Agricultural verification at portfolio scale

One engine, three orders of magnitude

The portfolio behind this work spans a 1,213-fold range of parcel size — from a 1.62 ha orchard to a 1,965 ha irrigation command area — across eleven sites in three provinces. Coastal delta, irrigated plain, arid rain-fed plain, foothill orchard and temperate mountain, all read by the same pipeline with the same tests.

That range is the point. A verification method that works only on large blocks cannot serve a smallholder portfolio, and one tuned to a single agro-ecological zone cannot serve a national programme. The determinations below have been produced at both ends of the range.

What gets determined, per parcel

• Cropped or fallow, by season, from the vegetation index series rather than a single date.
• Irrigated or rain-fed, from surface moisture behaviour against the local rainfall record.
• Cropping calendar — sowing and senescence timing against the zone's norm, which is what exposes a parcel that is out of step.
• Performance against its own history and against neighbouring parcels in the same zone.
• Boundary agreement — whether the cultivated extent matches the declared extent.
• Climate envelope — chill hours, frost risk, heat accumulation and water deficit from a 31-year reanalysis baseline.
• Terrain suitability — slope, aspect, drainage and flood exposure.
• Change since last season, so a portfolio is re-verified rather than re-surveyed.

Four layers, each priced against something already being paid for

The capability sells as a ladder, and two rules govern it. Every layer replaces a cost the client already carries — never an abstract "value of insight", because a credit committee can approve a substitution and cannot approve a valuation. And each layer builds on the one beneath it, so a client who starts at origination is already most of the way to portfolio-level analysis — the engagement compounds instead of starting over.

The economics of imagery invert with scale

Very-high-resolution satellite tasking is sold against minimum order areas — typically tens to hundreds of square kilometres. A single farm of a few hundred acres is far below that floor, so it pays for the minimum regardless: an effective penalty that our procurement study measured at between twelve and fifty times the notional per-area rate.

A lender's or a programme's served footprint is the opposite case. Aggregate the parcels and the same footprint clears the vendor minimum many times over, at which point a high-resolution baseline across the entire book becomes economically sensible for the first time. The constraint that puts premium imagery out of reach for one farm is dissolved by a portfolio — because the portfolio is the aggregation.

None of it is required for the baseline. Every determination listed above is produced from the public archive, so premium tasking stays what it should be — an upgrade for a specific question, scoped separately, never a dependency.