Incorporating evolutionary and threat processes into crop wild relatives conservation

The study addresses the challenge of conserving the genetic diversity of crop wild relatives, which is vital for breeding and food security under global change. Existing systematic conservation planning often emphasizes species representation over persistence and rarely accounts for range-wide intraspecific genetic diversity. Centers of domestication like Mexico harbor complex patterns of genetic differentiation driven by climatic, geologic, and human histories, and CWR frequently occur within agricultural mosaics with varying interactions with crops, including gene flow. The research proposes a framework to incorporate indicators of intraspecific genetic variation—proxies of genetic differentiation (PGD)—into spatial conservation planning, while also considering threat processes such as tolerance to human-modified habitats and extinction risk. The aim is to identify conservation areas that maximize representation of genetic diversity across CWR in Mexico and to support national and regional conservation strategies.

Recent CWR conservation planning has progressed in identifying priority sites for in situ and ex situ efforts, but most applications solve minimum set cover problems emphasizing taxon presence rather than genetic diversity. Few studies integrate genetic variation into conservation planning, and ecogeographic approaches that use environmental heterogeneity as proxies for genetic diversity may miss historical drivers of population structure (e.g., Pleistocene range shifts, isolation by topography) that generate distinct genetic lineages across similar environments. Isolation by distance alone is insufficient to capture broad-scale genetic diversity, underscoring the need to incorporate phylogeographic insights and historical isolation patterns. The authors build on systematic planning frameworks and global/national CWR analyses, addressing gaps by combining environmental life zones with literature-derived phylogeographic patterns to delineate spatial PGD applicable where genomic data are lacking.

Study scope: Focus on Mexico within the broader Mesoamerican region to cover full ranges of many assessed taxa, including Aridamerica and the Nearctic realm. Dataset: Inventory of 224 native Mesoamerican CWR taxa (210 species; 14 infraspecific taxa) related to nine crops (Capsicum, Cucurbita, Gossypium, Persea, Phaseolus, Physalis, Solanum sect. Petota, Zea/Tripsacum, Vanilla). IUCN Red List categories assessed for extinction risk. Species distribution models (SDM): Compiled >13,000 curated occurrences from multiple databases (GBIF, SNIB, expert datasets). Built SDMs for taxa with >20 unique 1 km² records using MaxEnt 3.3.1, with 19 bioclimatic and additional variables (PET, aridity, radiation, slope, altitude; bare soil/cultivated percent). Addressed collinearity (threshold 0.8; VIF). Model tuning with ENMeval: feature classes and regularization multipliers (0.5–4.0) across random k-folds (k=4). Selection via AICc, plus validation using withheld 30% AUC and 10th percentile omission rates. Binary thresholds using minimum training presence or 10th percentile, with expert review to trim overpredicted areas using Mexican ecoregions and watersheds. Retained SDMs with AUC>0.7. Final SDMs for 116 taxa at 1 km² resolution; taxa without SDM included via occurrence points as species of special interest (SSI) in Zonation. Proxies of genetic differentiation (PGD): Delimited 27 Holdridge life zones as environmental strata. Subdivided each life zone using general phylogeographic patterns derived from literature across plants, animals, fungi (excluding CWR to avoid circularity), translating to spatial units via biogeographic provinces, basins, topography, and edaphology. Targeted broad, congruent trends (e.g., differentiation among major mountain ranges; east/west of Isthmus of Tehuantepec). Result: 102 non-overlapping PGD for Mexico. Validation with empirical genetics: Tested PGD using Zea mays subsp. parviglumis SNP dataset (~30,000 SNPs; K=13 clusters). Compared representation of genetic clusters using standard SDM vs SDM subdivided by PGD (SDMPGD) in Zonation considering 20% of Mexico: PGD approach increased representation across all genetic groups relative to SDM-only, despite imperfect one-to-one matches. Habitat preference integration: Experts assigned per-taxon habitat preferences (weights 1.0, 0.5, 0.1) across seven land cover classes (primary vegetation; secondary vegetation; less intensive rainfed/moisture agriculture; intensive rainfed/moisture agriculture; irrigated agriculture; induced/cultivated grasslands/forests; urban). Land cover from national maps; agricultural intensity refined using municipal maize yield thresholds (≤3 t ha−1) to distinguish low vs high intensity. Generated 116 taxon-specific habitat maps to guide selection via Zonation’s landscape condition. Systematic conservation planning (Zonation v4): Core-area zonation (CAZ) removal rule. Compared five preliminary scenarios: (i) SDM (116 layers), (ii) SDM+LZ (116 SDM + 27 LZ), (iii) SDM+PGD (116 SDM + 102 PGD as features), (iv) SDMPGD (intersection, 5004 layers), (v) SDM and PGD as administrative units (ADMU). Evaluated performance at 20% area threshold (aligned with Aichi Target 11) via representation curves of taxa ranges and mean PGD-area representation within taxa ranges. Final configurations: Adopted SDMPGD scenario for core analyses. Inputs: 5004 SDMPGD layers for 116 taxa; SSI occurrences for 98 additional taxa; 116 habitat maps; weights by IUCN category for 116 taxa (CR=1, EN=1, VU=0.8, NT=0.5, DD=0.3, LC=0.2, NE=0.1). SSI taxa weighted 1. Three final scenarios: (a) all taxa; (b) taxa restricted to natural vegetation; (c) taxa tolerant of multiple land uses. Outputs: Hierarchical priority map and performance curves; a 20% conservation area proposal. Additional assessments: Overlap with federal protected areas and indigenous territories (not used as criteria), and land cover composition of selected areas.

• CWR richness and distribution: Using 116 SDMs, high CWR richness concentrated along the Trans-Mexican Volcanic Belt and montane regions of Oaxaca and Chiapas. Seven federal protected areas potentially host >20 taxa. - PGD development and validation: Defined 102 proxies of genetic differentiation nationwide. Validation with Zea mays subsp. parviglumis (13 genetic clusters) showed that subdividing SDMs by PGD improved representation of genetically differentiated groups in priority areas compared to SDM-only scenarios. - Scenario performance (20% of Mexico): The SDMPGD scenario maximized intraspecific diversity representation, with on average 41% of each taxon’s range and 76% of the area of each PGD-within-taxon represented. Threatened taxa were best represented due to weighting. Widely distributed taxa had ~25% of range included but still >50% average PGD-area representation within their ranges. Alternative scenarios were less efficient: SDM-only captured more taxon range on average (~48%) but less PGD representation (~54%); SDM and PGD as administrative units improved PGD representation (~66%) but underperformed overall compared to SDMPGD; SDM+PGD and SDM+LZ also underperformed relative to SDM*PGD. - Final 20% conservation proposal: Priority areas occur in temperate mountains of the Trans-Mexican Volcanic Belt; central Veracruz to Chiapas including Oaxaca, Tehuacán-Cuicatlán, and Chimalapas; northern Michoacán coastline; cloud and rainforests of southern Mexico (e.g., for Vanilla); and arid/semi-arid Sonora and Baja California (for Gossypium). Approximately half of the selected area overlaps regions where indigenous communities live; 11% overlaps federal protected areas. - Representation under final configuration: In 20% of the country, on average 50% of each taxon’s area within each PGD is represented. However, complete representation is unattainable due to habitat loss and fragmentation; for highly threatened taxa, >40% of potential range may lack suitable habitat. - Management implication: To cover at least 50% of all conservation features, ~80% of Mexico’s terrestrial surface would need sustainable management, emphasizing the necessity of conservation beyond protected areas.

The framework directly addresses the need to represent and safeguard intraspecific genetic diversity in conservation planning for CWR. By integrating PGD that reflect both environmental and historical drivers of population structure, the approach improves coverage of potential genetic lineages compared to traditional SDM-based prioritizations. Incorporating IUCN threat-based weights and taxon-specific habitat preferences ensures that vulnerable taxa and suitable habitats are emphasized, leading to more actionable and ecologically realistic priorities. The hierarchical prioritization supports both area-based conservation and sustainable management in agricultural landscapes, aligning conservation with rural livelihoods and acknowledging biocultural contexts, particularly in indigenous regions where CWR are managed and persist. The results guide National Strategic Action Plans, inform policy on GMO coexistence and agricultural subsidies, and identify gaps for ex situ sampling and monitoring. The method is broadly transferable to other taxa and regions, and PGD can be refined as genomic data accumulate, enhancing the capacity to conserve evolutionary resilience amid climate and land-use change.

This study introduces and applies a scalable, data-integrative conservation planning framework that explicitly represents intraspecific genetic diversity of CWR via proxies of genetic differentiation, while accounting for threat processes and habitat preferences. In Mexico, the approach identified priority areas that maximize genetic representation within a feasible fraction of land area and highlighted significant overlap with indigenous territories and existing protected areas. The methodology supports both in situ and ex situ strategies, informs cross-sectoral policy, and can be adapted for other taxa and countries. Future research should: (i) expand genomic datasets to validate and refine PGD delineations; (ii) incorporate dynamic scenarios of climate and land-use change; (iii) assess connectivity and effective population size requirements; and (iv) integrate socio-economic objectives and local community priorities to sustain coevolutionary processes in agroecosystems.

• Limited genomic data across taxa: PGD are surrogates and cannot perfectly capture true genetic structure; accuracy depends on coarse phylogeographic generalizations and available cartography. - Modeling and data constraints: SDMs depend on occurrence quality, environmental layers, and thresholding choices; some taxa lacked SDMs and were represented only by occurrences. - Habitat loss and land-use constraints: Inclusion of habitat preferences revealed that complete range and PGD representation is infeasible in current landscapes, especially for threatened taxa. - Computational complexity: Handling thousands of input layers required high-performance computing; results may be sensitive to parameter settings and weighting schemes. - Scope and criteria: Indigenous areas and protected areas were not used as prioritization criteria (only for overlap assessment); connectivity was not explicitly optimized though areas aggregated emergently. - Generalization of phylogeographic patterns: PGD capture broad trends, potentially missing fine-scale, taxon-specific differentiation; validation performed for one exemplar taxon (teosinte) suggests improvement but not perfect alignment.