132 G. P. WRIGHT — LOESS IN NORTHERN CHINA AND CENTRAL ASIA 



side, next to the Mongolian escarpment, consisting for the most part of 

 basaltic rocks belonging to a comparatively recent (probably Tertiary) 

 eruption. For a great extent all over this region the loess has accumu- 

 lated in level areas, which resemble lake basins. In many cases these 

 are without outlet, and contain remnants of larger bodies of water, which 

 are now drying up, leaving well marked terraces at elevations of con- 

 siderable height around the rim. In many of these level areas of loess 

 within the drainage basin of the Yangho there are numerous deep nar- 

 row ravines, with branching tributaries, cut to a depth of 100 feet or 

 more by retrograding erosion, the loess standing in perpendicular faces 

 on either side. Pumpelly describes one of these chasms as " more than 

 75 feet deep, with a width of only 4 feet between vertical walls of loam, 

 and winding in a crooked course for more than a mile." In many places, 

 especially near the bordering ledges of rock and near the center of the 

 larger valleys occupied by the main stream, there are distinct lines of 

 coarse gravel and rocky fragments interstratified with the loess. This 

 oftentimes continues for a long distance over a comparatively level area, 

 where it would seem impossible for superficial currents from local cloud- 

 bursts to have produced the results. 



On the other hand, it was noticed that in the narrower valleys run- 

 ning east from Kalgan to Shiwantse, between the lofty border of the 

 Mongolian plateau and the nearest border range, there were numerous 

 and extensive deposits of loess that had been very clearly drifted in by 

 the wind. The resemblance of these deposits to immense snowdrifts 

 accumulating on the lee side of the mountains was very striking. This 

 was especially the case at Shiwantse, where the entire village of 1,500 or 

 2,000 inhabitants finds shelter in commodious and comfortable houses dug 

 into the hillside of loess which flanks the eastern face of the mountain 

 range. These houses are excavated in successive receding stories one 

 above the other, the natural roof of one house serving as the front yard 

 of the house above it. These dwellings extend for 300 feet or more up 

 the slope of the loess, which continues upward for a considerably greater 

 distance. In this valley we saw many such villages, and in crossing the 

 mountain from west to east found extensive drifts of the loess up to a 

 height of 5,000 feet above the sea. But the greater accumulations of 

 loess were below a level of 3,000 or 3,500 feet above the sea, and in many 

 cases, even on the margins of the larger and deeper valleys, were spread 

 out in such extensive and level areas as to suggest a terrace deposit near 

 the margin of standing water. It became increasingly difficult for su 

 to believe that wind could have distributed the material with such an 

 even surface on the margin of such well marked and deep valleys as we, 

 repeatedly crossed. 



