Floodplains of large rivers: Weathering reactors or simple silos?

Chemical Geology 332–333 (2012) 166–184

Contents lists available at SciVerse ScienceDirect

Chemical Geology

j ourna l homepage: www.e lsev ie r .com/ locate /chemgeo

Floodplains of large rivers: Weathering reactors or simple silos?

Julien Bouchez a,⁎, Jérôme Gaillardet a, Maarten Lupker b, Pascale Louvat a, Christian France-Lanord b,Laurence Maurice d, Elisa Armijos c, Jean-Sébastien Moquet e

a Institut de Physique du Globe de Paris, Equipe de Géochimie-Cosmochimie, Université Paris Diderot, 1 rue Jussieu 75238 Paris cedex 05, Franceb Centre de Recherches Pétrographiques et Géochimiques, Centre National pour la Recherche Scientifique, 15 rue Notre-Dame-des-Pauvres, 54501 Vandoeuvre-lès Nancy, Francec Universidade Federal do Amazonas, Av. General Rodrigo Octávio Jordão Ramos, 3000, Campus Universitário, Coroado, Manaus, Brazild Géosciences Environnement Toulouse, Institut de Recherche pour le Développement, 14 avenue Edouard Belin, 31400 Toulouse, Francee Instituto de Geociências, Universidade de São Paolo, Rua do Lago, 562, Cidade Universitaria, CEP 05508-080 São Paolo, Brazil

⁎ Corresponding author now at: German Research CenCentre Potsdam, 14473 Potsdam, Germany.

E-mail address: [email protected] (J. Bouche

0009-2541/$ – see front matter © 2012 Elsevier B.V. Allhttp://dx.doi.org/10.1016/j.chemgeo.2012.09.032

a b s t r a c t

Article history:Received 21 February 2012Received in revised form 18 September 2012Accepted 21 September 2012Available online 1 October 2012

Editor: C.M. Koretsky

Keywords:Amazon RiverFloodplainsChemical weatheringGeochemical fluxesGrain size

Large river sediments are mostly derived from tectonically active mountain belts, but then undergo a series ofsedimentation, temporary storage and reworking episodes on their journey to the ocean. The long transfertime of these sediments through active floodplains might result in a significant chemical maturation viaweathering reactions, which is of critical importance for biogeochemical cycles at the Earth surface. Thisstudy reports the chemical composition of river sediments from different locations throughout the coursesof the main tributaries of the Amazon Basin. Sampling along river depth-profiles yields access to the wholegrain size and chemical composition range of river sediment. Here, weathering intensities (i.e. losses of Na,K, Mg and Ca) associated with chemical weathering in floodplains are (1) examined as a function of grainsize, and (2) integrated over the whole grain size range, for three selected long river reaches flowing throughthe foreland and the lowland of the Amazon Basin: the upper Marañon, the Beni and the lower Madeirarivers. A relatively small Na loss through plagioclase dissolution is observed in the Madeira reach, while animportant Ca loss due to carbonate dissolution occurs in the two foreland reaches (Marañon and Beni). Nosignificant loss of K and Mg is observed in any of the reaches, showing the low alterability of primary Kand Mg-bearing minerals and suggesting retention of K and Mg in the particulate phase by secondary min-erals. The combination of these findings with previously reported data on the downstream change ofdissolved Na, K, Mg and Ca fluxes suggests that chemical weathering of “stable” alluvial deposits could alsosignificantly contribute to theweathering flux generated in foreland and lowland areas. The comparison be-tween the three Amazon reaches and the Gangetic plain tends to show that the features observed in theAmazon are valid on a global scale. Finally, we show that, although resulting in a relatively small change inthe chemical composition of the river sediment, silicate weathering in the lower Madeira floodplain canlead to a CO2 drawdown equivalent to ca. 10% of the total CO2 consumption flux of the whole Madeira basin.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

Weathering of the continental crust acts as a major sink of atmo-spheric CO2 (e.g. Garrels et al., 1976) and partitions chemical ele-ments between dissolved compounds and residual solids that areboth transported by rivers to the ocean. Chemical weathering rateshave been demonstrated to be positively correlated to physical ero-sion rates in most settings, from the scale of soils to that of rivers(e.g. Riebe et al., 2003; West et al., 2005), thereby showing that thesupply of primary minerals is generally the limiting factor for chemi-cal weathering. This supply is the highest in high-relief, tectonicallyactive areas, such as collisional orogens, subduction arcs, or volcanic

tre for Geosciences, Helmholtz

z).

rights reserved.

islands (Milliman and Syvitski, 1992; Raymo and Ruddiman, 1992;Dessert et al., 2003).

However, continental-scale rivers, draining large areas fromhigh-relief orogens to passive margins, flow through vast foreland ba-sins and lowland areas. In these relatively flat regions, river sedimentsare continuously deposited and re-involved into riverine transportthrough a variety of geomorphological processes, resulting in contin-uous exchanges of sediments between the channel and its floodplain(e.g. Meade et al., 1984; Allison et al., 1998; Dunne et al., 1998). Thetransient storage of solid particles in these alluvial plains results inpresumably long sediment transfer times (i.e. the average time neededfor a grain to be transported from the entry to the outlet of the riverreach). The transfer time of river sediments in large river floodplains re-mains largely unknown, even if U-series disequilibria provide a first setof constraints that range from a few kyr for suspended load in theMackenzie, Amazon and Ganges systems (Vigier et al., 2001; Dosseto

167J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

et al., 2006a, 2006b; Granet et al., 2010), to several 100s of kyr for coarsesediments in the Gangetic plain (Granet et al., 2007). These studies alsosuggest that the floodplain transfer time is longer than the residencetime of sediment in the soils of the actively eroding orogens.

Therefore, large river floodplains constitute settings whereweathering reactions might proceed due to the long-lasting contactbetween mineral phases on the one hand, and water and the atmo-sphere on the other hand. In these settings, weathering processesare likely to be influenced by yet unidentified controls, such as riverand floodplain dynamics. Besides a handful of studies (e.g. Galy andFrance-Lanord, 1999; West et al., 2002; Dosseto et al., 2006a), onlylittle information is available to date on the potential contributionof weathering reactions during riverine transport and transientstorage to global weathering fluxes. However, recently, Lupker et al.(2012) showed that the principal locus of silicate weathering in theHimalaya–Ganga system is actually the Gangetic plain. This findinghas broad implications in terms of global weathering and deservesto be investigated on other large systems. Furthermore, Moquet etal. (2011) showed, using an extensive database of dissolved fluxesin the Amazon, that the relatively low-relief Andean foreland con-tributes significantly to the total Amazon silicate weathering flux.However, it is still unclear whether this flux is associated withweathering of sediments transported through the active floodplainor with chemical weathering affecting older, stable sedimentary for-mations (i.e. not part of the active floodplain) of the Andean foreland.A study of the downstream chemical change of Andean-derived sed-iments should help to disentangle these two processes.

Previous studies on weathering in the Amazon lowland showed adownstream mineralogical change of river sands (Franzinelli andPotter, 1983) in the Amazon mainstem. However, the authors suggestedthat most of the observed trend was due to the addition of plain-derivedriver sands, and not toweathering reactions. Johnsson andMeade (1990)showed that river sediments deposited on a point-bar on the SolimõesRiver were undergoing increasing mineralogical maturation with in-creasing deposit age.Martinelli et al. (1993) demonstrated, using a chem-ical mass-balance, that weathering reactions were responsible fordownstream chemical changes in sediments from the Solimões varzea(floodplain) through loss of plagioclase, smectite and illite, and formationof kaolinite. Similarly, a downstream shift in the mineral assemblagedominating the clay fraction of Amazon suspended sediments, fromillite–chlorite to kaolinite–smectite, was also reported by Guyot et al.(2007b). This observation was attributed to the introduction of maturebank material during river lateral migration. Nevertheless, while thesestudies provide indisputable evidences that weathering reactionsproceed to some extent in the Amazon plains, they focused on either(1) particular size fractions of river sediments or (2) floodplain deposits,of which the introduction rate to the river channel is still unconstrainedin most regions of the Amazon plain. Therefore, no quantitative estimateof the overall effects of these weathering reactions in terms of river sed-iment chemistry or in terms of geochemical fluxes is available to date.

In this paper, we address these questions using a set of river sedimentsamples from the main tributaries of the Amazon River, the world'slargest river in terms of drainage basin area andwater discharge. Our ap-proach consists in comparing the abundance of major chemical elementsthat are soluble during weathering (Na, K, Mg, Ca) in river sediments atvarious locations over the course of these rivers, and in calculating theweathering intensities associated with weathering during transport onthree long river reaches. The originality of our approach lies in theuse of river suspended sediment depth-profiles, which makes possiblethe covering of the whole grain size spectrum of river sediments. Theresulting weathering fluxes are compared with (1) those calculatedfrom amass-balance on the dissolved load, and (2) between the differentreaches and with results from the Gangetic plain (Lupker et al., 2012).These comparisons allow us to discuss the loci of weathering in theAmazon plain domain and the controls of weathering in large river flood-plains in general.

2. The Amazon system: setting, sampling, and methods

2.1. The Amazon Basin

The Amazon River annually discharges 6500 km3 of water(Callède et al., 2010) and between 500 and 1200×106 T of sediments(Meade, 1994; Dunne et al., 1998; Filizola and Guyot, 2009) to theAtlantic Ocean, thereby ranking first and third, respectively, amongthe world's rivers (Meybeck and Ragu, 1996). Its drainage area, theworld's largest, covers 6.4×106 km2, which can be divided into thefollowing main geomorphic units:

– The actively eroding eastern Cordillera of the Andes to the west,which altitudes range from 400 m to more than 6000 m, consistsof (1) a variety of Ordovician to Cretaceous meta-sedimentaryrocks (shales, sandstones, carbonates) and (2) relatively young,subduction-derived acid to intermediate igneous rocks. There, de-nudation rates are between 0.2 and 0.5 mm/yr (Wittmann et al.,2011a);

– The Guiana and Brazilian shields consist of relatively high eleva-tion (up to 3000 m), tectonically quiescent units of Precambrianmetamorphic granulite-facies rocks, with low erosion and lowweathering rates (Gaillardet et al., 1997; Wittmann et al., 2011a);

– The Amazon foreland bordering the Andes has elevations rangingfrom 400 m down to 120 m, and is formed of detrital sedimentsmostly of Cenozoic age, currently involved in folding-thrusting re-lated to Andean orogeny (Roddaz et al., 2005). There, rivers losesediment to the surrounding floodplains mostly via build-up ofchannel levees, point-bars and overbank deposits, while flood-plain sediment returns to the channel through bank erosion andmeander cut-off (Mertes et al., 1996; Gautier et al., 2006). On adecadal scale, as shown by sediment gauging, some parts of theAmazon foreland are eroding (e.g. the Napo basin, Guyot et al.,2007a), and other areas are currently accumulating sediments(e.g. the Beni basin, Guyot et al., 1996). This feature might betrue only for recent times (Wittmann et al., 2011a). The Andeanforeland generates approximately a third of the weathering fluxof the Amazon Basin (Moquet et al., 2011).

– The lowlands, representing areas with altitudes lower than 120 m,consist of a several kilometer-thick pile of detrital sediments(Dumont and Fournier, 1994). The main tributaries of the AmazonBasin flow in the lowlands at a very low gradient and exchangewater and sediments with the floodplain through a complex net-work of channels, lakes and periodically inundated areas (Bonnetet al., 2005). During high flow, the inundated area is estimatedto be around 300,000 km2 (Maurice-Bourgoin et al., 2007). Riversdraining exclusively the lowlands have low dissolved and particu-late loads, and high organic matter content (Gaillardet et al.,1997). The lowlands reaches studied by Moquet et al. (2011)(Purus, Orthon, Itenez) account for 5 to 10% of the weatheringflux of the Amazon.

This study focuses on the two main tributaries of the Amazon sys-tem in terms of sediment supply, the Solimões and the Madeira rivers,the headwaters of which drain the Andes. These two tributaries ac-count for 90% of the sedimentary budget of the Amazon River atmouth (Gibbs, 1967). Today, the Peruvian and Ecuadorian Andes sup-ply 450.106 T/yr of sediment to the Amazon plain through theSolimões Andean tributaries (Guyot et al., 2007a), whereas the Boliv-ian Andes deliver 400.106 T/yr, of which a significant part is trappedin the Madeira foreland basin (Guyot et al., 1996). Longer-term, cos-mogenic nuclide-derived sediment loads are, at first-order, comparableto these numbers (Wittmann et al., 2011a). U-series disequilibria showthat the sediment residence time is 5 kyr in the whole Solimões Basin(including the Andes), while the transfer time through the Madeiralowland plain is 14 kyr (Dosseto et al., 2006a, 2006b).

500 km

8000

6000

4000

2000

0

Elevation (m)

-1

-1

20

5

0

-5

0

5

275 280 285 290 295 300 305 310-

315

(A)

Mara

Mara

Beni

Beni

on @ Borja

on @ San Regis

Ucayali Amazonas

MouthHuallaga

Pastaza

Morona

Solim es system

Beni-Madeira system

@ Rurrenabaque

@ Riberalta

Madre de Dios

Mamor

Mouth

Basin limit Andes / plain transition

BeniM

. de

Dio

sM

adei

ra

Amazon

D) Madeirareach

Mamor

up

down

Mara on

Pastaza

Ucaya

li

Morona

Hualla

ga

B) Mara on reach

Solim es

Madeir

a

Amazon

Andes

up

down

Ben

i

M. de

Dios

C) Benireach

Rurrena-baque

Ribe-ralta

up

down

Fig. 1. The Amazon basin, sampling locations and studied river reaches. (A) Map of the Amazon Basin and location of the sampling sites and of the three reaches where a chemicalbudget carried out in this study (black rectangles). (B) Marañon reach. (C) Beni reach. (D) Madeira reach. In panels (A) to (C), the thick white lines labeled “up” and “down” locatethe upstream and downstream boundaries of the river reach, respectively.

168 J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

2.2. Sampling sites and river reaches

The sampling sites are shown in Fig. 1. The Solimões and Madeirarivers were sampled at their mouth in June 2005 and in March 2006,respectively corresponding to the Solimões high-water stage andMadeira low-water stage, and to the Solimões rising water stageand Madeira high-water stage (Bouchez et al., 2011a, 2011b). Twoof the three main Andean tributaries of the Solimões River in termsof sediment supply, i.e. the Marañon River and the Ucayali Riverwere sampled in April 2008 upstream from their confluence, respec-tively near the towns of San Regis and Jenaro Herrera. These rivers

Notes to Table 1:a Data from the HyBAm program (http://ore-hybam.org); the values for the Beni and the M

de Dios at Miraflores, respectively.b Values from Guyot et al. (1996).c Values from Guyot et al. (2007a); the values for the Huallaga at mouth and the Ucaya

Requeña, respectively.d The sediment fluxes of the Morona and the Pastaza are not available. To estimate the to

state for the Marañon reach between Borja and San Regis: the total input from tributaries (Regis). This yields a combined contribution for the Morona and Pastaza rivers of 23 MT/yreach). Although this calculation is associated with a relatively high uncertainty, these fluxethe overall mass-balance to a great extent.

were also sampled downstream from their confluence (Rio Amazonasnear Iquitos). The Marañon River was additionally sampled where itenters the Amazon foreland, near the town of Borja. The HuallagaRiver, a right-bankAndean tributary of theMarañonRiver, was sampledat its mouth, as well as the Morona and the Pastaza, two left-banktributaries. While the Huallaga River drains the Peruvian Andes andhas high sediment concentrations (Guyot et al., 2007a), the Moronaand Pastaza rivers, bothflowing through theNorth Amazonian ForelandBasin (Roddaz et al., 2005) and draining the actively eroding PastazaMegafan (Bernal et al., 2011), contribute only slightly to the sedimenta-ry budget of the Marañon River (Guyot et al., 2007a). The three

adre de Dios at Riberalta were taken as those of the Beni at Portachuelo and the Madre

li at Jenaro Herrera were taken as those of the Huallaga at Chazuta and the Ucayali at

tal contribution of these two rivers, the sediment budget was assumed to be at steady-Marañon at Borja, Huallaga, Pastaza, Morona) then equals the output (Marañon at San, that we assumed to be equally distributed between the two tributaries (11.5 MT/yrs are relatively small compared to the fluxes from other tributaries, and do not affect

Table 1Sample list with sample locations for the Bolivian and Peruvian sampling sites. The Brazilian sampling sites and samples are described in Bouchez et al. (2010) and Bouchez et al. (2011b). N.D.: not determined.

Sample Stream Location Annualdischargea

(m3/s)

Sedimentflux(MT/yr)

Date River width(m)

Sampling depth(m)

Longitude(W)

Latitude(S)

Distance fromthe left bank(m)

ADCP discharge(m3/s)

SPM conc.(mg/L)

Size distribution

Mode(μm)

D50

(μm)D10

(μm)D90

(μm)

AM-07-01 Beni Rurrenabaque 2153 212b 5/5/2007 340 4.5 67°32′05″ 14°28′27″ 170 1490 666 161 86 5 213AM-07-02 5/5/2007 340 3 67°32′05″ 14°28′27″ 170 236 7 11 2 55AM-07-03 5/5/2007 340 1.5 67°32′05″ 14°28′27″ 170 222 7 11 2 52AM-07-04 5/5/2007 340 0 67°32′05″ 14°28′27″ 170 109 7 9 2 44AM-07-05 5/5/2007 340 Bank N.D. N.D. 203 176 48 328AM-07-06 Beni Riberalta 3722 122b 5/7/2007 520 7 66°05′40″ 11°00′08″ 420 2530 2551 70 28 4 109AM-07-07 5/7/2007 520 5 66°05′40″ 11°00′08″ 420 3373 71 37 4 107AM-07-08 5/7/2007 520 2.5 66°05′40″ 11°00′08″ 420 2212 52 24 3 79AM-07-09 5/7/2007 520 0 66°05′40″ 11°00′08″ 420 1003 24 12 2 50AM-07-10 5/7/2007 520 Bank N.D. N.D. 110 106 51 175AM-07-11 Madre de Dios Riberalta 5602 71b 5/8/2007 660 7 66°05′54.5″ 10°57′45.8″ 460 4720 2135 83 35 4 122AM-07-12 5/8/2007 660 5 66°05′54″ 10°57′46″ 460 679 16 11 3 38AM-07-13 5/8/2007 660 2.5 66°05′54″ 10°57′46″ 460 592 15 13 3 62AM-07-14 5/8/2007 660 0 66°05′54″ 10°57′46″ 460 437 15 10 2 35AM-07-15 5/8/2007 660 Bank N.D. N.D. 257 259 155 380AM-07-16 Mamoré Guayaramerin 7916 66b 5/9/2007 880 13 65°18′43″ 10°50′16″ 440 15,460 631 128 83 10 228AM-07-17 5/9/2007 880 8 65°18′43″ 10°50′16″ 440 717 153 99 8 224AM-07-18 5/9/2007 880 4 65°18′43″ 10°50′16″ 440 371 97 49 6 148AM-07-19 5/9/2007 880 0 65°18′43″ 10°50′16″ 440 156 29 23 4 82AM-08-01 Amazonas Tamshiyacu 30,148 413c 4/23/2008 990 27 73°09′56″ 04°00′08″ 530 44,630 2529 N.D. 324 36 532AM-08-02 4/23/2008 990 21 73°09′50″ 04°00′15″ 720 925 N.D. 45 4 241AM-08-03 4/23/2008 990 14 73°09′52″ 04°00′16″ 750 457 N.D. 18 3 141AM-08-04 4/23/2008 990 7 73°09′48″ 04°00′16″ N.D. 497 N.D. 23 3 166AM-08-05 4/23/2008 990 0 73°09′48″ 04°00′16″ N.D. 344 N.D. 13 2 64AM-08-06 4/24/2008 990 Bed 73°10′09″ 04°30′09″ N.D. N.D. 462 331 618AM-08-07 4/25/2008 990 Triple (bottom) N.D. N.D. 690 1434 N.D. N.D. N.D. N.D.AM-08-08 4/25/2008 990 Triple (interm.) N.D. N.D. 690 922 N.D. N.D. N.D. N.D.AM-08-09 4/25/2008 990 Triple (top) N.D. N.D. 690 699 N.D. N.D. N.D. N.D.AM-08-10 Ucayali Jenaro Herrera 12,090 205c 4/25/2008 700 23 N.D. N.D. 270 19,550 641 N.D. 13 2 124AM-08-11 4/25/2008 700 16 N.D. N.D. N.D. N.D. N.D. N.D. N.D.AM-08-12 4/25/2008 700 7 73°40′28″ 04°54′28″ 350 599 N.D. 12 2 78AM-08-13 4/25/2008 700 0 N.D. N.D. 350 490 N.D. 10 2 50AM-08-15 4/26/2008 700 Bed N.D. N.D. N.D. N.D. 10 2 43AM-08-16 4/26/2008 700 Triple (bottom) N.D. N.D. 150 2738 N.D. 163 12 266AM-08-17 4/26/2008 700 Triple (interm.) N.D. N.D. 150 2065 N.D. 155 6 280AM-08-18 4/26/2008 700 Triple (top) N.D. N.D. 150 133 N.D. 55 3 265AM-08-19 Maranon San Regis 16,885 168c 4/27/2008 710 Bed N.D. N.D. N.D. 23,720 N.D. 407 252 583AM-08-20 4/27/2008 710 20 73°54′24″ 04°31′24″ 540 831 N.D. 125 7 273AM-08-21 4/27/2008 710 14 73°54′29″ 04°31′29″ 300 332 N.D. 22 3 231AM-08-22 4/27/2008 710 10 73°54′31″ 04°30′31″ 340 517 N.D. 40 4 247AM-08-23 4/27/2008 710 5 73°54′28″ 04°31′28″ 320 466 N.D. 29 4 210AM-08-24 4/27/2008 710 0 73°54′28″ 04°31′28″ 320 177 N.D. 16 3 50AM-08-25 4/28/2008 710 Triple (bottom) N.D. N.D. 570 1245 N.D. 78 6 189AM-08-26 4/28/2008 710 Triple (interm.) N.D. N.D. 570 1098 N.D. 79 7 193AM-08-27 4/28/2008 710 Triple (top) N.D. N.D. 570 1277 N.D. 96 9 208AM-08-29 Maranon Borja 4975 103c 5/3/2008 240 Bed N.D. N.D. N.D. 5910 N.D. 148 56 284AM-08-30 5/3/2008 240 14 73°32′44″ 04°28′44″ 120 1717 N.D. 95 4 352AM-08-31 5/3/2008 240 8 73°32′47″ 04°28′47″ 160 1251 N.D. 16 2 186AM-08-32 5/3/2008 240 4 73°32′45″ 04°28′45″ 110 1324 N.D. 34 3 259AM-08-33 5/3/2008 240 0 73°32′45″ 04°28′45″ 110 845 N.D. 11 2 75AM-08-34 Morona Mouth N.D. 11.5d 5/3/2008 210 0 N.D. N.D. 105 1650 170 N.D. 12 2 39AM-08-35 5/3/2008 210 Bed N.D. N.D. N.D. N.D. 413 280 598AM-08-36 Pastaza Mouth 704 11.5d 5/4/2008 510 0 N.D. N.D. 250 4150 100 N.D. 21 4 61AM-08-37 5/4/2008 510 Bed N.D. N.D. N.D. N.D. N.D. N.D. N.D.AM-08-38 Huallaga Mouth 3042 42c 5/4/2008 390 0 N.D. N.D. 190 2450 330 N.D. 11 2 40AM-08-39 5/4/2008 390 Bed N.D. N.D. N.D. N.D. 260 180 350

169J.Bouchez

etal./

ChemicalG

eology332

–333(2012)

166–184

170 J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

tributaries delivering sediments to the Madeira Basin (Guyot et al.,1996), i.e. the Beni, Madre de Dios and Mamoré rivers, were sampledin May 2007 upstream from their confluences (near the towns ofRiberalta and Guayaramerin). The Beni River was also sampled nearthe town of Rurrenabaque, at the outlet of the Andes.

The estimation of the change of river sediment chemistry and ofthe associated weathering fluxes can be made only in river reachesfor which the sediment flux and chemical composition is known(1) for each tributary and (2) at the outlet (see Section 4). With thepresent set of samples, three river reaches of the Amazon Basin fulfillthis requirement (Fig. 1):

– the Marañon River between Borja and San Regis. The chemicalcomposition of the lateral sediment contributions can be assessedfrom our sample set (Huallaga, Pastaza, Morona; Guyot et al.,2007a). This 600 km-long reach located in the North AmazonianForeland Basin meanders over the first 100 km, and mostly dis-plays an anabranching pattern in the downstream portion (Fig. 1).

– the Beni River between Rurrenabaque and Riberalta, for whichthere is no significant lateral sediment contribution from tribu-taries (Guyot et al., 1996). The reach is situated in the SouthAmazonian Foreland Basin. The valley length is about 450 km, andthe channel length is around 800 km, making this reach highlysinuous. The channel gradient is between 10−3 and 10−4, and theriver is actively meandering, producing a large number of oxbowlakes, in the first 200 km downstream of Rurrenabaque (Fig. 1).Channel sinuosity decreases after the junction with the Madidi(Gautier et al., 2007).

– the Madeira River between the Beni–Mamoré confluence and itsmouth. At the former location, the sediment input can be estimatedfrom the weighted average of the three Bolivian tributaries, sincethe Madre de Dios–Beni confluence is only ca. 150 km upstreamfrom the Mamoré–Beni confluence. Downstream from this conflu-ence, there is no significant sediment input from tributaries(Guyot et al., 1996). Most of this 1300 km-long reach consists ofrelatively straight river portions with gentle bends and scatteredvegetated islands (Fig. 1).

2.3. Sampling and analytical methods, and origin of hydrological data

At each site, river water was sampled following depth-profiles(one to four depth-profiles across the river for the Solimões andMadeirasampling sites, one single depth-profile at the middle of the channel forthe Peruvian and Bolivian sites). Each of the depth-profiles comprisedtwo tofive samples (Bouchez et al., 2011a). Samplingprocedures and lab-oratory analysis are explained in more details in Bouchez et al. (2011b).Briefly, river water was filtered at 0.22 μm-porosity, and sedimentswere recovered and then dried at 50 °C in the lab. Bed sediments weredredged when possible. For Bolivian tributaries, bank sediments weregrabbed and considered as being representative of the sediment coarsefraction. For some of the Peruvian tributaries, a triple-sampler was usedto sample the river water at 10, 40 and 70 cm above the bed of the chan-nel, allowing access to deep, coarse suspended sediments. The Huallaga,Morona, and Pastaza rivers were only sampled at the channel surfaceand bed. Chemical composition were obtained by ICP-OES and ICP-MSafter alkali fusion at the Service d'Analyse des Roches et Minéraux(SARM, Vandoeuvre-les-Nancy, France). Results for the Peruvian andBolivian tributaries are listed in Table 1; those for the Solimões andMadeira rivers at mouth are reported in Bouchez et al. (2011b).

The most adequate time scale over which of sediment fluxesshould be measured to be used in a floodplain weathering budget isthat of the sediment transfer time (i.e. millennial scale). Cosmogenicnuclides-derived fluxes reflect similar time scales, but such data isstill unavailable for many rivers considered in this study (Wittmannet al., 2011a). Hence, decadal estimates are used here. These numberswere obtained from multi-year records of sediment fluxes measured

by sediment gauging during the HyBAm program (http://www.ore-hybam.org, e.g. Filizola and Guyot, 2003; Guyot et al., 2007a;Laraque et al., 2009; Filizola and Guyot, 2009), and are summarizedin Table 1. The water discharge data used in this study is also fromdecadal estimates from the HyBAm program (as summarized byMoquet et al., 2011).

2.4. The hypothesis of a sedimentary steady-state for the Amazonianfloodplains

Different river sediment size fractions have different chemicalcomposition (Garzanti et al., 2010; Bouchez et al., 2011b; Lupkeret al., 2011). Therefore, during river transport, the overall chemicalcomposition of the sediment might be affected not only by weatheringreactions and bymixingwith sediments from compositionally differenttributaries, but also by preferential deposition or re-suspension of agiven size fraction. However, Guyot et al., (1999) reported the absenceof any evolution of suspended sediment grain size in the Madeira fore-land basin, downstream from the Andes-foreland transition. Further-more, Mertes and Meade, (1985) reported only little variation in bedsand grain size on the 2000 km-long reach on the Amazonas–Solimõesbetween Vargem Grande and Óbidos. Following what is suggested bythese studies, we consider in the following that, in the Amazon Basin,the average sediment grain size does not evolve significantly in agiven river channel once the sediments leave the Andes, neither bypreferential deposition or re-suspension of a certain grain size, nor bygrain comminution. This hypothesis implies that, should primary min-erals undergo weathering and subsequent fining (for example throughthe dissolution of coarse feldspars and the formation offine clays), theseminerals are relatively scarce in a given river sediment sample, and donot exert a first-order control on this sample's grain size parameters.This is in agreement with the observation that the most abundant min-eral in Amazon sediments is quartz, which is unlikely to dissolve signif-icantly in floodplains because of its relatively slow dissolution kinetics(Section 4.1). Therefore, potential weathering reactions in the Amazonfloodplains will likely not result in a downstream change of sedimentgrain size.

Moreover, we consider that on the millennial time scale (that of thesediment transfer time), the floodplains of the three reaches are not ac-cumulating nor exporting sediments. Even though an imbalance be-tween incoming and outgoing sediment fluxes was reported for someparts of the foreland floodplain (e.g. in the Beni reach, Guyot et al.,1996), or in the Central Amazon floodplain (Dunne et al., 1998), itshould be emphasized that this sediment gauging data reflect, by na-ture, decadal trends. Conversely, a good agreement was observed inthe Central Amazon between (1) decadal-time scale sediment fluxesfrom sediment gauging and (2) millennial-time scale, cosmogenicnuclides-derived sediment fluxes. The discrepancy between the twotime scales was the highest at the outlet of the Andes (Wittmann etal., 2011a). These observations suggest that the significantly higher sed-iment fluxes from the Andes are a relatively recent feature that has notbeen fully propagated downstream to the lowerAmazon reaches.Whilewe acknowledge that an imbalance might exist, the budget carried outin the present studywould be seriously affected only if a particular grainsizewas significantly lost or gained by the channel (see Section 4.1). Theimportant parameter to translate floodplain weathering intensities intofluxes is the magnitude of the sediment mass transferred throughthe Amazon floodplain, which is relatively well-known (e.g. a few100 MT/yr, see Table 1 and Wittmann et al., 2011a), independently ofthe estimates on sediment accumulation or loss by the floodplain.

It should also be emphasized that this hypothesis is not contradic-tory with the findings of Gaillardet et al. (1999a) and Bouchez et al.(2011c), who showed, on the basis of geochemical steady-state bud-gets, that a large fraction of the sediments exported at the mouth ofthe Solimões, Madeira and Amazon rivers are likely derived from aweathered component of the continental crust. A potential

0.0 0.1 0.2 0.3 0.4 0.5Al/Si

10

100

1000

D90

(µm

)

Solim es mouth

Solim es tributaries

Madeira mouth

Madeira tributaries

Fig. 2. Grain size distribution D90 (grain size under which 90% of the sediment volumeis found) vs. the Al/Si ratio in Amazon Basin river sediments. Bouchez et al. (2011b) ar-gued that a single relationship could describe sediments from the whole AmazonBasin, based on the sole Solimões and Madeira rivers at their mouth. When datafrom upstream tributaries is added, significantly different relationships can be distin-guished between the Solimões and the Madeira basins. The coarse end-member ofthe Madeira River appears to be quartz-richer than that of the Solimões River. In thisdiagram, the dataset of the Solimões River and Peruvian tributaries is satisfactorilydescribed by the linear best fit: logD90 ¼ −3:58� Al=Sið Þ þ 3:06 (R2=0.87), whilethe dataset of the Madeira River and its Bolivian tributaries is best-fitted by logD90 ¼−2:65� Al=Sið Þ þ 2:57 (R2=0.85).

171J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

explanation for this weathered component is the supply of sedimentfrom the lowland plain, meaning that the plain is affected by sedi-mentary imbalance. However, the inferred weathered component iswell accounted for by the sedimentary formations that underlaymost of the Solimões and the totality of the Madeira Andean headwa-ter areas.

To summarize, we consider the three studied reaches as undergoingneither gain nor loss of sediment, for any grain size fraction. This as-sumption means that preferential deposition of fine particles throughoverbank flows and of coarse particles on meander point-bars arebalanced by their incorporation into the river during channel lateralmigration (Lauer and Parker, 2008), or at least that the net floodplainaccumulation or loss is small compared to the flux of transiting sedi-ment. This quasi-steady-state is more likely to be achieved for riverreaches that are sufficiently long and actively reworking their flood-plain, which is supposedly the case for the three reaches of the presentstudy. Noticeably, Mertes et al., (1996) reported that the BrazilianAmazon can recycle its entire floodplain in less than 5 kyr, with higherrecycling rates in upstream reaches than in downstream reaches.

2.5. Chemical proxies for grain size and weathering

In general, a weathering intensity, quantified by a so-calledweathering index, describes the elemental or totalmass loss that resultsfromweathering reactions having affected a soil or sediment sample, ascompared to its parent material, such as bedrock. Weathering intensi-ties and associated weathering fluxes can be inferred from chemical el-ements that are soluble during weathering reactions. For the presentstudy, the concentration of these soluble elements in river sedimentsare compared between different locations over the course of a givenriver reach, so as to estimate changes in weathering intensity andweathering fluxes associated with floodplains. However, the chemicaland grain size variability of river sediments have to be considered toproperly estimate a weathering budget in the floodplains. Samplingriver sediments along depth-profiles grants access to the different sed-iment size fractions by taking advantage of the effect of hydrodynamicsorting within channels (Galy et al., 2008; Bouchez et al., 2011a,2011b; Lupker et al., 2011). Fine sediments are enriched near thechannel surface while coarse sediments are enriched near the channelbottom and in bed sediments. Sediments from intermediate depthsare mixtures between these two end-members. In the Amazon River,the Al/Si ratio of river sediment decreases from channel surface tobottom, and is tightly linked to the sediment grain size distributionand chemical composition (Fig. 2). This relationship results from themixing between (1) a coarse, Si-rich end-member, best representedby the bed sediment that is hydrodynamically enriched in quartz, and(2) a fine, Al-rich end-member, best represented by channel surfacesediments, hydrodynamically enriched in clay minerals. Fig. 2 showsthat Al/Si is a chemical proxy for grain size distribution against whichthe chemical composition of river sediments can be examined and com-pared between different locations over the river course. However, themajor process affecting chemical concentrations in river sedimentdepth-profiles is dilution by quartz, the effect of which increases atdepth in the channel. Alternatively, concentration ratios between asoluble and an insoluble element will be sensitive to weathering, butinsensitive to dilution by quartz.

Sodium (Na), K, Mg and Ca are soluble elements that arepartitioned between the dissolved phase and the solid residues dur-ing silicate weathering (Dupré et al., 1996; Gaillardet et al., 1999b).In this paper, we normalize Na, K, Mg and Ca concentrations to thatof Al to obtain a weathering index that is not sensitive to dilutionby quartz. Al is a major insoluble element (Dupré et al., 1996), andit is not preferentially carried by any accessory mineral in AmazonRiver sediments (Bouchez et al., 2011b), making it fairly insensitiveto normalization artifacts. The conceptual version of a Na/Al–, K/Al–,Mg/Al– or Ca/Al–Al/Si diagram is shown in Fig. 3. Anymixture between

a coarse, Si-rich end-member and a fine, Al-rich end-member, such as aseries of samples from a given depth-profile, will plot on a mixing hy-perbola in Fig. 3. Mineral dissolution during weathering of sedimentwill result in a downward offset between upstream and downstreamdata (loss of Na, K, Mg or Ca compared to Al, which is immobile). Inthe following, chemical data from all our sampling sites is first qualita-tively examined in the framework of Fig. 3, then weathering fluxes arecomputed using a mass balance of sedimentary fluxes through thethree studied reaches.

3. Results

3.1. Relationships for each sampling site

Weathering indexes (X/Al ratios) are shown for the whole set ofsediment samples as a function of the Al/Si ratio, in Figs. 4 and 5. Inall rivers, Na/Al and K/Al decrease with increasing Al/Si (i.e. decreas-ing grain size), similar to what was reported for Na/Th and K/Th byBouchez et al. (2011b), who used so-called fan-shaped diagrams toexamine these ratios. The only exception is K/Al in low-Al/Si samplesof the Amazon River. Mg/Al and Ca/Al generally decrease with in-creasing Al/Si as well, although Mg/Al increases with Al/Si in theUcayali, Mamoré, and Madre de Dios, and Ca/Al increases with Al/Siin the Ucayali and Mamoré rivers. Even in the cases where all X/Alratios decrease with increasing Al/Si, lower depletion in fineend-members, as compared to coarse end-members, are observedfor K and Mg than for Ca and Na. As noticed by Bouchez et al.(2011b), K and Mg are known to be incorporated into secondary min-erals (respectively mainly into illite and smectite), and are therebyenriched in high Al/Si samples as compared to Ca and Na.

3.2. Comparison between sampling sites

Values of Na/Al, K/Al, Mg/Al and Ca/Al are higher and more variablein the Solimões Basin than in the Madeira Basin, for the whole range ofgrain size, as observed by Bouchez et al. (2011b) for the mouth of thesame systems. At first order, the two large rivers at mouth plot well

X/Al

Al/Sisortingm

ixing

Mineral dissolution

Si lossNo

Si loss

Coarseend-member

Fineend-member

finercoarser

Upstreamlocation

(or mixture between

tributaries)

Downstreamlocation

Fig. 3. Conceptual diagram of an Al-normalized soluble element concentration (X/Al, used as a weathering index) vs. Al/Si (used as a chemical proxy for grain size). The brown starrepresents the fine end-member of the sediment carried by a given river (best represented by sediments sampled at the channel surface) and the yellow star is the coarseend-member (best represented by bed, bank sediment or suspended sediments sampled near the channel bottom). These end-members are physically separated in riverdepth-profiles because of hydrodynamic sorting at play in the river channel. Then, all intermediate samples from the river depth-profile lie on the hyperbolic “sorting/mixing”trend between these end-members. The overall input to the reach by several tributaries can be predicted by “mixing” their contributions for each grain size. To do so, the relativecontributions have to be known for each tributaries, as well as the X/Al–Al/Si relationships they define (Section 4.1). Weathering induces a change of the X/Al ratio along the courseof a river (between the “upstream location” and the “downstream location”); for example, mineral dissolution makes the X/Al ratios decrease. While evaporite and carbonatedissolution result in a strict downward trend in this diagram, silicate weathering might be associated with loss of Si. We show in Section 4.1 that this Si loss can be ignored inthe Amazon lowlands. Conversely, incorporation of soluble cations into secondary minerals leads to an upward shift in this diagram (not shown in this example). Finally, chemicalweathering might affect different grain size fractions to different extents.

172 J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

within the frame defined by their respective upstream tributaries. Thisis especially true for Na/Al, for which the trends from different riversare very similar. However, the Marañon at Borja is slightly Na-poorerthan the rest of the rivers of the Solimões Basin (Fig. 4), and the Beniat Riberalta is Na-richer than at Rurrenabaque (Fig. 5). The highest K/Al of the Solimões system are obtained for the Ucayali and Huallagarivers, while the lowest are obtained in the Pastaza and the Moronarivers. The Marañon is slightly K-richer at Borja than at San Regis. LowAl/Si samples are slightly K-richer in the Beni at Rurrenabaque than atRiberalta. The Marañon at San Regis displays higher Mg/Al than atBorja. The Beni at Rurrenabaque exhibits higherMg/Al than at Riberalta,and compared to the Madre de Dios. In the Marañon, Ca/Al is signifi-cantly higher at Borja than at San Regis. For the Beni, Ca/Al is higher atRurrenabaque than at Riberalta, and compared to the Mamoré. TheMadre de Dios is significantly Ca-richer than the other rivers of theMa-deira Basin.

To summarize, the differences between the Solimões and theMadeira rivers at mouth observed by Bouchez et al. (2011b) arealso observed for the upstream tributaries. In each of the two largebasins, for a given grain size, Na/Al is fairly constant across the sam-pled rivers, while K/Al, Ca/Al and Mg/Al show significantly differentratios from a tributary to another. Qualitatively speaking, no obviouschange of soluble element content is observed between upstreamand downstream locations, except for Ca in the Marañon River. Thisis in contrast to what was reported recently for the Ganges River byLupker et al.(2012). However, it should be borne in mind that in the

case where several tributaries are contributing to the reach, theweathering signal might be difficult to extract from the qualitativeexamination of Figs. 4 and 5. It appears that a closer quantitativelook is needed to unravel potential chemical changes of river sedi-ments during their transport in the Amazon Basin.

4. Discussion

4.1. Grain size-specific floodplain weathering intensities

In order to retrieve the weathering signal (i.e. loss or gain of solubleelements) from the comparison of sediment chemical compositionbetween two points of the river course, a steady-state mass balanceis established between sediment inputs to and outputs from a riverreach, for a given sediment size fraction and for a given elementX (Fig. 6):

∑I

_F IX− _FD

X þ _F EX ¼ _FM

X ; ð1Þ

where _FX is the particulate flux of the element X carried by theconsidered grain size fraction. In the following, grain size-specificparameters are overlain by a dot. Superscripts I, D, E and M, respec-tively stand for tributaries input, sediment deposition from thechannel to the floodplain, floodplain sediment incorporation tothe channel, and sediment export at the mouth of the river reach(in MT/yr). We consider here that there is no imbalance between

0 .0 0 .1 0 .2 0 .3 0 .4 0 .5

Al/Si

0 .0

0 .2

0 .4

0 .6

0 .8

Ca

/Al

0 .0 0 .1 0 .2 0 .3 0 .4 0 .5

Al/Si

0 .0 5

0 .1 5

0 .2 5

0 .3 5

Mg

/Al

0 .0 0 .1 0 .2 0 .3 0 .4 0 .5

Al/Si

K/A

l

0 .0 0 .1 0 .2 0 .3 0 .4 0 .5

Al/Si

0.0

0.1

0.2

0.3

0.4

Na

/Al

0 .1 5

0 .2 5

0 .3 5

0 .4 5

Mara on@ Borja

Morona Pastaza

Huallaga

Napo

Ucayali

Solim esAmazonas

Mara on@ SanRegis

Mara onreach

Andes Foreland Lowland

Fig. 4. Na/Al, K/Al, Ca/Al, and Mg/Al vs. Al/Si in the Solimões basin. The symbols description also features a simplified hydrographic network of the Solimões Basin, where bar widthis proportional to the sediment flux contribution.

173J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

deposition and incorporation of sediments into the channel(Section 2.4). However, while sediments are stored in the floodplain,soluble elements are lost (through mineral dissolution) or gained (e.g.through clay formation) by the sediment:

_F EX− _FD

X ¼ _WX : ð2Þ

The elemental weathering flux _WX results from a change in thechemical composition of sediments through chemical weathering ofthe considered sediment size fraction during storage (in MT/yr). Anegative value of _WX corresponds to a loss of the soluble element Xfrom the particles, which can result from mineral dissolution. _WX

can also be positive, in which case X is being incorporated into thesediment during storage in the floodplain. In other words, _WX issimply the net loss of X for the considered sediment grain size.

Combining Eqs. (1) and (2) and introducing the X/Al ratio (specificto each grain size fraction):

_WX ¼ _FM⋅ _Alh iM⋅ _X

Al

!M

−∑I

_F I⋅ _Alh iI⋅ _X

Al

!I

; ð3Þ

where _F and _Alh i

are respectively the sediment flux and the Al partic-

ulate concentration for the considered grain size (respectively inMT/yrand in g/g). The interest of Eq. (3) is that it features the grain size-specificweathering index X/Al ratio _X

Al

� �, which can be read in Fig. 3. Fur-

thermore, since Al is not soluble during weathering, and in the absenceof deposition and erosion, the grain size-specific particulate Al flux doesnot change during transfer of sediment through the floodplain:

_FM⋅ _Alh iM ¼ ∑

I

_F I⋅ _Alh iI ¼ _F ⋅ _Al

h i; ð4Þ

Al/Si

0 .20

0 .25

0 .30

0 .35

0 .40

K/A

l

0 .0 0 .1 0 .2 0 .3 0 .4 0 .5

Al/Si

0 .02

0 .04

0 .06

0 .08

0 .10

Ca

/Al

Beni @ Rurrenabaque

Beni @ RiberaltaMadre de Dios

Mamoré

Madeira

0 .0 0 .1 0 .2 0 .3 0 .4 0 .5

Al/Si

0 .08

0 .09

0 .10

Mg

/Al

0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5

Al/Si

0 .00

0 .05

0 .10

0 .15

Na

/Al

Madeirareach

Benireach

AndesForeland

Lowland

Fig. 5. Na/Al, K/Al, Ca/Al, and Mg/Al vs. Al/Si in the Madeira basin. The symbols description also features a simplified hydrographic network of the Madeira Basin, where bar width isproportional to the sediment flux contribution.

174 J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

where _F ⋅ _Alh i

is simply the flux of Al through the reach (in MT/yr).

Hence, Eq. (3) can be written:

_WX ¼ _F ⋅ _Alh i

⋅ _Γ X=Al; ð5Þ

where _Γ X=Al is the grain size-specific change in the X/Al ratio betweenthe incoming and outgoing sediment (in g/g), defined as:

_Γ X=Al ¼_XAl

!M

−∑I

_XAl

!I

: ð6Þ

_Γ X=Al actually corresponds to a vertical offset in Fig. 3 at a given grainsize, or equivalently at a given Al/Si (Fig. 2). The input to the floodplain

reach term,∑I_XAl

� �Iis a sediment flux-weighted X/Al ratio, which can

be calculated using sediment fluxes from each tributaries (Table 1)along with their chemical compositions (Fig. 6). It should be kept inmind that _Γ X is a difference between two large fluxes, from which aweathering signal might be difficult to extract in certain cases.

Finally, as compared to the sediment “initial” (i.e. input to the riverreach) chemical composition, the relative change in weathering index_γX=Al (in %) can be defined as:

_γX=Al ¼_Γ X=Al

∑I_F I⋅ _Alh iI⋅ _X

Al

� �I � 100 ð7Þ

In the absence of grain comminution, the concentration of Al in theconsidered grain size cannot change downstream. In this case, relative

0.0

0.2

0.4

0.6

0.8

Ca/

Al

“MMPH” mixture

Marañon (mouth)

ΓCa/Al

Huallaga

Marañon (Andes)

Pastaza

Morona

C

0.0 0.1 0.2 0.3 0.4 0.5

Al/Si0.0 0.1 0.2 0.3 0.4 0.5

Al/Si

0.0

0.1

0.2

0.3

0.4

Na/

Al

Marañon (m

outh)

Pastaza

Huallaga

Morona

Marañon (Andes)

“MMPH” mixture

B

A

Coarse particles

Fineparticles

River channel

Overbank deposits

Point bars - levees

F M

Tributaries inputs

Output at mouth

River floodplain

River reach

WX

WX

I

X

.

Σ F IX

.

F DX

.

F DX

.

F EX

.

F EX

.

.

.

Fig. 6. Steady-state mass-balance model used to determine weathering intensities in floodplains. (A): schematic depicting the mass-balance. The sediment coarse fraction isenriched near the bottom of the river channel, on channel levees and on point-bars; the sediment fine fraction is enriched near the surface of the channel and in overbank deposits.Fluxes are named following Section 4.1. The steady-state mass-balance in the floodplain dictates that for each grain size, the difference between deposition fluxes _F

DX and export

fluxes _FEX of a soluble element X is equal to the weathering flux in floodplain _WX (Eq. (2)). The steady-state mass-balance in the river channel dictates that for each grain size,

the difference between the sum of inputs from tributaries _FIX and the export at mouth _F

MX is also equal to _WX (Eq. (3)). (B) and (C): examples for Na/Al and Ca/Al in the Solimões

Basin (symbols are the same as in Fig. 4). The thick red line is the modeled mixture of the Marañon at Borja (Andes), Morona, Pastaza and Huallaga rivers (“MMPH mixture”). Thevertical offset between this mixture and the Marañon at San Regis (mouth) is _Γ X=Al . For _Γ Na=Al , this offset is too small to be indicated on the diagram. The gray symbol represents thebed sediment sample AM-08-19, which was not taken into account when fitting the X/Al–Al/Si relationships (Appendix A).

175J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

changes in theX/Al ratio canonly stem fromrelative changes in theX con-centration, _wX , therefore:

_wX ¼ _γX=Al: ð8Þ

Eq. (8) shows that the relative change in X content in the sedimentcan be estimated using the relative change in the X/Al ratio, whichcan be read using Fig. 3. _wX and _γX=Al are actually grain size- andelement-specific weathering intensities, using the chemical composi-tion of the sediment supplied to the river reach as a reference.

Values of _wX are plotted in Fig. 7 as a function of Al/Si. _wNa values aresignificant regardingmethod uncertainties (i.e. b15%, Appendix A) in theMarañon reach at high Al/Si ( _wNa≃50%) and in the Madeira for all grainsizes ( _wNa slightly above 15%)). In all reaches, _wK and _wMg are small forall grain sizes. _wCa is significantly negative (≃−30%) at very low Al/Si ofthe Beni reach and significantly positive at low Al/Si in theMadeira reach(≃+50%). _wCa is significantly negative for all grain sizes in the Marañonreach ( _wCa from−45% and−60%).

4.2. Grain size-integrated floodplain weathering intensities

As shown by Bouchez et al. (2011c) and Lupker et al. (2011), thevariability in chemical composition of river sediments with grain

size can be integrated using sampling along depth-profiles. Similarto the grain size-specific weathering intensities _Γ X=Al and _WX , grainsize-integrated weathering intensities ΓX/Al and WX can be defined(respectively in g/g and in MT/yr). Weathering processes leadingto the transfer of a soluble element from a grain size fraction to an-other (for example, K solubilization from a coarse primary mineral,followed by a complete incorporation of solubilized K into clays)have no effect on ΓX/Al and WX. Graphically, ΓX/Al corresponds to thevertical offset between two trends in Fig. 3 (each reflecting a sam-pling location on given river) read at an Al/Si ratio that is representa-tive of the grain size-integrated sediment, hereafter referred to as (Al/Si)int. In other words, the grain size-integrated metric ΓX can also beread as _Γ X at (Al/Si)int.

Since we assume that there is no preferential deposition of a givengrain size and no downstream fining, the grain size-integrated Al/Si ofriver sediment does not change downstream through a change ofgrain size, as would be expected from the relationship between thesediment grain size and the Al/Si ratio (Fig. 2). However, it couldalso be argued that (Al/Si)int changes downstream due to weatheringreactions that solubilize Si. This effect can be assessed using publishedvalues of dissolved Si at different points of river courses in the Ama-zon basin. Moquet et al. (2011) report decadal estimates for the Siconcentration for the tributaries considered in the present study.

-100

-50

0

50

100

0 0.1 0.2 0.3 0.4 0.5

Al/Si

.

Loss

Gai

n

(Al/Si)int

Mara–on reach

wx

(%)

0.1 0.2 0.3 0.4 0.5

Al/Si

Loss

Gai

n

Beni reach

(Al/Si)int

0.1 0.2 0.3 0.4 0.5Al/Si

Loss

Gai

n

Madeira reach

(Al/Si)int

?

Na K Mg Ca

Fig. 7. Grain size-specific weathering indexes of river sediment in floodplains (using the upstream sediment supplied to the river reach as a reference), _wX , for the three studiedreaches, as a function of the Al/Si ratio. For Al/Si ratios lower than 0.1, the uncertainty was considered to be too high for the data to be reliable, and the results were not plotted(Appendix A). The shaded gray area represents the limit under which _wX is not considered as significant, owing to the uncertainties of the calculation (Appendix A). (Al/Si)int isthe grain size-integrated Al/Si ratio determined by Bouchez et al. (2011c) for the Solimões (green area) and for the Madeira (orange area) sediments at mouth, ±15% (see Section 4.2).This Al/Si value serves to determine, for each reach, the grain size-integrated weathering intensities (Fig. 8).

176 J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

Balancing between tributaries inputs and reach output of dissolved Si(i.e. combining the dissolved Si concentration reported by Moquet etal. (2011) and Mortatti and Probst (2003) with water discharges fromthe HyBAm program, listed in Table 1) is used to estimate the amountof Si solubilized from sediments during transport and transient stor-age. This is likely to be an overestimate of the Si solubilized from sed-iments since it takes into account Si supplied by small tributariesdraining only the plain, which deliver significant amounts ofdissolved Si to the main river channels. Then, this amount can becompared to the amount of Si transported by sediments, using sedi-ments fluxes (Table 1) and representative Si concentrations in parti-cles (Table 2). This would correspond to a change in the Al/Si ratioby less than 2% ratio during transport through the three consideredreaches. Therefore, (Al/Si)int can be safely assumed to be constantdownstream.

Importantly, since a relationship exists between the Al/Si ratio andAl concentration in Amazon river sediments (Bouchez et al., 2011c), aconstant (Al/Si)int also means that the grain size-integrated Al con-centration is invariant downstream. Moreover, as we assume thatthere is no floodplain aggradation on long time-scales, the grainsize-integrated sediment flux is also constant downstream. Hence, itcan be shown that (Appendix B):

WX ¼ Fint⋅ Al½ �int⋅ΓX=Al; ð9Þ

where Fint and [Al]int are respectively the grain size-integrated sedi-ment flux and Al concentration, the values of which are given byBouchez et al. (2011a, 2011c). Eq. (9) is the grain size-integratedequivalent of Eq. (5). Finally, the relative changes in the grainsize-integrated X concentration (wX) and the relative change in thegrain size-integrated X/Al ratio (γX/Al, both in %) can be defined.Since Fint and [Al]int do not change downstream:

wX ¼ γX=Al: ð10Þ

Bouchez et al. (2011c) provide an estimate of (Al/Si)int for theSolimões and Madeira rivers, based on depth-integrated fluxes of Aland Si at two distinct times of the hydrological cycle. The average be-tween these two sampling campaigns is 0.25 for the Solimões atmouth and 0.34 for the Madeira at mouth. Hydrodynamic conditionsduring sampling (i.e. water flow velocity) are a first order control on

the integrated Al/Si of river sediments (Bouchez et al., 2011c;Lupker et al., 2011). However this variability of flow velocity is limitedat themouth of these large rivers as the large drainage area averages outthe seasonal variability and results in a dampened hydrograph (Richeyet al., 1989). Hence, at these locations, the two available sampling cam-paigns offer a reliable insight into the long-term (Al/Si)int of the wholebasin. As a sensitivity test, two different values of (Al/Si)int and the cor-responding results are reported in Table 3; these two values correspondto those obtained at the mouth of the large rivers, ±15%.

Values of ΓX/Al, WX and wX are reported in Table 3 and wX valuesare plotted in Fig. 8.wX values are discussed in terms of their potentialimpact on the dissolved load hereafter (Section 4.4). In most cases,these values are lower for the Beni and Madeira reaches than forthe Marañon reach. The wNa values are significantly positive in theMarañon reach (wNa=16 to 19%), small in the Beni reach (wNa=6%), and significantly negative (wNa=−16 to−17%) in the Madeirareach. K change is relatively small (wK between −6% and 2%). wMg

is also small for the three reaches (between −6% and 9%). No signif-icant change in the Ca content is observed in the Madeira and Benireaches (wCa between −2% and 7%), but wCa is between −44 and −48% in the Marañon.

4.3. Weathering processes during transfer of sediments through theAmazon floodplains

In the absence of grain comminution, grain size-specific changesin the concentration of soluble elements during transfer through theriver plains can be interpreted in terms of weathering processes.These processes are dissolution of silicate, carbonate and evaporiteminerals, and incorporation of soluble elements into secondaryminerals. In the Amazon Basin, the high solubility of evaporite min-erals as well as their fast dissolution kinetics is likely to result intheir complete dissolution within the Andean domain (Moquetet al., 2011). Therefore, evaporite dissolution should not contributeto any downstream trend of sediment chemical composition in theAmazon foreland–lowland.

In most cases, during transfer through the Amazon plains, Figs. 7and 8 show either no change or a decrease in the soluble elementcontent. The only two exceptions are the observed enrichment ofNa in the Marañon reach at Al/Si ratios >0.20 and of Ca in the Madeirareach at Al/Si ratios b0.25. Such enrichments could suggest scavenging

Table 2Chemical composition of river sediments from the Bolivian and Peruvian sampling sites. Concentrations in ppm. See http://helium.crpg.cnrs-nancy.fr/SARM/ for uncertainties. N.D.: not determined.

Sample Si Al Fe Mg Ca Na K Ti Ba Ce Cs Dy Er Eu Gd Ho La Lu Nd Pr Rb Sm Sr Tb Th Tm Yb Zr

AM-07-01 349,421 58,225 33,579 5382 2164 4125 17,201 3522 372 68 7.0 4.5 2.5 1.20 5.02 0.86 32.7 0.40 29.5 7.8 99 5.9 69 0.79 10.6 0.37 2.52 206AM-07-02 298,027 91,466 46,249 7950 2407 5743 26,072 4722 540 94 12.4 6.1 3.3 1.63 7.00 1.18 45.7 0.52 41.3 10.9 156 8.3 96 1.07 15.7 0.50 3.40 214AM-07-03 289,123 92,928 46,543 8022 2493 5809 26,503 4734 569 99 13.1 6.5 3.5 1.72 7.25 1.25 48.4 0.56 43.2 11.5 164 8.5 100 1.12 16.6 0.53 3.60 215AM-07-04 289,193 95,289 47,229 8208 2679 6024 27,117 4878 581 103 13.6 6.7 3.7 1.80 7.54 1.32 50.3 0.58 45.3 12.0 166 9.0 102 1.19 17.2 0.57 3.75 231AM-07-05 391,375 37,991 24,948 3636 1743 3005 11,858 2958 263 68 3.8 3.7 2 1.11 4.57 0.69 32.3 0.32 29.4 7.8 66 5.9 52 0.65 8.1 0.30 2.05 228AM-07-06 335,487 68,469 36,582 6024 1936 5023 20,578 4560 423 84 8.4 6.5 3.8 1.44 6.61 1.31 40.4 0.61 36.6 9.7 117 7.4 79 1.08 13.6 0.58 4.00 473AM-07-07 341,768 64,212 34,244 5670 1857 4971 19,674 4230 407 81 7.7 5.7 3.3 1.32 6.01 1.12 38.7 0.52 34.4 9.2 110 6.9 76 0.96 14.2 0.49 3.42 419AM-07-08 322,919 72,805 38,633 6516 2114 5208 21,591 4746 464 90 9.2 6.9 4.0 1.58 6.96 1.38 43.5 0.63 39.2 10.5 128 8.0 85 1.14 14.8 0.60 4.07 454AM-07-09 302,321 83,319 42,973 7230 2207 5231 24,321 4926 494 91 10.8 6.6 3.7 1.58 6.81 1.32 43.9 0.59 39.1 10.5 143 7.9 90 1.10 15.3 0.57 3.82 322AM-07-10 394,053 35,248 20,930 2994 1029 3123 12,040 3270 260 75 3.6 5.5 3.4 1.13 5.43 1.14 36.5 0.64 32.0 8.6 63 6.3 52 0.87 10.5 0.55 3.81 1032AM-07-11 375,475 37,789 18,571 3192 2743 5438 13,700 2532 496 83 6.2 5.7 3.2 1.56 6.24 1.11 39.9 0.49 36.6 9.7 99 7.4 106 0.97 12 0.47 3.20 296AM-07-12 286,696 86,808 49,350 8436 5379 5409 20,487 5952 575 96 8.3 6.8 3.8 1.79 7.36 1.34 46.7 0.59 42.5 11.3 123 8.6 115 1.14 14.8 0.56 3.81 333AM-07-13 283,122 89,047 50,344 8616 5464 5468 20,919 6120 611 99 8.7 7.0 3.9 1.86 7.62 1.39 48.6 0.61 44.4 11.7 127 9.0 117 1.18 15.4 0.58 3.93 313AM-07-14 275,049 94,447 52,164 9000 5464 5342 21,881 6024 624 98 9.4 6.8 3.7 1.84 7.51 1.32 47.8 0.58 42.9 11.4 135 8.8 120 1.17 15.3 0.55 3.72 261AM-07-15 387,170 38,615 18,732 3246 2850 5691 13,982 2616 368 46 2.6 2.8 1.5 0.89 3.26 0.53 22.2 0.24 20.0 5.3 62 3.9 74 0.49 5.2 0.22 1.54 128AM-07-16 389,587 37,260 18,781 3006 1057 3480 13,467 2658 315 50 3.4 3.4 1.9 0.82 3.57 0.67 23.4 0.33 21.3 5.7 64 4.2 49 0.56 7.9 0.30 2.12 333AM-07-17 386,806 37,122 18,193 2982 1021 3413 13,675 2622 316 53 3.3 3.7 2.1 0.83 3.90 0.75 24.7 0.36 22.5 6.0 65 4.5 50 0.62 9.1 0.33 2.31 354AM-07-18 353,337 53,174 26,880 4512 1629 4585 17,326 3624 401 67 5.3 5.2 3.0 1.16 5.12 1.03 31.8 0.49 28.5 7.6 91 5.7 65 0.84 11.1 0.47 3.14 358AM-07-19 330,064 65,123 33,635 5682 2114 5275 20,048 4248 443 77 6.6 5.6 3.2 1.30 5.69 1.10 36.2 0.54 32.9 8.8 107 6.6 76 0.94 12.7 0.50 3.42 356AM-08-01 355,549 54,715 29,974 9972 14,850 12,598 15,135 3408 471 57 2.6 3.8 2.1 1.21 4.46 0.74 27.6 0.32 26.4 6.8 62 5.4 218 0.67 7.3 0.32 2.10 184AM-08-02 321,473 68,479 34,034 9540 12,936 12,212 18,703 4122 527 67 4.8 4.6 2.6 1.29 4.93 0.90 32.9 0.42 30.4 7.9 88 6.0 193 0.78 10.4 0.41 2.73 228AM-08-03 287,854 78,533 40,957 10,770 12,507 10,446 19,359 4572 532 69 6.4 4.9 2.8 1.37 5.32 0.97 33.8 0.45 31.1 8.1 100 6.2 172 0.82 11.0 0.43 2.90 224AM-08-04 287,719 73,938 38,220 10,344 12,136 10,728 18,479 4206 537 68 5.9 4.9 2.7 1.35 5.19 0.95 33.4 0.43 31.0 8.1 97 6.2 175 0.82 10.6 0.42 2.82 191AM-08-05 284,578 87,390 45,136 11,646 12,636 9964 21,633 5148 548 78 7.5 5.6 3.1 1.46 5.82 1.07 38.0 0.49 34.4 9.2 110 6.9 160 0.92 13.8 0.47 3.22 233AM-08-06 371,093 39,150 26,992 11,826 18,393 11,715 9211 2250 430 48 0.9 3.4 1.7 1.20 4.18 0.62 22.1 0.23 23.6 5.8 32 5.0 260 0.61 4.1 0.24 1.52 81AM-08-07 329,145 62,179 31,304 9042 12,957 13,013 17,616 3930 494 57 3.9 4.1 2.3 1.18 4.31 0.79 27.6 0.35 26.4 6.8 79 5.3 198 0.68 8.3 0.34 2.32 239AM-08-08 313,502 70,179 35,049 9618 12,621 12,027 18,952 4380 524 64 5.1 4.9 2.8 1.25 5.10 0.99 31.6 0.46 29.3 7.6 89 6.0 186 0.82 10.4 0.44 3.01 263AM-08-09 310,151 75,653 38,479 10,248 12,479 11,671 19,840 4668 539 72 5.7 5.1 2.9 1.38 5.42 1.01 35.4 0.47 32.5 8.5 96 6.5 179 0.86 10.9 0.45 3.05 278AM-08-10 288,885 82,239 42,413 11,256 11,336 10,127 23,002 4950 569 84 8.3 5.9 3.3 1.56 6.37 1.14 41.2 0.52 37.5 9.9 121 7.5 150 1.00 13.0 0.50 3.29 213AM-08-12 288,269 84,229 44,044 11,724 12,043 10,120 23,350 5262 587 86 8.6 6.0 3.4 1.62 6.58 1.17 42.0 0.54 37.9 10.1 125 7.6 154 1.02 13.4 0.52 3.51 259AM-08-13 278,353 85,664 44,639 11,976 11,971 9964 23,267 5244 593 89 9.2 6.3 3.4 1.69 6.72 1.23 43.6 0.55 39.8 10.6 127 8.2 152 1.06 14.2 0.55 3.58 226AM-08-15 288,260 74,838 43,281 10,482 7700 2782 17,849 4560 291 78 6.9 5.3 3.1 1.33 5.50 1.08 38.8 0.52 33.5 9.0 109 6.6 96 0.88 13.1 0.49 3.33 304AM-08-16 348,413 62,010 28,308 7140 8643 13,451 20,803 3702 542 65 4.1 5.1 3.0 1.27 5.23 1.04 31.4 0.49 29.2 7.7 96 5.9 158 0.85 8.9 0.45 3.10 221AM-08-17 340,102 63,805 30,401 7548 9086 12,902 20,828 4146 542 70 4.6 5.2 2.9 1.36 5.33 1.01 34.2 0.45 31.6 8.3 97 6.2 155 0.86 9.0 0.44 2.89 261AM-08-18 320,707 72,111 35,392 9246 9850 11,767 21,906 4452 540 75 6.0 5.1 2.7 1.41 5.49 0.98 36.7 0.45 33.5 8.8 105 6.7 152 0.85 10.8 0.42 2.81 211AM-08-19 351,517 50,887 35,308 15,132 22,793 14,290 11,534 3312 529 47 1.3 3.1 1.6 1.11 3.74 0.59 22.0 0.24 22.5 5.6 41 4.7 324 0.55 5.1 0.24 1.57 113AM-08-20 321,622 66,208 34,230 11,148 18,079 13,140 15,558 3828 493 50 3.4 3.6 2.0 1.80 3.85 0.69 24.2 0.33 22.7 5.9 66 4.6 231 0.59 8.1 0.30 2.05 168AM-08-21 292,679 80,926 39,956 11,094 16,421 11,455 18,222 4422 545 66 5.3 4.6 2.6 1.31 4.94 0.91 32.2 0.42 29.7 7.8 86 6.0 202 0.78 11.2 0.39 2.64 170AM-08-22 307,809 74,345 37,079 11,004 17,886 12,138 17,276 4200 538 62 4.5 4.5 2.5 1.23 4.78 0.88 30.2 0.41 27.8 7.2 80 5.6 221 0.75 10.2 0.39 2.61 197AM-08-23 305,429 76,309 38,640 11,310 17,714 11,915 17,384 4506 523 71 4.6 4.6 2.6 1.31 5.01 0.90 34.7 0.42 31.8 8.3 80 6.2 212 0.78 11.3 0.39 2.68 206AM-08-24 275,534 89,947 44,576 11,676 15,457 10,083 19,641 4794 559 75 6.5 5.3 3.0 1.46 5.59 1.06 36.7 0.49 33.3 8.7 100 6.7 171 0.90 13.4 0.46 3.11 196AM-08-25 315,485 67,166 34,664 10,518 17,514 12,643 16,023 4068 475 52 3.5 3.9 2.2 1.11 4.12 0.78 25.3 0.37 23.9 6.2 69 4.9 217 0.65 8.5 0.34 2.35 213AM-08-26 316,909 67,304 34,244 10,404 17,150 12,502 16,123 4098 514 61 4.0 4.5 2.6 1.23 4.73 0.90 29.6 0.43 27.7 7.2 73 5.5 226 0.76 9.6 0.39 2.64 241AM-08-27 323,633 66,123 33,446 10,248 17,243 12,828 16,048 3978 502 54 3.6 3.7 2.1 1.12 4.15 0.74 26.2 0.35 24.9 6.4 71 5.0 228 0.63 9.0 0.32 2.21 217AM-08-29 331,361 50,665 27,636 7746 39,479 10,936 13,774 3348 402 45 2.7 3.6 2.1 0.92 3.66 0.73 21.4 0.37 20.3 5.3 60 4.2 170 0.60 7.4 0.33 2.29 335AM-08-30 302,680 65,382 30,443 7566 37,214 8302 15,865 3582 379 56 5.0 4.0 2.3 1.10 4.32 0.80 27.1 0.37 25.0 6.5 72 5.1 145 0.67 9.2 0.35 2.34 157AM-08-31 279,561 76,188 35,721 8364 38,986 7879 17,923 4014 413 65 6.6 4.5 2.6 1.27 4.81 0.89 31.5 0.43 29.0 7.6 85 5.8 153 0.75 10.9 0.40 2.71 167AM-08-32 280,849 73,440 34,741 8358 39,786 8087 17,484 3972 425 68 6.2 4.8 2.7 1.28 5.08 0.95 32.7 0.45 29.9 7.9 85 6.1 160 0.80 11.5 0.42 2.81 193AM-08-33 252,271 88,893 41,566 9264 39,179 6759 19,658 4752 422 79 7.7 5.3 3.0 1.47 5.74 1.03 37.8 0.49 34.6 9.1 92 6.9 151 0.89 13.0 0.46 3.10 205AM-08-34 272,697 95,702 44,786 7842 8007 5275 13,401 5016 474 77 6.4 5.2 2.9 1.40 5.49 1.03 38.2 0.49 33.7 9.0 84 6.6 144 0.88 13.8 0.45 3.11 219AM-08-35 380,823 39,616 24,696 9846 16,564 11,314 7410 2346 415 41 1.1 2.9 1.4 1.14 3.75 0.52 18.1 0.20 21.3 5.2 26 4.6 249 0.53 3.7 0.20 1.30 106AM-08-36 246,652 97,031 50,610 13,074 23,557 15,840 12,837 5406 709 53 3.3 3.9 2.1 1.34 4.28 0.73 26.2 0.34 25.2 6.4 57 5.2 374 0.66 8.7 0.31 2.11 161AM-08-37 291,835 79,814 45,017 19,914 37,586 26,954 15,633 4488 836 44 1.8 2.7 1.4 1.16 3.26 0.51 21.3 0.22 21.2 5.4 49 4.4 615 0.49 6.7 0.21 1.37 131AM-08-38 272,300 86,225 41,566 12,504 14,307 8458 22,578 4872 564 96 6.8 6.4 3.6 1.75 6.69 1.24 47.2 0.58 42.4 11.2 119 8.3 136 1.08 15.3 0.54 3.65 229AM-08-39 401,968 35,264 19,250 4560 6671 9578 13,741 3006 388 51 1.4 3.2 1.9 0.86 3.38 0.63 25.4 0.30 21.6 5.8 54 4.1 99 0.53 7.5 0.28 1.92 146 177

J.Bouchezet

al./Chem

icalGeology

332–333

(2012)166

–184

Table 3Net loss of soluble elements (X=Na, K, Mg or Ca, negative values mean loss of X) in the river sediment over the three reaches. See text for symbol definitions (Section 4). Twoextreme values of the grain size-integrated Al/Si ratio, (Al/Si)int, are used for each of the large basins (Solimões and Madeira), as a sensitivity test.

Reach Parameter (Al/Si)int=0.20 (Al/Si)int=0.28

Na K Mg Ca Na K Mg Ca

Marañon ΓX/Al (g/g) 0.027 −0.014 0.012 −0.219 0.022 −0.008 0.012 −0.187wX (%) 16 −6 8 −44 19 −3 9 −48WX (MT/yr) 0.36 −0.19 0.17 −2.89 0.29 −0.11 0.15 −2.47

Reach Parameter (Al/Si)int=0.29 (Al/Si)int=0.39

Na K Mg Ca Na K Mg Ca

Beni ΓX/Al (g/g) 0.004 0.005 0.001 0.000 0.003 0.003 0.003 0.002wX (%) 6 2 2 −2 6 1 3 7WX (MT/yr) 0.06 0.08 0.02 −0.01 0.05 0.04 0.04 0.03

Madeira ΓX/Al (g/g) −0.011 −0.015 −0.003 0.002 −0.011 −0.012 −0.005 −0.002wX (%) −16 −5 −4 5 −17 −5 −6 −4WX (MT/yr) −0.40 −0.52 −0.12 0.06 −0.39 −0.43 −0.17 −0.05

178 J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

of Na and Ca by sediments during transfer through the plain, for exam-ple by clay precipitation (such as Na- and Ca-smectites) or by adsorp-tion onto clays or organic matter. It is difficult to explain why thisprocess should be prominent in theMarañon basin for Na. Furthermore,the Ca enrichment in the Madeira reach is observed for coarse sedi-ments, showing that clay precipitation cannot explain this Ca enrich-ment. Carbonate precipitation is also unlikely in the Madeira reach,sincewaters in this region are undersaturatedwith respect to carbonateminerals. It should also be noticed that the Ca/Al–Al/Si relationship ofthe Madeira at mouth is driven towards high Ca/Al ratios for coarsesamples by one bed sediment sample. If this sample is ignored, muchlower values of _wCa are obtained at low Al/Si.

Besides these exceptions, the only significant trends are (1) Na is lostin the Madeira reach and (2) Ca is lost in the Marañon and the Benireaches. These changes are due to mineral dissolution during transport.In the sediments of the lowland Amazon Basin, Na is mainly carried byplagioclase, which is enriched in low Al/Si samples (Bouchez et al.,2011b). Hence, the loss of Na in the Madeira reach occurs most probablythrough the dissolution of plagioclase. It is worth noticing that this pro-cess of plagioclase dissolution is not accompanied by a significant lossof Ca in the Madeira reach. Such effect might result from either (1) thelack of precision of our calculations (since Ca concentrations are extreme-ly low in theMadeira sediments), or (2) a decoupling between Ca and Nabehavior during plagioclase dissolution (Lupker et al., 2012), which couldresult fromapreferential release of Ca overNa fromplagioclase in theAn-dean domain, or (3) a low plagioclase Ca/Na ratio in the crust drained bythe Madeira River.

However, the fact that Ca loss in theMarañon and Beni reaches is notassociated with Na loss shows that, in this case, Ca loss is not controlledby plagioclase dissolution. Although Ca could also be lost through disso-lution of other primary minerals such as pyroxene or epidote, thisprocess is unlikely to explain the observed Ca loss in the Beni reach, asthe Beni headwaters do not drain large volcanic or metamorphicareas, and as our own inspection of the Beni and Marañon samplesdid not reveal the presence of these minerals. Moreover, the fact that_wCa is negative for all grain sizes in the Marañon tends to show thatCa loss is controlled by amineralwhich is not strongly sorted. Therefore,the observed Ca loss is a result from the dissolution of carbonate min-erals during sediment transport. Although carbonate particles are virtu-ally absent in the sediment of Amazon lowland Basins (where totalcarbonate weight proportion of particles is lower than 0.08 wt.%; Galyet al., 2007; Bouchez et al., 2011b), carbonates are likely to be presentin trace amounts in the upstream areas, closer to the sediment source.The Solimões headwaters are known to drain carbonate outcrops inthe Peruvian Andes (Putzer, 1984), and the Amazon foreland is alsolikely to contain carbonate minerals as detrital particles, or as cement.Similarly, following our own observations, the black shales formationsdrained by the Beni headwaters contain disseminated carbonate. A

part of the particulate carbonates could be dolomite, which would inturn also affect the Mg budget. However, dolomite was not detectedin the sediments from the Amazon lowland by Galy et al. (2007), andno significant Mg loss could be calculated in the studied reaches(Figs. 7 and 8). Although large fluxes of carbonate weathering in theAndes were reported by Moquet et al. (2011) on the basis of dissolvedload budgets, undissolved carbonate particles are likely to remain andto be exported at the outlet of the Andes. Such particles are likely tobe later dissolved in the Amazon plain, where the waters are undersat-urated with respect to carbonate minerals. Interestingly, Moquet et al.(2011) reported significant carbonateweathering fluxes in the forelandarea of the Solimões basin (which comprises theMarañon reach studiedhere, the foreland parts of theHuallaga basin and the Pastaza basin) andin the Beni basin. In these two areas, the carbonate weathering flux is 5and 3 times higher than the silicate weathering flux, respectively(Moquet et al., 2011).

4.4. Comparison with floodplain weathering fluxes determined from riverdissolved loads

The weathering processes resulting in changes of the sedimentchemical composition should also have an impact on the compositionof the dissolved load. In the following, weathering fluxes calculatedfrom dissolved load budgets are compared with those calculated inSection 4.2 for the sediment chemical composition. A steady-statemass-balance can be established on the dissolved load, by calculatingthe difference between the incoming and outgoing dissolved fluxes toand from the reach:

WX;diss ¼ QM⋅ X½ �Mdiss−∑IQ I⋅ X½ �Idiss; ð11Þ

where Q is the water discharge and [X]diss is the time-averaged,atmospheric-corrected concentration of X in the dissolved phase. IfWX,diss is positive, the considered element is gained by the dissolvedphase. To calculate WX,diss from Eq. (11), we used the data reportedby Moquet et al. (2011), Mortatti and Probst (2003), and Gaillardetet al. (1997). Results are listed in Table 4. All WX,diss values are posi-tive, showing that the Andean foreland and the Amazon plain areloci of active silicate and carbonate weathering (Moquet et al., 2011).

AlthoughWX,diss andWX agree, within uncertainties (Appendix A),very broadly, most WX,diss values are larger than WX (Fig. 9). This ob-servation would suggest that weathering of actively transported sed-iments in not the sole source of cations in the foreland and the plain.Another potential source of major soluble cations is the weathering ofstable sedimentary formations that actually make up most of theAmazon plain surface area (such as the alluvial terraces called terrafirme in the Amazon), by opposition to the active floodplain that

+30

+10

-10

-30

-50

Na K Mg Ca

wX (

%)

Mara on reach

Na K Mg Ca

Madeira reach

Na K Mg Ca

Beni reach

Gai

nLo

ss

Fig. 8. Grain size-integrated weathering indexes of river sediment in floodplains (using the upstream sediment supplied to the river reach as a reference), wX, for the three studiedreaches. The shaded gray area represents the limit under which wX is not considered as significant, owing to the uncertainties of the method (Appendix A).

1

2

3

Na K

Mg Ca

Element

ReachMara–onBeniMadeira

solv

ed lo

ad c

hem

ical

com

posi

tion

m M

oque

t et a

l., 2

011)

179J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

continuously exchanges sediment with the large river channels. Al-though weathering rates are very low in these settings, as shown bythe analysis of rivers draining exclusively lowland settings(Gaillardet et al., 1997; Markewitz et al., 2001), the integration ofthese low weathering rates over large areas could result in a signifi-cant modification of the dissolved flux. A first-order test of this hy-pothesis can be made by predicting the concentration of dissolvedspecies in the water draining these stable sedimentary formations:

X½ �predictplain ¼ WX;diss−WX

QM−∑IQI : ð12Þ

The denominator of Eq. (12) is the runoff associated with thereach only, and [X]plainpredict is thus the concentration of dissolved X in ariver draining only the plain or the foreland (such as a plain tributary)that would explain the difference between WX,diss andWX. Results arereported in Table 4. It should also be reminded that some of thisdissolved contribution might be supplied to the large tributaries bygroundwaters, that we assume here to have a chemical compositionsimilar to that of plain tributaries.

Negative values of [X]plainpredict (Ca in the Marañon, Na and K in theMadeira) might reflect issues with estimates of WX,diss and WX. Thesecases correspond to those where WX,diss is lower than WX (Eq. (12),Fig. 9). However, all these negative values are relatively small, and wecan consider atfirst order that they are equal to 0. Conversely, predictedconcentrations much higher than the typical concentrations measured

Table 4Weathering fluxes in the Amazon plain using dissolved load budgets (WX,diss), and pre-dicted dissolved concentrations in the plain tributaries in the case where the supply byplain tributaries accounts for the discrepancy between WX,diss and WX (Table 3). Seetext for details (Section 4.4). Number in italics represent data where the weatheringflux in the stable sedimentary formations of the Amazon plains is apparently negative,but can be considered as being 0, given the uncertainties of the calculations. Data fromthe Purus River, an Amazon tributary draining the Andean foreland and the plain, isshown for comparison.

Reach Parameter Na K Mg Ca

Marañon WX,diss (MT/yr) 2.62 0.22 0.20 2.48[X]plainpredict (μmol/L) 673 16 49 −26

Beni WX,diss (T/yr) 0.06 0.10 0.03 0.18[X]plainpredict (μmol/L) 99 129 30 92

Madeira WX,diss (T/yr) 0.20 0.45 0.42 1.81[X]plainpredict (μmol/L) −23 −3 18 121

Purus (Moquetet al., 2011)

[X]diss (μmol/L) 72±27 28±6 47±17 135±52

in plain tributaries, are obtained for Na in the Marañon reach and K inthe Beni reach (Table 4). Interestingly, the first case corresponds tothatwhere a significantNa gain is observed from the sediment chemicalcomposition (Section 4.1). Hence, the supply of dissolved Na by plaintributaries does not explain the increase in Na content in the river sed-iment, through the Marañon reach. Therefore, this discrepancy is bestexplained by (1) a yet unidentified source in the Marañon foreland,characterized by an extremely high Na concentration, or (2) a failureof one of the hypotheses used in the mass-balance in the case of thisreach. Regarding explanation (2), it is interesting to notice from Fig. 4that, should the grain size-integrated Al/Si ratio for the Marañon atBorja be lower than at San Regis, Nawould be themost affected elementsince its Al-normalized ratio displays the steeper relationship with

-1

0

-1 0 1 2 3

WX

,dis

s ca

lcul

ated

from

dis

-WX calculated from sediment chemical composition(MT/yr, data from this study)

(MT

/yr,

dat

a fr

o

Fig. 9. Comparison of floodplain weathering fluxes as calculated from the particulateload chemical composition (WX, data from this study) and from the dissolved loadchemical composition (WX,diss, calculated from the data reported by Moquet et al.,2011) in the three studied reaches, plotted for Na, K, Mg and Ca. −WX is plottedhere so that a positive value represents a gain for the dissolved phase on both axes.Concerning the Marañon reach, no dissolved nor water discharge data is available forthe Morona, but the Tigre has been taken into account. The budgets represented onthe X- and Y-axis are thus slightly different systems in those cases. The diagonalstapled line is the line for which WX=WX,diss (in absolute value).

Table 5Comparison of relative mass loss for the different major soluble element during trans-fer through floodplain (wX in the present study, the average between the two estimatesgiven in Table 3 was used), between different river systems. Geomorphological param-eters were estimated from satellite images.

System Marañon reach(this study)

Beni reach(this study)

Madeira reach(this study)

Gangetic plain(Lupker et al.,2012)

Na 17% 6% −17% −25%K −5% 1% −5% −10%Mg 9% 2% −5% −3%a

Ca −46% 3% 0% −47%b

Valley length(km)

550 450 1200 1300

Channel length(km)

600 800 1300 1000

Channel sinuosity 1.1 1.9 1.1 1.3Residence time(kyr)c

1 4.6 3 10

a The reported loss is only silicate-bound Mg (a large loss of Mg from dolomite wasalso actually observed by Lupker et al., 2012).

b The number reported is actually the loss of CaCO3 (calcite), since no loss of Ca fromsilicates was observed by Lupker et al. (2012) in the Ganges.

c Residence times are derived from U-series disequilibria, from Granet et al. (2010)for the Ganges River, and from Dosseto et al. (2006a, 2006b) for the Amazon Basin.For the Marañon reach, a first-order estimation was made using one fifth of the resi-dence time reported by Dosseto et al. (2006a) for the whole Solimões basin.

180 J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

the Al/Si ratio. Hence, preferential deposition of coarse, low-Al/Si ratiosediment between Borja and San Regis is a potential explanation ofthe positive value of wNa (Fig. 8) and of the discrepancy between WNa

and WNa,diss in the Marañon reach, although it fails to explain why finesediments are more Na-enriched than coarse sediments during thetransfer (Fig. 7). For the case of K in the Beni reach, this explanation ismost likely not valid since a downstream change in the Al/Si ratioshould affect Na as well, which is not the case (Table 4). The release ofK could be due to long-term changes in the inventory of the vegetationpool in the Beni basin, but this hypothesiswould need further investiga-tion. In any case, these significant differences between WX,diss and WX

emphasize potential issues associated with the assumptions on whichour steady-statemass-balance relies, and stresses the need for indepen-dent assessments of elemental budgets in large river reaches.

For half of the calculations, [X]plainpredict are commensurate to X con-centrations of a plain-foreland tributary such as the Purus (Table 4;Moquet et al., 2011). This is the case for Mg in the three reaches, Cain the Beni and Madeira reaches, Na in the Beni reach and K in theMarañon reach. For these elements and in these river reaches, thesupply of dissolvedmatter by plain tributaries explains the discrepancybetweenWX,diss and WX.

This partial agreement between predicted and observed dissolvedconcentrations in plain tributaries anyway shows that further investiga-tion is needed on the chemistry of plain tributaries and potentialgroundwater inputs in the Amazon (Markewitz et al., 2001). However,the analysis also suggests that the contribution of theweathering of sta-ble sedimentary formations to the Amazon dissolved load should not beignored. In a system like the Amazon, as in many other large river sys-tems, most of these formations consist of alluvial sediments derivedfrom the past erosion of the orogen producing the sediments currentlytransported by the river. Therefore, over geologic time scales, the chem-ical weathering of these sediments is a part of the same system as the ac-tive floodplain, i.e. a weathering system constituted by the sedimenteroded from the orogen, and the river transporting this sediment to theocean. A complete model of steady-state weathering in the foreland andthe lowland could be used to test this hypothesis, provided that the fluxand chemical composition of the solid phase exported by the forelandand lowland areas are known. However, these constraints are usuallynot available mainly because, in these settings, (1) the concentration ofsuspended sediment is low (Gaillardet et al., 1997), thereby renderingsediment sampling difficult; (2) it is likely that bedload, which remainsdifficult to estimate, represents a significant component of the sedimentflux. In any case, the steady-state concept imposes that if these sedimen-tary formations are currently weathering, they should also export solids.The presence of “old” (i.e. buried in floodplain deposits for more than1 Myr) sediments in the Amazon bed sediment has been recently re-vealed (Wittmann et al., 2011b), which suggests that the sedimentfrom stable sedimentary formations finds its way back into the activeriver channels. In this case, “old” sediment means that it was depositedand abandoned by the active floodplain several “floodplain reworkingcycles” ago. At the scale of the river reach, in addition to the actual ero-sion of foreland areas, the main process permitting sediment from low-land areas to return to the active river channel is likely to be riveravulsion. Although this type of event is relatively rare, it might consti-tute the means for abandoned sedimentary formations to export sedi-ment again. The chemical characterization and quantification ofsediment exported by foreland and lowland areas, as well as the identi-fication of the geomorphological processes generating these fluxes ofmatter, represent a future direction of research that will help to unravelthe weathering processes at play in large river basins.

4.5. Comparison between different river systems and potential controlson chemical weathering in the plains

The comparison of theweathering indexeswX between river reacheshaving different geomorphological, climatic and tectonic characteristics

may shed light onto the potential controls on weathering processes inriver floodplains, and allows us to evaluate whether the features ob-served here are valid on a larger scale. Lupker et al. (2012) reported rel-ative changes in Na, K, Mg, and Ca during transfer of sediment throughthe whole Gangetic plain, between the Himalayan front and confluencewith the Brahmaputra, which can be compared to the three riverreaches studied in the present contribution, due to their broadly similarscale (i.e. channel length between 500 and 1500 km, Table 5).

Consistently across the four considered systems, relative losses ofK and Mg are small (wK and wMgb10%). The relative loss of Na duringsediment transport varies greatly amongst the different systems: it isthe highest for the Gangetic plain, significant for the Madeira, smallfor the Beni, and a Na gain is calculated for the Marañon (Section 4.2).Finally, although Ca is not significantly lost in the Beni or the Madeirareach, the sediments from the Marañon and Ganges reaches lose halfof the Ca they initially contain (calculated from the total pool of Ca inthe Marañon and from the calcite pool in the Ganges).

As explained earlier (Section 4.3), the apparent gain of Na by sedi-ments in the Marañon is difficult to interpret and might be due to thenon-validity of the hypotheses used for the steady-state mass-balance.However, the relative Na losses shown by the Ganga andMadeira sedi-ment are comparable, and are explained in both cases by plagioclasedissolution. The fact that plagioclase dissolution is detectable in theMadeira reach and in the Gangetic plain, and not in the Marañon andBeni reaches, might be due to different channel length and residencetimes, but no significant trend is observable (Table 5). As proposed byLupker et al. (2012), the limited loss of K by sediment in the plainscan be due to the slow dissolution rates of K-bearing minerals such asK-feldspar and biotite as compared to those of plagioclase (Clow andDrever, 1996). Moreover, solubilized K and Mg are likely to be incorpo-rated into secondary minerals and thus to not appear in the overallchemical net budget of sediments.

Therefore, the limited loss of silicate-bound Mg and of K is a wide-spread feature across the different systems, and plagioclase/micadissolution occurs to the same extent in two large systems. Theseobservations shows first that, despite their different sizes and hencepotentially their different transfer times (Table 5), the transfer timeis too short for K-feldspar weathering to occur to a great extent,but sufficient for plagioclase weathering to proceed. Second, themechanisms of K and Mg retention in the solid phase, although yetlargely unconstrained and potentially different (because of different

181J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

clay mineralogy, different temperature, or different water residencetime in the floodplains), lead to comparable limited losses of these el-ements to the dissolved phase. At the current state of knowledge, andcomparing river reaches from two of the largest river systems in theworld, and K and Mg retention in particles during floodplain transferseems to be a widespread process.

Similarly, Ca loss through carbonate dissolution occurs to a largeextent in the Marañon basin and in the Gangetic plain, showing thatthe residence time in both systems is long enough so that carbonateminerals have time to dissolve, at least partially. The fact that thisloss is virtually zero in the Beni and Madeira reaches, despite relativelylong residence times, is probably due to the fact that carbonates are tooscarce in this basin to contribute to the particulate Ca. Moreover, in theMarañon, the Ca loss reported in Table 5 is likely a lower bound on theproportion of carbonate dissolved during transport since this Ca loss iscalculated on the total Ca pool, and includes silicate-bound Ca that isalso released, albeit much more slowly. It thus appears that, despiteits twice shorter valley length, the Marañon reach is able to dissolve agreater proportion the carbonate entering the floodplain than thewhole Gangetic plain does. Calcite dissolution kinetics is greatly depen-dent on the water saturation state (Morse and Arvidson, 2002):the water saturation index with respect to calcite (ratio between themeasured product [Ca2+]·[CO3

2−] and the thermodynamic solubilityproduct of calcite) is 1 to 10 in the Ganges (calculated from datareported by Galy and France-Lanord, 1999) and lower than 0.05 in theMarañon (as can be calculated from the data reported by Moquetet al., 2011), hence most likely explaining the difference between thetwo systems.

To summarize, the behavior of particulate Na during transferthrough the floodplain depends greatly on the setting, following yetunexplained climatic or mineralogical controls. Contrarily, K and Mgare not significantly lost by the sediment during transport, and thisfeature is relatively independent on the setting. Finally, carbonateweathering leads to an important loss of Ca (provided that carbonateare present) that is influenced by the water saturation state withrespect to carbonate minerals.

4.6. CO2 flux associated with silicate weathering of river sedimentsin floodplains

Weathering of silicate minerals being the major sink of atmosphericCO2 on geological time scales, the chemical weathering of river sedi-ment during transfer through the Amazon floodplains should resultin a flux of CO2 uptake that can be evaluated from the chemicalweathering fluxes estimated in this study.

Weathering of 1 mol of Ca- or Mg-bearing silicate mineral byatmosphere-derived CO2 leads, over geological time scales, to theconsumption of 1 mol of CO2 and its sequestration as carbonate(Berner et al., 1984). Weathering of 1 mol of Na- or K-bearing silicatemineral leads to the consumption of atmospheric CO2 only if dissolvedNa and K exchange with Ca and Mg from terrigenous sediment orfrom the oceanic crust (France-Lanord andDerry, 1997). Herewe calcu-late the flux of CO2 associatedwith chemical weathering in theMadeirareach only, since the behavior of Na is ambiguous in the Marañon reachand that of Ca is governed by carbonate dissolution (Section 4.3), andsince no significant change in Ca, Mg, Na or K fluxes is detected in theBeni reach (Section 4.2).

Following France-Lanord and Derry (1997) and the results fromSayles and Mangelsdorf (1977), we assume that 30% of Na lost fromsediments during transfer through the floodplain exchanges with Caor Mg upon entry in seawater. This corresponds to 0.12 MT Na/yr(Table 3). For each mole of Na participating to the exchange process,1/2 mole of Ca or Mg is released into the ocean, and hence 1/2 mole ofCO2 is sequestered into carbonates. The calculation yields a sequestra-tion of 0.03 MT C/yr in carbonates following weathering of silicates infloodplains of the lower Madeira. Gaillardet et al. (1997) reports a

CO2 consumption derived from silicate weathering equivalent onthe long-term to 0.4 MT C/yr for the whole Madeira Basin, based onthe dissolved load budget. Although the river reach consideredhere represents only a relatively minor portion of the Madeira basinarea, the associated CO2 consumption flux by silicate weatheringrepresents a significant part (10%) of the same flux over the wholebasin.

It has to be noticed that the relative change of Na content in sedi-ment (wNa) during transfer through the Madeira reach is just aboveour limit of detection (15%, Appendix A). In other words, a weatheringsignal is difficult to extract from a system dominated by mixing suchas the reaches studied here. However, the CO2 drawdown given aboverepresents anyway an order of magnitude of what weathering of sedi-ment during transport can result in, even if only 15 wt.% of theNa-bearing silicate minerals were weathered. This contribution is thussignificant, even for a relatively small change of sediment chemistryduring transport.

5. Conclusion

This study makes use of river sediment sampled at differentlocations throughout the Amazon Basin to evaluate the effects ofchemical weathering during riverine transport. The comparison ofthe content of major soluble elements (Na, K, Mg, Ca) in river sedi-ment makes it possible to calculate the weathering fluxes duringtransient storage of sediments in the floodplain. Sampling alongdepth-profiles allows us to consider the whole range of sedimentgrain size transported by the river, from coarse bed sediment toclay-enriched sediment at the channel surface. Concentrations ofmajor soluble elements are normalized to Al to correct for the effectof quartz dilution.

The chemistry of sediment does not change drastically down-stream in the two large sub-basins delivering Andean sediment tothe Amazon system. In more detail, a quantitative analysis carriedout on three river reaches (Marañon, Beni, Madeira) shows that(1) K and Mg are not significantly lost by the sediment, (2) plagioclaseweathering generates a relatively limited loss of Na and (3) a significantCa loss is induced by carbonate dissolution, especially in the SolimõesBasin. A comparison with the Gangetic plain suggests that this latterprocess is controlled by water chemistry and saturation state withrespect to carbonate minerals, but also that our observations are wide-spread and not just valid at the scale of the three studied reaches.Further comparisons to river systems with contrasted climatic condi-tions and tectonic settings will lead to a better understanding of thecontrols on weathering in floodplains, and how these controls affectdifferent elements to various extents.

Although the changes due to chemical weathering in floodplainsare small in terms of sediment chemical composition (except for Cabecause of carbonate dissolution), they should result in significantchanges in the dissolved load, due to the overall smaller fluxes car-ried by the dissolved load. In other words, the river dissolved loadshould be more sensitive than the particulate load to weathering re-actions in the floodplains, and the great wealth of data availablethroughout the world's largest river systems could be used to inferthese weathering fluxes. However, the river dissolved load in plainareas is sensitive to weathering of both actively transported sedi-ments and of old alluvial deposits, which are not regularly visitedby the channel.

The estimation of the CO2 drawdown associated with silicateweathering during riverine transport emphasizes the potential im-portance of this latter process for the global CO2 cycle. Given the un-certainties on the different geochemical mechanisms in large riverfloodplains relevant to the long-term C cycle (Bouchez et al., 2010),it is yet unclear whether these settings represent sinks or sources ofCO2 to the atmosphere.

182 J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

Acknowledgments

This work was funded by the “Reliefs de la Terre” INSU-CNRS pro-gram. The field campaigns, supported by the French Institute for Devel-opment (IRD),were realized in the frameof theHyBAm (Hydrology andGeochemistry of the Amazonian Basin, http://www.mpl.ird/hybam/)which cooperation agreement with the Brazilian Research Centre(CNPq) is no. 492685/2004-5. We sincerely acknowledge the BrazilianInstitutions and Universities (ANA, UnB, UFF, CPRM) which collaboratein this project, as well as the Bolivian (SENHAMI, UniversidadMayor deSan Andres) and Peruvian (SENHAMI) institutions which helped us toorganize the field work. Philippe Vauchel, Pascal Fraizy, Alain Craveand the staff of the ships are greatly acknowledged for their invaluablehelp. JoshWest and Albert Galy are thanked for their thorough reviews.This is IPGP contribution no. 3334.

Appendix A. Regression of X/Al–Al/Si relationshipsand uncertainties

X/Al–Al/Si relationships were fitted with distinct hyperboliccurves for the Madeira (at mouth), Marañon (at Borja and at SanRegis), Beni (at Rurrenabaque and at Riberalta), Huallaga, Morona,Pastaza, Madre de Dios and Mamoré rivers. For the rivers wheremore than two samples were available, linear correlation coefficients(R2) above 0.85 were obtained for all soluble elements considered(X=Na, K, Mg and Ca), except for Mg and Ca in the Beni atRurrenabaque. The two only sample that were excluded from the re-gressions were:

– The bed sediment sample from the Marañon at San Regis(AM-08-19), since it was difficult to achieve a good regressionfor the K/Al–Al/Si relationship when this sample was included(Fig. 4). Regressions were similar for Na/Al–Al/Si and Ca/Al–Al/Sirelationships with or without this sample, but greatly differedfor Mg/Al–Al/Si at low Al/Si values.

– A bed sediment sample from the Madeira River at mouth(AM-05-20). This sample is significantly enriched in Na and Kand depleted in Mg as compared to the general trend describedby the other samples from this site (Fig. 5). Since it was impossibleto achieve reasonable regressions when this sample was included,it was ignored for the four major soluble elements.

Bed sediments are likely to be affected by selective entrainmenteffects, and a single sample cannot account for the mineralogicaland chemical variability described by the bed of a large river. Weattribute the particular chemistry of these two bed sediments tothis variability, and consider that these samples are enriched or de-pleted in heavy minerals to an extent that is not representative ofthe sediment spectrum.

The uncertainty on _γX=Al (and on _wX , Eq. (8)) is mostly derivedfrom the uncertainties on the regressions of X/Al–Al/Si relationships.The regression coefficient being fairly high (see above), the uncer-tainty is mainly derived from the uncertainties on data points.We carried out a Monte Carlo simulation which allowed us to esti-mate that the relative uncertainty on the hyperbolic regressionparameters is lower than 15%. We use this number of 15% as a relativeuncertainty for _wX and _γX=Al, and as a benchmark to test whether _wX

is significantly different from 0% (Figs. 7 and 8). Because of (1) thelack of samples at low Al/Si ratios, and (2) the hyperbolic nature ofthe regressions in the X/Al–Al/Si space, larger uncertainties are asso-ciated at low Al/Si than at highAl/Si. This is the reasonwhywe considerthat the values of _γX=Al and of _wX are not reliable at Al/Si lower than 0.1(Figs. 7 and 8).

Regarding the grain size-integrated values, the high uncertainty at lowAl/Si ratios does not have a large impact since (Al/Si)int values are above0.2 (i.e. in the Al/Si-range where the hyperbola are well-constrained) forthe three considered reaches. The grain size-integrated chemical changes

WX and ΓX/Al are also associatedwith a relative uncertainty of 15%, andwX

is considered significant if it is higher than 15% (or lower than −15%,Fig. 8).

Additional sources of uncertainty for theweathering fluxesWX are thedepth-integrated sediment flux Fint and Al concentration [Al]int (Eq. (9)).The uncertainty on the latter parameter is ca. 40% (Bouchez et al.,2011c), and the uncertainty on the sediment flux is difficult to estimatefrom river gauging only (e.g. Guyot et al., 1996), since the fluxes relevantto weathering processes in floodplains are those over time scales compa-rable to the sediment transfer time. Wittmann et al. (2011a) reported afactor of 2 as a maximum difference between sediment fluxes derivedfrom river gauging and those calculated from cosmogenic nuclides;based upon this number, we consider that the relative uncertainty onthe sediment flux is ca. 50%. Altogether, it can be considered that thenumbers given forWX are valid within a factor of 2.

Appendix B. Grain size-integrated parameters indexing thechanges of chemical composition of river sediment

Integrating Eq. (3) over grain size S:

WX ¼ ∫S_FM⋅ _Alh iM⋅ _X

Al

!M

⋅dS−∑I

∫S_F I⋅ _Alh iI⋅ _X

Al

!I

⋅dS" #

; ðA1Þ

where

WX ¼ ∫S_WX⋅dS ðA2Þ

is the total net loss (or gain, if positive) of X by chemical weatheringfrom the sediment stored in the plain (in MT/yr). For each grainsize, the particulate flux of Al, _FAl, is equal to _F ⋅ Al½ �, which yields:

WX ¼ ∫S_FMAl⋅

_XAl

!M

⋅dS−∑I

∫S_FI

Al⋅

_XAl

!I

⋅dS" #

: ðA3Þ

The grain size-integrated X/Al ratio, (X/Al)int, is as defined byBouchez et al. (2011c):

XAl

� �int

¼∫

S_FX⋅dS

∫S_FAl⋅dS

¼∫

S_FAl⋅ _X

Al

� �⋅dS

∫S_FAl⋅dS

: ðA4Þ

Eq. (A4) simply means that (X/Al)int equals the grain size-integrated flux of X divided by the grain size-integrated flux of Al.Eqs. (A3) and (A4) can be combined:

WX ¼ XAl

� �M

int⋅∫

S_FMAl⋅dS−∑

I

XAl

� �I

int⋅∫

S_F IAl⋅dS

� �: ðA5Þ

Similar to Eq. (16), the grain size-integrated Al concentration,[Al]int, can be defined as:

Al½ �int ¼∫

S_FAl⋅dS

∫S_F ⋅dS

¼∫

S_FAl⋅dSFint

; ðA6Þ

where Fint is the grain size-integrated sediment flux (in MT/yr), ascalculated in Bouchez et al. (2011a). Combining Eqs. (A5) and (A6):

WX ¼ FMint⋅ Al½ �Mint⋅XAl

� �M

int−∑

IFIint⋅ Al½ �Iint⋅

XAl

� �I

int

� �: ðA7Þ

Eq. (A7) can be simplified as we assume that the grain size-integrated Al flux does not change downstream (Section 4):

WX ¼ FMint⋅ Al½ �int⋅ΓX=Al: ðA8Þ

183J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

References

Allison, M.A., Kuehl, S.A., Martin, T.C., Hassan, A., 1998. Importance of flood-plain sed-imentation for river sediment budgets and terrigenous input to the oceans: in-sights from the Brahmaputra–Jamuna River. Geology 26, 175–178.

Bernal, C., Christophoul, C., Darrozes, J., Soula, J.-C., Baby, P., Burgos, J., 2011. Late glacialand Holocene avulsions of the Rio Pastaza Megafan (Ecuador–Peru): frequency andcontrolling factors. International Journal of Earth Sciences 100, 1759–1782.

Berner, R.A., Lasaga, A.C., Garrels, R.M., 1984. The carbonate–silicate geochemical cycleand its effect on atmospheric carbon dioxide over the past 100 million years.American Journal of Science 284, 1183–1192.

Bonnet, M., Barroux, G., Seyler, P., Pecly, G., Moreira-Turcq, P., Lagane, C., Cochonneau, G.,Viers, J., Seyler, F., Guyot, J.-L., 2005. Seasonal links between the Amazon corridor andits floodplain: the case of the várzea of Curuaí. IAHS Publication 294, 69–77.

Bouchez, J., Beyssac, O., Galy, V., Gaillardet, J., France-Lanord, C., Maurice, L., Moreira-Turcq, P., 2010. Oxidation of petrogenic organic carbon in the Amazon floodplainas a source of atmospheric CO2. Geology 38, 255–258.

Bouchez, J., Métivier, F., Lupker, M., Gaillardet, J., France-Lanord, C., Perez, M., Maurice, L.,2011a. Prediction of depth-integrated sedimentary fluxes in large rivers: particleaggregation as a complicating factor. Hydrological Procedure 25, 778–794.

Bouchez, J., Gaillardet, J., France-Lanord, C., Maurice, L., Dutra-Maia, P., 2011b. Grain-size controls of river suspended sediment geochemistry: insights from AmazonRiver depth-profiles. Geochemistry, Geophysics, Geosystems 12, GC00380.

Bouchez, J., Lupker, M., Gaillardet, J., France-Lanord, C., Maurice, L., 2011c. How impor-tant is it to integrate river suspended sediment chemical composition with depth?Clues from Amazon River depth-profiles. Geochimica et Cosmochimica Acta 75,6955–6970.

Callède, J., Cochonneau, G., Ronchail, J., Vieira, Alves F., Guyot, J.-L., Santos, Guimarães V.,De Oliveira, E., 2010. Les apports en eau de l'Amazone à l'Océan Atlantique. Journalof Water Science 23, 247–273.

Clow, D.W., Drever, J.I., 1996. Weathering rates as a function of flow through an alpinesoil. Chemical Geology 132, 131–141.

Dessert, C., Dupré, B., Gaillardet, J., François, L.M., Allègre, C.J., 2003. Basalt weatheringlaws and the impact of basalt weathering on the global carbon cycle. ChemicalGeology 202, 257–273.

Dosseto, A., Bourdon, B., Gaillardet, J., Allègre, C.J., Filizola, N., 2006a. Time-scale andconditions of weathering under tropical climate: study of the Amazon basin withU-series. Geochimica et Cosmochimica Acta 70, 71–89.

Dosseto, A., Bourdon, B., Gaillardet, J., Maurice-Bourgoin, L., Allègre, C.J., 2006b.Weathering and transport of sediments in the Bolivian Andes: time constraintsfrom uranium-series isotopes. Earth and Planetary Science Letters 248, 759–771.

Dumont, J.F., Fournier, M., 1994. Geodynamic environment of quaternary morphostructuresof the Subandean Foreland basins of Peru and Bolivia: characteristics and studymethods. Quaternary International 21, 129–142.

Dunne, T., Mertes, L.A.K., Meade, R.H., Richey, J.E., Forsberg, B.R., 1998. Exchangesof sediment between the flood plain and channel of the Amazon River in Brazil.Geological Society of America Bulletin 110, 450–467.

Dupré, B., Gaillardet, J., Rousseau, D., Allègre, C.J., 1996. Major and trace elements of river-borne material: the Congo Basin. Geochimica et Cosmochimica Acta 60, 1301–1321.

Filizola, N., Guyot, J.-L., 2003. The use of Doppler technology for suspended sedimentdischarge determinations on the River Amazon at Óbidos. Hydrological SciencesJournal 49, 143–153.

Filizola, N., Guyot, J.-L., 2009. Suspended sediments yields in the Amazon basin: anassessment using the Brazilian national data set. Hydrological Procedure 23, 3207–3215.

France-Lanord, C., Derry, L.A., 1997. Organic carbon burial forcing of the carbon cyclefrom Himalayan erosion. Nature 390, 65–67.

Franzinelli, E., Potter, P.E., 1983. Petrology, chemistry, and texture of modern riversands, Amazon River system. Journal of Geology 9, 23–39.

Gaillardet, J., Dupré, B., Allègre, C.J., Négrel, P., 1997. Chemical and physical denudationin the Amazon River Basin. Chemical Geology 142, 141–173.

Gaillardet, J., Dupré, B., Allègre, C.J., 1999a. Geochemistry of large river suspendedsediments: silicate weathering or recycling tracer? Geochimica et CosmochimicaActa 63, 4037–4051.

Gaillardet, J., Dupré, B., Louvat, P., Allègre, C.J., 1999b. Global silicate weathering andCO2 consumption rates deduced from the chemistry of large rivers. Chemical Geology159, 2–3.

Galy, A., France-Lanord, C., 1999. Weathering processes in the Ganges–Brahmaputrabasin and the riverine alkalinity budget. Chemical Geology 159, 31–60.

Galy, V., Bouchez, J., France-Lanord, C., 2007. Determination of total organic carboncontent and δ13C in carbonate-rich detrital sediments. Geostandards andGeoanalytical Research 31, 199–207.

Galy, V., France-Lanord, C., Lartiges, B., 2008. Loading and fate of particulate organiccarbon from the Himalaya to the Ganga–Brahmaputra delta. Geochimica etCosmochimica Acta 72, 1767–1787.

Garrels, R.M., Lerman, A., Mackenzie, F.T., 1976. Controls of atmospheric O2 and CO2 —past, present and future. American Scientist 63, 306–315.

Garzanti, E., Andó, S., France-Lanord, C., Censi, P., Vignola, P., Galy, V., Lupker, M., 2010.Mineralogical and chemical variability of fluvial sediments 2. Suspended load(Ganga–Brahmaputra, Bangladesh). Earth and Planetary Science Letters 302, 107–120.

Gautier, E., Brunstein, D., Vauchel, P., Roulet, M., Fuertes, O., Guyot, J.-L., Darozzes, J.,Bourrel, L., 2007. Temporal relations between meander deformation, water dis-charge and sediment fluxes in the floodplain of the Rio Beni (Bolivian Amazonia).Earth Surface Processes and Landforms 32, 230–248.

Gibbs, R.J., 1967. Amazon River system: environmental factors that control its dissolvedand suspended load. Science 156, 1734–1737.

Granet, M., Chabaux, F., France-Lanord, C., Stille, P., Pelt, E., 2007. Time-scales of sedi-mentary transfer and weathering processes from U-series nuclides: clues fromthe Himalayan rivers. Earth and Planetary Science Letters 261, 389–406.

Granet, M., Chabaux, F., France-Lanord, C., Stille, P., Pelt, E., 2010. U-series disequilibriain suspended river sediments and implication for sediment transfer time in alluvialplains: the case of the Himalayan rivers. Geochimica et Cosmochimica Acta 74,2851–2865.

Guyot, J.-L., Filizola, N., Quintanilla, J., Cortez, J., 1996. Dissolved solids and suspendedsediment yields in the Rio Madeira basin, from the Bolivian Andes to the Amazon.IAHS Publication 236, 55–63.

Guyot, J.-L., Jouanneau, J.-M., Wasson, J.G., 1999. Characterisation of river bed andsuspended sediments in the RioMadeira drainage basin (Bolivian Amazonia). Journalof South American Earth Sciences 12, 401–410.

Guyot, J.-L., Bazan, H., Fraizy, P., Ordonez, J.J., Armijos, E., Laraque, A., 2007a. Suspendedsediment yields in the Amazon basin of Peru: a first estimation. IAHS Publication314, 1–8.

Guyot, J.-L., Jouanneau, J.-M., Suares, L., Boaventura, G.R.,Maillet, N., Lagane, C., 2007b. Claymineral composition of river sediments in the Amazon Basin. Catena 71, 340–356.

Johnsson, M.J., Meade, R.H., 1990. Chemical weathering of fluvial sediments duringalluvial storage: the Macuapanim Island point bar, Solimões River, Brazil. Journalof Sedimentary Petrology 60, 827–842.

Laraque, A., Bernal, C., Bourrel, L., Darrozes, J., Christophoul, F., Armijos, E., Fraizy, P.,Pombosa, R., Guyot, J.L., 2009. Sediment budget of the Napo River, Amazon basin,Ecuador and Peru. Hydrological Procedure 23, 3509–3524.

Lauer, J.W., Parker, G., 2008. Net local removal of floodplain sediment by river meandermigration. Geomorphology 96, 123–149.

Lupker, M., France-Lanord, C., Galy, V., Lavé, J., Bouchez, J., Métivier, F., Gaillardet, J.,Lartiges, B., 2011. A Rouse-based method to integrate the chemical composition ofriver sediments: application to the Ganga basin. Journal of Geophysical Research,Earth Surface 116, F04012.

Lupker, M., France-Lanord, C., Lavé, J., Gaillardet, J., Gajurel, A.P., Guilmette, C., Rahman,M., Singh, S.K., Sinha, R., 2012. Predominant floodplain over mountain weatheringof Himalayan sediments (Ganga Basin). Geochimica et Cosmichimica Acta 84,410–432.

Markewitz, D., Davidson, E.A., Figueiredo, R. de O., Victoria, R.L., Krusche, A.V., 2001.Control of cation concentrations in stream waters by surface soil processes in anAmazonian watershed. Nature 410, 802–805.

Martinelli, L.A., Victoria, R.L., Dematte, J.L.I., Richey, J.E., Devol, A.H., 1993. Chemical andmineralogical composition of Amazon River floodplain sediments, Brazil. AppliedGeochemistry 8, 391–402.

Maurice-Bourgoin, L., Bonnet, M.-P., Martinez, J.-M., Kosuth, P., Cochonneau, G.,Moreira-Turcq, P., Guyot, J.-L., Vauchel, P., Filizola, N., Seyler, P., 2007. Temporaldynamics of water and sediments exchanges between the Curuaí floodplain andthe Amazon River, Brazil. Journal of Hydrology 335, 140–156.

Meade, R.H., Dunne, T., Richey, J.E., Santos, U. de M., Salati, E., 1984. Storage andremobilization of suspended sediments in the lower Amazon River of Brazil.Science 228, 488–490.

Meade, R.H., 1994. Suspended sediments of the modern Amazon and Orinoco rivers.Quaternary International 21, 29–39.

Mertes, L.A.K., Meade, R.H., 1985. Particle size of sands collected from the bed of theAmazon River and its tributaries in Brazil during 1982–84. USGS Open File Report,pp. 85–333.

Mertes, L.A.K., Dunne, T., Martinelli, L., 1996. Channel-floodplain geomorphology alongthe Solimões–Amazon River, Brazil. Bulletin of the Geological Society of America108, 1089–1107.

Meybeck, M., Ragu, A., 1996. River discharge to the oceans. An assessment of suspendedsolids, major ions and nutrients. Environment Information and Assessment UNEP/WH. UNEP, Nairobi.

Milliman, J.D., Syvitski, J.P.M., 1992. Geomorphic/tectonic control of sediment discharge tothe ocean: the importance of small mountainous rivers. Journal of Geology 100,525–544.

Morse, J.W., Arvidson, R.S., 2002. The dissolution kinetics of major sedimentary carbonateminerals. Earth-Science Reviews 58, 51–84.

Moquet, J.-S., Crave, A., Viers, J., Seyler, P., Armijos, E., Bourrel, L., Chavarri, E., Lagane, C.,Laraque, A., Lavado Casimiro, W.S., Pombosa, R., Noriega, L., Vera, A., Guyot, J.-L.,2011. Chemical weathering and atmospheric/soil CO2 uptake in the Andean andForeland Amazon basins. Chemical Geology 287, 1–26.

Mortatti, J., Probst, J.-L., 2003. Silicate rock weathering and atmospheric/soil CO2

uptake in the Amazon basin estimated from river water chemistry: seasonal andspatial variations. Chemical Geology 197, 177–196.

Putzer, H., 1984. The geological evolution of the Amazon basin and its mineralresources. In: Sioli, H. (Ed.), The Amazon: Limnology and Landscape Ecology of aMighty Tropical River and Its Basin. Dr. W. Junk, Dordrecht, pp. 15–46.

Raymo, M.E., Ruddiman, W.F., 1992. Tectonic forcing of late Cenozoic climate. Nature359, 117–122.

Riebe, C.S., Kirchner, J.W., Finkel, R.C., 2003. Long-term rates of chemical weatheringand physical erosion from cosmogenic nuclides and geochemical mass balance.Geochimica et Cosmochimica Acta 67, 4411–4427.

Roddaz, M., Viers, J., Brusset, S., Baby, P., Hérail, G., 2005. Sediment provenances anddrainage evolution of the Neogene Amazonian foreland basin. Earth and PlanetaryScience Letters 239, 57–78.

Richey, J.E., Mertes, L.A.K., Dunne, T., Victoria, R.L., Forsberg, B.R., Tancredi, A.C.N.S.,Oliveira, E., 1989. Sources and routing of the Amazon River flood wave. GlobalBiogeochemical Cycles 3, 191–204.

Sayles, F.L., Mangelsdorf Jr., P.C., 1977. The equilibration of clay minerals with seawater:exchange reactions. Geochimica et Cosmochimica Acta 41, 951–960.

184 J. Bouchez et al. / Chemical Geology 332–333 (2012) 166–184

Vigier, N., Bourdon, B., Turner, S., Allègre, C.J., 2001. Erosion timescales derived fromU-decayseries measurements in rivers. Earth and Planetary Science Letters 193, 549–563.

West, A.J., Bickle, M., Collins, R., Brasington, J., 2002. Tectonic and climatic controls onsilicate weathering. Geology 30 (4), 355–358.

West, A.J., Galy, A., Bickle, M., 2005. Tectonic and climatic controls on silicateweathering. Earth and Planetary Science Letters 235, 211–228.

Wittmann, H., von Blanckenburg, F., Maurice, L., Guyot, J.-L., Filizola, N., Kubik, P.W.,2011a. Sediment production and delivery in the Amazon River basin quantified

by in situ-produced cosmogenic nuclides and recent river loads. Geological Societyof America Bulletin 123, 934–950.

Wittmann, H., von Blanckenburg, F., Maurice, L., Guyot, J.-L., Filizola, N., Kubik, P.W.,2011b. Recycling of Amazon floodplain sediment quantified by cosmogenic 26Aland 10Be. Geology 39, 467–470.