A harmonized panel of United States compression-ignition engine certification data, with resolved certification lineage

Abstract

The United States Environmental Protection Agency (EPA) publishes emission certification records for compression-ignition engines, but distributes them as two spreadsheets whose internal structures are mutually incompatible and whose true column headers are not on the first row. The result is a public record that is open in principle and costly to use in practice: every researcher must independently reverse-engineer two layouts, and the characteristic failure is not an error message but a quietly wrong number.

We present CIDEX, a harmonized engine-family-level panel reconciling both sources into a single schema: 8,627 engine families, 9,794 engine configurations and 94,869 emission records from 108 manufacturers, covering model years 2011–2027.

Beyond reorganisation, we resolve a structure present in the source but not previously published in usable form: 5,982 nonroad families declare a predecessor family, and these declarations chain. Resolving them yields 2,541 distinct certification lineages with a maximum depth of 14 model years, allowing a family’s regulatory history to be followed across more than a decade. We show the lineage graph is a forest, and that a naive traversal reports 30 spurious cycles arising from an undocumented self-reference convention.

We further identify 50 records carrying physically impossible negative certification results, and four model years whose apparent decline is an artefact of EPA’s archive boundaries rather than a change in the fleet. Both are published as flagged data rather than silently corrected or removed. CIDEX preserves certification units as issued and performs no silent conversion: highway results are in g/bhp-hr and nonroad results in g/kW-hr, and we argue that normalising them would convert a visible incompatibility into an invisible one.