🗺️ Geospatial data to map the food moneys
- stephanielwalton
- Dec 13, 2024
- 7 min read
Updated: Feb 11
One of my big tasks for the PhD is to build a dataset of all the major meatpacking plants in the USA, how many cattle they slaughter a day and their 'investment cadence' - the timelines on which their owners invest a big chunk of capital in upgrading or expanding them.
To do this, I'm relying heavily on satellite imagery. Much of my time ATM is spent outlining meatpacking plants to track how the spatial footprint of plants change over time (PhDs are so glamorous) and, aside from giving me carpal tunnel, this is giving me lots of time to ponder upon the wonder that is geospatial data.
Because of the inherently land-based properties of agriculture, geospatial data has been used for agricultural and land use research for quite a while. But these uses have yet to create a major link-up between the food and the moneys 💸 but the field is moving very fast. Let's see how!
🛰️ Some brief plain-speak on geospatial stuff for the uninitiated-but-curious
Geospatial data and analysis is not the same as disaggregating data by an administration boundary and then visualising it on a map. For geospatial data, imagine a spreadsheet with lots of rows for observations (farmers' field or, say, a meatpacking plant) and columns for variables - and there is a column at the end that literally says 'geometry' and the column is full of latitudes and longitudes that pinpoint where that observation is in space.
With that geometry column, you can import your data into the miracle that is QGIS and literally place that observation on a map. Then you can do all kinds of interesting and fancy analysis with it - everything from classifying land use types, measuring distances and monitoring changes over time (either manually or with AI) to running spatial regressions or kernel density estimations. You can also combine it with other non-spatial data, like census data or weather forecasts or COP attendance data. There is lots of fancy and visually-pleasing fun to be had.
There are several sources to go to for geospatial data - the most obvious being satellite imagery (for the truly curious, this is a great explainer on how satellite data works). When most people think of satellite imagery, they think of Google Maps. Google Maps is truly one of the greatest wonders of our modern age. If you ever have some free time, download the old school Google Earth Pro and click through some of their historical satellite imagery.
But alas, Google Maps doesn't let you download their imagery so you can't import it into the aforementioned miracle that is QGIS to do all your fancy analysis. For that, we go to NASA and the European Space Agency (ESA) who are using their satellites to make geospatial data free to use for us peasant academics who don't have a contract with Maxar. The only bummer is that, while Google Maps spoils us with images that capture earthly details at up to 30cm, NASA and the ESA are in about the 10m range. Still a miracle, but not as fine as we might want for some things.
A relatively recent major glow-up is that the ESA added some jiggery-whatnot to their latest satellite, the beloved LANDSAT 5, that can monitor some types and sources of greenhouse gas emissions from space. If you're thinking, "OMG - let's use LANDSAT 5 to monitor methane emissions from cows!" - imma stop you right there. I have spent an inordinate amount of time trying to figure out if this is possible and, like, it kind of is and it kind of isn't. This is a big emerging topic and get in touch if you want to discuss more but, suffice it to say, we can't use it in that way you're thinking of - but this is a fast-moving field.
🐄 Geospatial data for food systems: my personal favourites
Ok, there are tons of ways geospatial data is being used for agricultural production - like I said above, the affinity is obvious. I'm not gonna go into satellite data for monitoring crop yields or soil water content or all of that unquestionably amazing stuff. Agricultural production is obviously a kind of important part of food systems - but food systems are also much more! So here are a few examples of my favourite non-agriculture related uses.
ClimateTrace (with Earth Genome)
They've wrangled a truly impressive amount of national-level data and combined it with the auto-identification of dairies (like in the UK) and feedlots (like in Brazil), measurements and emissions factors to predict emissions form individual agricultural businesses. It's amazing.
Global Forest Watch
This one have been around forever but it's a fav. GFW uses satellite imagery to monitor both deforestation and afforestation and links it to the location of palm oil mills and .
The UK's Gridded Emissions
The UK provides data on methane, nitrous oxide and ammonia emissions from agriculture on a 1km x 1km grid using a fancy model that I have spent way too much time figuring out how it works. It is not a spatially refined as ClimateTrace (to protect farmer privacy) but the emissions are probably much more accurate because they are working with farm and livestock census data that no one has access to except Defra. The data is free to download!
The Plotline's Food Twin, led by Zia Mehrabi
An amazing aggregation of data from all the places used to visualise where food is grown and where it goes in the USA - and which supply chains are exposed to environmental stressors.
Iceye's Port Monitoring
Ok this is not free sadly - but it should be! And is a great example of what is possible with satellite imagery for those who have the money to pay for it. Imagine being able to monitor the comings and goings of bulk commodities carriers around the world........
💸 Spatial finance for food systems transitions
Finance peoples are famously data-obsessed. No one can build a model like a quant. And in some cases (e.g. insurance, physical risk modeling), the industry have been quick on the draw with geospatial data. But in lots of cases, not. Finance is kind of a de-spatialised industry. What, for example, is the lat/long of an share in a multinational company? It has none. If floats on the breeze like the Forrest Gump feather 🪶. In some cases, an investment might 'land', if you will, at a location - say like when Macquarie buys up the UK's gas infrastructure. But even when an investment is physically located somewhere, it does not mean that geospatial data was used to make that investment decision or for any subsequent analysis
This should and is changing (slowly) because (1) spatial data is fun and finance people are allowed to have fun and (2) spatial data is really important for helping finance contribute to their client's and investee's sustainability transitions and for their own sustainable finance transition.
Ben Caldecott, who runs our Sustainable Finance Group at Oxford (and my supervisor), published a piece introducing the concept of 'spatial finance' to the academic finance literature and argued for its many practical applications - my personal favourite of which is the use of asset-level data for monitoring the firms that financial institutions work with.
'Asset-level data' 📍 is data on the location and characteristics of 'assets' to which financial institutions are linked through investments or financing. Think your meatpacking plants, your steel plants, your mines, your farms, your waste dump sites, your infrastructure 🏭. Asset-level data is where the de-spatialised finance economy meets the spatialised real economy. It grounds finance in physical reality and so is a particularly important element in giving investors and lenders insight into how they shape real-world outcomes.
There are two obvious use cases for asset-level data in food systems transition. The first is 📈 using asset-level data to measure financed emissions and other environmental impacts. Just like Tyson and Danone, financial institutions have set net zero targets for their own lending and investment portfolios - which means they need to be able to measure how much emissions their loans and investments are financing. The kind of work that ClimateTrace is doing with spatial data and that we're working on at OxSFG is critical and very helpful for the first. To do this with high accuracy, especially for industries as dis-consolidated as agriculture, is very very very difficult (although, how much more accurate does data need to get for us to act, amiright?)
The second use case is using 🔎 asset-level data to monitor firms' investment behaviour and using it to assess the credibility of their client's transition plans and how their capital is being put to work. If Tyson has a net zero transition plan but then satellite imagery shows they just expanded the size of their meatpacking plant in the Cerrado by 1000 square meters with the line of credit you gave them, what does that say about the credibility of their plan?
We can put spatial data to work to answer these types of questions. The pre-requisite for impact, of course, is that investors' and lenders' own transition plans are credible and that they are serious about driving forward the food systems transition in both the real and financial economies - which sadly grows increasingly tenuous. But for those investors and lenders who are serious and credible, asset-level data is critical and exciting and fun and pretty and cool.
