Page 54

of poorly bleached quartz grains. This last value was
choosen as the mean equivalent dose (tab. 2). However,
because of the observed scatter, it cannot be totally excluded
that this value is over or under-estimated. In
Table 2 are summarized the dosimetric data relevant to
the sediment samples. All errors are given at 1 σ level;
age uncertainties include statistical and systematic
errors.
4.3 - URANIUM/THORIUM DATES (tab. 3 and fig. 4)
Three travertine samples (no3, 5 and 35) belong to the
Fort Oued Noun main section, and two ( a calcrust, no3-
5, and 33) came from the left bank of the wadi, in equivalent
stratigraphic succession. The collected samples
thus come from the base (no3, 3-5, 5) and from the top
(no33, 35) of Unit 2. All samples, except no3-5, are true
travertines, of hard cemented laminated facies.
All samples were analyzed after total dissolution and
addition of a 232U/228Th tracer solution using a method
similar to that described by Ku (1976). Alpha counting
was carried out with solid-state detectors and errors are
expressed as one standard deviation based on counting
statistics.
Impure carbonates are generally considered as unreliable
species for Th/U dating (Ku & Liang 1984; Bischoff
& Fitzpatrick, 1991; Kaufmann, 1993) for two
main reasons: (i) presence of detrital 230Th in dirty samples
will increase the 230Th/234U ratio and make the
sample look older; and (ii) post deposition addition of
Uranium from groundwaters gives rejuvenated ages.
Most samples in Table 3 were corrected for a detrital
contribution using a 230 Th/232Th ratio of 1, measured in
sediments sampled in the studied area. For sample no5,
an isochron technique was attempted, using total sample
dissolution (Bischoff & Fitzpatrick, 1991). The
isochron age of 92.2 ka compares well with the corrected
age of 87,6 ka of no3-5, in the same stratigraphic
situation. Sample no3 was dated on the bulk sample and
on a leached (HNO3 1N) aliquot, both giving concordant
ages.
5 - DISCUSSION (fig. 4)
5.1 - UNIT 2, UPPER PART (Layers 29 - 35)
The 14C dates, from charcoal, show a securely dated
channel deposit (no 31) with an age of about 28 ka B.P,
but more imprecise results for the other channels (sampled
beds 29, 33 and 35) using Melanopsis shells,
whose ages lie between ≥ 28 ka B.P. – and ≥ 18,5 ka
B.P. The two OSL dates of channel 31 are concordant
with 14C results, but the date of 30,9 +/– 2,5 ka obtained
for layer 35 overlying channel deposit 33 is too old
when compared to 14C (≥ 24 ka) and U/Th
(21,6 + 4,3 ka and 24,4 + 1,6 – 1,5 ka). The spread of
dates from the three techniques might arise through an
overestimation of the OSL age; because of insufficient
bleaching as noted previously, and from the fact that
the carbonated secondary mineral dated by U/Th had
developed within that sediment after its deposition
(Nanson et al., 1991).
Despite these problems, the radiometric dates clearly
indicate that the upper part of Unit 2 (layer numbers 30
to 35) was deposited between ca 30 and 21 ka B.P., and
perhaps even up to 18,5 ka B.P. These ages fit well with
those obtained for the overlying Unit 3 dated by OSL at
17 ka B.P. at a neighbouring site (Mathieu, to be published).
5.2 - UNIT 2, MIDDLE PART (Layers 9 – 28)
OSL results indicate that the most important part of
the silty sedimentation (more than 10 m thick between
layer numbers 14 and 28) occured between 39,0 +/– 3,2
and 36,8 +/– 3,2 ka B.P. The errormargins don’t allow us
to choose between two hypotheses of duration: long (between
42 and 33 ka) or short (between 39 and 36 ka).
5.3 - UNIT 2, BOTTOM PART (Layers 2 – 8)
OSL and U/Th dates of this basal portion of Unit 2
give two groups of results. U/Th ages for layer no3
(laminated travertine) and an OSL age for layer no6
silts suggest an age range of 55,9 + 9,3 –8,6 ka to
49,1 +/– 4,3 ka B.P. However, the U/Th ages for calcrete/
travertine in layer no 5, of 92,2 ka, and in layer
no3-5, of 87,6 + 9,1 – 8,5 ka B.P., are not concordant
with the other dates (fig. 5). Sample layer no3-5 came
from the opposite side of the valley and was not in the
same stratigraphic succession as the other samples.
Therefore, either this date is correct and the stratigraphic
interpretation of layer no3-5 is wrong, or the
date is unsound. The U/Th isochron age of 92.2 ka for
sample no 5 remains inexplicable, although it is known
that the ages of indurated secondary carbonated deposits
may not necessarily fall in strict chronostratigraphic
order (Nanson et al., 1991). So, presently, we prefer to
consider the dates from layer numbers 3 (U/Th) and 6
(OSL), which are stratigraphically coherent, as
representating the time between carbonate closure in
the travertine (defined by U/Th dating) and overlying
silt sedimentation (defined by OSL dating). Therefore,
the age span of this lower part of Unit 2 lies possibly
before 55 ka and after 49 ka. This possibly means that
the O.I.S period 4 was characterised by a higher frequency
of channel forming sediment events than
O.I.S.3.
In spite of their uncertainty, the two results about
90 ka could not be abandoned: we have recorded similar
dates on calcrete/travertine deposits in apparently
similar stratigraphic positions in neighbouring sites,
as has M. Thorp further East (in litteris). These deposits
are possibly preserved remnants of older formations
at the base of the Soltanians, and may imply a beginning
of the infilling of the valley at O.I.S. 5b-5a,
as previously suggested with regard to the relation