Reconstructing terrestrial temperatures in the Australian sub-tropics and tropics: A chironomid based transfer function approach

摘要

This paper presents chironomid-based transfer functions for temperature reconstructions intended for use in the Australian tropics and sub-tropics. Temperature was the primary independent control of chironomid assemblages. Two transfer functions were created to infer mean annual (MAT) and mean summer temperature as represented by February temperatures (MFT) as it was not reasonable to automatically assume that chironomid populations in the tropics are controlled by summer temperatures. Weighted average partial least square (WA-PLS) models were the most robust. For MAT they produced a coefficient of determination (r(jack)(2)) of 0.83, a maximum bias of 1.74 degrees C and a root mean squared error of prediction (RMSEP) of 2.35 degrees C while for MFT an r(jack)(2) of 0.76, maximum bias of 2.43 degrees C, and RMSEP of 2.43 degrees C was achieved. Based on the statistics we recommend that the MAT model is used for chironomid work in tropical regions.