SDS 264: Data Science in Agriculture

Welcome back to the SuperDataScience podcast ladies and gentlemen, today we’re talking about data science in agriculture. This is part of our series of episodes about data science in various industries. If you’ve missed the previous ones, you can always find them at www.www.superdatascience.com/podcast. We’ve covered off quite a few interesting examples in a multitude of different industries. And today we’re moving on to agriculture.

So what is agriculture? Well, agriculture is defined as the science and art of cultivating on soil and the rearing of livestock. It is the key development in the rise of the human civilization. And by the way, if you want to learn more about how it’s actually contributed so much to the rise of our civilization, a great book to check out is Sapiens by Yuval Noah Harari. Amazing book and he spends quite a bit of time describing the whole process of domestication of animals and crops and how, what role that played in our development. So if you’re interested, check that out.

But other than that, right now, agriculture, it has grown a lot. In 2014 the revenue in the agriculture industry was 2.400 billion. So that’s 2.4 trillion with a T US dollars and agriculture employs over a billion people worldwide. So as you can imagine, some massive, massive sector of the economy. The percentage of the total GDP that’s attributed to agriculture is 5.9% globally. So almost 6%. It’s higher in developing countries, about 17.4% in India for example. And it’s lower in developed countries. So it’s only about 1% in the EU and US.

And today as usually we’re going to look at 5 use cases of data science and artificial intelligence in agriculture. So get your seat belts buckled up. And lets head off straight into it.

Use case number one, precision agriculture. So this term of precision agriculture was actually coined in the 1980s when gps made it possible to do geo referencing of data. And that allowed more data driven approach to agriculture. And in the modern day, we can take this even further with data science, by applying certain models and learning from past experiments or past experiences in order to inform, for instance, how to distribute fertilizer or pesticides or other additives according to the needs of all plants, soil, weather conditions and so on based on a geo referenced data point.

And the example here is PrecisionAg as a company that has developed a standardized system that they offer to small farmers all over the world. They are reporting 10% increases in yields for various crops with up to 30% increases for more demanding crops. So as you can imagine, that can be a huge uplift, especially if you’re talking about a large quantities of crops and large volumes of output, goods that are going to be produced by those farmers.

Use case number two, biogenetics. So genetics is the study of the effects of genetic variation to encourage valuable traits and crop plants and farm animals. And interestingly enough, genetics or biogenetics isn’t something that’s absolutely new. This process of encouraging favorable variations and mutations in animals and plants has been used for over 10,000 years. Ever since the first farmer chose which seeds to plant, which seeds to eat, which animals to continue breeding, which animals not to continue breeding. This process has actually been going on for ages. And now with the help of data science and more advanced genetics, we can take this to the next level.

And the example here is that back in 2000, Swiss scientists produced a breed of rice that also contains vitamin A. And this is important because the lack of vitamin A causes about 250 to 500,000, so that’s half a million children, to go blind every year. Just because they don’t have sufficient vitamin A in their diet. And therefore this breed of rice is a cheap alternative to vitamin A supplements in developing countries. And that can be a huge lifesaver. So, again, here we can use data science and data analytics to further inform and enhance certain biogenetics that need to take place.

Use case number three, robotic weed control. This is a very interesting one. So controlling weeds is a very important priority for farmers because there’s about 250 weeds species which are now resistant to herbicides. And that causes farmers in the US alone to report annual losses of about 43 billion US dollars due to weeds. So you imagine it’s a massive opportunity, like the world’s biggest problems are the world’s biggest opportunities as Peter Diamandis says.

And here you got $43 billion being lost due to weeds. And that’s despite the fact that over 400,000 tons of herbicides are used every year. So now companies are jumping onto this opportunity and developing automated solutions to protect crops such as robots or automated machines that move through the field. And they can actually pull out these weeds or identify and spray these weeds individually through technology such as computer vision.

And the example here is Blue River Technology claims that its robot called See & Spray eliminates 80% of the volume of chemicals normally sprayed on crops and can reduce herbicide expenditure by 90%. So, that’s a massive reduction, 80 and 90%. And it’s just using more advanced technology than was used before. And that is possible due to how far, how far we’ve advanced. So yeah, opportunity’s huge. 43 billion US dollars, no wonder companies are jumping onto this and leveraging the power of technology rather than chemicals in order to accomplish the same goals.

Use case number four, automatic crop harvesting. So even though we think that harvesting is a process done mostly by machines or huge massive machines, a lot of it is actually still done by hand. For example, in Florida alone 10,000 – 11,000 acres of strawberries are picked by hand every year. And at the same time it is predicted that the number of agriculture workers will reduce by 6% by 2024. So farmers are increasingly looking at robots that can perform harvesting work instead.

And the example here is that Harvest Croo robotics has developed a robot that can harvest eight acres per day. That’s the equivalent of 30 human workers. So another part of this industry that can still be further disrupted by automation and that is harvesting.

And finally use case number five, our favorite, weather prediction. So data science has been used to predict weather for a long time but accurate weather predictions are more important in agriculture then pretty much in any other industry. So that allows us to know when to plant, fertilize, dust, harvest. When to expect hail when to expect drought and other things. And all those are crucial for food security. Weather forecasting is a strong fit for machine learning.

And the example here is that IBM has actually been involved in weather predictions since 1996 but only recently have they purchased the weather company. And what this allows them is to combine the computing capabilities of IBM Watson with over a hundred terabytes of third party data every day to provide extremely accurate weather predictions down to resolution 300 meters. So again, a massive companies jumping onto this because it’s a huge opportunity. Whenever we see an industry of this size with like for instance 2.4 trillion US dollars in revenue, and that was back in 2014, so that is most likely grown even greater from then. Whenever we see an industry that size, any opportunity, because as long as you can scale it to multiple customers, multiple businesses, multiple regions, any opportunity becomes very lucrative. And that’s why even companies the size of IBM are jumping onto these opportunities to seize them while they are still available, while they’re still fresh and ready for disruption.

So there we go. That’s data science and AI in agriculture. Hope you enjoyed this podcast and if you are in agriculture then hopefully this gave you some ideas. If you’re not, then it’s still a great way to pick up some ideas for your industry because data science skills are so transferable. As usual, you can find all the show notes at www.www.superdatascience.com/264 and there you can also find the transcript for this episode and any links that we used in this research. And on that note, I look forward to seeing you back here next time. And until then, happy analyzing.