Testate amoeba-based hydrological reconstruction in West Siberian mires: Machine learning improves modern predictive performance but does not guarantee robust fossil reconstruction
Year: 2026
Type: dataset
Abstract: This dataset supports the study “Testate amoeba-based hydrological reconstruction in West Siberian mires: Machine learning improves modern predictive performance but does not guarantee robust fossil reconstruction.” It contains two comma-separated values (CSV) files used to evaluate conventional transfer functions and machine-learning models for reconstructing water table depth from testate amoeba assemblages in West Siberian mires.1. Modern_Testate amoeba_data.csvThis file contains modern testate amoeba assemblage data and associated water table depth, substrate water content measurements. It was used to develop and evaluate the reconstruction models, including comparisons between conventional transfer functions and machine learning algorithms. Spatially independent training and test subsets were used to assess model performance on samples from different mire sites.2. Great Vasugan Mire_core.csvThis file contains data from a fossil peat core collected from the Great Vasyugan Mire. It was used to apply selected models to fossil testate amoeba assemblages and reconstruct past water table depth. The fossil application enabled an assessment of whether strong predictive performance on modern samples translates into stable and reliable paleohydrological reconstructions.
Source: Figshare
Language: English
Subfield: Ecology
Field: Environmental Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7224959523