292 THE SEARCH FOR THE ANCIENT CHINESE HIGH ROAD [Chap. VIII
of our track averaged only twenty feet or so in height and were correspondingly narrow. Their axis lay uniformly from S. 2o° W. to N. 20° E., and thus differed distinctly from that of all the
Yardangs observed in the Lou-lan area. The same applied to the bearing of successive rows of
Mesas as observed with prismatic compass from our route? The difference of bearing in this belt of Mesas is all the more noteworthy that the Mesas farther west which our route from Altmish-
bulak crossed iri the vicinity of C. xcix, and which probably connect with this belt, approximately
conform in their direction to that of the Yardangs of the Lou-lan area.3 The slopes of the Mesas actually examined were covered, like the ground adjoining, with flakes of disintegrated gypsum. From a Mesa where a ` fixing ' was made after we had covered fifteen and a half miles, I could see that the belt of Mesas was thinning out northward on ground that seemed to rise slightly.
But a strong south-west wind had by then obscured the horizon with haze, which prevented us from seeing the foot of the gravel glacis lying in that direction. At a distance of eighteen miles from C. c we reached the last Mesa, rising to over forty feet with a Stüpa-like dome in the centre. Before us a bare plain, almost perfectly level, stretched to the east and north-east. To the north-north-west the edge of the Mesa-covered area was seen to curve round towards the point where Afraz-gul on his subsequent survey of February 15th, 1915, reached the lowermost edge of the Sai below Yetim-bulak. In order to make the most of the good ground for the camels we pushed on another four miles, the soil remaining much the same as before, but for a slight hardening in places owing to a thin outcrop of salt. The camels had kept up remarkably well all day, the worst sufferer of the preceding day being led at their head unladen ; they reached the patch of soft ground selected for our camp within eleven hours from the start.
The day's march, apart from the important find of coins and arrow-heads beyond L.J., had not furnished other relics of the ancient Chinese route. But the ground traversed had shown natural features that proved of material help in elucidating an interesting point in the ancient topography of the Lou-lan region. I had occasion, when dealing in Serindia with the Chinese historical records concerning Lou-lan, to discuss fully the important details to be found in the commentary on the
Shui thing, composed by Li Tao-yuan -f some time before A.D. 527, from which extracts
were first rendered accessible by M. Chavannes's translation.4 I there showed that the information given by Li Tao-yuan, undoubtedly from earlier sources, about the course of the ` River of the
north ' towards the ` town of Lou-lan ' 4; ; f and its final outflow into the ` lake Fu-ch`ang'
, agrees remarkably well with what our surveys have shown of the early hydrography of the region once watered by the Kuruk-daryâ.5 I indicated in the same work my belief that the ` Town of the Dragon ', to which the concluding portion of Li Tao-yuan's account refers, can safely be located in the area covered with big Mesas past which the march just described had taken us. Before reproducing Li Tao-yuan's notice of the ` Town of the Dragon ' and explaining the considerations which determine my location of it, it will be convenient here briefly to summarize once more the data furnished by the passages immediately preceding.
The identity of the river to which Li Tao-yuan referred with the Kuruk-dary5. is made perfectly clear ab initio by its eastward course being described as lying south of the kingdom of Mo-shan
2 A closely similar deviation from the regular Yardang 3 The delineation of the Mesa belt in Map No. 32. A. 3
direction was observed by Dr. Hedin in the clay ridges of is too conventional to bring out the relative bearing of Mesa
a belt which he crossed on his march of February 17, 1901, to Mesa in each row ; our plane-table, however, correctly
to the west of a shbr-covered depression forming probably records it.
an extreme north-eastern inlet of the Lop sea-bed ; cf. 4 See M. Chavannes's Note additionnelle to Les pays
Hedin, Central Asia, ii. p. 114. Its position corresponds d'occident d'après le Wei-lio ', T `oung-pao, 1905, pp. 563
approximately to that of the salt-encrusted bay shown in Serindia, i. pp. 324 sqq., 419 sqq.
Map No. 32. c. 2 to the north of Lai Singh's Camp C. 90. 5 Cf. Serindia, i. pp. 42o sqq.