A regional assessment of the water embedded in the US electricity system

Water consumption from electricity systems can be large, and it varies greatly by region. As electricity systems change, understanding the implications for water demand is important, given differential water availability. This letter presents regional water consumption and consumptive intensities for the United States electric grid by region using a 2014 base year, based on the 26 regions in the Environmental Protection Agency’s Emissions & Generation Resource Integrated Database. Estimates encompass operational (i.e. not embodied in fixed assets) water consumption from fuel extraction through conversion, calculated as the sum of induced water consumption for processes upstream of the point of generation (PoG) and water consumed at the PoG. Absolute water consumption and consumptive intensity is driven by thermal power plant cooling requirements. Regional consumption intensities vary by roughly a factor of 20. This variability is largely attributed to water consumption upstream of the PoG, particularly evaporation from reservoirs associated with hydroelectricity. Solar and wind generation, which are expected to continue to grow rapidly, consume very little water and could drive lower water consumption over time. As the electricity grid continues to change in response to policy, economic, and climatic drivers, understanding potential impacts on local water resources can inform changes.


Introduction
Electricity generation is responsible for substantial water resource consumption in the United States (US), especially when accounting for water used for processes upstream of the point of electricity generation (PoG). Water consumption for fuel extraction, processing, transportation, and conversion into electricity varies a great deal across fuel types, production techniques, and electricity generation technologies [1]. Water consumption for electricity also varies in space and time, even for similar fuels and generation technologies. Furthermore, electricity as a commodity is a product of a diverse generation portfolio, not an individual power plant, which means that quantifying water consumption associated with electricity is especially challenging [2]. Understanding the water intensity of electricity is relevant for diverse stakeholders, including utilities [3], sustainability planners and environmental analysts [4], and electricity consumers trying to understand the water footprints of their consumption patterns [5]. Specifically, understanding the water intensity of electricity at a regional level is important because of the regional nature of water availability and water management practices [6], and because thoughtful co-management of energy and water resources can lead to improved environmental outcomes [7,8].
Previous work on water use for electricity has focused on quantifying water use or water use rates for electricity generating facilities (i.e. power plants) using literature sources and government reports [9], heat budget models [10], and self-reported data [11]. One recent study from Argonne National Laboratory presents regional water consumption factors for North American Electric Reliability Corporation regions, accounting for 63% of US thermal generation in 2014 [12]. These studies represent valuable estimates for the electricity sector's water use, but they neglect water use occurring during production, processing, or transportation of fuels (referred to as 'upstream' stages in this letter). These upstream stages can be important contributors to the overall water intensity of electricity [13]. Other studies have addressed upstream stages [1,14], but they often rely on coarse or generic data or do not specifically characterize the water intensity of electricity.
This letter builds on prior work [1] to present regional estimates for the consumptive upstream and PoG water intensity of electricity in the US, using a 2014 base year and a regionalization consistent with common emissions intensity estimates. Specifically, we analyze the US electricity system at the generator level to estimate operational (i.e. water consumed as a result of ongoing electricity production) life cycle water consumption from resource capture through electricity generation for each generator reporting generation to the Energy Information Administration (EIA) [15,16], then aggregate water consumption across the regions used by the Environmental Protection Agency's (EPA's) Emissions & Generation Resource Integrated Database (eGRID) [17]. This database, which tracks greenhouse gas and other air pollutant emissions at the PoG for the US, is used extensively for modeling and analysis. Example applications include estimating the life cycle emissions impacts of fuels used to generate electricity [18], evaluating emissions trade-offs of fuel transitions for electricity [19][20][21][22][23], electrification studies [24][25][26], and urbanization [27] studies.
Notably, our use of eGRID regionalization means that the results presented here can easily be integrated to Life Cycle Assessment (LCA) studies in the US. LCA is a common environmental assessment method used for decision support, and it is heavily reliant on data. Regionalization [28] and data on water consumption [4] are both areas of need in LCA. Our work is compatible with life cycle inventory (LCI) data from the US LCI database [29], which is commonly used both directly and as a source for US data in other LCI databases. This analysis addresses essentially all US electricity generation, enabling our contribution of regionalized consumptive water intensities and total water consumption for both upstream and PoG uses. We also present details on the types of water being consumed (e.g. by source and quality), which aids in evaluation of the environmental impact of water use [30], and technological variability across regions that can inform similar analyses elsewhere.

Methods
We characterize total operational water consumption for electricity by US eGRID region, with a base year of 2014 (consistent with national estimates of water use for energy from Grubert and Sanders [1]). We note that temporal variability in water consumption is more associated with technology changes than, e.g. precipitation. That is, our technology-specific water intensity findings are unlikely to change substantially, but our regional water intensity findings will not apply if fuel and technology mixes change dramatically. Here, 'total' refers both to the electricity's full life cycle, from resource capture to power plant, and to water of all types. We distinguish among twelve water classifications combining water type (surface, ground, and reuse) and water salinity (fresh, brackish, saline, and not treatable by reverse osmosis) [1]. The focus on operational water consumption means that water embedded in infrastructure, such as power plants, pipelines, and other fixed assets, is not included. We focus on water consumption (water that is removed from its source and not directly returned) rather than water withdrawals (water that is removed from its source, whether or not it is directly returned) largely because withdrawals, which are driven by increasingly rare once-through cooling systems at power plants [1], are more consistently reported and are relatively well understood (though note that as of the 2015 data year, the US Geological Survey now reports consumptive use by thermoelectric power plants [31]). Also, consumption is a metric more consistent with other water footprinting exercises [2], as withdrawals are rarely used as a metric outside of the context of thermoelectric power plant cooling. To assess regional consumptive water intensity, we rely on a bottom up analysis of US electricity generation at the level of individual generators at power plants. Using federally reported data for grid-connected US generators with capacity greater than one megawatt [15,16] and recent studies of US consumptive water intensity by fuel [1,32] and generator type [11], we assign water consumption associated with the fuel (upstream of the plant) and direct use (at the PoG) to each generator. Then, we aggregate water consumption embedded in electricity over all reported generators and generation in a given eGRID region, using the eGRID 2016 boundaries [33]. This allows us to report both total water consumption and consumptive intensity (i.e. water consumption per unit of net electricity generation exported to the grid) by eGRID region.
We first calculate water consumption associated with producing, processing, and transporting power plant fuels (upstream processes), which might or might not occur in the same eGRID region as the power generation. In general, water consumption from upstream processes is calculated at the generator level by multiplying the generator's 2014 primary fuel consumption for electricity (from EIA Form 923 [16]) by a fuel-specific upstream consumptive water intensity factor based on values from Grubert and Sanders (2018) [1], which also uses a 2014 base year. Specifically, we multiply 2014 fuel consumption (FC) for generator, i, by the water intensity factor given by dividing total water consumption (WC) across water types, S, for upstream life cycle stages, j, for electricity fuel, F, [1] table 3) is calculated for each fuel, including fuel classifications and allocation schema (e.g. when a fuel is used for purposes in addition to electricity generation).
After calculating upstream water consumption associated with each generator's fuel, we calculate water consumption at the PoG-that is, consumption by the generators at each power plant. PoG water consumption occurs in the eGRID region where the power plant is located and is primarily associated with cooling, but it can also include water use e.g. scrubbers, human use, and fire suppression [1]. PoG water consumption (WC * ) is calculated by multiplying fueland technology-specific 2014 US water consumption rates (WCR) for generator classifications, c, from Peer and Sanders (2016) [11] and Grubert and Sanders (2018) [1] by net 2014 generation (GEN) per generator, i, per equation (2): Generator classifications, c, account for fuel, cooling system, and prime mover [11,15,16]. Data from [11] are used for natural gas, coal, and nuclear plants, while data from [1] are used for other technologies.
(Note that values from [11] are source data for natural gas, coal, and nuclear conversion values in [1]; this work recalculates due to a focus on regionalization rather than fuel averages as in [1].) Individual generators are related to cooling systems via relations to the boiler(s) at each power plant using EIA Form 860, Schedule 6-2 [15]. Generators linked to multiple cooling systems can have multiple classifications, c. For these generators, cooling water consumption is calculated by separating the fraction of generation cooled by each cooling system and using corresponding water consumption intensities from [11] for technology configurations matching the fuel, prime mover, and cooling systems at each multicooled generator. SM section 3 provides additional details, and SM table 11 shows water intensity by generator classification. To remain consistent with boundaries in LCA analysis e.g. emissions, our results do not account for water consumption downstream of the power plant (e.g. for waste management past the power plant gate) or water production from combustion of hydrocarbons, which can be non-negligible volumes in some cases, though see [1,34] for further discussion. Results from this study should be interpreted carefully with this boundary in mind. After calculating total generator-level water consumption (summing the results of equations (1) and (2)), we spatially join latitude and longitude data from EIA Form 860, Schedule 2 [15] with EPA shapefiles [33] to aggregate over eGRID regions.
Hydroelectric facilities are assigned to eGRID regions slightly differently, described further in SM section 2.1. Regional water consumption (WC R , in cubic meters, m 3 ) and regional water consumption intensity (WCI R , in m 3 per megawatt-hour (MWh) delivered to the grid ( GEN R S )) for each eGRID region, R, are reported based on the location of electricity generators, i, in region, R: Notably, these values do not correspond to totals and intensities associated with the location of water consumption or the location of electricity consumption. The nature of electricity systems is such that water is often virtually transferred as embodied water in a unit of input fuel or electricity [2], which means there are multiple ways to express water intensity by region. In this letter, we report consumptive water intensity as volume of water per unit of net electricity generation, or electricity first exported to the grid in a given region-this can also be seen as the virtual plus local water footprint of electricity at the power plant. That is, if water is consumed during fuel extraction in eGRID region 1, but that fuel is then exported to eGRID region 2, the water physically consumed in region 1 is assigned as embedded water for region 2 generation. Similarly, if electricity produced in eGRID region 2 is exported to eGRID region 3, embedded water in the electricity remains assigned to region 2 based on the power plant's location (figure 1).

Results and discussion
Overall, we find that consumptive water intensity associated with US electricity varies by a factor greater than 20 (i.e. from a low of 0.42 m 3 MWh −1 in AKMS to a high of 9.2 m 3 MWh −1 in AZNM) across eGRID regions (figure 2, table 1), confirming that spatial specificity is relevant for environmental assessments of electricity that include water consumption. SM table 1 defines eGRID region acronyms, and SM table 12 presents total volumetric water consumption for electricity by eGRID region.
Thermal power plants accounted for the majority (89%) of US electricity generation in 2014 [16]. These generators are fundamentally more water intensive than most non-thermal generating technologies, as they rely on cooling, which is typically provided with water. Coal, natural gas, and nuclear-powered electricity (97% of 2014 US thermal generation) collectively accounted for 71% of life cycle water consumption associated with US electricity production, about half of which is attributable to coal (38% of total life cycle water consumption and 39% of total generation). Reservoir-associated hydroelectricity also induced substantial water consumption through evaporation and accounted for essentially all of the water consumption associated with non-thermal generators (23% of electricity-associated water consumption and 6% of US electricity generation in 2014) (figure 3).
Although thermal plants drive total water consumption, fuel mix is the primary driver of regional variability in electricity-associated consumptive water intensity (here defined as m 3 of water consumed from fuel extraction through conversion per net MWh generated in the region) (table 1). A grid powered solely by wind and solar photovoltaic resources, for example, would not be as consumptively water intensive as a grid powered solely by water-cooled thermal generators. In general, thermal electricity is associated with higher consumptive water intensity than renewable resources typically eligible under renewable portfolio standards (RPS), like wind and solar, but not renewables in general or renewables relying on combustion.
This outcome is largely driven by high consumptive water intensity of reservoir-associated hydropower (non-thermal, no combustion, and renewable, but not generally accepted as an RPS resource) and nuclear power (thermal, no combustion, and nonrenewable) [1]. Fuel mix also drives regional variability within fuel categories. For example, regions with higher penetration by thermal renewables tend to have higher RPS renewable consumptive water intensity than those primarily reliant on low-water resources like wind and solar photovoltaics. SM tables 13, 14, and 15 show fuel mix and consumptive water intensities upstream, at PoG, and overall for each eGRID region.
Another main finding of this work is that water consumption upstream of the PoG is both non-negligible and highly variable by region (figures 3, 4). Using a net generation-weighted average, upstream water consumption accounts for about 38% of US water consumption for electricity generation, ranging from essentially none to essentially all of the consumptive water footprint of electricity in a given eGRID region (i.e. from 3% in SPNO to 100% in AKMS).
Part of the reason for variability shown in figure 4 is that upstream consumptive water intensity tends to be more site specific than PoG intensity. PoG intensities are usually thermally driven (see SM figures 3 and 4 for a view of generation and PoG water consumption by cooling type), while upstream intensities are often geologically or geographically driven [1]. For example, water consumption for oil, natural gas, coal, and uranium mining depends on the characteristics of the resource deposit, including rock permeability and water saturation levels, and evaporation from reservoirs for hydroelectricity or irrigation needs for biofuel crops depend on local climate. One major implication is that when water consumption is an environmental indicator of interest, consumption upstream of the power plant cannot be considered negligible without confirmation. This contrasts with typical practice regarding water withdrawals, as withdrawal intensities are dominated by power plant cooling systems [1].
A final general finding is that most (80%-95%) water consumption for electricity is fresh ( figure 5). Consistent with previous findings [35], this reliance on freshwater suggests that the electricity sector competes with other freshwater users (e.g. agriculture, ecosystems, and municipalities) for water access. Fresh surface water comprises the majority of water consumed for power plant cooling and all the water evaporated from reservoirs for hydroelectricity. When power plants use other water sources, it is typically because of Figure 1. Regional water consumption allocation approach, where α and β are fractions of water consumed for primary energy preparation (WC 1 and WC 2 ) and electricity generation (WC 3 ), respectively, and the red (upstream water consumption) and blue (PoG water consumption) arrows show water consumption attributed to a power plant from inside (solid) or outside (dashed) a given region. Each eGRID regional water consumption is calculated as all water embedded in electricity generated within the eGRID region's borders, which is not the same as water consumed within the region for electricity or water embedded in electricity consumed within the region's borders.
fresh surface water scarcity (e.g. groundwater or recycled water in AZNM, CAMX, ERCT, FRCC, and SRMV) or proximity to an ocean (e.g. FRCC, SRVC, RFCE, NEWE). Although surface water still comprises the majority of upstream water consumption, groundwater consumption is more common, largely because of geologically driven needs for water removal from resource deposits like coal mines and natural gas reservoirs [1]. Water consumption is defined in this work as the water removed from a source and not returned. Thus, groundwater removed during fuel extraction processes and not returned (e.g. water removed from aquifers during natural gas or coal extraction that is released to surface water or injected into different aquifers) is considered consumed, even if it is a nondiscretionary byproduct rather than a required water input.
3.1. eGRID regional variability in electricityassociated consumptive water intensity Average regional consumptive water intensity (m 3 of water consumed from fuel extraction through conversion per net MWh generated in the region) is highest for AZNM and lowest for AKMS. This range demonstrates the influence of the dynamics of evaporation from reservoir-associated hydropower plants on Figure 2. Regional water consumption intensity is highly variable across the country and is a function of not only technology mix, but also geography and climate (i.e. consumption intensity does not necessarily scale with regional generation). Each region is represented by a bar, colored by water consumption intensity (location displayed on the map with matching color). Total water consumption intensity and generation of each region are shown by the height and width of each bar, respectively, where a 1x1 square on the chart represents a volume of 1x10 9 m 3 . The bars are ordered from left to right in decreasing total volumetric consumption. overall consumptive water intensity (figure 2). Consumption from reservoir-associated hydropower, defined as evaporation that would not otherwise have occurred from the land associated with the reservoir (that is, net evaporation), is driven by regionally variable factors like weather and land cover. This work uses net evaporation values calculated for the United States, based on gross evaporation volumes calculated using a Penman-Monteith model less estimated evapotranspiration (ET) volumes associated with the most common landcover at proxy facility locations, based on National Land Cover Database mapping and landcover-specific ET coefficients. More discussion, and the full models, can be found in [32]. Hydropower thus has a broader range of possible consumptive water intensities than other resources in part because it can result in lower water consumption than would have otherwise occurred in a region, as when reservoirs replace water intensive land cover like wetlands. This negative consumption is observed in the northeast (NEWE, NYUP, RFCE), the forested southeast (SRSO), and the northwest (NWPP, RMPA) (table 1). By contrast, reservoirs in regions with limited regional landcover ET and high evaporation potential, notably the arid southwest (AZNM, ERCT) and the midwest (MROW, MROE, RFCM, SRMV, SRMW), experience very high evaporative losses from reservoirs (table 1). Note that these results reflect allocation of the evaporative burden of reservoirs across multiple purposes: see [32] for details.
Hydroelectricity is a major driver of variability on its own. A more typical profile for a given fuel is to have either consistently high or consistently low consumptive intensity, and regional variability is driven by variation in use of different types of fuel. For example, oil, nuclear, and coal are typically large water consumers (SM table 15), which is reflected in higherthan-average consumptive water intensity in eGRID regions with high contributions by those fuels (e.g. HIOA, HIMS, MROE, MROW, and SRTV, table 1). Regions with high relative fractions of natural gas generation (e.g. CAMX, NEWE, NYCW, RFCM) have noticeably lower water consumption intensities for fossil fuel generation (table 1), largely because many natural gas plants rely at least partly on gas turbines (versus steam turbines) that do not need to be water cooled.
High consumptive water intensity for electricity does not always mean high local water consumption or even high freshwater consumption, though. For example, Hawaii (HIOA, HIMS) is unusually reliant on oil, which is not produced locally and is less reliant on freshwater than most resources at the point of primary fuel production [1]. Similarly, high penetration by typically water-intensive fuels does not always mean high overall consumptive water intensity for the region. SRMW and SPNO both have high penetration by coal and nuclear facilities, but neither has higher than average overall consumptive water intensity due to the relatively low upstream water consumption intensity of coal used in these regions (see SM  table 13). . Water consumption (upstream, point of generation, and total), separated by fuel type for each eGRID region (ordered from west to east). Regional generation mix (i.e. fuel type) is a driver of regional water consumption variability at the PoG, whereas consumption upstream is driven more by geology (e.g. coal, gas, and oil resource extraction) and climate (e.g. consumption from hydroelectric reservoirs). For SPNO, high wind penetration (15% of 2014 generation) also lowers the overall average water intensity of generation, given that wind consumes essentially no operational water [1]. In general, just as some fuels tend to drive consistently high water consumption, wind and solar photovoltaic use tends to drive low water consumption (SM table 15). With the exception of HIMS (with its oil drivers) and MROW (with its hydroelectric drivers), all eGRID regions with wind penetration above 10% in 2014 had lower-thanaverage consumptive water intensity (ERCT, RMPA, SPNO, SPSO). For solar, the effect of higher penetration in lowering water consumption is somewhat masked by the fact that the largest solar-using regions (AZNM and CAMX, at 5% and 3% of 2014 generation, respectively) also use unusually water-intensive hydroelectric and geothermal resources. Growth in use of wind and solar photovoltaic resources is expected to continue to outpace growth in use of other resources [36,37], however, so these fuels might more clearly lower regional consumptive water intensity of electricity in the future. Variability in regional cooling system profiles is a secondary driver of the consumptive water intensity of electricity. Wet-cooled generation facilities display relatively low variation (i.e. same order of magnitude) in consumptive water intensity across generating technologies, but hybrid and dry-cooled facilities consume significantly less water per unit of output (SM table 11). Note that some variability in water consumption at the PoG is fuel mix driven, i.e. some generation does not require cooling, including wind, solar photovoltaics, and gas turbines using natural gas or biogas (SM table 11). Additional context can be found in SM section 4.1.

Data uncertainties and limitations
This work is largely based on modern estimates of water intensity from [1,11], which represent up-to-date, bestguess assessments of consumptive water use for the energy system. Comparing data from these publications to other published estimates is challenged by the fact that many other estimates are based on older information that does not reflect current technology. Thus, a precise quantitative estimate of uncertainty in the values presented in this letter is not possible. Major data limitations in this work are thus similar to limitations articulated by [1,11]. Most significantly, information about water consumption in general is not measured, centrally tracked, or published, so data are based on a combination of empirical, self-reported, and derived values based on physical relationships. Uncertainty is high, but the nature of that uncertainty is unknown given serious limits to data availability and quality. In general, estimates for processes that are not thermodynamically driven (e.g. process water for coal cleaning) Figure 5. Water consumption (upstream, point of generation, and total), separated by water quality (fresh, brackish, saline, not RO treatable) and type (RU=reuse, GW=groundwater, SW=surface water) for each eGRID region (ordered from west to east). Water consumption in every region (and nationally) is largely fresh, surface water, which is most heavily influenced by hydroelectric (upstream) and thermoelectric (at the PoG) for power plant cooling. Groundwater (mostly fresh) and reclaimed water (fresh and brackish) are consumed largely for resource extraction (e.g. coal and gas) and as an alternative source for power plant cooling, respectively. are more uncertain than estimates for processes that are thermodynamically driven, as the physical relationship between process and water consumption is less clear. Hydroelectricity-related consumption is thermodynamically driven, but data are highly sensitive to assumptions about allocation across multicriteria purposes and dam location [32]. PoG water consumption estimates are likely the most accurate, given some reporting requirements and multiple methods for triangulating true values based on physical and other relationships. For a given fuel and a given technology, estimates can be checked against values reported for other similar plants, thermodynamic laws, and EIA reports.
In addition to data quality-related uncertainty, this work is affected by the choice in [1] to conservatively overestimate freshwater consumption versus consumption of other water qualities, by assuming water quality is fresh when it is not known. The impact of this choice is estimated to be small, based on a general preference for freshwater in industrial and other contexts. Further, uncertainty associated with carriedthrough ambiguity related to units and definitions in the source data, discussed in more depth in [1,11], is also present in this work. Thus, overall data uncertainty is high but difficult to productively quantify. Note that given reporting requirements related to fuel use and outputs from electricity generators, uncertainty related to regional electricity production is very low.

Policy drivers and implications of variability in water consumption from electricity
The values presented in this work are based on technology and fuel mix conditions from 2014. These conditions are likely to change, however, in part due to changing policies. Policy can influence consumptive water intensity associated with electricity. As this work describes, fuel mix and cooling systems at thermal power plants affect water consumption intensity. Both are policy targets. For example, fuel mixes are increasingly affected by RPS in many US states, which aim to increase the amount and/or share of renewable generation. Federal tax credits similarly foster investment in various fuels. Cooling systems are affected by policy like section 316(b) of the Clean Water Act, which aims to reduce the impact of water withdrawals on aquatic organisms [38]. State implementation varies, leading to further differentiation. For example, California has implemented 316(b) by establishing cooling system technology-based standards aimed at reducing withdrawals of ocean and estuarine surface water, which has tended to shift once-through ocean water cooling to non-water cooled or recirculating systems in the state [39]. Through section 316(a), the Clean Water Act also sets operationally relevant limits on temperature increases due to water use for cooling [40,41], which influences choices about cooling system technologies. Although variability in regional consumptive water intensity for electricity is related to circumstance (e.g. fuel availability), geology (e.g. fossil fuel deposit characteristics), and geography (e.g. air and water temperature), it is also affected by policy.
Although this letter specifically investigates the US, many of the same conditions affecting water consumption from electricity in the US apply elsewhere [14,38,42]. Just as policy can drive water consumption associated with electricity, water consumption can drive policy, particularly in response to pressures like drought [43], strained water supply [44], rising water and air temperatures [38,41,45], and others. In considering potential future policy changes, one implication of this work is that water consumption from processes upstream of generators is both large enough and regionally variable enough to motivate specific consideration of localized data. PoG water consumption is relatively similar from place to place after accounting for fuel and technology, but accurately assessing upstream consumption requires more specific data. Future policy-motivated analysis might productively investigate regionally specific temporal patterns of water consumption from electricity, higher spatial resolution, different intensity metrics (e.g. water consumption physically in a region or water consumption per unit of electricity consumed in a region), and more regions (and see, e.g. [42,[46][47][48][49]) to expand on the existing literature on water for electricity.

Conclusion
Electricity systems consume large volumes of water. Based on this analysis and [1], upstream and PoG water consumption associated with electricity systems collectively accounts for about 5% of total US water consumption. Given that agriculture accounts for about 75% of total (nonrainfall) US water consumption [50], and given that much of the water associated with electricity systems is fresh surface water in relatively populated areas, this is a substantial volume. The volume of water consumed by electricity systems is driven by thermal power plants, and particularly by PoG water consumption from cooling systems. Consumptive water intensity varies by region, however, and this variability is largely driven by fuel mix. The consumptive water intensity of hydroelectricity in particular is highly regionally variable because of the influence of geography and climate on evaporation from reservoirs. Beyond hydroelectricity, fuel mix differentiates regions due to variability in water required for a given fuel (e.g. coal versus wind). Notably, water consumption upstream of the power plant is highly regionalized, while water consumption at the PoG tends to be similar for similar technologies. One implication is that studies of electricity's water intensity cannot easily adapt upstream water intensity values from one context to another even for the same fuel, while they can adapt PoG water intensity values if the technologies are similar. Policy affects consumptive water intensity of electricity, particularly via regulation about fuel mix and access to cooling water. In turn, changing drought, scarcity, temperature, and other conditions might tend to affect policy making regarding water consumption for electricity in the future.