A · ACCURACY & VALIDATION
How good is it, and how would I know?
We don't lead with a single accuracy percentage. Instead, every district value is delivered as a calibrated prediction interval built with conformal methods, so the band carries a guaranteed coverage rate — when we say a value sits inside its interval with a stated probability, that frequency actually holds out of sample. So you underwrite against the margin, not against a point estimate you're asked to trust blindly.
- Independent reference. Downscaled water-storage trends are benchmarked against national geological-survey monitoring-well networks, district by district.
- Honest failure mode. A district that fails the agreement check is flagged "insufficient agreement" and is not scored — rather than shipped with a flattering number.
- Continuous, not one-off. The benchmark is re-run with each new data release and ships as a validation pack, kept current for the life of the contract.
B · RESOLUTION & UPDATE FREQUENCY
How fine, how often?
Spatial: delivered at administrative-district (LAU) resolution, downscaled from a native orbital footprint that spans hundreds of kilometres.
Temporal: refreshed on the cadence of the underlying orbital signal — a new measurement each cycle, with no fabricated values invented between releases.
C · INTEGRATION
How does it reach my systems?
- API — district indicator plus interval, queried by code or region ID.
- GeoTIFF — georeferenced raster for direct load into GIS.
- PDF report — portfolio or district brief for committee and disclosure files.
- Ready-to-read map layers — for analysts who do not run a GIS; files and API, not a portal.
Vector and raster exports import into standard GIS and risk platforms. Tabular feeds map onto existing scoring pipelines.
D · TECHNOLOGICAL LIMITATIONS — HANDLED HONESTLY
Where the method is weak, and what we do about it
The native orbital footprint is coarse. A single raw cell can span an area the size of half a country, and we don't pretend otherwise — the downscaling step is physically constrained and reports its own uncertainty, so a district interpolated from a sparse signal carries a visibly wider band than one anchored by dense ground reference. Districts where the signal can't be resolved with confidence get flagged instead of silently filled in.
From orbit what we're actually reading is the change in water mass itself, and on its own that can't separate deep aquifer change from soil and surface water. We mitigate this by combining it with independent surface-balance signals and calibrating against monitoring wells, which lets the delivered indicator isolate the structural groundwater component rather than reporting a raw mass total.
E · COMPLIANCE & STANDARDS
Where it fits disclosure
Outputs are designed to support CSRD and EU Taxonomy water-risk disclosure with an auditable chain from signal to figure. Processing and delivery run under EU data residency.
What we claim is relevance, not certification. We don't hold a regulatory seal, and we say so plainly. The validation pack is built so your auditors can replay any figure we issue.
F · DELIVERY FORMATS
What you actually receive
- API endpoint with value + interval + validation flag
- GeoTIFF raster layers
- Vector district tables (CSV / GeoPackage)
- PDF portfolio and per-district briefs
- API and map-layer access
G · HORIZON
What time frame the indicator speaks to
This is a structural trend product, not a short-horizon forecast. It quantifies the multi-year slope of water storage — how fast reserves are gaining or draining once the season is removed — along with the statistical significance of that slope. A multi-year loan, a renewing treaty, a long-lived abstraction licence: that slope is exactly what each of them is exposed to.
