The human-decision layer existing networks lack
CMEMS marine, EUMETNET atmospheric, EUCLID lightning, EEA air quality and the national avalanche services already cover the meteorological side. What none of them hold is the response layer: given the forecast, did the session run, and what did the operator decide?
Every score carries a session_id. When an operator reports the outcome, the forecast and the human decision are paired. That paired signal, an observed operational response under real constraints, is one no raw-data vendor or B2C app produces.
Forecasts are everywhere. What is rarely captured is how real operators respond to them, and what happens next.
Score, operator outcome, de-identified derivative
Score
Each /v1/score call is audit-logged with its raw weather samples, provider chain and forecast horizon. This record stays tenant-scoped.
Operator outcome
The operator reports the observed decision, session ran or was cancelled, plus declared equipment category, against the session_id.
De-identified derivative
Forecast and outcome are joined, then reduced to a de-identified derivative before any research release. Raw commercial detail does not travel with it.
Operational audit records remain tenant-scoped. Only eligible, consented and de-identified derivatives can enter a research release, so a plain score call never exposes an operator's commercial demand.
What it measures, and what it does not
An observed operational decision, paired with the forecast that preceded it.
- Whether a session ran or was cancelled, and its metadata.
- The forecast and conditions recorded just before that decision.
Ground truth about the forecast, or the intrinsic quality of the score.
- Whether the forecast was objectively correct.
- A ran or cancelled outcome is a human and operational response under many confounders: demand, staffing, equipment availability, logistics, policy, perceived risk, customer preference and unmeasured local conditions.
The honesty is the point: the signal is a human and operational response, not a verdict on the forecast. That is exactly why the record is scarce, and why it is worth building carefully.
Opt-in consent, tenant isolation, disclosure control
Participation is set by the tenant, not by Goable. research_consent defaults false, is opt-in and remains revocable. Operational audit records stay tenant-scoped; only consented, eligible rows become candidates for a de-identified release. The controls below are disclosure-control measures, not a claim that released data is unconditionally anonymous.
Minimum distinct contributors behind any released cell, so no single operator is identifiable.
Coordinates are generalised to a grid before release; no exact spot is disclosed.
Releases trail real time so no live commercial signal is exposed.
research_consent defaults false and is set by the tenant; participation is opt-in and revocable.
Released datasets are subject to documented de-identification controls: spatial generalisation, contributor thresholds and a release delay. Each release is assessed for residual re-identification risk before publication. Erasure requests are honoured where applicable under GDPR Art. 17.
Six research streams, with testable questions
Suitability calibration across participating activity families
Paired forecast and observed outcome across participating activity families. Intended to back CMEMS, Horizon Europe and Interreg grant applications.
In which activities, regions and forecast horizons do verdicts correlate best with sessions run or operator review?
Equipment and propulsion transition
Tracks declared equipment and propulsion categories where supplied by operators, enabling modal comparison across zones without a carbon-quality judgement.
Across participating operations, what is the declared mix of propulsion categories, and how complete is the reporting?
Forecast verification
The most testable stream. Verification is kept distinct across: weather variables against observations; suitability probability against operational outcomes; verdict against operator review; and provider skill against a reference observation or reanalysis. Probabilistic metrics (Brier, CRPS, reliability) apply only where a probabilistic forecast and a well-defined target exist.
How does provider skill vary by variable, regime and horizon against defined reference sources?
Equipment transition
A longitudinal dataset for observing equipment-transition patterns in participating outdoor operations, across marine, snow and air sports.
How do declared equipment categories shift over time, with what coverage and bias?
Inverse Suitability: skill and difficulty
A research track on separating condition difficulty from rider skill using behavioural outcomes, via continuous-item IRT adapted to weather (δ(x) difficulty, θ_r skill). The current paper (arXiv 2607.01961) demonstrates recovery on synthetic data; real-world validation is a future research step.
Does the model recover difficulty and skill on real data and stay calibrated out-of-sample?
Drift Events
Per-cell forecast-skill regime shifts surfaced by the Concept-Drift Guardian's SPC monitor, recorded in a tamper-evident, versioned drift-event log behind the SPC charter shared with MGAs and reinsurance underwriters.
Do CUSUM alerts anticipate significant degradations in verification metrics?
Does the dataset exist yet?
No public release yet. The research registry is open for design partners.
Public releases begin only after the first dataset meets the stated consent, sample-size and disclosure-control thresholds. Until then there is no dataset to buy or cite, and we say so plainly.
Potential utility for ESG reporting, not certification
Operational sustainability signals, not an emissions inventory. Goable can report observed equipment and activity mix where tenants provide it. It does not calculate verified corporate GHG inventories or certify CSRD, GRI or GSTC compliance.
The map below shows where these signals could support the frameworks operators report against. It is potential utility, not coverage, compliance or certification.
| Framework | Topic | Potential Goable input |
|---|---|---|
| CSRD / ESRS E1 / E5 | Climate + circular economy | declared low-carbon activity share and equipment-type observability, where tenants supply it |
| GRI 301 / 302 / 305 | Materials / energy / emissions | declared equipment dependency, combustion-session energy proxy, modal mix |
| GSTC D2 / D4 | Sustainability + visitor management | seasonal evenness, group-size disclosure (not scored); protected-area proximity remains a future signal, not yet captured |
| Travelife / Blue Flag | Operator certification (marine) | per-trip session log + eco metadata; coastal decision support |
Carbon reported as intensity and mix, not absolute tCO₂e, until the data earns it.
Institutions whose expertise aligns with the programme
Potential research collaborators, not confirmed partners. Listing an institution indicates topical alignment with the programme, not any affiliation, endorsement or existing agreement.
IMEDEA · ICM-CSIC · Ifremer · OGS · IPMA
SLF Davos · IGE Grenoble · Innsbruck · Eurac
Lab. d'Aérologie Toulouse · DLR
CMCC · WMO · ISGlobal · EEA · CAMS