWC was an essential factor influencing the SV of R_s in the typical steppe. The typical steppe in this study was distributed in the arid and semi-arid regions of China, where soil moisture is a crucial limiting factor of various ecosystem processes (Jia, Zhou & Yuan, 2007; Li et al., 2021). Shi et al. (2020b) discovered that soil water characteristics limited the soil CO₂ release rate during the dry season. Increasing soil water availability, which promotes microbial community and fine root activity (Berry & Kulmatiski, 2017), accelerates substrate diffusion and soil organic matter decomposition (Suseela et al., 2012), thus increasing soil CO₂ emission. Therefore, the significant effect of SWC on the SV in R_s during the early growing season of the typical steppe is entirely expected. However, the same phenomenon did not occur in the desert steppe, probably because the factors affecting R_s changes during the early growing season were not independent. For example, the MNDVI and ST correlated with the SWC (Fig. 6A), which partly weakened the effect of SWC on R_s in the desert steppe.

Based on field observations of R_s in 42 sample plots in the alpine meadow on the Tibetan Plateau, a previous study discovered that the SV in R_s during the peaking growing season of alpine meadow was affected by the belowground root biomass rather than the SWC (Geng et al., 2012; Huang, He & Niu, 2013). The inconsistency between this finding and our study might be due to the difference in the study time (mid-growing season vs. early growing season). This temporal difference would have led to changes in the proportion of the two major components of R_s in grassland (i.e., soil autotrophic respiration and heterotrophic respiration) and their responses to impact factors (Li et al., 2010; Nissan et al., 2023; Shi et al., 2022). The surface vegetation is sparse during the early growing season compared to the mid-growing season. Therefore, R_s was dominated by soil heterotrophic respiration during the early growing season than soil autotrophic respiration (Li et al., 2018; Yang et al., 2020).