Water footprint assessment in space and time to support local and global sustainability

Crop production vastly dominates global freshwater use, accounting for nearly 70% of the total withdrawal and around 90% of the total consumption. Human beings are currently using 30% of precipitation-recharged soil moisture and less than 10% (i.e., 3800 km3yr-1}) of the maximum available renewable freshwater resources in the word. Notwithstanding, water resource availability is highly variable in space and time, and different studies have shown a significant mismatch between water use and availability. Accordingly, two-third of global population live under conditions of sever water scarcity for at least one month per year. Moreover, as a consequence of larger food demand and changing living standards, toward more caloric and protein intense diets, global water use has increased by 6-8 times during the past century. At the same time, areas equipped for irrigation have doubled with actual irrigation having unavoidable consequences for aquifers and river ecosystems. Future scenarios of climate change are expected to worsen this picture. Indeed, the rising trends of water demand may continue in the future, harshening the conditions in areas reaching critical thresholds of acceptable water balance. In this context, the goals of this thesis are (i) to identify the main determinants of water use efficiency in agriculture; (ii) to introduce a link prediction algorithm applied to the international trade of agricultural goods; (iii) to introduce a novel indicator to monitor the (mis)match between water use and supply. This thesis quantifies the crop water footprint (CWF, or amount of water use per unit weight of crop) of nine major crops (i.e., wheat, rice, maize, soybean, barley, potatoes, sugar cane, sugar beet, and cotton) through a daily soil water balance run on a grid with a 5’x5’ spatial resolution. The model considers scenarios of rainfed and irrigated crops, also exploring multi-cropping patterns. Quantitative assessments of green and blue (separated into surface and ground) CWF are mapped and analysed in order to identify and monitor the major local drivers of water use, such as climatic conditions, precipitation rate during the growing season, cropping calendar, soil properties, crop yields and agricultural management practises. Results show that crop yield is the most important determinant of the total CWF. Moreover, results of a first-order sensitivity analysis show that, e.g., wheat CWF is mostly sensitive to the length of the growing period, rice CWF to the reference evapotranspiration depth, soybean and maize CWF to the planting date. The CWF model has been adopted also to validate a Fast Track approach, recently developed to study the CWF changes in time, which are generally kept aside in Water Footprint assessments. This approach ascribes the temporal CWF changes only to the yield variations, while it assumes the evapotranspiration depth as time-invariant. This thesis shows the good performance of this approach and also provides an uncertainty analysis. Accordingly, the Fast Track approach shows an error three times smaller than the uncertainty associated with the CWF model. Following the yields patterns, CWF has significantly decreased along the period 1961-2013, but with different rates depending on the crop and the location of the production sites. In the second part of the thesis, the crop water footprint is compared to the local water availability, to assess the sustainability of crop production. In order to understand the size of local (mis)match between crop water use and available water resources, we introduce a water debt repayment time indicator (WD). The WD quantifies the time the hydrological cycle takes to replenish the water resources used for annual crop production, distinguishing the different sustainability levels of soil-, surface-, and ground-water. This indicator highlights the locations and typology of threats imposed by agricultural production on water resources. On a global average, we found that wheat and rice production critically overuses ground water resources, especially in China and the US, and cotton production overuses both surface -and ground-water, particularly in the US. Locally, unsustainable annual crop production is found over the Sabarmati basin (due to wheat) in India, and in the Chao Phraya basin (due to rice and sugarcane) in Thailand, where the water debt repayment time exceeds 5 years in many cultivated areas. Including in the same framework analyses on water use efficiencies (through the CWF) and measure of water use (un)sustainability (through the WD) enables screening analyses at finding specific solutions in cases of low water use efficiencies and/or in critical situation of overuses. While local drivers monitor the water use for production, global drivers attempt to explore the globalization of water resources that happens through the international trade of agricultural goods. Why do countries become trade patterns, hence establishing a more or less stable relation, which implies externalization of water resources use? The third part of this thesis answers to this question through the elaboration of a threshold-based link prediction algorithm, aiming at finding the drivers behind link activation. Accordingly, a link is expected to exist depending on the predicted virtual water volume traded from the source node to the target node: the link is modelled as active when the volume is higher than 1000 m3y-1, non-active otherwise. This algorithm is able to capture 84% of the currently active links and 93% of non-active links. Country population, geographical distance between countries and fertilizers use are the major drivers to explain link existence. The link prediction model may be applied to build future scenarios of virtual water trade, in order to understand how local consumption and production patterns could affect the trade network. Finally, in order to understand how close water demand to water availability is, we introduce a water debt (WD) indicator. The WD quantifies the payback time the hydrological cycle takes to replenish the water resources used for annual crop production. Hence, it highlights the locations and typology of threats imposed by agricultural production on water resources. E.g., the annual production of the nine study crops arise a WD of 10 years with the ground water resources of the US High Plain aquifer, mostly as a consequence of maize and soybean production. This indicator intends to connect and integrate water resource management with other environmental issues, such as the carbon footprint. In short, the thesis contributes advancing our knowledge in the spatio-temporal explicit water footprint assessments, virtual water trade network, sustainable water use. The models developed in this thesis and the results shown in the following chapters allow (i) to explore pathways toward improved water use efficiencies and more sustainable water withdrawals, (ii) to model backward and forward trade network dynamics, and (iii) to project future water use scenarios.