Consistent replacement of small- by large-ranged plant species across habitats

The direction and magnitude of long-term changes in local plant species richness are highly variable among studies, while species turnover is ubiquitous. However, it is unknown whether the nature of species turnover is idiosyncratic or whether certain types of species are consistently gained or lost 126 across different habitats. To address this question, we analyzed the trajectories of 1,827 vascular plant species over time intervals of up to 78 years at 141 sites in three habitats in Europe – mountain summits, 128 forests, and lowland grasslands. Consistent across all habitats, we found that plant species with small geographic ranges tended to be replaced by species with large ranges, despite habitat-specific trends in 130 species richness. Our results point to a predictable component of species turnover, likely explained by 131 aspects of species’ niches correlated with geographic range size. Species with larger ranges tend to be 132 associated with nutrient-rich sites and we found community composition shifts towards more nutrient- 133 demanding species in all three habitats. Global changes involving increased resource availability are thus 134 likely to favor large-ranged, nutrient-demanding species, which are typically strong competitors. 135 Declines of small-ranged species could reflect not only abiotic drivers of global change, but also biotic 136 pressure from increased competition. Our study highlights the need to consider the traits of species 137 such as the geographic range size when predicting how ecological communities will respond to global 138 change. reveal systematic temporal turnover of species, despite variable trends in species richness. Large-ranged, nutrient-demanding species are consistently replacing species with 144 small ranges, thus homogenizing vegetation between dominant habitats across Europe. Our cross- continent comparison highlights that such of increase and contribute to directional loss. Our predictions of community change and prioritization 147 of species conservation during the Anthropocene.


Fondation J.-M.Aubert, Champex-Lac, Switzerland
Abstract 123 The direction and magnitude of long-term changes in local plant species richness are highly variable 124 among studies, while species turnover is ubiquitous. However, it is unknown whether the nature of 125 species turnover is idiosyncratic or whether certain types of species are consistently gained or lost 126 across different habitats. To address this question, we analyzed the trajectories of 1,827 vascular plant 127 species over time intervals of up to 78 years at 141 sites in three habitats in Europe -mountain summits, 128 forests, and lowland grasslands. Consistent across all habitats, we found that plant species with small 129 geographic ranges tended to be replaced by species with large ranges, despite habitat-specific trends in 130 species richness. Our results point to a predictable component of species turnover, likely explained by 131 aspects of species' niches correlated with geographic range size. Species with larger ranges tend to be 132 associated with nutrient-rich sites and we found community composition shifts towards more nutrient-133 demanding species in all three habitats. Global changes involving increased resource availability are thus 134 likely to favor large-ranged, nutrient-demanding species, which are typically strong competitors. 135 Declines of small-ranged species could reflect not only abiotic drivers of global change, but also biotic 136 pressure from increased competition. Our study highlights the need to consider the traits of species 137 such as the geographic range size when predicting how ecological communities will respond to global 138 change. 139 140

Significance Statement 141
Vegetation resurveys at intervals of up to 78 years spanning mountain summits, forests and grasslands 142 in Europe, reveal systematic temporal turnover of vascular plant species, despite variable trends in 143 species richness. Large-ranged, nutrient-demanding species are consistently replacing species with 144 small ranges, thus homogenizing vegetation between dominant habitats across Europe. Our cross-145 continent comparison highlights that such gains of species might increase competition and contribute 146 to directional species loss. Our findings inform predictions of plant community change and prioritization 147 of species conservation during the Anthropocene. 148 149

Main text 150
Introduction 151 Long-term studies of changes in local plant species richness do not show systematic evidence of 152 decline (1)(2)(3). However, local richness changes provide only a limited picture of the extent of ongoing 153 biodiversity change, as they do not capture species turnover and changes in community composition 154 over time (4). While human activities have accelerated species turnover beyond background rates (2,155 3), it remains unclear whether the identities of "winner" and "loser" species represent the 156 idiosyncratic local outcomes of drivers of change (e.g., disturbance or climate warming), or whether 157 there are consistent patterns across systems (5). In order to gain a general understanding of why and 158 how plant diversity is changing, we need to ask whether similar types of plant species are consistently 159 lost and gained in communities in different habitats. 160 For plant species, studying temporal turnover in relation to their geographic range size can provide 161 insights on why and how species diversity is changing. On the one hand, the geographic range size of 162 plant species is a key synthetic measure of their ecological profile (6). Range size reflects the ability of 163 species to disperse and colonize (7,8), as well as their niche breadth (9,10) and niche position (11)(12)(13), thus capturing multiple factors relating to a species' vulnerability to global environmental 165 changes. On the other hand, range size links temporal with spatial turnover of species, as communities 166 that lose small-ranged while gaining widespread species become more similar over time (14). 167 Therefore, understanding the link between range size and a species' trajectory over time will not only 168 shed light on why certain species "win" or "lose", but also on the consequences of these shifts for the 169 distinctiveness of plant communities, an important component of biodiversity. 170 Here, we analyze individual trajectories of 1,827 vascular plant species over time in relation to their 171 range size at 141 study sites across three habitats in Europe -mountain summits, deciduous and 172 coniferous forests, and lowland grasslands (Figure 1), using vegetation resurveys spanning intervals of 173 12 to 78 years. Temporal trends in local species richness and drivers of change are known to vary 174 among these habitats, with climate warming increasing local species richness on summits (15,16), 175 eutrophication and changes in management reducing richness in grasslands (17,18), and a 176 combination of these drivers leading to both increases and decreases in richness in forests (13, 19-177 21). We hypothesize that, regardless of the richness trend in a habitat, smaller-ranged species are 178 consistently replaced by larger-ranged species, as environmental changes (such as increasing 179 temperatures, land-use change and eutrophication) alter ecological selection processes in favor of 180 widespread species; species that are expected to be more resilient, more nutrient-demanding and 181 better dispersed (13, 22). Our study explores whether the temporal turnover of species of vascular 182 plants is systematic, and whether it acts to homogenize vegetation between habitats. 183 184

Results and Discussion 185
We found that vascular plant species with larger ranges consistently emerged as winners and those 186 with smaller ranges as losers over time across all three habitats, regardless of trends in species 187 richness. While on mountain summits, species gains were clearly more prominent than species losses, 188 there was substantial species loss in forests and grasslands (Figure 2a). Losses and gains, however, 189 balanced out in forests, whereas in grasslands losses outweighed gains ( Figure 2b). Thus, the average 190 species richness increased on summits, showed no clear trend in forest and decreased in grasslands 191 (Figure 2c and Fig. S1), in accordance with single-habitat studies from each of these habitats (summits: 192 (15, 16), forests: (13, 19) and grasslands: (17, 18)). Despite variable trends in richness, species 193 turnover was systematic. We tested whether species with smaller ranges have been lost preferentially 194 at a study site. Even after accounting for demographic effects (i.e., due to the likelihood that small-195 ranged species are lost simply because of a smaller local population size; see Methods), range size was 196 negatively associated with loss probability in all three habitats, although on summits the association 197 was not statistically clear as the 66% credible interval overlapped with zero ( Figure 3a and Table S2). occupancies below 5% in the baseline survey) from the data (Table S2). We then asked whether 200 changes in site-occupancy of persisting species were related to range size. In all three habitats, 201 persisting species increasing in occupancy had larger ranges on average than species decreasing in 202 occupancy (Fig. S3 and Table S3). This relationship persisted after accounting for species baseline 203 occupancy ( Figure 3b, see Methods). Finally, we compared range sizes of species gained to species 204 lost. In all three habitats, species that were newly gained at a study site had, on average, larger ranges 205 than species lost (Figure 3c, d and Table S4). Together, these findings indicate commonalities between 206 contrasting habitats with respect to the nature of biodiversity change based on species range size. 207 Across habitats, plant species with larger ranges gained ground. The success of large-ranged species 208 could be due to previously limiting resources (e.g., nutrients) becoming more available as a result of 209 global changes such as eutrophication and warming (23,24). A greater availability of limiting resources 210 allows less specialized species to colonize, where larger-ranged species may be more likely to colonize 211 simply because they disperse from more sites. Larger-ranged species may also be more likely to persist 212 because they naturally face a larger gradient in environmental conditions and may thus exhibit a 213 greater niche breadth and phenotypic plasticity, making them more resilient to global changes (10, 22, 214 25). Furthermore, global changes may even favor large-ranged species, as they tend to be species with 215 resource-acquisitive strategies and might therefore benefit more from an increase in resources (12, 216 13, 22). We found support for this hypothesis in our data; species with larger ranges were associated 217 with higher nutrient demands ( Figure 4a) and community weighted means of species niche positions 218 for nutrients indicated community shifts towards more nutrient-demanding species (Figure 4b, see 219 Methods), in accordance with other studies in these habitats (summits: (26), forests: (19) and 220 grasslands: (18)). These findings suggest that a higher prevalence of larger-ranged species, often also 221 more resource-acquisitive species, is likely to exert increased biotic pressure on extant species. 222 In contrast to large-ranged species, small-ranged species tend to be adapted to lower nutrient 223 availability ( Figure 4) and thus are likely to grow more slowly (27), presenting a particular risk of 224 competitive exclusion by faster growing species. The loss of small-ranged species could therefore be a 225 result of the increase in less specialized, more competitive, larger-ranged species (i.e. biotic filtering). 226 Furthermore, small-ranged species tend to have adaptations to the stresses specific to their habitat 227 and therefore possibly a lower tolerance to new types of stress, such as stoichiometric imbalances in 228 resource supply from eutrophication (28). Thus, the decline in small-ranged species could also be due 229 to direct effects of environmental change (i.e. abiotic filtering). Importantly, we can largely exclude the 230 potential explanation that the higher loss probability of small-ranged species is due only to stochastic, 231 demographic effects (Table S2, see Methods). Also, if small-ranged species were simply more prone to 232 demographic fluctuations and therefore had a more variable presence, we would expect comparable 233 range sizes of species lost and gained, which we do not see in the data (Figure 3c and d). Thus, the preferential loss of small-ranged species is likely due not only to demographic stochasticity, but also to 235 aspects of species niche that confer a higher vulnerability to both abiotic and biotic pressures. 236 Despite the congruence across habitats of small-ranged species being replaced by large-ranged 237 species, our results also indicate differences in the effect of range size on temporal species turnover 238 between habitats. On summits, the effect of range size on species loss probability was weakest and 239 not clearly different from zero ( Figure 3a). Moreover, species gained on summits had larger ranges 240 than both persisting and lost species, whereas in forests and grasslands the main distinction was that 241 species lost had smaller ranges than both persisting and gained species (Figure 3c and Table S4). In 242 addition, on summits, species gains dominated and species losses were less important for driving 243 turnover compared with forests and grasslands (Figure 2a and b). These results suggest that the 244 directional turnover on summits in relation to species range size could be mainly due to species 245 differences in dispersal and colonization ability. On summits, warming may allow the colonization of 246 species from lower elevations, which tend to have larger ranges (Fig. S5), while extant species may 247 persist and escape changes in abiotic and biotic filters due to a high variation of micro-habitats (29, 248 30) and a still sparse or less tall-growing vegetation (31, 32). In forests and grasslands, the vegetation 249 is typically denser than on summits. Environmental changes, such as eutrophication or declines in 250 traditional land use, are thus likely to lead, in addition to abiotic changes, to higher biotic pressure (33, 251 34). We hypothesize that a greater relevance of biotic filtering in forests and grasslands could 252 contribute to the more directional loss of small-ranged species in these habitats ( Figure 3). Although 253 any cross-habitat comparison is limited due to inherent differences between habitats, we can rule out 254 that differences in the relationship of range size and loss probability simply arise from evident 255 differences in sampling methods among study sites. The number of plots, plot size, site area and time 256 span between surveys did not change the effect of range size on the probability of loss (see Methods 257 and Table S5). Our results thus support the potential role of indirect, biotic effects of global change in 258 understanding the preferential loss of small-ranged species. 259 Altogether, our results suggest that temporal species turnover has a predictable component based on 260 species range size. Regardless of whether site-level trends show increases or decreases in species 261 richness, larger-ranged species replaced smaller-ranged species. This has at least two implications. 262 First, as sites gain species that are already widespread and lose small-ranged species, cumulatively this 263 may lead to shifts from characteristic, often rare vegetation types to more widespread vegetation 264 types -a form of biotic homogenization (14). Indeed, we found that an average pair of study sites 265 became more similar in species composition and, moreover, that the total species pools of the three 266 habitats became more similar over time ( Fig. S6a and b). Second, small-ranged species may be doubly fewer sites, and because they can also be more vulnerable to being lost within each site, as we have 269 shown here. While the patterns found in our study suggest that the loss of small-ranged species within 270 sites is partially explained by species niches, it remains a future challenge to disentangle how much of 271 this loss is driven by indirect effects due to altered competitive interactions (i.e. biotic filtering) versus 272 direct effects due to environmental changes (i.e. abiotic filtering) in different habitats. Our study 273 demonstrates that even in seminatural habitats, biodiversity is systematically changing and that this 274 change can be predicted by the geographic range size of species. Thus, our results inform predictions 275 of how plant communities will respond to accelerating global change and the prioritization of 276 conservation efforts towards the species that are more likely to be lost. Insights on the relative 277 importance of biotic versus abiotic filtering will be essential when prioritizing measures to reverse the 278 declines of the most vulnerable species in the Anthropocene. 279 280

Materials and Methods 281
Databases. We synthesized data from three databases, each of which is a collation of vegetation 282 resurveys in a specific habitat in Europe. Mountain summits are represented by 52 sites from the 283 Global Observation Research Initiative in Alpine environments (GLORIA, gloria.ac.at, (36)), deciduous 284 and coniferous forests understories by 68 sites from the forestREplot database (forestreplot.ugent.be, 285 (37)) and lowland grasslands by 21 sites from the GRACE database (18) ( Figure 1 and Table S1). At 286 each site, plant communities were surveyed across multiple permanent or quasi-permanent plots in 287 either natural vegetation (summits) or semi-natural vegetation (forests and grasslands) at two points 288 in time (baseline and resurvey, further details available in (18,36,37)). The median time spans 289 between surveys were 14, 42 and 34 years for summits, forests and grasslands, respectively (Fig. S7a). 290 In forest and grassland surveys, the median number of plots per site was 43 and 36, and the median 291 size of plots was 400 m 2 and 25 m 2 , respectively ( Fig. S8a and b). Summits were always resurveyed in 292 eight spatial sections that together covered the entire area from the highest summit point to the 293 contour line 10 m in elevation below this point in a pie slice shape. The median summit area was 0.25 294 ha. In forests and grasslands, the median study area was 1,700 ha and 1,000 ha, respectively ( Figure  295 S8b). 296 Species data. Taxonomy. We accounted for within-and among-study variation in taxonomy by 297 determining the accepted species name for each species using the Global Biodiversity Information 298 Facility's (GBIF) backbone taxonomy (gbif.org). Harmonization thus ensured no double-counting of our data comprises 1,827 accepted vascular plant species (see Data Table 1 at 301 figshare.com/s/b37f6167b13ad5da9e9c). 302 Range size. We estimated species range sizes as area of occupancy (AOO) (6) using all point 303 occurrence records of the species in GBIF (gbif.org, May 2020; (38)). After excluding incomplete, 304 impossible and unlikely coordinates (e.g., country centroids) (39), there were c. 131 million 305 geographically referenced records available for the species in our database. Records were aggregated 306 to a hexagonal grid (ISEA3H) at a spatial grain of 10.7 km 2 (40), where the number of cells that a 307 species occupies on this grid represents its AOO estimate (see Data Table 1 for species AOO estimates 308 and GBIF urls). The species with the largest AOO in all three habitats were Achillea millefolium and 309 Trifolium repens (both with ca. 1.1x10 6 km 2 ), the species with the smallest AOO were the highly 310 endemic Draba dolomitica (c. 11 km 2 ) on summits, Galium abaujense (c. 21 km 2 ;endemic to the 311 Carpathians) in forests, and Pentanema germanicum (c. 503 km 2 ; critically endangered in Germany 312 and Austria (41)) in grasslands (Fig. S9). For plant species in Europe, range sizes calculated from GBIF 313 correlate strongly with expert drawn range maps but are available for many more species (13). 314 However, it is important to note that AOO ranges differ from expert maps, which measure species 315 extent of occurrence (EOO), in that they do not include areas that are unoccupied by species. Thus, 316 species with disjunct distributions, e.g., orchid species that occur throughout Europe but only in very 317 fragmented, well-conserved habitat, can have a very small AOO but a large EOO. AOO is therefore a 318 markedly better representation of species population sizes and differences related to habitat use and 319 species niche than is EOO, and provides a general measure of species vulnerabilities to stochastic and 320 directional threatening processes (6). 321 Occupancy. Measures of plot-level species abundance varied across studies (e.g., frequencies, 322 percentage cover, and categorical cover-abundance scales) and were often not available if only 323 species presence/absence was recorded. In order to estimate species abundance in a consistent way, 324 we estimated species occupancy at the spatial scale of a study. We therefore divided the number of 325 plots (grasslands, forests) or sections (summits) a species occupied at a given study site by the total 326 number of plots/sections in that study. This was done separately for the baseline survey and the 327 resurvey. Occupancy has been shown to correlate strongly and positively with abundance at local to 328 regional scales (42,43). 329 pseudo-gains and losses that can be inflated for rare species (44, 45). We account for this bias by 335 adjusting for species baseline abundances, which is strongly correlated with any such bias (46), as 336 explained below. 337 Analysis. The brms package (47) in R was used for all statistical analyses. R code for all analyses and 338 data visualization is available on figshare at https://figshare.com/s/b37f6167b13ad5da9e9c. 339 Species gains and losses. Using species trajectories we quantified the number of lost and gained 340 species on the spatial scale of a study site (Extended Data Table 2). The highest losses (126 species) 341 occurred in Hungarian forest-steppe landscapes, the highest gains (102 species) occurred in 342 acidic/mesic oak woods in the Czech Republic. We assessed changes in species richness (i.e., the 343 change in the total number of species per study site) by calculating the difference, d, between species 344 richness in the resurvey (t2) and species richness in the baseline survey (t1). Although species richness 345 at a given time period will be affected by sampling effort, d is not because it is a relative change in 346 species richness with sampling effort being the same for both time periods (baseline surveys vs. 347 resurveys). For each habitat, we modelled d using a Gaussian distribution to compute the posterior 348 distribution of the expected value of d (Fig. S1). 349 Probability of loss. We estimated the effect of species range size on the probability that a species 350 being present at the baseline survey is lost from a study site by the time of the resurvey. The effect of 351 range size can be confounded by species baseline occupancy if small-ranged species also tend to have 352 a lower abundance at a study site. Species with small population sizes are more likely to be lost owing 353 to 1) stochastic demographic processes and 2) an observer error, where rare species are more likely to 354 be overlooked in resurveys. Therefore, we tested first for a positive range size -site occupancy 355 relationship in our data (see Methods below). Range size and occupancy were not related on summits 356 and weakly positively related in forests and grasslands (Fig. S2). To estimate the effect of range size 357 that is not due to demographic effects, we statistically controlled for variation in species baseline 358 occupancies by including it as a covariate in our model (13). Furthermore, species with small ranges 359 may be disproportionately vulnerable at low abundances. This could be the case if range size covaries 360 with specific traits, such as, for example, height, where small plants would be expected to be more 361 vulnerable than tall plants at low site occupancy. To account for this possible further confounding 362 effect, we also include an interaction effect between range size and occupancy in our model. Finally, 363 the effect of species occupancy on species loss probability is likely to vary with the number of plots per 364 study site. For example, a species with 10% occupancy in a study of 10 plots, is more likely to be lost 365 than a species with 10% occupancy in a study of 100 plots. We therefore allow the effect of occupancy 366 to vary by study site.
Our model thus predicts a Bernoulli indicator variable that a given species was lost or persisted ( ) 368 with two fixed effects ( for range size ( ) and for occupancy ( ), where both and were 369 log10-transformed and scaled within habitats to have a mean of zero and a standard deviation of one) 370 and an interaction effect between the two fixed effects ( ). We allowed the intercept and the effect 371 of occupancy to vary by study site ( [ ] and , [ ] , respectively). Also, we included species as 372 an additional crossed varying effect ( [ ] ), since many species occur in more than one study site. 373 We ran this model for each habitat (see Table S2 for model R syntax, sample settings and convergence 374 diagnostics). The resulting model in mathematical form is: 375 As a further means to test whether demographic effects confound estimates of , we ran the same 377 model but excluded rare species (with site occupancies below 5%) from our data (Table S3). Since we 378 only had data on the species that were newly gained at a study site but not on all those that tried to 379 colonize, we were not able to directly calculate probabilities of gain in relation to range size. 380 Occupancy trends of persisting species. Here we only evaluate species that have persisted over time, 381 since species lost and gained necessarily decrease and increase in occupancy, respectively. We first 382 tested whether persisting species that increased in occupancy at a study site have on average larger 383 range sizes than persisting species that decreased in occupancy at a study site. We therefore 384 predicted range size (log10-transformed) with the categorical variable "decrease/increase" 385 Since changes in occupancy may depend on species baseline occupancy (e.g., species with a higher 387 baseline occupancy could be more likely to increase in occupancy due to a higher propagule pressure), 388 we also estimated the effect of species range size on the probability that a persisting species increases 389 in occupancy, controlling for variation in species baseline occupancies. For this logistic model, we 390 recoded the difference in occupancy at the resurvey and the baseline survey (d) into a binary variable 391 with d > 0 being "1", d ≤ 0 being "0" (ℎ ) and predicted ℎ with range size, including baseline 392 occupancy as a covariate. Since baseline occupancy ranges from 0 to 1, species with an occupancy of 1 393 cannot increase in occupancy. These species were therefore excluded from the model. The model in 394 math form is:

398
, where parameters are defined as in the model for species loss probability. However, we did not 399 include the interaction effect between occupancy and range size ( ) in this model, as a potentially 400 greater vulnerability of small-ranged species at low occupancy is likely to not be very relevant to 401 explain increases in occupancy (see Table S3

408
In order to test whether range sizes of species gained differ from those being lost, we calculated the 409 posterior difference in mean range size between gained and lost species in each habitat. Since the 410 posterior difference between gained and lost species is in the log10-scale, this gives a ratio of range 411 size of species gained/lost after back-transformed to the original scale (see Table S4 for model R 412 syntax, sample settings and convergence diagnostics). 413 Range size and nutrient demand. We used Ellenberg's indicator values for nutrient (N-number) to 414 approximate species niche position for nutrients (27,48,49). These values describe each species' 415 niche position on a scale from 1 to 9 (adapted to unproductive, nutrient-poor soils) to 9 (adapted to 416 fertile soils). We obtained N-numbers from sci.muni.cz/botany/juice/ELLENB.TXT and harmonized the 417 taxonomy with our data. If an accepted species had more than one N-number (either due to synonyms 418 or subspecies, e.g., Melampyrum pratense ssp. paludosum has an N-number of 1, while Melampyrum 419 pratense has an N-number of 2), we calculated the average. 1,297 species of the 1,827 species in our 420 data also had N-numbers (71%). For the species in each habitat, we calculated Pearson's correlation 421 coefficient between range size (log10-transformed and scaled) and N-number (scaled).

430
To gain insight into how much of the change in CWM-N is due to changes in species occupancy or 431 species composition, we also calculated community unweighted means by simply averaging N-432 numbers across species at a study site for both the baseline survey and resurvey, and tested for 433 changes over time using the same model as above. The comparison of weighted and unweighted 434 means showed that in forests and grasslands, the clear shift towards more nutrient-demanding 435 species was largely due to changes in species composition, while on summits the much weaker shift 436 was due to changes in species occupancy (Fig. S4).

443
Relationship between mean range size and elevation. We tested whether montane species from lower 444 elevations have larger ranges than alpine ones. Therefore, we regressed mean range size ( , 445 averaged across species occurring at a summit site at the baseline survey) against summit elevation 446 Effects of site-characteristics on the effect of range size. While the above model for species loss 449 probability provides estimates for range size-effects within habitats, different sampling methods 450 between habitats make it difficult to compare effect estimates across habitats. Summits are inherently 451 limited in size and were surveyed in always eight sections, while forest and grasslands areas were 452 sampled with differing number of plots of different sizes across differently large study areas (Fig. S7  453 and S8). Moreover, time intervals between surveys varied among habitats, with the shortest tested whether the effect of range size, , changed with plot number, plot size, site area (log10-456 transformed) and survey interval ( , 2 , , and # , respectively) We tested this in forests, where we 457 had most study sites and sampling varied the most, by including interaction effects between range size 458 and sampling characteristics (there was no strong collinearity between sampling characteristics (Fig. 459 Table 1) are available on figshare at https://figshare.com/s/b37f6167b13ad5da9e9c. Species composition 472 data for grasslands is available from published literature compiled in (18); for forest and alpine summits these 473 data are available upon request from forestreplot.ugent.be and gloria.ac.at, respectively.

Acknowledgements 476
This paper is an outcome of the sREplot working group supported by sDiv, the Synthesis Centre of the German

497
Tomaselli and numerous helpers for data originating from the GLORIA network.

50.
A. Baselga, C. D. L. Orme, betapart: an R package for the study of beta diversity. Methods Ecol. 602 Evol. 3, 808-812 (2012). 603  Figure 2. Species losses and gains vary across habitats. a, Species gains (white) and losses (grey) at each study site (numbers stacked, each bar represents a study site). b, Relative frequency (density) of the number of species lost and gained across sites. c, Density across study sites of the difference in species richness (S) between the baseline survey and resurvey. Dotted horizontal line represents zero change in S. Colours (blue, green, yellow) refer to habitats as in Figure 1. Posterior distribution of the mean difference in S is shown in Supplementary  Figure 1. Range size of species gained / lost d Figure 3. Consistent replacement of small-by large-ranged species across habitats. Posterior distribution of the effect of range size on a, the probability (Pr) of a species being lost at a study site and b, the probability (Pr) of a persisting species increasing in occupancy at a study site, after having accounted for demographic effects (see Methods). c, Posterior distribution of the mean range size of gained, persisting, and lost species. d, Comparison between the mean range sizes of species gained and lost, derived from the posterior distributions in c (persisting vs gained/lost comparison in Supplementary Table 4). Point and lines in a -d are the median and its 66% and 95% credible interval. Dotted vertical line in d represents no difference in mean range size. In a and b, range size was log10-transformed and scaled to have a mean of zero and a standard deviation of one, effect estimates (x-axis) are in the logit scale. Model summaries and sample sizes for panels a-d are in Supplementary  Table 2 Figure 4. Species with larger ranges tend to have higher nutrient demands and communities shift towards species with higher nutrient demands over time. a, Relationship between species range size and Ellenberg indicator values for nutrients (N-numbers) across species in each habitat. Line and transparent ribbon represent the mean regression line and 95% credible interval, ρ is the estimated correlation coefficient, σ is the standard deviation of ρ. b, Boxplot and density plot of the community weighted mean (CWM) niche position for nutrients (N-number) at the baseline survey (t 1 ) and resurvey (t 2 ). CWM is weighted by species occupancies at the study site. Triangles represent mean values. δ is the mean (pairwise) difference, σ is the standard deviation of δ.

Grassland
Forest Summit -20 -10 0 10 Mean difference in species richness Baseline occupancy (log10 and scaled) Range size (log10 and scaled) Fig. S2. Relationship between species range size and baseline site-occupancy accounting for the structure of our data. Colors present study sites, transparent dots present species, transparent lines represent the relationship between range size and site occupancy within a single study site, black straight line is the mean regression line across study sites resulting from a linear varying effect model with regression coefficients (slope and intercept) allowed to vary by study site, black dashed line is the mean regression line from a general additive model without varying effects. β is the slope and σ is the standard deviation of β from the linear varying effect model.   Table S2. Summary of the model predicting species loss probability with species range size and baseline abundance (Fig. 3a). Model syntax, sampling settings, parameter estimates, their standard deviation (sd) and 95% credible interval (CI). Rhat is the Gelman-Rubin convergence diagnostic, bulkand tail-ESS are the number of independent samples (i.e. effective sample sizes). Model includes species present at the baseline survey. Also, we ran the model excluding rare species (with a site-occupancy below 5% in the baseline survey) in forest and grassland to test for robustness of the range-size effect within these habitats.  Table S3. Summaries of models for occupancy trends of persisting species in relation to range size (Supplementary Figure 3 and Fig. 3b Model syntax, sampling settings, parameter estimates, their standard deviation (sd) and 95% credible interval (CI). Rhat is the Gelman-Rubin convergence diagnostic, bulk-and tail-ESS are the number of independent samples (i.e. effective sample sizes). Models only includes persisting species. "∆ to decreasing" presents the posterior difference in estimated mean range size between increasing and decreasing species. Model on probability of increasing controls for species baseline occupancy, and excludes species with a baseline occupancy of 1 (see Methods).  Table S4. Summary of the model predicting species range size with species trajectory (Fig. 3c and d). Model syntax, sampling settings, parameter estimates, their standard deviation (sd) and 95% credible interval (CI). Rhat is the Gelman-Rubin convergence diagnostic, bulk-and tail-ESS are the number of independent samples (i.e. effective sample sizes). Model includes all 1,827 species in our database. "∆ to gained" presents the posterior difference in estimated mean range size between lost/persisting and gained species. Contrasts are calculated as differences between the posterior distribution of mean range sizes of trajectories (as opposed to Figure 3c and d, model estimates are here in the log10-scale).  Table S5. Summary of the model testing for effects of sampling methods on the effect of range size on species loss probability. Model syntax, sampling settings, parameter estimates, their standard deviation (sd) and 95% credible interval (CI). Rhat is the Gelman-Rubin convergence diagnostic, bulk-and tail-ESS are the number of independent samples (i.e. effective sample sizes). Model is on forests, as forest study sites are most numerous and sampling characters (i.e. site areas, plot sizes/numbers and time intervals) varied here the most.