Refining RuFaS: Predictions are only as good as the data behind them

RuFaS is referred to as a prototype of a dairy farm “digital twin.” What does that mean?

RuFaS is a digital replica of a physical farm that allows the user to simulate and optimize various scenarios. Strictly speaking, it doesn’t fully qualify as a digital twin yet because that requires automated data streams between RuFaS and farms. We’re actively working toward that end.

However, farmers and advisors currently can ask RuFaS “what-if” questions, like: “What happens to milk output and methane emissions if I change my herd’s diet?” Or, “What if I switch from a storage tank to an anaerobic digester and lagoon?”  Unlike most carbon calculators, RuFaS doesn’t just give environmental numbers; it also predicts milk yields and economic outcomes. It helps farmers see the bigger picture tailored to their own farms.

You’ve been focusing on two of RuFaS’s modules: the animal module and the manure module. Why are these important?

Together, these two modules drive many of the farm’s emissions. One part of my work compares RuFaS with COMET-Farm, a greenhouse gas accounting tool often used by researchers and policymakers. By benchmarking the two tools side by side with real farm data, we can see not only whether RuFaS’s results line up but also give farmers and researchers a fair, objective way to interpret outputs from both tools.

I’ve also worked on refining how RuFaS predicts milk production. This makes herd-level milk predictions more accurate and gives farmers confidence that the model is realistic when testing new strategies.

And then there’s the piece of my research that focuses on ammonia emissions from barns. Ammonia harms both people and animals when it builds up indoors. It also contributes to nitrous oxide, which is a potent greenhouse gas, and it represents a loss of valuable nitrogen that farmers could otherwise use as fertilizer. 

Right now, RuFaS predicts emissions using a few key factors: how much manure is in the barn, how much ammonium nitrogen it contains and the temperature. But predictions are only as good as the data behind them, and the problem is that very limited barn-level data exists. 

How has working closely with the Cornell Agricultural Systems Testbed and Demonstration Site (CAST) for the Farm of the Future helped you with your research?

CAST focuses on technology and practices that can help farmers make informed decisions. Part of that is the integration and analysis of data from many diverse equipment and software systems that have been set up at the Cornell University Ruminant Center (CURC).

The RuFaS ammonia project is a direct partnership with CAST. I’ve worked with Martin Perez [CAST program and operations manager] and Jason Oliver [senior extension associate and dairy environmental systems engineer with PRO-DAIRY] to set up a system of gas sensors and wind-speed meters inside the CURC barns. These let us measure ammonia concentrations in real time and track how the air moves through the barn. We also pair this with lab measurements of the manure itself.

CAST and RuFaS are a natural fit because RuFaS needs good data on everything from feed rations to manure handling to give realistic outputs, and CAST is working on integrating those streams of farm data into one place. When you combine the two, you get a system where data flows seamlessly into the model and comes back out as clear, farm-specific insights.

How will the emission data you gather at CURC ultimately benefit farmers?

We will use this real-world data to tune the RuFaS model so its predictions better match what actually happens. If we succeed, farmers will have a more reliable tool for estimating ammonia losses and planning how much fertilizer they really need to apply, saving money and reducing environmental impact at the same time.

Looking forward, what are the next steps? 

I’ve already shared parts of this work with the scientific community. I presented a sensitivity analysis of RuFaS’s current ammonia predictions at a precision livestock farming workshop in Brazil in October 2024. Then I also presented a framework for refining those predictions at the American Dairy Science Association annual meeting in Kentucky this past June. The next step is to finish collecting and analyzing ammonia data at the CURC farm. With those results, we’ll refine the model and share findings that help both researchers and farmers. 

Jackie Swift is the communications specialist for the Cornell CALS Department of Animal Science and the communications manager for CAST.