Artificial intelligence approach to palaeogeography and evolutionary trend analysis of Laurentian brachiopod fauna in the Rhynchotrema-Hiscobeccus lineage

Abstract

Previous studies, based on the qualitative morphological description on the Late Ordovician brachiopod fauna hypothesized that the earliest species of Hiscobeccus, which is one of the diagnostic taxa in North American epicontinental brachiopod fauna, evolved from Rhynchotrema, a well-known rhynchonellide brachiopod, which lived in pericratonic regions of Laurentia. A quantitative approach conducting multivariate analysis, which was based on biometric characters of rhynchonellide brachiopod specimens from the Late Ordovician (upper Sandbian–upper Katian) in North America, supported this hypothesis that the younger and larger species of Hiscobeccus evolved from older and smaller species of Rhynchotrema during the Late Ordovician.The current study proposes the first report of employing an artificial intelligence model based on neural networks to map a set of quantitative morphological measurements of the Rhynchotrema-Hiscobeccus lineage of North America to their locality and evolutionary trend. For this purpose, a total of 171 morphometric measurements of the Late Ordovician brachiopods from 11 localities of North America were divided into 114 training and validation samples and 57 testing sets. The input morphometric parameters of the studied brachiopods include shell length (L), shell width (W), shell thickness (T), sulcus depth (T1), sulcus maximum width (W1), sulcus floor width (W2), apical angle (AA), lamella-covered shell length (L1), lamella number (Ln). Artificial neural network tries to find a mathematical formulation between the mentioned morphometric parameters and their corresponding localities from North America. The results showed that the accuracy of the neural network approach in estimating the locality of the testing brachiopod samples is 81%. In the light of satisfactory results of neural networks, having a set of the morphometric data from the unseen Late Ordovician brachiopods, their localities can be estimated.