SAR Basics

Much of the imagery discussed elsewhere on this site is optical: it records sunlight that the surface reflects, split into spectral bands. Synthetic aperture radar, or SAR, works on an entirely different principle. It is an active sensor, carrying its own source of illumination. The instrument transmits pulses of microwave energy toward the ground and measures the fraction that scatters back to the antenna. Because it supplies its own signal, SAR needs no sunlight and can image at night; because microwaves pass through clouds and smoke, and most rain (only intense rain troubles the shortest radar wavelengths), it can see the surface when optical sensors are blocked. That combination — day or night, cloud or clear — is what makes radar a distinct and complementary way of observing the Earth, and it comes at the price of a data type that must be read quite differently from a photograph.

Backscatter, not brightness

The quantity a SAR image records is backscatter: how strongly a patch of ground returns the transmitted microwave pulse toward the sensor. A bright pixel means a strong return, a dark pixel a weak one, but this brightness has nothing to do with color or sunlight. It is governed by the geometry and material of the surface — its roughness at the scale of the radar wavelength, its structure, and its electrical properties. Reading a radar scene therefore means thinking about how a surface scatters an incoming wave, not how it looks to the eye. A calm lake is almost black in radar, just as it is dark in most optical bands — but for a different reason: its smooth surface acts as a mirror and reflects the pulse away from the sensor, so almost none returns. A rough field or a forest canopy scatters energy in all directions and sends a measurable fraction back, so it appears bright.

Backscatter is normally reported as a calibrated coefficient — sigma nought (σ⁰), or gamma nought (γ⁰) when normalized for the viewing angle — and on a logarithmic decibel (dB) scale, so most values are negative and a change of 3 dB is roughly a doubling or halving of the returned power.

Wavelength, polarization, and viewing geometry

Three properties of the radar shape what it sees. The first is wavelength. Radar bands are much longer than optical ones — centimeters rather than micrometers — and the wavelength sets the scale at which the surface counts as “rough” and how far the signal penetrates vegetation or dry soil. Longer wavelengths pass through foliage to interact with branches, trunks, and the ground beneath, while shorter ones interact mainly with the top of the canopy, so the same scene can look different at different bands. The bands most often met are X band (wavelengths around 3 centimeters), C band (around 5 to 6), and L band (around 23).

The second is polarization, the orientation of the transmitted and received wave. A radar can send and receive in horizontal (H) or vertical (V) polarization, and the combination it uses is written as a pair of letters — HH for transmit and receive horizontal, HV for transmit horizontal and receive vertical, and likewise VV and VH. Each combination is sensitive to different scattering behavior. Volume scattering from vegetation tends to depolarize the signal, so cross-polarized channels highlight structure that co-polarized channels miss. Comparing polarizations is one of the main ways structure and surface type are distinguished in radar.

The third is viewing geometry. Unlike a nadir-looking optical camera, a SAR looks to the side at an incidence angle, and that angle strongly affects the return: the same surface gives a different backscatter when viewed steeply than when viewed at a shallow angle. The side-looking geometry is the source of the geometric distortions described below, so the incidence angle is always part of interpreting a scene.

Radar builds an image from timing, not from a lens, and that is why it must look to the side. Across the track (range), position comes from how long each echo takes to return, which only works if points at different distances return at different times. Along the track (azimuth), a small antenna would blur detail over a wide beam; instead the platform’s motion is used to combine many pulses received along its path, as if from one very long antenna. That synthesized aperture is what gives SAR fine resolution from a modest antenna, and it relies on the radar recording the phase of each echo — the same coherence that causes speckle.

What the surface does to the signal

Three surface characteristics dominate the response. Roughness relative to the wavelength controls how much energy scatters back: surfaces smooth at the radar scale act as mirrors and appear dark, while rough surfaces scatter diffusely and appear brighter. Structure matters because three-dimensional arrangements can redirect the signal strongly — a building wall meeting the ground, or the trunks of a flooded forest, can form a corner that reflects the pulse straight back — a double bounce — and produces very bright returns, which is why urban areas and some wetlands stand out dramatically in radar.

Moisture is the third factor. Backscatter depends on the surface’s dielectric properties, which change sharply with water content, so wetter soil and vegetation generally return more energy than dry — until water pools into a smooth surface, which turns dark. Optical shortwave-infrared bands also respond to moisture, but only at the very surface and only when the sky is clear; radar’s sensitivity reaches a few centimeters into bare soil and works through cloud, which is why it is the workhorse for soil-moisture retrieval. (Flooding beneath a canopy is detected mainly through the double-bounce effect above.)

Speckle and geometric distortion

Two artifacts must be understood before a radar image can be read correctly. The first is speckle, the grainy salt-and-pepper texture that gives raw SAR its characteristic look. Speckle arises because each resolution cell contains many tiny scatterers whose returns add together with random phase, so even a uniform surface produces a fluctuating brightness. It is inherent to coherent radar, not sensor noise, and it is usually reduced by averaging — spatially, or across multiple acquisitions in a time series — which trades some spatial or temporal detail for a cleaner estimate of backscatter.

The second is geometric distortion from the side-looking geometry over terrain. Because the radar measures distance to the sensor, slopes facing the radar are compressed, an effect called foreshortening, and in steep terrain the top of a feature can return before its base so that the slope appears to fall toward the sensor — layover. On the far side, slopes angled away can be hidden entirely in radar shadow, where no signal is received. These effects depend on the incidence angle and the terrain, and correcting them requires an elevation model and careful use of the scene’s coordinate system; without such terrain correction, radar geometry does not line up with a map or with optical imagery.

SAR alongside optical imagery

SAR and optical imagery are complementary because they measure different things. Optical sensors report reflected sunlight and color, which the eye and most vegetation indices are built around, but they are blind through cloud and at night. Radar reports structure, roughness, and moisture, sees through cloud and darkness, but produces scenes that need more interpretation and carry speckle and geometric quirks. The two are strongest together: radar can fill the gaps in an optical time series whenever cloud cover blocks the optical view, and it adds information — surface moisture, vertical structure, reliable all-weather coverage — that reflected light alone cannot supply. Treating radar not as a harder substitute for optical imagery but as a different measurement of the same ground is what makes it valuable, and understanding backscatter, geometry, and speckle is what makes a radar scene readable rather than merely grainy.

Sources

• Synthetic Aperture Radar (SAR) — NASA Earthdata