Data for Selection-Controlled Evaluation of ML-Readiness under Incomplete Environmental Digital Twin Observations

This dataset supports the study “Selection-Controlled Evaluation of ML-Readiness under Incomplete Observations in an Environmental Digital Twin.” It contains a deduplicated observational dataset derived from archived urban air-quality records from three London monitoring sites and the authoritative exact one-hour NO₂ forecasting baseline used in the study. The deposited analytical baseline contains 2,067 exact-horizon samples and enables deterministic reconstruction of the P1, P2, and P3 completeness-policy populations used in the selection-controlled evaluation. The accompanying README documents data provenance, variables, units, policy definitions, missing-data interpretation, and the relationship between the two deposited datasets.