The Fine Print on Beef's Water Use

By Brian Tomasik

First published: 19 Mar 2017. Last nontrivial update: 19 Mar 2017.

Summary

Fierce debates are fought over the question of how much fresh water it takes to produce a kilogram of beef. These statistics are widely cited, but less often is it explained what these numbers mean or to what extent we should care about them. A lot of the debate involves disputing definitions and a failure to clearly distinguish among so-called blue, green, and grey water. In addition, even if we focus on just blue water, there are major regional differences in blue-water use. The blue-water footprint of beef in arid Western US states is indeed quite high.

Side note: I probably oppose blue-water use to irrigate pasture and feed-crop fields because irrigation presumably increases total plant growth, thereby probably creating larger populations of invertebrates than would have existed otherwise. For this reason, beef produced in dry regions is probably worse or less good from the standpoint of wild-animal suffering than beef produced in wetter regions.

Contents

Introduction

Red cow on limestone pasture in BohoThe 2014 documentary Cowspiracy begins with a discussion of the water consumption involved in beef production. At around 6 mins, 30 secs into the film, narrator Kip Andersen says:

I found out that one quarter-pound hamburger requires over 660 gallons of water to produce. Here I've been taking these short showers trying to save water, and to find out just eating one hamburger is the equivalent of showering two entire months.

One problem I have with this comparison is that a main reason environmentalists should care about showering is not water use but energy use to heat water. A better comparison would have been with toilet flushing or some other use of non-heated water.

However, there's further ambiguity about what exactly the "660 gallons" figure means, which I'll discuss in this piece.

Interpreting water footprints

Why does water use matter at all? Obviously, the answer to this question depends on your perspective. For example:

The impacts of water use thus depend on one's goals and the local context. This casts doubt on the appropriateness of unidimensional metrics like a water footprint. There are several existing critiques of water footprints that echo this point.

We should also distinguish water use vs. water consumption, where "use" means total water withdrawn, while "consumption" means water withdrawn that doesn't quickly return to the original water source, such as because of evapotranspiration. From the perspective of reducing the killing of zooplankton, water use (at least for non-thermoelectric uses) is more important, while from the anthropocentric perspective of water shortages, water consumption is usually more relevant.

Finally, we need to distinguish blue, green, and grey water footprints. Mekonnen and Hoekstra (2010) explain:

The blue water footprint refers to consumption of blue water resources (surface and groundwater) along the supply chain of a product. [...] The green water footprint refers to consumption of green water resources (rainwater in so far as it does not become run-off). The grey water footprint refers to pollution and is defined as the volume of freshwater that is required to assimilate the load of pollutants given natural background concentrations and existing ambient water quality standards.

In simple terms, blue water is water that humans withdraw, while green water is plant-utilized rainwater that falls naturally on an area.

Lack of context in water footprints

Frontier Economics (2008) criticizes the "virtual water" (i.e., water-footprint) metric because, among other things (p. i), it "Implicitly assumes that water that would be released by reducing a high water use activity would necessarily be available for use in a less water-intensive activity." In fact (pp. 7-8, 11):

Water is heavy and costly to transport and therefore opportunities to shift water around in response to changes in production patterns, although important in some circumstances, is not unlimited, and in some cases is completely infeasible.

Water that is available to rangeland agriculture would not generally be available to be used to support other production systems since there is no infrastructure to capture and transport this water. The implication is that this water will have little or no alternative use, or in the language of economics, this water has a low opportunity cost. As we discuss in the following section it therefore makes no sense to make a direct comparison between this water and water that has a multitude of possible uses. [...]

This variation in the opportunity cost of different volumes of water is sometimes acknowledged in some of the virtual water literature. Hoekstra and Chapagain (2007, pp. 46–47), for example, note tellingly:

Also one has to realise that some parts of the total water footprint concern use of water for which no alternative use is possible, while other parts relate to water that could have been used for other purposes with higher added value. There is a difference for instance between beef produced in extensively grazed grasslands of Botswana (use of [rainfall] water without alternative use) and beef produced in an industrial livestock farm in the Netherlands (partially fed with imported irrigated feed crops).

However, these important caveats are seldom brought to the fore in the presentation and analysis of virtual water estimates – or when drawing policy conclusion from such analysis.

Should we count green water?

Personally, I think inclusion of green water in a water footprint is very misleading. Why are we counting rainwater that produces crops or pasture but not rainwater that feeds natural ecosystems? If you want to count rainfed plant growth for accounting purposes, fine. But including rainfed plants in a water footprint, seemingly on par with human-withdrawn blue water, makes no sense to me. It seems like a recipe for confusion, and as we'll see, this point is at the root of the debate over beef's water use.

Mekonnen and Hoekstra (2010) acknowledge that green-water is generally less important (p. 24): "freshwater problems generally relate to blue water scarcity and water pollution and to a lesser extent to competition over green water". Unfortunately, when people report water footprints in a shallow way, they usually just report the sum of blue, green, and grey water amounts.

Mekonnen and Hoekstra (2010) mention some justification for counting green water (p. 19):

In some cases the roughages eaten by ruminants are produced with land and water resources that cannot alternatively be allocated to crop production for human consumption (e.g. in the case of grazing in dry or wetlands), but often the land and water resources used for roughages supply can alternatively be used for crop growth for human consumption, so that ruminants compete with humans for food also through consumption of roughages.

But if we're counting evapotranspiration on land that could be used for crop cultivation, then shouldn't we count the green-water footprint of wilderness too? Environmentalists should seemingly be appalled at how much green water natural vegetation is using.

Perry (2014), pp. 123-24:

Deurer et al. (2011) and Herath et al. (2011) examine water footprints in New Zealand, comparing procedures in the 2009 [water-footprint] manual to methods that reflect traditional hydrologic principles. In their study of kiwi fruit, they find that the water footprint methodology assesses water consumption at 100 L per tray of fruit, while a hydrologic analysis estimates a net contribution to recharge of 500 L per tray. The hydrologic approach recognizes that kiwi fruit production consumes less water than the natural vegetation it replaces, while the water footprint method does not. Other authors have noted the importance of adjusting the datum, in order to measure the change in water consumption from an alternative [evapotranspiration] ET scenario, rather than a zero base (Scanlon et al., 2007; Pfister et al., 2009). Such an adjustment is not described in the Water Footprint Manual.

[...] most of the landscape in the UK would be forested today, if the trees had not been removed long ago to enable agriculture. The hydrologic impact of the conversion to crop and livestock production has been to decrease in situ [evapotranspiration] ET (green water consumption), while increasing runoff (blue water production).

Several authors have suggested that meat production consumes more water than grain production. [...] The implication is that meat production is more water-intensive than grain production. Closer examination reveals a much more complicated picture. As most livestock production occurs on land that might otherwise have been consuming more water in its natural, forested state, the down-stream effect of expanding any agriculture in such areas is likely to be an increase in the available runoff (blue water).

This last point about cattle production possibly increasing blue water is interesting, but I won't pursue it here. In this piece, I discuss differing estimates of beef's water footprint that have been debated over the past few decades.

A "pro-beef" water-footprint estimate

Beckett and Oltjen (1993) estimate water use for beef production in the USA. The authors mention in a footnote (p. 818): "Work supported in part by the California Beef Council." Robbins (n.d.) gives more backstory:

In 1978, Herb Schulbach (Soil and Water Specialist, University of California Agricultural Extension), along with livestock farm advisors Tom Aldrich, Richard E. Johnson, and Ken Mueller, published extensive research on water use in California agriculture in the journal Soil and Water (no. 38, fall 1978). They concluded that the average pound of beef produced in California required 5,214 gallons of water.

The livestock industry took great offense at this. Schulbach told me that they “turned a scientific project into political football.” Subsequently, at the behest of the cattlemen, Jim Oltjen and colleagues in the Department of Animal Science at U.C. Davis came out with a very different calculation, asserting the requirements for a pound of beef to be 441 gallons of water. Jim Oltjen’s work, along with similar work by Gerald Ward (Department of Animal Science, Colorado State University) forms the basis for the figures that the National Cattlemen’s Beef Association have used ever since to rebut the arguments of environmentalists who point to the enormous waste of water involved in modern beef production. (How identified Jim Oltjen is with the industry can be glimpsed from his official portrait at the University of California, where he wears a cowboy hat.)

I agree with the principle of caution when encountering industry-funded research. But after reading the entire Beckett and Oltjen (1993) paper, as well as other data I'll discuss later on, I think Beckett and Oltjen (1993) is possibly the best estimate of beef's water use in the USA that has been done. (If you find a better study on this topic, let me know.)

Beckett and Oltjen (1993) estimate the blue-water footprint of beef, although they don't use that terminology. That is, they count irrigation of crops and pastures but not rainwater that would have fallen anyway. Their estimate of water use to produce a kg of boneless beef in the USA is 3,682 liters (p. 818). Since much of the debate on this topic is done in the USA, this number is often expressed as 441 gallons of water per pound of beef (such as here).

Water used per cow

Following are two useful tables from Beckett and Oltjen (1993) (pp. 824-25), where the numbers are per year:





Dividing 25,325,109 million liters by (28,397,470 feedlot cattle + 5,747,100 cull cows) gives 742,000 liters per bovine killed, i.e., each bovine requires an average of 742,000 liters of blue water over its lifetime.

Nelson (2001) criticism

Nelson (2001) criticizes Beckett and Oltjen (1993)'s 441 gal/lb number, although he doesn't cite or mention their paper explicitly:

Not surprisingly, the beef industry promotes a study that determined, using highly suspect calculations, that only 441 gallons of water are required to produce a pound of beef.

(The cattlemen's study applied liberal deductions from water actually used, reasoning that water was evaporated at points during the process, or was "returned" to the water table after being used to grow plant feed, or was returned to the water table via urea and excrement from cows. Thus, study authors reasoned these waters were not "lost" but "recycled" and therefore could be subtracted from gross amount of water actually used in beef production. Of course, evaporation and cow dung don't go very far in replenishing water pumped from acquifers [sic] which took thousands of years to fill. It's interesting to consider that if the same fuzzy math were applied to calculating how much water it takes to grow vegetables, potatoes would probably only require about 2 gallons of water per pound.)

I'm puzzled by the second paragraph of this quote, because on my reading, Beckett and Oltjen (1993) don't discuss the alleged deductions for evaporation, urea, etc. at all. As best I can tell, Nelson (2001) was misreading the paper or maybe just heard about the complaint secondhand without verifying it. The closest I can find to something in the neighborhood of these "liberal deductions" is the following (Beckett and Oltjen 1993, pp. 825-26):

The model takes into account all the water that is applied to irrigated feedstuff crops and irrigated pasture. This total water is then used to calculate the water use efficiency for each crop by state and any feed that is fed to cattle is assigned that water use efficiency. However, 10 to 20% of the water that is applied to crops during irrigation runs off the fields and returns to water sources such as irrigation ditches, reservoirs, or streams (Richard Pruitt, personal communication). This water is then available for human consumption again and is considered developed water. [...] Considering a 15% runoff of applied water for both irrigated pasture and feedstuff crops, the adjusted water requirement for beef would be 3,151 L/kg of boneless beef. This is 529 L (14.4%) less than when runoff water is not considered.

But this 15% reduction in water use is not part of the main figures. Rather, the authors mention it in their "Results and Discussion" as a caveat that one might apply to their main figures.

Withdrawal vs. consumption

Nelson (2001)'s complaint, while apparently erroneous, raises the general question of whether to count water withdrawn or water consumed (i.e., evaporated or otherwise lost from the local area). Mekonnen and Hoekstra (2010) focus on consumption (p. 9):

The blue water footprint refers to consumption of blue water resources (surface and groundwater) along the supply chain of a product. ‘Consumption’ refers to loss of water from the available ground-surface water body in a catchment area. Losses occur when water evaporates, returns to another catchment area or the sea or is incorporated into a product.

In contrast, Beckett and Oltjen (1993)'s default numbers tallied up something closer to water withdrawals, because they didn't subtract off water that "returns to water sources such as irrigation ditches, reservoirs, or streams".

In this piece, I've ignored the distinction between withdrawal vs. consumption, and I use the term "blue water" to refer to either withdrawal or consumption, because I would guess that in the case of irrigation (the most important factor in beef's water footprint), water consumption is usually a pretty sizeable percentage of water withdrawal. The withdrawal vs. consumption distinction probably isn't big enough to explain most of the differences in water-footprint estimates among sources.

Environmentalist water-footprint estimates

Borgstrom (1981)

Robbins (n.d.) explains:

The figure of 2,500 gallons to produce a pound of meat that I used in Diet For A New America comes from a statement by the renowned scientist Dr. Georg Borgstrom at the 1981 annual meeting of the American Association for the Advancement of Science, in a presentation titled “Impacts On Demand For And Quality Of Land And Water.”

I haven't found Borgstrom's presentation online, but the "2500 gallons" figure is similar to other numbers we'll see later, so I assume it was calculated similarly.

Pimentel et al. (1997)

Pimentel et al. (1997) says (p. 100):

Producing 1 kg of beef requires approximately 100 kg of hay-forage and 4 kg of grain (Pimentel et al. 1980). Producing this much forage and grain requires approximately 100,000 liters of water to produce approximately 100 kg of plant biomass plus 5400 liters to produce 4 kg of grain (Falkenmark 1994). On rangeland, more than 200,000 liters of water are needed to produce 1 kg of beef (Thomas 1987).

So almost all the water use here is from growing forage. But except in the case of pasture irrigation (blue water), this water would fall anyway (green water).

Fairlie (2010), commenting on Pimentel et al. (1997), says (p. 64): "Pimentel's calculation takes into account every scrap of precipitation that falls upon the area of land that a beef cow might occupy."

If we take out Pimentel et al. (1997)'s forage water numbers and only use the 5400 liters of water used to produce the grain for 1 kg of beef, we get 5400 liters/kg = 650 gal/lb, which is pretty close to the castigated Beckett and Oltjen (1993) figure of 441 gal/lb.

Mekonnen and Hoekstra (2010)

Mekonnen and Hoekstra (2010) calculate blue, green, and grey water footprints for various animal products. They use units of m3 of water per ton of product, but this is the same as a liter per kg. Table 4 (p. 25) gives the following water footprints for boneless beefa (averaged over different production systems):

Region Blue water (L/kg) Green water (L/kg) Grey water (L/kg)
USA 525 12,933 733
Global average 550 14,414 451

Adding up the "Global average" numbers gives 15,415 L/kg, which is the same as the estimate given for beef in the Water Footprint Network's "Product gallery": "The global average water footprint of beef is 15400 litre/kg. This is predominantly green water (94%)." 15,415 litre/kg = 1847 gal/lb.

If we just look at blue water in the USA, we see 525 L/kg, which is 63 gal/lb. This is an order of magnitude less than the Beckett and Oltjen (1993) figure of 441 gal/lb. How can this be? Did I make a mistake somewhere?

Capper (2014) criticizes the Water Footprint Network figure (which she quotes as 15,500 L/kg) because

the authors used global averages to calculate water usage, which were then assumed to be representative of individual beef production systems, regardless of region or productivity. By contrast, the thorough analysis of water consumption within beef production published by Beckett and Oltjen (1993) with system boundaries extending from feed production to processing reports the aforementioned water-use figure of 3,682 L per kilogram of boneless beef.

Actually, it's not true that Mekonnen and Hoekstra (2010) only "used global averages".b In fact, Mekonnen and Hoekstra (2010) explain (p. 8): "We have estimated the amount of feed consumed per animal category, per production system and per country based on estimates of feed conversion efficiencies and statistics on the annual production of animal products." And as you can see in the above table, Mekonnen and Hoekstra (2010)'s water footprint for the USA is pretty similar to the global average, and their blue water footprint is, mysteriously, much lower than the Beckett and Oltjen (1993) number.

Mekonnen and Hoekstra (2010) mention (pp. 36-37) some other estimates of the global total green+blue water footprint of grazing. Three of these estimates are roughly an order of magnitude higher than Mekonnen and Hoekstra (2010)'s own estimate. Why? "unlike the current study, the estimates in these three studies refer to the total evapotranspiration from grazing lands rather than to the evaporation related to the grass actually consumed" (p. 36).

Finally, Mekonnen and Hoekstra (2010) acknowledge that blue water generally matters more than green water (p. 39):

The total water footprint of an animal product is generally larger when obtained from a grazing system than when produced from an industrial system, because of a larger green water footprint component. The blue and grey water footprints of animal products are largest for industrial systems (with an exception for chicken products). From a freshwater perspective, animal products from grazing systems are therefore to be preferred above products from industrial systems[.]

Cowspiracy

This page provides a helpful collection of statistics from Cowspiracy and sources for them.

For the statement that "One hamburger requires 660 gallons of water to produce", we find these citations:

For the statement that "2,500 gallons of water are required to produce 1 pound of beef", we find these citations:

Finally, we have this statement: "Animal agriculture uses 34-76 trillion gallons of water annually." The Cowspiracy page describes how the 34-trillion figure is derived. That methodology seems fairly reasonable to me, but it's hard to compare with the other figures because it's for all of animal agriculture, not just beef. Still, I'll play around with this number a bit. Beckett and Oltjen (1993) reported (in their Table 17, see above) an annual production of 5,873,301,157 kg boneless beef from feedlot cattle + 1,004,135,490 kg boneless beef from cull cows = 6,877,436,647 kg = 15 billion pounds. So (34 trillion gallons) / (15 billion pounds) = 2200 gal/lb boneless beef, which is quite similar to the widely cited 2500 gal/lb. However, this number is too high because it's counting water use for all of animal agriculture. Perhaps the actual number is a few times smaller, given that non-beef animal products require roughly comparable amounts of blue water as beef does (see the "Comparing beef with other animal products" section below).

Disputing definitions

As we saw above, beef water-footprint estimates depend a lot on whether green water (and sometimes grey water) is counted in addition to blue water. This basic point is usually omitted from the debate, although a few sources mention it:

Regional differences

While people typically focus on the national-average water-use figure from Beckett and Oltjen (1993), that paper showed some interesting differences in water use per region of the USA. For example, Table 4 (p. 820) shows that the national-average water use for alfalfa hay is 257.1 L/kg. But for the states CA, AZ, and NV, this average is more than three times the national average, at 893.1 L/kg, and 98.0% of alfalfa area was irrigated. Meanwhile, irrigation water use for alfalfa in more Eastern states (MO, IA, MN, WI, IL, MI, IN, OH, NY, PA) was a mere 1.9 L/kg because only 0.5% of land in those states was irrigated. Maybe these differences are a bit too extreme to take at face value, but the general trend that blue-water use varies dramatically by location seems plausible.

A similar point applies for other crop and pasture numbers. For example, Table 6 (p. 821) shows national-average water use for grain sorghum at 118.4 L/kg, but for CA, AZ, and NV, the number is 1,645.4 L/kg. Thus, it may be that raising cattle in certain Western US states uses several times more water per kg of beef than the national average. This may be a second source of variability among numbers for beef's water use.

Robbins (n.d.) touches on the same point: "Of course, beef produced in different parts of the country will take different amounts of water. Beef produced in the Southeast takes much less water because you don’t need to irrigate nearly as much thanks to so much more rain during the growing season. Arizona and Colorado beef, on the other hand, take even more water than California’s."

Oppenlander (2013) estimates (pp. 94-95) that a fully grass-fed beef cow in California requires 4,210 gallons of blue water per pound of beef using a calculation that I don't find totally implausible. This is because California pastures can require so much irrigation.

Lustgarten and Sadasivam (2015) quote John Bredehoeft, "the former manager of the western water program for the U.S. Geological Survey", as saying that "half of the water use in the [US] West is to raise cows".

The point of my piece is not that beef production uniformly requires less water than environmentalists claim. Rather, my point is that context matters a lot, and what's true in certain parts of California isn't necessarily true everywhere.

Comparing beef with other animal products

While it's common to consider beef to be "King of the Big Water Footprints", this is less true when we focus on blue water rather than also counting green water. In particular, following are numbers I collected from Mekonnen and Hoekstra (2010), Table 4, p. 25, focusing on weighted averages over all production systems in the USA:

Animal product Blue water footprint (L/kg) in the USA
Leather (bovine) 658
Pig meat 645
Beef 525
Sheep meat 315
Cheese 310
Chicken meat 187
Egg 130
Milk 60

I include this table only to add context to this discussion. I don't recommend consuming animal products lower on this list. For example, chicken and eggs arguably cause much more farm-animal suffering per kg than beef does (Tomasik 2007).

Why did I write this piece?

My research for this piece started because I was curious how many zooplankton are killed by the water used in beef production, so I wanted an accurate estimate of beef's water use. I continued exploring the issue in depth both to check my initial impressions and because I was annoyed by the creation and propagation of somewhat misleading water-footprint statistics.

I'm generally opposed to animal agriculture, and I don't get any industry funding. But I do feel that discussions of these topics should be honest and transparent.

Further reading

I haven't read any of the following sources. Following is from Mekonnen and Hoekstra (2010), p. 8:

A study by FAO has quantified the global blue water use for feed production, animal drinking and servicing (Steinfeld et al., 2006). De Fraiture et al. (2007) have estimated the global water use for animal feed production, both green and blue but not distinguishing between the two. They considered water use for two lumped categories: feed crops and grazing. Zimmer and Renault (2003) made a rough estimation of the global water consumption for producing meat and other animal products, not showing details per country, animal category or product. [...] Peden et al. (2007) made an estimate of the global water consumption for producing the feed for farm animals. In addition to the studies mentioned there have been a few more specific studies for the Nile River Basin (Van Breugel et al., 2010) and for the USA (Renault and Wallender, 2000; Pimentel et al., 2004).

Footnotes

  1. Mekonnen and Hoekstra (2010) don't use the adjective "boneless" specifically, but the calculations at the end of Table 5 on p. 28 suggest that boneless beef is their "functional unit" (i.e., the thing being measured in the denominator of L/kg numbers), just as it was for Beckett and Oltjen (1993).

    In particular, Mekonnen and Hoekstra (2010) begin with "a total carcass weight of 143 kg" and say that it yields 101 kg of beef. That implies that the percent of carcass consisting of boneless beef is 71%. In contrast, Beckett and Oltjen (1993) take this percentage to be 66.7% (Table 17, p. 825).

    Meanwhile, Mekonnen and Hoekstra (2010) (p. 28) use a live beef cow weight of 253 kg, which implies a 143/253 = 57% dressing percentage, in contrast to the estimate of 62% for that number by Beckett and Oltjen (1993) (Table 17, p. 825).  (back)

  2. Even though Capper (2014)'s paper was published in 2014, the article was first submitted to the journal in 2010, the same year as the Mekonnen and Hoekstra (2010) publication. Maybe Capper (2014) was working with older information that was correct at the time?  (back)