Why did the Indus cities empty?

A 2012 study reconstructed the rivers of the region using satellite topography, field survey and dating of sediments, and mapped settlements against them.

The heartland sat on rivers fed by monsoon rain, not by Himalayan meltwater. That is the study’s central negative result: the large glacier-fed river some identify with the mythical Sarasvati did not flow across that ground while the cities stood. As precipitation fell, lowest after about 2000 BC, the rain-fed rivers dried or became seasonal. The Indus and its Himalayan tributaries kept flowing, and flooded less rather than more.

Settlement, in the study’s reading, follows the water: sites shift east towards regions where monsoon rain was more reliable, and the settlements along the drying Ghaggar-Hakra are largely abandoned.

Not everyone reads the rivers that way. A 2017 study of the same valley found that the Sutlej ran down it until about 8,000 years ago, and that the monsoon-fed flow which replaced it had already declined before the cities there were founded. Its authors say that rules out a causal link in that region, and that the river dates are too coarse to test the climate account of the decline.

A 2014 study of a dried lake bed in Haryana recorded a sharp weakening of the summer monsoon about 4,100 years ago, from oxygen isotopes in the shells of snails preserved in the sediment. Earlier climate records for the question had all come from well outside the area the Indus cities occupied. This one is next to them.

That date falls inside the radiocarbon range for the beginning of de-urbanisation. It is a good fit and the authors do not claim more than that: their conclusion is that climate may have played a role, and that both the climate dates and the archaeological ones carry uncertainty.

Abrupt drying around 4,200 years ago has been read across west Asia and the eastern Mediterranean too, with societies in west Asia adapting to smaller harvests. Whether that was one event is now disputed. A 2024 synthesis of 1,142 climate records found little support for a drought centred on 4,200 years ago in the Mediterranean and the Middle East; the largest dry region it found at that time was in southern Asia.

Earlier accounts invoked invasion, on the strength of a few skeletons at Mohenjo-daro and of later textual traditions. The published evidence for those skeletons was re-evaluated in 1964: about thirty-seven of them from nine years of excavation, many incomplete, found in contexts that do not establish they belong to one moment, and not one of them inside the citadel, which is where a last defence could be expected. There is no destruction level over the latest period of the city and no sign of extensive burning.

The positive account that re-evaluation offered was a sudden rise of the coastline, worse floods and salted soil. The river work points the other way: the landscape grew more stable and the floods less intense as the region dried.

The drying and the dispersal overlap broadly in time. The mechanism connecting them, in detail, is inferred rather than documented, and the monsoon weakening at about 4,100 years ago sits a couple of centuries ahead of the settlement decline that begins about 3,900 years ago. The 2012 river study reads the drought of 4,200 years ago as falling while the cities flourished, and puts the damage in the further drying that followed.

A 2025 study that ran climate simulations through a river model finds likely river droughts, lasting decades to centuries, across the Indus basin between about 4,400 and 3,400 years ago. Its authors argue that less water may have dispersed people from the cities, and say climate was one pressure among social and economic ones.

Trade with Mesopotamia contracted over the same period: Meluhha, the Indus, is named regularly in Mesopotamian records from 2400 to 2000 BC, then largely stops being mentioned, and the material evidence of exchange thins with it. Internal social change is hard to read without texts, though the skeletons carry some of it: at Harappa, violence and infectious disease both rose after the cities declined.

The Indus script has not been deciphered, so there is no administrative record anyone can read, no chronicle and no statement by anyone who lived through it. It disappears with the cities.

Everything above is inferred from objects and sediments. A civilisation spread over a million square kilometres has left no readable sentence.