Sediment Fluxes and Buffering in the Postglacial Indus Basin

EAGE Basin Research (2014) 26, 369–386, doi: 10.1111/bre.12038 Sediment fluxes and buffering in the post-glacial Indus Basin P. D. Clift*,† and L. Giosan‡ *Department of Geology and Geophysics, Louisiana State University, Baton Rouge, LA, USA †South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China ‡Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, MA, USA

ABSTRACT The Indus drainage has experienced major variations in climate since the Last Glacial Maximum (LGM) that have affected the volumes and compositions of the sediment reaching the ocean since that time. We here present a comprehensive first-order source-to-sink budget spanning the time since the LGM. We show that buffering of sediment in the floodplain accounts for ca. 20–25% of the mass flux. Sedimentation rates have varied greatly and must have been on average three times the recent, predamming rates. Much of the sediment was released by incision of fluvial terraces con- structed behind landslide dams within the mountains, and especially along the major river valleys. New bedrock erosion is estimated to supply around 45% of the sedimentation. Around 50% of deposited sediment lies under the southern floodplains, with 50% offshore in large shelf clinoforms. Provenance indicators show a change of erosional focus during the Early Holocene, but no change in the Mid–Late Holocene because of further reworking from the floodplains. While suspended loads travel rapidly from source-to-sink, zircon grains in the bedload show travel times of 7–14 kyr. The largest lag times are anticipated in the Indus submarine fan where sedimentation lags erosion by at least 10 kyr.

INTRODUCTION (Metivier & Gaudemer, 1999). While this has been dis- puted over timescales >106 years (Clift, 2006), there is Marine sediments represent the longest and most com- presently little control on how effective this buffering plete archives of continental environmental evolution, process might be on shorter timescales. and they have been used to reconstruct past patterns, rates In this study, we present a new sediment budget for the of erosion and changing environments at the time of post-glacial Indus River system in order to quantify the deposition. These in turn can be used to assess the influ- degree of buffering in a large drainage system and to see ence that climate or tectonics have over continental envi- how the deep-sea sediment record might be used to ronments. However, if the marine record is to be utilized understand erosion in the mountain sources. We use sedi- to its full potential, we must first constrain the transport mentary provenance data to assess the speed of sediment of clastic sedimentary particles from continental sources transport from the source ranges to the ocean and further to the marine sink. to the deep-sea. We focus on the post-Last Glacial Maxi- To what extent are sediments deposited in the deep mum (LGM) period because the sediment stored on the ocean representative of the erosion in the source shelf clearly sits on a glacial unconformity, thus allowing regions at the time of their sedimentation? Clastic sedi- ready estimation of the total deposited volume. This is also ments do not relocate instantly into the sea after ero- an appropriate time period for estimating eroded volumes sion in mountains, but their transport to the ocean onshore because a number of incised fluvial terraces in the could vary in efficiency and speed, spanning from days floodplains (Giosan et al., 2012a) and mountains are or weeks, to >105–106 years, because of storage and known to date from ca. 10 ka (Bookhagen et al., 2006; He- reactivation en route. Indeed, it has been argued that witt, 2009). It has been recognized that the Early Holocene many Asian river systems have maintained a relatively was a period of rapid sediment flux to South Asian deltas constant mass flux to the ocean over at least the past (Goodbred & Kuehl, 2000a; Giosan et al., 2006a), but it is 2 Ma because of sediment buffering in the floodplain not clear to what extent this sediment was freshly eroded bedrock rather than older fluvial sediments reworked from Correspondence: P. D. Clift, Department of Geology and the floodplains and terraces in the mountains. While some Geophysics, Louisiana State University, E235 Howe-Russell, sediment was released by retreating glaciers after the Baton Rouge, LA 70803, USA. E-mail: [email protected] Liviu Giosan, Department of Geology and Geophysics, Woods LGM, this affected the central and eastern Himalaya more Hole Oceanographic Institution, Woods Hole, MA 20543, USA than the Indus watershed because of the relatively dry cli- E-mail: [email protected] mate of this western region (Owen et al., 2008). Hedrick

© 2013 The Authors Basin Research © 2013 John Wiley & Sons Ltd, European Association of Geoscientists & Engineers and International Association of Sedimentologists 369 P.D. Clift and L. Giosan et al. (2011) noted that while those regions north of and in provenance quantification simpler than in many drain- the rain shadow of the Greater Himalaya experienced only ages (Clift et al., 2004; Garzanti et al., 2005). Crucially, moderate glacial advance at the LGM, there was signifi- it is known that the origin of the sediment reaching the cant glacial advance in the more monsoonal regions within ocean has changed sharply since the LGM (Clift et al., and south of the Greater Himalaya. Retreat of those gla- 2008). The Indus is a sediment-productive river, which ciers would have released significant sediment quantities used to supply at least 250 Mt yearÀ1 to the Arabian that could then be transported to the ocean. Sea prior to modern damming (Milliman & Syvitski, The Indus River basin has a strongly erosive mon- 1992), although other estimates emphasize a larger range soonal climate over parts of the drainage, as well as from 100 to 675 Mt yearÀ1 (Ali & De Boer, 2008). steep topography and well-defined areas of rapid rock Modern and relict wide floodplains flank the Indus and uplift within the source mountains [e.g. Nanga Parbat, stretch >1300 km from the mountain front to the delta, south Karakoram metamorphic domes (Zeitler et al., providing ample potential opportunity for storage and 1993; Maheo et al., 2004)]. In addition, there is a strong reworking. rainfall gradient across the mountains from wet in the south to the dry regions of Ladakh, Karakoram and Tibet in the north (Bookhagen & Burbank, 2006) ERODED VOLUMES (Fig. 1). The sediment sources in the Himalaya and Ka- rakoram are very diverse in composition, and thermo- Eroded and incised fluvial terraces are recognized across chronologic ages that make large-scale sediment the northern half of the Indus floodplains, as well as in

Figure 3 Karakoram 36˚ Gilgit Hindu Kush

Kabul Ladakh 34˚

Indus Ravi Chenab 32˚ Jhellum Beas Sutlej Greater Himalaya Lesser Himalaya Harappa Punjab 30˚ Sulaiman Ranges Cholistan

Sukkur Figure 2 Thar Desert 28˚ Moenjodaro

Sindh Kirthar Ranges 26˚

K Matli Figure 12 Karachi Thatta L A B Jati Keti Gul Bandar Rann of Kutch 24˚

66˚ 68˚ 70˚ 72˚ 74˚ 76˚ 78˚ 80˚ 82˚

Fig. 1. Shaded topographic map of the Indus River basin showing the divisions between the major tributaries. Trunk streams are marked in white, with the drainage divides depicted with black lines. The map also shows the major geographical regions and mountain ranges mentioned in the text. Image is from NASA WorldWind (http://worldwind.arc.nasa.gov/). Our borehole locations (Gio- san et al., 2006b) are shown in the lower Indus plain and delta (i.e. Keti Bandar, Jati, Thatta, Gul, Matli) together with transects A-B and K-L (Fig. 11) using our data and published borehole info from Kazmi (1984). The dotted black lines along the Indus indicate the presumed boundaries of the incised valley (Kazmi, 1984).

NW SE 230 33˚ X 220 W 210 V 200 Y Ravi 190 (X) Jhellum Chenab 180

Indus 0 50 100 150 200 250 300 150 31˚ sand dunes Z 140 130

Multan 120 Sutlej 30˚ 110 (Z) 100 0 50 100 150 200 250 300

29˚29 300 Thar Desert 280 71˚ 72˚ 73˚ 74˚ 75˚

260 200

190 240 Interfluve (W) 180 220 170 (V) 160 200 River valley 150 (Y) 180 140 0 50 100 150 200 250 300 350 400

Elevation above sealevel (m) sealevel above Elevation 0 50 100 150 200 Distance (km) Distance (km)

Fig. 2. Shaded topographic map of the northern half of the Indus River floodplains showing the modern courses of the major rivers and the ‘interfluve’ terraces between them. Optically stimulated luminescence dating has shown that these terraces formed up to 10 ka and has been incised. Transects across the floodplain show that the river now occupy incised valleys. We calculate the maximum and minimum limits for eroded sediment as the regions under the dotted lines interpreted as a smooth surfaces at 10 ka. Topography is from NASA’s Shuttle Radar Topography Mission via GeoMapAppTM. Stars show the age control points from Giosan et al.(2012b). the mountains, and constitute a potentially significant incision, we can estimate how much material has been source of sediment for the river. Reworking of sediment excavated from the valleys. Using maximum and mini- stored in these terraces is likely a major source of sedi- mum estimates of the total area of the incised valleys ment to the Indus since the LGM (Fig. 1). Giosan et al. (Fig. 2) and assuming a linear reduction in depth from (2012a) noted that the alluvial plain in Punjab shows inci- north to south, we estimate that the amount of sediment sion by the modern Indus and its tributary rivers forming reworked further south since 10 ka is ca. 925 km3 if we valleys only 10–20 m deep, but tens of kilometres wide use the volume under the upper presumed interfluve sur- (Fig. 2). The valleys are separated from each other by face in Fig. 2. A minimum estimate can be derived using elevated plateau regions or interfluves. Optically stimu- the lower interfluve surface, which implies erosion of ca. lated luminescence (OSL) dating of the sediments at the 370 km3 (Table 1). The upper figure for the eroded vol- top of the interfluves has revealed that sedimentation was ume assumes that relief on the interfluves is largely the ongoing until ca. 10 ka and that the valleys have thus result of fluvial erosion, whereas the lower figure esti- been cut after that time (Giosan et al., 2012a). A high- mates only the main valley erosion of the Indus and its resolution digital elevation model of the region reveals Himalayan tributaries assuming that the interfluve relief that the incision is deepest closest to the mountain front is caused by aeolian remodelling. These two estimates and decreases southward, becoming insignificant close to also give a measure of the uncertainty for this source of the final confluence of the Indus and the Sutlej Rivers sediment, although we do expect that the larger value is (Figs 1 and 2). By assuming that the northern floodplains closer to reality because active dunes are mostly limited were accretionary and had a smooth topography prior to to the Thal Desert (the Indus–Jhelum–Chenab inter-

Table 1. Estimates of the eroded volumes in different sectors Table 1 (continued) of the Indus floodplains and Himalayan source regions since the Volume (km3) Percentage Last Glacial Maximum Maximum Minimum Maximum Minimum Volume eroded (km3) Area of 2 Total 925 370 23 21 Erosion sources incision (km ) Maximum Minimum Floodplains Floodplains 92 545 925 370 Total around 1305 703 32 40 Rumbak drainage 162 0.077 0.051 Nanga Volumes eroded 510 609 242 161 Parbat outside main valleys Total from 137 39 3 2 Volumes in Kirthar 195 402 93 62 rain and Sulaiman shadow Ranges Total from 363 195 9 11 side valleys Length (km) Maximum Minimum Grand total 4104 1752 Indus in Ladakh 70 7 4 Indus upstream 670 71 38 of Gilgit fluve) and in the Thar Desert, which is largely south of Indus around 220 601 323 our study area (the interfluve south of the Sutlej–Indus Nanga Parbat courses; Fig. 2). Trunk 425 53 15 Assessing the volumes eroded from the mountains is Indus – Himalaya more complex and several sectors need to be considered realm separately if a realistic figure of remobilized sediment is to Shyok River 270 33 11 be achieved for this region. We divide the mountains into Hunza River 225 27 9 four sectors based on climate and geology. One sector Shigar River 93 254 137 covers the Greater and Lesser Himalayan ranges that are Gilgit River 160 20 7 Zanskar River 180 22 7 influenced by heavy summer monsoon rains. These are Swat Valley 280 34 11 expected to be more sediment productive and potentially Chitral Valley 236 29 10 more tectonically active than those areas to the north that Kohistan River 98 12 4 lie effectively on the western edge of the Tibetan Plateau Nanga Parbat River 165 450 243 and that experienced major uplift earlier in the Palaeogene Trunk 290 36 10 (Harris, 2006). The main tributaries that join the Indus Sutlej – Himalaya from the east are largely sourcing their sediment in this realm Himalayan regime (Fig. 3). The trunk Indus, the Zanskar – Trunk Sutlej Tibetan 230 29 8 and the northern parts of the Sutlej River drain a separate realm region lying in the rain shadow of the Himalaya. We infer Trunk Jhelum 213 27 8 similar conditions for the western parts of the Kabul Jhelum tributary 169 21 6 Trunk Chenab 365 46 13 River, and those streams draining the Karakoram and Chenab Tributary 155 19 5 Hindu Kush (i.e. the Shyok, Hunza, Gilgit, Swat and Trunk Ravi 217 27 8 Chitral Rivers) because these regions are also not heavily Trunk Beas 176 22 6 monsoonal in character. The third major division lies Trunk Kabul – East 300 38 11 around the Nanga Parbat massif (Fig. 3) where it has Branch been recognized that especially strong landsliding has Trunk Kabul – West 400 51 14 caused large-scale damming of the Indus in the past Branch 20 kyr (Korup et al., 2010). We follow the study of Total 1929 908 Hewitt (2009) in defining this region to be 100 km around Volume (km3) Percentage the peak of Nanga Parbat itself, including the major rivers passing around the peak to the north and east, as well as Maximum Minimum Maximum Minimum the Shigar River (Fig. 3). A separate estimate was made Total 252 71 6 4 for a fourth sector in which sediment is stored behind Himalayan landslide rock dams in the Karakoram. Hewitt (1998) rivers mapped a series of large dams and the extent of lake ter- Total 1122 374 27 21 races behind them in Baltistan, east of the Hunza River Karakoram, (Fig. 3). Not all the lakes were completely filled with sedi- Kohistan ment at the point that the dam was breached, but we use and Hindu the estimates of the breached height to calculate how Kush much material may have been reworked, at least from (continued) those dams defined by Hewitt (1998) (Table 1). The

37˚

Gilgit R. Hunza 36˚ Gilgit Kohistan Shigar

Chitral 35˚ Nanga Kabul Parbat Shyok Swat Khalsi 34˚ Kyber Pass Leh Attock Rumbak Zanskar Islamabad 33˚ Pang Indus Ravi Chenab 32˚ Beas Jhellum Lahore

31˚

Sutlej

30˚

70˚ 71˚ 72˚ 73˚ 74˚ 75˚ 76˚ 77˚ 78˚ 79˚ 80˚ 81˚

Fig. 3. Shaded topographic map of western Himalaya, Karakoram and Hindu Kush showing the major rivers feeding the Indus. Dot- ted lines show divisions between major region of contrasting climate and geomorphology. North of the line on Sutlej, we assume La- dakh-type dry sediment storage, while between the lines along the main Indus around Nanga Parbat large-scale landslide damming is assumed.

34˚ 8’ uncertainties on these numbers are high if poorly deﬁned, Indus but it is clear that damming and rock slides are important sediment buffers in the Karakoram, and that major vol-

5 umes of sediment have been stored in the mountain val- 34˚ 1 6’ leys. We use an uncertainty of Æ50% for the eroded 13 4 volumes in these rock-dammed valleys, following the logic for the large Himalayan rivers outlined below.

11 34˚ In contrast to the sediment stored behind dams in the 7 4’ Rumbak major river valleys, we recognize that sediment is also 12 3 stored within the tributary valleys to these large trunk rivers, as well as on the wider landscape in the upper 6 34˚ parts of the drainage. In an attempt to quantify how 2 8 2’ 10 much sediment may be released from these areas, we 2 9 consider the drainage that covers ca. 160 km located SW of Leh and south of the Indus River in Ladakh 34˚ 0’ (Figs 3 and 4). We do not choose the Rumbak drainage 5 km for any particular reason and indeed aim to use it as a

77˚ 77˚ 77˚ 77˚ 77˚ typical example of a catchment in the rain shadow of 20’ 22’ 24’ 26’ 28’ the Greater Himalaya. It was selected for study based Fig. 4. Satellite image of the Rumbak drainage near Leh on its moderate size that allowed it to be surveyed across showing the main tributaries considered in this study and most of its extent and because it is relatively accessible. whose sediment volumes are provided in Table 2. See Fig. 3 Comparison of the Rumbak with other valleys in for location. Ladakh suggests that it is typical in its form and

Table 2. Estimates of the sediment volumes eroded from the individual side valleys within the Rumbak drainage used to determine sediment storage within the upper part of the Indus source regions. See Fig. 4 for location

Length of segment Distance from Maximum incision Maximum channel width Volume eroded Streams (km) conﬂuence (km) (km) (km) (km3)

1 15.20 0.00 0.0500 0.0500 0.0190 2 5.30 9.30 0.0194 0.0388 0.0020 3 3.40 10.90 0.0141 0.0283 0.0007 4 4.20 5.10 0.0332 0.0664 0.0046 5 9.60 1.00 0.0467 0.0934 0.0209 6 6.10 7.90 0.0240 0.0480 0.0035 7 5.20 5.60 0.0316 0.0632 0.0052 8 5.85 8.70 0.0214 0.0428 0.0027 9 2.20 12.50 0.0089 0.0178 0.0002 10 2.20 11.00 0.0138 0.0276 0.0004 11 1.90 7.20 0.0263 0.0526 0.0013 12 1.70 8.50 0.0220 0.0441 0.0008 13 2.45 5.10 0.0332 0.0664 0.0027 Total 0.0641

(a) (b)

Fig. 5. Field photographs showing examples of terracing in the Indus basin (a) in the Rumbak drainage (see Figs 3 and 4) where the terrace is around 20 m high, and (b) the Pang gorge, south of the more plains in Ladakh (location on Fig. 3). Terrace here shows ca. 250 m of relief. location. In the future, surveying of wider areas will test NASA Shuttle Radar Topography Mission (SRTM) whether the Rumbak is indeed representative as we data. Sediment is presumed to have filled the valley to argue. The wider region is made up of a series of smal- the level of the terrace top, so that the sediment volume ler catchments of this type, and this is therefore a rea- eroded can be estimated by projecting the terrace level sonable starting point for making a budget in this part across the valley and calculating the area of sediment of the Indus basin. Surveying of this drainage revealed missing. A volume can then be derived by multiplying the presence of an incised terrace system running all the that area by the length of the stream. In the case of the way from the confluence with the Indus to the upper Rumbak drainage where the terraces are known to thin reaches (Fig. 5a). While the incision is greatest towards up the valley to zero, we can estimate the volume by the base of the drainage, this reduces to zero in the assuming a linear decrease from the downstream end of upper reaches. We used field measurements of the inci- the valley where the maximum incision is noted and the sion to calculate the volume of sediment removed since top of the valley where zero erosion has occurred the incision of the terrace, which is presumed to be (Fig. 6b). post-glacial, based on comparison with other Himalaya We estimated 0.064 km3 (64.1 9 106 m3) of erosion terraces (Table 2). over the 160 km2 of the Rumbak Valley, and we used this Figure 6a shows the erosion calculation approach. A value to extrapolate how much material might be held in terrace dating to ca. 10 ka or older is identified, and a similar systems across the entire mountainous parts of the profile across the valley is measured from maps or from Indus catchment, including the Kirthar and Sulaiman

3 À1 (a) lost 77.1 m km of stream in that study area. To understand the potential uncertainties in extrapolating this stor- Eroded material age capacity along the length of the river, we measured Terrace dated ~10 ka five sections across the Sutlej upstream and downstream of the Bookhagen et al. study area and estimated the possible amount of material that would have been eroded had the valley been filled up the level of the terrace dated at 10 ka by Bookhagen et al. (2006). Figure 7 shows that Modern the valley is much narrower over significant lengths com- river pared with where the terraces are well developed. None- theless, the average eroded estimate from these five (b) sections was still 80.4 m3 kmÀ1 of stream, but with a standard deviation equivalent to Æ56%. Other significant sources of uncertainty to the eroded volumes in other parts of the Himalaya, which have not been the subject of Y detailed study, are the height of the terraces above the river and the age of the terraces, as the length and width of the river valleys are reasonably easily measured. Based on the existing study of terraces, we estimate height Z uncertainties of not more than 20% and a similar value for the age, although terraces dating at 8–10 ka are found where studies have been performed in the monsoonally affected Himalaya (Bookhagen et al., 2006; Dortch et al., Fig. 6. Schematic diagram showing how the volumes of eroded 2008). sediment are calculated. (a) In areas where a 10 ka or other post- 3 À1 glacial terrace are known the area of sediment removed since the We apply a value of 80.4 m km of incision to the start of incision can be used to calculated volumes along a river. entire length of the Beas, Ravi, Chenab and Jhelum Riv- (b) In a side valley, such as the Rumbak drainage, the depth of ers, as well as their major tributaries in order to estimate incision is normally at a maximum where the stream reaches the the possible released volumes. Because incision of terraces trunk river (at point Z) and reduces to zero at the head of the val- is known to be high immediately into the frontal Siwalik ley (point Y). Hills in Nepal (Lave & Avouac, 2001), there is no reason to consider a reduced incision adjacent to the floodplains. We further apply this rate to the Himalayan parts of the Ranges in the west. These latter ranges are not part of the Sutlej (i.e. those south of the Himalayan rain shadow), the Himalaya, and they experience a drier climate than the Indus downstream of the Nanga Parbat region and the monsoonal southern face of the Himalaya. Until they can eastern parts of the Kabul River that lie in the wetter be surveyed themselves, we use the sediment yield from regions east of the Khyber Pass (Fig. 3). Unlike the Rum- the Rumbak Valley because this is located in the Himalaya bak drainage, we assume a constant terrace incision along rain shadow and is also subject to a more arid environ- the length of the valley between the range front and the ment. The amount of material derived from such settings northern side of the Greater Himalaya. We follow the ear- appears to be much smaller than other sources and is esti- lier studies in considering that this terracing has resulted mated here at 279 km3, although clearly the uncertainties from landsliding and damming of the river followed by in such an estimates must be significant if difficult to back-filling (Bookhagen et al., 2006). The common occur- quantify at the present time (Table 1). While we recog- rence of landsliding in the Himalaya (Dortch et al., 2008) nize that in some locations much greater thicknesses have and the record of localized lake sediment (Phartiyal et al., been stored and released, for example the Pang gorge 2005) testifies to this being an active and important pro- south of the more plains in Ladakh (Fig. 5b), such areas cess in controlling sediment flux in that part of the river are of generally limited areal extent and are not believed basin. to make a large difference to the basin-wide budget. Simi- We estimate sediment storage along the trunk Indus in larly, we recognize that in some locations, such as the Nu- Ladakh using a different data set because of the contrast- bra–Shyok Valley in Ladakh, glacial advance resulted in ing drier climate and less tectonically active geological set- formation of a large lake system, but sedimentary facies ting here that might result in different conditions analysis has shown that this was largely a water-filled compared with the monsoonal Sutlej. Figure 8 shows a basin and did not release large volumes of sediment as the transect along the Indus gorge between Leh and Khalsi glaciers retreated (Dortch et al., 2010). (Fig. 3) with a number of terrace levels identified at a To estimate the amount of sediment released from the range of heights above the Indus stream. We note that Himalayan valleys, we use the terrace mapping and dating because no single terrace extends all the way along the work of Bookhagen et al. (2006) in the Sutlej Valley as a 70 km survey, we infer that the valley has been filled type example. This study estimated that the Sutlej has locally by landslides damming the river, especially in

E 31˚ 28’

31˚ 26’ C

31˚ 24’ B A D 31˚ 22’

31˚ 20’ 5 km

77˚ 27’ 77˚ 30’ 77˚ 33’ 77˚ 36’ 77˚ 39’

(a) 1200 (b)

1000 10 ka 1000 10 ka

3 3 800 120 m /km 45 m /km

1200 (c)1200 (d) Elevation (masl) Elevation

1000 10 ka 10 ka 1000 104 m3/km 111 m3/km

0 1.0 2.0 3.0 0 1.0 2.0 3.0

1400 (e)

1200 10 ka 19 m3/km Elevation (masl) Elevation 1000 0 1.0 2.0 3.0

Fig. 7. Shaded topographic map of a section of the Sutlej River showing the variation in storage capacity along a stretch of the main- stream. Sections show that depending on the valley geometry significant differences in storage capacity exist. Section D corresponds to the study area of Bookhagen et al. (2006). regions where the gorge is narrow and prone to blocking. was likely incised prior to the LGM. Terrace A, B and C Although landslides are common, they are usually rapidly are 30 ka or older, allowing us to assume that the lower breached and destroyed, but when large rock landslides terraces (i.e. D–I) post-date the glacial maximum and block the river, then the possibility exists for the forma- provide an estimate of the amount of sediment released tion of long-lived lakes and their subsequent infilling by from the trunk Indus in Ladakh (Table 3). Although bet- sediment (Costa & Schuster, 1988). That such events ter age control would be required for a more precise esti- have occurred along the upper Indus is testified by the mate, our approach here will provide a first-order presence of laminated lake deposits interbedded with flu- approximation for the amount of sediment reworked dur- vial and alluvial fan sediments in the Indus Valley around ing the Holocene. We estimate that volume to be Leh (Phartiyal & Sharma, 2009). 14C dating of these sedi- 5.69 km3 over the 70-km transect, allowing us to calculate ments shows that they roughly predate the LGM and how much material could have been stored in the Indus were formed between 31 and 51 ka, during a period of Valley upstream of the Nanga Parbat region. Although stronger monsoon. We thus know that Terrace B (Fig. 8) the region of study is reasonably well defined in terms of

Khalsi Nurla Ule Tokpo SaspolAlchi Basgo Nimu Zanskar Phey SpitokEast West Confluence 76˚50’ 77˚00’ 77˚10’ 77˚20’ 77˚30’ 3400 ? 3400

A 3300 B 3300 C 3200 3200 D E F 3100 3100 Altitude (m) GA Terrace sediment 3000 AH 3000 AI Lacustrine sediment

2900 2900 01020304050 60 70 Distance (km)

Fig. 8. Section along the Indus River in Ladakh showing the terraces and their extent along the trunk river. The lacustrine sediments at Spitok are roughly dated at 33–51 ka (Phartiyal et al., 2005) so that the next lowest terrace is presumed to be the post-glacial.

Table 3. Estimates of the sediment volumes eroded from under mapped ﬂuvial terraces in Ladakh along the trunk Indus River divided into sectors based on morphology of the valley and extent of terraces

Volume (km3)

Elevation (m) 3320 m 3285 m 3250 m 3180 m 3160 m 3120 m 3070 m 3000 m 2975 m Terrace Terrace Terrace Terrace Terrace Terrace Terrace Terrace Sector Length (km) A B C D E F G H Terrace I

Leh Valley 40 0.00 12.83 6.95 0.00 0.00 0.00 0.00 0.00 0.00 Zanskar 5 0.39 0.00 0.15 0.00 0.03 0.02 0.00 0.00 0.00 confluence Basgo–Nimu 8 3.30 0.00 1.39 0.00 0.41 0.19 0.00 0.00 0.00 Valley Basgo–Alchi 12 3.39 0.00 1.89 0.00 0.65 0.30 0.00 0.00 0.00 Ule Tokpo 14 0.00 0.00 0.00 0.78 0.00 0.42 0.11 0.00 0.00 Nurla 8 0.00 0.00 0.00 1.09 0.00 0.00 0.25 0.00 0.00 Khalsi 10 0.00 0.00 0.00 1.24 0.00 0.00 0.00 0.14 0.06 Total 97 7.08 12.83 10.39 3.11 1.08 0.93 0.36 0.14 0.06 volume and to a lesser extent in age, we assign an uncer- distributed between the Karakoram–Hindu Kush (21– tainty of Æ30% to the total estimate, reflecting along 27%) and the northern floodplains (21–23%) (Fig. 9). strike possible variability in storage capacity. This value is Overall, uncertainties are estimated at Æ40% reflecting less than that assigned to the Sutlej because the surveyed the relatively well-defined volumes in the floodplains, but region of the Indus spans ca. 70 km, rather than being the less well-known values from the large river valleys, or focused in a restricted zone as was the case with the eroded from the high-mountain topography. Sutlej. This is considerably less than the sediment storage capacity of the Himalayan Sutlej; 0.081 km3 kmÀ1 com- 3 À1 pared with 0.771 km km (Bookhagen et al., 2006). We DEPOSITED VOLUMES also apply this ‘rain shadow’ rate to those lengths of the Kabul River west of the Khyber Pass, as well as to the The Indus alluvial plain gradually becomes depositional Shyok, Hunza, Gilgit, Swat, Chitral, Kohistan and Zans- south of the confluence with the Punjabi tributaries and kar Rivers because they all lie north of the Himalayan shows maximum aggradation near the modern channel range crest and thus are generally much drier (Bookhagen belt (Giosan et al., 2012a). The resulting landscape is a & Burbank, 2006). unique fluvial mega-ridge of subdued relief extending The total eroded volume is estimated at 1752– from the lower Punjab to the Arabian Sea. Under the 4104 km3 (Table 1). The regional breakdown shows that deposits of the mega-ridge, post-glacial sediments fill up the largest single source of sediment is the peri-Nanga the Indus’ Pleistocene incised-valley system (Kazmi, Parbat region (32–40%), with other major contributions 1984; Giosan et al., 2006a).

Indus Holocene Depocenters Erosion sources Indus

Side Himalayan valleys rivers Ladakh and Karakoram Sindh western Tibet Kohistan Kutch alluvial plain clinoform Hindu Kush

Nanga Parbat Western clinoform

Eastern clinoform Delta plain Flood Plains

Fig. 9. Pie diagrams showing the origin and fate of sediment in the post-glacial Indus River system. Note equal division between on and offshore sedimentation and the moderate contribution from reworking of the ﬂoodplains.

We estimated the volumes stored by the incised valley by assuming rectangular and triangular cross sections for and the mega-ridge as irregular prisms (Fig. 10) with the incised-valley system, respectively (Table 4; average widths of 100 and 150 km for the Indus alluvial Fig. 10a). Within these parameters, together, the alluvial plain and the delta, respectively. We employed average plain and the delta have stored between ca. 2100 and gradients for the modern alluvial plain (7 cm kmÀ1) 3600 km3 of post-glacial sediment. (Giosan et al., 2012a) and the Pleistocene incised-valley In front of the Indus delta, post-glacial Indus sedi- profile (25 cm kmÀ1) (Giosan et al., 2006a) starting from ments have been stored within the Indus canyon and as a deposit thickness of 90 m at the coast. The depth mid-shelf clinoforms on both sides of the canyon assumption is supported by interceptions of the valley (Fig. 12) (Giosan et al., 2006b; Limmer et al., 2012). A bottom by our drill sites at the coast (Figs 1 and 11) at third clinoform has developed along the Kutch coast with Keti Bandar as well as inland (Jati and Gul at ca. 70 km most sediment probably advected eastward alongshore from the coast and Matli at ca. 170 km from the coast). from the Indus delta and redeposited on the mid-shelf by These values should provide a minimum estimate of the the offshore-directed tidal jet at the Gulf of Kutch mouth valley fill volume with the incised valley extending ca. (Giosan et al., 2006a). However, it is possible that the 250 km from the present coast. Then, using Kazmi’s Kutch clinoform may only be a surficial feature as the interpretation of an incised-valley extending over 700 km incised valley of the Indus may have not extended so far upstream, we obtained a maximum volume estimate. east. We calculated the maximum volumes stored within Based on our drill core data, we assume Kazmi’s view the clinoforms below the modern shelf bathymetry down of an incised valley that is very wide and deep with a low to À90 m water depth assuming a lowstand erosion of the gradient that allows the valley to reach far inland. This shelf to that depth (Table 4). This assumption is sup- view is supported by available data so far but will need to ported by the depth and width of the incised valley at the be confirmed by future chronologies of the infill in the coast (Fig. 11) as well as by available seismic profiles Sindh plain. However, the Indus incised valley as such (Giosan et al., 2006b; Limmer et al., 2012). envisioned is well within the empirical range for other Our recent survey of the Indus canyon shows a transi- large palaeovalley dimensions (Blum et al., 2013). tion at ca. 800 m water depth from an active depositional Although the proposed flanking of the incised valley by region in the upper canyon to a meandering channel mor- patches of previous highstand deposits (Kazmi, 1984) is phology indicative of intermittent turbidity flow (Clift in line with some existing models for palaeovalleys (see et al., 2013). Sediments recovered from the upper canyon e.g. Blum et al., 2013), this is not supported by our Matli document extremely high rates of sedimentation, while site where Holocene sediments overlie Eocene limestone the lower Indus canyon shows instead that turbidite at ca. 60 m depth rather than being entirely Pleistocene sedimentation ended shortly after 6.9 ka (Clift et al., (Fig. 11) (Giosan et al., 2006b). Nevertheless, to allow 2013). Previous coring in the canyon dated a cessation of for a first-order uncertainty estimate, we calculated mini- turbidite sedimentation after 11 ka ca. 75 km from the mum and maximum stored volumes in the alluvial plain canyon head (Prins et al., 2000). We calculated the

(a)

100 km

Mega-ridge deposits desert post-glacial Indus Bedrock older Indus Bedrock deposits

(b) Starved Shelf clinoforms Delta Alluvial plain shelf

Post-glacial Indus deposits 90 m

Bedrock or older indus deposits Fig. 10. Schematic figures showing how the volumes of sediments stored in the various parts of the southern floodplains and delta are estimated. (a) Diagram showing a cross section through the major depocentres in the Sindh flood- Upper canyon Delta Alluvial plain plains showing the maximum (solid line) (c) and minimum (dashed line) eroded volumes, (b) diagram showing the longitudi- nal cross section through the Sindh Post-glacial Indus deposits floodplains, delta plain and offshore shelf 90 m regions, (c) diagram of the stored sediment in the region of the Indus canyon. Thicknesses of post-Last Glacial Maxi- mum sediment in the canyon is Bedrock or older indus deposits unknown, but coring in the mid-canyon has shown no sediment post-dating ca. 11 ka at 75 km from the head of the canyon (Prins et al., 2000). 800 m volume of the sediment prism accumulated in the upper used to reconstruct erosion patterns in the mountains that Indus canyon using a 4 km average width, the present can be correlated with changes in climate. Available average slope of the canyon to 800 m water depth and a U-series dating of sediment in the Ganges River, for predepositional lowstand profile extending from 800 m example, suggests long transport times of ca. 100 kyr for water depth to À90 m subsurface at the coast (Fig. 10c). sediments travelling between the range front and the Overall, the maximum total volume of sediment stored confluence with the Brahmaputra near the delta (ca. offshore in the canyon and clinoforms is ca. 2800 or 1650 km) (Chabaux et al., 2006; Granet et al., 2007), 1900 km3 when we do not consider the Kutch clinoform, although residence times were estimated to be shorter ca. values that are of the same order as the Indus inland stor- 25 kyr for finer-grained sediment (Granet et al., 2010) age (Table 4). First-order uncertainties vary between and most recently 10Be cosmogenic dating argues for sedi- 55% for the inland part of the system and 35% offshore. ment transfer times of ca. 1400 years in the Ganges (Lup- ker et al., 2012) consistent with the similar short residence times deduced in the Andean foreland (Witt- MINERALTRANSPORT TIMES mann et al., 2011). Furthermore, different mineral phases will have different lag times depending on their density, The transport times of mineral grains need to be con- shape and grain size. Radiometric dating of mica by strained for the Indus system if different phases are to be 39Ar-40Ar methods and zircons by U-Pb methods in the

Transect A-B Keti Bandar Jati Gul

–40 m ~9 ka ~12 ka –80 m Depth below sealevel (m)

–120 m ~14 ka ~130 km

Transect K-L Mirpur Thatta Matli Batoro Badin Digh Nabisar

~7.5 ka –40 m

12.4 ka –80 m

–120 m ~100 km

Fig. 11. Simplified sedimentary logs for two transect of boreholes in the Indus delta and lower Indus alluvial plain (location for transects in Fig. 1). Mud (brown) and sand (yellow) fill a wide incised valley floored by gravel or Indurated rock (orange). Available data suggest that the fill is latest Deglacial to Holocene in age.

Table 4. Estimates of the deposited volumes in different sec- reflecting the slower passage of zircon grains through the tors of the Indus floodplains, delta plain and marine regions river compared with the mica or clay fractions. Zircon is since the Last Glacial Maximum documented to be concentrated in the bedload of large Himalayan rivers (Garzanti et al., 2010; Lupker et al., Depositional environment Volume (km3) 2011). Sindh alluvial plain 725–2500 Suspended sediments tend to be transported quickly Delta plain 1410 from source to the ocean, but the grains of the river Shelf delta 2790 bedload would be expected to move more slowly and may Eastern clinoform 1260 undergo significant transport only during the strongest Western clinoform 630 flood events (Williams, 1989; Syvitski et al., 2000). Kutch clinoform 900 Bedload not only moves more slower than suspended Canyon 25 load, but stepwise with potentially very significant storage Deposition 4950–6725 Deposition without Kutch 4050–5825 intervals in fluvial bars and floodplain. Here we compare the evolution in bulk sediment Nd isotopes with the changing zircon age populations in the same sediments cored at Keti Bandar at the river mouth. Red River has highlighted how different mineral groups Nd is a useful sediment proxy because it is controlled by may indicate quite different provenances for the same the average age of the crust from, which it is eroded and is river sediment samples (Hoang et al., 2009, 2010). In the generally not affected by weathering processes (Goldstein Red River example, the discrepancy was explained as et al., 1984). A shift in eNd to more negative values

25˚

Karachi Matli K L Delta Plain

Keti Jati Gul westernBandar Rann of Kutch clinoform B A 24˚

n eastern o clinoform y Fig. 12. Shaded bathymetric and topo- n a graphic map of the Indus delta area C s u d showing the major depositional areas esti- In 20 Kutch mated in our budget. Map is contoured clinoform 23˚ 40 between 200 m water depth and 200 m Arabian Sea 80 60 elevation above sea level at 20-m inter- 100 vals. Topography is from NASA’s Shut- tle Radar Topography Mission via GeoMapApp. 66˚ 67˚ 68˚ 69˚ 70˚

Nanga Parbat Transhimalaya-Karakoram Greater Himalaya Lesser Himalaya GISP icecore Monsoon strength ε Proportion of total δ18O δ18O (VPDB) Nd sediment (%) –42 –38 –34 –5 –6 –7 –8 –11 –13 –15 0 1020304050 0 (a) (b) (c) (d) 1 Indus delta Indus canyon 2 Drier Wetter Maximum 3 change in 4 Primitive Continental zircons Fig. 13. Diagram showing changing 5 sediment provenance compared with 6 evolving continental climate in SW Asia. 7 (a) The GISP2 ice core climate record Qunf (Stuiver & Grootes, 2000), (b) the inten- 8 9

sity of the SW monsoon traced by speleo- Age (ka) them records from Qunf and Timta 10 Caves (Fleitmann et al., 2003; Sinha 11 et al., 2005) in Oman and by pollen Timta Maximum 12 Younger change in Nd (Herzschuh, 2006) from across Asia Dryas (black line), (c) Nd isotopic evolution of 13 sediments from the Indus delta and Indus 14 canyon (Clift et al., 2008) and (d) chang- 15 ing zircon age populations from the delta colder warmer at Keti Bandar (Clift et al., 2010). 20 between 14 and 9 ka was interpreted by Clift et al. (2008) grained fraction (Goldstein et al., 1984) and because these as reflecting stronger erosion of older, more continental usually travel rapidly as suspended load it is not surprising rocks compared with younger more primitive ones during that the change in eNd values is synchronous with the gen- the Early Holocene (Fig. 13). Such a trend would indicate eral increase in summer monsoon strength over the same a shift of erosion away from arc-like rocks in the Indus time period (Fleitmann et al., 2003; Herzschuh, 2006). suture zone and towards the range front of the Himalaya, What is more difficult to explain is why the subsequent as these are composed of deformed but very old Indian reduction in summer monsoon rains is not paralleled by a plate rocks. Today, the strongest monsoon rains are in the matching rise in eNd values as the erosion of the Lesser Lesser Himalaya, and the change in eNd values is consis- Himalaya weakened. The lack of a response suggests a tent with a climatic trigger for that shift, not least because buffer in the system. We suggest here that the stability of the speed of the change precludes tectonic explanations eNd values analysed at the Indus delta since 7 ka reflects a and drainage capture can be ruled out in this case. Nd dominant reworking of sediment since that time, and that isotopes are dominated by clay minerals in the fine- although the discharge from the Indus tributaries now

(a) Modern discharge from himalaya (b) Modern sediment based on water budget at thatta

Fig. 14. Pie diagram showing (a) the estimated zircon age population of Indus River sands now being discharged from the mountains based on water discharge and the known composition of the major tributaries (Alizai et al., 2011), and (b) 750-1250 Ma 300-750 Ma the observed zircon age populations from close to the delta at Thatta (Clift et al., 0-300 Ma 1500-2300 Ma 2004). favours a stronger erosion from the Karakoram and Koh- size (>100 lm required for the laser dating method). istan compared with that in the Early Holocene (Alizai The platiness of phyllosilicates including micas means et al., 2011), this signal has yet to be communicated to the that these are preferentially carried in the suspended delta, at least in terms of Nd isotopes (Fig. 14). Indeed, load compared with the more rounded aspect of zircon the most recent change in eNd values near the delta is a grains. We suggest that the time difference between the decrease from À15 to À16, the opposite of the change change in Nd and that seen in zircons reflects the trans- expected from the water discharge (Fig. 13). We interpret port time of the zircon in the bedload, that is, ca. 7– this most recent change to be linked partly to reworking 14 kyr. This number is significantly less than the in the floodplain and also to the effect of synthetic dam- 100 kyr estimated by Granet et al. (2007), although the ming on the Indus since the 1930s (Inam et al., 2007). course of the Ganges and the Indus have comparable U-Pb dating of zircon sand grains can be used to assess lengths: ca. 1300 km from the end of the Sutlej gorge to the provenance of the bedload dominated fluvial fraction, the Indus delta and ca. 1650 km from the end of the and it has been shown that this method is particularly Ganges gorge to the confluence with the Brahmaputra. suitable to the Indus because of the diversity of ages in the Our estimate is longer than the 10Be-derived rates of potential source terrains and the documented variation Lupker et al. (2012), although because that 1400 year the age spectra of zircons in the tributaries (Alizai et al., estimate is based on a quartz-dominated mineralogy, it 2011). We note that while the zircon populations seen at might be expected to be shorter than the zircon travel the delta change with time in a way that is consistent with time. the variation in Nd isotopes (i.e. with more Karakoram– Trans-himalayan grains (<300 Ma) at the LGM compared with the Holocene), these changes are not syn- DISCUSSION chronous for the two proxies. Figure 13 highlights the fact that while eNd values had fallen to a minimum by 9 ka The sediment budget components presented above are and stabilized at a new low value of À13 to À14 after clearly subject to significant uncertainties, but it is 7 ka, the zircon populations remain relatively constant nonetheless noteworthy that the total sediment volumes over that entire time period. The most obvious change in deposited, 4050–6725 km3, are only moderately greater zircon ages occurs sometimes after 7 ka, when the propor- than the volumes eroded 1752–4104 km3. This means tion of Lesser Himalayan zircons (1500–2300 Ma) rose that much (26–99%, average 54%) of the sediment depos- from 22% to 40%, at the same time that Karakoram– ited in the Indus delta, on the Indus shelf and within the Trans-himalayan grains (<300 Ma) dropped from 37% to Indus canyon since the LGM can be explained in terms of 22%. reworking from pre-existing sediments stored in the allu- We propose that the two provenance proxies are out vial plains, rather than from newly generated sediment of phase with one another because they represent differ- from bedrock erosion. Because it seems unlikely that ero- ent parts of the sediment load. While the bulk sediment sion would simply cease for any significant length of time, Nd signature is dominated by sediment transported in our result implies that the products of new erosion may be the suspended load, especially clays and mica, the zir- stored in terraces or landslides and that erosion is much cons are necessarily part of the bedload because of the slower in the absence of extensive glaciers. Our budget À high density (ca. 4.6 g cm 3 compared with 2.75 for reconstruction shows that the floodplains are one of the À illite or 2.68 g cm 3 for plagioclase) and coarse grain most important depocentres and sediment sources, but

© 2013 The Authors 382 Basin Research © 2013 John Wiley & Sons Ltd, European Association of Geoscientists & Engineers and International Association of Sedimentologists Sediment flux in the Indus Basin the volumes stored and released are just 21–23% of the zircon populations seen at Thatta and Keti Bandar. Our total transport. The concept of sediment flux to the ocean eroded budget suggests that 21–27% of the total sediment being buffered by the floodplains alone clearly does not load is being derived from the Karakoram/Hindu Kush apply to the Indus system (Metivier & Gaudemer, 1999). and another 32–40% from ‘greater Nanga Parbat’ The 4050–6725 km3 of sediment deposited since ca. (including the Kohistan Arc), a total of 53–67%. This 14 ka (the age of the base of the post-glacial delta as dated proportion is significantly higher than the 24% of zircons at Keti Bandar; Clift et al., 2008) corresponds to an aver- with U-Pb ages less than 300 Ma seen at Thatta (Clift age rate of 506–841 Mt yearÀ1, assuming an average den- et al., 2004), but closer to the 44% of grains dated at sity of 1.75 g cmÀ3. Considering that the 725–2500 km3 <300 Ma from the LGM sediments at Keti Bandar (Clift stored under the Sindh alluvial plain post-dates 14 ka, the et al., 2010). Zircons with this <300 Ma age range are average rate of mass delivery would be 325– considered typical of erosion from Karakoram, Kohistan 1120 Mt yearÀ1. We note that this is considerably greater and Nanga Parbat sources. Given the zircon transport than the 250 Mt yearÀ1 estimated as predamming flux by time, we would expect those sediments eroded from ter- Milliman & Syvitski (1992) but lies in the range from 100 races in the mountains since 10 ka to have now reached to 675 Mt yearÀ1 estimated by Ali & De Boer (2008). the delta and be represented by the modern sample from Even in a more conservative estimate with 20 ka as the Thatta. However, Alizai et al. (2011) noted that the trunk lower limiting age for the budget, the average rate stays Indus is much less rich in Zr, and thus zircon, compared high, at 420 Mt yearÀ1. This implies a large degree of with most of the Himalayan tributaries. Indus sands temporal variation in sediment flux: low rates in more yielded only 18 ppm in Zr compared with as much as recent times would have to be balanced by higher rates in 63 ppm in the Jhelum. These concentrations are pre- the early Holocene, and this would be consistent with the sumed to be representative of the zircon load in the mod- observation of the delta prograding seawards from 14 to ern rivers, yet we note that the Zr concentrations are 8 ka despite the rapidly rising sea level of that time much less than the 162 Æ 47 ppm reported from the (Giosan et al., 2006a). Such a pulse mirrors the observa- delta by Clift et al. (2010). The analysed sands from the tion of such an event in the Bengal delta (Goodbred & delta may have been enriched in zircon as a result of Kuehl, 2000b) and further argues against major sediment hydraulic sorting concentrating these heavy minerals in buffering in the net flux to the ocean. the samples taken, although the concentration is within The eroded material derived from the Nanga Parbat the range of 152–238 ppm reported by Hu & Gao (2008) region is one sediment flux anomaly that emerges from for the continental upper crust. Alternatively, hydraulic our budget: while sediment eroded from that region sorting may have caused dilution of heavy minerals in the accounts for 32–40% of the total erosional side of the samples analysed by Alizai et al. (2011). In the absence of budget, only 3% of zircons can be clearly tied to Nanga alternative ways of correcting for zircon abundance in the Parbat at the delta (Clift et al., 2010). This apparent con- different streams feeding the delta, we use the ratios of Zr tradiction is related to the definition of Nanga Parbat. in the different streams as a proxy for weighting their con- The very young (<12 Ma) U-Pb ages of zircons that are centrations. considered characteristic of the massif are found within Using the modern water discharge values of the main the gneisses and granites in the core of the uplift (Zeitler eastern Himalaya tributaries as a proxy for relative dis- & Chamberlain, 1991; Zeitler et al., 1993). However, the charge, we can calculate a weighted average Zr concentra- region of heaviest landsliding and sediment storage is tion for sediment from these rivers of 37 ppm. This in more widely defined within 100 km of the peak (Hewitt, turn allows us to calculate the predicted proportion of 2009) and thus encompasses areas of the Kohistan Arc <300 Ma vs. Himalayan zircons in the mixed sands. If and the southern Karakoram that are experiencing some 59–61% of the sediment since the LGM were from the additional rock uplift centred around the peak. These trunk Indus, then we calculate that ca. 30% of the zircons areas may not be experiencing the degree of uplift and would be <300 Ma in the mixed sediment. This is closer exhumation seen at Nanga Parbat itself, but they are also to the 24% of such zircons recorded by Clift et al. (2004) clearly more strongly uplifted and subject to seismic shak- at Thatta and suggests that our eroded sediment budget ing that other regions in the mountain source regions. If lies within error of the predicted budget based on zircon this wider definition of the Nanga Parbat syntaxis is used, ages. If the zircon concentration correction is in error, then the apparent discrepancy with the eastern Namche then there may be a significant mismatch between our Barwa syntaxis is significantly reduced. Petrology and sediment budget and the provenance of Holocene delta geochemical indicators suggest that ca. 35% of the Brah- sands. maputra sediment downstream of Namche Barwa is On millennial timescales, sediment flux rates and eroded from that massif (Garzanti et al., 2004), far more sources vary significantly, indicating that several buffers than the values normally assigned for the western syntaxis are active in the Indus sediment transport system. at 3% based on zircons (Clift et al., 2010) and 6% based Although the flux must have been much higher earlier on heavy minerals (Garzanti et al., 2005). in the Holocene than in the pre-industrial recent, the We note that there is a reasonable agreement between composition of sediment varies in ways that makes simple the estimated sources of the eroded sediments and the linkage between erosion and climate records difficult to

© 2013 The Authors Basin Research © 2013 John Wiley & Sons Ltd, European Association of Geoscientists & Engineers and International Association of Sedimentologists 383 P.D. Clift and L. Giosan make using marine sediments. This is particularly true if CONCLUSIONS deep-sea fan sediments are considered because the flux of sediment to the final depocentre ceased during the Holo- The sediment flux to the Arabian Sea from the Indus cene as a normal consequence of rising sea level and River has experienced significant variability after the following the classical sequence stratigraphic models of LGM. We estimate that 4050–5675 km3 of sediment has Vail et al. (1977). Prins et al. (2000) dated the youngest been deposited in the incised valley, subaerial delta, shelf turbidite on the upper fan at 11 ka, while our coring of clinoforms and the upper Indus canyon. About half of the the Indus canyon itself shows that turbidite sedimentation stored sediment lies offshore on the shelf and in the can- in the lower parts ended shortly after 6.9 ka (unpublished yon. While ca. 21–23% of the total depositional volume data), during a period of rapid sea level rise and when can- comes from incision of the upper alluvial plain after yon sedimentation would normally have finished. Con- 10 ka, the bulk of the sediment is derived from the deep versely, sedimentation has been rapid both on the gorges around the Nanga Parbat syntaxis that account for clinoforms of the Indus shelf and likely offshore the Rann 32–40% of the sediment flux despite comprising only ca. of Kutch, as well as at the head of the Indus canyon where 5% of the area of the mountain sources in the Himalaya– the most rapid recent sedimentation has been occurring Karakoram–Hindu Kush region. The Karakoram is also (Giosan et al., 2006a). Presumably, if the sea level an important source, accounting for ca. 21–27% of the remains stable and sediment delivery continued at a high total sediment released. Sediment buffering in the moun- rate, then sediment flux to the deep canyon and submar- tains appears to be largely controlled by landsliding, ine fan would be re-established in the near future (Bur- which is in part controlled by climate (Bookhagen et al., gess & Hovius, 1998). However, in most glacial cycles, sea 2005). Only 5% is eroded from terraces in major river val- level has not been as stable for as long as during the Holo- leys within the monsoonal regions of the Lesser and cene but instead fell again shortly after reaching a maxi- Greater Himalaya. mum (Stuiver & Grootes, 2000). Our budget implies that primary weathering of bedrock This scenario suggests that the sediment flux to the supplied ca. 46% of the sediment reaching the delta since deep-sea would have been re-established on the fan after the LGM. Although climate seems to control the source a moderate nondepositional break, but the source of the and rate of sediment supply in the Early Holocene, strong sediment reaching the fan is primarily from reworking. reworking from terraces and floodplains during the Mid– Since 7 ka, sediment transported to the delta appears to Late Holocene means that changes in erosion patterns be dominated by sediment reworked from terraces in after ca. 8 ka are not recorded in the changing composi- intramountain tributary valleys with a subordinate but tion of delta sediments. We estimate that zircon grains in significant reworking of the lower alluvial plains. As sea the bedload take between 7 and 14 kyr to travel to the level fell, erosion of the shelf clinoforms and incision of delta compared with shorter travel times for the sus- the neighbouring delta plain would have resulted in a pended load. This further inhibits our ability to relate mixed array of sources and ages for the sedimentary changes in provenance to climatic events as the bedload grains delivered to the submarine fan during the regres- and suspended load travel times are decoupled and lag the sion phase. It thus seems unlikely that detailed prove- initial climatic trigger. nance work on fan turbidite sands could be correlated in any meaningful way with the monsoon climate record on millennial timescales or shorter because the initial erosion of these grains may have occurred during the LGM. An ACKNOWLEDGEMENTS initial reworking phase from mountain terraces occurred PC thanks the Leverhulme Trust, the University of Aber- in the Early Holocene. This material might then be deen and NERC for support for the science project deposited in the northern floodplains from where it may underlying this study. LG acknowledges support from be reworked again in the latter parts of the Holocene. A the National Science Foundation and Woods Hole lag of at least 10 kyr would seem to be implied for sedi- Oceanographic Institution. The authors thank colleagues mentary particles travelling from the peaks of the Karak- at National Institute of Oceanography, Karachi for their oram to the bottom of the Arabian Sea. This is not help over many years of work on and offshore Pakistan significant if the record is to be used to look at erosion on and Bodo Bookhagen for advice on sediment storage in tectonic timescales (106 year). If the goal is to look at ero- the Sutlej Valley. We thank editor Peter van der Beek and sion on millennial timescales, records on the shelf front- reviewers Eduardo Garzanti and Steven Goodbred for ing the delta may be of greater use because the delta their helpful advice in improving the initial submission. stratigraphy does show a clear change in provenance during the 14–9 ka period synchronous with the climate change, at least in the fast-travelling suspended sediment load. However, even in these deposits, further reworking REFERENCES of the lower alluvial plain masks any erosional change dri- ALI, K.F. & De BOER, D.H. (2008) Factors controlling specific ven by a weakening monsoon in the upper parts of the sediment yield in the upper Indus River basin, northern Paki- drainage since 8 ka. stan. Hydrol. Process., 22, 3102–3114.

ALIZAI, A., CARTER, A., CLIFT, P.D., VANLANINGHAM, S., WIL- of the Indian monsoon recorded in a stalagmite from South- LIAMS, J.C. & KUMAR, R. (2011) Sediment provenance, ern Oman. Science, 300, 1737–1739. reworking and transport processes in the Indus River by U- GARZANTI, E., VEZZOLI, G., ANDO, S., FRANCE-LANORD, C., Pb dating of detrital zircon grains. Global Planet. Change, 76, SINGH, S.K. & FOSTER, G. (2004) Sand petrology and focused 33–55. doi:10.1016/j.gloplacha.2010.11.008 erosion in collision orogens: The Brahmaputra case. Earth BLUM, M., MARTIN, J., MILLIKEN,K.&GARVIN, M. (2013) Pa- Planet. Sci. Lett., 220, 157–174. leovalley Systems: insights from Quaternary analogs and GARZANTI, E., VEZZOLI, G., ANDO, S., PAPARELLA,P.&CLIFT, experiments. Earth-Sci. Rev., 116, 128–169. doi:10.1016/j. P.D. (2005) Petrology of Indus River sands: a key to interpret earscirev.2012.09.003 erosion history of the Western Himalayan Syntaxis. Earth BOOKHAGEN,B.&BURBANK, D.W. (2006) Topography, Relief, Planet. Sci. Lett., 229, 287–302. doi:10.1016/j.epsl.2004.11. and TRMM-derived rainfall variations along the Himalaya. 008 Geophys. Res. Lett., 33, L08405. doi:10.1029/2006GL026037 GARZANTI, E., ANDO, S., FRANCE-LANORD, C., VEZZOLI, G., BOOKHAGEN, B., THIEDE, R.C. & STRECKER, M.R. (2005) Late CENSI, P., GALY,V.&NAJMAN, Y. (2010) Mineralogical and Quaternary intensified monsoon phases control landscape chemical variability of fluvial sediments: 1. Bedload sand evolution in the northwest Himalaya. Geology, 33, 149–152. (Ganga–Brahmaputra, Bangladesh). Earth Planet. Sci. Lett., BOOKHAGEN, B., FLEITMANN, D., NISHIIZUMI, K., STRECKER, 299, 368–381. doi:10.1016/j.epsl.2010.09.017 M.R. & THIEDE, R.C. (2006) Holocene monsoonal dynamics GIOSAN, L., CLIFT, P.D., BLUSZTAJN, J., TABREZ, A., CONSTAN- and fluvial terrace formation in the northwest Himalaya, TINESCU,S.&FILIP, F. (2006a) On the control of climate- and India. Geology (Boulder), 34, 601–604. human-modulated fluvial sediment delivery on river delta BURGESS, P.M. & HOVIUS, N. (1998) Rates of delta progradation development: the Indus. EOS Trans. Am. Geophys. Union, 87, during highstands; consequences for timing of deposition in OS14A-04. deep-marine systems. J. Geol. Soc., 155, 217–222. GIOSAN, L., CONSTANTINESCU, S., CLIFT, P.D., TABREZ, A.R., CHABAUX, F., GRANET, M., PELT, E., FRANCE-LANORD,C.& DANISH,M.&INAM, A. (2006b) Recent morphodynamics of GALY, V. (2006) U-238-U-234-Th-230 disequilibria and the Indus delta shore and shelf. Cont. Shelf Res., 26, 1668– timescale of sedimentary transfers in rivers: clues from the 1684. Gangetic plain rivers. J. Geochem. Explor., 88, 373–375. GIOSAN, L., CLIFT, P.D., MACKLIN, M.G., FULLER, D.Q., CON- CLIFT, P.D. (2006) Controls on the erosion of Cenozoic Asia and STANTINESCU, S., DURCAN, J.A., STEVENS, T., DULLER, the flux of clastic sediment to the ocean. Earth Planet. Sci. G.A.T., TABREZ, A., ADHIKARI, R., GANGAL, K., ALIZAI, A., Lett., 241, 571–580. FILIP, F., VANLANINGHAM,S.&SYVITSKI, J.P.M. (2012a) CLIFT, P.D., CAMPBELL, I.H., PRINGLE, M.S., CARTER, A., Fluvial landscapes of the Harappan civilization. Proc. Natl. ZHANG, X., HODGES, K.V., KHAN, A.A. & ALLEN, C.M. Acad. Sci. USA, 109, 1688–1694. doi:10.1073/pnas. (2004) Thermochronology of the Modern Indus River Bed- 1112743109 load: new insight into the control on the marine stratigraphic GIOSAN, L., FULLER, D.Q., NICOLL, K., FLAD, R.K. & CLIFT, record. Tectonics, 23, TC5013. doi:10.1029/2003TC001559 P.D., editors (2012b) Climates, Landscapes, and Civilizations. CLIFT, P.D., GIOSAN, L., BLUSZTAJN, J., CAMPBELL, I.H., ALLEN, Geophysical Monograph Series. American Geophysical Union, C.M., PRINGLE, M., TABREZ, A., DANISH, M., RABBANI, Washington, D. C. M.M., CARTER,A.&LUCKGE€ , A. (2008) Holocene erosion of GOLDSTEIN, S.L., O’NIONS, R.K. & HAMILTON, P.J. (1984) A the Lesser Himalaya triggered by intensified summer mon- Sm-Nd isotopic study of atmospheric dusts and particulates soon. Geology, 36,79–82. doi:10.1130/G24315A.1 from major river systems. Earth Planet. Sci. Lett., 70, 221– CLIFT, P.D., GIOSAN, L., CARTER, A., GARZANTI, E., GALY, 236. V., TABREZ, A.R., PRINGLE, M., CAMPBELL, I.H., FRANCE- GOODBRED, S.L. & KUEHL, S.A. (2000a) The significance of LANORD, C., BLUSZTAJN, J., ALLEN, C., ALIZAI, A., LUCKGE€ , large sediment supply, active tectonism, and eustasy on mar- A., DANISH,M.&RABBANI, M.M. (2010) Monsoon control gin sequence development; Late Quaternary stratigraphy and over erosion patterns in the western himalaya: possible evolution of the Ganges-Brahmaputra delta. Sed. Geol., 133, feed-backs into the tectonic evolution. In: Monsoon Evolu- 227–248. tion and Tectonic-Climate Linkage in Asia (Ed. by P.D. GOODBRED, S.L. & KUEHL, S.A. (2000b) Enormous Ganges- Clift, R. Tada & H. Zheng) Geol. Soc. London Spec. Publ., Brahmaputra sediment discharge during strengthened early 342, 181–213. Holocene monsoon. Geology (Boulder), 28, 1083–1086. CLIFT, P.D., GIOSAN, L., HENSTOCK,T.&TABREZ, A.R. (2013) GRANET, M., CHABAUX, F., STILLE, P., FRANCE-LANORD,C. Sediment Storage and reworking on the shelf and in the can- &PELT, E. (2007) Time-scales of sedimentary transfer and yon of the Indus River-Fan system since the Last Glacial weathering processes from U-series nuclides: clues from Maximum. Basin Res. In press. the Himalayan rivers. Earth Planet. Sci. Lett., 261, 389– COSTA, J.E. & SCHUSTER, R.L. (1988) The formation and failure 406. of natural dams. Geol. Soc. Am. Bull., 100, 1054–1068. GRANET, M., CHABAUX, F., STILLE, P., DOSSETO, A., FRANCE- DORTCH, J., OWEN, L.A., HANEBERG, W.C., CAFFEE, M.W., LANORD,C.&BLAES, E. (2010) U-series disequilibria in sus- DIETSCH,C.&KAMP, D.U. (2008) Nature and timing of pended river sediments and implication for sediment transfer large-landslides in the Himalaya and Transhimalaya of north- time in alluvial plains: the case of the Himalayan rivers. Geo- ern india. Quatern. Sci. Rev., 28, 1037–1054. chim. Cosmochim. Acta, 74, 2851–2865. DORTCH, J.M., OWEN, L.A. & CAFFEE, M.W. (2010) Quaternary HARRIS, N.B.W. (2006) The elevation of the Tibetan Plateau glaciation in the Nubra and Shyok Valley confluence, north- and its impact on the monsoon. Palaeogeogr. Palaeoclimatol. ernmost Ladakh, India. Quatern. Res., 74, 132–144. Palaeoecol., 241,4–15. FLEITMANN, D., BURNS, S.J., MUDELSEE, M., NEFF, U., KRA- HEDRICK, K.A., SEONG, Y.B., OWEN, L.A., CAFFEE, M.W. & DI- MERS, J., MANGINI,A.&MATTER, A. (2003) Holocene forcing ETSCH, C. (2011) Towards defining the transition in style and

timing of quaternary glaciation between the monsoon-influ- between oblique crustal boundaries in south Karakorum, enced Greater Himalaya and the semi-arid Transhimalaya of northwest Himalaya, Pakistan. Geol. Soc. Am. Spec. Pap., 380, Northern India. Quatern. Int., 236,21–33. doi:10.1016/j. 141–154. quaint.2010.07.023 METIVIER ,F.&GAUDEMER, Y. (1999) Stability of output fluxes HERZSCHUH, U. (2006) Palaeo-moisture evolution in monsoonal of large rivers in South and East Asia during the last 2 million Central Asia during the last 50,000 years. Quatern. Sci. Rev., years; implications of floodplain processes. Basin Res., 11, 25, 163–178. 293–303. HEWITT, K. (1998) Catastrophic landslides and their effects on MILLIMAN, J.D. & SYVITSKI, J.P.M. (1992) Geomorphic/tec- the Upper Indus streams, Karakoram Himalaya, northern tonic control of sediment discharge to the ocean; the impor- Pakistan. Geomorphology, 26,47–80. tance of small mountainous rivers. J. Geol., 100, 525–544. HEWITT, K. (2009) Catastrophic rock slope failures and late OWEN, L.A., CAFFEE, M., FINKEL, R.C. & SEONG, Y.B. (2008) Quaternary developments in the Nanga Parbat-Haramosh Quaternary glaciation of the Himalayan–Tibetan orogen. J. Massif, Upper Indus basin, Northern Pakistan. Quatern. Sci. Quatern. Sci., 23, 513–531. doi:10.1002/jqs.1203 Rev., 28, 1055–1069. PHARTIYAL,B.&SHARMA, A. (2009) Soft-sediment deforma- HOANG, L.V., WU, F.Y., CLIFT, P.D., WYSOCKA,A.&SWIERC- tion structures in the late Quaternary sediments of Ladakh: ZEWSKA, A. (2009) Evaluating the evolution of the Red River evidence for multiple phases of seismic tremors in the system based on in-Situ U-Pb dating and Hf isotope analysis northwestern Himalayan region. J. Asian Earth Sci., 34, of zircons. Geochem. Geophys. Geosyst., 10, Q11008. doi:10. 761–770. 1029/2009GC002819 PHARTIYAL, B., SHARMA, A., UPADHYAY, R., RAM-AWATAR &SIN- HOANG, L.V., CLIFT, P.D., MARK, D., ZHENG,H.&TAN, M.T. HA, A.K. (2005) Quaternary geology, tectonics and distribu- (2010) Ar-Ar muscovite dating as a constraint on sediment tion of palaeo- and present fluvio/glacio lacustrine deposits in provenance and erosion processes in the Red and Yangtze Ladakh, NW Indian Himalaya – a study based on field obser- River systems, SE Asia. Earth Planet. Sci. Lett., 295, 379– vations. Geomorphology, 65, 241–256. 389. doi:10.1016/j.epsl.2010.04.012 PRINS, M.A., POSTMA, G., CLEVERINGA, J., CRAMP,A.&KEN- HU,Z.&GAO, S. (2008) Upper crustal abundances of trace ele- YON, N.H. (2000) Controls on terrigenous sediment supply to ments: a revision and update. Chem. Geol., 253, 205–221. the Arabian Sea during the late Quaternary: the Indus Fan. doi:10.1016/j.chemgeo.2008.05.010 Mar. Geol., 169, 327–349. doi:10.1016/S0025-3227(00) INAM, A., CLIFT, P.D., GIOSAN, L., TABREZ, A.R., TAHIR, M., 00086-4 RABBANI, M.M. & DANISH, M. (2007) The geographic, geo- SINHA, A., CANNARIATO, K.G., STOTT, L.D., LI, H.-C., YOU, logical and oceanographic setting of the Indus River. In: Large C.-F., CHENG, H., EDWARDS, R.L. & SINGH, I.B. (2005) Vari- Rivers: Geomorphology and Management (Ed. by A. Gupta), ability of Southwest Indian summer monsoon precipitation pp. 333–345. John Wiley and Sons, Chichester, UK. during the Bolling-Allerod. Geology, 33, 813–816. KAZMI, A.H. (1984) Geology of the Indus delta. In: Marine Geol- STUIVER,M.&GROOTES, P.M. (2000) Gisp2 Oxygen isotope ogy and Oceanography of Arabian Sea and Coastal Pakistan ratios. Quatern. Res. (New York), 53, 277–284. (Ed. by B.U. Haq, J.D. Milliman), pp. 65–70. Van Nostrand SYVITSKI, J.P.M., MOREHEAD, M.D., BAHR, D.B. & MULDER,T. Reinhold, New York. (2000) Estimating fluvial sediment transport: the rating KORUP, O., MONTGOMERY, D.R. & HEWITT, K. (2010) Gla- parameters. Water Resour. Res., 36, 2747–2760. cier and landslide feedbacks to topographic relief in the VAIL, P.R., MITCHUM, R.M., TODD, R.G., WIDMIER, J.M., Himalayan syntaxes. Proc. Natl. Acad. Sci. USA, 107, THOMPSON, S.I., SANGREE, J.B., BUBB, J.N. & HATLELID, 5317–5322. W.G. (1977) Seismic stratigraphy and global changes of sea- LAVE,J.&AVOUAC, J.P. (2001) Fluvial incision and tectonic level. In: Seismic Stratigraphy – Applications to Hydrocarbon uplift across the Himalaya of Central Nepal. J. Geophys. Res., Exploration (Ed. by C.E. Payton) AAPG Mem., 26,49–212. 106, 561–526, 592. doi: 10.1029/2001JB000359 WILLIAMS, G.P. (1989) Sediment concentration versus water LIMMER, D.R., HENSTOCK, T.J., GIOSAN, L., PONTON, C., TAB- discharge during single hydrologic events in rivers. J. Hydrol., REZ, A.R., MACDONALD, D.I.M. & CLIFT, P.D. (2012) Impacts 111,89–106. of sediment supply and local tectonics on clinoform distribu- WITTMANN, H., von BLANCKENBURG, F., MAURICE, L., GUYOT, tion: the seismic stratigraphy of the mid Pleistocene-Holocene J.-L., FILIZOLA,N.&KUBIK, P.W. (2011) Sediment produc- Indus Shelf. Mar. Geophys. Res., 33, 251–267. doi:10.1007/ tion and delivery in the Amazon River basin quantified by in s11001-012-9160-6 situ-produced cosmogenic nuclides and recent river loads. LUPKER, M., FRANCE-LANORD, C., LAVE, J., BOUCHEZ, J., GALY, Geol. Soc. Am. Bull., 123, 934–950. doi:10.1130/B30317.1 V., METIVIER , F., GAILLARDET, J., LARTIGES,B.&MUGNIER, ZEITLER, P.K. & CHAMBERLAIN, C.P. (1991) Petrogenetic and J.L. (2011) A Rouse-based method to integrate the chemical tectonic significance of young leukogranites from the north- composition of river sediments: application to the Ganga western Himalaya, Pakistan. Tectonics, 10, 729–741. Basin. J. Geophys. Res., 116, F04012. doi:10.1029/ ZEITLER, P.K., CHAMBERLAIN, C.P. & SMITH, H.A. (1993) Syn- 2010JF001947 chronous anatexis, metamorphism, and rapid denudation at LUPKER, M., BLARD, P.-H., LAVE, J., FRANCE-LANORD, C., LE- Nanga-Parbat (Pakistan Himalaya). Geology, 21, 347–350. ANNI, L., PUCHOL, N., CHARREAU,J.&BOURLES, D. (2012) 10be-Derived Himalayan denudation rates and sediment budgets in the Ganga Basin. Earth Planet. Sci. Lett., 333–334, Manuscript received 08 October 2012; In revised form 146–156. doi:10.1016/j.epsl.2012.04.020 04 May 2013; Manuscript accepted 08 July 2013. MAHEO, G., PECHER, A., GUILLOT, S., ROLLAND,Y.&DELA- COURT, C. (2004) Exhumation of Neogene gneiss domes