A chironomid based transfer function for reconstructing summer temperatures in southeastern Australia

Abstract

We present a new chironomid based temperature transfer function which was developed from a training set of 33 natural and artificial lakes from southeast Australia from subtropical Queensland to cool temperate Tasmania. Multivariate statistical analyses (CCA, pCCA) were used to study the distribution of chironomids in relation to the environmental and climatic variables. Seven out of eighteen available variables were significantly (p < 0.05) related to chironomid species variation and these were mean February temperature (9.5%), pH (9.5%), specific conductance (8.2%), total phosphorous (8%), potential evapotranspiration (8%), chlorophyll a (6.9%) and water depth (6.2%). Further pCCA analyses show that mean February temperature (MFT) is the most robust and independent variable explaining chironomid species variation. The best MFT transfer function was a partial least squares (PLS) model with a coefficient of determination (r2jackknifed) of 0.69, a root mean squared error of prediction (RMSEP) of 2.33 °C, and maximum bias of 2.15 °C. Chironomid assemblages from actively managed reservoirs appear to match assemblages from equivalent natural lakes in similar climates and therefore can be included in the development of the chironomid transfer function. Although we cannot completely rule out some degree of endemism in the Tasmanian chironomid fauna, our analyses show that the degree of endemism is greatly reduced. Therefore, integrating the existing chironomid transfer function for Tasmania (Rees et al., 2008) with this new model is a real possibility.