The LIM-only and development

Hydroelectric reservoirs have been under sampled to establish them as sources or sinks of the atmospheric carbon dioxide (CO 2 ). Such poor coverage is well known for subtropic, particularly monsoon driven reservoirs in China. Our study presented the spatiotemporal changes of the carbonate system and CO 2 flux in a hydroelectric reservoir (Dangjiankou Reservoir) locating in a subtropical monsoon climate region. Our 21 filed surveys conducted during 2004–2011 revealed significantly spatial and monthly variations of surface water partial pressure of CO 2 ( p CO 2 ) in the Reservoir. p CO 2 , showing higher concentrations in the wet and warm seasons, averaged 595 ± 545 µatm (ranging from 53–3751 µatm) in the reservoir surface, while substantially higher p CO 2 (1132 ± 1220 µatm) was observed in the river downstream the dam. A clear p CO 2 drawdown in the reservoir as water flows demonstrated a significantly descending order of Dan Reservoir > site close to dam > Han Reservoir. This spatial contrast can also be seen in the distributions of dissolved inorganic carbon and total alkalinity. Pronounced seasonality in p CO 2 was controlled by seasonal monsoon rainfall, while photosynthetic CO 2 uptake dominated spatial patterns and dry-month variability of p CO 2 . We further related p CO 2 to water chemical properties and indicated that p CO 2 had strong positive correlations with Si, TP and DOC, negative correlations with DO saturation, TN and Chl a , while weak correlations with other variables including biogenic elements. CO 2 flux from the Reservoir surface showed a bottom average of 9 mmol m –2 d –2 in comparison with other hydroelectric reservoir in China. River downstream the dam had quite high flux of CO 2 (119 mmol m –2 d –2 ), which was intermediate between temperate rivers and compared to global rivers' average. This means that water releasing from reservoir would be an important channel for atmospheric CO 2 sources. The annual CO 2 emission from the Danjiangkou Reservoir was estimated to be 3.4 × 10 9 mol C. Remarkably spatial and temporal heterogeneities in CO 2 flux from China's hydroelectric reservoirs are urgently included for advancing global models of reservoirs' carbon emissions.

The more recent work by Barros et al. (2011)  tification due to more than 80 000 reservoirs in operation and many projected in China, as reflected by Barros et al. (2011) that carbon fluxes from reservoirs correlate well to reservoir age, size, latitude, and environmental factors.This was corroborated by distinct sample size and incomplete coverage of spatial cases in particular, which contributed to large variability of CO 2 releases from global reservoirs.For example, the areal flux of CO 2 in the temperate reservoirs averaged 387-1400 mg CO 2 m −2 d −1 (St. Louis et al., 2000;Barros et al., 2011).A broad ranges were also found for CH 4 release from global hydroelectric reservoirs, i.e., from 4 Tg CH 4 yr −1 (Barros et al., 2011) to 100 Tg CH 4 yr −1 (Lima et al., 2008).Thus, more data particularly in China is urgently fueled for precise assessment on the liberation of GHG from hydroelectric reservoirs.
China, one of the largest hydroelectricity producer has a hydropower capacity of 654 × 10 9 kWh yr −1 , contributing 14 % to the total national electricity.More than 40 000 reservoirs are located in the Yangtze basin with subtropical monsoon climate, while very few reports concern their GHG emissions.Until now, methane efflux has only been studied for the Three Gorges Reservoir (TGR) (Chen et al., 2009(Chen et al., , 2011;;Yang et al., 2012;Zhao et al., 2013), Miyun (Yang et al., 2011) and Ertan (Zheng et al., 2011), however, more limited studies were conducted on CO 2 emissions from hydropower reservoirs, i.e., two papers from ISI-listed journals (e.g., cascade reservoirs (Hongfeng, Baihua, Xiuwen and Hongyan) in the Wujiang of the Yangtze basin by Wang et al., 2011, as  Similar to reservoirs in the tropical zone, Chinese hydroelectric reservoirs also receive world-wide concerns mainly because of very high methane flux (6.7 ± 13.3 mg CH 4 m −2 h −1 ) and CO 2 flux (88-175 mmol m −2 d −2 ) from the Three Gorges Reservoir (TGR) (Chen et al., 2009;Zhao et al., 2013) and consequently have been described as "GHG menace" reported in Nature News (Qiu, 2009).However, this perspective is challenging by the drastically downward revision of the previous emission rate.For example, the subsequent measurements of 0.29 and 0.18 mg CH 4 m −2 h −1 (Yang et al., 2012), 3-4 % of the previous CH 4 flux in the drawdown area of the TGR (Chen et al., 2011) implied carbon emission from hydroelectricity could be largely overestimated.Spatiotemporal sampling and varied calculated methods caused large differences in CO 2 flux from monsoonal rivers such as Yangtze and Pearl River systems (Yao et al., 2007;Wang et al., 2011;Li et al., 2012).These could potentially cause an over-estimate of CO 2 flux from TGR, where extremely high level occurs when compared to other China's hydropower reservoirs and is comparable to the tropical reservoirs (e.g., Petit Saut and Balbina) (cf.Zhao et al., 2013;Guerin et al., 2006).Therefore, more cases should be developed to better constrain the GHG emission from China's hydroelectricity Consequently, our study here focuses on the Danjiangkou Reservoir in the Han River, which belongs to Yangtze drainage basin and has a subtropical monsoon climate.The main objectivities are to (1) examine the spatial and temporal changes of partial pressure CO 2 (pCO 2 ), (2) unravel the mechanisms controlling the variability of pCO 2 and (3) quantify the water-air interface CO 2 flux in the surface water of the reservoir.

Study area
The Danjiangkou Reservoir (32  area of approx.95 000 km 2 which includes the upper Han River and Dan River basins (Fig. 1; Li et al., 2008bLi et al., , 2009a)).The Reservoir drains a region of northern subtropical monsoon climate with distinct transitional climatic characteristics.The annual mean temperature is 15-16 • C. The average annual precipitation is 800-1000 mm with large inter and intra-annual variability, and 80 % of which concentrates in the time period of May through October (Li et al., 2009a;Li and Zhang, 2010).Similar to the rainfall patterns, the 41.1 × 10 9 m 3 yr −1 runoff from its upper basin to the Reservoir also shows large annual and interannual variability.For example, in the historical records, the flood peak was 34, 300 m 3 s −1 , the mean maximum monthly flow was 7500 m 3 s −1 in the wet season and the minimum flow in the dry season was 64 m 3 s −1 (cf.Li et al., 2009a).
Currently, the water level of the Reservoir is 157 m, and its corresponding water surface area and storage capacity are 745 km 2 and 17.5 × 10 9 m 3 , respectively.Due to the China's South to North Water Transfer Project, the water level will be 170 m with the dam height increasing from the current 162 m to 176.6 m, the water surface area will reach 1050 km 2 when the project is finished in 2014.In our study, five sampling sites during 2004-2006 while eight sites for 2007-2011 were geolocated in the Reservoir using a portable Global Position System (Table 1 and Fig. 1).

Water sampling and analyses
21 field surveys were conducted for water samplings in the Danjiangkou Reservoir during 2004-2011 (Table 2).A total of 432 grab water samples (3 L), consisting of three replicates within a 200 m diameter circle, were taken from 50-100 cm depth below water surface using previously acid-washed high-density polyethylene (HDPE) containers.Thus, a total of 144 samples were pretreated for laboratory analysis after mixing with replicates.The acid-washed containers were rinsed thrice with sampled water on site, and the rinsing process was carried out downwind.A 500 mL subsample was filtered in situ through a previously acid-washed 0.45 µm pore Millipore nitrocellulose membrane filter.The initial portion of the filtration was discarded to clean the membrane.Introduction

Conclusions References
Tables Figures

Back Close
Full A small portion of the filtrate was used for measuring major anions Cl − , F − and SO 2− 4 , while another portion for major cation analyses was acidified to pH < 2 using ultra-pure concentrated nitric acid.Filtrates for measurements of dissolved organic carbon (DOC), dissolved nitrogen (DN) and solute reactive phosphate (SRP) were also prepared.Raw water samples for determination of total phosphorus (TP), total nitrogen (TN), total organic carbon (TOC), biogeochemical oxygen demand (BOD) and chemical oxygen demand (COD) were acidified with 10 % (v/v) sulphuric acid to pH < 2. Filtrates and raw samples were all stored in pre-cleaned HDPE bottles, and transported to the laboratory in a refrigerator at 4 • C.
Water temperature (T ), pH, dissolved oxygen (DO), oxidation-reduction potential (ORP), electrical conductivity (EC), turbidity, nitrate-nitrogen (NO − 3 -N) and ammoniumnitrogen (NH − 4 -N) were determined on site using a YSI 6920 (YSI Inc., Yellow Springs, Ohio, USA).The instrument was calibrated at 0 and 100 % oxygen saturation before and after usage for DO measurement.The pH electrode was calibrated at 7 and 10, and turbidity electrode at 123 and 0 before sampling.The nitrogen sondes were both calibrated at 100 and 1 mg L −1 before sampling.The membranes used for filtration were dried at 63 • C to constant weight, and total suspended solid (TSS) was calculated from the difference in the filter paper weights before and after filtering.Alkalinity, a measurement of bicarbonate buffering components in solution, was determined by titration of hydrochloric acid (0.020 M) in situ to pH 4.5 using Methyl orange (Li and Zhang, 2008, 2009, 2010).Total phosphorus (TP) was analysed with acidified molybdate to form reduced phosphor-molybdenum blue and measured spectrophotometrically at 700 nm, with the method detection limit (MDL) of 0.01 mg L −1 (CSEPB, 2002).SRP was determined by the same analytical method as TP, but measured spectrophotometrically at 882 nm (CSEPB, 2002).TN and DN were determined by alkaline potassium persulfate digestion-UV spectrophotometric method, and TOC and DOC were determined using TOC analyzer (MultiN/C2100 TOC/TN, Jena).Chlorophyll a (Chl a) was analysed using spectrophotometer.Determinations of BOD and COD were following Li et al. (2008a).Introduction

Conclusions References
Tables Figures

Back Close
Full The methods given by the Chinese State Environment Protection Bureau were followed in all the procedures (CSEPB, 2002).Anions (Cl − , F − and SO 2− 4 ) were measured using a Dionex Ion Chromatograph (Dionex Corporation, Sunnyvale, California, USA).An inductively-coupled plasma optical emission spectrometer (ICP-OES) (IRIS Intrepid II XSP DUO, USA) was used to determine the concentrations of major cations (K + , Ca 2+ , Na + and Mg 2+ ), Si and dissolved phosphorus.Reagent and procedural blanks were determined in parallel to the sample treatment using identical procedures.Each calibration curve was evaluated by analyses of these quality control standards before, during and after the analyses of a set of samples.Our analytical precision for major cations was better than ±10 %, and ±5 % for major anions.

pCO 2 calculations
There are direct and indirect methods using acidimetric titrations for pCO 2 determinations.The direct method, named the headspace technique, has been described in detail by Hope et al. (1995).The indirect method, much simplified procedure of alkalinity determination by titration to pH 4.5, has been widely used, i.e., pCO 2 is calculated via pH/DIC and pH/alkalinity in particular (Neal et al., 1998a,b;Telmer and Veizer, 1999;Yao et al., 2007;Butman and Raymond, 2011;Wang et al., 2011;Li et al., 2012).This titration based method for measurements of dissolved CO 2 are well acknowledged especially for natural rivers with pH > 6 and low organic carbon content.

Conclusions References
Tables Figures

Back Close
Full

Water-air interface CO 2 flux calculations
The diffusion flux of CO 2 (F ) across the water-air interface can be estimated based on a theoretical diffusion model: Where k is the gas transfer velocity of CO 2 (also referred to as piston velocity), pCO 2water in µatm is the partial pressure of CO 2 in the surface water, and pCO 2air in µatm is the CO 2 concentration in equilibrium with atmosphere.K h is the solubility of CO 2 corrected using temperature.This model has been widely used (i.e., Telmer and Veizer, 1999;Richey et al., 2002;Yao et al., 2007;Teodoru et al., 2009;Alin et al., 2011;Wang et al., 2011;Li el., 2012).
The piston velocity (k in cm h −1 ) of CO 2 at the water-air interface is affected by different basin physical factors but largely by wind speed and water turbulence (Telmer and Veizer, 1999;Alin et al., 2011).Extensive efforts have been made to improve the reliability of k term for highly precise and accurate methods of water-air CO 2 evasion, such as k values from a suit of empirical functions of wind speed (i.e., Wanninkhof, 1992;Raymond and Cole, 2001;Alin et al., 2011).However, k values from their models showed large differences (Zhai et al., 2007).
The riverine k values showed wide ranges of 3 cm h −1 to 115 cm h −1 (Aucour et al., 1999;Raymond and Cole, 2001), however, 8-15 cm h −1 for k values has been widely adopted for large rivers, for instance k values (8-15) in the River Rhone and Saone (Aucour et al., 1999), 10 cm h −1 for Amazon mainstream (Raymond and Cole, 2001) and 15 cm h  2011).Thus, we estimated an averaged k of 15 cm h −1 in the river downstream the dam considering the normal k level of 10 cm h −1 due to the similar hydrographic features with other Yangtze River system and world rivers (Aucour et al., 1999;Yao et al., 2007;Wang et al., 2011), and 20 cm h −1 for reservoir discharge with high turbulence.Considering the mean wind speed in the Dnajiangkou Reservoir is around 2.5 m s −1 , thus, the k level of water surface in the Danjiangkou Reservoir can be calculated using a function of wind speed and temperature (Cole and Caraco, 1998): (1) (2) where k T is the measured values in situ temperature (T ), Sc CT is the Schmidt number of temperature T (unit in • C), and U 10 is the wind speed at 10 m above waters (m s −1 ).

Statistical analyses
Multivariate statistics such as correlation analyses and analysis of variance (ANOVA) were used in our study.Correlation analyses was employed for relations between pCO 2 and environmental parameters, while ANOVA was performed for differences of spatial and monthly DIC system with significance at p < 0.05.The statistical processes were conducted using SPSS 16.0 (Li and Zhang, 2009;Li et al., 2013).Introduction

Conclusions References
Tables Figures

Back Close
Full  (November 2008) (Table 2).In total, water temperature, pH and alkalinity averaged 18.75 • C, 8.22 and 2385 µmol L −1 , respectively and all of them exhibited significant monthly differences (p < 0.01 by ANOVA; Table 4a).Contrary to the monthly variations, T and pH in the Danjiangkou Reservoir showed insignificant variability among sampling sites (Table 4b).T and pH levels in the reservoir surface were higher than the site downstream the dam (DJK5) with lowest averages of 16.4±8 • C for T and 8.01±0.38 for pH (Table 3).However, alkalinity showed significant variations among sites with highest and lowest levels of 2736 ± 437 µmol L −1 (DJK1 in the Dan Reservoir) and 2117 ± 390 µmol L −1 (DJK4 in the Han Reservoir), respectively.Generally, alkalinity significantly decreased in the Dan Reservoir as water flows (R 2 = 0.68, p < 0.01) till the site (DJK 4) in the Han Reservoir, and then increased significantly onward till the downstream of the dam (R 2 = 0.93, p < 0.01).In total, alkalinity in the different zones of the Reservoir was following in an order of Dan Reservoir > river downstream the Reservoir > site close to dam > Han Reservoir.

Dissolved inorganic carbon (DIC) species
The DIC species calculated using CO 2 SYS were presented in Figs.As for pCO 2 downstream the Reservoir, highest level (4764 µatm) occurred in July, and lowest level (155 µatm) in January 2010 (Fig. 3).pCO 2 concentrations in the site downstream the dam were consistently higher in relation to those in the reservoir surface when November 2008 was excluded (Figs. 2 and 3).Spatial patterns indicated great variations from 53-3751 µatm and 155-4764 µatm of pCO 2 upstream and downstream the dam, respectively (Fig. 4).The box-whiskers plots illustrated that pCO 2 median levels significantly decreased as water flows in the Dan Reservoir (R 2 = 0.41, p < 0.05).There was a significant decrease (R 2 = 0.61, p < 0.05) in terms of pCO 2 medians from Dan Reservoir-Han Reservoir-Dam with lowest level (320 µatm) in the site close to Dam.The maximal median (702 µatm) was observed downstream the dam (Fig. 4). ) (Fig. 5b).F CO 2 downstream the dam was much higher than the reservoir surface, i.e., averages of 119 vs 9 mmol m −2 d −1 for downstream and upstream the dam, respectively (Fig. 5).In total, the Danjiangkou Reservoir emitted 2.4×10 9 mol CO 2 yr −1 in the current situation while 3.4×10 9 mol CO 2 yr −1 from 2014 onward, respectively representing 8.3 % and 11.7 % the CO 2 emission from Three Gorges Reservoir (ca.29.5 × 10 9 mol CO 2 yr −1 ; Zhao et al., 2013).

Controls on aqueous pCO 2 (CO 2 flux)
Potential processes such as soil CO 2 by runoff and in situ aquatic respiration of organic carbon would elevate pCO 2 in water, and resulted in oversaturated CO 2 in the rivers worldwide with respect to atmosphere (cf.Cole and Caraco, 2001;Richey et al., 2002;Yao et al., 2007;Sarma et al., 2011;Wang et al., 2007Wang et al., , 2011;;Li et al., 2012), albeit photosynthesis and water-air interface CO 2 evasion decrease the aqueous pCO 2 (Cole and Caraco, 2001;Zeng and Masiello, 2010).Temperature and monsoonal precipitation in the upstream of the Danjiangkou Reservoir (Fig. 6) thus have large effects Introduction

Conclusions References
Tables Figures

Back Close
Full on aquatic pCO 2 levels due to that rainfall alters the biogeochemical processes in the terrestrial ecosystem and organic carbon composition in the Reservoir (Yao et al., 2007;Zeng and Masiello, 2010).This hypothesis was corroborated by pronounced seasonality in pCO 2 concentrations (Figs. 2 and 3).Previous researches reported that negative or positive contributions of rainfall events to pCO 2 depend on rainfall intensity and basin impermeability (Ho et al., 1997;Yao et al., 2007;Zeng and Masiello, 2010;Li et al., 2012), for example, initial rainfall enhances the export of soil CO 2 and soil organic carbon load to river system, while continued rain especially the concentrated precipitation lower soil permeability and consequently shows dominant dilution effects.These mechanisms could explain large variability in pCO 2 occurring in the wet months (ca.August 2005 vs August 2009) (Fig. 2).In the year of 2009, persistent precipitation (22 July to 5 August) and particularly the extreme storm (i.e., 62 mm d −1 ) from the onset of August (Fig. 6) had flushed out the soil CO 2 and hence storm dilution effects dominated, which resulted in the trough of pCO 2 on 11 August (Fig. 2).
Samplings in 2005 allowed us analyze seasonal controls on pCO 2 in a hydrological year (Fig. 7).Little rainfall occurred between the date of sampling in January and April in 2005, while rise in temperature and algae blooming increased photosynthesis in the Spring season would reduce aquatic pCO 2 (from 372 to294 µatm).From May afterward, more rainfall events occurred due to monsoonal effects, and enhanced pCO 2 in May could be assumed albeit no sampling in May 2005.However, this could be supported by the pCO 2 concentrations observed in May 2007 (582 µatm) and May 2011 (650 µatm), which was likely caused by export of soil CO 2 by rain runoff.Weaker biological CO 2 uptake through photosynthesis due to lower sun angle and high-turbid water might be additional reasons for the rapid pCO 2 increase.The continued rain especially the storm water with lower permeability or without infiltration into soil contributed to the trough of pCO 2 (150 µatm) in June.However, from the late of July to the mid-August, proper temperature and wetted soils by lower precipitation favored soil bacterial respiration and thus higher soil CO 2 content, consequently making elevated aqueous pCO 2 via baseflow and interflow albeit more rainfall events in 13-19 August partially con-Introduction

Conclusions References
Tables Figures

Back Close
Full tributed to dilution effects.As expected, precipitation showed a quick decrease from the mid-October onward, evenly distributed precipitation and appropriate rainfall, however, provided optimum environment for soil respiration (cf.Zeng and Masiello, 2010;Li et al., 2012).This allowed the rain water to infiltrate and flush out soil CO 2 to the river with limited dilution effects, and therefore contributed to the crest level of pCO 2 in November (Fig. 7), as indicated by other two major pCO 2 peaks in November in 2008 and 2009 (Fig. 2).Henceforth, little rainfall and lowest temperature in December through January limited the export of soil CO 2 to rivers, which resulted in very low pCO 2 in January (the most cold and dry month) (Fig. 7).
Consistent with previous studies (Richey et al., 2002;Yao et al., 2007;Zeng and Masiello, 2010;Li et al., 2012), pCO 2 in the reservoir surface was generally higher in the warm and wet season (May-November) than in cold and dry season (December-April) (Figs. 2 and 7).We also concluded that enhanced photosynthesis due to abundant phytoplankton and high water temperature could not explain clearly pCO 2 fluctuations by a factor of more than 10 in the monsoon season (Fig. 2) (Cole et al., 1992).This was corroborated by previous studies (cf.Zeng and Masiello, 2010) that diurnal pCO 2 showed little changes in the day with little rain while large differences in the rainier day, as well as little differences of diurnal pCO 2 in the Nihe Reservoir, China (Lu et al., 2010).Different monthly patterns of pCO 2 occurred in the river downstream the Danjiangkou Dam when compared to reservoir surface waters (Fig. 3).This could be responsible for artificial water regulation, for example, water discharges via deep turbines before monsoonal flooding arrival for flood control strategy made the extremepCO 2 in July.However, the second pCO 2 peak level in November 2009 was due to autumn flooding.
As water flows in the reservoir, aquatic ecosystem gradually transformed from "heterotrophy" to "autotrophy" with more photosynthetic uptake of CO 2 close to dam (Saito et al., 2001).This could explain the significant decrease as water flows in the Dan Reservoir (R 2 = 0.41, p < 0.05 for pCO 2 medians).However, both pCO 2 medians and Introduction

Conclusions References
Tables Figures

Back Close
Full means were higher in the Dan Reservoir than in the Han Reservoir, which could be contributable to higher pollution load and thus relatively lower pH in the Dan Reservoir zone (Li et al., 2009a).Shifts of hydrological dynamics, nutrient structure and complicated biogeochemical process in the reservoir generally reduced the pCO 2 close to the dam, as indicated by significant decreases in pCO 2 medians from Dan Reservoir-Han Reservoir-Dam (R 2 = 0.61, p < 0.05; Fig. 4), as well as other cases in China such as Wan'an (Mei et al., 2011) andXin'anjiang Reservoirs (Yao et al., 2010).Turbines with intake of deep water in the Danjiangkou Reservoir produce a hydroelectricity of 45 × 10 8 kWh yr −1 , this water releases had hypoxic environment with lower pH and higher CO 2 concentration (cf.Wang et al., 2011), as well as considerably enhanced pCO 2 in the river downstream the dam.Similar results were obtained in China's other hydroelectric reservoirs such as Hongjiadu (Yu et al., 2008) and Xin'anjiang (Yao et al., 2010).

Correlations between water quality and pCO 2
Apart from precipitation prevailing on monthly pCO 2 variations, as well as hydrological and biogeochemical processes on spatial pCO 2 variations in the reservoir area, water chemical variables including nutrients (ammonium-N, nitrate-N, total-N, dissolved N, soluble reactive phosphorus, dissolved phosphorus, total phosphorus (TP), DO%), organic carbon (TOC, DOC), biogenic elements (Cl − , SO 2− 4 , Na + , K + , Ca 2+ , Mg 2+ and Si), water pollution parameters (chemical oxygen demand (COD), biochemical oxygen demand (BOD), total suspended solid (TSS), turbidity) and Chl a were correlated to pCO 2 to gain insights of the possible controls (Fig. 8; Table S2).Our data indicated that pCO 2 was significantly and positively related to Si, TP and DOC and negatively to DO saturation, TN and Chl a, while slightly to other water variables including biogenic elements.
The increased nitrogen could enhance aquatic photosynthesis and thus O 2 production, while reduce dissolved CO 2 level (e.g., Cole and Caraco, 2011;Wang et al., 2007), which was relevant to the negative relationships between pCO 2 and nitrogen, Chl a and 10070 Introduction

Conclusions References
Tables Figures

Back Close
Full DO concentration (Table S2).Higher organic carbon appeared to facilitate high bacteria respiration (cf.Sarma et al., 2009Sarma et al., , 2011;;Zeng and Masiello, 2010), and thus led to increase in pCO 2 , while O 2 was consumed simultaneously.This process was consistent with the algal photosynthesis process that caused opposite changes between pCO 2 and O 2 The results were in excellent agreement with the observations in other China's hydropower reservoirs, e.g., Shuibuya (Zhao et al., 2012), Wan'an (Mei et al., 2011) and Three Gorges Reservoir (Zhao et al., 2011(Zhao et al., , 2013)).Moreover, aquatic ecosystems tend to have an optimum nutrient stoichiometry of phytoplankton (e.g., Redfield Ratio; Redfield, 1958).Phytoplankton via photosynthesis assimilates dissolved CO 2 and nutrients such as nitrogen, phosphorus, and silica in a stable element stoichiometric ratio.Thus, dramatic increase in nitrogen load due to anthropogenic inputs in the upper basin of the reservoir (Li et al., 2009b(Li et al., , 2013) ) could stimulate the assimilation of dissolved CO 2 and P and Si by aquatic photosynthesis.Slightly increase in P and exclusive source from silicate weathering for Si in the upper Han River (Li et al., 2009c) understandably resulted in positive correlations between pCO 2 and Si and P.These trends were, in general, in concurrence with the results elsewhere (cf.Wang et al., 2007).Yet these processes could not explain the associations between pCO 2 and nitrogen species and biogenic elements, this could be attributed to multi-collinear effects of water quality properties, for example, varied effects of nitrate on pCO 2 were observed (cf.Wang et al., 2007;Li et al., 2012).which resulted in that total CO 2 emission from the Danjiangkou was an order of magnitude lower than the TGR albeit they have same water surface area (Zhao et al., 2013).Our estimated CO 2 emission rate was much higher than some temperate reservoirs (e.g., Wallula, New Melones and Dworkshak): these three reservoirs acted as the atmospheric CO 2 sink, however, much lower than those from boreal and tropic reser-10071 Introduction

Conclusions References
Tables Figures

Back Close
Full voirs.For example, the estimate of CO 2 flux was around 40 % of that from Arrow-Upper (boreal) and only 1 % that from Samuel (tropic) (see Table 5).
On the basis of the available literature, Barros et al. (2011) recently estimated an average of CO 2 flux ∼ 8.8 mmol m −2 d −1 from temperate hydroelectric reservoirs (25-50 • latitudinal belt).Our calculated CO 2 flux from the Danjiangkou Reservoir equaled to the currently estimated global average for the mid-latitude reservoirs, however, there exited significantly geographical heterogeneities in CO 2 flux from China's hydroelectric reservoirs (see Table 5).For the global estimates, data on CO 2 emission China's reservoirs have been excluded (cf.St Louis et al., 2000;McCully, 2006;Aufdenkampe et al., 2011;Barros et al., 2011).This indicated that the current global estimation of CO 2 emission from reservoirs could be under-estimated or overestimated, as reflected by the wide ranges (see Table 5).It is clear therefore that more data particularly in China, where a large number of hydropower dams exist with many planned for the future, are mandatory to better constrain the global CO 2 emission from hydroelectric reservoirs.Worldwide riverine pCO 2 was primarily 2-20 times supersaturated in CO 2 relative to the atmosphere (Cole and Caraco, 2001), our calculations of 1132±1220 µatm (ranges of 155-4764 µatm) in the river downstream of the Reservoir was three times lower than global river's average (3230 µatm calculated from world 47 large rivers; Cole and Caraco, 2001).However, our results were comparable to the Chinese Rivers such as Yangtze (1297 ± 901 µatm; Wang et al., 2007), Pearl (450-2360 µatm;Zhang et al., 2007), Yellow (1137±189 µatm; Cole and Caraco, 2001) Our areal flux of CO 2 in the river downstream the dam was intermediate relative to rivers in China (Table 5).For example, the CO 2 emission rate in the river downstream the Danjiangkou Reservoir was 8 times higher than that in the main channel of the Yangtze (ca.14 mmol m −2 d −2 ), while it was much lower with respect to the river reach

Conclusions References
Tables Figures

Back Close
Full  −2 ).When compared to the world rivers, the estimated CO 2 emission flux per unit area was much lower than those in tropic rivers (e.g., Amazon), while was intermediate between those of temperate rivers and comparable to the global average (e.g., 147 mmol m −2 d −2 ).
Similar to other reservoirs in China, downstream waters had quite high flux of CO 2 (cf.Yu et al., 2008;Wang et al., 2011).This could be explained as follows: turbines for hydroelectricity generation are located in the deep water areas, and this deep water with prevailing respiration property shows very high dissolved CO 2 concentration due to low permeability light and thermocline in particular.For example, CO 2 flux in the reservoir downstream was 13 fold that from reservoir surface in the Danjiangkou, and this could reach as high as 15 for Hongjiadu (Yu et al., 2008) and 33 for Hongfeng (Wang et al., 2011) (see Table 5).This suggested deep water releases via turbines is an important sources of atmospheric CO 2 (cf.Kemenes et al., 2011).

Conclusions
CO 2 outgassing flux across water-air interface in the Danjiangkou Reservoir was 1/13 lower than the flux from the river downstream the dam.Average CO 2 flux and annual CO 2 emission from the Reservoir were 9 mmol m −2 d −1 and 3.4 × 10 9 mol C yr −1 , respectively.CO 2 flux from the Danjiangkou Reservoir was near the lower end of the ranges from hydroelectric reservoirs in China, clearly lower than other global estimates for reservoirs (e.g., St Louis et al., 2000;McCully, 2006;Aufdenkampe et al., 2011) and natural lakes (Barros et al., 2011), while similar to global average for temperate hydroelectric reservoirs estimated by Barros et al. (2011).Substantially monthly and spatial variations in pCO 2 and CO 2 flux demonstrated the dominant control of rainfall events particularly in the monsoon seasons, while biologic CO 2 uptake through aquatic photosynthesis dominated the spatial distributions in the reservoir area.The much higher CO 2 flux in the river downstream the reservoir was due 10073 Introduction

Conclusions References
Tables Figures

Back Close
Full    Full  Full  Full     S2).
estimated 48 Tg C as CO 2 and 3 Tg C as CH 4 annually from hydroelectricity, respectively, a downgrade from earlier estimates of 321 Tg C yr −1 .The estimate was based on 85 globally distributed hydroelectric reservoirs from the Americas and Northern Europe.No reservoirs from Asia especially from China were included, which could be a key component that limits the C emission quan-Discussion Paper | Discussion Paper | Discussion Paper | well as TGR by Zhao et al., 2013), and five Chinese journal papers (Hongjiadu by Yu et al., 2008; Xin'anjiang by Yao et al., 2010; Nihe by Lu et al., 2010; Wan'an by Mei et al., 2011; Shuibuya by Zhao et al., 2012, which are not readable by international scholars).Therefore, carbon emission from reservoirs in China largely lags behind.Combining very limited studies with clearly spatial (cf.10-90 mg CO 2 m −2 h −1 for) and temporal (cf.−22-330 mg CO 2 m −2 h −1 ) heterogeneities in CO 2 flux relating to China's hydroelectricity, we can urge that current data paucity of carbon release from China's reservoirs is constraining the accurate quantification of global carbon emission from hydroelectric reservoirs.Discussion Paper | Discussion Paper | Discussion Paper | 36 -33 • 48 N, 110 • 59 -111 • 49 E) built in 1970s is situated in the juncture of Hubei and Henan provinces, Central China.It has a drainage Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | −1 for St Lawrence (Yang et al., 1996).Albeit 8-15 cm h −1 of k value was designated in Xijiang River and Wujiang River (Yao et al., 2007; Wang et al., 2011), Alin et al. (2011) got the k 600 level of 22 cm h −1 at a temperature of 20 • C for Asian monsoon river (e.g., Mekong), and 20 cm h −1 for temperate rivers (Aufdenkampe et al., Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 29 ± 1.4 2, 3 and 4. HCO − 3 ,showing similar trends with DIC in space and month, was the dominant component of DIC for all samples, accounting for ∼ 93.5 % of DIC on average, followed by dissolved CO 2− 3 and CO 2 , accounting for 5.3 % and 1.2 %, respectively.Also, HCO − 3 accounted for almost all of the alkalinity, up to an average of 97.4 %.There were pronounced monthly and spatial variations in pCO 2 illustrated by boxwhiskers plots (Figs. 2, 3, and 4).Monthly statistics of pCO 2 in the Reservoir area generally showed same medians and means with the ratio of 0.83-1.17(median/mean), while extremely spatial variations resulted in that the mean concentration (491 µatm) Discussion Paper | Discussion Paper | Discussion Paper | was three-fold the median (154 µatm) in July 2006 (Fig. 2).Both of them displayed the minimal and maximal concentrations in the flooding season (ca.June-November), i.e., pCO 2 median values ranged 90 µatm in August 2009 to 1530 µatm in November 2008, while 110 (August 2009)-1580 µatm (November 2008) for pCO 2 mean values (Fig. 2).
pCO 2 mean values significantly decreased from the Dan Reservoir to Han Reservoir (R 2 = 0.42, p < 0.05), then increased quickly till the downstream of dam (R 2 = 0.88, p < 0.05) with lowest (477 µatm) and highest levels (1132 µatm) in sites D2 (Han Reservoir) and DJK5 (downstream of the dam), respectively.The pCO 2 mean values were consistently higher relative to pCO 2 medians, reflected by the ratio of mean to median from 1.14 (D1 in the Dan Reservoir)-1.87(D3 close to the dam).Regarding to respective sampling site, large variation factors of pCO 2 (max./min.)varied from 9 at DJK2 in the Dan Reservoir to 31 at the river downstream the Reservoir (Fig. 4).pCO 2 totally averaged 595±545 µatm (mean± S.D) in the reservoir surface area, which was around 1/2 of the average downstream the Reservoir (1132 ± 1220 µatm)Discussion Paper | Discussion Paper | Discussion Paper | 3.3 CO 2 diffusion fluxes of the Danjiangkou Reservoir The average CO 2 diffusion fluxes (F CO 2 ) were calculated, as illustrated in Fig. 5. F CO 2 was generally higher in the wet season, with averages of −8.2-31.4mmol m −2 d −1 , and in the dry season, it ranged from −5.8-10.6 mmol m −2 d −1 in the reservoir surface waters.Similar to the distribution of pCO 2 , F CO 2 demonstrated remarkable variations upstream and downstream the dam in each sampling month.Generally, F CO 2 downstream the dam was quite high, with ranges of −20-703 mmol m −2 d −1 in the wet season, and −38-125 mmol m −2 d −1 in the dry season (Fig. 5a).Spatial F CO 2 in the reservoir surface demonstrated the following descending order: Dan Reservoir (3.2-13.7 mmol m −2 d −2 ) > the site close to dam (6 mmol m −2 d −2 ) > Han Reservoir (3-3.4 mmol m −2 d −2 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | from the Danjiangkou Reservoir surface area was at the bottom in contrast to other China's hydroelectric reservoirs (e.g., 6-96 mmol m −2 d −2 ), Discussion Paper | Discussion Paper | Discussion Paper | , and other Asian river such as Mekong (703-1475 µatm, Alin et al., 2011; 1090 ± 290 µatm; unpublished), and other world Rivers, i.e., Hudson (1125 ± 403 µatm, Raymond et al., 1997), St. Lawrence (1300 µatm, Helli et al., 2002), Ottawa rivers (1200 µatm, Telmer and Veizer, 1999) and Mississippi (1335 ± 130 µatm, Dubois et al., 2010).

Fig. 1 .
Fig. 1.Locations of eight sampling sites in the Danjiangkou Reservoir, China (DJK1, DJK2, D1 and DJK3 are located in the Dan Reservoir zone, DJK4 and D2 in the Han Reservoir zone, and D3 and DJK5 in the upstream and downstream of the dam, respectively.D1, D2 and D3 were added from 2007 afterward).

Fig. 8 .
Fig. 8. Scatter plots between pCO 2 and some key water chemicals in the Danjiangkou Reservoir, China (Spearman's rho coefficients for pCO 2 and all the hydrogeochemical variables were tabulated in TableS2).
• C with minimum and maximum in winter (January 2010) and summer (July 2008), respectively.pH showed lower values in the monsoon rainy season, for example, June through November, ranging from 7.81 ± 0.18 (November 2009) to 8.65 ± 0.09 (January 2010).Alkalinity ranged from 1795±456 (August 2010) to 3241±332 µmol L to deep water releasing for power generation.Similar to other hydroelectric reservoirs in China, reservoir surface waters were generally supersaturated with CO 2 relative to atmosphere, while lower values of pCO 2 than the atmospheric level of CO 2 often occurred in the concentrated rainfall months and dry season.We concluded remarkably spatiotemporal variability in pCO 2 and CO 2 fluxes from China's reservoirs are urgently included for a substantial revision of global estimate of carbon emission, and water discharge from reservoirs should deserve extra attention for assessment of reservoirs' source and sink effects on atmospheric CO 2 .Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | reservoirs as regulators of carbon cycling and climate, Limnol.Oceanogr., 54, 2298-2314, 2009.Wang, F. S., Wang, B. L., Liu, C. Q., Wang, Y. C., Guan, J., Liu, X. L., and Yu, X. L.: Carbon dioxide emission from surface water in cascade reservoirs-river system on the Maotiao River, southwest of China, Atmos.Environ., 45, 3827-3834, 2011.Lu, F., Wang, X. K., Duan, X. N., Song, W. Z., Sun, B. F., Chen, S., Zhang, Q. Q.

Table 1 .
The main features of the Danjiangkou Reservoir, China.

Table 2 .
Descriptive statistics of monthly variability in T ( • C), pH and alkalinity (µmol L −1 ) of the Danjiangkou Reservoir, China.

Table 4 .
ANOVA for T , pH and alkalinity in month (a) and space (b) of the Danjiangkou Reservoir, China.