Mapping variability in CO2 saturation in Amazonian rivers reveals undersaturated areas

Introduction: Amazonian rivers outgas high amounts of carbon dioxide (CO2). The outgassing estimate, however, remains uncertain due to the limited spatial distribution of data on water-atmosphere CO2 fluxes. So far, the vast extent of the basin and the difficult access to some regions have hampered full mapping. To obtain insight in the spatial distribution of partial pressure of CO2 (pCO2), an important determinant of CO2 fluxes, we analyzed water sampled by a specially equipped hydroplane at an unprecedented large spatial scale in the Amazon basin.

Methods: 1 Study region and sampling All locations (n= 419 for CO2 and n=361 for O2) were sampled within the Brazilian territory of the Amazon basin in 2003 and 2004 as part of the Brasil das Águas project (www.brasildasaguas.com.br). All samples were taken between 11:12 and 20:39 with the vast majority being taken between 13:00 and 18:00. Locations were selected as to cover as much variety in water quality as possible (Tables S1 and S2). River orders ranged from 3 to 10. Samples were taken through a tube 20 cm below the water surface connected to an autosampler on board of a hydroplane (Lake Renegade, model LA-250, USA). The system was flushed with river water four times before samples were taken. Oxygen (O2; % saturation and concentration), temperature, pH and conductivity were measured directly on board of the hydroplane with a YSI multiparameter water quality sonde, model 6600 (Yellow Springs, USA). The sonde contained a YSI 6562 oxygen rapid pulse sensor, a YSI 6561 pH sensor and a YSI 6560 conductivity sensor. The sensors were calibrated daily before the cruise flights. The pH sensor was calibrated with YSI certified standards with a pH of 4, 7 and 10 units. The conductivity sensor was calibrated with a 100 mS/cm YSI certified solution. Samples for the remaining analyses were frozen immediately after sampling using liquid nitrogen. Upon arrival in the laboratory they were kept at -20 oC and analyzed within 20 days after sampling.

2 Sample analysis and GIS data sources Dissolved inorganic carbon (DIC) and dissolved organic carbon (DOC) were measured using a Shimadzu TOC analyzer, model 5000 after carefully filtering through a GF/F Whatman filter. Data on the organic carbon density in the soil (kg C m-2 to 1 m depth) were obtained from The Nelson Institute Center for Sustainability and the Global Environment (IGBP-DIS, 1998; 2015). The net primary production (NPP) data were obtained from Numerical Terradynamic Simulation Group (NTSG) from the University of Montana (product MOD17A3) (n=407 due to missing data for some sampling locations). Stream orders were retrieved from Mayorga et al. (2012). Boundaries of hydrographic sub basins as well as their wetland area and vegetation coverwere retrieved from Melack and Hess (2010).

3 Calculations and data analysis pCO2 (partial CO2 pressure) in water was calculated based on DIC and pH as in Cole et al. (1994), adjusting the equilibrium constants for water temperature according to (Butler, 1991) and ionic strength according to Davies and Shedlovsky (1964). Ionic strength (in moles/L) was estimated based on the electric conductivity (in µS/cm) using a conversion factor of 16 (Ponnamperuma et al., 1966). When DIC was below detection limit (2.5 µM) a DIC concentration equal to the detection limit was used for further calculations (n=47) which results in an overestimation of the pCO2 value. To visualize the spatial distribution of locations with pCO2 values above and below atmospheric equilibrium, we used an equilibrium concentration of 380 ppm. Locations that deviated less than 10% of this value were considered in equilibrium with the atmosphere. Samples for DIC analyses are typically not frozen as this may alter the inorganic carbon concentrations. We tested the effect of our freezing method and found that DIC as well as the pCO2 values before and after freezing were strongly correlated (R2=0.94; p<0.001; n=20 for both correlations). Moreover, pCO2 values below atmospheric levels increased due to sample treatment but stayed below equilibrium concentrations (R2=0.90; p<0.001 n=9; Fig. S1). pCO2 values above atmospheric levels decreased but stayed above equilibrium concentrations (R2=0.92; p<0.001; n=11; Fig. S1). In other words, freezing led to a loss of CO2 in samples above atmospheric levels and an increase in CO2 in samples below atmospheric equilibrium. This indicates that we likely have underestimated the number of locations with pCO2 values below and above equilibrium concentrations and overestimated the number of locations that are in equilibrium with the atmosphere. Samples were classified as taken during high or low-waters as follows. Samples in the larger rivers were classified as taken during low (LW) and high-water (HW) based on the sampling date and the hydrographs of the river where the sample was taken. For the smaller rivers, no hydrographs are available. Here we used the hydrograph of nearby rivers. Hydrographs were obtained from Junk et al. (2011) (2014), Roland et al. (unpublished data) and from the National Water Agency, ANA, Brasil (ANA, 2016; http://hidroweb.ana.gov.br/). Next, we extracted the minimum and maximum water levels and calculated the average. Sampling dates with water levels above the average were classified as HW and below the average as LW. To screen for potential relationships between pCO2 and environmental conditions, we worked on two scales. First, we used individual pCO2 sampling points and local environmental variables: water temperature, DOC concentration, NPP and the soil organic carbon density (pCO2, and DOC were log transformed to approach normality). We used Pearson correlation (one-tailed) to test for significant correlations. Subsequently, we used linear regression to quantify the relationship between pCO2 and the environmental variables that correlated most strongly. In addition, we categorized the individual data points by stream order and used a one-way ANOVA and a post-hoc Tukey test to evaluate potential differences between the stream orders. Second, we used aggregated data on a sub-basin scale. For each sub-basin in which at least four water samples were taken (seven sub-basins for the high-water period and 13 for the low-water period) we determined the median pCO2. Next, we again used Pearson correlation to disclose possible relationships between pCO2 on the one hand and basin characteristics on the other: mean soil organic carbon density, mean NPP, the percentage the subbasin area that is floodable (as a proxy for the relative importance of wetlands in a sub-basin), the percent of the floodable area made of herbaceous and woody vegetation (to further zoom in on potential important characteristics of the wetlands). All analyses were performed in IBM SPSS Statistics 21.

4 references ANA. (2016). Agência Nacional de Águas, Sistema de Acompanhamento de Reservatórios Retrieved from http://ana.gov.br/sar Butler, J. N. (1991). Carbon dioxide equilibria and their applications: CRC Press. Cole, J. J., Caraco, N. F., Kling, G. W., & Kratz, T. K. (1994). Carbon-dioxide supersaturation in the surface waters of lakes. Science, 265(5178), 1568-1570. ://A1994PF33600031 Davies, C. W., & Shedlovsky, T. (1964). Ion association. Journal of The Electrochemical Society, 111(3), 85C-86C. IGBP-DIS. (1998). SoilData(V.0) A program for creating global soil-property databases IGBP Global Soils Data Task, France. Junk, W. J., Piedade, M., Lourival, R., Wittmann, F., Kandus, P., Lacerda, L., et al. (2014). Brazilian wetlands: their definition, delineation, and classification for research, sustainable management, and protection. Aquatic Conservation: Marine and Freshwater Ecosystems, 24(1), 5-22. Junk, W. J., Piedade, M. T. F., Schöngart, J., Cohn-Haft, M., Adeney, J. M., & Wittmann, F. (2011). A classification of major naturally-occurring Amazonian lowland wetlands. Wetlands, 31(4), 623-640. Mayorga, E., Logsdon, M., Ballester, M., & Richey, J. (2012). LBA-ECO CD-06 Amazon River basin land and stream drainage direction maps. ORNL DAAC( Data set. Available on-line [http://daac.ornl.gov] from Oak Ridge National Laboratory Distributed Active Archive Center, Oak Ridge, Tennessee, U.S.A. http://dx.doi.org/10.3334/ORNLDAAC/1086). Melack, J. M., & Hess, L. L. (2010). Remote sensing of the distribution and extent of wetlands in the Amazon basin. In Amazonian floodplain forests (pp. 43-59): Springer. Ponnamperuma, F., Tianco, E. M., & Loy, T. A. (1966). Ionic strengths of the solutions of flooded soils and other natural aqueous solutions from specific conductance. Soil Science, 102(6), 408-413. The Nelson Institute Center for Sustainability and the Global Environment, U. o. W.-M. (2015). Atlas of the Biosphere.

Identifier
DOI https://doi.org/10.17026/dans-zmd-stsg
PID https://nbn-resolving.org/urn:nbn:nl:ui:13-dr-dq1v
Metadata Access https://easy.dans.knaw.nl/oai?verb=GetRecord&metadataPrefix=oai_datacite&identifier=oai:easy.dans.knaw.nl:easy-dataset:163102
Provenance
Creator Kosten, S.; Mendonça, R.; Abe, D.S.; Nes, E.H. van; Roland, F.; Huszar, V.L.M.
Publisher Data Archiving and Networked Services (DANS)
Contributor Vera Huszar, Laboratory of Phycology, Nacional Museum of Rio de Janeiro, Federal University of Rio de Janeiro, Rio de Janeiro, RJ Brazil
Publication Year 2024
Rights info:eu-repo/semantics/openAccess; License: http://creativecommons.org/licenses/by/4.0; http://creativecommons.org/licenses/by/4.0
OpenAccess true
Representation
Resource Type Dataset
Format txt; pdf; csv
Discipline Biology; Life Sciences
Spatial Coverage north=-2.7565043855432503; east=-57.42797206907587Brazilian territory of the Amazon Basin