Methodology
AtlasCore produces deterministic climate outputs from authoritative Australian government data. This page explains how data is selected, composed, and validated.
Principle: government data in, structured evidence out
AtlasCore does not model or estimate emissions. It retrieves, normalises, and composes data from authoritative government publications — the same sources that auditors reference under AASB S2.
Data composition
Company climate profiles
A company climate profile is composed from:
- Entity resolution — the ABN is resolved against CER NGER corporate registrations using Splink probabilistic matching, with ABR as the fallback
- Emission data — annual Scope 1 and 2 emissions from NGER reporting
- Grid intensity — AEMO CDEII emission intensity for the relevant NEM region
- Physical climate — BOM SILO gridded observations at the company's primary or site locations
All inputs are persisted before composition. The profile is computed deterministically from stored data — no live API calls occur at query time.
Emission factors
NGA factors are ingested from the DCCEEW workbook:
- Each sheet is parsed, handling merged cells and multi-level headers
- Raw rows are normalised to canonical categories, scopes, grids, and units
- Constituent gases (CO2, CH4, N2O) and combined factors are stored separately
- Provenance metadata (source table, sheet, row number, quality grade) is preserved per value
Physical risk assessment
Physical risk uses BOM SILO gridded data (~5km resolution):
- Temperature — mean temperature as a chronic physical risk indicator
- Rainfall — rainfall deficit as a chronic physical risk indicator
- Extreme heat — days exceeding thresholds as an acute physical risk indicator
Risk bands use a 1981–2010 baseline period. Drought modifiers escalate bands when rainfall falls below historical thresholds.
Determinism
AtlasCore guarantees deterministic outputs: the same persisted inputs always produce the same result. This is achieved through:
- Snapshot isolation — queries reference specific data snapshots, not live sources
- Evidence hashing — SHA-256 hashes computed from canonical inputs
- No stochastic models — all computations are deterministic functions of stored data
- Amendment tracking — factor set changes (corrections, restatements, new editions) are versioned and auditable
Quality controls
| Control | Mechanism |
|---|---|
| Source licence audit | Every data source is verified for commercial repackaging rights before inclusion |
| Snapshot versioning | Every extraction creates a numbered snapshot with quality metrics |
| Data quality grade | Factor values carry A/B/C quality grades from source metadata |
| Evidence hashing | Per-value and per-response SHA-256 integrity hashes |
| Five-file bundles | Every report preserves full provenance regardless of requested format |
| Amendment tracking | Factor set corrections and restatements recorded with diff counts |
Limitations
- Scope 3 — reference factors plus indicative screening-grade estimates for key categories (Cat 3, 5, 6); not disclosure-grade
- Governance and Strategy — disclosure sections require entity-authored narrative
- Scenario analysis — CMIP6 projection support implemented; populates when projection data is ingested
- Market-based Scope 2 — evidence-gated; requires tenant-held contractual instruments
- Filing — AtlasCore produces disclosure inputs, not regulatory filings
Source authority
All quantitative outputs trace to these government publications:
| Source | Authority |
|---|---|
| NGA Factors Workbook | DCCEEW (Department of Climate Change, Energy, the Environment and Water) |
| NGER emissions | Clean Energy Regulator |
| Grid intensity (CDEII) | Australian Energy Market Operator |
| Climate data (SILO) | Bureau of Meteorology |
| Entity data | Australian Business Register |
| Statistical data | Australian Bureau of Statistics |