166 ON THE WAY TO LOP-NOR [Chap. V
of any definite evidence I must content myself with the observation that the remains brought to light at the ruins, scanty as they are, including the fragments of manuscripts, point to the site having been occupied during early Tang times.'
The second site to the south, known as Bash-koyumal, also proved to be small, but showed some features of interest. It lay about one mile and three-quarters to the SSW. of Koyumal, beyond some shallow branches of the river in which the water, under the protection of its cover of ice, was then spreading over the gravel plain. Above the western bank of what evidently is a main bed of the Charkhlik river, about 25 yards wide with water flowing freely down its centre, there rises very steeply a plateau of gravel to a height of about 25 feet. The north-eastern end of this plateau overlooks the head of a canal carrying water to the lands of Tatran opposite Charkhlik proper. At that end there survives part of a circumvallation, semilunar in its present form and extending from one point of the plateau edge to another some 205 feet farther south (see sketch plan, Pl. 9). The extant portion of the wall line is not exactly the segment of a circle, but shows six facets, each about 45 feet long on the outside. Assuming the plan originally to have been that of a sixteen-sided polygon, the diameter of the whole circumvallation may have been approximately 2I0 feet.
The enclosing wall built of sun-dried bricks, 17" x 9" x 4", showed a thickness of 4 feet 9 inches and on the west, where best preserved, still stood to a height of over 6 feet. Elsewhere it had decayed into the appearance of a low mound, but its original thickness could be traced even there on excavation. Outside, a well-marked depression running along the wall indicates the former existence of a ditch. The enclosing wall both on the north and south breaks off abruptly where it reaches the plateau edge. As this descends very steeply to the river-bed, the conclusion suggests itself that the remainder of the circumvallation, together with the eastern half of the area enclosed, has been washed away by prolonged encroachment of the river as it sweeps against its right bank.
This conclusion is supported by the survival of a massive wall (marked II in plan), about Io feet thick and over 5o feet long, close to the edge of the plateau and towards the centre of the area (Fig. 109). Its position suggests that it belonged to a central keep or tower of the ruined fort, probably square in shape, the other walls of which have disappeared owing to erosion of the ground. At the meeting-point of two facets of the enclosing wall on the west, a gate about 7 feet wide leads into the interior of the small fort. Immediately to the south of a broken wall running from this gate towards the supposed central structure stands a much-decayed mound, which on excavation was found to contain the square base of a Stûpa or shrine with an enclosing passage (marked 1 in plan, Pl. 9).
This base, badly broken on all sides, as Fig. 107 shows, probably by treasure-seekers' operations, measured 12 feet square and at its centre still rose to about 9 feet in height. It was built of sun-dried bricks measuring 17" x 9" x 4", like those found at Koyumal. Its faces, except on the east where the foundation of stairs was traceable, were found to retain at their foot remains of relievo decoration in plaster, consisting of rows of niches, five on each face. These niches, much injured and nowhere rising to more than a foot and a half, were 2 feet wide and were once divided by pilasters, probably resembling those in the Mirân shrine M. 11.8 In some of the niches on the northern and western sides the feet of small stucco figures could still be distinguished.