Grassland ecosystems account for more than 10% of the global CH₄ sink in soils. A 4-year field experiment found that addition of P alone did not affect CH₄ uptake and experimental addition of N alone significantly suppressed CH₄ uptake, while concurrent N and P additions suppressed CH₄ uptake to a lesser degree. A meta-analysis including 382 data points in global grasslands corroborated these findings. Global extrapolation with an empirical modeling approach estimated that contemporary N addition suppresses CH₄ sink in global grassland by 11.4% and concurrent N and P deposition alleviates this suppression to 5.8%. The P alleviation of N-suppressed CH₄ sink is primarily attributed to substrate competition, defined as the competition between ammonium and CH₄ for the methane monooxygenase enzyme. The N and P impacts on CH₄ uptake indicate that projected increases in N and P depositions might substantially affect CH₄ uptake and alter the global CH₄ cycle.

Field Experiment

The field experiment was conducted in Duolun County (42°02’N, 116°70’E, 1324 m a.s.l.), a semiarid temperate steppe in Inner Mongolia, northern China. The multiple-year mean annual precipitation and mean annual temperature are approximately 385 mm and 2.1°C, respectively. The topography is characterized by low foothills at elevations of 1150-1800 m. Soil is Calcis-orthic Aridisol (the US Soil Taxonomy classification), with 62.75 ± 0.04% sand, 20.30 ± 0.01% silt and 16.95 ± 0.01% clay (Wu et al. 2010). The dominant plant species include Stipa krylovii, Artimesia frigida, Potentilla acaulis, Cleistogenes squarrosa, Allium bidentatum and Agropyron cristattum.

To assess the nitrogen (N) and phosphorus (P) impacts on CH₄ flux, a block design experiment with different combinations of N and P additions was established in early 2013 and run to 2016. A complete random design, with three block replicates, was adopted to address the high spatial heterogeneity. There were four experimental treatments (i.e., control (CK), N addition (N), P addition (P), both N and P additions (N+P)), which were randomly assigned to four 6 m × 6 m plots in each block. Two chambers for greenhouse gas measurement were set up in each plot. All blocks were separated with a 3-m buffer zone. The N and P were applied twice per month from May to September during 2013-2016; we sprayed the fertilizer solution to ensure that the application was evenly distributed in the plots. The dose of N addition was 100 kg N ha^-1 y^-1 as NH₄NO₃ solution; this dose was selected because it meets the N required to sustain the local maximum vegetation productivity (Bai et al. 2010). We monitored the atmospheric N deposition and estimated it to be 20.4 kg N ha^-1 y^-1 in our experimental site in 2012. The dose of P was 100 kg P ha^-1 y^-1, equivalent to previous nutrient addition experiments in grasslands (Phoenix et al. 2003). The plots with both N and P additions received the same amounts of N and P as in the N-only or P-only addition treatments. Control plots were not fertilized, but rather watered with the same amount of water as used in the fertilizer solutions; the water used to dissolve N and P was approximately 800 ml, which is equivalent to 0.8 mm for a 1 m² plot, a pre-treatment field experiment found that this small amount of water addition did not cause any significant changes in water supply, thus unlikely to have altered the ecosystem functions.

Over the study period, soil and plant properties were measured (Supplementary Online Materials). Soil samples were collected to a depth of 10 cm once per year during 2013-2016. Six soil cores were randomly taken in each plot and mixed completely. The measured soil variables included ammonium (NH₄⁺) and nitrate (NO₃^-), soil pH, soil organic carbon (SOC), soil total nitrogen (STN), soil total phosphorus (STP), soil microbial biomass carbon (MBC) and soil microbial biomass nitrogen (MBN). It should be noted that no nitrite was detected at our field site. The measured vegetation variables included aboveground biomass, plant total carbon (PTC) and total nitrogen (PTN).

Field Measurements of CH₄ Flux

A static chamber technique was used to measure the CH₄ flux (Song et al. 2009; Zhang et al. 2017). Stainless steel permanent bases (50 × 50 × 12 cm) with a 3-cm-deep groove for water sealing were inserted into soil down to a depth of 12 cm in the plots approximately one month before the experiment started in the first year. The chamber base was left in the field for one month before any flux measurement, avoiding potential disturbances. An opaque stainless-steel top chamber (50 cm height) with heat-isolation was installed over the base, with bottom in the groove. The grooves were filled with water for sealing the chamber. Two electric fans were installed inside the chamber to mix the air in the headspace continuously and thoroughly during the measurements (Fan et al. 2007). Sixty-ml gas samples were collected at 10-min intervals for 30 min using sixty-ml syringes with airtight stopcocks. Simultaneously, the air temperature and pressure in the chamber were measured, and the soil temperature and moisture were measured at a 5 to 10-cm depth using a thermometer (Spectrum Technologies, Inc. East - Plainfield, Illinois) and a portable soil moisture measuring kit ML2x (ThetaKit, Delta-T Devices, Cambridge, UK), respectively. The CH₄ concentrations of the gas samples were analyzed using a gas chromatograph (Agilent 7890A, Agilent Technologies Limited Co., USA) equipped with flame ionization detector (FID) using 60–80 mesh 13 XMS column (2 mm inner diameter and 2 m long), with an oven temperature of 55°C. Pure nitrogen was used as carrier gas at a flow speed of 30 ml min^-1, and the CH₄ flux was determined from the linear slope of the mixing ratio changes in four samples taken at 0, 10, 20 and 30 min after chamber closure. The CH₄ flux was measured weekly from May to September in 2013, 2014, 2015 and 2016, respectively.

The CH₄ fluxes were linearly interpolated and sequentially cumulated to estimate the total flux over growing seasons (Zhang et al. 2017). The linear interpolation was adopted due to two reasons: (1) it has been widely used in previous studies and has been suggested to be effective for seasonal interpolation of ecological variables (Song et al. 2009; Nikiema et al. 2011); (2) the auxiliary data to assist annually interpolation was lacking. Throughout this analysis, positive fluxes represent gas uptake by the grassland ecosystems.

Meta-analysis

A meta-analysis was conducted to investigate the N and P impacts on CH₄ across global grasslands. We collected publications by searching for “nitrogen and phosphorus”, “methane”, “grassland”, and “upland” in Google Scholar in July 2016, and later updated in September 2018. The search returned 6230 publications. A few criteria were then used to screen these publications for our purpose. The criteria applied to determine whether or not to use the data were: (1) it must be manipulation experiments with either external N and/or P addition; (2) the field measurements must cover at least one full growing season, which makes the estimation of annual budget more reliable; (3) the studies report clear information for the field site that is useful when extracting edaphic and meteorological data from global datasets. For sites without clear latitude and longitude, we Googled and found their geographic coordinates with country and site names. Finally, 35 papers were selected for data extraction. When the data were presented in figures, we extracted mean values and standard errors using GraphClick (http://www.arizona-software.ch/graphclick/). For studies with measurements from different N and P addition levels in one paper, they were extracted and treated as independent data points.

Finally, we compiled a comprehensive dataset of 382 data points for meta-analysis (Figs. S4 and S5). The dataset covers the major grassland types across the globe and all experiments were carried out between 1980-2017. For all field experiments, the atmospheric depositions of N and P from the global dataset were treated as background rate and were added to the reported N and/or P addition rates. Since a majority of the data points from Yu et al. did not contain N or P deposition (Yu et al. 2017), we used the extracted contemporary nutrient deposition (Mahowald et al. 2008), based on latitudes and longitudes, as nutrient inputs. Every CH₄ flux rate corresponds to one site with auxiliary information including latitude, longitude, and factors such as N deposition, P deposition, soil temperature (ST), soil moisture (SM), soil pH, soil organic carbon (SOC), bulk density (BD), and clay content (CL). Across the studies in our data set, the N and P treatments fell within the ranges of 0~200 kg N ha^-1 yr^-1 and 0~200 kg P ha^-1 yr^-1 (Mori et al. 2013; Zheng et al. 2016), respectively. The non-growing season CH₄ flux is normally not available for most field experiments, therefore, the annual rate of CH₄ uptake (kg C-CH₄ ha^-1 yr^-1) thereafter was calculated by the ratio of the growing season to non-growing season CH₄ uptake as reported by a few studies (Li et al. 2012; Yue et al. 2016).

The soil properties for each point data site were extracted from global datasets according to their latitudes and longitudes. Specifically, the soil pH, SOC, BD and clay were retrieved from the Re-gridded Harmonized World Soil Database v1.2 in the Oak Ridge National Laboratory Distributed Active Archive Center for Biogeochemical Dynamics (available online: https://daac.ornl.gov/SOILS/guides/HWSD.html); soil temperature and moisture for the top 10 cm were from the NCEP/DOE AMIP-II Reanalysis (Reanalysis-2): http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.derived.surfaceflux.html).