Don Kirkup

Comparative morphology of populations of Monstera Adans. (Araceae) from natural forest fragments in Northeast Brazil using elliptic Fourier Analysis of leaf outlines

SummaryA comparative study of the leaf outline morphometrics of Monstera adansonii var. klotzschiana, M. adansonii var. laniata and M. praetermissa was carried out. The study focused on populations in isolated montane humid (brejo) forests of Ceará state in Northeast Brazil and compared them with populations from Amazonia and the Brazilian Atlantic Forest. Digitised outlines were prepared from a total of 1,695 field-collected leaf images from 20 populations, and elliptic Fourier analysis was used to generate matrices of coefficients, from which six shape variables (principal components) were extracted using Principal Components Analysis. Intra-population variability and inter-population differences were analysed with multivariate distance methods. Separate analyses were carried out for each of three leaf size classes (juvenile, submature, mature) because of the strong heteroblasty typical of this genus. Juvenile leaves were the least variable size class within populations of M. adansonii var. klotzschiana. The shape variables expressed very similar types of variation in all three size classes. The Ceará brejo populations of M. adansonii var. klotzschiana showed significant differences between mature leaf outlines in all pairwise comparisons; the Pacatuba population was the most distinct. The Ceará populations did not cluster together exclusively. In all three size classes, populations clustered together into their taxonomic groups, most clearly so in mature leaves. No correlation between morphological and geographic distance matrices was found, nor between morphological and molecular distance. The study showed that leaf outline shape is a practicable and useful quantitative trait for studying morphological variability at species, varietal and population levels.

Enriched biodiversity data as a resource and service

Abstract Background: Recent years have seen a surge in projects that produce large volumes of structured, machine-readable biodiversity data. To make these data amenable to processing by generic, open source “data enrichment” workflows, they are increasingly being represented in a variety of standards-compliant interchange formats. Here, we report on an initiative in which software developers and taxonomists came together to address the challenges and highlight the opportunities in the enrichment of such biodiversity data by engaging in intensive, collaborative software development: The Biodiversity Data Enrichment Hackathon. Results: The hackathon brought together 37 participants (including developers and taxonomists, i.e. scientific professionals that gather, identify, name and classify species) from 10 countries: Belgium, Bulgaria, Canada, Finland, Germany, Italy, the Netherlands, New Zealand, the UK, and the US. The participants brought expertise in processing structured data, text mining, development of ontologies, digital identification keys, geographic information systems, niche modeling, natural language processing, provenance annotation, semantic integration, taxonomic name resolution, web service interfaces, workflow tools and visualisation. Most use cases and exemplar data were provided by taxonomists. One goal of the meeting was to facilitate re-use and enhancement of biodiversity knowledge by a broad range of stakeholders, such as taxonomists, systematists, ecologists, niche modelers, informaticians and ontologists. The suggested use cases resulted in nine breakout groups addressing three main themes: i) mobilising heritage biodiversity knowledge; ii) formalising and linking concepts; and iii) addressing interoperability between service platforms. Another goal was to further foster a community of experts in biodiversity informatics and to build human links between research projects and institutions, in response to recent calls to further such integration in this research domain. Conclusions: Beyond deriving prototype solutions for each use case, areas of inadequacy were discussed and are being pursued further. It was striking how many possible applications for biodiversity data there were and how quickly solutions could be put together when the normal constraints to collaboration were broken down for a week. Conversely, mobilising biodiversity knowledge from their silos in heritage literature and natural history collections will continue to require formalisation of the concepts (and the links between them) that define the research domain, as well as increased interoperability between the software platforms that operate on these concepts.

Elliptic Fourier Analysis of leaf outline shape in forest fragment populations of Anthurium sinuatum and A. pentaphyllum (Araceae) from Northeast Brazil

SummaryA field-based comparative study of leaflet outline shape in five populations of Anthurium sinuatum Benth. ex Schott from Amazonia and humid brejo forests of Ceará, Northeast Brazil, and ten populations of A. pentaphyllum (Aubl.) G. Don var. pentaphyllum from Amazonia and the Brazilian Atlantic forest is reported. 1,120 leaflets from 561 individuals in 15 populations were sampled. Using images, leaflet outlines were captured by digitisation and subjected to Elliptic Fourier Analysis (EFA) to produce matrices of Fourier coefficients based on 40 harmonics; lateral and central leaflets were analysed separately. Twelve shape variables, extracted from the Fourier coefficient matrices, described the leaflet outlines of 531 individuals. Between-population relationships were estimated using cluster analysis, Canonical Variates Analysis and non-parametric MANOVA. Amazonian populations showed the greatest within-population variability. In A. sinuatum the Pacatuba population was the most distinct in Ceará and the Pará population (Amazonia) was not significantly different from those at Ibiapaba and Maranguape (both Ceará). In A. pentaphyllum 80% of population pairs were significantly different and three significantly different groups of populations were observed. Although the populations of A. pentaphyllum displayed considerable variation, the morphological patterns did not correlate with previously reported molecular ones. The two species were found to be slightly but significantly different on leaf outline data.

Elliptic Fourier Analysis of leaf outlines in five species of Heteropsis (Araceae) from the Reserva Florestal Adolpho Ducke, Manaus, Amazonas, Brazil

SummaryA pilot study of leaf outline morphometrics was carried out on populations of five species of Heteropsis located in the Reserva Florestal Adolpho Ducke near Manaus, Amazonas, Brazil: H. flexuosa (Kunth) G. S. Bunting, H. macrophylla A. C. Sm., H. spruceana Schott, H. steyermarkii G. S. Bunting and H. tenuispadix G. S. Bunting. The aim of the study was to investigate quantitative methods for discriminating species within a local area based on vegetative morphology in a genus where fertile parts are often difficult to find in the field; this study focussed on leaf outline shape. Using digital images of 347 leaves, outlines were captured as coordinates using the TpsDig software and analysed using Elliptic Fourier Analysis (EFA). Principal Component Analysis (PCA) was used to reduce the 160 elliptic Fourier coefficient descriptors to a smaller number of independent shape variables corresponding to the principal component axes. The first nineteen shape variables (constrained by the smallest species sample, N = 20) were then subjected to Linear Discriminant Analysis (LDA) to find new axes that would discriminate the five species to the maximum extent. The first six shape variables were visualised by a reification procedure in order to illustrate their characteristic contribution to the total shape variation within all five species. The results showed that the mean shapes of all five species were significantly different, but shape variation within each species overlapped with the others. Percentage assignment of individuals to their correct species was encouragingly high given that only outline shape was used, but was not high enough overall to provide reliable identification. Elimination of one species (H. steyermarkii) using easily observed qualitative vegetative characters improved discrimination of the remaining four. The investigation of new approaches to identification is potentially valuable for conservation of natural populations — the root fibre of Heteropsis is extracted from primary forest and is a valuable non-timber forest product that forms the basis for an important local industry in Amazonia.

Variation of theanine, phenolic, and methylxanthine compounds in 21 cultivars of Camellia sinensis harvested in different seasons

This is the first study to use chemometric methods to differentiate among 21 cultivars of Camellia sinensis from China and between leaves harvested at different times of the year using 30 compounds implicated in the taste and quality of tea. Unique patterns of catechin derivatives were observed among cultivars and across harvest seasons. C. sinensis var. pubilimba (You 510) differed from the cultivars of C. sinensis var. sinensis, with higher levels of theobromine, (+)-catechin, gallocatechin, gallocatechin gallate and theasinensin B, and lower levels of (-)-epicatechin, (-)-epigallocatechin (EGC) and (-)-epigallocatechin gallate (EGCG), respectively. Three cultivars of C. sinensis var. sinensis, Fuyun 7, Qiancha 7 and Zijuan contained significantly more caffeoylquinic acids than others cultivars. A Linear Discriminant Analysis model based on the abundance of 12 compounds was able to discriminate amongst all 21 tea cultivars. Harvest time impacted the abundance of EGC, theanine and afzelechin gallate.

Machine Learning for Plant Leaf Analysis

One of the key challenges for automated analysis of plant leaves lies in the range of variation presented by a species and even by a single plant (Fig. 2.1). As well as the natural variation to be expected from any organic object, the variation of a leaf can arise from a number of sources, for example, its age and developmental stage. Shape varies during development, with early growth phases occurring primarily length-wise and increase in width coming later. In some species with lobed leaves, the leaf lobes are not apparent until after a certain stage in development. In others, like many Eucalyptus taxa, the leaves of young shoots are morphologically very distinct from those of mature ones. Margin characteristics such as teeth may not develop until the leaf has reached full size, often appearing first near the apex and then gradually extending further back towards the insertion point. Pigmentation often changes as the leaf develops.

Morphometrics: A Brief Review

In the Introduction (Chap. 1) an account was given of the rationale for using the morphology of leaves as the source of raw material for automatic systems for the identification of plant species. In this chapter we focus on how morphological features have been used in statistical analysis to define taxa like species, test pre-existing species classifications and carry out identification. Our aim here is to survey briefly the progress and development of the statistical analysis of morphological variation. This general area of scientific research is called “morphometrics”, a term which is a compound of the classical Greek words morphe: “form”, and metron: “that by which anything is measured” (http://www.lexilogos.com/english/greek_ancient_dictionary.htm). “Morphometrics” can thus be construed as “the measurement of form”. “Form”, in turn, can be understood in a narrow sense as referring only to shape, or more broadly as including shape, structure (which includes size, architecture and internal anatomy) and other aspects of external appearance. It is in the broad sense that we will use it here.

IntroductionOpen image in new window

One of the most immediately obvious features of organic life is its extraordinary diversity — Charles Darwin’s “endless forms” (Darwin 1859). Evolutionary biology attempts to understand how nature’s diversity arose and the sciences of ecology, genetics and physiology investigate the interactions of organisms and seek to understand how they function as part of higher level systems within the cosmos as a whole.

Botanists’ Vision Open image in new window

First-rate botanists have accumulated extensive experience in studying and identifying plants and this forms the foundation for their greater intuitive powers of species recognition. This suggests that it may be possible to improve on computational methods by investigating how botanists gather classificatory information from visual scanning of leaves, either from images or from actual specimens. Such data needs to be captured through the use of eye-tracking technology, since for the botanist this process is mostly intuitive.

Feature Extraction Open image in new window

As discussed in Chap. 2, plant leaves contain many different morphological components (characters) that can be used in analysis for identification, such as outline shape, margin details and venation type. It is important to extract these components accurately from the leaves and generate appropriate descriptors for them.