White Paper: State of AI for Earth Observation
Maral Bayaraa
![](https://sa.catapult.org.uk/app/themes/catapult/assets/img/hero-bg.jpeg)
Authors: Freddie Kalaitzis, Maral Bayaraa & Cristian Rossi
Introduction: This work is a brief survey of AI for Earth Observation, broken down into the following themes: sensor data, machine learning methods, and applications.
Earth Observation, Remote Sensing, Machine Learning—are all independent fields of study with their own research communities. Many textbooks have been written on each topic; entire physics and engineering courses are being taught on each sensor. Despite this, the conglomeration of Machine Learning, Remote Sensing and Earth Observation—henceforth referred to as AI4EO— raises basic questions that are rarely motivated in isolated fields.
For example:
- How can we tell what happens on Earth based on observations from space?
- How can we let data tell the story of a natural or anthropogenic phenomenon?
- How can we meaningfully combine sensors of fundamentally different mechanics?
- How can we place all data streams on the globe continuously and harmoniously?
- How can we do all of the above, mindful of noise, errors and observation gaps?
- Finally, how do we walk away with knowledge of what we don’t yet know?