State-space models and imputation of environmental time series
Environmental records are frequently irregular and incomplete. Latent state-space models separate an underlying process from observation noise, providing a natural framework for smoothing, interpolation, forecasting and uncertainty-aware reconstruction.
Lake Ohrid
A latent Gaussian state-space model and SSVS reconstruct missing stable-isotope observations.
Structural time series
Bayesian structural time-series models construct counterfactual trajectories for testing a proposed climatic event.
Practical missing-data methods
A CIESM chapter compares hierarchical, change-point and Kalman-filter approaches.
Related concepts
Bayesian inference · MCMC · environmental time series · palaeoclimate proxy data · uncertainty quantification · reproducible quantitative analysis