Running a city, a building site or a logistics chain means answering the same questions every day: what changed, where and when? Satellites answer those questions from orbit, reliably, repeatedly and without driving a single kilometre. Modern commercial constellations now deliver clear optical images down to 30 cm detail, radar views that see through cloud, and temperature maps that expose hidden heat. Archive scenes start at about € 2 km²; a freshly tasked, cloud-free shot in the sharpest resolution costs nearer € 15 km². EO has therefore shifted from scientific novelty to routine data subscription.

What is Satellite Earth Observation?

Picture a camera orbiting roughly 600 km above Earth in a near-polar path that crosses the Equator at the same local time on every pass. That sun-sync timing keeps lighting consistent, vital for change detection. Agile attitude motors let the craft swivel sideways, so a city might be revisited twice a day instead of once a week. Light (or microwave energy, for radar) strikes the sensor, turns into digital numbers and is stored onboard. When the satellite flies over a ground station, it beams the data down; cloud software then corrects the geometry, so each pixel lines up with latitude and longitude. Only after that “clean-up” step do analysts or algorithms see the file.

How does a satellite take a photo?

A single overhead snapshot (mono) is perfect for mapping “what is here now?”, but some questions need height or richer change cues. Operators, therefore, schedule satellites to collect images from multiple angles during one orbit or over successive passes:

Stereo pairs two images – one taken looking forward, one backward – allowing software can triangulate building heights or terrain models at metre-scale accuracy.

Tri-stereo adds a third look (often off to the side), improving elevation accuracy and reducing occluded “blind spots” behind tall structures.

Choosing mono, stereo, or tri-stereo – and whether the extra looks come along-track or across-track – depends on the end goal: flat maps need only one clear shot, while 3-D city models, landslide risk maps or volumetric stockpile estimates require multi-angle acquisition.

What types of photos can EO satellites take?

Satellites do not all carry the same “camera”. Four sensor families dominate, each giving a different slice of reality.

Think of optical as eyes, radar as touch, thermal as a thermometer and spectral as a chemical sniffer. Choosing the right sensor means matching the business question to the slice of observation that reveals it.

Five ways EO satellites describe the world

“Resolution” is only part of the picture. Five attributes govern what you ultimately see and pay for.

RULE OF THUMB: Buy the lowest resolution and least-frequent revisit that still answers the business question.

What can actually be identified?

Once the right data is in hand, both analysts and machine-learning models can outline every roof, road and parking bay from 30 cm optical imagery.

• Detect ground movement of a few centimetres by stacking SAR scenes (InSAR).
• Map floodwater within hours of rain using cloud-penetrating radar.
• Reveal insulation leaks in factory roofs through night-time thermal passes.
• Flag nutrient stress in parkland or crops via subtle spectral colour shifts.

Each result is a GIS-ready layer (points, polygons, alerts) that plugs straight into dashboards or digital twins.

From raw pixels to decision-ready layers

EO data travels through four shorthand levels, each adding value:

1. L0 – Raw packets: engineering only.
2. L1 – Calibrated & geo-referenced: pixels now sit in the right place on Earth.
3. L2 – Analysis Ready: terrain-corrected, cloud-masked, tiled.
4. L3 – Analytics: finished outputs such as “new-build polygons” or “ship counts”.

Most newcomers start consuming at L2 (run their own models) or L3 (subscribe to turnkey insights). Either way, cloud catalogues expose the data via simple REST or STAC APIs, so your code fetches just the tiles it needs.

Six ways to buy insights

After pixels are converted into layers, you still need to decide how to obtain them. Satellite-data vendors package their products in a number formats; knowing the differences saves both budget and integration effort.

A useful mental rule: the farther along you move in this table, the less raw data you handle and the faster you reach an actionable answer – but the less flexibility you have to tweak methods later.

INSIGHT-AS-A-SERVICE

When EO arrives as ready-made answers

A growing slice of the EO market no longer sells pictures at all. Instead, providers run the entire chain – from satellite tasking to machine-learning analytics, and deliver only the decision trigger: a polygon, a score, or a yes/no flag that slips directly into corporate software. To understand how this works, think of three building blocks layered on top of the raw imagery stream:

EVENT DETECTION ENGINES
Algorithms search every new scene for a predefined pattern: standing water, fresh earthworks, smoke plumes, parked aircraft. When the signature appears, the service posts a time-stamped GeoJSON payload to a webhook or message queue. The client system consumes the alert exactly as it would ingest a sensor reading – there is no imagery to store or interpret.

CHANGE-MONITORING SERIES
Some questions rely on trends rather than one-off events. Here, the provider assembles a time-ordered stack of scenes, registers them to sub-pixel precision, and publishes a rolling raster that expresses movement or accumulation: millimetre ground subsidence, vegetation stress trajectories, port-yard congestion indices. Subscribers pull the fresh layer nightly into BI dashboards or digital-twin models.

RISK OR PERFORMANCE SCORING
The highest level abstracts even further, combining EO-derived layers with third-party and customer data to output a single metric per asset: flood-exposure probability, construction-progress score, roof-condition index, supply chain dwell-time KPI. Because the score travels as a lightweight JSON record, insurers, asset managers or ERP modules can drop it straight into pricing or workflow rules.Across these blocks the common thread is automation: data travels through an API, not an analyst, and lands inside systems that already make operational decisions. Technical integration usually follows one of two patterns – REST endpoints for pull workflows, or webhooks for push – and therefore fits comfortably alongside weather feeds, IoT telemetry and financial market ticks.

Image Credits: