**4. Discussion and Policy Implications**

The United Nations Sustainable Development Goal 12 highlights the need to pursue responsible consumption and production to address global environmental challenges such as water scarcity (Target 6.4 of Reference [56]). Pakistan, with water withdrawals for the agricultural sector of 94%, ranks among the highest in the world, compared to 69% at the global level [33]. Resolving water scarcity in developing countries like Pakistan could be achieved by focused research on the agricultural sector and the food system. In addition to devising localized, sustainable production and consumption systems, the trade in agricultural commodities (through tele-consumption of food) can also play a crucial role in reducing water-scarcity at local and global levels.

Literature is available on the amount of virtual water traded between Pakistan and other countries, but there is a lack of research on Pakistan's blue and green virtual water trade, the relative economic value of blue and green virtual water and the implications for national and global water resources, especially the scarcer blue water. Also, the network properties of a Pakistan-centered virtual water trade network have never been analyzed. Differentiating the analysis into blue and green water components is particularly crucial because blue water has a higher opportunity cost [7] and different environmental effects than green water [57].

Similar to the earlier work in the field [7,58,59], we find that during the 1990–2016 period, Pakistan's total virtual water trade has been increasing. However, the net virtual water import has been small. Our future scenario of Pakistan's trade in agricultural commodities, however, shows that Pakistan will lose a significant amount of virtual water by 2030. At first glance, it appears that Pakistan's historical trade in agricultural commodities neither benefits nor harms the domestic water resources. However, once we separate virtual water trade into blue and green virtual water, we discover a large and increasing net export of blue water from Pakistan, which means that Pakistan indeed exported more expensive and scarce blue water. The future trade scenario shows that the losses of blue virtual water will further increase between 2016 and 2030. Our results clearly show the importance of considering the trade in blue and green water separately.

Our analysis shows that the unit value of blue virtual water export through rice crop (the dominant export crop) 0.10 US\$/m<sup>3</sup> has been lower than the corresponding value of green water import through palm oil (the dominant import crop) 0.15 US\$/m<sup>3</sup> during last the decade. This further demonstrates that, although unconsciously, Pakistan has not only been exporting its more precious water resource to other countries, it has been doing so by fetching a relatively lower price for the blue water export thus making the water trade even less economically viable. Moreover, the costs of construing and running extensive irrigation infrastructure for the blue water used in agricultural exports further calls into question the actual benefits of exporting blue virtual water embodied in the commodities. One of the main reasons for these inefficient production and trade decisions is that farmers have been using irrigation water quite inefficiently in Pakistan [60], mainly due to subsidized irrigation water supply.

Pakistan's situation becomes even more worrisome when we consider the increasing scarcity of blue water in the country. Pakistan has been a net exporter of blue virtual water despite being a water-deficient country [61]. Other studies have shown that most of the water deficit countries have compensated for their domestic water scarcity by importing increasing imports of cereals [62,63]. The authors assert that upon reaching water availability of roughly 1500 m3/capita/year, an inverse relationship can be identified between a country's cereal import and its per capita renewable water resources. Pakistan's annual water availability per inhabitant, which had already dropped to 1391 m<sup>3</sup> [33], is expected to reach around 800 m<sup>3</sup> by 2050 [64,65]. Looking ahead, the compounding water shortage in Pakistan would thus make the export of blue virtual water more unsustainable and questionable.

In contrast, the other water-stressed countries have been relying, although partially, on food imports to alleviate some pressure on domestic natural resources. For example, China's food trade saved 215.5 billion m3 of domestic water in 2015, which was equivalent to about 55% of the country's irrigation water used in the same year [27]. Libya and Israel, having extremely scarce water resources, almost entirely rely on global markets for their cereal supply, with small domestic production [3]. Similarly, for Spain, higher net imports of virtual water during a dry year (8415 million m<sup>3</sup> in 2005) as compared to a wet year (3420 million m<sup>3</sup> in 1997) are consistent with the country's relative water scarcity [66]. These countries seem to have steered their domestic production and trade in agricultural commodities in line with the relative scarcity of water resources, which does not seem quite true for Pakistan.

A silver lining to Pakistan's net exports of domestic blue water resources is that Pakistan has induced a small but positive global saving of blue water over the years. The savings resulted from higher blue water use efficiency in the production of exported commodities (mostly rice) than the water use efficiency of the export destinations. The net imports of green water, on the other hand, saved considerable volumes of domestic green water for Pakistan, while there were quite modest global savings.

Our findings suggest that Pakistan has consistently engaged in exports of much scarcer blue VW, while the VW trade is often determined by factors other than water, like economic and political ones. Virtual water trade alone, therefore, cannot be taken as a yardstick to steer the trade and production policies. It is the combination of agricultural structure adjustment towards high water use-value commodities and active promotion of trade in these commodities that can optimize agricultural water use. By adopting such adjustments, Pakistan can also engage its rural labor force into agriculture, in which the country is better endowed than many other countries in the world [67,68]. Thus, Pakistan can not only make full use of its comparative international advantage in the production of labor-intensive commodities, such as fruits and vegetables [69] but also can save water by improved irrigation technologies, for example, drip irrigation and spray irrigation systems, which are more suitable for horticultural crops [70]. Our suggestion is also in line with the ones found in existing literature where the authors suggest that when water becomes a significant constraint and economic cost factor to agricultural production, improving per unit water value through agricultural structure adjustment is one of the rational options to increase water use efficiency and farmer's income [22].

Importantly, we do not take these results to argue that Pakistan should stop the production or export of blue water-intensive crops like rice. On the contrary, we suggest that the relative scarcity of blue water should be included in the production of agricultural commodities. Such policy measures will not only improve the domestic water use efficiency but will also increase the economic benefits of virtual water exports.

Although the notion of virtual water does not provide unambiguous conclusions about international trade efficiency from a water resources perspective, it might foster cooperation among countries for improving water and land management globally. This is especially relevant when considering adaptation to climate change, together with production and consumption patterns. Virtual water could, therefore, encourage discussions on transboundary water resource management strategies.

A few caveats of our study need to be highlighted here. We used VW values from global model studies, whereas VW data from a national model for Pakistan might be a better choice for future studies on VW trade and footprint. The results for fruits and vegetables should be read with caution because the grouping and reporting for the two sectors are significantly broader than all the other commodities in this study. Future studies should include more detailed analysis for both the groups of commodities, although greywater footprint is also a pertinent topic covered in many studies on VW trade. However, because the environmental side of the virtual water trade (covering greywater) is beyond the scope of this study, future studies should include grey virtual water as well. Due to the unavailability of data on the cost of production for the studied commodities at the national level in Pakistan, we had to use production value for estimating the value of virtual water. Although this is not the ideal measure to account for differences in cost/benefit of various agricultural commodities, the measure still provides the closest proxy for evaluating the economic value of virtual water in agricultural production. Our study did not include the social costs that are often associated with blue water use in exportable commodities. Adding these costs is expected to further support our assertion regarding the inclusion of virtual water trade into domestic water policies. The analysis of water use in agriculture within the country is also an essential topic for prioritizing water use in the country. However, it is out of the scope of this study and should be covered in the future work
