Untangling the European Space Market

When you find yourself at a place that is non-transparent and has many scatter players on various levels yet with overlapping fields and funny acronyms, then you most likely landed in the European space market. It takes a little while to cut to the chase.

But don’t worry: In this article, I try to untangle the earth observation market by splitting it into mission owners, data vendors, and service providers to give you a clearer picture.

Before we start, consider the image above.

Did you get it?

Well, at a very (very) high level this is it. The entire data value chain: from satellite to end-consumer.

Easy. All done, let’s call it a day……not.

Things are a bit more complicated than that.

The space market in Europe is relatively young, compared to e.g. the US, and is therefore dominated by SMEs and start-ups. This is also manifested in public spending, with the US spending roughly €40 billion in 2019 (or 60% of the global market), compared to a total European public investment of a mere €10.5 billion (or 16% of the global market share). On the contrary private funding in the US reached €5 billion compared with €188 million in Europe (for start-ups only) in 2019.

The global space industry could reach generate revenues of US $1 trillion or more in 2040, up from US $350 billion currently. Looking at the earth observation (EO) segment only, the market for value-added services is forecasted to double from roughly €2.8 billion to over €5.5 billion over the next decade. From a supply perspective, the EO market is jointly led by the United States and Europe with market shares of 42% and 41% respectively. Europe plays a leading role in the market of Analysis, Insights and Decision Support (the subset of value-added services closest to end users) with a 50% market share covering all segments, contributing to its overall market share above. So the market is huge, the potential is there, and the impact on global challenges given.

Before we drill a bit deeper it is important to understand what vertical integration and downstream focus mean. Vertical integration means a company covers the entire data value chain, from operating satellite missions to data processing. Downstream focus refers to companies providing data processing and other services. But let’s get started, and see what each of the points in the image above — (1) mission owner & data producer (2) data vendors, (3) data pre-processors & service providers and (4) the market for EO services — has to offer

(1) Mission owner & data producer

Copernicus is the earth observation program jointly established by the European Commission (EC) and the European Space Agency (ESA) in 1998. As an independent European observation system, Copernicus is operational since 2014, providing up-to-date information for environmental and security-related issues based on open data and open algorithm policies.

ESA’s flagship missions are the Sentinel missions, carrying a range of technologies, such as radar and multi-spectral imaging instruments for land, ocean, and atmospheric monitoring:

• Sentinel 1 — constellation of two radar satellites — 1A & 1B. Unfortunately, Sentinel-1B experienced a malfunction on 23 December 2021 and is no longer able to provide data. To get back on track, ESA plans to launch Sentinel 1C soon.
• Sentinel 2 — constellation of two a multispectral high-resolution satellites — 2A & 2B — imaging providing imagery for a wide range of applications (e.g. vegetation, soil and water cover, inland waterways, and coastal areas).
• Sentinel 3 — constellation of two multi-instrument satellites — 3A & 3B, measuring sea-surface topography, sea- and land-surface temperature, ocean color and land color.
• Sentinel 4 — mission to monitor the atmosphere from a geostationary orbit.
• Sentinel 5P — this precursor mission provides data on a multitude of trace gases and aerosols affecting air quality and climate.
• Sentinel 5- mission to monitor the atmosphere from a polar orbit.
• Sentinel 6 — mission that carries a radar altimeter to measure global sea-surface height.

Here is an overview of all ESA missions and planned missions.

The Landsat-Programm is the second mission I would like to mention, a joint effort of NASA / USGS operating since 1972 and continuously acquiring space-based images of the Earth’s land surface. The opening of the Landsat data archive in 2008 and the launch of ESA’s Sentinel Missions formed the basis for cost-efficient and effective implementation of remote sensing applications.

Despite the Sentinel missions and the Landsat program, there is a vivid market of satellite manufacturers, that either focus on manufacturing or integrate vertically. Here is an excerpt (a very short one) of satellite manufacturers that integrate vertically in the EO market and downstream service providers:

(2) Data vendors & service providers

The European Commission facilitates standardized data access and has funded the deployment of five cloud-based platforms — the DIAS (Copernicus Data and Information Access Services). The DIAS allow centralized access to Copernicus data and processing tools. There are five DIAS: CreoDIAS, Mundi Web Wervices, ONDA DIAS, WekEO and Sobloo. The DIAS provide full archives of Sentinel-1 GRD, Sentinel-2 L1C, Sentinel-3 OLCI, and SLSTR, Sentinel-5P, European coverage of Landsat-5, -7 and -8, Envisat MERIS as well as significant parts of Sentinel-2 L2A.

The Euro Data Cube (EDC) is an ongoing project funded by ESA, focusing on integration, upgrade of several operational and prototype services, and mass processing. So the EDC tries to cut through the weeds, by connecting the dots along the line. EDC is based on data archives and cloud platforms such as the DIAS and Amazon Web Services. EDC provides managed compute and storage environments, advanced EOdata manipulation capabilities and feature management via GeoDB. Data access to both open and commercial satellite missions is enabled via Sentinel Hub — fully integrated into EDC (we will learn more about Sentinal Hub a bit later).

Google Earth Engine — the elephant in the room — offers a wide range o of data products and processing power. It is an excellent way to get a sense of the vast possibilities remote sensing has to offer, is free of charge for scientific purposes and is therefore a good place to start. Recently, Google Earth Engine joined forces with Google Cloud service and offers a professional subscription for around 2000$. In terms of pricing, this is very similar to the service package offered by DIAS. I am not sure if Google offers some perks. Maybe you have some ideas? To me, it is also not entirely clear how IP rights are handled when using Google Earth Engine. I tried several times to reach out to Google but was not successful. Again, maybe some of you can shed light on this?

Microsoft Planetary Computer too offers a broad selection of data products, managed development environments and a few applications. The HUB is Microsoft’s focal point that lets you spin up several managed environments, such as R, Notebook, PyTorch, Tensorflow and a QGIS preview. Among the applications, you can choose is a Land Use and Land Cover classifier. I tried to get in the module up and running, based on the documentation, but could not even execute the first code cell from the documentation (dependency issues in a managed env…). However, this application feels a lot like a slim version of Google’s Dynamic World app. The Planetary Computer seems quite interesting, but it is not quite there yet. To get access, you need to fill out some sort of application form (don’t know what the criteria are). There are some free resources you can use, but the underlying cost structure is a bit untransparent.

(3) Data pre-processors & service providers

Several private data vendors exploit these satellite data archives and offer globally available analysis-ready data and tools for on-demand processing in a cloud-agnostic manner. Those platforms also offer purchasing and processing options for satellite data from commercial providers like Planet.

There are several data vendors out there (including the DIAS), but from my perspective one stands out: Sentinel Hub, a subsidiary of the Slovenian company Singergise, established in 2017. Sentinel Hub provides a wide range of apps and utilities including data access to several could platforms, such as the DIAS, AWS, and others, They develop and maintain their own python library to work with geospatial and satellite data and provide very powerful APIs, e.g. the process API and the batch process API. The process API is the most commonly-used API, as it provides data from satellite imagery as an image, as source data, or as simple-band combinations. An important limitation of the process API is that one can query a maximum of 2500*2500 pixels, so requests need to be adapted accordingly. The batch API enables requests for larger areas and/or longer time periods for any Sentinel Hub supported collection, including BYOC (bring your own data). Another important point, not covered by Sentinel Hub is the display of results over a web application. The use of a Geoserver could play a vital role here. Overall more than 10 petabytes of data from the different sensors can be accessed.

(4) Market for EO services

The potential for EO applications is enormous. However, for quite some time I have been looking for use cases that go beyond observing and pointing out risks and threats, and in fact, focus on creating opportunities. I feel like this way of thinking has not trickled through yet.

I am very curious about what YOU think about that.

Any ideas, sources, initiatives,…come to mind?

Conclusion

It takes some time to understand the mechanics of the EO market. Yet every time I go on the internet I find something new since it is such a vibrant environment.

However, I wonder what EDC and other providers are aiming at? Notebooks are a good place to start, play around and maybe build a prototype, and seem to be the standard feature of most platform providers. But: running complex computations in these notebooks will likely cause memory issues and time-outs. How do you roll out an EO product in such an environment? And what about scaling?

There is a distinct difference between platforms targeting scientific research and production. In research, you try to find evidence that supports or objects your hypothesis. The things like efficiency, scalability, reliability, stability, and agility are of secondary importance but are vital for any business venture. Before TensorFlow, executing a machine learning project was a pain. Tools, software, ideas, prest practice,….it was all over the place. Creating a toolchain for your ML project was truly a tedious undertaking (it still is a tedious process, but nothing in comparison). I feel like EO is in a similar place right now. But we are getting there…….

This article leans more toward the European space market and — by far — is not comprehensive. If you want to dig deeper into any of the topics mentioned in this article or find new directions → Google is your friend. Just remember to share your insights.

I would like to hear your feedback. Drop me a message in the comment section below, or via LinkedIn.

Stay tuned,

this is Matthias