1138 THE NORTH-WEST FRONTIER OF KAN-SU [Chap. XXVII
Shih-êrh-tun and Hu-hai-tzû runs more or less parallel to a route still occasionally followed to the present clay by Mongols and others who wish to proceed to Hâmi by a direct track crossing the Pei-shan east of the An-hsi-Hâmi high road.G For those who follow this track from the side of Hâmi, Shih-êrh-tun is the first inhabited place reached on the Kan-su borders. Hence it is likely to have already had its own `Gate' station in ancient times. But whether this fact had anything to do with the later transference of the name Yü-mtu to the district headquarters now located at Yü-mênhsien remains doubtful.
From Shih-êrh-tun I regained the present highway at the village of San-tao-kou to the northwest of Yü-mên-hsien (Map No. 83. D. 3). The march which thence took me on September 22 to Bulungir led over ground abundantly watered from branching beds of the Su-lo 1-io, as well as by marshy springs gathering at the foot of the gravel glacis further west. Extensive areas of abandoned cultivation now overrun by scrubby jungle, as far as Chi-tao-kou (` the seventh canal '), attest the ravages here produced by the great Tungan rebellion. Beyond this there spreads a wide grassy steppe traversed by marshy depressions and affording fine grazing, which in times gone by must have had its attractions for nomadic invaders. From the high ground to which the road keeps here I sighted far away to the north a line of ruined towers, extending along the right bank of the deep-cut Su-lo Ho bed. In view of the definite proof gained at Shih-êrh-tun there could be little doubt that they marked the line of the Han Limes. The necessity forced upon me by various practical reasons of quickly gaining An-hsi prevented my examining these ruined posts at, the time and searching for remains of the Limes wall. But subsequently, when Surveyor Rai Ram Singh had been relieved by R. B. Lâl Singh at An-hsi, I was able to send back the latter for a rapid reconnaissance. This rendered it possible to indicate in Map No. 83. B, C. 2 the position of the least decayed of the towers. The careful survey I effected myself in 1914 along the whole line right through to Shih-êrh-tun has supplemented the evidence thus gained in various ways. But it has shown also that, owing to excessive wind-erosion in some places and the effect of moisture in others, the traces surviving of the Limes agger are very scanty indeed until its line passes on to firm gravel soil west of the abandoned station of Chiao-wan-ch`êng (Map No. 83. D. 2).G
The massive walls of Bulungir, enclosing an area about 1,ioo yards square, appear to have sheltered during the seventeenth-eighteenth centuries an important Chinese garrison. The place, now almost deserted, had evidently served as an advanced base for the operations by which the Chinese, under the great Manchu. Emperors K`ang-hsi and Chien-lung, pushed back the threatening power of the Dzungars and finally conquered the ' New Dominion '.7 A strong garrison holding
6 For a detailed account and carefully executed map of this direct route from Su-chou to Hâmi, see Futterer, Geographische Skizze der Wilsie Gobi, loc. cil., pp. 3 sqq. Prof. Futterer duly noticed the ruined fort of Shih-êrh-tun on his way from the Su-lo Ho to Hu-hai-tzû (cf. p. 25), without, however, becoming aware of its significance or of the remains of the ancient Limes line.
The continued use of the route from Hu-hai-tzti to Shihêrh-tun probably explains the notice taken of the ancient watch-towers along it in calling the hamlets at its western end
after ' Tower X Tower xii '. It also accounts for the re-
pairs effected to those towers which are close to the route. The popular explanation I heard was that Shih-êrh-tun was
twelve towers', i. e. 12o li, distant from Hu-hai-tzti, a watchtower, or rather a much-reduced representation of one, being the usual road-mark on modern Chinese highways placed at every to li or what is conventionally taken for that distance.
In Chinese Turkestân 120 li would at the present day approximately correspond to 24 miles. The actual distance between Shih-êrh-tun and Hu-hai-tzû is well over 4o miles.
6 The representation of the Limes north of the river as an unbroken line from circ. 96° 25' to 96° 40' long. is an error, due to R. B. Lai Singh, then new to such work, having in places mistaken low Yârdang • ridges for remains of the wall. But the line of the latter could in 1914 be fixed with certainty all the same by the ancient pottery débris, etc., that marks the position of numerous intermediary towers now completely decayed.
Clfiao-wan-ch`êng proved to be a well-built small fortress erected under the Emperor Chien-lung and destroyed during the Tungan rebellion.
7 Cf. Ritter, Asien, ii. 371 sq. The name, also pronounced I3rrlun jir, or Pu-kung-chi by Chinese, is assumed to be of Mongol origin.
Road from San-tao-kou to Bulungir.
Limes line along north bank of Su-lo Ho.
Fortified station of Bulungir.