months was attested by a regular flood bed, about 300 feet across and well marked to a height of 3 feet above the surface of the actual flow.
This latter observation fully confirmed the statement recorded on my second journey that the difficulty experienced in maintaining the present small terminal oasis of Endere is due mainly to the frequent changes in the river's course lower down, consequent on the heavy summer floods.3 I found no salinity in the kara-su water that filled the winter channel and no shôr on its banks. Taking all these facts into consideration I see no reason to modify the views previously expressed as to the conditions that in ancient times determined cultivation near the end of the Endere river, conditions which have also affected the successive later attempts at resumed occupation after prolonged abandonment.4
It is these repeated changes, to which its occupation has in historical times been subject, that lend a special interest to ancient remains in this district. I had therefore already, before my arrival at Kashgar, arranged through Badruddin Khan for a renewed search to be made by certain of my old Niya diggers for previously unexplored ruins. The report received from one of them while I was at Khotan had not sounded encouraging ; for they had found only some rough timber and reed huts such as shepherds still build to-day, in the forest belt at some distance above Baba-koidi on the Charchan road (Map No. 19. D. 1) and evidently not far from an old bed of the Endere river, and another little group of rough structures in the jungle of Shûdan (Map No. 23. A. 1). They also had heard from a Niya shepherd of a site with more substantial ruins including a ` Pao-t`ai which he called Kök-tin (` the green tower ') and which he said he had come upon by chance some ` five marches ' south of Shûdan. On being sent back to look for this ` Kök-tim ' the party had not been able to secure Ahmad, the pretended guide, and in consequence all but one of them failed to meet me at Endere-langar, as arranged. The two ` old houses ' which that man was prepared to indicate to the north-west of Kamaghaz were manifestly identical with the small ruins already visited in 1906 near the ancient watch station at the southern end of the site, and when subsequently asked to accompany us to Shûdan he, too, decamped.
In view of the negative result of these inquiries and those made here in 1905 by Professor Huntington, I consider it very improbable that there exist any remains of ancient settlements, other than mere shepherds' huts and the like, in the desert separating the riverine belts of Endere and Charchan. I had already followed the caravan route passing through it in 1906,6 and no further account of the ground or of early travellers' references to it is needed. I found, however, signs of increasing traffic and of some improvement of travel conditions in the shape of newly built ` Langars ' at the wells where we halted. Exceptionally clear atmospheric conditions allowed me on the last two marches before Charchan to sight the great snow-covered range of the K`un-lun, and thus to fix exactly our positions on the plane table from high peaks triangulated in 1906.
In the report on my second journey I have fully discussed the physical aspects of the Charchan oasis and its importance as a link on the southern trade route of the Tarim Basin. I have also reviewed there the scanty remains of its early occupation, and the historical records bearing on the peripeteias through which it has passed since Han times.' My renewed brief stay at Charchan, necessary for the purpose of securing fresh camels and supplies, furnished distinct evidence of an increase of cultivation and prosperity since my first visit. The western limit of the oasis now approached close to the small ruin known as Tam, and its southern limit had encroached still
3 Cf. Serindia, i. pp. 274 sq. of the earlier settlement that I-Isüan-tsang noted here in
4 See in particular Serindia, i. pp. 286 sqq., concerning the ` old Tu-huo-lo country '.
reoccupation of Endere Fort in Tang times and the ruins 5 See Serindia, i. pp. 284 sq. See ibid., i. pp. 293 sq.
7 Cf. ibid., i. pp. 294 sqq.