Global hotspots of butterfly diversity in a warming world

doi:10.21203/rs.3.rs-4437399/v1

Download PDF

Biological Sciences - Article

Global hotspots of butterfly diversity in a warming world

https://doi.org/10.21203/rs.3.rs-4437399/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Insects and their many ecosystem functions are in decline and threatened by climate change^1,2, yet lack of globally comprehensive information limits the understanding and management of this crisis³. Here we use butterflies as a global model insect system^4–7 and uncover a strong coincidence of their diversity and threat. Integrating comprehensive phylogenetic and geographic range data for 12,119 species, we find that global centers of butterfly richness, rarity, and phylogenetic diversity are unusually concentrated in tropical and sub-tropical mountain systems. Mountains⁸ hold 3.5 times more butterfly hotspots (top 5%) than lowlands and two thirds of the species are primarily mountain-dwelling. Only a small portion (14%-54%) of these diversity centers overlap with those of ants, terrestrial vertebrates and vascular plants, and this spatial coincidence rapidly decreases above 2,000 m elevation where butterflies are uniquely concentrated. The geographically restricted temperature conditions of these mountain locations now put butterflies at extreme risk from global warming. We project that 64% of butterflies’ temperature niche space in tropical realms will erode by 2070. Our study identifies critical conservation needs for butterflies and illustrates how the consideration of global insect systems is key for assessing and managing biodiversity loss in a rapidly warming world.

Biological sciences/Ecology/Conservation biology

Biological sciences/Ecology/Biodiversity

Biological sciences/Zoology/Entomology

Biological sciences/Ecology/Climate-change ecology

Earth and environmental sciences/Climate sciences/Climate change/Climate-change impacts

Understanding the distribution of biodiversity on Earth is the prerequisite for effective conservation and for mitigating the loss of species and their functions under rapid environmental change^9–13. Recent work has highlighted the range of implications arising from an uneven geographic distribution of biodiversity, such as vast differences in countries’ conservation responsibilities¹⁴ and heterogenous representation in protected areas^15,16. These geographic differences in cause and consequence extend to, and are often exacerbated for, the functional and phylogenetic aspects of biodiversity¹⁷, which are recognized as central for supporting ecosystem resilience and preserving critical evolutionary heritage^18–20.

The same motivation for a more comprehensive understanding extends to different taxa. For terrestrial vertebrates, the thus far dominant model system for global ecology and conservation^10,13,21,22, prior work has documented marked differences in diversity patterns of endothermic and ectothermic taxa. While birds and mammals (endotherms) share 75% of their richness and rarity hotspots, their respective overlap with amphibians and reptiles (ectotherms) is not strong^20,21. The extent to which these differences apply to plants and invertebrates remains poorly understood. This is especially true for insects, despite their essential ecosystem functions^2,5, outstanding diversity, and alarming decline³. For ants, species richness and rarity hotspots are uniquely concentrated in regions with dry conditions that reflect the xeric preference and advantages of the groups’ social organization²³. These recognized differences exemplify the precarious foundation of the current knowledgebase for global biodiversity at large scales and leave insect biodiversity poorly represented.

The uneven distribution of biodiversity takes on an additional weight when hotspots cluster into distinct portions of the climate space that are particularly strongly impacted by anthropogenic change^22,24. One prominent case are mountains. Owing to their historical role in supporting species’ survival and speciation through long-term climate dynamics and their own isolation, many mountain systems are centers of biodiversity^25–27. Over 40% of terrestrial vertebrate species are mountain endemics²⁸, but despite the strong implication for conservation whether and how this extends to insects remains unknown. At the same time, temperature-induced up-slope shifts and losses in habitat combined with strong geographic isolation put mountain biodiversity in great peril^29–32. Impacts from warming might be particularly severe for insects due to their known strong temperature sensitivity^33–36, but to date a global assessment of the geographic coincidence of diversity, rarity, and climate change threats for an insect system does not exist.

Here we address these issues using butterflies as global insect system. Given their ecological importance⁵, high surrogacy for insect diversity⁴, and their uniquely comprehensive distributional and phylogenetic information^6,7, butterflies offer pivotal insights into complementary priorities for insect conservation. To achieve global coverage, we mobilized, modelled, and validated species distribution information for 12,119 species based on over 8 million occurrence records and newly digitized expert range maps (Supplementary Figs. 1, 2 and Table 1).

Richness, rarity and phylogenetic diversity

Across our 110 km-resolution analysis grid, butterfly species richness (SR) increases toward lower latitudes and higher elevation (Fig. 1a), peaking near the equator and at approximately 2500 m mean elevation (see methods and supplement for additional results addressing data gap species). In contrast, average range rarity (RR), defined as the average inverse of species’ global range sizes per assemblage – a continuous measure of endemism, bimodally peaks at around 20° and 40° latitude and 3500 m elevation (Fig. 1b). Notably in most realms, both metrics show limited congruence, particularly for peaks in narrow-ranged species, which are usually of greatest conservation concern (Spearman’s rank SR-RR: ρ = 0.40, n = 12,515). These differences become more evident in an assessment of global hotspots, defined as the top 5% of assemblages of the six major biogeographical realms (Fig. 2). Only 10% of richness and range rarity hotspots are shared globally, with no overlap in the Indomalayan and Afrotropic realms and as few as 2% in the Afrotropic and 4% in the Palearctic realms. This suggests that the hotspot disparity previously documented for vertebrates^10,11,13, and hence the limited value of richness hotspots for identifying places of greatest global conservation concern, also extends to butterflies.

A recently completed comprehensive phylogenetic framework⁷ enables us to consider the geography of butterfly phylogenetic diversity. Assessments of the phylogenetic aspect of insect biodiversity at large scale are still sparse, but offer important opportunities to determine the degree of congruence with traditional diversity measures²⁰. We find that phylogenetic diversity (PD, defined as deviation from a realm-level null expectation) peaks at about 20° latitude and at elevations of 3000 m and 4000 m (Fig 1c). This PD only weakly correlates with species richness and range rarity (SR-PD: ρ = 0.17, RR-PD: ρ = 0.29; n = 12,515), and respective global hotspots show only limited congruence (SR-PD: 29%; RR-PD: 15%). The mismatch is particularly strong in the central Neotropics, south-east Palearctic and central Indomalaya (Fig. 2), all representing global centers of PD. These centers of PD (Fig. 1c; see also Supplementary Fig. 3 for an alternative PD measure) match the previously argued hotspots of deep lineage diversification and potential evolutionary origin of butterflies based on phylogeographical reconstructions⁷. The limited congruence between PD and other measures of diversity we uncover, highlights the importance of geographically targeted efforts to protect the evolutionary heritage of butterflies.

Cross-taxon congruence

Research on global priorities for biodiversity conservation has offered important insights, but remains mostly limited to terrestrial vertebrates^10,11,13,20. Because of their ectothermic physiology and thus direct dependence on ambient temperature^34,37, we expected butterfly richness patterns to be more similar with those of amphibians, reptiles, and ants than with birds and mammals. Within ectotherm taxa, we expect the diversity of butterflies and ants to be more congruent with one another than with vertebrates, given that insects are orders of magnitude smaller and differ in both scale of habitat selection and physiological rates from vertebrates^34,35. Due to the known co-evolutionary association⁷, we also predict a particularly high association between the richness patterns of butterflies and plants. Given a much lesser dominance of a minority of wide-ranging species in driving range rarity³⁸ and a stronger signature of idiosyncratic biogeographic histories, we expect overall weaker associations for this measure than for species richness patterns³³.

We find that across the assessed 110 km-resolution assemblages, the species richness of butterflies is moderately positively associated with those of plants, mammals, birds, and amphibians (all ρ > 0.70, n = 12,515), but less so with that of ants (0.67) and reptiles (ρ = 0.55; Fig. 3). This suggests that differences in thermal strategies or ecophysiology play a limited role for driving richness differences at this scale. Instead, the environmental drivers of immigration and survival that ultimately underpin richness patterns appear shared among taxa. While the exact contributions of evolutionary, abiotic, and biotic factors shaping cross-taxon richness similarities remain largely unknown³³, our results for insects, vertebrates, and plants emphasize the potential to derive generalities of broad relevance for biodiversity^28,39,40.

Range rarity hotspots harbor many species not occurring anywhere else, and their recognition is thus of extraordinary conservation concern. Compared to geographical congruence in species richness, cross-taxon similarity in range rarity patterns is much weaker (all ρ > 0.50, n = 12,515), and particularly poor with amphibians (ρ = 0.39). Even stronger cross-taxon mismatches emerge when comparing the top 5% diversity centers per realm and taxon (Fig. 3, 4). For these hotspots, overlap among taxa is generally poor, varying from 19% to 36% for species richness and 14% to 33% for range rarity. Notably, over 41% of the butterfly range rarity hotspots – approximately 4% of the land surface – are not shared with a single terrestrial vertebrate group (Fig. 4). Only 16% of butterfly range rarity hotspots overlap with a currently recognized rarity hotspot of ants. The hotspots areas distinct to butterflies include parts of the SW US, W Madagascar, S Asian highlands and select other locations. These results indicate that priority areas identified based on vertebrates miss critical places needed to safeguard the diversity of insects. Advancing the information foundation and conservation actions for these newly recognized butterfly hotspots should be of highest urgency^17,20 (Fig. 4).

The unique role of mountains

Owing to their geographically isolated nature, strong topographic heterogeneity, and rapid environmental turnover, mountains are well-recognized for their role as catalysts for speciation and for offering refugia for species survival^27,28,41,42. All of these factors are expected to contribute to the large number of butterfly species found to be concentrated in, or endemic to, mountain regions. In addition, a broad range of insects possesses physiological adaptations to cold conditions^43–46. We thus expected mountains to play an important role in supporting exceptionally high levels of species richness and range rarity of butterflies compared to vertebrates. Our results confirm several of these expectations. Although mountains represent only 38% of the world’s terrestrial surface outside the polar regions at our study grain, they harbor 72% and 76% of global hotspots of species richness and range rarity, respectively (Figs. 2, 4; odds ratio of 4.29 and 5.24, respectively). This proportion is lower for phylogenetic diversity (59%), but still more than twice as high than expected by chance alone (odd ratio of 2.35). On the species level, these differences are reflected in the much greater portion of butterflies’ geographic ranges in mountains than elsewhere (violin plots in Fig. 2). The importance of mountains differs among realms, with generally lower mountain association in tropical compared to temperate realms (all hotspots combined: 1.43 to 7.09; bar plots in Fig. 2).

Our findings corroborate existing work recognizing the evolutionary and conservation significance of mountains for biodiversity^26–28,42, but also highlight that their role might be even greater than previously thought. Greater concentrations of insect compared to vertebrate diversity toward higher elevations have been documented at local scales⁴⁷, but at broad scales remain untested. We find that the concentration of butterfly diversity in mountains substantially exceeds that of almost all globally studied taxa, especially so for range rarity (Fig. 5a). Toward higher average elevations, butterfly richness decreases weakly, and butterfly rarity increases much more strongly than found in ants and terrestrial vertebrates. Above 2000 m, plant and butterfly richness remain relatively high, and butterfly range rarity markedly exceeds that of other taxa. We attribute this to several factors, including physiological adaptations such as color- and size-based thermoregulation of butterflies^43,46 supporting short windows of activity⁴⁸ even in cold, high-elevation settings. The remarkably strong association of butterflies with mountains seems to be further driven by their strong co-evolution with plants⁷ and its impact on resource availability in colder climates⁴⁹. Specifically, grasslands and open habitats above the tree line are known to support mountain-top endemics, including Erebia and Parnassius species highly specialized to grasses (mostly Poaceae) and other alpine plants (e.g. Sedum)^50,51. For instance, in the European Alpes alone at least 30% (1,489) of all European Lepidoptera (butterflies and moths) species occur and 15% (220) are endemic to the alpine region⁵¹. This extraordinary concentration in narrow parts of geographical and environmental space³⁹ highlights a precarious situation of butterflies in a rapidly changing world.

Projected erosion of butterfly niches

Through their role as buffers for past climatic change, their diversity of microclimatic conditions, and their isolation, mountains have repeatedly served as fundamental refugium for terrestrial global biodiversity^{26–28,31,33,42}. During the projected upcoming period of rapid global warming, the same attributes have been hypothesized to convert mountain habitats from safe havens²⁶ to graves, especially in combination with growing land-use pressures^52–54. To gauge this threat, we assess the relative temperature exposure and niche erosion of butterfly hotspots in a future warmer world using ensemble predictions of change in mean annual temperature between now and 2070 (see methods).

Assessing the availability of temperatures within realms while accounting for the inherently small portion of hotspot assemblages⁵⁵ reveals that even minor warming can have severe impacts on hotspots of species richness, range rarity, and phylogenetic diversity (Fig. 6). For example, the rare cold temperatures of Afrotropical richness hotspots are predicted to erode by 60% despite a comparatively minor temperature increase of 2.6°C (RCP 8.5). This temperature niche loss is five times greater than for Afrotropical non-hotspots, which are mainly encompassing warmer conditions. Similar trends underpin other realms, especially in the tropics. Butterfly hotspot temperature niche loss ranges from 13% to 64% (mean: 31%) for species richness, from 6% to 14% (mean: 11%) for range rarity, and from 9% to 60% (mean: 33%) for PD, respectively (Fig. 6a, Extended Data Fig. 4).

In almost all cases, the niche loss is greater for hotspots compared to non-hotspots (Fig. 6a) even though non-hotspots usually experience greater projected increase in absolute temperature (Extended Data Fig. 5). However, because temperature regimes of biodiversity hotspots are distinct and within a realm geographically much rarer compared to non-hotspots, they are more susceptible to niche loss (Fig. 6 and Extended Data Fig. 2, 3). For butterfly hotspots, we found a negative relationship between projected warming and resulting niche loss (RCP 4.5 and 8.5: Spearman’s ρ = –0.43 and –0.53; P = 0.072 and 0.023; Fig. 6c and Supplementary Fig. 4). Together, these findings reveal that under the anticipated rapid global warming, mountains do not function as safe havens, but instead might be traps for butterfly biodiversity.

Assessments of climate change impacts often neglect niche availability, for instance because they are based on absolute climate changes in focal areas^56–58, or rely on simple, binary scenarios of whether or not species might track their niches^59,60. Our results suggest that quantifying geographic niche availability is crucial for understanding threats to mountain biodiversity, due to the nuanced and gradual decline of specific temperature regimes at upslope locations following warming. These trends are mirrored by local population declines at highest elevations despite concurrent up-slope range shifts in the last decades^29,61. However, we note that additional dispersal constraints not accounted for in our analysis might cause yet greater niche losses in island endemics (e.g. range rarity hotspots in Australasia and Indomalaya) and in regions where the latitudinal orientation of mountain ranges hinders northward shifts (Fig. 6 and Extended Data Fig. 4).

Despite a growing recognition of their worldwide populations declines³, less than 1% of all insects and under 8% of butterflies have so far been assessed for their global threat status). Our study uncovered that the global conservation hotspots for butterflies vary strongly across different aspects of diversity, show limited overlap with other taxa, and are unusually strongly exposed to global warming. This threat arises from butterflies’ strong concentration at higher elevations that differs markedly from other species groups assessed globally to date. Mountains play a pivotal role for this species group yet due to their geographically rare and isolated environmental conditions are now bound to become ecological dead ends. There is an urgent need for targeted conservation strategies that address the connectivity, protection, and restoration of mountain areas where identified rarity and threat most strongly coincide. Our findings sound a clarion call for a more comprehensive global biogeographic knowledgebase to ensure the recognition of imminent threats to biodiversity and guide effective biodiversity conservation and management.

All analyses were conducted in R v.3.5.1.

Distributional data

Despite the rich natural history records for butterflies, the limited availability of expert range maps thus far hampered global-scale assessments of their diversity and conservation prioritization. We therefore used a broad spectrum of approaches and types of distributional data to fill this knowledge gap. Specifically, globally comprehensive country-level distribution information was used in cleaning all available occurrence records and digitized records from publications to model species’ distributions and generate ecoregional ranges of species (for sources see Supplementary Table 1). This information was complemented with digitized regional expert range maps. All data were taxonomically harmonized using the most up-to-date taxonomic reference^6,62. We considered Hesperiidae, Lycaenidae, Nymphalidae, Papilionidae, Pieridae, Riodinidae, but excluded Hedylidae (American moth-butterflies) due to their limited species diversity, unique ecology among butterflies, and poor representation in literature⁶³. We preferred species distribution models due to their high accuracy and resolution and used expert range maps only for species with fewer than five records or those with SDMs of poor accuracy. In the absence of expert range maps, records for species with fewer than five records or those with SDMs of poor quality were intersected with terrestrial ecoregions. The final analyses were based on 6,650 species distribution models, expert range maps for additional 881 species and ecoregional range maps for additional 5,137 species, covering a total of 65% of all butterfly species⁶. All distribution data were resampled to the same grid of assemblages with an approximate size 110 km × 110 km as a compromise between the high accuracy of species distribution models (here 1 km-resolution) and the rather low accuracy of range maps⁶⁴.

Data integration

Occurrence records. For generating SDMs and ecoregional ranges, we used 8,160,747 taxonomically harmonized and spatially cleaned records of 10,565 species (original data retrieved from gbif.org February 9, 2021; data and download filters at DOI: 10.15468/dl.92t83z)⁶. Cleaning steps for these data, included the removal of duplicated records, of non-terrestrial records and of spatial outliers. To improve the accuracy of occurrence records based on expert knowledge, we employed a validation step based on country-level occurrences from a near-complete checklist⁶. In addition, we included 11,839 digitized records of 593 species from literature for the Iran and Cambodia for which only ecoregional range maps were generated due to their lower spatial accuracy (Supplementary Fig. 2 and Table 1).

Species distribution models. Spatially explicit and cleaned occurrences recorded after the year 1970 were thinned at a 10 km-resolution⁶⁵. These records and twelve biologically relevant environmental variables⁶⁶, were used to build Maximum Entropy-based species distribution models⁶⁷ with functions of the R-package dismo. Five variables that describe annual and seasonality trends in temperature and precipitation were retrieved from chelsa.org (Bio1, Bio4, Bio10, Bio12, Bio15; CHELSA v2 current condition records^68,69). Species distribution models also included the coefficient of variation in elevation and average elevation⁷⁰, mean annual EVI (Enhanced Vegetation Index), Winter EVI and Summer EVI⁷¹, and the standard deviation of interannual variation in MODIS-based cloud cover⁷² downloaded from EarthEnv.org. In addition, as a proxy for paleoclimatic stability we used the standard deviation of mean annual temperatures across the Pleistocene calculated based on climate simulations in 10-thousand-year intervals⁷³. All variables were cropped to the same extent and if necessary resampled to a 1 km-resolution. MaxEnt models were fitted using 10,000 randomly sampled background points as suggested by ref.⁷⁴ and default settings. Note that the inclusion of more variables than occurrence records is unproblematic for MaxEnt SDMs as they evaluate predictive gain and incorporate an overfitting penalty for each predictor⁶⁷. Models were evaluated on a held-out test set consisting of a fifth of the original presences and sampled pseudo-absences. The raw MaxEnt output of species’ habitat suitability at 1 km-resolution was then converted to binary data, by using the 95% quantile of the suitability values extracted from the underlying occurrences records as presence threshold. Finally, these binary distribution data were masked with terrestrial ecoregions⁷⁵ (retrieved from OneEarth.org) that intersected with occurrence records of the species⁶⁶. 336 species distribution models with poor accuracy as determined by a low area under the curve (i.e. an AUC lower than 0.5) were visually checked by experts. 51 of them were discarded because the predictions were evaluated as unrealistic. The median AUC of the remaining 6,650 species distribution models was 0.93 (mean: 0.87). The final binary species distribution maps were reassigned to our equal-area grid⁷⁶ and only range fragments overlapping more than 50% with an assemblage were considered presences.

Expert and ecoregional range maps. 10,314 vector distribution maps for 5,303 butterfly species from 35 field guides were georeferenced, quality controlled, taxonomically harmonized^6,62 and spatially merged (Supplementary Table 1). However, for many species both sufficient records for SDMs as well as expert range maps were lacking, and these data shortfall was most severe in tropical regions and taxa (Supplementary Fig. 6). We therefore adopted a regionalization scheme commonly used in global-scale conservation for these species, if at least one record was available. This scheme, the ecoregions of the world, represents a downscaling of biogeographical realms based on expert synthesis and species turnover⁷⁵. The 5,137 generated ecoregional range maps⁶⁶ were highly spatially and taxonomically complementary to SDMs and expert range maps, particularly for tropical regions and taxa (Supplementary Fig. 2). We preferred ecoregional range maps over alternative approaches, such as alpha hulls and geographical buffers, as they have been shown to be more concordant with expert range maps and SDMs, particularly for data-poor species and because they span ecologically meaningful extents rather than being based on spatial proximity alone⁶⁶. Both expert and ecoregional range maps were reassigned to our equal-area grid and only range fragments overlapping more than 50% with an assemblage were considered presences.

Diversity aspects. Species richness (SR) was calculated as the count of co-occurring species per assemblage. As a measure of endemism, we calculated the range rarity (‘RR’) as the mean of the inverse occupancy (total count of assemblages globally/ count of assemblages occupied by a species) of co-occurring species. Our choice of this measure, in contrast to the one employed in pioneering studies on vertebrates and plants^9,10, was driven by the absence of established thresholds for endemic species and regions crucial for butterfly conservation. Additionally, range rarity proved to be less depended on species richness³⁸, making it a more effective complement for the conservation of biogeographical unique and highly threatened species¹¹.

To also incorporate concentrations of distantly related species into our prioritization, we calculated the phylogenetic distinctiveness of co-occurring butterfly species (hereafter ‘phylogenetic diversity’ or ‘PD’ for simplicity). We used a recently published genus-level phylogeny⁷ and the taxonomic names of all accepted butterfly species⁶, to add all butterfly species for which distribution data was available at their respective genus and randomly resolved the intra-genus relationships with functions of the R-package phytools⁷⁷. Species were not added if they were represented by only one taxon in the phylogeny or if the genus to which they belong was paraphyletic. Note that this phylogenetic tree included only 11,143 species, while species richness and range rarity were calculated based on the distribution data for 12,119 species. For the calculation of phylogenetic diversity, the tree was pruned to include only the regional (realm-specific) species pool. Phylogenetic diversity was calculated as the standardized effect sizes of mean pairwise distances (MPD) across co-occurring species, i.e., the observed MPD minus the random MPD divided by the standard deviation of the random MPD values, to account for the effect of species richness⁷⁸. Random MPD represents a null model expectation of the phylogenetic composition of an assemblage based on the average of 10,000 calculations of the MPD for the same number, but a randomly drawn sample, of species per assemblage. As for range rarity the choice of this measure was driven by its empirical independence of species richness and range rarity, compared to, for instance, Faith’s phylogenetic diversity (Supplementary Fig. 3) or phylogenetic endemism²⁰, and consequently its greater emphasis on unique phylogenetic aspects. We restricted the species pool to the respective biogeographical realms to accommodate species turnover resulting from past and present geographical dispersal barriers. Using Chao’s dissimilarity index with species’ range size as weight, we confirmed that the selected realm definition⁷⁵ effectively delimited highly dissimilar species pools (Supplementary Table 2). Only Oceania was included in the Australasian realm due to its small extent and a similar species composition of the two realms.

Hotspots of diversity. Hotspots of diversity patterns for all taxa were arbitrarily defined as the 5% assemblages with the highest SR, RR, and PD per realm, respectively. PD was transformed to positive values for this purpose [PD+(min(PD*-1))]. We identified hotspots at the realm scale, instead of the global scale, to ensure accurate representation of the three complementary aspects of diversity within and between regions²⁰, while minimizing the impact of potential sampling biases. Leveraging coarse, but near-complete country-level distribution data, we document the species coverage of our high-resolution data (Supplementary Figs. 5, 6). With this information on data shortfalls, we also assessed the robustness of our realm-level hotspot definition. Specifically, we multiplied the respective diversity estimate per assemblage by the inverse of the country-specific species coverage and compared our core hotspots with those derived based on the weighted diversity estimates (Supplementary Fig. 8). For instance, while the species richness for a given assemblage that falls into a country with 100% species coverage remains the same that of an assemblage with a lower coverage in a data-poor country is upweighted by the factor 1.67 (1/0.60 in the most extreme case) before hotspots are defined. This allowed an evaluation of the robustness of hotspot locations for all diversity aspects, based on the assumption that the magnitude, but not the relative within-country variation of diversity, will change with the addition of species. Thus, greater emphasis is placed on data-poor countries across each realm. We thereby confirm that only 19% of the combined hotspots of SR, RR and PD do not overlap with any of the coverage weighted ones (17% for SR, 11% for RR, and 37% for PD, separately; Supplementary Fig. 8). Strongest mismatches highlight potential shifts in PD hotspot locations from lowland Japan to highlands in southern China and from lowlands of Borneo to highlands of New Guinea. However, note that our measure of PD is richness-corrected and should be hence robust to the sampling of additional species that by chance alone add phylogenetic diversity to an assemblage. Following ref.¹⁰, we also evaluated the influence of different hotspot definitions in percent top-ranking assemblages by recalculating overlap of diversity aspects for the top 1–100% assemblages in 1%-steps. This analysis showed that the increase in congruence among the three diversity aspects is proportional to the increase in area until approximately 20% of all assemblages are considered (Supplementary Fig. 9). For butterflies both congruence and overlap were compared between SR, RR, and PD. To assess to congruence and hotspot overlap of butterfly diversity with terrestrial vertebrate diversity, expert range maps for 32,851 amphibians (https://www.iucnredlist.org/resources/spatial-data-download), reptiles²¹, mammals⁷⁹, and birds⁸⁰ were retrieved, and then resampled to our grid. We also leveraged ant richness and rarity data based on 14,324 (sub-)species²³ as well as a prediction of plant richness based on more than 10,000 species (maximum regional estimate)⁸¹. Layers of plant and ant diversity data had a two times finer resolution and were therefore averaged per intersecting cell of our grid. Although thus far, this ant and plant data have not been explicitly factored into the establishment of conservation priorities, they serve as valuable benchmarks for comparing elevational gradients and assessing the prevalence of hotspots in mountainous regions. This is particularly important due to the co-evolution of butterflies with plants⁷ and the potential ecological parallels among insects²³. Hotspots of all taxa were calculated in the same way as for butterflies, yielding in total 626 hotspot assemblage for each vertebrate taxon-diversity aspect combination, but fewer for ants and plants (622 and 592, respectively) due a slightly lower spatial coverage of the underlying data.

Mountain diversity

To evaluate the likelihood that a hotspot was located in mountains, we took into account that mountain regions cover a smaller proportion (38% at our grain) of the terrestrial area of the Earth and that hotspot regions represent only 5% of all assemblages within each realm, as per our definition. We intersected our grid with the most up-to-date classification of mountains from a global inventory (GMBA)⁸, considering cells as mountainous if they were covered by more than 5% of the area. We chose this threshold to exclude smaller mountain fragments, but also accommodate the rather coarse resolution of the distribution data. For each realm and diversity aspect the proportion of hotspot areas in mountains was divided by the proportion of hotspot areas in lowlands. For instance, the likelihood of range rarity hotspots being located in mountains is 5.24 times greater (472 of 4,696 grid cells ≈ 10.0%) than that for lowlands (150 of 7,819 grid cells ≈ 1.9%). For species-level analyses, the proportion of species’ ranges in mountains was calculated based on those species that occur in each respective type of hotspot and realm. As the regional baseline for this measure, we calculated the proportion of grid cells that are mountains on the total number of grid cells per realm. Species whose ranges cover a higher proportion of mountains than the baseline (disproportionately more mountain) were considered primarily mountain-dwelling.

Temperature difference and niche loss under global change scenarios

For analysis of the impact of temperature changes on species assemblages in hotspots and non-hotspots, we generated ensembles of mean annual temperature across predictions from complementary climate models for the period of 2061–2080. These models provided predictions for two representative carbon pathways, one with net carbon emissions of zero until 2050 and one with continued emissions until the end of this century (RCP 4.5 and 8.5 scenario), as defined by the Intergovernmental Panel on Climate Change⁸². We focused on mean annual temperature as it is the most important and general determinant of species’ distributions and diversity patterns^28,33,59. The selection of eight models (CESM1-BGC, CESM1-CAM5, CMCC-CM, FIO-ESM, INMCM4, IPSL-CM5A, MIROC5, MPI-ESM-MR) from climatologies available at chelsa.org (v1.2 CMIP5) followed ref.⁸³ in using predictions with low interdependence. Temperature values were averaged across predictions for each of the two RCP’s and aggregated to mean values per grid cell. The temperature difference was calculated from these ensembles by subtracting the current mean annual temperature from that of future temperature ensembles (Extended Data Fig. 1). Absolute temperature differences are a common measure to identify regions with high potential threat to biodiversity under future climate change (i.e. climate change exposure). However, here we argue that even small difference in temperature can have drastic impacts on biodiversity as certain climates and particularly hotspot climates might be rarer than those of non-hotspots. To address this issue, we calculated the loss of the current temperature niche of hotspots and their corresponding non-hotspots for each realm and diversity aspect based on density overlap in a Bayesian framework⁵⁵. Frequency distributions were converted to densities with a sample size of 1,000 samples and the default probability region size of 95% was used for 1,000 repeated calculations (Extended Data Fig. 4). Overlap was calculated as the probability that a given current temperature of an assemblage is found in the probability region of future temperatures, and the non-overlapping part of the current conditions we denominate the ‘temperature niche loss’ (see Fig. 6c). Unlike other niche overlap metrics, this probabilistic measure explicitly accounted for uncertainty in the niche regions definition and for an uneven number of compared entities (i.e. in our case grid cells). Note that the direction of comparison matters for this calculation, and we did not discuss the gain in temperature niche (compared future with current temperature) because novel very warm climates will not maintain current biodiversity. In addition to that of hotspots and non-hotspots, we also calculated the niche loss for a random subset of 5% of the number of assemblages per realm to explore whether niche loss is affected by the inherently lower count of hotspots. This analysis confirmed the robustness of our niche loss estimate to differences in sample size, being similar between the small set of 5% random assemblages and the 95% of non-hotspot assemblages per realm (Extended Data Fig. 4).

Data gaps

Despite extensive data mobilization and integration efforts (Supplementary Figs. 1, 2 and Table 1) one third of the described butterfly species lacked sufficient data to be included, particularly for the Afrotropic (42% missing), Indomalayan, and Neotropic realms (34% and 32%; Supplementary Figs. 5, 6). We therefore assessed hotspots within realms to limit the impact of these main differences in data coverage²⁰, but also evaluated the robustness of within-realm patterns. We expect regions with particularly large portions of unmapped species to see refinements to their hotspot locations using coarse but near-complete country-level distribution data⁶. We find that among countries, species-coverage is negatively correlated with the count of identified hotspots, contrasting the expected effect of sampling bias (Spearman’s ρ = − 0.43, P = 0.001; Supplementary Fig. 7). Adapting the hotspot definition with a country-level species coverage weight based on these data, our supplementary analysis confirmed only minor changes in the locations of species richness and rarity hotspots (83% and 89% remain unchanged, Supplementary Fig. 8). Hence, although many species are added to certain countries this generally does not impact realm-level priorities. Stronger mismatch (37% globally) is observed for hotspots of phylogenetic diversity, highlighting mountains in New Guinean and southern China. These areas are important targets for additional sampling of phylogenetic data, but note that their confirmation would shift priorities mainly from lowlands to highlands.

Data availability

All produced hotspots and diversity maps for butterflies will be made available through map.half-earthproject.org and mol.org upon acceptance. Distribution data for amphibians can be requested from IUCN (https://www.iucnredlist.org/resources/spatial-data-download). Distribution data for birds⁸⁰, mammals⁷⁹ and reptiles²¹ can be requested from authors of the respective publications. Plant species richness predictions⁸¹ are available from the authors upon request. Ant richness and rarity data²³ is available at DataDryad DOI:10.5061/dryad.wstqjq2pp. The layer of terrestrial ecoregions used to generate ecoregional range maps for butterfly species can be retrieved from OneEarth.org. Digitized expert range maps are available from mol.org through the source links in the Supplementary Information. Global country-level distribution information for butterflies⁶ can be retrieved from https://mol.org/datasets/149345d9-eb27-4140-ba60-eaa32ff08ff8. The phylogenetic tree for butterflies is available from the authors⁷. Any additional data not listed can be made available upon reasonable request from the authors of this article.

Code availability

No custom code or mathematical algorithm was used in the analysis. Code to analyze niche loss and the proportion of hotspots in mountains will be uploaded to figshare upon acceptance.

Competing interests The authors declare no competing interests.

Seibold, S. et al. Arthropod decline in grasslands and forests is associated with landscape-level drivers. Nature 574, 671–674 (2019).
Seibold, S. et al. The contribution of insects to global forest deadwood decomposition. Nature 597, 77–81 (2021).
Wagner, D. L., Grames, E. M., Forister, M. L., Berenbaum, M. R. & Stopak, D. Insect decline in the Anthropocene: Death by a thousand cuts. Proc. Natl. Acad. Sci. 118, (2021).
Thomas, J. A. Monitoring change in the abundance and distribution of insects using butterflies and other indicator groups. Philos. Trans. R. Soc. B Biol. Sci. 360, 339–357 (2005).
IPBES. The Assessment Report on Pollinators, Pollination and Food Production: Summary for Policymakers. (2016).
Pinkert, S., Barve, V., Guralnick, R. & Jetz, W. Global geographical and latitudinal variation in butterfly species richness captured through a comprehensive country‐level occurrence database. Glob. Ecol. Biogeogr. 31, 830–839 (2022).
Kawahara, A. Y. et al. A global phylogeny of butterflies reveals their evolutionary history, ancestral hosts and biogeographic origins. Nat. Ecol. Evol. 7, 903–913 (2023).
Snethlage, M. A. et al. A hierarchical inventory of the world’s mountains for global comparative mountain science. Sci. Data 9, 149 (2022).
Myers, N., Mittermeier, R. A., Mittermeier, C. G., Da Fonseca, G. A. & Kent, J. Biodiversity hotspots for conservation priorities. Nature 403, 853–858 (2000).
Orme, C. D. L. et al. Global hotspots of species richness are not congruent with endemism or threat. Nature 436, 1016–1019 (2005).
Lamoreux, J. F. et al. Global tests of biodiversity concordance and the importance of endemism. Nature 440, 212–214 (2006).
Jetz, W. et al. Essential biodiversity variables for mapping and monitoring species populations. Nat. Ecol. Evol. 3, 539–551 (2019).
Grenyer, R. et al. Global distribution and conservation of rare and threatened vertebrates. Nature 444, 93–96 (2006).
Oliver, R. Y., Meyer, C., Ranipeta, A., Winner, K. & Jetz, W. Global and national trends, gaps, and opportunities in documenting and monitoring species distributions. PLOS Biol. 19, e3001336 (2021).
Jetz, W. et al. Include biodiversity representation indicators in area-based conservation targets. Nat. Ecol. Evol. 6, 123–126 (2022).
Jung, M. et al. Areas of global importance for conserving terrestrial biodiversity, carbon and water. Nat. Ecol. Evol. 5, 1499–1509 (2021).
Pollock, L. J., Thuiller, W. & Jetz, W. Large conservation gains possible for global biodiversity facets. Nature 546, 141–144 (2017).
Owen, N. R., Gumbs, R., Gray, C. L. & Faith, D. P. Global conservation of phylogenetic diversity captures more than just functional diversity. Nat. Commun. 10, 859 (2019).
Brodie, J. F., Williams, S. & Garner, B. The decline of mammal functional and evolutionary diversity worldwide. Proc. Natl. Acad. Sci. 118, (2021).
Voskamp, A. et al. Utilizing multi-objective decision support tools for protected area selection. One Earth 6, 1143–1156 (2023).
Roll, U. et al. The global distribution of tetrapods reveals a need for targeted reptile conservation. Nat. Ecol. Evol. 1, 1677–1682 (2017).
Powers, R. P. & Jetz, W. Global habitat loss and extinction risk of terrestrial vertebrates under future land-use-change scenarios. Nat. Clim. Change 9, 323–329 (2019).
Kass, J. M. et al. The global distribution of known and undiscovered ant biodiversity. Sci. Adv. 8, eabp9908 (2022).
Pigot, A. L. et al. Macroevolutionary convergence connects morphological form to ecological function in birds. Nat. Ecol. Evol. 4, 230–239 (2020).
Sandel, B. et al. The influence of late Quaternary climate-change velocity on species endemism. Science 334, 660–664 (2011).
Keppel, G. et al. Refugia: identifying and understanding safe havens for biodiversity under climate change. Glob. Ecol. Biogeogr. 21, 393–404 (2012).
Antonelli, A. et al. Geological and climatic influences on mountain biodiversity. Nat. Geosci. 11, 718–725 (2018).
Rahbek, C. et al. Humboldt’s enigma: What causes global patterns of mountain biodiversity? Science 365, 1108–1113 (2019).
Freeman, B. G., Scholer, M. N., Ruiz-Gutierrez, V. & Fitzpatrick, J. W. Climate change causes upslope shifts and mountaintop extirpations in a tropical bird community. Proc. Natl. Acad. Sci. 115, 11982–11987 (2018).
Elsen, P. R. & Tingley, M. W. Global mountain topography and the fate of montane species under climate change. Nat. Clim. Change 5, 772–776 (2015).
Chan, W.-P. et al. Climate velocities and species tracking in global mountain regions. Nature 1–7 (2024) doi:10.1038/s41586-024-07264-9.
La Sorte, F. A. & Jetz, W. Projected range contractions of montane biodiversity under global warming. Proc. Biol. Sci. 277, 3401–3410 (2010).
Pinkert, S. et al. Climate–diversity relationships underlying cross‐taxon diversity of the African fauna and their implications for conservation. Divers. Distrib. 10, 1330–1342 (2020).
Gillooly, J. F., Brown, J. H., West, G. B., Savage, V. M. & Charnov, E. L. Effects of size and temperature on metabolic rate. Science 293, 2248–2251 (2001).
Gillooly, J. F., Charnov, E. L., West, G. B., Savage, V. M. & Brown, J. H. Effects of size and temperature on developmental time. Nature 417, 70–73 (2002).
Buckley, L. B., Hurlbert, A. H. & Jetz, W. Broad-scale ecological implications of ectothermy and endothermy in changing environments. Glob. Ecol. Biogeogr. 21, 873–885 (2012).
Rohde, K. Latitudinal gradients in species diversity: The search for the primary cause. Oikos 65, 514 (1992).
Jetz, W., Rahbek, C. & Colwell, R. K. The coincidence of rarity and richness and the potential signature of history in centres of endemism. Ecol. Lett. 7, 1180–1191 (2004).
Coelho, M. T. P. et al. The geography of climate and the global patterns of species diversity. Nature 622, 537–544 (2023).
Hawkins, B. A. et al. Different evolutionary histories underlie congruent species richness gradients of birds and mammals: Bird and mammal richness gradients. J. Biogeogr. 39, 825–841 (2012).
Körner, C. & Spehn, E. A Humboldtian view of mountains. Science 365, 1061–1061 (2019).
Quintero, I. & Jetz, W. Global elevational diversity and diversification of birds. Nature 555, 246–250 (2018).
Pinkert, S. & Zeuss, D. Thermal biology: Melanin-based energy harvesting across the tree of life. Curr. Biol. 884–887 (2018).
Brehm, G., Zeuss, D. & Colwell, R. K. Moth body size increases with elevation along a complete tropical elevational gradient for two hyperdiverse clades. Ecography 42, 632–642 (2019).
Wu, S. et al. Artificial intelligence reveals environmental constraints on colour diversity in insects. Nat. Commun. 10, 1–9 (2019).
Heidrich, L. et al. Noctuid and geometrid moth assemblages show divergent elevational gradients in body size and color lightness. Ecography 44, 1169–1179 (2021).
Peters, M. K. et al. Predictors of elevational biodiversity gradients change from single taxa to the multi-taxa community level. Nat. Commun. 7, 13736 (2016).
Novella-Fernandez, R., Brandl, R., Pinkert, S., Zeuss, D. & Hof, C. Seasonal variation in dragonfly assemblage colouration suggests a link between thermal melanism and phenology. Nat. Commun. 14, 8427 (2023).
Chazot, N. et al. Conserved ancestral tropical niche but different continental histories explain the latitudinal diversity gradient in brush-footed butterflies. Nat. Commun. 12, 5717 (2021).
Clarke, H. E. A provisional checklist of European butterfly larval foodplants. Nota Lepidopterol. 45, 139–167 (2022).
Glerean, P., Deutsch, H., Morandini, C., Morin, L. & Huemer, P. Lepidoptera of the Prealpi Giulie natural park (Friuli Venezia Giulia, north-east Italy). Gortania Bot. Zool. 44, 29–72 (2022).
Williams, J. N. Humans and biodiversity: Population and demographic trends in the hotspots. Popul. Environ. 34, 510–523 (2013).
Peters, M. K. et al. Climate–land-use interactions shape tropical mountain biodiversity and ecosystem functions. Nature 568, 88–92 (2019).
Li, B. V., Jenkins, C. N. & Xu, W. Strategic protection of landslide vulnerable mountains for biodiversity conservation under land-cover and climate change impacts. Proc. Natl. Acad. Sci. 119, e2113416118 (2022).
Swanson, H. K. et al. A new probabilistic method for quantifying n-dimensional ecological niches and niche overlap. Ecology 96, 318–324 (2015).
Bellard, C. et al. Vulnerability of biodiversity hotspots to global change. Glob. Ecol. Biogeogr. 23, 1376–1386 (2014).
Brito-Morales, I. et al. Climate velocity can inform conservation in a warming world. Trends Ecol. Evol. 33, 441–457 (2018).
Dobrowski, S. Z. et al. Protected-area targets could be undermined by climate change-driven shifts in ecoregions and biomes. Commun. Earth Environ. 2, 1–11 (2021).
Trisos, C. H., Merow, C. & Pigot, A. L. The projected timing of abrupt ecological disruption from climate change. Nature 580, 496–501 (2020).
Biber, M. F., Voskamp, A. & Hof, C. Potential effects of future climate change on global reptile distributions and diversity. Glob. Ecol. Biogeogr. 32, 519–534 (2023).
Halsch, C. A. et al. Insects and recent climate change. Proc. Natl. Acad. Sci. 118, e2002543117 (2021).
Sandall, E. et al. A globally integrated structure of taxonomy supporting biodiversity science and conservation. Trends Ecol. Evol. 38, 1143–1153 (2023).
Kawahara, A. Y. et al. Phylogenetics of moth-like butterflies (Papilionoidea: Hedylidae) based on a new 13-locus target capture probe set. Mol. Phylogenet. Evol. 127, 600–605 (2018).
Hurlbert, A. H. & Jetz, W. Species richness, hotspots, and the scale dependence of range maps in ecology and conservation. Proc. Natl. Acad. Sci. 104, 13384–13389 (2007).
Aiello-Lammens, M. E. et al. spThin: Functions for spatial thinning of species occurrence records for use in ecological models. (2019).
Pinkert, S., Sica, Y. V., Winner, K. & Jetz, W. The potential of ecoregional range maps for boosting taxonomic coverage in ecology and conservation. Ecography 2023, e06794 (2023).
Phillips, S. J., Anderson, R. P. & Schapire, R. E. Maximum entropy modeling of species geographic distributions. Ecol. Model. 190, 231–259 (2006).
Karger, D. N. et al. Climatologies at high resolution for the earth’s land surface areas. Sci. Data 4, 170122 (2017).
Karger, D. N. et al. Data from: Climatologies at high resolution for the earth’s land surface areas - Dryad. https://datadryad.org/resource/doi:10.5061/dryad.kd1d4 (2018).
Amatulli, G. et al. A suite of global, cross-scale topographic variables for environmental and biodiversity modeling. Sci. Data 5, 180040 (2018).
Tuanmu, M.-N. & Jetz, W. A global, remote sensing-based characterization of terrestrial habitat heterogeneity for biodiversity and ecosystem modelling. Glob. Ecol. Biogeogr. 24, 1329–1339 (2015).
Wilson, A. M. & Jetz, W. Remotely sensed high-resolution global cloud dynamics for predicting ecosystem and biodiversity distributions. PLOS Biol. 14, e1002415 (2016).
Gamisch, A. Oscillayers: A dataset for the study of climatic oscillations over Plio-Pleistocene time-scales at high spatial-temporal resolution. Glob. Ecol. Biogeogr. 28, 1552–1560 (2019).
Valavi, R., Guillera-Arroita, G., Lahoz-Monfort, J. J. & Elith, J. Predictive performance of presence-only species distribution models: a benchmark study with reproducible code. Ecol. Monogr. 92, e01486 (2022).
Dinerstein, E. et al. An ecoregion-based approach to protecting half the terrestrial realm. BioScience 67, 534–545 (2017).
Pebesma, E. et al. sp: Classes and methods for spatial data. (2022).
Revell, L. J. phytools: Phylogenetic tools for comparative biology. (2017).
Webb, C. O., Ackerly, D. D., McPeek, M. A. & Donoghue, M. J. Phylogenies and community ecology. Annu. Rev. Ecol. Syst. 33, 475–505 (2002).
Marsh, C. J. et al. Expert range maps of global mammal distributions harmonised to three taxonomic authorities. J. Biogeogr. 49, 979–992 (2022).
Jetz, W., Thomas, G. H., Joy, J. B., Hartmann, K. & Mooers, A. O. The global diversity of birds in space and time. Nature 491, 444–448 (2012).
Kreft, H. & Jetz, W. Global patterns and determinants of vascular plant diversity. Proc. Natl. Acad. Sci. 104, 5925–5930 (2007).
Stocker, T. F. et al. IPCC - Intergovernmental Panel on Climate Change Report 2013 - The Physical Science Basis. 1535 pp https://www.ipcc.ch/report/ar5/wg3/ (2013).
Sanderson, B. M., Knutti, R. & Caldwell, P. A representative democracy to reduce interdependency in a multimodel ensemble. J. Clim. 28, 5171–5194 (2015).

There is NO Competing Interest.

SuppInfoGlobalbutterflydiversity.docx
Supplementary Information
EXTENDEDDATA.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Global hotspots of butterfly diversity in a warming world

Status:

Version 1

Abstract

Figures

Main

Methods

Distributional data

Data integration

Mountain diversity

Temperature difference and niche loss under global change scenarios

Data gaps

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1