There's been a good deal of research recently on 'embedded' or 'virtual' water.
My understanding of this is: the quantity of water used to grow a crop, per harvested unit of that crop. Some data on common crops is available. For wheat for example one estimate is that 1,400 litres of water are needed to grow a kg of wheat (never mind the specifics for now). For rice the figure is 3,000 litres of water for a kg of rice (harvested).
The question is: how 'true' to the idea of water use is the concept of embedded water? Obviously, the kg of rice that we buy at the store doesn't have a truck outside full of water that represents what that kg has consumed. But that is the sort of imagery the idea of 'embedded' or 'virtual' water seeks to evoke.
What this concept seems not to be able to tell us is that the water, like almost all water on the planet, is in fact part of the hydrological cycle.
Would it be more accurate to say: a kg of harvested rice has 'rented' the use of 3,000 litres of water in order to become consumable? This would imply that the water has returned to the cycle (in whatever form, it could have made its way into a neighbouring vegetable field where it was again 'rented').
Does that sound more realistic?