the ravine, where a considerable quantity of slag was scattered over the surface between bits of smelted ore, half-burned bricks and stones apparently containing iron-ore. Fragments of charcoal, too, could be traced among this refuse, evidently marking the spot where a smelting-furnace had once existed.
More interesting than these scanty remains, and at first decidedly puzzling, were the Conditions conditions in which they presented themselves. I had, however, occasion to observe the same of Tatis'. conditions again and again at similar old sites which I subsequently traced at numerous points beyond the limits of the present cultivated area in the Khotan region, and which are all known locally by the general designation of ` Tati' 6. And this renewed observation of identical features, together with the lessons taught by my explorations at far better preserved sites, such as those at Dandan-Uiliq and beyond the river of Niya and Endere, gradually furnished a convincing explanation.
The most striking feature noticed at Kakshal Tati, as well as at all other ` Tatis ', was Débris rest-
that the débris rests on nothing but natural loess, either firm or more or less disintegrated into ing on
the condition of dust or extremely fine sand'. It was always easy to ascertain that the soil underneath contained neither walls nor other structural remains of any kind ; for the isolated small terraces or banks of loess which rise here and there above the general level of a Tati, and on the top of which the fragments usually lie thickest, invariably displayed on their bare sides the natural soil without any trace of ancient deposits or distinctive strata. At Kakshal Tati the highest of these flat-topped banks rose about fifteen feet above the dust-filled depressions around them ; the average elevation of most was, however, from eight to twelve feet.
In the formation of these banks, as in all other features of such sites, it was impossible Results of
not to see evidence of the powerful erosive action of the winds and sand-storms which sweep nd-
the desert and its outskirts with great frequency during the spring and summer. The remarkable force of these desert storms or ` Burans ' has been commented upon by all travellers in Eastern Turkestan. Though my stay in the desert region did not continue beyond the early season of March and April, I had ample opportunities, as my Personal Narrative shows, of gaining practical experience of their vehemence'. Only the materials above described could, by their hardness and weight, survive, sinking lower and lower as the ground beneath gets more and more eroded, while everything in the shape of mud walls, timber, &c., used in the construction of Turkestan houses, has long ago decayed and been swept away. Even the potsherds and other fragments which have withstood destruction, bear plain evidence of the ever recurring onset to which they have been exposed, in their small size and the peculiarly rough surface already noted.
It is evident that such a process of erosion at sites of old settlements built on friable loess Lowering
soil could not have gone on during the long centuries since their abandonment without also of ground
considerably lowering the ground-level. But the erosion could never proceed uniformly over erosion.
a whole area, and of this we have evidence in the banks of loess already referred to, which are seen rising like small plateaus or islands above the more disintegrated parts of a ` Tati '. They may owe their relative protection to a variety of special features, such as the greater density of the débris with which they are ordinarily covered, or to their having been once occupied by structures which, though now completely vanished, may yet by the weight of their crumbling ruins have effectively shielded the soil beneath. We shall see hereafter that the