The lost river and the Indus Civilisation

How river dynamics influenced a Bronze-age civilisation

During the early to mid-third millennium BCE, the Indus Civilisation developed one of the most extensive urban cultures in the Old World. This civilisation was established on the alluvial plains of the Indo-Gangetic basin in northwestern India and Pakistan, with an urban phase commencing ~4.6-4.5 ka B.P. It was contemporaneous with and more extensive in area than the earliest urban societies of Egypt and Mesopotamia, encompassing an area estimated at ~1 million km2.
The largest concentration of Indus Civilisation settlements is located near the divide between the Ganges-Yamuna and Indus river systems in India and Pakistan, far from major active rivers. Why numerous Indus settlements should have been located in a region now devoid of large perennial rivers has been the subject of vigorous debate and controversy.
In a newly published paper Counter-intuitive influence of Himalayan river morphodynamics on Indus Civilisation urban settlements in Nature Communications we investigate this problem.

CF Oldham's 1874 map of dry river beds in 
NW India and Pakistan

For well over 100 years, the traces of a large dry river bed on the northwest Indo-Gangetic Plains have been ascribed to flow of a now defunct major Himalayan river. The discovery of numerous Indus urban settlements apparently clustered along this purported river channel led to suggestions that perennial Himalayan river flow sustained settlements in this region and was crucial to onset of urbanisation. Moreover it has been proposed that drying up of the river led to the end of the urban phase in the civilization at about 3.9 ka BP.

Today only a sluggish monsoon-fed seasonal river flows along the trace of the palaeochannel.

So what formed the palaeochannel and what was its relationship to Indus Civilisation urban sites?

An ancient street at Kalibangan

Kalibangan in Rajasthan is one of the most important Indus urban sites in NW India. Built along the margin of the Ghaggar-Hakra palaeochannel, it is comprised of two major walled mounds containing regular house plans, and a grid of streets. The site is located topographically above the palaeochannel floor on the southern edge of the Ghaggar-Hakra palaeochannel.

Kalibangan is an incredible archaeological site. The main site is now buried to preserve it from disintegration but wandering across it, the surface is littered with fragments of pottery and bangles.

Other sites such as Banawali and Kunal also show a close spatial relationship to the palaeochannel.

It has generally been assumed that a large urban site such as Kalibangan must have been sustained by a large Himalayan river that flowed along the Ghaggar-Hakra palaeochannel.

We aimed to investigate this question.

Mapping the lost river by satellite

Trace of Ghaggar-Hakra palaeochannel on northwestern Indo-Gangetic plain. Landsat 5 TM colour composite satellite image mosaic.

To map the lost river afresh, we began by mapping the surface geomorphology. For such an extensive region, remote sensing (also known as Earth Observation or satellite imaging) is the best tool to use. NASA's Landsat program has made freely available its 40 year near-continuous global archive of satellite images (

The Landsat satellite digitally images the Earth, from its orbit 700 km above, capturing the brightness of the suns reflectance from the ground across a range of wavelengths. It measures reflected light in the visible part of the spectrum (blue, green and red light) and in the infra-red (near, short-wave and thermal infra-red) as a series of discrete spectral bands.

These Landsat multi-spectral images can be manipulated to reveal the chemical signatures of different rock and soil types, vegetation species, wet and dry ground and much more.

Each Landsat image is large (185 x185 km) but the area we are interested in is very large indeed and so we needed 12 Landsat images to map it all.

We selected Landsat images acquired in 1998 (at a time before the most noticeable increase in urban development and expansion of intensive cultivation), and from the months immediately following the monsoon rains (it can be challenging to find images that are not covered in thick clouds).

Next we constructed some colour composite images, using spectral bands carefully chosen to discriminate cool, damp, dark soils from warm, dry, bright ones. When we found the right combination of bands (spectral bands 4, 5 & 6 RGB), we repeated the process for all 16 images and then merged them to form a mosaic of NW India.

The result was spectacular and showed us, very clearly, that the channels of all the major rivers of this region were characterised by cool, damp and relatively dark soils, and that the soils were different from the surrounding alluvial fan sediments. One of these cool, damp channels was devoid of a major active river but had the similar geomorphological characteristics as the others: the Ghaggar-Hakra palaeochannel. This single 456 RGB image mosaic provided an essential tool to guide our drilling and sampling program that followed.

To better understand the topography of the area, we used the Shuttle Radar Topographic Mission (SRTM) 1 arc second (30 m) Digital Elevation Model or DEM ( This dataset provides unparalleled detail of the topography and was invaluable for mapping and visualising the terrain of the entire area and specifically of the palaeochannel.

At first glance the topography of the Indus-Ganga plain seems very flat and a little featureless, and it is hard to see the subtle shapes of the various river channels. However the overall topographic relief between the Himalayan front and the coast is about 300 m, over a distance of 1200 km, and it became quickly apparent that the dominant component of the topography across this region is the gentle slope, of about one degree, from the NE to the SW.

To compensate for this slope effect we calculated the average elevation, i.e. the gradient, over a moving window of ca 500 m in size, and subtracted it from the raw elevation value, on a pixel-by-pixel basis, to derive a map of relative elevation change. This slope-corrected relative elevation shows vividly that the 'lost river' palaeochannel forms a clearly and sinuous depression in the ground along its entire length.

Is the 'lost river' a lost river?

To test the hypotheses that (1) the Ghaggar-Hakra palaeochannel hosted a major Himalayan river, and (2) that its abandonment coincided with Indus urban settlement decline, we drilled five cores perpendicular to the axis of the palaeochannel adjacent to the important Indus site of Kalibangan in Rajasthan

We identified five sites for sediment coring along a 12 km long transect across the palaeochannel. The sites were located on the palaeochannel and also few sites outside the palaeochannel to cover cross-section of the palaeovalley.

Core drilling was performed by rotary drilling using a diamond core bit mounted on the core tube barrel installed on a calyx drilling rig and rotated using a electric power generator.

Sediment cores were retrieved in PVC pipes (~63 mm diameter) inserted in the core tube. After a drilling run of about 1 to 1.5 meters, depending on the lithology, the whole drilling assembly was taken out and the drill bit was removed to pull the PVC pipe out of the drilling core tube barrel.

Generally in textbooks and research articles, a sedimentary succession is described bottom to top to build up sequence of events observed in a stratigraphic succession. However, sediment coring and drilling progresses top to bottom and as a field geologist at a drilling site our first interaction with stratigraphy also progresses the same way. That is we collect core cuttings in the drill bit every 1 to 1.5 meters and the first sedimentary succession that we see is from top and then we move bottom wards as the drilling progresses.

Yet it is equally interesting as we unfold the story of sedimentary succession starting from present and going back in time as we see more of older sediment from base of the drill hole. The cores located at centre of Ghaggar-Harkra palaeochannel begin with core cutting red-brown silty clays that quite similar in appearance with modern Ghaggar river. This became interesting after a drilling to depth of 8 meters when we begin to see silty sand and then fine sand in the core cuttings. This first observation of first major change in characteristic of sediment beneath the palaeochannel was encouraging in the direction of search of deposits of a large river.

With further drilling we began to observe medium to coarse grey micaceous sand that reminds of modern sand in present day river bed of Ganga, Yamuna or Sutlej with characteristic salt and pepper texture. As the drilling progressed we knew we were looking into fining upward fluvial successions which were continuously present to the depth of 35 to 40 meters after which drilling cutting changed to finer sediment. This yellow silty sand deposit was quite different from the grey micaceous sand and had more resemblance to modern dune sand abundantly present in the area nearby the drilling site. Again this was an equally exciting observation as it suggested that we had touched the base of the major fluvial succession.

Probing the lost river

Analysis of the sedimentology of the Ghaggar-Hakra palaeochannel enables us to understand the direct connection between river morphodynamics and Indus settlements.

The cores are dominated by a ~30-m-thick fining-up succession of unconsolidated, dark grey, mica-rich, coarse- to fine-grained sand. The grain size, poor to moderate sorting and abundance of angular grains in the sands indicate high-energy fluvial channel deposits.

Beneath the surface trace of the palaeochannel, in cores GS7 and GS10, the grey fluvial sands are overlain by an ~8-m-thick fining-up succession that shows upward transition from brown very fine sand and silt into reddish-brown silty clay. The abrupt grain size change from the grey sand likely records a cessation of high-energy fluvial deposition and the onset of low-energy fluvial activity and suspension fall-out from standing, ponded water on floodplains.

Where did the lost river come from?

What was the source of the river sands in our cores? Were they delivered by a big Himalayan river or did these sands come from small, seasonal rivers derived from the foothills of the Himalaya? 
To determine what geologists call the 'provenance' of the river sands we use isotopic ‘fingerprinting’ to identify source areas for river sands using the minerals zircon and mica.

Dating zircons

Zircon is a stable mineral that is common to a wide range of rock types which makes it ideal as an indicator of sediment provenance. Zircon also contains levels of uranium and lead suitable for measuring geological ages and detrital zircon uranium-lead (U-Pb) dating has become the tool of choice for geologists who want to learn about where sand grains came from.

Zircon analyses for this study was carried out at the London Geochronology Centre and used Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA-ICPMS). This involved using an ultraviolet laser to vapourise part of a zircon grain and transporting the zircon aerosol into an ion source by carrier gas. An argon plasma at temperatures comparable to the surface of the sun, then breaks any remaining molecular bonds and converts them into ions which are channelled into a quadrupole mass analyser to detect ions by their mass-to-charge ratio. A detector measured abundances.

To prepare samples for analysis, and avoid bias from hand picking, heavy mineral concentrates rich in zircon were mounted in epoxy resin on glass slides, polished and placed in a laser cell. Slides were systematically scanned to locate zircons for laser drilling using a typical spot size of 25 microns. Typically 100-150 zircons were analysed for each sample, with reference to age standards to correct for elemental fractionation and any drift in machine sensitivity.

Dating Micas

To further constrain the source of the sand in cores, we also analysed the age of mica grains extracted from core samples from the Ghaggar-Hakra palaeochannel at the NERC Argon Isotope Facility at the Scottish Universities Environmental Research Centre (SUERC) in East Kilbride. To work out the age of the mica grains we employed a radioisotopic dating approach. Mica contains abundant potassium and an isotope of potassium, 40K, is naturally unstable, radioactively decaying to an isotope of argon (40Ar) at a known rate.

Determining the age is as simple as working out how much 40K is left in the mineral, how much 40Ar has been produced in the mineral, and using the known decay rate to calculate an age. All of these measurements are made using sophisticated noble gas mass spectrometers that isolate the minerals from our atmosphere. Because the atmosphere contains 1% argon it would swamp the argon signal we are trying to measure in the mica grains and lead to an erroneous age. Therefore we have to use large pumps to suck all the atmosphere out of our mass spectrometers before we can fuse the mica grain using a laser and release its argon budget for measurement. 

It took the NERC Argon Isotope Facility approximately 6 weeks of constant 24 hours per day operation to measure all the grains that are reported in our study.

Our zircon and mica provenance analysis revealed that the coarser-grained river sands were actually sands derived from the Sutlej River, which now flows some 150 km to the west. Our isotopic 'fingerprinting' suggest that the Sutlej River once flowed along the Ghaggar-Hakra palaeochannel.

When did the lost river flow?

Dating sand grains using optically stimulated luminescence at the DTU-Aarhus Risø laboratory

To find out when the sediments had been deposited and thus the timing of major river flow we used optically stimulated luminescence (OSL) dating of mineral grains extracted from the sediment. When sediments containing minerals such as quartz and feldspar are buried beneath the ground, background environmental radiation results in energy being stored in mineral grains. Before burial, the mineral grains are exposed to daylight in the river, and this empties any previously stored energy. Then, after burial, the grains are hidden from light, energy is stored, and the amount of energy built up represents the amount of time since burial. This stored energy was measured using OSL in the laboratory and used to pinpoint when the layers of sediment were buried, which told us the timing of river flow and abandonment.

Optically stimulated luminescence (OSL) dating determines the time elapsed since the last exposure of sediment grains to daylight (the burial age); this light exposure resets any prior latent OSL signal. During burial, mineral grains are exposed to environmental ionising radiation resulting in the creation of free charge. Some part of this free charge is trapped at defects and the net trapped charge is related to the total radiation exposure. In the case of quartz and feldspar, optical stimulation can release this trapped charge resulting in the emission of luminescence called OSL. The OSL signal can be calibrated in terms of absorbed dose (the equivalent dose, De; expressed in Gy). The burial age is then determined by dividing the De by the independently measured environmental dose rate (expressed in Gy ka-1).

After several years of careful analysis, we were able to pull together a detailed chronology of our cores based on the feldspar luminescence ages. The ages reveal that the Himalayan Sutlej River had abandoned the Ghaggar-Hakra palaeochannel several thousand years before the Indus Civilisation settlements.

Instead a low-energy river system deposited very fine-grained sediments in the centre of the palaeochannel.

We deduce that no Himalayan river flowed adjacent to the abundant Indus urban settlements along the Ghaggar-Hakra palaeochannel. The Indus people must have been reliant on monsoon rainfall-derived river flow.
So what was this palaeochannel that is so distinct in the satellite images?

Our analysis suggests that the palaeochannel is a former course of the Sutlej River. Between 15 000 and 8000 years ago the Sutlej avulsed (migrated) to its present day course to the northwest. The abandonment of the former course left a topographic low formed by the former channel in the landscape.

We suggest this provided an ideal environment for the Indus people to have inhabited and built their settlements. The abandoned channel would have captured monsoon-derived river flow and enabled ponding of water. Muddy sediment would have settled out of suspension providing rich soils. And importantly with the big Himalayan river hundreds of kilometres away, the palaeochannel would not have been prone to catastrophic floods.

The former valley of the Sutlej likely offered favourable conditions that led Indus populations to preferentially settle and flourish along its banks. 

This offers an alternative model for how rivers can nucleate the development of ancient urban settlements.

See Counter-intuitive influence of Himalayan river morphodynamics on Indus Civilisation urban settlements

Ajit Singh, Kristina J. Thomsen, Rajiv Sinha, Jan-Pieter Buylaert, Andrew Carter, Darren F. Mark, Philippa J. Mason, Alexander L. Densmore, Andrew S. Murray, Mayank Jain, Debajyoti Paul & Sanjeev Gupta

Nature Communications 8, Article number: 1617 (2017) doi:10.1038/s41467-017-01643-9


Philippa Mason - Remote Sensing

Ajit Singh - Drilling and sedimentology

Andy Carter & Darren Mark - Provenance analysis

Kristina Thomsen & Andrew Murray - OSL dating