Scholarly article on topic 'Silicate weathering in the Ganges alluvial plain'

Silicate weathering in the Ganges alluvial plain Academic research paper on "Earth and related environmental sciences"

CC BY
0
0
Share paper
Keywords
{"silicon isotopes" / "silicate weathering" / "dissolved silica" / Ganges / "lowland weathering"}

Abstract of research paper on Earth and related environmental sciences, author of scientific article — Patrick J. Frings, Wim Clymans, Guillaume Fontorbe, William Gray, Govind J. Chakrapani, et al.

Abstract The Ganges is one of the world's largest rivers and lies at the heart of a body of literature that investigates the interaction between mountain orogeny, weathering and global climate change. Three regions can be recognised in the Ganges basin, with the Himalayan orogeny to the north and the plateaus of peninsular India to the south together delimiting the Ganges alluvial plain. Despite constituting approximately 80% of the basin, weathering processes in the peninsula and alluvial plain have received little attention. Here we present an analysis of 51 water samples along a transect of the alluvial plain, including all major tributaries. We focus on the geochemistry of silicon and its isotopes. Area normalised dissolved Si yields are approximately twice as high in rivers of Himalaya origin than the plain and peninsular tributaries (82, 51 and 32 kmol SiO2 km−2 yr−1, respectively). Such dissolved Si fluxes are not widely used as weathering rate indicators because a large but variable fraction of the DSi mobilised during the initial weathering process is retained in secondary clay minerals. However, the silicon isotopic composition of dissolved Si (expressed as δ 30Si) varies from + 0.8 ‰ in the Ganges mainstem at the Himalaya front to + 3.0 ‰ in alluvial plain streams and appears to be controlled by weathering congruency, i.e. by the degree of incorporation of Si into secondary phases. The higher δ 30Si values therefore reflect decreasing weathering congruency in the lowland river catchments. This is exploited to quantify the degree of removal using a Rayleigh isotope mass balance model, and consequently derive initial silica mobilisation rates of 200, 150 and 107 kmol SiO2 km−2 yr−1, for the Himalaya, peninsular India and the alluvial plain, respectively. Because the non-Himalayan regions dominate the catchment area, the majority of initial silica mobilisation from primary minerals occurs in the alluvial plain and peninsular catchment (41% and 34%, respectively).

Academic research paper on topic "Silicate weathering in the Ganges alluvial plain"

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

& PI .AN I "l'A li Y

Silicate weathering in the Ganges alluvial plain

Patrick J. Fringsa *. Wim Clymansa, Guillaume Fontorbea, William Grayb, Govind J. Chakrapanic, Daniel J. Conleya, Christina De La Rochad1

a Department of Geology, Lund University, Sölvegaten 12,223 62, Lund, Sweden

b Department of Earth Science and Marine Science Institute, University of California, Santa Barbara, CA 93106, USA c Department of Earth Sciences, Indian Institute of Technology, Roorkee, India d IUEM, CNRS UMR 5639, University of West Brittany, Plouzane, France

CrossMark

A R T I C L E I N F 0

Article history:

Received 17 February 2015

Received in revised form 20 June 2015

Accepted 24 June 2015

Available online 16 July 2015

Editor: M. Frank

Keywords: silicon isotopes silicate weathering dissolved silica Ganges

lowland weathering

A B S T R A C T

The Ganges is one of the world's largest rivers and lies at the heart of a body of literature that investigates the interaction between mountain orogeny, weathering and global climate change. Three regions can be recognised in the Ganges basin, with the Himalayan orogeny to the north and the plateaus of peninsular India to the south together delimiting the Ganges alluvial plain. Despite constituting approximately 80% of the basin, weathering processes in the peninsula and alluvial plain have received little attention. Here we present an analysis of 51 water samples along a transect of the alluvial plain, including all major tributaries. We focus on the geochemistry of silicon and its isotopes. Area normalised dissolved Si yields are approximately twice as high in rivers of Himalaya origin than the plain and peninsular tributaries (82, 51 and 32 kmol SiO2 km-2yr-1, respectively). Such dissolved Si fluxes are not widely used as weathering rate indicators because a large but variable fraction of the DSi mobilised during the initial weathering process is retained in secondary clay minerals. However, the silicon isotopic composition of dissolved Si (expressed as 530Si) varies from +0.8%o in the Ganges mainstem at the Himalaya front to +3.0%o in alluvial plain streams and appears to be controlled by weathering congruency, i.e. by the degree of incorporation of Si into secondary phases. The higher 530Si values therefore reflect decreasing weathering congruency in the lowland river catchments. This is exploited to quantify the degree of removal using a Rayleigh isotope mass balance model, and consequently derive initial silica mobilisation

respectively. Because the non-Himalayan regions dominate the catchment area, the majority of initial silica mobilisation from primary minerals occurs in the alluvial plain and peninsular catchment (41% and 34%, respectively).

© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license

(http://creativecommons.org/licenses/by/4.0/).

rates of 200, 150 and 107 kmol SiO2 km 2yr , for the Himalaya, peninsular India and the alluvial plain

1. Introduction

On geological timescales, Earth's climate is regulated by a balance between silicate weathering reactions that consume atmospheric CO2 and a continuous input of carbon from volcanic and metamorphic degassing (Walker et al., 1981). Given that degassing rates vary on long timescales but climate has remained broadly stable, a negative feedback between the rate of CO2 removal and atmospheric CO2 concentrations must exist (Berner and Caldeira, 1997). The dependency of silicate weathering rates on climate is the strongest contender for such a feedback, although the exact nature of this dependency remains elusive. River geochemistry is an

* Corresponding author.

E-mail address: patrick.frings@geol.lu.se (P.J. Frings). 1 Presently no affiliation.

integrative function of catchment solute inputs and biogeochem-ical cycling, and so is an excellent tool with which to quantify and understand weathering rates, provided inputs of solutes from non-silicate sources, biological cycling and human activity can be successfully de-convolved.

Driven by the hypothesis that the long-term global cooling since the early Eocene, illustrated, for example, by marine sediment records (Zachos et al., 2001) can be attributed to the Himalayan orogenesis over the same period (Raymo and Ruddiman, 1992), many investigations focus on the major and trace element and isotope geochemistry of the mountainous headwaters of large rivers. As weathering processes in these regions become better understood their role as important long-term CO2 sinks is increasingly questioned due to their apparently modest silicate weathering rates and small surface areas (see e.g. Moore et al., 2013). However, there is also a growing awareness that lowland regions

http://dx.doi.org/10.1016/j.epsl.2015.06.049

0012-821X/© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

of large rivers potentially play an important role in the generation of river weathering fluxes. The deposition and active reworking of freshly eroded and highly weatherable material in a system with water and sediment residence times substantially longer than the upstream source areas creates potential for additional silicate weathering and CO2 consumption (Bouchez et al., 2012; Lupker et al., 2012). Unfortunately, the capacity of the lowland regions to modify weathering fluxes at the whole-basin scale is poorly understood. For example, the Ganges is one of the world's largest river systems, lies at the heart of the uplift-climate hypothesis (Raymo and Ruddiman, 1992), and has a vast alluvial plain in the foreland basin of the Himalaya. Yet few studies have attempted to budget silicate weathering in the alluvial plain although it constitutes the majority of the Ganges basin. Those that have are often complicated by the anthropogenically perturbed solute and sediment budgets, or the presence of evaporite soils (Rengarajan et al., 2009).

Examination of the biogeochemical cycling of silicon (Si) can provide insights into weathering processes in the Ganges alluvial plain (hereafter 'GAP') because Si derives ultimately - and exclusively - from silicate minerals. Indeed, the flux of dissolved Si (DSi) from a catchment has often been used as a surrogate for silicate weathering rates (Edmond and Huh, 1997; Rengarajan et al., 2009; White and Blum, 1995). However, DSi fluxes are not widely used since a large but variable amount of DSi mobilised during solu-bilisation of primary minerals typically ends up locked into clay minerals or cycling biologically, meaning what is observed in a river does not necessarily directly reflect the initial weathering. If we are to quantify weathering fluxes, the information we really want is the rate at which Si is mobilised from primary minerals, before some fraction of it is incorporated into secondary clays and biogenic silica. This could be further converted into a silicate-cation weathering rate if the Si/cation ratio of the parent material is known (although the fraction of silicate hydrolysis driven by inorganic acids and the fraction of silicate-hosted cations incorporated into secondary clays must be accounted for by other methods). Previous work on Si geochemistry in the Ganges basin has focused mostly on the behaviour of Si isotopes in the upper end of the basin (Fontorbe et al., 2013), or in groundwater of the Ganges-Brahmaputra delta (Georg et al., 2009); knowledge of Si geochemistry and its relation to silicate weathering in the GAP is limited.

Here, based on samples from 51 sites in the Ganges basin, we use an approach that exploits the fractionation of silicon isotopes during removal of Si from solution to constrain initial DSi mobilisation rates. Our results show that a majority of initial Si mobilisation occurs in lowland regions of the catchment, albeit at lower surface area-normalised rates than the Himalaya. Overall, this suggests that any role the Himalayan orogeny plays as a driver and determinant of global climate change has been modulated by the evolution and functioning of its alluvial plains over time.

2. Study area

The Ganges (Fig. 1) drains a basin of ~0.98 x 106 km2 and has a mean annual discharge in Bangladesh of ~490 x 109 m3 yr-1. Conventionally the Ganges basin is subdivided into three broad physiographic regions: the Himalayan orogeny to the north, the hills and plateaus of Peninsular India to the south, with the GAP lying between. Climate within the Ganges basin varies widely, from tropical/sub-tropical in the south to areas with sub-zero mean annual temperatures in the high Himalaya (Tripathy and Singh, 2010). The most important hydrological aspect is the annual monsoon which occurs progressively from the southwest between approximately June and September and provides ~80% of the rainfall in the basin (Dalai et al., 2002; Tripathy and Singh, 2010). Rainfall is

not evenly divided between the three regions; the Himalaya have average runoff of ~100 cmyr-1, Peninsular India ~30 cmyr-1 and the GAP ~46 cmyr-1.

Ganges catchment streams draining the Himalaya are well studied (e.g. Bickle et al., 2003; Chakrapani, 2005; Dalai et al., 2002; Fontorbe et al., 2013; Galy and France-Lanord, 1999) and are generally characterised by rapid physical and chemical erosion rates, weathering fluxes dominated by carbonate weathering and minor contributions from silicate weathering and hot springs. Besides the headwaters of the Ganges itself, major tributaries draining into the Ganges from the Himalayan orogeny include the Yamuna, Ram-ganga, Ghaghara (also known as the Karnali), the Gandak (the Narayani) and the Kosi. The Himalaya account for 17% of the total surface area of the basin, and 34% of total runoff, and are thought to supply ~44% of dissolved Si (Galy and France-Lanord, 1999).

The southern part of the Ganges basin partly drains a plateau of the Bundelkhand crystalline granites and Vindhyan Precambrian, shales, sandstones and sedimentary carbonates. Together, they constitute a peripheral cratonic bulge that forms part of the Indian shield. The south-western basin also partly drains the Deccan Traps, solidified flood basalts from the late Cretaceous. The shield is relatively low-elevation and is covered in part by deeply weathered soils (laterites) as well as saline and alkaline soils. As a result, the headwaters of southern tributaries of the Ganges have different chemistries compared to the Himalayan streams, e.g. their cation compositions are often dominated by Na+ (Rengarajan et al., 2009) instead of Ca2+. Major rivers draining peninsular India include the Chambal, Betwa and Ken, tributaries to the Yamuna with their headwaters in the Deccan Traps, and the Tons (also known as the Tamas) and the Son (the Sone) (Fig. 1). Peninsular India accounts for ~31% of the Ganges basin surface area, and 19% of the runoff. Galy and France-Lanord (1999) estimate that peninsular catchments supply ~30% of total exported DSi.

The GAP is essentially the result of the infilling of the active foreland basin by fluvial sediments, of both Himalayan and cratonic origin, up to several kilometres thick. The ~300 km wide GAP lies in a south-easterly direction and consists of massive beds of clays, sand and gravels, mostly Himalayan, and currently aggrades by ~65 x 106 tonsyr-1 (Sinha, 2005). It is a region of low elevation, limited topography and high population density. Many streams drain the GAP directly; the largest of these is the 900 km long Gomati (also referred to as the Gomti), which originates ~50 km from the Himalaya and drains an interfluve between the Ramganga and the Ghaghara. Other notable tributaries include the Punpun from the south and the Buri Gandak from the north. The GAP accounts for 52% of the total surface area of the Ganges basin. The alluvial plain supplies about 33% of total Ganga discharge and ~26% of DSi (Galy and France-Lanord, 1999), and disproportionately contribute Na, Cl and Sr (Rai et al., 2010).

An unusual feature of the GAP is the presence of saline/alkali soil salts and carbonates, locally called kankar, that are thought to form through repeated annual wetting and drying of depressions or seasonally endorheic areas. Locally, they can strongly influence the water chemistry but their composition is poorly constrained (Galy and France-Lanord, 1999; Rengarajan et al., 2009; Sarin et al., 1989). Large-scale human activity has altered the hydrology of the region through irrigation, including the construction of canals, groundwater abstraction and damming for hydropower or water management, although this is still somewhat limited in scope.

3. Materials and methods

3.1. Fieldwork

Samples were collected at 51 sites within the Ganges fluvial network, including 19 on the Ganges mainstem itself. They span

30°N-

27°N-

24° N-

Yamuna

Farra kah

f ) Bangladesh and the Bay of Bengal

^Calcutta The Bay of Bengal

10000 11300

■ Mainstem/Major tributary □ Minor tributary/water body • Significant town/city ► River {Discharge, xl&m3yr')

Fig. 1. Representations of the Ganges basin, including the major tributaries and sampling points. A: The major tributaries and their drainage basins, derived using GIS capabilities from SRTM topographic data (Jarvis et al., 2008). B: A schematic of the hydrographic network showing the location of samples on the major tributaries (filled red squares) and minor streams or water bodies (open red squares). For reference, large cities are indicated on both representations. The discharges of the major tributaries are given in italics in 10 m3yr-1. The uncoloured areas in panel A we assume are direct, unaccounted for, alluvial plain inputs to the Ganges. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

from the base of the Himalaya at Rishikesh to the Hooghly distributary system at Calcutta (Fig. 1) and include all major tributaries of Himalayan, shield and GAP origin, plus several small streams directly draining the active floodplains. In some cases (the Yamuna, Ramganga and Chambal), tributaries were sampled at multiple distances from their confluence with the Ganges. The mainstem and a tributary were generally sampled independently just upstream of their confluence, and then approximately 20 km downstream to allow quantification of their mixing. Sampling oc-

curred between the 10th and 28th September 2013, at the end of the annual flood pulse. The channel distance upstream of the mouth of the Hooghly in the Bay of Bengal and the catchment delineations were defined using a river network derived within the geographical information system (GIS) software ESRI ArcGIS 10.3 using the shuttle radar topography mission (SRTM) digital elevation model (DEM) at a 3 arc-second resolution (Fig. 1).

At each site, 5 L surface water (~0.25 m) samples were collected from a boat positioned near the quickest flowing water, or

directly from the bank or a bridge when no boat was available. When a boat was available, water was also collected at selected points in the water column with a ballasted 3-L Van Dorn horizontal water sampler (or, for the first seven sites, a ballasted 5-L polyethylene (PE) container). Samples were stored in 5-L pre-rinsed PE containers and filtered within a few hours of collection through 142 mm diameter, 0.4 jam pore-size polycarbonate membrane filters using a Teflon-coated pressure-filtration unit (MaxFil) at a maximum of ~0.42 x 106 Pa (ca. 60 psi). A 100 ml sample from this filtration was retained in an HDPE bottle for analysis of major anions and silicon isotope composition (530Si) of the dissolved Si. For the surface waters (and deeper waters, when taken), a subsample of 30 ml was filtered immediately after collection into HDPE bottles using a syringe and a 0.4 jam pore cartridge filter (Millipore Sterivex) and acidified by addition of ~50 jal of 65% w/w HNO3 for the analysis of Si and major cations. All bottles were sealed with Parafilm and stored in the dark at ambient temperatures.

3.2. Major ion geochemistry

Major anions were measured by ion-chromatography at the Department of Biology, Lund University while Si and major cations were measured with a Varian 720 ICP-AES, at the Department of Geography, University College London, with estimated precision <5%.

3.3. Silicon isotope geochemistry

All isotope ratios are presented in delta notation (530Si, in per mil), relative to the standard NBS28:

5 (in %o) =

— Rstd

x 1000,

where Rsam and Rstd are the ratios of the heavy (30Si or 29 Si) to the light isotope (28Si) in the sample and in NBS28, respectively.

Silicon isotope ratios of DSi were measured on a Thermo Finni-gan Neptune MC-ICP-MS at the Pole Spectrométrie Océan (PSO, Ifremer, Brest). Sample purification followed a two-step procedure outlined in Frings et al. (2014). First, silicon was precipitated with triethylamine molybdate ("TEA-Moly") to form a triethylamine sil-icomolybdate complex, which was collected by filtration and combusted for 10 h at 1000 ° C to form a nearly-pure SiO2 polymorph (a mixture of cristobalite and tridymite) (De la Rocha et al., 1996). Second, this silica was dissolved in HF at an Si:F ratio of 1:100 before anion exchange column chemistry following Engstrom et al. (2005) to eliminate remaining contaminants. The silicon was eluted at 11.2 ppm in a 0.15 M HNO3 and 5.5 mM HF matrix. Column yields were assessed colourmetrically with matrix-matched standards. The solution was diluted to 2 ppm Si with 2% nitric acid before doping to 0.1 ppm Mg for the mass-spectrometry. All acids were Merck Suprapur grade and all dilutions were made with MilliQ deionised distilled water (18.2 Mfl cm-1)

530Si was assessed on the 2 ppm Si solution on the Neptune operating in dry-plasma mode with an Apex desolvating nebulizer for sample introduction. The Neptune was operating in medium resolution (resolving power, m/Am & 4000), and samples were bracketed with precisely matrix-matched standards. Any individual measurements with sample beam intensities outside ±15% of the bracketing standards were rejected. Mg isotopes were measured simultaneously to monitor and correct for instrumental mass-bias following Cardinal et al. (2003). Internal measurement precisions are typically <0.1%o. A long-term precision (1a SD), including column chemistry, was ±0.07%o, based on 29 separate analyses of an NBS28 preparation between July 2009 and November 2013.

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

630Si (%o)

Fig. 2. Three-isotope plot (529Si vs. 530Si) for all silicon isotope data points included in this manuscript. The calculated regression line (529Si = 0.512S30Si; dashed line) is indistinguishable from a theoretical kinetic fractionation line (529Si = 0.5092 x 530Si), emphasising the successful removal of polyatomic interferences during mass spectrometry.

A conservative value of ±0.1 %o is used as the uncertainty estimate in this manuscript, unless the internal precision associated with an individual measurement is larger. All results plot on a mass-dependent fractionation line 529 Si = 0.512 x 530 Si in a three-isotope plot (Fig. 2), indistinguishable from the theoretical kinetic gradient, emphasising the successful removal of polyatomic interferences during measurement. Machine accuracy was assessed in a separate analytical session using the secondary standards Diatomite and Big-Batch prepared following the same protocols. These yielded values in agreement with the results from an inter-lab comparison (Reynolds et al., 2007), i.e. +1.20%o vs. +1.26 ± 0.1 %o for Diatomite (n = 3) and -10.48%o vs. -10.48 ± 0.27%o for Big Batch (n = 2).

4. Results

4.1. Water geochemistry

Silicon and major ion concentrations and 530Si values are presented in Supplementary Table 1. Dissolved Si (DSi) concentrations in the Ganges range from 72 to 332 ^M (Table 1; Fig. 3) with a mean of 135 ^M, just below the global mean river water concentration of ca. 160 iaM (Dürr et al., 2011). There is some longitudinal variation in the Ganges mainstem, with concentrations increasing from —75 iaM at the base of the Himalaya to a plateau of —135 iaM around the Allahabad-Varanasi region, before decreasing to —80 iaM at Calcutta (Fig. 3a), almost identical to that seen previously (e.g. Fontorbe et al., 2013). Downstream variation occurs relatively smoothly, with step changes at confluences. This pattern is not mirrored in the Yamuna mainstem, which exhibits relatively constant [DSi] of —150 ^M. Of the major tributaries, the eastern Himalayan tributaries (the Ghaghara, Gandak and Kosi) have similar DSi concentrations of —90 ^M, while the non-Deccan southern tributaries (the Ken, Son and Tons) are slightly elevated, at —150 iaM. The Ramganga, a partially Himalayan river that drains the Siwaliks only, has higher [DSi] (190 and 184 ^M) that are closer to concentrations in samples from the rivers draining the Deccan traps (Chambal, Betwa; mean = 219 ^M) and of small floodplain systems (variable, but up to >300 ^M), perhaps reflecting that the majority of the Ramganga catchment lies within the GAP.

With a mean concentration of 1210 ^M, Ca2+ is the dominant cation in all the streams sampled here. On average, Ca2+ is 64% of cations (by concentration), dropping below 50% only at four

Table 1

Measured concentrations of Si and other major elements in waters of the Ganges basin at the most downstream point of each respective system, and calculated fluxes (mass time-1) and yields (mass area-1 time-1) based on measured concentrations, derived areas (Fig. 1) and literature discharge values. Total yields are area weighted.

Subbasin fQa Concentration (observed) Fluxes Yield

(-) (|iM) (Gmolesyr-1) (kmolkm-2yr-1)

Si Mg Ca Na K Sr Si Mg Ca Na K Sr Si Mg Ca Na K Sr

Northern Himalayan

Ganges at Rishikesh 0.052 80 107 689 70 122 2.6 1.9 2.6 16.5 1.7 2.9 0.1 88 117 756 77 134 2.9

Ramganga 0.034 184 396 1982 423 295 13.8 2.9 6.2 31.0 6.6 4.6 0.2 108 231 1158 247 172 8.1

Ghaghara 0.206 85 200 1263 124 121 6.8 8.0 18.9 119.2 11.7 11.4 0.6 55 128 810 80 78 4.4

Gandak 0.114 72 173 1066 65 135 5.7 3.8 9.0 55.7 3.4 7.1 0.3 75 181 1116 68 141 6.0

Kosi 0.149 98 75 725 85 101 2.1 6.7 5.1 49.6 5.8 6.9 0.1 103 79 759 89 106 2.2

Yamuna at Him. Front 0.023 154 314 1663 636 195 17 1.7 3.4 17.9 6.8 2.1 0.2 172 351 1863 712 219 19.0

Yamuna and Tributaries

Yamuna at Allahabad 0.203 152 229 1417 570 128 17.4 14.1 21.3 131.8 53.0 11.9 1.6 35 53 329 132 30 4.0

Chambal 0.065 208 295 1168 529 96 20.7 6.2 8.9 35.1 15.9 2.9 0.6 39 56 220 100 18 3.9

Betwa 0.022 215 233 1341 503 76 16.1 2.2 2.3 13.4 5.0 0.8 0.2 41 45 257 96 15 3.1

Ken 0.025 192 200 1418 350 56 13.2 2.2 2.3 16.0 4.0 0.6 0.2 64 67 474 117 18 4.4

Unaccounted forb 0.067 152 269 1379 602 168 16 4.7 8.3 42.6 18.6 5.2 0.5 32 57 292 128 36 3.4

Southern Peninsular

Tons 0.013 110 167 1216 232 66 12.4 0.7 1.0 7.2 1.4 0.4 0.1 32 49 356 68 19 3.6

Son 0.069 157 121 801 201 56 4.6 5.0 3.9 25.5 6.4 1.8 0.2 65 50 332 83 23 1.9

Alluvial Plain

Gomati 0.016 111 316 1186 542 136 12.2 0.8 2.3 8.8 4.0 1.0 0.1 23 65 245 112 28 2.5

Buri Gandak 0.015 131 163 1508 188 135 7.3 0.9 1.2 10.7 1.3 1.0 0.1 22 28 255 32 23 1.2

Unaccounted forb 0.129 152 269 1379 602 168 16 9.0 16.0 81.9 35.7 10.0 1.0 85 151 773 337 94 9.0

Weighted averages

Estimated at Farrakah 123 199 1156 280 128 9

Observed at Farrakah 98 145 1068 186 123 7

Discrepancy (%) 21 27 8 33 4 25

Total fluxes andyieldsc

Himalayan rivers 24.9 45.2 289.7 36.0 35.0 1.5 82 140 915 115 114 5

Southern Peninsular 16.2 18.3 97.2 32.6 6.5 1.1 32 59 303 125 36 3

Alluvial Plain 15.5 27.8 144.0 59.7 17.1 1.6 51 53 305 93 20 3

a fQ = fraction of total discharge at Farrakah, based on compilation in Jain et al. (2007). fQ sums to more than unity because the Yamuna tributaries are included twice; this has been accounted for in the flux and yield calculations. Unaccounted for discharge refers to the fraction of discharge for a given point on the mainstem or Yamuna that is in excess of the sum of all the upstream tributaries with available discharge data, and is assumed to derive from the areas not part of one of the tributaries as defined in Fig. 1.

b Major element geochemistry taken as mean of alluvial plain water samples (supplementary Table 1).

c Himalayan rivers are defined as the sum of Yamuna at Delhi, Ganges at Rishikesh, and Ramganga, Ghaghara, Gandak and Kosi at their confluence with the Ganges, which implicitly blurs the distinction between Himalaya and GAP. Southern Peninsular rivers are the sum of the Chambal, Betwa, Ken, Tons and Son and Alluvial Plain rivers are the sum of the Gomati and Buri Gandak, plus the discharge at Farrakah unaccounted for by the above rivers.

0 Data from Fontorbe et al. (2013)

★ Ganges mainstem

• Yamuna mainstem

□ Major tributaries of Yamuna

A Southern tributaries

▼ Northern tributaries

& 0 Alluvial plain waters

Major alluvial plain tributaries

3000 2000 1000 0

Distance upstream from Bay of Bengal (km)

3000 2000 1000 0

Distance upstream from Bay of Bengal (km)

Fig. 3. Downstream variation in A: dissolved Si (DSi) concentrations (|iM) and B: silicon isotopic composition (S30Si) of DSi in the Ganges basin. Data from Fontorbe et al. (2013), collected mostly in mountain streams and the alluvial Gomati River, are shown for comparison.

sites. The first three of these sites correspond to closely spaced samples with unusually elevated [Na] (~1500 |aM) along the Yamuna and the fourth to the Varuna, a small, highly polluted stream draining the alluvial plain before its confluence with the Ganges at Varanasi. Bicarbonate (HCO-) is the dominant anion, with an average (calculated) concentration of 2420 |aM. The next most dominant anion after bicarbonate is Cl- with a mean concentration, after rainwater-correction (see below) of 500 |aM. Some extremely high Cl- concentrations in excess of Na (>4000 |aM Cl) occur in several Yamuna mainstem samples.

4.1.1. Solute mass balance

Taking discharges for the major rivers from Kothyari and Garde (1991) and assuming our measured concentrations are representative, and that the relative discharges between tributaries are constant, the solute budgets of the major elements (i.e. Si, Mg, Ca, Na, K and Sr) broadly balance in the Ganges basin (Table 1), with the concentrations at the lowermost sample (sample 051; Farrakah Barrage, near the Bangladeshi border) tending to lie a little below the weighted sum of the upstream inputs. Calculated fluxes (kmolyr-1) and yields (Gmolkm-2yr-1) emphasise the importance of the solutes of Himalayan origin at Farrakah. The majority of Ca, Mg and K derive from the Himalaya, while Si derives in a roughly 50/50 ratio from the Himalaya and the rest of the catchment. Na and Sr originate predominantly from non-Himalayan areas.

Dissolved Si yields in the Ganges range from a mean of 100 kmolkm-2yr-1 in the Himalayan rivers to 48 and 41 kmolkm-2yr-1 in the peninsular and alluvial plain rivers, respectively (Table 1). These values cluster around the global mean DSi yield of c. 55 kmolkm-2yr-1 (3.3 tkm-2yr-1; Dürr et al., 2011), and are in good agreement with previous work in the Go-mati and Son rivers (Rai et al., 2010).

3.5 3.0 2.5

1.0 0.5 0.0

0 100 200 300 400 Dissolved Si concentration (pM)

Fig. 4. Silicon isotope composition (S30Si) vs. dissolved Si concentrations (|iM) in Ganges basin river waters.

two joining streams, arguing against large in-stream processes impacting the solute geochemistry. One exception may be the penultimate sample, (#050), collected immediately downstream of the Farrakah Barrage, which is slightly higher than the two nearest upstream samples - the Kosi River and the Ganges upstream of its confluence with the Kosi (Fig. 1; Supplementary Table 1). This may be due to conversion of river DSi to diatom silica in the less turbid waters behind the barrage, though this is speculative.

o Fontorbe et al. (2013) if Ganges mainstem • Yamuna mainstem

R Major tributaries of Yamuna Southern tributaries

¥ Northern tributaries Alluvial plain waters O Major alluvial plain tributaries

4.2. Silicon isotope geochemistry

All DSi samples, with 530Si ranging from +0.81%o to +3.04%o (Fig. 5a), are enriched in 30Si relative to silicate rocks (530Si of mean upper continental crust (UCC) = -0.25 ± 0.16%o, Savage et al., 2013). Samples from the Ganges mainstem increase from < +1%o at base of the Himalaya to a relatively constant value of ~ + 1.7%o after ca. 500 km. This downstream increase is consistent with 530Si of DSi in the Chiangjiang and Huang He (the Yangtze and Yellow Rivers), two other large rivers draining the Himalayan-Tibetan plateau (Ding et al., 2004, 2011). Smaller rivers draining the GAP have generally higher 530Si (mean = 2%o) (Fig. 3b). Taking the samples as a whole, there is no clear relationship with [DSi] (Fig. 4) and the points fall within a roughly triangular array, and well within the range previously observed in river waters worldwide of -0.14%o to 4.66%o (mean = 1.38%o, n = 528) (Frings, 2014).

The major southern tributaries (the Chambal, Betwa, Ken, Son and Tons) have similar 530Si values, falling between +1.48%o and +2.04%o while the major northern tributaries draining the Himalaya (the Ramganga, Ghaghara, Gandak and Kosi) have slightly lower 530Si (+1.18 to +1.73%o). The highest but most variable 530Si occurs in the small streams draining the alluvial plain directly: + 1.03 to +2.38%o in the Ganges subbasin and +1.53 to +2.09%o in the Yamuna/Chambal subbasins. The highest measured value of +3.04 derives from an inundated region in the active floodplain, hydrologically isolated from the main channel. This high DSi, high 530Si value is problematic to explain but may relate to a high degree of evaporation in concert with high biogenic silica production, as seen in e.g. the Nile river (Cockerton et al., 2013). River elemental concentrations and 530Si value after a confluence are generally within error of the discharge-weighted sum of the

5. Discussion

We can conceptualise silicate weathering fluxes as the sum of two processes: the initial breakdown at the solid-solution interface, and the formation of secondary (alumino-)silicates - i.e. clays and related minerals - that removes some fraction of the new solutes. The ratio of primary mineral breakdown to secondary mineral formation can be seen as an index of weathering con-gruency (Pogge von Strandmann and Henderson, 2014). Congruent weathering produces solutes with elemental stoichiometries and isotopic compositions close to the parent material (i.e. lower 530Si), whereas incongruent weathering causes the composition of the solutes to deviate from that of the parent material (i.e. higher 530Si, as more Si is stored in clays). The use of DSi fluxes in the calculation of weathering mass-budgets has been limited to date because the incorporation of mobilised Si into secondary minerals or bio-genic silica - i.e. the degree of weathering congruency - is hard to account for. In the following, we suggest the 530Si value of river DSi primarily reflects weathering congruency, and can therefore be used to correct observed Si yields and derive initial silicate-weathering rates.

5.1. Riverine behaviour of silicon isotopes

Consideration of published river silicon isotope data supports the interpretation that a primary driver is weathering congru-ency. Unweathered silicates at the Earths surface define a relatively constrained range of 530Si - a compilation of rock 530Si values, weighted by abundance, yields a global upper continental crust (UCC) value of -0.25 ± 0.16%o (Savage et al., 2013), and conservation of mass requires that there can be no net isotopic fractionation during their dissolution (Ziegler et al., 2005),

"if Ganges mainstem o

• Yamuna mainstem #

□ Major tributaries of Yamuna x

A Southern tributaries +

T Northern tributaries Alluvial plain waters

Q Major alluvial plain tributaries

Ganges headwaters (ref 1) Amazonian rivers (ref 2) Icelandic rivers (ref 3) Congo basin rivers (ref 4) Kalix, Sweden (ref 5) Yangtze (ref 6) Tana, Kenya (ref 7)

Rayleigh model

Batch model

fr* ** *

# ^sSs» x

1 2 DSi/(Na+K)

1 2 DSi/(Na+K)*

Fig. 5. Si is increasingly removed from solution at higher solute concentrations, which imparts a greater observable fractionation. A: Increasing retention of Si into secondary phases (i.e. decreasing Si/(Na + K)*, where (Na + K)* indicates the sum of Na and K concentrations corrected for cyclic inputs; see main text for details) at increasing solute (total cation) concentrations. B: Increasing fractionation relative to bedrock with increasing Si retention (decreasing Si/(Na + K)* ). Panel B also shows expected evolutions of S30Si as DSi as it is progressively removed relative to (Na + K), for a Rayleigh model (solid grey line) and a steady-state model (dashed brown line), assuming an initial solution with S30Si = —0.25%o and Si/(Na + K) of 3.53 (Hughes et al., 2013). A range of river waters worldwide are plotted for comparison in both panels. Element concentrations have only been corrected for rainfall inputs if performed in the original publication. References: 1 - Fontorbe et al. (2013), 2 - Hughes et al. (2013), 3 - Georg et al. (2007), 4 - Cardinal et al. (2010), 5 - Engstrôm et al. (2010), 6 - Ding et al. (2004), 7 - Hughes et al. (2012). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

so variability in source rock 530Si is unlikely to impart significant variability to the 530Si of the initial solution, especially at large spatial scales. There are then two remaining pathways to altering a 530Si value of DSi. The first is to create new solid phases, either abiotically (e.g. clay precipitation from a saturated solution) or biologically (e.g. diatom growth or plant phytolith production). These processes consistently discriminate against the heavy 30Si isotope, creating lower 530Si products and leaving the residual solution with higher S30Si (De la Rocha et al., 1997; Oelze et al., 2014; Opfergelt et al., 2010; Ziegler et al., 2005). The second pathway is to add new DSi through the dissolution of isotopically distinct secondary minerals (including biogenic silica).

Since at steady-state, over an annual cycle, any dissolution of a secondary phase must be balanced by its prior production, the second pathway is unlikely to have a major effect at large scale, though small scale or transient impacts are plausible. A number of observations worldwide suggest that removal from solution (i.e. pathway 1) is the most important mechanism in determining river 530Si. The samples with the lowest values yet measured (c. 0.0%o to — 0.1%o) derive from locations that exhibit near-congruent primary mineral dissolution. These include 'blackwater' rivers in tropical environments (Cardinal et al., 2010; Frings, 2014; Hughes et al., 2013), but also some glacial rivers in Iceland (Georg et al., 2007). Correspondingly, of all reported measurements, the highest 530 Si values come from systems where a high degree of Si removal into clays or biogenic silica is expected. These include the semi-arid Nile basin (Cockerton et al., 2013) and the highly agricultural Yangtze river (Ding et al., 2004), where 530Si reaches +4.7%o and +3.4%o, respectively. For rivers with 530Si monitored over an annual cycle, there is a tendency for the highest 530Si values to coincide with the periods of lowest discharge and DSi export (Delvaux et al., 2013; Engstrôm et al., 2010; Georg et al., 2006; Hughes et al., 2011, 2013; Pokrovsky et al., 2013). This is observed in Ganges basin itself: dry-season wa-

ter samples from several rivers presented by Georg et al. (2009) are consistently offset to higher values than those we have measured on the same or nearby rivers. This implies that at greater mean water residence times, clay mineral formation (or biogenic silica production) is increased, pushing the 530Si of residual DSi higher. A similar observation can be made based on changes in 530Si along river longitudinal transects (Cardinal et al., 2010; Cockerton et al., 2013; Ding et al., 2004, 2011; Fontorbe et al., 2013; Hughes et al., 2012): as with the data presented here, there is a tendency for 530Si to increase downstream as water and sediment residence times increase. Again, the greater interaction times implied should increase the degree of clay formation as the limits to clay precipitation (and therefore Si isotope fractionation) are more likely to be overcome, whether these limitations are primarily kinetic, thermodynamic, or a mixture. Pogge von Strandmann and Henderson (2014) reach a similar conclusion with respect to lithium isotopes, based on a negative relationship between the iso-topic composition of dissolved lithium (57Li) and uplift rates in the New Zealand Alps.

5.2. Silicon geochemistry in the Ganges alluvial plain: first order controls and potential for use as a weathering index

Overall, the behaviour of silicon in the Ganges basin corroborates the above. At lower solute concentrations, silicate weathering generally becomes more congruent, i.e. more stoichiometric conversion of the available minerals to solutes. Conversely, at higher solute concentrations, the fraction of Si incorporated into secondary phases becomes larger. To a first degree, this is seen in a plot of total cations against DSi/(Na + K)* (Fig. 5a, where the asterisk denotes values corrected for rainfall solute inputs following Galy and France-Lanord, 1999 using a larger rainwater chemistry database included in the supplementary online material). Na and K are hosted predominantly in silicates, so higher DSi/(Na + K)*

(i.e. approaching source material) reflects more congruent weathering, whereas lower values reflect greater relative retention (as clays or biogenic silica) of the DSi released from dissolution of primary minerals. DSi/(Na + K)* ranges between 0.05 and 0.8, consistent with previous work in the Ganges basin (Sarin et al., 1989), but unfortunately a number of caveats to the use of (Na + K)* for normalisation preclude using this as a quantitative index of weathering congruency (cf. Hughes et al., 2013). These include i) imprecisely known bedrock Si/(Na + K) ratios (and whether or not to include the very slowly reactive quartz component), ii) the incorporation of K into secondary phases, though this is likely small in comparison to its mobilisation from the K-rich Himalayan bedrock, iii) other, non-silicate inputs of Na or K from e.g. soil-salts.

Incorporation of silicon into secondary solids by either biogenic or abiotic means imparts a silicon isotope fractionation (e.g. Oelze et al., 2014; Opfergelt et al., 2006). Correspondingly, 530Si increases with decreasing DSi/(Na + K)* (Fig. 5b), both in the Ganges and in a range of global rivers. In other words, the greater the fraction of Si retained into secondary minerals (or biogenic silica), the greater the observed deviation from a typical primary mineral value (~-0.25%o; Savage et al., 2013). This is consistent with previous Si isotope work in the Ganges basin (Fontorbe et al., 2013) which also inferred that weathering became less congruent at higher solute concentrations.

Although the dominant mechanism of DSi removal from solution is likely incorporation into neoformed clay minerals, one potentially important anthropogenically induced DSi removal mechanism may be the retention of biogenic silica remains in reservoir sediments (Frings et al., 2014), with an associated biological fractionation. However, the Ganges basin is not yet excessively dammed. The Global Reservoir and Dam (GRanD) database lists 80 reservoirs within the Ganges basin, of which only three have a surface area greater than 100 km2 (Lehner et al., 2011). The total area of natural lakes and wetlands is larger (Lehner and Doll, 2004), so any net 'natural' biogenic silica sequestration here is likely more important. The construction of large diversionary irrigation canals (Jain et al., 2007) may similarly have increased the potential for enhanced production of biogenic silica. Note that even if biological activity is the main mechanism by which DSi is removed from solution relative to Na and K, the impact on 530 Si of the residual DSi will be very similar, and our attempt below to quantify this removal is independent of the actual DSi sink.

5.2.1. A weathering index based on silicon isotope geochemistry

Experimental and theoretical evidence suggests the range of 530Si observed in nature can be mostly attributed to unidirectional kinetic isotope effects (Dupuis et al., 2015; Oelze et al., 2014), i.e. there is little evidence for equilibrium isotope fractionation between low temperature precipitates (including clays) and solutions (which should tend to leave the clays isotopically heavier). Given this kinetic dominance (i.e. mostly forward reactions), two simple conceptual models are commonly applied to the evolution of the isotopic composition of a given mass of DSi as it is removed from solution. Firstly, in a system that is closed with respect to further DSi inputs and the product has no further interactions, the 530Si of the residual DSi follows a Rayleigh distillation curve:

$ Siresidual — $ Sijnitial + ^removal • ln( f )

Where 530Siresjduai is the isotopic composition of DSi after a fraction f has been removed with a per mil enrichment, £ removal, from DSi with an initial isotopic composition 530Sijnitiai. Rearranging to find f :

where A530Si is the difference between the 530Si of the initial and residual DSi. We will refer to this as a Rayleigh model. Alternatively, the DSi can exist in an open system, such that it constantly receives 'fresh' DSi (e.g. from primary mineral weathering). If a dynamic equilibrium is reached such that removal into secondary phases balances the supply of new DSi (Reynolds et al., 2006), the isotopic composition of the residual water:

& Siresidual — & Siinitial — £ removal (1 — f ) where f is:

f — 1 - (A530Si/30£removal)

f — e(AS30Si/30£removal)

This is sometimes referred to as a 'steady-state' or 'batch equilibrium' model (Bouchez et al., 2013; Georg et al., 2007; Hughes et al., 2013), since the evolution of isotopic composition as a function of f is the same as for a reversible reaction at equilibrium in a closed system; we will refer to it as a batch model. In the case that the supply of new DSi is large relative to the removal, f & 1, 530Siresiduai & 530Sijnitial and the product is simply 530Sijnitial + 30 £removal. In both models, there are two unknowns: 530Sijnitial and 30eremoval, the per mil fractionation associated with DSi removal from solution. For the large-scale, mixed lithologies of the Ganges subbasins, a mean UCC 530Si of -0.25%o (Savage et al., 2013) can satisfactorily approximate the 530Si of the DSi initially released from primary minerals (discounting any net fractionation during dissolution). Because the DSi in any given sample reflects the integration of a number of different flowpaths of water that have experienced different histories, a single, lumped fractionation is also reasonable, and has been applied successfully at a range of scales (Bouchez et al., 2013). 30eremoval is therefore an integrative function of all DSi-consuming reactions that is likely in the region of -1.50%o based on compiled values for both biological and abiotic processes (Opfergelt and Delmelle, 2012), and in line with recent experimental work (Oelze et al., 2014).

A choice needs to be made between the two isotope evolution models. Some evidence suggests application of the batch fractionation model, as used by e.g. Bouchez et al. (2013) is inappropriate here, since the greatest allowable fractionation of DSi relative to the parent material under an open-system model with 30eremoval = - 1.50%o is +1.25%o, which is often exceeded, both in the Ganges and elsewhere. We could therefore argue that isotope fractionation for a given flowpath is dominantly driven by Rayleigh style fractionation. Fig. 5b shows the isotope evolution as a function of DSi/(Na + K)* as predicted by the two models. Visually, the Rayleigh model seems to capture the overall behaviour of global rivers. However, work on silicon isotope sys-tematics in Amazonian and Icelandic streams (Georg et al., 2007; Hughes et al., 2013) and lithium isotopes, again in the Amazon (Dellinger et al., 2015) suggests some individual rivers have a tendency to plot along a line defined by the batch model, whilst other rivers will follow the Rayleigh pathway. A more quantitative index of f is needed to discriminate between the two. Further, both models (Rayleigh or batch) strictly apply to the evolution of iso-topic compositions in a simple, discrete package of DSi. We overcome this limitation by conceptualising the DSi in any given river sample as resulting from the summation of many small flowpaths, each contributing a fraction of the total, and with its own isotopic evolution.

Interestingly, in the case of systems following the Rayleigh model, because of the mixing of multiple flowpaths, one cannot simply apply equation (3) to a river 530Si value and obtain an accurate estimate of f, our proposed index of silica-weathering congruency. The mixing of multiple flowpaths imposes the problem that f as calculated from equation (3) will not necessarily correspond to the ftrue, the 'true' fraction remaining in solution.

Rayleigh Model

Batch model

IS ¡7)

1.0 0.2 0.4 0.6 0.8 1. Fraction Si remaining in solution

0.0 0.2 0.4 0.6 0.8 1.0 Fraction Si remaining in solution

Fig. 6. Simulation of possible ranges of S30Si for a given f Si, for A: a Rayleigh-style system and B: a steady-state system. In both cases, the S30Si-fSi calibration is derived using 1500 iterations (grey dots) per step in fSi and using values of —0.25%o for S30Siinitiai and -1.5 ± 0.5%o for 30£removal (see main text for details). The shaded areas represent the 5th to 95th percentile of simulated data, while the histogram on the y-axis of panel B displays the relative frequency of measured river DSi S30Si in the Ganges basin at 0.2%o bin intervals, including data from Fontorbe et al. (2013) and Georg et al. (2009). These models can be seen as sensitivity.

This is because the observed 530Si is the flux-weighted average of the individual flowpath 530Si values, which themselves are a non-linear function of the individual flowpath's f values (cf. equation (2)):

^3°Siobs — ^^ Xi • [530Si¡n¡tial + 3°£removaI,i • ln( fi )]

for i flowpaths contributing a fraction Xi of the total Si. Conversely, ftrue is simply the (linear) weighted average of the individual f values:

ftrue — ^^ Xi • fi

This imposes the constraint that, for systems that experience Rayleigh fractionation, f inferred from a measured 530Si value via equation (3) will be less than ftrue.

To account for the departure from ideal behaviour caused by mixing of multiple flowpaths and to propagate the uncertainties associated with imprecisely known fractionations, we performed Monte-Carlo style simulations to derive the expected range of 530Si values in river water DSi for any given ftrue (Fig. 6). We achieve this using artificially generated data with n flowpaths, where each flowpath i is randomly assigned (i) a fraction of the total flux (Xi), (ii) a per mil solid-solution enrichment factor, (30ei) and (iii) a fraction remaining (fi, from 0 to 1). Xi follows an exponential distribution, to account for the concept that some flowpaths will be more important than others. For each iteration, a lumped frac-tionation factor is randomly assigned, normally distributed around a mean of — 1.50%o with a standard deviation of 0.5%o. Within that iteration, the individual flowpath 30ei values are further distributed around that lumped value, again with standard deviation = 0.50 . Note that this is a conservatively large error estimate, and that our simple approach doesn't account for any correlation between these variables. Preliminary investigation showed the approach is insensitive to the number of flowpaths included above n & 15 so in the following all simulations take place with n = 100. These simulations are repeated twice, for the Rayleigh and batch fractionation models.

By varying the mean of the fraction remaining in solution, we obtain the range of 530Si values for a corresponding mean ftrue. We achieve this by prescribing fi to follow a beta distribution (i.e. constrained between 0 and 1) with mean = ftrue. Fig. 6 shows the output of the simulations for 30eremoval = —1.50%o ± 0.50%o and 530Sijnitial = —0.25%o. The lower and upper limits of the range

if- - Ganges mainstem

• Yamuna mainstem

□ Major tributaries of Yamuna

A Southern tributaries

▼ Northern tributaries

v Alluvial plain waters

Q Major alluvial plain tributaries

o Fontorbe etal. (2013)_

3000 2000 1000 0

Distance upstream from Bay of Bengal (km)

Fig. 7. The congruency of weathering with respect to Si, assuming Rayleigh style behaviour. Weathering congruency decreases with distance downstream, i.e. the fraction of initially solubilised Si remaining in solution decreases (alternatively: the fraction incorporated into clays/biogenic silica increases). The corresponding S30Si value for a given fSi is shown on the right-hand axis; note that this scale is derived from the calibrated S30Si/fSi relationship in Fig. 5a and is not linear. Uncertainties represent the 5th and 95th percentiles of the calibrated data.

of 530Si produced for a given ftrue is taken as the 5th and 95th percentiles of 1500 individual iterations (the shaded envelopes in Figs. 6a and 6b) and includes both the uncertainty associated with imprecise knowledge of the true fractionation factor and the departure from ideal Rayleigh behaviour caused by mixing of multiple flowpaths (for the Rayleigh model only). Calibrating the observed 530Si against the mean of the simulations and these limits, f values in the Ganges basin fall between 0.54 (+0.14/—0.26) and 0.19 (+0.13/—0.13) (Fig. 7) if the rivers follow a Rayleigh model, or between 0 (i.e. near-total removal) and 0.46 if the batch model better describes their behaviour. The batch model produces results that are difficult to meaningfully interpret; all samples could imply near-total removal of Si from solution if the mean fractionation is towards the lower end of estimates. Conversely, even at the greater

Table 2

Summary of measured DSi yields and calculated initial mobilisation rates in the subbasins and the three domains of the Ganges basin (Himalaya, Plain, Southern Peninsula).

Observed Rayleigh model Batch model

River _ _-_-___

Qa x106 DSi yieldb S30Si jc 'Initial' DSi yield jc 'Initial' DSi yield

(m3yr-1) (kmolkm-2yr-1) % ± 1sd (5th to 95th (kmol km-2 yr-1) (-) (kmol km-2 yr-1)

percentile)

Himalayan Tributaries

Ganges at Rishikesh 23 900 87 1.03 ± 0.10 0.48 + 0.15/- -0.26 183 + 211/-42 <0.46 >189

Ramganga 15620 107 1.72 ± 0.10 0.34 + 0.16/- -0.21 317 + 516/-99 < 0.16 >676

Ghaghara 94400 54 1.47 ± 0.10 0.39 + 0.15/- -0.23 142 + 212/-39 <0.26 >206

Gandak 52 200 75 1.18 ± 0.10 0.44 + 0.15/- -0.25 170 + 215/-42 <0.39 >192

Kosi 68340 102 1.36 ± 0.10 0.41 + 0.14/- -0.24 253 + 351/-66 <0.30 >342

Yamuna at HF 10 750 172 1.27 ± 0.10 0.43 + 0.14/- -0.24 125 + 165/-32 <0.35 >499

Yamuna and Tributaries

Yamuna at Allahabad 93 020 35 1.72 ± 0.10 0.34 + 0.16/- -0.21 104 + 169/-33 < 0.16 >222

Chambal 30050 39 2.04 ± 0.10 0.29 + 0.15/- -0.19 134 + 256/-46 < 0.05 >1001

Betwa 10000 41 1.48 ± 0.10 0.38 + 0.15/- -0.23 108 + 162/-30 <0.26 >159

Ken 11300 64 1.73 ± 0.10 0.34 + 0.15/- -0.21 189 + 307/-59 < 0.16 >411

Unaccounted ford 30920 36 2.00 ± 0.10e 0.29 + 0.15/- -0.19 110 + 180/-35 <0.05 >620

Southern Tributaries

Tons 5910 32 2.04 ± 0.10 0.29 + 0.15/- -0.19 111 + 211/-38 <0.04 >825

Son 31 800 65 1.54 ± 0.10 0.37 + 0.15/- -0.23 175 + 270/-51 <0.24 >276

Alluvial Plain

Gomati 7390 22 1.96 ± 0.10 0.3 + 0.15/- 0.20 76 + 138/-25 <0.07 >347

Buri Gandak 7100 22 1.55 ± 0.10 0.37 + 0.15/- -0.22 60 + 93/-17 <0.23 >96

Unaccounted ford 26476 38 2.00 ± 0.10e 0.29 + 0.15/- -0.19 113 + 186/-36 <0.05 >731

Ganges at Farrakah

Farrakah Barage 459040 51 1.93 ± 0.10 0.31 + 0.15/- -0.20 170 + 303/-56 <0.08 >1077

Discharge weighted means Si mobilisation

rate (Gmolyr-1 )f

Himalayan 82 1.35 0.41 200 33.3 <0.31 >277

Alluvial Plain 32 1.95 0.30 107 54.5 <0.07 >581

Peninsular 51 1.76 0.34 150 45.6 < 0.15 >561

a From |ain et al. (2007). b From Table 1.

c Calculated as described in main text and Fig. 5, where 530Siinitial = -0.25%o and mean 30£removal = -1.50 ± 0.50%o. d As in Table 1.

e S30Si of 2.00%o from average of alluvial plain waters (see supplementary Table 1).

f Assuming the Himalaya = 17%, peninsular India = 31% and the alluvial plain = 52% of the total catchment (0.98 x 106 km2) (see Section 2).

end of fractionation estimates (30eremovai < —2%o), some samples do not overlap with the produced calibration, as shown by the 530Si-histogram in Fig. 6b. We therefore focus here primarily on the results from the Rayleigh model, but suggest that in the future, approaches that combine the two models with an index of flowpath isolation could improve our understanding.

The 530Si values interpreted using a Rayleigh model indicate between 46% and 81% of Si released during weathering is retained in secondary phases (either aluminosilicates or biogenic silica; their formation is isotopically indistinguishable). In general, these values are comparable to values derived in Iceland, the Amazon and the Orinoco, using elemental stoichiometries (Georg et al., 2007; Hughes et al., 2013; Murnane and Stallard, 1990). Although the estimated uncertainties generally considerably overlap (Fig. 7), these are largely systematic in that they reflect our imprecise knowledge of the fractionating factors and processes associated with DSi removal from solution. It is therefore still possible to meaningfully interpret the overall trend shown in Fig. 7.

The Himalayan rivers (i.e. the Ganges headwaters, Ghaghara, Gandak and Kosi) have the greatest weathering congruency (~45% DSi remains in solution), and become progressively more incon-gruent (i.e. a greater proportion of Si retained in secondary minerals plus biogenic silica) with increasing distance from the Himalayan front (Table 2 and Fig. 7). Some small Himalayan catchments reported by Fontorbe et al. (2013) have f as high as 0.65 (+0.11/—0.26) (530Si = 0.49%o), while all their alluvial plain samples have f < 0.48 (Fig. 7). Here, the GAP and southern tributaries have similar weathering congruencies, with a mean of about 30%

of initially mobilised DSi remaining in solution. We also note that the large Himalaya rivers (the Ghaghara, Gandak and Kosi; Fig. 1), at the point of sampling, have large portions, sometimes majorities, of their catchments in the GAP, blurring the distinction between mountain and lowland regions: samples from these rivers at the Himalayan Front would be beneficial but are not available. Our data show that Himalayan catchments weather more congru-ently, regardless of the fractionation model chosen. This is counter to a conventional interpretation that sees mountainous regions as areas of extremely incongruent weathering and high proportions of clay formation (e.g. Misra and Froelich, 2012). However, it is consistent with the 2D reactive transport model proposed by Fontorbe et al. (2013), in which clay formation is kinetically controlled, and also with the behaviour of lithium isotopes in uplifting areas (Pogge von Strandmann and Henderson, 2014). It is also consistent with mineralogical analyses of sediment carried by the Ganges in Himalayan streams: the clay size fraction is dominated (~80%) by detrital micas (Chakrapani et al., 1995), a major group of rock forming minerals. Only in the alluvial plain do authigenic clays become dominant components of the riverine clay minerals (Chakrapani et al., 1995). Finally, the interpretation that the alluvial plain is the chief locus of weathering and clay formation corroborates previous conclusions based on i) solute mass balance (Galy and France-Lanord, 1999) and ii) the geochemistry of sediments in the Bay of Bengal: progressive rain-out of the monsoon induces large latitudinal and longitudinal gradients in the O and H iso-topic composition of rainwater, and clay assemblages since the late Miocene have O and H isotope ratios consistent with formation in

the palaeo-Gangetic plain (Derry and France-Lanord, 1996). Overall, some combination of low temperatures, low solute concentrations and short sediment and solution residence times combine to hinder secondary mineral formation in the mountainous regions. Seen the other way, the high solute concentrations and long sediment-water interaction times promote secondary clay formation and/or biological utilisation of DSi in the alluvial plain.

5.3. Mass balance, weathering and CO2 consumption in the Ganges alluvial plain

Finally, an initial Si mobilisation rate (SMR, mol Si km-2yr-1) for rivers where discharge estimates are available, is simply the quotient of the observed Si yield, Ydsi (mol Si km-2yr-1) and the calculated (unitless) jRayleigh value:

SMR = Y DSi/ /Rayleigh (8)

and is shown in Table 2 for rivers of the Ganges basin. Normalised to catchment area, SMRs are lower in the GAP and peninsular rivers (107 and 150 kmol DSikm-2yr-1 respectively) than the Himalaya (200 kmol DSikm-2yr-1). Net DSi export (i.e. as observed in the river; Section 4.1.1 and Table 1) derives in a roughly 50/50 ratio from the Himalaya, and the plain and peninsular regions combined, in good agreement with previous work (Galy and France-Lanord, 1999). However, the initial release of Si during weathering, after accounting for reincorporation of Si into secondary phases following the approach above, is largely skewed towards the southern and alluvial plain catchments (Table 2) as a result of the (i) greater retention of Si and (ii) larger surface area. Our calculations (Table 2) suggest a total of c. 1300 Gmol Si yr-1 (i.e. ~77 x 106 tons SiO2 yr-1) is initially mobilised within the Ganges basin, of which 24% derives from the Himalaya, 35% from the Indian Peninsula and 41% from the GAP. In other words, the majority of initial silicate weathering (~75%) takes place outside the mountains, consistent with the interpretation of West et al. (2002) on the basis of chemical to total denudation rates.

In terms of long-term CO2 consumption via silicate weathering (Walker et al., 1981), the mobilisation of Si is unimportant - what matters is the release of silicate-hosted cations, predominantly Ca and Mg, and also Na and K to a lesser extent. The rationale for attempting to infer initial Si solubilisation rates is that if the Si-cation ratio of the source material is known, and provided it releases elements to solution congruently, Si isotope geochemistry potentially provides a powerful means of determining silicate-hosted Ca and Mg fluxes. As a first step, we assume the composition of sediment, bedload and banks in Himalayan Front rivers (Lupker et al., 2012) is representative of the material deposited in the GAP, These data suggest typical Ca/Si and Mg/Si molar ratios of ~0.10 and ~0.06, although the presence of detrital quartz or carbonate minerals will lower or increase, respectively, the cation/Si ratio of the silicate material susceptible to weathering. Crudely, our initial GAP Si mobilisation rate estimate of 55 x 109 mol Si yr-1 (Table 2) suggests that 5.5 x 109 mol Ca yr-1 and 3.3 x 109 mol Mg yr-1 are weathered from silicates in the GAP. These estimates overlap with those based on sediment composition change (<10 x 109 molyr-1 Ca and 17 ± 23 x 109 molyr-1 Mg, respectively) from Lupker et al. (2012), demonstrating that an Si-based approach can provide sensible values, and is readily improvable in future by i) more realistic silicate cation/Si ratios and ii) a better understanding of the processes determining the manifestation of Si isotope fractionation.

A growing body of work argues that the impact of Himalayan orogeny on global climate change is minimal since the surface area is small (Li et al., 2014; Moore et al., 2013). The work presented here adds another nuance to this argument: the presence of a vast

alluvial plain with greater net weathering, albeit at lower weathering rates, means the influence of the Himalaya extends beyond their geographical boundaries (West et al., 2002). Although at the whole basin scale DSi export rates are relatively modest, they may still be greater than the situation before the onset of Himalayan orogenesis. The silicon isotopic composition of the Ganges very quickly loses its Himalayan character as it traverses the GAP, with the final DSi exported at Farrakah isotopically closer to Si derived from an alluvial plain source than a Himalayan source. Such lowland processes may also need to be considered in interpreting e.g. marine isotope curves for elements that dominantly reflect frac-tionation during weathering (e.g. Li) since the GAP has co-evolved with the Himalaya. Since the late Paleogene to early Neogene (Singh, 1996), downward flexing of the Indian plate after collision with Eurasia and subsequent sedimentary infilling has created the modern GAP. During its evolution, the sediment accommodation space, the rate and composition of sediment supply, and regional hydrology will inevitably have changed. These changes can have affected the capacity of the (palaeo-)GAP to store and weather Himalayan material, and thus the net weathering capacity attributable to the Himalayan orogeny will not have been constant. Some debate has focused on whether the Himalayan uplift hypothesis negates the possibility of a climate-weathering feedback. Since a majority of silicate weathering apparently occurs outside the Himalaya, this suggests that any climate-weathering feedback must operate primarily via the lowland regions.

6. Conclusions

At a whole-basin scale, the Ganges at Farrakah has dissolved Si yields (52 kmol km-2 yr-1) and 530Si (1.93%o) similar to global average values (Section 4.1.1). The peninsular and alluvial plain tributaries have lower area-normalised silica-mobilisation rates than Himalaya rivers, but due to their larger surface area, the majority of Si mobilised during decomposition of primary minerals (75%) occurs outside the Himalaya. The contribution of the mountain vs. non-mountain to the measured net DSi export is approximately equal (i.e. 43% vs. 57%, respectively; Table 1) because a greater proportion of Si is retained in secondary phases in the alluvial plain and Indian shield catchments than in mountain catchments (Table 2). The basin-wide rate of silicate-weathering (and by extension, long-term CO2 consumption), is therefore partially a function of whatever controls the supply of sediment to the alluvial plain, and the capacity of the alluvial plain to chemically weather this sediment. The co-evolution of the Himalayan orogeny and associated alluvial plains, together with hydrological changes that have occurred over the Neogene may have altered the efficiency and climate dependency of silicate weathering in the catchment.

Acknowledgements

This work was primarily funded by a grant from Lund Royal Physiographic Society (No. 31436) to PJF, plus support from the Knut and Alice Wallenberg Foundation to D|C and a LEFE/CYBER grant ("SiMS") to CDLR. We thank Carolina Funkey at Lund University for assistance with laboratory analysis, and the staff and students at IIT Roorkee for their help and enthusiasm. Julian Bouchez, Bastian Georg and an anonymous reviewer are thanked for their insightful comments on an earlier version of this manuscript.

Appendix A. Supplementary material

Supplementary material related to this article can be found online at http://dx.doi.org/10.1016/jxpsl.2015.06.049.

References

Berner, R.A., Caldeira, K., 1997. The need for mass balance and feedback in the geo-chemical carbon cycle. Geology 25, 955-956.

Bickle, M.J., Bunbury, J., Chapman, H.J., Harris, N.B.W., Fairchild, I.J., Ahmad, T., 2003. Fluxes of Sr into the headwaters of the Ganges. Geochim. Cosmochim. Acta 67, 2567-2584.

Bouchez, J., Gaillardet, J., Lupker, M., Louvat, P., France-Lanord, C., Maurice, L., Armi-jos, E., Moquet, J.-S., 2012. Floodplains of large rivers: weathering reactors or simple silos? Chem. Geol. 332-333, 166-184.

Bouchez, J., von Blanckenburg, F., Schuessler, J.A., 2013. Modeling novel stable isotope ratios in the weathering zone. Am. J. Sci. 313, 267-308.

Cardinal, D., Alleman, L.Y., de Jong, J., Ziegler, K., Andre, L., 2003. Isotopic composition of silicon measured by multicollector plasma source mass spectrometry in dry plasma mode. J. Anal. At. Spectrom. 18, 213-218.

Cardinal, D., Gaillardet, J., Hughes, H.J., Opfergelt, S., André, L., 2010. Contrasting silicon isotope signatures in rivers from the Congo Basin and the specific behaviour of organic-rich waters. Geophys. Res. Lett. 37, L12403.

Chakrapani, G.J., 2005. Major and trace element geochemistry in upper Ganga river in the Himalayas, India. Environ. Geol. 48, 189-201.

Chakrapani, G.J., Subramanian, V., Gibbs, R.J., Jha, P.K., 1995. Size characteristics and mineralogy of suspended sediments of the Ganges river, India. Environ. Geol. 25, 192-196.

Cockerton, H.E., Street-Perrott, F.A., Leng, M.J., Barker, P.A., Horstwood, M.S.A., Pash-ley, V., 2013. Stable-isotope (H, O, and Si) evidence for seasonal variations in hydrology and Si cycling from modern waters in the Nile Basin: implications for interpreting the Quaternary record. Quat. Sci. Rev. 66, 4-21.

Dalai, T.K., Krishnaswami, S., Sarin, M.M., 2002. Major ion chemistry in the headwaters of the Yamuna river system: chemical weathering, its temperature dependence and CO2 consumption in the Himalaya. Geochim. Cosmochim. Acta 66, 3397-3416.

De La Rocha, C.L., Brzezinski, M.A., DeNiro, M.J., 1996. Purification, recovery, and laser-driven fluorination of silicon from dissolved and particulate silica for the measurement of natural stable isotope abundances. Anal. Chem. 68, 3746-3750.

De La Rocha, C.L., Brzezinski, M.A., DeNiro, M.J., 1997. Fractionation of silicon isotopes by marine diatoms during biogenic silica formation. Geochim. Cosmochim. Acta 61, 5051-5056.

Dellinger, M., Gaillardet, J., Bouchez, J., Calmels, D., Louvat, P., Dosseto, A., Gorge, C., Alanoca, L., Maurice, L., 2015. Riverine Li isotope fractionation in the Amazon River basin controlled by the weathering regimes. Geochim. Cosmochim. Acta 164, 71-93.

Delvaux, C., Cardinal, D., Carbonnel, V., Chou, L., Hughes, H.J., André, L., 2013. Controls on riverine 530Si signatures in a temperate watershed under high anthropogenic pressure (Scheldt-Belgium). J. Mar. Syst. 128, 40-51.

Derry, L.A., France-Lanord, C., 1996. Neogene Himalayan weathering history and river 87Sr/86Sr: impact on the marine Sr record. Earth Planet. Sci. Lett. 142, 59-74.

Ding, T., Wan, D., Wang, C., Zhang, F., 2004. Silicon isotope compositions of dissolved silicon and suspended matter in the Yangtze River, China. Geochim. Cosmochim. Acta 68, 205-216.

Ding, T.P., Gao, J.F., Tian, S.H., Wang, H.B., Li, M., 2011. Silicon isotopic composition of dissolved silicon and suspended particulate matter in the Yellow River, China, with implications for the global silicon cycle. Geochim. Cosmochim. Acta 75, 6672-6689.

Dupuis, R., Benoit, M., Nardin, E., Méheut, M., 2015. Fractionation of silicon isotopes in liquids: the importance of configurational disorder. Chem. Geol. 396, 239-254.

Dürr, H.H., Meybeck, M., Hartmann, J., Laruelle, G.G., Roubeix, V., 2011. Global spatial distribution of natural riverine silica inputs to the coastal zone. Biogeo-sciences 8, 597-620.

Edmond, J., Huh, Y., 1997. Chemical weathering yields from basement and orogenic terrains in hot and cold climates. In: Ruddiman, W. (Ed.), Tectonic Uplift and Climate Change. Springer, US, pp. 329-351.

Engström, E., Rodushkin, I., Baxter, D.C., Öhlander, B., 2005. Chromatographic purification for the determination of dissolved silicon isotopic compositions in natural waters by high-resolution multicollector inductively coupled plasma mass spec-trometry. Anal. Chem. 78, 250-257.

Engström, E., Rodushkin, I., Ingri, J., Baxter, D.C., Ecke, F., Osterlund, H., Ohlander, B., 2010. Temporal isotopic variations of dissolved silicon in a pristine boreal river. Chem. Geol. 271, 142-152.

Fontorbe, G., De La Rocha, C.L., Chapman, H.J., Bickle, M.J., 2013. The silicon isotopic composition of the Ganges and its tributaries. Earth Planet. Sci. Lett. 381, 21-30.

Frings, P., Clymans, W., Jeppesen, E., Lauridsen, T., Struyf, E., Conley, D., 2014. Lack of steady-state in the global biogeochemical Si cycle: emerging evidence from lake Si sequestration. Biogeochemistry 117, 255-277.

Frings, P.J., 2014. Integrating Fluvial Processes into the Global Si Cycle. Department of Geology, Lund University, Lund, p. 161.

Galy, A., France-Lanord, C., 1999. Weathering processes in the Ganges-Brahmaputra basin and the riverine alkalinity budget. Chem. Geol. 159, 31-60.

Georg, R.B., Reynolds, B.C., Frank, M., Halliday, A.N., 2006. Mechanisms controlling the silicon isotopic compositions of river waters. Earth Planet. Sci. Lett. 249, 290-306.

Georg, R.B., Reynolds, B.C., West, A.J., Burton, K.W., Halliday, A.N., 2007. Silicon isotope variations accompanying basalt weathering in Iceland. Earth Planet. Sci. Lett. 261, 476-490.

Georg, R.B., West, A.J., Basu, A.R., Halliday, A.N., 2009. Silicon fluxes and isotope composition of direct groundwater discharge into the Bay of Bengal and the effect on the global ocean silicon isotope budget. Earth Planet. Sci. Lett. 283, 67-74.

Hughes, H.J., Bouillon, S., André, L., Cardinal, D., 2012. The effects of weathering variability and anthropogenic pressures upon silicon cycling in an intertropical watershed (Tana River, Kenya). Chem. Geol. 308-309, 18-25.

Hughes, H.J., Sondag, F., Cocquyt, C., Laraque, A., Pandi, A., Andre, L., Cardinal, D., 2011. Effect of seasonal biogenic silica variations on dissolved silicon fluxes and isotopic signatures in the Congo River. Limnol. Oceanogr. 56, 551-561.

Hughes, H.J., Sondag, F., Santos, R.V., André, L., Cardinal, D., 2013. The riverine silicon isotope composition of the Amazon Basin. Geochim. Cosmochim. Acta 121, 637-651.

Jain, S., Agarwal, P., Singh, V., 2007. Ganga basin. In: Jain, S.K., Agarwal, P.K., Singh, V.P. (Eds.), Hydrology and Water Resources of India. Springer Netherlands, Dordecht, pp. 333-418.

Jarvis, A., Reuter, H., Nelson, A., Guevara, E., 2008. Hole-filled SRTM for the globe Version 4, available from the CGIAR-CSI SRTM 90m Database. http://srtm. csi.cgiar.org.

Kothyari, U., Garde, R., 1991. Annual runoff estimation for catchments in India. J. Water Resour. Plan. Manag. 117, 1-10.

Lehner, B., Doll, P., 2004. Development and validation of a global database of lakes, reservoirs and wetlands. J. Hydrol. 296, 1-22.

Lehner, B., Liermann, C.R., Revenga, C., Vörösmarty, C., Fekete, B., Crouzet, P., Döll, P., Endejan, M., Frenken, K., Magome, J., Nilsson, C., Robertson, J.C., Rödel, R., Sindorf, N., Wisser, D., 2011. High-resolution mapping of the world's reservoirs and dams for sustainable river-flow management. Front. Ecol. Environ. 9, 494-502.

Li, D.D., Jacobson, A.D., McInerney, D.J., 2014. A reactive-transport model for examining tectonic and climatic controls on chemical weathering and atmospheric CO2 consumption in granitic regolith. Chem. Geol. 365, 30-42.

Lupker, M., France-Lanord, C., Galy, V., Lavé, J., Gaillardet, J., Gajurel, A.P., Guilmette, C., Rahman, M., Singh, S.K., Sinha, R., 2012. Predominant floodplain over mountain weathering of Himalayan sediments (Ganga basin). Geochim. Cosmochim. Acta 84, 410-432.

Misra, S., Froelich, P.N., 2012. Lithium isotope history of Cenozoic seawater: changes in silicate weathering and reverse weathering. Science 335, 818-823.

Moore, J., Jacobson, A.D., Holmden, C., Craw, D., 2013. Tracking the relationship between mountain uplift, silicate weathering, and long-term CO2 consumption with Ca isotopes: Southern Alps, New Zealand. Chem. Geol. 341, 110-127.

Murnane, R.J., Stallard, R.F., 1990. Germanium and silicon in rivers of the Orinoco drainage-basin. Nature 344, 749-752.

Oelze, M., von Blanckenburg, F., Hoellen, D., Dietzel, M., Bouchez, J., 2014. Si stable isotope fractionation during adsorption and the competition between kinetic and equilibrium isotope fractionation: implications for weathering systems. Chem. Geol. 380, 161-171.

Opfergelt, S., Cardinal, D., André, L., Delvigne, C., Bremond, L., Delvaux, B., 2010. Variations of S30Si and Ge/Si with weathering and biogenic input in tropical basaltic ash soils under monoculture. Geochim. Cosmochim. Acta 74, 225-240.

Opfergelt, S., Cardinal, D., Henriet, C., Draye, X., André, L., Delvaux, B., 2006. Silicon isotopic fractionation by Banana (Musa spp.) grown in a continuous nutrient flow device. Plant Soil 285, 333-345.

Opfergelt, S., Delmelle, P., 2012. Silicon isotopes and continental weathering processes: assessing controls on Si transfer to the ocean. C. R. Géosci. 344, 723-738.

Pogge von Strandmann, P.A.E., Henderson, G.M., 2014. The Li isotope response to mountain uplift. Geology 43, 67-70.

Pokrovsky, O.S., Reynolds, B.C., Prokushkin, A.S., Schott, J., Viers, J., 2013. Silicon isotope variations in Central Siberian rivers during basalt weathering in permafrost-dominated larch forests. Chem. Geol. 355, 103-116.

Rai, S.K., Singh, S.K., Krishnaswami, S., 2010. Chemical weathering in the plain and peninsular sub-basins of the Ganga: impact on major ion chemistry and elemental fluxes. Geochim. Cosmochim. Acta 74, 2340-2355.

Raymo, M.E., Ruddiman, W.F., 1992. Tectonic forcing of late Cenozoic climate. Nature 359, 117-122.

Rengarajan, R., Singh, S.K., Sarin, M.M., Krishnaswami, S., 2009. Strontium isotopes and major ion chemistry in the Chambal River system, India: implications to silicate erosion rates of the Ganga. Chem. Geol. 260, 87-101.

Reynolds, B.C., Aggarwal, J., Andre, L., Baxter, D., Beucher, C., Brzezinski, M.A., Engstrom, E., Georg, R.B., Land, M., Leng, M.J., Opfergelt, S., Rodushkin, I., Sloane, H.J., van den Boorn, S., Vroon, P.Z., Cardinal, D., 2007. An inter-laboratory comparison of Si isotope reference materials. J. Anal. At. Spectrom. 22, 561-568.

Reynolds, B.C., Frank, M., Halliday, A.N., 2006. Silicon isotope fractionation during nutrient utilization in the North Pacific. Earth Planet. Sci. Lett. 244, 431-443.

Sarin, M.M., Krishnaswami, S., Dilli, K., Somayajulu, B.L.K., Moore, W.S., 1989. Major ion chemistry of the Ganga-Brahmaputra river system: weathering processes

and fluxes to the Bay of Bengal. Geochim. Cosmochim. Acta 53, 997-1009.

Savage, P.S., Georg, R.B., Williams, H.M., Halliday, A.N., 2013. The silicon isotope composition of the upper continental crust. Geochim. Cosmochim. Acta 109, 384-399.

Singh, I.B., 1996. Geological evolution of Ganga Plain - an overview.

Sinha, R., 2005. Why do Gangetic rivers aggrade or degrade? Curr. Sci. 89, 836.

Tripathy, G.R., Singh, S.K., 2010. Chemical erosion rates of river basins of the Ganga system in the Himalaya: reanalysis based on inversion of dissolved major ions, Sr, and 87Sr/86Sr. Geochem. Geophys. Geosyst. 11, Q03013.

Walker, J.C.G., Hays, P.B., Kasting, J.F., 1981. A negative feedback mechanism for the

long-term stabilization of earths surface-temperature. J. Geophys. Res., Oceans Atmos. 86, 9776-9782.

West, A.J., Bickle, M.J., Collins, R., Brasington, J., 2002. Small-catchment perspective on Himalayan weathering fluxes. Geology 30, 355-358.

White, A.F., Blum, A.E., 1995. Effects of climate on chemical-weathering in watersheds. Geochim. Cosmochim. Acta 59, 1729-1747.

Zachos, J., Pagani, M., Sloan, L., Thomas, E., Billups, K., 2001. Trends, rhythms, and aberrations in global climate 65 Ma to present. Science 292, 686-693.

Ziegler, K., Chadwick, O.A., Brzezinski, M.A., Kelly, E.F., 2005. Natural variations of S30Si ratios during progressive basalt weathering, Hawaiian Islands. Geochim. Cosmochim. Acta 69, 4597-4610.