the two roads from Sa-tscheo meet at Jol-arelisch; but the map shows the two routes in question meeting on the south-west shore of the lake of Urghu-khoitu and at Nukitu-qia-tarim, and both Ike-ghaschon and Bagha-ghaschon lie a long way east of that point. In all probability what the text ought to have told us is, that from Nukitu-schan-k'ou the road proceeded north-west to Nukitu-qia-tarim, as it is made to do on the map.
Having advanced so far in the interpretation of the text and the map, we have no difficulty in imagining the rest. Had the Chinese author visited the region himself, he would never have entered the old Lop-nor on his map at all, for of that ancient lake there certainly survived at the time he wrote nothing more than traces, nor would he have put in the Kuruk-darja either. These he has therefore borrowed from one of the old documents for which he entertained such a profound respect, and in fixing their position he was guided by some ancient map, on which the lake occupied its former site in lat. 401;12 N. On the accompanying sketch-map (vide pl. 34). I have attempted to reconcile the data of the Chinese map with the conceptions which we Europeans, with our better and more accurate methods of topographical determination, have formed of the Lop-nor region. Let us take a general glance at the Wu-tschang map. The Bostang-nor or Baghrasch-köl is the lake which lies nearest to its proper latitude and longitude. Along the road that proceeds west from that lake we find Korla, Tschadir, Bugur, Kutschar, and Baj, all a little too far to the south and west, though Kôrla lies in approximately its proper longitude; but the error increases the farther we go west, until in the case of Baj the displacement is three times as great as for Tschadir. There is an almost equivalent error for the towns on the southern East Turkestan route. Kerija, whilst put at almost the proper latitude, has been shifted a little less than 1/3 too far to the west. With the latitude of Chotan there is not much fault to find, but its longitude is one degree too far to the west, an error exactly equal to that of Baj.
The Mapa-kisil-kül, which still haunts our maps under the names of Baba-köl and Sarik-kamisch, has probably experienced a displacement to the south similar to that of the towns which lie to the north of it. Nevertheless its position with regard to the river Tarim is plausible. Presumably this marsh has now disappeared, no matter whether it was formed by a river from the Tien-schan or was only one of the numerous marginal lakes along the Tarim. As for the Tarim itself, it would appear to be placed, on the whole, a little too far to the south. Here again the error of latitude would appear to increase from east to west, the river being drawn in the west fully I/2 too far to the south; whereas on long. 851/2 it is only Viz° out in point of latitude. South of Bostang-nor the two rivers intersect one another, and farther east the Tarim of the Wu-tschang map lies north of the existing Tarim. Nevertheless it is very possible, nay probable, that on the Wu-tschang map the western part of the Tarim, and indeed in general its course throughout, is perfectly correctly drawn, especially as it may be assumed, that the material from which the river was drawn was, like that for the Lop-nor, of ancient date. To me after mapping the whole of the Tarim, as I have done, and ascertaining that not a single year passes without changes in the situation of the river-bed at various points, it would have appeared much more wonderful had the Tarim on the Wu-tschang
IIIF Color HighRes
