Climate change and land use threaten global hotspots of phylogenetic endemism for trees

The study addresses where global hotspots of phylogenetic endemism (PE) for trees are located, what environmental factors (current and past climates, topography) drive these patterns, and how vulnerable these hotspots are to human pressures and future climate change. Trees underpin ecosystems and human well-being but many species—especially range-restricted, evolutionarily distinct endemics—face high risks from habitat loss, overexploitation, and invasions. PE captures how evolutionary history is geographically restricted and, via CANAPE, distinguishes paleo- (ancient lineages) and neo-endemism (recently diverged lineages). Although PE centers are known for some taxa, a global assessment for trees has been lacking. Prior work highlights current climate as a key driver of endemism, with long-term climatic stability promoting lineage persistence and diversification, while high climatic oscillations can foster turnover and neo-endemism. The authors hypothesize that tree PE hotspots are linked to both present climate and paleoclimatic stability, and that these hotspots are inadequately protected yet exposed to high human modification and substantial future climate change.

The paper synthesizes work showing: (1) widespread anthropogenic pressures on trees and insufficient protection for range-restricted species; (2) the conceptual development of phylogenetic diversity and endemism, and the CANAPE framework to categorize paleo- vs. neo-endemism; (3) mixed evidence across taxa on the geography of endemism centers and examples from butterflies, beetles, vertebrates, and select tree groups; (4) the roles of current climate and paleoclimate in shaping biodiversity, with stable climates over geological times favoring diversification and persistence, while glacial-interglacial oscillations promote turnover and neo-endemism; and (5) the conservation relevance of identifying PE hotspots given their evolutionary uniqueness and functional importance. The gap identified is the absence of a comprehensive, global assessment of tree PE centers and their exposure to human and future climate threats.

- Data: Compiled range maps for 41,835 tree species (65.1% of known species) using GlobalTreeSearch v1.6, standardized via TNRS, with occurrences from GBIF, BIEN, DRYFLOR, RAINBIO, and Atlas of Living Australia. Range maps were generated using alpha hull algorithms and validated against independent datasets. Data were rasterized to 110×110 km equal-area grid cells, retaining cells with ≥5 species. - Phylogeny: Constructed a dated seed-plant megaphylogeny (ALLMB), integrating a sequence-based backbone and Open Tree of Life taxonomy. Missing species were grafted at genus/family level. The tree was pruned to species with distributions, and split into angiosperms (n=41,275) and gymnosperms (n=560). - Spatial phylogenetics: Computed phylogenetic endemism (PE) per grid cell in Biodiverse, scaling branch lengths by descendant clade range size. Calculated relative PE (RPE = PEoriginal/PEequal). Conducted 999-tip-shuffle randomizations preserving richness and range-size structures to identify cells significantly higher/lower than expected (two-tailed α=0.05). Classified endemism types via CANAPE: paleo- (significantly high RPE with PEoriginal > PEequal), neo- (high RPE with PEoriginal < PEequal), mixed (both PEoriginal and PEequal high without significant RPE), and non-hotspots. - Environmental predictors: Present climate (CHELSA, 1970–2000) MAT and AP; paleoclimate anomalies (past minus present) for the Late Miocene (11.61–7.25 Mya) and Last Glacial Maximum (21 kya); topography via elevation range (SRTM 90 m), aggregated to 110×110 km. - Modeling drivers: Ordinary least squares and spatial simultaneous autoregressive error models (SAR) with standardized predictors (MAT, AP, four anomaly variables, elevation range). Checked multicollinearity (VIF<4) and residual spatial autocorrelation (Moran’s I). Reported SAR results (better AIC/BIC/R²). - Hotspot vs. non-hotspot contrasts: Two-sample Fisher-Pitman permutation tests (100,000 permutations) comparing environmental variables between hotspots and non-hotspots; further comparisons among hotspot types. - Human pressure and future climate: Human Modification Index (median year 2016) at 1 km² resolution. Future MAT and AP (2071–2100) from CHELSA v2 CMIP6 (5 GCMs × SSP126/370/585). Computed anomalies (future minus present), focusing on SSP370 due to high inter-scenario correlations. - Protection analyses: Existing protected areas from WDPA (Nov 2021), filtered to terrestrial IUCN I–VI, statuses designated/inscribed/established, excluding UNESCO MAB and points. Assessed coverage of hotspots by existing PAs and by three tree-diversity prioritization frameworks (top 17%, 30%, 50%) that optimize taxonomic, phylogenetic, and functional tree diversity. - Sensitivity: Parallel analyses at 50×50 km and 220×220 km (gymnosperms at 50 km completed; general consistency across resolutions).

- Geography of PE: High angiosperm PE in Central America, northern Andes, east Madagascar, southwestern China, northern Borneo and Peninsular Malaysia; high gymnosperm PE in southern China, northeastern Borneo, Japan, Tasmania, and Fiji. - Hotspot prevalence and types: 24.3% of angiosperm cells (1508/6198) and 12.0% of gymnosperm cells (145/1200) are PE hotspots (neo-, paleo-, or mixed). Angiosperms: mixed-endemism dominates (22.9%, 1419 cells); pure neo- 0.8% (49), pure paleo- 0.7% (40). Gymnosperms: mixed 10.3% (123), neo 1.2% (14), paleo 0.7% (8). There are 128 shared hotspot cells, constituting 8.5% of angiosperm hotspots and 88.3% of gymnosperm hotspots. Overall, about 20.8% of grid cells (with ≥5 species) are hotspots for angiosperms, gymnosperms, or both. - Environmental drivers (SAR models): Explained variance: >90% (angiosperms) and 84% (gymnosperms). Present-day AP is a dominant positive predictor for both groups; MAT shows an even stronger positive effect for angiosperms. Elevation range is positively associated with PE. LGM anomalies (AP and MAT; past minus present) relate positively to PE for both groups, indicating higher PE where LGM conditions were relatively warm and wet. Miocene anomalies show no relation for gymnosperms; for angiosperms, Miocene MAT anomaly is positive and Miocene AP anomaly negative, suggesting higher PE where Miocene was warmer and drier relative to today. - Hotspot environments vs. non-hotspots: Both angiosperm and gymnosperm hotspots have greater elevation ranges, higher MAT and AP, and generally smaller paleoclimatic anomalies (lower long-term climatic variability) than non-hotspots (p<0.001). Neo-endemism hotspots tend to occur in cooler, less wet environments with colder LGM MAT and, for angiosperms, higher Miocene AP, compared to mixed/paleo hotspots. - Human pressure: Hotspots have higher Human Modification Index values than non-hotspots (p<0.001); gymnosperm-only hotspots are particularly impacted. - Future climate exposure: Angiosperm-only hotspots are projected to experience greater warming (p<0.001) than other hotspot categories; both separate (angiosperm-only or gymnosperm-only) and joint hotspots are projected to see greater precipitation increases than non-hotspots (p<0.001). - Protection status: Existing protected areas cover only 7.4% of angiosperm-only hotspots and 8.7% of joint hotspots; gymnosperm-only hotspots have 0% coverage. Prioritization frameworks would markedly improve coverage: Top 17% areas would cover 67.9% (angiosperm-only), 36.0% (gymnosperm-only), 78.6% (joint); Top 30%: 90.4%, 68%, 92.1%; Top 50%: ≥96% for all hotspot types. Non-hotspot coverage would also rise from 6.1% (existing) to 60.9% (top 50% priority areas).

Findings confirm that global tree PE hotspots concentrate at low-to-mid latitudes and are shaped by both current climate (especially temperature and precipitation) and long-term climatic stability. The strong prevalence of mixed-endemism centers underscores the role of stable environments in both preserving ancient lineages and fostering recent diversification. Positive associations with warm/wet LGM conditions and Miocene anomalies (for angiosperms) indicate deep-time legacies beyond Pleistocene dynamics. Hotspots’ elevated human modification and higher exposure to near-future climate change (greater warming for angiosperm-only hotspots and higher precipitation increases broadly) heighten extinction risks for range-restricted, evolutionarily unique tree lineages. Protection gaps are stark—particularly for gymnosperm-only hotspots—which, alongside projected climate shifts, necessitate ambitious conservation responses. The congruence of tree PE hotspots with those of other taxa highlights that tree-focused conservation can deliver broad biodiversity co-benefits across ecosystems.

This work provides a global map of tree phylogenetic endemism, identifies its main macroecological drivers across present and paleo-timescales, and quantifies hotspots’ exposure to human pressure and future climate change. Hotspots are predominantly in low-to-mid latitudes, linked to high current temperature and precipitation, complex topography, and long-term climatic stability. Yet they are insufficiently covered by existing protected areas and face strong land-use and climate threats. Implementing tree-diversity priority area frameworks (top 17%, 30%, 50%) would greatly improve hotspot coverage, with near-complete protection at the 50% target. Effective protected area expansion, coupled with complementary measures such as assisted colonization and ex-situ conservation where appropriate, is urgently needed to safeguard evolutionarily unique, range-restricted tree species and the ecosystem functions they support.

- Incomplete species coverage: analyses include 41,835 species (~65.1% of known tree species), potentially missing patterns in under-sampled regions/taxa. - Range estimation uncertainties: alpha hull–based range maps derived from occurrence records may contain sampling and methodological biases despite external validations. - Phylogenetic uncertainties: the megaphylogeny includes polytomies and grafted species at higher taxonomic levels; while prior work suggests minimal bias for broad-scale analyses, fine-scale relationships remain uncertain. - Spatial resolution: primary analyses at 110×110 km (with sensitivity checks at 50 and 220 km); patterns at finer scales may differ, and angiosperm analyses at 50 km were computationally infeasible. - Climate reconstructions: paleoclimate and future climate data carry model uncertainties; future projections based on five CMIP6 GCMs and focus on SSP370 anomalies may not capture full scenario/model spread. - Protected area data: WDPA (Nov 2021) filtering choices (e.g., exclusion of point PAs, UNESCO MAB) and resampling to coarse grids may under/overestimate coverage. - Thresholding: exclusion of cells with <5 species may bias results in species-poor regions.