Upward elevation and northwest range shifts for alpine Meconopsis species in the Himalaya–Hengduan Mountains region

Abstract Climate change may impact the distribution of species by shifting their ranges to higher elevations or higher latitudes. The impacts on alpine plant species may be particularly profound due to a potential lack of availability of future suitable habitat. To identify how alpine species have responded to climate change during the past century as well as to predict how they may react to possible global climate change scenarios in the future, we investigate the climatic responses of seven species of Meconopsis, a representative genus endemic in the alpine meadow and subnival region of the Himalaya–Hengduan Mountains. We analyzed past elevational shifts, as well as projected shifts in longitude, latitude, elevation, and range size using historical specimen records and species distribution modeling under optimistic (RCP 4.5) and pessimistic (RCP 8.5) scenarios across three general circulation models for 2070. Our results indicate that across all seven species, there has been an upward shift in mean elevation of 302.3 m between the pre‐1970s (1922–1969) and the post‐1970s (1970–2016). The model predictions suggest that the future suitable climate space will continue to shift upwards in elevation (as well as northwards and westwards) by 2070. While for most of the analyzed species, the area of suitable climate space is predicted to expand under the optimistic emission scenario, the area contracts, or, at best, shows little change under the pessimistic scenario. Species such as M. punicea, which already occupy high latitudes, are consistently predicted to experience a contraction of suitable climate space across all the models by 2070 and may consequently deserve particular attention by conservation strategies. Collectively, our results suggest that the alpine high‐latitude species analyzed here have already been significantly impacted by climate change and that these trends may continue over the coming decades.


| INTRODUC TI ON
Species may respond to climate change by shifting their ecological niche through plastic changes (Nicotra et al., 2010) and evolutionary adaptation (Visser, 2008), and/or by shifting their range to track original climatic conditions (Hickling, Roy, Hill, Fox, & Thomas, 2006;Holt, 1990). Evidence suggests that the rate of species climatic niche evolution may be slow compared to the rate of climate change (Quintero & Wiens, 2013), and failure to respond to the changing abiotic and biotic conditions may lead to range contractions (Giménez-Benavides, Albert, Iriondo, & Escudero, 2011) and/or local extinctions (Moritz & Agudo, 2013;Wiens, 2016).
Due to their restricted distribution range and high levels of endemism, alpine species, in particular, are generally highly sensitive to climate change (Jump, Huang, & Chou, 2012;Lenoir et al., 2008).
Cold-adapted species (mainly nival and subnival species) endemic to the summit region of mountain systems tend to decline in abundance or contract in range size (Pauli, Gottfried, Reiter, Klettner, & Grabherr, 2007;Rumpf et al., 2018), while low-elevation species adapted to warmer temperatures encroach. For example, numerous studies in Europe have documented how species adapted to warmer temperatures have occupied habitats previously occupied by the cryophilous subnival flora, which, as a result of increased competition for cooler habitats and limited space for new habitat expansion, is then restricted to small patches "trapped" in remaining habitats where cooler conditions persist (Gottfried et al., 2012;Gottfried, Pauli, Reiter, & Grabherr, 1999;Pauli et al., 2003). However, the extent of distributional range shifts due to climatic change for alpine species remains poorly understood.
The Himalaya-Hengduan Mountains region is located in a global biodiversity hotspot which, due to its recent geological history and diversity of habitats, supports alpine regions containing relatively high levels of nival and subnival plant diversity and endemism (Xu, Li, & Sun, 2014). Recent photographic comparisons have shown that climate change during the past several decades has caused glacier retreat and subsequent upward shifts of the alpine tree line in the area (Baker & Moseley, 2007). This is in line with findings for the wider Asian mountain region, where there is widespread evidence for climate-related glacier shrinkage and tree and shrub line advancement (Cogley, 2016;Du et al., 2018;Myers-Smith & Hik, 2017), potentially threatening regional endemism in alpine communities.
Meconopsis, commonly known as Himalayan blue poppies, is a genus of the Papaveraceae with ~60 species confined to alpine meadow or subnival habitats ( Figure 1) in the Himalaya-Hengduan Mountains region. It is verified by recent molecular phylogenies (Liu, Liu, Yang, & Wang, 2014;Xiao & Simpson, 2017) with a conserved type, M. regia (Grey-Wilson, 2012). Meconopsis species are entomophilous plants, and they mainly attract flies as pollinators by providing them with a warm shelter (Wu et al., 2015). Due to their restricted ranges and limited pollinators in high-elevation habitats, species of Meconopsis may be particularly sensitive to climate change and are an ideal model to investigate the climatic responses of plants in this biodiversity hotspot. Here, we investigate the climatic responses of seven species of Meconopsis during the past century F I G U R E 1 One of the species studied: Meconopsis punicea, photographed in Gansu province, China (2016) to predict how they may react to possible global climate change in the coming decades. We use specimen records over the past one hundred years to see whether significant historical shifts in elevation have occurred. To explore potential future shifts in longitude, latitude, elevation, and range size, we used a species distribution modeling (SDM) framework to project the future distributions of the species under optimistic and pessimistic greenhouse gas scenarios in 2070. In addition, we compared historical rates of shifts in elevation with future projections to evaluate the validity of model projections and to evaluate species persistence under climate warming.

| Occurrence data
Occurrence data were obtained from three sources: (a) speci-  Shang et al., 2015;Yang, Qin, Li, & Wang, 2012;Yang et al., 2010) and from field collections of our colleagues in the last decade.
We first collected the distribution information of all the species of Meconopsis that had occurrence data in the Himalaya-Hengduan

Mountains region (3,745 samples for 35 species). Seven species of
Meconopsis that had a representative number of specimens ranging between 147 and 807 (N = 2,911; Supporting information Table S1) were included in our analysis to ensure that we would have enough valid data for subsequent analyses. We removed duplicated specimens with the same collection number, specimens with problematic identification and/or potentially erroneous locality information, and specimens without collection year.
Two sets of data were generated respectively following different criteria. (a) For data used in SDM, we restricted our analysis to specimens collected after 1950 that had a detailed description of their collection localities, which were then used to search on Google Earth for precise spatial coordinates. We combined occurrence data from specimens, field collections, and the published literature and removed duplicate occurrences within a 5-kilometer range (yielding a total of N = 793 records with numbers of records per species ranging from 19 to 252, Figure 2; Supporting information Table S1) in order to lower the potential autocorrelation through spatial filtering. (b) For analysis of historical shifts in elevation, only those specimens with detailed elevation and collection date (year) were used.
With data from field collections and published literature combined, 2,541 records remained spanning from year 1922 to 2016 (Table 1; Supporting information Figure S1).

| Historical shifts in elevation
We divided the specimens with elevational records into two equal time periods of c. 50 years each (pre-1970s: 1922-1969 and post-1970s: 1970-2016). This split reflects a turning point in global (and local Himalayan) temperature trends, which exhibit a steady increase since 1970 (IPCC, 2014;Shrestha, Wake, Mayewski, & Dibb, 1999). It should be noted that due to the vacancy of refined tools to estimate the precise elevation in the early years, the pre-1970s elevational records normally have an approximate accuracy of 50 or 100 m compared to post-1970s data, which in contrast have much finer and precise values. In total, there were 874 elevational records for the pre-1970s and 1667 for the post-1970s, and the mean year of occurrences was 1950 and 1990, respectively. The number of elevational records per species and time period ranged from a minimum of 29 to a maximum of 477 (Table 1). Shifts in mean elevation within and between time periods were then compared for each species separately as well as across all species collectively, and significance was established using paired t tests.

| Species distribution modeling
To project the species distributions to different climate change scenarios in 2070, bioclimatic variables of current conditions (Current: Pathway (RCP 4.5), and a pessimistic scenario whereby emissions continue to rise throughout the century (RCP 8.5). Thus, there were six potential future scenarios in total. The three GCMs were chosen as they were evaluated to perform best in terms of both temperature and precipitation in Himalaya-Hengduan Mountains region (Wu, Jiang, & Xie, 2017;Zhang, Zhang, & Fan, 2015). A Pearson's correlation test was implemented for each pair of the 19 climatic variables downloaded (Supporting information Table S2), and we removed the highly correlated variables with correlation coefficients above 0.90.
After this procedure, 8 bioclimatic variables, including four variables associated with temperature (bio1: annual mean temperature; bio2: mean diurnal range; bio3: isothermality; and bio4: temperature seasonality) and four variables associated with precipitation (bio12: annual precipitation; bio14: precipitation of driest month; bio15: precipitation seasonality; and bio18: precipitation of warmest quarter), were used in our analyses. All the layers were cut and standardized to the same resolution (30 arc-seconds) using a mask fitted to the species distribution region (78°E-117°E, 22°N-45°N) with the same coordinate system (WGS 1984) and transferred to ASCII format to enable the operation in the model.

Our species distribution models were based on Maximum
Entropy Modeling (MaxEnt), which has been shown to be the most appropriate technique for modeling presence-only data (Elith et al., 2011;Phillips, Anderson, & Schapire, 2006). We set the regularization multiplier value as "2" to reduce overfitting (Radosavljevic, Anderson, & Araújo, 2014) and the maximum iterations as "1,000" to allow more time for convergence. We used the average output (based on ten replicate cross-validation runs for each species) for subsequent analyses. We reclassified the MaxEnt output file using the 10-percentile training presence logistic threshold value to define a species potential distribution region, above which species were considered "present" in the region, a method widely recognized for distinguishing suitable from unsuitable regions (Deb, Phinn, Butt, & McAlpine, 2017;Hughes, 2017;Kramer-Schadt et al., 2013;Radosavljevic et al., 2014). We then calculated the longitude, latitude, elevation, and range size of each cell of potential presence,   (Table 1; Supporting information Figure S2b).

| Projected distributions of species
Projections of current climate preferences onto six climate scenarios for 2070 (three GCMs combined with two RCPs) showed similar trends to the ones established using historical records. For all species and all scenarios, the models suggest that there will be shifts TA B L E 1 Mean elevation (m ± standard error) of the seven species of Meconopsis species in the Himalaya-Hengduan Mountains Notes. Elevational records are listed for the two time periods (pre-1970s: 1922-1969; post-1970s: 1970-2016) in this study. # Significant historical shifts in elevation between the two time periods. *p < 0.05, **p < 0.01,***p <0.001.
in suitable climate to higher elevations, latitudes, and more west-  TA B L E 2 The RCP 4.5 and RCP 8.5 scenarios of ACCESS1-0 (AC) model projections for the average distribution in elevation (m ± standard deviation), the range size (km 2 ), and the proportion of range size shift (%) between the current time period and the year 2070 for the seven Meconopsis species in the Himalaya-Hengduan Mountains

| Historical shifts in elevation
We found that all seven sampled species There has been a steady increase in global temperature since 1970, and the Himalaya-Hengduan Mountains region has an alarming warming rate of 0.6°C per decade, which is considerably higher than the global average (IPCC, 2014;Shrestha, Gautam, & Bawa, 2012). Species of Meconopsis are perennial herbs that mainly occur in alpine or subnival habitats, which may be more sensitive to the climate warming and subsequent upward shifts to cooler habitats. Our results contribute to a growing literature base that increasingly suggests that the process of tracking suitable climatic niches through dispersal to relatively cooler habitats may be a ubiquitous response of alpine plant species to climate change at local, regional, and global scales.

| Projections in distributional shifts
For all seven species in our study, the modeled predictions of suitable climate consistently indicated that species would need to move to higher elevation and latitudes to track the currently Mountains; whereby, species are often predicted to experience an expansion of suitable climate space upwards and northwards (Liang et al., 2018;You et al., 2018). An increase in suitable habitat does of course not necessarily mean that the species will be able to track it in complex mountain systems. M. punicea, the species that was projected to show range contraction among all the models, is a species distributed in relatively high latitude in this region. The species was projected to lose area in the east and southwest margin of the distribution range, and the new potential habitat in the northwest is limited. Notwithstanding these potential limitations, our results suggest that Meconopsis will be impacted by climate change and that the im- Lian-Ming Gao https://orcid.org/0000-0001-9047-2658