An earlier Fact of the Day noted that about 15,000 km3 of rain falls on the North American continent every year. Where do all those raindrops go? One approach to measuring and tracking water flows is “water footprint” analysis. This approach to the question would seem on a cursory examination to indicate that more than one in every twenty-five raindrops in all of North America goes to producing beef, chicken, soybeans, wheat, and corn in the United States. Here’s a table of some water footprints for various agricultural products multiplied by annual US production.
Water Footprints of agricultural products and production volume in the U.S.
Does that number seem high? It does to me. In fact I don’t believe them at all! The main problem is in the way the water footprint analysis is done. Some of the problems I have with water footprint analysis I have discussed before
, at least in the abstract. I’d like to offer more specific examples of what I find misleading about the water footprint.
1. Inconsistent time frame of analysis. The water footprint for beef, as calculated, stems mainly from the water inputs into feed production over the entire lifespan of the animal, from birth to slaughter. Waterfootprint.org assumes cattle take three years to reach maturity, which sounds reasonable enough. Certainly it takes water to grow feed for the cattle in all three years of its life. But the failure to base the water footprint of a product on some consistent unit of time leads to some strange consquences. Take a calf named Betsy. Some of the same water molecules that grows the grass that feeds Betsy in the year after her birth get evapotranspirated back into the atmosphere; others get diverted to streams and rivers and flow to the ocean, from where the evaporate again. Eventually they fall again as rain. The turnover time for this process is much less than a year. For example, water draining to the Mississippi River basin reaches the Gulf of Mexico in less than one month. The mean residence time of moisture in soil before evaporation might be ~90 days. (These numbers are coming from Figures 5.5 and 9.1 of this book.) Thus, those water molecules that we counted as part of the year I water footprint for production of our hypothetical calf might be the same molecules that are counted again in year 2!
COUNTERPOINT AND RESPONSE: Prof. A. Y. Hoekstra, perhaps the chief proponent of water footprint analysis, points to FAQ Question 4 at the waterfootprint.org site:
…But in a certain period one cannot use more water than is available. A river can be emptied and in the long term one cannot take more water from lakes and groundwater reservoirs than the rate with which they are recharged. The water footprint measures the amount of water available in a certain period that is consumed (i.e. evaporated) or polluted. In this way, it provides a measure of the amount of available water appropriated by humans. The remainder is left for nature.
To me, this “answer” acknowledges the problem but provides no indication that water footprint analysis solves it. Water footprint analysis does not reveal the “amount of available water” that humans could appropriate “in a certain period”, it reveals the amount of water that humans do appropriate in a certain period. And it appears that in many cases, the period used for water footprint analysis is longer than the time it takes the water cycle to recharge itself.
2. Failure to consider substitution effects. At waterfootprint.org, note the following text:
… to produce one kilogram of boneless beef, we use about 6.5 kg of grain, 36 kg of roughages, and 155 litres of water (only for drinking and servicing). Producing the volume of feed requires about 15300 litres of water in average.
In the US, the grain is probably mostly corn. Even if it were the ostensibly more water-intensive wheat, the water footprint per kg of beef from grain would be only 8450 L (from 6.5 kg grain per kg of beef × 1300 L of water per kg of wheat). That’s only about half of the feed water. The remainder, 7100 L, must come from the roughage. At waterfootprint.org, roughage is described as “pasture, dry hay, silage and other roughages”. Fair enough. But what happens if the beef industry decides it doesn’t need to produce Betsy, and as a result, Betsy is never born? Do we save 7100 L of water for every kg that Betsy would have weighed in year 3 of her life? Possibly, but only if none of the pasture, dry hay, and silage that would have gone to feed Betsy is produced in her absence. If Betsy’s portion of hay is grown anyway, it will still evapotranspirate water that could have been directed to other uses, even if the hay is left on the field instead of harvested. In short, it seems far easier to change the water footprint attributed to a product than it does to change the perturbations to the water cycle caused by the production of that product.
COUNTERPOINT AND RESPONSE: Prof. Hoekstra says: “If not appropriated for human consumption (e,g, hay as input of cows that provide meat), then the water is available to sustain natural vegetation (or in the case of river water: sustain aquatic life)”. True enough…but why design a method of water use accounting that stacks the deck in favor of “natural” vegetation so highly? From a water use perspective, what is the difference between a prairie grazed by a “natural” bison population which people do not eat, and a pasture grazed by cattle, which people do eat, other than in the former case people have less food?
3. Double-counting. Even if Betsy does get produced, and even if we ignore the incommensurate time scales of Betsy’s life and the terrestrial water cycle, there’s yet another problem in adding up the water footprints of various agricultural commodities, as I did above. In the US, about 55% of the US corn crop is used as feed for meat production. That means that 55% of the water we’ve attributed to corn production gets counted again when we tabulate up numbers for beef, chicken, or other meats.
COUNTERPOINT AND RESPONSE: Prof. Hoekstra agrees with me that water footprints cannot be added in the way I attempted. Opinions may differ, I suppose, on whether this property is a feature or a bug. But in my view, since mixing a gallon of water with a gallon of water results in two gallons of water it would be nice if water footprints had the same property.
I was very curious to hear a proponent of water footrprint analysis point out flaws in my reasoning or defend waterfootprint.org’s use of these numbers. To that end, I contacted Professor A. Y. Hoekstra, whom I believe to be the primary exponent of water footprint analysis, and shared with him my three concerns. Graciously, he has responded, and I have included his rebuttals (and my response to them) inline with my arguments above.