Research article · 2026

Bayesian stochastic imputation of discontinuous paleoenvironmental proxy records: a case study from Lake Ohrid stable isotopes

Zeki Bora Ön · Environmental and Ecological Statistics 33, 129–148 (2026)

What this study does

This study develops a modular Bayesian workflow for discontinuous proxy records. A latent Gaussian state-space model regularises irregularly sampled series and represents uncertainty; Bayesian multivariate regression with stochastic search variable selection then uses informative coeval proxies to reconstruct missing observations.

Main finding

The framework reconstructs gaps in Lake Ohrid δ18O and δ13C over a long Quaternary sequence and returns posterior uncertainty for the imputed values rather than a single deterministic interpolation.

When this paper may be useful

• Missing values in palaeoclimate or environmental proxy records
• Irregularly sampled multivariate time series
• Latent state-space modelling
• Stochastic search variable selection (SSVS)
• Propagation of reconstruction uncertainty

Methods and keywords

Suggested citation

This page is a concise research guide. For complete methods, results and wording, use the version of record linked above.