MkulimaScore Geointelligence

Methodological Monograph · MS-GEO-METH-2026

Evidence, defined by sensor, scale, and provenance.

Every spatial and agronomic output published in MkulimaScore GEO declares its underlying satellite sensor or field enumeration source, spatial resolution, temporal cadence, and operational limits.

Optical Sensor Baseline
Copernicus Sentinel-2 L2A (10 m)
Hydro-Climatic Baseline
CHIRPS v2.0 (0.05° / ~5.5 km)
Topographic Baseline
NASA SRTM 1-ArcSec (30 m)
Published Cohort
1,147 Polygons · 6 Kenyan Counties

01 — Evidentiary Pipeline

Observe. Verify. Audit Scale. Publish.

How raw field geometries and multi-spectral satellite acquisitions are transformed into institutional intelligence dossiers.

  1. STAGE 01

    Boundary & Ground Capture

    Surveyed farm polygons are ingested with perimeter-to-area geometry checks, vertex counts, and ground-enumerated enterprise records.

    Audited Gate · MS-GEO
  2. STAGE 02

    Multi-Sensor EO Extraction

    Sentinel-2 L2A (10 m optical/NIR/SWIR), CHIRPS v2.0 pentadal rainfall, and NASA SRTM 30 m terrain layers are extracted over validated geometries.

    Audited Gate · MS-GEO
  3. STAGE 03

    Scale & Quality Auditing

    Scene classification (SCL) cloud masks, mixed-pixel edge buffers, and minimum valid parcel size thresholds are applied before index computation.

    Audited Gate · MS-GEO
  4. STAGE 04

    Institutional Publication

    Every published metric carries its provider, acquisition window, spatial scale, and explicit interpretive limits inside the 7-section farm dossier.

    Audited Gate · MS-GEO

02 — Radiometric & Spectral Formulations

Five calibrated indices across the 7-section farm dossier.

Rather than relying on a single greenness proxy, MkulimaScore GEO evaluates five complementary spectral dimensions so analysts can separate canopy density, leaf water content, soil background reflectance, and surface moisture.

IndexFull NameBand FormulationSensor & ResolutionTypical RangeAgronomic Interpretation
NDVINormalized Difference Vegetation Index(B08 − B04) / (B08 + B04)Sentinel-2 NIR (842 nm) & Red (665 nm) · 10 m−0.20 to +0.92Primary indicator of chlorophyll density and active green biomass across highland perennials (coffee, tea) and seasonal cereal canopies.
NDMINormalized Difference Moisture Index(B08 − B11) / (B08 + B11)Sentinel-2 NIR (842 nm) & SWIR-1 (1610 nm) · 10/20 m−0.40 to +0.60Measures canopy water content and early-onset moisture stress before visible yellowing occurs in the red spectrum.
EVIEnhanced Vegetation Index2.5 × (B08 − B04) / (B08 + 6·B04 − 7.5·B02 + 1)Sentinel-2 NIR, Red & Blue (490 nm) · 10 m0.00 to +0.85Corrects for atmospheric aerosol scattering and canopy background saturation in dense highland coffee and tea zones (Nyeri, Meru).
NDWINormalized Difference Water Index(B03 − B08) / (B03 + B08)Sentinel-2 Green (560 nm) & NIR (842 nm) · 10 m−0.50 to +0.45Detects surface water ponding, riparian buffer proximity, and irrigated plot saturation in lowland and valley-bottom parcels.
SAVISoil-Adjusted Vegetation Index1.5 × (B08 − B04) / (B08 + B04 + 0.5)Sentinel-2 NIR & Red with L=0.5 canopy factor · 10 m0.00 to +0.75Minimises bright soil background reflectance during early emergence and in semi-arid corridors (Laikipia, Isiolo) where canopy closure is below 40%.

03 — Operational Status by Evidence Stage

Distinguishing operational readiness from ongoing validation.

The published 1,147-farm cohort establishes spatial boundary readiness and multi-sensor environmental integration today. Outcome-linked loan default backtesting is maintained as a separate, explicitly labelled validation track.

Evidence StageInstitutional Scope & Responsible ClaimCurrent Status
01 · Spatial Boundary Readiness1,147 surveyed farm polygons across 6 Kenyan counties with area, perimeter, vertex count, and administrative hierarchy.Operational · Published
02 · Multi-Sensor EO & Climate ExtractionSentinel-2 L2A (NDVI, NDMI, EVI, NDWI, SAVI), CHIRPS v2.0 rainfall anomalies, and SRTM 30 m terrain profiles.Operational · AlmaEarth Adapter
03 · Seven-Section Farm Intelligence DossierUnified parcel inspection drawer combining overview, geometry, EO time series, climate, terrain, agronomy, and role-gated risk.Operational · v1.0 Blueprint
04 · Multi-Season Loan Outcome BacktestingOut-of-time statistical validation linking pre-loan spectral trajectories to realised repayment performance.Active Research Cohort

04 — Canonical Measure Specifications

Public measure sheets and provenance metadata.

Each measure below documents its primary data source, spatial resolution, update cadence, aggregation rule, and known interpretive limits.

Registered farmers

count

Distinct farmer identities in the approved publication snapshot for the selected scope.

Primary Source
Approved public publication snapshot
Spatial Resolution
County (publication); subcounty/ward navigation is demonstration-only
Refresh Cadence
Monthly publication refresh target
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Publication as-of date
Validation Status
Demonstration dataset — not operational registrations
Aggregation Method
Distinct identity count within scope
Quality Exclusions
Private identities withheld; duplicates collapsed in publication build
Applicable Enterprises
All published value chains in scope
Known Failure Cases
County totals can exceed mapped farm points when locations are missing
Governance Licence
Synthetic demonstration licence (evaluation only)

Registered farms

count

Distinct farm records; one farmer may contribute multiple farms.

Primary Source
Approved public publication snapshot
Spatial Resolution
County
Refresh Cadence
Monthly publication refresh target
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Publication as-of date
Validation Status
Demonstration dataset
Aggregation Method
Distinct farm_id count
Quality Exclusions
Farms without publication approval are omitted
Applicable Enterprises
All published value chains in scope
Known Failure Cases
Intercropped farms still count once as a farm record
Governance Licence
Synthetic demonstration licence (evaluation only)

Mapped locations

count

Farms with a usable map point in the publication. A usable coordinate is not the same as a field-verified location.

Primary Source
Approved public publication snapshot
Spatial Resolution
Point location when present; otherwise unmapped
Refresh Cadence
Monthly publication refresh target
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Publication as-of date
Validation Status
Demonstration dataset; field verification is a separate metric
Aggregation Method
Count of farms with usable coordinates / total farms
Quality Exclusions
Points failing coordinate QA are treated as unmapped
Applicable Enterprises
All published value chains
Known Failure Cases
Centroid placeholders can inflate mapped counts if QA is weak
Governance Licence
Synthetic demonstration licence (evaluation only)

Enterprises

count

Distinct enterprise activities (value-chain instances) linked to farms in scope.

Primary Source
Approved public publication snapshot
Spatial Resolution
County
Refresh Cadence
Monthly publication refresh target
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Publication as-of date
Validation Status
Demonstration dataset
Aggregation Method
Distinct enterprise records; not physical land area
Quality Exclusions
Unpublished enterprise types omitted
Applicable Enterprises
Catalogue value chains only
Known Failure Cases
Mixed-chain counties show multiple enterprises per farm
Governance Licence
Synthetic demonstration licence (evaluation only)

Reviewed boundaries

count

Farm or plot boundaries that passed a documented review step.

Primary Source
Verification workflow outcomes in the publication build
Spatial Resolution
Farm/plot geometry when reviewed
Refresh Cadence
With each publication refresh
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Review event date collapsed into publication as-of
Validation Status
Demonstration workflow outcomes
Aggregation Method
Count of reviewed boundary records
Quality Exclusions
Provisional corrections pending review are excluded
Applicable Enterprises
All
Known Failure Cases
Reviewed count can lag mapped points
Governance Licence
Synthetic demonstration licence (evaluation only)

Usable observations

count

Publication metric: farms (or observation readiness rows) marked as having usable environmental observation readiness in the coverage snapshot. This is not the same population as usable satellite composites shown in the Conditions evidence panel.

Primary Source
Connected observation providers or demonstration observation set
Spatial Resolution
County composite or provider footprint
Refresh Cadence
Provider-dependent; publication lists available dates
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Observation date / composite window
Validation Status
Separate from forecasts
Aggregation Method
Count of usable observation rows in scope
Quality Exclusions
Clouded, failed QA, or out-of-season rows excluded
Applicable Enterprises
Depends on indicator (e.g. NDVI not for all livestock)
Known Failure Cases
Availability of a provider is not the same as a condition change
Governance Licence
Provider-specific; demonstration set is labelled

Vegetation index (NDVI 16-day)

index

Sixteen-day NDVI composite for the selected geography, with prior composite and delta when available.

Primary Source
EO adapter (demonstration composites labelled synthetic; live mode when connected)
Spatial Resolution
County centroid / composite footprint — not farm-plot NDVI unless secure join is authorised
Refresh Cadence
Adapter available-dates list
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Composite end date; prior composite for change
Validation Status
Observation — not a yield estimate
Aggregation Method
Per-geography composite value; change = current − prior
Quality Exclusions
Cloud-contaminated or low usable spatial coverage flagged separately
Applicable Enterprises
Crop / vegetation contexts; not a dairy, poultry, livestock or aquaculture health measure
Known Failure Cases
Missing prior composite prevents change; cloud can null the value
Governance Licence
Provider-specific; demonstration set is labelled

Precipitation outlook

mm

Daily precipitation outlook for the selected geography. Labels distinguish forecast occurrence from observed or verified impact.

Primary Source
Open-Meteo forecast adapter when live; otherwise disconnected
Spatial Resolution
County centroid point forecast — not farm- or ward-level rainfall fields
Refresh Cadence
Retrieved on request with cache TTL
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Seven-day forecast period; application validity may be shorter (e.g. three-day refresh window)
Validation Status
Forecast — separate from incidents and verified impacts
Aggregation Method
Summarised over the stated forecast period; production alert rules are not published.
Quality Exclusions
Not an official Kenya authority warning; not verified impact
Applicable Enterprises
Context for all agricultural inventory under the outlook
Known Failure Cases
Centroid representativeness fails in spatially heterogeneous rainfall
Governance Licence
Open-Meteo terms; indicative meteorology only

County agricultural inventory under outlook

count

Demonstration farm and enterprise counts for a county that sits under an active meteorological outlook. Not physical susceptibility from geometry intersection with a hazard footprint.

Primary Source
Public publication snapshot joined to outlook by county
Spatial Resolution
County
Refresh Cadence
With outlook and publication refresh
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Outlook as-of / publication as-of
Validation Status
Illustrative inventory context
Aggregation Method
County farm and enterprise totals under outlook
Quality Exclusions
Does not claim footprint intersection or verified impact
Applicable Enterprises
Published enterprises in the county
Known Failure Cases
Misreading inventory as susceptibility or damage
Governance Licence
Synthetic demonstration licence (evaluation only)

Opportunity selection criteria

brief

Documented gaps and concentrations that meet stated criteria in the publication opportunity brief — including need, evidence, proposed intervention, verification-first steps, success measures and requested partner contribution.

Primary Source
Approved opportunity briefs in the publication snapshot
Spatial Resolution
As stated on the brief (typically county × value chain)
Refresh Cadence
With publication refresh
Expand Full Technical Provenance & Failure Cases
Temporal Coverage
Brief status date within publication
Validation Status
Diligence required before operational decisions
Aggregation Method
Matching briefs for current county and value-chain selection
Quality Exclusions
Draft/demonstration briefs labelled; not investment-ready labels
Applicable Enterprises
Brief value chain only
Known Failure Cases
Treating a title alone as a funding recommendation
Governance Licence
Synthetic demonstration licence (evaluation only)

MkulimaScore GEO · Institutional Engagement

Inspect these methods on live mapped farms.

Open the Nyeri Arabica coffee corridor or Nakuru highland cohort to examine real Sentinel-2 L2A, CHIRPS, and SRTM dossiers.