Almost every serious agricultural decision rests on knowing what’s planted where and how it’s faring. Shortfall warnings, yield forecasts, insurance payouts: all of it traces back to a map. Optical satellite archive is the main source for these maps, but sometimes it is not enough.
Optical and SAR sensors work on different principles. Optical instruments measure sunlight reflected off the canopy, so they read chemistry: chlorophyll, pigment, moisture inside the plant. The drawback: it needs daylight and a clear sky. If you purchase SAR satellite imagery, it will supply its own microwave signal, times what bounces back, and measures structure instead — stem geometry, biomass, water content. Cloud and darkness are irrelevant to it.
Optical Remote Sensing: Rich Spectral Detail, Hostage to the Weather
Sentinel-2, Landsat and MODIS have done the heavy lifting in agricultural monitoring for decades, and they’ve earned it. Each one measures sunlight bouncing back off the surface in separate spectral bands, and that’s what makes crop physiology legible from orbit — how much chlorophyll a canopy holds, how fast biomass is piling on, whether the plants are short of water, when they start to die back.
NDVI still sits at the center of most of this work. Rouse and his co-authors worked it out in 1974 for Landsat-1, and it has survived fifty years because the physics behind it is hard to argue with. Photosynthesizing plants swallow red light and throw near-infrared straight back; a wilting or diseased canopy doesn’t. Put those two bands against each other on a scale from -1 to 1, and the contrast holds up well enough for fertility maps, early disease flags, even water monitoring — Landsat-8 followed the shrinking of China’s Poyang Lake through recent drought years on exactly that arithmetic.
Beyond NDVI
NDVI isn’t the ultimate tool all the season long because it can’t say much about soil. Narrowband hyperspectral data, reviewed by Thenkabail and co-authors in the ISPRS Journal of Photogrammetry and Remote Sensing (2022), extends the diagnostic range considerably:
- soil type and moisture saturation
- leaf chlorophyll content and leaf area index
- crop identification by spectral signature
- early warning of pest and disease outbreaks
Then the clouds arrive, and the record simply stops. Whitcraft, Becker-Reshef, and Justice (2015, Remote Sensing of Environment) mapped clear-view probability across global cropland and found that in large agricultural regions — monsoon Asia and equatorial Africa in particular — the odds of obtaining a usable cloud-free image during peak growing months drop below 50%.
That’s not an inconvenience, that can be fatal. Crop-type classification depends on the shape of the seasonal growth curve, and the transitions that distinguish maize from soybean or rice from sugarcane occur in a matter of weeks. Lose those weeks, and the map degrades, regardless of how good the sensor’s spectral resolution is.
Synthetic Aperture Radar: Structure, Moisture, and No Excuses About Weather
Radar carries its own light. That single fact removes cloud and nightfall from the equation. The satellite sends microwave pulses down at the surface and times whatever comes back up. Because the spacecraft is moving, thousands of those returns can be combined along the flight path into a synthetic aperture much wider than the physical antenna — the trick that lets a small instrument in orbit resolve features a few metres across.
One correction worth making: backscatter isn’t a processing stage. It is the measurement. How strongly a pulse returns depends on the shape of the canopy, the amount of biomass standing in the field, and the water held inside it. Colour never enters into it.
For crop mapping, that’s the benefit. Wheat stems, a broadleaf canopy and a flooded rice paddy each throw C-band back differently, so SAR images can distinguish crops that look similar to an optical sensor. Everything practical follows from that:
- Yield and acreage — backscatter time series show planted area and biomass build-up, which is what underwriters and acreage reporting need.
- Flood damage — calm open water reflects almost nothing back, so drowned fields stand out within hours of an overpass.
- Soil moisture — dielectric response at C- and L-band puts numbers on surface wetness across entire basins.
- Claims verification — order SAR imagery from before and after a reported loss and you can see whether the field was ever planted.
For instance, RIICE, the rice monitoring program, coordinated by IRRI, used Sentinel-1 data to get insurance money to smallholders in Tamil Nadu after the 2015–16 drought, before field surveys had even wrapped up. That’s the case for radar in a sentence: it delivers evidence while optical sensors are still waiting for a clear sky.
Where This Leaves Us
Orynbaikyzy, Gessner, and Conrad reviewed dozens of combined-sensor studies for the International Journal of Remote Sensing in 2019. The same result kept coming up: adding radar to an optical time series lifts overall accuracy by a few percentage points, and the best fused results land in the low-to-mid 90s.
We have all the technologies available for this precise remote analysis: Sentinel-1 and Sentinel-2 are free, commercial SAR satellite imagery is also available at sub-meter resolution on demand, and the classification methods have been published in full. The only thing that is holding us back is subsidy agencies and crop insurers. They still build their processes around optical archives, and consider a SAR data request like a special favor. So let’s turn that around and start with radar as the backbone, bringing in spectral detail where it must be.
Editor’s Note: The opinions expressed here by the authors are their own, not those of impakter.com — Cover Photo Credit: K.
