446 KARA-DONG AND THE SEARCH FOR HSÜAN-TSANG'S P`I-MO [Chap. XIII
declared that we were near the high tamarisk-covered cones that had given the site its name, Kara-dong ` the Black Hillocks'. But as in the blinding dust they could not make sure of the exact direction I let them go ahead, while we sought shelter under the lee of a large sand-cone. It was interesting to watch the sand being driven in a thick spray over the crest line of the dunes just as if they were storm-tossed waves. After half an hour Muhammad Shah returned with the news that the ruins were due west of us, and not far off. In proof he brought a piece of old pottery he had picked up. So the march was resumed just as the force of the storm showed signs of abating, and after another two miles over dunes rising from about 15 to 25 ft. in height we reached easier ground, where pottery fragments appeared in occasional patches between dunes only 5 to io ft. high. Going north-west for about half a mile we arrived at a series of remarkably high sand-cones close to the western edge of this relatively open area, and hidden between them sighted the ruins I had come in search of 7.
The remains of Kara-dong proved to consist mainly of the ruined quadrangle shown by the plan on Plate XXXVIII. Portions of its southern and eastern faces are seen in Fig. 52, while the photograph (Fig. 53) reproduces the appearance of the ruin from the east near what proved to be the entrance gate. Closer examination showed that the quadrangle was formed by a mud rampart, which was once occupied on its top by rows of rooms built of timber and plaster. Except for lines of posts still rising near the south-east corner (see Fig. 52), splintered beams indicating foundations of cross walls, and similar scanty remains, the walls of the structures once lining the top of the rampart had so badly decayed through erosion that the approximate dimensions and disposition of the quadrangle could be ascertained only after a prolonged examination of all details and repeated surveys. These showed that the quadrangle, as marked by the outer walls of the rooms still traceable on the top of the rampart, had measured approximately 236 ft. square. There were indications both on the northern and western faces that the rows of rooms had been originally double. But as even the foundations of the walls of the second row could be traced only in the form of rotten timber débris strewing the slopes of the rampart where not hidden away by the dunes, no measurements could be obtained. Only in the north-west corner could the dimensions of two small rooms, set back to back, and nearly I I ft. square each, be exactly ascertained 8. Here, too, the timber and plaster work of the walls reached nowhere higher than about i ft. above the mud flooring.
But while all extant remains plainly indicated the havoc wrought by excessive erosion, to which the ruin must have been exposed for a prolonged period, its survey was rendered still more difficult by the heavy sand that had subsequently invaded it. Two large dunes stretched diagonally across the quadrangle, rising within it to heights of more than 20 ft. above the original ground-level, and connecting outside with the neighbouring sand-cones, which showed elevations up to 5o ft. (see plan). Where the axis of these dunes lay within the quadrangle it was impossible to see whether they covered any structural remains, and excavation in any
case would have been an affair of weeks. But in the receding angle between the two dunes, where the sand was lower, there appeared the timber framework of a building (K. i. in plan),
measuring about 48 by 26 ft. It is seen, after excavation, in the centre of Fig. 52. Its
clearing proved very difficult, as the plaster of the walls had completely disappeared, and the sand from the slope of the dune behind kept pouring down into the area excavated. The