Comparison of an optimal regression method for climate reconstruction with the compare_methods() function from the dendroTools R package

The selection of a regression technique for climate reconstruction may have an important impact on reconstructed values. In this paper, we introduce the compare_methods() function from the dendroTools R package. This function compares different regression algorithms and returns validation results for each. In addition to mean validation metrics and ranks derived from these, transfer functions should have a key role in the evaluation of different regression algorithms. These are also returned as the output of compare_methods(). Our methodology is introduced on two case studies, one using a mean vessel area (MVA) chronology and one using a standardised tree-ring width (TRW) chronology. The nonlinear machine learning methods compared in our study provided relatively small (if any) improvements in terms of explaining climatic variance. However, they do offer different treatments of extreme values, and if providing more plausible climate reconstructions, this could make them a useful tool for climate reconstruction. We propose the use of the compare_methods() function as a standard methodological check before performing climate reconstruction.