suppose that the rampart conforms pretty nearly to the shape of the original marsh and of the tract of irrigated land which has replaced the marsh. The capital importance of ground-moisture in determining the formation of a dune-tract, or dune-massif, throughout the sandy districts which I visited became clearer to me towards the end of my stay. This subject, to which my attention was drawn by M. Doyen, is dealt with more fully at the end of the paper.

The surface of this dune-massif is modelled by two sets of waves, of which the crests are nearly at right angles, the ridges running from a little west of north and a little south of west respectively, these directions, and others mentioned in this paper, being compass-bearings taken with a common pocket-compass. Viewed from the cultivated land on the east, the summit of the rampart appears in waves of the familiar ripple form, the longer and flatter slope to the north, the shorter and steeper slope to the south.

Sample 4 was taken from the surface of a heap of material turned out from a drainage-canal which crosses the cultivated land. It consists mainly of quartz-looking particles, with some dust. The grains are not so well rounded as those of the dunes.

Returning to Ismailia, I went to the limit of the Casuarina plantations on the north of the town. Here a dune is in process of formation close to the road which follows the shore of Lake Timsah. It is about 140 feet broad and 8 feet high; the ends are slightly curved forward, and have no cliff. The central part of the lee slope is a cliff, but the summit of the dune is some distance to windward of the top of the cliff (compare the figure given by Captain Younghusband, on p. 99 of 'The Heart of a Continent'). April 16.-It will be convenient under this date to describe the observations made on several days of a dune to the north-west of Ismailia, which we came to know as "Doyen's Dune." A part of the crest forms the sky-line on the left in Fig. 25 of the paper on Sand-dunes in the Geographical Journal, March, 1897. The accompanying figure (Fig. 2) is a general view of this dune taken on May 13, 1899. I estimate its greatest height at about 40 feet. Its eastern end is on the summit of the rampart of sand which flanks the plantation of Abu Racan, shown in Fig. 25. At its western extremity, on the northern side, it abuts upon a large pool, which is a part of the marsh of Abu Racan. M. Doyen, at my request, had seventeen reeds put in along the top of the cliff at intervals of ten metres, commencing near the eastern, or plantation, end. This line of reeds extended from the eastern extremity of the cliff beyond the highest part of the dune, but did not reach by 70 paces the western extremity of the cliff. Beyond this, again, the dune extends as a rounded swell, until it merges in the undulating surface of the desert further west. stay at Ismailia a northerly wind blew steadily for strength in the afternoon being that of a stiff breeze.

Each day of our several hours, its At that time the

haze caused by sand flying from the top of the cliff was sufficient to spoil my photograph. The photograph (Fig. 2) was taken in the early morning. The following numbers show the march of the top of the cliff before the northerly wind. The numbers in the first (left-hand) column refer to

FIG. 2. DOYEN'S DUNE, ISMAILIA.

the upright reeds, No. 1 being at the eastern, and No. 17 towards the western end, the whole distance, from 1 to 17, being 170 metres in a line following the sinuosities of the cliff-top. The highest point is near No. 12.

The average rate of advance of the cliff-top is 0.925 inch per hour, or, excluding the one exceptional advance noted near the western extremity, 0-717 inch per hour. The results may be expressed with

sufficient precision if we say that the cliff-top advanced about threequarters inch per hour. This is the readiest means of obtaining a first notion of the rate of progression of the dune; but, as indicated by Fig. 3,

the advance of the cliff-top is not necessarily at all points identical with the real progress of the dune. The direction of the cliff-front varies at different points between reed No. 1 and reed No. 12 by 45°.

FIG. 5.-TWO PARALLEL RIDGES.

The

variation of direction of the ripples on the windward side of the cliff at the same points between reed 1 and reed 12 was 56°. This shows how greatly the direction of the wind is locally modified even on the windIward side of the dune by its various height at different points. The deviation from parallelism between ripplefront and cliff-front between reed 1 and reed 12 was from 1 to 4 points of the compass -that is to say, from 111° to 45°. This illustrates the fact that the march of the sand is not necessarily at right angles to the face of the cliff.

This is a good example of a dune the greater extension of which is transverse to the wind. On the lee side the direction of the wind is, locally, deflected through 90° in the horizontal plane. The sand comes sweeping round the ends, especially round the western end, and travels towards the middle, at the same time sweeping upwards and backwards. The activity of the wind on the lee side, and the increased capability for sorting which is conferred by the upward action of the eddying motion, is very striking. The lower part of the lee slope has a fairly hard surface, and is well rippled, the ripples facing in the reverse direction to those on the windward slope of the dune. The upper part of the

lee slope of the more developed and middle part of the dune is the wellknown straight cliff which has a soft, unrippled, slipping, plane surface. It has been customary to ascribe this slipping cliff to the pushing of sand-grains over the crest by the forward action of the wind and their downward rolling. This explanation implies that the cliff is constantly refaced by sand-grains thus driven forward from the surface of the windward slope. What I saw on this and other dunes during my stay in Egypt convinced me that this notion is incorrect. The cliff is due to undercutting by the backward-acting eddy. Were the sand coherent, an overhanging cornice would be formed, as I have seen happen during an easterly gale on a sandhill at Poole Haven. I have watched the same process when lying under a bank of drifting snow on the hills above Innsbruck (January, 1898). The loose dry sand, however, slips, so that

the spraying sand-wave has not the cusped form, but presents the curious anomaly of a straight line interpolated between curves moulded by the wind.

Fig. 3 is from a drawing made at the western end of Doyen's Dune, somewhat generalized to represent what appeared to be the aspect of the sand-dune profiles at successive stages of development. It is not precisely a diagram, for the intent was to represent the appearance of the slopes, which is generally steeper than reality. I hope in the future to replace this figure by careful measurements; meanwhile the following description of the probable course of development of the dune profile must suffice. The case considered is that in which the eddy on the lee has not cut down to hard rock. At first both windward and lee slope are very gentle; the highest point is near the middle. The summit probably moves to the leeward of the middle point, the lee slope certainly becomes steeper. A slipping cliff is formed on the upper part of the lee slope when the eddy has gained sufficient strength. The top of the cliff is worked back towards the summit; the windward slope meanwhile