Machine learning algorithms improve the power of phytolith analysis: A case study of the tribe Oryzeae (Poaceae)

Abstract

Phytoliths, as one of the important sources of microfossils, have been widely used in paleobotanyrelated studies, especially in the grass family (Poaceae) where abundant phytoliths are found. Despite great efforts, several challenges remain when phytoliths are used in various studies, including the accurate description of phytolith morphology and the effective utilization of phytolith traits in taxon identification or discrimination. In this study, we analyzed over 1000 phytolith samples from 18 taxa representing seven main genera in the tribe Oryzeae (subfamily Ehrhartoideae) and five taxa in the subfamilies Bambusoideae and Pooideae. By focusing on Oryzeae, which has been extensively investigated in terms of taxonomy and phylogeny, we were able to evaluate the discrimination power of phytoliths at lower taxonomic levels in grasses. With the help of morphometric analysis and by introducing several machine learning algorithms, we found that 87.7% of the phytolith samples could be classified correctly at the genus level. In spite of slightly different performances, all four machine learning algorithms significantly increased the resolving power of phytolith evidence in taxon identification and discrimination compared with the traditional phytolith analysis. Therefore, we propose a pipeline of phytolith analyses based on machine learning algorithms, including data collection, morphometric analysis, model building, and taxon discrimination. The methodology and pipeline presented here should be applied to various studies across different groups of plants. This study provides new insights into the utilization of phytoliths in evolutionary and ecology studies involving grasses and plants in general.

Grass habitat analysis and phytolith-based quantitative reconstruction of Asian monsoon climate change in the sand-loess transitional zone, northern China

“…For example, Gu et al (2016) illustrated different genera of bamboos can be distinguished by the detailed shape of saddle phytoliths. Cai and Ge (2017) used machine learning to distinguish short cell phytoliths from different linages of Poaceae. In addition, with the development of a biogeographical database and a database of climatological observations, it is possible to calculate the climate range (ecological amplitude) of each species, which provides a robust foundation for the CA.…”

“You Are Not My Type”: An Evaluation of Classification Methods for Automatic Phytolith Identification

“…Images of four examples of incorrect classification results are illustrated in Figure 4. Cai & Ge (2017), instead of working with general types, focused their research specifically on the classification of short cell phytoliths, aiming at taxonomical identification and obtaining results comparable to ours, with SVM performing the best of the models they tested.…”

“…There have been several studies that used quantitative phytolith morphometric size and shape parameters for the identification of morphometric characteristics (Ball et al, 2016;Out & Madella, 2016;Portillo et al, 2019). Recently, researchers have started to design computing methods for the automatic identification of phytoliths (Evett & Cuthrell, 2016;Cai & Ge, 2017;Gallaher et al, 2020). Evett & Cuthrell (2016) established the conceptual basis for the application of semi-automated classification methods to morphometric phytolith analysis, describing detailed procedures and strategies to be tested while acknowledging current technical limitations.…”

“…The authors examined functional aspects regarding image acquisition, morphometric parametrisation (e.g., geometric parameters and elliptic Fourier analysis), classification techniques (multivariate statistics versus supervised learning models), and the development of a semi-automated phytolith analysis system. Cai & Ge (2017) extracted grass short cell phytoliths from the leaves of 23 taxa belonging to the subfamilies Ehrhartoideae, Bambusoideae, and Pooideae. They used morphometric data from scanning electron microscopy (SEM) images to train a classifier to successfully distinguish different genera within the Oryzeae even though they all produce the same phytolith morphotypes.…”

JSE IBC Special Issue on Frontiers in Plant Systematics and Evolution

“…They explain how the sex chromosomes of haploid‐dominant organisms are distinct from the X‐Y and Z‐W systems, and provide a summary on their distribution and genetic composition. Cai & Ge () propose a pipeline for analyzing phytoliths based on machine learning algorithm, including data collection, morphometric analysis, model building, and taxon discrimination. Their methodology and pipeline are developed based on a case study on the rice tribe Oryzieae, but should be applied to studies across different groups of grasses and other plants that utilize phytoliths in evolutionary and ecology studies.…”