Discussion

This study identified two diverged genetic lineages (WTIP and CTIP) in the seagrass T. hemprichii across the tropical Indo-Pacific. The observed niche differentiation between the two lineages suggests a violation of the niche conservatism assumption for species-level SDMs, and our lineage-level predictions of present and future range importantly avoid this assumption. Despite differences between the habitat suitability predictions of the lineage-level and species-level SDMs, they consistently predict that the CTIP lineage is at greater risk of range contraction in the future. For this seagrass, but also for other taxa with intraspecific genetic differentiation, incorporating information about phylogeographical structure when modelling the impacts of climate change provides more realistic predictions to better understand future range shifts (Smith et al. 2019; Zhang et al. 2021).

Critical marine predictor variables for seagrasses

Both the lineage-level and species-level SDMs showed that distance to land, water depth, and annual mean SST represent the most essential factors in explaining the distributional patterns of T. hemprichii . The predominant roles of the two geographical predictors and the negligible roles of marine environmental predictors in the WTIP lineage-level model (Table 2) may partially explain the marginal impacts of climate change predicted for this region. The importance of these three predictors has been emphasized in previous studies ofThalassia species (e.g., Duarte 1991; Lapointe et al.1994; Fourqurean & Zieman 2002; Zhang et al. 2014) and other seagrasses (e.g., Baumstark et al. 2016; Jayathilake & Costello 2018). For instance, Jayathilake & Costello (2018) used a set of 13 predictors and developed SDMs for 60 seagrass species including T. hemprichii . They reported the important roles of distance to land and mean SST in explaining geographical distributions of seagrasses. Unexpectedly, maximum SST was reported to be critical, but water depth was less important in their study (Jayathilake & Costello 2018). This inconsistency in our study might be attributed to (a) different sets of predictors, and/or (b) different roles of marine predictors in different seagrass species.

Incorporating intraspecific variation into SDMs for seagrasses

Seagrasses provide vital ecological services in marine ecosystems and SDMs have been applied to this taxonomic group for multiple purposes (see reviews by Robinson et al. 2011; Robinson et al.2017; Melo-Merino et al. 2020). Nonetheless, all previously reported SDMs on seagrasses were built at the species level and thus have not considered possible intraspecific variation. For instance, Chefaoui et al. (2018) developed species-level SDMs for two seagrasses (Posidonia oceanica and Cymodocea nodosa ) in the Mediterranean Sea and predicted that the two species are likely to experience dramatic habitat loss in the future. We fully agree that species-level SDMs are by definition informative; but given the high prevalence of intraspecific variation in marine macrophytes (e.g., Kinget al. 2018), and the significance of intraspecific variation in SDMs (Benito Garzón et al. 2019; Smith et al. 2019; Zhanget al. 2021; Collart et al. 2021), incorporating intraspecific genetic variation into forecasts of seagrass distribution should result in more realistic scenarios of the potential consequences of climate change.
The importance of taxonomic resolution in SDMs has been addressed in several terrestrial and freshwater species, but much more rarely for marine species (see review by Smith et al. 2019; Collart et al. 2021). Species-level SDMs that disregard existing intraspecific variation can either over- or under-estimate climate change impact on distributional change. For instance, species-level models for the lodgepole pine Pinus contorta consistently predicted more extreme habitat loss than subspecies-level models, regardless of the dispersal scenario (i.e., no or unlimited dispersal ability) (Oney et al.2013). As another example, although a species-level model for the reef-building coral Porites lobata predicted over 5% habitat expansion, when modelling this species as five genetically isolated subpopulations the prediction was ca. 50% habitat loss (Cacciapaglia & van Woesik 2018). In the present study, the species model consistently predicted low impacts of climate change in the CTIP region in comparison to the lineage model (e.g., the habitat loss vs. stability in the Sunda Shelf in Fig. 4c vs. Fig. 4d). As for the WTIP region, we found the opposite pattern. Here, the lineage model predicted stable future habitats in the southern Red Sea (Fig. 4c), whereas the species model predicted habitat loss, including to the north of Mauritius (Fig. 4d). In addition, both species and lineage models predict a southward range expansion in the southern CTIP, but only the species model clearly predicts this in the WTIP. Southern expansion is likely correlated with future temperature increases in areas which are now too cold (Supporting Information Fig. S5a). We should note that MESS values in the equatorial regions were slightly negative, which indicates novel future environmental conditions. This is due in part to higher future SST values for this region than those used by the present-day SDM (Supporting Information Fig. S5b)—thus, SDM projections in this region should be associated with more uncertainty. Further studies involving both field investigations and associated data updates and methodological developments for models [e.g., developing ensembles of small models (Breiner et al. 2018) or using smaller study extent] would further improve our predictions for climate change impacts on T. hemprichii in the Tropical Indo-Pacific.

Intraspecific variation and local adaptation in seagrass

Differences in response to thermal changes related to intraspecific variation, whether eco-physiological or evolutionary, are well-documented in seagrasses (King et al. 2018). This variation, partly based on phenotypic plasticity or local adaptation, ultimately might permit seagrasses to acclimatize and adapt to changes in climate (Duarte et al. 2018). The marine predictor variables that played a predominant role in our SDMs (e.g., annual mean SST and water depth) could be responsible for both long- and short-term local adaptation ofT. hemprichii to a changing climate (King et al. 2018; Jahnke et al. 2019b). In support of this, common-garden experiments have revealed a clear local adaptation to increased temperatures in Zostera marina (Franssen et al. 2011; 2014), and to a depth gradient in Posidonia oceanica(Marín-Guirao et al. 2016; Jahnke et al. 2019b). Further, parallel adaptation of Z. marina to thermal clines along the American and European coasts was demonstrated using a space-for-time substitution design and gene expression profiling (Jueterbock et al. 2016). Such adaptive local differentiation induced by divergent environmental forces (e.g., light, depth and temperature) has led to structured populations and lineages in seagrasses at various spatial scales (Dattolo et al. 2014; Jueterbock et al. 2016; Jahnke et al. 2019b), suggesting that adaptation to local conditions is a key mechanism for seagrasses to face global climate change.
In T. hemprichii , natural selection imposed by environmental heterogeneity might have resulted in the evolution of locally adapted populations with considerable variation in productivity, growth rate and competitive interactions (Martins & Bandeira 2001; Lyimo et al.2006; Larkum et al. 2018). Despite clear genetic differentiation identified between the WTIP and CTIP lineages, we did not ascertain the adaptive and non-adaptive components of divergence in a common landscape of the tropical Indo-Pacific. Future studies should focus on distinguishing neutral genetic differentiation from local adaptation using reciprocal transplant trials (e.g., common gardens and provenance trials) (see Joyce & Rehfeldt 2013; Ralph et al. 2018). Also, it is most important to assess the sub-lethal susceptibility of T. hemprichii to thermal stress before the strongest impacts of future climate change are sustained. Intraspecific genetic diversity across populations can increase a species’ adaptive capacity and result in cascading effects to the entire ecosystem (Evans et al. 2017). It is thus important to identify the most temperature-tolerant genotypes from the WTIP and CTIP lineages, perhaps by manipulating temperature to quantify the performance of individual genotypes of T. hemprichiiacross thermal gradients. It is also essential to clarify whether genotype complementarity or dominance enhance the adaptive capacity in a population (Hughes & Stachowicz 2011).

Conservation implications

The challenge of designing effective actions for seagrass conservation in the Indo-Pacific exists in the gap between science, policy, and practice (Fortes 2018). In this study, the separation in geographic distribution and high niche differentiation between the CTIP and WTIP lineages suggest that T. hemprichii populations may be locally adapted (Merilä & Hendry 2014). For species with significant intraspecific genetic diversity, it is crucial to help maintain the species’ potential for adaptive responses to climate change by conserving this diversity (D’Amen et al. 2013). In particular, lineage differentiation can be explained by recruitment rate (Lyimoet al. 2006; Sherman et al. 2018), nutrient resorption (Martins & Bandeira 2001), and evolutionary history from the origin center to the distributional margins (Mukai 1993). Dramatic future habitat loss in the CTIP was predicted by both the species- and lineage-level models (Fig. 4), stressing the urgency to develop monitoring programs to rescue evolutionary and/or ecologically important units in T. hemprichii , particularly the populations and gene pools that have persisted through past long-term climate change because of local adaptation (Bell 2017; Hernawan et al. 2017). Furthermore, the recognition of high niche differentiation between the WTIP and CTIP lineages may help to establish coherent principles and regulating practices by which the different areas that T. hemprichii inhabits can be protected efficiently.
The biomass, abundance, and productivity of seagrasses are highly correlated with both habitat suitability (Martins & Bandeira 2001; Saunders et al. 2013) and epiphytic species biodiversity (Lyimoet al. 2008). Optimizing productivity of T. hemprichii in a given site or population can help to increase associated community diversity (Eklöf et al. 2006; Lyimo et al. 2008). Thus, it is necessary to explore how community diversity and structure correlate with the genetic composition and structure of the foundational speciesT. hemprichii . Such research can help validate the results of SDMs in this study and quantify the relationship between T. hemprichii and its relevant community components (Ikeda et al.2017). Since populations in each of the CTIP and WTIP lineages are locally adapted, policymakers and stakeholders are encouraged to use local seed sources of T. hemprichii to ensure management strategies for successful restoration and conservation purposes. However, as the WTIP lineage may be more resilient to future climate change, WTIP seeds could possibly be used to restore CTIP seagrass beds which are predicted to disappear in the future.
Finally, apart from marine geographical and environmental predictors, geographical distributions of seagrasses are also determined by other factors including biotic interactions. For instance, Hyndes et al. (2016) predicted that accelerating tropicalization can lead to a potential shift both among the seagrass themselves and among their associated communities, thereby affecting ecosystem services that seagrasses provide in this region. The importance of incorporating biotic interactions into SDMs has long been recognized but it is still poorly addressed in the marine realm. More mechanistic studies underlying thermal adaptation by linking ecology to genetics should be done to better understand how T. hemprichii will adapt to climate change (Duarte et al. 2018; Hu et al. 2020).