Sec. iv] THE TERRITORY OF WEI-LI AND THE MODERN KARA-KUM 783
on these heights overlooking the defile through which the Konche-darya debouches might have served most conveniently for a signal station, and this was strongly suggested to me by my observations in December, 1908, relating to the two undoubtedly ancient watch-towers found in a line to the north and south-west of the Ming-oi site above Shôrchuk (Map No. 25. A. 1)14
In my description of these ruined towers in Serindia, I have fully indicated those structural details which link them very closely both with the watch-towers traced on the Ying-p`an—Korla route and those to be found on the Tun-huang Limes. I have in the same place called attention to the obvious advantages to be derived from a line of signal stations pushed out from Korla into the Kara-shahr valley. It would afford timely warning of danger threatening from what at all times must have been a main gate for the irruptions of Huns and other foes into the Tarim basin. A position of great military importance is offered by the narrow defile in which the Konche-darya has cut its way through, between the last outlier of the Kuruk-tagh and the foot of the big spur of the Tien-shan flanking the valley of the Kara-shahr from the west. Under the name of the ` Iron
Gate ', p9 T`ieh-mên, the defile figures in the Chin shag's account of the exploit by which
in A. D. 345 a Chinese expedition dispatched by Chang Chan, the local ruler of westernmost Kan-su, forced its way through from Yen-ch`i (Kara-shahr) and conquered Wei-li.15 Here, too, Yaqûb Beg during the months preceding his death at Korla had hoped to stem the advance of the Chinese reconquest in 1877. The direct distance separating the watch-tower south-west of the Ming-oi site from that of Y. ix at Suget-bulak is not more than about twenty-two miles, and with a point of such great strategical importance between the two it seems difficult to believe that the facilities for semaphoric communication offered by the intervening hill chain were left unused. But this conjecture could be established only as the result of a close and systematic search on the ground, for which, to my great regret, I was unable to spare the time during my busy stay at Korla.
After visiting the remains of the tower at Suget-bulak I regained the northern edge of the cultivation of Shinega,'6 a flourishing little oasis watered from the same large canal that carries water to Kara-kum. Together with Bash-engiz, a hamlet farther up, it counts over fifty households, all families from Korla who settled here in pre-rebellion times and are now,well established. The fine orchards and vineyards amidst which the prosperous looking farms were ensconced testified to the fertility of the soil and the abundance of water. Bash-engiz lies in a gap about a mile wide which separates a gravel-covered plateau, contiguous with the glacis of the hill chain, from a small isolated peneplain about forty feet high stretching westwards and falling off in steep cliffs on all sides. It seemed clear to the eye that this peneplain is but a continuation of the plateau and the Sai glacis behind it, and that the wide gap which now separates it from the latter was carved out by a branch of the Konche-darya which at an earlier period cut its way through at this spot and deposited its alluvium all over the plain of Kara-kum.
The canal which winds round the foot of the plateau and then passes on through this gap towards Kara-kum thus in reality merely follows an earlier river-bed. It was small until it was enlarged in 1900 to serve the newly colonized Kara-kum as well as Shinega. Its discharge as measured at the latter place amounts to over 200 cubic feet per second, and this could be greatly increased by taking off at the canal head above Korla more of the superabundant supply of water in the Konche-daryâ, which now flows past the town and after passing through the marshy basin of the Boto-köl west of the Korla oasis meanders in a wide semicircle round to Konche-mazar
14 See Serindia, iii. pp. 1199, 1226 sq.
15 Cf. Chavannes, Anc. Khotan, i. pp. 543 sq. For a mention of the ` Iron Gate ' in the Tang Annals, see also
Chavannes, Turcs occid., pp. 7, 304.
16 This is the form of the name as I heard it, not Shinalga as printed in Serindia, iii. p. 1230, note 1.