Making satellite data work for us all

A new blog from the team at Imago, our Data Service for Imagery

Satellite imagery could help solve some of the UK’s most pressing problems. But right now, few in social research or policymaking can actually use it.

Satellites capture detailed and comprehensive images of the Earth. They allow us to see and understand our planet in unprecedented detail. Imago, the Data Service for Imagery, part of Smart Data Research UK’s family of six data services, exists to make satellite imagery more useful, usable and used.

What do we mean by that?

Delivering data products that researchers and policymakers need but can’t currently use
Using formats that are familiar and easier to work with
Training, community building, and providing examples that demonstrate real value for UK research and policymaking.

What makes satellite imagery so great?

There are three trends converging to make this an opportunity we don’t want to miss.

1. We have more and better imagery from satellites than ever before

Launching stuff into space is much cheaper than it used to be. As introductory economics would predict, lower the price of something and more of it will be bought. For satellites, this means more placed in orbit (and more data). Newer satellites also benefit from the revolution in consumer electronics of the last few decades, so the sensors onboard can capture much more than before (better data).

Historically, we had to be content with 30m resolution imagery, typically used to capture land use, vegetation, or impermeable surfaces. These days resolution is higher and sensors are built to ‘see’ so much more about the Earth’s surface – think air pollution or temperature.

2. We have much better ways of extracting information from imagery

Images may contain lots of useful information, but it’s encoded in pixels. They don’t mean anything by themselves. To make them useful, you need a ‘magic decoder’. This turns pixels into information such as buildings, roads, or land uses. Luckily, the last fifteen years of progress in computer vision has delivered just that. In the early 2010s, we had the first deep learning models. By the end of the decade, we had robust and more advanced neural nets. Nowadays, the foundation models for vision, how we teach computers to see things is undergoing a new revolution.

3. The technology required to access the algorithms and work with the data has been democratised

Compute is cheap(ish) and software is free(ish). It’s a brave new world. It’s what makes all this possible at scale.

What does this mean for social, economic, and health research and policy?

The confluence of these three trends means we can do things now that we couldn’t do before. This is invaluable for measuring, understanding, and improving the world around us. We can measure stuff we haven’t yet been able to measure, particularly at scale. And we can measure better, sooner, and more often many things that, up until now, we’ve measured slowly, late, and sparsely.

Sometimes we simply can’t measure things that matter, like temperature or air quality, at the level of detail and scale our work needs. Other times, we have data, but it’s so limited that we only use it because there’s nothing better. It’s not good data; it’s just the best we have.

For many of these problems, we believe satellite imagery can change what counts as good enough data for research and policy decisions.

Why should we try to realise such promise? It’s not enough just to say ‘because we can’. The world is becoming more uncertain and more unpredictable. Think climate change and regional inequalities. We can’t wait for the next census to know how society and the built environment are changing and adapting right now. Decisions need to be made before then. Actions taken. For these to be meaningful, there’s a whole lot of understanding we need to unlock first.

It’s important to note that our vision does not clash with traditional sources of data like the census or surveys. We need both. Satellites can leverage the value of censuses by providing additional context. And most of the value held by imagery relies almost entirely on our ability to relate them to existing measurements of phenomena. Our best shot at that ‘labelling’ in the social realm are traditional sources. We need them more than ever.

Why hasn’t the satellite imagery revolution already happened?

If challenges and solutions are so readily apparent, why aren’t we already realising this promise? Our educated guess is that at least three barriers get in the way.

Imagery is big. It takes a lot of space on storage disks. It contains a lot of information. It requires a lot of expertise to work with. In fact, it’s so much to deal with that you need a plan just to store, access, and analyse these data – the ‘it doesn’t fit in your laptop’ problem.

Imagery is hard. And while ‘magic decoders’ have become much more accessible, the science underpinning it all remains tricky.

Imagery is different. We don’t train social scientists to work with images. Local government officers and government analysts don’t know where to start to make sense of it all. This contrasts with other disciplines where imagery is front and centre. Social science and health have, for a long time, spent all their training allowance on other skills and techniques that are fundamentally different.

These barriers are at the core of what we are trying to change with Imago.