Satellite Crop Monitoring in Africa: NDVI, NDWI and NDMI

Published by Faminga. Prepared with AI-assisted research and drafting; sources checked on 5 September 2026. Product descriptions are attributed to their providers. Examples and evaluation checklists are editorial guidance, not results of a comparative field trial.

Satellite crop monitoring helps African farmers and cooperatives see patterns across fields and decide where to inspect first. It is most useful when the image date, resolution and limitations are visible. A green or red patch is an observation to investigate, not a complete explanation of crop health.

This guide explains NDVI, NDMI and the ambiguous term NDWI, then shows how to connect a satellite signal with a field visit and soil data. The aim is a decision that somebody can verify on the ground.

What NDVI measures

The Normalized Difference Vegetation Index, or NDVI, compares near-infrared and red reflectance: NDVI = (NIR − red) / (NIR + red). USGS describes it as an indicator of vegetation greenness, while noting limitations including sensitivity to soil background and saturation in dense vegetation. See the USGS NDVI explanation.

That makes NDVI useful for examining change, but it is not a direct disease label. Interpretation needs crop context. A change after harvesting has a different meaning from an unexpected change during active growth. Comparing fields planted on different dates can be misleading without accounting for their growth stages.

Ask the service to show a time series rather than a single coloured map. Keep the underlying image dates accessible. The date when the dashboard refreshed is not necessarily the date when the satellite observed the field.

NDMI, NDWI and vegetation moisture

USGS defines the Normalized Difference Moisture Index, or NDMI, using near-infrared and shortwave-infrared reflectance: NDMI = (NIR − SWIR) / (NIR + SWIR). It is used to describe vegetation water content. It does not directly provide a root-zone soil moisture percentage.

Names require care. A USGS-listed paper on index terminology notes that the NIR–SWIR ratio has several names, including NDWI and NDMI. Other water-index formulations use different bands. Ask the provider which formula and bands its “NDWI” layer uses before interpreting the legend.

For a farmer, the practical requirement is plain language beside the index: what is being observed, what might affect it and what needs checking. An unexplained decimal is not an irrigation instruction.

Why pixel size matters on a small plot

ESA's Sentinel-2 documentation lists bands with 10, 20 and 60 metre spatial resolutions. The red and broad near-infrared bands are at 10 metres; the shortwave-infrared bands used in common moisture calculations are at 20 metres.

A 10-by-10-metre pixel covers 100 square metres. A 20-by-20-metre pixel covers 400 square metres. As an illustrative calculation, a square half-hectare plot has about 50 or 12.5 pixel-area equivalents at those respective resolutions. Actual usable pixels depend on the field's shape, grid alignment, boundary and image quality; those figures are not a count of independent measurements inside a real plot.

A narrow or irregular plot can contain many mixed pixels that include paths, trees or neighbouring crops. Enlarging the picture on a phone does not create new observations. Ask whether the service reports how much usable imagery actually falls within the field.

Clouds, shadows and missing observations

An optical satellite schedule does not guarantee a usable farm image on every pass. Check whether clouds and shadows were excluded and how much of the field remains. A report should distinguish “no reliable observation” from “no change.”

For a practical workflow, set a maximum acceptable age for the decision being made, with agronomic input where appropriate. If the newest reliable image is too old, the next step may be a field visit rather than another automated conclusion.

When a provider combines observations into a composite, ask what time period it represents. A composite can be helpful, but its interpretation differs from a photograph taken at one moment.

Turn a map into a field inspection

Start by checking the field boundary and crop record. Then identify a meaningful change or unusual area. Ask someone to inspect that location and a comparison area, recording photographs and observations with dates. Keep the satellite image that triggered the visit.

The visit should investigate plausible explanations rather than confirm the software's first suggestion. Record recent harvesting, planting, irrigation, disturbance and other relevant events. If the observation cannot be explained, retain it as unresolved instead of assigning a convenient label.

A cooperative can use this process to prioritise visits when staff time is limited. The value is the quality of the prioritisation and follow-up, not the number of maps produced.

Combine satellites with soil sensors carefully

A satellite observes spatial patterns; a soil sensor reports at a particular point and depth. They provide different perspectives. To compare them, connect both to the correct field and timestamp, and record the sensor's location.

If the signals disagree, first check the measurement context. Was the image captured before irrigation? Is the sensor in a different management area? Is either observation stale? Disagreement can guide an investigation, but it should not be hidden by averaging unrelated values.

Use our soil sensor guide for the ground measurement side and the smart irrigation guide for turning evidence into water-management decisions.

A Rwanda example and a useful supplier checklist

FAO's KnoWat programme in Rwanda used WaPOR-based remote sensing to assess water use and productivity in the Yanze, Muvumba and Lower Akagera catchments. This is an example of geospatial information informing agricultural water assessment. It does not imply that the same product resolves every individual smallholder plot.

Ask any provider for the data source, band formula, observation date, cloud handling, pixel size, field coverage and export options. Request an example where the system could not make a reliable assessment. That reveals how uncertainty reaches the user.

For Faminga or another service, evaluate the complete field workflow rather than assuming that a satellite layer includes a validated yield model or automatic disease diagnosis. Our Rwanda precision agriculture guide and farm data guide explain how to connect these observations with accountable decisions.