2.1 Experimental design
Healthy green leaves were sampled from the trees of K. obovataA. corniculatum , A. marina, and S. alterniflora in the same forest. Thus, all leaves were exposed to similar climatic conditions, which facilitated interspecies comparison of litter mass loss. After being taken to the laboratory, the leaves were gently washed and all dirt particles were then removed by using a soft brush followed by rinsing in distilled water. Five replicates were set up for each litter or litter mixture.
All leaf litters were air-dried to constant weight, 5 g of litter for each single species and 5 g of each litter mixture (2.5 g per species) was placed in each 75 × 75 mm nylon bag with 1 mm2mesh size. There were 40 bags in total, including 15 bags of litter mixture (3 pairs) and 25 for single species (5 single species). All the litter bags were placed on the top of the soil in plastic containers (500 ml), all the containers consisted of 200 g soil and 200 ml water. Based on the previous observation that the decomposition of mangrove leaves generally become slow and steady after 7 weeks (Chanda et al., 2016), this experiment was then conducted for 90 days. The duration of 90 days is often representing a short-term decomposition experiment (Keuskamp et al., 2013).

2.2 Element analysis

After 90 days of decomposition, litter and water were collected to the laboratory for mass and nutrient analyses. The water on the surface of soils was collected with a syringe and then stored in a polyethylene bottle. Litters were determined C and N concentrations with an elemental analyzer (Vario MAX, Vario MACRO, Germany Elementar) after being grounded through 60 mesh sieves. Phosphorus (P) in litters was quantified with ICP-OES (Inductively Coupled Plasma Optical Emission Spectrometer, Optimal 7000DV, PerkinElmer USA) after digested with nitric and perchloric acid. The ammonium–nitrogen (NH4-N), nitrate-nitrogen (NO3-N) and nitrate-nitrogen (NO2-N) were quantified by ultraviolet spectrophotometer (UV6100, mapada instrument, China) following the standard methods of water and waste monitoring and analysis method (Wei et al., 2002). Digestion of the water by alkaline potassium persulfate and potassium persulfate was carried out after filtration of the water sample, and then the total N and total P were determined respectively according to the standard methods (Wei et al., 2002), NH4-N, NO3-N and NO2-N were then analyzed by ultraviolet spectrophotometer (UV6100, ma pada instruments, China). Due to the specific characters of mangrove leaves (Lin et al., 2007), the condensed tannin of the leaves of the four species were also determined before decomposition experiment is conducted. The total condensed tannin was the sum of the extractable, protein-bound and fiber-bound concentrations. The extraction process was followed by the method of Lin et al. (2007).

2.3 Calculation

The litter mass loss was calculated by comparing the initial mass and the mass after decomposition. The initial mass remaining after decomposition was calculated based on the following formula (Wu et al., 2013):
\begin{equation} X_{r}=\frac{X_{i}}{X_{t}}\times 100\nonumber \\ \end{equation}
Where Xr is the percentage (%) of mass remaining after decomposition, Xi is the initial litter mass, Xt is the mass of the remained litter in the litter bags after a given time period (t) of decomposition.
The decomposition rate (k) was calculated by exponential decay model, the litter mass remaining after decomposition and initial litter mass:
\begin{equation} \frac{X_{t}}{X_{i}}=\mathbb{e}^{-kt}\nonumber \\ \end{equation}
Where, k is the decomposition rate coefficient; t is the time duration of decomposition.
The percentage of the initial litter nutrient remaining (Nr) during decomposition was calculated by the following equation (Zhang et al., 2014):
\begin{equation} N_{r}=\left(X_{t}\times{[N}_{t}]+X_{i}\times[X_{i}]\right)\times 100\nonumber \\ \end{equation}
We also calculated the expected rate of the litter mixture after decomposition using the following formula:
\begin{equation} X_{\exp}=\left(\frac{X_{i1}}{X_{i1}+X_{i2}}\times X_{r_{1}}+\frac{X_{i2}}{X_{i1}+X_{i2}}\times X_{r_{2}}\right)\times 100\nonumber \\ \end{equation}
Where Xexp is the percentage of the expected mass remaining (MR) after decomposition, X1 and X2 are the initial dry masses in single species which was 2.5 g, and Xr1 and Xr2 are the mass remaining from the single decomposition species. The effect strength was calculated by the difference between the observed mass remaining (%) and the expected mass remaining (%) (O-E): additive (no significant difference between observed and expected values), synergistic non-additive (negative value, meaning an acceleration of litter decomposition), antagonistic non-additive (positive value, meaning a deceleration of litter decomposition) (Lecerf et al., 2011).
The community-weighted mean value of traits (CWM) was calculated as the mean value of each species in the mixture, because the two litters mixed as 1: 1 of mass in this study (Roscher et al., 2018).

2.4 Statistical analysis

Statistical analysis was performed with IBM SPSS Statistics 23. All data were removed outlier and checked for the normal distribution and homogeneity of variances before the statistical analysis was carried out. One-way ANOVA was used to detect the differences in decomposition rate and nutrients in water among litters, as well as in trait dissimilarity and CWM among each trait. The differences in mass remaining and nutrients in water between observed and expected values were analysed by independent t-test. The variation of the difference in mass remaining between observed and expected values from zero (i.e the observed equals to expected value) was detected by one-sample t-test analysis. The linear correlations of decomposition rate or mass remaining with leaf trait dissimilarity or CWM were assessed through the correlation process. This process was also used to identify the correlation between nutrients in water and in leaves.

3 Result

Non-additive effect on litter decomposition were found in the litter mixtures from three mangrove species. Nutrients in water, the initial trait dissimilarity and CWM of litter mixtures varied with species composition.

3.1 Decomposition rate (k) and mass remaining

Non-additive effect on litter decomposition was detected in the litter mixtures from mangrove species, but not in the mixture of A. corniculatum vs. S. alterniflora (Fig. 1a, b). The observed mass remaining of the mixture of A. corniculatum vs. K obovatawas substantially higher than the expected value, whereas the observed mass remaining of the mixture of A. corniculatum vs. A. marina was lower than the expected one (p < 0.05), only the mixture of A. corniculatum vs. S. alterniflora  did not show significant difference between the observed and the expected values (p > 0.05; Fig. 1b).