Diffusive emission of methane and carbon dioxide from two hydropower reservoirs in Brazil

The role of greenhouse gas emissions from freshwater reservoirs and their contribution to increase greenhouse gas concentrations in the atmosphere is currently under discussion in many parts of the world. We studied CO2 and CH4 diffusive fluxes from two large neotropical hydropower reservoirs with different climate conditions. We used floating closed-chambers to estimate diffusive fluxes of these gaseous species. Sampling campaigns showed that the reservoirs studied were sources of greenhouse gases to the atmosphere. In the Serra da Mesa Reservoir, the CH4 emissions ranged from 0.530 to 396.96 mg.m–2.d–1 and CO2 emissions ranged from –1,738.33 to 11,166.61 mg.m –2.d–1 and in Três Marias Reservoir the CH4 fluxes ranged 0.720 to 2,578.03 mg.m –2.d–1 and CO2 emission ranged from -3,037.80 to 11,516.64 to mg.m–2.d–1. There were no statistically significant differences of CH4 fluxes between the reservoirs, but CO2 fluxes from the two reservoirs studied were significantly different. The CO2 emissions measured over the periods studied in Serra da Mesa showed some seasonality with distinctions between the wet and dry transition season. In Três Marias Reservoir the CO2 fluxes showed no seasonal variability. In both reservoirs, CH4 emissions showed a tendency to increase during the study periods but this was not statistically significant. These results contributed to increase knowledge about the magnitude of CO2 and CH4 emission in hydroelectric reservoirs, however due to natural variability of the data future sampling campaigns will be needed to better elucidate the seasonal influences on the fluxes of greenhouse gases.


Introduction
Methane (CH 4 ) is the most abundant organic gas in Earth's atmosphere and has an important role to tropospheric and stratospheric chemistry, affecting for example, tropospheric ozone, hydroxyl radicals and carbon monoxide concentrations, stratospheric chlorine and ozone chemistry and, through its infrared properties, Earth's energy balance (Cicerone and Oremland, 1988).Wuebbles and Hayhoe (2002) have estimated that up to 0.6 Gt of methane are emitted annually into the atmosphere; moreover about 75% of this is produced exclusively by strictly anaerobic methanogenic microorganisms present in anoxic environments (Segers, 1998;Whitman et al., 2006).
In the same way CO 2 plays an important role not only for atmospheric chemistry but also to the chemistry of the biosphere due to its availability as a carbon source for photosynthesis.CH 4 is the third most important greenhouse gas after water vapor and CO 2 and has a Global Warming Potential (GWP) 25 times greater than CO 2 on a 100 year timescale (IPCC, 2007).According Dlugokencky and Tans (2012) and IPCC (2007) global concentrations of CH 4 and CO 2 in the atmosphere were 1,775 ppb and 394 ppm while in pre-industrial era no more than 715 ppb and 280 ppm, respectively.This trend of increased concentration in the atmosphere is more and more linked to anthropogenic activities such as livestock, changes in land use and mainly energy use (IPCC, 2007).
Hydro power reservoirs as artificial aquatic systems represent an important part of the Earth's continental territory.They have an important role in the aquatic biogeochemistry and have also many effects on the environment.Recently another important negative impact of dam construction has been reported: emission of greenhouse gases generated by flooding organic matter during reservoir formation.Since the beginning of the 1990's several scientists have argued that hydropower reservoirs, as well as natural ecosystems, emit biogenic gases by bubbling and by molecular diffusion (Rudd et al., 1993;Bartlett and Harriss, 1993;Kelly et al., 1997;Hamilton et al., 1995;Abril et al., 2005).
Knowledge of greenhouse gases emissions from hydroelectric reservoirs in Brazil becomes important since 83% of Brazilian electricity is produced by hydraulic sources (Brasil, 2012) and Brazil is the second largest producer of hydroelectricity, after China (IEA, 2012).
Research conducted by national and international teams has given successive contributions to the understanding of greenhouse gases emissions from Brazilian hydroelectric reservoirs (Rosa et al., 1994(Rosa et al., , 2003;;Guerin et al., 2006;Santos et al., 2006;Roland et al., 2010).
This study presents the results of measurements of CH 4 and CO 2 diffusive emissions from two large hydroelectric reservoirs at in the Brazilian Cerrado, in an attempt to improve quantity and quality of data available.

Experimental Methods
Four sampling campaigns were conducted for each reservoir in order to collect data covering all the hydrologic periods.Sampling sites was undertaken in Três Marias and Serra da Mesa in different seasons (Table 1).
In order to determine the CH 4 and CO 2 diffusive flux, a PVC chamber with a volume of 1000 mL and area of 0.047 m 2 was placed floating on the water surface.The method was described by Devol (1988Devol ( , 1990) ) and Bartlett et al. (1988Bartlett et al. ( , 1990)).All the samples were taken in vegetation-free areas both in the middle of reservoir and near the edges.One gas sample was taken from the chamber initially after 2, 4 and 8 minutes, counting from the initial moment when the chamber was placed on the water/air interface.A single sampling was used for each floating chamber point.The air samples inside the chambers (30mL) were collected by 60 mL polyethylene syringes and transferred to glass gasometric ampoules.All samples were taken between 9:00 and 17:00 h, local time.
CH 4 and CO 2 concentrations were determined in a field laboratory within 8 hours after collection, using a Varian CP-3800 chromatograph, with a thermal conductivity detector (TCD), FID (Flame Ionization Detector) and a PoraPLOT column.The chromatograph was calibrated using certified standards purchased from White Martins (Praxair).We use three calibration ranges for each gas: certified standard n.2432/11 (1,98 mg/L for CH 4 and 400 mg/L for CO 2 ), certified standard n.2440/11 (20,1 mg/L for CH 4 and 602 mg/L for CO 2 ) and certified standard n.2442/11 (50,2 mg/L for CH 4 and 998 mg/L for CO 2 ) The rate of gas concentration increase within the chamber, and thus the diffusive flux, was determined by linear regression of concentration/time data sets (IEA, 2012).According to the IEA guidelines, fluxes were considered valid only when the regression coefficient (R 2 ) was greater than 0.85 the root-mean-square error was less than 0.11 (IEA, 2012).The samples that not meet these requirements were discarded.
The Kruskal-Wallis test was used to verify possible differences in emissions between the two reservoirs and to check for differences among the sampling campaigns of each reservoir."R statistic" software was used for statistical assessment (The R Foundation, 2012).

Results
Of 162 fluxes for each gas has measured at Serra da Mesa Reservoir, 5% of fluxes of CH 4 and 9% of CO 2 were discarded.Considering the whole sample period, CH 4 emissions ranged from 0.530 to 396.96 mg.m -2 .d - and CO 2 emissions ranged from -1,738.33 to 11,166.61mg.m - .d - .
In Três Marias Reservoir we have measured 186 CH 4 fluxes for each gas, of which, 10% of fluxes of CH 4 and 13% of CO 2 were discarded.CH 4 emissions in Três Marias ranged from 0.720 to 2,578.03mg.m -2 .d - and CO 2 emission ranged from -3,037.80 to 11,516.64 mg.m -2 .d - .The fluxes measurements from four field campaigns are shown in Table 2.
Figure 2 shows historical data series of the rainfall distribution 17 years (from 1975 to 1992 and 2011 to  in regions of the reservoirs as well as the median emission measurements.And as shown in Figure 3 we can see the median values and the outliers of CH 4 emissions in both reservoirs.The use of median results as robust description of gas fluxes and comparison of others central tendency statistical descriptors can be read in (Damazio et al., 2013).
In this current study we have made comparisons of measured fluxes among the period studied.Regarding the comparison of CH 4 fluxes, statistically significant distinctions between the periods studied were not found in Três Marias Reservoir (see Table 3).
We can say the same thing for the comparisons of CH 4 fluxes among sampling campaigns in the Serra da Mesa Reservoir.An exception was observed in the fluxes   measured in April, which was particularly lower than in other periods (see Table 4).
The Figure 4 suggest a certain seasonality of CO 2 emission in the Serra da Mesa Reservoir', due to differences among the fluxes from rainy-transition (January vs. April and October) and dry-transition (April vs July and October).Furthermore, emissions measured in transition months are different between themselves (April vs. October).(see Table 5).
The Três Marias Reservoir showed no seasonality with regards to CO 2 emission, since we found no statistically significant difference, except for the emissions measured in May and August (see Table 6).

Discussion
In 1998 and 1999, Santos et al. (2006) measured diffusive emission at Serra da Mesa (range from -6,048 to 10,178 mg.m -2 .d - to CH 4 and -5,360 to 5,903 mg.m -2 .d - to CO 2 ) and Três Marias Reservoir (range from 0.660 to 241 mg.m -2 .d - to CH 4 and -10,060 to 7,346 mg.m -2 .d - to CO 2 ).Thus, the highest emissions that we found in the present study were higher than those found by Santos et al. (2006) in the previous study, with the exception of CH 4 emissions in Serra da Mesa, which in this study had lower values.
The emissions peak in the first years after filling a reservoir tends to decrease and to stabilize over the subsequent years.In older reservoirs (over 10 years) in boreal and temperate regions, emissions of greenhouse gases are similar to natural lakes.However, in the tropics, the time to return to natural values may be longer, depending on the water quality (Tremblay et al., 2005).We suggest that both the natural variations and external anthropogenic factors, such as the organic material supply, are contributing to maintain high value in the Serra da Mesa Reservoir and Três Marias Reservoir, even 13 years after these early studies (Santos et al., 2009;Fonseca, 2010;Chandrasekera, 2000).
The results shown in Figure 2 suggest that there is a general trend of increase in median values of CH 4 emissions    No difference Jan-Oct 11.018162 27.19567No difference Jul- Oct 4.127289 26.16606No difference in both reservoirs throughout the year, recording the lowest in April and highest in October despite being in the same hydrologic period and confirmed by the statistically significant differences in the flow of CH 4 .However in Três Marias Reservoir, the lowest value was recorded in March (end of the rainy season and very close to the rainy-dry transition season) and the highest in November (beginning of the rainy season and the end of the period of dry-rainy transition).We suggest that this trend of emissions are somehow related to transitional periods due to changes in the pattern of temperature and rainfall.When we compare the CH 4 emission shown in Figure 3, considering the significance level of <0.05, we found no statistically significant difference between the two reservoirs studied (Kruskal-Wallis chi-squared = 3.8217, df = 1, p-value = 0.0509).However, this value was considered borderline for the test, since the observed value of the test statistic is slightly smaller than the critical value and this must be exceeded to be considered a statistically significant difference (SMR-TMR Difference observed = 20.4066critical difference = 20.45932Result = There are difference).On the other hand the CO 2 fluxes between the reservoirs studied showed statistically significant differences (Kruskal-Wallis chi-squared = 21.7085,df = 1, p-value = 3.17 -6 ) possibly due to the median values obtained in the months of August in Três Marias Reservoir and July in Serra da Mesa reservoir (Table 2).
In the present study, the CO 2 fluxes measured in January (rainy season) and July (dry season) in Serra Mesa Reservoir (as shown in Figure 4) proved to be indistinguishable from each other, but they are different when compared to April and October which are transition months from wet to dry and from dry to wet season, respectively.Moreover, the CO 2 emission measured in April and October also showed differences between themselves.We attribute this large natural range of data as well as specific characteristics of each study period, for example, by the fact that it rains more in October than in April, even though these two months are in transition periods.
Thus, we believe that this natural variability of the phenomenon of gas emissions in the air-water interface contributes to find results that are discordant at first glance, like an apparent lack of seasonality of CH 4 emission in both reservoirs, even though they almost doubled over the months analyzed.Also relevant was the fact that the Serra da Mesa reservoir and the Três Marias Resevoir showed negative CO 2 emissions by 3 of the 4 campaigns in Serra da Mesa and all periods in Três Marias (Table 2).This fact is linked to the intense metabolism of CO 2 convert it to organic matter by photosynthetic organisms and thus they influence the chemical gradient of CO 2 in the air-water interface.

Conclusion
We concluded that the CH 4 fluxes were statistically indistinguishable in all analyzed hydrological periods, although the median have increased over the periods studied in both reservoirs.However during the month of April, which is a transition period in Serra da Mesa, fluxes were shown to be distinct from other periods studied, suggesting that there may be some component in this period that somehow influences the changes in CO 2 emissions standards.
Corroborating with this idea, the CO 2 fluxes measured in Serra da Mesa reservoir were distinct when comparing the periods of transition versus rainy and dry periods.We believe that perhaps this is due to seasonal influences changes in rainfall and temperatures.
Finally, the hydropower reservoirs are emitters or absorbers of carbon as CO 2 , which may in the long term balance the positive emissions beginning of the filling period.We believe that further measurements in greenhouse gas emissions are needed in order to better understand the variability of emissions.
In addition, other factors must be better analyzed as the input of different carbon fractions and their concentrations in the lake, the influence of meteorological factors, the human interventions such as land use basin which can exert influence and contribution with this allocthonus organic matter on greenhouse gases emissions.

Figure 1 .
Figure 1.Geographical locations of Serra da Mesa and Três Marias Reservoirs.

Figure 2 .
Figure 2. (A) and (B) refer to Serra da Mesa Reservoir while (C) and (D) Tres Marias Reservoir.In the horizontal axes are the months of the year.Solid lines represent monthly average rainfall and the lines segmented monthly average temperature.The blacks circles represent the medians of CH 4 emissions and the open circles the median of CO 2 emissions.

Figure 3 .
Figure 3. Box plot showing median CH 4 emissions from the sampling campaigns in the two reservoirs studied.

Figure 4 .
Figure 4. Box plot showing CO 2 emissions along the sampling campaigns in the two reservoirs studied.

Table 1 .
Sampling sites of reservoirs studied.
( ) The numbers in parentheses represent samples valid in each sampling campaign.

Table 3 .
Results of comparisons of CH 4 fluxes between the sampling campaigns conducted in Três Marias Reservoir.P-value 0.06.

Table 4 .
Results of comparisons of CH 4 fluxes among the sampling campaigns in the Serra da Mesa Reservoir.

Table 5 .
Comparison of CO 2 emissions among sampling campaigns from Serra da Mesa Reservoir.p-value < 0.05.

Wallis multiple comparison test to CO 2 emissions from Três Marias Reservoir Period Difference observed critical difference Result Aug
-Mar 18.08718 28.63825No difference Aug-May 34.28263 27.05063There are differences Aug-Nov 11.92904 27.19802No difference Mar-May 16.19545 27.85801No difference Mar-Nov 6.15814 28.00116No difference May-Nov 22.35359 26.37522No difference