biogeografía-ago2026

Protocolo de OCR selectivo para la digitalización masiva de etiquetas de ejemplares de herbario
Miguel Murguía-Romero ^a, *, Diana G. Flores-Camargo ^b

a Universidad Nacional Autónoma de México, Instituto de Biología, Unidad de Informática para la Biodiversidad, Tercer Circuito Exterior s/n, Ciudad Universitaria, Coyoacán, 04510 Ciudad de México, Mexico
b Universidad Nacional Autónoma de México, Instituto de Biología, Jardín Botánico, Tercer Circuito Exterior s/n, Ciudad Universitaria, Coyoacán, 04510 Ciudad de México, Mexico

*Corresponding author: miguel.murguia@ib.unam.mx (M. Murguía-Romero)

Received: 07 October 2025; accepted: 20 March 2026

Abstract

It still remains to digitize label data of a high percentage of specimens of vascular plants in the herbariums. There are experiences of the use of the OCR technique to support the process of digitization of specimens, however, it is still necessary to explore and describe in greater detail its limitations and strengths. The digitization of some of the data of the labels from herbarium specimens can be done massively and automatically by applying optical character recognition techniques (OCR). Five target fields (geographic and taxonomic super-primary fields) were selected to obtain their information through OCR applied to 8,451 images of herbarium specimens, guided by a label information architecture and with human intervention. The information contained in the 5 target fields was identified in 70.6% of the labels, 23.7% in 1–4 of the target fields, and only in 5.7% none could be identified. Mistakes ranged from 0.8 to 3.3% depending on the field. OCR cannot automatically identify all information fields of herbarium specimen labels; however, it is possible in a high percentage to retrieve the information of the most consulted fields.

Keywords: Biodiversity informatics; Biological collections; Databases; Image processing; MEXU

Resumen

Aún falta digitalizar los datos de las etiquetas de un alto porcentaje de especímenes de plantas vasculares en los herbarios. Existen experiencias del uso de la técnica OCR para apoyar el proceso de digitalización de especímenes; sin embargo, aún es necesario explorar y describir con mayor detalle sus limitaciones y fortalezas. La digitalización de algunos de los datos de las etiquetas de ejemplares de herbario se puede realizar de forma masiva y automática mediante la aplicación de técnicas de reconocimiento óptico de caracteres (OCR). Se seleccionaron 5 campos objetivo (campos superprimarios geográficos y taxonómicos) para obtener su información mediante OCR aplicado a 8,451 imágenes de especímenes de herbario, guiado por una arquitectura de la información de las etiquetas y con intervención humana. La información contenida en los 5 campos objetivo se identificó en 70.6% de las etiquetas, en 23.7% en 1-4 de los campos objetivo y solo en 5.7% no se pudo identificar ninguna. Los errores variaron de 0.8 a 3.3% según el campo. El OCR no puede identificar automáticamente todos los campos de información de las etiquetas de organismos de herbario, sin embargo, es posible obtener la información en un alto porcentaje de los campos más consultados.

Palabras clave: Informática de la biodiversidad; Colecciones biológicas; Bases de datos; Procesamiento de imágenes; MEXU

Introduction

Having records of herbarium specimens in databases is important from 2 perspectives. First, it allows users to have information regarding global flora in different geographical scales in order to propose more effective and precise strategies for its utilization and conservation (Soltis et al., 2018). Secondly, it facilitates the administration and curation of the collection, because doing an efficient curation of a collection with thousands of specimens as well as maintaining the services that they provide to different types of users is practically impossible now without a database of its specimens.

Currently, we have a great number of herbarium specimens already digitized and available in public databases. For example, the website GBIF (www.gbif.org) has 102.5 million records of vascular plant specimens available. Nevertheless, despite the digitization efforts of the last 3 decades, a great number of specimens from a variety of collections have not been digitized yet, therefore, the information is not available in databases (Corlett, 2022). It has been estimated that at the present speed of the cataloging process, 1,500 years are needed to complete the digitization of the global specimens (Blagoderov et al., 2012). This showcases the need to generate more efficient technological and administrative strategies in order to have, in a shorter period, complete data on the global flora and build comprehensive databases that can be used for analysis.

The optical character recognition technology (OCR) has been proposed to regain the information from herbarium’s labels massively, but the exercises in this regard achieve recognition rates below 50% (Takano et al., 2024). However, to increase the recognition percentage, the OCR can be combined with the appropriate use of the database technology, particularly, through the use of catalogs and SQL language, as well as an adequate definition of the information architecture of a specimen that guides the text recognition.

This work proposes a protocol that defines a strategy to fill the gap in information on herborized specimens that have not yet been digitized. The proposed protocol, the Selective-OCR protocol (Appendix 1), addresses 2 aspects in a novel manner. First, it explicitly recognizes that OCR technology fails in the recognition of some words or text. Secondly, it proposes the selective digitization of only 5 of the primary fields (Conabio, 2019), referred to here as super-primary fields. The information of these fields is useful to delineate searches of interest of specimens for users that look for the data. This, when associated with photographs, makes it possible to complete the other data capture with additional efforts based on specific needs, for example, it will be possible to locate the records of specimens collected from a particular state or province, or for a specific species, in the database.

Materials and methods

The Mexican National Herbarium of the Instituto de Biología of the Universidad Nacional Autónoma de México (MEXU) has around 1.6 million herborized specimens of vascular plants, 1.3 million have been digitized. It is estimated that 300,000 specimens from this collection have yet to be digitized (Murguía-Romero et al., 2024). We have selected 8,451 images from herborized specimens to process them via OCR to obtain the data from their labels. The images come from specimens from diverse families and collection years. They were photographed from January to September 2023 by the herbarium personnel. This activity constitutes a previous step to the capture of the information of its labels, which is usually carried out by the same staff and volunteers using the IBdata system’s capture interface (https://ibdata4.ib.unam.mx, Murguía-Romero, Serrano-Estrada et al., 2023; Murguía-Romero et al., 2024). IBdata is the web system developed by the Instituto de Biología of the Universidad Nacional Autónoma de México through which the database of specimens of the biological collections (zoological, botanical and mycological) can be consulted, which stores the information following the Darwin Core standard (https://dwc.tdwg.org/terms/); in addition to the visualization of the specimen records, it integrates a capture and editing interface with validation rules.

Because the photography process includes controls that ensure that work is not duplicated, this also ensures that there are no duplicate specimens in the image sample used to apply the OCR protocol. Two geographic fields (country and state) and 3 taxonomic fields (family, genus, and species) were defined as super-primary fields, which have also been called GeoTax fields (Murguía-Romero et al., 2024; Fig. 1). To quantify the percentage of error of assignment of values to each of the 5 fields, 30% of the processed records were reviewed along with the image of their labels. The purpose of applying the protocol in this exercise was to assign 5 super-primary fields of the label of each specimen and to make them available for their public consultations in the IBdata institutional web system.

The only information about the specimens that can be determined from their images prior to processing is the catalog number, as this is part of the file name. Therefore, stratified sampling, for example by taxonomic family, collection date, or state, is not possible. To verify that the sample was unbiased, a Wilcoxon test was performed post-processing, that is, after OCR processing, to compare the percentages by state in the sample and across all specimens in the MEXU database. The image sample included specimens from more than 50 herbaria, ensuring a high level of representativeness in the structure of the processed labels.

The specimen’s images were subjected to the Selective-OCR protocol (Fig. 2), which can be considered a specialization of the object-to-image-to-data workflow (Nelson et al., 2015). The result of applying OCR to each image was stored in a text file. The average number of lines, after excluding lines with 1 to 2 characters, was 16.5 (SD = 5.11) with a range of 1 to 57. The working database was integrated with 4 tables. The center table “dat_line” was integrated by exporting the lines of text obtained with OCR with 2 additional fields: the row number (sequential) and the name of the image file. The catalog of species, genera, and families “cat_tax” (according to the names accepted by the Kew’s World Check List of Vascular Plants, https://powo.science.kew.org; Murguía-Romero, Ortiz et al., 2023), and the catalog from countries and states “cat_geo” (states or provinces) formed 2 other tables. A fourth table “dat_specimen” was created with the list of image names and by adding the 5 super-primary fields (country, state/province, family, genus, and species) and the catalog number to store the identified value in the text lines of the OCR.

Through queries to the database, and by using regular expressions (Friedl, 2006), the names of the species and states were identified in the OCR text table by using the respective catalog tables. The values from the super-primary fields were stored in the particular fields of each record of each label. These searches assume a standard information architecture of a specimen (Morville, 2006), meaning, a pattern of the order and place of each type of information on the surface of the specimen and the label can be proposed. The information architecture proposed includes a label from the specimen, the determination correction labels, the catalog number, and the institutional seals, among other elements (Appendix 2).

Results

The 5 super-primary fields were recovered in 70.6% of the 8,451 OCR analyzed images, in 23.7% at least 1 of the 5 were recovered, and only in 5.7% of cases no fields were recovered (Fig. 3). A manual capture was performed of records corresponding to labels whose primary fields could not be recovered by processing the OCR lines database. In total, the values of the super-primary fields of 2,483 specimen label records were manually recovered: 2,002 in which the OCR did not identify a super-primary field and 481 in which none were identified.

After processing, it was determined that the specimens in the sample came from 24 different countries, and in the case of Mexico, all 32 states were represented, and more than 200 botanical families were included, which ensures a broad representation of the type of labels on the specimens and reflects the variability of labels in herbaria around the world. No significant differences were found when comparing the percentages of specimens in the 32 Mexican states, which indicates that the sample is not biased in this respect.

One of the main aspects that must be addressed to move towards the evaluation and improvement of data quality is to estimate the error of the proposed protocol (Chapman et al., 2020). To do this, the values of the records that involved manual capture were compared before and after being reviewed and corrected, so the error was estimated based on a 30% sample. The estimated error rate varied from 0.8% to 3.3% depending on the field (Table 1).

Super-primary field	% assigned values	% estimated error (95% CI)	Overall success rate (95% CI)
Country	90.30	0.62 (0.20-1.04)	89.74 (88.13-91.35)
State/Province	83.10	0.78 (0.31-1.25)	82.45 (80.43-84.47)
Family	84.10	1.50 (0.85-2.15)	82.88 (80.88-84.88)
Genus	84.10	3.30 (2.35-4.25)	81.31 (79.24-83.38)
Species	79.30	0.80 (0.33-1.27)	78.67 (76.49-80.85)

Among the main causes identified as to why the values from the super-primary fields could not be recovered with OCR include the badly oriented images, images with degraded letters, text in manuscript, and dot matrix printed labels, among others.

When the values of the 5 super-primary fields for the 8,451 records were registered in the “dat_specimen” table, the catalog identifiers from the IBdata database were assigned and imported into the web system for public consultation. Of this set, 1,145 had already been captured by the herbarium staff and some photographs corresponded to the same specimen, therefore, only 6,667 new records were imported into IBdata. In the IBdata interface (https://ibdata4.ib.unam.mx) the list of these records can be obtained using the Simple Search with the word ocr2023.

Discussion

Using the Selective-OCR protocol, applied by a 2-person team, it was possible to integrate data from 5 fields with more than 8,000 herborized specimens and assign them to their respective Darwin Core fields in a searchable online database. Measuring the method’s efficiency is an important parameter for comparing different approaches. In the proposed protocol, the success rate of assignment to the 5 target fields was estimated at 70.6%, while in 5.7% of cases none of the fields could be assigned. The result of the digitization through the Selective-OCR protocol is available in a public database where the data of the super-primary fields of the specimens that were analyzed can be consulted, which provides to those interested in digitization a context of the strengths and weaknesses of this tool.

Other efforts to systematize the use of OCR for digitizing data from herbarium specimen labels are reported in the literature. For example, Guralnick et al. (2024) reported the digitization of labels from 7,995 specimens using OCR and the participation of 177 volunteers; however, the method they propose focuses on the correct identification of the type of label on the specimen; it does not include the classification of information into Darwin Core fields, which, from the perspective of the authors, should be performed by humans. The digitization result cannot be consulted in a public database, which makes it difficult to evaluate the results. Therefore, the success rate in the automatic identification of information from specific fields is not reported either.

The Selective-OCR protocol provides a methodological framework for organizing biological collection digitization projects, as it is structured in sequential steps that allow for clear management of resources and objectives. Another strength of this procedure is that the information architecture of the herborized specimen and the label constitute a tool that addresses the diversity of label formats and can be adapted to different label models. The proposed information architecture is a tool that guides in the allocation of the Darwin Core field. In fact, this architecture is inspired by the knowledge representation called “scripts” (Schank & Abelson, 1975, 1977) in the context of Artificial Intelligence, proposed to represent paradigmatic situations, which, when framed within a general structure, facilitate information processing in specific contexts. The information architecture can be further improved by incorporating more elements considered in the scripts construct. Kirchhoff et al. (2018) have proposed a service-based workflow for automated information extraction from herbarium specimens in which the ‘text operator’ conducts and coordinates all text processing activities after text extraction by OCR, and is assisted by means of a proper user interface; the protocol proposed here addresses the need to reduce the role of the ‘text operator’ so that the identification of fields with Darwin Core is performed automatically through database queries guided by the information architecture. After identifying the type of information, for example, taxonomic names, the corresponding Darwin Core field must be identified, in this case, if it is the identification of the specimen or associated taxa, or in the case of personal names, if it is the person who determined the specimen or the one who collected it.

One of the weaknesses of the protocol we present here is that it did not estimate productivity or costs. This is because the proposed protocol requires human intervention in 6 of the 7 steps. Just as the speed and quality of records of manually captured specimens depend largely on the profile of the person capturing them, the same is true for human intervention in the proposed protocol.

Although the primary objective of the protocol is to digitize specimens whose data are not yet available in databases, the protocol can also be used to validate the integrity and quality of the information captured manually. For example, we have detected that in some database records there are entries without values in the geographic coordinate fields, however, the labels of the corresponding specimens do specify them, indicating that there was an omission in the capture process. In these cases, although the OCR cannot accurately transcribe the coordinates from the labels in the images, it is possible to detect whether they are coordinates. Therefore, the presence or absence of coordinates in the database records can be compared, thus estimating the degree of completeness in the capture of this field.

This work did not investigate or evaluate Large Language Models (LLM; Shanahan, 2024) tools, which are colloquially known as Generative Artificial Intelligences (GAI; Pavlick, 2025). These tools can be used to assign each line of text extracted from the OCR to a specific field; however, in cases where they are applied, the assignment error must also be estimated in order to report the error percentage in queries that access records constructed with this method. The information architecture of the herborized specimen and the label used for the Selective-OCR protocol can be used as a prompt for the GAI tools (Korzynski et al., 2023).

The use of OCR technology is necessarily linked to relational database technology, which is the artifact in which the scan results will ultimately be stored. Therefore, one of the weaknesses of the proposed protocol is that its implementation requires personnel specialized in database management with knowledge of biodiversity data. For example, it will be difficult for a person to review and validate OCR outputs using only spreadsheets without resorting to database management. Likewise, it will be difficult to do it for an IT professional without knowledge of biodiversity as a scientific name, accepted name, synonym, taxonomic hierarchy, among many others. It is therefore important to train personnel specialized in biodiversity informatics, as well as the integration of multidisciplinary work teams.

The Selective-OCR protocol deals with the fact that the use of OCR to recover information from herbarium specimen labels is not all-encompassing. With OCR and its subsequent processing in a relational database using regular expressions, and guided by an information architecture of the specimen, recovering the fields that can be corroborated by catalogs, such as taxonomic and geographic catalogs, is massively feasible and more efficient. The effectiveness and the margin of error of the Selective-OCR protocol reported here can be useful for comparison when applied to more specimens and in other collections.

Acknowledgements

The authors thank the entire MEXU herbarium staff who maintain the collection and participate in photographing the specimens. José Luis Villaseñor reviewed a preliminary version of the manuscript, and his suggestions substantially improved the final version. We are deeply grateful for the comments from the Editor in Chief of RMB and from an anonymous reviewer who substantially improved the final version of the manuscript. Fernanda Y. Bernal Bonilla reviewed and edited the English of the preliminary version.

Appendix 1. Steps of the Selective-OCR protocol

Step 1. Image preparation

The optical character recognition (OCR) process is carried out on the specimen photographic image. It is important to verify the correct orientation (vertical or horizontal), as well as a minimum resolution of 300 dpi (or even 600 dpi if the font size is very small) to allow recognition using the OCR tool. Two aspects must be decided or taken into account; the first is to define whether character recognition will be performed on the entire image area or only on the area where the data label(s) is (are) located. By choosing the area of the image where only the data label is located, information that can cause noise in the identification of the target information, such as stamps and catalog numbers, can be eliminated. On the other hand, the advantage of applying OCR to the entire image area is that additional information can be identified. What was noise in the context of objective fields, now acquires value, such as the ability to recover catalog numbers. The second aspect is the nomenclature of the image files. It is important that from the creation of the files, names that facilitate their management are assigned, for example, the initials of the person who took the picture, the date of capture, or the room or drawer where the specimens come from, in this last case it might be possible to recognize from the beginning the family or taxonomic group of the specimen.

Step 2. OCR application to the images

OCR will be applied to each of the specimen images, storing in text files all recognized lines. The result will be text files for each image. Diverse programs that do OCR exist, the most adequate one can be chosen depending on the specimen characteristics and the images that will be processed.

Step 3. Working database preparation

The text files coming from the OCR (step 2) will be imported into a database table (“dat_line”). It is recommended to include the specimen or photograph identifier, the line number, and the OCR text line, as fields. The line number is useful when an information architecture is considered for the herborized specimen (Appendix 2), because, when more than one scientific name of a species appears on the label, the order in which they appear is useful to identify the last identification of the specimen and, above all, differences from other scientific names that are indicated as associated taxa (associatedTaxa in Darwin Core, Darwin Core Maintenance Group, https://dwc.tdwg.org/terms/). In addition to the table of OCR lines, it is important to include tables of geographic (county, and state/province) and taxonomic (families, genera, and species) catalogs, such as “cat_geo” and “cat_tax.” As well as a fourth table “dat_specimen” where the values from the five super-primary fields will be stored, that can include the specimen identifier or image of the digitized specimen as a primary key.

Step 4. Recovery of values from the super-primary fields

In the “dat_specimen” table, the values from the five super-primary fields will be stored and identified by database queries that use regular expressions in the geographic and taxonomic catalogs. The information architecture of the labels is a key element in differentiating homonym values, for example, the name of the country can be associated with the place of the collector’s affiliated institution instead of the collection place or locality (e.g. “Herbario de la Universidad de Sonora”). A taxonomic name may refer to a taxon associated with the locality of collection and not to the taxon identity of the specimen, for instance, “Associated with Pachycereus weberi.” The order and disposition of the information architecture will determine the type of information it will deal with.

Step 5. Content validation and error estimate

Once the “dat_specimen” table has been completed with the values that the OCR identified, a validation of the allocated data will be carried out by quantifying those cases in which they have been erroneously assigned. This evaluation can be done by either comparing previously captured records done by persons or by selecting a subset of 200 to 400 records and revising them manually with the specimen labels. From this evaluation, the percentage of incorrect allocations of values can be estimated and reported to document the process of digitization via OCR in the respective collection.

Step 6. Manual capture to recover missing data

Optionally, the values from the fields that could not be identified by the OCR will be completed. This is the purpose of identifying the possible causes of error; so that the whole set of processed fields has the values from the five super-primary fields. That way, the users can filter out and select the subsets of interest according to geographic and taxonomic criteria.

Step 7. Publication of data from specimen records

Finally, from the “dat_specimen” table, the analyzed specimen data will be published in repositories that allow for public consultation.

Appendix 2. Information architecture of the herborized specimen and the label used for the Selective-OCR protocol

The information architecture of a specimen (Fig. A2.1) describes the topology of the elements in a paradigmatic herbarium sample that is used as a guide to identify the correct values of the objective fields of the OCR. Nevertheless, each element could not be in the place described by the information architecture, it serves as a general guide that works for most records.

As part of the information architecture of a specimen, the information architecture of a specimen label is also included (Fig. A2.2), which describes the topology of the information within the main label of the specimen.

In the case that the information is preceded by a category (for example, “Det”, “Collection date”, etc.), it can be identified easily by the analysis of the texts obtained from the OCR in the database. But, if it is the opposite case, the information architecture helps in this process since the elements corresponding to each field, by the order or apparition on the label, can be excluded or selected. For example, the architecture indicates that the associated taxa are positioned after the taxonomic identification, that way, the first taxonomic name that appears on the texts can be assigned as the identity of the specimen (scientificName), while the subsequent ones, to the “associated taxa” field.

Figure A2.1. Information architecture of a herbarium specimen.

Figure A2.2. Information architecture of a herbarium specimen label.

References

Blagoderov, V., Kitching, I. J., Livermore, L., Simonsen, T. J., & Smith, V. S. (2012). No specimen left behind: industrial scale digitization of natural history collections. Zookeys, 209, 133–146. https://doi:10.3897/zookeys.209.3178

Chapman, A. D., Belbin, L., Zermoglio, P. F., Wieczorek, J., Morris, P. J., Nicholls, M. et al. (2020). Developing standards for improved data quality and for selecting fit for use biodiversity data. Biodiversity Information Science and Standards, 4, e50889. https://doi.org/10.3897/biss.4.50889

Conabio (Comisión Nacional para el Conocimiento y Uso de la Biodiversidad). (2019). Datos primarios de ejemplares del Sistema Nacional sobre Biodiversidad (SNIB) – características y reglas –. Ciudad de México. Last access June 3, 2025. www.snib.mx/ejemplares/docs/CONABIO-SNIB-ProtocoloCalidadI.pdf

Corlett, R. T. (2022). We need to accelerate the digitization of existing botanical information and complete the global plant inventory. Integrative Conservation, 1, 6–7. https://doi.org/10.1002/inc3.12

Friedl, J. E. (2006). Mastering regular expressions. Sebastopol, CA: O’Reilly Media, Inc.

Guralnick, R., LaFrance, R., Denslow, M., Blickhan, S., Bouslog, S. M., Yost, J. et al. (2024). Humans in the loop: community science and machine learning synergies for overcoming herbarium digitization bottlenecks. Applications in Plant Sciences, 12, e11560. https://doi.org/10.1002/aps3.11560

Kirchhoff, A., Bügel, U., & Santamari, E. (2018). Toward a service-based workflow for automated information extraction from herbarium specimens. Database, 2018, bay103. https://doi.org/10.1093/database/bay103

Korzynski, P., Mazurek, G., Krzypkowska, P., & Kurasinski, A. (2023). Artificial intelligence prompt engineering as a new digital competence: analysis of generative AI technologies such as ChatGPT. Entrepreneurial Business and Economics Review, 11, 25–37. https://doi.org/10.15678/EBER.2023.110302

Morville, P., & Rosenfeld, L. (2006). Information architecture for the World Wide Web: designing large-scale web sites. Sebastopol, CA: O’Reilly Media, Inc.

Murguía-Romero, M., Ortiz, E., Serrano-Estrada, B., & Villaseñor, J. L. (2023). The Kew’s” World Checklist of Vascular Plants” and its relevance to the knowledge of the flora of Mexico. Botanical Sciences, 101, 632–653. https://doi.org/10.17129/botsci.3223

Murguía-Romero, M., Serrano-Estrada, B., Salazar, G., Gernandt, D. S., Melo-Samper-Palacios, U., Sánchez- González, G. E. et al. (2023). IBdata v3 “Helia Bravo Hollis”. Manual de uso. Ciudad de México: Instituto de Biología, Universidad Nacional Autónoma de México.

Murguía-Romero, M., Serrano-Estrada, B., Salazar, G. A., Sánchez-González, G. E., Melo-Samper-Palacios, U., Gernandt, D. S. et al. (2024). The IBdata web system for biological collections: design focused on usability. Biodiversity Informatics, 18, 1–12. https://doi.org/10.17161/bi.v18i.20516

Nelson, G., Sweeney, P., Wallace, L. E., Rabeler, R. K., Allard, D., Brown, H. et al. (2015). Digitization workflows for flat sheets and packets of plants, algae, and fungi. Applications in Plant Sciences, 3, 1500065. https://doi.org/10.3732/apps.1500065

Pavlick, G. (2025). What is Generative AI (GenAI)? How does it work? Oracle. Accessed June 6, 2025: https://www.oracle.com/uk/artificial-intelligence/generative-ai/what-is-generative-ai/

Schank, R. C., & Abelson, R. P. (1975). Scripts, plans, and knowledge. In IJCAI’75: Proceedings of the 4th International Joint Conference on Artificial Intelligence, Volume 1 (pp. 151–157). San Francisco, CA: Morgan Kaufmann Publishers Inc. https://dl.acm.org/doi/10.5555/1624626.1624649

Schank, R. C., & Abelson, R. P. (1977). Scripts, plans, goals, and understanding: an inquiry into human knowledge structures. Hillsdale, New Jersey: Lawrence Eribaum Associates Inc. Publishers.

Shanahan, M. (2024). Talking about large language models. Communications of the ACM, 67, 68–79. https://doi.org/10.1145/3624724

Soltis, P. S., Nelson, G., & James, S. A. (2018). Green digitization: online botanical collections data answering real-world questions. Applications in Plant Sciences, 6, e1028. https://doi.org/10.1002/aps3.1028

Takano, A., Cole, T. C., & Konagai, H. (2024). A novel automated label data extraction and database generation system from herbarium specimen images using OCR and NER. Scientific Reports, 14, 112. https://doi.org/10.1038/s41598-023-50179-0

Las montañas del centro de México: un enigma fitogeográfico resuelto mediante análisis de similitud florística

José Luis Villaseñor *, Enrique Ortiz

Universidad Nacional Autónoma de México, Instituto de Biología, Departamento de Botánica, Apartado postal 70-233, 04510 Ciudad de México, Mexico

*Corresponding author: vrios@ib.unam.mx (J.L. Villaseñor)

Received: 07 October 2025; accepted: 29 March 2026

Abstract

Approximately 24.9% of the national territory is classified as mountainous, characterized by high floristic diversity and by the challenges in its biogeographic regionalization. The objective of this study was to determine whether the mountain ranges of central Mexico constitute a single phytogeographic province or comprise independent provinces. To resolve this enigma, a quantitative floristic analysis was conducted. Floristic composition was determined in 51 grid cells, each measuring 1° in latitude and longitude; subsequently, a floristic similarity analysis was performed using the β-Simpson index, followed by a cluster analysis to regionalize the study area. A total of 10,431 native species were documented in the mountains, representing 44.5% of Mexico’s entire flora. Of these, 3,723 species are exclusive to temperate regions, while 5,562 are also distributed in lowland tropical zones. Furthermore, 7,024 species endemic to the mountains were recorded. In resolving the enigma, 5 floristic provinces were identified, each supported by a significant number of endemic species, thereby demonstrating their distinct identity as biogeographic regions. The analysis presented here lends quantitative rigor to the traditionally qualitative debate regarding the regionalization of central Mexico’s mountains, thereby enabling an explicit comparison with previous proposals.

Keywords: Biogeography; Endemism; Floristics; Regionalization; Richness; Vascular plants

Resumen

Aproximadamente 24.9% del territorio nacional se clasifica como montañoso, se destaca por su gran diversidad florística y por los desafíos de su regionalización biogeográfica. El enfoque de este trabajo fue determinar si las sierras del centro de México forman una sola provincia fitogeográfica o constituyen provincias independientes. Para resolver este enigma se realizó un análisis florístico cuantitativo. Se determinó la composición florística en 51 celdas de 1° de latitud y longitud, y se realizó un análisis de similitud florística utilizando el índice β-Simpson para, posteriormente, regionalizar el área de estudio mediante un análisis de agrupamiento. Se documentaron 10,431 especies nativas en las montañas, lo que representa 44.5% de la flora total de México. De ellas, 3,723 especies son exclusivas de las regiones templadas, mientras que 5,562 también se distribuyen en zonas tropicales bajas. Se registraron 7,024 especies endémicas de las montañas. En la resolución del enigma se identificaron 5 provincias florísticas, cada una respaldada por un número significativo de especies endémicas, lo que demuestra su identidad como regiones biogeográficas. El análisis presentado aporta solidez cuantitativa al debate sobre la regionalización de las montañas del centro de México, tradicionalmente de carácter cualitativo, lo que permite contrastar explícitamente las propuestas previas.

Palabras clave: Biogeografía; Endemismo; Florística; Regionalización; Riqueza; Plantas vasculares

Introduction

A mountain is any orographic feature that stands out from the plain, is surrounded by valleys, and is characterized by complex relief and steep slopes (Hoorn et al., 2018). This relief and slope can range from near sea level to high elevations with snowy peaks (Fig. 1). It is challenging to determine which parts of this extent can be appropriately called mountains; often, we refer to “mountains” only when considering the temperate forests typical of regions we call mountainous, which in Mexico are known as “sierras”. Mountains influence the organization of vegetation zones along altitudinal gradients, with each gradient displaying unique sets of species that are adapted to it (see, for example, Hernández-Rojas et al., 2020; Jiménez-López et al., 2020; Karger et al., 2011; McCain & Grytnes, 2010; Salas-Morales & Meave, 2012). However, the continuous nature of altitudinal gradients makes it difficult to define precise boundaries between mountains and lowlands.

One potential defining criterion could be elevation, but given this wide altitudinal gradient, how high should we consider a mountain? It is clear that, in general, elevation per se is not a reliable criterion for characterizing a mountainous area. For instance, in the state of Chiapas, tropical forests, which are rarely considered a mountain vegetation type, can occur at higher elevations than the same type of forest in Sinaloa or Sonora. Thus, although elevation is an informative parameter, it is not sufficient as a sole determinant of a mountain’s definition.

Rzedowski (1978) indicates that the tropical deciduous forest has the widest elevational distribution among tropical plant communities in Mexico, extending from sea level to 1,900 m, although it is more frequent below 1,500 m. A recent study of this forest type reports that its average upper elevation in the Balsas River Basin is 1,600 m (Flores-Tolentino et al., 2023). Thus, this is the elevation at which, on average, this vegetation type gives way to temperate forests, such as oak or pine forests, providing a general elevational threshold between mountain and lowland tropical forests.

Perhaps a more rigorous way to define the boundaries of a mountain is to consider the specific relationships between its aspect or topography and altitude. One example is the roughness index, which quantifies the altitudinal difference in the terrain. For example, Körner et al. (2011) classify any terrain with a roughness index greater than 200 as a mountain. Figure 2A shows the roughness index values across Mexico, with white areas indicating values below 200, thereby categorizing them as flat regions. Figure 2B segregates all localities in the country with a roughness index above 200 into those above and below 1,600 m elevation, which marks the boundary between tropical plant communities typical of lowlands and temperate communities found at higher elevations (typically montane). Based on the roughness criterion (Fig. 2), 488,592.3 km² of Mexico’s area would be classified as mountains, representing 24.9% of the national territory. Of this area, 299,762.4 km² fall below 1,600 m elevation, while 188,829.9 km² are above that elevation.

Figure 3 illustrates the distribution of mountainous areas across Mexico’s biogeographic provinces (Conabio, 1997), highlighting regions located at elevations below or above 1,600 m. As noted, not all biogeographic provinces identified as mountain ranges (Sierra Madre Occidental, Sierra Madre Oriental, Sierra Madre del Sur, or Trans-Mexican Volcanic Belt) consist solely of mountainous areas. Most of the land classified as mountainous in Mexico consists of flat regions (valleys or plateaus) that do not genuinely qualify as mountainous territory. These flat areas, characterized by a roughness index of less than 200, are interspersed with sections classified as mountains, which have a higher roughness index. Therefore, it is not surprising that floristic elements with varying adaptive strategies and ecological and biogeographic affinities are found intermixed among valleys and slopes throughout a mountain range (Fig. 1).

The Trans-Mexican Volcanic Belt (TMVB, known in Spanish as the “Eje Volcánico Transversal”) is considered one of the most significant mountain ranges in central Mexico (Luna et al., 2008). Alongside the TMVB, several other crucial mountain ranges converge in this area: from the west, the Sierra Madre Occidental (SMOC, González-Elizondo et al., 2012) and the Sierra Madre del Sur (SMS, Luna-Vega et al., 2016); and from the east, the Sierra Madre Oriental (SMOR, Suárez-Mota et al., 2017; Salinas-Rodríguez et al., 2022; Villaseñor & Ortiz, 2022; Villaseñor et al., 2023) and the mountains of northern Oaxaca (OAX, Lorence et al., 2009). Each of these ranges possesses distinct floristic richness (see cited references) and many endemic species confined to their territory. Still, many species are shared across ranges, creating challenges in defining the biogeographical boundaries of each range.

Rzedowski (1978) emphasized the difficulties in defining the mountain ranges of central Mexico in his biogeographic regionalization. Consequently, he grouped them into a single unit, the Serranías Meridionales Floristic Province, which includes the TMVB, the Sierra Madre del Sur (SMS), and the complex montane system of Oaxaca (OAX). More recently, authors such as Conabio (1997) and Morrone (2019) have regarded each mountain range as an independent biogeographic province. However, there has not yet been a formal evaluation of which of these biogeographic proposals is more suitable for assessing diversity patterns in this region, a region of significant floristic richness and endemism. We therefore aimed to test 2 hypotheses in this context: First, based on floristic composition and number of endemisms, do the mountain ranges of central Mexico form a single phytogeographic province (Rzedowski hypothesis) or constitute independent provinces (Conabio or Morrone hypotheses)? Secondly, where are the geographic boundaries of the floristic units (districts, if a single province is accepted, or provinces, if each range is a separate province)?

To test these hypotheses, we defined the ecological affinities and geographic distribution of the species recorded throughout the central Mexican mountain ranges. The ecological affinities were determined by the degree of conservatism observed among species of the same genus in tropical or temperate environments, as well as by distribution, measured by the number of geographic units (OGUs) in which each species is recorded. We predict that each mountain range is characterized by several endemic species, as well as a set of species adapted primarily to temperate affinity.

Materials and methods

The first step was to determine the floristic composition of the central Mexican mountain ranges. For this purpose, we used shapefile maps of the biotic and biogeographic regions proposed by Ferrusquía-Villafranca (1990), Rzedowski and Reyna-Trujillo (1990), Conabio (1997), and Morrone et al. (2017). From each map, we delineated a polygon and subdivided it into 1° latitude-by-longitude squares, numbered according to a national schedule (see, for example, Cruz-Cárdenas et al., 2013, or Villaseñor et al., 2024) (Fig. 4). The various proposals regarding the mountains of the central region place their point of contact with the SMOR and SMOC at differing latitudes; therefore, to delineate the boundaries between the distinct mountain ranges within our analysis, we have slightly expanded the limits of each chain (TMVB, SMS, OAX) to encompass the southern limits of the SMOC to the west and the SMOR to the east. This approach resulted in 51 grid cells being used in the following analyses (Fig. 4).

We determined the total number of native plant species (excluding introduced or exotic species) in each grid cell, enabling us to calculate alpha diversity per cell. Floristic information was compiled and entered into a database using the same strategy as in previous studies (Villaseñor & Ortiz, 2025). Introduced or exotic species were excluded at this stage, although most have already become naturalized and form part of the biodiversity of the mountain range ecosystems.

Once we had compiled the checklist of species in the mountainous regions of central Mexico, we compared the total number of species in the country with the number present in the mountain ranges for each genus. The ecological fidelity of the genus to tropical or temperate environments was assessed by examining the differences in species proportions within each genus across both environments nationwide. When more than 50% of species were recorded in tropical lowlands (Majority method; Murphy et al., 2019), we classified the genus as having a tropical affinity. In comparison, if more than 50% of species within the genus were found in mountainous regions, this indicated a temperate affinity. For instance, the genus Ageratina comprises 169 species, of which 154 (91.1%) are recorded from montane environments, indicating a temperate affinity. In contrast, the genus Bursera comprises 93 species in the country, of which 91 (97.8%) are recorded in the tropical lowlands, indicating a tropical affinity.

We used clustering analysis at the grid cell level to identify potential floristic regions (assemblages, floristic units, or phytochoria) and the main floristic elements (species) that characterize each floristic unit (Kreft & Jetz, 2010). To achieve this, the region was divided into 51 grid cells. We then used the species records to construct an incidence matrix for each grid cell and generated a β-Simpson dissimilarity matrix across grid cells. We used this index because it is not sensitive to differences in species richness among grid cells (Tuomisto, 2010). We then used the dissimilarity matrix to perform agglomerative clustering with the WPGMA method (weighted pair-group method using arithmetic means), which weights cluster contributions by the number of cells they contain, ensuring that each cell contributes equally to each agglomerate (González-Orozco et al., 2013). A threshold of 7 groups was determined to form phytochoria, corresponding to the regionalization of the study region proposed by Rzedowski and Reyna-Trujillo (1990; Fig. 4B). The clustering analysis was performed separately for the total flora and the tropical and temperate elements. Finally, we mapped and explored the floristic dissimilarities and identified agglomerates to identify grid cells with similar species assemblages. These analyses were performed in Biodiverse 4.3 (Laffan et al., 2010). We performed an Analysis of Similarities (ANOSIM) to assess whether the groups identified in the cluster analysis differ significantly in species composition. ANOSIM is a nonparametric technique that allows statistical comparisons between and within groups and tests the significance of differences among 2 or more groups of sampling units (Clarke & Green, 1988). The ANOSIM statistic “R” compares the mean of ranked dissimilarities between groups to the mean of ranked dissimilarities within groups. R varies from +1 to -1. R values greater than 0 indicate a higher dissimilarity between groups than within groups; R values equaling zero indicate an equal level of between-group and within-group average dissimilarity. Negative values of R indicate that dissimilarities within groups are greater than dissimilarities between groups (Clarke & Green, 1988). Chiarucci et al. (2019) classify R values as follows: sharp separability (R > 0.75); good separability (0.5 < R ≤ 0.75); and separated but overlapping (0.25 < R ≤ 0.5). These analyses were carried out in the R environment (R Core Team, 2025) with the vegan package (Oksanen et al., 2025).

Results

Taxonomic dispersion of the mountain flora of central Mexico. The total floristic richness of Mexico’s mountain ranges was 14,068 species (Villaseñor, unpublished data); this count excludes 490 known introduced or exotic species, most of which have become naturalized. The mountains of central Mexico contained records of 10,431 species, distributed among 1,816 genera and 256 families (Table 1), representing 74% of the country’s total native montane flora. Of these species, 3,723 are known to be distributed exclusively in the temperate regions of these mountain ranges (generally above 1,600 m elevation). At the same time, 5,562 are also distributed toward the lower tropical parts of the mountains. Asterids and Rosids account for most of this diversity, and Monocots are the third-most-represented group.

	Families	Genera	Species
Ferns and monilophytes	33	113	629
Gymnosperms	6	12	98
Basal angiosperms	2	3	3
Magnoliids	11	30	346
Monocots	33	362	2,273
Eudicots	10	27	91
Rosids	84	523	2,759
Asterids	77	746	4,232
Total	256	1,816	10,431

Species richness throughout the central Mexican mountains. Table 2 presents the species richness in each state in central Mexico, indicating the number of species exclusive to temperate regions and those whose distribution also includes lowlands with more tropical environments. The supplementary material comprises the species that constitute the exclusive mountain flora of central Mexico (3,723 species), indicating the states and grid cells where they are recorded. The state with the fewest number of species was Tlaxcala (991), and the richest was Oaxaca (6,846) (Table 2). The median number of species per state was 3,690. Similar richness patterns are observed among species typically found in highlands, with notable differences when considering widespread species (Table 2).

State	Total species	Exclusive highlands	Distributed in tropical lowlands
Colima (COL)	1,593	171	1,374
Guanajuato (GTO)	2,194	368	1,242
Guerrero (GRO)	4,707	1,103	3,255
Hidalgo (HGO)	3,690	771	2,176
Jalisco (JAL)	4,929	1,320	3,054
Mexico (MEX)	3,837	956	2,370
Mexico City (CDMX)	1,344	347	660
Michoacán (MICH)	4,180	1,022	2,719
Morelos (MOR)	2,534	474	1,780
Nayarit (NAY)	2,763	565	2,008
Oaxaca (OAX)	6,846	1,715	4,543
Puebla (PUE)	4,574	869	3,062
Queretaro (QRO)	3,096	535	1,856
Tlaxcala (TLAX)	991	236	455
Veracruz (VER)	5,702	1,155	3,870
Total	10,431	3,723	5,562

Table S1, in the Supplementary material, presents the species by grid cell for the mountainous region under study. The highest values occurred in grid cell 228, located in Oaxaca, and grid cell 147, on the border between Nayarit and Sinaloa, had the lowest richness. The median number of species per grid cell was 1,532 species.

Figure 5A illustrates the alpha diversity patterns across the central Mexican mountain ranges. There are major richness centers in the west and east of the region, particularly along the TMVB. Another significant grid cell with exceptionally high richness is in the OAX range, in northern central Oaxaca. There is another prominent grid cell in eastern-central Mexico, at the convergence zone between the SMOR and the TMVB.

Species ecological affinities. When comparing the total number of species recorded in the country (Villaseñor, 2016, and updates not published since then) with the number in the mountains under study, it is found that of the 10,431 species recorded, only 3,723 were exclusively distributed in the highlands (temperate) (Tables 2, 3). Of the 1,815 genera, 457 had temperate affinity, and 108 had tropical affinity, while the remaining 1,250 genera included species whose ecological range did not show precise fidelity to either environment. Table S2 (Supplementary material) presents examples of genera that primarily exhibited temperate or tropical affinity.

Some of the characteristic genera of temperate mountain regions are Salvia L. (330 species in all of Mexico, 176 found in temperate highlands, and 52 in tropical lowlands), Tillandsia L. (232, 129, and 69, respectively), Ageratina Spach (169, 126, 69), Quercus L. (159, 98, 28), and Stevia Cav. (116, 90, 21). On the other hand, Tetramerium Nees (26 species in all of Mexico, 3 found in temperate highlands, and 25 in tropical lowlands), Desmopsis Saff. (22, 3, and 21, respectively), Bourreria P. Browne (17, 2, 17), Pouteria Aubl. (14, 3, 14), or Mortoniodendron Standl. & Steyerm. (8, 1, 8) are examples of genera with more species found in tropical than temperate environments, despite having some highland species.

Floristic similarities. Figure 5 shows alternative interpretations of the regionalization of the mountain regions in central Mexico. In addition to the mountain ranges that were the focus of this study (TMVB, SMS, and OAX), we show some contiguous provinces. Figure 5B considers the complete temperate flora of central Mexico (N = 10,431 species), while Figure 5C only includes the set of species known exclusively from highland environments (above 1,600 m elevation) of the mountains (N = 3,723 species). Figure 5D includes only species that are considered tropical due to their distribution extending to the lower parts of the mountains, primarily in their most tropical and thermophilic environments (N = 5,562 species). ANOSIM indicated that the species composition among the groups identified in the cluster analysis differed sharply and significantly. For all species, R = 0.7621, p < 0.001; for exclusively temperate species, R = 0.7624, p < 0.001; for exclusively tropical species, R = 0.7606, p < 0.001.

The groupings shown in Figure 5 align geographically with the previously defined provinces (Fig. 4). However, the mountainous regions become more evident when only the temperate flora is evaluated (Fig. 5C). In this case, the clustered grid cells define 5 phytogeographic regions, known as phytochoria. One is located towards the northwest (group 1), which characterizes the southern portion of the SMOC. The central one (group 3) defines the core of the TMVB, which converges towards the northeast with the SMOR (group 2). To the east, there is a sector corresponding to the Mexican Gulf slope formed by the eastern end of the TMVB and the OAX mountains (group 4). Finally, group 5 corresponds to the SMS. Four cells in the southeastern part of the study area did not group with the 5 phytochoria mentioned. Their isolation results from the small number of temperate-affinity species recorded in their territory (Supplementary material: Table S1).

Considering only species primarily distributed in the lowlands of the mountains (with a tropical affinity) yielded 7 clusters (Fig. 5D). One of these is the western region (group 1), which, in Figure 5C, is identified as the SMOC. This area now forms a continuous region that encompasses the SMOC, the southern part of the TMVB (group 3 in 5C), and a large portion of the SMS. Group 3 in this scheme corresponds to the SMOR (group 2 in 5B, C). Group 4 characterizes the lowlands of the Gulf of Mexico slope, and group 5 constitutes the core TMVB, sharing elements with the drier sectors of the southern Mexican highlands (ALTS, Fig. 3). The fifth group (group 6, Fig. 5D) encompasses a significant region of south Puebla and northern Oaxaca, known for its rich flora and numerous endemisms, specifically the Tehuacán-Cuicatlán Valley. Grid cell 240 is linked to this region (group 6) due to a small area with semiarid environments that recent explorations indicate have a strong floristic relationship with those to the north. Group 7 (Fig. 5D), located in southern Oaxaca, is considered an extension of the coastal grid cells of group 1, which characterize the Pacific floristic province (PACIF, Fig. 3). Grid cells 224 and 239 can be regarded as outliers of this continuous tropical lowland region.

Table 3 shows the floristic richness of each of the 5 floristic units (phytochoria) of the temperate mountainous region in central Mexico. Phytochoria 3 (TMVB), 4 (SE Mexico), and 5 (SMS) had higher species richness than Phytochoria 1 and 2 (SMOC and SMOR, respectively).

Floristic group (Phytochorion)	Species	Endemic to Mexico
1 (SMOC)	1,244 (494)	980 (452)
2 (SMOR)	908 (208)	597 (184)
3 (TMVB)	1,614 (382)	1,182 (346)
4 (SE Mexico)	1,385 (830)	830 (283)
5 (SMS)	1,373 (558)	976 (480)

Discussion

The 10,431 species documented as the flora of the mountainous regions of central Mexico represent 44.5% of the 23,412 species reported in the latest assessment of Mexico’s complete flora (Villaseñor & Ortiz, 2025). This high percentage is undoubtedly essential, but it is more interesting that only one-third of this mountain flora is restricted to the highlands of the Sierras (Table 2). These values are undoubtedly explained by the small proportion of territory occupied by the mountain highlands (38.6% of the total mountain surface, as defined by a roughness index greater than 200).

The most represented groups in the mountainous flora of central Mexico are the Monocots, Rosids, and Asterids (Table 1), which corresponds with richness patterns at both the national and regional levels (Villaseñor, 2016; Villaseñor et al., 2023; Jiménez-López et al., 2023). These 3 groups were also dominant in species richness in evergreen tropical forests, which are characteristic of the country’s low, humid tropical regions (Villaseñor et al., 2025).

We recorded 7,024 Mexican endemic species in the central Mexican mountains. This represents 67.3% of total species richness, which is higher than the national average endemism rate (Villaseñor, 2016). Such a high level of endemism is unknown in any other region or ecosystem in the country. Similar levels of endemism have been observed in the country’s high-diversity taxonomic groups, such as Asteraceae (Villaseñor, 2018).

Interestingly, of the 3,723 species known to be exclusive to highland areas, 1,180 are endemic to Mexico. In addition, 978 of these endemic species are known only from such highlands. This low number of endemism restricted to the highlands suggests a wider species distribution, likely due to historical factors, as discussed by Mastretta-Yanes et al. (2015) for the TMVB.

Species ecological affinities. Our results show that genera considered typical of mountain environments, because they are distributed in areas with high roughness index values, are not necessarily restricted to temperate, montane-type environments; instead, they may include a significant number of species distributed in lower-elevation tropical environments. Likewise, in any grid cell in central Mexico, the number of montane species distributed across both the high and lower parts of the mountain ranges is as essential as that of species known only from the highest, most temperate parts (Table 2).

As mentioned above, only 35.7% of the total richness documented in the montane forests of central Mexico is restricted to the high-mountainous part; 53.3% of the species are distributed over a wider gradient, also occupying the lower parts of the mountains, where forests with tropical affinities predominate. These results should not surprise us, as indicated above, since the territory considered mountainous in Mexico, based on the roughness index criterion (which represents 24.9% of the national territory), is predominantly located below 1,600 m, where vegetation with a tropical affinity is commonly observed.

The ecological breadth of many genera (and their species) does not support the hypothesis of environmental conservatism, which posits stricter fidelity of descendants to their ancestors’ environments. Among the mountain flora studied, only 457 of 1,815 genera (25.2%) support this hypothesis, as most of their species exhibit an ecological preference for mountain/temperate environments. Similarly, on the lower slopes of the mountains, only 108 genera (5.9%) had a strictly tropical environmental affinity. The species of the other genera were relatively evenly divided between temperate and tropical preferences, thus showing no apparent genus-level affinity. A few studies have evaluated whether species’ distributions are restricted to temperate or tropical environments or span the elevational gradient (Hernández-Rojas et al., 2020; Jiménez-López et al., 2020; Salas-Morales & Meave, 2012). Further studies of this kind are needed to deepen our understanding of ecological conservatism among species.

Floristic similarities. The 5 groups identified for temperate species (Fig. 5C) align with well-established previous circumscriptions. For example, group 1 (in the northwest) corresponds with the southern edge of the SMOC and the piedmonts of the Pacific coast (PACIF) provinces. Group 2 corresponds with the southern part of the SMOR, along with the south edge of the Southern Highland (Altiplano Sur). Group 3 corresponds to the core of the TMVB, including lowlands oriented towards the Pacific and the Balsas (BALS) provinces. BALS is a tropical region that disjoints the TMVB and SMS (Flores-Tolentino et al., 2023; Villaseñor et al., 2021). Group 4 encompasses the easternmost edge of the TMVB, the southernmost part of SMOR, and the mountain system of northern Oaxaca (OAX). Finally, group 5 corresponds to the SMS, which begins in Guerrero and extends into central and southern Oaxaca (Fig. 5C). There are 4 grid cells considered to be outliers; cell 248 is easily assigned to the SMS, representing the lowlands of the Sierra oriented towards the PACIF province. The other 3 grid cells (230, 241, and 242) are located primarily in the lowland, where the elevation of the SMS and OAX decreases toward the characteristic lowlands of the Isthmus of Tehuantepec, where the lowland tropical provinces merge (Gulf of Mexico (GOLF) and Pacific (PACIF)). These cells likely represent part of the transition zone between the mountains of central Mexico and the mountain systems of Chiapas.

Our results do not support extending the delimitation of the SMS beyond the borders of Guerrero; the parts considered by Morrone (2019) as disjoint portions of this province in Michoacán and Jalisco are more floristically similar to the TMVB (Fig. 5C).

Biogeographic implications

Floristic regions are part of the Earth’s geography. They share a similar floristic composition, as defined by the number and distribution of species, genera, and families. Phytogeographic categories are determined using a hierarchical classification; for example, Takhtajan (1986) considers the kingdom, region, province, and district as the primary categories. Rzedowski (1978) used several of these to regionalize Mexico.

The region level is characterized by the occurrence of endemic genera and a high proportion of endemic species. Meanwhile, the province level exhibits less characteristic genus-level endemism, and the endemic genera present may include few or even a single representative species (Takhtajan, 1986). However, the most crucial factor in defining this province level is the floristic similarity among the geographical units (OGUs) that constitute it, which can be determined quantitatively (for example, using clustering methods) or statistically.

In the mountain ranges studied (Figs. 4, 5), 43 Mexican endemic genera are recorded. However, as with many species (Supplementary material: Table S2), several are not exclusive to temperate mountain regions but are also distributed in other biomes characteristic of the lower parts of the mountains. They may even extend into non-mountainous neighboring areas of adjacent floristic provinces that occupy sites within the grid cells included in the study area.

In phytochorion 1 (SMOC), no genus was restricted to the territory, as this part corresponds to the southern end of a floristic province that is better represented in the northwestern part of Mexico. Examples of endemic genera that are more widely distributed in the SMOC and whose southern limits fall within the study area include Bolanosa A. Gray and Pippenalia McVaugh. Phytochorion 2 (SMOR) contained the endemic genus Velascoa Calderón & Rzed., as well as 4 other near-endemic genera that are shared with the Southern Altiplano province and are found in more xeric environments (scrublands and grasslands): Chichimecactus Bárcenas, H.M. Hern. & P. Hern.-Led., Kroenleinia Lodé, Pseudonemacladus McVaugh, and Strombocactus Britton & Rose. In phytochorion 3 (TMVB), there was no typical genus from its temperate regions. However, a genus that occupies lowland mountainous areas, bordering the more tropical parts of the Balsas River Basin province, is Beiselia Forman. Phytochorion 4 (SE Mexico) is characterized by the genera Habroneuron Standl., Placocarpa Hook. f., and Stramentopappus H. Rob. & V.A. Funk, as well as Jeronimoa A. Vázquez, Islas & Rosales, which is also distributed towards more tropical and semi-arid areas. Phytochorion 5 (SMS) contains 2 endemic genera (Dahliaphyllum Constance & Breedlove and Eremitilla Yatsk. & J.L. Contr.). Five additional genera are documented at the foothills of the mountains, sharing temperate environments but more commonly found in more tropical ones towards the interior of the continent (Balsas River province) and on the Pacific Ocean slope (Pacific floristic province): Anotea (DC.) Kunth, Mexipedium V.A. Albert & M.W. Chase, Mezcala C.E. Hughes & J.L. Contr., Mixtecalia Redonda-Mart., García-Mend. & D. Sandoval, Petronymphe H.E. Moore, and Wimmeranthus Rzed.

To date, the genera mentioned above are known only from a single phytochorion; however, we have documented 23 additional endemic genera whose distributions are shared between 2 or more floristic groups. For example, the genus Davilanthus E.E. Schill. & Panero is documented from 4 phytochoria (TMVB, SE Mexico, SMOR, and SMS). Another 13 genera are known from 3 floristic groups, while 9 more occur in the territory shared by 2 phytochoria. In addition to the restricted genera, each of the 5 phytochoria identified in this study contains a significant number of characteristic species (Table 3). More importantly, each of them includes a substantial number of Mexican endemic species, many of which have restricted distributions within the phytochorion.

The large number of endemic genera restricted to this central mountainous region of Mexico supports Rzedowski’s (1978) hypothesis of treating it as a single phytogeographic unit (the Serranías Meridionales Province). However, each phytochorion, in addition to having one or more distinctive genera, includes significant numbers of restricted endemic species. Therefore, following Takhtajan’s (1986) proposal, each of them should be treated as a floristic province. There is no defined criterion for determining how many endemic species are required to establish a regionalization at the provincial level. Morrone (2019), for example, supports his regionalization with a few examples but does not indicate the total number of species characterizing each of his provinces. In this work, Table 3 indicates that each phytogeographic province contains more than 100 species restricted to its territory, supporting the idea that each should be treated as a distinct floristic province.

The 5 floristic provinces recovered in this study for the mountain system in central Mexico (Table 3) align most closely with the Conabio (1997) proposal. However, there are some differences in how parts of the SMS in Jalisco and Michoacán are classified as TMVB members. Similarly, Morrone’s (2019) regionalization includes the Sierras del Tuito and Manantlán in Jalisco (the Districts Jaliscience and Jalisco-Manantlán, respectively), as well as the Sierra de Coalcomán (District of Michoacán) in Michoacán, all of which form part of the SMS. In this study, the floristic data support classifying these districts as part of the TMVB rather than the SMS. Their placement within the TMVB is also consistent with the proposed geological evolution of the TMVB discussed by Ferrari et al. (2012).

The separation of species based on ecological preferences provides a more rigorous method for identifying floristic units in biogeographic regionalization. Most of the time, species with broad ecological tolerances are used as examples for particular environments, such as the temperate forest evaluated in this case. We recommend considering the species’ ecological affinities when evaluating further biogeographical proposals in the heterogeneous Mexican landscape.

The mountainous region of central Mexico is complex and difficult to characterize. Its orogenesis and the evolution of its biota have not been fully clarified, which, as noted above, has led to different biogeographical proposals. Using the floristic composition, we identified 5 floristic provinces, some of which have also been analyzed by other approaches as independent units, similar to those found here (Aragón-Parada et al., 2021, 2023; Suárez-Mota et al., 2013; Villaseñor et al., 2021).

We hope that analyses like the one presented here will motivate continued accumulation of data and analysis that will provide ever clearer insight into the origins of the rich biodiversity found in this important region of Mexico. However, equally important, this information enables us to clarify the limits of the floristic units that constitute the southern boundary of the significant Mexican Transition Zone (Villaseñor et al., 2020).

Acknowledgments

We thank Socorro González-Elizondo and Miguel Murguía-Romero for their helpful comments. We appreciate Lynna M. Kiere for reviewing the English version and her thoughtful comments.

References

Aragón-Parada, J., Rodríguez, A., Munguía-Lino, G., De-Nova, J. A., Salinas-Rodríguez, M. M., & Carrillo-Reyes, P. (2021). Las plantas vasculares endémicas de la Sierra Madre del Sur, México. Botanical Sciences, 99, 643–660. https://doi.org/10.17129/botsci.2682

Aragón-Parada, J., Carrillo-Reyes, P., Rodríguez, A., Munguía-Lino, G., Salinas-Rodríguez, M. M., & De-Nova, J. A. (2023). Spatial phylogenetics of the flora in the Sierra Madre del Sur, Mexico: Evolutionary puzzles in tropical mountains. Journal of Biogeography, 50, 1679–1691. https://doi.org/10.1111/jbi.14679

Chiarucci, A., Nascimbene, J., Campetella, G., Chelli, S., Dainese, M., Giorgini, D. et al. (2019). Exploring patterns of beta-diversity to test the consistency of biogeographical boundaries: a case study across forest plant communities of Italy. Ecology and Evolution, 9, 11716–11723. https://doi.org/10.1002/ece3.5669

Clarke, K. R. & Green, R. H. (1988). Statistical design and analysis for a ‘biological effects’ study. Marine Ecology Progress Series, 46, 213–226. https://doi.org/10.3354/meps046213

Conabio (Comisión Nacional para el Conocimiento y Uso de la Biodiversidad). (1997). Provincias biogeográficas de México, escala 1:4000000. México. Accessed January 15, 2025 from: http://www.conabio.gob.mx/informacion/gis/layouts/rbiog4mgw.png

Cruz-Cárdenas, G., Villaseñor, J. L., López-Mata, L., & Ortiz, E. (2013). Distribución espacial de la riqueza de especies de plantas vasculares en México. Revista Mexicana de Biodiversidad, 84, 1189–1199. https://doi.org/10.7550/rmb.31811

Ferrari, L., Orozco-Esquivel, T., Manea, V., & Manea, M. (2012). The dynamic history of the Trans-Mexican Volcanic Belt and the Mexico subduction zone. Tectonophysics, 522–523, 122–149. https://doi.org/10.1016/j.tecto.2011.09.018

Ferrusquía-Villafranca, I. (1990). Provincias bióticas (con énfasis en criterios morfotectónicos). Escala 1:4 000 000. In Regionalización biogeográfica. Atlas Nacional de México. Vol. II. Instituto de Geografía, UNAM. México. Accessed January 15, 2025 from: http://www.conabio.gob.mx/informacion/gis/?vns=gis_root/region/fisica/rmorf4mgw

Flores-Tolentino, M., Ramírez-Rodríguez, J. R., Morales-Linares, J., Ibarra-Manríquez, G., Dorado, O., & Villaseñor, J. L. (2023). Delimitación geográfica y florística de la provincia fisiográfica de la Depresión del Balsas, México, con énfasis en el bosque tropical estacionalmente seco. Revista Mexicana de Biodiversidad, 94, e944985. https://doi.org/10.22201/ib.20078706e.2023.94.4985

González-Elizondo, M. S., González-Elizondo, M., Tena-Flores, J. A., Ruacho-González, L., & López-Enríquez, I. L. (2012). Vegetación de la Sierra Madre Occidental, México: una síntesis. Acta Botanica Mexicana, 100, 351–403. https://doi.org/10.21829/abm100.2012.40

González-Orozco, C. E., Laffan, S. W., Knerr, N., & Miller, J. T. (2013). A biogeographical regionalization of Australian Acacia species. Journal of Biogeography, 40, 2156–2166. https://doi.org/10.1111/jbi.12153

Hernández-Rojas, A. C., Kluge, J., Krömer, T., Carvajal-Hernández, C., Silva-Mijangos, L., Miehe, G. et al. (2020). Latitudinal patterns of species richness and range size of ferns along elevational gradients at the transition from tropics to subtropics. Journal of Biogeography, 47, 1383–1397. https://doi.org/10.1111/jbi.13841

Hoorn, C., Perrigo, A., & Antonelli, A. (2018). Mountains, climate, and biodiversity: an introduction. In C. Hoorn, A. Perrigo, & A. Antonelli (Eds.), Mountains, climate and biodiversity (pp. 1–13). Newark: Wiley-Blackwell.

Jiménez-López, D. A., Martínez-Camilo, R., Martínez-Meléndez, N., & Kessler, M. (2020). Diversity of epiphyte ferns along an elevational gradient in El Triunfo Biosphere Reserve, southern Mexico. Plant Ecology and Evolution, 153, 12–21. https://doi.org/10.5091/plecevo.2020.1759

Jiménez-López, D. A., Gallardo-Cruz, J. A., Véliz, M. E., Martínez-Camilo, R., Méndez, C., Solórzano, J. V. et al. (2023). High vascular plant species richness in the Usumacinta River Basin: a comprehensive floristic checklist for a natural region in the Mesoamerican biodiversity hotspot. Botanical Sciences, 101, 908–930. https://doi.org/10.17129/botsci.3253

Karger, D. N., Kluge, J., Krömer, T., Hemp, A., Lehnert, M., & Kessler, M. (2011). The effect of area on local and regional elevational patterns of species richness. Journal of Biogeography, 38, 1177–1185. https://doi.org/10.1111/j.1365-2699.2010.02468.x

Körner, C., Spehn, E. M., & Paulsen, J. (2011). A definition of mountains and their bioclimatic belts for global comparisons of biodiversity data. Alpine Botany, 121, 73–78. https://doi.org/10.1007/s00035-011-0094-4

Kreft, H., & Jetz, W. (2010). A framework for delineating biogeographical regions based on species distributions. Journal of Biogeography, 37, 2029–2053. https://doi.org/10.1111/j.1365-2699.2010.02375.x

Laffan, S. W., Lubarsky, E., & Rosauer, D. F. (2010). Biodiverse, a tool for the spatial analysis of biological and related diversity. Ecography, 33, 643–647. https://doi.org/10.1111/j.1600-0587.2010.06237.x

Lorence, D., Ramírez-de Anda, M., Torres-Colín, R., & Villa-Arce, R. (2009). Listados florísticos de México. XXV. Flora de la Sierra de Juárez, Oaxaca: Distrito de Ixtlán y áreas adyacentes (Sierra Norte de Oaxaca). Ciudad de México: Universidad Nacional Autónoma de México.

Luna-Vega, I., Espinosa, D., & Contreras-Medina, R. (2016). Biodiversidad de la Sierra Madre del Sur, una síntesis preliminar. Ciudad de México: Universidad Nacional Autónoma de México.

Luna, I., Morrone, J. J., & Espinosa, D. (2008). Biodiversidad de la Faja Volcánica Transmexicana. México D.F.: Universidad Nacional Autónoma de México/ Comisión Nacional para el Conocimiento y Uso de la Biodiversidad.

Mastretta-Yanes, A., Moreno-Letelier, A., Piñero, D., Jorgensen, T. H., & Emerson, B. C. (2015). Biodiversity in the Mexican highlands and the interaction of geology, geography, and climate within the Trans-Mexican Volcanic Belt. Journal of Biogeography, 42, 1586–1600. https://doi.org/10.1111/jbi.12546

McCain, C. M., & Grytnes, J. A. (2010). Elevational gradients in species richness. Encyclopedia of Life Sciences (Eds.), (pp. 1–10). John Wiley & Sons. https://doi.org/10.1002/9780470015902.a0022548

Morrone, J. J. (2019). Regionalización biogeográfica y evolución biótica de México: encrucijada de la biodiversidad del Nuevo Mundo. Revista Mexicana de Biodiversidad, 90, e902980. https://doi.org/10.22201/ib.20078706e.2019.90.2980

Morrone, J. J., Escalante, T., & Rodríguez-Tapia, G. (2017). Mexican biogeographic provinces: map and shapefiles. Zootaxa, 4277, 277–279. https://doi.org/10.11646/zootaxa.4277.2.8

Murphy, D. J., Ebach, M. C., Miller, J. T., Laffan, S. W., Cassis, G., Ung, V. et al. (2019). Do phytogeographic patterns reveal biomes or biotic regions? Cladistics, 35, 654–670. https://doi.org/10.1111/cla.12381

Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn, D. et al. (2025). “Vegan: community ecology package”. R Package Versions 2.6-8 and 2.7-1. https://cran.r-project.org/web/packages/vegan/

R Core Team (2025). R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/

Rzedowski, J. (1978). Vegetación de México. México D.F.: Limusa.

Rzedowski, J., & Reyna-Trujillo, T. (1990). Atlas Nacional de México. Vol. II. Divisiones florísticas. Escala 1:8000000. En Instituto de Geografía (Ed.). México D.F.: Universidad Nacional Autónoma de México. Accessed January 15, 2025, from: http://www.conabio.gob.mx/informacion/gis/?vns=gis_root/region/biotic/pfito8mgw

Salas-Morales, S. H., & Meave, J. A. (2012). Elevational patterns in the vascular flora of a highly diverse region in southern Mexico. Plant Ecology, 213, 1209–1220. https://doi.org/10.1007/s11258-012-0077-6

Salinas-Rodríguez, M. M., Hernández-Sandoval, L., Carrillo-Reyes, P., Castillo-Gómez, H. A., Castro-Castro, A., Estrada-Castillón, E. et al. (2022). Diversidad de plantas vasculares de la provincia fisiográfica de la Sierra Madre Oriental, México. Botanical Sciences, 100, 469–492. https://doi.org/10.17129/botsci.2864

Suárez-Mota, M. E., Téllez-Valdés, O., Lira-Saade, R., & Villaseñor, J. L. (2013). Una regionalización de la Faja Volcánica Transmexicana con base en su riqueza florística. Botanical Sciences, 91, 93–105. https://doi.org/10.17129/botsci.405

Suárez-Mota, M. E., Villaseñor, J. L., & López-Mata, L. (2017). Dominios climáticos de la Sierra Madre Oriental y su relación con la diversidad florística. Revista Mexicana de Biodiversidad, 88, 224–233. https://doi.org/10.1016/j.rmb.2017.01.020

Takhtajan, A. (1986). Floristic regions of the world. Berkeley: University of California Press.

Tuomisto, H. (2010). A diversity of beta diversities: straightening up a concept gone awry. Part 1. Defining beta diversity as a function of alpha and gamma diversity. Ecography, 33, 2–22. https://doi.org/10.1111/j.1600-0587.2009.05880.x

Villaseñor, J. L. (2016). Checklist of the native vascular plants of Mexico. Revista Mexicana de Biodiversidad, 87, 559–902. https://doi.org/10.1016/j.rmb.2016.06.017

Villaseñor, J. L. (2018). Diversidad y distribución de la familia Asteraceae en México. Botanical Sciences, 96, 332–358. https://doi.org/10.17129/botsci.1872

Villaseñor, J. L., Encina-Domínguez, J. A., Estrada-Castillón, E., Hinton, G. S., Mora-Olivo, A., Ortiz, E. et al. (2023). Diversidad florística de la región noreste de México, estados de Coahuila, Nuevo León y Tamaulipas. Botanical Sciences, 101, 1301–1319. https://doi.org/10.17129/botsci.3328

Villaseñor, J. L., & Ortiz, E. (2022). A phytogeographic assessment of the Sierra Madre Oriental physiographic province, Mexico. Botanical Sciences, 100, 1102–1123. https://doi.org/10.17129/botsci.3086

Villaseñor, J. L., & Ortiz, E. (2025). Floristic richness comparison among the Mexican states. Revista Mexicana de Biodiversidad, 96, e965505. https://doi.org/10.22201/ib.20078706e.2025.96.5505

Villaseñor, J. L., Ortiz, E., Delgadillo-Moya, C., & Juárez, D. (2020). The breadth of the Mexican Transition Zone as defined by its flowering plant generic flora. Plos One, 15, e0235267. https://doi.org/10.1371/journal.pone.0235267

Villaseñor, J. L., Ortiz, E. & Juárez, D. (2021). Transition zones and biogeographic characterization of endemism in three biogeographic provinces of central Mexico. Botanical Sciences, 99, 938–954. https://doi.org/10.17129/botsci.2768

Villaseñor, J. L., Ortiz, E., & Meave, J. A. (2025). Floristic diversity of the Tropical Rainforest in Mexico. Botanical Sciences, 103, 899–916. https://doi.org/10.17129/botsci.3681

Villaseñor, J. L., Ortiz, E., & Murguía-Romero, M. (2024). The rarest of rarities in the flora of Mexico. Botanical Sciences, 102, 1300–1317. https://doi.org/10.17129/botsci.3498