This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 182.73.193.34

This content was downloaded on 13/10/2016 at 11:55

Please note that terms and conditions apply.

You may also be interested in:

Thermal effluent from the power sector: an analysis of once-through cooling system impacts on

surface water temperature

N Madden, A Lewis and M Davis

Horizontal cooling towers: riverine ecosystem services and the fate of thermoelectric heat in the

contemporary Northeast US

Robert J Stewart, Wilfred M Wollheim, Ariel Miara et al.

Riverine ecosystem services and the thermoelectric sector: strategic issues facing the Northeastern

United States

Ariel Miara, Charles J Vörösmarty, Robert J Stewart et al.

Energy sector water use implications of a 2 °C climate policy

Oliver Fricko, Simon C Parkinson, Nils Johnson et al.

Marginal costs of water savings from cooling system retrofits: a case study for Texas Power Plants

Aviva Loew, Paulina Jaramillo and Haibo Zhai

The ‘thirsty’ water-electricity nexus: field data on the scale and seasonality of thermoelectric

power generation’s water intensity in China

Daqian Jiang and Anuradha Ramaswami

Global thermal pollution of rivers from thermoelectric power plants

View the table of contents for this issue, or go to the journal homepage for more

2016 Environ. Res. Lett. 11 104011

(http://iopscience.iop.org/1748-9326/11/10/104011)

Home Search Collections Journals About Contact us My IOPscience

Environ. Res. Lett. 11 (2016) 104011 doi:10.1088/1748-9326/11/10/104011

LETTER

Global thermal pollution of rivers from thermoelectric power plants

CERaptis1,MTHvanVliet2,3 and SPfister1

1 Ecological SystemsDesign group, Institute of Environmental Engineering, ETHZurich, Zurich 8093, Switzerland2 Water Systems andGlobal Change group,WageningenUniversity, 6700AAWageningen, TheNetherlands3 International Institute for Applied SystemsAnalysis (IIASA), Schlossplatz 1, A-2361 Laxenburg, Austria

E-mail: raptis@ifu.baug.ethz.ch

Keywords: heat emissions, water temperature model, water temperature increase, grid-based, global, electricity generation, once-throughcooling

Supplementarymaterial for this article is available online

AbstractWorldwide riverine thermal pollution patterns were investigated by combiningmean annual heatrejection rates frompower plants with once-through cooling systemswith the global hydrological-water temperaturemodel variable infiltration capacity (VIC)-RBM.Themodel simulates bothstreamflow andwater temperature on 0.5°×0.5° spatial resolutionworldwide and by capturing theireffect, identifiesmultiple thermal pollution hotspots. TheMississippi receives the highest totalamount of heat emissions (62% and 28%ofwhich come from coal-fuelled and nuclear power plants,respectively) and presents the highest number of instances where the commonly set 3 °C temperatureincrease limit is equalled or exceeded. The Rhine receives 20%of the thermal emissions compared totheMississippi (predominantly due to nuclear power plants), but is the thermallymost polluted basinin relation to the totalflowperwatershed, with one third of its total flow experiencing a temperatureincrease�5 °Con average over the year. In other smaller basins in Europe, such as theWeser and thePo, the share of the total streamflowwith a temperature increase�3 °Cgoes up to 49%and 81%,respectively, during July–September. As the first global analysis of its kind, this work points towardsareas of high riverine thermal pollution, where temporally finer thermal emission data could becoupledwith a spatiallyfinermodel to better investigate water temperature increase and its effect onaquatic ecosystems.

1. Introduction

Riverine ecosystems worldwide face multiple pres-sures due to human interventions such as dams,channelization, deforestation, and irrigation, to namebut a few, as well as a plethora of industrial andagricultural waterborne emissions (Brooker 1985,Meybeck 1989, Bunn and Arthington 2002, Sweeneyet al 2004). When combined, physical and chemicalstressors can have grave impacts on aquatic ecosystems(Malmqvist and Rundle 2002). One major physicalstressor on riverine ecosystems is thermal pollution, astressor not only problematic in and of itself, but onethat can also aggravate the effects of chemical pollution(Heugens et al 2002, Holmstrup et al 2010). Conse-quently, there is much motivation to identify areas ofhigh thermal pollution.

One of the largest sources of freshwater thermalpollution can be found in the thermoelectric powersector (Hester and Doyle 2011). Power plants sta-tioned along rivers employ two main types of coolingsystems, namely once-through and recirculation(tower) cooling. In once-through cooling systems theheat absorbed by the cooling water during the steamcycle is directly rejected back into the river. In a cool-ing tower setup, on the other hand, most of the absor-bed heat is removed via evaporation and dissipatedinto the atmosphere. The heat contained in the peri-odic cooling tower blowdown is negligible comparedto the heat released in once-through cooling emissions(Stewart et al 2013).

In some rivers, especially those with a series oflarge thermal emissions sources, the water temper-ature increase resulting from cooling water emissions

OPEN ACCESS

RECEIVED

29 February 2016

REVISED

15August 2016

ACCEPTED FOR PUBLICATION

28 September 2016

PUBLISHED

12October 2016

Original content from thisworkmay be used underthe terms of the CreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2016 IOPPublishing Ltd

can be substantial (Madden et al 2013). Legislativemeasures in place in the US and in Europe, whichimpose thresholds for surface water temperature ortemperature increase, mean that numerous power sta-tions are forced to restrict their electricity output,often at times of peak demand, and are predicted to doso even more under climate change scenarios (Försterand Lilliestam 2010, van Vliet et al 2012b). Specifically,for the purpose of protecting the aquatic ecosystems,many US states enforce an upper temperature limit of32 °C for surface water (Madden et al 2013), while inthe European Union, water temperatures downstreamfrom the point of discharge should not exceed 1.5 °Cand 3 °C above natural temperatures (or 21.5 °C and28 °C) in salmonid and cyprinid waters, respectively(European Parliament and Council of the EuropeanUnion 2006).

Previous studies on the impacts of cooling waterthermal releases have focused either on the impactsfrom a single power plant (e.g. Wu et al 2001, Con-tador 2005, Verones et al 2010) or from multipleplants over large watersheds (e.g. Stewart et al 2013,Pfister and Suh 2015). In a recent study, the coolingsystems for the great majority of thermal power plantsworldwide were identified (covering 92% of the globalthermoelectric power installed capacity), and the ther-mal emissions for those stations with once-throughcooling systems were modelled (Raptis and Pfis-ter 2016). The available data from that work provide aunique opportunity to model power-related fresh-water thermal pollution by using these estimates as aninput to water temperature models, be they on a smallor large geographical scale. The objective of this workis to utilise these heat emission data together with themacroscale hydrological-river temperature modelvariable infiltration capacity (VIC)-RBM (Lianget al 1994, Yearsley 2012, van Vliet et al 2012b) toobtain the first global view of river temperatureincrease due to electricity generation.

2.Methods

2.1. Global heat emission datasetRaptis and Pfister (2016) based their work on the PlattsUDI WEPP (World Electric Power Plants Database)version March 2012 (Platts 2012), and analysed theheat emissions from all power plants with once-through freshwater cooling systems worldwide, interms of the technological, geographical and chron-ological (age and first year of commercial operation)patterns behind the stations. Vassolo and Döll (2005)used a 12 years older version of the WEPP database(2000) in their study and filled in the missing coolingsystem information by area-specific statistical andregression analyses. Raptis and Pfister (2016) usedGoogle Earth imagery (Google Inc. 2013) and followedamore laborious procedure similar to that outlined bythe USGS for cooling system identification from aerial

imagery (Diehl et al 2013). Having covered 92% of theglobal thermoelectric power installed capacity asreported in the WEPP database, there is confidencethat the subset characterised as once-through fresh-water and analysed by the authors (amounting to 19%of the global thermoelectric installed capacity) is, todate, the most complete global dataset of power plantsemploying once-through cooling. The authors char-acterise the type of freshwater body receiving theemissions, but also provide plant coordinates, deter-mined via individual power plant station informationand confirmed via Google Earth imagery (GoogleInc. 2013). The thermal emissions were obtained as adirect output of systematically solving the thermody-namic cycle for every generating unit (∼2400 unitsworldwide) based on the design thermodynamic datain each case, and differentiating between simple,reheat and cogenerative Rankine cycles. Raptis andPfister (2016) provide maximum heat rejection rates,as well as mean annual ones calculated by multiplyingthe maximum by mean annual capacity factors. Nosingle source of data that would enable the calculationof monthly capacity factors worldwide was found (theInternational Energy Agency2011 provides data thatwould permit the calculation of monthly capacityfactors only for OECD countries on a country and fuelgroup level). Clearly, monthly thermal emission rateswould have been preferable, but for the sake ofconsistency, mean annual heat rejection rates wereused in this work, assumed constant over themonths.

For the purpose of investigating the thermal pollu-tion of rivers globally, a subset of power stations reject-ing heat into rivers was taken from the datasetavailable in Raptis and Pfister (2016), excluding emis-sions into lakes (see section 2.2). A total of 1750 gen-erating units were selected, pertaining to 565 powerstations, which account for 12% of the global thermo-electric power installed capacity.

On a country-level, the share of the total thermo-electric power installed capacity represented by theselected power plants varies considerably (figure 1).The United States, China and France are the countrieswith the highest combined rates of riverine thermalemissions, occupying shares of 26%, 16%, and 12% ofthe global thermal emission rate respectively, but thepower stations with once-through cooling systemsalong rivers contribute only 17%, 8%, and 25% to therespective national thermoelectric installed capacity(which is dominated by stations employing tower andseawater once-through cooling). In countries wherethe capacity of the riverine once-through power plantsamounts to over 50% of the country-wide thermo-electric installed capacity, the national contribution tothe global heat emissions power stations is less than2%, and the combined contribution sums up to 10%of the global total.While the global and national sharesin terms of total thermoelectric capacity occupied bythe selected power stationsmight appear to be small insome cases, considering the completeness of the

2

Environ. Res. Lett. 11 (2016) 104011

dataset produced in the work of Raptis and Pfister(2016) and the accuracy of the georeferencing, thedataset used in this study in fact provides the to datemost complete quantification of global heat emissionsfrom the thermoelectric sector into rivers.

Figure 2 shows the spatial distribution of thepower plants and the magnitude of the thermal emis-sions, differentiating between the fuels used in eachstation (fuel group breakdowns are provided in Raptisand Pfister2016). Globally, over 46% of the thermalemissions into rivers are due to coal-fuelled power

plants and almost one third due to nuclear powerplants. According to the analysis carried out by Raptisand Pfister (2016) based on the WEPP database, themajority of power plants with once-through coolingsystems are identified in the northern hemisphere andalong major rivers. This is in line with the findings ofVassolo and Döll (2005), whose method for populat-ing cooling system data gaps resulted in few power sta-tions being assigned once-through cooling in Asia,Africa, Latin America and Oceania, except for the caseof China, where numerous power plants with this type

Figure 1.Country-based share of capacity of the riverside power stationswith once-through cooling systems selected for this study interms of the total national thermoelectric power capacity, colour-coded according to the total thermal emission rate resulting fromthese power plants in each country. The pie chart in the figure inlet shows the fraction these total thermal emissions account for interms of the global thermal emission rate for the top 12 countries and regions; countries appearing in a category of their own are notincluded in the regions they belong to (no double counting).

Figure 2.The global distribution of riverside power plants with once-through cooling systems and their thermal emission rate ascalculated in Raptis and Pfister (2016). The stations are colour-coded according to their fuel group and the pie chart in thefigure inletpresents the combined share of global thermal emission rate per fuel group.Other fuel groups includewaste heat, waste, biofuel andgeothermal,more details in Raptis and Pfister (2016).

3

Environ. Res. Lett. 11 (2016) 104011

of cooling systems were reported by Raptis and Pfister(2016). Rivers in central-east United States and Europeappear to be heavily burdened by thermal effluents, asdo the Nile and the Yangtze. 14 out of the 15 top ther-mally polluting stations are nuclear power plants, 11 ofwhich are located in Europe, while the greatest ther-mal emissions from a coal-fuelled power plant comefroma station inChina.

2.2. Global water temperature impactmodellingVIC-RBM was chosen for this global study assessingthe impacts on water temperature of cooling wateremissions from the thermoelectric industry, because itis appropriate for large-scale river temperaturemodel-ling. It comprises the VIC macro-scale hydrologicalmodel (Liang et al 1994) combined with the one-dimensional river temperature model RBM (Years-ley 2009). VIC-RBM is suitable for advection-domi-nated rivers, it allows for point-sources of heatemissions, and accounts for reservoir impacts onstreamflow and water temperature (van Vlietet al 2012a). The modelling framework has previouslybeen applied on a global level and on a 0.5°×0.5°spatial resolution (van Vliet et al 2013). Water temper-ature is modelled by solving the one-dimensional heatadvection equation in a semi-Lagrangian framework(Yearsley 2009). In this grid-based approach thevariables necessary for the solution of the thermalenergy budget equation are aggregated per grid cell innodes that are connected in a river network created bya digital elevation model (Yearsley 2012). The one-dimensional nature of the model means that onlyemissions into rivers should be considered since themodel assumes full mixing in each grid cell. Impacts ofthermal stratification of lakes and reservoirs are notincluded in this version of RBM. For this reasonemissions into lakes and reservoirs were excluded inthis study, as this would lead to an overestimation ofthe water temperature increase in the grid cells underquestion. Therefore, as input to VIC-RBM, riverineemissions were selected from the dataset of Raptis andPfister (2016).

The VIC-RBMmodel was applied using the eleva-tion, vegetation and soil characteristics as described in(Nijssen et al 2001), which was later implemented at0.5°×0.5° and using the DDM30 routing network(Döll and Lehner 2002) for lateral routing of stream-flow. The headwater temperatures (upstream bound-ary conditions) were estimated using the nonlinearwater temperature regression model of (Mohseniet al 1998). Heat emissions are included in RBM aspoint sources of advected heat as described in (vanVliet et al 2012b). In their study, impacts of heat efflu-ents and reservoirs were included with the aim ofimproving water temperature simulations in riversworldwide. Water temperature simulations of RBMwere previously evaluated by using observed recordsof daily water temperature of monitoring stations

worldwide from the UN Global Monitoring System(GEMS/Water) combined with other sources forselected river basins (see van Vliet et al (2012a) fordetails and results of model validation). In this globalstudy, thermal emission rates as calculated by Raptisand Pfister (2016) were aggregated on 0.5°×0.5°resolution and used as input to VIC-RBM. As an out-put, the mean monthly river temperature increaseabove the naturalised water temperature per grid cellwas obtained, where the naturalised water tempera-tures are mean monthly values of water temperature,averaged over the period 1971–2000 (van Vlietet al 2012a).

3. Results

Both globally and in the northern hemisphere thelowest and highest global mean monthly temperatureincreases were observed in April and October, so thesemonths were selected to showcase the results infigure 3. Large impacts of cooling water emissions onsimulated water temperatures were found in manyareas in theUS and in Europe (figures 3(b)(i–ii) and (c)(i–ii)). In October, when low, if not always minimumannual discharge is observed (see figure S1 in thesupplementary information (SI) for examples), a highconcentration of heat-rejecting power plants can leadto areas of extensive thermal pollution. In the US, theMississippi watershed stands out, in particular, theUpper Mississippi, Ohio and Tennessee subbasins(figures 3(b)(i) and (c)(i)). In Europe, the watertemperature increase equals or goes above the limit setby the EU (3 °C) over large stretches of the Rhine andWeser watersheds in October (figure 3(c)(ii)), in fact,this temperature increase threshold is met or exceededover half the year round in these rivers (figure 3(d)(ii)).Moreover, given that the entire area under considera-tion in Europe is marked as salmonid habitat by theIUCN (2014), an exceedance of 1.5 °C is alreadyconsidered a problem. Other rivers along which thereappears to be considerable increase above the naturalbackground temperature include the Danube andthe Po.

In some river basins with great amounts of heatemissions the impacts on the river temperature weremuch smaller, for example in the Nile and the Yangtze(figures 3(b) and (c)). To a large extent this can beattributed to the very high discharge of these rivers atthe location of the rejected heat and distinct impacts oflarge reservoirs, which diminish water temperaturerises. Strong effects of evaporative cooling might alsohave led to smaller water temperature increases inbasins with warm atmospheric conditions, such as theNile, compared to basins in the temperate cli-mate zone.

Figure 4 provides a comprehensive view of theimpact of thermal effluents in themost affected basins.Figure 4(a) presents the sum over months and grid

4

Environ. Res. Lett. 11 (2016) 104011

cells that the 3 °C temperature increase limit is equal-led or exceeded per basin. In essence it shows in abso-lute terms the cumulative number of instances over

the year where significant thermal pollution isobserved per basin, while not accounting for theextent of the river section affected. Figure S2(a) in the

Figure 3. For the selected areas with high emission densities: (i)Central-EasternUS and part of Canada, (ii)Europe andNorth Africa,(iii)China, the following analysis sequence is presented: (a) themean thermal emission rate aggregated on 0.5°×0.5° grid cells; theresultingmeanmonthly river temperature increase above the naturalisedwater temperature per grid cell in (b)October and (c)April;and (d) a representation of the number ofmonths the 3 °C temperature increase threshold set by EU regulation is equalled orexceeded, scaled by themaximum temperature reached (over the entire year) in the given grid cell. Pink areasmark salmonid habitats(IUCN2014), which are consideredmore vulnerable to temperature increase (European Parliament andCouncil of the EuropeanUnion 2006).

5

Environ. Res. Lett. 11 (2016) 104011

SI presents the same figure but for all watershedswhere an effect on river temperature was observed,that is, with at least a 0.1 °C temperature increase inany given grid cell in any given month. Figure 4(b)(also S2(b)) shows a breakdown in terms of the power-ing fuel of the total thermal emission rate received ineach basin, and figure 4(c) (also S2(c)) presents the

distribution in the form of boxplots of the temper-ature increase over the entire year in the grid cells thatwere affected in each watershed. Figures 4(a)–(c) coverdifferent aspects of the thermal pollution cause andeffect and are completed by figure 4(d) (also S3), whichpresents the variation over the months of the relativeextent of thermal pollution per basin, as a percentage

Figure 4. For themost affected basins (at least one instance of water temperature increase�3 °C in any given grid cell at any givenmonth of the year): (a) the cumulative number of instances (sumovermonths and grid cells) a 3 °Cor higher temperature increasewas recorded over the year; (b) the total thermal emission rate colour-coded according to the relevant fuel group; (c) the distributionof recorded temperature increases over the entire year for the grid cells where an impact was observed (i.e. at least one instance oftemperature increase) in the formof Tukey boxplots (McGill et al 1978); (d) themonthly variation of portions of the flowunaffectedand affected by thermal pollution (according to the defined temperature increase grades) as a fraction of the total watershed flow (thetemporal legend is provided in the plot for theMississippi basin). Figures S2 and S3 present the same plots for the complete range ofbasins affected (at least one instance of water temperature increase�0.1 °C in any given grid cell in any givenmonth).

6

Environ. Res. Lett. 11 (2016) 104011

of the total thermally affected flow in each watershed.Specifically, on a monthly scale, the discharge in allunaffected (0 °C temperature increase) and affected(�0.1 °C temperature increase and defined intervalsthereafter) grid cells was summed and related to thetotalflowof each basin.

According to the synthesis of figure 4, the Mis-sissippi receives the largest combined thermal emis-sions, with 62% of them coming from coal-fuelledpower plants and 28% coming from nuclear powerplant heat effluents. Evidently, due to the scale of theemissions and the basin itself, the absolute number ofinstances with significant thermal pollution is alsolarge (figure 4(a)). However, in relative terms, themore severely affected portion of the entireMississippiflow (temperature increase �3 °C) never goes beyond9% of the total flow over the entire year (figure 4(d)).The Rhine receives 20% of the heat emissions theMis-sissippi receives, almost 3/4 of which come fromnuclear power plants, and accumulates many instan-ces of severe thermal pollution over the year, namelyover half the number accumulated in the Mississippi(figures 4(a) and (b)). Being a smaller basin, it is there-fore more heavily affected than the Mississippi, with1/3 of its flow experiencing a temperature increase�5 °C and only 14% remaining entirely unaffected onaverage over the year (figure 4(d)). A total of∼3 GWofheat is released into both the smaller European basinsof the Schelde and the Po.However, the thermal pollu-tion intensity patterns vary between the two basins:46% of the total flow of the Schelde experiences temp-erature increases �3 °C consistently throughout theyear. The Po, on the other hand, experiences a surge inthe fraction of thermally polluted flow between July–September, when temperatures increase by 3 °C–5 °Cand 5 °C–10 °C in 35% and 14% of the total flow,respectively (figures 4(b) and (d)). A high thermal pol-lution intensity is noted between July–September alsoin theWeser basin, where over 80% of the total flow isaffected by temperature increases �3 °C as a result ofthermal emissions predominantly due to coal-fuelledpower plants (figure 4(b)). These results show that theimpact of thermal emissions is of varying magnitude,depending, among others factors, on the geographicconditions and the effect of watermanagement in eachbasin. In some river basins (in particular basins withlarge reservoir impacts like the Colorado) heat emis-sions have very little impacts on simulated watertemperature. Figure S3 presents the proportion ofthermally unaffected and affected flow for all basinsworldwide where an impact (even a small one) wasobserved, ordered in terms of decreasing share of dis-charge unaffected by thermal emissions. According tothis figure, 3 out of the 5 proportionally most ther-mally polluted watersheds are located in central Eur-ope, a fact that is troubling given that this area isdesignated as a salmonid habitat and almost half theproportion of flow in all of these basins over the entireyear is affected by temperature increases equal to or

exceeding the 1.5 °C temperature increase limitimposed by the EU for the protection of salmonids(European Parliament and Council of the EuropeanUnion 2006).

4.Discussion

The modelled global thermal pollution must be putinto the context of the uncertainties and limitationsbehind both the heat emissions as well as the modelused. As explained in the methods section, in terms ofcoverage, Raptis and Pfister (2016) obtained a near-complete view of the thermal power plants with once-through freshwater cooling systems, including exactlocations, allowing the subset of riverside stations to beidentified. However, as the authors point out in theirwork, the dataset these power plants were identifiedfrom, the WEPP database, itself has some weak points(Raptis and Pfister 2016).While formost countries thecoverage of the relevant thermoelectric sector ischaracterised as ‘complete’ (>95% of facilities), cover-age of fossil fuel fuelled power plants with a capacity�50MW in China is characterised as ‘comprehen-sive’, that is, including >75% of the facilities in thiscategory (Platts 2012). With coal power plants beingthe dominant means of thermoelectric generation inChina, this under-representation constitutes themajor shortcoming with regard to this work. A recentpublication gives some insight into what it mightactually mean specifically in relation to thermal powerplants with once-through cooling systems (Zhanget al 2016). The authors calculate the total installedcapacity of thermoelectric power plants with once-through cooling systems in China at 116.1 GW for theyear 2011, which is ∼40% higher than the 68.5 GW ofthe stations identified and analysed for their freshwaterheat emissions by Raptis and Pfister (2016). Conse-quently, there is an underestimation of the thermalemissions and of the modelled river temperatureincrease in theChinese basins too.

Electricity production and any ensuing thermaleffluents vary greatly over the months as a result ofsocietal demand in response to climatic conditionsand more. Reports are plentiful of power stationsbeing forced to regulate their operation to prevent theoverheating of freshwater bodies during heat waves(Förster and Lilliestam 2010, Cook et al 2015). TheThermoelectric Power and Thermal Pollution Model(TP2M) is a dynamic model that permits the model-ling of power plant adaptation strategies to watertemperature and flow variations (Miara and Vör-ösmarty 2013) and has been applied in studies centredin Northeastern United States (Miara et al 2013, Stew-art et al 2013). The use of mean annual heat emissionrates as an input to VIC-RBM means that such fluc-tuations were not captured in this work. This is a clearlimitation, as it means that there may well be caseswhere (a) the river temperature increase has been

7

Environ. Res. Lett. 11 (2016) 104011

underestimated, since months of peak electricity out-put have not been modelled (especially in countrieswhere these coincide with low river flow), or (b) theriver temperature increase has been overestimated,due to power plant operational adjustments enforcedby local legislation. Nevertheless, the results of thiswork are still valuable, since they constitute the firstconcerted effort to model in a consistent way theimpact of heat emissions from the thermoelectricpower sector on rivers globally, and point to severalthermally pollutedwatersheds.

In terms of the model used, it is likely that the0.5°×0.5° grid cell resolution of VIC-RBM was toocoarse to capture the effect of thermal emissions ininstances where the river discharge is particularly high.This might be the case in the Yangtze basin, forinstance, where the total thermal emission rate is the2nd highest globally, but the modelled river temper-ature increase appears to beminimal (figures 3(iii) andS2), while in reality the rejected heat might sig-nificantly affect certain sections of the river over lim-ited stretches from the point of release, which could becaptured by amodel withfiner spatial resolution.

5. Conclusions

As the first attempt to model the river temperatureincrease induced by cooling water emissions on aglobal scale, this work identifies several areas of highconcern. The modelled water temperature increasesare not trivial if one considers the length covered by ariver in a 0.5°×0.5° grid cell. Large, medium andsmall river basins are singled out as hot-spots ofthermal pollution with varying proportional andtemporal patterns. The Mississippi is prominent interms of absolute occurrences of thermal pollutionover its large basin,while smaller Europeanwatershedssuch as the Rhine, the Weser and the Po see a largefraction of their flow affected by thermal pollutionover the entire year, and particularly between July–October. Based on these findings more detailed andlocalised analyses are recommended for futureresearch, in which riverine thermal pollution patternsare investigated via a hydrological-river temperaturemodel operating on a finer spatial resolution. Thiswould enable the tracing back of the river temperatureincrease to the responsible power plants and allow forfurther patterns of thermal pollution to be revealed, interms of power plant technical characteristics, such asthe fuel used, the type of steam cycle employed, theconstruction year, and more. Such analyses couldassist in the compilation of recommendations andstrategies for environmental pollution mitigation,particularly if the scope of power plant environmentalimpacts is broadened, e.g. to include air emissions.Moreover, refining the annual thermal emission ratesto a shorter temporal scale (e.g. monthly) before usingthem in any hydrological-river temperature model,

would provide a more realistic pattern of the variationof thermal pollution over the year in many basins andis also recommended for future work. It would alsoenable the development of adaptation strategies forplant operation to mitigate the effects of thermalemissions during peakmonths.

It is important to note that the thermal pollutionimpactsmodelled in this work are constrained to thosecaused by the electricity industry. In urbanised water-sheds, heat contained in effluents from wastewatertreatment plants can also lead to significant streamtemperature increases (Kinouchi et al 2007), and socan the effluent discharges from other industries. Assuch, the results of this study become even more pro-minent, since the modelled thermal pollution con-stitutes the minimum expected anthropogenicperturbation of river temperature through the pro-duction of electricity.

It is also worth pointing out that, while the major-ity of the global heat emissions originate from powerplants entering commercial operation during the1980s or before, and while the use of once-throughcooling is generally being phased out in the UnitedStates and Europe (Raptis and Pfister 2016), this is notthe case everywhere around the globe. The use ofonce-through cooling has been increasing steadily inChina since the 1980s and the current estimation of itsshare in terms of installed capacity seems even to beunderestimated (Raptis and Pfister 2016, Zhanget al 2016). Finally, it is crucial to have current, regio-nalised estimates of all related impacts, includingfreshwater thermal pollution, in order to conduct aproper assessment of the trade-off in impacts if andwhen the once-through cooling systems are replacedby alternative technologies (Sanders 2015, Frickoet al 2016). Among some options, recirculating (tower)cooling overall increases water consumption, drycooling results in reduced power output (or increasedfuel consumption), and once-through cooling withseawater can be problematic when heat dilution at thecoast is limited or where reefs are concerned. Quanti-fying all current impacts, then, is crucial in order toavoid problem shifting, also when switching the powergeneration technology completely. Ultimately, thiswork points to hotspots of thermal pollution, wherethe replacement of old power plants might beprioritised.

Acknowledgments

We would like to thank the anonymous reviewers fortheir constructive comments and suggestions. We alsothank Stefanie Hellweg for her useful suggestions andEileen Raptis for proofreading the manuscript.Michelle van Vliet was supported by a Veni-grant(project 863.14.008) of NWO Earth and LifeSciences (ALW).

8

Environ. Res. Lett. 11 (2016) 104011

References

BrookerMP1985The ecological effects of channelizationGeogr. J.151 63–9

Bunn S E andArthingtonAH2002Basic principles and ecologicalconsequences of altered flow regimes for aquatic biodiversityEnviron.Manage. 30 492–507

Contador J F L 2005Adaptivemanagement,monitoring, and theecological sustainability of a thermal-pollutedwaterecosystem: a case in Sw Spain Environ.Monit. Assess. 10419–35

CookMA,KingCW,Davidson FT andWebberME2015Assessing the impacts of droughts and heat waves atthermoelectric power plants in theUnited States usingintegrated regression, thermodynamic, and climatemodelsEnergy Rep. 1 193–203

Diehl TH,HarrisMA,Murphy J C,Hutson S S and LaddDE 2013Methods for estimatingwater consumption forthermoelectric power plants in theUnited States ScientificInvestigations Report 2013–5188 U.S. Department of theInterior; U.S. Geological Survey

Döll P and Lehner B 2002Validation of a new global 30 mindrainage directionmap J. Hydrol. 258 214–31

European Parliament andCouncil of the EuropeanUnion 2006Directive 2006/44/EC of the European Parliament

FörsterH and Lilliestam J 2010Modeling thermoelectric powergeneration in view of climate changeReg. Environ. Change 10327–38

FrickoO, Parkinson SC, JohnsonN, StrubeggerM,vanVlietMTHandRiahi K 2016 Energy sector water useimplications of a 2 °C climate policyEnviron. Res. Lett. 11034011

GEMS/WaterGlobal EnvironmentMonitoring System(http://gemstat.org)

Google Inc. 2013Google EarthHester ET andDoyleMW2011Human impacts to river

temperature and their effects on biological processes: aquantitative synthesis J. Am.Water Resour. Assoc. 47 571–87

Heugens EHW,Hendriks A J, Dekker T, Van SNMandAdmiraalW2002A review of the effects ofmultiple stressorson aquatic organisms and analysis of uncertainty factors foruse in risk assessmentCrit. Rev. Toxicol. 31 247–84

HolmstrupM et al 2010 Interactions between effects ofenvironmental chemicals and natural stressors: a review Sci.Total Environ. 408 3746–62

International EnergyAgency 2011 Electricity Information Statistics(http://oecd-ilibrary.org/energy/data/iea-electricity-information-statistics_elect-data-en)

IUCN2014 Freshwater fish, the IUCNRed List of ThreatenedSpecies, Version 2014.4 (http://iucnredlist.org) (Accessed:11 September 2015)

Kinouchi T, YagiH andMiyamotoM2007 Increase in streamtemperature related to anthropogenic heat input fromurbanwastewater J. Hydrol. 335 78–88

LiangX, LettenmaierDP,WoodE F andBurges S J 1994A simplehydrologically basedmodel of land surfacewater and energyfluxes for general circulationmodels J. Geophys. Res. 99 14415

MaddenN, Lewis A andDavisM2013Thermal effluent from thepower sector: an analysis of once-through cooling systemimpacts on surfacewater temperature Environ. Res. Lett. 8035006

Malmqvist B andRundle S 2002Threats to the runningwaterecosystems of theworldEnviron. Conserv. 29 134–53

McGill R, Tukey JWand LarsenWA1978Variation of box plotsAm. Stat. 32 12–6 (www.jstor.org/stable/2683468)

MeybeckM1989The quality of rivers: frompristine stage to globalpollutionGlob. Planet. Change 1 283–309

Miara A andVörösmarty C J 2013Adynamicmodel to assesstradeoffs in power production and riverine ecosystemprotection Environ. Sci. Process. Impacts 15 1113–26

Miara A, Vörösmarty C J, Stewart R J,WollheimWMandRosenzweig B 2013Riverine ecosystem services and thethermoelectric sector: strategic issues facing theNortheasternUnited States Environ. Res. Lett. 8 025017

MohseniO, StefanHG andEricksonTR 1998Anonlinearregressionmodel forweekly stream temperaturesWaterResour. Res. 34 2685

Nijssen B,O’Donnell GM, LettenmaierDP, LohmannD andWoodE F 2001 Predicting the discharge of global riversJ. Clim. 14 3307–23

Pfister S and Suh S 2015Environmental impacts of thermalemissions to freshwater: spatially explicit fate and effectmodeling for life cycle assessment andwater footprinting Int.J. Life Cycle Assess. 20 927–36

Platts UPG2012UDIWorld Electric Power PlantsData Base,March 2012

Raptis C E and Pfister S 2016Global freshwater thermal emissionsfrom steam-electric power plants with once-through coolingsystemsEnergy 97 46–57

Sanders KT 2015Critical review: unchartedwaters? The future ofthe electricity- water nexusAm.Chem. Soc. 49 51–66

Stewart R J,WollheimWM,Miara A, Vörösmarty C J, Fekete B,Lammers RB andRosenzweig B 2013Horizontal coolingtowers: riverine ecosystem services and the fate ofthermoelectric heat in the contemporaryNortheastUSEnviron. Res. Lett. 8 025010

Sweeney BW, Bott T L, Jackson J K, Kaplan LA,Newbold JD,Standley L J, HessionWCandHorwitz R J 2004Ripariandeforestation, streamnarrowing, and loss of streamecosystem services Proc. Natl Acad. Sci. USA 101 14132–7

Vassolo S andDöll P 2005Global-scale gridded estimates ofthermoelectric power andmanufacturingwater useWaterResour. Res. 41W04010

Verones F,HanafiahMM, Pfister S,HuijbregtsMA J,Pelletier G J andKoehler A 2010Characterization factors forthermal pollution in freshwater aquatic environmentsEnviron. Sci. Technol. 44 9364–9

vanVlietMTH, FranssenWHP,Yearsley J R, Ludwig F,Haddeland I, LettenmaierDP andKabat P 2013Global riverdischarge andwater temperature under climate changeGlob.Environ. Change 23 450–64

vanVlietMTH, Yearsley J R, FranssenWHP, Ludwig F,Haddeland I, LettenmaierDP andKabat P 2012aCoupleddaily streamflow andwater temperaturemodelling in largeriver basinsHydrol. Earth Syst. Sci. 16 4303–21

vanVlietMTH, Yearsley J R, Ludwig F, Vögele S,LettenmaierDP andKabat P 2012bVulnerability of US andEuropean electricity supply to climate changeNat. Clim.Chang. 2 676–81

Wu J, Buchak EM, Edinger J E andKolluruV S 2001 Simulation ofcooling-water discharges frompower plants J. Environ.Manage. 61 77–92

Yearsley J 2012A grid-based approach for simulating streamtemperatureWater Resour. Res. 48W03506

Yearsley J R 2009A semi-lagrangianwater temperaturemodel foradvection-dominated river systemsWater Resour. Res. 45W12405

ZhangC, Zhong L, FuX,Wang J andWuZ2016Revealingwaterstress by the thermal power industry inChina based on a highspatial resolutionwater withdrawal and consumptioninventory Environ. Sci. Technol. 50 1642–52

9

Environ. Res. Lett. 11 (2016) 104011