The experiment was established in a semi-open greenhouse on the campus of Wenzhou University (28°01′N, 120°65e;E, Wenzhou City, Zhejiang Province, China). A greenhouse with a transparent plastic roof and no walls was used; this kept off rainwater while ensuring that light intensity, air temperature, and humidity inside the greenhouse were similar to those of outside. According to meteorological data from the Zhejiang Provincial Bureau of Statistics, the average temperature in Wenzhou in October was 21.1°C, and the average minimum temperature was 17°C. The average temperature in November was 18.3°C, and the average minimum temperature was 14°C. The average temperature in December was 11°C, the average minimum temperature was 7°C. The microcosms were constructed using PVC (51 cm length × 38 cm width × 29 cm height) and filled with sand to a depth of 25 cm in October 2020 (Fig. 1).

We selected two common local aquatic plants that could grow in the cold season: Acorus calamus L. and Reineckea carnea (Andrews) Kunth. The experiment was a two-factor study of plant species diversity and density. Plant species diversity includes two species richness levels (1 or 2) and three plant community compositions: two monocultures and one two-species mixture. Each composition was tested at three densities: 4, 8, and 12 individuals per microcosm (Fig. 1). The experiment used a randomized complete block design with three replicates. In total, 27 microcosms were constructed, and seedlings were transplanted into the microcosms in October 2020.

Wastewater was simulated by the Hoagland nutrient solution [33] with a minor modification (Table 1). Due to the high nitrogen concentration of the simulated catering wastewater, the experimental total inorganic nitrogen (TIN) concentration was 336 mg L⁻¹ [20, 25]. Before each irrigation event, all microcosms were emptied of water. Then, each microcosm was supplied with 6 L of simulated wastewater every seven days for eleven weeks, simulating the batch watering of CWs [34].

Water samples were collected weekly over the 11-week experiment. The microcosms were drained at each sampling. Effluent samples (250 ml) were collected weekly from each microcosm and stored at −20°C. Before analysis, the water samples were thawed at 4°C and filtered using a membrane syringe filter (pore size 0.45 μm). The concentrations of ammonium nitrogen (NH₄⁺-N) and nitrate nitrogen (NO₃⁻-N) were determined by Nessler’s reagent spectrophotometry and ultraviolet spectrophotometry, respectively. Total inorganic nitrogen (TIN) concentration was calculated as the sum of these concentrations.

All plants were harvested by species after the final effluent collection. The aboveground and belowground biomass were measured after the plant samples were dried at 65°C for 72 hours and summed to determine the total biomass. Then, the aboveground and belowground plant samples were dried and ground into powder, and the N concentration was measured by isotope mass spectrometry (Flash HT2000, Thermo Finnigan, Bremen, Germany). The plant N pool was calculated as the plant biomass of each species multiplied by the plant N concentration.

The N removal efficiency (RE) in the CWs system was calculated as follows:

where C_i is the influent N concentration of each microcosm (mg L⁻¹) and C_e is the effluent N concentration of each microcosm (mg L⁻¹).

The contribution efficiency of plant uptake to N removal (CE) was calculated by:

where N_p is the N uptake by plants (mg), N_i is the influent microcosm N (mg), and N_e is the effluent microcosm N (mg).

Two-way ANOVA was used to test the effects of plant species diversity (species richness and composition) and plant density on N removal efficiency, plant biomass, and plant N pool. One-way ANOVA was used to detect differences in various parameters between plant species combinations and planting densities. If there were significant differences, multiple additional comparisons were carried out (Turkey method). Before statistical analysis, the data were analyzed for normality (Kolmogorov–Smirnov test) and equality of variance (Levene’s test). All statistical analyses were conducted using software R 4.1.1 with a statistical significance level 0.05.

Plant density is one of the key factors affecting ecosystem functioning and plays an important role in plant productivity [28, 29]. In this study, plant density had a significant effect on plant aboveground, belowground, and total biomass (Table 2). The aboveground biomass of R. carnea in monoculture and mixed cultures were significantly higher under high-density treatment than those of low densities (Fig. 2). Both the belowground biomass and total biomass of A. calamus and R. carnea monocultures under high-density treatment were significantly higher than those of low density (Fig. 2). This may be because the increase of plant density can increase plant biomass. However, exceedingly high density may increase competition among species and decrease the biomass of individual plants [35], which may explain the lack of significant differences between biomass at medium density and high density.

Over the past few decades, the number of studies on the relationship between plant species diversity and ecosystem function has increased. Most of these studies have shown that high plant species diversity improves plant productivity in CWs [36]. However, in this study, plant species richness had no significant effect on plant aboveground, belowground, and total biomass across plant density treatments (Table 2). Different plant species prefer different forms of N, resulting in a positive effect of plant species diversity on productivity [37,38]. Existing studies were mainly carried out in high ammonium or high nitrate systems, but this study was carried out under a NH₄⁺/NO₃⁻-N ratio of 1:1. Previous studies also showed no significant difference in plant biomass and species richness under the mixed NH₄⁺/NO₃⁻-N treatments [21]. Beyond this possible explanation, temperature can affect the photosynthetic and metabolic activities of plants [39]. Under low-temperature stress, the enzymatic activity of plants decreases significantly, resulting in slow uptake of external substances [40]. The data from the Plant Science Data Center indicates that Acorus calamus prefers a warm and humid environment, with the most optimum temperature for growth being 20–25°C and stopping growing below 10°C; Reineckea carnea has good cold resistance, with the most optimum temperature for growth being 15–28°C. Since this experiment was conducted in the cold season, plants were inhibited by low temperature, which may have prevented plant species richness from affecting plant biomass. In addition, this study only sets two richness levels, which limited the effect of species richness to a certain extent.

As an important part of plant species diversity, species composition affects plant biomass [20–23]. Our results showed that plant species composition significantly affected plant belowground and total biomasses (Table 2). Under medium and high-density conditions, the plant belowground and total biomasses of A. calamus monoculture were significantly higher than those of R. carnea monoculture (Fig. 2). Studies have shown that the growth of A. calamus depends on the application dose of N [41]. Under field conditions such as nutrient-rich wetland habitats, A. calamus grows vigorously [42]. In this experiment, simulated wastewater with high N concentration met the needs of A. calamus and permitted a greater biomass accumulation. These results indicate that the effect of plant species diversity on productivity was limited in the cold season, but appropriately increasing plant density can promote growth, and N can be removed by harvesting specific plants, such as A. calamus.

As a key factor of ecosystem functioning, temperature plays an important role in plant growth and microbial activity in CW systems [43]. Studies have shown that along with increased plant growth, the average N removal efficiency of CWs in the warm season is 6–11% higher than in the cold season [9–11]. In this study, the average removal efficiencies of NO₃⁻-N, NH₄⁺-N and TIN in the CWs in the cold season were 83.02%, 79.27%, and 80.21%, respectively. These values are equivalent to the N removal efficiency of the CWs in the warm season [44,45], indicating that high N efficiency can be maintained by selecting appropriate cold-tolerant plants.

Previous studies have shown that plant species diversity has a positive effect on N removal efficiency in CWs, and these effects can be enhanced by increased plant density [22, 24, 25]. This study focused on the effects of plant species diversity and density on N removal of CWs. We found that plant density has a significant impact on the removal efficiencies of NO₃⁻-N and TIN from CWs (Table 3). NO₃⁻-N removal efficiency of R. carnea monoculture under high and medium density is significantly higher than at a low density (Fig. 3a). TIN removal efficiency of R. carnea monoculture at medium density was significantly higher than at a low density (Fig. 3c). Plant uptake is a main form of N removal in CWs [46–48], so the improved N removal at medium and high densities may be due to its increased uptake by R. carnea monoculture. Plant biomass was negatively correlated with effluent N concentration in CWs (Fig. 5). At medium and high densities, the belowground N pools of R. carnea monoculture were significantly higher than at a low density (Fig. 4b). N is additionally removed through nitrification and denitrification [49]. Higher plant density can increase root length density and provide more microbial attachment sites, which improves the N removal efficiency of CWs [29].

N removal often increases in systems with higher plant species richness [50]. Many studies have shown that species composition has a significant effect on N removal efficiency in CWs [21,24,25]. However, in this experiment, plant species richness and composition had no significant effect on N removal efficiency (Table 3). A common form of N removal is plant uptake. Plant N uptake is related to biomass and N concentration [51]. In this study, plant species diversity did not affect plant aboveground biomass or N pool (Fig. 2, Fig. 4a). In this study, the average contribution efficiency of plant uptake was 5.39%. This is far below 100%, indicating that other N removal pathways played a large role; this likely included microbial nitrification and denitrification. There was also no difference in plant N removal between the two species at medium and high density (Fig. 4d). Some studies have shown that nitrification and denitrification intensity differ by species [52]. Future research into nitrification and denitrification intensities is required.

N removal is the main purpose of CWs [44]. This experiment illustrated that plant density significantly affected the N removal efficiency of CWs in the cold season; this cold season N removal efficiency could be improved by increasing plant density. There is no significant N removal efficiency difference between A. calamus and R. carnea; both are excellent species for N removal in the cold season. Plant biomass is an important index of plant growth, and biomass accumulation also accumulates N in plants. After harvesting, plants can be used as a sustainable form of bio-energy [53]. In this experiment, the biomass and N pool of A. calamus were significantly higher than that of R. carnea. Therefore, A. calamus is a better choice for N removal in CWs. In the future we can improve the N removal efficiency of CWs in the cold season by increasing plant density and sowing specific plants, such as A. calamus.