To assess the impacts of drought hazard, the first step is to inventory and analyze the environment that can be damaged (Di Mauro 2014). In general, exposure data identify the different types of physical entities that are on the ground, including built-up assets, infrastructures, agricultural land, and people, to cite but a few (Peduzzi et al. 2009). Drought exposure is very different from that of other hazard types. First, unlike earthquakes, floods, or tsunamis, which occur along generally well-defined fault lines, river valleys, or coastlines, drought can impact extended areas and can occur in most parts of the world (even in wet and humid regions), with the exception of desert regions, where it does not have meaning (Dai 2011; Goddard et al. 2003). Second, drought develops slowly, resulting from a prolonged period (from weeks to years) of precipitation below average or expected value at a particular location (Dracup et al. 1980; Wilhite and Glantz 1985). Therefore, droughts have an impact on different water use sectors as a function of the timing, duration, and amount of a precipitation deficit. For example, the immediate impacts of short-term (i.e., a few weeks) droughts might be a fall in crop production, poor pasture growth, or a decline in fodder supplies from crop residues for livestock farming. Prolonged water shortages (i.e., several months or years) may lead to effects such as lower earnings from agriculture, reduced energy production (e.g., reduced hydropower production, reduced cooling capacities for nuclear plants), problems in public water supply (both quantity and quality), reduced inland water transport, problems for tourism, job losses, food insecurity, and human casualties (Downing and Bakker 2000; Mishra and Singh 2010).
To address the diversity of drought impacts, we compute exposure by means of a nonparametric and noncompensatory Data Envelopment Analysis (DEA) (Cook et al. 2014; Lovell and Pastor 1999), as recently proposed by Carrao et al. (2016). This approach to drought exposure is multivariate and takes into account the spatial distribution of human population and numerous physical assets (proxy indicators) characterizing agriculture and primary sector activities, namely: crop areas (agricultural drought), livestock (agricultural drought), industrial/domestic water use (hydrological drought), and human population (socioeconomic drought). In the DEA methodology, the exposure of each region to drought is relative and is determined by a normalized multivariate statistical distance to the most exposed region.
Currently, drought exposure is computed on the basis of four spatially explicit geographic layers that completely cover the global land surface, namely: global agricultural lands in the year 2000 (Ramankutty et al. 2008); gridded population of the world, version 4 (GPWv4) (Balk et al. 2006; Deichmann et al. 2001; Tobler et al. 1997); gridded livestock of the world (GLW) v2.0 (Robinson et al. 2014); and baseline water stress (BWS) (Gassert et al. 2014a, 2014b).