What is Earth observation?

How it works

The European Space Agency's Newcomers Earth Observation Guide defines Earth observation (EO) as acquiring observations of the planet's surface and atmosphere with remote sensing instruments, usually as digital imagery. NASA Earthdata describes remote sensing as acquiring information from a distance. Instruments on satellites and aircraft detect and record reflected or emitted energy, it says.

The ESA guide describes two groups of sensors. Passive instruments detect existing emissions, such as reflected sunlight, so they usually depend on the day-night cycle and can be blocked by cloud. Active instruments transmit their own signal and measure what comes back. The guide calls synthetic aperture radar (SAR) the most common active sensor in EO. It says radar imaging is unaffected by the day-night cycle and, most of the time, by weather. NASA lists four types of resolution for any dataset: radiometric, spatial, spectral and temporal.

Who runs it

The Landsat missions are part of the U.S. Geological Survey's (USGS) National Land Imaging Program. The first satellite launched on July 23, 1972. USGS describes the series as the world's longest space-based record of Earth's land surface. As of October 2026, USGS says Landsat 10 is expected to launch in 2031. In Europe, Copernicus is the Earth observation component of the European Union's space program. The European Commission says it draws on dedicated Sentinel satellites and contributing missions, plus in situ (non-space) sensors. It says its Copernicus Data Space Ecosystem gives free access to Sentinel data.

Where things stand in 2026

Space agencies are applying artificial intelligence (AI) to this data in orbit and on the ground. In July 2025, NASA's Jet Propulsion Laboratory reported flight-testing software that lets a satellite avoid cloudy scenes without human involvement. It said clouds can block optical sensors as much as two-thirds of the time. On June 4, 2026, ESA reported that Tessera had been made widely available to researchers. It said University of Cambridge researchers and partners developed the foundation model, first launched in 2025. NASA Earthdata describes foundation models as AI models trained on broad sets of unlabeled data that can be used for different tasks. ESA says Tessera was trained on Sentinel-1 radar and Sentinel-2 optical data. It says Tessera produces compressed yearly representations of the surface, called embeddings, at 10 m resolution for each year from 2017 to 2025. ESA lists uses from crop monitoring to measuring burned areas.