348 TO TUN-HUANG AND AN-HSI [Chap. X
allowed me to consult, the text was composed, A. D. 886, as an ` official memorandum ' on the Tunhuang territory and the regions adjoining it to the west and north-west. It shows close agreement in many points with the Tun-huang Lu and is manifestly based on information collected locally."
The notice, which I think must refer to Chien-ch`üan-tzû, runs in Dr. Giles's translation as follows : ` Hsing-hu lake, i io li NW. of the hsien. The other wàter [in the vicinity] is all brackish ; this water alone is drinkable. Hu traders on their journey to or from the Yü-mên Barrier all stop here. The Sha-chou chih says : " The water is brackish ; only the spring is fit for drinking." It gives its dimensions as follows : 19 li east and west, 9 li north and south, depth 5 feet.' Judging from other local references in the text it is clear that by the hsien or district head-quarters is here meant the walled city of Tang times, situated about a mile to the west of the present Tun-huanghsien. Now the distance of i io li to the north-west which the text indicates from this point brings us exactly to the position of the Chien-ch`üan-tzû lakelet ; for it is the equivalent of the 22 miles which the map marks, taking the li at the value of about one-fifth of a mile, which on Central-Asian ground is proved by abundant evidence to be approximately correct.20
This location of the ` Hsing-hu lake ' is strongly supported by the statement that its water alone is drinkable in that vicinity, and by its being described as the regular halting-place of Hu, i.e. ` barbarian ', traders on their way to or from the Yü-mên barrier ; for Chien-ch`üan-tzû or Shôr-bulak is still invariably so used by all caravans moving to, or from, the west past the ancient ` Jade Gate '. Now, in order to understand correctly the passage quoted from the Sha-chou chih, a somewhat older description of the Tun-huang region,21 we must take into account what the same manuscript Ch. 917 tells us in a preceding passage of the ` Western Salt Lake, 117 li northwest of the hsien. It is popularly known as Sha-ch`üian [the sand spring]. The salt is of the same kind, but it has a nice taste and is of a pink colour.' Considering that the direction named is the same as that to the Hsing-hu lake and the distance only 7 li farther, I think we may safely identify this ` Western Salt Lake ', which obviously was sought as a place for the supply of salt,'22 with the area of dried-up salt marsh which is crossed by the caravan route from the Limes. In Tang times it may well have been subject to seasonal inundation, just as the ground farther north towards T. xxIII. c, d is still.
I.t is to this area, now encrusted with hard salt, that the dimensions quoted from the Sha-chou chih must reasonably be held to apply. The extent of 19 li from east to west and 9 li across manifestly cannot refer to the ` spring ' which alone is fit for drinking, since the pool or lakelet fed by it is quite small—scarcely thirty to forty yards across. But it agrees remarkably well with the extent of the dried-up marsh bed that is crossed by the route a couple of miles to the north-west. This identification, if correct, may claim some geographical interest ; for it would furnish us with an approximate estimate of the length of time required before a salt marsh, holding water for at least a portion of the year, assumed that form of a dry bed, covered with hummocks of hard salt, which we find to-day to the north-west of Chien-ch`üan-tzii and which was already in existence in Han times over the greater portion of the dried-up Lop sea, as it is still to-day.
About 21 miles to the ENE. from T. xxiii. a, the ruined watch-tower, T. xxiii. b, was traced on the top of an eroded ridge of clay about 5o feet high. The line of the Limes wall, decayed into a mere
19 For other data drawn from this manuscript, cf. Pelliot, J. Asiat., 1916, jan.—fév., pp. 116 sq. ; also Serindia, i.
20 Cf. Serindia, ii. p. 735 and the evidence quoted there in note 28 a. It must be kept in view that the route between the two points lies in a practically straight line and, with the exception of 5 or 6 miles, passes over bare gravel desert.
21 For its approximate dating, cf. Serindia, ii. p. 716.
22 This may be concluded from the fact that our manuscript Ch. 917 mentions this ` Western Salt Lake ' immediately after ` Eastern Salt Lake ' where ` salt forms in natural lumps in the water ; men strain away the water and dry the salt, which is all in crystalline form '.