to the river's capacity for depositing sediment. From what is said above we see how ready the Tarim is to shift its bed in this region. Not a year passes without long stretches being destroyed and new channels being formed in their places. The farther we advance, the more restless and capricious grows the stream, and the more numerous become the I oiza-darjas or eski-tarims on both sides of the main channel. Here for centuries past the Tarim has been depositing vast quantities of sand, which it has shorn off the edge of the desert and carried clown with it in suspension. The reason why the river flings itself out of one bed into another is that it keeps on successively raising their levels. After these oscillations have continued for centuries, the face of the country becomes eventually chequered with a number of more or less parallel swellings or ridges, running north-west and south-east. When an overflow takes place, it is quite easy to see how the space between two of these fluvial ramparts may become filled with water, the result being that the lakes, which are thus formed, inevitably tend to assume the same direction as the river; and this, as we have just seen, is actually the case. Indeed all the depressions in the Lop country exhibit an inconceivable regularity in the matter of having their longer axes disposed from north-east to southwest. In the case of the lakes situated amongst the desert sand on the right of the Tarim, in the case of the Kara-koschun lakes, the dry depressions of the Lop Desert, the bajirs of the Desert of Tschertschen — everywhere the same north-east and south-west direction prevails. But the two lake regions we have just traversed, both of them originating directly in the water of the Tarim, have their long axes disposed at right angles to this, or from the north-west to the south-east. Now for this there must of course be a special cause, and this cause is very probably to be found in the river's levelling tendencies, its peculiar property of forming ramparts by means of its sedimentary deposits, and the effects of the riparian vegetation in binding these deposits together. Later on, however, we shall see that the lakes of Avullu-köl, Kara-köl and others must have been formed in exactly the same way as the lakes and bajirs of the Desert of Tschertschen.
The eastern shore of the lake of Nias I3ek consists of low connected dunes, with tamarisks growing amongst them on their characteristic cones; amongst these lies the deserted village of Abdul Baki-uji, containing four huts. In the district of Nuruma, beside the Tarim, which is now some distance to the east, the barren sand is said to plunge steeply down into the river; though probably the dunes there are low in elevation. The lake of Kakel-dijni-clschaji lies beside the river, and to the east of it is another lake, the Ullugh-köl, into which there is an inflow at the season of high flood only. On the right of our route, that is to the west, there is said to lie amongst the sand an old bed known as the Tale-jatghan-tarim, now dried up, except that when the Tarim rises exceptionally high, it sends a small stream along this old channel. On the other side of Nias Bek's lake, we were taken by a small ilek, in which there was a distinct current, as well as alluvial deposits, into the lake of Örtäng-dschajir. This is quite a small sheet of water, and on its opposite side we again penetrated a narrow, shallow channel fenced in by rushes. In this, as in the final outlet from the previous lake-complex, there were cataracts, though the fall here was certainly not more than I or 2 dm. Nor is it at all sur-
IIIF Color HighRes
