I was explaining my work in agriculture to my kids the other day. I pointed to our fish tank and asked, “What would happen if we never cleaned the fish poop out of the tank? It would become fertilizer for algae, which would take over the tank, and the fish would die.” My kids looked at me wide-eyed, alarmed. Something similar is happening on Earth today: fertilizer practices in agriculture are leading to life-altering chemical changes to our planet. The good news is that there are now digital tools that can help reduce fertilizer loss into the environment by increasing fertilizer use efficiency. However, after spending 10 years implementing these tools, it has been more challenging than I anticipated because of what I call “the fertilizer paradox”: the agrochemical industry, which is most capable of building and implementing these solutions, is not financially incentivized to get them adopted. In this post I hope to explain my view of the current state of technologies, raise awareness of the challenges, and build support for adoption of these tools.
The human population is ten times larger than it was at the start of the industrial revolution, and the world’s farms have grown to meet the demand for food. The nutrients that these crops need have long outstripped the natural supplies provided by animal manure and other natural processes. This is why artificial fertilizers were developed. Without artificial fertilizer, farming practices today would struggle to feed even half of the world’s population. But recent scientific reviews suggest that the chemical changes resulting from global use of artificial fertilizer may be more significant than the changes in the carbon cycle. We have more than doubled the abundance of plant nutrients on Earth’s surface from pre-1900 levels. Those chemicals are highly reactive; they include many of the chemicals used for making bombs. These fertilizers run into drinking water, and when water in farming areas cannot be treated, fertilizers can outcompete oxygen in blood, sometimes causing a deadly blood disorder called “blue baby syndrome”. Fertilizer dramatically improves plant growth in fields, but they also lead to algae blooms in lakes, the Gulf of Mexico, and thousands of other bodies of water around the world. Those algae blooms create dead zones where animals cannot live. The earth is now like my kids’ fish tank. We are adding nutrients faster than they are being used by natural processes with no end in sight.
The “fertilizer challenge” is well known in the environmental science community and has been listed as a ”grand challenge” by the National Academy of Engineers. One engineering solution is to simply be smarter and more precise about how and when fertilizer is applied.
The problem is not with farmers. The farmers I know are dedicated to being good stewards of their land and have often managed it for generations. When surveyed, farmers say that one of their biggest motivations is to leave their land in good shape for future generations. That said, they also need to run their operations to make money, which isn’t easy. In today’s economy, farmers can easily work a full year and not return a profit. In order to survive, farmers have been forced to consolidate and scale up. The word “acre” originally defined an area that one farmer could till in one day, but it is now common for family farms to operate thousands of acres. Modern equipment allows a farmer to complete complex operations such as planting and fertilizing at more than 20 acres per hour.
To manage logistics, farmers often apply the same amount of fertilizer to every acre, every year. Despite fertilizer’s hefty expense, the hedge of applying a generous amount everywhere has long been viewed as worth the cost, especially since previously available tools, like those provided by university systems, have not been able to predict variable needs. But by applying the same amount everywhere, every year, farmers waste money and inadvertently harm the environment when excess fertilizer is applied. Farmers know there are large variations in soil quality: some acres have deep soil that is abundantly rich in nutrients, while in other areas, like on a sandy knoll, the soil’s ability to provide or even hold fertilizer has completely washed away. Optimal fertilizer amounts often vary by a factor of two within the same field. Weather effects can change the optimal fertilizer for a given year by a factor of three: some years have heavy rains early in the season that wash away fertilizer, while other years have moderate weather, where the supply of nutrients naturally released by the soil perfectly matches the crops’ need for food. The advent of digital tools that allow farmers to calculate and monitor their need for fertilizer can make them more profitable while also helping them to be better stewards of their land.
I co-founded a company called Solum in 2009 to bring digital solutions to agriculture. Back then, venture capital investments in agriculture were few and far between, but since then the so-called “AgTech” industry was born. The widespread adoption of the iPad among farmers, along with GPS technology, has made it easier to control planting and other operations in a field, on a meter-by-meter basis. Digital devices have generated a dizzying array of data from farm machinery. Farmers also now have access to satellite images, drones, weather stations, advanced crop models, and more. Many tools were created for digital agriculture, and big companies invested heavily to help them grow. Granular built successful planning and financial support tools and was later acquired by Dupont, where it expanded into fertility management. Solum’s software offering, along with its soil measurement tools and operations, was acquired by The Climate Corporation, which had recently been acquired by Monsanto and is now owned by Bayer. I joined Climate to apply data science to the fertilizer challenge and worked there for four years.
In order to predict optimal fertilizer application for a given plot of land, data scientists need training sets, or “ground truth” of known optimal fertilizer amounts across various weather, soil and cropping environment. Acquiring such data requires field trials where varying levels of fertilizer are applied and crop yield is measured. While such datasets are growing, they are still too small for robust predictions using simple forms of machine learning algorithms. Data scientists bring in other forms of data, including national soil surveys, local soil measurements, precision maps of elevation data, public weather models, and laboratory results. By skillfully leveraging models of the physical environment with advanced machine learning approaches, researchers have made substantial improvements in their ability to predict optimal fertilizer rates that maximize profit with high yields.
Applying data science to fertilizer decisions requires interdisciplinary teams with specialized talents. My colleagues at The Climate Corporation included experts in soil processes, plant modeling, statistics, and applied math. Specialized teams handled the data, and software engineers implemented models that work seamlessly with a smooth customer interface. The company has developed new soil measurement tools and new weather models, and it operates research farms around the United States. A network of research collaborators in academic institutions and private partnerships make the development process even more robust. Over my four years at The Climate Corporation, my colleagues and I found we could write software that would help farmers spend less money and get the same reward in terms of crop yield (Our study with Environmental Defense Fund), and sometimes we even helped them increase yields, growing much more food. Those results were produced with a small subset of the most advanced farmers. Because well designed software can be easy to use, tools like these can now be extended to a much broader set of less advanced farmers worldwide.
Although new software solutions for precision agriculture appear promising, there are major challenges that limit adoption. Saving farmers money seems like an easy proposition, but incentives are not always aligned. Digital farming advice is primarily sold through fertilizer dealers, many of whom provide digital tools for free or reduced cost in order to gain a better relationship with their farmer-customers. These relationships help them to sell more physical products, such as fertilizer. But when digital tools recommend selling less fertilizer, there is conflict with their business. Although they do not invest heavily in figuring out the science of how and when fertilizer should be applied, these fertilizer sellers do own the relationship with the farmers. The fertilizer sellers often even apply the fertilizer so that their farmers can focus on other parts of their business. The misaligned incentives of fertilizer sellers have made the pace of innovation for new digital fertility tools slow.
There are reasons for hope. The relatively new ability to plan out fertilizer rates and measure a crop’s yield on a meter-by-meter basis provides a new level of transparency in results. Farmers no longer need to trust the people who sell them products: they can program their tractors to place two fertilizer amounts side-by-side and observe which one results in the greatest profit. Sellers that don’t provide honest advice will eventually lose business. Actions that make it easier to test and interpret such results will benefit both farmers and the environment.
External incentives could break down misalignments and speed up adoption of smarter, data-driven fertilizer choices. In the United States, we spend more than $20 billion each year on direct farm subsidies. Farmers who receive their full portion of such subsidies could be required calculate how much fertilizer they need on each acre of their land. Such calculations would allow farmers to increase profit by managing the specific needs of each acre instead of the average needs of a multi-acre operation, or instead of a rate that was determined with no calculation at all. Good calculations would also demonstrate the benefits of fertilizer applications at different times or with different equipment, and reveal benefits of new practices like planting cover crops that retain fertilizer and keep the soil healthy all year. If farmers don’t like the results of an approved calculation method, they could choose their own rate but they should be required to test that rate on a small fraction of their land, so that they can learn from their choices. The problem with misuse of fertilizer is a largely invisible problem. There is limited public discussion about the challenge and, unlike issues with weeds or disease, farmers cannot visibly see a problem when fertilizer is over applied. Requiring calculations and testing of new rates would make the problem with fertilizer visible and would enable valuable solutions to the problem along the way.
The world needs fertilizer to feed the human population, but we also need to protect our local water supplies and the environment as a whole. Smarter fertilizer decisions will dramatically improve the situation, but widespread adoption of such tools may require new incentives. Incentives that require farmers to calculate fertilizer application rates would not only help farmers and their trusted advisors become more efficient, they would also improve the environment for the remaining 98% of the US population that do not work in agriculture. And the earth won’t end up like my kids’ poopy fish tank.