A quantitative review of water footprint accounting and simulation for crop production based on publications during 2002–2018

Abstract The water footprint (WF) of crop production has been widely accepted as a comprehensive indicator of agricultural water consumption. Rationality and accuracy in crop WF accounting are thus prerequisites for implementation of WF assessments that yield sustainable regional agricultural water management. However, few studies have focused on the rationality of multiple quantitative approaches and the associated differences in crop WF accounting among different studies. Here we, focusing on maize, wheat and rice, review quantitatively the effects of different quantification approaches and scales on the results of crop WF accounting and simulations in relevant published research during 2002–2018 worldwide. Results show that (i) The number of studies on crop WF accounting has increased by 17 times since 2002 (~the year of creation of WF concepts); the research direction is focused on improvement of quantification and resolution in both time and space. (ii) The current approaches to WF calculation can be divided into five main types: the field crop water requirement (FCWR) approach, field soil water balance (FSWB) approach, regional water balance (RWB) approach, remote sensing (RS) approach and field measured water balance (FMWB) approach. The FCWR and FSWB approaches are more widely adopted than the other three. (iii) There were non-negligible differences in the WF accounting results among approaches and scales. At the global level, the deviations in WF for maize, wheat, and rice were relatively low among different studies, with the world average values of 0.73 m3 kg−1 ± 14.9%, 1.136 m3 kg−1 ± 13.5%, and 1.269 m3 kg−1 ± 27%, respectively. The ranges of uncertainty varied significantly when downscaling to specific countries and provinces. The maximum coefficients of variation (CV) of WF for maize, wheat, and rice in different regions were up to 40%, 49%, and 50%, respectively. (iv) The WF simulations showed very reasonable agreement and lower deviation between the FCWR and FSWB approaches.