Articles https://doi.org/10.1038/s41477-022-01144-8 State of ex situ conservation of landrace groups of 25 major crops Julian Ramirez-Villegas   1,2,3,25 ✉, Colin K. Khoury   1,4,25 ✉, Harold A. Achicanoy1,25, Maria Victoria Diaz1,25, Andres C. Mendez1,25, Chrystian C. Sosa   1,5,6,25, Zakaria Kehel7, Luigi Guarino8, Michael Abberton9, Jorrel Aunario10, Bashir Al Awar11, Juan Carlos Alarcon12, Ahmed Amri7, Noelle L. Anglin13,14, Vania Azevedo13,15, Khadija Aziz7, Grace Lee Capilit10, Oswaldo Chavez13, Dmytro Chebotarov10, Denise E. Costich   12, Daniel G. Debouck1, David Ellis13, Hamidou Falalou16, Albert Fiu17, Michel Edmond Ghanem   18, Peter Giovannini8, Alphonse J. Goungoulou19, Badara Gueye   9, Amal Ibn El Hobyb7, Ramni Jamnadass   20, Chris S. Jones   21, Bienvenu Kpeki   19, Jae-Sung Lee10, Kenneth L. McNally   10, Alice Muchugi21, Marie-Noelle Ndjiondjop19, Olaniyi Oyatomi   9, Thomas S. Payne12, Senthil Ramachandran15, Genoveva Rossel13, Nicolas Roux22, Max Ruas22, Carolina Sansaloni   12, Julie Sardos22, Tri Deri Setiyono10,23, Marimagne Tchamba9, Ines van den Houwe24, J. Alejandro Velazquez12, Ramaiah Venuprasad10, Peter Wenzl1, Mariana Yazbek11 and Cristian Zavala12 Crop landraces have unique local agroecological and societal functions and offer important genetic resources for plant breeding. Recognition of the value of landrace diversity and concern about its erosion on farms have led to sustained efforts to establish ex situ collections worldwide. The degree to which these efforts have succeeded in conserving landraces has not been compre- hensively assessed. Here we modelled the potential distributions of eco-geographically distinguishable groups of landraces of 25 cereal, pulse and starchy root/tuber/fruit crops within their geographic regions of diversity. We then analysed the extent to which these landrace groups are represented in genebank collections, using geographic and ecological coverage metrics as a proxy for genetic diversity. We find that ex situ conservation of landrace groups is currently moderately comprehensive on aver- age, with substantial variation among crops; a mean of 63% ± 12.6% of distributions is currently represented in genebanks. Breadfruit, bananas and plantains, lentils, common beans, chickpeas, barley and bread wheat landrace groups are among the most fully represented, whereas the largest conservation gaps persist for pearl millet, yams, finger millet, groundnut, potatoes and peas. Geographic regions prioritized for further collection of landrace groups for ex situ conservation include South Asia, the Mediterranean and West Asia, Mesoamerica, sub-Saharan Africa, the Andean mountains of South America and Central to East Asia. With further progress to fill these gaps, a high degree of representation of landrace group diversity in genebanks is feasible globally, thus fulfilling international targets for their ex situ conservation. Crop landraces, also known as farmers’ traditional, heritage, their unique agroecological and societal functions and services1,2. folk or heirloom varieties, are cultivated plant populations These typically genetically heterogeneous populations are com-developed and managed by Indigenous or traditional agrar- monly planted in a mosaic of different crop species and varieties, ian cultures through cultivation, selection and diffusion1. Having in combinations sustaining local agricultural resilience and adap- recognizable characteristics and geographic origins, landraces con- tive capacity, human nutrition and cultural needs1,2. Farmer-based tinue to be cultivated by these communities in many regions for exchange3 and gene flow among landrace populations, occasionally 1International Center for Tropical Agriculture (CIAT), Cali, Colombia. 2CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS), Cali, Colombia. 3Wageningen University & Research (WUR), Plant Production Systems Group, Wageningen, The Netherlands. 4San Diego Botanic Garden, Encinitas, CA, USA. 5Pontificia Universidad Javeriana Cali, Cali, Colombia. 6Universidad del Quindío, Armenia, Colombia. 7International Center for Agricultural Research in the Dry Areas (ICARDA), Rabat, Morocco. 8Global Crop Diversity Trust, Bonn, Germany. 9International Institute of Tropical Agriculture (IITA), Ibadan, Nigeria. 10International Rice Research Institute (IRRI), Los Baños, Philippines. 11International Center for Agricultural Research in the Dry Areas (ICARDA), Beirut, Lebanon. 12International Maize and Wheat Improvement Center (CIMMYT), Texcoco, México. 13International Potato Center (CIP), Lima, Peru. 14United States Department of Agriculture (USDA), Agricultural Research Service, Aberdeen, ID, USA. 15International Crops Research Institute for the Semi-arid Tropics (ICRISAT), Hyderabad, India. 16International Crops Research Institute for the Semi-arid Tropics (ICRISAT), Niamey, Niger. 17Centre for Pacific Crops and Trees (CePaCT), Narere, Fiji. 18Mohammed VI Polytechnic University (UM6P), Benguerir, Morocco. 19Africa Rice Center (AfricaRice), Bouaké, Côte d’Ivoire. 20World Agroforestry (ICRAF), Nairobi, Kenya. 21International Livestock Research Institute (ILRI), Addis Ababa, Ethiopia. 22Bioversity International, Montpellier, France. 23Louisiana State University, Baton Rouge, LA, USA. 24Bioversity International, Leuven, Belgium. 25These authors contributed equally: Julian Ramirez-Villegas, Colin K. Khoury, Harold Achicanoy, Maria Victoria Diaz, Andres Mendez, Chrystian C. Sosa. ✉e-mail: j.r.villegas@cgiar.org; c.khoury@cgiar.org NATuRE PLANTS | www.nature.com/natureplants Articles Nature PlaNts Number of crops 1 2 3 4 5 6 7 8 9 10 11 12 Fig. 1 | Richness map of the predicted distributions of landrace groups of 25 cereal, pulse and starchy root/tuber/fruit crops within their geographic regions of diversity. Darker colours indicate greater numbers of crop landrace groups potentially overlapping in the same 2.5-arc-minute cells, quantified in terms of number of crops. See Extended Data Fig. 1 for richness across all 71 landrace groups within the 25 crops. also involving modern cultivars4 or wild progenitors5, encourage Results the development of new variation, while longstanding cultiva- Geographic distributions of crop landrace groups. On the basis tion and selection lead to adaptation to local environmental and of correlations among 93,269 landrace occurrences of 25 crops societal conditions6. (61.9% of occurrences having pre-assigned landrace group assign- Landrace diversity is an essential genetic resource for modern ments and the rest inferred) and 50 environmental and socioeco- crop breeding7 and is key to understanding agricultural origins nomic predictor variables, landraces as a whole were predicted to be and domestication processes8. Landraces are typically accessed distributed on all inhabited continents, including throughout most via ex situ repositories, called genebanks, for these research pur- of the world’s tropical and subtropical lands (Fig. 1 and Extended poses. Efforts to collect landraces for genebank conservation Data Fig. 1). Regions with particularly high levels of richness have often prioritized sampling from geographic regions and across crops were projected in East and Southern Africa, South and cultures wherein crops were domesticated and/or have been cul- Central Asia, the Mediterranean and West Asia, West Africa and the tivated for a very long time, in recognition of the extraordinary Andean mountains of South America and Mesoamerica, with land- genetic variation in landraces found in these environments1,7,9. races of up to 12 of the 25 crops potentially cultivated within single These activities have gained urgency since the 1960s as economic, 2.5-arc-minute grid cells in Bangladesh, Ethiopia, India, Nepal and agricultural, demographic, environmental and climatic changes Pakistan. These geographic concentrations of landrace group diver- increasingly impact in situ populations1,7. The result of these col- sity align well with the historically recognized centres of origin and lection efforts has been the assemblage of approximately three primary regions of diversity of the world’s major crops14,15. Notably million landrace samples in international, regional, national and less landrace diversity across crops was predicted to be cultivated in subnational genebanks10. most temperate regions, in some very arid zones such as the Saharan Despite these extensive efforts, landrace diversity is not com- Desert and in a few highly mesic areas such as the Amazon Basin. monly considered to be comprehensively represented ex situ, The predicted distributions of the five major races of sorghum are and major international agreements, including the Convention provided in Fig. 2a as an example of landrace-group-level results (the on Biological Diversity (CBD) Aichi Target 13 (ref. 11) and the Supplementary Information presents the occurrences and predicted Sustainable Development Goals (SDGs) Target 2.5 (ref. 12), urgently distributions of landrace groups for all assessed crops). Sorghum prioritize the resolution of this conservation gap. To reach these tar- landrace group ranges were modelled throughout the crop’s main gets, information about the current distributions of landraces and regions of diversity in Africa, South Asia, the Mediterranean and their degree of representation in genebanks is needed. To respond West Asia. Its races inhabit distinct eco-geographic ranges but also to this need, in this Article we employ a conservation gap analysis overlap in specific areas, particularly in Southern and West Africa methodology13 to predict the distributions and quantify the current and in South Asia. Regarding different types of assessed crops, cereal ex situ conservation status of 71 eco-geographically distinguishable and pulse landrace group diversity was predicted to be particularly groups of landraces within 25 cereal, pulse and starchy root/tuber/ rich in South and Central Asia; West, East and Southern Africa; fruit crops whose genetic resources are researched and conserved by the Mediterranean and West Asia; Europe; the Andean mountains; CGIAR international agricultural research centres or by the Centre and Mesoamerica. Meanwhile, starchy root/tuber/fruit crop land- for Pacific Crops and Trees (CePaCT) of the Pacific Community race group richness was concentrated in Mesoamerica, Southeast (SPC). We identify gaps in existing ex situ collections to inform fur- Asia and the Pacific, South America, West Africa and South Asia ther collecting efforts. (Extended Data Figs. 2–4). NATuRE PLANTS | www.nature.com/natureplants Nature PlaNts Articles a b Number of landrace groups Number of landrace groups 1 2 3 4 5 1 2 3 4 5 Bicolor Caudatum Durra Bicolor Caudatum Durra Guinea Kafir Guinea Kafir Fig. 2 | Richness maps of sorghum landrace group distributions and ex situ conservation gaps. a,b, Predicted distributions (a) and ex situ conservation gaps (b) for five landrace groups of sorghum in Africa, South Asia, the Mediterranean, and West Asia—namely, the races bicolor, caudatum, durra, guinea and kafir. Small maps, individual distributions of each landrace group; large maps, richness at the crop level. Ex situ conservation status and gaps for crop landrace groups. numbers of genebank samples (an average of 1,052.7 accessions ver- On average, ex situ conservation of crop landrace groups—mea- sus 2,827.4 of pulses and 5,796.4 of cereals); these typically clonally sured in terms of the extent of current cultivated geographic range propagated crops often require higher ex situ conservation expendi- and ecological variation in the range that has previously been col- tures per sample and present more substantial challenges from pests lected from and is now conserved in genebanks—was estimated to and diseases16. Nonetheless, cereal, pulse and starchy root/tuber/ be moderately comprehensive at present, with substantial variation fruit crop types all included members with some of the least and among crops; an average of 63% ± 12.6% of distributions was repre- most comprehensive conservation scores (Fig. 3). Moreover, these sented ex situ (Fig. 3 and Supplementary Table 1). Measured as the scores were not correlated with importance of the crop to global mean of the estimated minimum and maximum extent of represen- food supplies, production and trade (r = 0.064) tation in genebanks, geographic and ecological variation in landrace At the landrace group level, considerable differences in current groups of the following crops was among the most comprehensively conservation status were identified among groups within many represented: breadfruit at 81.6% conserved, bananas and plantains crops (Supplementary Table 2). For example, geographic and eco- at 81.5%, lentils at 78.3%, common beans at 77.4%, chickpeas at logical variation in barley landraces with covered (with hull) grains 75.8%, barley at 75.5% and bread wheat at 71.3%. Conversely, the was estimated to be 89.1% conserved, while diversity in landraces largest conservation gaps persist for pearl millet at 32.7%, yams at with naked or hull-less grains was only 31.3% represented ex situ. 43.0%, finger millet at 45.4%, groundnut at 46.5%, potatoes at 50.3% Asian rice, finger millet, potato, sorghum and yam landrace groups and peas at 52.4%. The maximum potential representation metrics also varied rather widely regarding current conservation in gene- indicate that breadfruit, lentil, banana and plantain, grasspea and banks, while cassava, chickpea, common bean, cowpea, groundnut, chickpea landrace group variation may already be very well con- lentil, maize, pea, pearl millet, African rice, sweetpotato and bread served, since all have maximum current ex situ conservation scores wheat landrace groups had more similar within-crop ex situ repre- above 90%, while the minimum coverage metrics warn that some sentation estimates. crops may still face extensive conservation gaps, such as pearl millet Taking sorghum landrace groups as an example, high-confidence at 15.2%, groundnut at 22.6%, finger millet at 25.3%, peas at 28.1% gaps in current ex situ conservation in terms of geographic and and yams at 29.0%. ecological variation were identified for all five major races in Regarding types of assessed crops, the average degree of ex situ sub-Saharan Africa, with overlapping gaps concentrated in Central, conservation of cereal, pulse and starchy root/tuber/fruit landrace West and Southern Africa, including in Madagascar (Fig. 2b). The groups did not differ significantly (P = 0.69), measuring 59.9%, Supplementary Information provides conservation gap maps for all 64.6% and 64.9%, respectively. At 45.0%, 45.6% and 50.4%, their assessed crops. mean minimum potential representation values were also similar, Across the landrace groups of all 25 crops, geographic areas iden- as were their maximum potential representation values of 74.8%, tified as hotspots requiring further collecting for ex situ conserva- 83.6% and 79.3%. For the final crop-type category, this finding is tion were concentrated in South Asia; the Mediterranean and West remarkable because these plants are represented by lower overall Asia; Mesoamerica; West, East and Southern Africa; the Andean NATuRE PLANTS | www.nature.com/natureplants Articles Nature PlaNts 1 100.0 Wheat (durum) Wheat (bread) climatic and political change . Recent decades of progress in clari- Rice (Asia) fying and, in some cases, expediting the terms and conditions of Rice (Africa—glaberrima) genetic resources sampling and exchange21,22 bolster anticipation Maize that such gaps can be filled through international collaboration. Potato With further concerted efforts to collect crop landraces of these Groundnut and other crops, a high degree of representation of their diversity 10.0 Cassava Barley Banana and plantain in genebanks appears to be feasible, and, thus, the fulfilment of Sorghum Common bean the international targets of the CBD11 and SDGs12 regarding their Yam Sweet potato ex situ conservation also seems achievable. Conducted periodically Pearl millet Finger millet over time, the gap analysis offers a more holistic approach to assess Pea Chickpea the state of landrace conservation than simply reporting changes in 1.0 Cowpea counts of accessions held in genebanks23 and, thus, may also repre- Faba bean Lentil Taro sent a useful complement to the current indicators for these targets. Pigeonpea The landrace group classification and modelling processes described here demonstrate the potential to associate genetic, mor- phological, physiological, chemical, nomenclatural and other char- 0.1 Grasspea acteristics of cultivated plant populations with environmental and socioeconomic predictors within the regions of origin and diversity of crop taxa. These processes can be performed across a spectrum of infraspecific groups and geographic scales, depending on avail- Breadfruit able knowledge and occurrence and characterization information. While our processes are based on openly available data and tools 0 25 50 75 100 that undergo continual updating, they involve several limitations. Degree of representation in ex situ conservation (%) First, our methods are vulnerable to deficiencies in the quality, completeness and availability of occurrence and infraspecific group- Fig. 3 | The current representation of crop landrace groups in ex situ ing information. Many cultivated plants are insufficiently sampled, conservation. Conservation metrics provide a scale from the lower to potentially due to a historical emphasis on wild rather than farming the upper estimates of current ex situ conservation status per crop with landscapes within biodiversity initiatives and persisting disconnects the averages denoted by circles. The crop importance metric indicates between biodiversity conservation and agricultural research com- the current significance of the crop, averaged across global food supply, munities24. Robust landrace classifications based on genetic struc- production and trade metrics (Supplementary Information). Gold, cereals; ture, geography and other attributes also require further resolution green, pulses; purple, starchy roots/tubers/fruits. for many crops. The major biodiversity and conservation repository databases that we utilize here do not yet represent all pertinent national and mountains; and Central and East Asia (Fig. 4 and Extended Data subnational institutions worldwide; those institutions that do par- Fig. 5; online results at https://ciat.shinyapps.io/LGA_dashboard/). ticipate may not report all holdings and locality and characteriza- Currently, uncollected landrace groups of up to nine crops are poten- tion information is incomplete for many existing records13,20. Some tially cultivated within single 2.5-arc-minute grid cells in India and additional information is probably present in other, smaller online Morocco and of up to eight crops in Algeria, Greece, Iran, Mexico, databases or in offline or undigitized datasets. These gaps increase Pakistan, Sierra Leone and Turkey. Regarding types of assessed the uncertainty in our results, possibly leading to underestima- crops, cereal and pulse landrace group diversity was predicted to tions of the true degree of ex situ landrace conservation. On the be particularly in need of further collecting in the Mediterranean other hand, the accessibility and long-term security of many such and West Asia; South Asia; West, East and Southern Africa; Europe; low-visibility collections are often equally uncertain13,19. Several the Andean mountains; and Mesoamerica. Conversely, starchy processes would strengthen the conservation and potential useful- root/tuber/fruit crop landrace group ex situ conservation gaps were ness of these genetic resources and the accuracy of analyses such as concentrated in East and Southeast Asia, South Asia, West Africa, ours: the generation of characterization information and knowledge South America and Mesoamerica (Extended Data Figs. 6–8). about infraspecific groups, improvements in the quality and com- pleteness of existing occurrence information and better availabil- Discussion ity of landrace samples and their associated data, including safety Our analysis of the ex situ conservation status of landrace groups duplication to better ensure long-term persistence. within 25 staple crops suggests that their representation in gene- Second, because our modelling method is based on statistical banks is most often substantial, a finding that highlights the impact relationships between occurrences and environmental and socio- of extensive international, national and subnational efforts world- economic predictor variables, it is also sensitive to the quality and wide over more than a half-century, both individually and via col- comprehensiveness of these predictor datasets. Factors lacking pre- laborative networks and initiatives1,7,17,18. Conservation of landraces dictor information or acting at finer scales than currently available of these crops—or at least their eco-geographically distinguishable data reflect will not be well incorporated into modelling processes. groups—appears to be considerably further advanced than equiva- These may include environmental factors—both abiotic, such as lent protection for crop wild relatives (Extended Data Fig. 9)19,20. soil characteristics or supplemental irrigation in small plots, and However, the findings also reveal that ex situ conservation gaps biotic, such as pathogen pressures or pollinator distributions—and in terms of uncollected geographic and environmental variation socioeconomic drivers such as farm sizes, agronomic practices and across the distributions of landrace groups of these crops persist. seed system dynamics. Further, the models are unlikely to account Our quantitative and spatial results can aid in priority setting across for relatively recent disappearances of landraces unless such losses these crops, their landrace groups and geographic regions, con- are associated with available predictors. The increasing generation tributing to conservation targeting, planning and action. Further of land-use-change information25 may partially resolve this chal- prioritization may be applied based on known or perceived threats lenge. In all cases, further development of high-resolution pre- related to economic, agricultural, technological, demographic, dictor datasets with global scope will improve modelling. Deeper NATuRE PLANTS | www.nature.com/natureplants Crop importance (log10) Nature PlaNts Articles a Number of crops 1 2 3 4 5 6 7 8 9 b c d Fig. 4 | Geographic hotspots for further collection for the ex situ conservation of crop landrace groups. a, Global map of ‘gap richness’ across the predicted distributions of landrace groups of 25 cereal, pulse and starchy root/tuber/fruit crops within their geographic regions of diversity, indicating where landraces are expected to occur and have not yet been collected and conserved in genebanks. Darker colours indicate greater numbers of uncollected crop landrace groups potentially overlapping in the same 2.5-arc-minute cells, quantified in terms of numbers of crops. b–d, Examples of regions with particularly high gap richness in South Asia (b), the Mediterranean and West Asia (c) and Mesoamerica (d). See Extended Data Fig. 5 for gap richness across the 71 landrace groups within the 25 crops. understanding of the wide range of factors affecting farmer choices among their populations, which may be influenced by reproduc- regarding landrace cultivation, including apparent stochasticity26, tive biology, such as by outcrossing versus inbreeding species and may also be important to improved modelling, while the limits to by the mode of pollination; by mode of propagation, such as by predicting distributions of populations whose ranges are driven seed versus clonally; and by other ecological and cultural factors3. by human preferences as much as environmental factors must be Moreover, our conservation gap analysis methodology is based on acknowledged. the assumption that the existence of an ex situ accession from a Third, while geographic and ecological variation within pre- site indicates that the targeted landrace group has been adequately dicted native ranges of plants has been shown to be an effective sampled there. In reality, landrace distinctions at finer resolution surrogate for direct measures of genetic diversity27,28, the modelling than their modelled groupings may be ignored and, thus, not fully and conservation metrics used here may not fully reflect the dis- conserved. Previous field collecting may also not have comprehen- tributions of and gaps in genetic variation within crop landraces. sively sampled populations at the resolution needed for all conser- Further, our standardized method may not take into account the vation, plant breeding or other research aims. This drawback may differences between crop species in genetic diversity within and be particularly applicable to landraces that are typically genetically NATuRE PLANTS | www.nature.com/natureplants Articles Nature PlaNts heterogeneous and, thus, may require large sample sizes to repre- the International Transit Centre and Musa Germplasm Information System of sent their diversity and, in particular, rare alleles. Finally, because Bioversity International 35, CePaCT, International Center for Tropical Agriculture in situ crop diversity constantly changes, developing novel variation (CIAT), International Maize and Wheat Improvement Center (CIMMYT), International Potato Center (CIP), International Center for Agricultural Research from gene flow, recombination and mutation6, valuable new forms in the Dry Areas (ICARDA), International Crops Research Institute for the may have arisen in previously collected areas. Further sampling for Semi-arid Tropics (ICRISAT), International Institute of Tropical Agriculture (IITA) ex situ conservation may, therefore, be warranted within or near and International Rice Research Institute (IRRI), as well as from the United States previously collected sites. Department of Agriculture (USDA) Genetic Resources Information Network 36 The combination of these vulnerabilities reinforces the impor- (GRIN)–Global and the Comisión Nacional para el Conocimiento y Uso de la Biodiversidad (CONABIO)37. Occurrences were compiled from the Global tance of field reconnaissance and of partnering with Indigenous Biodiversity Information Facility (GBIF), with ‘living specimen’ records classified and traditional agrarian communities and associated organizations as ex situ conservation records and the remaining serving as reference sightings to inform further collecting activities. In this sense, our results are for use in distribution modelling. Reference occurrences were also drawn from best considered as support tools, useful for guiding rather than pre- published literature (Supplementary Information). Duplicated observations within scribing taxonomic and geographic priorities13. Additional essential or between data sources were eliminated, with a preference to utilize the most original data. Coordinates were corrected or removed when latitude and longitude steps include ensuring adherence to international, national and local were equal to zero or inverted, located in water bodies or in the wrong country or sampling and exchange policies21,22; assessing field work risks, par- had poor resolution (<2 decimal places). Occurrences were clipped to study areas ticularly in regions affected by war and civil strife29; and determin- per crop. The complete occurrence dataset is available in Supplementary Dataset 2. ing the most appropriate timing to maximize the harvest of viable seeds and other propagules. The capacity of pertinent genebanks to Spatial predictors. We compiled and calculated spatially explicit gridded information for 50 potential environmental and cultural predictors of landrace receive, adequately maintain and distribute landrace diversity must distributions, including climatic, topographic, evolutionary history and be preconfirmed1,7, and the logistics of getting collected material socioeconomic variables (Supplementary Table 3)13. For climate data, we gathered into relevant genebanks in a timely fashion must be established. or derived 39 variables from WorldClim version 2 (ref. 38) and Environmental Further development and mobilization of landrace modelling Rasters for Ecological Modeling (ENVIREM)39. We included elevation from the and conservation gap analysis would ideally assess a wider range Shuttle Radar Topography Mission (SRTM) dataset of the CGIAR–Consortium on Geospatial Information portal40,41. Two crop evolutionary history proxies of crops, including fruits and vegetables, nuts and other groups were included: distance to human settlements before the year ad 1500 (ref. 42) of importance to human nutrition and agricultural livelihoods30. and distance to known wild progenitor populations13. The eight socioeconomic It is probable that many other crops, especially those that have variables included population density43, distance to navigable rivers44, percentage 45 not received primary focus in international or national genetic of the area under irrigation , population accessibility 46,47, geographic distributions resources conservation and crop improvement efforts over the past of ethnic or cultural groups 48 and crop harvested area, production quantity and yield49. All predictor data were scaled to 2.5-arc-minute resolution with World half-century, are less well represented in ex situ conservation reposi- Geodetic System (WGS) 84 as a datum. Extended descriptions of the sources tories1,10; thankfully, erosion of the in situ genetic diversity of these and their justification for inclusion are provided in Ramirez-Villegas et al.13. The crops may be less severe thus far than in major staples1. complete spatial predictor dataset is available in Supplementary Dataset 3. Geographic expansion of the analyses beyond historical regions of diversity9,14,15 may also aid in identifying novel variation, although Crop landrace group classifications. Crop landraces are cultivated plant further assessment of the correlation between landrace groups and populations managed by Indigenous or traditional farmers through cultivation, selection and diffusion1. They are typically genetically heterogeneous, although spatial predictors in such regions will be necessary. To more fully some types, such as clonally propagated populations, may be relatively address the scope of international conservation targets for landra- homogeneous. They have recognizable characteristics, identities and geographic ces, these analyses must also assess the state of their in situ (on-farm) origins are in an ongoing process of adaptation to their local environments and 1,2,31 conservation; this task presents substantial challenges because societal conditions . For most crops, landraces number in the thousands, with emphasis in this context falls on the conditions and processes that major global staple cereals such as rice and wheat potentially represented by hundreds of thousands of landraces50,51, although precise numbers and consensus foster landrace diversity rather than on the persistence of particular regarding differentiations among landraces within crops have not been established. populations1,2,31. Given further development and expansion of the Given the diversity of landraces and the complexity of environmental and cultural methods and scope, and the combination of the results with parallel drivers differentiating them, our method seeks a compromise between, on the analyses of crop wild relatives19,20 and other socioeconomically and one hand, acknowledgement of this diversity and, on the other, the feasibility and culturally valuable plants32, a significantly improved understanding performance of distribution modelling and conservation gap analysis.For each crop, we, therefore, conducted an extensive literature review to of distributions, protection status and conservation gaps across the identify recognized infraspecific groups with distinct morphological, physiological, major forms of crop diversity prioritized by the CBD and the SDGs chemical, genetic, nomenclatural or other characteristics that could be tested should be achievable. for environmental and cultural associations (Supplementary Table 1 and Supplementary Information). The nature of these groups varied by crop and included genepools, races, genetic clusters and geographic or environmental Methods groupings. Crops often had more than one proposed grouping or classification. Crops and their landrace study areas. Food crops whose genetic resources are We then built and tested classification models to determine how well the researched and conserved by CGIAR international agricultural research centres or proposed groups could be predicted and distinguished based on spatial predictors, by the CePaCT of the SPC were included in this study. Crop landrace distributions drawing from the occurrence database and training datasets compiled from the were modelled and conservation analyses conducted within recognized primary literature review. A random forest52, a support vector machine53, the K-nearest and, for some crops, secondary regions of diversity, where these crops were neighbour (KNN) algorithm54 and artificial neural networks55 were used to domesticated and/or have been cultivated for very long periods, and where determine classification performance. The response variable was the group they are, thus, expected to feature high genetic diversity and adaptation to local to which a given occurrence was assigned, whereas the explanatory variables environmental and cultural factors (Supplementary Tables 1 and 2)9,13. These were the spatial predictors. Models were combined into an ensemble using the regions were identified through literature review (Supplementary Information) and mode—that is, the most frequent predicted value among the models—and tested confirmed by crop experts. using 15-fold cross-validation with 80% training and 20% testing. We accepted a given classification if each of its components was predicted with an average Occurrence data. Our crop landrace group distribution modelling and cross-validated accuracy of at least 80%. In the case of multiple classification conservation gap analysis rely on occurrence data, including coordinates of proposals per crop, we selected the one with the best overall performance. Finally, locations where landraces were previously collected for ex situ conservation we used the trained models to predict the corresponding group for occurrences and reference sightings. For ex situ conservation records, occurrences marked missing such information. All landrace groups for all crops are provided in as landraces were retrieved from two major online databases: the Genesys Plant Supplementary Table 2, with the best-performing groups identified. Genetic Resources portal33 and the World Information and Early Warning System on Plant Genetic Resources for Food and Agriculture (WIEWS) of the Food and Crop landrace group distribution modelling. To predict the probability of Agriculture Organization of the United Nations34. Occurrences were also obtained geographic occurrence for each landrace group within each crop, we generated directly from individual international genebank information systems: AfricaRice, MaxEnt models56,57 using the ‘maxnet’ R package58. Group-specific spatial NATuRE PLANTS | www.nature.com/natureplants Nature PlaNts Articles predictors were selected using a combination of the variance inflation factor estimate of current representation; the complement of the proportion considered (VIF) and a principal component analysis (PCA) to control for excessive by all three scores as a gap, which is to say high-confidence gap areas, represents model complexity and variable collinearity59. We removed variables that did the maximum estimate (Supplementary Tables 1 and 2). not contribute significantly to the variance in the PCA, defined as contributing While distribution modelling and conservation gap analyses were conducted at less than 15% to the first component, and we further discarded variables with a the crop landrace group level and results are presented in full in the Supplementary VIF > 10 (ref. 60). The predictors and whether they were selected for the modelling Information, for ease of comparison of results across crops, and to avoid bias of each landrace group are presented in Supplementary Table 4. towards crops with many landrace groups, we also calculated summary results at the We generated a random sample of pseudo-absences as background points crop level. Crops that had been assessed with geographic differentiations, including in areas that (1) were within the same ecological land units61 as the occurrence maize in Africa and Latin America and yams in the New World and the Old World, points, (2) were deemed potentially suitable according to a support vector machine were also combined. For spatial results, the pixels in crop landrace group models classifier that uses all occurrences and predictor variables and (3) were farther than were summed—that is, constituent landrace group models were combined. The 5 km from any occurrence62. The number of pseudo-absences generated per crop minimum and maximum current conservation representation estimations at the group was ten times its number of unique occurrences. crop level were then calculated based on combined spatial models. MaxEnt models were fitted through five-fold (K = 5) cross-validation with 80% training and 20% testing. For each fold, we calculated the area under the receiving GBIF occurrence downloads. The following occurrence downloads from the operating characteristic curve (AUC), sensitivity, specificity and Cohen’s kappa as Global Biodiversity Information Facility (GBIF; https://www.gbif.org/, 2017−2021) measures of model performance. To create a single prediction that represents the were used: 10.15468/dl.rrntfr, 10.15468/dl.2f2v4h, 10.15468/dl.2ywlb7, 10.15468/ probability of occurrence for the landrace group, we computed the median across dl.lnfelh, 10.15468/dl.ryrmfj, 10.15468/dl.8adf61, 10.15468/dl.nff5ys, 10.15468/ K models. Geographic areas in the form of pixels with probability values above dl.erxs6e, 10.15468/dl.vbfgho, 10.15468/dl.mjjk3x, 10.15468/dl.uppz1n, 10.15468/ the maximum sum of sensitivity and specificity were treated as the final area of dl.938bgm, 10.15468/dl.hr87hm, 10.15468/dl.k1va80, 10.15468/dl.coqpu2, predicted presence13. 10.15468/dl.lkoo9u, 10.15468/dl.e998mp, 10.15468/dl.vfbmm7, 10.15468/ dl.tnp478, 10.15468/dl.6zxsea, 10.15468/dl.0lray8, 10.15468/dl.5sjgsw, 10.15468/ dl.wkju6h, 10.15468/dl.7xzfvc, 10.15468/dl.autlf5, 10.15468/dl.fe2amw, Ex situ conservation status and gaps. Three separate but complementary metrics 10.15468/dl.2zblvz, 10.15468/dl.ddplkj, 10.15468/dl.jbzejg, 10.15468/dl.ej5bha, were developed to compare the geographic and environmental diversity in 10.15468/dl.905pxd, 10.15468/dl.pim1vs, 10.15468/dl.vdridc, 10.15468/ current ex situ conservation collections to the total geographic and environmental dl.b43gyv, 10.15468/dl.nnw3z7, 10.15468/dl.bnt9jc, 10.15468/dl.f5x2cg, 10.15468/ variation across the crop landrace group distribution model and, thus, to identify 13 dl.ub7zbg, 10.15468/dl.sggf2v, 10.15468/dl.ath5ve, 10.15468/dl.23k3ug, 10.15468/and quantify ex situ conservation gaps . dl.cym376, 10.15468/dl.53bwzk, 10.15468/dl.fsad7h and 10.15468/dl.fm6p7z. A connectivity gap score (SCON) was calculated for each 2.5-arc-minute pixel within the distribution model by drawing a triangle63,64 around each pixel using Reporting Summary. Further information on research design is available in the the three closest genebank accession occurrence locations as vertices and then Nature Research Reporting Summary linked to this article. deriving normalized values for the pixel based on distance to the triangle centroid and vertices13. The SCON of a pixel is high—closer to 1 on a scale of 0–1—when its corresponding triangle is large, when the pixel is close to the centroid of the Data availability triangle or when the distance to the vertices is large. A high S represents a Occurrence data, including spatial predictor variable results (at 2.5-arc-minute CON greater probability of the pixel location being a gap in existing ex situ collections. resolution) for each occurrence (Supplementary Dataset 2) and the global spatial An accessibility gap score (S ) was calculated for each 2.5-arc-minute pixel predictor dataset (2.5-arc-minute resolution, all 50 variables) (Supplementary ACC in the distribution model by computing travel time from each pixel to its nearest Dataset 3) are available at https://doi.org/10.7910/DVN/J8WAPH. Source data are genebank accession occurrence location based both on distance and the speed provided with this paper. of travel, defined by a friction surface13,45. Travel time scores were normalized by dividing pixel values by the longest travel time within the distribution model, with Code availability the final score ranging from 0 to 1. A high SACC value for a pixel reflects long travel Code for the crop landrace group classification testing, distribution modelling times from existing genebank collection occurrences and, thus, represents a higher and conservation gap analysis is available at https://github.com/CIAT-DAPA/ probability of the pixel location being a gap in existing ex situ collections. gap_analysis_landraces. An environmental gap score (SENV) was calculated for each 2.5-arc-minute pixel in the distribution model by conducting a hierarchical clustering analysis using Ward’s method with all the predictor variables from the distribution modelling. Received: 13 October 2021; Accepted: 28 March 2022; The Mahalanobis distance between each pixel and the environmentally closest Published: xx xx xxxx genebank accession occurrence location was then computed13. Environmental distance scores were normalized between 0 and 1. A high SENV value for a pixel References reflects a large distance to areas with similar environments where landraces have 1. Khoury, C. K. et al. Crop genetic erosion: understanding and responding to previously been collected for genebank conservation and, thus, represents a higher loss of crop diversity. New Phytol. 233, 84–118 (2021). probability of the pixel location being a gap in existing ex situ collections. 2. Jarvis, D. I. et al. A global perspective of the richness and evenness of Spatial ex situ conservation gaps were determined from the conservation gap traditional crop-variety diversity maintained by farming communities. scores using a cross-validation procedure to derive a threshold for each score. Proc. Natl Acad. Sci. USA 105, 5326–5331 (2008). We created synthetic gaps by removing existing genebank occurrences in five 3. Allinne, C. et al. Role of seed flow on the pattern and dynamics of pearl randomly chosen circular areas with a 100 km radius within the distribution millet (Pennisetum glaucum [L.] R. Br.) genetic diversity assessed by AFLP model. We then tested whether these artificial gaps could be predicted by our gap markers: a study in south-western Niger. Genetica 133, 167–178 (2007). analysis, identifying the threshold value of each score that would maximize the 4. Rojas-Barrera, I. C. et al. Contemporary evolution of maize landraces and prediction of these synthetic gaps. Performance for each of the five gap areas was their wild relatives influenced by gene flow with modern maize varieties. assessed using AUC, sensitivity and specificity. The average cross-area threshold Proc. Natl Acad. Sci. USA 116, 21302–21311 (2019). value was calculated for each score to discern pixels with a high likelihood of 5. Jarvis, D. I. & Hodgkin, T. Wild relatives and crop cultivars: detecting natural finding ex situ conservation gaps and that, thus, were higher priority for further introgression and farmer selection of new genetic combinations in field sampling. These were pixels with combined gap scores above the threshold, agroecosystems. Mol. Ecol. 8, S159–S173 (1999). assigned a value of 1, as opposed to the relatively well-conserved areas below the 6. Mercer, K. L. & Perales, H. R. Evolutionary response of landraces to climate threshold, which were assigned a value of 0. change in centers of crop diversity. Evol. Appl. 3, 480–493 (2010). The three binary conservation gap scores were then mapped in combination, 7. Gepts, P. Plant genetic resources conservation and utilization: the resulting in pixels across the distribution model with gap values ranging from 0 to accomplishments and future of a societal insurance policy. Crop Sci. 46, 3. Pixels with a value of 0 display no connectivity, accessibility or environmental 2278–2292 (2006). gaps and are considered well represented ex situ. Pixels with a value of 1 8. Meyer, R. S., DuVal, A. E. & Jensen, H. R. Patterns and processes in crop indicate a conservation gap in connectivity, accessibility or the environment; we domestication: an historical review and quantitative analysis of 203 global consider these ‘low-confidence’ gaps. Pixels with a value of 2 indicate gaps in food crops. New Phytol. 196, 29–48 (2012). two metrics or ‘medium-confidence’ gaps, and values of 3 indicate gaps across 9. Khoury, C. K. et al. Origins of food crops connect countries worldwide. Proc. all metrics or ‘high-confidence’ gaps. High-confidence gap areas are displayed R. Soc. B 283, 20160792 (2016). on crop-conservation-gap maps (Fig. 2b and Supplementary Information) and 10. Second Report on the State of the World’s Plant Genetic Resources for Food and conservation hotspot maps across crops (Fig. 4 and Extended Data Figs. 5–8). Agriculture (Food and Agriculture Organization of the United Nations, 2010); The representation of crop landrace groups in current ex situ conservation https://www.fao.org/3/i1500e/i1500e.pdf collections was calculated based on the final 1–3 value conservation-gap maps. 11. Convention on Biological Diversity. Strategic Plan for Biodiversity 2011-2020, The complement of the proportion of the modelled distribution considered as including Aichi Biodiversity Targets (Convention on Biological Diversity, 2010); a potential conservation gap by any single gap score represents the minimum https://www.cbd.int/doc/strategic-plan/2011-2020/Aichi-Targets-EN.pdf NATuRE PLANTS | www.nature.com/natureplants Articles Nature PlaNts 12. Sustainable development goals. United Nations https://sdgs.un.org/2030agenda 44. Rivers + lake centerlines. Natural Earth https://www.naturalearthdata.com/do (2015). wnloads/10m-physical-vectors/10m-rivers-lake-centerlines/ (2019). 13. Ramirez‐Villegas, J. et al. A gap analysis modelling framework to prioritize 45. Siebert, S., Henrich, V., Frenken, K., & Burke, J. Update of the Global Map of collecting for ex situ conservation of crop landraces. Divers. Distrib. 26, Irrigation Areas to Version 5: Project Report (Food and Agriculture 730–742 (2020). Organization of the United Nations, 2013). 14. Vavilov, N. I. Tzentry proiskhozhdeniya kulturnykh rastenii (The centres of 46. Weiss, D. J. et al. A global map of travel time to cities to assess inequalities in origin of cultivated plants). Works Appl. Bot. Plant Breed. 16, 1–248 (1926). accessibility in 2015. Nature 553, 333–336 (2018). 15. Ladizinsky, G. Plant Evolution under Domestication (Kluwer Academic, 1998). 47. Travel time to major cities. Publications Office of the European Commission 16. Halewood, M. et al. Germplasm acquisition and distribution by CGIAR Joint Research Centre—Institute for Environment and Sustainability https:// genebanks. Plants 9, 1296 (2020). op.europa.eu/en/publication-detail/-/publication/20a3a771-15b3-45ac-9606- 17. Plucknett, D. L., Smith N. J. H., Williams, J. T. & Murthi-Anishetty, N. Gene 7575b9df740a/language-en (2008). Banks and the World’s Food (Princeton Univ. Press, 1987). 48. Weidmann, N. B., Rød, J. K. & Cederman, L.-E. Representing ethnic groups 18. Thormann, I., Engels, J. M. M. & Halewood, M. Are the old International in space: a new dataset. J. Peace Res. 47, 491–499 (2010). Board for Plant Genetic Resources (IBPGR) base collections available through 49. You, L. et al. Spatial Production Allocation Model (SPAM) 2005 v3.2 the Plant Treaty’s multilateral system of access and benefit sharing? A review. (MapSPAM, 2019). https://www.mapspam.info/ Genet. Resour. Crop Evol. 66, 291–310 (2019). 50. Harlan, J. Crops and Man (American Society of Agronomy, 1975). 19. Castañeda-Álvarez, N. P. et al. Global conservation priorities for crop wild 51. Jones, H. et al. Approaches and constraints of using existing landrace and relatives. Nat. Plants 2, 16022 (2016). extant plant material to understand agricultural spread in prehistory. Plant 20. Khoury, C. K. et al. Crop wild relatives of the United States require urgent Genet. Resour. Charact. Util. 6, 98–112 (2008). conservation action. Proc. Natl Acad. Sci. USA 117, 33351–33357 (2020). 52. Pal, M. Random forest classifier for remote sensing classification. Int. J. 21. The International Treaty on Plant Genetic Resources for Food and Agriculture Remote Sens. 26, 217–222 (2005). (Food and Agriculture Organization of the United Nations, 2002). 53. Meyer, D., Leisch, F. & Hornik, K. The support vector machine under test. 22. Nagoya Protocol on Access and Benefit-sharing (Convention on Biological Neurocomputing 55, 169–186 (2003). Diversity, 2011). 54. Guo, G., Wang, H., Bell, D., Bi, Y. & Greer, K. On the Move to Meaningful 23. SDG Indicators, Metadata Repository, Goal 2. End hunger, achieve food Internet Systems 2003: CoopIS, DOA, and ODBASE Vol. 2888 (eds Meersman, security and improved nutrition and promote sustainable agriculture. United R. et al.) 986–996 (Springer, 2003). Nations https://unstats.un.org/sdgs/metadata/?Text=&Goal=2&Target (2021). 55. Dreiseitl, S. & Ohno-Machado, L. Logistic regression and artificial neural 24. Scherr, S. J. & McNeely, J. A. Biodiversity conservation and agricultural network classification models: a methodology review. J. Biomed. Inform. 35, sustainability: towards a new paradigm of ‘ecoagriculture’ landscapes. 352–359 (2002). Phil. Trans. R. Soc. B 363, 477–494 (2008). 56. Phillips, S. J., Anderson, R. P. & Schapire, R. E. Maximum entropy 25. Winkler, K., Fuchs, R., Rounsevell, M. & Herold, M. Global land use changes modeling of species geographic distributions. Ecol. Model. 190, are four times greater than previously estimated. Nat. Commun. 12, 2501 (2021). 231–259 (2006). 26. Zeven, A. C. The traditional inexplicable replacement of seed and seed ware 57. Elith, J. et al. A statistical explanation of MaxEnt for ecologists. Divers. of landraces and cultivars: a review. Euphytica 110, 181–191 (1999). Distrib. 17, 43–57 (2011). 27. Hanson, J. O., Rhodes, J. R., Riginos, C. & Fuller, R. A. Environmental and 58. Phillips, S. J., Anderson, R. P., Dudík, M., Schapire, R. E. & Blair, M. E. geographic variables are effective surrogates for genetic variation in Opening the black box: an open-source release of Maxent. Ecography 40, conservation planning. Proc. Natl Acad. Sci. USA 114, 12755–12760 (2017). 887–893 (2017). 28. Hoban, S., Kallow, S. & Trivedi, C. Implementing a new approach to effective 59. Warren, D. L. & Seifert, S. N. Ecological niche modeling in Maxent: the conservation of genetic diversity, with ash (Fraxinus excelsior) in the UK as a importance of model complexity and the performance of model selection case study. Biol. Conserv. 225, 10–21 (2018). criteria. Ecol. Appl. 21, 335–342 (2011). 29. Sperling, L. The effect of the civil war on Rwandas bean seed systems and 60. Braunisch, V. et al. Selecting from correlated climate variables: a major source unusual bean diversity. Biodivers. Conserv. 10, 989–1010 (2001). of uncertainty for predicting species distributions under climate change. 30. Willett, W. et al. Food in the Anthropocene: the EAT–Lancet Commission on Ecography 36, 971–983 (2013). healthy diets from sustainable food systems. Lancet 393, 447–492 (2019). 61. Sayre, R. et al. A New Map of Global Ecological Land Units—An 31. Brush, S. B. In situ conservation of landraces in centers of crop diversity. Ecophysiographic Stratification Approach (Association of American Crop Sci. 35, 346 (1995). Geographers, 2014). 32. Khoury, C. K. et al. Comprehensiveness of conservation of useful wild plants: 62. Senay, S. D., Worner, S. P. & Ikeda, T. Novel three-step pseudo-absence an operational indicator for biodiversity and sustainable development targets. selection technique for improved species distribution modelling. PLoS ONE 8, Ecol. Indic. 98, 420–429 (2019). e71218 (2013). 33. Genesys-PGR: a gateway to genetic resources. Global Crop Diversity Trust 63. Lee, D. T. & Schachter, B. J. Two algorithms for constructing a Delaunay https://www.genesys-pgr.org/ (2021). triangulation. Int. J. Comput. Inf. Sci. 9, 219–242 (1980). 34. United Nations Food and Agriculture Organization World Information and 64. Turner, R. deldir: Delaunay triangulation and Dirichlet (Voronoi) tessellation. Early Warning System on Plant Genetic Resources for Food and Agriculture. R package version 0.1-16. https://cran.r-project.org/web/packages/deldir/ Food and Agricultural Organization of the United Nations http://www.fao.org/ deldir.pdf (2021). wiews/en/ (2021). 35. Ruas, M. et al. MGIS: managing banana (Musa spp.) genetic resources information and high-throughput genotyping data Database https://doi. Acknowledgements org/10.1093/database/bax046 (2017). Support was provided by the CGIAR Genebank Platform (J.R.-V., C.K.K., H.A., M.V.D., 36. National Plant Germplasm System, GRIN-Global Accessions. US Department A.M., C.C.S., Z.K., L.G., M.A., J.A., B.A.A., J.C.A., A.A., N.L.A., V.A., K.A., G.L.C., O.C., of Agriculture Agricultural Research Service https://npgsweb.ars-grin.gov/ D.C., D.E.C., D.G.D., D.E., H.F., A.F., M.E.G., P.G., A.J.G., B.G., A.I.E.H., R.J., C.S.J., gringlobal/search (2021). B.K., J.-S.L., K.L.M., A.M., M.-N.N., O.O., T.S.P., S.R., G.R., N.R., M.R., C.S., J.S., T.D.S., 37. Portal de Geoinformacion. Comisión Nacional para el Conocimiento y Uso M.T., I.v.d.H., J.A.V., R.V., P.W., M.Y. and C.Z.) and by grant number 2019-67012-29733/ de la Biodiversidad http://www.conabio.gob.mx/informacion/gis/ (2021). project accession number 1019405 from the USDA National Institute of Food and 38. Fick, S. E. & Hijmans, R. J. WorldClim 2: new 1‐km spatial resolution climate Agriculture (C.K.K.). surfaces for global land areas. Int. J. Climatol. 37, 4302–4315 (2017). 39. Title, P. O. & Bemmels, J. B. ENVIREM: an expanded set of bioclimatic and Author contributions topographic variables increases flexibility and improves performance of J.R.-V., C.K.K., H.A., M.V.D., C.C.S., Z.K. and L.G. conceived and designed the study. ecological niche modeling. Ecography 41, 291–307 (2018). J.R.-V., C.K.K., H.A., M.V.D., A.M., C.C.S., Z.K., L.G., M.A., J.A., B.A.A., J.C.A., A.A., 40. Reuter, H. I., Nelson, A. & Jarvis, A. An evaluation of void‐filling interpola- N.L.A., V.A., K.A., G.L.C., O.C., D.C., D.E.C., D.G.D., D.E., H.F., A.F., M.E.G., P.G., tion methods for SRTM data. Int. J. Geogr. Inf. Sci. 21, 983–1008 (2007). A.J.G., B.G., A.I.E.H., R.J., C.S.J., B.K., J.-S.L., K.L.M., A.M., M.-N.N., O.O., T.S.P., 41. Jarvis, A., Reuter, H. I., Nelson, A., & Guevara, E. Hole-filled Seamless SRTM S.R., G.R., N.R., M.R., C.S., J.S., T.D.S., M.T., I.v.d.H., J.A.V., R.V., P.W., M.Y. and C.Z. Data V4 (International Center for Tropical Agriculture, 2008); https://cgiarcsi. contributed data and conceptual inputs. J.R.-V., C.K.K., H.A., M.V.D., A.M., C.C.S., Z.K., community/data/srtm-90m-digital-elevation-database-v4-1/ K.A., O.C., A.I.E.H. and S.R. compiled and processed data and wrote and ran the code. 42. Reba, M., Reitsma, F. & Seto, K. C. Spatializing 6,000 years of global J.R.-V., C.K.K., H.A., M.V.D., A.M., C.C.S., Z.K. and L.G. interpreted the results. J.R.-V., urbanization from 3700 bc to ad 2000. Sci. Data 3, 160034 (2016). C.K.K., M.V.D. and A.M. wrote the paper. J.R.-V., C.K.K., H.A., M.V.D., A.M., C.C.S., 43. Gridded Population of the World, Version 4 (GPWv4): Population Density, Z.K., L.G., M.A., J.A., B.A.A., J.C.A., A.A., N.L.A., V.A., K.A., G.L.C., O.C., D.C., D.E.C., Revision 11 (Center for International Earth Science Information Network, D.G.D., D.E., H.F., A.F., M.E.G., P.G., A.J.G., B.G., A.I.E.H., R.J., C.S.J., B.K., J.-S.L., 2018); https://sedac.ciesin.columbia.edu/data/set/gpw-v4-population- K.L.M., A.M., M.-N.N., O.O., T.S.P., S.R., G.R., N.R., M.R., C.S., J.S., T.D.S., M.T., I.v.d.H., density-rev11 J.A.V., R.V., P.W., M.Y. and C.Z. edited the paper. NATuRE PLANTS | www.nature.com/natureplants Nature PlaNts Articles Competing interests Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in The authors declare no competing interests. published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Additional information Attribution 4.0 International License, which permits use, sharing, adap-tation, distribution and reproduction in any medium or format, as long Extended data is available for this paper at https://doi.org/10.1038/s41477-022-01144-8. as you give appropriate credit to the original author(s) and the source, provide a link to Supplementary information The online version contains supplementary material the Creative Commons license, and indicate if changes were made. The images or other available at https://doi.org/10.1038/s41477-022-01144-8. third party material in this article are included in the article’s Creative Commons license, Correspondence and requests for materials should be addressed to unless indicated otherwise in a credit line to the material. If material is not included in Julian Ramirez-Villegas or Colin K. Khoury. the article’s Creative Commons license and your intended use is not permitted by statu- tory regulation or exceeds the permitted use, you will need to obtain permission directly Peer review information Nature Plants thanks James Borell, Gayle Volk and the other, from the copyright holder. To view a copy of this license, visit http://creativecommons. anonymous, reviewer(s) for their contribution to the peer review of this work. org/licenses/by/4.0/. Reprints and permissions information is available at www.nature.com/reprints. © The Author(s) 2022 NATuRE PLANTS | www.nature.com/natureplants Articles Nature PlaNts Extended Data Fig. 1 | Richness map of the predicted distributions of 71 landrace groups of 25 cereal, pulse, and starchy root/tuber/fruit crops within their geographic regions of diversity. Richness map of the predicted distributions of 71 landrace groups of 25 cereal, pulse, and starchy root/tuber/fruit crops within their geographic regions of diversity. Darker colors indicate greater numbers of crop landrace groups potentially overlapping in the same 2.5 arc-minute cells. NATuRE PLANTS | www.nature.com/natureplants Nature PlaNts Articles Extended Data Fig. 2 | Richness map of the predicted distributions of landrace groups of 9 cereal crops within their geographic regions of diversity. Richness map of the predicted distributions of landrace groups of 9 cereal crops within their geographic regions of diversity. Darker colors indicate greater numbers of crop landraces potentially overlapping in the same 2.5 arc-minute cells, quantified in terms of numbers of crops. NATuRE PLANTS | www.nature.com/natureplants Articles Nature PlaNts Extended Data Fig. 3 | Richness map of the predicted distributions of landrace groups of 9 pulse crops within their geographic regions of diversity. Richness map of the predicted distributions of landrace groups of 9 pulse crops within their geographic regions of diversity. Darker colors indicate greater numbers of crop landraces potentially overlapping in the same 2.5 arc-minute cells, quantified in terms of numbers of crops. NATuRE PLANTS | www.nature.com/natureplants Nature PlaNts Articles Extended Data Fig. 4 | Richness map of the predicted distributions of landrace groups of 7 starchy root, tuber, and fruit crops within their geographic regions of diversity. Richness map of the predicted distributions of landrace groups of 7 starchy root, tuber, and fruit crops within their geographic regions of diversity. Darker colors indicate greater numbers of crop landraces potentially overlapping in the same 2.5 arc-minute cells, quantified in terms of numbers of crops. NATuRE PLANTS | www.nature.com/natureplants Articles Nature PlaNts Extended Data Fig. 5 | Geographic hotspots for further collection for the ex situ conservation of crop landrace groups. Geographic hotspots for further collection for the ex situ conservation of crop landrace groups. The map displays ‘gap richness’ across the predicted worldwide distributions of 71 landrace groups of 25 cereal, pulse, and starchy root/tuber/fruit crops within their geographic regions of diversity, indicating where landrace groups are expected to occur and have not yet been collected and conserved in genebanks. Darker colors indicate greater numbers of un-collected crop landrace groups potentially overlapping in the same 2.5 arc-minute cells. NATuRE PLANTS | www.nature.com/natureplants Nature PlaNts Articles Extended Data Fig. 6 | Geographic hotspots for further collection for the ex situ conservation of landrace groups of cereal crops. Geographic hotspots for further collection for the ex situ conservation of landrace groups of cereal crops. The map displays ‘gap richness’ across the predicted distributions of landrace groups of 9 cereal crops within their geographic regions of diversity, indicating where landrace groups are expected to occur and have not yet been collected and conserved in genebanks. Darker colors indicate greater numbers of un-collected cereal crop landrace groups potentially overlapping in the same 2.5 arc-minute cells, quantified in terms of numbers of crops. NATuRE PLANTS | www.nature.com/natureplants Articles Nature PlaNts Extended Data Fig. 7 | Geographic hotspots for further collection for the ex situ conservation of landrace groups of pulse crops. Geographic hotspots for further collection for the ex situ conservation of landrace groups of pulse crops. The map displays ‘gap richness’ across the predicted distributions of landrace groups of 9 pulse crops within their geographic regions of diversity, indicating where landrace groups are expected to occur and have not yet been collected and conserved in genebanks. Darker colors indicate greater numbers of un-collected pulse crop landrace groups potentially overlapping in the same 2.5 arc-minute cells, quantified in terms of numbers of crops. NATuRE PLANTS | www.nature.com/natureplants Nature PlaNts Articles Extended Data Fig. 8 | Geographic hotspots for further collection for the ex situ conservation of crop landrace groups of starchy root, tuber, and fruit crops. Geographic hotspots for further collection for the ex situ conservation of crop landrace groups of starchy root, tuber, and fruit crops. The map displays ‘gap richness’ across the predicted distributions of landrace groups of 7 starchy root, tuber, and fruit crops within their geographic regions of diversity, indicating where landrace groups are expected to occur and have not yet been collected and conserved in genebanks. Darker colors indicate greater numbers of un-collected starchy root, tuber, and fruit crop landrace groups potentially overlapping in the same 2.5 arc-minute cells, quantified in terms of numbers of crops. NATuRE PLANTS | www.nature.com/natureplants Articles Nature PlaNts Extended Data Fig. 9 | Comparison of ex situ conservation representation of crop landrace groups and crop wild relative (CWR) for 25 cereal, pulse, and starchy root/tuber/fruit crops. Comparison of ex situ conservation representation of crop landrace groups and crop wild relative (CWR) for 25 cereal, pulse, and starchy root/tuber/fruit crops. For CWR, conservation representation results were first averaged across CWR taxa in each crop genepool19. The summary results were also averaged across related crops assessed here; for example, the results for three yam crop genepools were averaged to form a single result for the global yam genepool. The crop genepool results were then transformed to the crop landrace scale and format used here, and are compared to the crop aggregated-level conservation representation average (%) estimate. Crop wild relatives of taro were not assessed in Castaneda-Alvarez et al. (2016)19; for this figure the pertinent score was set to zero. Cereals are displayed in gold, pulses in green, and starchy roots, tubers, and fruits in purple. NATuRE PLANTS | www.nature.com/natureplants