Earth science
Earth science research, evaluated in the field's own terms
You cannot rerun the Earth. Evidence here means observation, models validated against it, and uncertainty stated plainly — the draft rubric below is built around exactly that.
What counts as evidence of earth science skill
Earth science runs on observation at scale: field campaigns, sensor networks, remote-sensing pipelines that turn raw satellite radiances into usable records. Skill is visible in the processing chain itself — calibration choices, gap handling, honest flags on suspect data — and much of it is public through archives like Copernicus, NOAA, and USGS.
The field's models make claims about a system that allows no controlled rerun, so validation strategy is where research judgment concentrates: hindcasts against held-out records, comparison across independent observation types, sensitivity shown rather than asserted. A modeling paper that treats validation as an afterthought reads very differently from one built around it.
Applied earth science — hazard assessment, resource evaluation, climate-risk analysis — produces reports where uncertainty has consequences and overstatement has a cost. Writing them well is real evidence of skill, and almost none of it is citable.
Contribution to community datasets and model intercomparisons is first-class evidence as well: quality-controlling a station record, submitting to an intercomparison project and documenting honestly where your model diverged, publishing a correction to a widely used product. This is the field's infrastructure work — high judgment, low citation — and evaluators here will read it as seriously as any paper.
The earth science evaluation rubric, first draft
Earth science evaluates claims about a system that cannot be rerun, so this rubric prizes validation strategy and uncertainty candor over model sophistication.
- Observational grounding
- Claims trace to observations through a documented, inspectable processing chain, and data provenance is explicit at every step.
- Validation strategy
- Models are tested against records they were not tuned on, across more than one observation type where possible, with disagreements reported.
- Uncertainty communication
- Scenario is distinguished from probability, ranges are defended rather than decorated, and what is unknown is stated as plainly as what is known.
- Scale judgment
- Spatial and temporal scales of data, model, and claim actually match, and extrapolations across scales are flagged as such.
Founding earth-science evaluators will trial this draft on public datasets, model papers, and assessment reports before it hardens.
What founding earth science evaluators will do
Test the rubric across the field's full width — a paleoclimate reconstruction, a seismic-hazard model, and a land-cover pipeline should all be scoreable without bending the standard.
Score public artifacts independently — processing chains, validation sections, assessment reports — and map where expert readers disagree before any score is treated as settled.
Start the calibration records that will weight earth-science evaluation when the community opens, so measured accuracy anchors the discipline from its first day.
Who this is for
The cohort needs earth scientists who respect the distance between a measurement and a claim:
- Field and observational scientists who know what a clean data record costs to produce.
- Modelers who treat validation as the work rather than the garnish.
- Remote-sensing and geospatial researchers whose pipelines are their principal artifact.
- Applied scientists in hazard, water, energy, and climate services whose reports carry real consequences.
Who this is not for
And the honest exclusions, before you spend the time:
- Anyone wanting their own model or dataset certified — evaluation here is outward-facing.
- Researchers allergic to uncertainty language; candor about ranges is scored, not penalized.
- Anyone who prefers an untracked scoring record; measured calibration is the mechanism.
- Anyone expecting live evaluation today rather than a standard being built in public first.
Apply to evaluate earth science
Earth science is pre-selected on the application. Link an ORCID profile, a data or code repository, or published assessment work.