depressions, could no longer be reached at the next two camps. But some miles before Camp 320, at a point marked in Map No. 36. B. 2, I noticed the first indication of an ancient river-bed in a narrow steep-banked depression, running from south to north, half smothered by dunes but undoubtedly cut by water action.
On the evening of the following day, February 6, having shaped our course now to S. 190° W., we reached a big ` Dawan', rising to 30o feet or so. Surmounting it we found ourselves suddenly before a belt of dead forest that unmistakably marked the northern end of an old dried-up delta of the Keriya River. Here at Camp 321 (Map No. 36. A. 2) a well yielding fresh water could be dug in a depression eroded ages ago by the dying river. There could be no doubt here that we had rightly struck, as expected, the ancient termination of the dune-covered delta in which the Keriya River loses itself. Yet it was also at this point that the most trying portion of our desert crossing commenced. For the difficulties encountered through want of water and guidance in this most deceptive dead delta I may refer to my Personal Narrative. It will suffice here to state that it took us six more marches before we discovered the shallow ice sheet of the dying river (Map No. 30. D. 2). It had taken a new course, as it proved, some four years earlier, a considerable distance to the west of the one which Dr. Hedin had followed in 1896 (see Map No. 30. D. 2).
It was at Camp 323, some 25 miles in a straight line from our entry into this forbidding dry delta, that the first sign of human presence at a past epoch was. met with in the shape of a small worked flake of flint,'$ similar to those found so plentifully on the way to the Lou-lan Sites It was picked up on bare eroded ground within an ancient river-bed, and was certainly of palaeolithic origin. Some nine miles further south and by the same ancient bed a piece of charred wood turned up, also a sign of some human's visit which on such ground might be centuries old. But it was not until after tramping for another 24 miles or so that we came upon definite proof of temporary human occupation in times gone by in the shape of the badly decayed remains of a herdsman's hut (Map No. 30. D. I). By ascending the actual river-bed for two marches we passed out of the area of bare dunes and dead jungle into a riverine belt of living vegetation. Finally we found ourselves, near the grazing-ground of Yoghan-kum (Map No. 30. D. 3), at the point which I had recognized already in 19o1 as the head of the Keriya River delta. On the way we had passed within less than four miles of the ruined site of Kara-dong (Map No. 30. D. 2) without knowing it.
It was in connexion with the ruins of a fortified station there explored in 1901 that I had occasion to discuss before the possibility of a regular route having in ancient times led from the end of the Keriya River to the Tarim, and thus on to Kucha.sa The personal experience gained and the survey effected on the desert crossing just briefly recorded have led me to modify to some extent the conclusion then arrived at. I still believe that the opening up of a route by maintaining wells and small posts along a line roughly corresponding to the one we followed, and by marking the track where the line is particularly liable to be lost, would be practicable. The belief of the energetic Keriya Amban who actually planned such a route, according to M. Grenarcl's report in 1893,' was well founded. But in view of the almost total absence of camel-grazing between the southernmost dry bed of the Tarim and a march or two below the shepherd station of Tonguz-baste (Map No. 30. D. 3), and taking into account the great distance intervening, I do not think that such a route could nowadays ever become a regular trade route.
Conditions may possibly have been somewhat better in I-Ian or even in Tang times, assuming that the terminal course of the Keriya River has since then shrunk considerably through desiccation.
6 See above, pp. 356 sq.
6a Cf. Ancient Kholan, pp. 449 sq.
7 Cf. Grenard, Mission Dulreuil de Rhins, i. p. 172.