Towards variance-conserving reconstructions of climate indices with Gaussian Process Regression in an embedding space
Abstract. We present a new framework for the reconstruction of climate indices based on proxy data such as tree rings. The framework is based on the supervised learning method Gaussian Process Regression (GPR) and designed to preserve the amplitude of past climate variability and to adequately handle noise-contaminated proxies and variable proxy availability in time. To this end, the GPR is performed in a modified input space, termed embedding space. We test the new framework for the reconstruction of the Atlantic Multi-decadal Variability (AMV) in a controlled environment with pseudoproxies derived from coupled climate-model simulations. In this test environment, the GPR outperforms benchmark reconstructions based on multi-linear Principle Component Regression. On AMV-relevant timescales, i.e., multi-decadal timescales, the GPR is able to reconstruct the true magnitude of variability even if the proxies contain a non-climatic noise signal and become sparser back in time. Thus, we conclude that the embedded GPR framework is a highly promising tool for climate-index reconstructions.
Viewed (geographical distribution)