Global warming has exerted a profound influence on plant species diversity at the community level (Arft et al., 1999; Klein et al., 2004; Elmendorf et al., 2012; Pauli et al., 2012; Steinbauer et al., 2018), and will continue to influence ecosystem processes, functioning, and ecosystem services to human society (Hector and Bagchi, 2007; Naeem et al., 2010; Hooper et al., 2012; Tilman et al., 2014; Isbell et al., 2017; Liu et al., 2018). For example, plant species loss may significantly reduce primary productivity and plant litter decomposition rates (Hector and Bagchi, 2007; Hooper et al., 2012). Consequently, examining the responses of plant communities to warming is critical to accurately predict future changes in ecosystem properties.

Prior studies have revealed significant changes in species richness and relative abundance of plant communities in response to warming particularly in alpine and arctic ecosystems (Grabherr et al., 1994; Klein et al., 2004; Walker et al., 2006; Elmendorf et al., 2012). For example, experimental warming is reported to increase primary productivity and the relative abundance of grasses at the expense of the sedges and forbs in alpine meadows (Ganjurjav et al., 2016; Liu et al., 2018). A rapid loss of plant species has also been recorded in alpine meadows (Klein et al., 2004; Wang et al., 2012; Zhang et al., 2015, 2017). Analogous results have been obtained in arctic regions (Chapin et al., 1995; Walker et al., 2006). However, most of these reports have focused on changes in plant productivity or species diversity at the community level (Klein et al., 2004), or on changes in relative abundance at the plant functional type level (Chapin et al., 1996; Ganjurjav et al., 2016; Liu et al., 2018). Moreover, most of these studies have reported the number of species lost but not the identity of the species that were lost.

As a consequence, the effect of global warming on species diversity at the species level is incompletely understood. The paucity of information at the species level is likely to result in misleading conclusions for three reasons. First, it fosters the assumption that changes at the species level are consistent with changes at the plant functional type level. This assumption is most probably incorrect because it is unlikely that all of the species within a specific functional type will change in abundance in the same way. For example, it is possible that one species will increase its relative abundance even if the functional type to which it belongs decreases in overall abundance. Likewise, it is also possible that some species will increase in abundance while others will decrease, such that the abundance of the functional type to which they belong remains unchanged (Klanderud and Totland, 2005). Second, the assumption that changes at the species level are consistent with changes at the plant functional type level leads to the speculation that species loss occurs only in the plant functional types that decrease in relative abundance. Theoretically, this speculation is likely to be true only if the species decreasing in abundance are more likely to be lost, a speculation that thus far has not been tested. Third, in the context of global warming, it leads to the speculation that species loss is the result of warming-induced changes in species richness. This would be untrue if warming-induced changes in species identity differ among treatments. Specifically, if only species richness is considered, species loss could be masked by species invasion. Consequently, overlooking information at the species level can limit our ability to accurately predict future ecosystem functioning, particularly when the selection effect (i.e., species identity effect) of species diversity on ecosystem functioning dominates the biodiversity—ecosystem functioning relationship.

Here, we present data concerning changes in community species diversity, the relative abundance for four plant functional types (i.e., legumes, grasses, sedges, and forbs) and each plant species, and species identity in an alpine meadow in eastern Tibetan Plateau, which is considered among the most sensitive ecosystems to climate change (Piao et al., 2012; Zhang et al., 2017). We use the data to test the null hypothesis that warming-induced changes are consistent among the community level (as indicated by aboveground plant biomass, species richness, species diversity, and evenness), the functional type level, and the species level. To test this hypothesis, we examined (1) whether changes in relative abundance are consistent among different plant functional types, (2) whether changes in relative abundance are consistent among different plant species within specific plant functional types, and (3) whether species loss occurs in the functional types whose relative abundances decreased by warming.

Our results indicate that a warmed alpine meadow community has a higher primary productivity (as indicated by aboveground plant biomass), and warming-induced changes in the relative abundance differ significantly among plant functional types. These responses are consistent with previous studies addressing the effect of experimental warming on grassland communities (Arft et al., 1999; Wang et al., 2012; Zhang et al., 2015; Ganjurjav et al., 2016). Importantly, data show that species within the same plant functional type can respond differently to warming. These observations indicate that reporting changes at the community or functional type level is insufficient to predict changes at the species level. In addition, species loss occurred in both types of chambers and different species were lost in different chamber types, such that reporting a change in species richness is insufficient for describing the change at the plant community level. Thus, reporting the identity of lost species and understanding how they are lost are important when attempting to predict what species are more likely to go locally extinct under warming climatic conditions, a finding that has significance regarding conservation biology.

Although the warming-induced increase in aboveground plant biomass is consistent with previous warming studies in similar alpine meadows (Ganjurjav et al., 2016; Liu et al., 2018), changes in species diversity are less pronounced in this as opposed to other studies. For example, Chapin et al. (1995) showed that experimental warming reduced species richness by 30–50% in a tussock tundra; Klein et al. (2004) reported that warming over a period of 4 years resulted in a 26–36% decrease in species richness in Tibetan alpine meadow; and Sternberg et al. (1999) reported that manipulated warming decreased species richness by 10% in a calcareous grassland. In our study, warming did not significantly change species richness, Shannon diversity, or Shannon evenness over a 4-year period. This result contrasts with those of previous warming experiments conducted in alpine and arctic regions (Klein et al., 2004; Wang et al., 2012; Zhang et al., 2017). There are three possible explanations for why the warming-induced decrease in species richness is much smaller in our study system. First, the temperature increase in our experiment is smaller than that in other studies. Greater temperature increases often induce more pronounced changes in plant species diversity (Wang et al., 2019). Second, our experimental chambers are much larger than those used in previous studies. Large chambers might buffer the changes in species richness occurring at smaller and more homogeneous scales, i.e., larger chambers are likely to be more heterogeneous in their microclimate and micro-soil conditions compared to smaller ones (Vellend et al., 2013; Suggitt et al., 2018). Third, in the present study, the experimental control was non-warmed chambers compared to dispersal-unlimited field sites used in many other studies (e.g., Klein et al., 2004; Zhang et al., 2017). These chambers (including the warmed ones) were taller than those used in other warming studies (Klein et al., 2004, 0.4 m; Zhang et al., 2017, 0.4 m; Ganjurjav et al., 2016, 0.45 m), which could have significantly reduced or even prevented seed dispersal. Thus, seed dispersal would be equally confined in this study, whereas it was more confined in the warmed chambers than open sites in the other studies.

The difference in relative abundance among functional types reported here is also consistent with similar studies in our study region. Ganjurjav et al. (2016) and Liu et al. (2018) both reported a community shift from one dominated by forbs to one dominated by grasses in response to warming, whereas other studies indicated that sedge and legume species remain relatively unchanged in response to warming (Klanderud and Totland, 2005; Ganjurjav et al., 2016; Liu et al., 2018). These observations suggest that plant growth is normally limited by the low temperatures in alpine meadows and that increased temperatures likely increase aboveground plant biomass and plant height, which collectively increases interspecific competition for light. Tall grasses can competitively shade and exclude short forb and sedge species that tend to be rosette-like and short. Moreover, although soil moisture was on average high in our meadow study site, warming reduced soil moisture to a low level (<20% v/v) during the dry season (usually in August). This reduction in soil moisture might have limited the growth of sedge species and some forb species that are adapted to moist soil conditions (e.g., B. sinocompressus and S. stella; see also Klanderud, 2008; Little et al., 2015; Cao et al., 2017), but favored the growth of drought-tolerant species such as D. caespitosa and Festuca spp. (this study; see also Klanderud, 2008; Little et al., 2015).

Importantly, individual species differed in their responses to warming both in direction and amplitude even for those within the same plant functional type. As noted, contrasting responses to warming have been observed in all plant functional types. This observation is attributable to differences in plant traits even those within the same functional type. For example, most forb species decreased in relative abundance probably because of competitive exclusion by tall grass species. Yet, three forbs (i.e., G. pylzowianum, P. anserine, and E. pectinate) increased substantially in their relative abundance in the warmed chambers. The abundance increase in G. pylzowianum and P. anserine is consistent with previous studies (Klanderud, 2008; Li et al., 2011), and can be attributed to their belowground storage organ (tubers) and the habit of climbing growing, which allows individuals to vegetatively overgrow other plants, thereby avoiding shade compared to plants growing in non-warmed chambers. The increase in the abundance of E. pectinate is also consistent with previous findings (Nyléhn and Totland, 1999; Klanderud, 2008). The underlying mechanism for its success is that this species is an annual root hemi-parasitic forb that can absorb water and nutrients from monocot species. Consequently, it increases in abundance in tandem with the abundance of grass species. It could be argued that species with different traits should not be placed within the same functional group. However, assigning species to too many different functional groups presents additional problems. Nevertheless, it is clear that subtle differences among the species assigned to any functional species group can result in substantive differences in response to warming, whether induced experimentally or experienced under natural field conditions.

Although differences in species richness were statistically indistinguishable between warmed and non-warmed treatments, species loss was demonstrably manifest as indicated by the fact that both warmed and non-warmed chambers experienced species loss and that different species were lost in the two different types of chambers. A number of factors may contribute to this observation. For example, in both warmed and non-warmed chambers, the loss of the forb species Carum carvi and Microula sikkimensis might have been the result of reduced disturbance and increased plant density during the experiment, because these species usually grow in ruderal or bare areas. The loss of the forbs Tibetia himalaica and Sanguisorba filiformis might be due to the shading of tall grasses because these two species are short and rare. The loss of the forb species Saussurea stella and the sedge species Blysmus sinocompressus in warmed chambers may perhaps be the result of decreasing soil moisture and their competitive exclusion by tall plants. Both of these species are short and often grow in high-moisture microhabitats (Xiaoli Hu, Personal Observation; Flora of China, http://www.iplant.cn/). Finally, the loss of the forb species Angelica apaensis in non-warmed chambers may reflect its demographic stochasticity because this species is extremely rare.

In summary, the results presented here show that a small increase in mean annual temperature and a winter-biased higher temperature during the non-growing season induces significant changes in an alpine meadow plant community, as evidenced by shifts in the relative abundance and biomass of different plant functional types. The results also indicate that changes in species relative abundance and biomass are not necessarily consistent with those of plant functional types, indicating that the data from plant functional types can obscure important information about the differential responses of individual species to climate change. Perhaps more important, whether species loss occurs cannot be deduced from changes in species richness. Collectively, species-level information is necessary to accurately predict plant community responses to global change. Future studies should consider the behavior of individual species when predicting changes in ecosystem functioning or service, or making policy decisions concerning biodiversity protection and conservation.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

SS started the project and designed research, and drafted the manuscript. XH, WZ, and XL performed research. SS, XH, and KN analyzed data and wrote the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.