Data Source & Methodology
Where the data comes from and how the atlas processes it — short and transparent.
Data source
All emissions data comes from Climate TRACE — a coalition that estimates emissions from satellite data and machine learning, not from self-reporting. We use the Germany package v5.8.0, metric CO₂-equivalent (100-year GWP). Licence: CC BY 4.0.
What the atlas shows
67,129 asset-level emitters across Germany — individual facilities with coordinates, estimated capacity and emission factor. Shown is the annual total of the last full calendar year (2025), summed from Climate TRACE's monthly values.
Processing
Raw data is read via a DuckDB pipeline, aggregated to one year per source, sectors normalised, and exported as a lean Parquet (~1 MB). The map reads that Parquet directly in the browser — no backend in the read path. The global climate dataset is never loaded, only the pre-processed Germany artefact.
Asset-level vs. sector estimate
Not every sector is available asset-level. Power, manufacturing, transport (incl. airports), fossil fuels, waste and most of agriculture carry individual sources with coordinates. Where Climate TRACE only provides regional estimates, these appear as aggregated points (e.g. “… District”).
Trend
The per-facility trend arrow compares the annual total 2025 with 2024 (▲ up / ▼ down). It reflects the change in estimated emissions, not measurement precision.
Limits & uncertainty
Climate TRACE values are estimates with uncertainty (confidence, where available, is in the detail popup). Cattle operations only carry coded names in the dataset — we label them “Dairy farm” / “Cattle farm (beef)”. Mapping facility→operator is incomplete per Climate TRACE itself and is not shown here. Embedded third-party sources (EDGAR, CEDS) have their own terms.
Scientific guardrail
The link emissions → weather is cumulative and statistical — like loaded dice. Never single-event causation from one source to a specific weather event.
Data source: Climate TRACE Emissions Inventory v5.8.0 — CC BY 4.0. Aggregated and regionally filtered from the original dataset.