Blog

How many cloud-free satellite passes before a crop-type call is reliable

7 October 2026

A ripe wheat field under a wide sky with scattered clouds.

Every reviewing season the same question comes up in the file review room: how many clear passes does a parcel need before you can sign off on a crop-type call instead of flagging it for a field visit. There's no single number that works everywhere, but there's a defensible range, and it comes down to what the crop is doing, not how many satellites happen to fly overhead.

What one pass tells you

A single cloud-free pass gives you one reflectance signature on one date. That's a snapshot, and for anything beyond "is this bare soil or is this green," a snapshot isn't enough to separate maize from sunflower, or winter wheat from winter barley. Spectral signatures overlap heavily at certain growth stages, mid-vegetative growth especially, when half the crops on your declared-area list look like the same shade of green from orbit.

What separates one crop from another is the shape of the curve across a season: when it greens up, how fast it peaks, when it senesces. That shape only shows up across multiple dates, and the count that matters is how many of those dates fall across distinct growth stages, from emergence through peak cover to senescence.

The practical minimum

For most temperate row and cereal crops, three to five cloud-free passes spaced across the growing season is the practical floor: one near emergence, one at or near peak vegetative cover, one during senescence or harvest prep. Miss a stage entirely and the classifier is guessing from the two it has, which is where confidence drops and ambiguous parcels start piling up on your desk.

The complication paying agencies run into every year is that "cloud-free" and "scheduled" aren't the same thing. A wide-swath multispectral sensor might revisit a region every few days, but if a wet spring parks cloud cover over the same block for three straight weeks, you lose an entire phenological window no matter how often the satellite flies over. That's the gap that actually drives rework, not the nominal revisit rate on a spec sheet.

Why cloud gaps don't average out

A common assumption is that enough total passes in a season compensates for a few cloudy weeks. That assumption breaks down unevenly: losing the emergence window on a parcel costs more than losing a mid-season pass, because emergence timing is one of the stronger signals for telling spring-sown crops apart from autumn-sown ones. A parcel can have six clear passes on the calendar and still carry a weak crop-type call if all six land in the same six-week stretch of the season.

This is also why coverage numbers quoted on their own, five passes, ten passes, don't settle the question. A reliable call depends on whether the passes that did come through land across distinct growth stages. Five clustered passes in the same six-week window carry less signal than three passes spread across emergence, peak cover, and senescence.

What this means for the annual check

For a compliance review running against the full declared-area file, set the pass-count threshold against phenology coverage instead of a flat number. Treat parcels where cloud gaps swallowed a whole growth stage as weaker calls that deserve a second look, even if the nominal pass count looks fine on paper.

Crop Type Mapping builds each parcel's call from its full seasonal reflectance signature rather than a single date, pulling from wide-swath passes spread across the season instead of betting on one clear day. You can see how the deliverable sits next to a claim file on the Crop Type Mapping overview.

If you're trying to work out whether this season's parcel file needs a second pass before sign-off, that's worth a look.