• Amon, R. M. W., , and R. Benner, 1996a. Bacterial utilization of different size classes of dissolved organic matter. Limnol. Oceanogr. 41:4151.

    • Search Google Scholar
    • Export Citation
  • Amon, R. M. W., , and R. Benner, 1996b. Photochemical and microbial consumption of dissolved organic carbon and dissolved oxygen in the Amazon River system. Geochim. Cosmochim. Acta 60:17831792.

    • Search Google Scholar
    • Export Citation
  • Aufdenkampe, A. K., 2002. The role of sorptive processes in the organic carbon and nitrogen cycles of the Amazon River basin. Ph.D. thesis, University of Washington, 164 pp.

  • Aufdenkampe, A. K., , J. I. Hedges, , A. V. Krusche, , C. Llerena, , and J. E. Richey, 2001. Sorptive fractionation of dissolved organic nitrogen and amino acids onto sediments within the Amazon basin. Limnol. Oceanogr. 46:19211935.

    • Search Google Scholar
    • Export Citation
  • Bustillo, V., 2005. Hydroclimatologie et Biogéochimie appliquées à l’aménagement des bassins fluviaux. Modèles de mélange. Diagnostic et prévision. PhD thesis, Institut National Polytechnique de Toulouse, 426 pp.

  • Bustillo, V., 2007. A large-scale synthetic model applied to the hydroclimatology and eco-geodynamics of the Amazonian basin. FAPESP Rep., Post-doctoral fellowship 2005-58884-5, 93 pp.

    • Search Google Scholar
    • Export Citation
  • Dunne, T., , L. A. K. Mertes, , R. H. Meade, , J. E. Richey, , and B. R. Forsberg, 1998. Exchanges of sediment between the flood plain and channel of the Amazon River in Brazil. Geol. Soc. Amer. Bull. 110:450467.

    • Search Google Scholar
    • Export Citation
  • Gonzales, A. L., , J. Nonner, , J. Heijkers, , and S. Uhlenbrook, 2009. Comparison of different base flow separation methods in a lowland catchment. Hydrol. Earth Syst. Sci. 13:20552068.

    • Search Google Scholar
    • Export Citation
  • Guyot, J. L., 1993. Hydrogéochimie des Fleuves de l’Amazonie Bolivienne. Editions de l’ORSTOM, 261 pp.

  • Hamilton, S. K., , S. J. Sippel, , and J. M. Melack, 2002. Comparison of inundation patterns among major South American floodplains. J. Geophys. Res. 107:8038. doi:10.1029/2000JD000306.

    • Search Google Scholar
    • Export Citation
  • Hedges, J. I., , W. A. Clark, , P. D. Quay, , J. E. Richey, , A. H. Devol, , and U. M. Santos, 1986. Compositions and fluxes for particulate organic material in the Amazon River. Limnol. Oceanogr. 31:717738.

    • Search Google Scholar
    • Export Citation
  • Hedges, J. I., , G. L. Cowie, , J. E. Richey, , P. D. Quay, , R. Benner, , and M. Strom, 1994. Origins and processing of organic matter in the Amazon River as indicated by carbohydrates and amino acids. Limnol. Oceanogr. 39:743761.

    • Search Google Scholar
    • Export Citation
  • Hooper, R. P., , N. Christophersen, , and J. Peters, 1990. End-member mixing analysis (EMMA): An analytical framework for the interpretation of streamwater chemistry. J. Hydrol. 116:321345.

    • Search Google Scholar
    • Export Citation
  • Irion, G., , W. J. Junk, , and J. A. S. N. Mello, 1997. The large central Amazonian river floodplains near Manaus: Geological, climatological, hydrological and morphological aspects. The Central Amazon Floodplain, W. J. Junk, Ed., Springer-Verlag, 23–44.

    • Search Google Scholar
    • Export Citation
  • Johnsson, M. J., , and R. H. Meade, 1990. Chemical weathering of fluvial sediments during alluvial storage: The Macuapanim Island point bar, Solimões River, Brazil. J. Sediment. Petrol. 60:827842.

    • Search Google Scholar
    • Export Citation
  • Junk, W. J., , and M. T. Piedade, 1997. Plant life in the floodplain with special reference to herbaceous plants. The Central Amazon Floodplain, W. J. Junk, Ed., Springer-Verlag, 147–181.

    • Search Google Scholar
    • Export Citation
  • Marengo, J. A., , and R. L. Victoria, 1998. Pre-Large-Scale Biosphere-Atmosphere Experiment in Amazonia Data Sets Initiative, 3 Vols. Center for Weather Forecasting and Climate Study, National Institute for Space Research, CD-ROM.

    • Search Google Scholar
    • Export Citation
  • Martinelli, L. A., , R. L. Victoria, , J. L. I. Dematte, , J. E. Richey, , and A. H. Devol, 1993. Chemical and mineralogical composition of Amazon River floodplain sediments, Brazil. Appl. Geochem. 8:391402.

    • Search Google Scholar
    • Export Citation
  • Martinelli, L. A., , R. L. Victoria, , P. B. Camargo, , M. Piccolo, , L. Mertes, , J. E. Richey, , A. H. Devol, , and B. R. Forsberg, 2003. Inland variability of carbon-nitrogen concentrations and δ13C in Amazon floodplain (várzea) vegetation and sediment. Hydrol. Proc. 17:14191430.

    • Search Google Scholar
    • Export Citation
  • McClain, M. E., , J. E. Richey, , and R. L. Victoria, 1995. Andean contributions to the biogeochemistry of the Amazon River system. Bull. Inst. Fr. Etud. Andines 24:425437.

    • Search Google Scholar
    • Export Citation
  • Meade, R. H., , T. Dunne, , J. E. Richey, , U. M. Santos, , and E. Salati, 1985. Storage and remobilization of sediment in the lower Amazon River of Brazil. Science 228:488490.

    • Search Google Scholar
    • Export Citation
  • Meybeck, M., , and C. Vörösmarty, 2005. Fluvial filtering of land-to-ocean fluxes: From natural Holocene variations to Anthropocene. C. R. Geosci. 337:(1–2). 107123.

    • Search Google Scholar
    • Export Citation
  • Mortatti, J., 1995. Erosão na Amazônia: Processos, modelos e balanço. Ph.D. thesis, University of São Paulo, 150 pp.

  • Quay, P. D., , D. O. Wilbur, , J. E. Richey, , J. I. Hedges, , A. H. Devol, , and R. L. Victoria, 1992. Carbon cycling in the Amazon River: Implications from the 13C composition of particles and solutes. Limnol. Oceanogr. 37:857871.

    • Search Google Scholar
    • Export Citation
  • Redfield, A. C., 1958. The biological control of chemical factors in the environment. Amer. Sci. 46:206226.

  • Richey, J. E., , J. I. Hedges, , A. H. Devol, , P. D. Quay, , R. L. Victoria, , L. A. Martinelli, , and B. R. Forsberg, 1990. Biogeochemistry of carbon in the Amazon River. Limnol. Oceanogr. 35:352371.

    • Search Google Scholar
    • Export Citation
  • Richey, J. E., , S. R. Wilhem, , M. E. McClain, , R. L. Victoria, , J. M. Melack, , and C. Araujo-Lima, 1997. Organic matter and nutrient dynamics in river corridors of the Amazon basin and their response to anthropogenic change. Cienc. Cult. 49:98110.

    • Search Google Scholar
    • Export Citation
  • Richey, J. E., , J. M. Melack, , A. K. Aufdenkampe, , M. V. Ballester, , and L. L. Hess, 2002. Outgassing from Amazonian rivers and wetlands as a large tropical source of atmospheric CO2. Nature 416:617620.

    • Search Google Scholar
    • Export Citation
  • Richey, J. E., , R. L. Victoria, , J. I. Hedges, , T. Dunne, , L. A. Martinelli, , L. Mertes, , and J. Adams, 2008. Pre-LBA Carbon in the Amazon River Experiment (CAMREX) data. Oak Ridge National Laboratory Distributed Active Archive Center dataset. [Available online at http://daac.ornl.gov/cgi-bin/dsviewer.pl?ds_id=904].

    • Search Google Scholar
    • Export Citation
  • Tardy, Y., , V. Bustillo, , and J-L. Boeglin, 2004. Geochemistry applied to the watershed survey: hydrograph separation, erosion and soil dynamics. A case study: The basin of the Niger River, Africa. Appl. Geochem. 19:469518.

    • Search Google Scholar
    • Export Citation
  • Tardy, Y., , V. Bustillo, , C. Roquin, , J. Mortatti, , and R. Victoria, 2005. The Amazon. Bio-geochemistry applied to the river basin management: Part 1. Hydro-climatology, hydrograph separation, mass transfer balance, stable isotopes, and modelling. Appl. Geochem. 20:17461829.

    • Search Google Scholar
    • Export Citation
  • Tardy, Y., , C. Roquin, , V. Bustillo, , M. Moreira, , L. A. Martinelli, , and R. L. Victoria, 2009. Carbon and Water Cycles: Amazon River Basin, Applied Biogeochemistry. Atlantica, 479 pp.

    • Search Google Scholar
    • Export Citation
  • Victoria, R. L., , L. A. Martinelli, , P. C. O. Trivelin, , E. Matsui, , B. R. Forsberg, , J. E. Richey, , and A. H. Devol, 1992. The use of stable isotopes in studies of nutrient cycling: Carbon isotope composition of Amazon varzea sediments. Biotropica 24:240249.

    • Search Google Scholar
    • Export Citation
  • Weng, L. P., , L. K. Koopal, , T. Hiemstra, , J. C. L. Meeussen, , and W. H. Van Riemsdiejk, 2005. Interactions of calcium and fulvic acids at the goethite-water interface. Geochim. Cosmochim. Acta 69:325339.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Map of the Amazon basin upstream from Óbi showing the major tributaries and the geographical repartition of small tributaries (areas colored in gray) along the Amazon River main stem. Numbers in italics stand for the drainage area of major subbasins (expressed in km2).

  • View in gallery

    Compared performance of the five hydrochemical models (M1–M5) based on the probability of nonexceedance of the determination coefficient r2 established by confronting simulated and observed concentrations (or isotopic values). Data represented are obtained by gathering the results of 42 chemical parameters for the station of Óbi, outlet of the studied reach.

  • View in gallery

    (a) Composition of (i) the three individual runoffs RS, RI, and RB and (ii) the river (AVE) obtained by averaging the model M3’s outcomes of nine stations located on the studied Amazonian reach. Calculated data [calc; see Equation (16)] resulting from the discharge weighing of runoffs composition of the major tributaries are compared to observed data (obs) obtained by multilinear regression. Dissolved species and biogeochemical indices (see the appendix for a list of parameters). (b) Composition of (i) the three individual runoffs RS, RI, and RB and (ii) the river (AVE) obtained by averaging the model M3’s outcomes of nine stations located on the studied Amazonian reach. Calculated data [calc; see Equation (16)], resulting from the discharge weighing of runoffs composition of the major tributaries, are compared to observed data (obs.) obtained by multilinear regression. Suspended sediments, organic carbon and nitrogen, C/N molar ratios, and isotopic signature of carbon (δ13C) and water (δ18O).

  • View in gallery

    Influence of the hydrobiological regime, appreciated by for 12 chemical parameters (noted i) at the 11 sampling stations (noted j) over the Amazon River longitudinal profile: VG, SAI, Xib, Tup, Jut, Ita, Ano, Man, SJA, Pau, and Óbi (the outlet of the studied reach); [i] AVE,8R is the mean concentration of i, calculated by discharge weighing the inputs of the eight major tributaries upstream from Óbi; is a calibrated parameter [outcomes of the model M4; see Equation (11)] corresponding to the rate of uptake or release of each bioactive element (i) for each station (j) associated to biologically mediated processes in the river water and describing thus the response of chemical parameters to . Here, means that Cij rises with photosynthetic pathways , decreases with mineralization pathways and vice versa.

  • View in gallery

    Variation of the coefficients γij over the Amazon River (longitudinal profile between Jut and Óbi) obtained by the model M5 for a sample of 24 chemical parameters (i) and 7 sampling stations (j): Jut, Ita, Ano, Man, SJA, Pau, and Óbi (the outlet of the studied reach). The coefficients γij enable tracking of the influence of floodplains water balance on the compositional changes of water chemistry for a given parameter (i) at a given station (j): γij > 0 indicates that concentrations are higher (all other things being equal) when the floodplains drain and vice versa. For example, γij = 0.92 for fine suspended sediments (FSS) at Óbi, indicating that [FSS] in the outgoing flow increases by 92% compared to [FSS] in the incoming flow (data calculated by discharge-weighing chemical signals from the eight tributaries upstream from Óbi) when (i.e., outflow = 2 × inflow).

  • View in gallery

    Mean simulated variations of (a) δ13C (DIC), (b) [CSS], and (c) δ13C (POCC) as a function of the river outflow (Qt) and the floodplain water balance . The fluctuations modeled by M6 over an annual cycle, at the station of Óbi, are represented by arrows, showing (a) 13C/12C depletion during falling waters (path 3 → 4); (b) sedimentation patterns on the 1 → 2 → 3 paths and remobilization patterns on the 3→ 4 → 1 paths; and (c) the exportation of the várzeas grasses (13C/12C enriched) toward the main channel during the falling water stage (3 → 4 path). The hydrological sequence is 1) lowest waters with outflow = inflow (FBW = 0); then 2) rising waters, with outflow < inflow (FBW < 0); then 3) highest waters, with outflow = inflow (FBW = 0); then 4) falling waters, with outflow > inflow (FBW > 0); and then 1) lowest waters.

  • View in gallery

    Comparison between the fluctuations of discharge of the Amazon River (10 stations) and the theoretical compositional fluctuations (9 virtual stations) estimated by cumulating the inputs and inflows: total discharge; Qt; and river flow components superficial runoff QRS, interflow QRI, and baseflow QRB (all data given in mm yr−1).

  • View in gallery

    (Continued)

  • View in gallery

    (Continued)

  • View in gallery

    (Continued)

  • View in gallery

    M6 at the station of Óbi. Representation of the compositional variations (for all the parameters) as a function of the river flow (x axis) and the water balance of the floodplains appreciated by Δ Qt (I-O) (y axis). The input is supposed to exhibit a constant concentration corresponding to the value at the center of the diagram for which Qt = 1122 mm yr−1 and ΔQt (I-O) = 0.13. The stages 1, 2, 3, and 4 correspond to the lowest waters, rising waters, peak of discharge, and falling waters, respectively.

  • View in gallery

    (Continued)

  • View in gallery

    (Continued)

  • View in gallery

    (Continued)

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 66 66 5
PDF Downloads 47 47 3

Biogeochemistry of the Amazonian Floodplains: Insights from Six End-Member Mixing Models

View More View Less
  • * Centro de Energia Nuclear na Agricultura, Laboratório de Geoprocessamento e Tratamento de Imagens, Piracicaba, Brazil
  • | + Université François-Rabelais de Tours, UMR CNRS/INSU 6113 Institut des Sciences de la Terre d’Orléans, Université d’Orléans, Tours, France
  • | # USP–ESALQ, NUPEGEL, Piracicaba, Brazil
  • | 4 Embrapa Monitoramento por Satélite, Campinas, Brazil
  • | * * Universidade Federal de Mato Grosso, UFMT, Campus de Rondonópolis, Rodovia Rondonópolis-Guiratinga, Rondonópolis, Brazil
  • | ++ Universidade Federal do Tocantins, AgroUnitins, Palmas, Brazil
© Get Permissions
Full access

Abstract

The influence of Amazonian floodplains on the hydrological, sedimentary, and biogeochemical river budget was investigated along the Vargem Grande–Óbidos reach, by applying six mixing models based on variable regional and/or variable hydrological sources. By comparing the output of many different models designed for different purposes, the nature and the magnitude of processes linking water and biogeochemical budgets of the Amazonian floodplains were clarified. This study reveals that most of the chemical baseline of the Amazon River basin is acquired before the studied 2000-km Amazonian reach. However, the tight connection between the hydrograph stage of the river and the chemical signals provides insightful information on the dynamics of its floodplains. The chemical expression of biotic and abiotic processes occurring in the Amazonian floodplains can be particularly perceived during falling waters. It appears delayed in time compared to the maximum extension of submerged area, because the alternating water circulation polarity (filling versus emptying) between the main channel and the adjacent floodplains determines delayed emptying of floodplains during falling waters. It results also in a longer time of residence in the hydrograph network, which strengthens the rate of transformation of transiting materials and solutes. Biotic and biologically mediated processes tend to accentuate changes in river water chemistry initiated upstream, in each subbasin, along river corridors, indicating that processes operating downstream prolong those from upstream (e.g., floodplains of the large tributaries). Conversely, the flood wave propagation tends to lessen the seasonal variability as a result of the water storage in the floodplains, which admixes waters of distinct origins (in time and space). The morphology of floodplains, determining the deposition and the diagenesis of the sediments as well as the variable extension of submerged areas or the chronology of floodplains storage/emptying, appears to be the main factor controlling the floodplains biogeodynamics. By coupling classical end-member mixing models (providing insight on hydrological source) with a variable regional contribution scheme, relevant information on the biogeochemical budget of the Amazonian floodplains can be achieved.

* Corresponding author address: Vincent Bustillo, Université François Rabelais de Tours, Parc Grandmont, UFR Sciences et Techniques, Bâtiment E, 37200 Tours, France. bustillovincent@hotmail.com

Abstract

The influence of Amazonian floodplains on the hydrological, sedimentary, and biogeochemical river budget was investigated along the Vargem Grande–Óbidos reach, by applying six mixing models based on variable regional and/or variable hydrological sources. By comparing the output of many different models designed for different purposes, the nature and the magnitude of processes linking water and biogeochemical budgets of the Amazonian floodplains were clarified. This study reveals that most of the chemical baseline of the Amazon River basin is acquired before the studied 2000-km Amazonian reach. However, the tight connection between the hydrograph stage of the river and the chemical signals provides insightful information on the dynamics of its floodplains. The chemical expression of biotic and abiotic processes occurring in the Amazonian floodplains can be particularly perceived during falling waters. It appears delayed in time compared to the maximum extension of submerged area, because the alternating water circulation polarity (filling versus emptying) between the main channel and the adjacent floodplains determines delayed emptying of floodplains during falling waters. It results also in a longer time of residence in the hydrograph network, which strengthens the rate of transformation of transiting materials and solutes. Biotic and biologically mediated processes tend to accentuate changes in river water chemistry initiated upstream, in each subbasin, along river corridors, indicating that processes operating downstream prolong those from upstream (e.g., floodplains of the large tributaries). Conversely, the flood wave propagation tends to lessen the seasonal variability as a result of the water storage in the floodplains, which admixes waters of distinct origins (in time and space). The morphology of floodplains, determining the deposition and the diagenesis of the sediments as well as the variable extension of submerged areas or the chronology of floodplains storage/emptying, appears to be the main factor controlling the floodplains biogeodynamics. By coupling classical end-member mixing models (providing insight on hydrological source) with a variable regional contribution scheme, relevant information on the biogeochemical budget of the Amazonian floodplains can be achieved.

* Corresponding author address: Vincent Bustillo, Université François Rabelais de Tours, Parc Grandmont, UFR Sciences et Techniques, Bâtiment E, 37200 Tours, France. bustillovincent@hotmail.com

1. Introduction

This paper is dedicated to the study of coupled biogeochemical, sedimentary, and hydrological budgets of the Amazonian floodplains along a 2000-km reach extending between the stations of Vargem Grande (VG) and Óbidos (Óbi).

1.1. Preliminary work

Based on the chemical data of the Carbon in the Amazon River Experiment (CAMREX), biogeochemical mass balances over the studied reach were calculated at 10 sampling sites, well spatially distributed, by comparing incoming and outgoing signals and fluxes (Bustillo 2007) with respect to 44 physicochemical parameters. This approach, based essentially on empirical observations instead of modeling outputs, emphasized that the anomalies of mass balances were mainly related to hydrograph stages and to the hydrological balance of the floodplains. Geochemical and hydrological information were treated in a lumped way, providing thus a pertinent insight on the complex hydrological and chemical linkages normally present between floodplains and river channels. Deliberately based on facts instead of modeling outcomes, this preliminary work raised many intriguing questions with respect to the structure of flux and signal anomalies (e.g., the coarse fraction of particulate organic carbon is very significantly 13C enriched during falling waters). The calculations of mass balances were performed in an exhaustive way, at the 10 monitoring stations, involving 44 physicochemical parameters over 8 sampling cruises. However, the determinants of flux imbalances remain to be identified.

1.2. Research objectives and challenges

This study aims at deciphering the nature and magnitude of underlying processes driving the transport of particulate and dissolved species toward the ocean and the gaseous exchanges at the river–atmosphere interface, with a special focus on floodplain–channel linkages. This last component is especially significant in large river systems and is highly relevant to contemporary international debates on the human modification of floodplain land use, flood control, and the construction of levees and reservoirs, which all act to decouple floodplains from stream channel environments. Spatially linking tributary streams and longitudinal shifts in hydrology, water chemistry, and sedimentology is therefore a very challenging issue. Achieving this purpose requires testing the validity of the interpretative statements inspired by the empirical observations (preliminary work). The question addressed in this paper can be formulated as follows: what is the actual impact of the floodplains and of their hydrological functioning on the biogeochemical budget of the Amazon River basin?

To these ends, six hydrochemical modeling strategies, based on end-member mixing concept and using tracer-based separation methods, were implemented. These approaches aim to link hydrological pathways and chemical signals in order to couple hydrological and biogeochemical budgets. Comparing the outputs of these different models designed for different purposes is expected to provide a better sense of the whole by better constraining the range of possible interpretations given to the flux imbalances.

1.3. On the use of end-member mixing models

End-member mixing models provide comprehensive understanding of runoff generation processes with a special focus on hydrological pathways, contributive areas, and retention times (Gonzales et al. 2009). However, the direct measurement of each contributive runoff in a continuous way and at a sufficient number of locations is practically impossible (Tardy et al. 2004; Bustillo 2005). Hydrograph separation methods can be divided in two main categories: tracer-based and nontracer-based separation methods. Nontracer-based separation methods are based on the analysis of hydrographs, including a large variety of procedures, including graphical analysis of recession curves, low-pass filtering, unit hydrograph modeling with extrapolation to rising limb of hydrographs, and rating curve methods linking groundwater levels and river flow. Tracer-based separation methods are based on a mass balance approach determined by the conservative mixture in variable proportions of compositionally constant end members (or at least sufficiently stable and distinct from one end member to another to make the procedure achievable). They are usually recognized to deliver valuable information about the groundwater contribution to the river discharge, provided that adequate tracers are selected. The procedure proposed by Hooper et al. (Hooper et al. 1990), which was called end-member mixing analysis and based on the identification of end members by principal component analysis (PCA), was widely applied for studying the hydrology of small catchments. Its implementation supposes that the water chemistry within each hydrological component is known. At the scale of very large river basins, no such input data can be measured, particularly because of spatial heterogeneities of tracer concentrations and because the fluxes supplied by hydrological reservoirs are not systematically conservative because of in-stream processes and fluvial filtering in river corridors and floodplains (Meybeck and Vörösmarty 2005).

The synthesis of Mortatti (Mortatti 1995) attempted to provide new insights into Amazon River hydrology by gathering biogeochemical and hydrological approaches. However, the hydrograph separation is based on two reservoirs only; although it is very interesting, it proved to be insufficient, particularly because it did not allow capture of the very significant influence of the floodplains on the biogeochemical budget of the Amazon River basin. Despite peculiar cases that are easily corrected case by case, it appears that most large river basins, whatever their morphology and hydroclimatology, are reasonably modeled using the hydrograph separation concept, dividing the total discharge into—at least—three reservoirs.

1.4. Organization of the manuscript

To overcome the difficulties mentioned above, we proposed to investigate the hydrologic function of the Amazon River floodplains, between Vargem Grande (before the confluence of Rio Iça) and Óbidos (the outlet of the studied area), by applying six complementary modeling approaches (including end-member mixing models) to the successive sampling stations located along the main stem of the Amazon. These are based on 1) variable regional sources with (model M2) and without (model M1) correction of inputs by small tributaries; 2) variable hydrological sources with three end members (model M3) to determine their individual compositional evolution, with contrasted response depending on hydrograph stage, throughout their course in the floodplains; 3) variable hydrological sources with three end members, including a correction on the baseflow to account for in-stream biogenic transformations (model M4); and 4) mixed approaches (models M5 and M6) combining the regional variability of chemical signals (between river basins) and the variability related to hydrological source (between contributing runoffs or end members), taking into consideration the defaults of floodplains water balance.

The compositional changes of the chemical baseline in each individual reservoir, set in evidence by comparing their composition within incoming (tributaries) and outgoing (Amazon River reach) runoffs, are more particularly analyzed. By determining hydrological sources and the magnitude of their individual compositional changes, this approach delivers a valuable and original insight on the main factors [hydrological source, water budget of the floodplains, nature of hydrobiological pattern (e.g., photosynthesis versus mineralization, air–water gaseous exchanges, etc.)] driving the biogeochemical and sedimentary budgets of Amazonian floodplains.

2. Study area and dataset

The main physiographic structural elements of the basin include (i) the Precambrian, highly weathered Guyana and Brazilian shields; (ii) the Andean mountains to the west; (iii) the Andean alluvial foreland; and (iv) a large alluvial plain along the Amazon main stem. Soils in the lowlands are generally deep and highly weathered, with widespread covers of sandy podzols in the shields. The soils in floodplains (and alluvial regions around main stems draining the Andes) are much less weathered because of the continuous input of fresh sediments delivered by physical erosion. The Amazon River and most of its large tributaries have developed extensive floodplains, which are integral parts of the river systems (Richey et al. 1997). After leaving the Andean foothills, the tributaries of the Amazon converge into a large sedimentary plain, where they deposit large volumes of sediments (e.g., Guyot 1993) and inundate the floodplains via an extensive network of drainage channels called paranas. Two distinct types of floodplain channels are (i) tributary channels that drain upland terraces and (ii) distributary channels that transport main-stem Amazon water and sediment to floodplain lake basins. A synthetic view of the Amazon River basin upstream from Óbidos, including the delineation of major subbasins, the location of the sampling station along the Amazon River main stem, and the geographical repartition of small tributaries ungauged during the CAMREX project, is presented (Figure 1).

The samples were collected during the CAMREX project over the period 1982–84 (eight cruises), during contrasted hydrographic stages, completed by five additional cruises between 1985 and 1991 focusing on specific topics, for which an exhaustive dataset is not available (thus not considered in this paper). The objective of CAMREX project was to define by mass balances and direct measurements the processes that control the distribution of bioactive elements (C, N, P, and O) in the main stem of the Amazon River in Brazil. The CAMREX dataset represents a time series unique in its length and detail for very large river systems. The dataset, extracted from Pre-Large-Scale Biosphere-Atmosphere Experiment in Amazonia (PreLBA) compilation (Marengo and Victoria 1998; Richey et al. 2008), consists in representative flux-weighted water samples for comprehensive chemical analysis measured over 18 different sites within a 2000-km reach of the Brazilian Amazon main stem, including seven major tributaries. This dataset constitutes, until that date, the basis of more than 130 CAMREX publications, which have focused on understanding physical and biogeochemical dynamics throughout the basin using a large variety of approaches (e.g., Richey et al. 1990). Monitoring stations are located a few kilometers upstream of the confluence of the seven major tributaries with the Amazon River, Iça, Japura, Jutai, Jurua, Purus, Negro, and Madeira, and along the Amazon River at the 11 following stations: Vargem Grande (VG), Santo Antonio do Iça (SAI), Xibeco (Xib), Tupe (Tup), Jutica (Jut), Anori (Ano), Itapeua (Ita), Manacapurú (Man), São Jose da Amatari (SJA), Paurá (Pau), and Óbidos (Óbi; 4 619 000 km2; the outlet of the studied area). Thus, it becomes possible to compare the inputs from tributaries and the outputs of the Amazon River at different locations along the longitudinal profile of the main stem.

3. Modeling strategy

Six mixing models of increasing complexity are implemented (Table 1). They belong to three distinct categories. The first category (models M1 and M2) accounts the variable contribution of the subbasins to the biogeochemical budget. The second category relies on end-member mixing models (models M3 and M4), which allow the identification of source reservoirs, supposed to have constant composition but contributing in variable proportion to the river flow. A third category of model, taking into account the variability related to regional contrasts and hydrological source, is also explored (models M5 and M6).

3.1. Variable regional source

The first model (M1) fundamentally relies on the comparison between calculated and observed longitudinal profiles of concentrations. Upstream, the chemical composition of Rio Solimões constitutes 1) the starting point. Along its course, in low plains, several tributaries join the principal river and modify its chemical composition: the rivers 2) Içá; 3) Jutai; 4) Japurá; 5) Juruá; 6) Purus; 7) Negro; and 8) Madeira, the last one before 9) Óbidos, the outlet of the Amazon River basin chosen in this study. Considering the concentrations Cij of each chemical species i, the total discharge Qtj of the jth confluent, and before the confluence with the tributary (j + 1), the concentration in the Amazon River after the (j + 1)th confluent is established as follows:
i1087-3562-14-9-1-e1
Subscripts i and superscripts j correspond to the parameter analyzed and to the number of tributaries contributing to the Amazon River flow at each station considered (from j = 1 standing for Santo Antonio do Içá to j = 8 designating the station of Óbidos), respectively. Assuming conventionally that and , one obtains
i1087-3562-14-9-1-e2
After the jth confluent, becomes , and the cumulative runoff changes from .
For each parameter, the concordance between theoretical (i.e., calculated) and observed longitudinal profiles is appreciated by analyzing the fitting capability of the simple linear model,
i1087-3562-14-9-1-e3
The correlation coefficient r2 is calculated for each parameter (ni = 44) and each station (nj = 10). The results of calibration are given in the appendix (Table A2). Then, the slope αij (ideally close to 1) and the intercept to origin βij (ideally close to 0) are considered. Finally, the mean bias B is calculated as follows:
i1087-3562-14-9-1-e4
where k indexes the number of the sample and for the number of samples for each station and each parameter. This bias, calculated on average discharge-weighed values, allows estimating the mean chemical composition of floodplains and small tributaries Cij(υ), assuming that measured biases depend on their variable contribution to river flow. The model M2 corrects M1 by taking into account the mean composition of small tributaries and floodplains. The composition of the additional flow (small tributaries plus alluvial aquifers), noted Cij(υ), is estimated by the analysis of differences (composition and flow) between the sum of major tributaries (calculated input), Φij(in) = Cij(in) × Qtj(in), and the output (measured output), Φij(out) = Cij(out) × Qtj(out), corresponding to the river water composition at the considered station:
i1087-3562-14-9-1-e5
Finally, the concentrations obtained through the model M2 for each cruise are given by
i1087-3562-14-9-1-e6
To test the model fitting capability, linear equations comparable to those presented above for M1 are also calibrated for M2. By averaging the estimations of all the samples, it must be reminded that Bij = 0 for each station and each parameter in the case of M2.

3.2. Variable hydrological source

The third model (M3) relies on the hydrograph separation into three components: the superficial runoff RS, the interflow RI, and the baseflow RB. These reservoirs are meant as the expression of spatially organized tributary basins with vertical (top–bottom in soils) and upstream–downstream gradients involving three mixing end members:
  • RS tracking superficial and hypodermic pathways that arise more particularly in upstream areas, which provide most of the solid load transported by fluvial systems;
  • RI tracking superficial and hypodermic pathways that arise more particularly in downstream areas where deep leached soils provide waters of low dissolved (except for organic matter) and solid loads; and
  • RB tracking groundwater pathways, corresponding to the leaching of the soil horizon C (permeable saprolite) that occasionally emerge in the gleysols of lowlands, as inferred from the very characteristic 18O enrichment of waters (Tardy et al. 2009).
The identification of these three components relies upon chemical tracing. The concentrations of Na+ and fine suspended sediment [FSS] are selected as the best tracers (Tardy et al. 2005). The fluctuations of [Na+] track the processes of dissolution and evaporation, which tend to generate a concentration gradient from the superficial layer of soil to groundwaters that are directly at the contact with the chemical front of alteration. On the other hand, the fluctuations of [FSS] track the soil erosion, which is almost specific of surface runoff. The chemical tracers in each reservoir determine the contribution of these source reservoirs to the total river flow Qt by solving the system of equation composed of two equations of mass conservation for each tracer [Equation (7)] and the equation of flow conservation [Equation (8)],
i1087-3562-14-9-1-e7
i1087-3562-14-9-1-e8
where j indexes the considered subbasin and i the parameter used as a chemical tracer.
Here [Na+] and [FSS] within each reservoir correspond to the values established by Tardy et al. (Tardy et al. 2005). The next step consists in adjusting CijRS, CijRI, and CijRB to the whole dataset (42 parameters, excluding the 2 tracers) by performing multilinear regressions. As a result, we define statistically the most probable composition of each reservoir RS, RI, and RB. End-member mixing models are calibrated for each of the 10 stations located along the Amazon River main stem (from Saõ Antonio do Içá to Óbidos) but also for each of the eight major tributaries. Then, the decomposition of hydrograph is performed theoretically by cumulating the reservoir inflow QK of each tributary (trib) to each station (1 ≤ j ≤ 8),
i1087-3562-14-9-1-e9
where K is the index for hydrological reservoirs, RS, RI, or RB. Consequently, for each hydrological node, two models of repartition are implemented. The first one is calibrated using chemical data measured at each of the 10 stations, whereas the second one is calibrated using a calculated pool of chemical data, corresponding to the variable spatial contribution of subbasins to the Amazon River discharge. It is expected that differences between the chemical characteristics of reservoirs are good indicators of floodplains biogeodynamics. A full dataset is provided in the appendix (Table A3).

3.3. Biologically mediated processes

The biological control of chemical factors in river, popularized by Redfield (Redfield 1958), is evaluated in the model M4 by testing the influence of biotic processes on the composition of the baseflow RB. The protocol of calculation for evaluating QRS(j), QRI(j), and QRB(j) is identical to that of the model M3. The composition of baseflow is supposed to be variable as a function of biological pathway tracked with the synthetic variable IBIO,
i1087-3562-14-9-1-e10
In the case of intense photosynthesis, O2 is actively produced while CO2 is removed and consequently IBIO increases. Conversely, when the decomposition prevails, CO2 is actively produced while O2 is removed and consequently IBIO diminishes. The model M4 is formalized as follows:
i1087-3562-14-9-1-e11
with corresponding to the rate of uptake or release of each bioactive element (i) for each station (j) associated to biologically mediated processes in the river water. If , the concentration increases when the photosynthesis pathway prevails and decreases when the mineralization predominates . The mineralization leads to the removal of dissolved O2 and to the release of CO2. That is the reason why IBIO associated to mineralization paths is usually negative and potentially very negative. Therefore, indicates that the concentration increases when mineralization pathway prevails and decreases when photosynthesis predominates . In the case of isotopic data (δ18O, δ13C) that are all negative, the interpretation of is inverted. For simplification purposes, the signs of associated to isotopic values were systematically inverted to homogenize the deciphering for all the parameters. Full model outcomes relative to M4 are presented in the appendix (Table A4).

3.4. Composite approach

The model M5 is a composite approach that integrates both variable spatial contribution (M1) and variable hydrological source (end-member mixing models). First, we establish for each cruise k (Nk = 8) each parameter i and each hydrological node j, the relative difference, noted , between calculated (calc) and observed (obs) values, as follows:
i1087-3562-14-9-1-e12
The second step consists in relating this relative difference with several factors. We have selected three covariates corresponding to the relative differences Δ(QRS/Qt), Δ(QRI/Qt), and ΔQt:
i1087-3562-14-9-1-e13
with
i1087-3562-14-9-1-e14
The term δij in Equation (13) stands for the residual relative difference when the three following conditions are fulfilled:
i1087-3562-14-9-1-eq1
The coefficients αij, βij, and γij, estimated by multilinear regressions, provide qualitative information on river diagenesis in RS (surface runoff), RI (interflow), and Qt (total runoff). If the coefficient is positive, it indicates that the concentration increases in the correspondent runoff as the individual discharge QK increases. Considering the total river flow Qt, the sign of γij indicates whether the discharge of floodplains, roughly estimated by , contributes to increase or decrease the chemical concentration of the chemical parameter i in the river water at the station j. Values of QRS, QRI, QRB, and Qt are given in the appendix (Table A1 and Figure A1), with M3 corresponding to model-derived data and M1 corresponding to data calculated from upstream subbasins. A complementary approach (M6) consists in evaluating the combined effect of the total discharge and its excess or deficit ,
i1087-3562-14-9-1-e15
The calibration of these four coefficients for each sampling station and each parameter leads to synthetic 3D diagrams (see Figure A2), which allow describing the compositional fluctuations of the river water as a function of the river discharge and the default of water balance .

4. Results and discussion

After a brief comparison of the five models in term of statistical resolution capability, the information supplied individually by each method is analyzed. The analysis of their mutual consistency is more particularly performed.

4.1. Compared performance

The agreement between calculated and observed water composition is very significant for most of the parameters. The cumulative distribution of correlation coefficient for five tested models is presented in Figure 2. This indicates that 50% of the parameters modeled by M1 exhibit r2 > 0.75. Unexpectedly, M1 provides better results than M2, suggesting that floodplains are not of constant composition, contrary to the assumption underlying the approach M2. The comparison between M3 and M4 reveals a significant improvement of the performance of end-member mixing models by taking into account the “hydrobiological index” IBIO, which allows identifying the parameters influenced by in-stream processes. The level of performance remains deficient (threshold arbitrarily fixed to r2 < 0.60) on 25% of parameters for M4 versus 50% of parameters for M3. Finally, the mixed approach (M5) combining variable spatial contribution and end-member mixing models constitutes a very convenient compromise, which provides the best results.

4.2. Variable regional contribution (M1 and M2)

The simple approach consisting in correlating incoming (theoretical and calculated) and outgoing (measured) concentrations (M1) provides insightful information. It appears that most of linear calibrations are very significant, except for SO42−, HPO42−, coarse fraction of suspended sediment (CSS), particulate organic carbon (POCC), nitrogen (PONC), and C/N (in all fractions). These deficiencies reveal that substantial modifications occur in the floodplains.

Table 2 delivers the mean values of α (slope), β (intersect of line for x = 0), r2 (correlation coefficient), bias, and average for each parameter and for the 10 sampling stations located along the Amazon main stem. These linear calibrations indicate that the compositional fluctuations of river water in the Amazon reach might be greater than those impulsed by the tributaries inputs (α > 1 and β < 0) for pH, K+, Mg2+, NO3, CO2, NaSil, KSil, CaSil, MgSil, dolomite, and FR. For example, when the inflow defines a low pH, the outflow is still more acidic and conversely, when the inflow defines an elevated pH, the outflow is more basic. Considering the parameters listed before, the open system dynamics along the Amazon main stem (and in its floodplains) accentuate the chemical perturbations initiated upstream, in the subbasins. Conversely, the compositional fluctuations generated in the tributaries tend to be buffered in the outflow (α < 1 and β > 0) for other parameters such as Ca2+, HCO3, DIC, Cl, DOC, O2, CSS, POCC, C/N (all the fractions), dissolved and particulate organic nitrogen, and [CaCO3].

A complementary analysis of Table 2 consists in assessing the bias between the concentrations in the inflow and in the outflow. Positive values (bias > 0) indicate that concentrations in the inflow are superior to those measured in the outflow and vice versa. Negative biases are observed for Ca2+, HCO3, Cl, O2, CSS, PONF, PONC and CaCO3. We observe also a decrease of δ13C in all the carbon fractions: DIC, POCF, and to a lower extent POCC. In turn, positive biases are obtained for NO3, CO2, and C/N (in all the fractions), whereas δ18O gets less negative. Globally speaking, the measured composition of the Amazon River (outputs) follows the chemical baseline imprinted by the tributaries (inputs). Thus, in-stream processes arising in the studied reach do not fundamentally modify the chemical composition acquired in the tributaries. The mitigation of compositional fluctuations is probably related to the contribution of ungauged rivers, which influence substantially the chemical signal measured in the Amazon River (e.g., Ca2+, HCO3, and Cl) because of the very low salinity of small rivers draining thick, sandy soils in central Amazonia. Conversely, the accentuation of trends observed downstream seems to be due to organic matter decay, which is expected to take place in the floodplain as water slowly enters the stream channel from temporary storage. This leads to the release of CO2 (13C depleted) and nitrogenous dissolved species (NO3 and DON) and symmetrically to the removal of O2.

In the model M2, outputs are adjusted by prescribing ad hoc additional contribution of small rivers (whose average composition is not accurately known) that border the Amazon River. The reconstituted mean annual composition of small rivers and floodplains (Bustillo 2007) delivers reliable results for most of the parameters and provides valuable insight on the presumed impact of river processes. However, the correction proposed in the model M2, relying on the variable contribution but constant composition of small rivers and adjacent floodplains, unambiguously fails (Figure 2). This may be due to several factors: 1) the area of flooded areas, on which direct precipitation falls, is variable; 2) the biotic transformations undergone by transiting materials are not the same depending if floodplains fill or dry up; and 3) the respective contributions of small rivers draining lowlands [low total dissolved solids (TDS)] and groundwater (high TDS) fluctuate along the hydrological cycle.

4.3. Hydrograph separation into three reservoirs (model M3)

The detailed modeling outcomes relative to M3 are presented in the appendix (Table A3). A simple comparison can be made between inputs (calculated) and outputs (observed). Mean values are more accurately examined (Figures 3a,b). Observed values are established by averaging the results obtained for nine sampling stations (excluding Vargem Grande and Santo Antonio do Iça, which are located at and close to the upstream boundary) located along the Amazon River main stem. Calculated concentrations (Ĉi,k) are obtained as follows:
i1087-3562-14-9-1-e16
Notice that [C]RS × QRS/Qt + [C]RI × QRI/Qt +[C]RB × QRB/Qt = [C]AVE for all the chemical parameters.

The agreement between both datasets is good, except for C/N, DON, HPO42−, CO2, O2, and pH. Despite some unavoidable deviations resulting from the imprecisions of the chemical analyses and from simplifying assumptions required for modeling, the repartition of chemical species and isotopic signatures display the same pattern. The compositional contrasts between the three reservoirs tend to decrease in the outflow, suggesting that the intermittent storage of water in floodplains contributes to mix waters originating from different sources (e.g., hydrological reservoirs, subbasins). This tends to homogenize their chemical composition at the outlet of the system. The examples provided by CSS, δ18O, and DOC (Figure 3b) and by SO42− and Cl (Figure 3a) are particularly explicit. The greatest deviations are observed for the sand fraction CSS whose transport in the Amazonian reach is considerably delayed compared to solutes and water.

Concerning the major chemical species (anions and cations), concentrations in RS and RI tend to be lower in the outflow: this effect of dilution is very marked for SO42−, Cl, DOC, Ca2+, and HCO3 (Figure 3a). This is consistent with the biogeochemical balance calculated for the floodplains (Bustillo 2007), which did not reveal any dissolution of carbonates in central Amazonia. It is likely that a part of Ca2+ is adsorbed on transiting clay suspensions, fulvic acids, and/or goethite (Weng et al. 2005), whereas HCO3 might be partly converted into CO2 as a result of pH buffering of very acidic waters provided by small Amazonian rivers draining lowlands and Rio Negro. We observe also a very significant increase of weathering rate in the outflow (see CO2 SIL), attributed to the baseflow RB. The consequence is the correlative decrease of the lithological index FR and δ13C (DIC): −9.6 ‰ → −11.7 ‰. However, the values established for δ13C (DIC) and FR are not totally compatible, because a low contribution of carbonates on DIC release (FR low) should lead to a very negative δ13C (DIC). The unexpected heavy signature of δ13C (DIC) in the baseflow (Figure 3b) might be the consequence of 1) CO2 outgassing (Richey et al. 2002) and 2) aquatic photosynthesis, which both subtract preferentially 12C and thus concentrate 13C in the river water.

4.4. Focus on the sampling stations of Paurá and Óbidos

The outcomes of the end-member mixing model are more particularly analyzed downstream from the confluence of the eight major tributaries, at the monitoring stations of Paurá and Óbidos (Table 3). Chemical characteristics of the three reservoirs are almost similar for Paurá and Óbidos; they display distributions comparable to those resulting from the conservative mixing of the eight major tributaries. Even so, the statistical resolution might be deeply altered on specific parameters for one station and not for the other (e.g., Ca2+, Mg2+, and HCO3). Small errors of chemical analyses might have large repercussions on the outcomes of the models, especially when the number of samples is low and when the variability of the chemical baseline is moderate. For the sampling stations located at the outlet of large fluvial basins, the compositional fluctuations of the river water are often attenuated because of the water storage in the floodplains and the slow motion of flood wave, which mixes waters having resided for a short or long time in the surface network. Even when r2 values are low, the model outcomes are qualitatively very instructive to appreciate the dynamics within each reservoir and their heterogeneity. The similarity between the fictitious station (8 Rios) and the sampling stations (Paurá and Óbidos) indicates that the chemical composition (including isotopic composition) of the Amazon River water is essentially acquired before the waters supplied by tributaries reach the Amazonian floodplains. It seems that underlying processes driving biogeochemical budgets (chemical weathering, gas emissions toward the atmosphere, deposition versus remobilization of sediments, etc.) in the tributaries and in the Amazon main reach are of the same nature and define comparable chemical equilibria. The large-scale flooding of lowlands, occurring almost concomitantly over central Amazonia (Hamilton et al. 2002), provides autochthonous organic substrate for decomposition, leading subsequently to the creation of hypoxic and anoxic environments along river corridors. These reductive conditions influence considerably carbon and nutrient cycling resulting from the enhancement of gas emission (CH4, CO2, NO, N2O, and N2) toward the atmosphere and thus determine the isotopic signature of the dissolved inorganic carbon. The major chemical species released by chemical weathering in upstream reaches, where the bedrock outcrops, and in the floodplains, where coarse unweathered sediments deposit, seem to be transported (almost) conservatively within the river network.

4.5. Hydrobiological modeling (model M4)

The analysis focuses more particularly on the sign of the coefficient associated to the hydrobiological factor . Model outcomes, on several characteristic parameters and for the 11 monitoring stations of the Amazon profile, are represented in Figure 4 (full dataset in Table A4 and electronic supplementary material; supplements are available online at http://dx.doi.org/10.1175/2010EI326.s1.). Chemical responses to the hydrobiological factor may be roughly grouped into three categories: 1) , indicating that concentrations (or values) increase concomitantly to IBIO or that concentrations are higher when photosynthetic paths dominate; 2) , indicating that concentrations (or values) increase when the mineralization prevails on the photosynthesis; and 3) , indicating that hydrobiological processes do not significantly influence the chemical baseline. Among the parameters varying like , we have the following: pH, K+, NO3, SO42−, HPO42−, O2, δ13C (DIC), δ13C (POCF), and C/N (three fractions). Among the parameters varying in the opposite sense, we have the following: Ca2+, HCO3, DOC, DIC, CO2, POCF, PONF, PONC, and DON. Other parameters do not exhibit any significant and reproducible correlation with . A break is observed at Manaús, depending if we locate upstream or downstream from the confluent of Rio Negro. Roughly speaking, there is a loss of Ca2+, Mg2+, K+, and HCO3 under photosynthetic regime and conversely a gain of these solutes after the confluent of Rio Negro. Similarly, the decrease of [DOC] is clearly observed downstream from Manaús but remains negligible upstream.

When the photosynthetic paths dominate , CO2 is removed while O2 is released. As a result, the pH increases and influences the nature and magnitude of abiotic processes. First of all, the biological uptake of CO2 operates a carbon fractionation, which tends to make heavier, by mass effect, the value of δ13C (DIC) in the river water. Concomitantly, the decrease of [Ca2+] and [HCO3] supports the hypothesis that Ca2+ is adsorbed by clay minerals, which releases H+ and leads to the subsequent protonization of HCO3 (which frees CO2). After the confluence of Rio Madeira, the dynamics of Ca2+ is reversed: all other things being equal, [Ca2+] and [HCO3] tend to increase (photosynthetic path). The rise of pH and/or [O2] related to photosynthetic paths might promote the side-chain oxidation of nitrogenous functions contained in dissolved organic molecules (Aufdenkampe et al. 2001; Aufdenkampe 2002). Their condensation makes them get more hydrophobic (Tardy et al. 2009) and leads presumably to their sorption onto fine suspended sediments to form diagenetic POCF. The rise of POCF/PONF and POCC/PONC reflects the genesis of autochthonous molecules, which progressively obliterates the signal of soil-derived substances. Actually, riparian grasses and floating aquatic plants that grow in the floodplains exhibit a high atomic ratio C/N, evaluated to 42 by Victoria et al. (Victoria et al. 1992), suggesting that the uptake of NO3 is rather low and does not totally counterbalance the input associated to the sorption of DOM. However, the most significant rise is observed for [DOC]/[DON] because of the diagenesis of DOM, which tends to release NO3 and to concentrate carbon in DOM.

Conversely, when the mineralization paths prevail on photosynthesis , low 13C/12C source of DIC is released in the river while O2 is consumed. The chemical signals associated to mineralization are, roughly speaking, symmetrical to those imprinted by the photosynthesis. The Andean soil-derived POCC, mainly mobilized in surface runoff, is exposed to increasing temperature as transported downstream and subjected to mineralization in lowland environments (McClain et al. 1995). This source of unstable carbon provides a significant amount of carbon substrate that fuels the heterotrophic metabolism of the river, after deposition of large quantities of carbon. In addition to allochthonous POCC, the mechanism locally named “terras-caídas,” corresponding to the large-scale bank erosion during flood periods (Irion et al. 1997), promotes the large-scale destruction of well-developed floodplain forest communities (Junk and Piedade 1997) and provides large amounts of highly unstable organic substrate. The contribution of grass várzeas to the carbon budget of floodplains, appreciated by the isotopic composition of várzea sediments (Victoria et al. 1992; Martinelli et al. 2003), increases as we move downstream. This suggests that the impact of aquatic vegetation on the carbon budget of floodplain progressively increases (Quay et al. 1992). Because of preservation mechanisms during decomposition (such as adsorption-linked protection), the fine fraction POCF is clearly refractory: AVE and does not appear to be significantly influenced by in-stream mineralization.

4.6. The composite approaches (M5 and M6)

The model M5 exhibits an unexpected high performance for all monitoring stations and all chemical parameters, as shown in the appendix (Table A5). The reconstitution of isotopic signatures is very convincing for δ18O (SMOW) and δ13C (DIC), whereas the lowest levels of confidence are observed for POCC, CSS, PONC, and C/N atomic ratios for POCC and POCF. The significant improvement compared to M1 and M3 denotes the influence of floodplains on the chronological variations of the Amazon River composition, mainly because of 1) the polarity of water circulation in the floodplains and 2) the variable contribution of each individual runoff to the water budget of floodplains. The chemical response of river to the polarity of water circulation in the floodplains can be approached by analyzing the magnitude and the sign of γij [cf. Equation (13)] corresponding to the variation of concentration in the river water associated to the water balance of floodplains . The complete dataset compiling the values of γij is supplied in the electronic supplementary materials (Table A5); several selected values are represented in Figure 5. Despite some variations between stations, the magnitude and the sign of γij match quite well. The model M5, taking into account simultaneously the variables , ΔQRS/Qt, and ΔQRI/Qt, although providing reliable outcomes, might sometimes be difficult to interpret, notably because of covariations between variables. For example, and ΔQRI/Qt exhibit a positive correlation simply because the drainage of floodplains principally involves the hydrological reservoir RI. Moreover, the simple interpretation of γij does not allow investigating the additional impact of the river flow, whose magnitude influences floodplain dynamics.

To facilitate the deciphering process, the model M6 was implemented. The calibration of the four parameters [see Equation (15)] leads to the determination of 3D diagrams with spatial representation realized for (average = 1122 mm yr−1) and (average = 0.134). These figures, available in full in the appendix (Figure A2) for the station of Óbidos, represent the variations of concentration (or isotopic value) as a function of the river outflow and as a function of the water balance of floodplains, tracked by . The mean concentration of each parameter, calculated in the inflow (for eight cruises), is centered on and . The isolines represent the changes of chemical characteristics as we stray from the central point, assuming that the chemical composition of the inflow remains constant, whatever and are. In this section, the objective is not to predict the chemical composition of the outflow but rather the deviation to the inflow. As a consequence, the following diagrams must be read and interpreted only in terms of relative values. In each diagram are identified the main sequence of hydrological cycle: 1) lowest waters, characterized by low Qt and excess of outflow; then 2) rising waters, exhibiting intermediate Qt and a severe deficit of outflow; then 3) highest waters, with high Qt and moderate deficit of outflow; then 4) falling waters, with intermediate Qt and excess of outflow; and back to 1) lowest waters.

4.6.1. Major chemical species and chemical weathering

Concerning the major chemical species, a dilution effect is observed during the phase of water storage while concentrations rise when the waters stored in the floodplains join back the main channel. The influence of the circulation polarity is greater after the confluence of Rio Japurá, as the Amazon valley widens. Considering the poles of chemical erosion, it is noticeable that the apparent rate of chemical alteration is substantially greater (both for silicates and carbonates) when the floodplains empty, with the exception of dolomite, all other things being equal (Figure 5). On average, the lithological index FR (cf. list of parameters) exhibits lower values during the emptying of floodplains. It suggests that submerged low plains drain areas where the alteration of sedimentary minerals mimics the weathering of crystalline rocks. It is likely that chemical weathering in floodplains manifests itself sequentially, when the floodplains dry up: that is, when the water stored in low plains joins back the main channel. Following Johnsson and Meade (Johnsson and Meade 1990), these model outcomes support the idea that the chemical weathering of the additional flow is mainly driven by the diagenesis of unweathered sediments that are deposited during the filling of floodplains (Martinelli et al. 1993). The decrease of lithological index FR during the emptying of floodplains coincides with lower δ13C (DIC), compared to the stage of filling (Figure 6a). The parallel evolution of δ13C (DIC) and FR confirms that the isotopic signal of dissolved inorganic carbon is fundamentally determined by the pattern of weathering processes (synthesized by FR). Because of the hydrological dynamics of floodplains, following an annual immutable cycle, the chemical expression of weathering processes is sequential as well. It should also be noticed that the index Re (SiO2/Al2O3 in altered products inferred from the chemistry of river water; Tardy et al. 2004) is much higher during the emptying of floodplains, suggesting that SiO2 might be picked up and converted into a particulate form because of the growth of diatoms, which is encouraged in adjacent lakes and flooded areas.

4.6.2. Budget of sediments

The sand fraction CSS exhibits systematically lower concentrations during the phase of water storage (γij values are all positive in Figure 5) than during the emptying stage (Figure 6b). This result supports the idea that the sediments tend to deposit as the river inundates the low plains and tend to be remobilized as the extension of submerged areas lessens. The greatest contrasts are observed between Itapeua and São José da Amatari and the lowest is accredited to Paurá, after the confluence of Rio Madeira, which provides large amounts of coarse sediments. The trends for the silt–clay fraction are less explicit. Considering the silt–clay fraction FSS, no clear tendency can be outlined on the Vargem Grande–Manacapurú reach. In turn, between São José da Amatari and Óbidos, the pattern is very similar to the one described for CSS, suggesting that the flushing action of Rio Negro (Meade et al. 1985; Dunne et al. 1998) might promote the remobilization of fine sediments when the floodplains drain.

4.6.3. River metabolism

Considering the gaseous composition of the river, the computed γij values indicate higher [CO2] during the emptying of flooded area on the upstream reach and lower [CO2] on the downstream reach (from Manacapurú to Óbidos). All other things being equal, pH appears to be higher during the emptying of floodplains while [O2] is lower.

Considering nitrogenous species, the evolutions are the following: while the floodplains dry up, a generalized drop of [NO3] and a gain of [PONC] and [DON] are observed. The exceptions are Paurá and São José da Amatari, where sorption processes of dissolved organic matter arise (Aufdenkampe et al. 2001; Tardy et al. 2009).

Considering organic carbon species, γij values highlight a gain of POCF and POCC during the emptying of floodplains and conversely a deficit of DOC. The major exception corresponds to the station of Paurá, which is highly influenced by the forwarded contribution of Rio Madeira, which reverberates directly on the chronological evolution of processes.

As the waters stored in the floodplains join back the main channel, the atomic ratio POCF/PONF rise, whereas DOC/DON and POCC/PONC drop drastically. The drift observed along the Amazon profile involves the increasing contribution of submerged areas to the water budget of the Amazon River, as we move downstream. This amplifies the imprint of river diagenesis on the organic matter. As heterotrophic processes operate more and more intensely, the maturation of organic matter is accelerated and leads to lower and lower POCF/PONF in the water draining floodplains. The effects of these processes on the isotopic signature of carbon are not appreciable. Concerning POCC, the concentrations tend to be higher during the emptying of the floodplains. The effects on the isotopic signal of δ13C (POCC) are noticeable, leading unequivocally to a 13C enriched signature when the water stored in the floodplains rejoins the main channel (Figure 6c). The presumed influence of aquatic grasses, whose isotopic signature is heavy (−13 ‰, according to Victoria et al. 1992), on the isotopic composition of POCC, seems to be confirmed here.

The amounts of DOM, both those observed at Óbidos and those reconstituted by modeling (model M6), exhibit a drastic decline of [DON] and a correlative increase of [DOC]/[DON] when the inundation of the floodplains occurs: that is, during rising water stage. At this stage, the chemical nature of DOM, mobilized by surface runoff (Tardy et al. 2005), is mainly soil derived and refractory (Hedges et al. 1986). Low C/N molecules, conveyors of positive charge (e.g., amino acids) or hydrophobic (humic acids), are recognized to be the best candidates to sorption onto fine sediments (Aufdenkampe et al. 2001), leading subsequently to rising C/N in the remaining DOM fraction. Conversely, the emptying of floodplains coincides with a rise of [DOC] and [DON] and a decrease of [DOC]/[DON]. This supports the idea that additional DOC and DON observed during the emptying of floodplains is autochthonous and is released by the decay of aquatic biomass, which contains high proportion of amino acids (low C/N) compared to soil-derived DOM (Hedges et al. 1994). The high variability of C/N tends to confirm that, along a complete hydrological cycle, distinct pools of molecules (allochthonous soil derived versus autochthonous river derived) exhibiting contrasted reactivity and very dissimilar elemental composition are exported by the Amazonian rivers (Amon and Benner 1996a; Amon and Benner 1996b).

5. Summary and concluding remarks

The six hydrochemical models that were tested provide a valuable insight on the main factors [hydrological source, water budget of the floodplains, nature of hydrobiological pattern (e.g., photosynthesis versus mineralization, air–water gaseous exchanges, etc.)] controlling the biogeochemical and sedimentary budgets of the Amazonian floodplains. The influence of floodplain and additional flow (small rivers, alluvial groundwater, and direct precipitation) could be shown for most of the studied parameters (Bustillo 2007). Unfortunately, because of the lack of reliable data concerning the water chemistry of small tributaries, the influence of variable additional input (involving variable contribution of small rivers and alluvial groundwaters to the river flow) cannot be distinguished from the effects of the diagenesis operating in the floodplains. At the light of the results provided by the six mixing models, three main issues dealing with the biogeochemistry and hydrology of the floodplains are addressed:

  1. coupling between sediment deposition and biogeochemical diagenesis;
  2. organic metabolism of the river and its effects on the nature and intensity of biotic processes; and
  3. nature and intensity of abiotic processes, involving notably the sorption of DOM, the evaporation of wetlands, and the river outgassing.

A companion paper (V. Bustillo et al. 2010, unpublished manuscript) aims to investigate more in detail these three topics, which are intrinsically related and which determine most of the biogeochemical budget relative to Amazonian floodplains.

The magnitude and polarity of water exchanges between the Amazon River and its floodplains strongly influence the sedimentary and chemical signals measured in the river waters. The floodplains constitute widespread sites where major biotic and abiotic processes affecting the dynamics of transiting materials occur: sedimentation, remobilization of sediments, organic matter decay, CO2 outgassing, etc.

Unexpectedly, the chemical trends observed upstream are sometimes accentuated downstream, as shown by the model M1. It supports the idea that the processes operating downstream are of the same nature as those occurring upstream, prolonging therefore the imprint given by upstream rivers to the chemical baseline. Because of the increase in floodplain size as we move downstream, the impact of floodplain filling and draining on the biogeochemical qualities of the water are therefore amplified downstream. Taking into account the additional contribution of ungauged areas, using ad hoc constant characteristics to close the river budget (model M2) does not improve the performance of modeling compared to the simplest possible model M1. It means that the composition of the “additional” flow is probably very variable, particularly because the alluvial groundwaters draining unweathered sediments deposited alongside the Amazon River (high TDS) and the small tributaries draining thick, sandy soils (intensively leached, low TDS) do not contribute synchronously to the river flow (Bustillo 2007) and exhibit very distinct chemical characteristics.

The mixing model M3 shows a decrease of compositional differences between hydrological reservoirs as we move downstream. This homogenization might be the result of the mixture of waters having resided more or less durably in the hydrographic network. The main differences between incoming and outgoing compositions are attributable to the baseflow RB and, to a lesser extent, to the delayed direct runoff RI. This would be the result of in-stream biogeochemical processes: the aquatic photosynthesis impacts RB, whose contribution to river flow is at maximum during lowest waters stage (i.e., when autotrophic regime prevails), whereas organic matter decay more particularly impacts RI (↑ CO2, ↓ pH, ↓ O2, ↓ δ13C-DIC, ↓ DON, ↓ DOC, and ↓ DOC/DON), whose contribution is maximum when the emptying of floodplains (where heterotrophic regime prevails) occurs.

The model M4, which is also based on variable hydrological source, involves the hydrobiological index , used as a tracer of autotrophic versus heterotrophic regime. This improves considerably the performances of the simulations, compared to M3. The model M4 enables us 1) to identify the parameters significantly influenced by in-stream processes and 2) to determine their response, depending on the nature and magnitude of hydrobiological regime. Globally speaking, the hydrobiological regime promotes large variations of pH and [O2], which have direct repercussions on the biodynamics of other chemical variables. The autotrophic regime is dominant (IBIO > 0) during lowest waters stage, when 1) the river turbidity is minimum, 2) when the river–floodplain connectivity is interrupted, and 3) when the rate of incoming solar radiation reaching the water body is at maximum (flow concentrated within the well exposed main channel). The rises of pH and [O2] directly induced by aquatic photosynthesis coincide with losses of organic nitrogen to the benefit of mineral nitrogen, increases of 13C/12C for DIC (isotopic fractionation induced by aquatic photosynthesis), and losses of Ca2+ and HCO3. The heterotrophic regime is dominant over the annual cycle, except during lowest waters stage. The heterotrophic signal is hugely amplified when the waters stored in the floodplains rejoin the main channel. Falling waters constitute privileged moments to appreciate the biogeodynamics of the floodplains, because their discharge in the main channel is intermittent. The decrease of pH and [O2] related to the heterotrophic regime coincide with increase of [CO2]; decrease of δ13C(DIC); decrease of [DOC]/[DON]; and rise of δ13C(POCC), which is interpreted as the result of the sequential release of autochthonous carbon (C4 aquatic grasses).

The models M5 and M6 enable the impact of floodplains water balance (filling versus emptying) on the differences of chemical concentrations between the tributaries and the Amazon River to be tested more specifically. This test appears to be very conclusive, showing that chemical signals observed in the Amazon River waters are thoroughly influenced by the magnitude and polarity of water exchanges between the Amazon River main channel and its floodplains.

Acknowledgments

This work was funded by the Brazilian FAPESP agency by way of a postdoctoral fellowship (2005/58884-5) associated to the project entitled “A large-scale synthetic model applied to the hydroclimatology and eco-geodynamics of the Amazon River basin.” This study benefited from insightful comments from two anonymous referees, whose very careful review and far-reaching vision contributed indeed to substantially improve the quality of the language and the clarity of the thinking.

REFERENCES

  • Amon, R. M. W., , and R. Benner, 1996a. Bacterial utilization of different size classes of dissolved organic matter. Limnol. Oceanogr. 41:4151.

    • Search Google Scholar
    • Export Citation
  • Amon, R. M. W., , and R. Benner, 1996b. Photochemical and microbial consumption of dissolved organic carbon and dissolved oxygen in the Amazon River system. Geochim. Cosmochim. Acta 60:17831792.

    • Search Google Scholar
    • Export Citation
  • Aufdenkampe, A. K., 2002. The role of sorptive processes in the organic carbon and nitrogen cycles of the Amazon River basin. Ph.D. thesis, University of Washington, 164 pp.

  • Aufdenkampe, A. K., , J. I. Hedges, , A. V. Krusche, , C. Llerena, , and J. E. Richey, 2001. Sorptive fractionation of dissolved organic nitrogen and amino acids onto sediments within the Amazon basin. Limnol. Oceanogr. 46:19211935.

    • Search Google Scholar
    • Export Citation
  • Bustillo, V., 2005. Hydroclimatologie et Biogéochimie appliquées à l’aménagement des bassins fluviaux. Modèles de mélange. Diagnostic et prévision. PhD thesis, Institut National Polytechnique de Toulouse, 426 pp.

  • Bustillo, V., 2007. A large-scale synthetic model applied to the hydroclimatology and eco-geodynamics of the Amazonian basin. FAPESP Rep., Post-doctoral fellowship 2005-58884-5, 93 pp.

    • Search Google Scholar
    • Export Citation
  • Dunne, T., , L. A. K. Mertes, , R. H. Meade, , J. E. Richey, , and B. R. Forsberg, 1998. Exchanges of sediment between the flood plain and channel of the Amazon River in Brazil. Geol. Soc. Amer. Bull. 110:450467.

    • Search Google Scholar
    • Export Citation
  • Gonzales, A. L., , J. Nonner, , J. Heijkers, , and S. Uhlenbrook, 2009. Comparison of different base flow separation methods in a lowland catchment. Hydrol. Earth Syst. Sci. 13:20552068.

    • Search Google Scholar
    • Export Citation
  • Guyot, J. L., 1993. Hydrogéochimie des Fleuves de l’Amazonie Bolivienne. Editions de l’ORSTOM, 261 pp.

  • Hamilton, S. K., , S. J. Sippel, , and J. M. Melack, 2002. Comparison of inundation patterns among major South American floodplains. J. Geophys. Res. 107:8038. doi:10.1029/2000JD000306.

    • Search Google Scholar
    • Export Citation
  • Hedges, J. I., , W. A. Clark, , P. D. Quay, , J. E. Richey, , A. H. Devol, , and U. M. Santos, 1986. Compositions and fluxes for particulate organic material in the Amazon River. Limnol. Oceanogr. 31:717738.

    • Search Google Scholar
    • Export Citation
  • Hedges, J. I., , G. L. Cowie, , J. E. Richey, , P. D. Quay, , R. Benner, , and M. Strom, 1994. Origins and processing of organic matter in the Amazon River as indicated by carbohydrates and amino acids. Limnol. Oceanogr. 39:743761.

    • Search Google Scholar
    • Export Citation
  • Hooper, R. P., , N. Christophersen, , and J. Peters, 1990. End-member mixing analysis (EMMA): An analytical framework for the interpretation of streamwater chemistry. J. Hydrol. 116:321345.

    • Search Google Scholar
    • Export Citation
  • Irion, G., , W. J. Junk, , and J. A. S. N. Mello, 1997. The large central Amazonian river floodplains near Manaus: Geological, climatological, hydrological and morphological aspects. The Central Amazon Floodplain, W. J. Junk, Ed., Springer-Verlag, 23–44.

    • Search Google Scholar
    • Export Citation
  • Johnsson, M. J., , and R. H. Meade, 1990. Chemical weathering of fluvial sediments during alluvial storage: The Macuapanim Island point bar, Solimões River, Brazil. J. Sediment. Petrol. 60:827842.

    • Search Google Scholar
    • Export Citation
  • Junk, W. J., , and M. T. Piedade, 1997. Plant life in the floodplain with special reference to herbaceous plants. The Central Amazon Floodplain, W. J. Junk, Ed., Springer-Verlag, 147–181.

    • Search Google Scholar
    • Export Citation
  • Marengo, J. A., , and R. L. Victoria, 1998. Pre-Large-Scale Biosphere-Atmosphere Experiment in Amazonia Data Sets Initiative, 3 Vols. Center for Weather Forecasting and Climate Study, National Institute for Space Research, CD-ROM.

    • Search Google Scholar
    • Export Citation
  • Martinelli, L. A., , R. L. Victoria, , J. L. I. Dematte, , J. E. Richey, , and A. H. Devol, 1993. Chemical and mineralogical composition of Amazon River floodplain sediments, Brazil. Appl. Geochem. 8:391402.

    • Search Google Scholar
    • Export Citation
  • Martinelli, L. A., , R. L. Victoria, , P. B. Camargo, , M. Piccolo, , L. Mertes, , J. E. Richey, , A. H. Devol, , and B. R. Forsberg, 2003. Inland variability of carbon-nitrogen concentrations and δ13C in Amazon floodplain (várzea) vegetation and sediment. Hydrol. Proc. 17:14191430.

    • Search Google Scholar
    • Export Citation
  • McClain, M. E., , J. E. Richey, , and R. L. Victoria, 1995. Andean contributions to the biogeochemistry of the Amazon River system. Bull. Inst. Fr. Etud. Andines 24:425437.

    • Search Google Scholar
    • Export Citation
  • Meade, R. H., , T. Dunne, , J. E. Richey, , U. M. Santos, , and E. Salati, 1985. Storage and remobilization of sediment in the lower Amazon River of Brazil. Science 228:488490.

    • Search Google Scholar
    • Export Citation
  • Meybeck, M., , and C. Vörösmarty, 2005. Fluvial filtering of land-to-ocean fluxes: From natural Holocene variations to Anthropocene. C. R. Geosci. 337:(1–2). 107123.

    • Search Google Scholar
    • Export Citation
  • Mortatti, J., 1995. Erosão na Amazônia: Processos, modelos e balanço. Ph.D. thesis, University of São Paulo, 150 pp.

  • Quay, P. D., , D. O. Wilbur, , J. E. Richey, , J. I. Hedges, , A. H. Devol, , and R. L. Victoria, 1992. Carbon cycling in the Amazon River: Implications from the 13C composition of particles and solutes. Limnol. Oceanogr. 37:857871.

    • Search Google Scholar
    • Export Citation
  • Redfield, A. C., 1958. The biological control of chemical factors in the environment. Amer. Sci. 46:206226.

  • Richey, J. E., , J. I. Hedges, , A. H. Devol, , P. D. Quay, , R. L. Victoria, , L. A. Martinelli, , and B. R. Forsberg, 1990. Biogeochemistry of carbon in the Amazon River. Limnol. Oceanogr. 35:352371.

    • Search Google Scholar
    • Export Citation
  • Richey, J. E., , S. R. Wilhem, , M. E. McClain, , R. L. Victoria, , J. M. Melack, , and C. Araujo-Lima, 1997. Organic matter and nutrient dynamics in river corridors of the Amazon basin and their response to anthropogenic change. Cienc. Cult. 49:98110.

    • Search Google Scholar
    • Export Citation
  • Richey, J. E., , J. M. Melack, , A. K. Aufdenkampe, , M. V. Ballester, , and L. L. Hess, 2002. Outgassing from Amazonian rivers and wetlands as a large tropical source of atmospheric CO2. Nature 416:617620.

    • Search Google Scholar
    • Export Citation
  • Richey, J. E., , R. L. Victoria, , J. I. Hedges, , T. Dunne, , L. A. Martinelli, , L. Mertes, , and J. Adams, 2008. Pre-LBA Carbon in the Amazon River Experiment (CAMREX) data. Oak Ridge National Laboratory Distributed Active Archive Center dataset. [Available online at http://daac.ornl.gov/cgi-bin/dsviewer.pl?ds_id=904].

    • Search Google Scholar
    • Export Citation
  • Tardy, Y., , V. Bustillo, , and J-L. Boeglin, 2004. Geochemistry applied to the watershed survey: hydrograph separation, erosion and soil dynamics. A case study: The basin of the Niger River, Africa. Appl. Geochem. 19:469518.

    • Search Google Scholar
    • Export Citation
  • Tardy, Y., , V. Bustillo, , C. Roquin, , J. Mortatti, , and R. Victoria, 2005. The Amazon. Bio-geochemistry applied to the river basin management: Part 1. Hydro-climatology, hydrograph separation, mass transfer balance, stable isotopes, and modelling. Appl. Geochem. 20:17461829.

    • Search Google Scholar
    • Export Citation
  • Tardy, Y., , C. Roquin, , V. Bustillo, , M. Moreira, , L. A. Martinelli, , and R. L. Victoria, 2009. Carbon and Water Cycles: Amazon River Basin, Applied Biogeochemistry. Atlantica, 479 pp.

    • Search Google Scholar
    • Export Citation
  • Victoria, R. L., , L. A. Martinelli, , P. C. O. Trivelin, , E. Matsui, , B. R. Forsberg, , J. E. Richey, , and A. H. Devol, 1992. The use of stable isotopes in studies of nutrient cycling: Carbon isotope composition of Amazon varzea sediments. Biotropica 24:240249.

    • Search Google Scholar
    • Export Citation
  • Weng, L. P., , L. K. Koopal, , T. Hiemstra, , J. C. L. Meeussen, , and W. H. Van Riemsdiejk, 2005. Interactions of calcium and fulvic acids at the goethite-water interface. Geochim. Cosmochim. Acta 69:325339.

    • Search Google Scholar
    • Export Citation

APPENDIX

List of parameters

The full results of the end-member mixing models are provided in the appendices. Table A1 and Figure A1 report the simulated fluctuations of the three runoff components for monitoring [(11)] and virtual [(7)] stations. Tables A2 –A5 show the detailed results (calibrated parameters and performance criteria) of the models M1, M3, M4, and M5, respectively. The results of the model M6 are presented synthetically in Figure A2. The list of parameters shown below is intended to facilitate the self-exploration of the appendices.

a. Indexes

i chemical species

j sampling station

k number of the sample.

b. Hydroclimatic features

RS forwarded direct runoff

RI delayed direct runoff

RB baseflow

QK discharge of each individual runoff with K standing for RS, RI, or RB

c. Geochemical characteristics (for full details, see Tardy et al. 2004)

Re: SiO2/Al2O3 in altered products stoechiometry of clays formed by chemical weathering

CO2 CARB CO2 consumed by the alteration of carbonated rocks

CO2 SIL CO2 consumed by the alteration of crystalline rocks

CO2 TOT: CO2 SIL + 2.CO2 CARB DIC released by geochemical alteration

FR: CO2 CARB/CO2 TOT lithological index is the part of DIC originating from the dissolution of carbonates

WR chemical weathering rate (m Ma−1)

FCO2 rate of CO2 consumption (TC km−2 yr−1)

d. Hydrochemical modeling

hydrobiological index based on the gaseous composition of the river water

rate of uptake or release of chemical species associated to hydrobiological path

Qt/Qt)kj index of hydrograph stage (>0 during rising water, <0 during falling water, and =0 for highest waters and lowest waters)

water balance of floodplains (<0 if filling and >0 if emptying)

αij variation of concentration in the river water associated to the variation of QRS/Qt in the main channel

βij variation of concentration in the river water associated to the variation of QRI/Qt in the main channel

γij variation of concentration in the river water associated to the water balance of floodplains

δij residual variation of concentration in the river water, for ΔQRS/Qt = ΔQRI/Qt = ΔQt = 0

Figure 1.
Figure 1.

Map of the Amazon basin upstream from Óbi showing the major tributaries and the geographical repartition of small tributaries (areas colored in gray) along the Amazon River main stem. Numbers in italics stand for the drainage area of major subbasins (expressed in km2).

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure 2.
Figure 2.

Compared performance of the five hydrochemical models (M1–M5) based on the probability of nonexceedance of the determination coefficient r2 established by confronting simulated and observed concentrations (or isotopic values). Data represented are obtained by gathering the results of 42 chemical parameters for the station of Óbi, outlet of the studied reach.

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure 3.
Figure 3.

(a) Composition of (i) the three individual runoffs RS, RI, and RB and (ii) the river (AVE) obtained by averaging the model M3’s outcomes of nine stations located on the studied Amazonian reach. Calculated data [calc; see Equation (16)] resulting from the discharge weighing of runoffs composition of the major tributaries are compared to observed data (obs) obtained by multilinear regression. Dissolved species and biogeochemical indices (see the appendix for a list of parameters). (b) Composition of (i) the three individual runoffs RS, RI, and RB and (ii) the river (AVE) obtained by averaging the model M3’s outcomes of nine stations located on the studied Amazonian reach. Calculated data [calc; see Equation (16)], resulting from the discharge weighing of runoffs composition of the major tributaries, are compared to observed data (obs.) obtained by multilinear regression. Suspended sediments, organic carbon and nitrogen, C/N molar ratios, and isotopic signature of carbon (δ13C) and water (δ18O).

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure 4.
Figure 4.

Influence of the hydrobiological regime, appreciated by for 12 chemical parameters (noted i) at the 11 sampling stations (noted j) over the Amazon River longitudinal profile: VG, SAI, Xib, Tup, Jut, Ita, Ano, Man, SJA, Pau, and Óbi (the outlet of the studied reach); [i] AVE,8R is the mean concentration of i, calculated by discharge weighing the inputs of the eight major tributaries upstream from Óbi; is a calibrated parameter [outcomes of the model M4; see Equation (11)] corresponding to the rate of uptake or release of each bioactive element (i) for each station (j) associated to biologically mediated processes in the river water and describing thus the response of chemical parameters to . Here, means that Cij rises with photosynthetic pathways , decreases with mineralization pathways and vice versa.

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure 5.
Figure 5.

Variation of the coefficients γij over the Amazon River (longitudinal profile between Jut and Óbi) obtained by the model M5 for a sample of 24 chemical parameters (i) and 7 sampling stations (j): Jut, Ita, Ano, Man, SJA, Pau, and Óbi (the outlet of the studied reach). The coefficients γij enable tracking of the influence of floodplains water balance on the compositional changes of water chemistry for a given parameter (i) at a given station (j): γij > 0 indicates that concentrations are higher (all other things being equal) when the floodplains drain and vice versa. For example, γij = 0.92 for fine suspended sediments (FSS) at Óbi, indicating that [FSS] in the outgoing flow increases by 92% compared to [FSS] in the incoming flow (data calculated by discharge-weighing chemical signals from the eight tributaries upstream from Óbi) when (i.e., outflow = 2 × inflow).

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure 6.
Figure 6.

Mean simulated variations of (a) δ13C (DIC), (b) [CSS], and (c) δ13C (POCC) as a function of the river outflow (Qt) and the floodplain water balance . The fluctuations modeled by M6 over an annual cycle, at the station of Óbi, are represented by arrows, showing (a) 13C/12C depletion during falling waters (path 3 → 4); (b) sedimentation patterns on the 1 → 2 → 3 paths and remobilization patterns on the 3→ 4 → 1 paths; and (c) the exportation of the várzeas grasses (13C/12C enriched) toward the main channel during the falling water stage (3 → 4 path). The hydrological sequence is 1) lowest waters with outflow = inflow (FBW = 0); then 2) rising waters, with outflow < inflow (FBW < 0); then 3) highest waters, with outflow = inflow (FBW = 0); then 4) falling waters, with outflow > inflow (FBW > 0); and then 1) lowest waters.

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure A1.
Figure A1.

Comparison between the fluctuations of discharge of the Amazon River (10 stations) and the theoretical compositional fluctuations (9 virtual stations) estimated by cumulating the inputs and inflows: total discharge; Qt; and river flow components superficial runoff QRS, interflow QRI, and baseflow QRB (all data given in mm yr−1).

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure A1.
Figure A1.

(Continued)

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure A1.
Figure A1.

(Continued)

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure A1.
Figure A1.

(Continued)

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure A2.
Figure A2.

M6 at the station of Óbi. Representation of the compositional variations (for all the parameters) as a function of the river flow (x axis) and the water balance of the floodplains appreciated by Δ Qt (I-O) (y axis). The input is supposed to exhibit a constant concentration corresponding to the value at the center of the diagram for which Qt = 1122 mm yr−1 and ΔQt (I-O) = 0.13. The stages 1, 2, 3, and 4 correspond to the lowest waters, rising waters, peak of discharge, and falling waters, respectively.

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure A2.
Figure A2.

(Continued)

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure A2.
Figure A2.

(Continued)

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Figure A2.
Figure A2.

(Continued)

Citation: Earth Interactions 14, 9; 10.1175/2010EI326.1

Table 1.

Modeling of the Amazon River composition along a 2000-km longitudinal profile. Main principles and rules of calculations of six distinct mixing models: M1 and M2 are based on the variable contribution of regional sources to water and biogeochemical budgets; M3 and M4 rely on the variable contribution of hydrological sources (viz., forwarded direct runoff RS, delayed direct runoff RI, and baseflow RB) with a correction for M4 taking into account the influence of the river processes; and M5 and M6 are composite models taking into account the combined effects of the variable contributions relative to the regional and hydrological sources. Note that i is the index of the chemical species (ni = 44), j is the index of the monitoring station (nj =10), and k is the index of the sample (nk = 8).

Table 1.
Table 2.

Mean parameters of the linear equation (α, β, r2, Bias) relating incoming and outgoing concentration (model M1) for 44 chemical parameters at 10 sampling stations (j) of the Amazon main stem. Data are presented for each chemical parameter (index i) and correspond to the mean values from 10 equations: Cij(obs) = αij × Cij(calc) + βij(nj = 10). The mean biases and r2 are also given.

Table 2.
Table 3.

Compared composition of the three individual runoff RS, RI, and RB for the eight major tributaries (8 Rios, calculated data from discharge weighing of the contributing reservoirs) and the stations of Pau and Óbi (data established from measured concentrations by multilinear regressions).

Table 3.
Table 3.

(Continued)

Table 3.
Table 3.

(Continued)

Table 3.
Table A1.

Total discharge and river flow components (superficial runoff RS, interflow RI, and baseflow RB) as calculated by chemical tracing (model 3); in-long profile of discharge established for each individual cruise (n = 8) and for averages; and observed values on the Amazon River main stem (10 stations) compared to theoretical in-long profile (nine virtual stations) by cumulating the inputs and inflows.

Table A1.
Table A1.

(Continued)

Table A1.
Table A1.

(Continued)

Table A1.
Table A1.

(Continued)

Table A1.
Table A1.

(Continued)

Table A1.
Table A1.

(Continued)

Table A1.
Table A1.

(Continued)

Table A1.
Table A1.

(Continued)

Table A1.
Table A2.

Results of model M1. Calibration of linear coefficient (a = slope; b = x intersect) between the concentrations in the inflow (in) and in the outflow (out). View of correlation coefficient r2 and determination of mean bias for the 10 sampling stations located along the Amazon main stem.

Table A2.
Table A2.

(Continued)

Table A2.
Table A2.

(Continued)

Table A2.
Table A2.

(Continued)

Table A2.
Table A2.

(Continued)

Table A2.
Table A2.

(Continued)

Table A2.
Table A2.

(Continued)

Table A2.
Table A2.

(Continued)

Table A2.
Table A3.

Results of model M3. Chemical composition [C]K of hydrological reservoirs (indexed k) RS, RI, and RB, established by chemical tracing for all parameters (indexed i) for the 10 sampling stations (index j) located along the Amazon main stem and the seven virtual stations whose chemical composition is calculated by adding the successive inputs of major tributaries. View of R-squared value (r2) and mean value (Ave) for each parameter. Notice that QRS/Qt (ij) × [Cij]RS + QRI/Qt (ij) × [Cij]RI + QRB/Qt + [Cij]RB = [Cij]Ave.

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A3.

(Continued)

Table A3.
Table A4.

Results of model M4. Values of KBIO defining the chemical response to the river to the variations of the hydrobiological index IBIO. KBIO > 0 determines rise of concentrations or rise of values when river photosynthesis prevails and conversely KBIO < 0 reveals rise of concentrations or values when the organic matter decay predominates. Values established for the 10 sampling stations located along the Amazon main stem. Chemical composition of the aquifers RB* extrapolated for IBIO = 0. View of R- squared value (r2) and mean value (Ave) for each parameter. Notice that QRS/Qt (ij) × [Cij]RS + QR/Qt(ij) × [Cij]RI + QRB/Qt (ij) × {[Cij]RB* + KBIO(ij) × IBIO(j)} = [C]Ave.

Table A4.
Table A4.

(Continued)

Table A4.
Table A4.

(Continued)

Table A4.
Table A4.

(Continued)

Table A4.
Table A4.

(Continued)

Table A4.
Table A5.

Results of model M5, combining chemical tracing (implying variable contribution of hydrological reservoirs) and variable regional contribution (M1 × M3). Values of linear coefficients α(ij), β(ij), γ(ij), and δ(ij) are adjusted for each chemical parameter (indexed i) and each sampling station (noted j) is located on the Amazon River main stem, in reference to the following equation (see text): ΔC(ijk) = α (ij) × Δ(QRS/Qt), (jk) + β (ij) × Δ(QRI/Qt), (jk) + γ (ij) × ΔQt (I-O), (jk) + δ (ij) View of R-squared value (r2) is after reconstitution of water composition.

Table A5.
Table A5.

(Continued)

Table A5.
Table A5.

(Continued)

Table A5.
Table A5.

(Continued)

Table A5.
Table A5.

(Continued)

Table A5.
Table A5.

(Continued)

Table A5.
Table A5.

(Continued)

Table A5.
Table A5.

(Continued)

Table A5.
Table A5.

(Continued)

Table A5.

* Supplemental information related to this paper is available at the Journals Online Web site: http://dx.doi.org/10.1175/2010EI326.s1.

Save