Model based functional clustering of varved lake sediments

TL;DR

This paper introduces a model-based clustering method for functional data with covariates, allowing different covariance structures across clusters, and applies it to varved lake sediments to infer past climate changes.

Contribution

It extends existing models by allowing distinct covariance structures within clusters and provides an EM algorithm for parameter estimation.

Findings

Successfully applied to lake Kassjön sediments

Revealed insights into past climate variations

Enhanced clustering accuracy with covariance flexibility

Abstract

In this paper we propose a model-based method for clustering subjects for which functional data together with covariates are observed. The model allows the covariance structures within the different clusters to be different. The model thus extends a model proposed by James and Sugar (2003). We derive an EM algorithm to estimate the parameters. The method is applied to annually laminated (varved) sediment from lake Kassj\"on in northern Sweden, to infer on past climate changes.