Natural events of anoxia and low respiration index in oligotrophic lakes of the Atlantic Tropical Forest

Introduction Conclusions References


Introduction
Lakes are small but broadly distributed at low altitudes (Downing et al., 2006), representing a common fate for organic and inorganic inputs from large areas in the watershed (Tranvik et al., 2009).These ecosystems show intense metabolic activity supported by the availability of water, nutrients and organic matter in both pelagic (Biddanda et al., 2001) and benthic (Downing et al., 2008) compartments.Globally, important pool of carbon (C) fixed in organic compounds by terrestrial plants may be buried (von Wachenfeldt and Tranvik, 2008) or mineralized to C gases (Cole et al., 2007) within lakes, a crucial component of the C cycle.
Oxygenic photosynthesis and aerobic respiration are the major metabolic pathways by which organic matter is produced and destroyed in the biosphere (Cole et al., 2000), corresponding to the overall metabolic balance of an ecosystem (Howarth et al., 1996).Carbon dioxide (CO 2 ) and oxygen (O 2 ) are metabolic gases involved in both processes, as oxygenic photosynthesis absorbs CO 2 producing O 2 , while aerobic respiration demands O 2 releasing CO 2 (Clarke and Fraser, 2004).In this way, lakes may show net autotrophy uptaking atmospheric CO 2 , or net heterotrophy with subsequent CO 2 evasion to atmosphere.However, most lakes are heterotrophic due to terrestrial organic inputs subsiding the aquatic decomposition (Duarte and Prairie, 2005;Cole et al., 1994) and food web (Pace et al., 2004).
Respiration is the most efficient biological process of organic degradation, but is strongly limited by the O 2 supply (Sobek et al., 2009).The O 2 depletion following high respiration of the excessive organic loading is a typical cause of organism death and severe decline in the species (Vaquer-Sunyer and Duarte, 2008), which also stimulates the anaerobic organic decomposition in natural waters (Conrad et al., 2011).These anaerobic processes have important implications to global warming, producing more powerful greenhouse gases than CO 2 (Bastviken et al., 2011), as well to create "dead zones" by releasing toxic substances for major aquatic organisms (Diaz and Rosenberg, 2008).Besides aerobic conditions (Diaz and Rosenberg, 2008;Vaquer-Sunyer and Duarte, 2008), the high ratio between partial pressures of O 2 and CO 2 (pO 2 : pCO 2 ), named respiration index (RI), is also crucial to support biological diversity, as it provides a simple numerical constraint related to available energy in natural waters (Brewer and Peltzer, 2009).
Along the latitudinal gradient, warmer annual temperatures in ecosystems may contribute to higher diversity of organisms (Amarasinghe and Welcomme, 2002) and more intense metabolic processes (Brown et al., 2004;Davidson and Janssens, 2006), including those involved in the organic mineralization with subsequent production of greenhouse gases (Marotta et al., 2009a;Bastviken et al., 2010).The magnitude of metabolic responses following common changes in resource availability or conditions may be substantially enhanced under higher temperatures, resulting in a high variability either among (Marotta et al., 2009a) or within tropical lakes (Marotta et al., 2010a) and over time in these ecosystems (Marotta et al., 2010a, b).
The Atlantic Tropical Forest is a very productive and threatened biome in Brazil (Metzker et al., 2011).Lakes surrounded by this forest show a persistent CO 2 evasion to the atmosphere attributable to terrestrial C inputs (Marotta et al., 2009b), despite large changes related to seasonal events of water stratification and mixing, especially during the summer and winter, respectively (Tundisi, 1997).The aim of the present study was to assess pO 2 , pCO 2 and RI fluctuations following seasonal water column stratification and mixing periods over 19 months in two oligotrophic lakes of the Atlantic Tropical Forest.We tested the hypothesis that thermal stratification events could be coupled to natural hypoxia in deep waters of both lakes.

Barra
(19  1).This protected area includes one of most important conserved remnant of the Atlantic Tropical Forest in Brazil (36 000 ha). Lake Barra and Lake Aguapé are shallow (maximal depth of 10 m) and small (areas of 1350 and 1372 km 2 , respectively), showing organic and oligotrophic waters (total phosphorus around 1 µmol l −1 , chlorophyll a about 15 µg l −1 , and dissolved organic carbon above 5 mg l −1 during this study).Despite any human interference in the margins (abandoned eucalyptus plantation in regeneration to native forest and few field houses), both lakes receive natural inputs from the watershed dominated by the Atlantic Tropical Forest with low human use and preserved natural conditions.Terrestrial inputs from the protected tropical forest commonly affect aquatic organisms and metabolic processes in lakes of this region (Petrucio and Barbosa, 2004;Petrucio et al., 2006).
The climate of the study area is tropical wet and dry (Köppen climate classification Aw; Peel et al., 2007), characterized by a strong seasonality in rainfall (Metzker et al., 2011).This includes dry winters from June to September and rainy summers from December to March showing 25yr monthly mean precipitation (± SE) around, respectively, 10 (± 2) and 198 (± 13) mm (data of the National Institute of Meteorology for 1987-2011).Lakes of this region show a well-described seasonal stratification during the rainy season, caused by less water circulation, higher air temperatures and inputs of slightly colder and denser groundwaters, contrasting with a typical mixing during the dry winter by lower air temperatures and more windy conditions (Tundisi, 1997).

Study design
Water samples for O 2 , pH, alkalinity, temperature, nutrients, chlorophyll a and dissolved organic carbon (DOC) were collected in the morning, using a 3-l Van Dorn bottle, at approximately monthly intervals from March 2004 to October 2005 (19 months).Additionally, one daily variation in O 2 , pH, alkalinity and temperature without any nocturnal data (time of sampling at 24:00, 18:00, 06:00 and 24:00 LT the day after) was simultaneously assessed in three periods: Four sampling depths at the central station in both lakes were chosen assuming the light penetration by a 20-cm diameter Secchi disk: 100 % (surface waters), 10 % (the Secchi depth), 1 % (three times the Secchi depth), and 0 % (aphotic zone below the 1 % light penetration depth and above the bottom sediment).

Analytical methods
Dissolved O 2 concentrations by the Winkler method, pH using a pH meter Marconi PA-200 (precision of 0.01 unities of pH), and the total alkalinity by the Gran titration were immediately analyzed after the sampling (APHA, 1992).At the laboratory, pre-filtered (0.7 µm, Whatman GF/F) water samples were analyzed for chlorophyll a concentrations by extraction with 90 % acetone (Lorenzen, 1967), and for DOC concentrations by high-temperature catalytic oxidation using a TOC-5000A Shimadzu Analyzer (samples pre-acidified to pH < 2.0).CO 2 concentrations were estimated from measurements of pH and alkalinity (Stumm and Morgan, 1996) with corrections for temperature, altitude, and ionic strength (Cole et al., 1994).pCO 2 and pO 2 were calculated from the Henry's law with appropriated adjustments for temperature and salinity for CO 2 (Weiss, 1974) and O 2 (Garcia and Gordon, 1992) solubility.The respiration index was calculated as the ratio pO 2 : pCO 2 in log 10 following Brewer and Peltzer (2009).Negative values of RI (RI < 0) indicate a ratio pO 2 : pCO 2 < 1.0.

Statistical analyses
Log-transformed data of pO 2 and pCO 2 or raw data of RI from the same sampling depth or period for each lake showed significant Gaussian distribution (Kolmogorov-Smirnov, p < 0.05), homogeneity of variances (Bartlett, p > 0.05) and significant matching (F test, p < 0.05).Hence, these variables in different sampling depths and periods were compared using parametric tests (Zar, 1996), repeated measures of one-way ANOVA followed by Tukey-Kramer multiple comparisons (significance level set at p < 0.05).In contrast, non-parametric statistics were used to test for differences in pO 2 , pCO 2 and RI between the lakes, repeated measures of Friedman test followed by Dunn's post-test (significance level set at p < 0.05), as the transformed data set including all measurements of each variable from the same lake did not meet parametric assumptions.Consequently, pO 2 , pCO 2 and RI were correlated with chlorophyll a and DOC concentrations using Spearman correlations (significance level set at p < 0.05).All statistics were conducted using the software Statistica 7.0.2b and c).Lake Barra and Lake Aguapé showed similar fluctuations in pO 2 , pCO 2 and RI during 19 months (Fig. 3), and non-significant differences for these variables comparing monthly (from March 2004 to October 2005) or 24-h cycle (in March, June and September 2005) measurements (Paired t-test, p<0.05).In addition, non-significant difference was observed for gas fluxes between stratified and unstratified periods in both lakes (Paired t-test, p < 0.05).From the end of the rainy seasons to the beginning of the dry winters in 2004 and 2005, pO 2 and RI decreased in surface and increased in deep waters following the thermal mixing, while pCO 2 showed the opposite trend, increasing in surface and decreasing in deep waters (Fig. 3).Additionally, a reversal increase in surface and decline in deep pO 2 and RI followed the thermal stratification in both lakes during the end of the dry seasons and during the rainy summer in 2004 and 2005, also contrasting with the opposite trend observed for pCO 2 (Fig. 3).

Our
In this way, the stronger thermal stratification over the 24h cycle in the rainy summer (March 2005) was coupled to more intense differences for metabolic variables in the vertical profile, which showed, on average, pO 2 values about 50-fold higher, pCO 2 about six-fold lower, and RI around two-fold higher in surface (at 100 or 10 % light penetration) than in deep (at 1 % or 0 % light penetration) waters of both lakes (Fig. 4; Tukey-Kramer, p < 0.05).RI values were ≤ 0 in deep waters of both lakes during this strong thermal stratification (Fig. 4).On the other hand, a persistent thermal mixing over the 24-h cycle at the beginning of the dry winter (June 2005) was related to non-significant differences for pO 2 , pCO 2 and RI at both surface and deep depths (Fig. 5; repeated-measures one-way ANOVA, p > 0.05).Lastly, the initial thermal stratification over the daily variation at the end of the dry season (September 2005) significantly showed, on average, pO 2 about eight-fold higher, pCO 2 about 2.5-fold lower, and RI around two-fold higher comparing surface (100 or 10 % light penetration) and deep (1 % or 0 % light penetration) waters of both lakes (Fig. 6; Tukey-Kramer, p < 0.05), with RI values once again ≤ 0 in the deep waters (Figs. 4 and  6).
The negative relationship between pO 2 and pCO 2 was significant but weak for waters from Lake Barra and Lake Aguapé (R Spearman = −0.37 and −0.38 respectively; Spearman correlation, p < 0.05).Additionally, pO 2 , pCO 2 and RI in waters from all depths showed non-significant relationships with chlorophyll a and DOC in both lakes (Spearman correlation, p > 0.05), except weak significant correlations of pO 2 with chlorophyll a and DOC (R Spearman = 0.33 and −0.23, respectively) or RI with chlorophyll a (R Spearman = 0.17) in Lake Aguapé (Spearman correlations, p < 0.05).

Discussion
Overall, Lake Barra and Lake Aguapé showed a consistent prevalence of pCO 2 above and pO 2 below the equilibrium with the atmosphere, resulting in low RI to aquatic organisms.The persistence of CO 2 emissions to the atmosphere during the study reached, on average (± SE), 27.6 (± 3.6) mg C m −2 h −1 (assuming air-water fluxes calculated as Cole and Caraco, 1998, a pCO 2 in equilibrium with the atmosphere of 380 µatm and a mean 10-m wind speed over land of 3.28 m s −1 ; Archer and Jacobson, 2005).Using the conservative wind velocity applied in other studies of 0.5 m s −1 (Cole et al 1994), the average CO 2 efflux calculated for lakes Barra and Aguapé was 17.4 mg C m −2 h −1 , which is comparable to the average reported for 367 tropical lakes (20.1 mg C m −2 h −1 ) (Marotta et al., 2009).This confirms that the typical role of terrestrial C inputs subsiding the biological degradation and CO 2 supersaturation in organicenriched waters (Marotta et al., 2012) or in most lakes around the world (Cole et al., 1994(Cole et al., , 2007;;Duarte and Prairie, 2005) might be also found in inland waters of the Atlantic Tropical Forest.
Indeed, the negative relationship between pO 2 and pCO 2 also supported the potential role of the balance between aquatic respiration and photosynthesis in regulating the production and consumption of metabolic gases in waters of both lakes studied here.This was confirmed in the lake with higher chlorophyll a and DOC concentrations (Lake Aguapé), as expected significant relationships were found between chlorophyll a (algal biomass) and increases in O 2 and RI, probably from oxygenic photosynthesis (Carignan et al., 2000).However, all these significant correlations were weak, coupled to other non-significant relationships between chlorophyll a or DOC with pO 2 , pCO 2 or RI in Lake Barra and Lake Aguapé.These weakly significant and non-significant correlations suggest that dynamics other than the balance between aquatic photosynthesis and respiration might drive high fluctuations in metabolic gases, strongly reducing the negative relationship between metabolic gases, pO 2 and pCO 2 , in both lakes.The C inputs from the watershed (Marotta et al., 2010b), and anaerobic (Conrad et al., 2011) or physical-chemical (Amado et al., 2007) organic degradation processes may enhance CO 2 without consuming O 2 in natural waters.In addition, anoxygenic photosynthesis may be responsible for the decoupling between CO 2 fixation and O 2 production (Fontes et al., 2011).After the aerobic organic degradation reducing O 2 supply (Vaquer-Sunyer and Duarte, 2008;Sobek et al., 2009), intense anaerobic pathways subsidized by high allochthonous organic inputs may decrease RI to negative values (RI ≤ 0) (Brewer and Peltzer, 2009) or release toxic compounds (Diaz and Rosenberg, 2008) to critical levels to major organisms in aquatic ecosystems.
The prevalence of high pCO 2 and low pO 2 also revealed highly dynamic fluctuations in metabolic gases and RI in waters of both lakes during 19 months.Substantial changes in pCO 2 , pO 2 and RI were closely related to seasonal patterns of water stratification and mixing.Natural shifts from stratified and anoxic to oxic and mixed conditions were observed throughout the year in deep waters of both lakes.On the other hand, surface waters showed a contrasting decline in O 2 and RI following the mixing with deep waters during a mixing period at the beginning of the dry winter.Higher temperatures in the summer stimulating biological activity (Brown et al., 2004;Clarke and Fraser, 2004) might explain more intense increases reported here for O 2 and RI in the surface photic zone, probably by primary producers, and stronger decreases for both (O 2 and RI) in deep waters, probably by light limitation to oxygenic photosynthesis (Gu et al., 2011;Fontes et al., 2011).These results confirm the high temporal variability of metabolic gases described in previous studies on tropical lake waters, which related typical warmer temperatures at low latitudes to large shifts in biological processes, following common changes in meteorological and physicalchemical conditions over time (Marotta et al., 2010b, a).
In conclusion, we confirm the hypothesis, as thermal stratification events were coupled to hypoxia, reaching anoxia in deep waters of both studied lakes.Indeed, our results consistently suggest a natural susceptibility of deep waters in oligotrophic lakes of the Atlantic Tropical Forest to anoxia and low RI (reaching values < 0) mainly during the summer.These conditions in aquatic ecosystems typically result in low biological diversity (Brewer and Peltzer, 2009;Diaz and Rosenberg, 2008;Vaquer-Sunyer and Duarte, 2008) and high production of CO 2 and other more powerful greenhouse gases (Conrad et al., 2011;Bastviken et al., 2011).Here, the natural water mixing during the beginning of the dry winter showed a reversal oxygenation and increase of RI in deep waters, coupled to the opposite trend at the surface without reaching severe hypoxia throughout the water column.This illustrates that tropical lakes could be very dynamic, but also especially sensitive to organic inputs, which are commonly intensified by human activities in the watershed, like from untreated discharges of sewage and animal manure (Downing and McCauley, 1992).Natural events of anoxia under warm temperatures in tropical waters indicate, therefore, that human-induced organic inputs could potentially contribute to persistence of low O 2 supply and RI resulting in CO 2 evasion to the atmosphere.Studies on the fluctuations of metabolic gases, like O 2 and CO 2 , related to hypoxia at low latitudes are crucial to a better knowledge on the controls and feedbacks of two relevant topics that are often intensified by human activities in broad areas, the organism's death and greenhouse gas emissions in aquatic ecosystems.
(a) rainy season at the end of the summer (March 2005), (b) dry season at the beginning of the winter (June 2005) and (c) dry season at the end of the winter (September 2005).

Fig. 2 .
Fig. 2. Monthly fluctuations in 30-day precipitation (A) and water temperature in Lake Barra (B) and Lake Aguapé (C) during 19 months from March 2004 to October 2005.Four depths were sampled in the morning assuming the light penetration in lake waters: 100 % (unbroken line, filled circle), 10 % (unbroken line, filled triangle), 1 % (dashed line, open circle), and 0 % (dashed line, crosses).See material and methods for details on the determination of each sampling depth.Arrows point to the thermal mixing events during the dry season at the beginning of the winter (2004 and 2005), and the dotted frames indicate the rainy seasons (between 2004 and 2005) in each lake.

Fig. 3 .
Fig. 3. Monthly fluctuations in pO 2 (A and D), pCO 2 (B and E) and respiration index (C and F) in Lake Barra and Lake Aguapé, respectively, during 19 months from March 2004 to October 2005.Symbols, arrows and the dotted frame are as defined in Fig. 2. The dashed-dotted line represents the pO 2 (A and D) or pCO 2 (Band E) in equilibrium with the atmosphere and the critical limit for RI (C and F) to major aquatic organisms(Brewer and Peltzer, 2009).Negative values of RI indicate a ratio pO 2 : pCO 2 < 1.0.

Fig. 4 .
Fig. 4. Vertical profiles of daytime pO 2 , pCO 2 and respiration index from the lake surface to deep waters for Lake Barra (unbroken line, filled squares) and Lake Aguapé (dashed line, crosses) during a thermal stratification event in the rainy season (March 2005).Values are the average ± standard error (SE).The same letters indicate non-significant differences among treatments (p < 0.05, Tukey-Kramer).

Fig. 5 .
Fig. 5. Vertical profiles of daytime pO 2 , pCO 2 and respiration index from the lake surface to deep waters for both lakes during a thermal mixing event in the dry winter (June 2005).Values are the average ± SE.Symbols and letters are as defined in Fig. 4.

Fig. 6 .
Fig. 6.Vertical profiles of daytime pO 2 , pCO 2 and respiration index from the lake surface to deep waters for both lakes during initial thermal stratification events at the end of the dry season (September 2005).Values are the average ± SE.Symbols and letters are as defined in Fig. 4.