The mythical broken satellite business

The premise. Every now and then we read a dramatic story about how the Earth Observation (EO) business is broken and we find ourselves reflecting through a litany of broken promises and missed opportunities. The inevitable corollary is the following: the satellite business is broken irreparably and that only governments can splurge on this data of dubious commercial value.

I have no idea if the statement above is true or not. Trying to answer these questions triggers aggravation in communities of users trying to build businesses against this infrastructure, and in the communities of entrepreneurs, engineers, scientists and officials who day-to-day carry the water in making these systems work. The effort in making these missions happen is staggering and nothing short of miraculous — be it Landsat, Planet, ESA or MAXAR missions. But, even though the industry is made of human beings, we expect the organizations they form to behave in ways that are truly exceptional — be it in terms of talent, decisiveness, altruism or plain creativity. Nevertheless, people, companies and governments are fine: they are generally smart, embrace good values and are well funded. We have never had more assets in space or data to choose from.

Even small captures are made of many granular images that can be abstracted away using sensor fusion (credit: Ignacio Zuleta).

But data still not being perceived as being adopted fast enough: the “adoption gap”

A healthy industry should provide mechanisms for optimizing the economies of the trade, be it in the form of regulations, markets or clearing houses for assets and intellectual property. The EO and Remote Sensing communities are a wonderful cross-roads where several generations of practitioners stare across vast cultural divides and missions and still do work every day. Surveyors, conservationists, security agencies, investors, governments, NGOs, tech companies, startups, farmers, insurers form a fractured collection of verticals trying to leverage EO data according to disparate funding levels and timescales. On the provider side, countries and a burgeoning New Space sector race to differentiate themselves in bringing new datasets aligned with new and old problems. But, for all the busy work, the papers, the conferences, the business plans and the success stories — there is still a widespread feeling that the data is still yet to be fully exploited and adopted. If that is the case, one can argue that we have not agreed, as a market, to fund the mechanisms that bridge the adoption gap. Possibly because we have not successfully articulated the value of easily discovering/aggregating the data, even though we struggle to surface, distribute and curate it every day. Looking at other technology ecosystems indicates that we might need to tackle the adoption gap as an industry, and that no single entity can solve this on its own. The main obstacle this presents is that we don’t have a mechanism to price and surface the commercial successes, we only see the failures — even though missions are being invested in and customers are routinely still making do with and asking for more and better data. It all starts with setting up the tools to survey, size and discover the market — what does this entail?

Uncovering is enumerating — let’s size our library

There are lot’s of things we could do, but I wanted to focus on the some directions that help us size and or feel how much data is really out. This is not comprehensive but, in the meantime, here are some concrete (old) ideas about what to do next. And we only need to take inspiration from librarians — since what we are trying to build is a traceable archive of representations, stories about the Earth’s surface. The building blocks are almost at our fingertips:

• Standardize Product Levels: the vast majority of the assets in EO are varying degrees of repetitive calibration and correction of remote sensing data. The industry has been around in one form or another for a century and standarizing at least the broad strokes of L1/L2/L3/L4+ levels would open up the door to meaningful comparisons and bench-marking based on real-world customer scenarios.
• Abstract data from hardware: the definition of product levels would naturally lead to a sensor-fusion-centric stance where the quality of EO data and metadata would be determined by its ability to interoperate with other data during model-based fusion.
• Embrace cloud-based standards: move around only the data that matters when it matters using COGs and STAC catalogs, expand the standards to include labels and IoT data.
• Catalog in the open: much in the same way literature is cataloged, instantiate in the open a proposed Geospatial Digital Object Identifier (GDOI) containing the basic metadata for a given atomic unit of observation. Record, standarize and expose licensing terms, data life-cycle information, endpoints and enable the comprehensive enumeration of all assets collected by participating missions. Make publication of granular per-capture GDOIs a requirement for all licensed missions.
• Decouple licensing from provisioning: charge what actually costs to provision the data warehousing and decouple it from licensing. Implement DRM technology for phased exercise of licence rights and to further enable co-location of unlicensed assets wherever it is needed.
• Provide a tiered commercial path to Open-access Licenses: Open-access does not equal free, let’s create a mechanism for someone to pay for some data to be open. Same way you can pay more to make an article open-access, we could provide a path for entities to buy Open-access data for a suitable high price. Let’s enable licensing terms where Open-access Licenses can be auctioned, traded and/or paid for by governments or entities for large blocks of GDOIs. List Open Licenses in GDOI records for all to understand which data has already been paid for to be opened to be free. This would unblock large swaths of commercially inviable imagery for ML benchmarking and long term scientific monitoring.
• Create industry-wide patent pools: ecosystem building is usually stiffed by the fear litigation for a body of technology that has been for the most part stagnant for the last 20–30 years. The full tech stack in EO is arguably stale and a big part of it is incumbent IP positions from old players in the GIS, aerial imaging and defense spaces. Open source initiatives mitigate some of this risk for the industry but the latent risk of litigation hinders standarization efforts. Much the same way other industries, such as telecommunication and semiconductor, embraced patent pools EO could unblock innovation by aggressively populating IP into patent pools which capture the state of the art. Any algorithm or technology that has been commoditized and is low margin can be subsequently added to the canon of standards and reference designs everyone can build monetizable innovation on.
• Catalog what matters only! Sensor-fusion, GDOIs, open licenses, patent pools, virtualization/abstraction and product levels can be leveraged to only surface traceable L3+ products that are actually usable and can be ingested by the mythical long tail of data scientists who do not want to be EO/Remote Sensing experts. Moreover, by simplifying our lexicon and surfacing data we take a step ahead in cataloging and surfacing workflows and commercial successes thus enabling the sizing and pricing of the market.

So, is anything broken? No, I don’t think so — a lot of things are working great. They are just not connected, yet. Let’s just get to work.

The views and opinions expressed in this blog are those of the authors and do not necessarily reflect the official policy or position of any other agency, organization, employer or company.