The water implications of generating electricity: water use across the United States based on different electricity pathways through 2050

As described in Clemmer et al (2012), we consider four electricity scenarios: one reference scenario and three carbon-constrained scenarios emphasizing different low carbon technologies. The reference scenario (scenario 1) is patterned off of EIA's Annual Energy Outlook 2011 (AEO 2011) reference case. Scenario 2 assumes that the United States meets the electricity sector's share of a cumulative economy-wide carbon budget of 170 gigatons of CO₂eq from 2012 to 2050 through economic competition of low carbon technologies. Scenarios 3 and 4 include the electricity sector's share of the US CO₂eq emissions budget plus additional targets for specific low carbon technologies. For scenario 3, we assume nuclear generation would grow from approximately 20% of the US electricity mix today to 29% in 2035 and 36% in 2050, while coal with CCS would grow to 15% of the generation mix by 2035 and 30% by 2050. For scenario 4, we assume aggressive deployment of energy efficient technologies and buildings would reduce US electricity demand 20% by 2035 and 35% by 2050 versus the reference case, while generation from renewable energy technologies (wind, solar, geothermal, biomass and hydropower) increases from 10% in 2010 to 50% in 2035 and 80% by 2050.

The national electricity capacity expansion model used to analyze these scenarios, the regional electricity deployment system (ReEDS), forecasts the deployment of supply side generation and transmission capacity for the power sector in the contiguous United States in two year increments to the year 2050 (Short et al 2009). ReEDS is a long-term capacity expansion and dispatch model that represents all major generation technologies, including coal, natural gas combined cycle, natural gas combustion turbines, fossil fuels with carbon capture and storage (CCS), nuclear, hydropower, wind, solar, geothermal, biopower and storage. It is unique among capacity expansion models for its highly discretized regional structure and statistical treatment of the impact of variability of wind and solar resources on capacity planning and dispatch. Further details of the electricity scenarios and modeling assumptions can be found in Clemmer et al (2012).

Water withdrawal and consumption coefficients are applied in a consistent functional unit of gallons per MWh of electricity generated and are adapted from Macknick et al (2012) as utilized in Averyt et al (2011). Only operational freshwater uses are analyzed by region, but use of saline water is also tracked. Water use rates are assumed to be constant over the duration of the analysis, though any changes in technologies' thermal efficiencies could alter those rates. We aggregate fuel and cooling type technologies from existing plants to match ReEDS technologies along with fuel and cooling system categories discussed in Macknick et al (2012). Median withdrawal and consumption factors are utilized for all technologies.

All existing coal and nuclear plants are matched with the generic category in Macknick et al (2012) which includes estimates of water usage for a variety of different types and vintages of plants. For concentrating solar power (CSP) technologies, which include both parabolic trough and power tower systems in ReEDS, we use dry-cooled parabolic trough systems water use rates for this analysis, with exception of the cooling system sensitivity analysis, where we considered wet-cooled parabolic trough systems. Geothermal technologies have large ranges of water use values, depending on the specific technology and whether they use externally sourced freshwater or the on-site geothermal fluids for cooling. For purposes of this study, we assume all new geothermal technologies are dry-cooled binary systems that required additional freshwater for makeup due to operational geofluid loss, per assumptions in Clark et al (2011). For the cooling system sensitivity analysis we consider wet/dry hybrid-cooled binary systems, which use more freshwater as part of their cooling systems. Due to data constraints, we use dedicated biopower plant values for plants that co-fire biomass with coal.

ReEDS does not consider combined heat and power (CHP) systems. Water use from CHP technologies are thus omitted, though CHP could play an important role in meeting energy efficiency reductions such as those envisioned under Scenario 4. Whereas substantial evaporation can occur from reservoirs that produce hydroelectricity, we have elected to not consider this water usage (withdrawal or consumption) due to the complexities in attributing water use to particular demands on these reservoirs, such as drinking water supply, recreation and flood control (Gleick 1992, Torcellini et al 2003, Pasqualetti and Kelley 2008).

Water use coefficients were applied to technology-specific generation output from each ReEDS power control area (PCA) region, based on 2008 generation data by technologies with specific cooling systems, as compiled and described in Averyt et al (2012). We consider only volumes of inland freshwater resources and existing wastewater utilization for regional analysis. Saline water volumes utilized in electricity production are tracked but excluded from the regional analysis. In cases where EIA does not report cooling systems for certain fuel types within a particular PCA region, the cooling system makeup for the fuel type with most generation is applied. If there is no existing generation in a PCA, new thermoelectric generation (with the exception of geothermal and CSP technologies) is assumed to be cooled with recirculating cooling systems. In addition, reflecting Environmental Protection Agency guidelines (EPA 2011), all new thermal generation is assumed to be equipped with recirculating cooling towers or dry-cooled systems, depending on existing structures in each PCA. For retirements of thermal generation, once-through cooled thermal plants within a PCA are assumed to retire prior to plants with recirculating cooling technologies. In coastal areas, retiring once-through cooled facilities utilizing saline water are assumed to be replaced by power plants utilizing freshwater in recirculating cooling systems.

Water withdrawal and consumption values are displayed according to USGS hydrologic unit code 2 levels (HUC-2) (Seaber et al 1987). Water values are calculated based on electricity generation by PCA region as calculated by ReEDS and then distributed to HUC-2 regions on a land area basis. Figure 1 shows the mapping of ReEDS PCA regions to USGS HUC-2 regions. For a PCA entirely within a HUC-2 region, all generation and water use in the PCA is attributed to the HUC-2. For a PCA that overlaps with multiple HUC-2 regions, generation and water use from the PCA are apportioned to overlapping HUC-2 regions based on the per cent of the HUC-2 region inside each PCA region. No attempts are made to project water demands from other sectors or any metrics of water availability.

Zoom In Zoom Out Reset image size

Results from the ReEDS modeling show a range of electricity generation technologies deployed between 2010 and 2050 under the four scenarios (figure 2). Under the reference case (scenario 1), natural gas becomes the dominant fuel for generating electricity, while renewable energy experiences more modest growth, to meet the projected increase in electricity demand and replace coal and nuclear plants retired in the ReEDS model analysis. Under the carbon budget scenarios (scenarios 2–4), conventional coal generation is largely phased-out by 2030. Under scenario 2, conventional natural gas generation increases in the early years to replace coal and reduce power plant carbon emissions, while renewable generation (particularly wind and solar) and natural gas with CCS make significant contributions in the last half of the forecast. Under scenario 3, coal with CCS and nuclear steadily increase after 2020, providing approximately two-thirds of total US generation by 2050, while renewable energy technologies provide most of the remaining generation. Under scenario 4, energy efficiency more than eliminates the projected growth in electricity demand, while wind and solar increase to meet a large share of the renewable energy target of 80% by 2050.

Zoom In Zoom Out Reset image size

Based on the results of the ReEDS model analysis, national-level water withdrawals steadily decrease from 2010 values under all scenarios (figure 3). Compared with 2010 withdrawals, 2030 annual withdrawals decrease by 10.6 trillion gallons (26.6%), 27.6 trillion gallons (69.2%), 26.7 trillion gallons (67.0%) and 27.7 trillion gallons (69.5%) for scenarios 1–4, respectively. By 2050, these scenarios have reduced water withdrawals from 2010 by 32.2 trillion gallons (80.7%), 37.9 trillion gallons (95.1%), 29.9 trillion gallons (75.2%) and 38.7 trillion gallons (97.0%), respectively.

Zoom In Zoom Out Reset image size

The universal reduction in withdrawals is largely due to the retirement of once-through cooled thermal generation and the construction of new facilities utilizing recirculating cooling technologies. In addition, high penetration of renewable technologies with minimal water requirements and energy efficiency reduce water withdrawals. Specifically, scenario 1 reductions are primarily due to retirements of some coal and nuclear plants, including those that utilize once-through cooling, along with the construction of new natural gas combined cycle facilities, which have a lower water withdrawal requirement than coal and nuclear facilities. Scenario 2 reductions are primarily due to the gradual phase-out of coal power plants and the increase in new natural gas combined cycle facilities between 2010 and 2030, and further coal and nuclear retirements along with high renewable penetration between 2030 and 2050. Scenario 3 reductions are a result of existing once-through cooled coal and nuclear facilities being replaced by newer coal and nuclear facilities that utilize recirculating cooling technologies, which have lower withdrawal rates. As coal and nuclear technologies still have higher withdrawal rates than natural gas combined cycle plants, however, by 2050 scenario 3 withdrawals are higher than those of scenario 1. Scenario 4 reductions are driven by a combination of energy efficiency, which reduces overall electricity demand and associated water requirements, and the high penetration of renewable technologies, which generally require less water than non-renewable technologies.

Despite substantial national-level reductions in withdrawals, regional withdrawal impacts vary greatly by scenario (figure 4).

Zoom In Zoom Out Reset image size

Certain HUC-2 regions in the northeast (1, 2), southeast (6) and northwest (17) mirror national trends and show substantial reductions in withdrawals under all scenarios for both 2030 and 2050. Other HUC-2 regions in the southeast (3, 8), Midwest (4, 5, 7, 9) and central (10) parts of the nation show only modest reductions in withdrawals by 2030 for scenario 1, with more substantial reductions by 2050. The southwest (13, 14, 15) and south central (11, 12) regions of the nation show only modest reductions in withdrawals by 2030 and 2050 under scenario 1. Two HUC-2 regions in the west (16, 18) show increases in freshwater withdrawals in both 2030 and 2050. These western regions (including parts of California, Nevada and Utah) show increases in water withdrawals in scenario 1 largely due to new electricity demands in the region. In addition, California shows increases in freshwater withdrawals largely due to once-through coastal facilities (which withdraw saline water) retiring and being replaced by inland (freshwater-cooled) energy generating sources. Regions 12 and 15 (parts of Texas and Arizona, respectively), show decreases in withdrawals by 2030 for scenario 3, but substantial increases in withdrawals by 2050 as more coal with CCS and nuclear technologies are adopted in these regions. Under scenario 2, withdrawals increase from 2030 to 2050 for region 18, but the 2050 value is only 57.1% of the 2010 freshwater withdrawals. Scenario 4 shows substantial reductions in withdrawals for all regions in 2030 and 2050.

National-level water consumption trajectories vary widely depending on energy scenario (figure 5). Compared with 2010, scenario 1 shows an increase in national water consumption of 8.5 billion gallons (0.6%) by 2030, but a decrease of 460 billion gallons (34.2%) by 2050. Scenario 3 leads to a 470 billion gallon (35.0%) reduction by 2030, though subsequent increases in consumptive uses lead to a net increase of 190 billion gallons (21.7%) from 2010 values. Scenarios 2 and 4 follow a similar decreasing trajectory until 2030, reducing consumptive uses by 810 (60.0%) billion gallons by 2030, yet diverge from 2030 to 2050. In 2050, total reductions in consumption for scenario 2 are 750 billion gallons (55.4%), whereas scenario 4 leads to reductions of 1.1 trillion gallons (85.2%) from 2010 values.

Zoom In Zoom Out Reset image size

National consumption trends are slightly different than withdrawal trends due to different relative withdrawal and consumption factors for the energy technologies and cooling systems deployed. Under scenario 1, consumption in 2030 increases from 2010 as a result of increased electricity demand being met by primarily natural gas combined cycle plants, with no substantial reduction in coal and nuclear generation. By 2050, coal and nuclear generation is substantially reduced and replaced with natural gas combined cycle generation, which has a lower consumption rate than coal and nuclear generation. Scenario 2 consumption decreases greatly from 2010 to 2030 due to coal plant retirements, and then increases from 2030 to 2050 as a result of building new natural gas combined cycle plants with CCS. Scenario 3 consumption declines sharply from 2010 to 2030 due to the retirement of conventional coal facilities and additional natural gas combined cycle generation. From 2030 to 2050, consumption increases due to increased deployment of coal with CCS and nuclear facilities utilizing recirculating cooling technologies, which have higher water consumption rates than other technologies. Scenario 4 national water consumption declines steadily due to a reduction in total energy demand and increased penetration of renewable technologies.

Similar to withdrawal impacts, regional changes in consumptive uses vary greatly by energy scenario and may differ from national trends (figure 6).

Zoom In Zoom Out Reset image size

In particular, scenarios 1 and 3 lead to changes in water consumption in many regions (figure 6) that may differ from national trends (figure 5). Compared with 2010 values, scenario 1 leads to a slight increase in consumptive uses in 2030; this national-level increase is a result of increases in consumption for certain HUC-2 regions representing the southeast (3, 8), southwest (16, 18), northwest (17) and central (10, 11, 12) parts of the nation. The majority of regions in the US show reductions in expected consumptive use. By 2050, although national levels of consumptive uses decrease, HUC-2 regions in the northeast (1) and the southwest (16, 18) show increases in consumptive uses. Increases in consumption in the southeast (3, 8) and central (10, 11, 12) regions are largely due to the retirement of once-through coal and nuclear facilities, which are replaced by new power plants operating with recirculating cooling systems. The switch from technologies utilizing once-through systems to technologies utilizing recirculating systems contributes to the decrease in withdrawals in the southeast (3, 8) and central (10, 11, 12) regions, yet increased consumption levels. Increases in consumption in the southwest (16, 18) and the northwest (17) regions in scenario 1 are largely due to the construction of new recirculating cooling systems in areas that currently have no generation or have predominantly hydropower generation. Under scenario 3, national levels of consumption decrease in 2030 (figure 5), yet HUC-2 regions 6 and 8 in the southeast show increases in consumptive uses (figure 6). By 2050, after substantial increases in nuclear and coal with CCS generation, consumption increases from 2010 values in a variety of regions in the Mid-Atlantic (2), Great Lakes (4), southeastern (3, 6, 8), central (12) and southwestern (13, 15, 18) parts of the country. All other regions (1, 5, 7, 9, 10, 11, 14, 16, 17) show decreases in consumption in 2050 from 2010 values. Scenario 2 shows reductions in consumption for all regions except regions 1 and 18 in 2050. Scenario 4 leads to reductions in consumption for all regions in both 2030 and 2050.

Certain renewable energy technologies (e.g., concentrating solar power and geothermal) deployed in arid regions can have higher water consumption rates than non-renewable energy technologies (Macknick et al 2011). We conducted sensitivity analyses associated with the choice of cooling systems for geothermal and concentrating solar power technologies in scenario 4 in 2050, which achieves the highest level of penetration for these technologies in the ReEDS model analysis. Due to resource availability, deployment of these technologies only occurs in western states. Under a wet/hybrid-cooled variation of scenario 4, where all concentrating solar power systems are wet-cooled and all geothermal systems are cooled utilizing wet/dry hybrid-cooling systems, national-level water consumption in 2050 increases by 80 billion gallons (41.5%) from scenario 4 with the dry-cooling assumption. Regionally, water consumption changes depending on the level of geothermal and concentrating solar power penetration in each region, with regions experiencing more concentrating solar power deployment having higher consumptive impacts (figure 7).

Zoom In Zoom Out Reset image size

Overall, deploying geothermal and CSP with dry-cooling lead to substantial reductions in water consumption in 2050 from 2010 values. Utilizing wet-cooled systems for these technologies increases water consumption over dry-cooling, but overall water consumption is still reduced from a 2010 baseline. The only region where the wet/hybrid assumption pushes total consumption over 2010 values in the model is in region 18 (i.e., California), but the likelihood of that occurring is small as California prohibits the use of freshwater for cooling in desert regions (CEC 2009).