The farming toolkit is changing, with new computational techniques vying for a place in decision-making processes. Teresa Rush reports.
Matthew Smith is a computational ecologist with Microsoft Research, based in Cambridge. Much of his work is focused on ‘geotemporal analytics’ - analysis of data with a geographical and time component.
In a farming context this is leading him into areas such as the development of predictive models for minimising waste and improving supply scheduling in the fresh produce sector, using data generated on-farm.
He is interested in how new information useful to farm businesses can be generated from data already being produced, using ‘agricultural intelligence’.
He says: “It does really feel like we are on the cusp of some great new agricultural revolution, fuelled by agricultural intelligence – this ability to make more intelligent decisions on the back of being able to join up domain understanding, data and methods.”
And he is in no doubt this agricultural intelligence revolution will happen.
“Whether it happens next year or in 10 year’s time, inevitably it will happen and I think we all sense that,” he says. From its UK base in East Anglia’s ‘Silicon Fen’ Microsoft Research has been working with business-focused cluster organisation Agri-Tech East to identify businesses with which the software giant might work to test its agricultural intelligence capabilities.
It is currently involved in a project with Cambridgeshire- based G’s, one of the UK’s largest salad and vegetable producers.
“We have been working to reduce waste and predict supply, using geotemporal analysis, to see where we can make efficiency savings,” says Dr Smith. He was recently asked by Agri-Tech East to share his thoughts on where the software industry might have an impact on agri-tech.
Interestingly, despite high level concern over future global food supplies, the software corporations are not yet seeing agri-tech as a big area to move into, he adds
“What they want to do more is to be a platform to provide the pieces which enable information to be joined up, predictive analytics to be done.” He sees this technology developing into more ‘personalisable’ productivity software.
“Farm management systems which mean you can see much better the various key performance indicators in relation to your farm.
“Whether that is viewing your profit over time, whether it is seeing how your inputs contribute to your outputs, giving a systemic perspective, it is about being able to join up all that information within your software.
“We’ve got advances in mapping, in sensing, in data analytics and in visualisation and all these things are being made available individually, they are there to be brought together.
“There is going to be a lot of trial and error in this arena of making more personalisable software but I think in five to 10 years time, what you use will be quite different from what it is now; you will be able to see your yields on a map; you will be able to relate the information you have on a particular field with other geotemporal information, such as weather, soil properties, disease risk.
“I think we will undoubtedly get developments in the ability to join up data more effectively.
“Right now there is a huge range of individual data sets and it is very difficult to pull them all together. There is going to be value in making these easier to pull together. The big software companies are adopting standards and there is also going to be innovation in algorithms and code libraries which allow you to convert data in any one format to any other format.”
A third area with relevance to agriculture is predictive analytics - the branch of data mining concerned with the prediction of future probabilities and trends “I think it is going to be the key enabling technology to enable us to make more intelligent decisions,” says Dr Smith.
Developments in modelling and forecasting might help growers link information on crop maturity or crop health, for example, and use this information to predict crop supply. In turn, combining supply and demand predictions could help reduce crop wastage more effectively than is currently possible.
For those with a deep conviction farming is all about being out in the field, the very mention of ‘algorithms’ is enough to arouse at least a degree of wariness. But according to Dr Smith, algorithms will be tools of future farming.
They will be tailored to particular farming scenarios – perhaps being used to compare the performances of similar farms or enterprises. So-called ‘recommender’ algorithms, similar to those used by e-retailers to highlight comparable or associated products, may also have a place.
“They are matching like with like – what is the farm most similar to yours going and how is it performing?”
Combining cloud technology, already in use in agriculture, with algorithms also offers possibilities for benefiting from combined data.
“How about the idea of having multiple providers providing data which is held in a cloud? Imagine these are farms providing encrypted data and you’ve got someone providing an algorithm based on that, which might be trying to explain what is behind your yield variations or estimating the supply to a supermarket within the next week.
“The decryption is only done in the cloud and no identityrevealing information is ever given back to the provider.”
The farming industry has not shied away from information technology, despite some significant practical challenges, not least poor internet connectivity in many rural areas.
And, according to Dr Smith, there is a third challenge the sector must address, that of finding people equipped with the necessary farming and IT skills to make sure these emerging technologies are put to best use.
“We really do have a skills gap in making best use of agri-tech and so I think one of the other developments we need over the next five to 10 years is a real skilling-up,” he concludes.