Farming is the oldest industry on earth and arguably the most important. It feeds the world, employs millions and sustains life. Yet it's under attack because agriculture requires energy and nutrients, it emits 9% of UK greenhouse gas emissions and consumes 70% of our water. To produce one slice of bread for your toast in the morning takes 40 litres of water. Put another way, the water used to produce the bread for a city the size of Manchester in one year is enough to keep the whole of the UK in drinking water for four years.
Farmers rely on very diverse data and years of accumulated experience to produce our food. But farming food requires constant attention to detail, a lot of hard work and, operating margins are very slim. The UK Agriculture and Horticulture Advisory Board (AHDB) report, "Driving productivity together" describes how improving our farming productivity plays a significant part in UK competitiveness, and is critical to operating in a global market. But UK productivity in farming has not kept pace with our main overseas competitors, and the agriculture sector is three times less efficient than other areas of the economy. Given the resources required for agriculture, and the national need to increase productivity, our greatest challenge now is how we help agriculture to consume less and produce more.
If we want to understand how we bring data and diagnostics to traditional broadacre farming so that we grow more from less, we have to compare how we make decisions on the farm today with how we will exploit data in the future to make decisions.
The traditional unit of activity in crop production is a field. However, no particular field is ever homogenous, and farmers know; moisture varies across their fields, edge effects from hedgerows can change the environment, and soil nutrients can be affected by their farm operations. Yet operations treat the whole crop in a field the same way, regardless of the actual variability across the field.
If we truly want to optimise our resource consumption and maximise crop productivity, the grower needs to understand the individual needs of each plant and the in-field variability. For best performance, the treatment of each plant should be tailored to its unique requirements so that resource use is optimised. Rather than treating the whole population of plants in a field the same way, we could monitor and manage each plant individually, so that it can reach its yield potential - a kind of personal healthcare for plants. The data that enables this new 21st-century farming will pour in from irrigation systems, weather stations, imaging systems, soil activity, pest and disease sensors, all located in and around the crop, on-farm machinery, robots and satellites.
With this explosion of data sources, the amount of data we expect to collect for farm data and diagnostics will double each year. As technology and the internet of things (IoT) becomes ubiquitous in food production, we have the parallel challenges of managing that data and analysing it to help us drive down consumption of resources in agriculture and increase farm productivity.
Our technical challenge is to turn all that data into a real-time indicator of how the crop is doing. Imagine you are a farmer and can access an app that lets you see how each part of the field and plant is performing; where the moisture is, the microclimate, and the ripeness of the crop across the field. With this knowledge, a grower can plan crop protection activities, schedule machinery maintenance, and protect biodiversity in increasingly efficient ways. Rich predictive models to understand the disease and pest risk, can be created from this data, helping the farmer have confidence in spraying (and not spraying) pesticide or applying fertiliser.
A new generation of farming has started to emerge where the plants themselves request resources and treatment, and there is a system that automatically responds. These future systems of agriculture are being targetted by new technology startups, like the Small Robot Company. This "Per plant precision" enables Integrated Farm Management (IFM), being "kinder to the soil, kinder to the environment, more precise and more productive".
The farmer will be able to focus in on one particular area of the field or rows of plants and monitor their ripeness, nutrient content and instruct treatments using their experience and knowledge. This precision agriculture vision of our future relies on huge volumes of diverse data about the farm, the field, the crop and the plant. It is our data and diagnostics capabilities that predicate its success.
How do data and diagnostics capabilities work in crop and farm management?
The major stages in the data and diagnostics pipeline are; data capture, sampling and analysis, machine learning, management and automation. Each of these form part of the chain of custody and actions that capture field observations and turn them into actionable insights for the grower.
Data Capture: Observations, in increasingly higher volumes, will be captured as data, by sensors that use remote monitoring by networks of robots, UAVs, phones and satellites. Observations will continue to be made and recorded by experts such as the farmer and the agronomist. These observation data are all recorded electronically and passed back to data centres in the cloud via ubiquitous wireless networks.
Sampling and Assay: To augment data captured by sensors and remote systems, physical samples of plants, pests, and soil may also be taken and passed to laboratories for assay. Notoriously this can be a rate-determining step in the data and diagnostics pipeline. To address this concern CHAP has a developed a Lab to Field Capability of mobile crop science labs, that bring a diverse range of assay capabilities to the field, enabling rapid on-site assays and timely data collection.
Data Analysis: All captured data are accessed by data science teams to determine the current state of the crop ecosystem. Rich predictive algorithms can be applied to quantify the risk of pest and disease, the current health of the crop, and the likely yield and quality outcomes. These aggregated and predicted values then update digital representations of the farm, field and plant. Results are stored on the farmer's own farm management systems and used to update a digital-twin of the field, such as the representations in Agrimetrics Field Explorer.
Machine Learning (ML): ML is a branch of artificial intelligence that provides the ability for machines or computer programmes to learn and improve on what they "know". It has great promise in agriculture where diagnostics systems will generate huge volumes of data from which ML can learn and offer insight to the grower. Supervised ML is where the computer program is trained to recognise a thing of interest. It could perhaps be images, taken from in-field sensors, of Cabbage Flea Beetle present and not present in the crop. Over time it will learn to recognise when Cabbage Flea Beetle is present. This approach is used by Hummingbird technologies to create insights about crop health, disease risk in the field.
Management Systems & Automation: Decision support systems and tools based on the analytics and machine learning can be used by the grower to monitor the crop and make decisions, such as whether to apply crop protection chemistry. Crop Monitor and the BASF Water Stewardship Tool are both examples of this type of decision support tool. This comprehensive and accurate data about the state of the farm and diagnostics can be the basis for more extensive integrated farm management decisions. For example, the Syngenta and RSPB Bird Environmental Stewardship Tool (BEST) helps farmers make better land stewardship choices that will enhance farmland bird populations. It exploits detailed data about the sub-field location and the known effectiveness of stewardship options for boosting bird populations.
These examples are tools which help the farmer make decisions and initiate an action. But the bundle of data analytics plus insight from machine learning gives farming the opportunity to delegate decisions and actions to automated systems. Such automation is already being used with many established irrigation systems on farms in Europe and North America. Currently, decisions on weeding individual plants can also be handed-off to automated weeding systems like those offered by Ecorobotix and in the future, more multi-functional automated systems like Dick from the Small Robot Company.
The impact that data and diagnostics could have on the future of farming is immense. But there are some blockers which if allowed to remain could slow our path to clean growth in agriculture. In particular, the components in the data and diagnostics pipeline; the measurement systems, the data capture networks, and the analytics algorithms all need to be FAIR. FAIR stands for Findable, Accessible, Interoperable and Reusable principles (as published by Force 11). Just how critical it is that data and components are FAIR was described by GODAN, the 2016 summit report "A Global Data Ecosystem for Agriculture and Food".
The UN: Food and Agriculture Organisation (FAO) describes how 80% of the effort in these new data systems, goes into searching for data, curating it so that it works with other systems, and enhancing it for reusability. Many organisations are adopting FAIR principles to make their products more interoperable and helping them operate in a multi-vendor space and remove that 80% resource overhead.
All the UK Agritech Centres are champions for FAIR data. And the UK through the Open Data Institute has taken a leadership position in advocacy for FAIR data for agriculture and food. The partnership between Hummingbird and Agrovista, combining the data capture and machine learning capabilities of Hummingbird with the powerful agronomic decision support tools of Agrovista is a prominent example of achievements using FAIR principles in data and diagnostics.
Sustainable productivity in farming is critical for the UK to remain competitive. Productivity levels in neighbouring regions show that there is a yield gap which can be bridged. Smarter data capture in the field, combined with data science, and artificial intelligence has created a data and diagnostics pipeline, to build better decision support tools for farmers. With more accurate and timely data, growers can spend less time out in the field assessing and diagnosing conditions, and more time working on solutions. Automation will be enabled by the data and diagnostics pipeline, allowing farmers to scale their operations and be more efficient, but still input the human instincts and decision-making to run a modern farm.