Keeping Landsat Honest: How the USGS Calibrates Fifty Years of Earth Images

For more than 50 years Landsat satellites have photographed the Earth's surface, and the record is only useful if an image from 1985 can be compared with one from 2025. That depends on calibration: each pixel must land on the right spot on the ground (geometry) and report the right amount of light (radiometry). At the USGS Earth Resources Observation and Science (EROS) Center, the Calibration and Validation Center of Excellence (ECCOE) does this for every active Landsat mission — well enough that other government and commercial satellites around the world use Landsat as their reference, and the team works with U.S. and international agencies and companies to make their data agree.

The Landsat missions over five decades. USGS

Calibration also looks backwards: improvements made today are applied to past missions, so researchers can follow land change consistently over decades. The team keeps looking for better references, including calibrating against the Moon and validating on the ground with drones.

Putting each pixel in the right place

Geometric work runs in both life stages of a mission:

Correction combines dynamic measurements from telemetry (the spacecraft's position, velocity and attitude) with static ones (the focal-plane layout, the instruments' alignment), and uses a digital elevation model to remove the parallax of viewing hilly ground at an angle. On the Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS), the detectors are arranged on several Sensor Chip Assemblies (SCAs), and the raw image shows it: each chip's strip slightly offset, and each band slightly apart from the next.

A raw OLI image: the chips misaligned within a band, and bands 6, 5 and 4 apart within a chip. USGS

Corrected and registered to the Landsat ground-control library, the result is a terrain-corrected product in which every band, and the OLI and TIRS images, line up — including a seasonal drift in TIRS-to-OLI alignment on Landsats 8 and 9 that is measured, predicted and corrected. On orbit, the team verifies band registration, OLI-to-TIRS alignment, geometric and geodetic accuracy, and how well Landsat 9 lines up with Landsat 8.

Reading the right amount of light

The number a pixel records depends not only on the ground but on the sensor's quirks, the angles of sun and view, and the atmosphere. So the team states each measurement with its uncertainty — a range that should contain the true value — and aims to keep it as small as possible. Landsat 8's OLI is traceable to the standards of the National Institute of Standards and Technology (NIST); along the way, spectral response, nonlinearity, nonuniformity, stray light and the stability of the calibration sources all add uncertainty.

Before launch, the OLI's prelaunch radiance scale carried uncertainties of about 3–4.5% in radiance and 1.7–2.8% in reflectance, depending on band and brightness — the cirrus band the least certain.

Every day, for the life of the mission

The team trends the results of the radiometric and geometric checks on all Landsat 8 and 9 data, daily, and decides when to update the files that turn raw counts into images — the Calibration Parameter Files, Bias Parameter Files and response linearization tables — balancing a better product against consistency over time. On-orbit checks include the OLI's signal-to-noise ratio (measured monthly with onboard calibrators), TIRS noise, radiometric stability, detector-to-detector relative gains, and the cross-calibration of Landsat 9 with Landsat 8.

A TIRS band 10 image before and after reprocessing with updated parameters: the striping fades. USGS

The team publishes what it finds each quarter in the ECCOE Landsat Quarterly Calibration and Validation Report (for example, Haque and others, 2024).

Sources

Written from Landsat Geometric and Radiometric Calibration and Characterization, U.S. Geological Survey Fact Sheet 2024–3039; a work of the United States government in the public domain. It cites Haque and others (2024), ECCOE Landsat Quarterly Calibration and Validation Report—Quarter 1, 2024 (doi:10.3133/ofr20241058), and the USGS page on Landsat radiometric uncertainty.