Building a Sentinel-2 NDVI workflow for scouting

The workflow turns imagery into a decision layer

A Sentinel-2 NDVI workflow turns field imagery into an NDVI map that can guide scouting and later analysis. The core job is not flying a drone or applying a product. It is moving from imagery to a vegetation-index layer, then using that layer to plan field checks.

1. DJI Agras T100 100 L
2. DJI Agras T70P 70 L
3. XAG P150 70 L
4. DJI Agras T55 50 L
5. XAG P100 Pro 50 L
6. DJI Agras T50 40 L

Spray tank capacity (L)

Model	Value
DJI Agras T100	100 L
DJI Agras T70P	70 L
XAG P150	70 L
DJI Agras T55	50 L
XAG P100 Pro	50 L
DJI Agras T50	40 L

Sources: ag.dji.com, ag.dji.com, xa.com, ag.dji.com, xa.com, ag.dji.com

The supplied research shows this split in practice. Imagery was processed into RGB mapping and vegetation indices, including NDVI. The resulting growth map showed variation across a cotton farm and helped the grower place sampling points for further inspection.

That sequence matters. The NDVI map guided the next field task rather than replacing it. Our guide to what NDVI drones measure, and what they do not covers that boundary in more depth.

A Sentinel-2 NDVI workflow should therefore produce clear outputs for scouting:

• the source imagery;
• the processed NDVI layer;
• zones or areas marked for review;
• a record of field checks made from the map.

This scope stops before spraying, spreading or other field work. An NDVI layer may inform a later prescription map, but those are separate products.

The distinction prevents a common workflow error. A vegetation map shows spatial variation. It does not, by itself, set an application rate or tell an aircraft to act.

A useful map starts with a fixed data path

The best workflow is one that operators can repeat without changing its logic between field visits. Each run should use the same broad path: collect imagery, process it, inspect the output, mark zones and check them in the field.

The supplied material does not document Sentinel-2 band definitions, pixel size, revisit interval, product levels or a download route. It also does not set out a supported NDVI formula. Those details cannot be prescribed from the available sources.

That gap should shape the build. Record the choices made in the source-data stage rather than treating them as hidden software defaults. If a team later changes the imagery source or processing method, the record will show where the workflow changed.

Workflow stage	Input	Output	What to keep stable
Imagery intake	Sentinel-2 imagery or verified drone imagery	Accepted source set	Source type and field area
Map processing	Accepted imagery	NDVI layer	Processing sequence and settings
Map review	NDVI layer	Areas marked for checking	Review method
Field scouting	Marked areas	Field observations	Sampling approach
Archive	Imagery, map and observations	Comparable field record	File names and saved settings

Do not merge different source types without noting the change. Sentinel-2 imagery and drone imagery may both sit in the same scouting system, but they come through different collection routes. The practical comparison is covered in satellite versus drone NDVI.

A simple archive is part of the workflow, not an office task added later. Keep each accepted image set with its NDVI output and scouting notes. That gives the next map a known point of comparison.

Analytics should narrow the scouting task

The first NDVI map is a baseline layer. Its value grows when the same review method turns visible variation into a clear field task.

The research gives a useful model. A grower used an NDVI growth map to segment cotton fields into growth zones. Those zones then guided the placement of sampling points for closer inspection.

That is a sound role for analytics: reduce a whole-field image to areas that deserve attention. The map points the scout towards variation. The scout then checks what is present on the ground.

A repeatable review process can ask the same questions after each acquisition:

• Where does the current map show variation within the field?
• Which zones need a field check?
• Have the marked areas changed since the prior accepted map?
• Did the imagery source or processing path also change?
• Do field observations support the map-based reading?

The method used to make zones should stay fixed when maps are compared. If zone rules change between acquisitions, a different map may reflect a different method rather than a crop change.

The supplied research does not provide zone thresholds. It also does not set a rule for turning a zone into a treatment rate. Operators should not invent either and present it as a feature of Sentinel-2 or NDVI.

For the wider step from map to farm action, see turning drone imagery into decisions. The key hand-off is from a coloured layer to a check that someone can carry out in the field.

Processing belongs between collection and inspection

Imagery processing should be planned as part of field work. It sits after mission planning and collection, but before map-based scouting and any later prescription work.

Commercial drone systems reflect this split. The supplied product data lists flight planning for Wingtra WingtraOne GEN II through WingtraPilot. Other supplied material groups an imaging aircraft under mission planning and separates that role from crop protection.

That is a useful way to arrange the farm workflow:

Field phase	Main task	Decision produced
Before collection	Set the field area and collection plan	What imagery is needed
After collection	Check and process the imagery	Whether an NDVI layer can be accepted
Map review	Mark variation and growth zones	Where scouts should look
Field inspection	Check marked areas	What is present in the crop
Later field work	Build and check any prescription	Whether and how to act

Keep the map review separate from the collection flight. A completed flight does not mean a usable map exists. The imagery still has to pass through the chosen processing path and produce an output fit for review.

Keep actuation separate as well. The prescription remains another step, with its own checks.

For readers building the wider collection process, precision agriculture mapping with drones sets the imagery stage in context. This article stays with the narrower path from accepted imagery to repeatable NDVI review.

Sensor evidence should decide the aircraft shortlist

Choose collection hardware from verified sensor data, not from a general claim that an aircraft can map. Of the supplied products, DJI Mavic 3M has the clearest direct evidence for multispectral and NDVI work.

DJI Mavic 3M records green at 560±16 nm, red at 650±16 nm, red edge at 730±16 nm and NIR at 860±26 nm. Its multispectral sensor is a 1/2.8-inch CMOS unit with 5 MP effective resolution and a maximum image size of 2592×1944 in TIFF.

It also has a built-in light-sensor module. Its multispectral video content includes NDVI, GNDVI and NDRE. Fixed RTK positioning accuracy is listed as 1 cm + 1 ppm horizontally and 1.5 cm + 1 ppm vertically.

Those specifications establish a verified multispectral collection context. They do not document how Sentinel-2 imagery should be downloaded or processed.

Aircraft	Verified mapping evidence	NDVI sensor evidence in supplied data	Safe conclusion
DJI Mavic 3M	Multispectral and RGB sensors, RTK	Multispectral bands and NDVI video are listed	Directly relevant to NDVI collection context
Wingtra WingtraOne GEN II	WingtraPilot flight planning and mapped-area figures	No multispectral or NDVI sensor specification is supplied	Do not assume NDVI support
XAG P150	Survey capacity of up to 13.3 ha per flight	No multispectral or NDVI sensor specification is supplied	Mapping evidence alone is not NDVI evidence

RTK, flight time and mapped area may matter to a survey plan. They do not prove that an aircraft records the imagery needed for an NDVI product. The sensor must be checked separately.

The same rule applies to RGB cameras. A listed camera is not enough evidence for a multispectral workflow. See agricultural drone sensors explained before treating any camera specification as proof of NDVI support.

Stable processing makes change maps worth reading

Comparisons only work when the team can tell crop change from workflow change. Keep the imagery intake, processing path and review method stable wherever the operator has control.

Start by saving the source set used for every accepted map. Do not overwrite it with a later download or a reprocessed version. Keep the NDVI output beside the source, along with the settings and field notes tied to that run.

When a processing choice changes, mark the break clearly. The next map may still be useful for scouting, but it should not be treated as a clean continuation without review.

Use a short acceptance check before comparing maps:

• Confirm that the map covers the intended field area.
• Confirm that the expected processing path was used.
• Check whether the imagery source changed.
• Check whether the zone method changed.
• Save the field observations made from the map.
• Keep any later prescription as a separate file.

Repeatability is the practical test of the workflow. A map that cannot be traced to its source and settings may still look convincing, but its changes cannot be checked with confidence.

A sound Sentinel-2 NDVI workflow is therefore modest in scope. It accepts imagery, makes an NDVI layer, marks areas for scouting and keeps a clear record. Drone imagery can add another collection stream where the sensor evidence supports it, but it should not blur the line between mapping, field diagnosis and application.