In spring 2020, U.S. dairy producers were forced to dump millions of pounds of milk when the system around them failed. Schools closed, institutional buyers disappeared, processing plants couldn’t pivot and the disconnect between production and demand became painfully clear. That experience raised a critical question: Could similar system-wide disruptions happen again, driven not by markets but by disease?
That question helped drive a new proof-of-concept project from the Western Institute for Food Safety and Security (WIFSS) at UC Davis: a simulation model designed to examine what happens when H5N1 highly pathogenic avian influenza (HPAI) enters a dairy herd.
“This was more a proof of concept,” says David Goldenberg, food safety and security training coordinator for WIFSS at UC Davis. “Can we develop a model that would mimic a dairy farm and the resulting impacts [HPAI] would have not only on the farm but also elsewhere and down the road?”
Rather than attempting to predict the next outbreak, their goal was to understand what an outbreak would look like on a single dairy and how its impacts unfold over time.
What the Model Simulated
The team based their simulation on a small dairy herd of roughly 260 cows with the following assumptions:
- No animals were purchased from outside sources; replacements were born into the herd
- Labor, equipment and milking infrastructure functioned normally
- Cows were assumed healthy apart from H5N1 infection
- Milk from infected cows was discarded
Further, reinfections were not modeled, and the analysis focused on acute infection rather than chronic disease.
Speed of Spread Mattered More than Severity
One of the clearest lessons from the simulation was how fast H5N1 spreads through a herd might matter more than how sick individual cows appear.
The model evaluated low, medium and high infectivity scenarios. In high-infectivity cases, nearly the entire herd became sick within about 30 days. That rapid clustering overwhelmed treatment capacity, increasing the risk of dehydration, delayed care and mortality. This wasn’t because the disease was more severe but because too many animals required attention at once.
“That’s a tremendous effort to simultaneously try to treat every cow in your herd at the same exact time due to limited resources,” says Nelson Alfaro Rivas, simulation consultant with MOSIMTEC. “Unfortunately, some of the cows might succumb just because of dehydration from the disease just because you don’t have an unlimited number of veterinarians to try and hydrate the cows as they’re sick.”
In lower-infectivity scenarios, illness spread more slowly, peaking later and involving fewer animals simultaneously. The contrast underscored why early isolation, movement control and disease recognition can fundamentally change outcomes.
Milk Loss Didn’t End When Cows Recovered
Even when cows clinically recovered, milk production did not bounce back quickly.
The simulation assumed infected cows experienced either a 15% or 30% reduction in milk yield for the remainder of their lactation, figures drawn from field observations. In high-infectivity, worst-case scenarios, total milk production across the herd fell sharply within the first month.
Over time, those losses accumulated. In the most severe scenarios, the herd produced approximately 25% less milk over the modeled period compared with an uninfected baseline. Perhaps more striking, herd-level production did not return to baseline for almost a year, long after the active outbreak had resolved.
“How long does it take to recover from something like this?” Rivas asks. “All the cows were recovered by day 26, but what you don’t really see is that the herd that got infected doesn’t really recover and produce the same amount of milk as the non-infected herd until almost 300 days later.”
This gap matters not only for producers but also reframes recovery as an extended process rather than a clinical endpoint.
Recovery Didn’t Mean Economic Recovery
Because dairies run on thin margins, sustained milk loss drove decisions beyond treatment and recovery. Cows producing well below expectation after infection were more likely to be removed from the herd, even if they survived the disease itself.
“At the end of the day, farms are businesses, and you can’t keep an underproducing and therefore unprofitable cow,” Rivas says.
The model reinforced a familiar reality: Profitability, not survival alone, determines herd composition after disease events.
Where Biosecurity Fits
Biosecurity practices were not explicitly modeled as individual actions. Instead, their effects were represented indirectly through changes in infectivity. Lower infectivity scenarios approximated the benefit of practices such as isolating sick cows, cleaning equipment and controlling farm access.
Those measures come with costs — labor, time and disruption — but the simulation showed even modest reductions in spread speed dramatically altered outcomes. The model did not attempt to assign dollar values to biosecurity steps, but it made clear why reducing infectivity yields outsized returns.
Why This Matters Now
What this model ultimately provides is a clearer sense of risk timing, not new disease facts. By compressing complex outbreak dynamics into a single on-farm view, it shows how quickly routine management assumptions can be tested once disease pressure rises, particularly when multiple animals require attention at the same time. The practical consequence is that delays in recognition or response can carry operational costs that aren’t immediately visible.
The take away is not alarm but foresight. Decisions around monitoring, separation and communication that are made early shape how manageable an outbreak remains and how disruptive its aftermath becomes. By visualizing those downstream effects in advance, the model offers a way to stress-test response strategies before they’re needed, helping dairies prepare for uncertainty rather than react to it.
