How to monitor crop health using satellites

Step 1: Get the field boundary right

Every number you'll look at later is an average or a comparison inside the boundary, so a sloppy outline costs you accuracy all season. Drawing on a basemap is fine for a handful of fields. For more, export the boundaries from your farm management or machinery software as KML, KMZ or a shapefile (zip the .shp, .shx, .dbf and .prj files together) and upload them in one go.

Keep the edge out of your averages

A 10 m pixel on the boundary line usually mixes your crop with a headland, a track or the next field, which drags down the edge of almost every map. Ignore the outer ring of pixels when you look at field averages, or draw the boundary a few meters inside the real edge. On a 40-acre block you lose very little area.

Split fields that aren't managed as one

If half a field was sown two weeks later, or one end carries a different variety, map the parts separately. Otherwise the later half looks worse all season and the map keeps flagging a difference you already know about. Note the crop, variety and sowing date for each field too, since reading a map against growth stage depends on it.

Step 2: Choose the imagery and the index

For most fields the starting point is Sentinel-2, the European Space Agency's optical mission. It photographs most farmland about every five days, with 10 m pixels in the blue, green, red and near-infrared bands and 20 m pixels in the red-edge and shortwave-infrared bands. Landsat 8 and 9 add 30 m imagery and a thermal band. Sentinel-1 radar covers cloudy weeks, and commercial constellations sell near-daily imagery at around 3 m for high-value crops.

Use atmospherically corrected surface reflectance (Level-2A for Sentinel-2). If you skip that correction, a hazy day shifts your index values and two dates become hard to compare. Then pick the index that answers your question:

NDVI is (NIR − Red) / (NIR + Red) and runs from −1 to +1; the USGS explanation of NDVI and our walkthrough of the NDVI formula cover the math. For water, see the USGS page on NDMI and our guide to NDWI, NDMI and evapotranspiration. Whatever index you use, keep a true-color image open beside it. That's how you tell a weak crop from a cloud shadow, a flooded hollow or a strip that was just cut.

Step 3: Read the crop health map

Compare zones inside the same field first

Read an NDVI map in relative terms. A wheat field at 0.55 might be fine in early spring and a worry by late May, so the number alone says little. What matters is a zone clearly below the rest of the same field on the same date. Most tools stretch the colors across the field's own range, which makes differences easy to spot but can make a very even field look patchy. Check the scale before you worry: a field running from 0.62 to 0.68 is in good shape even if half of it is orange.

As a rough guide for NDVI, bare soil sits around 0.1 to 0.2, a young or sparse crop around 0.2 to 0.5, and a dense, healthy canopy between 0.6 and 0.9. Water and snow fall below zero, which our note on negative NDVI values explains.

Read the shape of the pattern

The outline of a weak zone is often the best clue to its cause:

Look at the trend over several passes

Thin cloud, smoke or a field irrigated that morning can pull a reading down for a day. Before you act, put the latest map next to the previous two passes and look at the zone's season curve. A real problem keeps falling away from the rest of the field. A glitch recovers on the next clear image.

Step 4: Decide how often to look and what triggers a visit

Satellite crop health monitoring works best on a routine. During rapid growth and around flowering, look at every usable pass. Early in the season and near harvest, once a week is enough, and a late-season drop in greenness is usually just the crop maturing.

Alerts help only if they're rare enough that people act on them. A reasonable starting rule: flag a zone that sits 10 to 15 percent below the field median on NDVI or NDRE for two clear passes in a row and covers at least half an acre. Adjust the rule once you see how often it fires for nothing.

Step 5: Scout to confirm what the map shows

The map shows you where to go, but only someone standing in the crop can say what's wrong. Open the map on your phone so your position shows over the weak zone, walk straight to it, and then:

• Visit the weak zone and a healthy zone nearby on the same day. The contrast between the two is what makes the diagnosis.
• Walk the edge of a spreading patch, where pests and disease are usually most active.
• Dig a hole. Look at root depth, any compaction layer and how moist the soil is at 15 and 30 cm.
• Check whether yellowing is on older leaves or new growth; each points to different nutrients.
• Look at the irrigation in that zone: emitters, nozzles, pressure.
• Take a geotagged photo and a short note so the observation sits on the same map as the image.

If you find nothing wrong, note it (a sandy patch that always reads low, say) and stop chasing it. For nutrient questions, separate tissue or soil samples from the weak and healthy zones are often the cheapest way to settle it. Our article on crop stress detection covers the common causes.

Step 6: Act on the zone and keep a record

Once you know the cause, treat the area that needs it where you can: a spot spray on the infested patch, new nozzles on one pivot span, or a variable-rate nitrogen map drawn from the NDRE zones. Then watch the next two or three passes. If the zone catches up, the fix worked. If it keeps falling, go back out.

Keep a simple log for each field (a spreadsheet is fine): image date, what the map showed, what you found and what you did. After one season you'll know which patterns are real problems and which are permanent features, like a sand lens or an old fence line. After two or three you can see whether the poor zones are shrinking. The same history is handy when a lender, insurer or buyer asks how the crop was managed, which is where satellite checks for crop loans and insurance come in.

Cloud, haze and when to use radar

Cloud is the biggest practical limit on optical crop health monitoring. A five-day revisit is a chance of a picture every five days, and in a tropical rainy season you can go weeks without a clear look. Three things help.

First, mask what you can't trust. Sentinel-2 Level-2A data comes with a scene classification layer that marks cloud, cloud shadow and snow. Shadows are the trap: they make a healthy crop look stressed, so a dark blob next to a bright one in true color deserves suspicion.

Second, use radar. Sentinel-1 carries a C-band radar that sends its own signal and records the echo, so it works through cloud and at night. The echo depends on the canopy's structure and water content and on surface soil moisture. Radar vegetation indices follow crop growth reasonably well but are noisier than optical ones and react to wet leaves and row direction, so read them as a trend over several dates.

Third, combine sources. Landsat and Sentinel-2 together give more chances of a clear image, and merging optical with radar keeps the record going through the wet season. Farmonaut does this for every field it monitors.

Common mistakes when monitoring crop health from satellite images

Most bad calls we see come from a few habits:

• Reading NDVI late in the season. Once a maize or sugarcane canopy closes, NDVI saturates and the whole field reads high. Switch to NDRE for in-season nitrogen decisions; our post on NDVI vs NDRE explains why.
• Comparing fields that aren't comparable. Different crops, varieties and sowing dates give different values on the same day.
• Acting on one image. Wait for the next clear pass, and check true color for cloud shadow before anyone goes out.
• Treating the map as the diagnosis. A red zone is a reason to walk the field. It isn't a reason to spray it.
• Not writing anything down. Without notes you'll scout the same sandy corner every season.
• Expecting real-time data. Public satellites give a new view a few times a week at best. For hourly soil moisture, use a probe; IoT crop monitoring pairs well with satellite maps.

Comparing tools for crop health monitoring using remote sensing

There are five broad ways to get the data, and many farms combine two. The right mix depends on your acreage, who reads the maps and whether you want advice or only imagery.

For whole farms watched all season, satellites are usually the cheaper base layer, with drones (see our piece on drone NDVI mapping) and probes added where they answer a specific question. Our roundup of crop scouting apps covers the wider app market.

What to ask before you choose a crop health monitoring system

Nearly every product shows a greenness map now. These questions separate them:

1. Does it mask cloud and cloud shadow automatically, or will you have to spot them yourself?
2. Can you see each field's season curve and compare any two dates?
3. How far back does the image history go? Past seasons tell a recurring weak zone from a new one.
4. Is there radar for cloudy weeks?
5. Does it suggest an action in usable units, such as millimeters of water, or only show colors?

Then try it on a field you know well. If the map matches what you know about that ground, you can start trusting it elsewhere.