Dongting Lake (28∘30′–30∘20′ N, 111∘40′–113∘10′ E) is the second-largest freshwater lake in China. It is connected to the Yangtze River via tributaries. Dongting Lake wetlands are characterised by large seasonal fluctuations in water level (≤ 15 m) and are completely flooded during June–October and exposed during November–May (Chen et al., 2016). Soil cores (40 cm diameter × 50 cm length) were taken from the wetland. Leaf litter was collected in May 2017 from an undisturbed Carex brevicuspis community at the sampling site (29∘27′2.02′′ N, 112∘47′32.28′′ E) of the Dongting Lake Station for Wetland Ecosystem Research, which is part of the China Ecosystem Research Network. The litter was cleaned with distilled water, oven-dried at 60 ∘C to a constant weight, and cut into pieces 5–10 cm long. Pre-weighed litter samples (5 g; 10.73 ± 0.28 g kg-1 N, 0.89 ± 0.04 g kg-1 P, 40.23 ± 2.6 % organic C, and 17.83 ± 0.25 % lignin) were placed into 10 cm × 15 cm 1 mm mesh nylon bags. This mesh size excluded macroinvertebrates but permitted microbial colonisation and litter fragment leaching (Xie et al., 2016a).

There were three water level treatments (-25, 0, and +25 cm relative to the soil surface) nested by two litter treatments (input vs. removal) and three replicates. The experiment was conducted in nine cement ponds (2 m × 2 m × 1 m) at the Dongting Lake Station for Wetland Ecosystem Research. For the -25 cm treatment, the water level was 25 cm below the soil surface. For the 0 cm treatment, the soil was fully wetted with belowground water (the belowground water was extracted from the well in the experiment site by a water pump) but without surface pooling. For the +25 cm treatment, the water level was 25 cm above the soil surface. Water levels were adjusted weekly using belowground water (total organic carbon: 3.44 mg L-1; total nitrogen: 0.001 mg L-1; total phosphorus: 0.018 mg L-1). Three soil core sets were placed in each pond. One was designated the litter removal control (S), the second was distributed on the soil surface with 15 litter bags to observe the effects of leaf litter input on soil carbon pool (L), and the third was distributed on the soil surface with 15 litter bags to monitor the litter decomposition rate and process (D) (Fig. 1). Litter bags were laid flat on the surface of the soil. Each litter bag was not filled, and there are a little overlap between the litter bags where there is no litter. All the litter bags were fixed to the soil surface with bamboo sticks. The experiment started on 20 August 2017 and lasted 280 d. By that time, no further significant change in litter dry weight was observed. Before incubation, three litter and three soil samples (SOC: 63.32 g kg-1) were collected to determine their initial quality. Litter bags were randomly collected from treatment D after 20, 40, 60, 80, 100, 130, 160, 190, 220, 250, and 280 d. After collection, the litter samples were separated, cleaned with distilled water, and oven-dried at 60 ∘C to a constant weight (±0.01 g). All samples were pulverised and passed through a 0.5 mm mesh screen for litter quality analysis. At the end of incubation, the surface soil (0–5 cm, ∼ 600 g in fresh weight) was collected to eliminate the influences of root decomposition on the soil organic pool. The soil samples were placed in aseptic sealed plastic bags and transported to the laboratory. The samples were sieved (<2 mm), thoroughly mixed, and divided into three subsamples. The first subsample (∼ 150 g) was stored at -20 ∘C and freeze-dried for phospholipid fatty acid (PLFA) analysis. The second one (∼ 150 g) was stored at 4 ∘C for MBC and DOC measurements. The third subsample (∼ 300 g) was air-dried for physicochemical analysis.

Litter organic carbon content was analysed by the H2SO4-K2Cr2O7 heat method. Litter nitrogen was extracted by Kjeldahl digestion and quantified with a flow injection analyser AA3 (SEAL, Germany) (Xie et al., 2017). Litter phosphorus content was quantified by the molybdenum–antimony anti-spectrophotometric method. The lignin content was measured by hydrolysis (72 % H2SO4) (Graça et al., 2005; Xie et al., 2017).

The percentage of litter dry weight losses and the instantaneous decomposition rates were compared among the three water levels by repeated ANOVA analyses. The water level was the main factor, and time was the repeated factor. The intrinsic litter decomposition rate-limiting factor was analysed by the stepwise regression method in a multiple regression model. The surface soil chemical components and the microbial community structure were compared by two-way ANOVA. Treatment (with or without litter input) and water level were the main factors. The percentage differences in litter dry weight loss, the instantaneous decomposition rates, the soil chemical components, and the microbial community structure were evaluated by least significant difference (LSD) at the 0.05 significance level. The data were expressed as means ± standard error. All statistical analyses were performed in SPSS 21 (IBM Corp., Armonk, NY, USA).

The percentage of litter dry weight loss was the highest for the +25 cm water level treatment through the entire litter decomposition period followed by the 0 cm water level treatment. The percentage of litter dry weight loss was the lowest for the -25 cm water level treatment (P<0.01; Fig. 2a). After 280 d decomposition, the percentage litter dry weight loss values under the +25, 0, and -25 cm water level treatments were 61.8 %, 49.8 % and 32.4 %, respectively.

The instantaneous decomposition rate at each measurement time point was calculated based on the Olson negative exponential attenuation model and double exponential decay model. The instantaneous decomposition rate was highest at initial and slowly decreased and stabilised for all three water levels. The maximum decomposition rates for the -25, 0, and +25 cm water levels were 0.00527, 0.00908, and 0.01307 d-1, respectively (Fig. 2b).

During the entire decomposition process, CRRI, NRRI, PRRI, and LRRI significantly increased with the water level. Litter carbon and lignin were always released at all three water levels whilst at -25 cm, nitrogen and phosphorus enrichment appeared in the middle stage (Fig. 3a–d). At the start of the experiment, neither the C/N nor the lignin/N ratio significantly differed at the three water levels. At the middle stage, however, both the C/N and lignin/N ratios were significantly lower at the -25 cm water level than they were at the 0 and -25 cm water levels (Fig. 3e–f).

The multiple regression model of the instantaneous litter decomposition rate and the litter properties showed that at the -25 cm water levels, the main decomposition rate-limiting factor was the lignin concentration whilst at the 0 and +25 cm water level, the main litter decomposition rate-limiting factor was the lignin/N ratio (Table 1).

Under both litter input and litter removal conditions, the bacterial, fungal, and microbial biomass levels were the highest under the 0 cm water level treatment; however, these parameters showed nonsignificant differences between +25 cm above and below water level treatments (P>0.05; Fig. 4a, b, and f). The actinomycete biomass was the highest under the -25 cm water level treatment, followed by that under the 0 cm water level treatment. It was the lowest under the +25 cm water level treatment (Fig. 4c). Litter input significantly stimulated fungal and microbial biomass at all three water levels but only significantly stimulated bacterial and actinomycete biomass at the -25 cm water level (P<0.05; Fig. 4a–c and e). Under litter input conditions, the fungal/bacteria ratio was the highest at the 0 cm water level, followed by the +25 cm water level. It was the lowest under the -25 cm water level treatment. Under litter removal conditions, however, the fungal/bacteria ratio was significantly higher under the -25 cm water level treatment than it was under the 0 cm and +25 cm water level treatments (P<0.05; Fig. 4d).

The SOC, MBC, and DOC concentrations were significantly affected by the water level. The SOC and MBC were the highest at the 0 cm water level and the lowest at the -25 cm water level (P<0.01; Fig. 5a and b). The DOC was the highest at the -25 cm water level and the lowest at the +25 cm water level (P<0.01; Fig. 5c).

Compared with the litter removal group, the SOC concentrations were significantly higher for the litter input group at the +25 and 0 cm water levels. Relative to the litter removal group, the DOC concentrations were significantly higher for the litter input group at the 0 and -25 cm water levels (P<0.001; Fig. 5a and c). The contribution of the litter-C input to the S-SOCP was the highest for the +25 cm water level treatment (16.93 %), intermediate for the 0 cm water level treatment (9.44 %), and the lowest for the -25 cm water level treatment (2.51 %) (P<0.001; Fig. 5d).

The water level significantly influenced C. brevicuspis leaf litter decomposition (P<0.001). The instantaneous decomposition rates (k) were the highest for the +25 cm water level treatment, intermediate for the 0 cm water level treatment, and the lowest for the -25 cm water level treatment (Fig. 2b). Hence, the percentage litter dry weight loss and the decomposition rate increased with the water level, which supported our first hypothesis. The wetland water level strongly affects litter leaching and microbial decomposition (Peltoniemi et al., 2012). Related research showed that the wetland water level strongly affects litter leaching and microbial decomposition (Peltoniemi et al., 2012). Molles et al. (1995) also found that compared with the terrestrial environment, in wetland, water promotes litter leaching and microbial metabolism, thereby accelerating litter decomposition. Moreover, water infiltration into litter also increases relative leaching loss (Molles et al., 1995). Here, the high litter decomposition rate measured for the +25 cm water level treatment may be explained primarily by litter leaching. This finding was consistent with results reported for Carex cinerascens litter decomposition in Poyang Lake (Zhang et al., 2019) and Calamagrostis angustifolia litter decomposition on the Sanjiang Plain (Sun et al., 2012).

The high soil total microbial, bacterial, and fungal biomass levels at the 0 cm water level could account for the rapid litter decomposition observed there. Certain microorganisms are vital to the decomposition process (Yarwood, 2018). Fungi are primary litter decomposers as they fragment dead plant tissues by breaking down lignin and cellulose. Bacteria are secondary decomposers that utilise the simpler compounds generated by fungal activity (de Boer et al., 2005; Bani et al., 2019). Microbial decomposers generally flourish in humid environments. At the 0 cm water level, microbial activity explains most of the litter decomposition. However, at the -25 cm water level, there are comparatively few microbial decomposers, and decomposition is very slow.

The instantaneous decomposition rate was highest at initial and slowly decreased and stabilised for all three water levels (Fig. 2b). Water-soluble components and non-lignin carbohydrates are preferentially and quickly decomposed at the initial of decomposition (Davis et al., 2003). Here, a multiple regression model of the instantaneous litter decomposition rate and litter properties showed that the internal limiting factors affecting the rate of C. brevicuspis leaf litter decomposition varied with the water level. The lignin concentration determined the litter decomposition rate for the -25 cm water level treatment whilst the lignin/N ratio regulated the litter decomposition rate for the 0 and +25 cm water level treatment. This discovery upheld our second hypothesis and was consistent with the findings of Zhang et al. (2019), who reported that wetland ecosystems decomposed Carex cinerascens lignin much earlier and faster than terrestrial ecosystems. Here, we found that the lignin content was the major internal limiting factor of the C. brevicuspis leaf litter decomposition rate at -25 cm water level. At the 0 and +25 cm water level, N is rapidly lost, and the L/N ratio significantly increases. Thus, L/N is the main internal limiting factor at the 0 and +25 cm water levels. A few studies have shown that the lignin content is a key factor limiting terrestrial plant and hygrophyte litter decomposition (Yue et al., 2016; Zhang et al., 2018). Therefore, the amount of carbon that the litter can return to the ecosystem is closely associated with the plant lignin content. The lignin content of C. brevicuspis leaf litters is ∼ 10 % less than that of other wetland plants such as Miscanthus sacchariflorus (∼ 30 %) (Xie et al., 2016a), Spartina alterniflora (∼ 40 %) (Yan et al., 2019), and terrestrial plants such as willow (∼ 25 %), larch (∼ 38 %), and cypress (∼ 28 %) (Yue et al., 2016), so the C. brevicuspis leaf litter is more easily leached and then contributes more to the SOC pool. Furthermore, in Dongting Lake wetland, the Carex genus covers a large area (∼ 23 950 hm2) and generates abundant litter (∼ 36 547 t) (Kang et al., 2009). Thus, C. brevicuspis litter may potentially return large amounts of carbon to the soil.

Litter decomposition is the main pathway by which nutrients are transferred from the plants to the soil. Litter affects the SOC, the stabilisation of which affects other soil properties such as sorption, nutrient availability, pH, and water-holding capacity (Liu et al., 2017). The results of this study showed that litter addition increases SOC in a manner that varies with the water level. The contribution of litter-C input to the S-SOCP was the highest under the +25 cm water level treatment (16.93 %), intermediate under the 0 cm water level treatment (9.44 %), and the lowest under the -25 cm water level treatment (2.51 %). For this reason, flooding conditions are conducive to litter carbon input into the soil. These findings corroborated our third hypothesis. In addition, litter input had a similar effect on soil DOC at the 0 and -25 cm water levels. Therefore, litter decomposition contributes mainly soluble carbon to the soil (Zhou et al., 2015). However, this DOC is also readily lost and decomposed (Sokol and Bradford, 2019; Gomez-Casanovas et al., 2020). This fact accounts for the significantly lower relative DOC under the +25 cm water level treatment here. Wetlands have comparatively larger but also more unstable S-SOCPs than terrestrial environments. In wetlands, water level fluctuations could readily cause carbon loss (Gao et al., 2016; Chen et al., 2018). The SOC differences among three water levels were caused by different soil mineralisation in different environments. Soil mineralisation in aerobic environment (-25 cm) was significantly higher than that in the flooded environment (0, +25 cm) (Qiu et al., 2018), so the SOC at -25 cm water level was lower than the other two water levels. Nevertheless, we considered mainly aboveground litter in this experiment. Hence, the influence of underground litter (root) decomposition on the SOC pool should be investigated in future research (Sokol and Bradford, 2019; Lyu et al., 2019).