Biodiversity Loss from Freshwater Use for China’s Electricity Generation

Electricity generation has two major, under-investigated impacts on freshwater biodiversity due to its water use: the consumption of freshwater and thermal emissions to freshwater. Here, we analyze the spatiotemporal freshwater biodiversity impacts of China’s electric power system and the driving factors for these impacts. We show that between 2008 and 2017, the freshwater consumption of electricity generation peaked in 2013 (13.6 Gm3). Meanwhile, the freshwater consumption factor of China’s electricity generation decreased from 3.2 to 2.0 L/kWh. However, due to increasing thermal emissions, the biodiversity loss via freshwater use increased from 1.1 × 108 in 2008 to 1.6 × 108 PDF m3 year. The overall biodiversity loss per unit of electricity generation decreased from 3.2 × 10–5 to 2.5 × 10–5 PDF m3 year/kWh. Biodiversity loss from thermal pollution is 60% higher than that driven by water consumption. Electricity transmission results in the shifting of biodiversity impacts across regions. The results show that 15% of total biodiversity loss was embedded in transmission networks. In terms of electrical power system drivers of biodiversity loss, the total generation was the main driving factor of the increase in loss (rather than shifts in generation type, for example). Our results indicate the necessity of assessing the biodiversity impacts of electricity generation and incorporating them into energy system planning.


S1 Estimates of water consumption
We assessed the provincial water consumption factors for thermal power and hydropower generation using the method in Jin et al. 1 . Based on the factors, the water consumption of electricity generation in each year is calculated as follows: In which, gives the national water consumption for electricity generation (m 3 );

S2 Characterization factors for water consumption
Water consumed for electricity generation is not returned to the river. The influence of reduced flow rates on aquatic biodiversity can be quantified with the global species-discharge model, an index of habitat space, feeding and reproductive opportunities. This model is developed based on native fish species and river discharges in various river basins 2 . This model assumes a positive correlation between the number of freshwater fish species and average river discharges at the mouth of river basins.
where R is the freshwater fish species richness and Q mouth,i is the annual average river discharge at the river mouth of basin i (m 3 /s).
The species-discharge relationship can be used to calculate characterization factors for water consumption that specify freshwater fish species loss per unit of reduced river discharge for river basins in different regions. Characterization factors (CF c ) for S3 water consumption reflect the impact of water use due to human activities on freshwater biodiversity loss.
• ( ℎ, • ) where FF i is the fate factor of river basin i, EF i is the effect factor of river basin i (PDF·m 3 ·yr·m -3 ), dQ mouth,i is the marginal change in water discharge at the river mouth in basin i (m 3 ·yr -1 ), dW i is the marginal change in water consumption by human activities in river basin i (m 3 ·yr -1 ), dPDF i is the marginal change in the potentially disappeared fraction of the freshwater fish species due to the marginal river discharge change dQ mouth,i and V i is the volume of river basin i (m 3 ). The dQ mouth,i /dW i is assumed to be equal to one, indicating that a change in water consumption is fully reflected in a change in water discharge at the mouth for that river basin.
The river volumes (m 3 ) for all river basins are calculated according to Hanafiah et al. 3 as follows: where V i is the water volume in river basin i (m 3 ), Q mouth,i is the discharge at the river mouth in basin i (m 3 ·yr -1 ), Li is the length of river i (m).
The characterization factors are calculated for these river basins. Specifically, Qiantang and Min rivers are the representatives of the Southeast river basin. The characterization factors of Qiantang and Min river basins are calculated for Zhejiang and Fujian provinces since they are the largest river basins of the two provinces, respectively 4, 5 . Talimu is the largest river in the Continental basin, and its characterization factor is calculated for this basin. In terms of Southwest, Brahmaputra is the largest river basin of Tibet, and its characterization factor is calculated for Tibet 6 . Nandu river is the largest river of Hainan province, and its S4 characterization factor is calculated for Hainan. The discharges at the river mouth and the river length are obtained from the Ministry of Water Resources 4, 7-9 .

S3.1 Thermal pollution from thermal power
In power plants with once-through cooling systems, water from a freshwater body is used to absorb heat from the working fluid in the condenser. The entire volume of heated water is then discharged back into the water body. In the Rankine cycle of steam-electric generating units, pumps and boilers add heat to liquid water, which is converted into steam during that process. The high-pressure steam then expands in the turbine producing power. Upon exiting the turbine, the steam passes through the condenser where heat is rejected from the system turning the working fluid into a saturated liquid, ready to re-enter the pump. To calculate the rate of heat rejected in each cycle, the difference in enthalpy of the working fluid on either side of the condenser must be multiplied by the steam flow rate. The thermal pollution to water bodies from thermal power is calculated using the method of Raptis and Pfister 10 .
The heat rejection rates of thermal power are assessed as follows: Where Q is the heat rejection rate (MW), LF is the load or capacity factor of electricity generating units, which are derived from Jin et al. 11 , m steam is the steam flow rate at the high-pressure turbine (kg/s), h b -h a is the difference in enthalpy of the working fluid on either side of the condenser (kJ/kg).
The steam flow rate can be calculated as: Reheat cycles can be added into the Rankine cycle, increasing the generation efficiency and thus reducing fuel inputs. When reheat cycles are employed, the steam passes first through a high-pressure turbine and, after being reheated, through a low-S5 pressure turbine. 94% of China's units with an installed capacity of 100-220 MW use a reheat system, whereas all 300-1000 MW units use a reheat system 12 . For a reheat system, the ratio (r) of the steam flow at the entry of low-pressure turbine to the steam flow at the entry of high-pressure turbine is inserted to scale the rejection rate: Where r=0.85 is used for China's units in this study, referring to Yan et al. 13 and Cheng et al. 14 . h a is related to the water temperature withdrawn for use in the Electric Power Plants Database 17 and our previous study 11 . The results of plant-level thermal pollution and its impacts are then aggregated to the provincial level.

S3.2 Thermal pollution from hydropower
Hydropower also produces heat during operation, though its thermal pollution is smaller than that of thermal power with a once-through cooling system because of its higher energy efficiency. The thermal emission of hydropower can be calculated as follows: Where HR is the heat emission rate of China's hydropower, referring to Xu et al. 18 and Yan and Hao 19 ; here, HR is 1.8%. The definitions of LF and C gross are the same as those in Equations 7-9.
There are approximately 47,000 hydropower stations in China 20 . It is infeasible to assess the thermal emission and biodiversity impacts at the plant level because of data limitations. We made assessments at the provincial level by changing equation 10 to S6 11: Where LF p is the provincial load or capacity factor of hydropower, C p is the provincial installed capacity of hydropower (MW). The values of LF p and C p are obtained from the National Bureau of Statistics 21 .

S4 Biodiversity impacts assessments
Biodiversity loss caused by freshwater consumption: Electricity generation can cause aquatic biodiversity loss because of its water use 22,23

S5 Estimates of water stress
The Water Stress Index is calculated according to Pfister et al. 24 , which is adapted from the water withdrawal-to-availability indicator by applying a logistic function to acquire continuous values between 0.01 and 1. The equation is as follows: Where WSI is the water stress index. WTA* is a modified WTA indicator considering the difference for watersheds with and without strongly regulated flows.

S6 Converting local impacts to global impacts
Kuipers et al. estimated global extinction probabilities (GEPs) based on species range sizes, species vulnerabilities, and species richness, indicating to what extent regional species loss in the respective area may contribute to global species loss. They generate them for marine, terrestrial, and freshwater species groups on the local (i.e., 0.05° × 0.05° grid) and ecoregion scale 26 . The regional fractions of freshwater species losses are then multiplied with the corresponding GEPs to calculate potential global fractions of extinctions: Where GBL i gives the potential global biodiversity loss in province i (PDF yr); V i is S8 the volume of the representative river in province i; GEP i is the global extinction probability in province i, calculated by aggregating the cell-level GEPs from Kuipers et al. 26 within province i.

Supplementary figures
Supplementary Figure S1. The decoupling state quadrant map corresponding to the decoupling degree. Here, BL r =ΔBL/BL t-1 , EG r =ΔEG/EG t-1 . This map is modified from Tapio 27 .