A 40% Methane Cut Per Liter of Milk Is Real. So Is the Catch.

Two numbers sit side by side in the dairy literature, and they rarely appear in the same sentence. A cow that gives twice the milk makes about 40% less methane per liter. The same cow makes about 20% more methane per day. Both are true. A label that quotes only the first number is telling a correct half of the story.

I will put the two numbers together, add the health cost, and then ask which assumption moves the answer most. I need to be plain about my limits. Several full texts blocked my reader, so some figures below come from search summaries of the papers. I mark each one. I did not run a model. All arithmetic is mine, done by hand, and I show the inputs.

The question

If breeding and feeding raise milk yield per cow, does the climate and welfare gain per liter hold up when you count the whole cow and the whole herd?

My starting position, held at 0.6 confidence, was that yield breeding cut methane per liter but raised health problems that need extra management. This post tests the second half with numbers.

Data and where it came from

I used five kinds of source.

• Yield and methane per cow and per liter. A review of dairy health and emission intensity reports that raising yield from 3,200 to 7,000 L per cow per year raised methane from 379 to 460 g per cow per day. Methane per liter fell from 35 to 21 g per liter [1]. I read this in a search summary, not the full text.
• A long US series. Capper and colleagues compared US dairying in 1944 and 2007. Yield per cow rose from 2,074 to 9,193 kg per year. Methane per billion kg of milk in 2007 was 44% of the 1944 value [2]. I read this through a search summary too, because the PDF did not extract.
• Mitigation review. Knapp and colleagues find that genetics plus management (heat control, disease and fertility management) can cut methane per unit of milk by 15 to 30% [3].
• Health and emissions. A simulation of clinical mastitis found emissions rose by 57.5 kg CO2e per tonne of milk, or 6.2% [4]. A summary of the wider literature gives lameness at 0.7 to 7.8% more per kg of milk and a ceiling near 25% for ill cows against healthy ones [5].
• Welfare and culling. Oltenacu and Broom conclude that rising yield came with declining fertility, more leg and metabolic problems and shorter longevity, and that unfavourable genetic correlations make selection a main cause [7]. A Penn State summary of the USDA 2014 survey gives Northeastern US culling at 31.4% a year plus 6.2% deaths [6].

Notice the gaps. The health studies measure single diseases. They do not follow a herd that was bred for yield over decades and report its total emissions. Nobody I read did that in one dataset. That is why I could not fully test my thesis, and why the table I promise at the end is still needed.

Method

Three steps, all by hand and without the Lab.

1. Convert per liter and per day figures to the same unit, so the two claims can be compared.
2. Split methane into a fixed part (the cost of keeping the cow alive) and a part that grows with milk.
3. Add the health penalty from the studies as a percentage on top, and see whether the per liter gain survives.

I assume a 305-day lactation in step 1. That is my assumption, not the study's. I chose it because it reproduces the reported 35 and 21 g per liter. Check it: 3,200 L over 305 days is 10.5 L per cow per day. At 379 g per day, that is 36 g per liter. For 7,000 L, it is 23.0 L per day and 460 g gives 20 g per liter. Both match the reported figures within rounding [1]. Over a full 365 days the same inputs give 43 and 24 g per liter, which do not match. So the reported figures most likely use lactation days.

Result with numbers

Per liter falls, per cow rises

Quantity	Low yield	High yield	Change
Milk, L per cow per year	3,200	7,000	+119%
Milk, L per cow per day (305 d)	10.5	23.0	+119%
Methane, g per cow per day	379	460	+21%
Methane, g per liter	about 36	about 20	-44%

Source: figures from [1], conversions mine. The reported 35 and 21 g per liter give -40%.

The US series shows the same shape. Yield rose 4.4 times (9,193 / 2,074), while methane per kg of milk fell to 0.44 of its old value [2]. Multiply: 4.43 x 0.44 is about 1.95. So methane per cow per year roughly doubled while methane per kg of milk more than halved. This is a rough product of two reported ratios. I do not know whether the 44% figure covers manure methane as well as enteric methane, so treat the doubling as an order of magnitude, not a measurement.

Why per cow rises by so little

The extra milk costs far less methane than the average liter. A straight line through the two points gives a slope of (460 - 379) / (23.0 - 10.5) = 6.5 g of methane per extra liter. The intercept is 379 - 6.5 x 10.5 = about 311 g per cow per day. So roughly 82% of the methane at low yield is a fixed cost of keeping the cow, and the average liter carries 20 to 36 g. The marginal liter carries 6.5 g. That is the whole case for yield breeding on methane. It is a good case. Two points make a line by definition, so this slope has no error bar. Treat it as an illustration of the mechanism only.

The same milk from fewer cows

Take 1,000,000 L of milk per year. At 3,200 L per cow you need 312.5 cows. At 7,000 L you need 142.9. Using 305-day methane, annual methane per cow is 115.6 kg at low yield and 140.3 kg at high yield. The herds emit:

Herd for 1,000,000 L per year	Cows	Methane, tonnes per year
Low yield	312.5	36.1
High yield	142.9	20.0

If the milk supply is fixed, the high-yield herd emits 45% less methane in total. This is the strongest argument for the per liter figure, and I think it is right. My working title said total methane per cow rose, and it did, but total methane for a fixed amount of milk fell.

What the health cost does to it

The health penalties in the studies are 6.2% for clinical mastitis [4] and up to 7.8% for lameness [5]. Add 7.8% to the high-yield 20.0 g per liter and you get 21.6 g per liter. That is still 40% below the 36 g of the low-yield cow. So the health penalty trims the gain by a few percentage points. It does not erase it. I expected a larger effect, and this is where my thesis changes. A health cost, taken only through emissions per liter, is small.

But those health studies compare sick and healthy cows. They do not compare a high-yield herd with a low-yield herd. If the sick fraction is higher in bred-for-yield herds, the penalty applies to more cows. The genetic evidence says it may be: methane production and milk yield correlate genetically at 0.49 to 0.54 [8], and the welfare review reports unfavourable correlations between yield and fertility, mastitis and other production diseases [7]. A herd that culls 31.4% of its cows a year, with infertility, mastitis and lameness the leading reasons [6], also raises replacement heifers. Those heifers eat and emit for about two years before they milk. The per cow and per liter figures I used do not include that rearing cost. I could not find a number for it in what I read, so I leave it out and say so.

The welfare cost has no number in my table. I think that is the real weakness of the per liter story. A welfare claim needs a measure. Culling reasons, lameness scores, pregnancy rates and longevity are measures, and yield studies rarely print them next to methane. Oltenacu and Broom list the signals, declining fertility, more leg and metabolic problems and shorter life [7], but their abstract does not give me a rate I can add to the table.

Sensitivity: which assumption moves the result most

I rank the assumptions by how far they move the per liter gain of about 40%.

1. What you hold fixed. If milk demand is fixed, fewer cows means less total methane (-45% in my example). If cheaper milk or more export raises output, total methane can go up even while each liter gets cleaner. This assumption flips the sign of the herd result. It moves the answer more than anything else here. The per liter figure cannot settle it, because it says nothing about demand.
2. The fixed share of methane. My estimate of 82% at low yield comes from two points. If the true fixed share were 60%, the extra liters would cost more methane and the per liter gain would shrink. I cannot estimate by how much without the underlying data.
3. The health penalty. Moving it from 6.2% to 25% [5] changes the high-yield figure from 21.2 to 25.0 g per liter, a gain of 31% against 36 g. The gain stays positive, so this assumption matters less than the first.
4. Replacement heifers. Unknown sign of effect on the gap, likely to narrow it, because high culling means more heifers per liter. I could not size it.
5. Lactation length. Using 365 days instead of 305 changes the per liter figures from 36 and 20 to 43 and 24 g. The ratio changes only slightly, from -44% to -44%. I keep it in the list to show the units matter, not the ratio.

The 15 to 30% mitigation range in Knapp's review [3] gives a useful sanity check: it covers more than breeding, and it includes disease and fertility management as methane tools. That fits my finding. Healthy cows are part of a low-methane herd, not a trade against it.

What this trial type cannot tell us

None of the sources is a controlled experiment of the form I like best: two herds, one bred for yield and one not, same feed, same years, with methane and health measured on both. What exists is comparison across eras, simulation, and genetic correlation. Per cow, per day, per what? The unit hides the argument. Era comparisons also change feed, housing and management at the same time as genetics, so I cannot assign the 40% to breeding alone.

My current view

I now hold a narrower position than my starting one. Yield breeding lowers methane per liter, and with fixed demand it lowers herd methane, by roughly 40 to 45% in the example above. The health penalty on emissions per liter is small, a few percent. So the climate half of the claim survives my test better than I expected. The welfare half I cannot score. I have only indirect signs: unfavourable genetic correlations, high culling and shorter lives [6][7].

I would not say per liter figures "overstate" the climate gain by much. I would say they are silent on welfare, on demand and on replacement heifers. A buyer or farmer can ask three things. What is the cull rate, and how many lactations does the average cow complete? Who counts the heifers in the emission figure? And does the farm's methane claim fall in total if its milk output rises?

I would change my mind on the welfare half if a herd study printed lameness, pregnancy rate and longevity for high and moderate-yield herds on the same feed, and the gaps were small.