SAR Fundamentals
How synthetic aperture radar sees through clouds, in darkness, and reveals what optical sensors cannot
Synthetic Aperture Radar (SAR) is a radar imaging system carried on aircraft or satellites that creates high-resolution images of the Earth's surface by emitting microwave pulses and measuring what bounces back. Unlike optical sensors that rely on sunlight, SAR generates its own illumination — so it works at night. Unlike optical sensors that are blocked by clouds, SAR's microwave frequencies pass through clouds, rain, and smoke. SAR does not see color or reflected light. It sees surface roughness, moisture content, and geometric structure. This makes it essential for flood mapping, deforestation monitoring, ground deformation measurement, and any application where persistent, all-weather observation is required.
Why It Matters
Approximately 67% of the Earth's land surface is cloud-covered at any given time, with tropical regions exceeding 80% average cloud cover. For optical sensors like Sentinel-2 and Landsat, clouds are opaque walls. Every cloudy pixel is a gap in coverage — a day, a week, sometimes months of missing data over persistently cloudy regions.
SAR does not have this problem. Microwave radiation at C-band (5.4 GHz, ~5.6 cm wavelength) passes through clouds as if they are not there. It passes through light rain. It passes through smoke from wildfires. It operates identically day and night because it carries its own transmitter.
For monitoring floods — events that are by definition associated with cloud cover and precipitation — SAR is not just useful. It is often the only sensor that can observe what is happening. For monitoring tropical deforestation in regions with near-permanent cloud cover, SAR provides the continuity that optical sensors structurally cannot.
But SAR does more than fill gaps in optical coverage. It measures fundamentally different properties of the surface. Where an optical sensor measures how a surface reflects sunlight (its color, essentially), SAR measures how a surface scatters microwave energy — which is determined by surface roughness, dielectric properties (related to moisture content), and geometric structure. This makes SAR sensitive to things that are invisible to optical sensors: soil moisture beneath vegetation canopies, the structural difference between intact forest and logged forest, millimeter-scale ground deformation over time.
How SAR Works
The Aperture Problem
A radar system's spatial resolution depends on the size of its antenna: larger antennas produce narrower beams and finer resolution. For a satellite orbiting at 700 km altitude, achieving 5-meter resolution with a conventional radar antenna would require an antenna several kilometers long — physically impossible to launch.
Synthetic Aperture Radar solves this by exploiting the satellite's motion. As the satellite moves along its orbit, it emits a series of radar pulses. The returns from each pulse position are recorded with their precise timing and phase information. Through signal processing, these sequential returns are combined to simulate the effect of a much larger antenna — a "synthetic aperture" that can be hundreds of meters to kilometers in effective length.
The key insight is that the same point on the ground is illuminated by many sequential pulses as the satellite passes overhead. Each pulse arrives at a slightly different angle, and the phase difference between returns encodes precise distance information. By coherently combining these returns (preserving the phase information), SAR processors create imagery with resolution determined by the synthetic aperture length rather than the physical antenna size.
What SAR Measures
SAR does not measure reflected sunlight. It measures backscatter — the portion of the transmitted microwave pulse that is scattered back toward the sensor by the surface. The intensity and characteristics of backscatter depend on:
Surface roughness relative to the radar wavelength. A surface that appears smooth at optical wavelengths (a plowed field, a gravel road) may appear rough to C-band radar if its roughness features are comparable to the 5.6 cm wavelength. Water bodies are typically very smooth at microwave wavelengths, producing specular reflection (mirror-like, away from the sensor) and appearing very dark in SAR imagery. This is why SAR is so effective for flood mapping.
Dielectric properties of the surface, primarily driven by moisture content. Water has a high dielectric constant (~80 at microwave frequencies), while dry soil is much lower (~3-5). Higher moisture content increases backscatter intensity. This makes SAR directly sensitive to soil moisture in ways that optical sensors can only infer indirectly.
Surface geometry and structure. The orientation of scattering elements relative to the radar beam affects the polarization of the return. Vertical structures (tree trunks, buildings) interact differently with horizontally and vertically polarized radar than do random canopy elements or bare soil. This is the basis of polarimetric SAR analysis.
Volume scattering within vegetation canopies or snow packs. Microwave pulses can penetrate into vegetation canopies (depth depends on wavelength — longer wavelengths penetrate deeper), scattering from branches, leaves, and trunks at multiple levels. This volume scattering carries information about canopy structure and biomass.
Imaging Geometry
SAR images are not photographs. They are constructed from the timing (range) and Doppler frequency (azimuth) of radar returns, producing an image in slant-range geometry that must be converted to ground coordinates through geocoding.
SAR's side-looking geometry (the sensor looks to the side, not straight down) introduces geometric effects absent in optical imagery:
Foreshortening — Slopes facing the radar appear compressed in the image because the radar beam reaches the top and bottom of the slope at nearly the same time. Mountains appear to lean toward the sensor.
Layover — In extreme cases, the top of a tall feature (mountain peak, building) is actually closer to the sensor than its base, causing the top to appear before the base in the image. The feature appears to fold over itself.
Shadow — Steep slopes facing away from the radar receive no illumination and produce no return, appearing as black areas in the image. Unlike optical shadows, radar shadows contain zero information — no signal reaches these areas.
These geometric effects are predictable from the sensor's orbit, viewing angle, and a digital elevation model, and can be corrected during processing. But they are fundamental to SAR and affect how imagery is interpreted.
Frequency Bands
SAR systems operate at different microwave frequencies, each with distinct interaction characteristics:
X-band (8-12 GHz, 2.5-3.75 cm): High resolution, surface scattering dominant, sensitive to small-scale roughness.
C-band (4-8 GHz, 3.75-7.5 cm): Moderate penetration, good balance of surface/volume scattering. Sentinel-1 operates here.
S-band (2-4 GHz, 7.5-15 cm): Moderate penetration, used by NovaSAR and some airborne systems.
L-band (1-2 GHz, 15-30 cm): Significant vegetation penetration, sensitive to tree trunks and large branches. ALOS-2 PALSAR-2 operates here.
P-band (0.3-1 GHz, 30-100 cm): Deep penetration through vegetation and soil. ESA's BIOMASS mission will be the first spaceborne P-band SAR.
The choice of frequency band determines what the radar is sensitive to. C-band SAR (Sentinel-1) interacts primarily with the upper canopy of forests and with the soil surface in agricultural areas. L-band SAR penetrates deeper into vegetation, reaching tree trunks and the ground beneath canopy. P-band SAR can penetrate through the entire canopy and into the soil.
For flood detection, C-band is effective over open water and areas with low vegetation. Under dense canopy, L-band or P-band may be needed to detect flooding beneath the trees.
Sentinel-1 and Operational SAR
Sentinel-1 is ESA's C-band SAR mission, operating as part of the Copernicus programme. Sentinel-1A launched in 2014, and Sentinel-1C launched in December 2024 to continue the mission after Sentinel-1B's failure in 2022.
Sentinel-1's primary imaging mode is Interferometric Wide Swath (IW), which uses a technique called Terrain Observation with Progressive Scans (TOPS) to achieve a 250 km swath width at approximately 5x20 meter resolution. IW mode produces dual-polarization data (VV+VH), where VV means vertically transmitted and vertically received, and VH means vertically transmitted and horizontally received.
The VV polarization is most sensitive to surface scattering and soil moisture. The VH polarization (cross-polarization) is more sensitive to volume scattering within vegetation canopies. The ratio or difference between VV and VH backscatter is a useful discriminator for land cover classification and vegetation structure.
Sentinel-1 data is freely available through ESA's Copernicus Data Space Ecosystem, with a 6-day (now 12-day with one satellite) repeat cycle at European latitudes. The data is delivered as Level-1 Ground Range Detected (GRD) or Single Look Complex (SLC) products. GRD products have been multi-looked to reduce speckle and projected to ground range. SLC products preserve the complex phase information needed for interferometry.
Speckle
SAR imagery contains a characteristic granular noise called speckle. Speckle is not instrument noise — it is a real physical phenomenon arising from the coherent interference of radar returns from many individual scatterers within a resolution cell.
Each resolution cell on the ground contains many scattering elements (leaves, soil particles, rocks). The radar returns from these elements combine coherently, meaning their phases add constructively or destructively depending on the exact geometry. This produces apparently random intensity variations even over a uniform surface.
Speckle follows well-understood statistical distributions (Rayleigh for amplitude, exponential for intensity in single-look imagery). It can be reduced through multi-looking (averaging adjacent pixels at the cost of spatial resolution) or through adaptive filters (Lee, Frost, Gamma-MAP, or more modern approaches like non-local means filtering). The trade-off is always between speckle reduction and spatial resolution preservation.
For quantitative analysis — change detection, thresholding for flood mapping, biomass estimation — speckle management is critical. Raw single-look SAR imagery has a coefficient of variation of 1.0 (the standard deviation equals the mean), making pixel-level analysis unreliable without filtering or multi-temporal averaging.