Developing diatom-based transfer functions for Central Mexican lakes

Abstract

This paper is the first attempt to produce diatom-based transfer functions for the northern tropical Americas. A dataset of 53 modern diatom samples and associated hydrochemical variables from 31 sites in the volcanic highlands of central Mexico is presented. The relationship between diatom species distribution and water chemistry is explored using canonical correspondence analysis (CCA) and partial CCA. Variance partitioning indicates that ionic strength and ion type both account for significant and independent portions of this variation. Transfer functions are developed for electrical conductivity (r 2 = 0.91) and alkalinity (as a percentage of total anions) (r 2 = 0.90), reflecting ionic strength and ionic composition respectively. Prediction errors, estimated using jack-knifing, are low for the conductivity model, but the carbonate transfer function performs less well. This study highlights the potential for diatom-based quantitative palaeoenvironmental reconstructions in central Mexico. However, a number of key diatom species found in fossil material are not represented in the modern flora. Sampling of additional sites may resolve this, but it is thought that the lack of modern analogues may reflect the high degree of anthropogenic disturbance in many of the catchments. This highlights the problem of trying to reconstruct pre-disturbance environmental changes in highly modified ecosystems. One possible solution is to merge the central Mexican data with the African dataset, which includes sites of similar chemical composition, but which have not suffered the same degree of disturbance.