2.1 Account for interrelationships among traits

Traits reflect complex patterns of co-adaptation, allocation trade-offs, and by-products of selection—the so-called trait syndromes (Mauro & Ghalambor, 2020; Salguero-Gómez et al., 2016; Stevens et al., 2014)—such that covariations among traits can obscure the mechanisms underlying climate-induced range shifts. For example, body size is viewed as a master trait or the ‘Swiss army knife’ trait that is integrative of a large number of physiological, life history, and behavioral processes (Box 3; Figure 2). Therefore, even if body size can be an efficient predictor of range shifts, it provides only weak mechanistic understanding owing to trait covariations (Visakorpi et al., 2023), especially as these trait syndromes may be inconsistent across taxonomic groups. For example, small-bodied species are usually associated with higher fecundity and shorter generation time among terrestrial animals (r-strategists) but not necessarily among aquatic animals (Winemiller & Rose, 1992). Trophic position increases with body size in fishes but not in mammals (Romanuk et al., 2011; Tucker & Rogers, 2014), and plants and animals appear to achieve demographic resilience through different combinations of life history traits (Capdevila et al., 2022). In this context, when pooling all taxonomic groups together, it is not entirely surprising that conflicting relationships between range shifts and these traits are frequently reported, sometimes in opposition with prevailing hypotheses (MacLean & Beissinger, 2017).

These results caution against using a single trait to clarify the mechanisms of range shifts, especially across diverse taxonomic groups. Instead, detailed hypotheses and methods about range shift–trait associations must be specified. First, it is important to implement a systematic and ecologically defensible approach of trait selection with clear links to range shift processes. Second, formal tests of hypotheses would benefit from being developed within a methodological framework addressing the interrelationships among traits. For example, advanced multivariate regression techniques (e.g., partial least square regression, geographically weighted regression, hierarchical partitioning, and structural equation modeling) lend themselves to understanding the complex relationships among a (very) large number of traits and their effects on ecological processes that can also take latent and non-stationary variables into consideration (Carrascal et al., 2009; Graham, 2003), thus providing useful tools to identify the potential mechanisms underlying range shifts (e.g., Bertrand, 2019; Bertrand et al., 2016; Pöyry et al., 2009).

2.2 Test stage-specific range shift-trait associations

Range shifts are by essence temporally dynamic and can be conceptualized as consecutive but interrelated stages, composed at the trailing edge of (i) decline in performance, (ii) population decrease, and (iii) local extirpation, while at the leading edge, it can be composed of (i) dispersal (including both passive and decision-mediated active dispersal), (ii) population growth, and (iii) persistence (Bates et al., 2014; Estrada et al., 2016). This suggests that developing and testing stage-specific range shift–trait association expectations could improve our ability to develop a mechanistic understanding of the drivers of range shifts that explicitly consider both the spatial and temporal components of range shift processes. For instance, different facets of ecological generalism or specialism may capture different stages of the range shift process, which may lead to conflicting hypotheses (and weak cross-study inferences) if the temporal dynamic of the range shift is ignored. Ecological generalism is often correlated with range size, and thus expected to be intrinsically linked with the evolution of dispersal ability over long timescales (Alzate & Onstein, 2022; Lancaster, 2022). Conversely, specialist species may have evolved better habitat-selection abilities (Jacob et al., 2018) and may therefore better track climatic conditions over short timescales; a hypothesis that has recently gained traction in contemporary climate-driven range shifting moths, butterflies, and birds at their leading edge (Hällfors et al., 2023). On examining coral-reef fish species, Monaco et al. (2020) also demonstrated that dispersal capacity contributed to range expansions only in the dispersal stage of redistribution [stage (i) above], whereas resource-use breadth was more important during the growth and establishment stages [stages (ii) and (iii) above]. As such, thermal specialists may display a higher dispersal propensity but being a generalist along other niche dimensions (e.g., tolerance to drought or generalist diet) may still be expected to facilitate dispersal and successful establishment. Understanding range shifts through the lens of traits might therefore benefit from looking at how traits act sequentially along the range shift pathway, and potentially interact beyond what might be expected from their additive effects (e.g., ‘extinction-promoting traits’; Davies et al., 2004).

2.3 Consider nonlinear range shift–trait associations

Finally, it is important to note that many associations between traits and ecological processes are intrinsically nonlinear. In the context of species redistribution, the unimodal allometric scaling of dispersal and maximum speed with body size (Hirt et al., 2017; Stevens et al., 2014) or the fact that even species with moderate dispersal abilities may be able to keep up with the velocity of isotherm shifts (Schloss et al., 2012; Urban et al., 2013) could translate into an asymptotic relationship between range expansion and body size. Likewise, given that the thresholds leading to extirpations are not expected to be reached for all species (e.g., Pinsky et al., 2019), range contractions might only be expected when temperatures greatly exceed thermal tolerance limits. Yet, hitherto, most studies have used linear responses to capture relationships between range shifts and species traits. Using models carefully informed by theory may therefore represent new opportunities to reveal the contribution of species traits to range shifts (Beissinger & Riddell, 2021; Stenseth & Mysterud, 2002).

3.1 Integrate traits that capture the mechanisms of range shifts

Although in some circumstances, the same suite of traits may be useful to predict both range contractions and expansions (although often in the opposite direction), the assumption that range expansions and contractions are simply the two sides of the same coin is likely to be an oversimplification (Table 1). Yet, many of the same traits have been used with the same frequency in range shift–trait studies, irrespective of the range position (trailing edge, centre, or leading edge; Box 2; Table 1). For example, studies remain dominated by traits related to range size, habitat specialization, and preference, whose underlying mechanisms might be sometimes difficult to interpret (as discussed above for body size; Box 3). In comparison, behavioral and physiological traits have been less explored. This reflects the scarcity of databases compiling organismal traits (measured on individuals) over distribution-based traits (estimated from species distributions) and highlights opportunities for new trait integration as data on physiological, dispersal, and behavioral traits become increasingly available (e.g., Bennett et al., 2018; Herberstein et al., 2022; Lososová et al., 2023). This coupled with new methodological approaches for trait imputation to fill in data gaps (e.g., Thorson et al., 2023), as well as transcriptome databases to detect significant gene–trait associations (e.g., Primmer et al., 2013), may help researchers to integrate traits that have been less explored but that are more likely to capture the mechanisms underlying range shifts.

Because of the scarcity of trait data, studies also often used markedly different proxies or estimates per trait, which may make findings on the importance of traits incomparable across studies. For instance, habitat breadth has been estimated using many different indices, including the number of habitat types (e.g., Powney et al., 2015), the coefficient of variation across habitat classes (e.g., Platts et al., 2019), and multivariate indices based on continuous environmental gradients (e.g., Comte et al., 2014). Yet, it is unclear whether these indices, estimated from a variety of variables and across different spatial resolutions, are inter-comparable and approximate the same hypothesized process, which makes it difficult to draw generalizations across systems and studies, either qualitatively or quantitatively through meta-analysis. Furthermore, most studies conduct cross-species comparisons using only adult traits, despite the fact that traits can vary across a species' lifespan and life history stages (e.g., ontogenetic niche shifts); this in turn can have direct consequences on our ability to test the role of traits. For example, the fact that thermal tolerances and plasticity can vary through ontogeny (e.g., plants: Caron et al., 2021; fishes: Dahlke et al., 2020; insects: Weaving et al., 2022) suggests that traits measured on non-adult life stages may be complementary and thus necessary to explain the rates of species redistribution. Likewise, traits can vary across species ranges for a variety of reasons (e.g., predation risk can alter the temperature dependence of life history traits: Luhring et al., 2018), and trailing edge populations that are often small and isolated often display unique intraspecific phenotypic adaptations to local conditions that can increase their resilience to climate change (Hampe & Petit, 2005). Linking range shifts to traits measured at specific range positions represents a promising path of inquiry. Although widely available and easier-to-get traits may prove sufficient for certain conservation applications (Gallagher et al., 2021), supporting a more mechanistic and predictive science of range shifts will likely require spending more time on difficult-to-get and time-consuming traits, including renewed considerations for cross-studies standardization and incorporation of intraspecific variability (e.g., across life stages or at different range positions). We do not suggest that there is a dichotomy between “bad” and “good” traits but rather that trait selection must be informed by strong a priori hypotheses based on the taxonomic and geographic scope of the study and the subsequent analytical framework developed using informed models accounting for (co)variation among traits.

3.2 Consider evolutionary potential alongside traits

An additional consideration is that species trait means and variances are not fixed but instead can change through evolutionary adaptation (Diamond, 2018; Donelson et al., 2019; Hoffmann & Sgrò, 2011). While recognizing that evolutionary potential can hardly be classified as a “trait,” by facilitating adaptive responses to climate change, trait evolution can strongly influence species' range shifts and could thus be more often considered alongside traits. A classic example is the evolution of dispersal ability at expanding range limits due to spatial sorting of the most dispersive individuals (Phillips et al., 2008; Travis et al., 2013). Climatic niche shifts have also been documented during range expansion, which may explain why some species have shifted faster than expected (Lustenhouwer & Parker, 2022). Providing accurate estimates of evolutionary potential remains challenging (Forester et al., 2022; Hoffmann & Sgrò, 2011), particularly as many factors such as population size, phenotypic plasticity, life history traits, genetic correlations, or spatial connectivity may either facilitate or hinder evolutionary responses to selection (Martin et al., 2023). Genetic diversity indices (e.g., allelic richness, heterozygosity) and effective population size or genetic markers identified through genotype–environment associations or adaptive landscape modeling may be effective proxies for evolutionary potential, but are difficult to gather across multiple taxa and at large spatial extents (Capblancq et al., 2020; Hoban et al., 2022; Razgour et al., 2019; see also Thompson et al., 2023). A complementary approach is to assess the degree to which species niches and underlying traits are conserved through evolutionary time (Bennett et al., 2021; Diamond, 2018; Lavergne et al., 2013). While macroecological approaches may not provide great precision, they could nonetheless help in providing a benchmark to estimate evolutionary potential (Diniz-Filho et al., 2019). Testing whether lineages that have experienced faster niche evolution in the past display a higher potential for evolutionary rescue that prevents range contraction in response to contemporary selective pressures remains an exciting and under-explored research question.

4.1 Account for interactions between traits and climate exposure

One obvious extrinsic factor likely to influence species' range shift-trait associations is climate exposure. At the simplest level, without any changes in climatic conditions, species would not be expected to show any range shifts despite displaying very different traits; conversely, if climate exposure varied across regions or time periods of study, climate exposure alone could cause high variation in range shifts, over and above the influence of species traits. More complex interactions between species traits and climate exposure are also likely to arise, yet they have not received the same level of attention compared with the individual effects of climate change exposure or species traits. For instance, using a process-based model to simulate, under a set of different climate change scenarios, the redistribution of virtual animal species with different dispersal abilities and lifespans (simRShift; see Appendix S2; Bertrand, 2012), we show that the velocity of range shifts at the trailing and leading edges are, at first glance, poorly correlated with species traits (Figure 4a,b). Indeed, lifespan and dispersal abilities alone explain 7.5% and 33.8%, respectively, of the simulated range shifts at the trailing and leading edges. However, once we illuminate the plots with information on the velocity at which isotherms are shifting (expressed in terms of shifts in mean annual temperature along a hypothetical latitudinal gradient), interactions between climate change exposure and species traits appear (Figure 4c,d). Accounting for these trait–environmental change interactions drastically increases the explanatory power of traits (up to 99.5%; Figure 4e,f). Consistent with theoretical expectations (Table 1), we found that species with short lifespans and high dispersal abilities shift their distribution limits faster at the trailing and leading edges, respectively, and especially so when the exposure to climate change is higher. At the trailing edge, the velocity of range contraction is negatively associated with lifespan and displays an interaction with climate velocity, such that range contractions are hastened when lifespan is short and climate exposure is high, indicating that climate impacts are more readily detectable in short-lived species (Figure 4c). At the leading edge, the velocity of range expansion is positively associated with maximum dispersal distance through an asymptotic relationship whose initial slope (at small dispersal distances) is steeper when the velocity of isotherm shifts increases, indicating that the expansion process is likely limited by dispersal ability (Figure 4d).

Interestingly, even if the individual effects of species traits can be weaker predictors of species range shifts than the individual effect of isotherm shift velocity, it is the interaction effect that captures most of the observed variation in species range shifts. Empirical studies in marine systems, where climate velocities are up to seven times higher than on land and where ocean currents strongly influence redistribution, also suggest that traits may be comparatively less important than extrinsic factors to explain variation in biotic velocity among species (García Molinos et al., 2022; Pinsky et al., 2013). Nonetheless, in accordance with our simulations, Sunday et al. (2015) demonstrated that considering the interactions between climate exposure and species traits (including mobility) more than doubled the percentage of explained variance in range expansion of marine fish and invertebrate species compared with a model accounting only for climate expectations. Hence, incorporating these complex interactions may help reveal the contributions of species traits to explain the observed variation in range shift patterns, particularly when the trait effect is weaker (e.g., such as when considering the effect of lifespan alone on range contraction). Obviously, species climatic niches are by essence multidimensional and considering only changes in temperature might also misrepresent range shift expectations (Tingley et al., 2012). Similarly, accounting for extreme events in addition to long-term climatic changes (e.g., heat waves as opposed to mean annual temperature trends) may better capture climate exposure and subsequent impacts on the distribution of populations and species (Auth et al., 2018). In this context, considering multiple climate axes simultaneously (Crimmins et al., 2011; Dobrowski et al., 2013), using model-based bioclimatic velocities (VanDerWal et al., 2013) or accounting for both climatic presses and pulses (Harris et al., 2018) could help refine range shift expectations based on the multifaceted effects of climate change, thereby providing a stronger conceptual framework to understand the contribution of species traits to range shifts.

4.2 Account for interactions between traits and other abiotic and biotic dependencies

Beyond climate exposure, a myriad of abiotic and biotic factors can influence the ability of species to track shifting climate conditions at a variety of temporal and spatial scales, as mediated through their traits. Although these environmental context dependencies must be considered on a case-by-case basis, here we discuss a few examples to illustrate the importance of considering environment–trait interactions to explain variability in species range shifts. Habitats differ in terms of spatio-temporal heterogeneity and human pressures, which will determine opportunities for dispersal and persistence under climate change (Pinsky et al., 2022). These variations in habitat characteristics may explain the faster rates of redistribution in the ocean than on land (Lenoir et al., 2020), but the importance of habitat characteristics to range shift processes will ultimately depend on the interactions with species traits. On land, landscape fragmentation (both natural and anthropogenic) is likely to represent a major constraint on species redistribution but its influence is expected to vary with species-specific habitat requirements, such that habitat generalists may be less constrained by habitat availability than more specialized species (Platts et al., 2019). In the ocean, the opportunities offered by habitat verticality in 3-D habitats can mediate the rate of latitudinal range shifts by allowing species to locally exploit deeper ocean layers (Brito-Morales et al., 2020), yet the rates of depth shifts likely depend on both the niche characteristics and fishing pressure exerted on the shifting species (Dahms & Killen, 2023).

Biotic factors of the environment may also affect the associations between species traits and range shifts. For example, in polyphagous insects, the ability to exploit alternative hosts may confer a survivorship advantage and decrease contraction at the trailing edge, but these benefits rely upon host plant availability now and in the future (Hellmann, 2002). Conversely, although in general, the higher mobility of animals compared with plants a priori suggests that plants should not be limited by dispersal limitations of their animal vectors to keep up with climate change at their leading edge, defaunation of seed-dispersing birds and mammals may drastically reduce the ability of plants to shift their ranges (Fricke et al., 2022). These environmental dependencies highlight the importance of examining realized as opposed to fundamental range shift capacity when testing the relationships between species traits and range shifts through trait–environment interactions to capture the extent to which extrinsic factors constrain the expression of the intrinsic determinants of the range shifts processes (Beever et al., 2016). The characteristics of the range shifting species with respect to the recipient communities can also influence range expansions via biotic resistance and ecological niche opportunities as in the case of biological invasions (Shea & Chesson, 2002), although such hypotheses have so far received mixed support (Alexander et al., 2015; Miller et al., 2023). There is still much to be learned by developing a trait-based framework that explicitly considers the biotic community context (Gilman et al., 2010; Lavergne et al., 2010; Schleuning et al., 2020), as well as capitalizing on the decades-old field of invasion biology to better apprehend the mechanisms underlying range shifts (Pauchard et al., 2016; Wallingford et al., 2020).

5.1 Account for methodological differences in cross-study comparisons

The distribution and abundance patterns of species within their range limits are not static but fluctuate due to demographic and environmental variability and can result in gaps and fragmented areas, particularly at range limits (Brown et al., 1996; Hampe & Petit, 2005). Estimating species range shifts, especially at the edges, may thus be particularly sensitive to the sampling method, number and consistency of studied locations, temporal coverage, spatial grain, as well as the analytical methods used (Loehle, 2020; Shoo et al., 2006). Noteworthy, beyond the noise introduced by the different methods used in the scientific literature, the traits and the estimation process can interact. For instance, species detectability can vary as a function of species morphological and behavioral characteristics (e.g., Sólymos et al., 2018), and the influence of the temporal resolution on perceived range shifts will likely depend on the pace of life of the organisms under study. Therefore, carefully accounting for methodological differences in cross-studies comparisons could improve our ability to detect ecologically meaningful relationships with species traits and thus our power to explain species redistribution.

5.2 Reduce functional biases in species redistribution research

Biases with respect to the trait coverage of the species for which range shifts have been documented may also lead to misrepresentations of the strength and direction of the relationships between range shifts and species traits, particularly if these relationships are nonlinear (“functional biases”; Figure 5a–c). Yet, a comparison of the degree of morphological trait space covered by BioShifts, a range shift database compiled from the scientific literature (Comte et al., 2020), for two widely studied taxonomic groups—birds and fishes—suggests that these functional biases are prevalent (Figure 5d,e). For instance, among terrestrial birds, range shift detections of functionally unique species such as flightless (e.g., ostrich, emu, and kiwi) or lightweight (e.g., hummingbirds) birds remain understudied. Similarly, a bias against small-bodied fishes displaying a high caudal peduncle depth is evident in both marine and freshwater species, with these attributes being strongly related to swimming ability (Fisher & Hogan, 2007; Radinger & Wolter, 2014). Although the consequences of these functional biases on our ability to detect meaningful relationships between range shifts and species traits remain to be formally investigated, to fully understand range shift processes, both functionally common and unique species would benefit from being studied in a shared framework. Functional uniqueness can represent adaptations to specific environmental conditions (e.g., reduced physical capacity for dispersal for flightless birds on islands due to the island syndrome; Wright et al., 2016), distinct alternative phenotypic adaptations to the same environmental challenges (e.g., active dispersal vs. in situ drought resistance forms in aquatic invertebrates; Osakabe et al., 2014) or competition-driven specialization (e.g., morphological and foraging microhabitat specialization in coral reef fishes; Brandl et al., 2015), and may thus hold key insights to contextualize the functions and evolutionary trajectories of trait syndromes (Munoz et al., 2023), including in the context of species redistribution.

5.3 Leverage differences across a variety of taxa and regions

In addition, geographic and phylogenetic biases in research effort may impair our ability to comprehend the importance of the intrinsic and extrinsic factors relevant to range shifts (Feeley et al., 2017; Lenoir et al., 2020). Overrepresentation of temperate species may, for instance, underestimate the vulnerability of tropical species that display particular combinations of behavioral and physiological adaptations, such as narrower thermal tolerances and lower dispersal abilities (Feeley et al., 2017; Tewksbury et al., 2008). Indeed, from an evolutionary perspective, climatic variations (from long-term geological to annual seasonality or daily fluctuations) in the tropics are very different from the ones in the temperate zone or at higher latitudes, with important implications for shaping the current distribution, traits, and genetic diversity of species and populations (De Kort et al., 2021; Hampe & Petit, 2005). These variations can influence their sensitivity and adaptive capacity to climate change (Chan et al., 2016; Steele et al., 2019). These types of bias call for more balanced comparative approaches across the tree of life and globe, leveraging differences across a variety of taxa and regions (including less researched taxa and areas as well as between mainland and islands), to better tease apart the relative contribution of intrinsic species traits and extrinsic factors to explain species redistribution (Pinsky et al., 2022). Recognizing that all the challenges cannot all be resolved at once, we also emphasize the opportunities for carefully targeted laboratory and mesocosm studies that can more directly focus on the mechanisms at play and improve our ability to tease apart the environmental context dependencies of range shift–trait associations (e.g., Luhring et al., 2018).