Radar Doesn't Care About Clouds: How SAR Keeps Field Monitoring Running All Year

Radar Doesn't Care About Clouds: How SAR Keeps Field Monitoring Running All Year


Every optical satellite has the same weakness: if there is a cloud between it and your field, it sees the cloud. In northern Europe that can mean weeks without a usable image right when crops are changing fastest. Synthetic Aperture Radar (SAR) solves this in a fundamentally different way — it doesn't wait for sunlight or clear skies, it brings its own signal. This post explains how SAR works, what it can and can't tell you about a field, and where you can get the data, much of it completely free.

The Cloud Problem in One Paragraph

Optical sensors like Sentinel-2 measure reflected sunlight in visible and infrared bands. That's what makes NDVI and other vegetation indices possible. But sunlight in those wavelengths is scattered and blocked by cloud droplets, and at night there is no sunlight at all. In our earlier post on seeing through the clouds we covered compositing and gap-filling, which help, but they only stretch the clear images you already have. If May and June are overcast, no amount of interpolation tells you what happened to the crop in between. To truly fill the gap you need a sensor that isn't stopped by cloud in the first place.

What SAR Actually Is

A radar satellite is an active sensor. Instead of passively recording sunlight, it fires microwave pulses at the ground, slightly to one side of its flight path, and records the echo that bounces back. Microwaves have wavelengths of centimetres rather than the fractions of a micrometre used by optical sensors, so they pass straight through clouds, haze, smoke and darkness almost unaffected.

Optical vs radar on a cloudy day
OPTICAL (e.g. Sentinel-2) RADAR / SAR (e.g. Sentinel-1) Signal blocked ✗ Field hidden: no usable image Microwave pulse Echo (backscatter) Field measured ✓ day or night

Optical sensors rely on sunlight that clouds block. SAR sends its own microwave pulse and measures the echo, so cloud cover and darkness barely matter.

Why "synthetic aperture"?

The sharpness of a radar image depends on the size of its antenna: the bigger the antenna, the finer the detail. To get field-scale detail from orbit at microwave wavelengths you would need an antenna kilometres long. SAR fakes it. As the satellite moves along its orbit it records echoes from the same patch of ground many times, then combines them in processing as if they had come from one giant antenna stretched along the flight path. That "synthesised" antenna is what turns a blurry radar footprint into an image with pixels of around 10 metres (Sentinel-1) or even below one metre (commercial constellations).

What the satellite actually measures

The core quantity in a SAR image is backscatter: how much of the pulse comes back towards the satellite. It depends on two things about the surface:

  • Structure and roughness. Smooth surfaces like calm water act like a mirror and bounce the pulse away, so they look dark. Rough surfaces, buildings and tall, dense vegetation scatter energy in every direction, and some of it comes back, so they look bright.
  • Moisture. Water changes how strongly a material reflects microwaves (its dielectric properties). Wetter soil and wetter plants generally return more signal than dry ones.

That is why SAR is often described as seeing the structure and water of a field, while optical sees its colour and greenness. They answer different questions, which is exactly why they work so well together.

Polarisation: VV and VH

Radar pulses are polarised, and the satellite can listen for echoes in the same or a different orientation. Sentinel-1 over land typically records two channels:

VV (sent vertical, received vertical)

  • Dominated by surface scattering
  • Strongly influenced by soil moisture and roughness
  • Useful for tillage, flooding, bare-soil conditions

VH (sent vertical, received horizontal)

  • Produced mostly by "volume" scattering inside a canopy
  • Grows as crops get taller and denser
  • The channel most linked to crop biomass

Because VH responds to vegetation and VV responds more to the ground, the ratio between them tracks crop development surprisingly well and partly cancels out effects that hit both channels equally.

Wavelength bands: C, L and X

📡

X-band (~3 cm)

  • Very fine detail possible
  • Interacts with the top of the canopy
  • Mostly commercial satellites
e.g. ICEYE, Capella, Umbra, TerraSAR-X
🛰

C-band (~5.5 cm)

  • The agricultural workhorse
  • Sensitive to leaves, stems and topsoil
  • Free, systematic global coverage
e.g. Sentinel-1, RADARSAT
🌳

L-band (~24 cm)

  • Penetrates deeper into canopy and soil
  • Better for forests and dense crops
  • Less saturated at high biomass
e.g. NISAR, ALOS-2/4

Rule of thumb: the longer the wavelength, the deeper the signal reaches into vegetation before bouncing back.

What SAR Can Tell You About a Field

SAR does not give you NDVI, and it's important to be honest about that. What it gives you is a dense, regular time series of structural and moisture signals that you can read in several useful ways.

Use 1 — Crop growth

Tracking the season under cloud

As a crop emerges, grows and closes its canopy, VH backscatter and the VH/VV ratio rise; at ripening and harvest they change sharply again. Plotted over a season, these curves follow crop development even through weeks when optical sensors see nothing.

Use 2 — Field events

Spotting sowing, mowing, tillage and harvest

Operations that change a field's structure overnight show up as sudden jumps in backscatter. Grass cuts, ploughing and harvest are among the most reliably detected events, which makes SAR great for "has anything happened in this field?" alerts.

Use 3 — Water

Waterlogging, flooding and surface moisture

Standing water is very dark in SAR, so flooded or waterlogged parts of a field stand out clearly. Over bare or sparsely covered soil, backscatter also tracks changes in topsoil moisture, though turning that into an absolute moisture value requires careful modelling.

Use 4 — Fusion

Filling optical gaps

The most powerful use is in combination: training a model on dates when both optical and radar are available, then using radar to estimate the optical signal on cloudy dates. This is an active research area and the natural next step after compositing.

A simple radar vegetation index

One common way to condense dual-polarisation Sentinel-1 data into a single vegetation signal is the dual-pol Radar Vegetation Index:

RVI = 4 × VH / (VV + VH)

In plain language: the larger the share of the echo that comes back cross-polarised, the more vegetation volume there is. Compute it on linear backscatter values, not on decibels, and treat it as a relative trend for one field over time rather than an absolute number to compare across regions.

The Honest Limitations

SAR is powerful, but it is not a drop-in replacement for optical imagery. Anyone relying on it should know these traps:

Challenge What happens How to handle it
Speckle Radar images have a grainy "salt and pepper" texture caused by interference between echoes. Average over the whole field or a zone, use multi-temporal filtering, never trust a single pixel.
Viewing geometry Backscatter changes with incidence angle and with whether the satellite was ascending or descending. Compare images from the same relative orbit, or use terrain-corrected (RTC) products.
Rain and dew A wet surface brightens the image even if the crop hasn't changed. Cross-check with weather data before calling an anomaly.
Frozen ground and snow Frozen soil behaves very differently from thawed soil, a real issue for Nordic winters. Flag or exclude frozen dates using temperature data.
Interpretation There is no "greenness" in radar; signals mix structure, moisture and geometry. Use crop-specific baselines, and fuse with optical whenever a clear image exists.
Our take: treat SAR as the always-on heartbeat of a field and optical as the detailed health check. Radar tells you that something changed and roughly what kind of change it was; the next clear optical image tells you how the crop is actually doing.

Where to Get SAR Data

The good news for anyone starting out: the most useful SAR data for agriculture is free and openly licensed. Here's the current landscape.

The free backbone: Sentinel-1

Sentinel-1 is the EU Copernicus radar mission and the default choice for field monitoring. After Sentinel-1D was launched in November 2025 and became fully operational in May 2026, the constellation moved to the two newest satellites, Sentinel-1C and Sentinel-1D, when Sentinel-1A ended its observation role at the end of June 2026. With two satellites, the nominal revisit returns to six days, and it is often better over Europe once you combine ascending and descending passes. All data is free, including for commercial use, under the Copernicus licence.

The newcomer: NISAR (L-band)

NISAR, the joint NASA–ISRO mission launched in July 2025, carries L-band and S-band radars. Its provisional L-band science products were opened to the public on 20 July 2026, covering acquisitions from 17 June 2026 onward, and all L-band products are distributed free through NASA's Alaska Satellite Facility. For dense crops and forests where C-band saturates, it is a very welcome addition. Treat early products as still maturing: some known issues remain, and the reprocessing of older acquisitions is expected to run until the end of 2026.

Free platforms and access points

Platform SAR data offered Best for Access
Copernicus Data Space EcosystemEU / ESA Sentinel-1
  • GRD and SLC scenes
  • Browser, APIs, openEO
  • Sentinel Hub processing APIs
The official source; on-the-fly processing per field via API Free with monthly processing quotas; paid tiers for heavy use
ASF Vertex & HyP3NASA Alaska Satellite Facility Sentinel-1 NISAR
  • Full Sentinel-1 archive
  • NISAR L-band
  • Historic ALOS PALSAR, RADARSAT-1
  • On-demand RTC and InSAR (HyP3)
Archive search, NISAR, and processing without your own SAR software Free with an Earthdata login
Google Earth EngineGoogle Sentinel-1
  • Pre-processed S1 GRD backscatter
  • OPERA RTC-S1 (30 m, terrain-corrected)
  • JAXA PALSAR-2 mosaics
Fast prototyping and large time-series analysis Free for research and non-commercial; paid licence for commercial use
Microsoft Planetary ComputerMicrosoft Sentinel-1
  • Sentinel-1 GRD
  • Sentinel-1 RTC
  • STAC API, cloud-optimised files
Python pipelines that stream data straight into analysis Free data catalogue
NASA OPERA productsNASA JPL via ASF Sentinel-1 derived
  • RTC backscatter
  • Surface displacement
  • Surface water
Analysis-ready data, no SAR processing needed Free, mainly North America plus near-global RTC
JAXA ALOS PALSAR mosaicsJAXA (Japan) L-band
  • Annual global mosaics
  • Forest / non-forest maps
Long-term land-cover context, not field-level timing Free (full-resolution scenes are commercial)

Commercial SAR

Commercial constellations offer far finer detail (often under a metre) and the ability to task a satellite over a specific location, sometimes several times a day. That's invaluable for insurance claims, disaster response or high-value crops, but generally priced per scene or per square kilometre, which is hard to justify for routine monitoring of ordinary arable fields.

Provider Band Strength Free sample data?
ICEYEFinland X-band Largest SAR constellation; fast revisit; strong in flood and insurance Yes — open dataset on AWS, CC BY 4.0
Capella SpaceUSA X-band Very high resolution, low-latency tasking Yes — open data on AWS, plus research grant tiers
UmbraUSA X-band Sub-metre spotlight imagery Yes — open data programme, no sign-up
Airbus (TerraSAR-X, PAZ)Germany / Spain X-band Long, well-calibrated archive Commercial
MDA (RADARSAT-2)Canada C-band Multiple polarisations, long agricultural track record Commercial
Marketplaces (UP42, SkyFi and others)Resellers Mixed One account to buy scenes from several operators Pay per order
Tip: the commercial open-data programmes are great for learning what very-high-resolution SAR looks like, but they cover selected sites, not your fields. For systematic monitoring, Sentinel-1 is still the only free option with regular, wall-to-wall coverage.

Getting Started: A Practical Path

Step 1

Look before you code

Open the Copernicus Browser, pick a field you know, and switch between Sentinel-2 and Sentinel-1 views across a cloudy month. Seeing the radar signal change with field operations builds intuition fast.

Step 2

Use analysis-ready data

Start with terrain-corrected backscatter (RTC) from Planetary Computer, Earth Engine or ASF HyP3 rather than raw scenes. You skip the hardest preprocessing and avoid the most common mistakes.

Step 3

Build field-level time series

Average VV, VH and the VH/VV ratio inside each field polygon, keep one relative orbit per series, and plot them next to your NDVI curve. The patterns become clear within a single season.

Step 4

Fuse and alert

Once you have both series, use radar to flag sudden changes and to bridge cloudy gaps in the optical record, and confirm with the next clear optical image.

The Bottom Line

Clouds are the biggest single reason satellite crop monitoring goes quiet at the worst possible moment. SAR removes that blind spot: it works through cloud, day and night, and thanks to Sentinel-1 and now NISAR, the core data costs nothing. It won't replace optical vegetation indices, and it takes care to interpret, but combined with optical imagery it turns patchy snapshots into a continuous record of what is happening in every field.

Back to blog