Sec. i] THE LIMES LINE NORTH OF THE SU-LO-HO 375
hsien and Su-chou may at one time have been occupied by a ` Gate ' station corresponding to the ancient Yii-mên or ` Jade Gate ' and to the modern Chia-yü-kuanl° In that hypothesis the position chosen for the ruined' Temple might, perhaps, be accounted for by lingering local worship such as we have found elsewhere clinging tenaciously to ` Gate ' sites in the Limes.n More light would perhaps be thrown on the question if it became possible to determine the exact location of Chin-ch`ang pi R. A passage of the Tang Annals mentions this place as occupied by the frontier station of Yü-mên kuan in A.D. 610, and Chinese antiquarians look for it to the east of An-hsi12
Meanwhile I must confine myself to pointing out the strategic importance of the Wan-shan-tzû defile for the defence of the high road leading from Kan-su to An-hsi and thence to Hâmi. This is clearly brought out by the fact that a large Chinese garrison had been stationed at Bulungir or Pu-lung-chi, some ten miles to the east of it (Map No. 40. B. 4), in the time of the Emperor K`anghsi and Chien-lung, when the marches of Kan-su had to be guarded against Dzungar inroads and preparations had to be made for the conquest from them of the ` New Dominion '.13 Troops, placed there in the centre of several small cultivated areas, were in a convenient position to guard both the Wan-shan-tzû defile and the line of the Su-lo-ho where it might be approached from the north by turning the former.
On resuming our eastward march on the right bank of the river we came, after proceeding less than half a mile from Camp 122, upon the line of the Limes where it passes between two low ridges of much-decayed rock, forming the south-eastern extremity of the spur bearing the tower T. xL. c. The line for a distance of about i 20 yards could be traced here in the shape of a double embankment covered with stones and detritus. The southern embankment or mound had a width of about 24 feet at the base and a height of io feet. The bank to the north was less wide and rose only to 5 or 6 feet, the distance between the centre lines of the summits of the mounds being some 44 feet. Beyond that stretch the line was continued for another 8o yards or so by a double agger. Here layers of tamarisk branches cropped out on both sides and the reduced width of the agger on the south, only 12 feet, indicated a different method of construction. The broken nature of the ground may account for the fact that the line was protected at this point by a double embankment or agger.
A short distance farther on the outcrop of rock disappeared, and all trace of the line was lost in a belt of soft alluvial loess fringing the bed of the river and about half a mile wide. Deep trenches fissured the ground in a direction parallel to the river and seemed to be due to wind-erosion. After having covered about three miles we passed into a belt evidently liable to occasional inundation and covered with luxuriant reeds and bushes. Here, too, no remains of the Limes line were to be seen, but five miles farther on we came upon the ruin of a tower ; it stood on a clay terrace overlooking the wide depression of the river-bed, partly covered with thickets of Toghraks and bushes.
This tower, T. xLi. a, had already been sighted by me in September, 1907, from the road west of Bulungir, and had actually been visited some days later by Lai Singh on a reconnaissance trip made under my instructions from An-hsi. But it was only now possible to make certain that it marked a watch-post on the ancient Limes of Han times. It was built of layers of stamped clay and measured about 20 feet square at its base. The northern face of the tower had fallen and the remaining portion had split into two parts. Yet in spite of this far-advanced decay the antiquity of the ruin was quickly established. Potsherds of the dark-grey mat-marked Han type covered the bare clay around, and definite confirmation was supplied by the fragment of a broad wooden
10 Cf. ibid., ii. p. 727 ; iii. p. 1099, note 20.
II See above, p. 369 ; Serindia, ii. pp. 602, 696 ; iii. Pp. 1094 sq.
12 Cf. Serindia, iii: p. 1099, note zo, for references to texts quoted by M. Chavannes.
13 See ibid., iii. pp. 1138 sq.