The research site was located at the Haibei Alpine Meadow Ecosystem Research Station (37°29′–37°45′N, 101°12′–101°23′E; 3,200–3,600 m a.s.l.), Menyuan County, Qinghai Province, China ( Figure 1 ). The station is characterized by a typical continental plateau climate, with cool, short summers and long, cold winters, a mean temperature of −1.7°C, and a mean annual precipitation of 580 mm (Chen et al., 2022). The local soil is gelic cambisol, which has relatively homogeneous physiochemical properties. The main soil body type is alpine meadow soil, which is a shallow layer of young, weakly alkaline soil formed through simple soil formation processes that is rich in stored nutrients but poor in fast-acting nutrients (Zhang et al., 2017).

The experiment was established on a flat and undisturbed site in 2012. For conservation purposes and to enable regular experimentation, the site was fenced. The targeted species were divided into four PFGs based on ecophysiological characteristics potentially associated with response variables, such as biomass production, resource use patterns, and nitrogen (N) fixation ability (Reich et al., 2001; Gross et al., 2007; McLaren and Turkington, 2010). Cyperaceae are the dominant species in the alpine meadow. Gramineae species exhibit a resource conservation strategy due to their high C:N ratio and leaf dry matter content (Chen et al., 2022). Legumes and other forbs promote species and functional diversity, offering different plant resource utilization pathways that enhance nutrient use and release, leading to higher N and P content in alpine plant communities (Chen et al., 2022). By contrast, species and functional diversity can be fostered by the growth of forbs and legumes, which provide access strategies to improve species resource use and release, leading to higher foliar N and P contents (Chen et al., 2022). In the study area, the species are typical of subalpine meadows, dominated by Cyperaceae, with a diverse forb community. Other PFGs such as Gramineae and legume species are rare, with about 10% Gramineae and 8% of the total coverage. The most abundant plant species are Carex alatauensis and Carex parva (Cyperaceae), accounting for 40% of the total coverage. All the rest are forbs.

The establishment of the Meadow Removal Experiment (IMGRE) at the Haibei Station was described by Chen et al. (2022). The experiment followed a completely randomized design. Five treatments were established, which include the control (CK), Gramineae removal (RG), Cyperaceae removal (RC), legumes removal (RL), and other forbs removal (RF). Each of the five treatment plots (1 m × 0.75 m) was replicated five times with a total of 25 plots spaced 1 m apart. To minimize physical disturbance to soil, plants were removed by cutting aboveground sections and tiller nodes to 3 cm soil depth, twice per month from May to August (Naeem, 2002; Chen et al., 2016, 2022). At the peak of vegetative biomass (August), we recorded the species composition, frequency, cover, height, and soil nutrients within each plot in 2015 and 2022 to examine short- and long-term effects, respectively. Plant species are identified and classified with reference to “Flora Reipublieae Popularis Sinicae” (http://www.iplant.cn/frps).

The importance value is a useful measure for species location in a community and is used to indicate the population dominance. The dominant species in the community with importance values more than 0.5% were selected for niche analysis (Camargo et al., 2021). The importance value is calculated as follows ( Tables 1 , 2 ):

The Levins index (B_L ) was used to determine niche width (Lu et al., 2020), and the Pianka index (O_ik ) was used to calculate niche overlap (Schoener, 1974). These indices were calculated as follows:

where P_{i j} and P_kj are the importance values of species i and species k within the j^th plot, respectively; and O_ik is the niche overlap index between species i and k, ranging from 0 to 1.

The variance ratio (VR) test was used to determine the overall interspecific association at the community level. VR indicates whether there is a significant relationship among multiple species in the selected area, and its significance was tested using the W statistic (Schluter, 1984). VR was calculated as follows:

where N is the total number of plots in the community, T_i is the total number of target species in plot i, t is the mean number of species observed in all plots, S represents the total number of species in the community, P_j represents the frequency of species j, and n_j represents the total number of plots occupied by species j (Schluter, 1984).

Under the null hypothesis of independence, the expected value of VR is 1; that is, when VR = 1, it means that there is no connection between the species. If VR > 1, then it means that there is a positive connection between the species. If VR < 1, then it means that there is a negative connection between the species. The statistic W was used to verify the significant degree of VR deviation from 1, W = VR × N. If W > χ² _0.05 (N) or W < χ² _0.95 (N), then it means that the overall connection between the species is significant. On the contrary, the overall connection between the species is not significant (Wang et al., 2018).

Interspecific association has been widely used in interspecific ecological studies to determine the significance of interspecific association by using the χ² test, to determine interspecific associations with the Association coefficient (AC), and to analyze the interspecific associations strength based on the Ochiai index (OI), and this methodology has also been applied to the study of interspecific relationships in grassland plants (Zhang et al., 2018; Liu et al., 2019). Pairwise interspecific association is one of the important quantitative and structural characteristics of plant communities, which is the basis for the formation and evolution of plant community structure (Zhong et al., 2010). OI analysis of the strength of connectivity between species pairs (Zhang et al., 2018). The χ² value, AC, and OI were calculated as follows:

where N represents the total number of quadrats, α represents the number of quadrats with two species, b and c represent the number of quadrats with only one species, whereas d represents the number of quadrats with neither of the two species. The interspecific association is not significant when χ² < 3.814 (P > 0.05), at which point it is considered largely independent; when χ² > 3.814, at which point it is considered significant (P < 0.05); when ad > bc, the interspecific association is considered positive, and vice versa.

All analyses were conducted in the R v4.2.1 statistical platform. The niche width, niche overlap, community-level interspecific association, AC, OI, and statistical analyses were performed using the “spaa” package, and semi-matrix plots were performed using the “ggplot” package.

Differences in the dominant species niche widths within the same experimental treatments at different time scales ( Figures 2 , 3 ; Table 1 ). In the short term (2015), there were 20 dominant species in the study area, including 5 Gramineae, 3 Cyperaceae, 3 legumes, and 9 other forbs. In the long term (2022), there were 18 dominant species in the study area, including 2 Gramineae, 2 Cyperaceae, 3 legumes, and 11 other forbs. The species with the niche width range in the short-term removal period was Poa araratica (5.02) in the control, Carex alatauensis (4.99) in the RG treatment, and Deschampsia cespitosa (3.02) in the RC treatment. The species with the greatest niche width in the RL treatment was Helictotrichon tibeticum (5.00), and, in the RF treatment, it was Helictotrichon tibeticum (4.00). In the long-term removal period, the species with the niche width range was Sibbaldianthe bifurca (4.72) in the control, Thalictrum alpinum (3.95) in the RG treatment, Potentilla nivea (4.84) in the RC treatment, Ranunculus pulchellus (4.57) in the RL treatment, and Oxytropis arctica (4.68) in the RF treatment. The maximum niche wide of the species in each treatment was lower in the short term compared to that in the control, whereas there was no regular change in the long term.

Differences in the dominant species niche overlaps after PFG removal ( Figure 4A ; Supplementary Figure S1 ). The species pairs with maximum niche overlap during the short-term removal were 85 for the control, 43 species pairs for the RG treatment, 57 species pairs for the RC treatment, 93 species pairs for the RL treatment, and 36 species pairs for the RF treatment. For the long-term removal, the species pairs with maximum niche overlap were 30 for the control, 27 species pairs for the RG treatment, 26 species pairs for the RC treatment, 26 species pairs for the RL treatment, and 3 species pairs for the RF treatment. Overall, the number of species pairs with maximum niche overlap decreased as removal time proceeded.

We evaluated overall interspecific associations between species within the community ( Table 2 ). Positive correlations (VR > 1) were observed among 14 species pairs in the RG treatment during the short-term time. Similarly, positive correlations (VR > 1) were present among seven species pairs in the RF treatment during the long-term time. Conversely, negative correlations were identified among all remaining species pairs in the other treatments (VR < 1). The W Statistics showed that the overall associations among dominant species in the treatment with negative correlations among species pairs were all significant.

The number of species pairs showing a positive correlation in the short-term removal time varied among treatments ( Table 3 ; Supplementary Figure S2 ). There were 89 species pairs (46.84%) in the control, 64 species pairs (60.95%) in the RG treatment, 101 species pairs (74.26%) in the RC treatment, 72 species pairs (52.94%) in the RL treatment, and 25 species pairs (45.45%) in the RF treatment. In the long-term removal time, the number of positively correlated species pairs were 72 (47.06%) in the control, 55 (45.83%) in the RG treatment, 61 (50.83%) in the RC treatment, 57 (54.28%) in the RL treatment, and 12 (57.14%) in the RF treatment. All other species pairs exhibited a negative correlation ( Table 3 ; Supplementary Figure S2 ). The number of species pairs that were positively correlated across treatments decreased with time of removal compared to the control.

In the short-term removal time, the number of species pairs within the maximum AC range among the dominant species in the treatments followed the order of CK (21) > RL (5) > RF (4) = RC (4) > RG (2) ( Supplementary Figure S3 ). In the long-term removal time, the number of species pairs within the maximum AC range of the dominant species in the treatments followed the order of RL (13) = RG (13) > RC (6) > CK (4) > RF (1) ( Supplementary Figure S3 ). Under the highest interspecific association coefficients, the CK and RL treatment had the highest number of species pairs within the community, and the niches between species were more complex.

The number of species pairs present in the conditions with the highest values of interspecific linkage strength varied among treatments ( Figure 4B ; Supplementary Figure S4 ). There were 63 species pairs in the control, 4 species pairs in the RG treatment, 18 species pairs in the RC treatment, 47 species pairs in the RL treatment, and 14 species pairs in the RF treatment during the short-term removal time. In the long-term removal time, there are 57 species pairs in the control, 28 species pairs in the RG treatment, 35 species pairs in the RC treatment, 38 species pairs in the RL treatment, and 5 species pairs in the RF treatment.

Niche width reflects the position and role of a population within a community (Carscadden et al., 2020). In this study, the number of dominant species differed across the experimental treatments, with a decrease in the number of Gramineae and Cyperaceae and an increase in the number of forbs other than legumes. Thus, long-term species diversity loss led to gradual reductions in the niche widths of Gramineae and Cyperaceae, slowly disadvantaging them in terms of resource competition. By contrast, the niche width of other forbs gradually increased, enhancing their competitiveness for resources. Our results also provided strong evidence of alpine meadow degradation. The present findings are consistent with a previous report that the combined dominance ratios of Gramineae, Cyperaceae, and forbs other than legumes are negatively correlated (Jiang et al., 2021). In this study, only when other forbs were removed did the remaining Gramineae and Cyperaceae attain larger niche widths. The niche widths of the remaining functional group treatments gradually shifted toward dominance by other forbs, indicating that these forbs were the most powerful competitors for environmental resources in the alpine meadow plant communities. Kong et al. (2011) indicated that PFG removal resulted in less aboveground biomass to provide the photosynthetic products needed to maintain root biomass, thereby affecting root nutrients; thus, the aboveground traits of forbs generally favor the uptake of more environmental resources, which in turn, can affect the survival of other PFGs to a greater extent. In this manner, the presence of forbs may be among the most important factors affecting the survival of other species. The niche width of the remaining PFGs after species loss varied by time scale; generally, longer intervals resulted in more realistic niche widths, as reflected by compensatory effects in the remaining species. Short-term species culling might have caused our results to be influenced by the compensatory effects of the remaining PFGs; because most of the biomass in alpine meadow is allocated to belowground plant parts, the species remaining after culling may not show significant compensation in the short term (Díaz et al., 2003; Bai et al., 2004), such that long-term monitoring would lead to more accurate results.

Niche overlap is a measure of the degree of species similarity in terms of environmental adaptation or resource use (Pastore et al., 2021). In this study, the niche overlap of the remaining dominant species in the community shifted from high to low, and the number of species pairs with the highest niche overlap shifted from high to low as the monitoring period was extended. Thus, our results indicated that species loss caused the remaining species to be less similar in terms of resource use. Previous studies have shown that species coexistence presupposes niche differentiation and that species with identical niches cannot coexist in the long term (Song, 2017); however, niche overlap does not necessarily lead to competition between species and may be necessary for exploitative competition, as competition is also related to resource amounts and population sizes (Letten et al., 2017). The high degree of niche overlap detected in this study cannot directly confirm intense competition between species but rather indicates high potential for competition.

The overall connectivity between community species is closely related to stability (Wittebolle et al., 2009). Based on previous studies, during positive community succession, species maximize resource use through competition and cooperation, species composition is gradually fixed, community structure becomes more stable and eventually reaches an apex, and coexisting species tend to be positively associated; however, the association between individual species pairs decreases over the course of succession (Guo et al., 2008; Haugaasen and Peres, 2009). In this study, overall species connectivity remained negative during the 10-year monitoring trial; however, overall species connectivity was positive in the treatment in which forbs other than legumes were completely removed, probably because Gramineae and Cyperaceae are dominant species in alpine meadows and thus determine the development direction of the community to some extent. As such, the community examined in this study may have reached a relatively stable late successional stage.

Interspecific associativity reflects interrelationships between the species comprising a community over a period of time (Ru et al., 2022). It is often assumed that positively associated species pairs have mutually beneficial and complementary relationships, whereas negatively associated species pairs are more likely to experience competition and exclusion (Levine et al., 2017). As time passes, most of the treatment trials showed negative associations, suggesting that species interactions tend to be mainly competitive and exclusionary; however, the treatment in which other forbs were completely removed showed the opposite trend, toward mutualism and complementarity, indicating that species interactions under this treatment were dominant in terms of resource use (Åkesson et al., 2021). Interspecific correlations are based on multidegree data between species pairs and can provide a basis for analyzing species extinction patterns within a community (Norberg et al., 2012). These species interrelationships are consistent with interspecific associations but differ somewhat in terms of sensitivity. Generally, when two species appear or disappear simultaneously in a sample, the association between these species is positive; however, the association could be negative if the number of individuals of one species increases while that of the other species decreases (Liu et al., 2019). In our study, this difference was reflected by the higher proportion of negatively correlated species pairs detected by the Pearson correlation test than by the χ² test and AC values in this study.

It is increasingly recognized that all organisms modify their environments (i.e., niche construction or ecosystem engineering) in response to various environmental disturbances (Odling-Smee et al., 2013). Such changes include effects on the distribution and abundance of organisms, dominant species, control of energy, material flows, ecosystem resilience, and specific nutrient relationships (Chapin et al., 1997). There is evolutionary evidence that, even when key resources are independently renewed or depleted, the effects of niche construction can override selection from external sources, thereby creating new evolutionary trajectories and equilibria, generating and eliminating polymorphisms, and producing time-lagged selection as well as other unusual dynamic responses (Laland et al., 1999). The present study revealed that long-term species loss resulted in larger niche widths for forbs other than legumes and smaller niche widths for dominant species in the community. Furthermore, niche overlapping pairs among dominant species in the community tended to decrease with removal treatment time, and the number of related species pairs decreased, indicating that competition intensity between species within the community tended to weaken. Interestingly, competitive intensity between species became stronger or weaker more rapidly when forbs other than legumes were removed from the community. The removal of different PFGs in this study was considered to reflect the process of reconstructing the niche of each species within the community under temporal turnover, including resource acquisition, resource use within changing communities, and interactions between species. Further investigation is required to explore how niche reconstruction among species in communities occurs over time, the key channels through which such reconstruction occurs, and changing patterns in the construction of linkages among species.

The niche overlap pairs of dominant species in each treatment were fewer compared to the control, and the niche overlap value in the control changed with the delay. The community’s species relationships exhibited a decline in positively correlated species pairs over time, with both the number of these pairs and the strength of associations in each treatment being lower than those in the control. Species interactions can both mitigate and exacerbate the effects of climate change on species and interact with other eco-evolutionary processes (Post, 2013). For example, species interactions can affect the evolutionary response to altered environmental conditions; dispersal can release species from negative interactions or increase them through migration or through invasions (Mazancourt et al., 2008; Walther et al., 2009; Van Eldijk et al., 2020). However, this phenomenon is not certainly caused by changes in the environment but is closely related to the number of potential interspecific competitors in the community or the so-called “diffuse competition,” and it has been argued that the maximum tolerable overlap should be inversely proportional to the intensity of the competition, which is really the main content of the niche overlap hypothesis (Pianka, 1974). More importantly, most of the studies have been conducted under fenced conditions, and the results of the experiments have limitations. However, an increasing number of studies pointed out that the species niches are affected by space and time and that quantifying species niches is needed to explore the relationship between species distributions and environment, as well as to predict the potential risk of species invasions and future species extinction rates (Jackson et al., 2009; Liu et al., 2020). In the future, we need to further explore the effects of plant diversity loss on species niches and ecosystem functioning at broader spatial scales and longer time scales.