Department Systems Analysis, Integrated Assessment and Modelling

Global food trade entails virtual flows of agricultural resources and pollution across countries. Here we performed a global-scale assessment of impacts of international food trade on blue water use, total water use, and nitrogen (N) inputs and on N losses in maize, rice, and wheat production. We simulated baseline conditions for the year 2000 and explored the impacts of an agricultural intensification scenario, in which low-input countries increase N and irrigation inputs to a greater extent than high-input countries. We combined a crop model with the Global Trade Analysis Project model. Results show that food exports generally occurred from regions with lower water and N use intensities, defined here as water and N uses in relation to crop yields, to regions with higher resources use intensities. Globally, food trade thus conserved a large amount of water resources and N applications, and also substantially reduced N losses. The trade-related conservation in blue water use reached 85 km3 y−1, accounting for more than half of total blue water use for producing the three crops. Food exported from the USA contributed the largest proportion of global water and N conservation as well as N loss reduction, but also led to substantial export-associated N losses in the country itself. Under the intensification scenario, the converging water and N use intensities across countries result in a more balanced world; crop trade will generally decrease, and global water resources conservation and N pollution reduction associated with the trade will reduce accordingly. The study provides useful information to understand the implications of agricultural intensification for international crop trade, crop water use and N pollution patterns in the world.

Trade of commodities can lead to virtual water flows between trading partners. When commoditiesflow from regions of high water productivity to regions of low water productivity, the trade has thepotential to generate water saving. However, this accounting of water saving does not account for thewater scarcity status in different regions. It could be that the water saving generated from this tradeoccurs at the expense of the intensified water scarcity in the exporting region, and exerts limited effecton water stress alleviation in importing regions. In this paper, we propose an approach to measure thescarce water saving associated with virtual water trade (measuring in water withdrawal/use). Thescarce water is quantified by multiplying the water use in production with the water stress index(WSI). We assessed the scarce water saving/loss through interprovincial trade within China using amulti-region input-output table from 2010. The results show that interprovincial trade resulted in14.2 km³ of water loss without considering water stress, but only 0.4 km³ scarce water loss using thescarce water concept. Among the 435 total connections of virtual water flows, 254 connectionscontributed to 20.2 km³ of scarce water saving. Most of these connections are virtual water flows fromprovinces with lower WSI to that with higher WSI. Conversely, 175 connections contributed to20.6 km³ of scarce water loss. The virtual water flow connections between Xinjiang and otherprovinces stood out as the biggest contributors, accounting for 66% of total scarce water loss. Theresults show the importance of assessing water savings generated from trade with consideration ofboth water scarcity status and water productivity across regions. Identifying key connections of scarcewater saving is useful in guiding interregional economic restructuring towards water stress alleviation,a major goal of China’s sustainable development strategy.

Aridity-the ratio of atmospheric water supply (precipitation;P) to demand (potential evapotranspiration; PET) - is projected to decrease (that is, areas will become drier) as a consequence of anthropogenic climate change, exacerbating land degradation and desertification^1-6. However, the timing of significant aridification relative to natural variability - defined here as the time of emergence for aridification (ToEA) - is unknown, despite its importance in designing and implementing mitigation policies ^7-10. Here we estimate ToEA from projections of 27 global climate models (GCMs) under representative concentration pathways (RCPs) RCP4.5 and RCP8.5, and in doing so, identify where emergence occurs before global mean warming reaches 1.5 °C and 2 °C above the pre-industrial level. On the basis of the ensemble median ToEA for each grid cell, aridification emerges over 32% (RCP4.5) and 24% (RCP8.5) of the total land surface before the ensemble median of global mean temperature change reaches 2 °C in each scenario. Moreover, ToEA is avoided in about two-thirds of the above regions if the maximum global warming level is limited to 1.5 °C. Early action for accomplishing the 1.5 °C temperature goal can therefore markedly reduce the likelihood that large regions will face substantial aridification and related impacts.

A large number of local and global databases for soil, land use, crops, and climate are now available from different sources, which often differ, even when addressing the same spatial and temporal resolutions. As the correct database is unknown, their impact on estimating water resource components (WRC) has mostly been ignored. Here, we study the uncertainty stemming from the use of multiple databases and their impacts on WRC estimates such as blue water and soil water for the Karkheh River Basin (KRB) in Iran. Four climate databases and two land use maps were used to build multiple configurations of the KRB model using the soil and water assessment tool (SWAT), which were similarly calibrated against monthly river discharges. We classified the configurations based on their calibration performances and estimated WRC for each one. The results showed significant differences in WRC estimates, even in models of the same class i.e., with similar performance after calibration. We concluded that a non-negligible level of uncertainty stems from the availability of different sources of input data. As the use of any one database among several produces questionable outputs, it is prudent for modelers to pay more attention to the selection of input data.

System identification, sensitivity analysis, optimization and control, require a large number of model evaluations. Accurate simulators are too slow for these applications. Fast emulators provide a solution to this efficiency demand, sacrificing unneeded accuracy for speed. There are many strategies for developing emulators but selecting one remains subjective. Herein we compare the performance of two kinds of emulators: mechanistic emulators that use knowledge of the simulator's equations, and purely data-driven emulators using matrix factorization. We borrow simulators from urban water management, because more stringent performance criteria on water utilities have made emulation a crucial tool within this field. Results suggest that naive data-driven emulation outperforms mechanistic emulation. We discuss scenarios in which mechanistic emulation seems favorable for extrapolation in time and dealing with sparse and unevenly sampled data. We also point to advances in Machine Learning that have not permeated yet into the environmental science community.