Democratization of space to offer more opportunities to understand Earth: Descartes Labs’...

Democratization of space to offer more opportunities to understand Earth: Descartes Labs’ Product Head

SHARE
Democratization of Space
Fritz Schlereth, Head of Product, Descartes Labs

With expansion in the commercial space industry, improvements in launch systems, sensors and other input technologies are transforming the satellite data value chain. This extends from the origination of data to the insights and analysis that represent value to businesses and public organizations, says Fritz Schlereth, Head of Product, Descartes Labs.

There is a lot of talk going around about revolution in satellite industry space and democratization of space. What exactly do they entail?

We are witnessing a convergence of technologies that will revolutionize remote sensing, generating more diverse data than ever before. Everyday we’re seeing new constellations going into space that are decreasing costs while increasing capabilities. Meanwhile, we are also improving our ability to process larger volumes of data and automatically extract insights. This is driven by the rapidly maturing cloud-based computing model, but also by leveraging recent advances in computer vision and machine learning. This conflux of technologies is unlocking more and more opportunities to understand the planet in greater detail and transform the way businesses use earth observation data.

It is said the ability of satellites to transform businesses and quality of life is significantly more relevant today than ever. Can you elaborate?

Decreased costs and increased capabilities are making earth observation satellites more relevant to business and the public good now more than ever before.  The state of technology within the satellite industry is evolving rapidly. On one hand, improvements in launch systems, sensors and other input technologies, and innovations such as the smallsat architecture are driving down costs. In parallel, more sensors and a greater diversity of sensor types mean greater spatial resolution, higher temporal cadence, and richer spectral coverage.  This combination of decreased cost and increased capabilities opens up new use cases, industries, and applications for businesses.

Do all these advances mean more and more space-based sensing and connectivity services with continual increases in image resolution and the area satellites can cover at lower cost?

Without question, the public investment in earth observation satellites and the resulting data sets is an incredible resource that continues to bootstrap the emerging commercial ecosystem. But, by leveraging the start-up model, which is geared for rapid iteration, the commerical space industry will push for scale and efficiency.

As one example, we’ve found that the distribution of satellite imagery is a big barrier to rapid iteration and solutions development. Satellite imagery is a massive data set and while some constrained problems require a small number of scenes – regional and global scale problems require data sets on the order of petabytes which needs super-computer equivalent compute capacity.  Furthermore, analysis at this scale must shift from the human analyst to a machine intelligence approach – where the extraction of analytical data is automated.

Startups and the commerical space industry will drive improvements and transform all steps of the satellite data value chain. This extends from the origination of the data (i.e., the satellites themselves) to the processing and refinement of the data, and finally to the insights and analysis that represent value to businesses and public organizations.

How far are we from creating a real-time digital twin of the Earth?

Change detection is one of the most valuable applications of satellite imagery. However, the traditional tradeoff (forced by technological limitations) between spatial resolution and temporal cadence has limited use cases to high frequency changes at low level detail or to higher detail but lower frequency changes. Lower fabrication, launch, and operational costs mean that more satellites are being deployed which allows us to chip away at the tension between resolution and cadence. This will allow us to observe and analyze more classes of human activity.

There is also a talk about a “sharing economy in space”? What does it entail?

At Descartes Labs, we believe that machine intelligence and data fusion are the keys to maximizing the value of earth observation data. Today, the transactional model for satellite imagery is largely optimized for just that – the sale and purchase of imagery. To accelerate progress, we must shift to a model that is better suited to data, which entails an ensemble of fundamental changes. First, transactions need to be instantaneous and programmatic.

Second, transactions need to reflect the value that the purchaser obtains from the imagery. For example, the purchaser should be able to easily evaluate the value of the data for his/her application and purchase scenes (or even pixels) in a piecemeal fashion. Third, the transactional space needs to be accompanied by a compute environment that allows the buyer to work with the imagery immediately and efficiently.

One of the major barriers to rapid development is simply the fact that data sets aren’t co-located in a high performance computing environment. Fourth, the transactional space needs to house multiple (really all) data sets in the same compute environment. Ultimately, the most interesting and valuable applications leverage multiple data sets to create a richer awareness (and greater predictive accuracy) of the observed phenomena.

How can we create a more humane and just world by democratizing access to space-based resources?

From the iconic blue marble photograph to our first experience zooming into our hometown on Google Earth, satellite imagery has always produced a startling effect in people. In part, this is because satellite imagery extends our perceptions in space and time.

In contrast to our unaided senses, satellites allow us to see entire regions (or the entire planet) in a single image or to witness changes that unfold over years (or even decades).  Of course, this expansive observational power extends beyond the visual – it enables scientific predictions and analytics that can be fed into decision-making for public policy and commercial interests.

Now, satellite technology is increasingly extending in the other direction – the ability to observe with greater spatial resolution and higher temporal cadence. These enhancements are revealing dynamics on the ground that are closer to the anthropomorphic scale. In addition to monitoring the U.S. corn belt throughout the growing season, from emergence to harvest, we can now see daily changes in a shipping yard.

On either end of the scale – from the very broad to the highly detailed – satellite imagery affords us visibility into changes and processes on the earth’s surface that would otherwise remain hidden. Without this technology, our ability to monitor forest loss, changes at the poles, the emergence of new population centers, the human impact following a natural disaster, and many more dynamics would be far more constrained.

Data is the raw input to realizing these benefits, and there are two major stumbling blocks today to unlocking it’s value. The first is technological – from both the generation of the data and the compute resources to turn the raw data into actionable insights. The second is the distribution model itself. The ability to rapidly test the data for a given application and transact will accelerate the iterative loop. At Descartes Labs, we are focused on solving both of these problems.