Interactive comment on “ River inflow and retention time affecting spatial heterogeneity of chlorophyll and water – air CO 2 fluxes in a tropical hydropower reservoir ”

Overall the authors present an analysis of field collected data designed to emphasize the importance of spatial heterogeneity on the calculation of carbon fluxes from a tropical reservoir. The authors take an interesting approach by attempting to constrain the drivers of both Chl and CO2 as a function of physical (riverine influence) and chemical parameters. With that said, the authors admit the hydrologic modeling was limited in its ability to constrain a small dataset for this analysis.


Introduction
Over the last two decades, hydropower reservoirs have been identified as potentially important sources of greenhouse gas (GHG) emissions (St Louis et al., 2000;Rosa et al., 2004;Demarty et al., 2011).In tropical region, high temperatures and the flood-Figures

Back Close
Full ing of large amounts of biomass, including primary forest, result to intense GHG emission (Abril et al., 2005;Fearnside and Pueyo, 2012).However, emissions are larger in tropical Amazonian (Abril et al., 2013) than in tropical Cerrado reservoirs (Ometto et al., 2011) and in younger than in older reservoirs (Barros et al., 2011).Large hydroelectric reservoirs, especially those created by impounding rivers, are morphometrically complex and spatially heterogeneous (Roland et al., 2010;Teodoru et al., 2011;Zhao et al., 2013).Different regions in terms of CO 2 may be observed in these systems because it is dependent on flooded biomass, river input of organic matter, primary production and dam operation regime.Furthermore, both heterotrophic and autotrophic activity influences CO 2 concentration along reservoirs and the role of these activities has been reported in subtropical (Di Siervi et al., 1995), tropical (Roland et al., 2010;Kemenes et al., 2011) and temperate areas (Richardot et al., 2000;Lauster et al., 2006;Finlay et al., 2009;Halbedel and Koschorreck, 2013).
As the sedimentation and light availability increase along the reservoir, biomass of primary producers may increase.The phytoplankton is distributed in patches along the reservoir due to differences in habitat conditions linked to nutrient distribution, light availability and stratification (Serra et al., 2007).Also, hydrodynamics factors as retention time and river inflow showed to influence the phytoplankton communities and growth (Vidal et al., 2012;Soares et al., 2008).Intense phytoplankton primary production has been identified as the main regulator of carbon (C) budget in temperate eutrophic lakes (Finlay et al., 2010;Pacheco et al., 2014), however the impact on tropical hydropower reservoir is still unclear.
River inflows may affect biogeochemical patterns in river valley reservoirs (Kennedy, 1999).Difference on density of the incoming stream and lake water, stream and lake hydraulics, strength of stratification and mixing are features that control how the river water will flow when it reaches the reservoir (Fischer and Smith, 1983;Fischer et al., 1979).As result of density differences between river and lake water, the river enters in the lake and can flow large distances as a gravity-driven density current (Ford, 1990;Martin and McCutcheon, 1998).The interaction of large nutrient load injected by river Introduction

Conclusions References
Tables Figures

Back Close
Full and the dynamic of river inflow can determine the spatial heterogeneity in phytoplankton distribution (Vidal et al., 2012).Consequently, river inflow may affect primary production along river/dam axis in hydropower reservoirs strongly influenced by river.
In this study, we investigate the relationships between phytoplanktonic biomass and water-air CO 2 fluxes in an old and stratified tropical reservoir (Funil, state of RJ, Brazil), where intense phytoplankton blooms and low pCO 2 are observed in the water.We combine fieldwork and modeling to analyze the respective impact of meteorological and hydrological factors on the spatial and temporal dynamics of phytoplankton and the intensity of CO 2 fluxes.We show the effect of the river inflow in the heterogeneity of pCO 2 and Chl in Funil reservoir.We also compare temporal data of pCO 2 collected near the dam with a high dense spatial data.Our hypothesis were that (1) seasonal and spatial variability of pCO 2 and Chl in Funil Reservoir is more related to river inflow and retention time than external environmental factor such as air temperature and solar radiation and, (2) very different conclusion can be drawn regarding carbon cycle in reservoir if spatial heterogeneity is not adequately considered.

Study site
Funil Reservoir is an old impoundment constructed at the end of the 1960s and is located on Paraíba do Sul River, in a southern city (Resende) of the Rio de Janeiro State, Brazil (22 • 30 S, 44 • 45 W, Fig. 1).It is 440 m a.s.l., with wet-warm summers and dry-cold winters (Cwa, Köppen system).The main purpose of Funil Reservoir is energy production, but the reservoir is also used for irrigation and recreation.It to September 2012, the difference between minimum and maximum water level was 15.6 m.Funil Reservoir has a catchment area of 12 800 km 2 that has a high demographic density.The river inflow is not restrictedly linked to the rainfall in the watershed because it also depends on the water supply-demand and operation of dams constructed upstream.There are around 2 million people living inside the catchment area and 39 cities depending on the Paraíba do Sul River for water supply.These cities comprises 2 % of Brazil's gross domestic product (GDP) (IBGE, 2010).In this area, 46 % of sewage is untreated (AGEVAP, 2011) and the Paraiba do Sul River receives a large portion of the sewage from one of the main Brazilian industrial areas crossing part of the São Paulo State.Consequently, the river has a large influence on reservoir's water quality that has experienced tragic eutrophication in recent decades, resulting in frequent and intense cyanobacterial blooms (Klapper, 1998;Branco et al., 2002;Rocha et al., 2002).In general, Funil Reservoir is a turbid, eutrophic system, with high phytoplankton (cyanobacteria) biomass (Soares et al., 2012;Rangel et al., 2012).

Field sampling
Spatial data -the water samples for Chl and pCO 2 were taken between 09:00 LT and 12:00 LT on 1 March 2012 (end of rainy season, high water lever) and 20 September 2012 (end of dry season, low water level).Samples were taken at the surface (0.3 m) at 42 stations in Funil Reservoir (28 located along the main body of the reservoir, Fig. 1) in the same day to limit the effect of diurnal variation on the results.
We measured the Chl using a compact version of PHYTO-PAM (Heinz Walz GmbH, PHYTO-ED, Effelrich, Germany).Functionally the PHYTO-ED displays the same features as the standard but has distinct advantages for fieldwork.In the PHYTO-PAM Phytoplankton Analyzer µsec measuring light pulses are generated by an array of lightemitting diodes (LED) featuring 4 different colors: blue (470 nm), green (520 nm), light red (645 nm) and dark red (665 nm).The differently colored measuring light pulses are applied alternatingly at a high frequency, such that quasi-simultaneous information on Figures

Back Close
Full Chl fluorescence excited at the 4 different wavelengths is obtained.Following proper calibration, this feature allows to differentiate between the contributions of the main types of phytoplankton (green, blue, brown algae) with different pigment systems.In addition, Chl content of the various types can be estimated.
The pCO 2 data were determine using water-air equilibration method.In a marbletype equilibrator (Abril et al., 2014(Abril et al., , 2006)), water pumped directly from the lake flows from the top to the bottom (0.8 liters per min), while a constant volume of air (0.4 liters per min) flows from the bottom to the top.The large gas exchange surface area promoted by the contact with the marbles accelerates the pCO 2 water-air equilibrium.
The air pump conduct the air from the top of the equilibrator through a drying tube containing a desiccant (Drierite), then to an infrared gas analyzer (IRGA, LI-840, LICOR, Lincoln, Nebraska, USA), and then back to the bottom of the equilibrator (closed air circuit, Abril et al., 2006).For each station, the lake water and air were pumped through this system for two minutes before the pCO 2 from the IRGA stabilized to a constant value.
Color maps were created to represent spatial distribution of Chl and pCO 2 (Fig. 2).We used variogram analysis to describe spatial correlation among samples and to spatially interpolate using Kriging methods (Bailey and Gatrell, 1995).The empirical variograms were fitted to different mathematical models using Akaike's information criterion (AIC, Akaike, 1974) to evaluate the best fit.The best model variogram were used for interpolation by ordinary kriging.The root mean-square error (RMSE), calculated comparing observed and calculated values, was 90 µatm and 15 µg L −1 for pCO 2 and Chl, respectively.We used the software Spring (Câmara et al., 1996)  monitoring of hydrological systems (Alcantara et al., 2013;Stevenson et al., 1993).
The SIMA consists of an independent system formed by an anchored buoy containing data storage systems, sensors (air temperature, wind direction and intensity, pressure, incoming and reflected radiation and a thermistor chain), a solar panel, a battery and a transmission antenna.A sonde (YSI model 6600, Yellow Spring, Ohio, USA) was attached to the SIMA buoy to collect hourly surface data on temperature, conductivity, pH, and oxygen.This sonde was calibrated every 15 days according to the YSI Environmental Operations Manual (http://www.ysi.com/ysi/support).
We calculated the pCO 2 in the surface water over one year near the dam from measured pH and alkalinity.The calculations include dependence on temperature for dissociation constants of carbonic acid (Millero et al., 2002) and solubility of CO 2 .We used pH and temperature collected by SIMA between 25 October 2011 and 25 October 2012 and monthly data of alkalinity determined by the titration method (APHA, 2005) at station S28 (Fig. 1).Samples for total phosphorous (TP) and nitrogen (TN) were taken monthly.For TP, the samples was oxidized by persulfate and then analyzed as soluble reactive phosphorus.TN was determined as the sum of organic fraction measured by Kjedahl method and the dissolved inorganic nutrients.Laboratory analysis for TP and NP was performed according to standard spectrophotometric techniques (Wetzel and Likens, 2010).
Where k is the gas transfer velocity of CO 2 (in cm h −1 ), α is the solubility coefficient of CO 2 (in mol kg −1 atm −1 ) as a function of water temperature (Weiss, 1974), and ∆pCO 2 is the air-water gradient of pCO 2 (in µatm).The atmospheric pCO 2 measured in the rainy and dry season was 375 µatm and this atmospheric value was used for all flux Introduction

Conclusions References
Tables Figures

Back Close
Full calculation.The gas transfer velocity k was calculated from the gas transfer velocity normalized to a Schmidt number of 600 that corresponds to CO 2 at 20 • C (Eq. 2) (Jahne et al., 1987).Positive values of CO 2 fluxes denotes net gas flux from the lake to the atmosphere.
Where k 600 is the normalized gas transfer velocity calculated according to the following equation (Cole and Caraco, 1998), Sc is the Schmidt number of a given gas at a given temperature (Wanninkhof, 1992).
Where U 10 is wind speed at 10 m height.The wind speed was obtained from the SIMA da at 3 m height and was calculated for 10 m height (Smith, 1985).
In riverine zone, we considered the k 600 as a function of wind and water current.The contribution of the water current to the gas transfer velocity was estimated using the water current (cm s −1 ), depth (m) and the equations in Borges et al. (2004).

Temperature profile
Temperature profiles were collected using thermistor chain deployed at station S09 in rainy season and station S14 in dry season to determine the thermal structure at transition zone.Eleven thermistors (Hobo, U22 Water Temp Pro v2, Bourne, Massachusetts, USA) were placed every 0.5 m up to 4 m and every 1 m from 5 to 7 m.We also deployed a thermistor chain at riverine zone at station S05 with thermistors placed every 2 m.The thermistors were programed to record temperature every 10 min.In the rainy season, the thermistor chain was deployed on 29 February 2012 at 18:30 LT and recovered after 40 h.In the dry season, the thermistor chain was deployed on 20 September 2012 In our analysis, temperature is considered as the factor controlling water density.The use of temperature is justified by the low conductivity and turbidity in the river.The values of turbidity measured in the field of 29 and 11 NTU in rainy and dry seasons, respectively, would have effected density < 5 % relative to that of temperature (Gippel, 1989).

Numerical Model description and setup
Numerical simulations of lake hydrodynamics were conducted with the Estuary and Lake Computer Model (ELCOM, Hodges et al., 2000), which solves the 3-D hydrostatic, Boussinesq, Reynolds-averaged Navier-Stokes and scalar transport equations, separating mixing of scalars and momentum from advection.The hydrodynamic algorithms that are implemented in ELCOM use an Euler-Lagrange approach for the advection of momentum adapted from the work of Casulli and Cheng, 1992, whereas the advection of scalars (i.e., tracers, conductivity and temperature) is based on the UL-TIMATE QUICKEST method proposed by Leonard, 1991).The thermodynamics model considers the penetrative (i.e., shortwave radiation) and non-penetrative components (i.e., longwave radiation, sensible and latent heat fluxes) (Hodges et al., 2000).The vertical mixing model uses the transport equations of turbulent kinetic energy (TKE) to compute the energy available from wind stirring and shear production for the mixing process (Spigel and Imberger, 1980).A complete description of the formulae and numerical methods used in ELCOM was given by Hodges et al. (2000).
Simulations of hydrodynamics of Funil reservoir were conducted with realistic forcing condition (e.g.inflow, outflow, atmospheric temperature, radiation).These simulations were aimed in order to test the hypothesis regarding the river inflows at transition zone in rainy and dry seasons in Funil reservoir.Simulations started 4 days before the date of the considered data.This is necessary to let the model equilibrate beyond the initial physical conditions.The numerical domain was discretized in a uniform horizontal grid containing 100 m × 100 m cells based on a depth sample.The depth data were collected from 27 to 29 February 2012.The vertical grid resolution was set to a uniform Introduction

Conclusions References
Tables Figures

Back Close
Full 1 m thickness, resulting in 72 vertical layers.The water albedo was set to 0.03 (Slater, 1980), and the bottom drag coefficient was set to 0.001 (Wüest and Lorke, 2003).The attenuation coefficient for PAR was set to 0.6 m −1 based on Secchi disc measurements.
A value of 5.25 m 2 s −1 , based on a previous study conducted in another tropical reservoir (Pacheco et al., 2011), was chosen for the horizontal diffusivity for temperature and for the horizontal momentum.
Because of the presence of persistent unstable atmospheric conditions over tropical reservoirs (Verburg and Antenucci, 2010), the atmospheric stability sub-model was activated during the simulation; this procedure is appropriate in the cases in which the meteorological sensors are located within the internal atmospheric boundary layer (ABL) over the surface of the lake and data are collected at sub-daily intervals (Imberger and Patterson, 1990).In this manner, the heat and momentum transfer coefficients were adjusted at each model time step based on the stability of the ABL.The Coriolis sub-model was also activated during the simulation.
We defined two sets of boundary cells to force the inflow (Paraíba do Sul River) and outflow: (the water intake at the dam).The meteorological driving forces over the free surface of the reservoir were considered uniform.The model was forced using hourly meteorological data acquired by SIMA, the daily inflow and outflow provided by Eletrobrás-Furnas and the cloud cover and river temperatures extracted from thermistor chain data and complemented with MODIS data.Two periods were simulated: one 3 Results

Spatial variability
Based on spatial data of Chl and pCO 2 , a typical zonation pattern usually found in reservoirs was observed in Funil main body (riverine, transition and lacustrine zones) Introduction

Conclusions References
Tables Figures

Back Close
Full (Fig. 2).Although the boundaries are influenced by many factors and are not easily determined, these regions have distinct physical, chemical and biological features.The riverine zone (RZ) has a high input of nutrients coming from terrestrial systems and human activities, but the primary production is limited by high turbidity and turbulence.As the sedimentation and light availability increase along the reservoir, biomass of primary producers increases in the transition zone (TZ).The lacustrine zone (LZ) is characterized by nutrient limitation and reduced phytoplankton biomass (Thornton, 1990).In this study, we considered the Chl to separate the reservoir in three zones.Riverine zone is characterized by low Chl (< 5 µg L −1 ).Transition zone begins where the Chl starts to increase and ends when Chl decrease to levels closely to Chl in Lacustrine zone.
Finally, Lacustrine zone is characterized by intermediate Chl.
Funil reservoir showed to be spatially heterogeneous with seasonal differences in Chl and pCO 2 (Fig. 2).The spatial data showed high spatial variation only in the main body of the reservoir, while the southern part was undersaturated in CO 2 in rainy and dry seasons (Fig. 2a and b).Spatially average of pCO 2 for rainy and dry season were 259 ± 221 and 881 ± 900 µatm, respectively.The pCO 2 varied from 140 to 1376 µatm in rainy season and from 43 to 2290 µatm in dry season.Higher values of pCO 2 in the riverine zone of the reservoir and a drastically decrease in the transition zone were observed in both sample periods (Fig. 3a and b).In the lacustrine zone, undersaturation on CO 2 was prevalent at all sample sites in rainy and dry season.Considering all sample sites, there was significant differences between rainy and dry seasons (t = 1.99, p < 0.05) and higher values of pCO 2 during dry season in Funil reservoir were previously reported (Roland et al., 2010).Chl was higher in the transition zone compared with others compartments (t = 2.01, p < 0.05, Fig. 3a and b).Further, average concentration in transition zone was 2.5 times higher than the reservoir average (129.2 and 52.0 µg L −1 , respectively).Unlike pCO 2 , Chl data showed no significant difference between rainy and dry season considering all spatial data (t = 1.99, p > 0.05).
The dry seasons, the maximum emission was observed in riverine zone and the minimum in the transition zone.The spatial average was −7.2 and 22.1 mmol m −2 d −1 in the rainy and dry season, respectively (Table 1).

Temporal variability
The pCO 2 calculated by multi-parameter sonde data (temperature and pH) and alkalinity showed a large seasonal variability over the year at the station near the dam (Table 2).The pCO 2 varied from 35 to 4058 µatm with average of 624 ± 829 µatm and median of 165 µatm.The pCO 2 supersaturation was prevalent between April and June, while pCO 2 undersaturation was prevalent in all other periods (Fig. 4a).Lowest median of pCO 2 was observed between October and December (43 µatm).Considering all temporal data over the year, 59.8 % of the data were below atmospheric equilibrium and 1.1 % were within 5 % of atmospheric equilibrium.
In Funil Reservoir, the seasonal pCO 2 variation over the year at the station near the dam agreed with variation of retention time (Fig. 4).The yearly average of the reservoir retention time was 32.6 days over the considered year.Lower retention time occurs between October and December when the water level is low and the reservoir is ready to stock water coming from the watershed and rain during the rainy season (October to March).
The CO 2 flux over the year at the station near the dam varied from −119.8 to 234.5 mmol m −2 d −1 .The average of flux was −0.9 ± 33.1 mmol m −2 d −1 and median of −6.2 mmol m −2 d −1 .We observed substantial uptake of CO 2 between October and December (rainy-spring) (Table 1).Uptake of CO 2 from the atmosphere was also prevalent between July and September (dry-winter).From January to July, the lake lost substantial CO 2 via degassing (Table 1).Summary of all other data collected over the studied period is shown in Table 2. Introduction

Conclusions References
Tables Figures

Back Close
Full

Thermal structure of transition zone and river
During the rainy season, thermal stratification occurred in the transition zone only during the daytime around 16:30 LT, when a maximum of 33.1 • C was observed at the surface for a minimum of 27.8 • C at the bottom (Fig. 5a); to the contrary, temperature was vertically homogeneous at nighttime.The daily range of temperature oscillation during rainy season at surface was up to 5 • C. In the dry season, water temperature were lower compared to the rainy season at transition zone.Stratification occurred around 14:00 LT in dry season, when we observed a maximum of 25.7 • C and a minimum of 23.1 • C at the bottom.The daily range of temperature oscillation was up to 3 • C at surface and stratus layers with different temperatures were observed every 2.5 m (Fig. 5b).
The river temperature varied from 27.7 to 28.7 • C and 23.6 to 24.1 • C in rainy and dry season, respectively (Table 3).The average temperature difference between river and reservoir surface water was 2.1 and 0.3 • C in rainy and dry season, respectively.

Simulations
We first compared the simulated and real temperature at station S09 and S14 for rainy and dry season, respectively.The RMSE, calculated by comparing data every 20 min were 1.4 • C for rainy season and 1.1 • C for dry season.These results obtained for both seasons are comparable with previous modelling exercises found in literature (Jin et al., 2000;Vidal et al., 2012).We also analyzed the ability of the model to reproduce the inflow using data from drifters released in the river and transition zone of the reservoir on 1 March and 20 September (data not shown).Although the vertical thermal structures observed in dry season (Fig. 5b) were not well represented, the model reproduced the behavior of the inflow as underflow in rainy season (Fig. 6a) and interflow and overflow in dry season (Fig. 6b) as anticipated by the schematic representation (Fig. 5c and d).
The river flowed mainly at 6 m depth near the bottom of Funil reservoir after the river plunging point in the rainy season.In the dry season, the river flowed mainly at 3 m depth at night and 4 m at daytime.Introduction

Conclusions References
Tables Figures

Back Close
Full The daily oscillation of the neutral buoyance observed occurs because of the variation of reservoir surface and river temperatures (Vidal et al., 2012;Curtarelli et al., 2013).The level of neutral buoyancy, where the densities of the flowing current and the ambient fluid are equal, represents the depth where the river water spreads laterally in the reservoir.In rainy season, the river flowed as underflow (Fig. 6a), however, when the river reached its maximum temperature around 21:00 LT (Table 3) the temperature difference between river and surface water decreased, the level of river neutral buoyance moved upward and the maximum flow was observed between 4 and 6 m (Fig. 6a).In the dry season, the river flowed as overflow, but it plunged down to 4 to 6 m depth when the high surface temperature during the day coincided with the period of lowest river temperature (Table 3) and neutral buoyance moved downward (Fig. 6b).The change in patterns observed in the river flow between 20 and 21 September occurred due to a decrease of the river temperature during a rainfall that occurred around 16:00 LT on 20 September 2012 (Fig. 6b).

pCO 2 driven by Phytoplankton
Primary production associated to the high Chl showed to be the main regulator of CO 2 concentration at the surface of the Funil Reservoir (Fig. 7).Spatially, pCO 2 were negatively correlated to the Chl (r 2 = 0.71).In old hydropower reservoir where C source from the flooded soil after impounding has become negligible, primary production may become a significant term in the C budget.Intense primary production fuelled by high levels of nutrients reduces CO 2 concentrations to levels below atmospheric equilibrium in transition zone and lacustrine zone (Fig. 3).To the contrary, high values of pCO 2 in riverine zone may be associated with suspended solids and turbulence and vertical mixing that inhibit primary production.Figures

Back Close
Full Low pCO 2 levels observed at the station near the dam over the year is associated with (1) high primary production due to higher temperature and solar radiation that promote water column stability and stratification, and (2) constant high nutrient availability.Since nutrient availability in Funil Reservoir is high during the entire year (Table 2), nutrients are never limited in the lacustrine zone and other factors that controls stability and stratification, such as temperature, wind and changes in mixing depth related to the seasonal variation are the main inhibitors of algal growth near the dam especially between April and June.
Due to phytoplankton productivity, we found net uptake of CO 2 over the year at the station near the dam, especially between October and December (Table 1).The fate of carbon fixed by the phytoplankton in Funil Reservoir is still unclear.The higher flux of methane (CH 4 ) from sediment to water found in Funil Reservoir compared to other tropical reservoir (Ometto et al., 2013) suggests that a substantial fraction of the fixed carbon reaches the sediment and is further mineralized in CH 4 .However, in lacustrine zone, the higher depth and high temperature may promote mineralization of part of carbon fixed by phytoplankton in the water column before it reaches the sediment.Outflow export the same amount of carbon that comes from the watershed (Ometto et al., 2013), so the water that flows through the turbine is not responsible to export the excess of carbon invaded from the atmosphere.Although there is no data to support this statement, as reported in natural eutrophic lakes (Downing et al., 2008), burial of organic carbon composed by phytoplankton and methanogenesis seems to be two important carbon pathways for the carbon fixed by the phytoplankton in Funil Reservoir.is full to ensure enough water to produce energy during entire dry season.Therefore, many ecological processes in natural lakes that are related with external factor (e.g.solar radiation, precipitation and temperature) may be regulated by dam operation in hydropower reservoir.The position of transition zone of the reservoir moves as result of the seasonally (Fig. 3).In the end of the rainy season, the retention time and water level was high, and the influence of the river in the surface water of the reservoir was restrict to a small area (Fig. 2a and c).Differently, when the water level and retention time was low, the transition zone moved toward the dam and the river inflow influenced the surface Chl and pCO 2 in more than 40 % of the total reservoir surface area (Fig. 2b and d).A reservoir can become a fluvial-dominated system when retention time is short or lacustrine when retention time is long (Straškraba, 1990).Size of the river-influenced area in the reservoir surface water also depends on water density.Differences on river and reservoir temperature, total dissolved solids, and suspended solids can cause density gradient in the water column.Depending on the water density differences between the inflow and reservoir, the river can flow into the downstream area as overflow, underflow, or interflow (Martin and McCutcheon, 1998).During the rainy season in Funil reservoir, due to the high difference between river and reservoir surface temperature (∼ 4

Physical feature and spatial distribution
• C), the river water progressively sinks down (underflow), and contributes to the thermal stability of the water column (Fig. 5a, Assireu et al., 2011).The denser river water flows under a lighter reservoir water and waves and billows develops along the interface due to shear velocity.This behavior is indicative of the Kelvin-Helmholtz instability, in which waves made up of fluid from the current (river) promote mixing with the reservoir water (Thorpe and Jiang, 1998;Corcos and Sherman, 2005) (Fig. 5c).This mixing and the high nutrient concentration coming from Paraíba do Sul River (Table 2) may explain the high levels of Chlorophyll a observed in the transition zone (Fig. 3).Many cold fronts pass through Brazilian middle-west and southeast in dry seasons.(Lorenzzetti et al., 2005;Alcântara et al., 2010).Thus, the decrease of reservoir surface Introduction

Conclusions References
Tables Figures

Back Close
Full temperature (Table 2) and consequent decrease in density difference between river and reservoir surface leads to river inflow characterized by inter-overflow (Fig. 5b and d).
In an inter-overflow, the riverine characteristic of high turbulence, pCO 2 and low Chl is observed in the reservoir surface 5 km toward the dam (Fig. 3a and b).Although there is high nutrient concentrations in the transition zone (Table 1) between S19 and the river, the surface water is dominated by river flow with low Chl concentrations (Fig. 3).Phytoplankton will not bloom until they get a certain distance down-reservoir and the inflow mixes with the reservoir and loses velocity (Vidal et al., 2012).
The ELCOM simulations using real data represented the river inflow during the rainy and dry season (Fig. 6).The results converge to the conclusion of a lower influence of the river on surface waters during the rainy season since a higher dense (colder and more turbid) river plunges when reach a warmer reservoir water.Although the model does not represented the intrusions of river water on different depths (every 2.5 m) suggested by temperature profile at transition zone in dry season (Fig. 5b), the model represented the overflow, especially at night (Fig. 6).The daily scale variation of the river inflow (Fig. 6) occur as response of the lagged change of temperature of the river and reservoir over the day.In rainy season, this oscillation facilitate the injection of nutrient in euphotic zone when the reservoir surface temperature decreases and river temperature reaches its maximum in the end of the day (Table 3).During the day, when the river temperature drops, the large peak of Chl in transition zone (Fig. 3a) could be result of diurnal stratification developing (Fig. 5).In dry season, the peak of Chl occurs five kilometers further downstream (Fig. 3b), since inflow never plunges due to lower temperature differences between river and reservoir surface.

Spatial and temporal heterogeneity
As a result of phytoplankton growth associated with these physical features, there are large spatial and temporal variation of CO 2 fluxes in the Funil Reservoir.Several studies of hydropower reservoir have suggested that significant CO 2 evade from these systems to the atmosphere at a global scale (St Louis et al., 2000;Roehm and Tremblay, 2006 Barros et al., 2011;Fearnside and Pueyo, 2012).However, recent studies have shown that the growing nutrient enrichment caused by human activities (eutrophication) can reverse this pattern in some hydropower reservoir (Roland et al., 2010) and natural lakes (Pacheco et al., 2014).Our study shows that Funil reservoir is spatially heterogeneous with high CO 2 emission in riverine zone and high CO 2 uptake in transition and lacustrine zone.Temporally, the reservoir near the dam is undersaturated in pCO 2 mainly between October and December, and supersaturated in pCO 2 between April and June (Table 1).
We might have different or opposite conclusions if spatial and temporal pCO 2 data are analyzed separately.Previous studies suggested that in natural small lakes, a single sample site should be adequate to determine if a lake is above or below equilibrium with the atmosphere and the intensity of the fluxes (Kelly et al., 2001).However, large spatial heterogeneity, regarding pCO 2 and CO 2 emission to atmosphere, was observed in boreal (Teodoru et al., 2011) and tropical (Roland et al., 2010) reservoir.Our temporal data at the dam station showed lower pCO 2 over October, November and December when the retention time is extremely low (Table 4), but this observation does not represent the entire reservoir.The spatial data collected at low water level showed low pCO 2 in the dam as well, however almost half reservoir is supersaturated due to the river influence (Fig. 2d).The average pCO 2 during low retention time was 881 µatm considering whole reservoir area, contrasting with only 69 µatm near the dam.Furthermore, if we considered only one station near the dam to estimate CO 2 flux between lake surface and atmosphere, the conclusion would be contradictory.For example, in periods of low retention time, calculated CO 2 flux shows that CO 2 flux would be −17.6 mmol m −2 d −1 (CO 2 sink) considering one spot temporal data, and 22.1 mmol m −2 d −1 (CO 2 source) considering whole reservoir (Table 4).
Same contradictory conclusion can be found when studies with low number of sample sites are considered in the spatial heterogeneity discussion.Previous studies about heterogeneity considering four stations in Funil reservoir showed that peak of phytoplankton biomass in transition zone was not observed (Soares et al., 2012).However, Introduction

Conclusions References
Tables Figures

Back Close
Full our Chl data collected every 1000 m, proximally, showed a clear transition zone with peaks of Chl.Furthermore, data analysis from only four stations showed that high spatial heterogeneity occurs in periods of high retention time (high water level).To the contrary, our data reveal high spatial heterogeneity in low retention time, corresponding to periods with high influence of the river in the surface water.Different conclusions found by Soares et al. (2012) may be explained by the variation of transition zone location.This zone is limited to a small area and its location varies depending on retention time and inflow (Fig. 2c and d).Therefore, low number of sample stations in Funil reservoir showed to limit the capacity to represent whole reservoir and to discuss about spatial heterogeneity.Since high number of stations and variables increases fieldwork time and costs, a balance between them must be considered to ensure acceptable representativeness.

Conclusion
In summary, the seasonal and spatial variability of Chl and CO 2 fluxes in Funil reservoir is mainly related to river inflow and retention time.However, the relationship between pCO 2 and Chl suggests that primary production regulates surface CO 2 fluxes in transition and lacustrine zone.Average of spatial data showed CO 2 evasion to the atmosphere in periods of low retention time (even with higher Chl) due to river influence on water surface, and CO 2 uptake in periods of high retention time when the river plunges and flows under the reservoir.However, the threshold of retention time that seal the transition between source and sink of CO 2 could not be determined.Comparison between spatial (42 stations) and temporal data (one station) showed that different conclusions can be drawn if spatial heterogeneity is not adequately considered.Moreover, change of transition zone location over the year must be considered when low number of stations is used to represent spatial heterogeneity.Lack of spatial information of CO 2 flux could lead to erroneous conclusion of the importance of hydropower reservoir to freshwater carbon cycle.Funil reservoir is a dynamic system where the Figures

Back Close
Full hydrodynamics linked to river inflow and retention time controls both pCO 2 and Chl spatial variability and seems to be the key that regulate most of ecological process.Introduction

Conclusions References
Tables Figures

Back Close
Full  Full  Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | has a surface area of 40 km 2 , mean and maximum depth of 22 and 74 m, respectively, and total volume of 890×10 6 m 3 .The maximum and minimum reservoir water level occurs in the end of rainy season (April) and dry season (October), respectively.From October 2011 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | version 5.1.8to conduct the spatial analysis and to produce the in situ pCO 2 and Chl maps.Time series data -wind speed and direction, solar radiation, pH, dissolved oxygen (DO), air temperature and temperature profiles (2 m, 5 m, 20 m and 40 m depth) were collected hourly at station S28 near the dam (Fig. 1) and transmitted by satellite in quasi-real time by the Integrated System for Environmental Monitoring (SIMA).The SIMA is a set of hardware and software developed for data acquisition and real-time Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | at 11:30 LT and recovered after 25 h.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | to represent rainy season (25 February 2012 to 4 March 2012) and one to represent dry season (15 to 23 September 2012).
Discussion Paper | Discussion Paper | Discussion Paper | calculated CO 2 fluxes from spatial data varied from −35.8 to 44.2 mmol m −2 d −1 and −47.7 to 87.3 mmol m −2 d −1 in rainy and dry season, respectively.In both rainy and Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | The retention time of Funil reservoir is strongly driven by the operation of the dam.The volume of water that flows through the turbine depends on the energy demands and inflow from Paraiba do Sul River.Periods of low retention time and water levels do not necessarily correspond to periods of low precipitation.In fact, highest retention time and water level is often observer in the middle of the dry season when the reservoir 8546 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Figure 1 .
Figure 1.Map of Funil reservoir showing geographic location and sampling stations.

Figure 3 .
Figure 3. Lotic-lentic gradient of pCO 2 and Chl along the 28 sampling station in the main reservoir body in rainy season (a) and dry season (b).The water level was 461.0 and 451.5 in rainy season and dry season.Three zones can clearly be defined (riverine, transition and lacustrine zone).The arrow shows that the transition zone starts 4.8 km down-reservoir in the period of low water level.

Figure 4 .Figure 5 .4
Figure 4. Box plot of pCO 2 at station S28 near the dam (a) and mean reservoir retention time (b) over the studied year.The dashed line represents the average of pCO 2 in the atmosphere (375 µatm).The data are subdivided in four seasons: rainy-spring (October-December), rainysummer (January-March), dry autumn (April-June) and dry winter (July-September).

Figure 5 .Figure 6 .Figure 6 .
Figure 5. Temperature profile collected at station S09 in rainy season (a) and at station S14 in dry season (b).Dashed line represent the depths where river flows as overflow or interflows.In rainy season the river plunges and flows under the reservoir (underflow) due to difference of density (c).Waves and billows develops along the interface due to shear velocity (Kelvin-Helmholtz instability) and facilitate vertical mixing (see text).In dry season the river flows as overflow or interflow (d) since the difference of density between river and reservoir is low.At this situation, the river can influence the reservoir surface water more 5 kilometers toward the dam.RZ = Riverine Zone, TZ = Transition Zone, LZ = Lacustrine Zone.

Figure 7 .
Figure 7. Relationship between spatial data of pCO 2 and Chl in Funil Reservoir.The regression is represented by dashed line (r 2 = 0.71, p < 0.001).

Table 1 .
Average CO 2 fluxes (mmol m −2 d −1 ) calculated using spatial and temporal data.Positive fluxes denotes net gas fluxes from the lake to the atmosphere.

Table 2 .
Average and standard deviation of environmental and chemical variable from the station S28 (near the dam) and river.a Cumulative precipitation over three months.Dev.Average Std.Dev.Average Std.Dev.Average Std.Dev.