Applying Information Science to agriculture

Sunset or sunrise, traditional or industrialised, niche or commodity? There sometimes appear to be as many opinions and forecasts for agriculture as there are commentators.

When my colleagues and I started Rezare Systems in 2004 we made the decision that we would focus on agriculture despite the forecasts, and on ways we can contribute to making farming systems smarter, more responsive and adaptable (whether that’s to market demands or environmental needs).

Our key lever in that commitment has been the way we apply information science and computer science to agriculture.

Information Science is a broad term that includes everything from libraries and classification techniques to information protection. We like it that information science challenges us to think about the organisation of information (hence our commitment to things like the Farm Data Standards), as well as the information lifecycle – capture and generation, storage, transformation, packaging and repackaging, protection, communication and presentation.

Computer Science is likewise broad, extending from the esoteric mathematics behind modern encryption and 4G communications protocols to computer programming, artificial intelligence, and the use of computational principles to manage and analyse big data sets.

We have great respect for those in the pure science end of these large domains. While we follow their advances, we work very much at the applied science end. We look to use the components and learnings of computer and information science to “move the needle” in agricultural forecasting, management, and product development.

Someone recently asked me what we look for when we recruit for our growing team. I’ve already written about some of the practical skills we look for such as communication and teamwork, but what of computer science and information science background?

Converting concepts to code

Much of the work we do involves turning conceptual models – how plants take water from the soil or how variation in a mob of animals changes as they grow – and converting this into a software algorithm. At first glance you might think this is just mathematics (and if it’s written in a scientific paper it may indeed be complete with formulae and symbols), but it requires understanding the context and concepts to be represented.

Our business analysts and many of our collaboration partners are great at explaining this sort of thing, but some of our best developers and modellers are also able to read through the literature and make the cognitive leaps themselves.

Comprehending correlation vs. causation

You’ve probably seen the wonderful charts that show how murder rates go up and down in line with ice-cream consumption, or the divorce rate in Maine correlates with per-capita consumption of margarine.

Our experience is more nuanced, but when building models and using principle component analysis, machine learning and pattern recognition tools, we value the ability to understand the weaknesses of our methods, question the data, and look for research into the underlying biological or physical processes – so we know if we are seeing something real, or a correlation that will vanish as soon as we use it in a commercial application.

Finishing in a reasonable time

Initial or “naïve” implementations of an algorithm or statistical process often produce the expected output, but can take a long time to run, or consume a lot of memory or processor time. As the size of data sets increases, limitations of the simplistic approach appear: response times become slow (or approach infinity), and servers can become unresponsive because of locks or lack of resources.

We employ computer science techniques to address this, including caching partly calculated results, splitting the task and using multi-threading, or employing more complex but memory and time efficient algorithms. Sometimes we can separate the process of analysing and calculating data from reporting, computing results in the background so they are ready when a user wants to look at them.

Our contribution to the planet may not be “rocket science”, but we take pride in applying smart techniques from computer and information science in support of more productive, sustainable, and profitable farming systems. Want to know more? Get in touch and let us know what you think.