Sec. x] THE LOU-LAN SITE IN CHINESE HISTORICAL REQORDS 423
Ts`ao Tsung, sent the Chan;-shih So Pan at the head of a thousand men to establish a military
colony at I-wu or Hami. The kings of Anterior Chü-shih (Turfâ.n) and of Shan-shan made their submission to So Pan, who, however, some months later perished in an attack of the .Huns and the tribe of Posterior Chu-shill (Guchen region). The first part of the names So Pan and So Man is identical, so also is the story of their dispatch from Tun-huang with a thousand men to establish a military colony. The locality indicated differs, it is true, but in each case it is the first cultivable tract beyond Tun-huang reached by the northern and central routes respectively. Is it possible that the name of an historical personage was foisted into a local anecdote about the military colony established at Lou-lan at approximately the same period, the founder of which was no longer remembered by name ? It must be left to others to clear up the question.
In any case our examination of Li Tao-yüan's passage about the ` town of Lou-lan ' and of his story of its foundation has made it quite clear that the source from which he drew his information knew of the existence of a Chinese military colony, occupying the same position in which we find the ruins of the Lou-lan Site and bearing the very name attested for this by the documents found there. The period from which this information approximately dates cannot be far removed from the time of the occupation of the Lou-lan Site as proved by our archaeological evidence.
It only remains for us to consider briefly what the concluding portion of Li Tao-yuan's above-quoted passage states about the final course of the Ho, i.e. the Kuruk-daryâ. Its waters after passing the town of Lou-lan are said there.' to empty themselves in the Yu marshes, which are those called by the [Shut' chine the Pu-ch`ang lake. The water accumulates in the north-east of Shan-shan and in the south-west of the Town of the Dragon.' My explorations of 1914 and 1915 have definitely proved that the ancient river:beds skirting the Lou-Ian Site found their termination. further east in dried-up marshes by the western shores of that great salt-encrusted lake-bed which in the Slaui chin; and also in the Former Han Annals bears the alternative names Pu-ch`ang lake or ` Marsh of Salt '.
It must be reserved for my detailed report on these explorations to show how closely the observations then made bear out the account which our earliest Chinese records give of the extent and character of the ancient ` Salt Marsh '. Here I shall confine myself to the mention of those topographical facts which help to explain the data contained in Li Tao-yuan's description of the Pu-ch`ang lake. Their interest reveals itself at once when we consider the statement made at the end of the passage just quoted, about the water accumulating ` in the north-east of Shan-shan and in the south-west of the Town of the Dragon '. The surveys effected in 1914 to the north-east of the Lou-lan Site, and briefly summarized in my paper A Third Tourney of Exploration in Central Asia, leave no doubt that by the legendary ` Town of the Dragon ' is meant the great belt of high, wind-eroded clay ridges, or Mesas, which I found extending from the vicinity of the ancient Han castruna (L.E.) north-eastwards for a distance of close on thirty miles." These Mesas, rising with precipitous walls to heights of a hundred feet and more, are quite distinct in appearance from the usual Yätrdang ridges, and go back in their origin to an earlier geological age, though the agency which created them, wind-erosion, is the same. By their bold shapes and fantastic outlines they constantly suggest visions of castles, bastioned town walls, Stûpas, and the like.35
' Cf. Wylie, Notes on Western Regions, J. Anthrop. Inst., x. p. zi.
34 See GI., xlviii. pp. 127 sq., with Fig. 14, which shows one of these Mesas (L.F.). For the ancient caslrum found near the south-west end of the Mesa belt, cf. ibid., p. 124. On the rough provisional sketch-map attached to this paper the position of this belt may be approximately indicated as
lying along and south of the letters DE in the entry LOP DESERT. A detailed delineation of this ground will be found in Map No. 32 of the r : 500,000 Atlas of our CentralAsian surveys now in preparation.
35 Mesas of the same type, but in a far thinner string and lower, are met with between the wells of Achchik-kuduk and Kum-kuduk on the caravan track skirting the southern