The Cinderella discipline: morphometrics and their use in botanical classification

Journal article

Between the 1960s and the present day, the use of morphology in plant taxonomy suffered a major decline, in part driven by the apparent superiority of DNA-based approaches to data generation. However, in recent years computer image recognition has re-kindled the interest in morphological techniques. Linear or geometric morphometric approaches have been employed to distinguish and classify a wide variety of organisms; each has strengths and weaknesses. Here we review these approaches with a focus on plant classification and present a case for the combination of morphometrics with statistical/machine learning. There are many classification techniques available for biological analysis and selecting the most appropriate is not trivial. Performance should be evaluated using standardized metrics such as accuracy, sensitivity and specificity. The gathering and storage of high-resolution images, combined with the processing power of desktop computers, makes morphometric approaches practical as a time- and cost-efficient way of non-destructive identification of plant samples.

Authors: Christodoulou, M; Clark, J; Culham, A

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Oxford University Press
• Journal: Botanical Journal of the Linnean Society
• Volume: 194
• Issue: 4
• Pages: 385–396
• Publication date: 2020-09-03
• Acceptance date: 2020-06-15
• DOI: 10.1093/botlinnean/boaa055
• EISSN: 1095-8339
• ISSN: 0024-4074
• Language: English
• Keywords: neural networks; machine learning; linear morphometrics; plant taxonomy; identification; classification; statistical learning; geometric morphometrics; FFR