Roda, Cairo · annual minima 622–1284
The Nilometer of Roda: 663 years of the Nile’s memory
On Roda island in Cairo, a graduated column stands in a stone well that tunnels once filled from the Nile. For every year from 622 to 1284, chronicles kept the river’s lowest level, most of it read at Roda. Pick a span of years, your age or a reign, and compare the averages over such spans with the same 663 years in random order.
A column in a well
The nilometer on Roda island dates from 861, successor to an earlier gauge finished there around 715. Three tunnels once let Nile water into its stone well, where a graduated column stands; the building has been repaired and restored many times. The record compiled by the Egyptian historian Omar Toussoun in 1925 gives the lowest level of each year from 622 to 1284: 663 numbers.
The readings were taken in cubits and fingers, 24 fingers to the cubit; Toussoun converted them to metres above sea level on the historical Roda scale. The 663 minima run from 9.35 to 14.66 metres, around a mean of 11.48. The earliest come from before Roda had a gauge: Toussoun thought they came from Memphis and put them on the same scale, and other gauges near Cairo have also been suggested. So the series is best read as the Nile’s low water near Cairo, mostly at Roda.
Low-water years come in runs
Shuffle those 663 years and each keeps its value; only the order changes. Year by year the two series have the same spread. Average over spans and they part. If the years were independent, averages over 30 years would be about 5.5 times less spread out than single years, since 5.5 is the square root of 30. The Nile’s averages narrow far more slowly: at 30 years they are more than three times as spread out as independence would give, and at 70 years about four times.
Low years follow low years. After a year below the median, the next year was below it too about 70 times in 100; in the shuffled series, about half. The longest run below the median lasted 33 years, from 758 to 790.
Hurst’s number
The British hydrologist Harold Edwin Hurst spent decades studying the Nile. In 1951 he showed why clusters of low years matter when sizing a reservoir to supply a given amount of water, and that natural series cluster more than chance would suggest. The exponent named after him, H, measures how slowly the spread of averages shrinks as the span grows: 0.5 for independent years, nearer 1 the slower it shrinks. In 1968 Benoît Mandelbrot and James Wallis called this kind of persistence the Joseph effect, after the years of plenty and famine in Genesis.
In a chapter written in 2004 and published in 2005, the hydrologist Demetris Koutsoyiannis plotted the spread of block totals for these 663 Roda minima and found H = 0.85. This page uses block averages, which give the same H, and estimates 0.852 over spans of 1 to 50 years. Five fixed shuffles of the same numbers give between 0.43 and 0.55; other shuffles vary.
What the record does not say
H describes this record over these spans; it does not explain it. Most of the Nile’s summer flood comes from rain over the Ethiopian highlands, and that rain varies from decade to decade, but the cause of the runs in this low-water record remains open. Nothing here is a prediction.
The readings were preserved in medieval chronicles and compiled by Toussoun, and the gauges changed over the centuries. Changes in the river bed, the gauge or its scale can add slow trends that look like memory. Nothing shows that those who read the column knew of the persistence in their numbers.
Sources
Every figure on this page is traced in the project’s source notes.
- Omar Toussoun, Mémoire sur l’histoire du Nil, Mémoires de l’Institut d’Égypte, 1925, pp. 366–385 (the minima); digitized as
NileMinin the R package longmemo (J. Beran), GPL. - Demetris Koutsoyiannis, “Hydrologic persistence and the Hurst phenomenon”, in Water Encyclopedia, Wiley, 2005 (2004 preprint, ITIA): the standard deviation of block totals for the Roda minima, H = 0.85.
- H. E. Hurst, “Long-term storage capacity of reservoirs”, Transactions of the American Society of Civil Engineers 116, 1951, 770–799.
- B. B. Mandelbrot and J. R. Wallis, “Noah, Joseph, and operational hydrology”, Water Resources Research 4 (5), 1968, 909–918.
- Jan Beran, Statistics for Long-Memory Processes, Chapman & Hall, 1994, data set 1.