New methods in Palaeopalynology: Classification of pollen through pollen chemistry

Abstract

Pollen grains are one of the primary tools of palaeoecologists to reconstruct vegetation changes in the past. The description, counting and analysis of pollen grains (palynology) has contributed to our understanding of establishment and dynamics of past and present plant communities. Advances in identification accuracy, precision and increased taxonomic resolution have greatly improved our understanding of biogeography and plant community interactions. Nevertheless, the techniques by which palynological studies are performed have not fundamentally changed. Taxonomic resolution and automation have been identified as some of the key challenges for palynology and palaeoecology. Chemical methods have been proposed as a potential alternative to morphological approaches and have demonstrated promising results in the classification of modern pollen grains and in the analysis of pollen chemical responses to UV-B radiation. The application of chemical methods for palynological needs have not been thoroughly explored, with analysis of (sub-)fossil pollen lagging behind their modern counterpart. Especially the application of infrared methods have gained popularity as an alternative to traditional morphological approaches.

In this thesis, I explore the use of infrared methods for palynological applications, by exploring the chemical variation in modern pollen grains and in the analysis of fossil pollen grains with IR microscope approaches. The objectives of this thesis are formulated into three research objectives:

* Collect modern pollen and explore the variation in chemical composition * Apply chemical methods to fossil material * Explore microscopy chemical methods on modern pollen

The thesis is structured into four studies to study these objectives. Papers I and II explore variation and classification based on the chemical composition of modern *Quercus* pollen using two IR approaches, Fourier transform infrared spectroscopy (FTIR) and Fourier transform Raman spectroscopy (FT-Raman). After exploring modern chemical composition of pollen, paper III investigates FTIR methods for the analysis of fossil pollen, in spectra of Holocene *Pinus* pollen. Additionally, the effects of acetolysis and density separation on *Pinus* pollen is described. Paper IV addresses the challenge of scattering signals when measuring small pollen grains of four *Quercus* species with FTIR microscopy and ways to surpress or weaken the scattering signals.

The results from paper I and II show classification success, surpassing traditional morphological approaches, at the *Quercus* section level and ~90% recall on species level with both IR approaches. Chemical bands most useful for classification are lipids, sporopollenin and proteins for both FT-Raman and FTIR. We observe differences in the importance of chemical functional groups for the classification. FT-Raman relies more on sporopollenin chemistry, while FTIR utilizes more variation in lipid bands. After finding considerable variation in sporopollenin chemistry in modern pollen samples, FTIR methods were applied to pollen from sediment cores spanning the Holocene. Paper III examines the differences between modern and sub-fossil pollen and reported large differences between them, mainly the removal of labile components, such as lipids and protein peaks from the sub-fossil spectra during diagenesis. Additionally, paper III finds changes to pollen chemistry caused by acetolysis in the 1200 - 1000 cm^-1^ region of the spectra, when comparing acetolysed spectra to non-acetolysed spectra. The paper concludes with findings of unwanted inorganic signals (BSi) and contamination from density separation media in the sediment pollen spectra. Paper IV demonstrates two successful methods of removing scattering signals from pollen spectra. Two approaches were examined, embedding and processing with signal correction algorithms. Spectra from embedded pollen have no scattering anomalies, but part of the spectra is unusable, because of absorbance of the embedding matrix (paraffin). The signal processing algorithm removes most of the scatter components and allows the scatter components to be extracted. Classification of the different data-sets (spectra without correction, embedded spectra, processed spectra, scatter parameters) reveals that scatter correction methods reduce classification success and that scatter parameters contain taxonomic information. This suggests that scatter corrections may not be the best approach for applications mainly focused on classification or identification, while reconstructions of, for example, UV-B radiation may benefit from scatter correction methods, when measuring single grain spectra.

This thesis shows that the performance of IR methods surpasses traditional morphological methods for pollen classification and that a considerable amount of taxonomic information is stored in functional groups associated with sporopollenin (phenylpropanoids). In a study on fossil pollen, this thesis demonstrates that conventional chemical extraction methods, such as acetolysis, alter the chemical composition of pollen and may not be ideal for palaeochemical purposes. Additionally, the scatter correction methods show that IR can provide non-chemical information in the form of scatter parameters, which contain taxonomic information. These results are useful additions to the growing knowledge on chemical methods for palaeoecological and palynological analyses.