Plant litter is known to affect soil, community, and ecosystem properties. However, we know little about the capacity of litter to modulate grassland responses to climate change. Using a 7-year litter removal experiment in a semiarid grassland, here we examined how litter removal interacts with consecutive droughts to affect soil environments, plant community composition, and ecosystem function. Litter loss exacerbates the negative impacts of drought on grasslands. Litter removal increased soil temperature but reduced soil moisture and nitrogen mineralization, which substantially increased the negative impacts of drought on primary productivity and the abundance of perennial rhizomatous graminoids. Moreover, complete litter removal shifted plant community composition from grass-dominated to forb-dominated, and reduced species and functional group asynchrony, resulting in lower ecosystem temporal stability. Our results suggest that ecological processes that lead to reduction in litter, such as burning, grazing, and haying, may render ecosystems more vulnerable and impair the capacity of grasslands to withstand drought events.

Study site and experimental treatments

The field experiment was conducted in a semiarid steppe at the Inner Mongolia Grassland Ecosystem Research Station (116°42'E, 43°38'N, and 1,250 m a.s.l.) of the Chinese Academy of Sciences, located in Inner Mongolia, northern China. This area has a semiarid temperate continental climate with mean annual precipitation (1982-2021) of 333.1 mm, of which 60–80% falls from May to August. Mean winter precipitation that falls as snow is 27.8 mm (November to April), accounting for 8.3% of the total annual precipitation (1982–2021). The mean annual temperature is 1.0°C, with mean monthly temperatures ranging from -21.2°C in January to 19.7°C in July. The soil at the study site is dark chestnut with loamy sand texture. The plant community in the study area is dominated by perennial herbaceous plants, including Leymus chinensis, Stipa grandis, Agropyron cristatum, Cleistogenes squarrosa, and Carex korshinskyi (Bai et al., 2010).

The plant litter removal experiment was conducted in a grassland enclosure that had been protected from grazing since 1999. A randomized complete block experiment was established in October 2014, including three plant litter removal treatments and six replicates for each treatment. Three 3 × 3 m plots were randomly laid out in each of six blocks (18 total plots) and then assigned to the 3 treatments: partial litter removal (-50%), complete litter removal (-100%), and no litter removal (control). Before the litter removal, there were no differences in plant community and soil properties among treatments. The litter removal treatments were applied annually at the end of October after plant natural senescence from 2014 to 2021. In each complete litter removal plot, all standing litter was removed by cutting with scissors and fallen plant litter was cleared by raking. In each partial litter removal plot, half of the standing and fallen litter were returned to the community in the first year and only the standing litter was manipulated in the following years.

Soil moisture, temperature, and nitrogen mineralization

Soil temperature was continuously measured every 60 min at a depth of 10 cm from November 2014 to September 2021.  Measurements were recorded automatically with a Datalogger (iButton DS1922L Wdsen Electronic Technology Co., Ltd).  Based on the daily average temperature data, we calculate the mean soil temperature in the growing season (ST) from May to September and in the non-growing season (NGST) from October to April of the following year.  Soil moisture was measured 3 times per month within the depth interval of 0–12 cm in each plot from June to August, using manual Time Domain Reflectometry (TDR) equipment (Soil Moisture Equipment Corp., Santa Barbara, CA, USA).

Soil net N mineralization rate was measured using the in situ soil core incubation method from mid-July to mid-August each year from 2015 to 2021, corresponding to the peak growing season.  For each plot, 2 sharp-edged PVC tubes (5 cm in diameter and 12 cm in length) were inserted into the soil at 10 cm depth to extract soil cores.  One core was sealed by a parafilm membrane to prevent water penetration while allowing gas exchange and was then incubated in the field for about 28 days.  The other core was taken back to the laboratory and stored in a refrigerator at 4°C for measuring the initial inorganic N contents.  Soil inorganic N (NH₄⁺-N and NO₃^--N) was extracted with a 2 mol/L KCl solution and subsequently measured using a continuous flow injection analyzer (AA3 HR, SEAL Analytical GmbH).

After harvesting the plant biomass, three soil cores (3 cm diameter, 0–10 cm depth) were randomly collected and mixed into one sample in each plot in mid-August 2021.  Soil total nitrogen (STN) was measured after soil samples were air-dried, ground, and passed through a 2 mm sieve.  STN was determined following Kjeldehl digestion by a Nitrogen Analyzer System (KJELTEC 2300 AUTO SYSTEM II, Foss Tecator AB, Sweden).  Roots and stones were removed prior to all soil sample processing.

Plant species composition and biomass

A permanent 1 × 1 m quadrat was randomly established in each of the 18 plots. The species abundance was determined by 1 × 1 m quadrat by individual counts, and the community abundance was quantified by the total number of plants in each quadrat. Species richness was measured by the number of species in mid-August at the peak of aboveground biomass of the plant community from 2015 to 2021. We randomly selected five individuals for each species present in the plot to measure plant height. If there were fewer than five individuals, all of them were measured. We then calculated the weighted average of plant height at the functional group level. Plant species were classified into 4 functional groups based on their life forms, including perennial bunchgrasses (PB), perennial rhizomatous graminoids (PR), perennial forbs (PF), and annual forbs (AF). The 4 groups consisted of 7, 2, 26, and 8 species, respectively. To estimate the biomass of the entire community and individual species, one 0.25 m × 1 m rectangular quadrat was randomly chosen within each plot in mid-August from 2015 to 2021, away from the 1 × 1m quadrats. All plants in each rectangular quadrats were clipped and sorted by species, and then weighed for biomass after drying at 65°C to a consistent weight for at least 48 hours. Aboveground primary productivity (ANPP) was estimated using the peak aboveground biomass of the plant community.