Conversations With the Coalition: Sam Schiller - Climate TRACE

As part of an ongoing Climate TRACE series, we are interviewing individual coalition members about their work. We recently caught up with Sam Schiller, Co-Founder and CEO of Carbon Yield, which works with more than 100,000 acres of croplands to help them access the resources required to transition to regenerative agriculture.

Which emissions sector are you focused on, and why? 

We focus on emissions from agriculture, primarily from croplands and livestock operations. Agriculture occupies a significant percentage of land globally — nearly 40% — making it a dominant and incredibly diverse set of activities that almost literally shape our environment. Agricultural soils are also an important reservoir for storing carbon, and can be managed in ways that draw down atmospheric carbon. This double-edged sword where agriculture is both a source of emissions and a potential climate solution makes it compelling to study, and the focus of new environmental markets and incentive programs. 

What we're trying to capture first at Climate TRACE are the most dominant sources of emissions from producing food, fiber, and ethanol for biofuel. These include livestock facilities (particularly beef and dairy systems), the use of synthetic fertilizers, and rice cultivation (flooded rice paddies produce methane-emitting microorganisms). 

What is your approach to estimating agricultural emissions?

Given agriculture’s decentralized nature, it has historically been difficult, time-consuming, and cost-prohibitive to gather solid emissions data from this sector. As one of the last of the major industries to digitize, a lot of efforts to document agricultural emissions have depended largely on information from surveys and other voluntary efforts. So not only are these assessments expensive, they also carry a lot of uncertainty and can place burdens on farmers.

At Carbon Yield, we felt like there was a real opportunity: If we could collect (and publicly release) emissions data through some more automated, independent means, then we could not only bring greater accuracy and transparency to these efforts — we could also decrease the cost of measurement in the first place.

How does your approach look in each of your three focus areas?

Let’s look at livestock facilities to start. [Interviewer’s note: We’ll take a closer look at emissions from synthetic fertilizer and rice cultivation in future articles.]

Beef and dairy operations are often relatively large emitters on the agricultural scale, but their air emissions are not regulated. While there’s movement in some areas toward regulation, broadly speaking there just isn't comprehensive data on these facilities. 

So, our first task in measuring livestock emissions is literally to find the facilities in question. Our process uses a machine learning model with a rapid detection system that can take in a small amount of training data on what a facility might look like and find all reasonable facsimiles within an area. We ingest geospatial data and satellite imagery to run this tool, and it's proven very good at finding these facilities. 

Thanks to areas like California where there is strong permitting data with both geolocated facility locations and cattle population, we have been able to build models that use the size of facilities and other variables to help estimate the number and type of animals at facilities where we do not have access to permitting data. And from there we can then apply emissions factors per head of animals that might be in an operation. Of course, not all systems look the same, so the model has to be tweaked and tuned as we move our search around the globe.

To date, Carbon Yield has contributed data from two U.S. states and from a region in Argentina to the broader Climate TRACE inventory. For livestock emissions, we expect the next Climate TRACE release will include parts of South Africa, Australia, China, Russia, Brazil, as well as additional U.S. states.

What challenges have you faced along the way?

One big challenge has been getting access to imagery to cover so much land, and then harnessing the kind of computing capacity necessary to evaluate all that imagery. Our initial plan with the livestock modeling, for example, was to take in all the imagery of the entire world. It created a huge backlog of servers running 24/7 to ingest even one country. We tried this for India, and it basically broke the system. 

It’s also challenging to assemble the right tiles to view (i.e., individual satellite images covering a given area of the Earth’s surface), such as those without issues like clouds or haze. From there each tile has to be perfectly arranged in order to create a seamless mapping process. So we had to figure out ways to downscale the work.

How can organizations use these data to support target setting and emissions reductions?

Multiple parties can find tremendous value from these data. For starters, procurement leaders can use it to support goals of sourcing commodities from regions with a lower carbon intensity. Governments can also use these data to better monitor emissions and respond to inventories appropriately, while scaling resources to hotspots and areas that have the greatest need. 

We also see deep value in public datasets that show where you can make interventions. For example, in livestock, innovative companies developing low-carbon products and projects can use these tools to identify hotspots where they might deploy feed additives or manure management technologies to reduce emissions. Overall, we believe these insights can help position low-carbon commodities as profitable for producers to create, and simple and affordable enough to monitor.

Interviewed by Daisy Simmons.