Estimation of virtual water contained in international trade products using nighttime imagery

Abstract

Freshwater that is consumed in the process of producing a commodity is called virtual water – it represents all water use contained in that commodity. In social systems, water resources can flow when commodities are traded from one region to another. Quantitative monitoring and assessing virtual water flow related to international trade products is an important issue to comprehensively understand the balance of global water resources. In this study we tested the potential of the Defense Meteorological Satellite Program's Operational Linescan System (DMSP-OLS) nighttime images in conjunction with the LandScan population dataset for estimation of virtual water contained in international trade products. Lit area (areal extent of night lights) and urban population were selected as proxies to estimate export virtual water (EVW), import virtual water (IVW), and traded virtual water (TVW) (summed EVW and IVW), respectively, on the national level. The results showed that IVW can be more accurately estimated than EVW regardless of lit area or urban population. Lit area is normally more appropriate for estimation of the virtual water of developed countries than those of developing countries, but urban population is more appropriate for estimation of the virtual water of developing countries than those of developed countries. Urban population is a better proxy than total population for estimations of virtual water. This study makes a negative finding in that there are relatively large underestimations for developed countries. Another negative finding is that neither lit area nor urban population can be used to estimate net import virtual water (NIVW).

Introduction

The rapid growth of global population has made limited natural resources become increasingly scarce. Water is one of the most vital resources to maintain human survival. The water-footprint theory that was first developed by Hoekstra and Hung (2002) and subsequently elaborated by other scholars (Chapagain and Hoekstra, 2004, Hoekstra and Chapagain, 2007, Chapagain et al., 2006) links water resources in the natural world with those of anthropogenic use. A country's water footprint is the total volume of freshwater used to produce commodities and services consumed by a country's residents (Chapagain and Hoekstra, 2007). Virtual water is one of the core concepts in the water footprint theory. The freshwater that is consumed to produce goods is called the virtual water contained in the goods (Hoekstra and Hung, 2005). For example, the virtual water of the agricultural production of wheat (usually expressed in m³/t) is the sum of the water volume demanded in the whole process of wheat production that is composed of three parts: the volume of blue water referring to the consumed ground and surface water, the volume of green water referring to the consumed water from rainfall, and the volume of grey water referring to the polluted water due to the production of wheat (Mekonnen and Hoekstra, 2010). When the goods are imported/exported from one country to another country via international trade, the virtual water resources begin to flow and are then redistributed spatially. Water resources flow not only in physical spheres of the Earth but also in human processes. Therefore, quantitatively monitoring and assessing virtual water flow is an essential issue in studies of the Earth's water resources.

Virtual water content (volume of virtual water required to produce a unit of product) varies greatly for different items (Liu and Savenije, 2008). The same products may vary greatly in virtual water content dependent on the products’ areas of origin (Chapagain and Hoekstra, 2007). For example, producing 1 t of wheat needs to consume 3710 m³ water in Morocco but only 566 m³ water in the UK (Mekonnen and Hoekstra, 2010). It becomes an arduous task to calculate virtual water related to international trade products at the global scale. Remote sensing is a powerful tool to quantitatively monitor global ecological and social systems. The Defense Meteorological Satellite Program (DMSP) is a department of defense (DoD) program and run by the Air Force Space and Missile Systems Center. The Operational Linescan System (OLS) is a group of visible and infrared sensors in DMSP satellites which can scan the global surface twice a day from an altitude of 830 km above the earth's surface with 3000 km wide swaths. The DMSP-OLS imagery was originally designed to monitor moonlit clouds (Doll, 2008). Croft (1973) first recognized the potential of using DMSP-OLS nighttime images to monitor large-scale anthropogenic effects. At present, the DMSP-OLS nighttime light imagery has been used to estimate gross domestic product (GDP), electric power consumption, fossil fuel carbon dioxide emission, and population on the national and regional levels (Chand et al., 2009, Doll et al., 2000, Doll et al., 2006, Ghosh et al., 2010a, Ghosh et al., 2010b, Letu et al., 2010, Lo, 2002, Raupach et al., 2010, Sutton, 1997, Sutton et al., 1997, Sutton et al., 2007). Zhao et al. (2011) found that lit area observed from the nighttime imagery has significant relationships with water footprints. Virtual water contained in international trade products is one of the major portions constituting a country's total water footprint (Chapagain and Hoekstra, 2007), so it should be feasible to assess virtual water's relationships to international trade products by using the DMSP-OLS nighttime imagery data.

Zhao et al. (2011) reported that nighttime light image data have potential to estimate water footprints but the estimate accuracy is not high enough for practical application. The LandScan population dataset is presently an intimate partner of the DMSP-OLS nighttime imagery when estimating the socioeconomic parameters (Ghosh et al., 2010b, Lo, 2002, Sutton et al., 2007). Therefore, the main objective of this study is to use the DMSP-OLS nighttime imagery data and the LandScan population dataset to estimate virtual water related to the international trade products. To accomplish this objective, we first describe a method of extracting two proxies, lit area and urban population, for estimating the virtual water. We then develop regression functions between lit area and urban population, and export virtual water (EVW), import virtual water (IVW), and traded virtual water (TVW) (summed EVW and IVW), respectively, on the national level. We analyze differences of estimation accuracy for different countries when using lit area and urban population as proxies, respectively. Finally, we discuss the reasons that lit area and urban population cannot be used to estimate net import virtual water (NIVW) and why higher estimation accuracy can be obtained when using the image-derived urban population data rather than official total population.