ABSTRACT Charcoal is widely accepted as evidence of the occurrence of palaeo-wildfire. Although fossil charcoal remains have been used in many studies, investigation into the anatomical changes occurring during charring are few. The present study analyses changes in selected anatomical characters during artificial charring of modern wood of three species of the genus Araucaria (i.e. Araucaria angustifolia, Araucaria bidwillii and Araucaria columnaris). Wood samples of the studied species was charred under controlled conditions at varying temperatures. Measurements of anatomical features of uncharred wood and artificial charcoal were statistically analysed. The anatomical changes were statistically correlated with charring temperatures and most of the parameters showed marked decreases with increasing charring temperature. Compared to the intrinsic variability in anatomical features, both within and between growth rings of an individual plant, the changes induced by temperature account only for a comparatively small percentage of the observed variability. Regarding Araucaria charcoal, it seems possible that at least general taxonomic and palaeoenvironmental implications can be drawn from such material. However, it is not clear so far whether these results and interpretations based on only three taxa, can be generalized for the entire family and anatomically similar fossil taxa or not.

The aim of the study is to evaluate the physical and chemical changes that occur on wood fragments submitted to different temperatures, verifying their influence on significant chemical characteristics in the forming process. Carbonization process in muffle associated to the techniques of Thermogravimetry, Fourier Transform Infrared Spectroscopy (FTIR) and multivariate analysis. The analyses were performed on 3 replicates for each burn temperature. The temperature of the thermal muffle process was in every 50 °C, from 200 °C to 600 °C. TGA were performed under nitrogen atmosphere, using a heating ramp from 25 °C min-1 to 995 °C. A range between 1900 to 650 cm-1 was used, with a resolution of 4 cm-1 and 64 scans. The PCA showed that it is possible to describe 95.73% of the data, grouping the samples into three main clusters. These clusters were used to build a SIMCA (Soft Independent Modeling of Class Analogy) model, enabling to predict with 100%. Results showed that technical associations, such as TGA, FTIR and multivariate analysis may help to characterize the natural carbonization process and, in future works, contribute to significant (paleo)environmental inferences.