The authors have declared that no competing interests exist.
Conceived and designed the experiments: RIM JJO WMM JF CR JVH JP. Performed the experiments: RIM JDO JJO. Analyzed the data: RIM JDO JJO WMM JF. Wrote the paper: RIM JDO JJO WMM JF JVH.
Rising energy consumption in coming decades, combined with a changing energy mix, have the potential to increase the impact of energy sector water use on freshwater biodiversity. We forecast changes in future water use based on various energy scenarios and examine implications for freshwater ecosystems. Annual water withdrawn/manipulated would increase by 18–24%, going from 1,993,000–2,628,000 Mm3 in 2010 to 2,359,000–3,271,000 Mm3 in 2035 under the Reference Case of the Energy Information Administration (EIA). Water consumption would more rapidly increase by 26% due to increased biofuel production, going from 16,700–46,400 Mm3 consumption in 2010 to 21,000–58,400 Mm3 consumption in 2035. Regionally, water use in the Southwest and Southeast may increase, with anticipated decreases in water use in some areas of the Midwest and Northeast. Policies that promote energy efficiency or conservation in the electric sector would reduce water withdrawn/manipulated by 27–36 m3GJ−1 (0.1–0.5 m3GJ−1 consumption), while such policies in the liquid fuel sector would reduce withdrawal/manipulation by 0.4–0.7 m3GJ−1 (0.2–0.3 m3GJ−1 consumption). The greatest energy sector withdrawal/manipulation are for hydropower and thermoelectric cooling, although potential new EPA rules that would require recirculating cooling for thermoelectric plants would reduce withdrawal/manipulation by 441,000 Mm3 (20,300 Mm3 consumption). The greatest consumptive energy sector use is evaporation from hydroelectric reservoirs, followed by irrigation water for biofuel feedstocks and water used for electricity generation from coal. Historical water use by the energy sector is related to patterns of fish species endangerment, where water resource regions with a greater fraction of available surface water withdrawn by hydropower or consumed by the energy sector correlated with higher probabilities of imperilment. Since future increases in energy-sector surface water use will occur in areas of high fish endemism (e.g., Southeast), additional management and policy actions will be needed to minimize further species imperilment.
In the United States (US), the energy sector is responsible for more than half of all water withdrawals
The future expansion of energy consumption and changes in the use of different sources could cause major changes in energy sector water use. Water withdrawal and consumption by the energy sector may increase in some areas
A number of recent studies have looked at how changes in the energy sector will affect water withdrawals or consumption
Synthesize information on the water-use intensity (m3GJ−1) of various energy production techniques, using high and low values of water-use intensity to provide a realistic range for each technique;
Present scenarios of future water withdrawal and consumption by the energy sector; and
Compare energy sector water-use with current patterns of threats to endangered fish species by major water resource region (
The 18 water resource regions of the United States, as defined by the 2-digit Hydrologic Unit Codes (HUC) of the USGS.
We present three scenarios of water use, based upon energy production scenarios developed by the US Energy Information Administration (EIA).
We first estimated water-use intensity for energy production techniques, drawing heavily from several of the published reviews of this topic
We recognized 12 energy production techniques (solar photovoltaic, solar thermal, wind, geothermal, biopower, hydropower, coal, nuclear, natural gas, power generated from municipal waste, biofuels, and petroleum), based upon the sectors used in the Annual Energy Outlook
In this paper, we define “water use” as any use of surface water or groundwater to produce energy, including water used for hydropower production and water used to irrigate bioenergy feedstocks. Note that other papers have categorized water use into three categories: “blue”, “green ,” and “gray”
We divide “water use” into two subcategories: water withdrawn/manipulated, the removal of water from a surface or groundwater source; and water consumption, the portion of water withdrawal that is not returned to the environment but is consumed by the process of energy production. This consumption can take several forms, including evaporation (e.g., in the cooling loop of a thermoelectric plant or from a reservoir), transpiration (e.g., irrigation water applied to energy crops), or incorporation into a product, byproduct or material of production (e.g., water used in biofuel production). Note that our definition of “withdrawal/manipulation” includes water that is removed from a river system only briefly, as water passed through a hydropower turbine located at a dam site. By classifying water used by hydropower as water manipulated, we are using a terminology that differs from the United States Geological Survey, which presents statistics on hydropower water use separate from water use for thermoelectric plant cooling. We adopted this different terminology because one of our primary goals in this paper is to present a full picture of the water used for energy production, including how that water use affects freshwater ecosystems. Dams can affect a river's flow regime, connectivity and water quality and are among of the leading sources of threat for aquatic species. Therefore, quantifying the volume of water that is run through hydropower dams' turbines (i.e., manipulated) provides relevant information on how water management by hydropower affects fish species.
We recognize two major parts of the production process in our estimate of water-use intensity (m3 of water per GJ of energy): material/resource acquisition and processing (e.g., mining coal and preparing it for use) (
Type | Withdrawal/manipulation (m3GJ−1) | Consumption (m3GJ−1) | Water-intensity varies by | EIA forecasts by | Notes | ||
|
|
|
|
||||
Solar PV | 0.486 | 0.549 | 0.061 | 0.161 | National | Elec. Producing regions | High withdrawal value: 60% mono-SI, 40% multi-SI. Proportional split based on |
Solar Thermal | 0.0825 | 0.36 | 0.024 | 0.105 | National | Elec. Producing regions | High values: 0.105 m3GJ−1 consumption for plant construction and O&M |
Wind | 0.047 | 0.089 | 0.011 | 0.019 | National | Elec. Producing regions | Withdrawal: Taken from |
Geothermal | 0.003 | 0.031 | 0.001 | 0.011 | National | Elec. Producing regions | Consumption data from |
Coal | 0.028 | 1.21 | 0.003 | 0.328 | National | Coal producing regions | Withdrawal: |
Nuclear | 0.083 | 0.392 | 0.047 | 0.159 | National | Elec. Producing regions | Withdrawal: |
Natural Gas | 0.033 | 0.153 | 0.025 | 0.036 | National | Natural gas producing regions | Withdrawal: |
Hydropower | 0.00 | 0.00 | 0.00 | 0.00 | National | Elec. Producing regions | Dam construction assumed trivial relative to water use for electricity production. |
Municipal Waste | 0.00 | 0.00 | 0.00 | 0.00 | National | Elec. Producing regions | Assumed municipal waste streams would have been created anyway, so no water for waste creation. |
Petroleum | 0.22 | 0.27 | 0.07 | 0.21 | National | Petroleum producing regions | |
Biopower | 0.00 | 0.00 | 0.00 | 0.00 | National except for energy crops, which are state-level | Biomass market | Assumed zero except for energy crops, because the waste would have been collected and stored anyway. Assumed all rain-fed for energy crop biomass market. |
Biofuel- corn | 19.4 | 24.3 | 16.4 | 19.7 | State-level |
Biomass market | High number is existing average, averaging across irrigated and non-irrigated acres. Low number is estimated value for 2035, with an increase in yield and a full-switch to pressure irrigation (and thus no gravity irrigation). |
Biofuel- soybean | 58.3 | 71.7 | 49.6 | 53.6 | State-level |
Biomass market | High number is existing average, averaging across irrigated and non-irrigated acres. Low number is estimated value for 2035, with an increase in yield and a full-switch to pressure irrigation (and thus no gravity irrigation). |
Biofuel- cellulosic | 0.0 | 0.0 | 0.0 | 0.0 | State-level |
Biomass market | Assumed all rain-fed. See text for details. |
Type | Withdrawal/manipulation (m3GJ−1) | Consumption (m3GJ−1) | Water-intensity varies by | EIA forecasts by | Notes | ||
|
|
|
|
||||
Solar PV | 0.004 | 0.021 | 0.004 | 0.021 | National | Elec. Producing regions | |
Solar Thermal | 0.58 | 1.06 | 0.58 | 1.06 | National | Elec. Producing regions | Withdrawal and consumption: range in |
Wind | 0 | 0.001 | 0 | 0.001 | National | Elec. Producing regions | |
Geothermal | 1.89 | 12.4 | 0.66 | 1.89 | Varies by mix of open and recirculating cool in each elec. Producing region | Elec. Producing regions |
|
Coal: once-through cooling | 21.1 | 52.6 | 0.35 | 1.23 | Varies by mix of open and recirculating cool in each elec. Producing region | Elec. Producing regions | Withdrawal: |
Coal: recirculating cooling | 0.35 | 1.23 | 0.31 | 1.23 | Varies by mix of open and recirculating cool in each elec. Producing region | Elec. Producing regions | Withdrawal: lower is from closed loop tower example in |
Nuclear: once-through cooling | 26.3 | 63.1 | 0.42 | 0.94 | Varies by mix of open and recirculating cool in each elec. Producing region | Elec. Producing regions | Withdrawal: |
Nuclear: recirculating cooling | 0.59 | 1.19 | 0.45 | 0.94 | Withdrawal: |
||
Natural Gas: once-through cooling | 7.89 | 52.6 | 0.11 | 0.35 | Varies by mix of open and recirculating cool in each elec. Producing region | Elec. Producing regions | Withdrawal: |
Natural Gas: recirculating cooling | 0.25 | 0.66 | 0.20 | 0.54 | Withdrawal: Low value from |
||
Hydropower | 1,811 (US mean) | 2,173 (US mean) | 4.6 (US mean) | 14.1 (US mean) | State-level | Elec. Producing regions | Manipulation: Calculated from head of dams listed in the National Inventory of Dams |
Municipal Waste | 6.6 | 16.7 | 0.1 | 0.5 | National | Elec. Producing regions | |
Petroleum: once-through cooling | 21.1 | 52.6 | 0.35 | 0.52 | Varies by mix of open and recirculating cool in each elec. Producing region | Elec. Producing regions | Withdrawal: |
Petroleum: recirculating cooling | 0.35 | 0.66 | 0.35 | 0.52 | Withdrawal: |
||
Biopower | 6.6 | 16.7 | 0.1 | 0.5 | Varies by mix of open and recirculating cool in each elec. Producing region | Elec. Producing regions |
For each part of the production process, we defined a high- and low-end estimate of water-use intensity (for both withdrawal/manipulation and consumption). Generally, the high-end estimate gives current levels of water-use intensity as reported in the literature, whereas the low-end estimate of water-use intensity is the lowest level reported in the literature or the lowest level likely with future technological changes (cf.
There is a large variation in water-use intensity, in terms of withdrawal/manipulation, between thermoelectric plants that use once-through cooling and those that use recirculating cooling (
We have included water used for the irrigation of bioenergy feedstocks in our calculations of material acquisition for biopower and biofuel production. There are three major feedstocks considered in our analysis: corn, soybean, and cellulosic.
Corn and soybean are the two feedstocks currently used for commercial biofuel production, and for these two our high-end estimate is simply a function of the current average state-level irrigation rate of the crops (m3/tonne). The advantage of using state-level estimates is that it accounts for the considerable variation in irrigation rates between corn raised in, for instance, Indiana (primarily rain-fed) and Nebraska (a significant portion irrigated). Our low-end estimate for corn and soybean in 2035 assumes continued gains in the yield of these crops with no additional inputs of water required, consistent with historical trends, plus a transition away from gravity-fed irrigation toward more efficient sprinkler systems, also consistent with historical trends. We note that rainfed biofuel crops also transpire significant amounts of water
The third major feedstock we consider is generic biomass used for either cellulosic ethanol or for biopower.Here, we recognized five sources of biomass, consistent with NREL research into price-supply curves for each of these sources
Finally, to estimate the amount of water saved when a unit of energy is not consumed, due to either efficiency gains or reductions in demand, we calculated the average water withdrawal/manipulation and water consumption per unit energy for both the liquid fuel sector and for the electricity sector. To derive these values, we first calculated total energy consumption for each sector in 2010 and divided by the total water use for each sector.
Our energy scenarios are taken from the EIA's Annual Energy Outlook (AEO) 2011
We present three water use scenarios, based upon energy production scenarios developed by the EIA. Much more information on these standards available in the AEO
The next phase of the analysis involved apportioning the water use predictions by various subregions among 18 major water resource regions for the contiguous US, as defined by the USGS Hydrologic Unit Code (HUC) system. Our general strategy was to use higher-resolution information on where material acquisition/resource processing or electricity generation occurs to partition the water use as accurately as possible. For projections into the future, we used information on both current and proposed energy facilities contained in the Ventyx database to partition the water use.
For material acquisition/resource processing, the method used varied by energy technique (
For electricity generation, we used the Ventyx database to calculate, for each energy production technique, the total electricity generation in each major water resource region. Since the location, as well as the capacity (MW), of most facilities is known with great precision, it is possible to calculate this accurately. Water-use was then partitioned among major water resource regions using this calculation.
Data on the status, source of imperilment, and geographic range of US freshwater fish species were taken from NatureServe
While water used in cooling of thermoelectric power plants is overwhelmingly from surface water, water used for irrigation is frequently obtained from groundwater
Water Resource Region | Average flow, million m3/yr (1901–2009) |
New England | 97,100 |
Mid-Atlantic | 133,600 |
South Atlantic-Gulf | 306,400 |
Great Lakes | 143,200 |
Ohio | 188,900 |
Tennessee | 58,000 |
Upper Mississippi | 96,700 |
Lower Mississippi | 111,300 |
Souris-Red-Rainy | 8,100 |
Missouri | 76,900 |
Arkansas-White-Red | 41,100 |
Texas-Gulf | 24,100 |
Rio Grande | 6,200 |
Upper Colorado | 16,700 |
Lower Colorado | 5,600 |
Great Basin | 15,600 |
Pacific Northwest | 223,700 |
California | 117,500 |
Next, we normalized surface water use by available water, to obtain the proportion of water used for energy production in each subregion. Specifically, we divided three metrics of water-use (hydropower water manipulation, non-hydropower energy sector surface water withdrawals, and energy sector surface water consumption) by average annual surface water availability in each water resource region
where
We tested two related hypotheses using logistic regression analysis. First, we tested to see if expert evaluation of the threats facing each species is consistent with our metrics of water use, simply examining whether fish species with a particular reported threat had higher relative water-use on the appropriate metric (e.g, species threatened by dams and hydropower water use). Second, we tested to see if the probability of fish species imperilment is positively correlated with one of our three metrics of water-use (hydropower water manipulation, non-hydropower energy sector surface water withdrawals, and energy sector surface water consumption), after accounting for species range size. For each logistic regression analysis, we first added the term for species range, then the term for normalized water use, and then tested for any interaction terms. At each step, the significance of each addition was tested using likelihood ratio tests. When comparing between models using different metrics of normalized water use (i.e., not nested models), we used Akaike's Information Criterion (AIC). To improve normality of variables and meet the assumptions of logistic regression, our metrics of normalized water-use and species area were log-transformed.
Current domestic energy production is dominated by coal and natural gas
US annual energy production (A) and electricity generation (B), in 2010 and in 2035 for three scenarios of future energy policy. Annual energy production is shown in petajoules and electricity generation is shown in terawatt-hours.
Coal, natural gas, and nuclear dominate current domestic electricity generation (
The water withdrawal/manipulation intensity of energy production techniques can vary over five orders of magnitude (
Water-use intensity (m3GJ−1) of US domestic energy production or energy conservation, in terms of water withdrawal (A) or water consumption (B). These water-use intensity estimates include water for material acquisition and processing, as well as for electricity generation where applicable. Errors bars indicate the range of our low and high water-use intensity estimates. The value labeled is the midpoint between these high and low estimates. The effect of energy conservation is shown using the energy mix in 2010. For hydropower, for display purposes typical consumption values are shown for more efficient and less efficient regions. Because hydropower water manipulation is more than an order of magnitude greater than water withdrawals for other technologies, hydropower is omitted in the top panel (A).
For thermoelectric power production (e.g., from coal, natural gas and nuclear energy sources), the major difference in water withdrawal/manipulation intensity is between once-through cooling (high water withdrawal/manipulation intensity) and recirculating cooling (low water withdrawal/manipulation intensity). Nuclear power has higher average water withdrawal/manipulation intensity because a higher proportion of nuclear plants are once-through cooling than other types of thermoelectric plants. Natural gas power has lower average water withdrawal/manipulation intensity because combined cycle gas turbine plants (the dominant natural gas power plant type) generally use less cooling water. Finally, renewable energy production technologies, such as solar and wind, have among the lowest water withdrawal/manipulation intensities of the technologies assessed.
Energy conservation (reduced energy consumption caused by increases in energy efficiency or reduced demand) would reduce US water withdrawal/manipulation. This effect is greatest for the electric sector, where every 1 GJ of electricity conserved would save 27–36 m3 of water withdrawal/manipulation (midpoint of range, 31.1 m3GJ−1). By contrast, the effect would be less for the liquid fuel sector (0.4–0.7 m3GJ−1 of liquid fuels saved, midpoint estimate 0.5 m3GJ−1). This results in part because so much of the energy in the US liquid fuel sector comes from petroleum, which is extracted abroad and hence does not figure into the calculations of US water withdrawal/manipulation. Another reason for this trend is that the use of cooling water in many thermoelectric plants is so high relative to the water used in the extraction of petroleum.
Water-consumption intensity trends among sectors have somewhat similar patterns to those for water withdrawal/manipulation intensity (
Biofuel production also has high water consumption intensities, due to the high fraction of irrigation water that is either lost to evapotranspiration or incorporated into plant biomass. Compared with the large differences in water withdrawal/manipulation intensities, there is little difference in water consumption intensities between once-through and recirculating cooling thermoelectric plants. The large differences in the intensity of water withdrawal/manipulation are offset because the vast majority of water used in once-through cooling is returned rather than consumed. Geothermal and solar thermal have similar water consumption intensities to fossil fuel technologies. However, solar PV and wind have much lower water consumption intensity.
Energy conservation would also reduce US water consumption. This effect would be similar in size for the liquid fuel sector, where every 1 GJ conserved would save 0.2–0.3 m3 of water consumption (midpoint of range, 0.25 m3GJ−1) and for the electricity sector (0.1–0.5 m3GJ−1 of electricity saved, midpoint estimate 0.3 m3GJ−1). The effect of energy conservation of liquid fuels on US water consumption is high, relative to the situation with US water withdrawal/manipulation, because a fraction of liquid fuels come from biofuels, and irrigation water used for biofuel production has a much larger consumption of water per unit energy than other energy production techniques.
Hydropower currently accounts for the largest total withdrawal/manipulation (1,851,000–2,222,000 Mm3) by far (
Water withdrawn (A) and consumed (B) for US domestic energy production, in Mm3, in 2010 and in 2035 for three scenarios of future energy policy. For each scenario, we show the value implied by our low and high water-use intensity estimates (
The rank ordering of energy technologies is different with regards to current water consumption (
National-level statistics of energy sector water use mask significant variation among major water use regions (
Water use by the energy sector in major water resource regions in 2010 (A) and 2035 (B), under the Reference Case. The size of the pie chart indicates the total water available (mean Mm3 per year) in major water resource regions. The pie chart is divided into three colors, based on energy sector water use (excluding hydropower production). Water not used by the energy sector is shown in blue, while water withdrawn but not consumed is shown in yellow, and water withdrawn and consumed is shown in red. Then, the number in each region indicates the amount of water used specifically for hydropower production divided by total water available.
Withdrawal/manipulation (million m3) | Consumption (million m3) | |||||
Water Resource Region |
|
|
|
|
|
|
New England | 4,340-11,216 | 213,207–255,848 | 0 | 116–310 | 96–294 | 0 |
Mid-Atlantic | 17,237-46,934 | 140,466–168,559 | 12–17 | 635–1,783 | 172–525 | 11–14 |
South Atlantic-Gulf | 23,497-60,122 | 123,343–148,011 | 57–69 | 789–2,604 | 423–1,294 | 49–56 |
Great Lakes | 14,512–38,488 | 116,730–140,076 | 28–38 | 403–1,340 | 222–678 | 25–31 |
Ohio | 20,255–57,933 | 45,794–54,953 | 16–22 | 1,163–4,797 | 235–717 | 14–20 |
Tennessee | 5,981–14,830 | 84,374–101,249 | 3 | 162–450 | 1,101–3,366 | 3 |
Upper Mississippi | 15,895–42,835 | 47,648–57,178 | 18–26 | 441–1,905 | 22–67 | 17–23 |
Lower Mississippi | 4,745–17,421 | 10,601–12,722 | 71–102 | 164–958 | 57–175 | 61–81 |
Souris-Red-Rainy | 48–127 | 1,128–354 | 10–13 | 6 | 1 | 8–11 |
Missouri | 7,752–26,302 | 165,476–198,572 | 1,331–1,716 | 351–2,557 | 139–426 | 1,156–1,395 |
Arkansas-White-Red | 3,709–12,868 | 63,121–75,745 | 213–270 | 219–1,493 | 690–2,110 | 184–219 |
Texas-Gulf | 8,936–31,733 | 17,192–20,630 | 124–156 | 327–1,569 | 78–237 | 108–127 |
Rio Grande | 3–29 | 4,613–5,536 | 46–63 | 8 | 30–91 | 40–51 |
Upper Colorado | 295–5,402 | 10,127–12,153 | 100–141 | 161–1,617 | 238–26 | 86–114 |
Lower Colorado | 855–2,353 | 14,333–17,199 | 9–13 | 52–209 | 510–1,558 | 8–10 |
Great Basin | 33–228 | 3,696–4,436 | 18–26 | 18–84 | 10–30 | 15–21 |
Pacific Northwest | 806–2,419 | 587,454–04,944 | 146–182 | 90–292 | 986–3,016 | 127–148 |
California | 5,129–19,232 | 202,466–42,959 | 63–82 | 214–504 | 818–2,502 | 55–67 |
As a share of available water, the largest withdrawal (excluding hydropower use) is in the Texas-Gulf and Lower Colorado water resource regions (
Relative to available water, consumption by the energy sector is generally low (
Our metric of greater normalized hydroelectric water manipulation (i.e., hydropower use through turbines divided by water availability) appears consistent with expert evaluation of the threat dams pose to fish species. Fish for which “dams/impoundments” were reported as a threat had an average normalized hydroelectric water manipulation of 0.85, whereas those fish species where it was not reported as a threat had an average normalized hydroelectric water manipulation of 0.76.
Water resource regions with greater normalized hydroelectric water manipulation have a greater proportion of imperiled fish species, after controlling for species range size and its interaction with normalized hydroelectric water manipulation (χ2 = 255.97, df = 3, P<0.001,
Probability of a fish species being imperiled, as a function of the normalized hydropower water manipulation (i.e., water used in turbines/available). Curves are shown for three range sizes (km2), corresponding to the 25, 50, and 75th percentile of fish species range sizes.
Predictor | β | SE β | Wald's χ2 | df | P | |
Intercept | 9.97 | 0.97 | 104.92 | 1 | <0.001 | NA |
LN(Normalized hydropower water manipulation) | −3.47 | 1.23 | 7.94 | 1 | 0.0048 | 0.031 |
LN(Total Range) | −1.06 | 0.095 | 126.32 | 1 | <0.001 | 0.347 |
Interaction | 0.37 | 0.12 | 9.05 | 1 | 0.0026 | 1.45 |
|
|
|
|
|||
Overall model evaluation: | ||||||
Likelihood ratio test | 255.97 | 3 | <0.001 | |||
Score test | 210.63 | 3 | <0.001 | |||
Wald test | 133.17 | 3 | <0.001 |
Note: Kendall's Tau-
Similarly, our metric of greater normalized energy sector water consumption (i.e., consumption divided by water availability) appears consistent with expert evaluation of the threat posed to fish species. Fish for which “excess withdrawals/consumption” were reported as a threat had an average normalized energy sector water consumption of 0.011, whereas those fish species where it was not reported as a threat had an average normalized energy sector water consumption of 0.008.
Water resource regions with greater normalized energy sector water consumption have a greater proportion of imperiled fish species, after controlling for species range size and its interaction with normalized energy sector water consumption (χ2 = 268.78, df = 3, P<0.001,
Probability of a fish species being imperiled, as a function of the normalized energy sector surface water consumption (i.e., consumption/available). Curves are shown for three range sizes (km2), corresponding to the 25, 50, and 75th percentile of fish species range sizes.
Predictor | β | SE β | Wald's χ2 | df | P | |
Intercept | −9.92 | 4.75 | 4.37 | 1 | 0.036 | NA |
LN(Normalized energy sector surface water consumption) | −4.44 | 1.07 | 17.30 | 1 | <0.001 | 0.012 |
LN(Total Range) | 1.08 | 0.47 | 5.20 | 1 | 0.023 | 2.94 |
Interaction | 0.48 | 0.11 | 19.92 | 1 | <0.001 | 1.6 |
|
|
|
|
|||
Overall model evaluation: | ||||||
Likelihood ratio test | 268.78 | 3 | <0.001 | |||
Score test | 221.34 | 3 | <0.001 | |||
Wald test | 128.97 | 3 | <0.001 |
Note: Kendall's Tau-
Care must be taken when extrapolating these correlative patterns to the future, because our regression model was fit to cross-sectional data not panel data. However, some insight into which types of species are most likely to be imperiled in the future can be gained by examining a projection of regression results. In general, species with small ranges that are in water resource regions with a projected increase in energy-sector water consumption are most at risk. Under the Reference Case, water resource regions with an increase in consumption include the Lower Mississippi, the Texas-Gulf, and the Lower Colorado. Note that a more realistic evaluation of the effect of new energy sector water use on a particular fish species would require an analysis of the new energy development within that species range, as well as knowledge of the species-specific sensitivity to alterations to hydrological conditions associated with energy sector water use.
Our study revealed strong correlative relationships between current energy-sector water use and the likelihood of fish species imperilment in the US. If this association holds into the future then we expect that projected increases in energy-sector water use have the potential to further threaten freshwater fish. Since federal and state energy policies affect the combination of technologies used to produce energy, which in turn affects the amount of water withdrawn and consumed for energy production, energy policy decisions plays an important role in the conservation of freshwater fish biodiversity into the future. Below, we discuss a few key aspects of energy policy that affect how much energy sector water use occurs, and then explore their implications for fish species imperilment.
Policies that limit total energy consumption, either through increased energy efficiency or incentives to reduce energy use, conserve water. Energy efficiency gains are greatest in the Extended Policies Case, where strict energy efficiency standards for buildings and appliances and tighter Corporate Average Fuel Economy (CAFE) standards would reduce energy production by 5,000 PJ below the Reference Case. Our calculations suggest this would save 33,900–54,300 Mm3 of annual withdrawal/manipulation (2,510–4,650 Mm3 of consumption). In general, policies that limit use of liquid fuels would cause a slightly greater reduction in water consumption per unit energy (m3GJ−1) than would policies that limit electricity use, largely because the consumptive water-use intensity of corn ethanol production is so high.
The effect of climate policy on energy-sector water use is complicated and varies by energy-production technology. The GHG Price Case would reduce total energy production by 4,750 PJ below the Reference Case, as higher electricity prices from fossil fuel sources would drive reductions in energy use. The GHG Price Case would also reduce water use by accelerating the retirement of old thermoelectric plants that disproportionately use once-through cooling. It would also shift some electricity production to renewable technologies that have relatively low water-use intensities. This would reduce withdrawals by thermoelectric plants by 1,740–91,000 Mm3 (1,180–9,000 Mm3 of consumption). Our study did not consider the effect of CCS technology on water use. Chandell et al.
All of our scenarios assume full implementation of the strong incentives for biofuel production mandated by the Energy Independence and Security Act of 2007 (EISA). Increased biofuel from corn and soybeans will increase the amount of water used for energy in the US, since both crops are occasionally irrigated. Note that our analysis assumed that new feedstocks for cellulosic ethanol are entirely rain-fed. If, however, feedstocks for cellulosic ethanol require irrigation, at least in some places in the US, then the water required for biofuel production may increase significantly.
The switch of thermoelectric plants from once-through cooling to recirculating cooling could substantially reduce water withdrawals. The biggest unknown here is how fast this shift will be, which is a function of market dynamics (i.e., how fast existing once-through cooled plants are retired) and policy regulation (i.e., if EPA regulations encourage existing plants to convert to recirculating cooling). Our calculations show that, in the Reference Case, conversion of all existing plants to recirculating cooling over a 20-year period would reduce annual water withdrawals in 2035 by 441,000 Mm3, relative to a continuation of the current gradual shift as existing once-through cooling plants are retired and new recirculating cooling ones come online. However, the switch from once-through cooling to recirculating cooling would not significantly reduce water consumption, which appears to have a greater impact, and is more predictive of native fish imperilment compared to water withdrawal. Moreover, the greatest reductions in withdrawals from the switch from once-through cooling to recirculating would be in the Northeast and Midwest, areas that have relatively low levels of fish endemism and imperilment.
The potential future impacts of energy-sector water use vary significantly by water resource region. Our statistical results show that fish species most likely to be imperiled have small ranges and are in water resource regions with high energy-sector normalized water consumption. The Southwestern US has both factors, with a large number of species with small ranges and high normalized water consumption, reinforcing the threat of water development on arid and semi-arid fish
Our results emphasize that policy decisions about energy are also decisions about water use, and that water sustainability and the health of freshwater ecosystems should be fully considered among the many factors that drive energy policy. The per-unit energy impacts of different energy technologies on water use, land use
We thank Joe Kiesecker and three anonymous reviewers for comments on earlier drafts of this manuscript. J. Slaats provided GIS support for this project. Staff at EIA were very helpful in answering technical questions. We thank all of the organizations that created data that made this analysis possible, including: NETL, DOE NREL, USGS, and Ventyx Corporation.