Drivers of compositional turnover in narrow‐ranged and widespread dragonflies and damselflies in Africa

We aim to explore what processes dominate community assembly of dragonflies (Odonata: Anisoptera) and damselflies (Odonata: Zygoptera) by differentiating the environmental and geographical drivers behind compositional turnover of narrow‐ranged versus widespread species. In this way, we further aim to describe patterns of species incidence and compositional turnover to expand upon the body of knowledge related to understanding biodiversity patterns and processes. We explored species turnover of dragonflies and damselflies separately, using zeta diversity to measure compositional turnover among multiple assemblages. Narrow‐ranged and widespread species within each suborder showed similar drivers. Specifically, both narrow‐ranged and widespread dragonflies show rapid turnover with small shifts in annual mean temperature, temperature seasonality and annual precipitation, whereas for damselflies, the major driver for turnover is distance between sites followed by climatic variables. Our results therefore show that odonate turnover is largely driven by climate, although the limited dispersal capabilities of damselflies also influences community assembly. Climate change could cause major changes in composition of odonates, presenting a challenge for conservation planning in Africa as species assemblages that were previously conserved may no longer be protected if their ranges shift outside protected areas. For damselflies, adaptation is a major concern, and with their limited dispersal capabilities and climate sensitivity, they may not be able to migrate effectively in response to changing climate conditions. The underlying assembly processes do not differ considerably for narrow‐ranged and widespread species within each suborder, suggesting that conservation planning tailored to each suborder may be sufficient in Africa.

Although regional scale drivers of species composition between Anisoptera and Zygoptera have been studied in the past, the focus was largely on how major rivers may restrict the distribution of taxa with lower dispersal abilities (Alves-Martins, Brasil, et al., 2019;Alves-Martins, Calatayud, et al., 2019).For example, lentic (standing water) and lotic (running water) species may also differ in distribution and range boundaries (Hof et al., 2006).Globally, major rivers often act as dispersal barriers for Zygoptera due to their limited dispersal capabilities, resulting in localised centres of endemism and diversity hotpots (Alves-Martins, Brasil, et al., 2019;Alves-Martins, Calatayud, et al., 2019).By contrast, Anisoptera adults are generally larger and more robust with a stronger flying capability than Zygoptera and are therefore able to disperse over longer distances (Alves-Martins, Brasil, et al., 2019;Alves-Martins, Calatayud, et al., 2019;Angelibert & Giani, 2003;Júnior et al., 2015;Mähn et al., 2023;Wootton, 2020).
These contrasting dispersal abilities may create distributional patterns more aligned with environmental variables for Zygoptera compared with Anisoptera, which may exhibit more random patterns (Alves-Martins, Brasil, et al., 2019;Alves-Martins, Calatayud, et al., 2019).

These factors backdrop the strong association of the distribution of all
Odonata with climatic variables (Bush et al., 2013;Pinkert et al., 2023;Simaika et al., 2013;Simaika & Samways, 2015) as well as the habitat and environmental conditions.Both mature and, particularly, immature odonates demonstrate a high degree of sensitivity to the integrity and variations in aquatic habitats (Kietzka et al., 2017;Mendes et al., 2018;Nasirian & Irvine, 2017;Šigutová et al., 2023).Other physical barriers include geographic features such as mountains, oceans and large expanses of unsuitable habitat (like urban areas or intensive agricultural lands), which can impede movement, isolate populations and prevent gene flow (C ordoba-Aguilar et al., 2023).
We expect that widespread and narrow-ranged odonates will exhibit distinct biological traits because of their varied habitat requirements and/or dispersal abilities.Narrow-ranged species are characterised by lower observed occupancy rates, being found in only a few sites, whereas widespread species exhibit higher observed occupancy rates, spanning multiple sites.The determinants of geographic range generally fall under one of two broad categories.The first determinant is linked to niche differentiation relating to environmental preferences as well as competition (Cornwell & Ackerly, 2010;Rabinowitz et al., 1984;Siqueira et al., 2012).For example, the distribution of widespread species may be less dependent on environmental heterogeneity because of their broader range of environmental tolerance, compared with narrow-ranged species (Lennon et al., 2011).Consequently, narrow-ranged species are typically habitat specialists, whereas widespread species are habitat generalists (Grant & Samways, 2007), although the level of habitat specialisation does not always correlate well with species occupancy, and additional factors, such as species interactions, may explain why some generalist species may be rare.The second determinant relates to differences in species dispersal abilities, as widespread species often have stronger dispersal ability (Siqueira et al., 2012).For example, Resh et al. (2005) found that in a long-term macroinvertebrate survey, widespread species are more likely to drift and exhibit a high female dispersal ability than narrow-ranged species.
Measures of spatial variation in the compositional similarity of assemblages are often represented by beta diversity (or species turnover), a fundamental metric based on pairwise comparisons between assemblages composing a metacommunity (Chao et al., 2012;Jost, 2007).For instance, a typical workflow for such pairwise comparison includes, first, measuring species turnover for all possible combinations of site pairs (as the compositional dissimilarity/distance matrix); then, the variation in rates of species turnover along site gradients-that is, between-site differences in explanatory variables for these combinations-can be explained by correspondence analysis (e.g., Grant & Samways, 2007) or dissimilarity modelling (Fitzpatrick et al., 2013).That is, widespread species were compared with closely related narrow-ranged species across pairs of sites (Kunin & Gaston, 1993) while ignoring the fact that the geographic ranges of species in a taxonomic group cover a continuous spectrum (McGeoch & Gaston, 2002).
Such compositional turnover analyses based on multiple pairwise comparisons provide useful indicators for site prioritisation ideal for conservation planning (Socolar et al., 2016).Consequently, past studies of odonates have focused on site-level or pairwise analyses of species richness and turnover (Basel et al., 2021;Bried & Siepielski, 2018;Dijkstra & Clausnitzer, 2006;Kalkman et al., 2007;Simaika et al., 2013), thereby disproportionately highlighting the contribution of narrow-ranged species to assemblage structures.This is because narrow-ranged species typically constitute a large portion in local assemblages, are less likely to appear in both sites of a randomly selected pair and thus contribute more to pairwise turnover than widespread species.Consequently, it generally undercuts the umbrella role of widespread species occurring among multiple communities (Hui et al., 2018;Latombe et al., 2017;McGeoch et al., 2019).Similarly, correlative analyses based on beta diversity can overemphasise the spatial and environmental drivers determining the presence or absence of narrow-ranged species while not accounting for the drivers of widespread species that often dominate ecosystem functioning (Gaston, 2010).As diversity components across sets of three or more sites cannot be expressed with metrics of alpha and beta diversity alone, we need to go beyond just pairwise analyses to estimate the contribution of widespread species to the compositional turnover of dragonfly species.
To address these limitations, we use increasing orders of zeta diversity (i.e., the average number of species shared by increasing number of sites; Hui & McGeoch, 2014) to gradually sift out the contribution of narrow-ranged species and increasingly accentuate the role of more widespread in driving species turnover.As the number of sites increases, the number of species shared between sites decline as narrow-ranged species become excluded from the computation of higher-orders of zeta diversity.Consequently, narrow-ranged species, shared by a small number of sites, will contribute disproportionately more to lower orders of zeta diversity, whereas widespread species contribute more to higher orders of zeta diversity.The relationship between compositional turnover and different gradients of environmental conditions can then be estimated using multi-site generalised dissimilarity modelling (MS-GDM; Latombe et al., 2017) of different orders of zeta diversity, to account for the full spectrum of species' geographic ranges.Furthermore, compositional turnover, typically computed using Sørensen or Jaccard dissimilarity between pairs of sites, can be partitioned into a nestedness and a species replacement component.The nestedness component captures the difference in species composition due to difference in species richness between sites, whereas the replacement component captures the difference in species composition between sites due to the checkerboard patterns of mismatched occurrences of species.Species replacement can be computed using Simpson dissimilarity index or Williams replacement index (Baselga & Leprieur, 2015).Differences in species richness and species composition are likely driven by different factors (Baselga, 2010;Bried & Siepielski, 2018).We thus used normalised zeta diversity metrics in the MS-GDM (Sørensen and Simpson equivalent of zeta diversity), designed for differentiating the replacement and nestedness components (Latombe, Richardson, et al., 2018).
Using MS-GDM for different orders of zeta, we compare the turnover of Anisoptera and Zygoptera suborders to explore, first, the turnover and potential community assembly processes of Anisoptera versus Zygoptera, and second, the (dis)similarity of environmental and geographical factors driving compositional differences of narrow-ranged versus widespread species.To distinguish whether compositional turnover between sites is a result of environmental filtering or geographical isolation, variables related to the environment and to the geographical distance between sites are included in the analyses.The results will allow for a better and more complete understanding of what drives biodiversity responses to environmental gradients in Odonata.Such an understanding can assist us in future selection of species (and subgroup focus) for ecological indicators and taxonomy-specific conservation prioritisation in Africa.

Odonata occurrence records
Georeferenced point locality records were obtained from the 'Odonata Database of Africa (ODA)', which consists of presence-only sitespecific records compiled from expert-reviewed records, literature, field notes and group members collections over the African continent (Clausnitzer et al., 2012;Kipping et al., 2009).It is the first continentwide, taxonomically verified database of tropical freshwater insects.
We excluded islands from the dataset and only considered mainland continental Africa.In total, our initial dataset comprised approximately 122,000 records.Despite the large number of records, there are clear sampling gaps in the database, with most records concentrated in southern Africa (Figure 1).

Environmental covariates
We obtained bioclimatic variables (annual mean temperature, temperature seasonality, annual precipitation and precipitation seasonality), from the WorldClim database (www.worldclim.org;Hijmans et al., 2005).Climatic variables, specifically linked to temperature and precipitation, are dominant factors driving insect distributions and are known to affect Odonata physiology, including phenology, seasonal regulation, immune function and pigment production used for thermoregulation (Hassall, 2015;Hassall & Thompson, 2008).
In addition, we obtained data characterising river length as well as minimum and maximum elevation from www.earthenv.org/streams(Domisch et al., 2015).We used river length as an indicator of habitat suitability because Odonata require water for breeding and the larval phase, making them dependent on water for much of their lifecycle (Simaika et al., 2016).We also included elevation as the altitudinal range of Odonata species is likely to result in differences in species composition (Niba & Samways, 2006;Samways, 1989).For instance, Niba and Samways (2006) found Anisoptera species richness and abundance increased with elevation; instead, although there was no change in Zygoptera species richness with elevation, Zygoptera abundance decreased significantly.Lastly, human population density data were used to quantify human disturbance and obtained from NASA's open data portal (beta.sedac.ciesin.columbia.edu/data;CIESIN, 2016).
Higher human population density often leads to increased habitat alterations and land use changes, which in turn could have negative effects on the habitat suitability and environmental conditions for Odonata species.As a result, increased human disturbance is expected to lead to a higher turnover of Odonata species, as it may disrupt the ecological communities and lead to changes in species composition.

Data preparation
All analyses were performed in R (version 4.1.2;R Development Core Team, 2019).The mainland African continent was divided into halfdegree cells based on preliminary assessment to identify sufficient data density for meaningful analysis.Each cell was considered a single site in subsequent analyses and was characterised by its central geographic coordinates, species composition (presence and pseudo-absence) and the mean of its associated environmental covariates described above when data were available at finer spatial grain.These were overlaid using the 'LetsR' package (Vilela & Villalobos, 2015).
The ODA records are adequate to represent a near-complete view of Anisoptera and Zygoptera assemblages, indicated by the level-offs of the site-based species accumulation curves (Figure S1).However, they are inadequate in many sites when exploring betweensite species turnover.Although stringent criteria for site selection are available (e.g., de Beer et al., 2023), they do not apply well here as biodiversity data are typically scarce in Africa.
To select well-sampled sites without overly discarding too many sites from the database, we designed a straightforward criterion for site selection.Specifically, we selected sites with record counts greater than a threshold, while the threshold was determined by the non-parametric Spearman correlation (ρ) between the number of species and the number of records within sites.Setting a higher threshold selects fewer sites and reduces the dependence of detected species on sampling effort (Figures S2 and S3).We found that Spearman's ρ declined from above 0.9 to below 0.5 with the increase in the threshold (minimum count in sites) and selected ρ = 0.5 to set the threshold (Appendix S1; Figures S1 and S2).Accordingly, for Anisoptera, we selected 160 sites with each more than 110 records, and for Zygoptera, 190 sites with each more than 60 records.In total, this represented approximately 70,000 cleaned data records.As Spearman correlation for these selected sites was still substantial, we kept record counts as a predictor in the analyses to directly interpret the role of sampling effort (Basel et al., 2021;Hortal & Lobo, 2005;Hui et al., 2011).
F I G U R E 1 Records of two odonate suborders across Africa.
We performed a square root transformation for covariates whose distributions were highly skewed (including stream length, human population density and sampling effort).All covariates were resampled to half-degree resolution ($ 50 Â 50 km) and tested for multicollinearity by computing the Variance Inflation Factor (VIF) (R package car; Fox et al., 2012).The VIF measures how much the variance of a regression coefficient is inflated due to collinearity with other coefficients within the model.Backward elimination was used to remove variables until all VIFs were below 5.As a result, nine variables were retained in the final model, namely annual mean temperature, mean diurnal range, temperature seasonality, annual precipitation, precipitation seasonality, maximum elevation, stream length, human population density and sampling effort.These were then used as environmental predictors in MS-GDM along with geographic distance (n = 10).A summary of the data layer sources, file formats and resolutions can be found in Table S1.

Zeta decline
Zeta diversity (ζ i ) is the average number of species in common among multiple i sites (Hui & McGeoch, 2014).ζ 1 is thus the average number of species per site (i.e., alpha diversity), ζ 2 is the number of species shared by two sites (i.e., complement of beta diversity), ζ 3 is the number of species shared by any three sites and so on, until the maximum number of sites is reached (Hui et al., 2018).Hereafter, the number of sites used for calculating zeta diversity is named the order of zeta.
Using the R package zetadiv (Latombe, McGeoch, et al., 2018), we calculated the zeta diversity decline for orders 1-5.The zeta decline is characterised by the form and the rate of decrease in the average number of species shared across an increasing number of sites (McGeoch et al., 2019).The shape of the zeta decline provides information on the structure of turnover and community assembly processes (Deane et al., 2023).A steep decline of zeta values at low orders reveals that turnover is mostly structured by differences in narrow-ranged species between sites, whereas a shallow decline reveals a structure that is largely driven by widespread species (McGeoch et al., 2019).

Multi-site generalised dissimilarity modelling
Standard generalised dissimilarity modelling is a regression method used to predict beta diversity from environmental differences and the distances between pairs of sites (Ferrier et al., 2007).MS-GDM extends this typical pairwise measure of beta diversity into multiplesite zeta diversity metrics (Latombe et al., 2017).MS-GDM therefore evaluates how the i th -order zeta diversity changes with the means of environmental differences and distances between C n i site pairs for any given n sites.MS-GDM was computed separately for each zeta order, which enables us to differentiate the drivers of species turnover for narrow-ranged versus widespread species.Specifically, we applied MS-GDM to Anisoptera and Zygoptera suborders separately for orders 2, 3, 4 and 5 using 10,000 random combinations.
As zeta diversity is calculated by counting the number of species shared by a specific combination of sites, its raw values are influenced by richness (order-1 zeta diversity).For this reason, we normalised the raw zeta values using the average and minimum number of species in the specified sites within a combination (Latombe, Richardson, et al., 2018) equivalent to the Sørensen and Simpson dissimilarity indices for beta diversity.Sørenson measures of beta diversity are not independent from richness differences (i.e., compositional nestedness), whereas Simpson beta diversity is a measure of species replacement.The difference of the two therefore reflects the role of nested occupancy structures (Baselga, 2010;Fischer & Lindenmayer, 2005).We thus applied MS-GDM for both the Simpson-and Sørensen-equivalent zeta diversity to determine the effect of compositional nestedness on turnover (see model settings for zetadiv; Latombe et al., 2017).Plots for the Simpson index (reflecting true turnover) are reported in the main text as well as Figure S4, and results for Sørenson index (reflecting total turnover) can be found in Figure S5.For the performance of MS-GDM, we calculated the variance explained as Pearson's R 2 between the observed zeta values and the zeta values predicted by the model.
The contributions of each environmental variable, sampling effort and geographic distance, to observed compositional turnover at different zeta orders, driven largely by narrow-ranged (ζ 2 ) versus widespread (ζ 5 ) species, were evaluated using MS-GDM for the two odonate suborders (Anisoptera vs. Zygoptera) separately.For each order of zeta, MS-GDM fits the log-transformed zeta diversity to I-splines of each of all predictors (with three internal knots to avoid overfitting) using generalised linear modelling to produce a monotonic partial response curve for each predictor.All predictors are rescaled between 0 and 1 before computing the I-splines.To visualise these outputs, the I-spline response curves were plotted for each order of zeta to highlight two important parts of information: first, the amplitude of the I-spline response curve indicates the overall effect of the variable on zeta diversity relative to the other covariates.In this case, high amplitude of an environmental variable would show that species presence and absence across sites of a specific order are largely filtered by this environmental condition, whereas a high amplitude for geographic distance between sites would demonstrate a notable distance decay of similarity and the role of geographic isolation.Second, the slope at a particular point in the response curve indicates the sensitivity of compositional turnover to local changes of an environmental condition around a particular value.For example, a steeper slope at lower temperatures would show that compositional turnover is more sensitive to temperature change in colder environments.

Patterns of zeta diversity
During the analysis, 443 Anisoptera and 334 Zygoptera species were considered.In the case of Anisoptera, the three most widespread Conversely, Anisoptera's decline in zeta similarity was more gradual, particularly across the first few orders of zeta (Figure 2a; zeta decline results are summarised in Table 1).
For Anisoptera, the main predictors of zeta diversity for orders 2 to 5 were temperature seasonality, annual precipitation and annual mean temperature (Figure 3a).Specifically, the turnover of narrowranged Anisoptera species (order 2) was more sensitive to changes in sites with low temperature seasonality but with moderate annual precipitation and mean temperature.For more widespread species (orders 3-5), compositional turnover was more sensitive to changes in sites with low temperature seasonality and moderate annual precipitation but with high annual mean temperature.The effect of precipitation seasonality on compositional turnover was moderate but slightly increasing with orders (from narrow-ranged to widespread).
Across zeta orders, sampling effort played a relatively trivial role, suggesting adequate sampling in selected sites.There was also a moderate but notable effect of human density on compositional turnover F I G U R E 2 Zeta diversity decline for two odonate subgroups across Africa.Zeta diversity was calculated with increases in zeta order (number of combinations of sites) for (a) Anisoptera and (b) Zygoptera.Means and ± standard deviation are represented by the solid and dashed lines.
T A B L E 1 Zeta diversity summary of odonate suborders Anisoptera (Ani) and Zygoptera (Zyg) in Africa.across zeta orders.Geographic distance between sites was the weakest predictor across all orders of zeta (i.e., from narrow-ranged to widespread species) (Figure 3a).
By contrast, for Zygoptera, geographic distance between sites was the most important predictor for all orders of zeta, whereas the second predictor was temperature seasonality followed by sampling effort (Figure 3b).Specifically, compositional turnover was more sensitive to changes in geographic distance between sites that are relatively close, less than 1/4 of the maximum average distance between sites, beyond which there could be a complete species turnover F I G U R E 3 I-splines plotted for odonate suborders (a) Anisoptera and (b) Zygoptera for zeta order 2, 3, 4 and 5 using Simpson normalised zeta.Predictors are rescaled between their minimal [0] and maximal [1] values to ease visualisation and comparison.
between sites.Temperature seasonality and annual mean temperature still played a moderate role in driving compositional turnover.Moreover, the effect of sampling effort was still notable, suggesting the need for increased sampling effort.The full statistical table of MS-GDM is provided in Tables S2 and S3.

Structures of community assembly
Among subgroups, the rate of the decline in zeta diversity was greater for Zygoptera compared with Anisoptera, indicating that turnover was twice as high in Zygoptera as in Anisoptera (Figure 2).The larger the change in the value of zeta across subsequent orders, the greater the relative difference in the numbers of narrow-ranged versus increasingly widespread species in the community (McGeoch et al., 2019).A steep decline of zeta values at low orders for Zygoptera reveals that turnover is mostly structured by differences in narrow-ranged species between sites (McGeoch et al., 2019).When comparing across datasets, species of Anisoptera are generally more widespread than Zygoptera.It has been argued that the distribution of widespread species is less dependent on environmental heterogeneity due to their broader environmental tolerance, contrary to narrow-ranged species, whose rarity is often due to habitat specialisation (Lennon et al., 2011).Being species with complex life histories, life stages such as the breeding status can further differentiate their niche breaths and geographical distributions (Patten et al., 2015).
Differentiating records to specific life stages could further refine our results to turnover of life stages along environmental gradients.
Environmental filtering or isolation by distance?
Our findings support the idea that Anisoptera and Zygoptera differ in their abiotic niche space, as their community drivers are clearly distinct.We see that for Anisoptera, environmental filtering, specifically climate, drives community assembly, whereas for Zygoptera, dispersal and colonisation (isolation by distance) plays a significantly larger role in driving community assembly.Indeed, for Anisoptera, geographic distance between sites is the weakest predictor, whereas for Zygoptera, geographic distance between sites was continuously the primary predictor of community assembly for narrow-ranged and widespread species (Figure 3).
Given the conservation, ecological and evolutionary significance of dispersal, research has been directed towards quantifying the dispersal ability of various species.Previous research has suggested that Anisoptera and Zygoptera tend to differ in dispersal ability in their adult stage (Alves-Martins, Brasil, et al., 2019;Alves-Martins, Calatayud, et al., 2019;Angelibert & Giani, 2003;Júnior et al., 2015).
In general, dispersal probability increases with species size because Anisoptera are generally larger in size than Zygoptera, and they are more robust and are thus are able to fly over longer distances (Mähn et al., 2023;Michiels & Dhondt, 1991).Although Anisoptera are sensitive to climate conditions, we can expect species to respond to climate change by shifting their distribution via dispersal.For Zygoptera, however, dispersal is more limited and so shifting across a heterogeneous landscape to find their ideal niche may be constrained.On a global level, major rivers have acted as dispersal barriers for Zygoptera with limited dispersal capabilities, resulting in centres of endemism and diversity hotpots (Alves-Martins, Brasil, et al., 2019;Alves-Martins, Calatayud, et al., 2019).These contrasting dispersal abilities may create non-random distributional patterns for Zygoptera compared to Anisoptera (Alves-Martins, Brasil, et al., 2019;Alves-Martins, Calatayud, et al., 2019).Future studies could explore the use of the Global Surface Water datasets by Pekel et al. (2016) and Pickens et al.
(2020) to capture the inland/surface water dynamics in more detail.
We find that for both suborders, climatic variables significantly influence the composition of species assemblages, especially for Anisoptera (Figure 3a).Climatic parameters are known to have a direct effect on insect population dynamics through the modulation of survival, development rates, fecundity and dispersal (Bush et al., 2013;Hassall & Thompson, 2008;Simaika et al., 2013;Simaika & Samways, 2015).As odonate fitness and survival is directly linked to the local climate of the area, small changes in temperature will likely cause species changes within assemblages (Simaika & Samways, 2015).In Basel et al. (2021), odonate distributions are predicted to show extensive reorganisation by 2050 as a result of climate change, whereas temporal turnover is expected to reach 80% in some areas of South Africa.With ongoing climate change, we can expect to see major shifts in assemblage structure.Nonetheless, these species may also be used to effectively track forefronts of climate change.

Considering rare and widespread species in conservation planning
Our results showed that for Anisoptera, temperature and precipitation gradients emerged as key drivers in turnover of narrow-ranged as well as widespread species (Figure 3a).For Zygoptera, geographic distance emerged as the primary predictor for both narrow-ranged and widespread species (Figure 3b).Species conservation efforts for narrowranged and widespread odonates within each subgroup will therefore focus on all levels of co-distribution so long as we address the subgroup specific drivers and threats.
For Anisoptera, small changes in annual mean temperature and precipitation, but also in temperature seasonality, are likely to have major implications for species turnover (Figure 3a).With both climate averages and extremes increasing in severity (Fischer et al., 2021), we are therefore likely to see rapid changes in the composition of Anisoptera assemblages.This presents a great challenge for conservation planning as species assemblages that previously occurred in protected areas may not stay intact as species move in response to climate change outside protected areas (Hannah et al., 2007).The notion that conservation areas are static geographical units for biodiversity conservation should be revised when planning for climate change adaptation.
The excessive rate of change of the environment may exceed the capacity of developmental, genetic and demographic mechanisms that populations have evolved to deal with (Chevin et al., 2010).For Zygoptera, adaptation is a major concern, with their limited dispersal capabilities (Figure 3b), they may not be able to migrate effectively in response to changing climate conditions.Climate change corridors and buffers may help reduce geographical boundaries that may inhibit their dispersal further (Samways, 2007).A strategy of stepping stones and corridors may offer the best opportunities for dispersal and migration (Samways, 2007).These corridors and buffers can also assist and benefit the more mobile and adaptable Anisoptera species.Furthermore, limiting climate change is an essential complement to adding protected areas for conservation of biodiversity.

CONCLUSION
Anisoptera and Zygoptera assemblages are driven by different factors.
For Anisoptera, environmental filtering plays a key role in species assemblage, whereas for Zygoptera, dispersal and colonisation exerts a greater influence on species assemblage.These differences reflect key ecological differences between these suborders and highlight a need to make subgroup specific monitoring programmes.
In particular, future changes in climate pose a major risk for Anisoptera because climate is a primary driver of turnover.Even small shifts in temperatures and precipitation will likely result in major changes in species distributions.As such, we can continue to rely on odonates as an important bioindicator species.In particular, Anisoptera species may be especially sensitive to tracking forefronts of climate shifts in the future.
For Zygoptera, limited dispersal ability coupled with climate sensitivity, emphasise the importance of ensuring these species can effectively move between the habitats should climates change.That is, limited dispersal ability can reduce their capacity to overcome barriers created by a heterogeneous environment, and further fragmentation in habitats may be fatal for the continuation of many Zygoptera species.A forward thinking strategy of building conservation corridors may offer the best opportunities for future dispersal and migration.
Understanding the drivers underlying patterns of species diversity is imperative for the effective protection of regional diversity and the preservation of ecosystem functioning.Robust and well-informed conservation management actions should encompass strategies that address the specific needs of each subgroup, taking into account the unique ecological factors that drive species assemblage within Anisoptera and Zygoptera.Here, we found that the differences in drivers between narrowranged and widespread subgroups is not pronounced.Consequently, species conservation efforts for narrow-ranged and widespread Odonata should not differ largely.Instead, when planning for conservation, we advocate to focus on subgroup specific drivers, which will have an umbrella effect on all levels of co-distribution.