What the field satellite indices holds
6 optical and 3 radar indices, averaged inside each field polygon every week. Optical: NDVI for greenness, EVI for greenness on dense canopies where NDVI itself saturates, SAVI for greenness with the soil signal damped down, NDWI for canopy and surface water, NDRE for chlorophyll in the red edge, GNDVI for greenness on the green band. Radar: backscatter across 2 polarisations in decibels, plus the radar vegetation index derived from them. Radar sees straight through cloud and responds to canopy structure and moisture rather than colour, so the 2 families disagree with each other in genuinely useful ways.
How a weekly value is made
Optical scenes over the polygon get cloud-masked and combined across a 15-day window, so a cloudy fortnight still produces a value; the index itself is then averaged inside the polygon. Radar scenes go through the same averaging. A 10 m optical pixel works out to 0.01 hectare, so a 2 hectare field holds about 200 pixels, plenty for a stable mean.
A year over the potato field
The chart tracks the Naivasha field across 49 weeks. NDVI climbs from 0.3 in April to 0.54 in June as the canopy closes, then falls through July and August with senescence and harvest, landing at 0.46 in September; NDRE follows the same shape later and lower, since the red edge saturates later on a dense canopy. Radar backscatter across the 2 polarisations converges as the crop grows and diverges again as it’s lifted. Rows before June 2026 carry the optical indices only, with no radar behind them yet.
Relationship to the county layer
The county-level demand indices run on this exact same optical product, just averaged over an administrative unit’s cropland instead of 1 field. The field is where you read a crop; the county is where you read a market. The anomaly against a field’s own history, specifically, is what the vegetation-anomaly advisory and the satellite tool in the connector both return; this dataset is simply where the underlying values live.