The field experiment was conducted in an alpine meadow at the Research Station of Alpine Meadow and Wetland Ecosystems of Lanzhou University in Maqu (N33°40′, E101°52′, altitude 3,550 m), Gansu Province, China, on the eastern Qinghai-Tibet Plateau. Average annual precipitation (1983-2013) is 620 mm, and it occurs mainly during the short, cool summer (Zhang et al., 2014). Mean annual temperature 2°C, and mean monthly temperature varies from −10.7°C in January to 11.7°C in July. The annual cloud-free solar radiation is ∼2580 h (Luo et al., 2006). There is an average of 270 frost days per year. The alpine meadow is mainly dominated by Anemone rivularis (Ranunculaceae), Carex kansuensis (Cyperaceae), Elymus nutans (Poaceae), Festuca ovina (Poaceae), Kobresia graminifolia (Cyperaceae) and Poa poophagorum (Poaceae). All of these six species, except P. poophagorum, were included in our study.

Five 10 m × 6 m blocks with the same topography and slope aspect separated by 2–3 m were established in a N-limited alpine meadow (Shen et al., 2002; Luo et al., 2006) in spring 2015. In each block, we established five 2 m × 4 m plots separated by 2 m and randomly allocated a control (N₀, 0 g N m^–2 year^–1), low (N₁, 5 g N m^–2 year^–1), intermediate (N₂, 10 g N m^–2 year^–1) and high (N₃, 20 g N m^–2 year^–1) plot. Thus, each of the four fertilization treatments included five replicates. The plots were fertilized on a day with light rainfall in early May 2015, 2016, and 2017 with granular slow-release ammonium nitrate (NH₄NO₃) applied by hand. Vegetation sampling was done at the peak of the growing season (July 2017). One 50 cm × 50 cm quadrat was sampled in each plot, and the presence and abundance (number of individuals of each species) of all species within this quadrat were recorded.

Mature seeds of the 63 species found in the vegetation sample (Appendix S1) were collected at their natural dispersal time during July to October 2017. In general, seeds of each species were collected from more than 20 plants and from more than five non-fertilized grasslands located near the experimental site. If there were less than 20 plants of a species, we collected seeds from all plants of the species found. The collected seeds were used to test the three aspects of germination traits most related to plant survival in the field: (a) range of temperatures over which seeds would germinate, which can determine time during the growing season on the Tibet Plateau when seeds can germinate; (b) temperature fluctuation requirement for germination, which can restrict seed germination to vegetation gaps; and (c) cold stratification requirement for seed germination, which can determine whether seeds germinate before or after winter. There were two germination experiments. One tested the first two aspects of germination traits (experiment 1), and the other tested the third aspect of seed germination traits (experiment 2). Previous studies on the emergence of seedlings in the field have shown that most species in grasslands on the eastern Tibet Plateau have a long mean emergence time (MET, the period of time from seed dispersal to seedling emergence), indicating that seeds of most species in these communities need a period of after-ripening for germination (Cao et al., 2018). Therefore, seeds treated by after-ripening instead of fresh seeds were used in germination experiment 1. Thus, the seeds were air dried and cleaned, and stored under warm-dry conditions at room temperatures (ca. 15°C to 20°C), until used in experiments.

For the fertilization experiment, we calculated the response of species abundance to fertilization (R_f). For the germination experiment at the species level, four germination indices were calculated: (1) germination percentage (GP); (2) germination rate (GR); (3) response of germination to alternating temperature (R_5/25); and (4) germination response to wet-cold storage (R_wc) (Table 1).

Three indices (Table 1) were calculated to describe the seed germination niche: BTN, breadth of temperature niche for seed germination; O_j, the extent to which a given species occupies the jth germination temperature condition (j = 5, 10, 15, 20, 25); P_j, proportion of temperature niche for seed germination at the jth temperature condition (j = 5, 10, 15, 20, 25).

To compare seed germination traits at the community level, community-weighted means (CWM) of these indices of germination traits were calculated by weighting these indices with species relative abundance to avoid amplifying the influence of rare species in the community (Niu et al., 2016a, c; Table 1). Germination trait diversity (the range and variation of germination indices in a community) was tested based on seven variables (seed germination percentage at 5, 10, 15, 20 and 25°C and response of seed germination to temperature fluctuation and wet-cold storage) using the ‘dbFD’ function of the R package ‘FD’ (Laliberté and Legendre, 2010). Three typical diversity indices (Mason et al., 2005; Villéger et al., 2008; Niu et al., 2016b) were estimated: 1) FRic, functional richness; 2) FEve, functional evenness; and 3) FDiv, functional divergence.

Beta regression, a model for response variable bounded in (0,1), was used to test the correlation of GP, GR, R_5/25, R_wc, BTN, O_j and P_j with R_f (Ferrari and Cribari-Neto, 2004). Simultaneously, to meet the requirements of Beta regression we removed the boundary values (i.e., −1 and 1) in the data and converted the data with (y + 1)/2 to shift our data interval from (−1, 1) to (0, 1) (Douma and Weedon, 2019). The Beta regression was performed with the ‘betareg’ function of the R package ‘betareg’ (Cribari-Neto and Zeileis, 2010). Generalized linear mixed model (GLMM) was used to evaluate the effects of N addition, temperature and their interaction on community-weighted mean of each germination trait (i.e., CWM_GP, CWM_GR, CWM_R5/25, CWM_Rwc, CWM_BTN, CWM_Oj and CMN_Pj) and germination trait diversity indices (FRic, FEve and FDiv), because of the lack of normality and homogeneity of the data. When GLMM was conducted, block was treated as a random variable. GLMM was conducted with the ‘glmer’ function of the R package ‘lme4’ (Bates et al., 2015) on communities with different N additions at different temperatures. Comparison of data between groups was performed by Tukey test.

To determine if shifts in seed germination traits occurred along the fertilization gradient, we used Principal Components Analysis (PCA), based on a correlation matrix, to identify species groupings or trends on the basis of comprehensive variables. CWM of 23 variables were included in PCA: CWM_GP at 5, 10, 15, 20 and 25°C; CWM_GR at 5, 10, 20 and 25°C; CWM of breadth of temperature niche for germination; CWM of occupation of temperature niche for germination at 5, 10, 15, 20 and 25°C; CWM of proportion of temperature niche for seed germination at 5, 10, 15, 20 and 25°C; CWM of R_5/25; and CWM of R_wc. PCA was conducted using the ‘princomp’ function of the R package ‘stats’ (R Core Team, 2021). All statistical analyses were conducted using R version 4.0.4.

Relative abundance of species along the fertilization gradient is shown in Appendix S2. The results of Beta regression analysis indicated that germination percentages at the five temperatures (5, 10, 15, 20 and 25°C) were significantly and positively related to R_f. Germination rates of seeds at 10, 15, 20 and 25°C were significantly related to R_f (Table 2), whereas at 5°C they were not (Table 2).

GLMM indicated that both temperature (5, 10, 15, 20 and 25°C) and fertilization (N₀, N₁, N₂ and N₃) had a significant effect on community-weighted means for germination percentages and rates (CWM_GP and CWM_GR) (Table 3). There were no interaction effects between temperature and fertilization on CWM_GP or CWM_Gr (Table 3). CWM_GP increased with increased N addition, and CWM_GP was significantly higher in the N₃ treatment than in the N₀ treatment (Figure 1). Only under low temperature (5 and 10°C), CWM_GP was significantly higher in the N₂ treatment than in the N₀ treatment (Figure 1). Similarly, CWM_GR increased along the fertilization gradient (except at 15°C), where there was a significant difference between N₃ and N₀ (Figure 1).

Beta regression analysis showed no significant relationship between R_5/25 and R_f (Table 2). Neither did R_wc have a significant relationship with R_f (Table 2). However, at the community level GLMM showed that fertilization significantly affected CWM_R5/25 (x² = 54.39, P < 0.001) and CWM_Rwc (x² = 15.29, P = 0.002). The strength of the positive response of seed germination to alternating temperature was significantly higher in N₁, N₂, and N₃ treatments than in N₀ treatment (Figure 2A). In contrast, at the community level the strength of the positive response of seed germination to wet-cold storage decreased with increase in amount of N added, and CWM_wc was significantly lower in N₃ treatments than in N₀ treatment (Figure 2B).

Beta regression analysis revealed a significant positive relationship between BTN and R_f (Figure 3A), and it also showed that proportion of temperature niche for seed germination at 10°C (P₁₀) was significantly and positively related to R_f (Table 4). However, there was no significantly positive relationship between R_f and proportion of temperature niche for seed germination at the other temperatures (Table 4). The results of Beta regression analysis indicated that occupation index of seed germination for all temperature conditions (5, 10, 15, 20 and 25°C) was always significantly and positively related to R_f (Table 4). At the community level, the CWM_BTN of the community treated with N₃ was significantly higher than that of the control (Figure 3B). The CWM_BTN of the community with N₁ or N₂ did not differ significantly from N₀ or N₃ (Figure 3B).

Based on the CWM of 23 variables, PCA revealed some clear trends. The first two principal components (PCs) accounted for 77.98% of the variation in the data set, with the first and second PCs explaining 53.02 and 24.96% of the variation, respectively. The two PCs represent the linear combinations of variables that are highly correlated with each other (Figure 4). The three variables most positively related to the first PC are CWM_GP5, CWM_P25 and CWM_BTN; however, the variable most negatively related to the first PC is CWM_RWC. The three variables most positively related to the second PC are CWM_O15, CWM_Gp15 and CWM_P15, while the three variables most negatively related to the second PC are CWM_P5, CWM_R5/25 and CWM_O25 (Figure 4). PCA revealed that shifts of germination strategies occurred along the entire fertilization gradient and especially in the high N addition treatment (N₃) (Figure 4).

Germination trait diversity indices along the N fertilization gradient are shown in Appendix S3. GLMM indicated that N addition had a significant effect on FRic (x² = 19.25, P < 0.001). FRic decreased with increased N, and FRic was significantly lower in N₃ treatments than in N₀ treatment (Figure 5A). However, N addition did not have a significant effect on either FEve or FDiv (P = 0.88 and P = 0.31, respectively) (Figures 5B,C).

Our hypothesis that changes occurring after N addition can change seed germination trait-environmental filter interactions and ultimately redistribute germination traits in the community and alter community composition was supported. At the species level, germination percentage at all temperatures, germination rate at all test temperatures except 5 °C and properties of the temperature niche for germination (BTN, P₅, O_j) were significantly related to R_f. Analyses at the community level also indicated significant differences between CWM_GP, CWM_GR, CWM_R5/25, CWM_Rwc and CWM_BTN of the alpine meadow with high level of N addition and those without N addition. The seed germination trait richness of the alpine meadow also decreased significantly after N addition. These effects of N addition on germination trait selection by the environmental filter is amount-dependent (i.e., low and medium level of N addition had less effect than the high level).

These results indicated that N addition alters filter-trait interactions. The environmental filter in the alpine meadows after N addition tended to select germination traits such as high germination percentage, fast germination rate (speed) and an alternating temperature requirement for germination. However, other germination traits were not selected by the environment filter (N addition), resulting in a decrease of seed germination trait richness in the meadow.

Previous studies indicated that fertilization can cause loss of biodiversity in the grassland community (Niu et al., 2008; Hautier et al., 2009). The decrease in germination traits richness in our study may provide a new explanation for biodiversity loss in grasslands. That is, the environmental filter (N fertilization) does not select for as many germination traits as found in grasslands without fertilization. Thus, grassland species with germination traits not selected by the fertilization filter may be lost from grasslands after fertilization.

Our research provides some evidence that seed germination traits at both the species and community levels play a role in plant survival/persistence in alpine meadows after N addition. Species with high germination percentages and rates are more likely to survive in fertilized alpine meadows than those with low germination percentages and rates. Previous studies indicated that germination traits such as high germination percentage and fast germination speed tend to be selected by a favorable environment (Baskin and Baskin, 2014). The increased plant height and density after N addition can increase soil moisture and reduce exposure to strong sunlight on the soil surface (Bakelaar and Odum, 1978; Ostertag and Verville, 2002; Fenner and Thompson, 2005; Song et al., 2013). In addition, N addition increases soil nutrition (Niu et al., 2008; Li et al., 2013). These habitat changes can promote seedling growth and survival and decrease the possibility of seedling death due to high temperature and drought conditions (Fenner and Thompson, 2005). Thus, favorable germination traits for species to occupy space and acquire resources, such as high germination percentage and fast germination speed, may be selected in fertilized alpine meadows.

Competition for light is a primary determinant of the survival of plant species in the community after N addition (Grime, 1973; Newman, 1973; Tilman, 1988). Since competition for light is plant size- and height-dependent, a newly-germinated seedling in a high-N community is at a great disadvantage with established plants in capturing light. Seed germination traits that can restrict germination to a time and place without competition for light are more likely be selected by the fertilization environmental filter in alpine meadows. Our study showed that the magnitude (slope) of the relationship between seed germination rate and response of species relative abundance to N addition is much greater at low than that at high temperatures. Thus, faster germination speed at low than at high temperature is important for survival of plant species in fertilized alpine meadows. Fast germination in early spring (when the temperature is low on the Tibet Plateau) is conducive to successful seedling establishment before the canopy of established plants becomes dense in the spring. It is a mechanism by which a plant species can avoid seedling death due to competition for light. Furthermore, the magnitude (slope) of the relationship between occupation of temperature niche for germination at low temperature (10°C) and the response of abundance to N fertilization (R_f) is larger than that at high temperature, and a proportion of germination temperature niche can positively relate to the response of abundance to fertilization (R_f), only at low temperature (10°C). These two results also prove that seed germinability at low temperatures is critical for plant species survival in fertilized alpine meadows. Previous studies also found that herbs in dense vegetation usually can germinate in early spring at low temperatures (Baskin and Baskin, 1988; Schütz, 1997; Vandelook et al., 2009).

As mentioned above, competition for light after formation of a dense canopy in spring is not conducive to seedling establishment. However, in Tibetan alpine meadows many vegetation gaps usually are created by burrowing animals such as pika, zokor and marmot and by overgrazing by sheep and yak (Liu et al., 2011). Light is sufficient for seedling establishment in these vegetation gaps. A fluctuating temperature requirement for germination, which restricts germination to vegetation gaps, would promote survival of plant species in fertilized alpine meadows. Our study indicates that at the community level, N addition significantly increased germination response to alternating temperature; seeds of more species in the fertilized than in the non-fertilized meadow required alternating temperatures to germinate. Since addition of N promotes plant growth and canopy closure, we would expect gap-sensing mechanisms for successful seedling establishment to be more important in fertilized than in non-fertilized alpine meadows. As expected, there was a stronger positive response of seed germination to alternating temperatures of species in fertilized than in non-fertilized alpine meadows.

A broad BTN is another germination trait that can promote survival of plant species in fertilized alpine meadows. In our research, Beta regression analysis revealed a significant positive relationship between BTN and R_f. One reason for this positive relationship is that a broad BTN allows seeds to germinate as soon as a vegetation gap is created, and a fluctuating temperature requirement for germination is fulfilled during the growing season, which results in an advantage for seedlings to occupy vegetation gaps, where the light environment is suitable for growth and establishment. Another reason is that seeds of species with a broad temperature niche for germination usually can germinate to high percentages at 5°C and 10°C, which may be beneficial for seedling establishment before the canopy becomes too dense in early spring.