A synthetic benchmark dataset of spatially correlated lithofacies logs with controlled missingness

This dataset provides a fully synthetic benchmark for testing lithofacies reconstruction and geological data-processing methods. It comprises 36 computationally generated stratigraphic models and 138,240 synthetic lithofacies cells, incorporating spatial correlation, controlled missing-data patterns and label perturbations. The package includes synthetic data, generation scripts, reproducibility information and internal validation outputs. It is designed to support reproducible benchmarking of lithofacies reconstruction algorithms. All observations are computationally simulated and do not represent field measurements, laboratory experiments or geological observations from specific locations. Internal consistency checks do not constitute independent geological validation.