Sec. ii] ACROSS THE EASTERNMOST TIEN-SHAN 531
passing this somewhat higher ground we came upon a dry deep-cut bed, about 3o yards across, which seemed to have its head in hills far away to the east. Its direction pointed to a junction lower down with the trough coming from Chin-êrh-ch`üan.
Beyond this the broken ground became level and then sloped down gently to a wide fan-like basin of gravel covered with scrub and furrowed by numerous shallow channels. These were all descending southward from the hills and thus carrying Tien-shan drainage. Unfortunately a haze, perhaps brought by the light north-westerly wind, had hid all distant mountains since the morning. Near the first larger channel the aneroid indicated an elevation of 4,100 feet, only slightly higher than that of Chin-êrh-ch`üan. The general direction followed by all these drainage beds was south-westerly, and this makes it quite certain that they, too, join the Yen-tun trough ', which the route from Ming-shui to Hâmi crosses near the halting-place of Wu-t`ung-wo-tzû (Map No. 37. c. 4), about twenty miles farther to the south-west.3 It will need further surveys to decide from which side this big dried-up river system of the Yen-tun trough ' received its chief tributaries. So much, however, is certain, that the area once drained by it extends much farther to the east than is shown by former maps.
We had moved to the north-west for more than twelve miles across the very gently rising glacis of gravel before we reached the first foot-hills of the Tien-shan. Passing up a defile about 300 yards wide at its bottom we came, after a vain search for water by our guide in a side valley northwestwards, upon a group of Toghraks and, close by, a spring amidst reed-beds. Its discovery was doubly welcome, for all traces of a track had disappeared since we left the gravel Sai. The view opening here northward across a plateau was wide. But the boldly serrated range that it showed in the distance raised doubts as to how, in the absence of reliable guidance, we should find a passage across it practicable for our already hard-tried camels.
On the morning of September 23rd our ` ta-lu-ti ' appeared to have recovered from his bewilderment and stoutly declared that he recognized the halting-place of Ta-hsi-k`ou, mentioned in the itinerary obtained at Mao-mei, in the spring at which we had camped. Stating that he now remembered his bearings, he led on across the plateau to the north-west. As this direction was evidently not such as to take us far away from Bai, our goal to the north of the range, I was ready to follow it. For seven miles we traversed a gently rising plateau, covered for the most part with detritus, from which there emerged clusters of low rocky ridges and knolls of what looked like trachyte. Then we struck a narrow Wadi-like nullah, with sandy bottom and abundant scrub along its banks. The cliffs on either side rose steeply to over a hundred feet and showed a strike from NW. to SE. with an almost vertical dip. Their faces, like those of the exposed rocks that we subsequently came upon, were far less decomposed than the hill-sides we had passed in the Pei-shan. We had advanced another three miles when the bottom of the valley widened, and the sand in what obviously was the bed of a temporary stream showed signs of moisture. On digging holes here we came upon deliciously fresh water at a depth of less than one foot. Equally encouraging was it to find good grazing for the ponies along the banks, and even some bushes of late-flowering wild roses.
As we ascended the valley farther, these signs of a comparatively moister climate continued, and when we had passed a reed-sheltered spring and some rough enclosures near it which looked like herdsmen's shelters, we began to hope that the faint track we were following would prove to