Alpine plants’ distribution is being pushed higher towards mountaintops due to global warming, finally diminishing their range and thereby increasing the risk of extinction. Plants with specialized ‘glasshouse’ structures have adapted well to harsh alpine environments, notably to the extremely low temperatures, which makes them vulnerable to global warming. However, their response to global warming is quite unexplored. Therefore, by compiling occurrences and several environmental strata, we utilized multiple ensemble species distribution modeling (eSDM) to estimate the historical, present-day, and future distribution of two alpine ‘glasshouse’ species Rheum nobile Hook. f. & Thomson and R. alexandrae Batalin. Rheum nobile was predicted to extend its distribution from the Eastern Himalaya (EH) to the Hengduan Mountains (HM), whereas R. alexandrae was restricted exclusively in the HM. Both species witnessed a northward expansion of suitable habitats followed by a southerly retreat in the HM region. Our findings reveal that both species have a considerable range shift under different climate change scenarios, mainly triggered by precipitation rather than temperature. The model predicted northward and upward migration for both species since the last glacial period which is mainly due to expected future climate change scenarios. Further, the observed niche overlap between the two species presented that they are more divergent depending on their habitat, except for certain regions in the HM. However, relocating appropriate habitats to the north and high elevation may not ensure the species’ survival, as it needs to adapt to the extreme climatic circumstances in alpine habitats. Therefore, we advocate for more conservation efforts in these biodiversity hotspots.

We focused on the entire geographic distribution ranges of the two species (Figure 1D). For this, the past field (1937–2020) (Supplementary Table S1) datasets were compiled and cross-checked with the herbarium records from the National Herbarium and Plant Laboratories (KATH, Nepal) and Kunming Institute of Botany, CAS (KUN, China) to construct the geographic distribution of both species. In addition to past field and herbarium collections, we also ground-validated occurrence points (from 2015 to 2022) through GPS in the field and cross-checked with the online databases of the Chinese National Herbarium (PE; http://pe.ibcas.ac.cn/en/), Chinese Virtual Herbarium (CVH; http://www.cvh.ac.cn/), the Global Biodiversity Information Facility (GBIF.org, 2021), the Royal Botanical Garden at Edinburgh (RBGE, United Kingdom; http://data.rbge.org.uk/search/herbarium/) and the Herbarium at the University of Tokyo (TI, Japan; http://umdb.um.u-tokyo.ac.jp/DShokubu/). While compiling the occurrence datasets, we mainly focused on geographic coordinates recorded by the different resource persons during the field. So, the occurrence points from the other herbarium specimens and online databases were discarded in the datasets. 

With the use of predictive environmental variables and a given species distribution model, we used the standardized approach in ODMAP (Overview, Data, Model, Assessment and Prediction) (Zurell et al., 2020) protocol structure (Supplementary Methods). The ensemble of SDMs was implemented using the R-package ‘Biomod2’ (Thuiller et al., 2020).