More than 17,000 tree species are at risk from rapid global change

Trees underpin ecosystem functioning and biodiversity, providing habitat for a large fraction of terrestrial species while delivering cultural, economic, and livelihood benefits. Rapid intensification of human activities in the Anthropocene has driven habitat loss, fragmentation, degradation, over-exploitation, and other pressures that elevate extinction risk for many tree species. Existing automated extinction risk assessments rarely incorporate temporal dynamics of threats, despite evidence that changing threat landscapes can substantially increase risk. The Global Tree Assessment (GTA) has identified nine major threats to trees and assessed conservation status for most species, but many species remain Data Deficient and some threats (e.g., climate change) are hard to diagnose. This study aims to quantify recent rates of change in exposure to six key threats (cropland expansion, tree cover decline, urban expansion, deforestation, burned area change, and climate change via temperature, precipitation, and vapor pressure deficit metrics) within species’ extents, evaluate congruence with current IUCN Red List categories, prioritize species for expert re-evaluation, and map global hotspots of species exposed to rapid change.

The authors situate their work within efforts such as the Global Tree Assessment (GTA), which enumerates nine principal threats to trees (agricultural expansion, over-exploitation, livestock farming, urban development, altered fire regimes, energy/mining, wood/pulp plantations, invasive/problematic species, and climate change). Prior studies indicate that temporal changes in threat exposure are often omitted from automated assessments, yet projections (e.g., for Chinese woody species) suggest large future increases in extinction risk when such changes are considered. There is recognized difficulty in diagnosing impacts of less direct threats like climate change, as effects may be mediated by dispersal limitations or changes to mutualists. The literature also highlights spatial and temporal variability, overlap, and potential synergies among threats, complicating consistent impact estimation across species, and underscores the need for systematic, data-driven approaches to aid expert-based IUCN Red List reassessments.

Study scope included 41,835 tree species (as defined by the GTA: woody plant with usually a single stem to ≥2 m, or multi-stemmed with ≥1 stem ≥5 cm DBH) drawn from Global TreeSearch v1.6 (names standardized via TNRS). Occurrence data (8,408,454 records) were compiled from GBIF, BIEN v3, DRYFLOR, RAINBIO, and Atlas of Living Australia, using quality labels AAA, AA, A, B, C from TREECHANGE (excluding records with combined geographic and environmental issues, unknown range, duplicates, or missing coordinates). Species’ extent of occurrence (EOO) was estimated as a minimum convex polygon (MCP) encompassing 95% of occurrences (adehabitatHR::mcp), projected to WGS84 at ~1 km resolution (0.01°). To better reflect potential presence, each MCP was refined by masking out waterbodies (MODIS MOD44W, resampled to 1 km) and climatically unsuitable areas using Köppen-Geiger climate zones (CHELSA 2.1, 1 km), retaining climate zones that constituted ≥5% of grid cells within a 1-km buffer around occurrences. The resulting refined polygons are termed species’ extents. Area calculations accounted for fractional cell coverage (exactextractr::exact_extract). Species with minimal occupied area (area of occupancy <10 km² on a 2×2 km grid or <5 unique occurrences) were flagged for prioritization owing to rarity or insufficient data (n=9,741). Threat layers and time windows: Four land-use threats (assumed unidirectional) and two bidirectional threats were analyzed for ~2000–2020: - Cropland expansion (Potapov et al.; GEE users/potapovpeter/Global_cropland, 30 m, 2003–2019): constructed post-2003 conversion mask; RRC computed as changed area / extent / 16 years × 100. - Tree cover decline (NASA MEASURES GFCC TC v3, 30 m, 2000–2015): differences in TC across 2000–2005, 2005–2010, 2010–2015; combined via maximum decline; RRC as changed area / extent / 15 × 100. - Built-up area expansion (GLAD GLCLU2020 Builtup_type, 30 m, 2000–2020): RRC as expansion area / extent / 20 × 100. - Deforestation (Hansen GFC 2021 v1.9, 30 m, 2000–2020): cells with forest loss from >50% tree cover to 0; RRC as loss area / extent / 20 × 100. - Burned area (ESA CCI FireCCI 5.1, 250 m, 2001–2020): annual burned area within extent; median-based linear regression (mblm::mblm, Siegel repeated medians) of burned area vs. year; if slope significant, RRC = slope / extent × 100; else RRC=0. - Climate change (CHELSA-BIOCLIM+, 4.6 km, 2000–2019): annual minimum and maximum temperature, mean annual vapor pressure deficit (VPD) and VPD seasonality (sd), mean annual precipitation and precipitation seasonality (sd). Annual extent-weighted means were computed (exactextractr) and analyzed with median-based linear models; significant slopes represent RRC in native units (°C yr⁻¹, Pa yr⁻¹, mm yr⁻¹); non-significant slopes set to 0. Threat maps were prepared in Google Earth Engine and resampled to 1 km via nearest neighbor. For land-use (unidirectional) threats, reversals were not counted to reflect lasting habitat alteration. Species prioritization used the 95th percentile threshold per threat’s RRC to flag highly exposed species (up to ~1605 species per threat), and species with minimal occupied area were also prioritized. Congruence with IUCN Red List categories (downloaded Jan 12, 2023 via taxize::iucn_summary) was evaluated by comparing RRC distributions across Threatened (EN, CR, and higher), Vulnerable, Not Threatened (NT, LC), and Data Deficient groups. Sensitivity of climate RRC to window length was assessed for 2000–2010 vs. 2010–2020 vs. 2000–2020.

- Coverage: Of 41,835 tree species, RRCs were computed for 32,090 species; 54.2% (17,393) were exposed to major and increasing threats. Only 8.7% of these candidate species are currently listed as threatened by the IUCN; 57.8% of Data Deficient species and 58.4% of Not Evaluated species fell among priority candidates. - Land-use threats: Tree cover decline had a median RRC of 1.86% of extent per year (95th quantile 3.55%); deforestation median 0.41% yr⁻¹ (95th quantile 1.85%). Maximum annual tree cover decline was 6.67% of extent (e.g., Rhodolaena macrocarpa, Arytera miniata, Syzygium sambogenese, Gluema korupensis, Weberbauerocereus madidiensis). Maximum annual deforestation was 5.0% (Eucalyptus redimiculifera). Cropland expansion median increased 0.95% of extent over 16 years; maximum annual expansion 2.51% (Premna richardsiae). Built-up area median expansion 0.80% over 20 years; maximum annual 1.48% (Leucadendron strobilinum). Burned area median annual change was 0; maxima: +1.74% (Eucalyptus ceracea) and −2.88% (Pyrostria lobulata) per year. - Climate change exposure (2000–2019/2020): 67.6% of species experienced increases in minimum temperature (max +0.31 °C yr⁻¹, Pinus peuce); 59.8% increases in maximum temperature (max +0.24 °C yr⁻¹, Trichilia bullata); 74.6% increases in VPD (max +15.5 Pa yr⁻¹, Caryodaphnopsis cogolloi); 39.7% increases in VPD seasonality (max +13.1 Pa yr⁻¹, Macrolobium urupaense); 23.0% increases in precipitation seasonality (max +20.2 mm yr⁻¹, Hypericum bequaertii). Precipitation trends: 22.2% increased (max +43.3 mm yr⁻¹, Magnolia lenticellata) and 28.5% decreased (max −37.9 mm yr⁻¹, Guatteria argentea). Climate change rates intensified in 2010–2020 relative to 2000–2010. - Correlations among threats: Positive association between cropland expansion and urban expansion (r=0.22); cropland expansion associated with decreased burned area (r=−0.29). Tree cover decline and deforestation RRCs were only 37% correlated, indicating distinct dynamics. - Distribution across IUCN categories: All status groups occurred across full exposure gradients. Among 2,623 species with >50% of extent degraded by tree cover decline/deforestation in the last two decades, 32.7% were Vulnerable/Threatened, 32.4% Not Threatened, and 34.8% Data Deficient/Not Evaluated. Of 11,645 species prioritized due to climate change exposure, only 9.0% were listed as Threatened; Data Deficient species represented 2.0% of that set but 43.7% of all DD in the study. - Prioritization outcomes: 17,393 unique species exceeded 95th percentile RRC for ≥1 threat. Among these, 8,119 are currently NT/LC (49.1% of Not Threatened group), 1,649 are Vulnerable (60.8% of VU), 1,521 are Threatened (63.9% of EN or worse), 312 Data Deficient (57.8% of DD), and 5,792 Not Evaluated (58.4% of NE). Additionally, 9,741 species had minimal occupied areas and, on average, co-occurred in locations with higher deforestation, tree cover decline, built-up expansion, and VPD change than non-rare locations. - Hotspots: High densities of highly exposed species occur mainly south of Tropic of Cancer, notably in South American and African (sub)tropical moist broadleaf forests and (sub)tropical regions of China, Tanzania, and Malaysia. Relative exposure hotspots (fraction of local flora exposed) include (sub)tropical China, Arctic Archipelago of Canada, and northern Russia. Threat-specific hotspots: cropland (Cerrado, Atlantic Forest), climate change (Amazon), tree cover decline (north coastal Central Africa), deforestation (Indo-Malaysia), built-up expansion (Central and South China), burned area change (Sub-Saharan Africa, Australia). 22.3% of priority species were highly exposed to >1 threat; three species were exposed to five of six threats. Regional overlay indicated areas with five or six threats within/near extents (Thailand, Cambodia, Vietnam, Borneo, West African coast, parts of Bolivia and Brazil).

The analyses reveal widespread and accelerating exposure of trees to multiple anthropogenic stressors, indicating that extinction risks are likely underestimated by current IUCN listings. Extensive tree cover decline and deforestation have degraded large portions of many species’ extents, yet only a minority are listed as Threatened, pointing to mismatches between observed threat dynamics and conservation status. Climate change pressures—rising temperatures, increasing atmospheric dryness (VPD), and shifting precipitation seasonality—are pervasive and intensifying, but their impacts are seldom reflected in Red List rationales, despite mounting evidence of forest dieback and declining resilience. Spatial hotspots of exposure concentrate in biodiverse (sub)tropical regions, with threat synergies (e.g., agriculture reducing burned area; warming and drought amplifying mortality where cover declines) likely to compound risks. Prioritizing species for reassessment using continuous rates of recent change offers a transparent, scalable means to target expert efforts, inform conservation planning, and potentially integrate dynamic threat metrics into future assessment decision trees.

Approximately 17,400 tree species are experiencing rapidly increasing exposure to global change stressors, many not currently recognized as at risk by the IUCN Red List. By quantifying recent change in key threats within refined species’ extents, the study provides a data-driven prioritization to expedite expert reassessments, identify geographic hotspots, and guide conservation action. The approach is transparent, flexible, and extendable to other taxa, serving as an early-warning tool—particularly for less conspicuous pressures like climate change. Future work should integrate dynamic threat metrics with species’ sensitivities and adaptive capacities, refine attribution among overlapping threats, improve coverage for underrepresented threats (e.g., invasives, plantations, mining/livestock), and investigate discrepancies where threatened species show lower recent change or NT/LC species show high exposure.

- Threat coverage gaps: Four GTA threats (livestock farming, energy production/mining, wood/pulp plantations, invasive/problematic species) lacked global, temporally resolved data and were excluded; deforestation served as a proxy but may not fully capture these drivers. - Overlap/double counting: Deforestation and cropland/urban expansion can represent overlapping conversions; counting both may inflate perceived cumulative impact, while cropland can also expand in non-forested areas. - Unidirectional assumption: Land-use changes were treated as irreversible over the window, potentially overlooking restoration or abandonment. - Range estimation and habitat suitability: MCP-based extents (95% occurrences) and climate-zone masking may include/exclude areas imperfectly; occurrence biases, outliers, or non-native occurrences could affect extents. Waterbody masking may underestimate extents of mangrove/marsh trees. - Data resolution and timing: Mixed spatial resolutions (30 m to ~4.6 km) and differing temporal spans (e.g., GFCC to 2015; cropland from 2003) can affect comparability; climate RRCs are in native units and assessed via significance testing, setting non-significant slopes to zero. - Sensitivity vs. exposure: RRCs quantify exposure, not species-specific sensitivity, adaptability, or demographic responses; some Threatened species may face high risk at lower recent change, and NT/LC species may still be resilient despite high exposure. - Small-range species: For species with minimal occupied area, rates were not directly computed; exposure inferred from co-occurring species may misrepresent true risk for these taxa.