presented a terribly eroded appearance, with the ground cut up into Ydrdang trenches and ridges, or, as along the west and north walls (Fig. 162), carved out into deep continuous hollows. No doubt the very resistance offered by the solid fort walls, obstructing the winds in their progress, was causing them to exert their destructive force with increased intensity on the ground within. A violent sandstorm from the east-north-east on the second night of my stay gave me a very trying personal experience of this. Regular whirlpools of sand raged around our bivouac in one of the hollows below the north wall, making it impossible to keep a fire alight.
It is easy to understand that under conditions such as these only the scantiest remains of the structures that once occupied the interior could survive. Quantities of red potsherds of great strength were strewn about the eroded soil both within the walls where this was clear of sand, and to a less extent outside, a proof of prolonged occupation. Clay ` witnesses ', rising here and there above the flat top of the Ydrdang ridges, probably indicated the position once occupied by quarters now completely effaced. But only at two places could slight remains of structures actually be traced. At a distance of about twenty-four yards from the north wall the brickwork foundations of a building, ii, could barely be traced on the top of a terrace rising more than twenty feet above the eroded hollows near by. Judging from what could be made out of the line of walls north and west, the area occupied by it measured at least about seventy by thirty-five feet. Fissured beams, about a foot square in section and up to twenty-six feet in length, lay about in utter confusion and suggested that the foundations and roof of the building had been of very massive construction. Any smaller debris had been completely disintegrated and blown away long ago, and the search of the ground here proved fruitless.
On the other hand, a narrow ledge of uneroded ground adjoining the north gate on the inside retained short portions of the walls of a structure built of timber and tamarisk wattle. An accumulation of refuse, mostly reed-straw and dung, had here protected the foundation beams together with a foot or so of the wattle wall rising above them. Timber debris strewing the slope below obviously belonged to the foundation of other walls that had been completely eroded. On clearing the refuse we recovered here three Chinese records on wood, L.E. i. 1, 2, 6, besides a rolled-up paper document, apparently complete, in Chinese, L.E. i. 3, and two Chinese paper fragments. The shape of the inscribed wooden tablets, resembling a ` wedge ' cover-tablet, with a raised seal socket at the square end and with the pointed end cut off, represents a new type. [M. Maspero's preliminary notes show that these tablets had served as address labels for official letters, L.E. i, 2 bearing dates corresponding to A. D. 266 and 267, respectively.] From this and from the presence of records on paper it is clear that the ancient castrum must have continued to see traffic down to approximately the same period as the Lou-lan station L.A. Apart from these records and a Chinese coin of the Wu-chu type the only finds from the site comprise two bronze arrow-heads, L.E. 01-2 (Pl. XXIV), and a few small stone implements including a leaf-shaped point, L.E. 05 (Pl. XXII), evidently neolithic. The latter were picked up in the immediate vicinity of the fort.
On the morning of February 16th, after taking a few photographs of the fort, intended for preliminary record only and reproduced in Figs. 160-63, I set out for the conspicuous Mesa visible to the north-east, on which Afrâz-gul on his reconnaissance had noticed remains of ancient occupation as well as some graves. It was reached after a march of about two and a half miles across ground covered with low Yardangs and in two places with patches of shôr.8 The salt crust had a soft surface as if due to deposit of no very distant date. Among the dead reeds and small tamarisk bushes met here and there between the Ydrdang ridges some looked as if they had died but recently.
8 The two Mesa symbols shown in Map No. 32. A. 3 to the SW. of L.F. should be replaced by Yârdang symbols.