What multispectral and radar imagery can – and can’t – tell us about a field
A satellite orbiting 500 kilometres overhead can now tell you more about the health of a wheat field than a person standing in it – not because it sees better, but because it sees differently, and it sees every field in a state on the same day. Satellite remote sensing has moved from a research curiosity to the backbone of how Indian agriculture is monitored at scale.
Before satellite monitoring, understanding crop condition across an entire state meant physically visiting a sample of fields and extrapolating. That approach is inherently limited by how many fields a finite number of field staff can visit in a season. Remote sensing removes that ceiling – the same satellite pass that observes one field observes millions of others simultaneously, at a fraction of the marginal cost per field. What used to require months of field survey work to get a rough state-wide picture can now be refreshed every few days, which changes not just the cost of monitoring but what’s actually possible to monitor – early-season stress, for instance, that would otherwise go unnoticed until it was too late to act on.
Not all satellite data is the same, and the distinction between sensor types explains most of the design choices in Indian agricultural monitoring programmes:
| Sensor type | What it captures | Key limitation |
|---|---|---|
|
Optical / multispectral (e.g. Resourcesat, Sentinel-2) |
Reflectance across visible and infrared bands – vegetation indices like NDVI, crop vigour, chlorophyll
|
Blocked by cloud cover, a major constraint during India’s monsoon (Kharif) season |
|
Radar / SAR (e.g. RISAT, Sentinel-1) |
Backscatter sensitive to vegetation structure and soil moisture, day or night, through cloud |
Interpretation is less intuitive than optical imagery; needs more specialised processing |
This is why most operational crop-monitoring systems in India combine both: optical imagery when skies are clear, radar imagery to keep monitoring continuous through the monsoon when optical data simply isn’t available.
ISRO’s Space Applications Centre has run agricultural remote sensing programmes for decades, including the long-running FASAL initiative (Forecasting Agricultural output using Space, Agro-meteorology and Land-based observations), which produces multiple in-season forecasts for major crops and is operated in coordination with the Mahalanobis National Crop Forecast Centre.
This isn’t purely a domestic effort, either. Indian crop monitoring increasingly draws on international collaboration and freely available global missions – part of a broader trend, reflected in initiatives like GEOGLAM (Group on Earth Observations Global Agricultural Monitoring), toward pooling satellite coverage across countries rather than each government building and operating an entirely separate constellation.
A raw satellite image isn’t directly useful – it has to be corrected for atmospheric interference and cloud contamination, converted into indices like NDVI (a measure of vegetation greenness) or soil moisture estimates, and only then applied to a specific question: is this crop stressed, is this district facing drought, what will this field yield.
WHY COMBINING RADAR AND OPTICAL OUTPERFORMS EITHER ALONEResearch on monsoon cropland mapping in India found that a radar-only approach achieved roughly 90% classification accuracy, while combining radar with optical vegetation indices pushed that to roughly 93% – a meaningful improvement particularly in regions where cropland is interspersed with forest or plantation cover that’s otherwise easy to misclassify. |
Satellite imagery isn’t always the right tool for a given question – it’s one point on a resolution-versus-cost-versus-frequency tradeoff that also includes drones and ground-based sensors.
| Data source | Typical resolution | Best suited for |
|---|---|---|
|
Satellite (optical/SAR) |
5–30m (medium-resolution missions) |
State or district-scale monitoring, trend tracking across a full season |
|
Drone imagery |
Centimetre-level |
Field-level scouting, verifying a specific stressed area flagged by satellite |
|
Ground sensors (soil, weather) |
Point measurement |
Precise, continuous local readings a satellite pass simply can’t capture |
In practice, an operational state monitoring system rarely relies on just one of these. Satellite imagery typically does the broad, continuous coverage; ground sensors and occasional drone flights fill in the detail where the satellite signal alone leaves a genuine question unanswered – confirming whether a flagged patch is pest damage, waterlogging, or simply a sensor artefact.
Remote sensing only creates value once its outputs reach the people making decisions – a state agriculture officer prioritising field visits, an insurer processing a claim, an advisory service pushing a irrigation alert to a farmer’s phone. This is the same closing-the-loop challenge covered in our piece on precision agriculture, just operating at state scale rather than field scale: the hardest part usually isn’t capturing the imagery, it’s building the institutional pipeline that turns a vegetation index into a specific action someone actually takes.
Even at 10-metre resolution, a single satellite pixel can span multiple smallholder plots in fragmented landholding regions, blending signals from different farmers’ fields into one averaged reading. Higher-resolution commercial imagery or drone data can help, but at a materially higher cost.
Optical satellites are only useful when skies are clear, and critical growth stages sometimes fall during the cloudiest weeks of the monsoon. This is precisely why radar has become a required part of the toolkit, rather than an optional enhancement, for continuous in-season monitoring.
Every operational satellite-based crop model – FASAL, YES-TECH, and others – is validated against ground-collected data, not run on satellite imagery alone. Remote sensing extends the reach of ground observation; it doesn’t yet fully replace it, and the programmes that treat it that way – as an extension of field capacity rather than a replacement for it – tend to produce more defensible results.
NDVI (Normalized Difference Vegetation Index) is a calculation derived from how much red and near-infrared light a surface reflects – healthy, dense vegetation reflects near-infrared strongly and absorbs red light, producing a high NDVI value. It’s one of the most widely used vegetation indices precisely because it’s simple to compute and correlates well with crop vigour.
Both models exist. ISRO’s own satellite data is typically made available to Indian government agencies at low or no direct cost for approved applications, while missions like Sentinel-1/2 are freely available globally under the Copernicus programme. Higher-resolution commercial imagery, where used, is licensed separately.
Revisit frequency depends on the satellite – some optical missions revisit the same location every few days, while combining multiple satellites (as most operational systems do) improves the effective revisit rate further. Radar satellites like Sentinel-1 are specifically valued for consistent revisit regardless of weather.
Indirectly. Vegetation stress from pest or disease pressure often shows up as a drop in NDVI or biomass estimates before it’s visually obvious on the ground, which can flag an area for field verification – but satellite imagery alone typically can’t distinguish the specific cause of stress without ground follow-up.
States don’t need to own satellites. Nearly all operational agricultural monitoring in India draws on data from ISRO’s existing fleet and freely available missions like Sentinel-1/2, accessed through processing platforms and analytics layers built on top – the investment is in the analysis and delivery pipeline, not in launching new satellites, which keeps the barrier to entry for a state agriculture department considerably lower than it might first appear.
Remote sensing that’s built into a working state platform fieldWISE combines optical and radar satellite data into the same crop monitoring and advisory pipeline used across Andhra Pradesh and Kerala – not a standalone imagery feed, but a working decision layer. → See how fieldWISE uses satellite data |
ISRO Space Applications Centre – Agriculture applications of satellite remote sensing: https://www.sac.gov.in/Vyom/Agriculture?lang=en
Related on this blog: Precision Agriculture Explained · YES-TECH, Explained · What Is the Digital Crop Survey (DCS)?
