research paper on plant taxonomy

Plant taxonomy learning and research: A systematics review

[email protected]

[email protected]

• Split-Screen
• Article contents
• Figures & tables
• Supplementary Data
• Peer Review
• Open the PDF for in another window
• Reprints and Permissions
• Cite Icon Cite
• Search Site

Wahyu Kusumawardani , Muzzazinah , Murni Ramli; Plant taxonomy learning and research: A systematics review. AIP Conf. Proc. 18 December 2019; 2194 (1): 020051. https://doi.org/10.1063/1.5139783

Download citation file:

• Ris (Zotero)
• Reference Manager

Concepts of plant identification and classification were important basic knowledge to be mastered by biology students. The research was aimed to find out what concepts and methods of learning plant taxonomy, and find out the objects and methods in plant taxonomy research. Seventeen articles published from 2005-2019 were selected as the review materials. Nine articles were about learning the plant taxonomy, and eight articles were about research on plant taxonomy. The articles were obtained from Journal of Biological Education and Science Direct. The results showed the common concepts learned about plant identification and classification. The prominent plant groups used in the learning were: the Bryophytes, Pteridophytes, Gymnosperms, and Angiosperm with the example of the native species and focal species. The learning methods and approaches were varied, including: using real plant specimens, dichotomous key method, word association exercise based on mnemonics approach, or pictorial card games for identification native plants. Others use an electronic multi-access key, iOS app on the iPod for plant identification guide, interactive multimedia dichotomous key for plant identification, labeled drawing and descriptive writing of native plant identification. Various aspects used as the object of the research on plants taxonomy, one of them was the leaves. Various methods used in the research on plant taxonomy, such as: FRT, LDC Linear, kNN, SIFT, Color moments, SFTA, ANNs, Deep learning techniques, hierarchical approach - NFC, and AIT.

Citing articles via

Publish with us - request a quote.

Sign up for alerts

• Online ISSN 1551-7616
• Print ISSN 0094-243X
• For Researchers
• For Librarians
• For Advertisers
• Our Publishing Partners  
• Physics Today
• Conference Proceedings
• Special Topics

pubs.aip.org

• Privacy Policy
• Terms of Use

Connect with AIP Publishing

This feature is available to subscribers only.

Sign In or Create an Account

• Utility Menu

GA4 Tracking Code

Harvard university herbaria & libraries.

IPNI A global authority of taxonomic names. Search the database here .

Malpighiaceae nomenclature Malpighiaceae are a family of tropical trees, shrubs, and vines that constitute an important element in the forests and savannas of both the New and Old World tropics and subtropics. Working with Professor William Anderson and Dr. Christiane Anderson (UMICH), Professor Charles Davis has helped to develop an interactive website  to make research in this family available to the broader community. This site provide updated nomenclature for hundreds of species in addition to a current phylogeny of the family, a list of the clades and genera recognized, descriptions and maps, and identification keys.

• Floristics & Monography
• Plant & Fungal Phylogenetics
• Paleobotany
• Plant Speciation and Local Adaptation

Related Researchers

• Charles Davis
• Tiago Vieira

Loading metrics

Open Access

Automated plant species identification—Trends and future directions

* E-mail: [email protected]

Affiliation Department of Biogeochemical Integration, Max Planck Institute for Biogeochemistry, Jena, Thuringia, Germany

Affiliation Software Engineering for Safety-Critical Systems Group, Technische Universität Ilmenau, Ilmenau, Thuringia, Germany

• Jana Wäldchen, 
• Michael Rzanny, 
• Marco Seeland, 
• Patrick Mäder

Published: April 5, 2018

• https://doi.org/10.1371/journal.pcbi.1005993
• Reader Comments

Current rates of species loss triggered numerous attempts to protect and conserve biodiversity. Species conservation, however, requires species identification skills, a competence obtained through intensive training and experience. Field researchers, land managers, educators, civil servants, and the interested public would greatly benefit from accessible, up-to-date tools automating the process of species identification. Currently, relevant technologies, such as digital cameras, mobile devices, and remote access to databases, are ubiquitously available, accompanied by significant advances in image processing and pattern recognition. The idea of automated species identification is approaching reality. We review the technical status quo on computer vision approaches for plant species identification, highlight the main research challenges to overcome in providing applicable tools, and conclude with a discussion of open and future research thrusts.

Author summary

Plant identification is not exclusively the job of botanists and plant ecologists. It is required or useful for large parts of society, from professionals (such as landscape architects, foresters, farmers, conservationists, and biologists) to the general public (like ecotourists, hikers, and nature lovers). But the identification of plants by conventional means is difficult, time consuming, and (due to the use of specific botanical terms) frustrating for novices. This creates a hard-to-overcome hurdle for novices interested in acquiring species knowledge. In recent years, computer science research, especially image processing and pattern recognition techniques, have been introduced into plant taxonomy to eventually make up for the deficiency in people's identification abilities. We review the technical status quo on computer vision approaches for plant species identification, highlight the main research challenges to overcome in providing applicable tools, and conclude with a discussion of open and future research thrusts.

Citation: Wäldchen J, Rzanny M, Seeland M, Mäder P (2018) Automated plant species identification—Trends and future directions. PLoS Comput Biol 14(4): e1005993. https://doi.org/10.1371/journal.pcbi.1005993

Editor: Alexander Bucksch, University of Georgia Warnell School of Forestry and Natural Resources, UNITED STATES

Copyright: © 2018 Wäldchen et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: We are funded by the German Ministry of Education and Research (BMBF) grants: 01LC1319A and 01LC1319B ( https://www.bmbf.de/ ); the German Federal Ministry for the Environment, Nature Conservation, Building and Nuclear Safety (BMUB) grant: 3514 685C19 ( https://www.bmub.bund.de/ ); and the Stiftung Naturschutz Thüringen (SNT) grant: SNT-082-248-03/2014 ( http://www.stiftung-naturschutz-thueringen.de/ ). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

One of the most obvious features of organic life is its remarkable diversity [ 1 ]. Despite the variation of organisms, a more experienced eye soon discerns that organisms can be grouped into taxa. Biology defines taxa as formal classes of living things consisting of the taxon's name and its description [ 2 ]. The assignment of an unknown living thing to a taxon is called identification [ 3 ]. This article specifically focuses on plant identification, which is the process of assigning an individual plant to a taxon based on the resemblance of discriminatory and morphological plant characters, ultimately arriving at a species or infraspecific name. These underlying characters can be qualitative or quantitative. Quantitative characters are features that can be counted or measured, such as plant height, flower width, or the number of petals per flower. Qualitative characters are features such as leaf shape, flower color, or ovary position. Individuals of the same species share a combination of relevant identification features. Since no two plants look exactly the same, it requires a certain degree of generalization to assign individuals to species (or, in other words, assign objects to a fuzzy prototype).

The world inherits a very large number of plant species. Current estimates of flowering plant species (angiosperms) range between 220,000 [ 4 , 5 ] and 420,000 [ 6 ]. Given the average 20,000 word vocabulary of an educated native English speaker, even teaching and learning the "taxon vocabulary" of a restricted region becomes a long-term endeavor [ 7 ]. In addition to the complexity of the task itself, taxonomic information is often captured in languages and formats hard to understand without specialized knowledge. As a consequence, taxonomic knowledge and plant identification skills are restricted to a limited number of persons today.

The dilemma is exacerbated since accurate plant identification is essential for ecological monitoring and thereby especially for biodiversity conservation [ 8 , 9 ]. Many activities, such as studying the biodiversity of a region, monitoring populations of endangered species, determining the impact of climate change on species distribution, payment of environmental services, and weed control actions are dependent upon accurate identification skills [ 8 , 10 ]. With the continuous loss of biodiversity [ 11 ], the demand for routine species identification is likely to further increase, while at the same time, the number of experienced experts is limited and declining [ 12 ].

Taxonomists are asking for more efficient methods to meet identification requirements. More than 10 years ago, Gaston and O’Neill [ 13 ] argued that developments in artificial intelligence and digital image processing will make automatic species identification based on digital images tangible in the near future. The rich development and ubiquity of relevant information technologies, such as digital cameras and portable devices, has brought these ideas closer to reality. Furthermore, considerable research in the field of computer vision and machine learning resulted in a plethora of papers developing and comparing methods for automated plant identification [ 14 – 17 ]. Recently, deep learning convolutional neural networks (CNNs) have seen a significant breakthrough in machine learning, especially in the field of visual object categorization. The latest studies on plant identification utilize these techniques and achieve significant improvements over methods developed in the decade before [ 18 – 23 ].

Given these radical changes in technology and methodology and the increasing demand for automated identification, it is time to analyze and discuss the status quo of a decade of research and to outline further research directions. In this article, we briefly review the workflow of applied machine learning techniques, discuss challenges of image based plant identification, elaborate on the importance of different plant organs and characters in the identification process, and highlight future research thrusts.

Machine learning for species identification

From a machine learning perspective, plant identification is a supervised classification problem, as outlined in Fig 1 . Solutions and algorithms for such identification problems are manifold and were comprehensively surveyed by Wäldchen and Mäder [ 16 ] and Cope et al. [ 17 ]. The majority of these methods are not applicable right away but rather require a training phase in which the classifier learns to distinguish classes of interest. For species identification, the training phase (orange in Fig 1 ) comprises the analysis of images that have been independently and accurately identified as taxa and are now used to determine a classifier's parameters for providing maximum discrimination between these trained taxa. In the application phase (green in Fig 1 ), the trained classifier is then exposed to new images depicting unidentified specimens and is supposed to assign them to one of the trained taxa.

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

https://doi.org/10.1371/journal.pcbi.1005993.g001

Images are usually composed of millions of pixels with associated color information. This information is too extensive and cluttered to be directly used by a machine learning algorithm. The high dimensionality of these images is therefore reduced by computing feature vectors, i.e., a quantified representation of the image that contains the relevant information for the classification problem. During the last decade, research on automated species identification mostly focused on the development of feature detection, extraction, and encoding methods for computing characteristic feature vectors. Initially, designing and orchestrating such methods was a problem-specific task, resulting in a model customized to the specific application, e.g., the studied plant parts like leaves or flowers. For example, Wu et al. [ 24 ] employ a processing chain comprised of image binarization to separate background and the leaf, image denoising, contour detection, and eventually extracting geometrical derivations of 12 leaf shape features. The approach was evaluated on 32 species and delivered an identification accuracy of 90%. However, this approach could only deal with species differing largely in their leaf shapes. Jin et al. [ 25 ] propose leaf tooth features extracted after binarization, segmentation, contour detection, and contour corner detection. The proposed method achieved an average classification rate of around 76% for the eight studied species but is not applicable to species with no significant appearances of leaf teeth [ 19 ]. The sole step from an image to a feature vector, however, typically required about 90% of the development time and extensive expert knowledge [ 19 ].

Model-free approaches aim to overcome the described limitations of model-based approaches. They do not employ application-specific knowledge and therefore promise a higher degree of generalization across different classes, i.e., species and their organs. The core concept of model-free approaches is the detection of characteristic interest points and their description using generic algorithms, such as scale-invariant feature transform (SIFT), speeded-up robust features (SURF), and histogram of gradients (HOG). These descriptors capture visual information in a patch around each interest point as orientation of gradients and have been successfully used for manifold plant classification studies, e.g., [ 26 – 28 ]. Seeland et al. [ 29 ] comparatively evaluate alternative parts of a model-free image classification pipeline for plant species identification. They found the SURF detector in combination with the SIFT local shape descriptor to be superior over other detector–descriptor combinations. For encoding interest points, in order to form an characteristic image descriptor for classification, they found the Fisher Kernel encoding to be superior.

The next obvious step in automated plant species identification and many other machine learning problems was removing an explicit decision about features to be described entirely. In the last years, deep learning CNNs have seen a significant breakthrough in computer vision due to the availability of efficient and massively parallel computing on graphics processing units (GPUs) and the availability of large-scale image data necessary for training deep CNNs with millions of parameters [ 19 ]. In contrast to model-based and model-free techniques, CNNs do not require explicit and hand-crafted feature detection and extraction steps. Instead, both become part of the iterative training process, which automatically discovers a statistically suitable image representation (similar to a feature vector) for a given problem. The fundamental concept of deep learning is a hierarchical image representation composed of building blocks with increasing complexity per layer. In a similar way, nature is compositional, i.e., small units form larger units, and each aggregation level increases the diversity of the resulting structure ( Fig 2 ). Such hierarchical representations achieve classification performances that were mostly unachievable using shallow learning methods with or without hand-crafted features (see Table 1 ).

https://doi.org/10.1371/journal.pcbi.1005993.g002

https://doi.org/10.1371/journal.pcbi.1005993.t001

Challenges in image-based taxa identification

In providing a reliable and applicable automated species identification process, researchers need to consider the following main challenges: (a) a vast number of taxa to be discriminated from one another; (b) individuals of the same species that vary hugely in their morphology; (c) different species that are extremely similar to one another; (d) specimen or other objects that are not covered by the trained classifier; and (e) large variation induced by the image acquisition process in the field.

Large number of taxa to be discriminated

The world exhibits a very large number of plant species. Distinguishing between a large number of classes is inherently more complex than distinguishing between just a few and typically requires substantially more training data to achieve satisfactory classification performance. Even when restricting the focus to the flora of a region, thousands of species need to be supported. For example, the flora of the German state of Thuringia exhibits about 1,600 flowering species [ 33 ]. Similarly, when restricting the focus to a single genus, this may still contain many species, e.g., the flowering plant genus Dioscorea aggregates over 600 species [ 17 ]. Only a few studies with such large numbers of categories have been conducted so far. For example, the important "ImageNet Large Scale Visual Recognition Challenge 2017" involves 1,000 categories that cover a wide variety of objects, animals, scenes, and even some abstract geometric concepts such as a hook or a spiral [ 34 ].

Large intraspecific visual variation

Plants belonging to the same species may show considerable differences in their morphological characteristics depending on their geographical location and different abiotic factors (e.g., moisture, nutrition, and light condition), their development stage (e.g., differences between a seedling and a fully developed plant), the season (e.g., early flowering stage to a withered flower), and the daytime (e.g., the flower is opening and closing during the day). These changes in morphological characteristics can occur on the scale of individual leaves (e.g., area, width, length, shape, orientation, and thickness), flowers (e.g., size, shape, and color), and fruits but may also affect the entire plant. Examples of visual differences of flowers during the daytime and the season are given in Fig 3 . In addition to the spatial and temporal variation, the shape of leaves and flowers may vary continuously or discretely along a single individual. For example, the leaf shape of field scabious ( Knautia arvensis ), a common plant in grassy places, ranges from large entire or dentate lanceolate ground leafs over deeply lobed and almost pinnate stem leafs to small and again lanceolate and entire upper stem leafs. Furthermore, diseases commonly affect the surface of leaves, ranging from discoloration to distinct marking, while insects often alter a leaf's shape by consuming parts of it. Some of this variation is systematic, particularly the allometric scaling of many features, but much variation is also idiosyncratic, reflecting the expression of individual genotypic and phenotypic variation related to the factors mentioned.

Visual variation of Lapsana communis 's flower throughout the day from two perspectives (left) and visual variation of Centaurea pseudophrygia 's flower throughout the season and flowering stage (right).

https://doi.org/10.1371/journal.pcbi.1005993.g003

Small interspecific visual variation

Closely related species may be extremely similar to one another. Even experienced botanists are challenged to safely distinguish species that can be identified only by almost invisible characteristics [ 35 ]. Detailed patterns in the form of particular morphological structures may be crucial and may not always be readily captured, e.g., in images of specimens. For example, the presence of flowers and fruits is often required for an accurate discrimination between species with high interspecific similarity, but these important characteristics are not present during the whole flowering season and therefore are missing in many images. Furthermore, particular morphological structures which are crucial for discrimination may not be captured in an image of a specimen, even when the particular organ is visible (e.g., the number of stamens or ovary position in the flower).

Rejecting untrained taxa

An automated taxon identification approach not only needs to be able to match an individual specimen to one of the known taxa, but should also be able to reject specimens that belong to a taxon that was not part of the training set. In order to reject unknown taxa, the classification method could produce low classification scores across all known classes for "new" taxa. However, aiming for a classifier with such characteristics conflicts with the goal of tolerating large intraspecific variation in classifying taxa. Finding a trade-off between sensitivity and specificity is a particular challenge in classifier design and training.

Variation induced by the acquisition process

Further variation is added to the images through the acquisition process itself. Living plants represent 3D objects, while images capture 2D projections, resulting in potentially large differences in shape and appearance, depending on the perspective from which the image is taken. Furthermore, image-capturing typically occurs in the field with limited control of external conditions, such as illumination, focus, zoom, resolution, and the image sensor itself [ 2 ]. These variations are especially relevant for an automated approach in contrast to human perception.

In the last decade, research in computer vision and machine learning has stimulated manifold methods for automated plant identification. Existing image-based plant identification approaches differ in three main aspects: (a) the analyzed plant organs, (b) the analyzed organ characters, and (c) the complexity of analyzed images. An extensive overview of studied methods is given by Wäldchen and Mäder [ 16 ] and is briefly summarized below.

Relevant organs for automated identification

Above the ground, plants may be composed of four visible organ types: stem, leaf, flower, and fruit. In a traditional identification process, people typically consider the plant as a whole, but also the characteristics of one or more of these organs to distinguish between taxa. In case of automated identification, organ characteristics were analyzed separately, too. For the following reasons one image alone is typically not sufficient: (a) organs may differ in scale and cannot be depicted in detail along with the whole plant or other organs; and (b) different organs require different optimal image perspectives (e.g., leaves are most descriptive from the top, while the stem is better depicted from the side).

A majority of previous studies solely utilized the leaf for discrimination [ 16 ]. The reason is a more methodological one, rather than meaning that leaves are a more discriminative part of plants from a botanical perspective. On the contrary, manual identification of plants in the vegetative state is considered much more challenging than in the flowering state. From a computer vision perspective, leaves have several advantages over other plant morphological structures, such as flowers, stems, or fruits. Leaves are available for examination throughout most of the year. They can easily be collected, preserved, and imaged due to their planar geometric properties. These aspects simplify the data acquisition process [ 17 ] and have made leaves the dominantly studied plant organ for automated identification methods in the past. In situ top-side leaf images in front of a natural background were shown to be the most effective nondestructive type of image acquisition [ 36 ]. Leaves usually refer only to broad leaves, while needles were neglected or treated separately.

Often, the visually most prominent and perceivable part of a plant is its flower . Traditional identification keys intensively refer to flowers and their parts for determination. In contrast, previous studies on automated identification rarely used flowers for discrimination [ 16 ]. Typically, flowers are only available during the blooming season, i.e., a short period of the year. Due to being complex 3D objects, there is a considerable number of variations in viewpoint, occlusions, and scale of flower images compared to leaf images. If captured in their habitat, images of flowers vary due to lighting conditions, time, date, and weather. All these aspects make flower-based classification a challenging task. However, accurate, automated identification supporting a realistic number of taxa will hardly be successful without the analysis of flowers.

Towards a more mature automated identification approach, solely analyzing one organ will often not be sufficient, especially when considering all the challenges discussed in the previous section. Therefore, more recent research started exploring multi-organ-based plant identification . The Cross Language Evaluation Forum (ImageCLEF) conference has organized a challenge dedicated to plant identification since 2011. The challenge is described as plant species retrieval based on multi-image plant observation queries and is accompanied by a dataset containing different organs of plants since 2014. Participating in the challenge, Joly et al. [ 37 ] proposed a multiview approach that analyzes up to five images of a plant in order to identify a species. This multiview approach allows classification at any period of the year, as opposed to purely leaf-based or flower-based approaches that rely on the supported organ to be visible. Initial experiments demonstrate that classification accuracy benefits from the complementarities of the different views, especially in discriminating ambiguous taxa [ 37 ]. A considerable burden in exploring this research direction is acquiring the necessary training data. However, by using mobile devices and customized apps (e.g., Pl@ntNet [ 38 ], Flora Capture [ 39 ]), it is possible to quickly capture multiple images of the same plant observed at the same time, by the same person, and with the same device. Each image, being part of such an observation, can be labeled with contextual metadata, such as the displayed organ (e.g., plant, branch, leaf, fruit, flower, or stem), time and date, and geolocation, as well as the observer.

It is beneficial if training images cover a large variety of scenarios, i.e., different organs from multiple perspective and at varying scale. This helps the model to learn adequate representations under varying circumstances. Furthermore, images of the same organ acquired from different perspectives often contain complementary visual information, improving accuracy in observation-based identification using multiple images. A structured observation approach with well defined image conditions (e.g., Flora Capture) is beneficial for finding a balance between a tedious observation process acquiring every possible scenario and a superficial acquisition that misses the characteristic images required for training.

Relevant characters for automated identification

A plant and its organs (i.e., objects in computer vision) can be described by various characters, such as color, shape, growing position, inflorescence of flowers, margin, pattern, texture, and vein structure of the leaves. These characters are extensively used for traditional identification, with many of them also being studied for automated identification. Previous research proposed numerous methods for describing general as well as domain-specific characteristics. Extensive overviews of the utilized characteristics, as well as of the methods used for capturing them in a formal description, are given by Wäldchen and Mäder [ 16 ] and Cope et al. [ 17 ].

Leaf shape is the most studied characteristic for plant identification. A plethora of methods for its description can be found in previous work [ 16 , 17 ]. Also, most traditional taxonomic keys involve leaf shape for discrimination, the reason being that, although species' leaf shape differs in detail, general shape types can easily be distinguished by people. However, while traditional identification categorizes leaf shape into classes (e.g., ovate, oblique, oblanceolate), computerized shape descriptors either analyze the contour or the whole region of a leaf. Initially, basic geometric descriptors, such as aspect ratio, rectangularity, circularity, and eccentricity, were used to describe a shape. Later, more sophisticated descriptions, such as center contour distance, Fourier descriptors, and invariant moments, were intensively studied [ 16 , 17 ].

In addition to the shape characteristic, various researchers also studied leaf texture , described by methods like Gabor filters, gray-level co-occurrence matrices (GLCM), and fractal dimensions [ 40 – 42 ]. Although texture is often overshadowed by shape as the dominant or more discriminative feature for leaf classification, it has been demonstrated to be of high discriminative power and complementary to shape information [ 16 , 43 ]. In particular, leaf texture captures leaf venation information as well as any eventual directional characteristics, and more generally allows describing fine nuances or micro-texture at the leaf surface [ 44 ]. Furthermore, leaf texture analysis allows to classify a plant by having only a portion of a leaf available without depending, e.g., on the shape of the full leaf or its color. Therefore, texture analysis can be beneficial for botanists and researchers that aim to identify damaged plants.

The vein structure as a leaf-specific characteristic also played a subordinate role in previous studies. Venation extraction is not trivial, mainly due to a possible low contrast between the venation and the rest of the leaf blade structure [ 45 ]. Some authors have simplified the task by using special equipment and treatments that render images with more clearly identified veins (mainly chemical leaf clarification) [ 45 , 46 ]. However, this defeats the goal of having users get an automated identification for specimens that they have photographed with ordinary digital cameras.

Leaf color is considered a less discriminative character than shape and texture. Leaves are mostly colored in some shade of green that varies greatly under different illumination [ 44 ], creating a low interclass color variability. In addition, there is high intraclass variability. For example, the leaves belonging to the same species or even the same plant can present a wide range of colors depending on the season and the plant's overall condition (e.g., nutrients and water). Regardless of the aforementioned complications, color may still contribute to plant identification, e.g., for considering leaves that exhibit an extraordinary hue [ 44 ]. However, further investigation on the leaf color character is required.

While the shape of the leaves is of very high relevance, flower shape has hardly been considered so far. Interestingly, flower shape is an important characteristic in the traditional identification process. It is dividing plants into families and genera and is thereby considerably narrowing the search space for identification. However, previous attempts for describing flower shape in a computable form did not find it to be very discriminative [ 47 ]. A major reason is the complex 3D structure of flowers, which makes its shape vary depending on the perspective from which an image was taken. Furthermore, flower petals are often soft and flexible, which is making them bend, curl or twist and letting the shape of the same flower appear very differently. A flower's shape also changes throughout the season [ 29 ] and with its age to the extent where petals even fall off [ 48 ], as visualized in Fig 3 .

Flower color is a more discriminative character [ 48 , 49 ]. Many traditional field guides divide plants into groups according to their flower color. For automated identification, color has been mostly described by color moments and color histograms [ 16 ]. Due to the low dimensionality and the low computational complexity of these descriptors, they are also suitable for real-time applications. However, solely analyzing color characters, without, e.g., considering flower shape, cannot classify flowers effectively [ 48 , 49 ]. Flowers are often transparent to some degree, i.e., the perceived color of a flower differs depending on whether the light comes from the back or the front of the flower. Since flower images are taken under different environmental conditions, the variation in illumination is greatly affecting analysis results. This motivated the beneficial usage of photometric invariant color characters [ 29 , 50 ].

Various previous studies showed that no single character may be sufficient to separate all desired taxa, making character selection and description a challenging problem. For example, whilst leaf shape may be sufficient to distinguish some taxa, others may look very similar to each other but have differently colored leaves or texture patterns. The same applies to flowers, where specimens of the same color may differ in their shape or texture. Therefore, various studies do not only consider one type of character but use a combination of characteristics for describing leaves and flowers [ 16 ]. The selection of characteristics is always specific for a certain set of taxa and might not be applicable to others. Meaningful characters for, e.g., flower shape can only be derived if there are flowers of sufficient size and potentially flat structure. The same applies to leaf shape and texture. This reflects a fundamental drawback of shallow learning methods using hand-crafted features for specific characters.

Deep learning

Deep artificial neural networks automate the critical feature extraction step by learning a suitable representation of the training data and by systematically developing a robust classification model. Since about 2010, extensive studies with folded neural networks have been conducted on various computer vision problems. In 2012, for the first time a deep learning network architecture with eight layers (AlexNet) won the prestigious ImageNet Challenge (ILSVRC) [ 51 ]. In the following years, the winning architectures grew in depth and provided more sophisticated mechanisms that centered around the design of layers, the skipping of connections, and on improving gradient flow. In 2015, ResNet [ 52 ] won ILSVRC with a 152 layer architecture and reached a top-5 classification error of 3.6%, being better than human performance (5.1%) [ 34 ]. As for many object classification problems, CNNs produce promising and constantly improving results on automated plant species identification. One of the first studies on plant identification utilizing CNNs is Lee et al.'s [ 53 , 54 ] leaf classifier that uses the AlexNet architecture pretrained on the ILSVRC2012 dataset and reached an average accuracy of 99.5% on a dataset covering 44 species. Zhang et al. [ 55 ] used a six-layer CNN to classify the Flavia dataset and obtained an accuracy of 94,69%. Barre et al. [ 19 ] further improved this result by using a 17-layer CNN and obtained an accuracy of 97.9%. Eventually, Sun et al. [ 31 ] study the ResNet architecture and found a 26-layer network to reach best performance with 99.65% on the Flavia dataset. Simon et al. [ 56 ] used CNNs (AlexNet and VGG19) for feature detection and extraction inside a part constellation modeling framework. Using Support Vector Machine (SVM) as classifier, they achieved 95.34% on the Oxford Flowers 102 dataset. Table 1 contrasts the best previously reported classification results of model-based, model-free and CNN-based approaches on benchmark plant image datasets. A comparison shows that CNN classification performance was unachievable using traditional and shallow learning approaches.

Training data and benchmarks

Merely half of the previous studies on automated plant identification evaluated the proposed method with established benchmark datasets allowing for replication of studies and comparison of methods (see Table 2 ). The other half solely used proprietary leaf image datasets not available to the public [ 16 ].

https://doi.org/10.1371/journal.pcbi.1005993.t002

The images contained in these datasets (proprietary as well as benchmark) fall into three categories: scans, pseudo-scans, and photos. While scan and pseudo-scan categories correspond respectively to leaf images obtained through scanning and photography in front of a simple background, the photo category corresponds to leaves or flowers photographed on natural background. The majority of utilized leaf images are scans and pseudo-scans [ 16 ]. Typically fresh material, i.e., simple, healthy, and not degraded leaves, were collected and imaged in the lab. This fact is interesting since it considerably simplifies the classification task. If the object of interest is imaged against a plain background, the often necessary segmentation for distinguishing foreground and background can be performed in a fully automated way with high accuracy.

Leaves imaged in the natural environment, as well as degraded leaves largely existing in nature, such as deformed, partial, overlapped, and compounded leaves (leaves consisting of two or more leaflets born on the same leafstalk), are largely avoided in the current studies. Segmenting the leaf with natural background is particularly difficult when the background shows a significant amount of overlapping, almost unicolor elements. This is often unavoidable when imaging leaves in their habitat. Interferences around the target leaves, such as small stones and ruderals may create confusion between the boundaries of adjacent leaves. Compound leaves are particularly difficult to recognize and existing studies that are designed for the recognition of simple leaves can hardly be applied directly to compound leaves. This is backed up by the variation of a compound leaf—it is not only caused by morphological differences of leaflets, but also by changes in the leaflet number and arrangements [ 57 ].

The lower part of Table 2 shows benchmark datasets containing flower images. The images of the Oxford Flower 17 and 102 datasets have been acquired by searching the internet and by selecting images of species with substantial variation in shape, scale, and viewpoint. The PlantCLEF2015/2016 dataset consists of images with different plant organs or plant views (i.e., entire plant, fruit, leaf, flower, stem, branch, and leaf scan). These images were submitted by a variety of users of the mobile Pl@ntNet application. The recently published Jena Flower 30 dataset [ 29 ] contains images acquired in the field as top-view flower images using an Apple iPhone 6 throughout an entire flowering season. All images of these flower benchmark datasets are photos taken in the natural environment.

Applicable identification tools

Despite intensive and elaborate research on automated plant species identification, only very few studies resulted in approaches that can be used by the general public, such as Leafsnap [ 61 ] and Pl@ntNet [ 37 ]. Leafsnap, developed by researchers from Columbia University, the University of Maryland, and the Smithsonian Institution, was the first widely distributed electronic field guide. Implemented as a mobile app, it uses computer vision techniques for identifying tree species of North America from photographs of their leaves on plain background. The app retrieves photos of leaves similar to the one in question. However, it is up to the user to make the final decision on what species matches the unknown one. LeafSnap achieves a top-1 recognition rate of about 73% and a top-5 recognition rate of 96.8% for 184 tree species [ 61 ]. The app has attracted a considerable number of downloads but has also received many critical user reviews [ 62 ] due to its inability to deal with cluttered backgrounds and within-class variance.

Pl@ntNet is an image retrieval and sharing application for the identification of plants. It is being developed in a collaboration of four French research organizations (French agricultural research and international cooperation organization [Cirad], French National Institute for Agricultural Research [INRA], French Institute for Research in Computer Science and Automation [Inria], and French National Research Institute for Sustainable Development [IRD]) and the Tela Botanica network. It offers three front-ends, an Android app, an iOS app, and a web interface, each allowing users to submit one or several pictures of a plant in order to get a list of the most likely species in return. The application is becoming more and more popular. The application has been downloaded by more than 3 million users in about 170 countries. It was initially restricted to a fraction of the European flora (in 2013) and has since been extended to the Indian Ocean and South American flora (in 2015) and the North African flora (in 2016). Since June 2015, Pl@ntNet applies deep learning techniques for image classification. The network is pretrained on the ImageNet dataset and periodically fine-tuned on steadily growing Pl@ntNet data. Joly et al. [ 63 ] evaluated the Pl@ntNet application, which supported the identification of 2,200 species at that time, and reported a 69% top-5 identification rate for single images. We could not find published evaluation results on the current performance of the image-based identification engine. However, reviews request better accuracy [ 15 ]. We conclude that computer vision solutions are still far from replacing the botanist in extracting plant characteristic information for identification. Improving the identification performance in any possible way remains an essential objective for future research. The following sections summarize important current research directions.

Open problems and future directions

Utilizing latest machine learning developments.

While the ResNet architecture is still state-of-the-art, evolutions are continuously being proposed, (e.g., [ 64 ]). Other researchers work on alternative architectures like ultra-deep (FractalNet) [ 65 ] and densely connected (DenseNet) [ 66 ] networks. These architectures have not yet been evaluated for plant species identification. New architectures and algorithms typically aim for higher classification accuracy, which is clearly a major goal for species identification; however, there are also interesting advances in reducing the substantial computational effort and footprint of CNN classifiers. For example, SqueezeNet [ 67 ] achieves accuracy comparable to AlexNet but with 50 times fewer parameters and a model that is 510 times smaller. Especially when aiming for identification systems that run on mobile devices, these developments are highly relevant and should be evaluated in this context.

Current studies still mostly operate on the small and nonrepresentative datasets used in the past. Only a few studies train CNN classifiers on large plant image datasets, demonstrating their applicability in automated plant species identification systems [ 68 ]. Given the typically "small" amounts of available training data and the computational effort for training a CNN, transfer learning has become an accepted procedure (meaning that a classifier will be pretrained on a large dataset, e.g., ImageNet, before the actual training begins). The classifier will then only be fine-tuned to the specific classification problem by training of a small number of high-level network layers proportional to the amount of available problem-specific training data. Researchers argue that this method is superior for problems with ≤ 1 M training images. Most previous studies on plant species identification utilized transfer learning, (e.g., [ 54 , 69 ]). Once a sufficiently large plant dataset has been acquired, it would be interesting to compare current classification results with those of a plant identification CNN solely trained on images depicting plant taxa.

Another approach tackling the issue of small datasets is using data augmentation schemes, commonly including simple modifications of images, such as rotation, translation, flipping, and scaling. Using augmentation for improving the training process has become a standard procedure in computer vision. However, the diversity that can be reached with traditional augmentation schemes is relatively small. This motivates the use of synthetic data samples, introducing more variability and enriching the dataset, in order to improve the training process. A promising approach in this regard are Generative Adversarial Networks (GANs) that are able to generate high-quality, realistic, natural images [ 70 ].

Without the complicated and time-consuming process for designing an image analysis pipeline, deep learning approaches can also be applied by domain experts directly, i.e., botanists and biologists with only a basic understanding of the underlying machine learning concepts. Large-scale organizations provide a competing and continuously improving set of openly available machine learning frameworks, such as Caffe2, MXNet, PyTorch, and TensorFlow. Developments like Keras specifically target newcomers in machine learning and provide add-ons to these frameworks that aim to simplify the setup of experiments and the analysis of results. Furthermore, it is mostly common practice that researchers make their models and architectures publicly available (model zoos), increasing visibility in their field but also facilitating their application in other studies.

Creating representative benchmarks

Todays benchmark datasets are limited both in the number of species and in the number of images (see Table 2 ) due to the tremendous effort for either collecting fresh specimens and imaging them in a lab or for taking images in the field. Taking a closer look at datasets, it becomes obvious that they were created with an application in computer vision and machine learning in mind. They are typically created by only a few people acquiring specimens or images in a short period of time, from a limited area, and following a rigid procedure for their imaging. As a result, the plants of a given species in those datasets are likely to represent only a few individual plants grown closely together at the same time. Considering the high variability explained before, these datasets do not reflect realistic conditions.

Using such training data in a real-world identification application has little chance to truly classify new images collected at different periods, at different places, and acquired differently [ 63 ]. Towards real-life applications, studies should utilize more realistic images, e.g., containing multiple, overlapped, and damaged leaves and flowers. Images should have real, complex backgrounds and should be taken under different lighting conditions. Large-scale, well-annotated training datasets with representative data distribution characteristics are crucial for the training of accurate and generalizable classifiers. This is especially true for the training of Deep Convolutional Neural Networks that require extensive training data to properly tune the large set of parameters. The research community working on the ImageNet dataset [ 71 ] and the related benchmark is particularly important in this regard. ImageNet aims to provide the most comprehensive and diverse coverage of the image world. It currently contains more than 14 million images categorized according to a hierarchy of almost 22,000 English nouns. The average number of training images per category is in the range of 600 and 1,200, being considerable larger than any existing plant image collection.

First efforts have been made recently to create datasets that are specifically designed for machine learning purposes—a huge amount of information, presorted in defined categories. The PlantCLEF plant identification challenge initially provided a dataset containing 71 tree species from the French Mediterranean area depicted in 5,436 images in 2011. This dataset has grown to 113,205 pictures of herb, tree, and fern specimens belonging to 1,000 species living in France and the neighboring countries in 2016. Encyclopedia Of Life (EOL) [ 72 ], being the world's largest data centralization effort concerning multimedia data for life on earth, currently provides about 3.8 million images for 1.3 million taxa. For angiosperms, there are currently 1.26 million images, but only 68% of them are reviewed and trusted with respect to the identified taxa [ 73 ].

Crowdsourcing training data

Upcoming trends in crowdsourcing and citizen science offer excellent opportunities to generate and continuously update large repositories of required information. Members of the public are able to contribute to scientific research projects by acquiring or processing data while having few prerequisite knowledge requirements. Crowdsourcing has benefited from Web 2.0 technologies that have enabled user-generated content and interactivity, such as wiki pages, web apps, and social media. iNaturalist and Pl@ntNET already successfully acquire data through such channels [ 37 ]. Plant image collections that acquire data through crowdsourcing and citizen science projects today often suffer from problems that prevent their effective use as training and benchmark data. First, the number of images per species in many datasets follows a long-tail distribution . Thousands of images are acquired for prominent taxa, while less prominent and rare taxa are represented by only a few and sometimes no images at all. The same fact applies to the number of images per organ per taxon. While prominent organs such as the flower of angiosperms are well populated, other organs such as fruits are often underrepresented or even missing. Second, collections contain a high degree of image and tag heterogeneity . As we elaborated in our discussion of identification challenges, the acquisition process is a main contributor of image variability. In a crowdsourcing environment, this fact is even exacerbated since contributors with very different backgrounds, motivations, and equipment contribute observations. Image collections today contain many examples not sufficient for an unambiguous identification of the displayed taxon. They may be too blurry or lack details. Collections also suffer from problems such as heterogeneous organ tags (e.g., "leaf" versus "leaves" versus "foliage"), manifold plant species synonyms used alternatively, and evolving and concurrent taxonomies. Third, nonexpert observations are more likely to contain image and metadata noise . Image noise refers to problems such as highly cluttered images, other plants depicted along with the intended species, and objects not belonging to the habitat (e.g., fingers or insects). Metadata noise refers to problems such as wrongly identified taxa, wrongly labeled organs, imprecise or incorrect location information, and incorrect observation time and date.

These problems show that crowdsourced content deserves more effort for maintaining sufficient data quality. An examination of a small number of randomly sampled images from the Pl@ntNET initiative and their taxa attributions indicated that misclassifications are in the range of 5% to 10%. In a first attempt to overcome these problems, Pl@ntNET introduced a star-based quality rating for each image and uses a community based review system for taxon annotations, whereas EOL offers a "trusted" tag for each taxon that has been identified within an image by an EOL curator. We argue that multimedia data should be based on common data standards and protocols, such as the Darwin Core [ 74 ], and that a rigorous review system and quality control workflows should be implemented for community based data assessment.

Analyzing the context of observations

We argue that it is hard to develop a plant identification approach for the worlds estimated 220,000 to 420,000 angiosperms that solely relies on image data. Additional information characterizing the context of a specimen should be taken into consideration. Today, mobile devices allow for high quality images acquired in well choreographed and adaptive procedures. Through software specifically developed for these devices, users can be guided and trained in acquiring characteristic images in situ. Given that mobile devices can geolocalize themselves, acquired data can be spatially referenced with high accuracy allowing to retrieve context information, such as topographic characteristics, climate factors, soil type, land-use type, and biotope. These factors explaining the presence or absence of species are already used to predict plant distribution and should also be considered for their identification. Temporal information, i.e., the date and the time of an observation, could allow adaptation of an identification approach to species' seasonal variations. For example, the flowering period can be of high discriminative power during an identification. Furthermore, recorded observations in public repositories (e.g., Global Biodiversity Information Facility GBIF) can provide valuable hypotheses as to which species are to expect or not to expect at a given location. Finally, additional and still-emerging sensors built into mobile devices allow for measuring environmental variables, such as temperature and air pressure. The latest cameras can acquire depth maps of specimens along with an image and provide additional characteristics of an observation and its context further supporting the identification.

From taxa-based to character-based training

In automated species identification, researchers solely aim to classify on the species level so far. An alternative approach could be classifying plant characteristics (e.g., leaf shape categories, leaf position, flower symmetry) and linking them to plant character databases such as the TRY Plant Trait Database [ 75 ] for identifying a wide range of taxa. In theory, untrained taxa could be identified by recognizing their characters. So far, it is uncertain whether automated approaches are able to generalize uniform characters from nonuniform visual information. Characters that are shared across different taxa are often differently developed per taxon, making their recognition a particular challenge.

Utilizing the treasure of herbarium specimens

Herbaria all over the world have invested large amounts of money and time in collecting samples of plants. Rather than going into the field for taking images or for collecting specimens anew, it would be considerably less expensive to use specimens of plants that have already been identified and conserved. Today, over 3,000 herbaria in 165 countries possess over 350 million specimens, collected in all parts of the world and over several centuries [ 76 ]. Currently, many are undertaking large-scale digitization projects to improve their access and to preserve delicate samples. For example, in the USA, more than 1.8 million imaged and georeferenced vascular plant specimens are digitally archived in the iDigBio portal, a nationally funded and primary aggregator of museum specimen data [ 76 ]. This activity is likely going to be expanded over the coming decade. We can look forward to a time when there will be huge repositories of taxonomic information, represented by specimen images, accessible publicly through the internet. However, very few previous researchers utilized herbaria sheets for generating a leaf image dataset [ 58 , 69 , 77 – 79 ]. On the other hand, analyzing herbaria specimens may not be suitable for training identification approaches applied in a real environment [ 69 ]. The material is dried, and thereby, original colors change drastically. Furthermore, all herbaria specimens are imaged flattened on a plain homogeneous background, altering their structure and arrangement. In conclusion, more research on the detection and extraction of characteristics from herbaria specimens is required. It is also an open research question (how to train classifiers on herbaria specimens that are applicable on fresh specimens).

Interdisciplinary collaborations

Twelve years ago, Gaston and O’Neill [ 13 ] argued that developing successful identification approaches requires novel collaborations between biologists and computer scientists with personnel that have significant knowledge of both biology and computing science. Interestingly, automated plant species identification is still mostly driven by academics specialized in computer vision, machine learning, and multimedia information retrieval. Very few studies were conducted by interdisciplinary groups of biologists and computer scientists in the previous decade [ 16 ]. Research should move towards more interdisciplinary endeavors. Biologists can apply machine learning methods more effectively with the help of computer scientists, and the latter are able to gain the required exhaustive understanding of the problem they are tacking by working with the former.

A vision of automated identification in the wild

We envision identification systems that enable users to take images of specimens in the field with a mobile device's built-in camera system, which are then analyzed by an installed application to identify the taxon or to at least get a list of candidate taxa. This approach is convenient, since the identification requires no work from the user except for taking an image and browsing through the best matching species. Furthermore, minimal expert knowledge is required, which is especially important given the ongoing shortage of skilled botanists. An accurate automated identification system also enables nonexperts with only limited botanical training and expertise to contribute to the survey of the world's biodiversity. Approaching trends and technologies, such as augmented reality, data glasses, and 3D-scans, give such applications a long-term research and application perspective. Furthermore, large-character datasets can be generated automatically (for instance, by taking measurements from thousands of specimens across a single taxon). We cannot only derive more accurate descriptions of a species and its typical character expressions, but also study the statistical distribution of each character, including variance and skew. Furthermore, image processing provides the possibility to extract not only the linear measurements typical of botanical descriptions (leaf length, leaf width, petal length, etc.), but also more sophisticated and precise descriptions such as mathematical models of leaf shapes.

• 1. Darwin Charles R. On the origin of species by means of natural selection, or the preservation of favoured races in the struggle for life. Murray, London. 1859.
• 2. Remagnino P, Mayo S, Wilkin P, Cope J, Kirkup D. Computational Botany: Methods for Automated Species Identification. Springer; 2016.
• 3. Hagedorn G, Rambold G, Martellos S. Types of identification keys. In: Tools for Identifying Biodiversity: Progress and Problems. EUT Edizioni Università di Trieste; 2010. pp. 59–64.
• View Article
• Google Scholar
• PubMed/NCBI
• 12. Hopkins G, Freckleton R. Declines in the numbers of amateur and professional taxonomists: implications for conservation. In: Animal Conservation forum. vol. 5. Cambridge University Press; 2002. p. 245–249.
• 14. Kumar N, Belhumeur P, Biswas A, Jacobs D, Kress WJ, Lopez I, et al. Leafsnap: A Computer Vision System for Automatic Plant Species Identification. In: Fitzgibbon A, Lazebnik S, Perona P, Sato Y, Schmid C, editors. Computer Vision–ECCV 2012. Lecture Notes in Computer Science. Springer Berlin Heidelberg; 2012. p. 502–516.
• 18. Pawara P, Okafor E, Schomaker L, Wiering M. Data Augmentation for Plant Classification. In: Proceedings of International ConferenceAdvanced Concepts for Intelligent Vision Systems. Springer International Publishing; 2017. pp. 615–626.
• 20. Pawara P, Okafor E, Surinta O, Schomaker L, Wiering M. Comparing Local Descriptors and Bags of Visual Words to Deep Convolutional Neural Networks for Plant Recognition. In: Proceedings of the 6th International Conference on Pattern Recognition Applications and Methods. ICPRAM; 2017. pp. 479–486.
• 24. Wu SG, Bao FS, Xu EY, Wang YX, Chang YF, Xiang QL. A Leaf Recognition Algorithm for Plant Classification Using Probabilistic Neural Network. In: Proceedings of the IEEE International Symposium on Signal Processing and Information Technology, 2007. pp. 11–16.
• 26. Xiao XY, Hu R, Zhang SW, Wang XF. HOG-based Approach for Leaf Classification. In: Proceedings of the Advanced Intelligent Computing Theories and Applications, and 6th International Conference on Intelligent Computing. ICIC'10. Berlin, Heidelberg: Springer-Verlag; 2010. pp.149–155.
• 27. Nguyen QK, Le TL, Pham NH. Leaf based plant identification system for Android using SURF features in combination with Bag of Words model and supervised learning. In: Proceedings of the International Conference on Advanced Technologies for Communications (ATC); 2013. pp. 404–407.
• 30. Söderkvist O. Computer Vision Classification of Leaves from Swedish Trees. Department of Electrical Engineering, Computer Vision, Linköping Universityping University; 2001.
• 32. Wang Z, Lu B, Chi Z, Feng D. Leaf Image Classification with Shape Context and SIFT Descriptors. In: Proceedings of the International Conference on Digital Image Computing Techniques and Applications (DICTA), 2011. pp. 650–654.
• 33. Zündorf H, Günther K, Korsch H, Westhus W. Flora von Thüringen. Weissdorn, Jena. 2006.
• 35. Müller F, Ritz CM, Welk E, Wesche K. Rothmaler-Exkursionsflora von Deutschland: Gefäßpflanzen: Kritischer Ergänzungsband. Springer-Verlag; 2016.
• 37. Joly A, Goëau H, Bonnet P, Bakić V, Barbe J, Selmi S, et al. Interactive plant identification based on social image data. Ecological Informatics. 2014;23:22–34.
• 38. Pl@ntNet; 2017. Available from: https://identify.plantnet-project.org/ . 1st October 2017
• 39. The Flora Incognita Project; 2017. Available from: http://floraincognita.com . 1st October 2017
• 40. Cope J, Remagnino P, Barman S, Wilkin P. Plant Texture Classification Using Gabor Co-occurrences. In: Bebis G, Boyle R, Parvin B, Koracin D, Chung R, Hammound R, et al., editors. Advances in Visual Computing. vol. 6454 of Lecture Notes in Computer Science. Springer Berlin Heidelberg; 2010. p. 669–677.
• 48. Nilsback ME, Zisserman A. A Visual Vocabulary for Flower Classification. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 2006. pp. 1447–1454.
• 49. Nilsback ME, Zisserman A. Automated Flower Classification over a Large Number of Classes. In: Proceedings of the IEEE Indian Conference on Computer Vision, Graphics and Image Processing. 2008. pp. 722–729.
• 50. Seeland M, Rzanny M, Alaqraa N, Thuille A, Boho D, Wäldchen J, et al. Description of Flower Colors for Image based Plant Species Classification. In: Proceedings of the 22nd German Color Workshop (FWS). Ilmenau, Germany: Zentrum für Bild- und Signalverarbeitung e.V; 2016. pp.145–1154.
• 52. He K, Zhang X, Ren S, Sun J. Deep Residual Learning for Image Recognition. In: Proceedings of the of the IEEE Conference on Computer Vision and Pattern Recognition. 2016. pp. 770–778.
• 53. Lee SH, Chan CS, Wilkin P, Remagnino P. Deep-plant: Plant identification with convolutional neural networks. In: Proceedings of the IEEE International Conference on Image Processing. 2015. pp. 452–456.
• 55. Zhang C, Zhou P, Li C, Liu L. A convolutional neural network for leaves recognition using data augmentation. In: Proceedings of the IEEE International Conference on Computer and Information Technology. 2015; 2143–2150.
• 56. Simon M, Rodner E. Neural Activation Constellations: Unsupervised Part Model Discovery with Convolutional Networks. In: Proceedings of IEEE International Conference on Computer Vision. 2015. pp. 1143–1151.
• 60. Goëau H, Bonnet P, Joly A. Plant identification in an open-world. In: CLEF 2016 Conference and Labs of the Evaluation forum. 2016. pp. 428–439
• 61. Kumar Mishra P, Kumar Maurya S, Kumar Singh R, Kumar Misra A. A semi automatic plant identification based on digital leaf and flower images. In: Proceedings of International Conference on Advances in Engineering, Science and Management, 2012. pp. 68–73.
• 62. Leafsnap; 2017. Available from: https://itunes.apple.com/us/app/leafsnap/id430649829 . 1st October 2017
• 63. Joly A, Müller H, Goëau H, Glotin H, Spampinato C, Rauber A, et al. Lifeclef: Multimedia life species identification. In: Proceedings of ACM Workshop on Environmental Multimedia Retrieval; 2014. pp. 1–7.
• 64. Huang G, Sun Y, Liu Z, Sedra D, Weinberger KQ. Deep Networks with Stochastic Depth. In: Proceedings of European Conference on Computer Vision. 2016. pp. 646–661.
• 68. Goëau H, Bonnet P, Joly A. Plant Identification Based on Noisy Web Data: the Amazing Performance of Deep Learning. In: Workshop Proceedings of Conference and Labs of the Evaluation Forum (CLEF 2017). 2017
• 70. Odena, A., Olah, C., Shlens, J. Conditional image synthesis with auxiliary classifier gans. arXiv preprint arXiv:1610.09585.
• 71. Deng J, Dong W, Socher R, Li L, Li K, Li F. ImageNet: A large-scale hierarchical image database. In: Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 2009. pp. 248–255.
• 72. Encyclopedia of Life (EOL); 2017. Available from: http://eol.org/statistics . 6th July 2017
• 73. Encyclopedia of Life (EOL); 2017. Available from: http://eol.org/pages/282/media . 6th July 2017
• 78. Grimm J, Hoffmann M, Stöver B, Müller K, Steinhage V. Image-Based Identification of Plant Species Using a Model-Free Approach and Active Learning. In: Proceedings of Annual German Conference on AI. 2016. pp. 169–176.

Plant Taxonomy: A Historical Perspective, Current Challenges, and Perspectives

• First Online: 11 December 2020

Cite this protocol

• Germinal Rouhan 3 &
• Myriam Gaudeul 3  

Part of the book series: Methods in Molecular Biology ((MIMB,volume 2222))

2094 Accesses

10 Citations

Taxonomy is the science that explores, describes, names, and classifies all organisms. In this introductory chapter, we highlight the major historical steps in the elaboration of this science, which provides baseline data for all fields of biology and plays a vital role for society but is also an independent, complex, and sound hypothesis-driven scientific discipline.

In a first part, we underline that plant taxonomy is one of the earliest scientific disciplines that emerged thousands of years ago, even before the important contributions of the Greeks and Romans (e.g., Theophrastus, Pliny the Elder, and Dioscorides). In the fifteenth–sixteenth centuries, plant taxonomy benefited from the Great Navigations, the invention of the printing press, the creation of botanic gardens, and the use of the drying technique to preserve plant specimens. In parallel with the growing body of morpho-anatomical data, subsequent major steps in the history of plant taxonomy include the emergence of the concept of natural classification , the adoption of the binomial naming system (with the major role of Linnaeus) and other universal rules for the naming of plants, the formulation of the principle of subordination of characters, and the advent of the evolutionary thought. More recently, the cladistic theory (initiated by Hennig) and the rapid advances in DNA technologies allowed to infer phylogenies and to propose true natural, genealogy-based classifications.

In a second part, we put the emphasis on the challenges that plant taxonomy faces nowadays. The still very incomplete taxonomic knowledge of the worldwide flora (the so-called taxonomic impediment) is seriously hampering conservation efforts that are especially crucial as biodiversity has entered its sixth extinction crisis. It appears mainly due to insufficient funding, lack of taxonomic expertise, and lack of communication and coordination. We then review recent initiatives to overcome these limitations and to anticipate how taxonomy should and could evolve. In particular, the use of molecular data has been era-splitting for taxonomy and may allow an accelerated pace of species discovery. We examine both strengths and limitations of such techniques in comparison to morphology-based investigations, we give broad recommendations on the use of molecular tools for plant taxonomy, and we highlight the need for an integrative taxonomy based on evidence from multiple sources.

Taxonomy can justly be called the pioneering exploration of life on a little known planet. —Wilson (2004). The goal of discovering, describing, and classifying the species of our planet assuredly qualifies as big science . —Wheeler et al. (2004).

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

• Available as PDF
• Read on any device
• Instant download
• Own it forever
• Available as EPUB and PDF
• Compact, lightweight edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info
• Durable hardcover edition

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Dobzhansky T (1973) Nothing in biology makes sense except in the light of evolution. Am Biol Teach 35:125–129

Article   Google Scholar  

de Carvalho MR, Bockmann FA, Amorim DS, de Vivo M, de Toledo-Piza M, Menezes NA, de Figueiredo JL, Castro RMC, Gill AC, McEachran JD, Compagno LJV, Schelly RC, Britz R, Lundberg JG, Vari RP, Nelson G (2005) Revisiting the taxonomic impediment. Science 307:353–353

Article   PubMed   Google Scholar  

Candolle AP (1813) Théorie Élémentaire de la botanique, ou Exposition des principes de la classification naturelle et de l'art de décrire et d'étudier les végétaux

Google Scholar  

Heywood VH, Watson RT (1995) Global biodiversity assessment. Cambridge University Press, Cambridge

Mayr E (1969) Principles of systematic zoology. Mcgraw-Hill, New. York

Simpson GG (1961) Principles of animal taxonomy. Columbia University Press, New York

Book   Google Scholar  

Tillier S (2000) Systématique - Ordonner la diversité du Vivant. Rapport sur la science et la technologie de l’Académie des sciences n°11. Éditions Tec & Doc

Small E (1989) Systematics of biological systematics (or taxonomy of taxonomy). Taxon 38:335–356

Sprague JL, Lanjouw J, Andreas CH (1948) Minutes of the Utrecht conference. Chronica Botanica 12(1/2):12

Morton CV (1957) The misuse of the term taxon. Taxon 6(5):155

Raven PH (2004) Taxonomy: where are we now? Philos Trans R Soc Lond B Biol Sci 359:729–730

Article   PubMed   PubMed Central   Google Scholar  

Pavord A (2005) The naming of names: the search for order in the world of plants. Bloomsbury, New York

Funk VA, Hoch PC, Prather LA, Wagner WL (2005) The importance of vouchers. Taxon 54:127–129

Knapp S (2012) What’s in a name? A history of taxonomy. http://www.nhm.ac.uk/nature-online/science-of-natural-history/taxonomy-systematics/history-taxonomy . Accessed Jan 2012

Griffing LR (2011) Who invented the dichotomous key? Richard Waller’s watercolors of the herbs of Britain. Am J Bot 98:1911–1923

Linnaeus C (1753) Species Plantarum. Stockholm

Linnaeus C (1758) Systema naturae, 10th edn. Stockholm

Candolle AP (1867) Lois de la nomenclature botanique adoptées par le Congrès international de botanique: tenu à Paris en août 1867. H. Georg, Geneva

Jussieu AL (1789) Genera plantarum. Herissant, Paris

Philippe H, Lecointre G, VanLe HL, LeGuyader H (1996) A critical study of homoplasy in molecular data with the use of a morphologically based cladogram, and its consequences for character weighting. Mol Biol Evol 13:1174–1186

Lamarck J-BPAM (1809) Philosophie zoologique. Dentu, Paris

Darwin C (1859) On the origin of species by means of natural selection, or the preservation of favoured races in the struggle for life. John Murray, London

Haeckel E (1866) Generelle Morphologie der Organismen. Reimer, Berlin

Dayrat B (2003) The roots of phylogeny: how did Haeckel build his trees? Syst Biol 52:515–527

Davis PH, Heywood PH (1963) Principles of angiosperm taxonomy. Oliver and Boyd, Edinburgh/London

Sneath PHA, Sokal RR (1963) Principles of numerical taxonomy, 7th edn. W. H. Freeman, San Francisco

Hennig W (1950) Grundzüge einer Theorie der phylogenetischen Systematik. Deutscher Zentralverlag, Berlin

Hennig W (1966) Phylogenetic systematics (tr. D. Davis and R. Zangerl), University of Illinois Press, Urbana

Godfray HCJ, Knapp S (2004) Taxonomy for the twenty-first century - introduction. Philos Trans R Soc Lond B Biol Sci 359:559–569

Article   CAS   PubMed   PubMed Central   Google Scholar  

Mullis K, Faloona F, Scharf S, Saiki R, Horn G, Erlich H (1986) Specific enzymatic amplification of DNA Invitro - the polymerase chain-reaction. Cold Spring Harb Symp Quant Biol 51:263–273

Article   CAS   PubMed   Google Scholar  

Bremer K, Chase MW, Stevens PF, Anderberg AA, Backlund A, Bremer B, Briggs BG, Endress PK, Fay MF, Goldblatt P, Gustafsson MHG, Hoot SB, Judd WS, Kallersjo M, Kellogg EA, Kron KA, Les DH, Morton CM, Nickrent DL, Olmstead RG, Price RA, Quinn CJ, Rodman JE, Rudall PJ, Savolainen V, Soltis DE, Soltis PS, Sytsma KJ, Thulin M, Grp AP (1998) An ordinal classification for the families of flowering plants. Ann Mo Bot Gard 85:531–553

Bremer B, Bremer K, Chase MW, Reveal JL, Soltis DE, Soltis PS, Stevens PF, Anderberg AA, Fay MF, Goldblatt P, Judd WS, Kallersjo M, Karehed J, Kron KA, Lundberg J, Nickrent DL, Olmstead RG, Oxelman B, Pires JC, Rodman JE, Rudall PJ, Savolainen V, Sytsma KJ, van der Bank M, Wurdack K, Xiang JQY, Zmarzty S, Grp AP (2003) An update of the angiosperm phylogeny group classification for the orders and families of flowering plants: APG II. Bot J Linn Soc 141:399–436

Bremer B, Bremer K, Chase MW, Fay MF, Reveal JL, Soltis DE, Soltis PS, Stevens PF, Anderberg AA, Moore MJ, Olmstead RG, Rudall PJ, Sytsma KJ, Tank DC, Wurdack K, Xiang JQY, Zmarzty S, Grp AP (2009) An update of the angiosperm phylogeny group classification for the orders and families of flowering plants: APG III. Bot J Linn Soc 161:105–121

The Angiosperm phylogeny group (2016) An update of the angiosperm phylogeny group classification for the orders and families of flowering plants: APG IV. Bot J Linn Soc 181:1–20. https://doi.org/10.1111/boj.12385

PPGI (2016) A community-derived classification for extant lycophytes and ferns. J Syst Evol 54:563–603. https://doi.org/10.1111/jse.12229

Pryer KM, Schneider H, Smith AR, Cranfill R, Wolf PG, Hunt JS (2001) Horsetails and ferns are a monophyletic group and the closest living relatives to seed plants. Nature 409:618–621. https://doi.org/10.1038/35054555

Bessey CE (1915) The phylogenetic taxonomy of flowering plants. Ann Mo Bot Gard 2:109–164

Cronquist A (1981) An integrated system of classification of flowering plants. Columbia University Press, New York

Stebbins GL (1974) Flowering plants: evolution above the species level. Belknap press, Cambridge

Takhtajan A (1997) Diversity and classification of flowering plants. Columbia University Press, New York

Thorne RF (1976) A phylogenetic classification of the Angiospermae. Evol Biol 9:35–106

Soltis DE, Soltis PS, Endress PK, Chase MW (2005) Phylogeny and evolution of angiosperms. Sinauer associates, Sunderland

Hebert PDN, Cywinska A, Ball SL, DeWaard JR (2003) Biological identifications through DNA barcodes. Proc R Soc Lond B Biol Sci 270:313–321

Article   CAS   Google Scholar  

May RM (2004) Tomorrow's taxonomy: collecting new species in the field will remain the rate-limiting step. Philos Trans R Soc Lond B Biol Sci 359:733–734

May RM (2011) Why worry about how many species and their loss? PLoS Biol 9:e1001130

Paton AJ, Brummitt N, Govaerts R, Harman K, Hinchcliffe S, Allkin B, Lughadha EN (2008) Towards target 1 of the global strategy for plant conservation: a working list of all known plant species - progress and prospects. Taxon 57:602–611

Wilson EO (2003) The encyclopedia of life. Trends Ecol Evol 18:77–80

Wilson EO (2004) Taxonomy as a fundamental discipline. Philos Trans R Soc Lond B Biol Sci 359:739–739

Candolle AP (1824-1873) Prodromus systematis naturalis regni vegetabilis. Sumptibus Sociorum Treuttel et Würtz, Parisii

International Plant Names Index (2012) Published on the Internet http://www.ipni.org . Accessed July 2019

Lughadha EN, Govaerts R, Belyaeva I, Black N, Lindon H, Allkin R, Magill RE, Nicolson N (2016) Counting counts: revised estimates of numbers of accepted 407 species of flowering plants, seed plants, vascular plants and land plants with a review 408 of other recent estimates. Phytotaxa 272:82–88

Scotland RW, Wortley AH (2003) How many species of seed plants are there? Taxon 52:101–104

The Plant List (2019) Version 1. Published on the Internet http://www.theplantlist.org/ . Accessed July 2019

Wortley AH, Scotland RW (2004) Synonymy, sampling and seed plant numbers. Taxon 53:478–480

Mallet J, Willmott K (2003) Taxonomy: renaissance or tower of babel? Trends Ecol Evol 18:57–59

Isaac NJB, Mallet J, Mace GM (2004) Taxonomic inflation: its influence on macroecology and conservation. Trends Ecol Evol 19:464–469

Meiri S, Mace GM (2007) New taxonomy and the origin of species. PLoS Biol 5:1385–1386

Pillon Y, Chase MW (2007) Taxonomic exaggeration and its effects on orchid conservation. Conserv Biol 21:263–265

Crane PR (2004) Documenting plant diversity: unfinished business. Philos Trans R Soc Lond B Biol Sci 359:735–737

Joppa LN, Roberts DL, Pimm SL (2011) How many species of flowering plants are there? Proc R Soc B Biol Sci 278:554–559

Mora C, Tittensor DP, Adl S, Simpson AGB, Worm B (2011) How many species are there on earth and in the ocean? PLoS Biol 9:e1001127

Bisby FA, Roskov YR, Orrell TM, Nicolson D, Paglinawan LE et al (2010) Species 2000 & ITIS Catalogue of Life: 2010 Annual Checklist. Digital resource at http://www.catalogueoflife.org/annual-checklist/2010 . Species 2000, Reading, UK

Caldecott JO, Jenkins MD, Johnson TH, Groombridge B (1996) Priorities for conserving global species richness and endemism. Biodivers Conserv 5:699–727

Joppa LN, Roberts DL, Myers N, Pimm SL (2011) Biodiversity hotspots house most undiscovered plant species. Proc Natl Acad Sci U S A 108:13171–13176

Callmander MW, Schatz GE, Lowry PP (2005) IUCN red list assessment and the global strategy for plant conservation: taxonomists must act now. Taxon 54:1047–1050

Godfray HCJ (2002) Challenges for taxonomy - the discipline will have to reinvent itself if it is to survive and flourish. Nature 417:17–19

Funk VA (2006) Floras: a model for biodiversity studies or a thing of the past? Taxon 55:581–588

Wheeler QD, Raven PH, Wilson EO (2004) Taxonomy: impediment or expedient? Science 303:285–285

Wheeler QD, Knapp S et al (2012) Mapping the biosphere: exploring species to understand the origin, organization and sustainability of biodiversity. Syst Biodivers 10(1):1–20

Ebach MC, Valdecasas AG, Wheeler QD (2011) Impediments to taxonomy and users of taxonomy: accessibility and impact evaluation. Cladistics 27:550–557

Ronquist F, Gardenfors U (2003) Taxonomy and biodiversity inventories: time to deliver. Trends Ecol Evol 18:269–270

Joly CA (2006) Taxonomy: programmes developing in the south too. Nature 440:24–24

Schatz GE, Lowry PP, Ramisamihantanirina A (1998) Takhtajania perrieri rediscovered. Nature 391:133–134

Jones WG, Hill KD, Allen JM (1995) Wollemia nobilis , a new living Australian genus and species in the Araucariaceae. Telopea 6:173–176

Mabberley DJ (2009) Exploring Terra Incognita. Science 324:472–472

Thulin M (2007) Acacia fumosa sp nov (Fabaceae) from eastern Ethiopia. Nord J Bot 25:272–274

Dransfield J, Rakotoarinivo M, Baker WJ, Bayton RP, Fisher JB, Horn JW, Leroy B, Metz X (2008) A new coryphoid palm genus from Madagascar. Bot J Linn Soc 156:79–91

Agnarsson I, Kuntner M (2007) Taxonomy in a changing world: seeking solutions for a science in crisis. Syst Biol 56:531–539

Crisci JV (2006) One-dimensional systematist: perils in a time of steady progress. Syst Bot 31:217–221

Joppa LN, Roberts DL, Pimm SL (2011) The population ecology and social behaviour of taxonomists. Trends Ecol Evol 26:551–553

Rodman JE, Cody JH (2003) The taxonomic impediment overcome: NSF's partnerships for enhancing expertise in taxonomy (PEET) as a model. Syst Biol 52:428–435

Bebber DP et al (2012) Big hitting collectors make massive and disproportionate contribution to the discovery of plant species. Proc R Soc Lond B Biol Sci 279(1736):2269–2274

Thiers B (2019) Index herbarium: a global directory of public herbaria and associated staff. New York Botanical Garden’s Virtual Herbarium. http://sweetgum.nybg.org/ih/

Bebber DP, Carine MA, Wood JRI, Wortley AH, Harris DJ, Prance GT, Davidse G, Paige J, Pennington TD, Robson NKB, Scotland RW (2010) Herbaria are a major frontier for species discovery. Proc Natl Acad Sci U S A 107:22169–22171

Fontaine B, Perrard A, Bouchet P (2012) 21 years of shelf life between discovery and description of new species. Curr Biol 22(22):R943–R944. https://doi.org/10.1016/j.cub.2012.10.029

Le Bras G, Pignal M, Jeanson ML, Muller S, Aupic C, Carré B, Flament G, Gaudeul M, Gonçalves C, Invernón VR, Jabbour F, Lerat E, Lowry PP, Offroy B, Pérez Pimparé E, Poncy O, Rouhan G, Haevermans T (2017) The French Muséum national d’histoire naturelle vascular plant herbarium collection dataset. Scientific Data 4:170016

Godfray HCJ, Clark BR, Kitching IJ, Mayo SJ, Scoble MJ (2007) The web and the structure of taxonomy. Syst Biol 56:943–955

Knapp S, McNeill J, Turland NJ (2011) Changes to publication requirements made at the XVIII International Botanical Congress in Melbourne - what does e-publication mean for you? PhytoKeys (6):5–11

Nicolson N, Challis K, Tucker A, Knapp S (2017) Impact of e-publication changes in the International Code of Nomenclature for algae, fungi and plants (Melbourne Code, 2012) - did we need to “run for our lives”? BMC Evol Biol 17:116. https://doi.org/10.1186/s12862-017-0961

Hebert PDN, Gregory TR (2005) The promise of DNA barcoding for taxonomy. Syst Biol 54:852–859

Savolainen V, Cowan RS, Vogler AP, Roderick GK, Lane R (2005) Towards writing the encyclopaedia of life: an introduction to DNA barcoding. Philos Trans R Soc B Biol Sci 360:1805–1811

Wiens JJ (2007) Species delimitation: new approaches for discovering diversity. Syst Biol 56:875–878

Pannell JR (2009) Mating-system evolution: succeeding by celibacy. Curr Biol 19:R983–R985

Hood ME, Antonovics J (2003) Plant species descriptions show signs of disease. Proc R Soc Lond B Biol Sci 270:S156–S158

Duminil J, Di Michele M (2009) Plant species delimitation: a comparison of morphological and molecular markers. Plant Biosystems 143:528–542

Bickford D, Lohman DJ, Sodhi NS, Ng PKL, Meier R, Winker K, Ingram KK, Das I (2007) Cryptic species as a window on diversity and conservation. Trends Ecol Evol 22:148–155

Grundt HH, Kjolner S, Borgen L, Rieseberg LH, Brochmann C (2006) High biological species diversity in the arctic flora. Proc Natl Acad Sci U S A 103:972–975

Pillon Y, Hopkins HCF, Munzinger J, Amir H, Chase MW (2009) Cryptic species, gene recombination and hybridization in the genus Spiraeanthemum (Cunoniaceae) from New Caledonia. Bot J Linn Soc 161:137–152

Dulvy NK, Reynolds JD (2009) BIODIVERSITY skates on thin ice. Nature 462:417–417

Robertson A, Newton AC, Ennos RA (2004) Multiple hybrid origins, genetic diversity and population genetic structure of two endemic Sorbus taxa on the Isle of Arran, Scotland. Mol Ecol 13:123–134

Squirrell J, Hollingsworth PM, Bateman RM, Tebbitt MC, Hollingsworth ML (2002) Taxonomic complexity and breeding system transitions: conservation genetics of the Epipactis leptochila complex (Orchidaceae). Mol Ecol 11:1957–1964

van Dijk PJ (2003) Ecological and evolutionary opportunities of apomixis: insights from Taraxacum and Chondrilla. Philos Trans R Soc Lond Ser B Biol Sci 358:1113–1121

Ennos RA, French GC, Hollingsworth PM (2005) Conserving taxonomic complexity. Trends Ecol Evol 20:164–168

Ennos RA, Whitlock R, Fay MF, Jones B, Neaves LE, Payne R, Taylor I, De Vere N, Hollingsworth PM (2012) Process-based species action plans: an approach to conserve contemporary evolutionary processes that sustain diversity in taxonomically complex groups. Bot J Linn Soc 168:194–203

Li FW, Tan BC, Buchbender V, Moran RC, Rouhan G, Wang CN, Quandt D (2009) Identifying a mysterious aquatic fern gametophyte. Plant Syst Evol 281:77–86

Van Deynze A, Stoffel K (2006) High-throughput DNA extraction from seeds. Seed Sci Technol 34:741–745

Asif MJ, Cannon CH (2005) DNA extraction from processed wood: a case study for the identification of an endangered timber species (Gonystylus bancanus). Plant Mol Biol Report 23:185–192

Colpaert N, Cavers S, Bandou E, Caron H, Gheysen G, Lowe AJ (2005) Sampling tissue for DNA analysis of trees: trunk cambium as an alternative to canopy leaves. Silvae Genetica 54:265–269

Rachmayanti Y, Leinemann L, Gailing O, Finkeldey R (2006) Extraction, amplification and characterization of wood DNA from Dipterocarpaceae. Plant Mol Biol Report 24:45–55

Tibbits JFG, McManus LJ, Spokevicius AV, Bossinger G (2006) A rapid method for tissue collection and high-throughput isolation of genomic DNA from mature trees. Plant Mol Biol Report 24:81–91

Novaes RML, Rodrigues JG, Lovato MB (2009) An efficient protocol for tissue sampling and DNA isolation from the stem bark of Leguminosae trees. Genet Mol Res 8:86–96

Deguilloux MF, Pemonge MH, Petit RJ (2002) Novel perspectives in wood certification and forensics: dry wood as a source of DNA. Proc R Soc B Biol Sci 269:1039–1046

Hiiesalu I, Opik M, Metsis M, Lilje L, Davison J, Vasar M, Moora M, Zobel M, Wilson SD, Partel M (2012) Plant species richness belowground: higher richness and new patterns revealed by next-generation sequencing. Mol Ecol 21:2004–2016

Kesanakurti PR, Fazekas AJ, Burgess KS, Percy DM, Newmaster SG, Graham SW, Barrett SCH, Hajibabaei M, Husband BC (2011) Spatial patterns of plant diversity below-ground as revealed by DNA barcoding. Mol Ecol 20:1289–1302

Dunn CP (2003) Keeping taxonomy based in morphology. Trends Ecol Evol 18:270–271

Santos LM, Faria LRR (2011) The taxonomy’s new clothes: a little more about the DNA-based taxonomy. Zootaxa 3025:66–68

Schaefer H, Carine MA, Rumsey FJ (2011) From European priority species to invasive weed: Marsilea azorica (Marsileaceae) is a misidentified alien. Syst Bot 36:845–853

Launert GOE, Paiva JAR (1983) Iconographia selecta florae Azoricae. Coimbra 2:159

Lipscomb D, Platnick N, Wheeler Q (2003) The intellectual content of taxonomy: a comment on DNA taxonomy. Trends Ecol Evol 18:65–66

Sites JW, Marshall JC (2004) Operational criteria for delimiting species. Annu Rev Ecol Evol Syst 35:199–227

Stace CA (2005) Plant taxonomy and biosystematics - does DNA provide all the answers? Taxon 54:999–1007

Linder CR, Rieseberg LH (2004) Reconstructing patterns of reticulate evolution UN plants. Am J Bot 91:1700–1708

Vriesendorp B, Bakker FT (2005) Reconstructing patterns of reticulate evolution in angiosperms: what can we do? Taxon 54:593–604

Egan AN, Schlueter J, Spooner DM (2012) Applications of next-generation sequencing in plant biology. Am J Bot 99:175–185

Harrison N, Kidner CA (2011) Next-generation sequencing and systematics: what can a billion base pairs of DNA sequence data do for you? Taxon 60:1552–1566

Straub SCK, Parks M, Weitemier K, Fishbein M, Cronn RC, Liston A (2012) Navigating the tip of the genomic iceberg: next-generation sequencing for plant systematics. Am J Bot 99:349–364

Bakker FT (2015) DNA sequences from plant herbarium tissue (Chapter 8). In: Hörandl E, Appelhans MS (eds) Next-generation sequencing in plant systematics. International Association for Plant Taxonomy (IAPT). https://doi.org/10.14630/000009

Richardson JE, Pennington RT, Pennington TD, Hollingsworth PM (2001) Rapid diversification of a species-rich genus of neotropical rain forest trees. Science 293:2242–2245

Baldwin BG, Sanderson MJ (1998) Age and rate of diversification of the Hawaiian silversword alliance (Compositae). Proc Natl Acad Sci U S A 95:9402–9406

Wang AL, Yang MH, Liu JQ (2005) Molecular phylogeny, recent radiation and evolution of gross morphology of the rhubarb genus Rheum (Polygonaceae) inferred from chloroplast DNA trnL-F sequences. Ann Bot 96:489–498

Hodges SA, Arnold ML (1994) Columbines - a geographically widespread species flock. Proc Natl Acad Sci U S A 91:5129–5132

Linder HP (2008) Plant species radiations: where, when, why? Philos Trans R Soc B Biol Sci 363:3097–3105

Tautz D, Arctander P, Minelli A, Thomas RH, Vogler AP (2003) A plea for DNA taxonomy. Trends Ecol Evol 18:70–74

Hey J, Pinho C (2012) Population genetics and objectivity in species diagnosis. Evolution 66:1413–1429

Knowles L, Carstens B (2007) Delimiting species without monophyletic gene trees. Syst Biol 56:887–895

Staats M, Cuenca A, Richardson JE, Vrielink-van Ginkel R, Petersen G, Seberg O, Bakker FT (2011) DNA damage in plant herbarium tissue. PLoS One 6:e28448

Seberg O, Humphries CJ, Knapp S, Stevenson DW, Petersen G, Scharff N, Andersen NM (2003) Shortcuts in systematics? A commentary on DNA-based taxonomy. Trends Ecol Evol 18:63–65

Lister DL, Bower MA, Howe CJ, Jones MK (2008) Extraction and amplification of nuclear DNA from herbarium specimens of emmer wheat: a method for assessing DNA preservation by maximum amplicon length recovery. Taxon 57:254–258

Wandeler P, Hoeck PEA, Keller LF (2007) Back to the future: museum specimens in population genetics. Trends Ecol Evol 22:634–642

Cozzolino S, Cafasso D, Pellegrino G, Musacchio A, Widmer A (2007) Genetic variation in time and space: the use of herbarium specimens to reconstruct patterns of genetic variation in the endangered orchid Anacamptis palustris. Conserv Genet 8:629–639

Erkens RHJ, Cross H, Maas JW, Hoenselaar K, Chatrou LW (2008) Assessment of age and greenness of herbarium specimens as predictors for successful extraction and amplification of DNA. Blumea 53:407–428

Drabkova L, Kirschner J, Vlcek C (2002) Comparison of seven DNA extraction and amplification protocols in historical herbarium specimens of Juncaceae. Plant Mol Biol Report 20:161–175

Jankowiak K, Buczkowska K, Szweykowska-Kulinska Z (2005) Successful extraction of DNA from 100-year-old herbarium specimens of the liverwort Bazzania trilobata. Taxon 54:335–336

Korpelainen H, Pietilainen M (2008) Effort to reconstruct past population history in the fern Blechnum spicant. J Plant Res 121:293–298

Savolainen V, Cuenoud P, Spichiger R, Martinez MDP, Crevecoeur M, Manen JF (1995) The use of herbarium specimens in DNA Phylogenetics - evaluation and improvement. Plant Syst Evol 197:87–98

Ribeiro RA, Lovato MB (2007) Comparative analysis of different DNA extraction protocols in fresh and herbarium specimens of the genus Dalbergia. Genet Mol Res 6:173–187

CAS   PubMed   Google Scholar  

Andreasen K, Manktelow M, Razafimandimbison SG (2009) Successful DNA amplification of a more than 200-year-old herbarium specimen: recovering genetic material from the Linnaean era. Taxon 58:959–962

Ames M, Spooner DM (2008) DNA from herbarium specimens settles a controversy about origins of the European potato. Am J Bot 95:252–257

Walters C, Reilley AA, Reeves PA, Baszczak J, Richards CM (2006) The utility of aged seeds in DNA banks. Seed Sci Res 16:169–178

Alves RJV, Machado MD (2007) Is classical taxonomy obsolete? Taxon 56:287–288

DeSalle R, Egan MG, Siddall M (2005) The unholy trinity: taxonomy, species delimitation and DNA barcoding. Philos Trans R Soc B Biol Sci 360:1905–1916

DeSalle R (2006) Species discovery versus species identification in DNA barcoding efforts: response to Rubinoff. Conserv Biol 20:1545–1547

Schlick-Steiner BC, Steiner FM, Seifert B, Stauffer C, Christian E, Crozier RH (2010) Integrative taxonomy: a multisource approach to exploring biodiversity. Annu Rev Entomol 55:421–438

Dayrat B (2005) Towards integrative taxonomy. Biol J Linn Soc 85:407–415

Wheeler QD (2005) Losing the plot: DNA “barcodes” and taxonomy. Cladistics 21:405–407

Corney DPA, Clark JY, Tang HT, Wilkin P (2012) Automatic extraction of leaf characters from herbarium specimens. Taxon 61(1):231–244

Raxworthy CJ, Ingram CM, Rabibisoa N, Pearson RG (2007) Applications of ecological niche modeling for species delimitation: a review and empirical evaluation using day geckos (Phelsuma) from Madagascar. Syst Biol 56:907–923

Rissler LJ, Apodaca JJ (2007) Adding more ecology into species delimitation: ecological niche models and phylogeography help define cryptic species in the black salamander (Aneides flavipunctatus). Syst Biol 56:924–942

Wiens JJ, Graham CH (2005) Niche conservatism: integrating evolution, ecology, and conservation biology. Annu Rev Ecol Evol Syst 36:519–539

Brautigam A, Gowik U (2010) What can next generation sequencing do for you? Next generation sequencing as a valuable tool in plant research. Plant Biol 12:831–841

Cronn R, Knaus BJ, Liston A, Maughan PJ, Parks M, Syring JV, Udall J (2012) Targeted enrichment strategies for next-generation plant biology. Am J Bot 99:291–311

Rodriguez-Fernandez JI, De Carvalho CJB, Pasquini C, De Lima KMG, Moura MO, Arizaga GGC (2011) Barcoding without DNA? Species identification using near infrared spectroscopy. Zootaxa 46–54

Munck L, Jespersen BM, Rinnan A, Seefeldt HF, Engelsen MM, Norgaard L, Engelsen SB (2010) A physiochemical theory on the applicability of soft mathematical models-experimentally interpreted. J Chemom 24:481–495

Cruickshank RH, Munck L (2011) It’s barcoding Jim, but not as we know it. Zootaxa 2933:55–56

Andres-Sanchez S, Rico E, Herrero A, Santos-Vicente M, Martinez-Ortega MM (2009) Combining traditional morphometrics and molecular markers in cryptic taxa: towards an updated integrative taxonomic treatment for Veronica subgenus Pentasepalae (Plantaginaceae sensu APG II) in the western Mediterranean. Bot J Linn Soc 159:68–87

Schlick-Steiner BC, Seifert B, Stauffer C, Christian E, Crozier RH, Steiner FM (2007) Without morphology, cryptic species stay in taxonomic crypsis following discovery. Trends Ecol Evol 22:391–392

Blaxter M, Mann J, Chapman T, Thomas F, Whitton C, Floyd R, Abebe E (2005) Defining operational taxonomic units using DNA barcode data. Philos Trans R Soc B Biol Sci 360:1935–1943

Markmann M, Tautz D (2005) Reverse taxonomy: an approach towards determining the diversity of meiobenthic organisms based on ribosomal RNA signature sequences. Philos Trans R Soc B Biol Sci 360:1917–1924

Pleijel F, Jondelius U, Norlinder E, Nygren A, Oxelman B, Schander C, Sundberg P, Thollesson M (2008) Phylogenies without roots? A plea for the use of vouchers in molecular phylogenetic studies. Mol Phylogenet Evol 48:369–371

Puillandre N, Bouchet P, Boisselier-Dubayle MC, Brisset J, Buge B, Castelin M, Chagnoux S, Christophe T, Corbari L, Lambourdiere J, Lozouet P, Marani G, Rivasseau A, Silva N, Terryn Y, Tillier S, Utge J, Samadi S (2012) New taxonomy and old collections: integrating DNA barcoding into the collection curation process. Mol Ecol Resour 12:396–402

Gemeinholzer B, Bachmann K (2005) Examining morphological and molecular diagnostic character states of Cichorium intybus L. (Asteraceae) and C-spinosum L. Plant Syst Evol 253:105–123

Bacon CD, McKenna MJ, Simmons MP, Wagner WL (2012) Evaluating multiple criteria for species delimitation: an empirical example using Hawaiian palms (Arecaceae: Pritchardia). BMC Evol Biol 12:23

Barrett CF, Freudenstein JV (2011) An integrative approach to delimiting species in a rare but widespread mycoheterotrophic orchid. Mol Ecol 20:2771–2786

Koffi KG, Heuertz M, Doumenge C, Onana JM, Gavory F, Hardy OJ (2010) A combined analysis of morphological traits, chloroplast and nuclear DNA sequences within Santiria trimera (Burseraceae) suggests several species following the biological species concept. Plant Ecol Evol 143:160–169

Ley AC, Hardy OJ (2010) Species delimitation in the central African herbs Haumania (Marantaceae) using georeferenced nuclear and chloroplastic DNA sequences. Mol Phylogenet Evol 57:859–867

Meudt HM, Lockhart PJ, Bryant D (2009) Species delimitation and phylogeny of a New Zealand plant species radiation. BMC Evol Biol 9:111

Article   PubMed   PubMed Central   CAS   Google Scholar  

Schmidt-Lebuhn AN (2007) Using amplified fragment length polymorphism (AFLP) to unravel species relationships and delimitations in Minthostachys (Labiatae). Bot J Linn Soc 153:9–19

Zeng YF, Liao WJ, Petit RJ, Zhang DY (2010) Exploring species limits in two closely related Chinese oaks. PLoS One 5:e15529

Rieseberg LH, Troy TE, Baack EJ (2006) The nature of plant species. Nature 440:524–527

de Queiroz K (2005) Ernst Mayr and the modern concept of species. Proc Natl Acad Sci U S A 102:6600–6607

de Queiroz K (2007) Species concepts and species delimitation. Syst Biol 56(6):879–886

Wright S (1940) The statistical consequences of Mendelian heredity in relation to speciation. In: Huxley J (ed) The new systematics. Oxford university press, London, pp 161–183

Mayr E (1942) Systematics and the origin of species. Columbia university press, New York

Dobzhansky T (1950) Mendelian populations and their evolution. Am Nat 84:401–418

Poulton EB (1904) What is a species? Proc Entomol Soc Lond 1903:lxxviicxvi

Dobzhansky T (1970) Genetics of the evolutionary process. Columbia University Press, New York

Sokal RR, Crovello TJ (1970) The biological species concept: a critical evaluation. Am Nat 104:107–123

Van Valen L (1976) Ecological species, multispecies, and oaks. Taxon 25:233–239

Simpson GG (1951) The species concept. Evolution 5:285–298

Wiley EO (1978) The evolutionary species concept reconsidered. Syst Zool 21:17–26

Cracraft J (1989) Speciation and its ontology: the empirical consequences of alternative species concepts for understanding patterns and processes of differentiation. In: Otte D, Endler JA (eds) Speciation and its consequences. Sinauer Associates, Sunderland, pp 28–59

Rosen DE (1979) Fishes from the uplands and intermontane basins of Guatemala: revisionary studies and comparative geography. Bull Am Mus Nat Hist 162:267–376

Donoghue MJ (1985) A critique of the biological species concept and recommendations for a phylogenetic alternative. Bryologist 88:172–181

Mishler BD (1985) The morphological, developmental, and phylogenetic basis of species concepts in bryophytes. Bryologist 88:207–214

Baum DA, Shaw KL (1995) Genealogical perspectives on the species problem. In: Hoch PC, Stephenson AG (eds) Experimental and molecular approaches to plant biosystematics. Missouri Botanical Garden, St. Louis, pp 289–303

Mallet J (1995) A species definition for the modern synthesis. Trends Ecol Evol 10:294–299

Templeton AR (1998) Species and speciation: geography, population structure, ecology, and gene trees. In: Howard DJ, Berlocher SH (eds) Endless forms: species and speciation. Oxford university press, New York, pp 32–43

Sites JW, Marshall JC (2003) Delimiting species: a renaissance issue in systematic biology. Trends Ecol Evol 18(9):462–470

Download references

Author information

Authors and affiliations.

Institut de Systématique, Evolution, Biodiversité (ISYEB), Muséum national d’Histoire naturelle, Sorbonne Université, Ecole Pratique des Hautes Etudes, Université des Antilles, CNRS, Paris, France

Germinal Rouhan & Myriam Gaudeul

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Germinal Rouhan .

Editor information

Editors and affiliations.

UMR PVBMT, Universite de la Reunion, St Pierre, Réunion, France

Pascale Besse

Rights and permissions

Reprints and permissions

Copyright information

© 2021 Springer Science+Business Media, LLC, part of Springer Nature

About this protocol

Rouhan, G., Gaudeul, M. (2021). Plant Taxonomy: A Historical Perspective, Current Challenges, and Perspectives. In: Besse, P. (eds) Molecular Plant Taxonomy. Methods in Molecular Biology, vol 2222. Humana, New York, NY. https://doi.org/10.1007/978-1-0716-0997-2_1

Download citation

DOI : https://doi.org/10.1007/978-1-0716-0997-2_1

Published : 11 December 2020

Publisher Name : Humana, New York, NY

Print ISBN : 978-1-0716-0996-5

Online ISBN : 978-1-0716-0997-2

eBook Packages : Springer Protocols

• Publish with us

Policies and ethics

• Find a journal
• Track your research

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals

Taxonomy articles from across Nature Portfolio

Taxonomy is the classification and description of living organisms. It includes the naming and defining of species, and the collation of data about their biology and biogeography.

Latest Research and Reviews

Circumtropical distribution and cryptic species of the meiofaunal enteropneust Meioglossus (Harrimaniidae, Hemichordata)

• Éloïse Defourneaux
• Maria Herranz
• Katrine Worsaae

Eocene amber provides the first fossil record and bridges distributional gap in the rare genus Robsonomyia  (Diptera: Keroplatidae)

• Alicja Pełczyńska
• Wiesław Krzemiński
• Agnieszka Soszyńska

The Miocene primate Pliobates is a pliopithecoid

Pliobates cataloniae is a small-bodied Miocene catarrhine primate with unclear systematic status. Here, the authors present additional dental remains from this species, conducting cladistic analyses that indicate it is a pliopithecoid convergent with apes in elbow and wrist morphology.

• Florian Bouchet
• Clément Zanolli
• David M. Alba

Low coverage of species constrains the use of DNA barcoding to assess mosquito biodiversity

• Maurício Moraes Zenker
• Tatiana Pineda Portella
• Pedro Manoel Galetti Jr.

Evolution, types, and distribution of flight control devices on wings and elytra in bark beetles

• Jakub Białkowski
• Robert Rossa
• Jakub Goczał

Oxytoxaceae are prorocentralean rather than peridinialean dinophytes and taxonomic clarification of heterotrophic Oxytoxum lohmannii (≡ “ Amphidinium ” crassum ) by epitypification

• Marc Gottschling
• Stephan Wietkamp
• Urban Tillmann

News and Comment

Change in biological nomenclature is overdue and possible.

• Mirjana Roksandic
• Charles Musiba
• Christopher J. Bae

John Macfarlane was the first to recognize Eukaryota as a group

• Yegor Shɨshkin

Science depends on nomenclature, but nomenclature is not science

The International Committee on Systematics of Prokaryotes (ICSP) has recently altered long-standing phylum names and given no guidance for taxonomy of uncultured or imperfectly cultured archaea and bacteria, disrupting progress towards a universal system of microbial taxonomy. Inclusion of new members into ICSP may help it to keep up to date.

• Karen G. Lloyd
• Guillaume Tahon

Decolonizing botanical genomics

By sieving through the plant genomic literature for the last 20 years, a study uncovered a disconnection between the research locales and plants’ native ranges. Colonialism, both past and present, might be behind this disparity.

Law, ethics, gems and fossils in Myanmar amber

• Paul M. Barrett
• Zerina Johanson
• Sarah L. Long

Taxonomy must engage with new technologies and evolve to face future challenges

• Michael C. Orr
• Rafael R. Ferrari
• Chao-Dong Zhu

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Academia.edu no longer supports Internet Explorer.

To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to  upgrade your browser .

•  We're Hiring!
•  Help Center

Plant Taxonomy

• Most Cited Papers
• Most Downloaded Papers
• Newest Papers
• Save to Library
• Last »
• Plant Taxonomy (Taxonomy) Follow Following
• Photocatalysts Follow Following
• Botany Follow Following
• Metal Nanoparticles Follow Following
• Threatened Plants Follow Following
• Endemic Plants Follow Following
• Plant Systematics Follow Following
• medicinal & Aromatic plants Follow Following
• Field botany Follow Following
• Ethnobotany Follow Following

Enter the email address you signed up with and we'll email you a reset link.

• Academia.edu Publishing
•   We're Hiring!
•   Help Center
• Find new research papers in:
• Health Sciences
• Earth Sciences
• Cognitive Science
• Mathematics
• Computer Science
• Academia ©2024

• Get involved

• Taxonomy of climate-attributable loss and damage and scalable responses related to DRR, health and human mobility pdf (1.5 MB)

Taxonomy of climate-attributable loss and damage and scalable responses related to DRR, health and human mobility

April 23, 2024, related publications, publications, small island digital states: how digital can catalyse sid....

Small Island Developing States (SIDS) are increasingly becoming Small Island Digital States. Digital is positively impacting lives and livelihoods across SIDS –...

State of Youth Entrepreneurship Ecosystem in Pakistan

The State of Youth Entrepreneurship in Pakistan reflects the country’s growing momentum in its entrepreneurship ecosystem since 2012, marked by an increasing nu...

Strengthening Integral Local Development

The overall objective of the project is to contribute to Timor-Leste's sustainable development. More specifically, it aims to support the deconcentration and De...

Strengthening Integral Local Government

The project is designed to complement and support other ongoing decentralization/local government projects and programmes, this UNDP project will support the Mi...

Gender equality barometer of Bosnia and Herzegovina

Conducted more than twenty years after the first Barometer, this comprehensive research endeavor delves into the intricate fabric of societal perceptions, shedd...

The Labs 2.0: Evolving into a global research and develop...

The UNDP Accelerator Labs Network 2023 Annual Report, “The Labs 2.0: Evolving into a global research and development capability for the SDGs,” highlights how th...