Landform and lithospheric development contribute to the assembly of mountain floras in China

Mountains act as both museums and cradles of biodiversity, concentrating a large share of global species, especially in the tropics. In China, mountainous hotspots harbor the majority of plant genera and terrestrial mammal species. Yet, the mechanisms generating and maintaining this exceptional diversity remain debated. Existing hypotheses emphasize climate stability, habitat heterogeneity (geodiversity), and energy availability (e.g., temperature, PET), with patterns varying across latitudes and regions. Contemporary climate alone often fails to explain observed diversity disparities, pointing to significant roles for geology, lithology, and evolutionary history. The mountain geobiodiversity hypothesis (MGH) posits that biodiversity arises from uplift, evolving geodiversity, and Neogene–Pleistocene climate change. The present study focuses on how landforms tied to bedrock and lithospheric processes shape floristic assembly in Chinese mountains by comparing species richness, phylogenetic diversity/structure, and species age structure across five bedrock-defined landform types.

The paper reviews major drivers of mountain biodiversity: climate (temperature, precipitation, energy), habitat heterogeneity/geodiversity, and evolutionary history. It highlights that geodiversity influences plant distributions and that bedrock composition can regulate vegetation and diversification. A substantial literature documents edaphic specialization (e.g., limestone, ultramafic soils), with 5–10% of species showing dependence on specific bedrock; many floras exhibit adaptive radiations tied to substrates (e.g., Cape, Teesdale, New Caledonia). Despite this, empirical links between edaphic diversity and plant diversity remain underexplored relative to climatic factors. At continental scales, mountain ranges sharing similar bedrock (e.g., granitic/metamorphic systems in eastern Asia and Appalachians) show floristic similarities, suggesting developmental landform processes may constrain assembly. The MGH provides an integrative perspective combining uplift, geodiversity, and climate history, but a refined framework focusing on bedrock-mediated assembly processes (beta diversity) is needed.

• Study design and sampling: The authors compiled checklists for 140 Chinese mountain floras (protected areas such as nature reserves/forest parks), each defined as all angiosperm species occurring on a mountain or well-delimited area. After reconciling taxonomy with the Leipzig Catalogue of Vascular Plants (LCVP), the dataset included 17,576 species in 2,585 genera, 251 families, 56 orders.
• Landform classification: Based on bedrock and stratigraphic denudation, five landform types were defined: karst (limestone), granitic (igneous), karst–granitic (mixed/transition), Danxia (continental clastic red beds/glutenite), and desert (arid region mountains often on sandshale). The 140 floras were classified into 19 karst, 14 karst–granitic, 84 granitic, 13 desert, 10 Danxia.
• Environmental and geographic predictors: 18 predictors captured geography (longitude, latitude, area [log-transformed], median elevation, elevation range), tectonics (craton vs orogenic belt), landform type, and climate (11 CHELSA bioclim variables retained after removing highly collinear variables). Variables were standardized to 0–1 except area (log-transformed).
• Phylogeny: A dated megaphylogeny was built using GBOTB.extended.LCVP as backbone and the R package V.PhyloMaker2 (Scenario 3). Missing genera were grafted as sisters to closest relatives; missing species added within genera with branch lengths via BLADJ. The resulting tree was used for community phylogenetic metrics and species divergence ages.
• Diversity and phylogenetic structure metrics: Species richness (SR); Faith’s phylogenetic diversity (PD); a standardized phylogenetic diversity index (PDI = [PD_observed − PD_random]/sd(PD_random)) computed with PhyloMeasures under a uniform null model; net relatedness index (NRI = [MPD_obs − MPD_rand]/sd(MPD_rand)) and nearest taxon index (NTI = −[MNTD_obs − MNTD_rand]/sd(MNTD_rand)) to capture deep vs tip-level structure (clustering > 0; overdispersion < 0).
• Species age structure: Mean divergence time (MDT) per flora from the dated tree; additionally MDT_oldest (oldest 25% of species) and MDT_youngest (youngest 25%). Four alternative divergence-time datasets were used to assess robustness, yielding consistent patterns.
• Statistical analyses: Two modeling frameworks per response variable: (1) landform-only generalized linear model (GLM) to test landform effects; (2) full model GLM including landform, tectonics, geography, and climate with stepwise AIC selection and leave-one-out importance checks. Interaction terms between landform and significant climatic/geographic variables were tested. Spatial autocorrelation of residuals was evaluated via Moran’s I; spatial error models (SEM; R spdep) were fitted to account for spatial structure. Standardized coefficients were used to compare predictor importance.

• Species richness and PD by landform:
  • Richness medians: granitic = 1456; karst–granitic = 1458; karst = 1137; Danxia = 1132; desert = 721. Igneous bedrock (granitic, karst–granitic) supports higher richness than sedimentary bedrock (karst, Danxia, desert).
  • Landform-only model explained 28.95% of deviance in SR (GLM; AIC = 116.21) and 31.40% (SEM; AIC = 115). Full model explained 62.8% (GLM) and 63.7% (SEM). Strong interactions detected between landform and mean temperature of the coldest quarter (TCQ); weaker with annual precipitation (PREC) and precipitation of coldest quarter (PCQ). Longitude, elevation range, and higher mean TCQ generally increased SR; TWQ and precipitation seasonality (Pvar) were negatively correlated with SR overall, with TWQ positive only in granitic landforms.
• Phylogenetic diversity index (PDI):
  • Medians: karst (−1.03; highest), karst–granitic (−1.27), granitic (−1.77), Danxia (−3.23), desert (−16.59; lowest). Deepest divergences occur in karst, shallowest in desert.
  • Landform effects were strong: landform model explained 70.89% (GLM) and 77.15% (SEM); full model 88.09% (GLM) and 88.25% (SEM). In full models, orogenic setting, higher latitude, higher temperature annual range (TAR), and higher TCQ were negatively related to PDI; higher TWQ was positively related to PDI.
• Phylogenetic structure (NRI/NTI):
  • NRI median: desert = 9.45 (strong clustering), granitic = −1.47 (overdispersion), Danxia = −0.41, karst = −0.42. Granitic floras tended toward overdispersion; sedimentary bedrock floras toward clustering (desert strongly clustered).
  • Landform effects explained 64.91% (GLM) and 77.43% (SEM) of variance in NRI in the landform-only model; in the full SEM, landform’s unique contribution to NRI was 1.22% within 81.51% total explained variance. TAR was positively correlated with NRI (more unstable climates → stronger clustering). NTI patterns were consistent with stronger clustering in sedimentary landforms and overdispersion in granitic landforms.
• Species age structure:
  • MDT, MDT_oldest, MDT_youngest were oldest in karst, then karst–granitic, granitic, Danxia, and youngest in desert. For MDT_oldest, mean age karst–granitic = 24.66 Mya slightly exceeded karst = 24.42 Mya.
  • Landform significantly affected MDT metrics in both landform-only and full models, with greater explanatory power for MDT_oldest than for MDT or MDT_youngest. In full models, TAR and TCQ were key climatic predictors, negatively correlated with divergence times; TAR had the strongest standardized effect on MDT_youngest, indicating climate more strongly structures younger species.
• Spatial and regional patterns: Higher richness and overdispersion were concentrated in the eastern monsoon region; northern/desert floras had younger MDTs, likely due to Pleistocene glacial impacts and recent recolonization.
• Synthesis: Landforms modulate water availability and erosion regimes (e.g., limestone permeability and underground drainage reduce moisture availability), interacting with temperature to shape richness and assembly. The study proposes the ‘floristic geo-lithology hypothesis’ attributing beta-diversity differentiation and assembly to lithospheric cycles (bedrock-constrained landform development), environmental filtering, and restricted cross-landform dispersal.

The results demonstrate that bedrock-derived landform types exert strong, independent influences on mountain floristic assembly beyond climate and geography. Igneous-derived granitic systems support higher species richness and more phylogenetic overdispersion, consistent with higher habitat heterogeneity and less severe water loss compared to sedimentary systems. In contrast, sedimentary landforms (karst, Danxia, desert) promote phylogenetic clustering, reflecting strong environmental filtering and/or local radiations of edaphic specialists. The clear landform signal in the oldest species quartile (MDT_oldest) indicates that lithospheric and landform development histories structure ancient lineages, whereas recent climate variability (e.g., TAR, TCQ) more strongly influences assembly of younger species. Interactions between landform and temperature (TWQ, TCQ) highlight how bedrock-mediated hydrology modulates the positive effects of energy on richness, with high TWQ reducing richness in permeable sedimentary landforms due to accelerated moisture loss. Regionally, monsoon-influenced eastern mountains exhibit richer and more overdispersed floras, while arid northern deserts show young floras with strong clustering, consistent with Pleistocene dynamics and harsh filtering. Collectively, these findings link geological processes to ecological assembly, underpinning the proposed ‘floristic geo-lithology hypothesis’ as a beta-diversity framework complementing the MGH’s alpha-diversity focus.

This study quantifies how landform type, rooted in bedrock and lithospheric processes, shapes the species richness, phylogenetic diversity/structure, and age composition of Chinese mountain floras. Igneous (granitic and karst–granitic) landforms harbor richer, more phylogenetically overdispersed floras, whereas sedimentary (karst, Danxia, desert) landforms show lower richness and stronger clustering. Landform history primarily structures older lineages, while contemporary climate increasingly structures younger lineages. The authors introduce the ‘floristic geo-lithology hypothesis’ to explain montane beta diversity through the combined effects of bedrock-driven landform development, environmental filtering, local speciation, and restricted cross-landform dispersal. Future research should (i) test this hypothesis across other mountain systems globally, (ii) integrate finer-scale lithological and hydrological data, (iii) link trait evolution and genomic adaptation to bedrock and landform transitions, and (iv) refine phylogenies and divergence-time estimates to improve assembly inference.

• Phylogenetic reconstruction required grafting missing genera/species and using a backbone megatree; divergence times for species are approximate and may be overestimated, though sensitivity analyses with multiple datasets showed consistent patterns.
• Landform classification into five types is coarse and may mask within-landform heterogeneity; some mountains include mixed substrates.
• Species lists were compiled from published checklists of protected areas, which may introduce sampling biases and uneven completeness across regions.
• Spatial autocorrelation was present (addressed with SEM), but residual spatial structure may still influence inferences.
• Strong interactions between landform and climate complicate disentangling their independent effects; some predictors (e.g., TWQ) showed landform-dependent relationships.