Elevation patterns of plant diversity and recent altitudinal range shifts in Sinai’s 2 high mountain flora 3

Questions: Is there evidence of recent altitudinal range shifts in a hyper-arid Middle Eastern desert mountain flora? How do the directions of shift for upper and lower altitudinal range limits of plants vary? Method: We tested for shifts in both upper and lower altitudinal range limits by comparing a 38 1970s dataset of recorded species’ limits with recent surveys using altitudinal transects across 39 36 sites. Altitudinal limits between 63 paired upper-limit and 22 paired lower-limit values from 40 the 1970s and 2014 were compared using paired t-tests; binomial tests were used to indicate 41 the dominant direction of change. The upper and lower limits of 22 species were considered 42 together to allow assessment of overall altitudinal range-size changes. In order to avoid the 43 potential effect of yearly environmental fluctuations on the distributions of annual species, 44 subsets of upper and lower limit shifts were taken for perennials, and trees and shrubs. 45 Results: Our results show significant overall upslope shifts in mean upper altitudinal limits 46 and significant overall downslope shifts in mean lower altitudinal limits. A majority of assessed 47 species expanded their altitudinal ranges, but the responses of individual species varied. Since 48 perennial herbs/graminoids, and trees and shrubs, show strong patterns of change, we suggest 49 there has been a long-term shift in altitudinal range in South Sinai’s mountain flora. Greater 50 research effort needs to be focussed upon the drivers of range-shift responses in arid regions.


Introduction
quadrat GPS locations, Table S4 for site photos and descriptions, and Table S5 for species lists   156 and abundances by quadrat). We were not able to revisit exact sites surveyed in the 1970s as were not differentiated in the earlier dataset, and therefore for the purposes of comparison the 181 records collected in 2014 were amalgamated for these species. In total, 81 species were 182 available with upper altitudinal limits from both the 1970s and 2014. The significantly greater 183 sampling effort required to establish accurately the lower altitudinal limits for the more 184 widespread species was beyond the scope of this study which deals specifically with the high- were selected (see Table S2). This selection allowed the upper limits of 63 and lower limits of 191 22 species to be identified. Subsets of upper-and lower-limit shifts were taken for perennials, 192 and trees and shrubs to allow comparisons to be made that avoided the potential effect of yearly 193 environmental (specifically rainfall) fluctuation on the distributions of annual species.  To estimate shifts in altitudinal ranges, the altitudinal limits between 63 paired upper-limit and 217 22 paired lower-limit values from the 1970s and 2014 were compared using paired t-tests to 218 test the null hypothesis that the mean difference was zero. Sign tests (i.e. binomial tests on the 219 numbers of negative and positive changes) were used to indicate the dominant direction of 220 change. 22 species had estimates of both upper and lower limits, and so were considered 221 together to allow assessment of overall altitudinal range-size changes. Species were categorised 222 as showing no change, expanded range, or contracted range ( Table 1). Movement of less than 223 100 m for either limit was regarded as stationary in view of the measurement resolution of the 224 1970s data. A binomial test was used to identify whether expansion or contraction of ranges 225 was the dominant pattern.

226
As an aid to interpretation, reasons for the changes were explored in a GLM by using 227 the differences in altitudinal limits between 2014 and the 1970s as the response variable, and a 228 variety of predictors: flowering season(s), basic growth-form (herb, shrub or tree), Raunkiaer life-form, and basic life-form (annual or perennial). The best fitting models and predictors were 230 selected by use of AICs.

233
Patterns of diversity in the new data 234 The overall patterns of diversity were indicated by the three Hill's numbers, but each followed 235 a distinct altitudinal pattern (see Fig. 2). The highest levels of species richness ( 0 D) were found 236 at higher altitudes, decreasing down a shallow concave curve with the lowest values at lower 237 altitude (approx. 1400-1600 m). The number of 'typical' (common) species, 1 D, was highest at 238 lower-middle elevations (approx. 1700-1800 m), and declined with increasing altitude. In 239 contrast, the number of abundant species, 2 D, was lowest at lower-middle elevations, with 240 highest values at the top of the altitude range. The summary data are in Tables S2 and S3.   this was particularly the case for species that differed by more than 100 m (9/9, binomial test 299 p=0.002).
The three Hill's number diversity indices provide a greater insight than a single measure (Chao accurately determining causes for the observed range shifts is beyond the scope of this study.

347
No good data on long term precipitation in the South Sinai mountains exist. It is therefore 348 difficult conclusively to attribute downward shifts of lower limits to increased precipitation.   Table 1 373 for detail). Therefore, whilst changes in grazing intensity may have affected downslope range 374 shifts, we suggest that climatic change explains the observed upwards range shifts better.

375
Here, in this arid mountain system, we have documented what we think is the first 376 record of significant downslope shifts of plant lower-altitudinal limits outside Europe. Despite the less-than-ideal quality of the historical data, mean upper limits have increased whilst lower 378 limits have decreased since the 1970s, leading to a divergent pattern of mean altitude limits.

379
When considering the upper and lower altitudinal limits of individual species, we found 380 heterogeneity in the joint responses with no clear predominant pattern. One must bear in mind 381 that these species are a subset of the selected group of high-mountain species that may not be 382 representative of all the species present in that environment. 383 We now know that there have been significant upwards shifts in the upper altitudinal  In this study we have presented the first recorded instance of contemporary altitudinal-  Table S2) will provide a baseline for future fine-scale monitoring. 405 We also stress how important it is to consider both upper and lower altitudinal limits to give an 406 accurate indication of overall altitudinal range changes. We need to focus on lower limits to 407 understand better the ecological drivers and dynamics underlying heterogeneous responses at 408 the range limits.

Figure 2
Hill's numbers (see Chao et al. 2012) for diversity by altitude with fitted GAM model with Normal errors and 95% confidence region. Ascending Hill's numbers give reducing weight to less-abundant species: (a) mean 0 D (= species richness); (b) mean 1 D (number of 'typical' common species); (c) mean 2 D (number of 'abundant' species).

Figure 3
Difference in upper altitude limit for each plant species between the 1970s and 2014.  Iphiona mucronata 17 34 Primula boveana 1 32