Résumé
While water requirement for livestock is widely perceived as daily drinking water consumption, similar to 100 times more water is required for daily feed production than for drinking water. Increasing livestock water productivity can be achieved through increasing the water-use efficiency (WUE) of feed production and utilisation. The current paper briefly reviews water requirements for meat and milk production and the extent of, and reason for, variations therein. Life-cycle analysis (LCA) can reveal these variations in WUE but LCA are not tools that can be employed routinely in designing and implementing water-use-efficient feed resourcing and feeding strategies. This can be achieved by (1) choosing agricultural by-products and crop residues where water applications are partitioned over several products for example grain and straw (or food and fodder) contrary to planted forage production where water and land have to be exclusively allocated to fodder production, (2) select and breed WUE crops and forages and exploit cultivar variations, (3) increase crop productivity by closing yield gaps; and (4) increase per animal productivity to reduce the proportion of feed (and therefore water) allocated for maintenance requirement rather than productive purposes. Feed-mediated WUE of dairy buffalo production on almost completely (94%) by-product-based feeding systems could be reduced from 2350 to 548 L of water per kg of milk by the combined effect of increasing basal ration quality in a total mixed ration, which resulted in increased milk yield of similar to 30%, and by increasing crop productivity from 1 t (actual crop yield) to 3 t (potential crop yield). Exemplary, multi-dimensional sorghum improvement using staygreen quantitative trait loci (QTL) introgression for concomitant improvement of WUE of grain and stover production and stover fodder quality showed opportunities for further linked improvement in WUE of crop and livestock production. Metabolisable energy (ME) yield under water stress conditions measured in lysimeters, (which measure crop water transpired) ranged QTL dependent from 16.47 to 23.93 MJ ME per m(3) H2O. This can be extrapolated to 8.23-11.97 MJ ME per m(3) H2O evapotranspired under field conditions. To mainstream improvement in WUE of feed resourcing and feeding, the paper suggests the combination of feed resource databases with crop-soil-meteorological data to calculate how much water is required to produce the feed at the available smallest spatial scale of crop-soil-meteorological data available. A framework is presented of how such a tool can be constructed from secondary datasets on land use, cropping patterns and spatially explicit crop-soil-meteorological datasets.