Reevaluation of pollen differentiation in Altingiaceae: Challenges in distinguishing deciduous (Liquidambar) and evergreen (Altingia) types using multivariate statistics and machine learning

Abstract

Altingiaceae, a family of woody plants, comprising evergreen Altingia and deciduous Liquidambar groups, exhibits distinct leaf morphology, yet both groups overlap in geographical range and climatic conditions. While some tropical Altingia species are confined to Myanmar, Thailand, Cambodia and India without Liquidambar, and some temperate Liquidambar species to northern China without Altingia. Their fossil pollen have significant implications in reconstructing palaeoclimate and historical biogeography, based on classification of Altingia-type and Liquidambar-type. However, the results of previous studies to differentiate pollen types of evergreen Altingia and deciduous Liquidambar were based on limited pollen specimens. Therefore pollen morphology of Altingiaceae and differentiation of above mentioned types needs reevaluation using more specimens from wider geographical range.

In this study, we present new findings on Altingiaceae pollen morphology from extensive collection of specimens and reassess the diagnostic features to distinguish evergreen and deciduous types. To improve the credibility of palaeoecological and palaeoclimatic interpretations, we applied multivariate statistical analyses to pollen size, number of pores, pollen wall thickness, and size and density of ornamental elements from light microscopy (LM) and Scanning Electronic Microscope (SEM) images. Additionally, random forest classification models were applied to test the accuracy of differentiating Altingiaceae pollen types. Our results reveal significant morphological overlap between the pollen of evergreen Altingia and deciduous Liquidambar, with classification models showing limited accuracy and explainability. Thus, fossil pollen of Altingiaceae cannot be confidently classified into evergreen or deciduous types, highlighting challenges in using their pollen morphology for taxonomic classification in palaeoecological research.