334 TO THE SU-LO-HO DELTA [Chap. IX
short remainder of its course the river is overlooked by the bare plateau of gravel and stone, dotted with low isolated hillocks, which, as we have seen above, cuts it off from the dry basin northward.
To the south of the final portion of the river's course and to the east of the terminal lake extends the large basin, marshy for the most part, the eastern and southern borders of which had already been surveyed in the course of my explorations of 1907 along the south-western flank of the Limes.3
This basin measures about twenty miles from north to south and close on thirty miles across at its widest. To the east and south it holds marshes which are fed by subterranean drainage passing
under the great gravel glacis of the mountains south-west of Nan-hu (Map No. 36. c, D. 2) and also, perhaps, by occasional rain floods descending from their slopes. There is reason to assume that the greater part of the basin is boggy ground, impracticable during the late spring and summer months.
Now, proceeding up the river above Toghrak-bulak, we find that there branch off from it, at distances of about eight and eighteen miles, respectively, two dry beds which undoubtedly formed
part of the Su-lo-ho delta as it existed at a comparatively recent period. They, too, lie in deep-cut
trenches between gravel-covered plateaus ; but judging from the survey of Ram Singh, who crossed them in three places, and from what I saw of the southern bed near the ruined watch-station
T. i, these trenches farther on are far wider than the bed which passes BEsh-toghrak. The valley-
like fosse containing the old bed was found at T. 1 to be about a mile broad and covered with reed-beds and scrub. The bank on the south rose very steeply about seventy feet above the bottom.
When, on March 16, 1914, I revisited the ruined tower on its edge, the dry salt pools I had noticed
below in 1907 were covered by large sheets of water. These were surrounded by salt-encrusted ground and looked as though they had been left behind by a big flood which in the preceding
summer, or perhaps a year or two before, had found its way into the old long-abandoned channel represented by this wide Nullah. Luxuriant reed-beds were similarly seen by Ram Singh in 1907, where he crossed the other old river-bed farther north.
Judging from the recorded direction of these two dry river-beds there can be, I think, no doubt that the water they once received was carried into the Mesa-girt basin westwards through the gap
marked by the above-mentioned Wadi. This gap divides the far-stretching plateau on the south from a corresponding outlier of the southernmost Pei-shan range which overlooks the trough of the terminal Su-to-ho. The defile thus formed has its exact counterpart in that trough which the Su-lo-ho passes between Bulungir and An-hsi.4
The fact that the basin thus reached by whatever water once passed down these old branches of the Su-lo-ho has no connexion with the present terminal basin to the south of it has its close
parallel in the case of more than one river ending within a drainageless basin. We have an example
of terminal bifurcation exactly corresponding in character within a region immediately adjacent to the drainage area of the Su-to-ho. The Etsin-gol, uniting the waters of the rivers of Su-chou
and Kan-chou, empties itself through deltaic branches, passing as close to each other as those of the Su-to-ho, into two distinct lake basins, the Gashun-nôr and Sokho-nôr. These do not communicate and apparently occupy different levels. Similarly it can be proved that the Oxus, down to comparatively recent historical times, fed two terminal basins as widely separated as the Caspian and the Sea of Aral.s We shall have occasion farther on to discuss an interesting instance of another such bifurcation in the case of the Su-lo-ho itself ; for it can be shown that the river, from