Transforming us into a society of food producers

Computer Science Professor Chandra Krintz recently gave a TEDx talk entitled Transforming us into a society of food producers. In her talk, which can be found here, she discusses how the same data analysis and prediction techniques used by tech companies such as Amazon and Google for targeted advertisement and marketing could be used to make farmers and ranchers more productive. The talk is based on the research she has done with other members of UCSB's RACElab, including Professor Rich Wolski and PhD students Nevena Golubovic and Varun Kulkarni. Their research is part of a $1.2 million NSF grant in collaboration with agricultural researchers at CSU Fresno and Cal Poly on research and education for sustainable agriculture practices. The work is also in collaboration with the Food, Water, Energy Nexus effort from the NIFA, NSF, USDA, and others.


Photos feature Professor Krintz on the TEDx stage and project members collecting multispectral image data (for plant health evaluations) using a drone.


We asked Professor Krintz some questions about the project.


What types of small sensors could be used by farmers and do these technologies already exist?

There is a vast diversity of sensors available to farmers and they range in price and what they do. This is where much of the precision farming advances have come from (sensing).

There are sensors on tractors that collect information about field operation and control input application (planting, harvesting, fertilizer/pesticide application). There are weather sensors on-farm and in the community that are available via the web (CIMIS, WeatherUnderground, Davis weather stations). There are soil sensors that you place in the ground to measure soil moisture and nutrient content. There are sensors for measuring the activity of most farm equipment (pumps, irrigation systems, fans, wells tanks). Multi/Hyper spectral imaging is used at various levels to record/track vegetative health of plants, soil quality, water use, disease/pest issues; taken manually by farm workers, drones, ground robots, fixed wing aircraft, and satellites. There are many manual records taken by farm workers that analyze specific plants/trees (planting/harvesting/applications history/method, pressure chambers to measure water potential of plants, etc.), and many more.


Who provides farmers with these sensors and how much do they cost?

Many different vendors provide growers with sensors and they all are different, proprietary (hardware and communications), and incompatible with each other (data access, cellular service/communications), which is part of the problem. Once a grower goes with one vendor, it's hard for that person to move their data or incorporate sensors from other vendors. In addition, vendors put up cell service (wireless is hard to come by farm-wide, especially in remote areas) independently of each other, which wastes energy, causes spectrum interference, and can be very costly.

More problematically, some vendors provide sensing at very low cost or for free in order to take data from a grower’s farm for their own benefit — vendors use the data to sell/market more products (or sell it to other vendors as a source of income). Growers, it turns out, give up ownership of this data (data about their farms) in these contracts. Vendors structure these deals/services this way because it has tremendous monetary value to vendors. Growers in many cases have no choice but give up this valuable asset because vendors require that they relinquish it in order to use their equipment, services, and products.

We have set out to address these issues. We believe industry cannot do it because they have no incentive to

1) share the data they collect

2) to NOT lock-in growers to their products

3) to make their devices compatible with others or even just work with others for competitive (IP, financial) reasons.


What is the current model for analyzing the data collected?

The model that all vendors use today is to collect all of this data and move it off-farm to the cloud (over individual cellular networks in most cases). This data can be enormous and costly to ship. We also don’t believe this model is the right one; they are moving huge amounts of data to code that processes it and serves it. It turns out that this “code” is tiny compared to the data. The model we are pursuing instead is to move the code to the data, like downloading an app from AppStore. Farmers can download an analytics or data visualization app from the Internet to their data on farm, to achieve the same analytics power and benefits that the vendors provide.​


What will your lab's role in the project be?

As the above list shows, there is a very large amount of data (retained for usefulness over many years) and it will require significant computation and storage to maintain and analyze it in ways that solve real problems for farmers. Our vision is that farmers can own their own computing resources, which our research team here at UCSB will design, and run them on-farm. We are developing a SmartFarm Appliance, like a refrigerator or Tivo box, which does complex things behind the scenes which the user doesn't need to worry about.

Our early SmartFarm Appliance prototype is a small distributed system (3ft x 3ft crate) that leverages recent advances in small, cheap commodity computers and cloud computing to produce an appliance that provides “lights-out” maintenance (i.e. it can run on its own, in harsh environments, tolerate failures, and not require system administration, for over a year). A repair man model will be used to have someone come in infrequently, and upgrade/fix/replace the appliance every other year or so. Otherwise the Appliance does everything else on its own to keep running and performing analytics and providing decision support to growers. Our prototype currently costs just under $5000 but we believe it will come down into the $3000 range in the next few years. It consists of multiple mini-PCs with multiple processors each, a local ethernet network, and a wireless tower for local (on-farm only) wireless communications to the sensors.

The SmartFarm Appliance can be attached to the Internet or not — if the farm is not connected, it will still work and rely on default analytics apps or those we send them to growers via USB keys. If the Appliance is attached to the Internet, analytics apps can be download from an AppStore as described above.

Our computer science research studies the various systems aspects related to making all of this happen in a computer/analytics system that is easy for growers to use.
We will also study analytics applications that solve problems and provide recommendations and tracking to help growers.

All of the software we develop is free and open source — anyone can contribute systems components, tools,  and analytics apps. By going open source, we are hoping to spur innovation and contribution of analytics apps by a broad technical community (a community we are developing as part of this project) so that more people can contribute to helping growers solve their problems, use resources more efficiently, increase their yields, and more sustainably feed the planet.


You are collaborating with agricultural science researchers, what is their role in the project?

Yes, we collaborate with these experts on the SmartFarm analytics. Our collaborators inform our research in terms of problems and solutions particular to different growers, properties, plants, and animals. They have tremendous experience in precision farming, plant phenology, sensor development, and geo-spatial analysis that they will bring to bear on this project to produce novel analytics applications that help growers extract actionable insights from their data.


What are the greatest challenges you have seen or expect to see while working on this project?

Our work requires close relationships with growers which take time to develop. Moreover, we no longer just sit in our labs/cubicles — this type of research requires that we get out in the field — literally!  Finally, this work is tremendously cross-disciplinary, requiring all of us to learn each  others “languages” and methods of pursuing research and forwarding innovation.  There is a learning curve that I’m sure is similar to other cross-disciplinary endeavors. However, we believe that doing overcoming these challenges will help us (as a community of scientists) solve a very challenging problem, making farms more efficient so that we are able to feed the planet as it grows in population. We also believe that to enable this we also have to teach and engage a new population of students and researchers so that they also have the tools, knowledge, and interest to contribute to this important problem. We believe that this project will lays the foundation for us to enable this.