454 THE ETSIN-GOL DELTA AND THE RUINS OF KHARA-KHOTO [Chap. XIII
not be determined. On visiting it myself, I noted the almost total absence of potsherds near it. Yet the line of what looked like the embankment of an old canal seemed to pass it north-eastwards.
On June Ist I spent a long day, fortunately undisturbed by the prevailing sandstorms, in examining the ruins discovered in the course of Afraz-gul's reconnaissance. Proceeding northeastwards from Khara-khoto, we crossed the eastern branch of the dry river-bed at a point where its width was about 375 yards. Though occupied partly by small tamarisk-cones, its course was quite well marked. Beyond it we soon found the bare clayey ground covered with pottery debris, the sign of prolonged occupation, and came upon traces of two canals, about Io feet across at the top, which had once watered it. Their direction was to NNW., suggesting that the channel once feeding them came from the south and was quite distinct from the river-bed we had crossed. Then we passed eastwards for about two miles across a belt of dunes, 3o to 4o feet high, which extended parallel to the river in exactly the same way as the ` Dawâns' of sand that line dead river-beds in the Taklamakan 1
Beyond this stretched, as far as the eye could see, ground covered, in parts thickly, in others less closely, with tamarisk-cones. Only here and there was it broken by short ridges of dunes or open patches of sandy soil. The whole vividly brought back to my mind all the aspects of that area, now deceptive desert but once occupied by flourishing settlements, which I had repeatedly visited between Uzun-tati and Achma, to the north of Domoko.2 As soon as the riverine ` Dawân' lay behind us, we came upon the first of the numerous ruins, some small, some fairly large, that are scattered over this extensive area and undoubtedly mark old farms and homesteads. This first ruin, E. of K.K. Ir, as seen in Fig. 255, though quite small, presented features characteristic of most of these old dwellings. From the side of a big tamarisk-cone emerged the walls of a fair-sized room, built of regularly set and fairly uniform lumps of clay, with reed layers between every four courses, and resting on a thick foundation of what looked like hemp matting. The ground in front, not protected by the accumulation of fine sand that the tamarisk roots held together, had undergone wind-erosion and been lowered 5 or 6 feet below the level of the foundation. This effect of wind-erosion was equally visible in small Yârdangs, from 3 to 5 feet in height, which appeared almost everywhere on open patches of ground, though not so close together as in the Lou-lan area. Their general direction was from north-west to south-east, indicating the prevailing direction of the winds.
The remains of dwellings, which could be traced at intervals for a distance of over four miles eastwards, were mostly larger, as appears from the plans of the structures marked E. of K.K. III, Iv, VII, VIII, x, in PI. 20, 22. But the constructive features were the same, and the conditions of the ground on which they survived similar. No detailed description of individual ruins is therefore needed here. The considerable number of rooms comprised in the better preserved of the ruins indicated a comparatively high standard of rustic comfort, such as I had found generally prevailing in the Chinese agriculturists' quarters that I had seen in Kan-su. Near most of them there still rose the dead trunks of elms and other cultivated trees, as seen in Fig. 253, which represents dwelling Iv. From the regular arrangement of the quarters, from the trees that had been planted near them and the abundance of pottery remains of superior type, including much glazed ware and porcelain, I was led to conclude that the agricultural settlement which once flourished here must have been essentially Chinese in its population.
This impression was greatly strengthened when we resumed our survey of the ruins eastwards, after a reconnaissance of some miles to the north-east of ruin vi had brought us to the apparent limit