HOW IT WORKS
Open at the bottom. Hosted at the top.
The scientific engine and the services built on it evolve separately. Every result carries a recipe and provenance object you can persist, export and — where licensing allows — recompute.
- Decision surfacesWeb app · Scenario Lab · reports
- ServicesStress-Test API · jobs · audit trail
- Registry & provenanceRecipes · lineage · validation
- Computeclimate_indices · Rust · xarray/Dask
- Public dataNOAA · NASA · ERA5 · PRISM · CMIP
HOSTED
Where we earn: scale, workflow, support
OPEN
Always: algorithms, specs, public data
A hazard index is not a loss estimate.
We keep each layer of climate risk explicit, so nobody mistakes a drought percentile for a dollar figure.
- HazardWhat Earthflux computes: published, validated physical and statistical indices.
- ExposureWhat you bring: asset locations, polygons, operating areas and thresholds.
- VulnerabilityDomain models from partners — always named, never hidden inside a score.
- LossNever implied by an index alone. Only where a defensible damage model exists.
What’s free, what’s hosted.
We charge for hosted compute, workflow, integration and support — never for the method or the result.
| Surface | Free & open | Hosted & paid | Built for |
|---|---|---|---|
| Registry | Read and query, with no account | None — the Registry stays free to read | Research orgs, agencies, enterprises |
| Scenario Lab | Examples on published Curated Datasets | A Place assessment and its report-ready export | Scientists, planners, consultants |
| Stress-Test API | Planned: published-data compute at documented rate limits | Higher scale, integration and support | Engineering, utilities, insurers, asset owners |
| Customer inputs | Your own Pipeline on your own infrastructure | Optional, planned Earthflux-hosted processing | Agencies, regulated and technical teams |
Commitments that don’t change.
Open science is how Earthflux earns trust. These hold whatever we build on top.
- 01climate_indices stays permissively open and never requires Earthflux to compute.
- 02Provenance and validation conventions are published as open specifications.
- 03Software releases and key derived datasets get DOIs and citation workflows.
- 04Uncertainty and validation status are shown, never collapsed into one confidence badge.
Built in the open, in this order.
NOW
Registry & provenance schema
Reproducible dataset pages and a published taxonomy for validation evidence.
NEXT
Scenario Lab beta
User workspaces, an intercomparison harness and scenario datasets.
LATER
Stress-Test API
Domain packages and compound-hazard methods with customer-specific validation.
Shape it with us.
Design partners help set validation priorities, reference datasets and the first workflows.