The cold-drought tolerance trade-off in temperate woody plants constrains range size, but not range filling

Interspecific differences in plant species’ ranges are shaped by complex mechanistic interactions, which have so far remained largely beyond the reach of comprehensive models and explanations. Previous attempts to find underlying mechanisms by examining physiological tolerances to cold and heat separately have yielded contradictory results. Here we test the hypothesis that, instead of examining single stressors, abiotic stress tolerance syndromes that involve trade-offs between multiple abiotic stressors (namely drought, cold, waterlogging and shade), will provide reliable explanations. We compiled a dataset of actual range size and range filling (the ratio between actual and potential species range) as range metrics for 331 temperate woody plants species from Europe and North America. Tolerance syndromes were expressed as two PCA axes. One axis reflects a drought-cold/waterlogging tolerance trade-off (cold/wet-drought trade-off), the second axis represents a shade tolerance spectrum. Phylogenetic generalized linear mixed models were used to model the range metric-tolerance axes relationships using latitude as an additional main effect, and phylogeny and plant functional type as random effects. Actual range scaled negatively with the cold/wet-drought tolerance trade-off axis, mostly independently of latitude and continent. Thus, cold/wet-tolerant species had the largest ranges and drought tolerant species the smallest. The sign (−) of the relationship was independent of phylogeny and plant functional type. In contrast, range filling depended on latitude. However, deciduous and evergreen species displayed different distributions of range metrics and tolerance syndromes. No significant relationships with the shade tolerance spectrum were found. Our findings demonstrate that the cold/wet-drought trade-off partly explains interspecific range size differences. However, this trade-off did not explain range filling. We also showed that fundamental adaptations of species also significantly influence range sizes – stress avoidance through the deciduous habit also explained interspecific differences in range size.


57
The identification of the factors shaping species range size and filling is a major focus of 58 biogeography (Brown, 1984;Gaston et al., 2009). In recent years, a number of theories aimed at 59 explaining species distribution patterns at the global scale have been developed. For example, the 60 mid-domain effect predicts that species close to the equator have larger distribution ranges 61 (Colwell & Hurtt, 1994), whereas Rapoport´s rule assumes the opposite (Stevens, 1989). Several 62 mechanistic hypotheses either addressing extrinsic factors (e.g. the climate variability hypothesis 63 (Stevens, 1989) to explain Rapoport's rule) or intrinsic factors (e.g. the dispersal (Hanski et al. 64 1993) or niche breadth (Brown, 1984) hypotheses) have been proposed (reviewed in Sheth et al., 65 2020). Most of the hypotheses formulated so far to explain interspecific differences in range size 66 have emerged from zoology (Fine, 2015). In comparison, this topic has been little investigated in 67 plant species (Sheth et al., 2020), for which the determinants of range size, and its global 68 variation, remain elusive. woody species tend to have larger ranges at higher latitudes (Morin & Chuine, 2006). 81 All these results are consistent with Rapoport´s rule, which predicts larger ranges and 82 lower species richness at higher latitudes, possibly due to high latitude species being more 83 tolerant of more variable environmental conditions (e.g., Morueta-Holme et al., 2013), or to an 84 increased frequency of pioneer species with broader niches at higher latitudes (Morin & Chuine,85 For example, in the Americas, woody species' range size has a bimodal distribution in relation to 87 latitude, being largest in both north temperate and tropical areas (Weiser et al., 2007). A similar 88 bimodal distribution has been observed for range filling in European (Svenning & Skov, 2004) 89 and North American tree species spanning sub-tropical to boreal climates (Seliger et al., 2021). 90 Thus, although a positive range size-latitude relationship is well documented for plants, it is not 91 without exceptions and a mechanistic explanation of this pattern is still missing. 92 Morin & Chuine (2006) proposed that the proximate driver behind Rapoport's rule is 93 abiotic stress tolerance. In particular, they argued that intrinsic differences in abiotic stress 94 tolerance between species, and thus in their ability to persist under given resource regimes, might  abiotic stress tolerance syndromes, as summarized by the Stress Space axes, and species actual 141 range sizes and range filling. We decided to consider both of these range metrics because they 142 reflect different aspects of species' ranges. Actual range size includes historical legacies (e.g. for 143 temperate species, the degree to which they have been able to expand since the Last Glacial dispersal, the regeneration niche theory (Grubb, 1977), also requires that a species is able to  In this study, we examined the relationships between temperate woody species range   proxy of realized range size for European species set a reasonable assumption that is strongly 192 needed to make the results between continental species sets comparable.

197
Potential range size and range filling calculation 198 To estimate species' potential range sizes, we used two presence-only models -i.e., Bioclim were available across tolerance scales. Cold tolerance data were extracted from USDA plant 251 hardiness data and represent species-specific averages gathered from multiple sources (see 252 Laanisto & Niinemets, 2015 for further details). In this respect, cold tolerance data do not show 253 any circularity with Tmin obtained from WorldClim as it is not based on any spatially-explicit 254 information, and being an average species-specific estimate, it is not affected by a species range 255 size (i.e. species with larger ranges are more likely to encounter extreme mean temperatures).

310
North America: slope = -0.18, R 2 = 0.14, p < 0.01, n = 201) (Fig. 1 a,c). A negative relationship 311 between actual range and cold-drought tolerance trade-off axis was also observed across the 312 considered quantiles of the response variable, but with differences between the species from the 313 two continents. For European species, the quantile regressions were mostly significant at average 314 to high values of actual range (Fig. 1 a). For North American species quantile regressions were 315 all significant, except the one fitted at the lowest quantile (Fig. 1 c). 316 Range filling residuals scaled negatively with species positioning along the cold-drought 317 tolerance trade-off axis only for European species (slope = -5.66, R 2 = 0.07, p < 0.01, n = 130) 318 and, as for actual range, the quantile regressions were mostly significant at average to high 319 values of range filling (Fig. 1 b). No significant relationship was found between range filling 320 residuals and the cold-drought tolerance trade-off axis for North American species, and the 321 relationship was not significant at any considered quantile (Fig. 1 d). lines represented not significant relationships. Sample size (n) is shown in panels (a,c) and applies also to 329 the relationships involving range filling as the response variable. The color gradient reflects the 330 progression from cold/wet-tolerant to dry/warm-tolerant species along the cold-drought trade-off. 331 The negative relationship between actual range and the cold-drought tolerance trade-off 332 axis was not affected by including latitude as an additional main effect (Table 1), and the cold-333 drought tolerance trade-off axis effects were always greater than that of latitude for all data 334 pooled (see Appendix S5, Table S5.3), and for each continent considered separately ( Table 1). 335 However, some differences between continents were detected. The cold-drought tolerance trade- American species sets having no species in common, we explored whether differences between 345 continental species sets could be driven by differences in terms of genera composition.

346
Out of 106 genera in our entire dataset, there were 29 genera in common between the two 347 continents. These genera contributed 70% of the species included in the entire dataset. When the 348 analyses were repeated after removing genera that were unique to one or other continent, the 349 result remained significant for North American species, but the relationship between range size 350 and the cold-drought tolerance trade-off axis changed from being marginally significant to 351 become highly significant for European species as well (R 2 = 0.12, see Appendix S5, Table   352 S5.4). This indicates that differences in results between continents are partly driven by 353 differences in genera composition of their constituent species.

354
There was no significant interaction between the cold-drought tolerance trade-off axis 355 and latitude in any model involving actual range as the response variable for European or North

359
The differences in results between continents were more pronounced for range filling 360 ( Table 1) (Table 1). In addition, the spline correlograms (see Appendix S6, Fig. S6.8) Fig. 2 a-f). As a general trend, deciduous angiosperms have larger actual range and range 397 filling values (p ≤ 0.05; Fig. 2 a,b,d,e), and they occupy the cold/wet side of the cold-drought 398 tolerance trade-off (i.e. more negative values along the cold-drought tolerance trade-off axis) 399 (Fig. 2 c,f), compared to the other plant functional types. However, multiple comparisons 400 sometimes differed between continents (Fig. 2 a-f). American woody species. Sample size and data points for each plant functional type are shown. *** 407 indicates significant differences between plant functional types (Kruskal-Wallis, p < 0.05). '*' indicates 408 marginal significance (p = 0.05). Only significant and marginally significant differences are shown. 409 Multiple comparisons between groups were carried out using the Dunn's test and Holm correction for 410 multiple testing. 411

413
Our results show that the cold-drought tolerance trade-off axis (hereafter cold-drought trade-off) 414 partly explains interspecific differences in actual range size of temperate woody plants (Fig. 1). 415 In particular, we found large-ranged species at the cold/wet tolerance end of the trade-off axis, 416 and small-ranged species at the drought tolerance end. Despite some nuanced variation discussed 417 below, this result was independent of continent, latitude, plant functional type and phylogeny, 418 indicating its generality. In contrast, range filling showed different patterns in Europe and North  Table 1). Consistent with this, large ranged, cold-tolerant North American tree species are 426 known to be generally absent from regions that are consistently warm and moist, such as the 427 southeastern regions of the continent (Pither, 2003). Palaeoecological records also provide 428 evidence for rapid northward range shifts in North American large-ranged trees after the latest weaker than for North American species (Table 1). This difference is caused by differences in 445 genera composition between the continental species sets (see Results section), and by a cluster of 446 European species with intermediate positions along the cold-drought trade-off having large 447 ranges (Fig. 1 a). across 38 European tree species could be due to covariation between seed mass and other factors 502 (e.g. drought tolerance), but further analysis including a larger number of species would be 503 needed in order to test this claim. 504 two continents, that contrarily to actual range, did not depend on differences in genera between 506 continents (see Appendix S5, Table S5.4). European species' range filling was driven by a 507 positive interaction between latitude and the cold-drought tolerance trade-off, while North 508 American species' range filling was influenced only by a negative effect of latitude. This European species ranges also occur outside Europe (e.g., species with a Eurasian distributions) 516 and this might have influenced model's projections (see Appendix S3). According to the above, 517 we therefore suggest that the interaction between these contrasting continental features and 518 intrinsic drivers, such as dispersal syndromes, might determine continent-specific levels of range 519 filling.  (Table 1). However, the distribution of range size, range filling and positioning 523 along the cold-drought trade-off axis differed between PFTs (Fig. 2 a-f). This probably explains 524 why PFTs generally accounted for more of the random effects variance (even though the amount 525 was generally low) for range metrics compared to phylogeny (Table 1) hemisphere, deciduous species tend to be more frequent in cold climates compared to evergreen 535 broad-leaved angiosperms, despite some overlap between PFTs at almost all latitudes (Zanne et 536 al., 2018). Conversely, while adaptations to tolerate drought closely match the distribution of 537 evergreen species, this is not always true for deciduous species (e.g. see Kunert et al., 2021), 538 suggesting that drought does not always limit deciduous species spatial distribution.

539
Deciduousness per se has been the common explanation for the ability of deciduous 540 species to colonize either cold or dry environments, as it is an adaptation that permits avoidance 541 of unfavorable environmental conditions. This explains why deciduous angiosperms in our 542 dataset have larger ranges than other PFTs, and are mostly cold-tolerant. However, if 543 deciduousness is also a successful drought avoidance strategy that could drive to large ranges in 544 drought tolerant species as well, the relationship between range size and the trade-off axis should 545 be not significant rather than negative. We can reconcile this by considering that woody species  To summarize, while we recognize that adaptations to very low temperatures are 555 complex, and involve a combination of avoidance and tolerance strategies together with 556 acclimation (Schubert et al., 2020), we argue that large range sizes at the cold tolerant end of the 557 trade-off axis is made viable because of the deciduous habit in temperate woody species. In this 558 instance, stress avoidance is more important as an asset than stress tolerance. However, we also 559 observed a greater overlap, in both range size and positioning along the trade-off axis, between 560 freezing-tolerant North American gymnosperms and deciduous angiosperms than with evergreen 561 angiosperms (Fig. 2 d,f). Thus, we do not exclude the additional role for freezing tolerance in 562 guaranteeing large ranges. 563 Our results demonstrate that the cold-drought tolerance trade-off partly explains interspecific 565 differences in range size across temperate woody plant species and that this relationship is 566 largely independent of latitude and consistent with woody species biogeographical histories in 567 the considered continents. Notably, our findings also suggest that accounting for species' abiotic 568 stress tolerance towards multiple stresses can reconcile macroecological and macrophysiological 569 theories aimed at explaining range size differences among woody plant species, supporting a 570 previous hypothesis by Morin & Chuine (2006). However, our results concerning the impact of 571 abiotic stress tolerance on range filling were inconclusive, suggesting that other factors not 572 studied here and that covary with latitude and/or abiotic stress tolerance syndromes are the main 573 drivers of range filling in temperate woody species. Finally, our results also demonstrate the 574 importance of considering trait syndromes, reflected here as differences between plant functional 575 types, in clarifying the ways in which species' adaptations influence broad interspecific 576 differences in range size in woody plant species in relation to abiotic stress tolerance.