Learning To Share Can Help Agriculture Endure Now and in the Future

Food is an important part of our everyday lives, and for many, the current COVID-19 pandemic and the related restrictions imposed have highlighted just how important having a robust food supply chain is. Right from growing and rearing through to the supermarket shelves, the supply chain must be able to cope with sudden change from many directions simultaneously. Ensuring enough seed is sown, in the right place at the right time, provided with sufficient irrigation and pest control and harvested at the right time is just one piece of the agricultural puzzle. So much interlocking information goes in to keeping the food production wheels turning. But how do you collate that much diverse data from so many sources, used by so many consumers? One company, Agrimetric, are trying to do just that.

We spoke to Dr Matthew Smith, Chief Product Officer at Agrimetrics, about the roots of the company, the way in which their system can work for agricultural data providers and users and how they are playing their part in managing the food system in the current pandemic situation.

Karen Steward (KS): What led you to found Agrimetrics? What are the key aims of the organization?

Matthew Smith (MS): Agrimetrics is one of four national Agri-tech Centres of Agricultural Innovation, and its formation was driven by Professors’ Richard Tiffin and John Crawford. Richard was formally director of the Centre for Food Security at Reading University. Richard says that the 2008 financial crisis helped illustrate for him dangers in our food system, i.e. the finance system collapsed because it was incredibly interconnected: interdependencies that no one fully understood compounded until there was cataclysmic collapse. The same is true of the agri-food system: it is incredibly complex and highly interconnected.

Connecting data was seen as essential for overcoming this risk, as well as increasing productivity (feeding 10 billion people in a sustainable way is a major challenge for agriculture). However, many sectors – agricultural in particular – are reluctant to share their data. A data marketplace was seen as a way to overcome this. Data owners, e.g. Airbus who derive crop analytics from satellite data, or the Centre for Ecology and Hydrology, Natural England and the Environment Office who collect ecological data, or farmers who have their own data, could connect their data to the platform and provide access in exchange for a new income stream, achieved through a revenue share model with Agrimetrics: monetization. Meanwhile, this data becomes available for other organizations who desperately need it: BASF for example has purchased data from us to enable the creation of a pesticide application decision support tool.

Our mission is to create a thriving data marketplace to transform the agri-food sector, where data owners can share and monetize their data, and data consumers can obtain the sets of data they need easily and affordably. Our aim with this is to both accelerate the sector in getting value from data but also to build more resilience and sustainability into the sector as a whole.

KS: The food system is very complex with many “moving parts”, where do you start in picking it apart to understand its strengths and weaknesses?

MS: A thriving data marketplace is a 5-10-year mission, but in order to get there we’ve decided it’s important to begin working on areas that really need the benefits of a data marketplace right now. So, we identify problem areas, “weaknesses” if you like, where our data marketplace can be harnessed to reduce or eliminate those weaknesses. These are areas where there is a real need to obtain and combine related datasets from different providers in order to gain improved actionable insights into how the specific piece of the system is functioning or is predicted to be. Key areas are 1) the accounting of a broad spectrum of sustainability metrics that indicate the long term viability of agricultural production to minimize shocks due to unsustainable practices and 2) the improved forecasting of primary production, such as crop yield estimation, to minimize over and under production.

KS: What do you perceive as the greatest dangers to the food system in the current pandemic status? What measures do you think can be taken to prevent the situation getting to crisis point?

MS: Under and over-production of key food items as a result of shocks in supply, demand, labor and finances. These are going to happen because multiple parts of the agri-food system are shutting down and starting up at different times around the world. If you imagine the spread of the epidemic and lockdowns travelling as a wave around the world then this is also going to cause a complex wave of supply chain disruption that will probably last even longer. To help prevent the situation we need to surface the data on what and where the key surpluses or deficiencies are or are looking to be, such as for example the shortage of farm labor in many UK production systems.

KS: Where does Agrimetrics fit in global food supply in the current pandemic?

MS: We are doing what we can to support making data available that can provide insight and foresight of impending agri-food system shocks. Some of this has been on our standard product roadmap anyway, such as making available field-linked data that can be used to characterize crop development and crop health throughout the UK. This sort of information provides foresight into how production patterns are being disrupted by the pandemic. Our long-term objective is to enable an agri-food system that is more resistant and resilient to shocks just as these, so it’s important we maintain our core development. We are also looking into how we can help by surfacing relevant information through our data marketplace in response to the crisis.

KS: Data sharing seems to be of key importance to the success of Agrimetrics. How much of a challenge is it to get diverse organizations sharing information and working together for the collective good?

MS: Mistakes by others in the past have led to some suspicion in the agri-food sector, particularly by farmers, that they will be exploited when it comes to data sharing. These historical challenging precedents mean we need to make sure people understand how we ensure that doesn’t happen. It is really important for our data providers to be in full control of how their data is used, the price they charge for it and that they can change these at any time if they wish. People also tend to be far more willing to share when they’re clear about what value they’re getting from any effort. Previously it hasn’t been clear which posed another challenge. So, a priority for us is identifying those key value-returning use cases that make the value clear to the organization.

Getting organizations to work together is less of a challenge than it might seem because the different data providers are not having to work together. They simply need to see value in their data being made available through our data marketplace and a key way we make that value is by enabling data consumers to understand the relationship between different datasets so that they can obtain the “data package” they need to solve their problems. So, we’re enabling the compatibility of the datasets from the different providers without them ever having to work together, whilst they are still in control of their data.

When a user wants to obtain information, they can do this via our data catalogues. A data catalogue is simply a way of viewing the datasets we have on offer. You can see for yourself here: it lists each dataset that is available for purchase on the data marketplace. Perhaps more interesting than the catalogue, however, is the ability to search our data by concept. This is possible because the data marketplace is built on top of a knowledge graph: a way of connecting data used by Google and Amazon. It works by explaining the relationships between the concepts. So, if I search for “soil”, the knowledge graph would pull soil data from multiple datasets, it would also flag data related to soil, e.g. crops or weather.

This is a far more efficient way for users to search our data because a) it’s clearly timesaving vs searching through independent datasets and b) its more cost effective, as you pay for only what you need rather than entire datasets that may contain irrelevant data. Finally, it’s really exciting because structuring data in this way is easily machine readable. We foresee AI being able to self-serve on the knowledge graph, uncovering insights and relationships that we didn’t even know to look for. In manufacturing, Smart Factories have exploited AI and 100% connectivity to get just these kinds of revolutionary insights.

KS: You mention the use of a “knowledge graph”. Are there currently any gaps in that “Knowledge Graph” that you are keen to fill?

MS: Yes. The key gaps in our knowledge graph that we’re actively filling are 1) including more sustainability – relevant information like carbon storage and emissions and priority habitats 2) More agri-landscape information like hedgerows and woodlands 3) more farm business information like net profitability for different land use options and 4) more predictive information like crop disease forecasts.

KS: Many minds are on the current situation in which we find ourselves, but thinking further ahead, what are some of your hopes and aims for the future of Agrimetrics?

MS: As stated, we want to create a thriving data marketplace for the sector and so our aims for the future are that our data marketplace is used as a key conduit for decision making throughout the agri food sector, and that the commercial model underpinning that is a sustainable one for Agrimetrics. The hope is that, as a consequence, the sector is overall far more resistant, resilient and sustainable as a result of always being able to harness value from the right data at the right time.

Dr Matthew Smith was speaking to Dr Karen Steward, Science Writer for Technology Networks.