Detecting patterns of change in a long pollen-stratigraphical sequence from Funza, Colombia – A comparison of new and traditional numerical approaches

Long continental pollen records from the northern Andes, interpreted by classical qualitative palynological approaches, have been shown to be rich sources of information about vegetational, environmental, and climatic change and biome evolution over long time periods. However, such data sets are large and complex, thereby restricting the type of specific research questions that can be addressed using classical qualitative approaches. Here, we use the upper part of a long pollen record (Funza09; 37 pollen types; 938 samples; mean temporal resolution 1250years) that reflects the period from 1.2 million years before present (1.2Ma) to 27,000years before present (27ka), corresponding to marine isotope stages (MIS) 37 to 3, to investigate and explore quantitatively the underlying patterns of change in the pollen stratigraphy. We show that numerical data analysis, both unsupervised and supervised methods, permit the exploration of such patterns in a large pollen data set and that the results generally support previous interpretations based on more classical qualitative and traditional quantitative approaches. However, modern numerical data-analytical methods open up new research questions, in particular about stratigraphical data clustering, changes in the taxonomic composition of montane forest, and rates of palynological change. This paper is the first attempt to re-visit several large data sets of tropical and temperate pollen records with the aim of improving data exploration in general, and addressing particular aspects of long-term community ecology. The last million years of the Pleistocene have witnessed an alternation of interglacial and glacial conditions. A better understanding of the underlying patterns in long pollen sequences can potentially contribute to a better appreciation of the possible consequences of global change.