Pollen analysis for the generation of environmental hypothesis

Abstract

Abstract Q-mode factor analysis combined with R- and Q-mode cluster analysis were used as hypothesis-generating tools on modern pollen dispersal data for Cerro Domuyo, Neuquén (Argentina). Q-mode cluster analysis partitioned the data matrix into three groups of samples that were consistent with plant geographic units (Subantartic forest, bogs, patagonian and andean xeric communities). Q-mode factor analysis produced three principal components that accounted fot 87% of the total variance and were assigned respectively to temperature, topographic gradient and precipitation. Three particular combinations of these components corresponded to the group of samples generated by Q-mode cluster analysis. When R-mode cluster analysed, each group of samples reflected a peculiar correlation of plant taxa. A model was postulated stating that the ‘causal’ structure of the system is the particular combination of those three physical components of the environmental complex and that the peculiar correlations of plant taxa are surface structures generated by biological ‘responses’ to the gradients of the physical parameters. The procedure is recommended for constructing explanatory modern analogs for Quaternary paleoecological research and as a fast and low cost survey aimed to planification of research work in plant ecology.