Enhanced Principal Tensor Analysis as a tool for 3-way geological data reconstructions.

Abstract Principal tensor analysis (PTA) is considered as a potentially useful tool in geosciences, particularly for reconstructions of multi-way (time (depth), space, and proxies) datasets. We introduce an extension of PTA: PTA enhanced with Singular Spectrum Analysis (SSA). Using this enhancement, we were able to isolate clear patterns in noisy multi-way data. As an illustrative example, the method has been applied to 4-way data tensor (time, space, proxies, and delay-time) constructed from marine sediment proxies. Possible restrictions and main reasons for the method's limitations are discussed. The algorithm has been implemented as an R function which is provided in the supplementary data and appendix. Highlights • Multi-way geo-data and tensor decomposition • How can we enhance tensor analysis using singular spectrum analysis • De-noising, de-trending, and orthogonal decomposition in one computer run [ABSTRACT FROM AUTHOR]