From the opening presentation at the hackathon, we learned a lot about the problems that a modern farmer operating a highly data-driven farm faces. We found it interesting that farmers are mindful about their contribution towards climate change, but often do not have the necessary tools to keep track of how sustainable their farming practices are.
We've entered a new decade with unprecedented challenges posed by climate change, and it's imperative to venture into new industries where there is a desire to be more conscious about individual and organizational contributions. Farmprint is a first step towards harnessing data and technology to analyze and reduce the carbon footprint of large farms.
What it does
Farmprint is a tool that allows farmers to access information about their farm's entire carbon footprint, in order to understand how to make their farms more sustainable. Farmprint analyzes data from farm equipment, ranging from electronics in barns to combines and harvesters in the field to get energy consumption data and fuel expended. Farmprint visualizes KPIs from a sustainability point of view. Further, farmers can compare how sustainable their farm's areas compared to others and global benchmarks. In addition, the dashboard also provides recommendations for reducing the farms' carbon footprint, by assessing equipment efficiency.
How we built it
Farmprint relies on a React.js frontend, and a Flask (a Python web micro-framework) backend, hosted on an AWS EC2 instance. We did a lot of data processing in pandas through jupyter notebooks, which have all been included in the GitHub repository.
Challenges we ran into
With the large amounts of data available for each equipment on the farm, it was difficult doing real-time processing, and even offline processing on historical data due to its sheer size. We were able to extract data from a couple of the available datasets. Another challenge was the lack of time, as this was a 24-hour hackathon, but we're happy with how we've executed our idea. Further, we were also not able to verify all our results against existing benchmarks, but with more time we would probably make sure that the values we obtain from our data processing make sense.
Accomplishments that we're proud of
With any web application, it's nice to see a frontend and backend, built separately, coming together and communicating and visualizing data. We were able to develop scripts to calculate overlaps in the paths of combines operating in the farm, and improving the efficiency and path planning for combines is integral to reducing the carbon footprint of the farm.
What we learned
On a qualitative level, it was really interesting to explore challenges that modern, data-driven farms face, and the size of the landscape an organization such as AGCO operates across, with so many diverse problems. It was interesting to see real-time raw CAN-style data from farm equipment, and finding tangible insight from such granular data was a challenge that we all enjoyed. We also got to learn about deploying applications to AWS, which turned out to be fairly simple, opening up opportunities for us to explore more advanced AWS solutions in the future.
What's next for Farmprint
There was a lot of farm equipment as data sources that we weren't able to integrate into our platform, and processing and including their data in calculating the carbon footprint of the farm. Furthermore, we also are keen to learn more about what KPIs that matter to farmers, and work towards finding the optimal sensors and data points that would allow us to create valuable insight.
Additionally, there's a lot of work still to be done on our recommendation system for how the carbon footprint of a farm can be reduced by replacing faulty equipment or optimizing the paths of combines. We found this problem to be very interesting and we definitely want to explore that in the future.