After 1898 the Nile carried 22% less water, and 99 years in 100 fell below the old median

How different was the Nile before and after 1898? Each row is one year of flow past Aswan, 1871 to 1970, in hundreds of millions of cubic metres. 1898 is the usual break point in this record: construction of the first Aswan dam began that year, and statisticians have used the series ever since as the textbook case of a level shift. The shift is not subtle. The two eras barely overlap.

Interactive chart available when JavaScript loads.

A step down, not a slide

The 27 years before 1898 averaged 1097.67 (10⁸ m³) a year. The 73 years from 1898 on averaged 853.40, a drop of 244, or 22%. The line does not drift down to the new level; it arrives there within two years of the break and stays for seven decades.

Interactive chart available when JavaScript loads.

The later era's typical year was a dry year before

Two boxes on one scale, with the pre-1898 median drawn across both. The median year before 1898 carried 1140; from 1898 on, 98.6% of years fell below that figure. Run it the other way and 92.6% of pre-1898 years sat above the later era's median of 845.

Interactive chart available when JavaScript loads.

The whole distribution moved, not just the middle

The cumulative curves barely touch. The driest year before 1898 carried 799, and 34.2% of later years came in below it. The wettest year from 1898 on carried 1170, and 29.6% of earlier years came in above it. A wet year after the break was an ordinary year before it.

Interactive chart available when JavaScript loads.

Takeaway

The Nile at Aswan carried about a fifth less water after 1898, and the change was a step: the eras' middle halves do not overlap, and nearly every year in the later era would have been a dry year in the earlier one. Any century-long "average flow" for this river is an average of two different rivers.

Limits

The break year is a convention from the literature, not an experiment; the dam, instrument changes, and a shift in East African rainfall are all candidates, and the data cannot separate them. Flow is measured at one point on the river, and 27 years is a short "before".