Earth Observation

Earth observation (EO) is the systematic collection of information about the physical, chemical, and biological systems of planet Earth using a combination of satellite sensors, airborne instruments, and ground-based measurements. While the term "remote sensing" refers specifically to the technology of measuring reflected or emitted energy from a distance, Earth observation is the broader applied discipline that encompasses the entire pipeline — from sensor design and data acquisition to processing, analysis, and the delivery of actionable information to decision-makers.

In practice, EO integrates data from optical and radar satellites, weather stations, ocean buoys, GPS networks, and airborne LiDAR into products that inform climate science, agriculture, disaster response, urban planning, defence, and environmental policy. It is the applied science of turning raw measurements of the Earth into knowledge that drives action.

Modern Earth observation operates through a layered system of complementary sensors and platforms. Satellite constellations form the backbone — optical satellites (Sentinel-2, Landsat 8/9, Planet) capture reflected sunlight across multiple spectral bands; Synthetic Aperture Radar (SAR) satellites (Sentinel-1, ICEYE, Capella) emit microwave pulses that penetrate clouds and work at night; atmospheric sensors (Sentinel-5P, OCO-2) measure trace gases and aerosol concentrations; and altimetry missions (Sentinel-6, Jason) track sea-level changes with millimetre precision.

Ground segment infrastructure handles the download, archiving, and processing of raw satellite data. Missions like Copernicus generate petabytes of data annually, distributed through platforms like the Copernicus Data Space Ecosystem, NASA Earthdata, and commercial cloud services. Analysis-ready data (ARD) products are then derived — atmospherically corrected surface reflectance, cloud-masked composites, and thematic maps. These feed into downstream applications: crop yield forecasts, deforestation alerts, air quality indices, flood extent maps, and climate indicators. Increasingly, machine learning models are applied directly to EO data to automate classification and change detection at global scale.

Despite its significant capabilities, Earth observation faces significant challenges. Optical sensors are blocked by persistent cloud cover, which affects up to 70% of the tropics on any given day. The sheer volume of satellite data — petabytes per year — creates storage, processing, and bandwidth bottlenecks that demand cloud computing infrastructure. Temporal gaps persist: even Sentinel-2's 5-day revisit can miss rapidly evolving events like flash floods. Spatial resolution remains a trade-off with coverage and revisit frequency — no single sensor optimises all three. Data continuity is also a concern: satellite missions have finite lifespans (typically 7–12 years), and gaps between successor missions can break long-term records critical for climate science.

Earth observation traces its origins to the space race. NASA launched TIROS-1 in 1960, the first weather satellite. The key moment came on 23 July 1972, with the launch of Landsat 1, which carried the first multispectral scanner designed for land surface monitoring. Through the 1980s and 1990s, France (SPOT), India (IRS), and Japan (JERS) launched their own EO programmes. The 2008 decision by USGS to make the entire Landsat archive freely available was a watershed moment, catalysing an explosion of research and commercial applications. ESA followed suit with Copernicus Sentinel data. Today, the field is being reshaped by commercial constellations (Planet, Maxar, ICEYE) offering daily revisits and sub-metre resolution, and by cloud platforms (Google Earth Engine, Microsoft Planetary Computer) that bring planetary-scale analysis to any researcher with a web browser.