w

THE RIVERS OF CHINESE TURKESTAN
AND THE DESICCATION OF ASIA.

By Ellswoeth Huntington.

EIB

{From ' The Geographical Journal ' for October, 1906.)

THE RIVERS OF CHINESE TURKESTAN AND THE
DESICCATION OF ASIA.
By ELLSWORTH HUNTINGTON.

(From 'The Geographical Journal' for October, 1906.)

" The desiccation of Asia " is a phrase frequently used but rarely well
defined. With some writers it means the retirement of Tertiary inland
seas because of warping of the Earth's crust — a process essentially com
pleted long before the appearance of man ; with others, who usually
disregard the division of the Glacial Period into alternate cold and warm
epochs through several of which man probably existed, it denotes the
change from the coldness or moisture of glacial times to the aridity of
to-day ; while with still a third group it indicates a gradual change of
climate, supposed to have taken place during the period covered by
history, and to have been a potent factor in causing great human migra
tions, such as the invasion of Europe by the Huns, or of western Asia by
the Turks. The reality of the first two types of desiccation is established
almost beyond question by geological evidence ; a study of the geo
graphy and archaeology of eastern Persia, Transcaspia, and Turkestan
during the past three years has led me to believe that the third type
is equally real, and that during' the last two thousand years there has

353 THE EIVEES OF CHINESE TUEKESTAN
been in; progress a desiccation of Asia which is the last faint undula
tion of the great climatic waves of the Glacial Period. In the reports
of the Pumpelly Expedition, I have considered the question as it applies
to Persia and Transcaspia. More recently, during 1905, as a member
of the Barrett Expedition to Chinese Turkestan, I have been able
to carry the investigation into regions where the evidence is more
conclusive. The rivers of East Turkestan fall into two main groups. One, in
cluding most of the large streams, comprises those which rise among the
lofty mountains on the south-west, west, and north of the Tarim, or Lop,
basin, from Khotan around through Kashgar to Korla, and join to form
the so-called Tarim terminating in the historic lake of Lob Nor. The
other comprises a multitude of smaller streams which rush down from
the mountains, chiefly in the southern and eastern parts of the basin,
and, for the most part, wither away in the desert. The latter belong
to a highly specialized type peculiarly sensitive to changes in the
water-supply. In the southern part of the basin, from Khotan east
ward to Lob Nor, where the type is most perfectly developed, the
larger withering streams, rising among snowy mountains, flow at first
in zigzags or in east-and-west courses among the shallow longitudinal
troughs of the northern part of the Tibetan plateau, and then, turning
northward, break through the great fault-scarp of the Kwen Lun
mountains in narrow gorges ; while the smaller, rising on the north
side of the Kwen Luns, flow down the face of the esoarpment as rushing
summer torrents. Beaching the sharply defined, squarely cut base of
the mountains, all the streams, large or small, enter the zone of Pied
mont gravels, a sloping plain from 15 to 50 miles wide. At first they
flow in deep terraced trenches cut in their own gravel fans, and often
difficult to cross; but soon the fans decrease in size, the terraces die
out, and the streams spread into such broad and numerous channels
that the water completely disappears, partly by evaporation, and still
more by sinking into the gravel. To judge from the map, this is the
end of the smaller streams. In reality, they continue underground as
a confluent sheet, deeply buried where the gravels prevail, but coming
close to the surface where the latter give place to clays and sands.
Here water, often saline, can be obtained in wells at a depth of from
3 to 6 feet, and its widespread presence is shown by the abundance of
reeds, tamarisks, and poplars which form a zone of vegetation from
1 to 20 miles wide, skirting the northern edge of the gravels for nearly
1000 miles. The larger streams succeed in crossing the Piedmont
gravels during the three or four months of the flood season from May
to August, and either lose themselves among reed-beds in the zone of
vegetation, or pass on and gradually dwindle to nothing in the sand
of the Takla-makan desert. The only exceptions are the Cherchen
river, which, being diverted eastward along the zone of vegetation,

AND THE DESICCATION OP ASIA. 354
succeeds in reaching Lob Nor, and the smaller streams farther east
which in extreme flood join the Cherchen. During the drier part of
the year all the streams, except those of Chira, Keriya, and Charklik,
disappear in the gravels, but reappear lower down in springs seeping
out in the river-beds. Typically each main stream irrigates an oasis
on the southern edge of the zone of vegetation, and formerly irrigated
another, which, because of salinity or lack of water, it cannot now
support. As a whole, the withering rivers show signs of having
decreased in size during the last two or three thousand years, the
evidence lying partly in diminished length, as shown by dead vege
tation, and partly in diminished volume and increased salinity, as
shown by ruins.
In the 700 miles eastward from Khotan to the eastern end of Lob
Nor, seventeen rivers are worthy of notice by reason of their size or
because they support oases. The test of diminished length as measured
Dy vegetation, however, can only be applied to five of the seventeen, or
possibly six, since the rest formerly united with one of the other five
or with the Cherchen, which reaches'Lob Nor. The possible sixth river
is the small Yes Yulghurj, the lower end of which has never been
explored. The Molja is also doubtful. It now terminates in the zone
of vegetation, but an old channel suggests that it formerly may have
joined the Endere to the west, although now it is too small to flow so
far. The Chira, the most westerly to which the test is applicable,
appears formerly to have been joined by the three or four streams to the
east of it, and to have flowed to the sand-buried towns of Dandan-
Uilik and Eawak described by Stein. A strip of dead tamarisks,
typical river jungle, extends even beyond Eawak, the northern town,
to a point 60 miles from the Chira's present terminus, and over 50 miles
from the limit of living vegetation. Farther east the Keriya river is
reported by Hedin and Stein to have dead jungle extending 20 miles
north of the present end of the stream. On the Niya and Yartungaz
rivers I found that similar dead vegetation extends equally far or farther,
and on the Endere at least 8 miles and possibly much more. In brief,
of the six rivers to which the test of the extent of dead vegetation can
be applied, one has not been explored ; and the other five all show that
within times so recent that the dead reeds, tamarisks, and poplars have
not yet disappeared, the rivers flowed from 8 to 60 miles farther than
to-day. The Endere and Yartungaz rivers are the most important,
for they each support but a handful of farmers, eight or ten families,
and the amount of water diverted for irrigation is insignificant. The
Yartungaz river is now artificially constrained to flow wholly in one
channel, although formerly during floods it spread into two or three.
Thus the main stream has been appreciably lengthened. Yet in spite
of this, the reeds, the form of vegetation responding most quickly to
changes in the water-supply, are dead for a distanoe of 25 miles along

96
 *- — r~
Sketch Map of
EAST TURKESTAN
to illustrate the paper "by
ELLSWORTH HUNTINGTON
latural Scale P7,50O,OOO or linch - IIBstatute miles
o 50 loo 200

THE EIVEES OF CHINESE TUEKESTAN. 356
the northward continuation of the main channel, which as yet is buried
in sand only in the narrower parts.
The minor streams, wherever I have examined their lower ends in
the zone of Piedmont gravel, have old channels never occupied now by
running water, and therefore in process of being filled with sand. The
channels indicate that the streams were once larger and ran farther, but
in the absence of vegetation and ruins, the time of this greater size
cannot be determined. The true ends of the small streams, however,
as has already been said, are found in the zone of vegetation. If
the streams have diminished within recent times, that is, within the
period covered by history, the vegetation ought to show it. Between
Khotan and Charklik, where the zone of vegetation varies in width
from 10 to 20 miles, I crossed it seven times in places several
miles from the larger rivers, and therefore dependent on the water
coming underground from the smaller streams. In every case the
vegetation of parts of the zone — usually the northern parts, more
remote from the mountains— is dead or dying ; and in some cases
half-dead vegetation extends over the whole width of the zone.
Frequently for many miles one traverses reed-beds, evidently vigorous
and thick at some former time, but now consisting only of a few
struggling shoots not a tenth as numerous as those of the past, the rest
being mere stubble undermined by the wind. Elsewhere there are
large tracts of dead and dying poplars, or of tamarisks, either dead or
forming a type of mound characteristic of places where the water-
supply is lessened. The death of the vegetation is often attributed to
the encroachment of sand, hut in the cases under consideration that
has little to do with it. Often the ground is almost free from sand,
there being only a few inches, or at most scattered dunes a foot or two
high. And even this in most cases is not derived from a distance, but
is the heavier residue left by the wind as it carries away the soil no
longer held in place by living vegetation. The very form of the stems
shows that death is not due to sand. In the limited region where sand
is manifestly encroaching upon living plants, the reeds and tamarisks
shoot up abnormally to two, three, or even four times their usual height,
in a vigorous and often successful attempt to keep their heads above the
dunes. In the far larger areas of little or no sand the dead tamarisks
and reeds show nothing of such tall slender growth, but tend, on the
contrary, to become short and stunted. The widespread death in the
zone of vegetation seems to be to the smaller streams what the dead
channels in the sand are to the larger — an expression of gradual de
siccation. The most unexpected and perhaps the most significant characteristic
of the typical river is its ancient towns. Thirteen of the seventeen
larger rivers have on their lower courses the ruins of towns dating
usually from the Buddhist era, a thousand or more years ago. On six

357 THE EIVEES OP CHINESE TUEKESTAN
of the thirteen rivers the ruins appear to be of only one age ; on five,
the Dumuka, Niya, Cherchen, Vash Sheri, and Miran, of two ages ;
and on the Chira and Endere rivers, of three ages. In general, the
older ruins are the larger, and lie farther down-stream. Also, on the
whole the ancient towns were larger than their modern successors, this
being notably true of Yartungaz, Endere, Cherchen, Vash Sheri, and
Miran. And lastly, with the exception of old Cherchen and Charklik,
and possibly of Kara Dong at the end of the Keriya river, which I
have not visited, the older ruins are situated so far out in the desert,
or upon rivers so small or so saline, that it would be impossible again
to locate towns of equal size in the same places, unless a far better
system of irrigation were introduced.
A brief description of the ruins of Niya, Endere, Vash Sheri, and
Miran will give a fair idea of the whole. Modern Niya is a town of
about three thousand people, located on the southern edge of the zone
of vegetation 60 miles east-north-east of Keriya. The town, like all
those east of Keriya, is young, having been founded in the early part
of the nineteenth century ; but though it has been fairly prosperous,
decay has already begun. During part of the season of irrigation the
water does not come direct from the mountains, but from springs a few
miles above the town. Accordingly it is slightly saline, so that the
fields farthest down-stream are gradually becoming unfit for culti
vation, and are being abandoned a few every year. The cultivated
area, acoording to Stein's map, amounts to 24 square miles, including
large lacunae. Below Niya the river-bed is almost dry, but water
seeps in from springs until a good-sized stream is formed, only to dis
appear again in the course of 20 to 25 miles. Forty miles from Niya,
however, at Imam Jafir Sadik, the most famous shrine in Central Asia,
the stream once more appears, but soon sinks into the earth again for
the fourth and last time. In its lower course the river becomes so
saline that permanent cultivation of the land along its banks is impos
sible. Even if there were no such town as Niya, the water, except
during the flood season, would still be saline here at the river's end,
and fields would become unfit for use after a short period of irrigation.
Fifty-five miles below Niya, and 4 miles below the present flood
limit of the river, one comes upon the first ruins— ancient houses
hidden among large tamarisk mounds, poplars, and sand. They belong
to the Buddhist epoch, and have been described in part by Stein. I
counted over a hundred houses extending northward for nearly 14
miles. There must be many more still to be discovered, for we found
a number which were new to the guides, and one can walk within
100 feet of a house and not see it among the tamarisk mounds and
sand. There must, too, be many houses buried under the latter, so
that the total pretty surely amounts to several hundred. In view of
the number of houses and of the large area— about 28 square miles—

AND THE DESICCATION OP ASIA. 358
apparently under cultivation — which in Chinese Turkestan always
means under irrigation — the town can scarcely have been smaller than
modern Niya, and must have demanded as abundant a water-supply.
At the northern end of the old town, and overlapping it for about
2 miles, there are traces of a still older ruin. No houses or walls remain,
only thickly strewn pottery, bones, and slag, the pottery being of a
oruder type than that of the later ruins, and broken into finer fragments.
The farthest outlier of this town, so far as I could discover, lies at a
distance of 24 miles from the present end of the river. Nothing is
known of its history, and there is nothing by whioh to date it. It may
represent an earlier epoch than the ruined houses, or it may merely
represent the part of a single long-lived town which was first
abandoned. In either case the question arises whether with the present water-
supply, which is completely utilized and is very slightly saline at Niya,
it would be possible for an equally large and permanent town to grow
up at a distance of from 55 to 70 miles out into the desert, and from
5 to 20 miles beyond the flood limit of to-day. Even if there were
no such town as modern Niya, it appears to be impossible. In the
first place, if the water at the shrine of Imam Jafir Sadik is to-day
too saline to allow of permanent cultivation, it would pretty surely
be as bad or worse farther down-stream. A dense population and
prolonged intensive culture such as the ruins indicate would hardly be
possible. In the second place, it is very doubtful whether the Niya
river would reach the lower ruins even if none of the water were
diverted for irrigation. A hundred years ago there were no inhabitants
along the Niya river, or, at most, only a few shepherds. The river ran
freely as far as it would. In the upper ruins the old channel is still
visible, and the poplars and tamarisks are only beginning to die.
Among the main ruins, however, and still more among the lower ones,
the vegetation has been dead hundreds of years, so long that even in
this dry climate the poplars are reduced to mere naked trunks, and the
tamarisks have either died or grown into mounds 20 to 30 feet high.
To put the matter in another way, if the Niya river was as large a
hundred years ago as it was at the time of the ruins, it ought at the
later epoch to have flowed not only to the ruins, but well on past the
most remote of them. Vigorous vegetation ought to have accompanied
it to a point at least 30 miles beyond the present end of the river — that
is, 5 or 6 miles beyond the most remote ruins. But, as we have already
seen, most of the vegetation among the ruins has been deprived of
water for much more than one hundred years, and that at the outer
limit has not only died and dried to tinder, but has been blown away
by the wind so that only the stumps remain, cut off level with the soil.
The weight of the evidence of dead vegetation appears even stronger
when the ruins of Niya are compared with those of old Dumuka between

359 THE EIVEES OF CHINESE TUEKESTAN
Chira and Keriya. In the early part of the nineteenth century the
water-supply of Dumuka gradually diminished, until about 1840 it
became so scanty that all the inhabitants moved at once to the present
site, 8 or 10 miles farther up-stream. Modern Niya was settled before
this, but as the town grew slowly, it is safe to say that the diversion of
the Niya river for irrigation did not take place on a large scale till
about the time of the abandonment of old Dumuka. Since that time no
surface water has reached the old site of the latter, but its vegetation is
still fresh and vigorous. If an abundant water-supply was cut off from
old Niya at most only twenty or thirty years earlier, it is hard to see
why the vegetation should show little or no sign of death in the one
case, at Dumuka, and should be completely dead and decayed in the
other. It is not probable that the Niya river reached the main ruins a
hundred years ago before the founding of the modern town, or that it
would do so now if it were not used for irrigation. That a river of the
present size could not only reach the ruins, but irrigate an area as large
as that now under cultivation is still more improbable. If, however, the
climate of antiquity was colder, so that evaporation was less, or moister,
so that the rainfall was greater, and if the river was correspondingly
larger, not only would this be possible, but the difficulty from salinity
would disappear.
The Endere river, 60 to 70 miles east of the Niya, though larger,
has no proper oasis. At its lower end, eight or ten shepherds have
cultivated a little land during the past thirty years ; but the river is too
saline to admit of permanent irrigation. The shepherds say that even
fresh land can only be cultivated once in three years, and any attempt at
prolonged cultivation would be futile. Nevertheless, the river has on its
right side — first, the ruins of a large Buddhist town of early date ; second,
of an almost equally large town of later Buddhist times ; and third, of
a small Mohammedan village ; while on the left or west side it once
supported a Buddhist hamlet and a fair-sized Mohammedan village.
The two larger Buddhist sites resemble those of Niya in relative age and
location, but are of greater size and more separated, and the older town
appears to have been the more populous. The larger Mohammedan
village, a walled enclosure within which the houses are still standing,
is much smaller than either Buddhist site, its houses numbering 150.
Like the Buddhist towns on the lower Niya and Chira rivers, and the
late Mohammedan sites of old Ponak and Dumuka, it appears to have
been abandoned deliberately and of set purpose, for the houses and
walls were left intact, and everything of value was carried away. Dis
regarding the two smaller sites, we thus have on the Endere river the
ruins of three towns of different ages, which grew up one after the
other, the later town in each case being smaller than its predecessor.
To-day there is water enough to support a large town, but it cannot be
used because of the salt. The uppermost ruins lie only 12 to 15 miles

AND THE DESICCATION OP ASIA. 360
further up-stream than the present shepherd hamlet, and the lowermost
lie quite as far down-stream. Even if the difficulty due to the entrench
ment of the river to a depth of 20 or 30 feet could be overcome, and the
water could again be brought to the ancient sites, it would be too saline
to support such large and permanent towns. If, however, the water-
supply was once enough greater so that surface water reached the lower
parts of the river three or four months each year instead of only one,
the difficulty would be largely obviated, and the conditions would be
like those of modern Niya, the fields would gradually become saline,
but it would be many years before they would have to be abandoned.
On any other supposition the location of the towns and their successive
abandonment seem inexplicable.
Still further to the east, 200 miles from Endere, the Vash Sheri
river rushes swiftly down from the mountains, and, after losing most
of its volume in the Piedmont gravels, supports the thirty people of
Vash Sheri. During the months of June, July, and August, the melt
ing snow on the mountains — 30 to 40 miles away — increases the river's
volume enormously. Hearing of this large amount of water wasted
among reed-beds, a considerable number of people have come to Vash
Sheri during the thirty years since it was settled, but most have gone
away after a year or two. In summer they had more than enough
water for their fields, but during the planting season of April and May
the supply often ran short. Under the present conditions of climate
and irrigation, a population of thirty or forty is all that the river can
support. Yet in ancient times the number of people must have been
many times as great. Formerly, the river followed a more westerly
course. At the end of the old channel, and at the same distance as the
modern hamlet from the mountains, the ruins of a Buddhist town cover
an area at least 2^ miles long by 1 wide. Here, as at Niya, the ruins
consist of two parts. The older village covered the whole area. Its
houses have completely disappeared ; but if it were settled as densely
as the abundant pottery indicates, or as densely as the modern villages,
its population must have numbered five hundred or more. The later
village, or the later part of the original village, as the case may be,
occupied only the upper portions of the ruins. Traces, sometimes very
faint, of sixteen houses can be seen, and some must certainly have dis
appeared or escaped notice. It is safe to say that near the time of its
abandonment the village must have had a population of a hundred
souls — three times as many as the present hamlet — and that earlier it
must have had far more : so large a number of people could not be
supported to-day without a radical change in the system of irrigation.
Fortunately, the river changed to its new course as soon as the abandon
ment of the village allowed the rude dam of small boulders at the
head of the old channel to fall into decay. Accordingly, the irrigation
works have not been subjected to the destructive influence of floods,

361 THE EIVEES OF CHINESE TUEKESTAN
and are still almost intaot. The difficulty of keeping the river in its
old bed may have been the immediate cause of the abandonment of the
village, but it had nothing to do with the sufficiency of the water-supply
of the past as compared with that of the present. The latter depends
upon the character and width of the river-bed, which are essentially
the same in both cases, the old channel being the wider, if anything,
and upon the nature of the irrigation system, which, again, was pre
cisely the same in the past as it is to-day. Then, as now, the in
numerable minor channels into which the river divides were1 simply
dammed with banks of gravel dug up close at hand. Thus all the water
was gathered in a single larger channel on one side of the mile-wide
flood-plain, and was led off into canals, mere ditches dug in gravel or
sandy clay, as the case might be. If the river should again be turned
into the old canals, the available supply of water at the ruins would
apparently be essentially the same as that at the village to-day.
A hundred miles east of Vash Sheri the Miran or Miyan river rises
in typical fashion in the main snowy range of Kwen Lun, breaks through
the front range in a deep gorge, traverses the Piedmont gravels in a
terraced trench, and disappears during much of the year in a broad
gravel flood-plain. Geographically Miran closely resembles Vash Sheri,
and the reasoning applied to the latter holds good with increased
emphasis for the former. Archsaologically Miran is more important,
for the ruins are not only far larger, but a new type of architecture
is developed, the chief structures being elevated 10 or 20 feet on solid
pedestals of sun-dried bricks. At present Miran is not permanently
inhabited, though the fisherfolk of Abdal, on Lob Nor, 20 miles to the
north, come in summer to cultivate the reedy fields, and, by using all
the water available in spring, raise grain for about fifteen families.
The fields can only be cultivated once in three years, for in a single
season of irrigation a cake of clay 2 or 3 inches thick is deposited, a
cake so stiff that crops cannot grow in it till it has been softened by two
years of sun and rain. The natives think that the clay is gathered after
the river begins to spread over the huge gravel flood-plain, 10 miles
long and 1 or 2 wide. In reality the river is charged with fine clay
when it leaves the mountains, where it is said to be equal to the river
which supports the large village of Charklik. On reaching the flood-
plain the water sinks rapidly into the coarse gravel, but as the current
is swift, the fine clay is all borne along, until at Miran the river is
literally a stream of mud.
Anciently, conditions must have been far different. The old
Buddhist Miran was neither a hamlet, such as to-day might be located
here, nor a village like old Vash Sheri, but a large and important town.
It covered an area of at least 5 square miles, all of which, judging from
the canals and pottery, and still more from the number and location of
public and religious structures, must have been thickly populated. The

AND THE DESICCATION OP ASIA. 362
houses, being made of clay apparently, have disappeared, with two
exceptions. Thirteen other structures remain, of which one is a fort,
400 feet square ; one a lamasery, the outer walls of which are adorned
with clay reliefs of Buddha ; two are stupas, or shrines ; and the other
nine are solid rectangular masses of sun-dried brick, capped in most cases
by the walls of what may, perhaps, have been monastic dwellings or
Buddhist temples. These establishments appear to have been kept up
after the rest of the town had become depopulated, or else were re-
occupied after a period of abandonment, during which the fort fell into
ruin, as is suggested by the flimsy repairs superposed upon the solid
older structure.
Ancient Miran, in its prime, must have required a water-supply
almost scores of times as large as that now available. It is reasonable
to suppose that, being the most important place for hundreds of miles
both from a religious and a military point of view, Miran had an irriga
tion system as good as the country afforded. The water at first came
from an old river-channel east of the town, and later, apparently, from
the present channel on the west. The dams and canals are preserved
just as at Vash Sheri, and do not differ from those of to-day. The main
older dam is composed of tamarisks and small boulders thrown loosely
together ; and there is said to be another, which I could not find, com
posed of boulders alone. The canals are either simple ditches, or are
raised a few feet on banks ; but in every case the material is that which
happened to lie close at hand. On the upper Miran river there are no
inhabitants, and no opportunity for the diversion of part of the water.
The present supply, sufficient for fifteen or twenty families, is all that
the river is capable of furnishing under the system of irrigation which
prevails now, and prevailed in Buddhist times a millennium or more
ago. If, however, the rainfall were increased, say, a third, the amount
of water reaching Miran would be multiplied vastly. Being confined
to a single channel, the water would suffer scarcely more loss from
evaporation and from sinking into the ground than it now does, and all
the extra supply would be available at Miran. The difficulty from the
excessive amount of sediment would vanish, for though the absolute
quantity might conceivably be more than now, it would be diluted with
a far greater amount of water, and spread over a much larger area. The
hard cake would be reduced from 2 or 3 inches to perhaps a quarter of
an inch, not enough to interfere with cultivation. Miran, even more
than the other ruins, seems to verify the hypothesis of a change of
climate during the historic era.
The ruins of Miran, Vash Sheri, Endere, and Niya are typical of
many others. Charklik and Cherchen belong to the same class as
Miran and Vash Sheri, the old towns having been located at nearly
the same place as the new, and having covered a larger area. This
is especially the case at Cherchen, where Lellik, the southern and

363 THE EIVEES OF CHINESE TUEKESTAN
apparently older ancient site, covered an area three times as large as
the modern town, and the northern later site may have been as large.
Neither Charklik nor Cherchen is conclusive, however, for the size
of the ruins at the former cannot be determined, and the whole
water-supply at the latter is not now utilized. Yartungaz resembles
Endere, but the river is much more saline. Accordingly, as might
be expected, the ruins are old and rather small, and lie at the highest
possible point, like most of the modern oases. The remaining ruins,
lying west of Niya, belong to the same type as the latter. Eawak,
Dandan-Uilik, and Uzun Tetti were successive towns on what appears
to have been the ancient extension of the Chira river, aided more or
less by the smaller streams to the east. The pottery-strewn site of
Lachinata lies beyond the end of the Gulakhma river ; the Buddhist
ruins of Payet Beg and Khadaluck and the recently abandoned towns
of old Dumuka and old Ponak indicate the former greater length of the
Dumuka and Ponak rivers, and an unnamed site in the sand shows
that the Karakir stream once extended 15 miles beyond its present
limit. On the northern border of the Tarim basin there is less opportunity
for the occurrence of ruins of the kind described above, for most of the
streams are either very small, or else, uniting in the highlands, issue
from behind the mountains as large rivers which join the Tarim, and
will be considered later. At Ying-pen, however, north-west of Lop-Nor,
the ruins of a small fort and village described by Hedin are now water
less. I followed the old canal northward for over 3 miles to its head,
and found that it was a simple ditch with two small ponds or reservoirs,
into which the water was turned from the little Shindi brook, the
largest stream in the Kuruk Tagh, or " Dry mountains." To-day no
water reaches the ruins during the greater part of the year, and the
place cannot be inhabited, yet formerly the water-supply sufficed to
support a small village.
Similar instances of ancient towns located where there is now no
sufficient water-supply are described by Stein as far west as Kashgar ;
and I found others as far north as the depression of Turf an.
Eastward beyond the end of Lob Nor, on the course which the united
Bulundsir and Tan-ho rivers once followed on their way to the lake,
two other ruined sites of the same sort are described by the natives.
The Chinese of Dung Khan (Sa Chau), higher up on the Tan Ho, say
that the ruins were abandoned by their ancestors long ago because the
water of the rivers would no longer reach them.
In the case of the rivers of the Tarim system, the evidence of desic
cation is less clear than among those of the withering type, for the
streams are so large that they still support extensive oases and termi
nate at a fixed point in Lob Nor. Still, the increasing salinity of the
lower Tarim, and the decreasing size of the terminal lake, seem to show

AND THE DESICCATION OF ASIA. 364
that the rivers of this group, like those of the other, have suffered
desiccation. The region along the lower Tarim or Yarkand river for
400 miles, and along its parallel tributary, the Konche, for 250 miles
above the mouth of the united streams in Lob Nor, is to-day praotically
uninhabited. The scattered little settlement of Tikkenlik is the only
real village, and almost the only place where agriculture of any sort is
practised ; and even its five hundred inhabitants could not Bubsist, so
they say, if it were not for their flocks and for the fish which they eat
in summer with young reeds and rushes. The great difficulty is salinity
in both soil and water. Indian corn, one of the great staples of Chinese
Turkestan, will scarcely grow at all, and is replaced by millet. Wheat
fares better, but only the first year. If a field is cultivated several
years in succession, the crops rapidly dwindle just as at Endere and
Yartungaz. The same difficulty prevails all along the lower course of the Tarim
and Konche rivers, and several recent attempts of the Chinese to found
towns have proved abortive. About 1889 some colonists settled at
Jan Kul, 30 miles west-north-west of Tikkenlik, on the Tarim. Next
year the population increased to over 2000, and Jan Kul, as the people
say, " became a town ; " that is, a bazaar was established, and an
"amban," or local governor, was installed. But the fields quickly
became saline, and after two years, in 1892, the place was abandoned,
most of the settlers returning whence they came. Meanwhile, however,
in 1891, about 1000 people had come from Turfan to Dural, 8 miles
south of Tikkenlik, and had begun to raise wheat. Accordingly a large
fort was built there, and in 1893 the amban of Jan Kul was removed to
Dural, which in turn " became a town." For three years the population
increased, but by 1897 the soil had become so salt and the crops so poor
that the colonists began to move away. In 1898 over 2000 rebellious
Dungans or Mohammedan Chinese were induced, or forced, to come
from Shi-ning, 800 miles to the east, and settle at Dural, but the land
and water were so bad that in 1900 they migrated 90 miles north-west
to Karakum, on the Konche, another site which had just been opened.
Thereupon, in 1901, the amban was removed to Karakum, and in 1904
the last of the people of Dural came to Tikkenlik. Karakum became
even larger than its predecessor, and from 1901 to 1903 had a population
of nearly 5000, but the salt was as bad as elsewhere, and in 1904 it too
was abandoned, though the amban and forty or fifty people still remain.
Jan Kul, Dural, Karakum, and Tikkenlik represent four abortive
attempts during sixteen years to utilize the water of the lower Tarim
and Konche rivers for irrigation. Tikkenlik, being but three or four
years old, still survives, but its people are moving away or else are
abandoning agriculture and betaking themselves to sheep-raising and
fishing, the means of livelihood of the former hamlet before the attempt
to establish a town.

365 THE EIVEES OF CHINESE TUEKESTAN
Turning now to the past, we find a wholly different state of affairs.
In Buddhist times the Tarim and Konche rivers were deflected east
ward in whole or in part between Karakum and Tikkenlik, and flowed
to the ancient ruins of Lulan, discovered by Hedin. In approaching
Lulan from the north-east, and in leaving toward the north-west, I
rode 15 miles each way through a region where houses, pottery, and
other evidences of human occupation show that the whole area was
once thickly settled by an agricultural population. Farther west, on
again entering the broad zone of dead vegetation which accompanies
the old river-courses, I found pottery, beads, slag, and bits of iron,
for a distance of nearly 40 miles, and there was a little ancient
pottery in a strip of dead vegetation between two arms of the Konche
only 25 miles east of Tikkenlik. It looks as though in ancient times
an area 100 miles east and west, by 10 or 15 miles north and south,
was thickly dotted with villages irrigated by the Tarim and Konche
rivers. Even if the present water-supply would suffice for so large
an area, a densely populated, flourishing district like the Lulan of
history would be impossible because of the water's salinity. Here, as
in so many other cases, the phenomena are easily explicable if the
rivers have decreased in size.
The lake of Lob Nor agrees with the rivers, for it shows signs of
having been larger at no very distant date. In this article it is not
possible to consider the controverted history of the " Great Salt Lake,"
as the ancient Chinese called it. It must suffice to say that in travelling
around it, and crossing the unexplored salt desert at its eastern end,
the presence of five, or possibly six, old shore-lines at altitudes of from
12 to 600 feet above the present level, and of a thick series of lacustrine
deposits, indicates that the lake has passed through a succession of
epochs of expansion and contraction corresponding to those of the
lake of Sistan in Eastern Persia, and less closely of Great Salt lake
in North America, and to the epochs of the Glacial Period all over the
world. The freshness of the dry lake-bed and bluffs, the distribution
of vegetation, the presence of an ancient road, now unused and unknown,
which makes a long dMour around a bay of the old lake-bed instead of
crossing it as the modern road does, and the well-known fact that
the present lake or swamp has diminished in size during the past
century, seem to indicate that the last notable expansion of the lake
took place, or at least had not yet passed away, when historic man
occupied the surrounding region.
The rivers which wither in the desert, together with those which
join the Tarim and terminate in Lob Nor, drain most of the mountains
around the Tarim basin, but there is one part so extremely dry that
there are practically no running streams whatever, only a few scattered
springs, mostly salt. This inhospitable distriot, as large as Great
Britain, lies between Lob Nor on the south, Bagrash Kul on the

AND THE DESICCATION OF ASIA. 366
west, and Turfan on the north, and stretches eastward 400 or 500
miles toward the desert of Gobi. It is a desolate region of naked
mountains half buried in plains of gravel, and is well named the
"Dry mountains" ("Kuruk Tagh") on the south, and the "Desert
mountains" ("Choi Tagh") on the north. It is so little known that
the best I can do is to quote what was said by my guide, a remark
ably intelligent camel-hunter of Kuzzil Singer, the only inhabited
place, who knows every spring and mountain for 100 miles in every
direction. I asked him whether there was more or less water now
than in the past, or just the same. In reply he told of the drought
during the last four or five years which every one speaks of, and then
went on, "And long, long ago, in the days of which neither my
grandfathers nor their grandfathers ever heard, there must have been
much more water. In the high mountains there are many places
where little stone shepherd's houses, with the roofs all fallen in, stand
in valleys where nobody has ever known of there being any water.
The nearest water is sometimes 5 or 10 miles away. Surely no one
would have built houses and kept flocks and herds in those places
unless there had been water. How many such houses have I seen ?
Oh, many ; I never counted, but the mountains are full of them. In
other places, away from the higher mountains, one often finds salty
deposits and a few reeds where once there must have been springs, and
on every side there are old paths coming in where once the wild camels
and antelopes used to come regularly to drink. Now, as there is no
water and only a very few reeds, scarcely two or three animals come in
a year. There were men here then, for near almost all of the dry
springs there are old shelters of rocks and pieces of tamarisk, just such
as we make now when we lie in wait for game. I never talked to any
one about this before, and I do not know whether I am right, but I
have seen these things when I have been hunting, and," as the Turki
idiom puts it, " that is what I know in my stomach." My own limited
observation agrees with that of the camel-hunter. Moreover, I found
traces of ancient roads and of villages where there is now no water, and
the basin of the Turfan, near by to the north, is full of evidences of an
historic decrease in the water-supply.
The phenomena of rivers, large and small, of springs, lakes, ruins,
and vegetation, all seem to point to a gradual desiccation of Chinese
Turkestan for nearly 1500 miles east and west, and 500 north and
south. The records of antiquity indicate that the Caspian and Aral
seas were once more extensive than now, and presumably that the
tributary rivers were much larger. Transcaspia, Eastern Persia, and
the neighbouring regions appear to have been subject to a desiccation
similar to that of Turkestan, and indicated by similar phenomena.
Thus all the more arid part of Asia, from the Caspian sea eastward
for over 2500 miles, appears to have been subject to a climatic

367 THE EIVEES OF CHINESE TUEKESTAN.
change whereby it has been growing less and less habitable for the
last two or three thousand years. Such a change, embracing Persia,
Baluchistan, Afghanistan, Kussian Turkestan, and Chinese Turkestan,
and probably an even larger area, and converting thousands of
square miles of habitable country into desert, can hardly fail to have
had some effect upon human distribution and history. In Chinese
Turkestan part, at least, of the ancient towns described above appear to
have been abandoned deliberately, presumably because the water-supply
became limited. Movements of the same sort are to-day taking place on
a small scale all over the arid parts of Asia ; as, for example, in the
cases already cited of the inhabitants of Dumuka and other villages who
moved in a body to a new location, or of the Dungans and people from
Turfan, who wandered here and there along the lower Tarim. In 1904,
500 Kossack nomads, with their flocks and herds, came to Lob Nor, and
would have settled among the reeds of the lower Cherchen river if
Chinese soldiers had not driven them out with bloodshed. The nomads
had come from the Altai mountains, where pasture had become too
scanty either because of increasing drought or increasing population,
or more probably both. Such movements only need to be magnified to
cause invasions like those of the Dark Ages. That they were magnified
in Turkestan when the Buddhist towns were abandoned appears highly
probable. The abandonment took place in the early centuries of the
Christian era, and for most of the time from that day to this the
whole country from Keriya eastward for 600 miles, almost to China
Proper, remained practically uninhabited, even more so than to-day.
Not only the inhabitants of the towns disappeared, but also the shep
herd tribes among the mountains, who, judging from names, traditions,
and ancient roads and dwellings, were formerly numerous, and who
would feel the pinch of aridity more quickly and keenly than the towns
people. Part of the population may have been killed in war, or have
died in the epidemics which rage when a country is over-populated, but
the rest must have migrated. If, though only in bands of a few
thousands, they invaded regions already well populated, where the means
of livelihood were growing less because of desiccation, distress and wars
would ensue, and the only relief would lie in further and greater migra
tions. And thus, all Central Asia being under the same ever-increasing
stress, a succession of dry years, or a change such as has taken place in
Chinese Turkestan since 1900 from an increasing to a decreasing
quantity of water in the rivers, might cause migrations equal to the
greatest recorded in history.

PRINTED BY WILLIAM CLOWES AND SONS, LIMITED, LONDON AND BECCLES.

YALE UNIVERSITY LIBRARY

3 9002 08837 1860