Milk Yield Losses Under Increasing Heat Load: What Automated Milking System Data Reveal

Most of us know a hot week when we see one. The cows bunch up in the shade, breathe fast, and the bulk tank comes up light. What an automated milking system (AMS) adds is a daily, cow-by-cow record that lets you ask how much yield falls, how fast, and at what point. The published answers are more nuanced than the usual rule of thumb, and they are worth understanding before you decide where to spend cooling money. For the basics of signs and cooling options, start with our guide to heat stress in dairy cows. This article is about what the robot data add.

Why robot data are different

A parlor herd usually gives you one milk weight per cow per milking, tied to a test day or a tank. A robot records every visit: yield, milking interval, flow, conductivity, and often rumination or activity from collars. That means heat effects can be separated into at least three pieces: how much milk each visit produces, how often cows choose to visit, and how the milk itself changes. These move on different schedules, and heat load does not hit all three at once.

Heat load is more than temperature

Researchers describe heat load with a temperature-humidity index (THI), which combines air temperature and relative humidity. Studies differ on how they compute it and which daily value they use, such as the daily maximum or the daily mean, so thresholds from one paper do not transfer cleanly to another. Treat any single "magic number" with suspicion.

A pasture-based robot herd in Australia

One useful dataset comes from a University of Melbourne research farm. In a published observational study, Talukder used three years of records (June 2016 to March 2019) from three Lely Astronaut robots serving up to 158 Holstein-Friesian cows on a pasture-based, voluntary-movement system. Rumination came from transponder collars, and THI was calculated from a local weather station using the daily maximum. The full paper is on the CSIRO publishing site.

The findings are a good corrective to simple stories:

• Milk yield rose with the daily maximum THI up to roughly 65, stayed about flat until roughly 85, and then declined.
• Milking frequency peaked around a THI of 90 and then dropped steadily; the paper says it fell significantly only under severe heat stress.
• Rumination was highest at mid-range THI and lower at both cool and very hot extremes.
• Fat percentage generally fell as THI rose, and the fat-to-protein ratio declined.
• Yield losses were smaller than in earlier Australian pasture studies, which the author links to cool nights and to cooling measures such as fans, sprinklers and shade.

The takeaway is not that heat is harmless. It is that the damage depends heavily on whether the cow gets a chance to recover overnight and whether the farm gives her tools to shed heat. Nights, shade and airflow are management levers.

A confined robot herd in Italy

A second paper looked at a very different setting. Zanchi and colleagues analysed AMS records from a commercial farm in northwestern Italy with four Lely Astronaut robots and about 240 Holstein Friesian cows, covering July 2016 to February 2023. Heat stress was defined as a THI of 68 or higher, calculated from an outdoor weather station. The full text is available through the University of Turin repository.

Two points stand out. First, milk temperature tracked THI closely (a maximum correlation of 0.68 with no lag), which is a reminder that the robot's own sensors can act as a cheap early warning of heat load. Second, the relationship between THI and milk yield changed sign over time: negative in the early years and positive in the later years. The authors suggest housing improvements might explain this, but they did not test it. They also caution that their weather data came from outside the barns, so finer effects inside the building may be hidden.

That second caveat matters for any farm. If your barn is cooled well, outdoor weather becomes a weaker predictor of what the cow actually feels.

Putting robot data to work on your own farm

1. Log barn conditions, not just the weather app. A cheap temperature-humidity logger at cow height in the barn and the holding area gives you a THI you can trust.
2. Plot yield, visits and rumination against that THI. Look for the point where your own herd bends, instead of borrowing a threshold from a paper.
3. Watch milk temperature and rumination as early flags. Both can move before yield does.
4. Separate cow groups. Fresh cows and high producers generally carry more metabolic heat, so check them first.
5. Check lying time and visits at night. If cows are not recovering after dark, add airflow in the resting area.

Do not forget the carryover

Heat can affect milk components and intake for days, so a bad week is not always over when the weather breaks. Compare components and rumination for several days after a hot spell before you declare a recovery. For the dollars side of the equation, our piece on the dairy enterprise budget shows how to translate a daily loss per cow into a cost per hundredweight.

Limits to keep in mind

These are observational studies on two commercial or research herds. They cannot prove that a particular cooling step caused a particular gain, and herds with different climates, cooling and feeding systems will see different curves. Use the papers to frame questions, then answer them with your own records. A vet, nutritionist or extension dairy specialist can help you read the pattern, particularly if components or fresh cow health shift at the same time.

If you want more background on the dairy sector as a whole, farm.fm keeps a running dairy section, and the Agriculture Directory is a handy place to look up extension and industry resources. This site is independent and is not affiliated with any breed association.