The Audit Trail
China manufacturing data, mapped to LCA datasets
HiQLCD carries tens of thousands of China-specific LCI records spanning every Chinese provincial grid. Thousands of expert-curated factory term mappings. Why a baseline mismatch in CBAM filings is not a rounding error.
A Chinese exporter receives a request from their European buyer: declare embedded GHG emissions for hot-rolled steel coil under CBAM Annex IV rules. The quarterly deadline is approaching.
They open an overseas LCA tool. Search “cold-rolled steel.” The results return a European Ecoinvent dataset — EAF process, low-carbon grid.
The problem: their Shanghai facility runs BF-BOF. Roughly 90% of Chinese steel production is BF-BOF (blast furnace–basic oxygen furnace, the integrated long route). The GWP difference between BF-BOF and EAF is 3–5×. Using European EAF data to represent a Chinese BF-BOF facility is not a rounding error. It is a baseline mismatch.
Cortex returns two results in parallel: the HiQLCD China BF-BOF dataset for the declaration, and the EF / Ecoinvent EU reference dataset for the buyer’s verification. Both in one query.
HiQLCD: 18.7× China regional coverage
| Dimension | Fact |
|---|---|
| Rows (with provincial variants) | tens of thousands |
| Provincial grid variants | every Chinese province (CN-BJ, CN-SH, CN-GD…) |
| China coverage multiplier vs. Ecoinvent | 18.7× (Ecoinvent: 2.3×) |
Ecoinvent is the global industry standard. Its China coverage relies primarily on GLO (global average) and RoW entries, with limited provincial grid data. HiQLCD’s China regional depth is more than 8× deeper.
The difference shows up sharpest in specific materials.
Primary aluminum: why the number changes with location
China primary aluminum GWP: 14–20 kgCO₂e/kg. Aluminum smelting is electricity-intensive; China’s grid is coal-heavy.
EU primary aluminum GWP: 8–10 kgCO₂e/kg. European smelters run on hydro-dominated grids.
In a CBAM filing or EPD comparison, this gap is decisive. A declaration built on EU aluminum data for a Chinese factory underestimates embedded emissions by roughly 50–60%. HiQLCD-AL, the aluminum-sector database, carries China-specific data across primary production, secondary (recycled), die casting, and rolling.
Steel: BF-BOF is the Chinese default
Ecoinvent’s Chinese steel data is sparse; its baseline reflects European EAF proportions. HiQLCD is built around the BF-BOF reality of Chinese steel production, with explicit BF-BOF process datasets at the provincial level.
This is a difference in data foundations, not a configurable parameter.
CBAM dual direction: one query, two reference points
CBAM involves two sides of the same supply chain. Cortex holds data for both.
Chinese exporter → EU declaration
“Our Shanghai facility produces 304 stainless steel sheet. We need CBAM-compliant emission factors.”
Cortex recognizes the CBAM context and applies the correct boundary: Scope 1 + Scope 2, not full cradle-to-gate. HiQLCD returns the Chinese BF-BOF production dataset. EF and Ecoinvent return the EU reference values the CBAM declaration requires. Both sets appear in the same response, with the GWP spread visible.
Dataset names return bilingual: “冷硬卷，304 不锈钢 (Cold-rolled annealed coil, 304 stainless steel).” The Chinese name helps the exporter cross-reference their factory records; the English name aligns with the buyer’s declaration form.
EU buyer → verifying supplier data
“Our Chinese supplier declared steel at 1.9 kgCO₂e/kg. Is that credible?”
Cortex retrieves the HiQLCD Chinese steel GWP range: BF-BOF approximately 1.7–2.5, EAF approximately 0.4–0.8. It places 1.9 in the BF-BOF range and explains why Chinese EAF steel has higher GWP than European EAF — because the Chinese grid’s carbon intensity is higher even for the same process route. Ecoinvent does not carry the provincial grid BF-BOF breakdown needed to give this answer.
Factory terms to dataset IDs: thousands of mappings
The search index carries a Chinese LCA terminology dictionary spanning tens of thousands of terms. It tokenizes Chinese material names as units of meaning, not as character sequences. “镍板” (nickel plate) is recognized as a single term and matched precisely — not split into “镍” (nickel) and “板” (sheet), which would match every nickel-containing dataset in the index.
Above that, thousands of expert-curated mappings — built over years of LCA practitioner decisions — translate Chinese factory terms into LCA dataset IDs.
背钩 (watch case back hook)
→ HiQ | Production | Cold-rolled annealed coil, 304 stainless steel, mixed technology, China
The mapping encodes an expert decision: the primary material for a watch back hook is 304 stainless cold-rolled annealed coil. The user does not need to know the decision chain — the mapping carries it.
IGBT模块 (IGBT module)
→ market for insulated gate bipolar transistor, electric vehicle application, Ecoinvent 3.12.0, Global
The specific factory term “IGBT模块” hits the EV-application dataset, not the generic transistor entry. The distinction matters: these are different datasets with different GWP profiles.
502胶水 (502 adhesive, cyanoacrylate)
→ market for polyurethane adhesive, Ecoinvent 3.12.0, Global (proxy substitution, flagged)
There is no 502 adhesive dataset in the LCA databases. The mapping captures an expert proxy decision: polyurethane adhesive as the closest equivalent. Cortex does not make this substitution silently — the result is labeled as a proxy, with the assumption named for audit purposes.
Cortex responds in the user’s query language. Chinese question, Chinese answer. English question, English answer. Dataset names always return bilingual.
Different data foundations, not configurable parameters
Overseas SaaS tools default to EU grid data. Switching to “China” typically means applying a regional parameter on top of Ecoinvent’s limited China entries. The underlying data is not built around Chinese manufacturing realities.
HiQLCD is built from the other direction: Chinese process routes, Chinese provincial grids, Chinese industrial mix as the starting point. Every Chinese provincial grid. BF-BOF as the default for Chinese steel. Coal-dominated grid intensity for Chinese aluminum smelting. These are database facts, not outputs of a configuration step.
Cortex searches 14 databases in parallel — HiQLCD 1.4.0, Ecoinvent 3.12.0, EF 3.1.0, HiQLCD-AL, TianGong, and nine more — in a single query. For China manufacturing scenarios, HiQLCD is the first match. For CBAM compliance, EF is the primary reference. Which path activates depends on the declared use case.
Cortex does not replace the LCA practitioner. It scales their judgment across Chinese supply chain data.
To run a query against the China dataset, open Cortex Chat. For full BOM processing across 14 databases, see BOM matching.
— HiQ Cortex Team