AI to Reduce the Gulf Dead Zone
Existing ag decision making tools won’t cut it. Future H2O will work on new analytics and mesoscale weather prediction capabilities.
Every year, from May to September, toxic nutrients from fertilizer enter the Mississippi River watershed and wash downstream to the Gulf of Mexico, feeding the second largest hypoxic (or “dead”) zone in the world. More than 235,000 tons of fish and other marine life is lost to hypoxia in the Gulf annually.
In short, the Gulf dead zone is a major problem — for marine health, human health, and local fishing economies. And climate change is making these dead zones worse. Increasing, rapid bursts of rainfall, especially after long periods of drought, are pushing even more nitrates from fertilizer into the Mississippi. Rather than reaching the soil, the nutrients farmers’ crops need get washed into waterways and ultimately the Gulf.
In an era of increasing climate extremes, current fertilizing practices and decision making tools won’t cut it anymore — for resilient crop yields or tackling our dead zone problem. They’ll only exacerbate the impacts of climate change on food production and watershed and ocean health.
Existing analytics and weather prediction tools aren’t sufficient
But right now we don’t have the analytics or weather prediction capabilities needed to help farmers and the agricultural industry make more informed, long-term business decisions that not only have the potential to improve crop yield, but mitigate the impacts of climate change, too. While we have the knowledge and tools necessary to make daily to weekly weather predictions as well as the skill to create robust climate projections at the 10 and 100-year scale, we need to be able to predict weather at the mesoscale — two to six weeks out — in order to time farm-scale applications of fertilizer so that they benefit crops instead of running off and creating dead zones. And we don’t yet have the resources anywhere in the world to do that.
The exciting news: Future H2O is working on it.
The power of creating mesoscale prediction tools and analytics around crop yield
Future H2O just received a two-year planning grant from the National Science Foundation’s Computer Science and Engineering division (CISE) focused on AI applications for the agricultural industry. Starting in hot-spot regions like Iowa or Indiana, where we know there are high contributions of fertilizer runoff, we plan to develop an early warning system for the Mississippi River Basin that relies on predicting weather two to six weeks out. This will empower us with the data needed, and in the timeframe needed, to alter farming operations and fertilizing schedules in a way that (1) potentially increases crop yield and profitability for farmers, and (2) reduces fertilizer runoff into waterways, and ultimately the Gulf, following rainfall.
In an era of increasing climate extremes, current fertilizing practices and decision making tools won’t cut it anymore — for resilient crop yields or tackling our dead zone problem.
— John Sabo
To do this, we’ll create mesoscale prediction tools and generate data on climate impacts related to water quality, waterways, and the Gulf dead zone. We’ll also generate analytics around yield differentials, assessing the yield benefits of delaying fertilizer application for various crops — because without these analytics, the farmer won’t be empowered to make optimized decisions about fertilizer timing and amount.
Our research might be able to demonstrate that by holding off on fertilizing until upcoming rain passes the crops will have a better chance of uptake, and therefore farmers will have a more profitable crop yield as a result of delaying.
(Help refine our work by taking this short survey on public views of agriculture and AI issues. The study is being conducted by researchers at the University of Oklahoma, a partner in the study.)
Using the data to incentivize changes in farming operations
But we know the technology and data alone are not enough. The real solution will be in how we’re able to communicate the data and use it to incentivize changes in farmers’ day-to-day operations.
For example: If we discover a positive correlation between delaying fertilizer application and crop yield, we’ll need to figure out how to successfully get the word out to farmers. Perhaps the findings will be incentive enough and the focus will be on how we regularly communicate mesoscale weather predictions.
We need to be able to predict weather at the mesoscale — two to six weeks out — in order to time farm-scale applications of fertilizer so that they benefit crops instead of running off and creating dead zones.
— John Sabo
Or if a positive correlation between fertilizer application delay and yield is not as clear, companies like Cargill, Monsanto, DuPont, or any organization with part of its value chain in the Mississippi River Basin (e.g., WalMart) might look at their enterprise-wide supplier list in the basin and say, “Hey, we’re going to set these incentives (like a payment) for you based on the weather data that’s coming in,” giving these companies the power to influence what’s going on at the supply chain scale. What’s in it for these big companies? Ultimately, they know data can guide them to better efficiency and higher profitability. And it could be a powerful and innovative way to meet ESG (Environmental, Social, Governance) standards by setting targets for their suppliers to provide their goods in the most sustainable way.
And if companies aren’t interested in dangling the carrot, states could empower a regulatory arm of their environmental quality agencies to issue penalties or fines if farmers don’t comply with recommendations to delay fertilizing.
In an era of increasing extremes from climate change, there isn’t anything that’s just about “water” or “pollution” or “floods” or “drought.” All factors are integrated and often reinforcing. And our science, data integration and decision-making tools need to reflect and stay ahead of this new reality.