Available online at www.sciencedirect.

com
JOURNAL OF
ENVIRONMENTAL
SCIENCES
ISSN 1001-0742
CN 11-2629/X
Journal of Environmental Sciences 2013, 25(11) 2229–2238 www.jesc.ac.cn

Spatial and seasonal variability of CO2 flux at the air-water interface of the
Three Gorges Reservoir
Le Yang1,2 , Fei Lu1, ∗, Xiaoke Wang1 , Xiaonan Duan3 ,
Lei Tong4 , Zhiyun Ouyang1 , Hepeng Li2

1. State Key Laboratory of Urban and Regional Ecology, Research Center for Eco-Environmental Sciences,
Chinese Academy of Sciences, Beijing 100085, China. E-mail: yangle3012@163.com
2. Zhejiang Forestry Academy, Hangzhou 310023, China
3. Bureau of Science and Technology for Resources and the Environment, Chinese Academy of Sciences, Beijing 100864, China
4. Ningbo Research Center for Urban Environment, Institute of Urban Environment, Chinese Academy of Sciences, Xiamen 361021, China

Received 09 January 2013; revised 07 March 2013; accepted 01 April 2013

Abstract
Diffusive carbon dioxide (CO2 ) emissions from the water surface of the Three Gorges Reservoir, currently the largest hydroelectric
reservoir in the world, were measured using floating static chambers over the course of a yearlong survey. The results showed that
the average annual CO2 flux was (163.3 ± 117.4) mg CO2 /(m2 ·hr) at the reservoir surface, which was larger than the CO2 flux in
most boreal and temperate reservoirs but lower than that in tropical reservoirs. Significant spatial variations in CO2 flux were observed
at four measured sites, with the largest flux measured at Wushan (221.9 mg CO2 /(m2 ·hr)) and the smallest flux measured at Zigui
(88.6 mg CO2 /(m2 ·hr)); these differences were probably related to the average water velocities at different sites. Seasonal variations
in CO2 flux were also observed at four sites, starting to increase in January, continuously rising until peaking in the summer (June–
August) and gradually decreasing thereafter. Seasonal variations in CO2 flux could reflect seasonal dynamics in pH, water velocity,
and temperature. Since the spatial and temporal variations in CO2 flux were significant and dependent on multiple physical, chemical,
and hydrological factors, it is suggested that long-term measurements should be made on a large spatial scale to assess the climatic
influence of hydropower in China, as well as the rest of the world.
Key words: CO2 emission; spatial variation; seasonal variation; water velocity; temperature; pH
DOI: 10.1016/S1001-0742(12)60291-5

Introduction Thomson, 2010). Thus it is necessary to investigate CO2

emission from hydroelectric reservoirs.
Carbon dioxide (CO2 ) is the foremost anthropogenic Spatial heterogeneity in CO2 flux was found for large
greenhouse gas (GHG) in terms of its radiative forcing reservoirs due to differences in the pre-flood landscapes
contribution to global warming, and has contributed about (Teodoru et al., 2011). Neglecting the spatial variability
70% of the enhanced greenhouse effect (Houghton, 2005). in CO2 emission from reservoirs may lead to more than
Hydropower has been regarded as green energy for a 25% error in estimation for tropical reservoirs (Roland et
long time (Chen et al., 2011), but this viewpoint was al., 2010). Moreover, CO2 emission was also found to be
tarnished in the last two decades (Demarty and Bastien, variable depending on the season of the year due to the
2011), because a large amount of CO2 would be generated variations in temperature (Xing et al., 2005; Wang et al.,
when rivers were dammed to create the reservoirs for 2011; Diem et al., 2012). CO2 emission from reservoir
hydroelectricity (St Louis et al., 2000; Fearnside, 2011). surfaces was often reported to be influenced by temper-
It was estimated that 48 Tg C-CO2 was emitted from ature. Respiration and primary production in the water
hydroelectric reservoirs every year, which accounted for column were strongly influenced by temperature, but the
9% of the total natural lake emissions (Barros et al., 2011). former was always reported to have a stronger response,
The average CO2 emission flux in hydroelectric reservoirs thus resulting in a net increase in CO2 production under
is 21.6% of that in soil respiration (Bond-Lamberty and warmer conditions (Mendonça et al., 2012). In addition,
CO2 flux was also reported to be influenced by pH (Soumis
* Corresponding author. E-mail: feilu@rcees.ac.cn
2230 Journal of Environmental Sciences 2013, 25(11) 2229–2238 / Le Yang et al. Vol. 25

et al., 2004; Tremblay et al., 2005; Therrien et al., 2005), N

chlorophyll-a (Zhao et al., 2011), rainfall (Abril et al.,
2005), and dissolved organic carbon (Tadonléké et al.,
2005). Therefore, the spatiotemporal variations in CO2
flux and their potential influencing factors need special Wuhan Sandouping
attention in CO2 emission from reservoirs. Yunyang Zigui
The Three Gorges Reservoir (TGR), which was im- Three Gorges Dam
pounded in 2003, has a length of 660 km and an area of
1084 km2 when the water level reach 175 m. The water
Legend
level fluctuates between 145–175 m throughout the year, Three Gorges Dam
and the water falls 80–110 m when passed through the Sampling site
turbines in the Three Gorges Dam (TGD), producing 8.47
× 1010 kWh of electricity in 2010, making the TGR the Fig. 1 Locations of the sampling sites in the Three Gorges Reservoir.
largest hydroelectric producer in the world (Yang et al.,
2012a). Greenhouse gas (such as CH4 ) emission from the three sampling points were selected at different water
TGR has aroused wide debate because of the remarkable depths on each site. To show the seasonal variation, CO2
spatial-temporal variations in CH4 emission (Chen et al., emissions from all the sampling points were measured
2009, 2011; Lu et al., 2011; Yang et al., 2013; Zhao et al., from 9:00 to 11:00 in the morning once or twice per
2013). CO2 , another important greenhouse gas, should also month, depending on weather, from January to December
be monitored from the TGR surface in consideration of of 2010. The previous studies (Xing et al., 2005; Iqbal et
spatial and temporal heterogeneity. The objectives of this al., 2009) and the preliminary experiments in the sites (data
article were to analyze temporal and spatial variations in not shown) all support the hypothesis that the CO2 flux at
CO2 flux and to determine the environmental factors that 9:00 to 11:00 am could represent the daily mean CO2 flux.
influence CO2 flux at the TGR. At the four transects perpendicular to the bank, we
selected five locations from the littoral zone to the pelagic
zone for water depths of 5, 10, 20, 50, and 100 m at
1 Materials and methods Sandouping, Zigui, Wushan, and Yunyang. However, mea-
surement was not made at 100 m at Sandouping because
1.1 Site description
the maximum water depth was less than 70 m. We selected
The TGR region is located in a subtropical monsoon June to July 2010 (summer) and January 2011 (winter)
climate zone, with an average annual temperature of 16.3– as the dates for comparing CO2 fluxes at different water
18.2°C and annual precipitation of 987–1326 mm, and depths in different seasons.
the relative humidity is about 70% (Guo et al., 2007).
1.2 Measurement of CO2 flux
The annual average water inflow from the upstream rivers
was 12,800 m3 /sec in 2010, with the maximum of 69,000 CO2 flux was measured at the air-water interface of the
m3 /sec on July 19 and the minimum of 3500 m3 /sec on TGR using floating static chambers (Kemenes et al., 2011).
February 21 (Yang et al., 2012b). Although the water The chamber (65 cm in length × 45 cm in width × 40
inflow is remarkably larger in the flood season than in the cm in height) consisted of a plastic box without a cover
non-flood season, the water level in TGR is controlled at that was wrapped in light-reflecting and heatproof films
145 m artificially during the flood season (June–August) to prevent temperature variation inside the chamber. In
to discharge the flood water and sediments. After the flood addition, plastic foam was fixed onto the opposite sides
season (in October), the reservoir begins to store clear of the chamber. Each chamber was inverted and immersed
water to a water level of 175 m (Lu et al., 2011; Yang et al., into the water, and the height of the headspace in the closed
2012a). The water quality during the non-flood season is chamber was about 30 cm. A silicone tube (with outer
grade III and is grade IV during the flood season, according and inner diameters of 0.6 and 0.4 cm, respectively) was
to the surface water environment quality standard of the inserted into the upper side of the chamber to collect gas
People’s Republic of China (Yang et al., 2012b). samples and another silicone tube was inserted into the
The study was conducted at four sites along the chamber in order to keep the air pressure balanced between
mainstream of the TGR (Fig. 1). Three sampling sites the inside and outside of the chamber (Lu et al., 2011).
upstream of the TGD were selected, i.e., Yunyang Three floating chambers were simultaneously deployed at
(30◦ 56′ N, 108◦ 39′ E), Wushan (31◦ 03′ N, 109◦ 51′ E), and the water surface of each sampling point by attaching them
Zigui (30◦ 51′ N, 110◦ 58′ E), which are about 240 km, 120 to a stationary boat during the measurement process.
km, and 2 km upstream from the TGD, respectively. An- The gases in the headspace of the closed chamber were
other sampling site was located at Sandouping (30◦ 49′ N, collected into air-sampling bags (0.5 L; Hedetech, Dalian,
110◦ 03′ E), which is 5 km downstream from the TGD. China) five times every 10 min over a 40-min period. CO2
At the four sites with different distances from the dam, concentrations in the air-sampling bags were analyzed us-
No. 11 Spatial and seasonal variability of CO2 flux at the air-water interface of the Three Gorges Reservoir 2231

ing a CO2 analyzer (LI-840 CO2 /H2 O Analyzer; LI-COR 2 Results

Co., Lincoln, USA) to determine the CO2 concentration.
CO2 flux (F, mg CO2 /(m2 ·hr)) was determined using Eq. 2.1 Spatial variation in CO2 flux
(1). Only samples with r2 values larger than 0.95 were
In 2010, the annual average CO2 flux at the TGR was
considered.
(163.3 ± 117.4) mg CO2 /(m2 ·hr) (range: 18.4–893.5 mg
CO2 /(m2 ·hr)). Significant spatial variations in CO2 flux
F = ρ × dc/dt × 273.15/(273.15 + T ) × H (1) were found between the four measured sites in the TGR
(last column in Table 1). The maximum average CO2 flux
where, ρ (kg/m3 ) is the density of CO2 under standard was found at Wushan (221.9 ± 130.3 mg CO2 /(m2 ·hr)),
temperature and pressure (1.96 kg/m3 ); dc/dt is the rate of and the lowest average CO2 flux was found at Zigui
concentration change in the chamber; H (m) is the height (88.6 ± 52.5 mg CO2 /(m2 ·hr)). There was no significant
from the top of the inversed chamber to the water surface difference between the CO2 flux measured at Sandouping
(0.3 m in this study); 273.15 is the absolute temperature at and Yunyang (189.2 ± 129.1 mg CO2 /(m2 ·hr) and 156.3 ±
0°C and; T (°C) is the air temperature inside the chamber. 88.6 mg CO2 /(m2 ·hr), respectively). The average CO2 flux
of the 3 sites in the upstream of TGD is (148.5 ± 110.8) mg
1.3 Measurement of environmental variables CO2 /(m2 ·hr), which was 78% of the CO2 flux in the river
Air and water temperatures were measured using alcohol downstream.
thermometers. The water depth at each measurement point Although CO2 flux increased with increasing water
was determined using a sounding rope. Water velocities, depth during the summer and winter, there was no signif-
pH, and turbidity were measured in situ using a velocity icant difference between CO2 flux at the different water
meter (LS1206B; Midwest Group, Beijing, China), pH depths, except at 50 m at Wushan in July 2010, 100 m at
meter (HI 8424; Microcomputer HANNA, Rome, Italy), Zigui in June 2010, and 50 m at Sandouping in June 2010
and turbidity meter (HI93703, Microcomputer HANNA, and January 2011 (Fig. 2). However, at the same water
Rome, Italy), respectively. These environmental variables depth of the same site, CO2 flux was significantly higher in
were simultaneously measured when the gas samples were June or July 2010 compared with that measured in January
collected. 2011, except at 100 m at Yunyang.

1.4 Statistical analysis 2.2 Seasonal variation in CO2 flux

The normal distribution of CO2 flux was tested using Obvious variations in seasonal CO2 flux were found at
the Kolmogorov-Smirnov Test. Because the variation in Yunyang, Wushan, Zigui, and Sandouping, with signif-
CO2 flux was skewed to the right, the data were first icantly higher CO2 flux measured in the summer and
normalized by natural logarithmic transformation so that lower CO2 flux measured in the winter (Table 1). CO2
the data would follow a normal distribution, and then flux began to increase from 47.2–186.1 mg CO2 /(m2 ·hr)
the spatial differences in CO2 flux at the four sites and in January 2010, reaching its maximum level (166.8–
different water depths were tested using one-way analysis 439.1 mg CO2 /(m2 ·hr)) in the summer (June–August)
of variance (ANOVA) in combination with the Tukey test. and declined thereafter (Fig. 3). Among the four sites,
In addition, CO2 flux was linearly regressed in terms of seasonal CO2 variation was the smallest at Zigui, being
pH, air temperature, air-water temperature difference, and in the range of 38.6–166.8 mg CO2 /(m2 ·hr). However,
water velocity. The multiple regression method was used differences between the maximum and minimum CO2
to find out the key environmental variables that influenced fluxes all exceeded 300 mg CO2 /(m2 ·hr) at the other three
CO2 emission. The data were analyzed with the SPSS 16.0 sites.
statistical package.
Table 1 Average CO2 flux at different sites in four seasons

Site CO2 flux (mg CO2 /(m2 ·hr))

Spring Summer Autumn Winter Average
(Mar–May) (Jun–Aug) (Sep–Nov) (Dec–Feb)

Upstream
Yunyang 117.9 ± 58.1 a 279.3 ± 97.1 b 125.1 ± 38.7 a 123.5 ± 35.8 a 156.3 ± 88.6 B
Wushan 229.0 ± 86.6 b 313.0 ± 161.4 c 166.7 ± 85.2 ab 124.0 ± 78.0 a 221.9 ± 130.3 C
Zigui 87.2 ± 47.7 b 129.8 ± 58.6 c 65.5 ± 26.6 ab 59.9 ± 33.1 a 88.6 ± 52.5 A
Downstream
Sandouping 191.9 ± 123 .1 b 339.1 ± 125.8 c 186.7 ± 97.6 b 77.7 ± 35.2 a 189.2 ± 129.1 B
Data are expressed as mean ± SD.
Different lower-case letters within a row indicate significant differences (p < 0.05), and different upper-case letters in the last column indicate significant
differences (p < 0.05).
2232 Journal of Environmental Sciences 2013, 25(11) 2229–2238 / Le Yang et al. Vol. 25

400 600
18 January 2011 B Yungang 21 January 2011 Wushan
AB B
29 June 2010 500 20 July 2010
CO2 flux (mg CO2/(m2.hr))

CO2 flux (mg CO2/(m2.hr))

300
400
AB b
ab AB
200 AB ab 300 a
A a
ab A
200 A
100 a a
a a
100

0 0
5 10 20 50 100 5 10 20 50 100
Water depth (m) Water depth (m)
350 1200
B Sandouping
11 January 2011 Zigui 6 January 2011
300 B
20 June 2010 1000 21 June 2010

CO2 flux (mg CO2/(m2.hr))

CO2 flux (mg CO2/(m2.hr))

250
800 AB
A
200 A
600
150
A A a
A 400
100 A b
a a
a a a
50 200 a
a
0 0
20 5 10
50 100 5 10 20 50
Water depth (m) Water depth (m)
Fig. 2 CO2 flux at different water depths in the summer (June or July, 2010) and the winter (January, 2011) at Yunyang, Wushan, Zigui, and Sandouping.
Different upper case letters indicate a significant level of difference (p < 0.05) in CO2 flux in June to July 2010, and different lower case letters indicate
a significant level of difference (p < 0.05) in January 2011.

2.3 Environmental factors influencing CO2 flux CO2 flux increased with increasing water depth at a
perpendicular transect to the bank (Fig. 2), and the high
A significant negative correlation was found between CO2
CO2 flux measured at deep water areas was probably due
flux and pH (p < 0.001) (Fig. 4a). Meanwhile, CO2 flux
to the high water velocities in such areas.
was positively correlated with water velocity, air-water
The average CO2 flux at Sandouping, 5 km downstream
temperature difference, and air temperature (p < 0.001;
from the TGD, was more than twice as large as that
Fig. 4 b, c, d). The results of the linear multiple regression
measured at Zigui, 2 km upstream of the TGD. Besides
analysis show that CH4 emission (F; mg CO2 /(m2 ·hr)) is
the difference in water velocity between the two places, the
significantly related to water velocity (V; m/sec), turbidity
dissolved CO2 in the hypolimnion at Zigui probably con-
(Turb; NTU); air-water temperature difference (∆T ; °C) ,
tributed significantly to the high CO2 flux at Sandouping
and water depth (D; m), as shown by Eq. (2) (r2 = 0.50,
because of the disturbance by turbines. Rivers downstream
d f =299, p < 0.001).
of dams have been considered as contributing a significant
fraction of the CO2 emission from hydroelectric systems
(Guérin et al., 2006; Kemenes et al., 2011). CO2 emissions
F = 68.24 + 342.57V + 0.37Turb + 4.00∆T + 0.35D (2) downstream from hydroelectric dams were often reported
to be significantly larger than that upstream from hydro-
3 Discussion electric dams (Roland et al., 2010; Wang et al., 2011).

3.1 Spatial variation in CO2 flux 3.2 Temporal variation in CO2 flux

Spatial variation in CO2 flux is commonplace in reservoirs. Seasonal variation in CO2 flux in the TGR could be
In the TGR, the spatial variation in CO2 flux was probably explained by the variation in water velocity, temperature,
related to the water velocities at different locations. The and pH. Firstly, the inlet water flow to the TGR drastically
average annual water velocity measured at Wushan (21.3 increased starting in June, causing floods during summer.
cm/sec) was considerably higher than that measured at The water velocity was faster in the flood season (June–
Zigui (5 cm/sec) and Yunyang (12.6 cm/sec). In addition, September) than the non-flood season, which contributed
No. 11 Spatial and seasonal variability of CO2 flux at the air-water interface of the Three Gorges Reservoir 2233

600 600
Yunyang Wushan
500 500
CO2 flux (mg CO2/(m2.hr))

CO2 flux (mg CO2/(m2.hr))

400 400

300 300

200 200

100 100

0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
Julian day Julian day
600 600
Zigui Sandouping
500 500

CO2 flux (mg CO2/(m2.hr))

CO2 flux (mg CO2/(m2.hr))

400 400

300 300

200 200

100 100

0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
Julian day Julian day
Fig. 3 Seasonal variations in CO2 flux at Yunyang, Wushan, Zigui, and Sandouping in 2010.

greatly to CO2 flux (Borges et al., 2004; Ferrón et al., et al., 2004). These negative fluxes were possibly related to
2007). Moreover, high air and water temperatures and high phytoplankton activity.
a high air-water temperature difference in the summer
3.3 Environmental variables influencing CO2 flux
stimulated CO2 production and transport. Furthermore, the
pH was significantly lower in the summer than in the Temperature was critical to CO2 emission because it regu-
winter (p < 0.05) because large amounts of acid rain in the lated microbial activities and influenced the decomposition
summer neutralized alkaline water (Larssen et al., 2006). of organic carbon. In this study, CO2 flux was 2–3 times
CO2 flux was negatively correlated with pH (Fig. 4a); higher in the summer than in the winter (Table 1), which
therefore, the low pH of the water was conducive to CO2 was probably related to the differences in water temper-
emission in the summer. ature. The results of controlled experiments showed that
CO2 emission flux in the TGR peaked in the summer CO2 production could increase by 3–5 times during peri-
(Fig. 3), which was different from the seasonal variations ods of high water temperature (20–22°C) compared with
in CO2 flux reported in natural lakes and other reservoirs periods of low water temperature (4–5°C) (Thérien et al.,
in China. CO2 was absorbed from the atmosphere in the 2005). The air-water temperature difference reflected the
spring and summer via photosynthesis by algal blooms thermal difference at the air-water interface. An increased
but emitted in large amounts in the winter, such as in air-water temperature difference promoted CO2 emission
Lake Donghu (Xing et al., 2005) and the four cascade from the reservoir surface under evaporative conditions
reservoirs on the Maotiao River in southwest China (Wang due to the destabilization of the water surface (Guérin et
et al., 2011). However, algal blooms hardly occurred in the al., 2007). For example, CO2 flux was 4% higher under
summer because of high water velocity in the mainstream evaporative conditions in the Pacific Ocean (Ward et al.,
of the TGR then (Xin et al., 2011). Negative CO2 flux 2004). In contrast, CO2 fluxes were reduced by 30% under
was often reported in other reservoirs in the world, such condensing conditions (Liss et al., 1981).
as the Três Marias Reservoir in Brazil (dos Santos et al., CO2 flux was negatively correlated with pH in this
2006), Petit Saut Reservoir in French Guiana (Abril et al., study (Fig. 4a), which was in agreement with other studies
2005), and F. D. Roosevelt, Dworshak, Wallula, and New (Gui, 2007; Peng et al., 2012). CO2 flux decreased as
Melones Reservoirs in the western United States (Soumis pH increased, and atmospheric CO2 was even absorbed
2234 Journal of Environmental Sciences 2013, 25(11) 2229–2238 / Le Yang et al. Vol. 25

1000 1000
a y = -159.7x + 1420.9, n = 265 b
y = 513.4x + 83.8, n = 226
800 r2 = 0.11, p < 0.001 800
CO2 flux (mg CO2/(m2.hr))

CO2 flux (mg CO2/(m2.hr))

r2 = 0.44, p < 0.001
600 600

400 400

200 200

0 0

-200 -200
7.0 7.5 8.0 8.5 9.0 0.0 0.2 0.4 0.6 0.8
pH Water velocity (m/sec)
1000 1000
c d
800 800

CO2 flux (mg CO2/(m2.hr))

CO2 flux (mg CO2/(m2.hr))

y = 139.6exp(0.05x), n = 295 y = 76.8exp(0.03x), n = 306

600 r2 = 0.13, p < 0.001 600 r2 = 0.13, p < 0.001

400 400

200 200

0 0

-200
-10 -5 -15 0 5 10 15 0 10 20 30 40
Air-water temperature difference (°C) Air temperature (°C)
Fig. 4 Relationships between CO2 flux and pH (a), water velocity (b), air-water temperature difference (c), and air temperature (d) at TGR.

at high pH values because high pH favored the formation factors: flooded organic carbon, allochthonous carbon,
of bicarbonate and led to the undersaturation of dissolved primary productivity, and dissolved oxygen (St Louis et
CO2 in the water (Tremblay et al., 2005). The critical al., 2000; Barros et al., 2011; Wang et al., 2011; Kemenes
pH value for CO2 absorption was 8.0–8.5 in boreal and et al., 2011). However, the relationships between CO2 flux
temperate reservoirs (Soumis et al., 2004; Therrien et al., and these factors were not found in our study because
2005; Tremblay et al., 2005). However, most of the CO2 every reservoir had its specific characteristics. In the TGR,
flux was positive in the study, probably because most of most of the biomass was cleared before flooding, and water
the pH values in the TGR ranged from 7.2 to 8.4, which quality was homogeneous after water storage. In addition,
did not surpass the abovementioned critical range. no macrophytes or phytoplankton were distributed in the
Water velocity, as a source of turbulence at the water water surface of the mainstream.
surface, had an effect on the gas-transfer rate at the air-
3.4 Carbon budget at the TGR
water interface (Ferrón et al., 2007). Water velocity had
significant spatial and temporal variations at the TGR, The average diffusive CO2 flux at the air-water interface
with high values in summer at Wushan, and low values of the TGR was (163.3 ± 117.4) mg CO2 /(m2 ·hr) (3919 ±
in other seasons and sites, which were almost coincident 2818 mg CO2 /(m2 ·day)) in 2010, which was higher than
with the spatiotemporal variations in CO2 flux (Table 1). the fluxes measured at boreal and temperate reservoirs
Water velocity was probably a crucial factor that drove the (except the Sainte-Marguerite Reservoir in eastern Québec,
variations in CO2 emission from the TGR, and a well- Canada) but lower than the fluxes of most of the tropical
correlated relationship between water velocity and CO2 reservoirs listed in Table 2.
flux also supported the assumption (Fig. 4b). Water veloc- To obtain the total CO2 emission from the TGR surface,
ity also influenced gas emissions from the water surface the CO2 emissions from the mainstream and the tributaries
in the shallow bay of Cádiz, Spain (Ferrón et al., 2007). were estimated separately. Our experiment was conducted
In addition, Borges et al. (2004) proposed an empirical on the mainstream, and the diffusive CO2 flux was 3919
equation that parameterized the cumulative effects of water mg CO2 /(m2 ·day). Considering the area of 528 km2 in
velocity on gas exchange at the air-water interface. the mainstream, the total diffusive CO2 emission from the
A wide variation in the intensity of the emissions was mainstream was estimated to be 0.21 Tg C/yr. Xiangxi
noted, indicating the influence of many other different River and Pengxi River are two representative tributaries in
No. 11 Spatial and seasonal variability of CO2 flux at the air-water interface of the Three Gorges Reservoir 2235

Table 2 CO2 flux in the boreal, temperate, and tropical reservoirs

Location Reservoir Age (yr) CO2 flux Reference

(mg CO2 /(m2 ·day))
Mean Range
Boreal
Finland Lokka 27 1520 –768∼6120 Huttunen et al., 2002
Porttipahta 25 1536 1296∼3456 Huttunen et al., 2002
Poland Solina 41 –914∼648 Gruca-Rokosz et al., 2010
Rzeszow 36 2042∼7150 Gruca-Rokosz et al., 2010
Wilcza Wola 21 3893∼4161 Gruca-Rokosz et al., 2010
Switzer- land 5 old reservoirs 42–85 970 132∼2516 Diem et al., 2012
Canada La Grande-2 18–20 1500 200∼6000 Duchemin et al., 1995
Laforge-1 2 2250 380∼6700 Duchemin et al., 1995
Old Québec reservoirs >10 1600 980∼4400 Tremblay et al., 2004
Sainte-Marguerite 2–4 4400 400∼8400 Tremblay et al., 2004
Old Manitoba reservoirs 35–73 3350 2000∼5560 Tremblay et al., 2004
Grand Rapids 624 –347∼7260 Demarty et al., 2009
Jenpeg 316 –926∼2126 Demarty et al., 2009
Kettle 514 –723∼5646 Demarty et al., 2009
McArthur Falls 367 –1057∼3556 Demarty et al., 2009
Eastmain-1 1–4 2426 –4∼19676 Demarty et al., 2009
Rivière-des-Prairies 665 –234∼9380 Demarty et al., 2009
Robert-Bourassa 28 661 45–6509 Demarty et al., 2009
Temperate
USA F. D. Roosevelt 59 –462 –852∼251 Soumis et al., 2004
Dworshak 28 –1195 –2278∼–720 Soumis et al., 2004
Wallula 47 –349 –1629∼1060 Soumis et al., 2004
Shasta 57 1247 351∼2150 Soumis et al., 2004
Oroville 33 1026 266∼2430 Soumis et al., 2004
New Melones 22 –1186 –3415∼–275 Soumis et al., 2004
China Xiangxi River (a tributary of TGR) 7 1836 –2400∼5053 Zhao et al., 2011
Hongfeng 50 660 –396∼3080 Wang et al., 2011
Baihua 48 1012 –352∼3388 Wang et al., 2011
Hongyan 37 1056 –264∼3916 Wang et al., 2011
Xiuwen 48 2077 –352∼3828 Wang et al., 2011
TGR 7 3919 442∼21444 This study
Tropical
Brazil Miranda 1–2 4388 16∼61182 dos Santos et al., 2006
Três Marias 36–37 1114 33∼10060 dos Santos et al., 2006
Barra Bonita 35–36 3986 83∼33424 dos Santos et al., 2006
Xingó 4–5 6139 29∼89203 dos Santos et al., 2006
Samuel 4 7447 2200∼24283 dos Santos et al., 2006
Tucuruı́ 8 8475 457∼142723 dos Santos et al., 2006
Balbina 18 13845 1258∼31273 Kemenes et al., 2011
French Guiana Petit Saut 10 5764 –440∼15840 Abril et al., 2005
Laos Nam Ngum 38 –704 –932.8∼–118.8 Chanudet et al., 2011
Nam Leuk 10 136.4 –466.4∼1680.8 Chanudet et al., 2011
Different lower-case letters within a row indicate significant differences (p < 0.05), and different upper-case letters in the last column indicate significant
differences (p < 0.05).

the TGR, and the average diffusive CO2 fluxes were 1836 99% of the total CO2 emission from the reservoir surface
mg CO2 /(m2 ·day) and 2196 mg CO2 /(m2 ·day) in 2010, (Rosa et al., 2003; Abril et al., 2005; Kemenes et al.,
respectively (Zhao et al., 2011; Guo et al., 2011). The 2011). The frequency of bubble occurrence was 5% in
total diffusive CO2 emission was estimated to be 0.10–0.12 our experiment. We considered the bubble occurrence
Tg C/yr from the tributaries, considering the total area of when CH4 concentration increased abruptly in the chamber
556 km2 in all the tributaries. Therefore, the total diffusive (Yang et al., 2012a, 2013). However, CO2 concentration
CO2 emission was estimated to be 0.31–0.33 Tg C/yr from did not increase obviously in the chambers when bubbles
the TGR surface. If adding in the diffusive CH4 emission were trapped in the chambers, indicating that bubbles had
from the surface of the TGR (2.46 Gg CH4 /yr; Chen et al., little contribution to CO2 emission from the TGR surface.
2011), the total diffusive emissions of C-CO2 and C-CH4 Thus CO2 bubbling flux was ignored in the estimation.
were 0.36–0.38 Tg C/yr from the reservoir surface (CO2 No measurement was conducted on degassing of CO2
equivalent) (Table 3). and CH4 at the turbine outflow at the TGR, which was the
The diffusive CO2 flux at the surface was the dominant major uncertainty in the estimation for the entire reservoir.
fraction in a hydroelectric system, which accounted for Based on the available literature, degassing flux at the
2236 Journal of Environmental Sciences 2013, 25(11) 2229–2238 / Le Yang et al. Vol. 25

Table 3 CO2 budget of the TGR in 2010

dumped into the TGR every year, which was estimated to
Carbon fluxesb be 1.98 Tg C/yr (Li and Huang, 2005, 2006). Judging from
(Tg C/yr) the above analysis, the TGR is a carbon sink (1.48–1.62 Tg
Carbon input
C/yr). This carbon loss might be buried in the sediments.
Allochthonous carbon from the upstream river +10.08c The TGD is located in the mainstream of the Yangtze
Water pollution sources +1.98d River, which intercepted 60% of the inflow sediments
Carbon output (about 3 × 108 tons) from the upstream rivers every year.
Carbon export to the downstream river –10.06e
Diffusive emission from reservoir surface –0.36∼0.38f
Most of the pollution could adhere to the sediments and
(C-CO2 equivalent)a deposit into the bottom of the TGR (Zhou et al., 2006).
Diffusive emission from downstream river –0.0093g CO2 and CH4 emissions from the TGR surface are mainly
(C-CO2 equivalent) derived from the input of all types of pollution by human
Degassing at the turbine outflow (C-CO2 equivalent) –0.006∼0.135h
Total carbon budget +1.48∼1.62
beings. Thus, it is necessary to control the water pollution
a
sources in the TGR region.
CO2 equivalent is calculated as the data for CO2 plus the CH4 data
multiplied by 25 according to IPCC (2007); b positive values indicate
carbon input to the TGR, while the negative values indicate carbon output 4 Conclusions
from the TGR; c calculated from the inlet flow of the TGR and the
concentration of total carbon at Cuntan (entrance to the TGR) every Based on our findings, the annual total emissions of C-CO2
month (unpublished data); d cited from (Li and Huang, 2005, 2006). Wa- and C-CH4 from the surface of the TGR were 0.38–0.52 Tg
ter pollution sources included point source pollution (household sewage,
industrial sewage), non-point source pollution (agriculture pollution, C/yr (1.39–1.91 Tg CO2 /yr). The CO2 emission density is
municipal waste and industrial waste pollution, ship-flowing pollution); 16.41–22.55 g/kWh when taking account of the fact that
e calculated from the outlet flow of the TGR and the concentration of
8.47 × 1010 kWh of electricity were produced by the TGR
total carbon at Sandouping (outlet of the TGR) every month (unpublished in 2010. The CO2 emission density of the hydropower
data); f total CO2 emission was estimated to be 0.31∼0.33 Tg C/yr from
the TGR surface in this study, and the total CH4 emission was estimated generated by the TGR is much less than that of the electric
to be 0.00185 Tg C/yr (Chen et al., 2011); g estimated from CO2 flux power currently generated by most power plants in China
and CH4 flux at Sandouping and the area of river downstream (19 km2 ). (807 g/kWh) (Lu et al., 2010). If the goal of generating
The total CO2 emission was estimated to be 0.0086 Tg C/yr and the
in total 1.216 × 1012 kWh of electricity is fulfilled in
total CH4 emission was estimated to be 0.000029 Tg C/yr (Yang et al.,
2013); h estimated by the proportions (CO2 : 1.6%–7%, CH4 : 1.5%–70%) 2020 (Zhou and Qian, 2011), 19.95–27.42 Tg CO2 will
of degassing at the turbine outflow account for the total from the available be emitted from hydroelectric reservoirs every year, which
literature. will be 0.22%–0.30% of the CO2 emission from fossil fuel
combustion (9051 Tg) in 2020 (Tang, 2009).
turbines accounted for about 1.6%–7% of the total CO2 In this study, a significant spatial variation in CO2 flux
emission and 1.5%–70% of the total CH4 emission in a was found in the TGR, which was attributed to differences
hydroelectric system (Soumis et al., 2004; Abril et al., in water velocities at the various sites and water depths un-
2005; Fearnside, 2005; Chanudet et al., 2011; Kemenes der investigation. Seasonal variation indicated that the CO2
et al., 2011). Referring to these two proportions, the flux was higher in the summer than in the other seasons,
degassing emission at the turbine outflow of the TGR which was mainly influenced by the seasonal changes in
was estimated to be 0.006–0.135 Tg C/yr. In addition, an temperature, pH, and water velocity. Other factors, such
emission of 0.0093 Tg C/yr from the river downstream of as local human activities and waves caused by the wind or
the TGD was estimated based on the data at Sandouping passing steamships, also influenced CO2 emission from the
(Table 3). TGR. More investigations need to be carried out to design
The exchange of CO2 across the air-water interface coupled hydrological and biogeochemical process models
is an important term in the carbon budget of aquatic that can better estimate CO2 emissions from reservoirs.
ecosystems. About 0.38–0.52 Tg C/yr of carbon would
be released from the TGR system based on the above Acknowledgments
analysis, including reservoir surface emission, degassing This work was supported by the National Natural Sci-
emission at the turbine outflow, and emission from the river ence Foundation of China (No. 50809067, 41303065), the
downstream of the TGD. As a net carbon source to the Strategic Priority Research Program of the Chinese Acade-
atmosphere, the TGR must be supported by some carbon my of Sciences (No. XDA05060102, XDA05050602), the
input. These inputs consisted of allochthonous carbon from Project of Zhejiang Key Scientific and Technological In-
the upstream rivers, water pollution sources in the TGR novation Team (No. 2010R50039) and the Project of Zhe-
region, and photosynthesis of the phytoplankton in the jiang Scientific and Technological Plan (No. 2011F20025).
reservoir surface. The carbon input from the upstream
rivers was estimated to be 10.08 Tg C/yr, which was References
almost balanced by carbon output to the downstream river
(10.06 Tg C/yr) (Table 3). Moreover, large amounts of Abril G, Guérin F, Richard S, Delmas R, Galy-Lacaux C, Gosse P
point source pollution and non-point source pollution were et al., 2005. Carbon dioxide and methane emissions and the
No. 11 Spatial and seasonal variability of CO2 flux at the air-water interface of the Three Gorges Reservoir 2237

carbon budget of a 10-year old tropical reservoir (Petit Saut, Gruca-Rokosz R, Tomaszek J, Koszelnik P, Czerwieniec E, 2010.
French Guiana). Global Biogeochemical Cycles, 19(4): 1– Methane and carbon dioxide emission from some reservoirs
16. in SE Poland. Limnological Review, 10(1): 15–21.
Barros N, Cole J, Tranvik L, Prairie Y, Bastviken D, Huszar V Guérin F, Abril G, Richard S, Burban B, Reynouard C, Seyler
et al., 2011. Carbon emission from hydroelectric reservoirs P et al., 2006. Methane and carbon dioxide emissions
linked to reservoir age and latitude. Nature Geoscience, from tropical reservoirs: Significance of downstream rivers.
4(9): 593–596. Geophysical Research Letters, 33(21): 1–6.
Bond-Lamberty B, Thomson A, 2010. Temperature-associated Guérin F, Abril G, Serça D, Delon C, Richard S, Delmas R
increases in the global soil respiration record. Nature Let- et al., 2007. Gas transfer velocities of CO2 and CH4 in
ters, 464(7288): 579–582. a tropical reservoir and its river downstream. Journal of
Borges A V, Vanderborght J, Schiettecatte L, Gazeau F, Ferrón S, Marine Systems, 66(1-4): 161–172.
Delille B et al., 2004. Variability of the gas transfer velocity Gui Z S, 2007. Study on the Distribution and influence mecha-
of CO2 in a macrotidal Estuary (the Scheldt). Estuaries, nism of pCO2 in the Yangtze River mainstream and estuary.
27(4): 593–603. Master Thesis. Ocean University of China, Qingdao. 23–29.
Chanudet V, Descloux S, Harby A, Sundt H, Hansen B H, Guo J S, Jiang T T, Li Z, Chen Y B, Sun Z Y, 2011. Analysis
Brakstad O et al., 2011. Gross CO2 and CH4 emissions from on partial pressure of CO2 and influencing factors during
the Nam Ngum and Nam Leuk sub-tropical reservoirs in spring phytoplankton bloom in the backwater area of Xiao-
Lao PDR. Science of the Total Environment, 409(24): 5382– jiang River in Three Gorges Reservoir. Advances in Water
5391. Science, 22(6): 829–838.
Chen H, Wu Y Y, Yuan X Z, Gao Y H, Wu N, Zhu D, 2009. Guo X Z, Huang X B, Xu K, 2007. The natural and geological
Methane emissions from newly created marshes in the environment in the Three Gorges Reservoir. In: Atlas of
drawdown area of the Three Gorges Reservoir. Journal of Prevention and Treatment of Landslide and Collapse Ge-
Geophysical Research, 114(D18): 1–7. ological Hazards in Three Gorges Reservoir Project Area
Chen H, Yuan X Z, Chen Z L, Wu Y Y, Liu X S, Zhu D (Wu G, Huang Z, eds.). China Water Power Press, Beijing.
et al., 2011. Methane emissions from the surface of the 1–16.
Three Gorges Reservoir. Journal of Geophysical Research, Houghton J, 2005. Global warming. Reports on Progress in
116(D21): 1–5. Physics, 68(6): 1343–1403.
Demarty M, Bastien J, 2011. GHG emissions from hydroelectric Huttunen J T, Väisänen T S, Hellsten S K, Heikkinen M, Nykänen
reservoirs in tropical and equatorial regions: Review of 20 H, Jungner H et al., 2002. Fluxes of CH4 , CO2 , and N2 O
years of CH4 emission measurements. Energy Policy, 39(7): in hydroelectric reservoirs Lokka and Porttipahta in the
4197–4206. northern boreal zone in Finland. Global Biogeochemical
Demarty M, Bastien J, Tremblay A, Hesslein R H, Gill R, Cycles, 16(1): 3-1–3-17.
2009. Greenhouse gas emissions from boreal reservoirs in IPCC (Intergovernmental Panel on Climate Change), 2007.
Manitoba and Québec, Canada, measured with automated Intergovernmental Panel on Climate Change’s Fourth As-
systems. Environmental Sciences and Technology, 43(23): sessment Report, Paris.
8908–8915. Iqbal J, Lin S, Hu R G, Feng M L, 2009. Temporal variability
Diem T, Koch S, Schwarzenbach S, Wehrli B, Schubert C J, of soil-atmospheric CO2 and CH4 fluxes from different land
2012. Greenhouse gas emissions (CO2 , CH4 , and N2 O) uses in mid-subtropical China. Atmospheric Environment,
from several perialpine and alpine hydropower reservoirs 43(37): 5865–5875.
by diffusion and loss in turbines. Aquatic Sciences, 74(3): Kemenes A, Forsberg B R, Melack J M, 2011. CO2 emissions
619–635. from a tropical hydroelectric reservoir (Balbina, Brazil).
dos Santos M A, Rosa L P, Sikar B, Sikar E, dos Santos E Journal of Geophysical Research, 116(G3): 1–11.
O, 2006. Gross greenhouse gas fluxes from hydro-power Larssen T, Lydersen E, Tang D G, He Y, Gao J X, Liu H Y et
reservoir compared to thermo-power plants. Energy Policy, al., 2006. Acid rain in China. Environmental Science and
34(4): 481–488. Technology, 40(2): 418–425.
Duchemin E, Lucotte M, Canuel R, Chamberland A, 1995. Li C M, Huang Z L, 2005. Study on the pollutant loads into
Production of the greenhouse gases CH4 and CO2 by Three Gorges Reservoir (I)–Pollutant load status before
hydroelectric reservoirs of the boreal region. Global Bio- impoundment. Resources and Environment in the Yangtze
geochemical Cycles, 9(4): 529–540. Basin, 14(5): 611–622.
Fearnside P M, 2005. Brazil’s Samuel Dam: Lessons for hy- Li C M, Huang Z L, 2006. Study on the pollutant loads into
droelectric development policy and the environment in Three Gorges Reservoir (II)-Pollutant load predictions after
Amazonia. Environmental Management, 35(1): 1–19. impoundment. Resources and Environment in the Yangtze
Fearnside P M, 2011. Greenhouse gas emissions from hydro- Basin, 15(1): 97–106.
electric dams in tropical forests. In: The Encyclopedia of Liss P S, Balls P W, Martinelli F N, Coantic M, 1981. The effect
Energy (Cleveland C J, Ayres R U, eds.). John Wiley & of evaporation and condensation on gas transfer across an
Sons Publishers, New York. 1–19. air-water interface. Oceanologica Acta, 4(2): 129–138.
Ferrón S, Ortega T, Gómez-Parra A, Forja J M, 2007. Seasonal Lu F, Wang X K, Han B, Ouyang Z Y, Duan X N, Zheng H,
study of dissolved CH4 , CO2 and N2 O in a shallow tidal 2010. Net mitigation potential of straw return to Chinese
system of the bay of Cádiz (SW Spain). Journal of Marine cropland: estimation with a full greenhouse gas budget
Systems, 66(1-4): 244–257. model. Ecological Applications, 20(3): 634–647.
2238 Journal of Environmental Sciences 2013, 25(11) 2229–2238 / Le Yang et al. Vol. 25

Lu F, Yang L, Wang X K, Duan X N, Mu Y J, Song W Z Tremblay A, Therrien J, Hamlin B, Wichmann E, LeDrew L,

et al., 2011. Preliminary report on methane emissions from 2005. GHG emissions from boreal reservoirs and natural
the Three Gorges Reservoir in the summer drainage period. aquatic ecosystems. In: Greenhouse Gas Emission-Fluxes
Journal of Environmental Sciences, 23(12): 2029–2033. and Processes: Hydroelectric Reservoirs and Natural Envi-
Mendonça R, Barros N, Vidal L, Pacheco F, Kosten S, Roland F, ronments (Tremblay A, Varfalvy L, Roehm C et al., eds.).
2012. Greenhouse gas emissions from hydroelectric reser- Springer, New York. 209–232.
voirs: What knowledge do we have and what is lacking. In: Wang F S, Wang B L, Liu C Q, Wang Y C, Guan J, Liu X D
Greenhouse Gas-Emission, Measurement and Management et al., 2011. Carbon dioxide emission from surface water
(Liu G X, ed.). InTech Press, Rijeka. 55–78. in cascade reservoirs-river system on the Maotiao River,
Peng X, Wang B L, Liu C Q, Wang F S, 2012. Diurnal variations southwest of China. Atmospheric Environment, 45(23):
of pCO2 in relation to environmental factors in the cascade 3827–3834.
reservoirs along the Wujiang River, China. Chinese Journal Ward B, Wanninkhof R, McGillis W R, Jessup A T, DeGrandpre
of Geochemistry, 31(1): 41–47. M D, Hare J E et al., 2004. Biases in the air-sea flux of CO2
Roland F, Vidal L O, Pacheco F S, Barros N O, Assireu A, resulting from ocean surface temperature gradients. Journal
Ometto J et al., 2010. Variability of carbon dioxide flux of Geophysical Research, 109(C8): 1–14.
from tropical (Cerrado) hydroelectric reservoirs. Aquatic Xin X K, Yin W, Ye M, 2011. Preliminary research on hydro-
Sciences, 72(3): 283–293. dynamic dispatch method of water blooms in Three Gorges
Rosa L P, dos Santos M A D, Matvienko B, Sikar E, Lourenço R Reservoir Region. Water Resources and Power, 29(7): 16–
S, Menezes C F, 2003. Biogenic gas production from major 19.
Amazon reservoirs, Brazil. Hydrological Processes, 17(7): Xing Y P, Xie P, Yang H, Ni L Y, Wang Y S, Rong K W,
1443–1450. 2005. Methane and carbon dioxide fluxes from a shallow
Soumis N, Duchemin É, Canuel R, Lucotte M, 2004. Greenhouse hypereutrophic subtropical Lake in China. Atmospheric
gas emissions from reservoirs of the western United States. Environment, 39(30): 5532–5540.
Global Biogeochemical Cycles, 18(3): 1–11. Yang L, Lu F, Wang X K, Duan X N, Song W Z, Sun B F
St Louis V L, Kelly C A, Duchemin É, Rudd J W M, Rosenberg D et al., 2012a. Surface methane emissions from different
M, 2000. Reservoir surfaces as sources of greenhouse gases land use types during various water levels in three major
to the atmosphere: A global estimate. BioScience, 50(9): drawdown areas of the Three Gorges Reservoir. Journal of
766–775. Geophysical Research-Atmosphere, 117(D10): 1–11.
Tadonléké R D, Planas D, Paquet S, 2005. Bacterial activity in Yang L, Lu F, Wang X K, Duan X N, Song W Z, Zhang Q Q et al.,
the water column and its impact on the CO2 efflux. In: 2013. Spatial and seasonal variability of diffusive methane
Greenhouse Gas Emission-Fluxes and Processes: Hydro- emissions from the Three Gorges Reservoir. Journal of
electric Reservoirs and Natural Environments (Tremblay A, Geophysical Research-Biogeosciences, 118(2): 471–481.
Varfalvy L, Roehm C et al., eds.). Springer, New York. 467– Yang L, Zhang Y, Hou P Q, Sun B F, Zhang Q Q, Wang X K
482. et al., 2012b. Temporal and spatial variation on nitrogen
Tang X, 2009. The target of carbon emission reduction was and phosphorus in the middle and lower reaches of the
confirmed in China, energy saving by industry. United Three Gorges Reservoir. Resources and Environment in the
Securities, 11: 1–5. Yangtze Basin, 21(6): 732–738.
Teodoru C R, Prairie Y T, del Giorgio P A, 2011. Spatial hetero- Zhao Y, Wu B F, Zeng Y, 2013. Spatial and temporal patterns of
geneity of surface CO2 fluxes in a newly created Eastmain-1 greenhouse gas emissions from Three Gorges Reservoir of
Reservoir in Northern Quebec, Canada. Ecosystems, 14(1): China. Biogeosciences, 10(2): 1219–1230.
28–46. Zhao Y, Zeng Y, Wu B F, Wang Q, Yuan C, Xu Z R, 2011.
Therrien J, Tremblay A, Jacques R B, 2005. CO2 emissions Observation on greenhouse gas emissions from Xiangxi
from semi-arid reservoirs and natural aquatic ecosystems. River in Three Gorges Region. Advances in Water Science,
In: Greenhouse Gas Emission-Fluxes and Processes: Hy- 22(4): 546–553.
droelectric Reservoirs and Natural Environments (Tremblay Zhou J J, Lin B N, Li Y L, 2006. On the control of internal source
A, Varfalvy L, Roehm C et al., eds.). Springer, New York. of pollution for the Three Gorges Reservoir. Science and
234–250. Technology Review, 24(10): 5–10.
Tremblay A, Lambert M, Gagnon L, 2004. Do hydroelectric Zhou J, Qian G, 2011. The planning of the thirteen largest
reservoirs emit greenhouse gases? Environmental Manage- hydroelectricity bases and development status quo. Journal
ment, 33(S1): S509–S517. of Water Resources and Hydropower Construction, 124(1):
1–7.