Greenhouse gas emissions of hydropower in the Mekong River Basin

The Mekong River region in Southeast Asia is undergoing rapid social and economic development (Grumbine et al 2012), which has led to increasing demand for energy. The region is abundant in water resources and therefore hydropower is seen as an attractive energy source. Although hydropower is often considered as a climate-friendly energy option (Kaygusuz 2004, Edenhofer et al 2011, Dincer and Acar 2015), reservoirs are known to produce greenhouse gases (GHG), such as methane (CH₄), carbon dioxide (CO₂) and nitrous oxide (N₂0) (Demarty and Bastien 2011).

These emissions originate from the degradation of organic matter in the reservoir and they enter the atmosphere via diffusive flux and bubbling through the reservoir water surface, via degassing and diffusion from the reservoir tail waters, and via the reservoir drawdown area (Demarty and Bastien 2011, Varis et al 2012). The emissions depend on the characteristics of the natural systems that are inundated, on organic matter entering the reservoir from the catchment, and on reservoir characteristics and climate conditions. The emissions are further distinguished between gross and net emissions. Gross emissions are those that are directly measurable from existing reservoirs and net emission consider also the emissions from the reservoir area before inundation, which can act as a GHG source (e.g. natural waters) or sink (e.g. forests).

In the Mekong, the construction of large dams (dam height >15 m) for hydropower and irrigation started in the 1960s, and became more intensive in the late 1990s. Currently the basin has at least 64 large dams and more than 100 are planned (MRC 2015, WLE 2015). The total hydropower capacity of all the existing and planned large dams is over 60 000 MW.

The impacts of hydropower development on various aspects are increasingly well understood in the Mekong River Basin; these include impacts on hydrology (Lauri et al 2012, Cochrane et al 2014, Räsänen et al 2017), ecosystems (Ziv et al 2012, Arias et al 2014), sediment (Kummu et al 2010, Kondolf et al 2014, Manh et al 2015), fisheries (Baran and Myschowoda 2009, Stone 2016) and riparian people (Wyatt and Baird 2007, Keskinen et al 2016). At the same time, the hydropower’s GHG emissions have received less attention and are not systematically assessed, although concerns on potentially high emissions have been raised (Yang and Flower 2012).

Globally, GHG emission measurements have been reported since the 1990s. Barros et al (2011) collected existing CO₂ and CH₄ gross emission data from 85 reservoirs worldwide and found that emissions varied considerably between regions, being highest in the tropics. They estimate that the reservoir emissions correspond to 4% of the global carbon emissions from inland waters.

Hertwich (2013) estimated that the global average emission is 85 kg CO₂ MWh⁻¹ and 3 kg CH₄ MWh⁻¹, the most important predictor for emissions being reservoir area per kWh. Scherer and Pfister (2016) developed another statistical model, which they applied to ∼1500 reservoirs, estimating the global average emissions to be 173 kg CO₂ MWh⁻¹ and 2.95 kg CH₄ MWh⁻¹. Both estimates are below the emissions from fossil fuel power plants (380–1300 kg CO₂e MWh⁻¹) (Turconi et al 2013), but there is a high variability between reservoirs.

A review of emission measurements from tropical and equatorial reservoirs by Demarty and Bastien (2011) suggests that emissions can be large in warm climates particularly in cases in which vegetation and other easily degradable matter such as peat was not cleared and thus submerged by a reservoir. They used measurements from 18 equatorial and tropical reservoirs in which emissions varied between 2 and 4100 kg CO₂e MWh⁻¹. Demarty and Bastien (2011) further note that the emission measurements are too limited to take global position on the emissions of tropical reservoirs, given that there is a large number of dams in the tropics, and that there is a need to develop unified measurement protocols (see also Goldenfum 2012).

In the case of the Mekong, the research on GHG emissions from the reservoirs is very limited. To our knowledge, there exist published GHG emission measurements only from three reservoirs in Lao PDR, namely Nam Ngum 1 and Nam Leuk reservoirs (Chanudet et al 2011) and Nam Theun 2 reservoir (Deshmukh et al 2012, Deshmukh et al 2013). These three cases provide an important starting point for quantifying reservoir GHG emissions in the Mekong Basin, but there is no basin-wide understanding of the potential emissions.

The methods for estimating the GHG emissions from reservoirs on regional scale are limited, particularly in situations when GHG measurements are scarce or not available. UNESCO/IHA (2012) developed a GHG risk assessment tool that provides an estimate of the vulnerability of a reservoir on GHG emissions. The tool is based on existing global reservoir emission measurements and used, for example, by Kumar and Sharma (2016) for analysing the Tehri hydropower project in India. Another approach was developed by de Faria et al (2015), who applied a combination of models and existing measurements from the Amazon region to estimate emissions for planned reservoirs. More detailed modelling methods also exist (e.g. Weissenberger et al 2010), but those are often data intensive and not feasible for regional scale studies with limited measurements.

The quantification of GHG emissions in the Mekong has clearly major research gaps, and scientific information to support decision-making is lacking. Therefore, in this paper we aim to conduct the first assessment of the gross GHG emissions of the hydropower development in the basin, with focus on gross emissions of CO₂ and CH₄ through the reservoir water surface. Our aim can be divided further into two objectives: to estimate emissions of hydropower per energy unit, and to estimate emission fluxes from the reservoirs.

We decided to achieve our objectives by estimating the GHG emissions of 141 existing and planned hydropower reservoirs in the Mekong Basin using global statistical models from Hertwich (2013) and Scherer and Pfister (2016), considering them the most robust and well-documented methods for data scarce area with climate zones ranging from cool continental to tropics. Further, in contrast to the global assessments for a single year, this is the first large-scale study to assess emissions over a lifetime of 100 years.

With this we aim to provide an improved understanding of the GHG emissions of the hydropower development in the Mekong and thus provide information for directing future research efforts and for climate-smart decision making. Since we are analysing GHG emissions of hydropower generation, our analysis includes only the reservoirs that have documented to be equipped for power generation, and leave other reservoirs for further studies.

Zoom In Zoom Out Reset image size

In this study, we focus on the atmospheric gross emissions of CO₂ and CH₄ and their combined CO₂ equivalent (CO₂e) through the reservoir water-air interface. We excluded other emission sources such as degassing and diffusion from the reservoir tail water, as well as dam construction. The results are reported as emissions per energy unit [CO₂e kg MWh⁻¹] and emission fluxes [t CO₂e yr⁻¹] averaged over a 100 year lifetime. In the Discussion section, we also provide results averaged over a 10 year lifetime for the purpose of comparison with emission estimates presented in the literature. Below, the data and methods used for estimations are described.

2.1. Reservoirs

2.2. Emission models

The GHG emissions were estimated using the models from Scherer and Pfister (2016) and Hertwich (2013). Both are based on linear statistical models for CO₂ and CH₄ that are fitted against emission data from about 100 reservoirs worldwide. For estimating emissions per energy unit we used the equation from Hertwich (2013) for CO₂ and the equation from Scherer and Pfister (2016) for CH₄, and for estimating emission fluxes we used the equations from Scherer and Pfister (2016) for both CO₂ and CH₄. There were two reasons for using a combination of models. First, the model from Scherer and Pfister (2016) for CO₂ emissions per energy unit lacks an age factor and thus considers the CO₂ emissions per energy unit to be constant in time. The constant CO₂ emissions, however, do not fit to the general understanding of reservoir emissions (St Louis et al 2000, Abril et al 2005, Barros et al 2011, Demarty and Bastien 2011, Miller et al 2011, Hertwich 2013). Second, Scherer and Pfister (2016) compare their model to the model of Hertwich (2013) using various indicators and found that their model outperformed the model of Hertwich (2013) in the case of CH₄ emissions. For further model comparison see Scherer and Pfister (2016).

The model, we used for estimating emissions per energy unit (EpEU model, kg MWh⁻¹), is based on the following equations

$$\begin{equation} \begin{array}{l@{}l@{}l@{}} \rm log_{10} (CO_{2}) = 0.8+0.97 \cdot log_{10} (ATE) -0.006 \cdot\\ \rm AGE + 0.737 \cdot log_{10} (NPP)\, \, (Hertwich\, 2013) \end{array} \end{equation} \tag{ 1 }$$

$$\begin{equation} \begin{array}{l@{}l@{}l@{}} \rm ln (CH_{4}) = -9.81-0.75 \cdot ln (AGE) + 1.18 \cdot ln (ATE)\\ \rm + 4.50 \cdot ln (TMAX)\, (Scherer\, and\, Pfister\, 2016) \end{array} \end{equation} \tag{ 2 }$$

where ATE [km² GWh yr⁻¹] is the reservoir area-to-electricity ratio, NPP [g C m² yr⁻¹] is the net primary production, AGE [yr] is the reservoir age, and TMAX [°C] is the temperature of the warmest month.

The model for estimating emission fluxes (EF model, mg C m² d⁻¹) is based on the following equations

$$\begin{equation} \begin{array}{l@{}l@{}l@{}} \rm CO_{2} = 494.46-4.07 \cdot AGE + 8.09 \cdot ERR\,\\ \rm (Scherer\, and\, Pfister\, 2016) \end{array} \end{equation} \tag{ 3 }$$

$$\begin{equation} \begin{array}{l@{}l@{}l@{}} \rm ln (CH_{4}) = -12.84-0.03 \cdot AGE + 0.21 \cdot \\ \rm ln (A)-0.01 \cdot ERR + 4.88 \cdot ln (TMAX)\,\\ \rm (Scherer\, and\, Pfister\, 2016) \end{array} \end{equation} \tag{ 4 }$$

where ERR [t ha yr⁻¹] is the annual erosion per hectare, A [km²] is the surface area of the reservoir.

Zoom In Zoom Out Reset image size

The spatial data used in the equations are listed in table 1, and the reservoir specific values derived from spatial data are given in online supplement 2. The NPP, ERR and TMAX were estimated using 5 km buffers at the dam location, if the reservoir area was not available. The calculated emissions were further corrected as in Scherer and Pfister (2016): the CO₂ emissions were reduced by multpying with a factor of 0.87 and the CH₄ emission increased by multipying with factor of 1.4. to consider the negligence of carbon burial and methane ebullition (bubbling) in the measurements.

In this paper, we present the results as combined CO₂e, and as averages of EpEU and EF models. The emission fluxes were further converted from mg C m² d⁻¹ to t CO₂e yr⁻¹. For transforming CH₄ to CO₂e we used a Global Warming Potential (GWP) of 34 over 100 years. As a comparison, we also calculated power densities (W m⁻²) for each reservoir. Power densities are used in the Clean Development Mechanism (CDM) of the Kyoto Protocol (UN 2017) for implementing emission-reduction projects in developing countries that can earn saleable certified emission reduction credits. Hydropower projects with power densities above 4 W m⁻² are eligible for the CDM.

We further provide 20–80 percentile uncertainty intervals for emission estimates. These intervals were derived by comparing here estimated emissions of 22 global low-latitude (33°N–33°S) reservoirs with measured emissions. In the comparison, we calculated relative errors and fitted a log-normal probability distribution to those. This was then used to characterize the uncertainty of the emission estimates in the Mekong according to probability quantiles of 0.2 and 0.8. The global low-latitude reservoirs were considered to provide a reasonable reference for model errors for the Mekong, as it is located in similar latitudes (33°N–8°N). The measurement data were collected from Scherer and Pfister (2016) and was supplemented with six reservoirs from Asia of which three are located in the Mekong Basin (Chanudet et al 2011, Deshmukh et al 2013, Wang et al 2013, Zhao et al 2013, Kumar and Sharma 2016). The emission range of these Asian reservoirs (10–336 kg CO₂e MWh⁻¹) is close and on both sides of global average (187–273 kg CO₂e MWh⁻¹) and median (84 kg CO₂e MWh⁻¹) emissions (Hertwich 2013, Scherer and Pfister 2016), which further supports the use of global emission models in the Mekong Basin. The uncertainty analysis method is presented in detail in supplement 1, while uncertainty intervals given in results section in appropriate place and for all reservoirs in supplement 2.

Table 2. Estimates of highest CO₂e emissions per energy unit and largest CO₂e fluxes of the reservoirs in the Mekong River Basin. Emission estimates are given as averages over a 100 year lifetime. The 60% uncertainty interval is given in parentheses.

A. Highest CO2e emissions per energy unit (119 hydropower reservoirs, reservoirs with irrigation excluded)
Reservoir	Country	Commission year	Annual energy [GWh]	Reservoir area [km2]	CO2e emission per energy unit [kg CO2e MWh−1]
Average/median of 119 reservoirs	—	—	1735/507	69/16	120/26 (1–114)
Xe Bang Nouan	Lao PDR	2021	79	87	1990 (299–3449)
Lower Sre Pok 3 (3A)	Cambodia	TBD	1201	721	1400 (210–2423)
Lower Sesan 3	Cambodia	TBD	1310	727	1380 (209–2394)
Xe Bang Hieng 2	Lao PDR	2022	73	46	1030 (154–1778)
Nam Ngum 1	Lao PDR	1971	1025	369	670 (100–1154)
Duc Xuyen	Vietnam	TBD	181	77	600 (89–1031)
Lower Sesan 2	Cambodia	2019	1954	334	370 (55–632)
Nam Feuang 1	Lao PDR	2022	113	26	360 (54–622)
Lower Sre Pok 4	Cambodia	TBD	221	33	350 (53–606)
Sekong	Cambodia	TBD	557	94	320 (48–557)
B. Largest CO2e fluxes (all 141 reservoirs)
Reservoir	Country	Commission year	Annual energy [GWh]	Reservoir area [km2]	CO2e fluxes [103 t CO2e y−1]
Average/median of 141 reservoirs	—	—	1715/485	77/22	133/28 (<1–109)
Lower Sesan 3	Cambodia	TBD	1310	727	1810 (272–3136)
Lower Srepok 3 (3A)	Cambodia	TBD	1201	721	1680 (252–2911)
Sambor	Cambodia	TBD	11 740	620	1 330 (200–2307)
Stung Sen	Cambodia	TBD	124	434	1150 (173–1992)
Ubol Ratana	Thailand	1966	56	401	1250 (187–2158)
Dachaoshan	China	2003	5500	826	970 (145–1673)
Sirindhorn	Thailand	1971	90	289	760 (113–1307)
Jinghong	China	2009	5570	510	740 (110–1272)
Lower Sesan 2	Cambodia	2019	1954	334	710 (107–1236)
Nam Theun 2	Lao PDR	2010	6000	450	700 (105–1213)

We estimated the unweighted average and median emissions per energy unit of all 141 reservoirs to be 419 and 30 (where 20–80 percentile uncertainty intervals for emissions are 1–161) kg CO₂e MWh⁻¹, respectively. For 119 hydropower reservoirs, the average and median emissions are 122 and 26 (1–114) kg CO₂e MWh⁻¹, respectively, while for the 22 hydropower reservoirs with irrigation those are 2031 and 85 (8–634) kg CO₂e MWh⁻¹, respectively. The emissions for individual reservoirs vary considerably, ranging from 0.2–22 272 kg CO₂e MWh⁻¹ (figure 1).

The frequency distribution of the emissions is highly skewed (figure 2(a)). Thus, a median, instead of a mean, provides a better description for the central tendency of the emissions. The skewed emission distribution suggests that a large number of the reservoirs have relatively low emissions per energy unit, but there are a number of reservoirs with high emissions, too. In the case of hydropower reservoirs, the ten highest emissions per energy unit range 322–1994 kg CO₂e MWh⁻¹ (table 2). The reservoirs with high emissions tend to have a large reservoir surface area in relation to power capacity and are located in the warmer parts of the basin (table 2).

The emission fluxes of all reservoirs (figures 1(b) and 2(b)) indicate that the emissions vary considerably, too. The average emission flux for all reservoirs is 133 000 t CO₂e yr⁻¹ with median of 28 000 (587–109 100) t CO₂e yr⁻¹. The range of the ten highest emission fluxes is 700 000–1 800 000 t CO₂e yr⁻¹ (100 yr) (table 2b). All of these ten reservoirs have a very large surface area.

The results further suggest that existing reservoirs have lower emissions than the planned reservoirs (figure 2). The median emission per energy unit for existing hydropower reservoirs (53 of 119) is 18 kg CO₂e MWh⁻¹ and for planned hydropower reservoirs (66 of 119) 31 kg CO₂e MWh⁻¹. There is, however, a large uncertainty in the characteristics of the planned reservoirs.

The comparison of emission estimates to power densities shows that they have a strong correlation (r = −0.96; p-value <0.01) (figure S3). The average and median power densities for the 119 hydropower reservoirs are 54.3 and 10.9 W m⁻², while for 22 hydropower reservoirs with irrigation those are 6.0 and 2.3 W m⁻², respectively. Altogether 84 out of 119 hydropower reservoirs and 8 out of 22 hydropower reservoirs with irrigation have a higher power density than the CDM threshold of 4 W m⁻² (figure 1). Out of the 77 planned reservoirs 27 are above the 4 W m⁻² threshold. This threshold corresponds to emissions per energy unit of 87 kg CO₂e MWh⁻¹.

Total reservoir emissions (figures 3(a)–(b)) illustrate well the different phases of hydropower construction in the basin. First reservoirs were completed in 1966 and 1971, and the second, very intensive construction phase started in the early 2000s. According to the used databases and our analysis, the growth in emissions will continue at least until the year 2023 when altogether 111 reservoirs are built, should all existing plans be implemented. There are plans for 30 more large dams for which commission years are not known—their emissions are not included in figure 3. The 111 reservoirs, with known commission year, continue to emit GHGs in the post-2023 era with a rather high rate but decreasing trend.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The median emission per energy unit for the hydropower reservoirs varies over time (figure 3(c)). In 2000–2005, when several new reservoirs were built, the median emission was 120 (1–344) kg CO₂e MWh⁻¹, while for 2015–2020 the median emission decreases to 41 (1–134) kg CO₂e MWh⁻¹. If no more reservoirs are built after 2023, the median emission is estimated to decrease to 26 (1–113) kg CO₂e MWh⁻¹ by the 2050s (see figure S4 for results for a situation where no more reservoirs are built after 2017).

In this article, we provide the first GHG emission estimates for hydropower in the Mekong Basin. We found that the emissions range from 0.2–1994 kg CO₂e MWh⁻¹ over a 100 year lifetime with a median of 26 (1–114) kg CO₂e MWh⁻¹. The emissions per energy unit and emission fluxes were most strongly related to the following model predictors: area-to-electricity ratio, surface area and air temperature (table S3). The power density (W m⁻²)—used in CDM—also showed a strong relationship with our estimated emissions per energy unit (figure S3).

4.1. Comparison to global, low-latitude and local emission estimates

4.2. Comparison to other energy forms

4.3. High-emission future hydropower projects

4.4. Limitations and ways forward

This paper provides the first assessment of the GHG emissions of hydropower reservoirs in the Mekong Basin. The basin is undergoing extensive hydropower development, yet the understanding of hydropower’s GHG emissions is limited. We estimated the emissions of 141 existing and planned reservoirs using statistical global emission models, with focus on gross CO₂ and CH₄ emissions through the reservoir water surface.

Our results show considerable variation in the estimated hydropower emissions. The hydropower was found to have an emission range of 0.2–1994 kg CO₂e MWh⁻¹ over a 100 year lifetime with a median of 26 (1–114) kg CO₂e MWh⁻¹. Altogether, 82% of hydropower reservoirs (119) and 45% of reservoirs facilitating also irrigation (22) have emissions comparable to other renewable energy sources (<190 kg CO₂e MWh⁻¹), while the rest have higher emissions equalling even the emissions from fossil fuel power plants (>380 kg CO₂e MWh⁻¹). Several of these high -emission reservoirs are still in the planning phase. The results further show that the total basin-wide emissions (t CO₂e) of the hydropower development are considerable.

Our findings indicate that, although the reservoir emissions per produced energy may be low in the Mekong, hydropower cannot be considered categorically as low-emission energy. The emissions can reach the emission levels from fossil fuels power plants, depending on the characteristics and location of the hydropower project. High emissions were related most strongly to low area-to-electricity ratios, large reservoir surface areas and high air temperature. Therefore, each hydropower project should be carefully analysed for its GHG emissions. It is also obvious that careful removal of vegetation and other easily degradable organic matter from the inundated area of a reservoir is fundamental in minimizing GHG emissions from it.

Our findings should be considered as tentative, given that they are based on global models with high uncertainty. To improve the estimates, more measurements and better models are needed. Besides geophysical, ecological and social impacts, this paper highlights the importance of considering the climate impacts of hydropower development.

Authors declare no conflicts of interest. TAR received funding from Maa- ja vesitekniikan tuki ry., MK from Academy of Finland funded projects WASCO (grant no. 305471) and Emil Aaltonen Foundation funded project ‘eat-less-water’ and OV from Aalto University. We are grateful for Dr. Joseph Guillaume for his support in uncertainty analysis, and Prof. Jamie Pittock for his valuable comments on the paper.