New single crystal 40Ar/39Ar ages improve time scale for deposition of the Omo Group, Omo–Turkana Basin, East Africa

Six tuffaceous beds within the Omo Group of the Omo–Turkana Basin have been dated using the 40Ar/39Ar single crystal total fusion method on anorthoclase, yielding eruption ages. The Omo Group constitutes up to 800 m of subaerially exposed sediments surrounding Lake Turkana within the East African Rift system in northern Kenya and southern Ethiopia. Rhyolitic explosive eruptions produced tuffs and pumice clasts that are considered to have been deposited shortly after eruption. The new age data on feldspars from the pumice clasts range from 4.02 ± 0.04 Ma for the Naibar Tuff of the Koobi Fora Formation to 1.53 ± 0.02 Ma for Tuff K of the Shungura Formation. The Orange Tuff in the KBS Member of the Koobi Fora Formation was dated at 1.76 ± 0.03 Ma, providing good control in this part of the sequence where formerly there was a >200 ka gap. Data are consistent with earlier measurements and significantly improve age resolution within the Omo Group, which has yielded many vertebrate fossils, including hominin fossils comprising a number of species. We suggest new age estimates for a limited number of hominin specimens. Supplementary material: Eleven tables and nine figures are available at www.geolsoc.org.uk/SUP18506.

The Omo-Turkana Basin in northern Kenya and southern Ethiopia comprises the region around the present-day closed basin of Lake Turkana. Dating of tuffaceous products within the sedimentary sequences shows that deposition within the basin was initiated about 4.24 Ma ago, and has continued since (McDougall & Brown 2008). Much of the sedimentary material deposited in the basin was brought in by the Omo River, which drains part of the Ethiopian highlands. The basin extends at least 350 km northsouth and up to 90 km east-west ( Fig. 1). It lies within the East African Rift system where the NNE-SSW-trending Ethiopian Rift intersects the north-south-trending Kenya Rift. Pliocene and Pleistocene sedimentary rocks exposed in the Omo-Turkana Basin have an aggregate thickness of up to 800 m subaerially, but interpretation of seismic data indicates much greater thicknesses beneath Lake Turkana ( Fig. 1; Morley et al. 1999). Initially the various regions were thought to comprise separate depocentres, which is reflected in the stratigraphic nomenclature. Thus, the Shungura Formation was defined and mapped north of Lake Turkana in southern Ethiopia (de Heinzelin 1983), with the Koobi Fora Formation east of Lake Turkana (Brown & Feibel 1986), followed by the Nachukui Formation west of Lake Turkana (Harris et al. 1988); all now collectively placed in the Omo Group of de Heinzelin (1983).
Correlating strata between regions is not straightforward because of major facies variations from fluvial to deltaic and lacustrine over short distances. Fortunately, tephras, with their distinctive elemental signatures of the contained volcanic glass, have allowed stratigraphical correlations from area to area. These tephras and tuffaceous sediments have been mapped and named as tuffs (see also Brown 1972). Chemical analysis of glass shards in these tuffs has shown that some are widespread. As many of the tuffs have been deposited by water it has become increasingly obvious that the sediments of the Omo-Turkana Basin are all part of a single depositional system . Thus, through successive improvements in correlation of tuffs within the basin, a robust stratigraphic framework has been developed (see Brown et al. 2006, and references therein). Numerous vertebrate fossils, including relatively large numbers of hominins, recovered from the sequences have been a major encouragement to further extending correlations throughout the basin.
Some tuffs locally contain pumice clasts, which usually have glass of the same composition as the enclosing tuff, indicating derivation from the same volcanic eruption. The pumice clasts commonly have alkali feldspar (anorthoclase) phenocrysts, which are ideal for isotopic dating by the K/Ar and the 40 Ar/ 39 Ar dating techniques. Thus, not only have the tuffs allowed correlations throughout the basin, but some have been isotopically dated, yielding a numerical time framework for the sequences. More than 30 units in the Omo-Turkana Basin have been dated over the last three decades, with most now having been measured using single crystal 40 Ar/ 39 Ar total fusion techniques , 2008. Virtually all these ages are consistent with the stratigraphic order, indicating not only that the ages record the time of eruption, but also that deposition occurred very shortly after eruption. Isotopic dating results already available mean that fossils can usually be assigned ages to better than 100 ka, without further direct dating, provided that the fossils can be placed within the stratigraphic sequence of tuffs. However, in some parts of the section gaps still exceed 400 ka between dated tuffs, so that considerable interpolation is required to assign an age to a single fossil, or to derive an age for a particular stratigraphic level. Here we present age data for several previously undated levels in the Omo-Turkana Basin to provide even better temporal control. This work demonstrates that linear stratigraphic interpolation between dated tuff levels in some cases provides only approximate answers.

Methods
For the present study we separated feldspar from pumice clasts in previously undated tuffaceous beds in the Shungura Formation and the Koobi Fora Formation, north and east of Lake Turkana, respectively (Fig. 1). These pumice clasts ranged in diameter from about 10 to 130 mm. A feldspar concentrate was prepared as previously described by , which was then treated ultrasonically for c. 5 min in 7% HF to help remove glass attached to the crystals. Hand picking under a binocular microscope followed, to prepare a high-purity concentrate for irradiation. Most of the concentrates were alkali feldspar (anorthoclase); however, pumice clasts from Tuff C-gamma yielded only plagioclase. Crystals were normally 0.4-1.4 mm in diameter. Samples were loaded into two 21-pit aluminium discs, along with the neutron fluence monitor Fish Canyon Tuff sanidine (92-176), following the geometry illustrated by Vasconcelos et al. (2002). The irradiation discs were closed with aluminium covers, wrapped in aluminium foil, vacuum heat sealed into quartz vials, and irradiated for 1 h in the cadmium-lined B-1 CLICIT facility, a TRIGA-type reactor, Oregon State university, uSA.
Following a decay period of c. 6 months after irradiation, samples were placed in wells in a copper tray and installed in the ultrahigh-vacuum extraction system for 40 Ar/ 39 Ar measurement at the university of Queensland (uQ). Before analysis, the mineral grains and fluence monitors were baked-out under vacuum at c. 200 °C for c. 12 h. Each aliquot was fused with a continuouswave argon-ion laser with a 0.2 mm wide focused beam. Single crystals were used for dating most samples from the Omo-Turkana Basin. In all cases, zero-age glass was added to the feldspar to assist fusion. The gas released was cleaned through a cryocooled cold-trap (T c. −120 °C) and two C-50 SAES Zr-V-Fe getters and analysed for argon isotopes in a MAP 215-50 mass spectrometer equipped with a third C-50 SAES Zr-V-Fe getter using procedures described by Vasconcelos et al. (2002). Measurement of ion beam intensity was through a Balzers multiplier (model SEV217) read on a Keithley electrometer (model 6512) operated in the current mode. Sensitivity of the mass spectrometer was about 3.1 × 10 −14 mol nA −1 for the largest ion beam, 40 Ar, which was usually in the range of 0.3-2.5 nA for the unknowns with 40 Ar*/ 39 Ar ratios of 1.2-8.4, depending upon age. Full system blanks were determined after every second or third run; argon from an air pipette was analysed at least once per day. Automation and analytical procedures followed are as described by Potts (1990) andVasconcelos et al. (2002).
Isotopic data from the mass spectrometer were corrected for full system blanks, mass discrimination, nucleogenic interferences, and atmospheric contamination following procedures of Vasconcelos et al. (2002), using the software Mass Spec Version 7.527 developed by Alan Deino of the Berkeley Geochronology Center, uSA. A 40 Ar/ 36 Ar value of 295.5 ± 0.5 for atmospheric argon was used for calculation of the mass spectrometer discrimination (Steiger & jäger 1977). The irradiation factor (J) for each aluminium disc was determined by laser total fusion analyses of between 12 and 15 pairs of crystals of the neutron fluence monitor, Fish Canyon sanidine 92-176, using a reference age of 28.10 ± 0.04 Ma (Spell & McDougall 2003). These measurements yielded J values of 2.689 (± 0.008) × 10 −4 (samples 5614-5635), and 2.690 (± 0.007) × 10 −4 (samples 5640-5661), where errors quoted are 1σ of the population. There was no detectable gradient of fast neutrons across an irradiation disc. All ages are reported using the decay constants of Steiger & jäger (1977).
Statistical treatment utilized to assess data followed that previously adopted by , 2008. Measurements made on gas from unfused samples or from quartz crystals were rejected, recognized usually from the very small ion beams. Results that appeared to be adequate for age determination were then arithmetically averaged, and any that were more than 2σ from the mean of the population were rejected as outliers. This approach was repeated as necessary until no outliers were detected. It is noteworthy that when analyses were pooled from more than one sample from a given stratigraphic level, the application of the 2σ test in some cases resulted in further rejections. justification for the use of an arithmetic mean age is that in most cases the uncertainties of the single ages are comparable. Where this is not the case it is pointed out in the text. The arithmetic mean age and its error generally is used in the discussion (see  for further justification). Table 1 gives a summary of results. The weighted mean age for each sample is also given in Table 1, based upon the statistical calculation from the ages themselves with quadratic addition of a multiplier of square root of the MSWD, where the latter exceeds 1.0, and the uncertainty in the J factor, given above. The age derived from analysis of the data in a 39 Ar/ 40 Ar vs. 36 Ar/ 40 Ar plot by least-squares regression (york 1969) is included in Table 1. The errors are 1σ and include the error in J, but not the errors in the irradiation correction factors, or the uncertainty in the potassium decay constants. These isochron ages agree fairly closely with the pooled results calculated directly from the single grain ages (Table 1), but the isochron ages are not used in the subsequent discussion. This is mainly because, in most cases, data are concentrated near the 39 Ar/ 40 Ar axis, so that the uncertainties in the derived composition for the trapped argon are often relatively high.
Age-probability diagrams (probability density plots, or ideograms) are based on the assumption that the errors for an age determination have a Gaussian distribution (Deino & Potts 1992).
A different Fish Canyon Tuff separate, FC-2, was irradiated together with the Fish Canyon sanidine fluence monitor 92-176 used in the present experiments. The ages of FC-2 were measured on single crystals. The mean arithmetic age found for 29 FC-2 crystals was 28.11 ± 0.12 Ma, where the error is 1σ of the population. Following successive rejection of data more than the 2σ from the mean, the mean arithmetic age becomes 28.09 ± 0.09 Ma, based on 21 analyses, which also yields a near-perfect Gaussian probability plot, giving a weighted mean age of 28.10 ± 0.12 Ma. All these ages are indistinguishable from the reference age used for fluence monitor 92-176 from the same unit, and provide great confidence that the ages on unknowns reported here are compatible between the ANu and uQ laboratories.  Analyses of Alder Creek Rhyolite sanidine (ACs) further increases our confidence in the results. The pooled measured age of 40 single crystals of ACs was 1.186 ± 0.017 Ma, or 1.184 ± 0.010 Ma (n = 32) after successively applying the 2σ rejection criterion ( Table 1). The ages yield a Gaussian probability plot with a weighted mean age and error of 1.186 ± 0.004 Ma (Fig. 2). The weighted mean age quoted for this standard by Nomade et al. (2005) is 1.196 ± 0.001 Ma, adjusted to the value of 28.1 Ma used in this study for the age of the Fish Canyon sanidine (92-176) fluence monitor. The age found here for ACs is younger than that given by Nomade et al. (2005) by 0.84%, although the ages are nearly in agreement when the errors are taken into account. Thus, the evidence is strong for the compatibility of results from the two laboratories as well as reasonable consistency with at least one other laboratory.

Naibar Tuff
This 3.55 m thick vitric tuff overlies a mudstone in Area 117 of the Koobi Fora region (Buchanan 2010). Here we name it the Naibar Tuff for its occurrence along Il Naibar, the principal ephemeral stream in Area 117, designating its type locality as the exposure south of Il Naibar at 4.0966°N, 36.2795°E. The tuff has two distinct layers separated by a thin (c. 2 cm) calcite-concreted layer with rhizoliths. The lower layer (1.05 m) is a pale green medium-grained vitric tuff with minor feldspar and quartz. The upper layer (2.50 m) is a grey reworked tuff with many cognate and detrital mineral grains, principally quartz and feldspar. A 10 cm thick ostracod-rich, calcite-concreted sandstone overlies the Naibar Tuff. In the Allia Bay region, the Naibar Tuff lies c. 13 m below the Moiti Tuff, and therefore belongs within the Lonyumun Member of the Koobi Fora Formation (Figs 3 and 4). Haileab (1995) reported analyses of glass from this tuff (samples K82-870 and K82-876), but described it as either unknown, or associated with the informally named Nabwal tuff. Brown & Feibel (1986) commented that glass from pumices from the Moiti Tuff is compositionally similar to that in the Naibar Tuff (sample 75-117A; Brown & Feibel 1986), but different from glass of the Moiti Tuff. For this study, well-rounded, calcite-concreted pumices 10-13 cm across were collected from the Naibar Tuff near its type locality. Glass of two of these pumice clasts is compositionally indistinguishable from glass shards from the Naibar Tuff, but it is distinct from glass shards in the Moiti Tuff, and also from glass in pumices from the Moiti Tuff in the Allia Bay region (Buchanan 2010).
In Area 117, the overlying Moiti Member is absent and presumably was eroded away before deposition of the Lokochot Tuff, the basal unit of the Lokochot Member of the Koobi Fora Formation (Fig. 3). The Lokochot Tuff (3.60 ± 0.05 Ma, McDougall & Brown 2008) occurs in the section about 3 m above the Naibar Tuff in Area 117 (Fig. 4). Elsewhere in the Koobi Fora region, for example near Snail Hill (3.693°N, 36.388°E), the Naibar Tuff occurs stratigraphically below the Moiti Tuff and it also lies below the Topernawi Tuff (Fig. 4).
Feldspar crystals were separated from six pumice clasts (08-036A to F) from the Naibar Tuff in Area 117, and 10 single anorthoclase crystals from each clast were analysed. The arithmetic mean ages of the six pumices are indistinguishable (Table 1). Thus, an overall average was calculated, yielding an arithmetic mean age of 4.023 ± 0.038 Ma, from 53 crystals out of 60 actually measured.    The main tuffaceous beds are shown in black, often named, and correlations between the stratigraphic columns are shown by grey lines. The Shungura Formation composite section is after de Heinzelin (1983), the Koobi Fora Formation is mainly from Brown & Feibel (1986), noting a significant hiatus, and that for the Nachukui Formation is derived from Harris et al. (1988) with minor modification. The sequence at Kanapoi, to the east of Loperot, is after Leakey et al. (1998) and Feibel (2003), and the sequence at Kibish to the north of Lake Turkana is after Brown & Fuller (2007). To the left of the stratigraphic columns for the Koobi Fora and Shungura formations are the measured magnetic polarities, where grey shading represents normal polarity and unshaded indicates reversed polarity. Data are from Hillhouse et al. (1986) and Lepre & Kent (2010) for the Koobi Fora Formation and after Brown et al. (1978) and Kidane et al. (2007) for the Shungura Formation. The ages shown for the polarity boundaries are from Gradstein et al. (2004). Most samples are single crystal alkali feldspar, except for feldspar from Tuff C-gamma, which is plagioclase. λ = 5.543 × 10 −10 a −1 . Fluence monitor 92-176 sanidine from Fish Canyon Tuff; reference age 28.10 Ma from Spell & McDougall (2003). Discs 300 and 301 in irradiation no. 902, April 2009, 1 h in CLICIT facilty, TRIGA reactor, Oregon State university. For disc 300 (08-035; 08-031A, B, C; 08-029; 08-032A, B; 08-034A, B; 08-036A), J = 2.689 × 10 −4 (± 0.30%), where quoted error is the standard deviation of the population. For disc 301 (08-024; 08-026A, B; 08-020A, B; 08-036B, C, D, E, F), J = 2.690 × 10 −4 (± 0.26%), where quoted error is the standard deviation of the population. Corrections used: ( 36 Ar/ 37 Ar) Ca = 2.57 (± 0.25) × 10 −4 ; ( 39 Ar/ 37 Ar) Ca = 6.91 (± 0.94) × 10 −4 ; ( 40 Ar/ 39 Ar) K = 8.0 (± 3.0) × 10 −4 . Numbers in parentheses after the isochron age refer to the number of analyses that were utilized in the calculation of the best-fit line; others have been eliminated as lying outside acceptable limits. The K/Ca ratios are imprecise mainly because at least 5 months elapsed between irradiation and measurement, so that much of the 37 Ar had decayed.

SECTION LOCATIONS
Results from three crystals had been eliminated at the single pumice level using the 2σ criterion, and a further four were discarded using the same criterion when all the results were brought together. The weighted mean age and the isochron age are concordant with the arithmetic mean age (Table 1, Fig. 5h).
The age for the Naibar Tuff is in keeping with the stratigraphy, as it is marginally older than ages determined by McDougall & Brown (2008) for the overlying Topernawi Tuff (3.987 ± 0.025 Ma) and Moiti Tuff (3.970 ± 0.032 Ma) (see Fig. 3). Although the Naibar Tuff lies in the lower part of the Koobi Fora Formation, it is not as old as silicic volcanism recorded in the Kanapoi region and at Lothagam, to the SW of Lake Turkana, where ages range from 4.11 ± 0.03 Ma to 4.24 ± 0.04 Ma (Leakey et al. 1998;McDougall & Brown 2008). At present Lothagam and Kanapoi have the earliest Pliocene sediments in the Omo-Turkana Basin for which ages have been established.
The measured age of the Naibar Tuff is consistent with its reversed magnetic polarity. The overlying Moiti and Topernawi tuffs also are reversely magnetized. Given the measured ages all should lie within the Gilbert Reversed Chron (3.596-6.033 Ma; Gradstein et al. 2004). The Cochiti normal polarity subchron, within the Gilbert Chron, has estimated boundary ages of 4.187 and 4.300 Ma (Gradstein et al. 2004), significantly older than the Naibar Tuff.

Tuff B-delta
Tuff B-delta in the Shungura Formation was identified and described by de Heinzelin & Haersaerts (1983) at a single locality extending over about 200 m. It is a cross-bedded coarse-to medium-grained tuffitic sand with sparse rounded pumice clasts <10 cm across deposited within a channel incised into Tuff B-gamma and Tuff B-beta (de Heinzelin & Haersaerts 1983). Tuff B-delta lies c. 80 m above the local base of the Shungura type sequence, and c. 13 m above the base of Member B. Tuff B correlates with the Tulu Bor Tuff of the Koobi Fora Formation (Brown & Cerling 1982). Arithmetic mean ages on alkali feldspar from four pumice clasts collected from Tuff B-delta range from 3.391 ± 0.025 Ma to 3.497 ± 0.064 Ma (Table 1), although three of the four pumice clasts have a much narrower mean age range from 3.391 ± 0.025 Ma to 3.427 ± 0.021 Ma (Table 1). However, the overall mean arithmetic age is 3.419 ± 0.036 Ma, based on results from 37 of 46 crystals measured. The pooled weighted mean age (Fig. 5g) and isochron age are similar but with smaller uncertainties, as they are standard errors. If results from sample 08-034B are omitted because their mean age is somewhat older at 3.497 ± 0.064 Ma, the arithmetic mean age changes only marginally to 3.412 ± 0.033 Ma (Table 1). We prefer this latter age estimate as the best that can be derived from the results on the assumption that pumice clast 08-034A is from an older eruption.
Ages were measured previously on the Tulu Bor Tuff (= Tuff B) by Walter & Aronson (1993) and McDougall & Brown (2008). A very precise estimate of 3.41 ± 0.01 Ma for the tuff was obtained by orbital tuning in the core at Ocean Drilling Program Site 722 in the Arabian Sea (deMenocal & Brown 1999). Our age of 3.412 ± 0.033 Ma on Tuff B-delta is regarded as the most precise direct isotopic age measurement currently available for the top of the Tuff B Complex (= Tulu Bor Tuff Complex).

Tuff C-gamma
In the type area for the Shungura Formation, Tuff C is located at the base of Member C. Tuff C, present only locally in fluvial channels (de Heinzelin & Haersaerts 1983), correlates with the Hasuma Tuff in the Koobi Fora Formation . Tuff C-alpha fills a channel eroded into Member B. Above Tuff C-alpha, up to 1 m of sand and silt occurs, followed by Tuff C-gamma, which is up to 1.6 m thick. Rare rounded pumice clasts 5-10 mm across are present in Tuff C-gamma west of L1-S (= L372) in Sector 15. A collection of these pumice clasts was divided into two groups, and feldspar was separated from each group. Electron microprobe analyses indicate that these feldspars are plagioclase (calcic oligoclase), and that the glass of the pumice clasts differs in composition from that of the enclosing Tuff C-gamma, so that the clasts are probably the product of a different volcanic eruption.
The 40 Ar/ 39 Ar ages measured (Table 1) were imprecise because of the small 39 Ar ion beams, and the relatively low and variable proportion of radiogenic argon (6-60%). Nevertheless, data obtained from the feldspars of the two groups of pumice clasts are in agreement, but with uncertainties on the arithmetic mean ages exceeding 10%. The arithmetic mean age on 17 crystals is 3.074 ± 0.376 Ma, with a weighted mean age of 2.990 ± 0.093 Ma (Fig. 5f). Because of the differences in the precision of single analyses, it is perhaps more appropriate to use a weighted mean age in this case.
Tuff C (= Hasuma Tuff) is stratigraphically younger than the Ninikaa Tuff of the Koobi Fora Formation (3.066 ± 0.017 Ma) and younger than Tuff B-10 of the Shungura Formation (2.965 ± 0.014 Ma, McDougall & Brown 2008), but older than the Burgi Tuff (2.630 ± 0.017 Ma; see below). Thus the imprecise data from Tuff C-gamma, reflected in the >12% uncertainty in the arithmetic mean age, are consistent with the much more precise age measurements on alkali feldspars from tuffaceous beds below and above it.

Burgi Tuff
The Burgi Tuff lies at the base of the Burgi Member of the Koobi Fora Formation (Brown & Feibel 1986). Two new pumice clast sam-ples labelled 08-032A and B from the tuff in Area 202 each provided enough alkali feldspar for single crystal dating. up to 10 crystals were measured from each sample with concordant results (Table 1), yielding an arithmetic mean age of 2.642 ± 0.042 Ma. One analysis was excluded from each set on the 2σ criterion, and another analysis was excluded when the data were combined. The age agrees closely with previously measured 40 Ar/ 39 Ar ages on single crystals from four pumice clasts from the type Burgi Tuff in Area 207 and two other pumice clasts collected from near Bura Hasuma, together yielding an arithmetic mean age of 2.622 ± 0.027 Ma (McDougall & Brown 2008). An overall arithmetic mean age for the Burgi Tuff is 2.630 ± 0.017 Ma, calculated from the means of the six previous measurements and the two new results reported here. All these pumice clasts from the Burgi Tuff come from the southern part of the Koobi Fora region within about 10 km of one another. Correlations between the various outcrops were authenticated by the stratigraphy as well as by chemical analyses of volcanic glass from the tuff. There

Tuff G
Tuff G defines the base of Member G of the Shungura Formation (de Heinzelin & Haersaerts 1983). At the type locality Tuff G is up to 6 m thick and Member G in this area is about 220 m thick. Brown et al. (1985) previously measured K/Ar ages on two K-feldspar separates, probably from the same pumice, at locality P162, Sector 9, yielding concordant ages averaging 2.33 ± 0.03 Ma. Pumice clasts in Tuff G are very rare (de Heinzelin & Haersaerts 1983). Where Tuff G crops out as a cross-bedded unit about 5 m thick filling a channel cut into Member F at the western edge of the exposure, it has lenses with sparsely distributed rounded pumice clasts < 20 mm in diameter. These pumice clasts have visible K-feldspar and bipyramidal quartz phenocrysts. From a composite sample of the pumice clasts (08-029) a mixture of K-feldspar and quartz was separated and utilized for single crystal 40 Ar/ 39 Ar dating. Of the 14 crystals analysed, only seven were alkali feldspar, with the remainder being quartz. The seven alkali feldspar crystals yielded essentially concordant ages, with an arithmetic mean age of 2.271 ± 0.041 Ma (Fig. 3), and a weighted mean age of 2.262 ± 0.012 Ma (Fig. 5d).
This measured age is slightly younger than the previously determined K/Ar age, but is our preferred age for this unit. The new age is consistent with the age on unit G-3 (2.188 ± 0.036 Ma, McDougall & Brown 2008), which lies c. 20 m above the base of Member G of the Shungura Formation from a locality nearby at 5.089°N, 36.028°E.

Orange Tuff
The Orange Tuff was defined by Harris et al. (1988, p. 10) with the type locality in Area 130 of the Koobi Fora region (see Brown et al. 2006) near the northern end of the Karari Ridge, and it also occurs in the Nachukui Formation. The Orange Tuff lies between the KBS Tuff (1.869 ± 0.021 Ma;  at the base of the KBS Member of the Koobi Fora Formation (Brown & Feibel 1986) and the Morutot Tuff (1.607 ± 0.019 Ma; ) essentially at the top of the KBS Member. It is stratigraphically below the Morutot Tuff ) but its position relative to the two dated tuffs, above and below, is not clear. Despite these difficulties, an estimated age of about 1.64 Ma was assigned to the Orange Tuff, based upon linear interpolation between the dated units ).
The Orange Tuff has not been directly dated previously. Pumice clasts from the Orange Tuff in Area 109, south of the western Koobi Fora Ridge, have large, fresh alkali feldspar phenocrysts. Single crystal age measurements on feldspars from three pumice clasts (08-031A, B, C) gave concordant results, with an overall mean arithmetic age of 1.760 ± 0.026 Ma (Table 1). This directly determined age is consistent with the existing older and younger age constraints, but is closer in age to the KBS Tuff and the Malbe Tuff than had been previously estimated. The difference between the predicted age and the directly measured age emphasizes that linear stratigraphic scaling must be treated with caution, even though it may be the only way to provide estimates of age when no direct age determination is possible. Brown et al. (2006) tentatively correlated the Orange Tuff with Tuff j-2 of the Shungura Formation, on the basis of similarity of glass compositions, but analyses of new samples show that the correlation should be with Tuff j. The type section for Tuff j has been defined in the section west of P832, Sector 19, in the type area of the Shungura Formation by de Heinzelin & Haersaerts (1983). They described Tuff j as being 6.6 m thick, beginning with 2 m of tuffite (i.e. tuffaceous siltstone), overlain by 1.2 m of grey vitric tuff with two or three more or less distinct layers, followed by 3.4 m of tuffite showing indistinct stratification. Brown et al. (2006) estimated the age of Tuff j at 1.76 Ma.
Based on the low-iron principal mode, Rogers (2010) correlated the Orange Tuff with Tuff j of the Shungura Formation, and also with the Kayle Tuff-1 of the Konso Formation. Tuff j is of reversed polarity (Brown et al. 1978), lying about 5 m above the top of the Olduvai Subchron (C2n), to which both Horng et al. (2002) and Gradstein et al. (2004) (1988,1995), but concluded that the tuffs represented separate stratigraphic beds because of the low similarity coefficient (0.83), and Haileab's positioning of the Orange Tuff below a bioclastic sandstone believed to be C4 that lies within the upper part of the Olduvai Subchron. Rogers (2010) negated identification of the bioclastic sandstone as C4, hence removing this objection to correlation. Katoh et al. (2000) reported an age for the Kayle Tuff-2 (2.5 m above Kayle Tuff-1) of 1.72 ± 0.03 Ma, where the error is the standard deviation of the weighted mean, and using an age of 27.84 Ma for the Fish Canyon sanidine fluence monitor. Adjusting the monitor age to 28.10 Ma as used here, an age of 1.74 ± 0.03 Ma is calculated for Kayle Tuff-2, or 1.74 ± 0.06 Ma if the error is given as the standard deviation of the population, as used throughout this paper. Thus within error, the age measurements agree with placing the Kayle Tuff-2 above the Orange Tuff.

Tuff K
The type area of the upper part of the Shungura Formation lies in the Kalam region, 25 km SSW of the Shungura Formation type area, and 15 km west of the Omo River. Indeed, Member K of the Shungura Formation has its type section at Errum (de Heinzelin & Haersaerts 1983), where Tuff K forms a cuesta. Tuff K is divided into Tuff K-alpha and Tuff K-beta, with 1.2 m of siltstone between them. Glasses from the two tuffs are compositionally very similar , so that the tuffs are probably products of the same eruption or closely related eruptions. Brown et al. (2006) showed that Tuff K is at a similar stratigraphic level to the Lower Koobi Fora Tuff in the Koobi Fora region.
Small pumice clasts, rarely up to 15 mm across, are found sporadically in Tuff K at Errum and a composite sample was collected from the two tuffs and in lenses in the intervening silty sediment.
Single crystal 40 Ar/ 39 Ar measurements were made on 10 feldspar crystals from this aggregate sample, yielding a concordant set of ages with an arithmetic mean of 1.526 ± 0.015 Ma (Table 1).  suggested a correlation with the Morte Tuff, but although the glass from the latter is similar in composition to that of Tuff K, it has significantly higher Al 2 O 3 and thus is unlikely to be a product of the same eruption. Nevertheless, the age obtained for Tuff K is very similar to ages given by  between 1.48 ± 0.01 Ma and 1.53 ± 0.01 Ma for the interval from the Lower Ileret Tuff to the Koobi Fora Tuff in the lower part of the Okote Member of the Koobi Fora Formation (Fig. 3). Thus, the suggested correlations based upon age are consistent with the stratigraphic and compositional correlations.

The Barrier
The Barrier, consisting essentially of volcanic rocks, named Barrier Volcanics by Ochieng' et al. (1988), separates Lake Turkana from the Suguta Valley to the south. It rises more than 600 m above Lake Turkana. Many of the volcanic units appear to be fairly youthful from the well-preserved and little eroded constructional landforms. Dodson (1963) and Ochieng' et al. (1988) stated that the Barrier consists of Quaternary volcanic extrusive units with volcanic activity continuing to nearly the present day. Dunkley et al. (1993) described the Barrier in terms of four overlapping volcanic centres, Likayu East, Likayu West, Kakorinya, and Kalolenyang from east to west. They reported 40 Ar/ 39 Ar ages of 1.37 ± 0.02 Ma and 1.34 ± 0.01 Ma for Likayu East, and 0.773 ± 0.007 Ma and 0.707 ± 0.006 Ma for trachytes of Kalolenyang. The youngest volcanic centre is Kakorinya, for which Dunkley et al. (1993) reported 40 Ar/ 39 Ar ages of 0.221 ± 0.004 Ma on a feldspar from pumice in the west wall of Kakorinya, 0.097 ± 0.003 Ma for the uppermost trachyte in the SE wall of the caldera, and 0.092 ± 0.002 Ma on pumice lapilli tuffs west of the caldera. The youngest age 40 Ar/ 39 Ar age reported on Kakorinya, 0.058 ± 0.004 Ma, is from a trachyte dome near the south wall of the caldera.
We obtained 40 Ar/ 39 Ar ages on alkali feldspar crystals from two additional rocks from the Barrier, a trachyte flow on the east flank of Kakorinya (K06-453; at 2.3213°N, 36.6098°E), and a dark trachyte from Likayu West (08-035; located at 2.3179°N, 36.6203°E). The arithmetic mean age derived from the Kakorinya sample from 12 single alkali feldspar crystals is 0.100 ± 0.015 Ma with a weighted mean age of 0.104 ± 0.009 Ma, and is in good agreement with the results of Dunkley et al. (1993). Results on 10 single alkali feldspar crystals from the trachyte of Likayu West form a concordant set giving an arithmetic mean age of 0.699 ± 0.053 Ma or a weighted mean age of 0.694 ± 0.009 Ma (Table 1)

New hominin age estimates
Even with the numerous new age measurements on volcanic ash layers from the Omo Group deposits since the study by Feibel et al. (1989) was published, their estimates of the ages of hominin specimens are generally still correct. We re-estimated ages for most specimens on the basis of our new information, and below we suggest new age estimates for crania, mandibles, and postcrania that  Feibel et al. (1989)  differ by more than 100 ka from the previous estimates. Generally, we do not discuss dental remains, but note that correlation of the tuff in unit u-11 of the usno Formation by Haileab (1995) with Tuff B-beta of the Shungura Formation confirms identification of the magnetozone in u-12 to u-14 as the Mammoth Subchron. Thus the 22 hominin teeth collected from the usno Formation, in addition to specimen O.20-4-1970-272 from submember B-2 of the Shungura Formation, should be assigned an age of 3.27 ± 0.06 Ma, rather than 3.05 ± 0.05 Ma as reported by Feibel et al. (1989).
All but three fossils of cranial, mandibular, and postcranial materials for which age corrections are ≥0.1 Ma are from the Koobi Fora Formation (Table 2). The outlying specimens are from the Nachukui Formation. Feibel et al. (1989) suggested an age of 1.70 Ma for specimens from Area 6A that were collected from a level 8 m below the Lower Ileret Tuff (1.527 ± 0.014 Ma; . Gathogo & Brown (2006) placed the Lower Ileret Tuff 82 m above the KBS Tuff, and an age of 1.56 Ma is found for a level 8 m below the Lower Ileret Tuff by linear scaling between these two endpoints. The Steel-grey Tuff lies 52 m above the KBS Tuff in the Ileret region between the Orange Tuff and the KBS Tuff, thus having an age between 1.76 and 1.87 Ma. Assigning the Steel-grey Tuff a maximum age of 1.87 Ma leads to an age estimate of 1.62 Ma for a level 8 m below the Lower Ileret Tuff. Thus an age of 1.59 ± 0.05 Ma for all these specimens is appropriate.
Six hominin specimens from Ileret for which we provide revised ages lie within the Ileret Tuff Complex near 1.50 ± 0.04 Ma (Table  2), and four others lie between the Ileret Tuff Complex and the Chari Tuff. Estimated ages range from 1.41 to 1.45 Ma for this latter group (Table 2). Another specimen, KNM-ER 807, lies 1.9 m below a lenticular tuff that is compositionally indistinguishable from the Lower Okote Tuff, and thus has an age of c. 1.6 Ma (see Brown et al. 2006).
Near Koobi Fora, specimens from Area 103 are related to either the Lower Koobi Fora Tuff or algal biolithites A6 or A7 (Table 2). Lepre et al. (2007)   . From other ages on tuffs in the Koobi Fora Tuff Complex, and stratigraphic arguments, we suggest that the base of the Koobi Fora Tuff Complex in Area 103 cannot be less than 1.51 Ma in age and is probably closer to 1.54 Ma, and therefore we estimate an age of 1.59 Ma for A6. The maximum range in age estimates for the interval below the Koobi Fora Tuff Complex from which hominins have been recovered is 1.54-1.61 Ma (Table 2), about 0.1-0.15 Ma younger than previously thought. Assuming that correlation of A6 between Areas 103, 121, and 123 is correct, then KNM-ER 1506 (Area 121) and KNM-ER 1821 (Area 123) should also lie within this temporal interval. Two other specimens from Area 103 (KNM-ER 1807 and 3892) derive from 21 and 5 m above algal biolithite A7, respectively. A7 lies c. 1 m above the Koobi Fora Tuff Complex. These two specimens have estimated ages of 1.42 and 1.43 ± 0.04 Ma, respectively, older than the Chari Tuff, but younger than the youngest age measured on a tuff in the Koobi Fora Tuff Complex. Other important hominin specimens from Area 123, including the relatively complete cranium KNM-ER 1813, are not discussed here, as debate continues as to the stratigraphy of that area. KNM-ER 3733, which Feibel et al. (1989) placed 6 m above algal biolithite A2, also deserves comment. Lepre et al. (2007) suggested an age of 1.748 Ma for A2, but Lepre & Kent (2010) provided new control through identification of the top of the Olduvai Subchron 20 m below it in Area 102, giving a new estimate of its age as 1.71 Ma. We derive ages for A2 of 1.71-1.73 Ma on the basis of simple linear scaling. KNM-ER 3733 is an important hominin cranium, assigned by Wood (1991) to Homo with affinities to H. erectus. It is regarded as one of the earliest fossils of this form so far known in East Africa.
The monument marking the find spot of KNM-ER 3733 still exists, and is located at 3.9639°N, 36.3070°E. About 230 m to the SE Tindall (1985) mapped a tuff (T9b) that he correlated with the White Tuff (see Brown et al. 2006) and showed that it lies 33-35 m above algal biolithite A2. At the top of the section (59 m) Tindall recorded another tuff (T10), which correlates with the Morte Tuff (1.510 ± 0.016 Ma, McDougall & Brown 2006). The White Tuff is underlain by a coarse, pale yellowish grey sandstone and an olive mudstone, the base of which can be traced laterally to the site of KNM-ER 3733. The base of the mudstone is 7.4 m below the White Tuff, whereas the monument marking 3733 is 5.9 m above the same level, so 3733 lies c. 1.5 m below the base of the White Tuff. This level is 31-33 m above algal biolithite A2, rather than 6 m as reported by Feibel et al. (1989). If we accept the identification of A2 in Area 104 and its age as being near 1.71 Ma (Lepre & Kent 2010), and further, identification of the Morte Tuff in Area 104 (analyses have been given by Tindall (1985)), then an age is derived for the White Tuff of 1.60 Ma. Within error, this is consistent with its placement c. 5 m below the Morutot Tuff (1.607 ± 0.019 Ma) in Area 131, and also with its reversed magnetic polarity. Lepre & Kent (2010) suggested revised limits of 1.78-1.48 Ma for the age of KNM-ER 3733, but its position below the White Tuff (c. 1.60 Ma) and above algal biolithite A2 provides closer constraints: 1.60-1.71 Ma or, as a single number, 1.65 ± 0.05 Ma. The more recent study by joordens et al. (2011) does not materially change this assessment.
Two specimens from the Karari escarpment, KNM-ER 1805 and 1806, also deserve mention. KNM-ER 1805 is an incomplete skull that Wood (1991, p. 85) assigned to an indeterminate species of Homo, whereas KNM-ER 1806 consists of robust mandible frag- ments assigned by Wood (1991, p. 190) to Australopithecus boisei. Both these important fossils are near the level of the Orange Tuff at archaeological site Fxjj 38 (Isaac & Behrensmeyer 1997, p. 20). Thus, the Orange Tuff (1.76 Ma) is probably closer to the age of these specimens, perhaps about 1.75 Ma, than the estimate of 1.85 Ma given by Feibel et al. (1989). Three important specimens lie in the upper Burgi Member in Area 131: KNM-ER 1470, 1472, and 1474. Hillhouse et al. (1977 showed that in this section normal palaeomagnetic polarity extends to c. 15 m below the KBS Tuff, followed by reversed polarity to c. 35 m below the tuff, and then normal polarity to the base of the section, which they gave as c. 42 m below the tuff (White (1976) gave the base as 41.5 m). The transition at c. 15 m below the KBS Tuff almost certainly represents the bottom of the Olduvai Subchron at 1.945 Ma (Gradstein et al. 2004), whereas the section with normal polarity still lower may either be spurious or it may represent a Reunion Subchron (Kidane et al. 2007) or the Huckleberry Ridge Subchron (Lanphere et al. 2002). KNM-ER 1472 lies in the reversed polarity section below the base of the Olduvai Subchron, whereas specimens 1474 and 1470 lie in the normal polarity section below the reversed interval. Thus all of these specimens are older than 1.945 Ma. Fine-grained strata near the base of the section in Area 131 reflect deposition above the disconformity that separates the lower Burgi from the upper Burgi Member. Elsewhere in the basin the Kangaki Tuff (2.06 ± 0.03 Ma;  predates deposition of these strata. Thus specimens 1474 and 1470 are younger than 2.06 Ma, but older than 1.945 Ma, and with current information, the best estimate of their age is between 2.01 and 2.05 ± 0.05 Ma (Table 2). Ages reported in Table 2 are computed assuming that the base of the section is 2.06 Ma old. KNM-ER 1470 is a very significant cranium, although Wood (1991, p. 76) could only assign it to Homo sp. indet.
Ages of two important specimens from the Nachukui Formation also require revision: KNM-WT 15000 and 16001. These lie <1 and 10 m above the Lower Koobi Fora Tuff, respectively, but below the Lokapetamoi Tuff (see Brown et al. 2006), giving age limits between about 1.48 and 1.42 Ma. The skeleton (15000), commonly assigned to early Homo erectus (Walker & Leakey 1993), is the older of the two, with a probable age of 1.47 ± 0.03 Ma, whereas the cranium (16001) has an estimated age of 1.44 ± 0.03 Ma. Another cranial fragment, KNM-WT 15001, lies less than 5 m below the Kangaki Tuff (2.06 Ma) of the Nachukui Formation, and hence is probably only slightly older than that, perhaps 2.09 ± 0.05 Ma. Although the assigned age changes little from that given by Feibel et al. (1989), the error estimate is much smaller.

Temporal changes in volcanic activity in source areas
Tuffs in the Omo Group provide a record of regional explosive volcanism for most of the past 4.3 Ma, with a significant gap between c. 0.7 and 0.2 Ma (Fig. 6). Brown (1972) and Martz & Brown (1981) suggested that source volcanoes for many of the tuffs lay in the Ethiopian highlands or on its flanks, but Feibel (1999) noted that some tuffs in the southern part of the basin may derive from sources further south in the Kenya Rift. Histograms showing the age distribution of tuffs (Martz & Brown 1981;Haileab 1995;Feibel 1999) had box widths of 0.2 or 0.25 Ma. using an updated catalogue of 360 tuffs, combined with new geochronological results (this study; , 2008, a frequency histogram was constructed with age resolution of 0.1 Ma (Fig. 6). All tuffs noted by de Heinzelin (1983) in the Shungura Formation are included, as are those reported by Feibel (1999) at Kanapoi and Lothagam, those documented in the Kibish, Nkalabong, and Mursi formations by Brown & Fuller (2007), and many unpublished tephra from the Nachukui and Koobi Fora formations. Pumice clasts that differ in composition from the tuff in which they occur, and that are not obviously related to the composition of the host tuff, were counted as recording separate eruptions.
The oldest tuffs, presumably from sources farther south in the Rift Valley of Kenya (Leakey et al. 1998;Feibel 2003), are from Kanapoi and Lothagam. Most of the remaining tuffs are thought to be from sources in Ethiopia, so that the number of tuffs preserved in the Omo Group should reflect the frequency of explosive volcanic eruptions in that region. Working only with the record from the Shungura Formation, Martz & Brown (1981) noted that tuffs were not uniformly distributed through time, and that the interval from 1.5 to 2.75 Ma contained the largest number of tuffs. With information from the Shungura, Nachukui and Koobi Fora formations, Haileab (1995) suggested three peaks of eruptive activity centred at c. 3.3, 2.5, and 1.5 Ma, supposing that these might reflect successive growth of different volcanic centres in Ethiopia. Feibel (1999) made the same observation using the same data, and he suggested that the three peaks corresponded to cycles of increasing eruptive activity each of which 'culminated in a major blow-out and a drop in activity'. Whether he envisaged a single volcanic centre or several is not known. With finer temporal resolution (Fig.  6) and inclusion of more data, it appears that earlier notions were oversimplified. The pronounced peak at 1.5 Ma is more sharply defined than before, and has a duration of about 300 ka. The reality of this peak is bolstered by the record from Konso in southern Ethiopia (WoldeGabriel et al. 2005; see Fig. 6). The peak is reasonably symmetrical, and does not suggest that a single 'blow-out' was involved in the reduction in activity. Such skewness as exists is perhaps more readily explained by the smaller area over which younger strata are known. Indeed, the sequence at Konso appears to record slightly younger times. The peak at 2.3 Ma remains, and it too is reasonably symmetrical, but the new compilation reveals a period of relatively high eruptive activity at 1.9 Ma. For the older part of the section, the single peak in frequency reported previously is no longer apparent; instead, smaller peaks lie at 3 and 3.4 Ma, both of which are symmetrical. In the much younger (< 200 ka) Kibish and Galana Boi formations there is also a pronounced peak, reflecting eruptions nearly as frequent as those recorded at 1.5 Ma, a strong reminder that the availability of exposed, well-dated sections strongly affects our perceptions of volcanic history in the region.

Summary and conclusions
An important consideration is that the present geochronological study was undertaken at the university of Queensland (uQ), whereas the previous measurements by , 2008 were made at the ANu. The new results and the earlier results are compatible, indicating that the two laboratories produce comparable and equally good and precise age measurements. This is shown through virtually all the new measurements fitting with our previously proposed time frame well, but more particularly through the agreement within error of the previous ages for the Burgi Tuff and new measurements on two additional pumice clasts from the same bed. Our confidence in the results from the two laboratories is confirmed in other cases as well, and these have been highlighted above.
As shown in Figure 3, the stratigraphy of the principal formations defined in the large Omo-Turkana Basin is well correlated through the tuffaceous beds within the sequences. The stratigraphic sections shown synthesize even more local and restricted sections and are an attempt to provide an overall synthesis of the stratigraphy, recognizing that marked facies variations are common even within a small area. The schematic sections illustrate the significance of the tuffs in correlating from one area to another in the basin, so that they are really the foundation of virtually all stratigraphic studies within the basin. They also facilitate correlations to other parts of East Africa and even into the deep sea off East Africa and in the Gulf of Aden (see Sarna-Wojcicki et al. 1985). Likewise the tuff beds, and especially those pumice clasts that may contain alkali feldspar phenocrysts, are the major source of material for the single crystal 40 Ar/ 39 Ar dating studies. This dating has allowed a comprehensive time framework to be developed, not only for the explosive eruptions that produced the volcanic material but also for their deposition , 2008. Our new data fill in significant time gaps in the previous time scale, with results that are consistent with earlier work. Our results also allow the age of vertebrate fossils, recovered from the sediments, to be placed in a time framework, independently of assumptions as to their evolutionary origins. Revision to the assigned ages for hominin fossils is provided where these exceed 100 ka; noteworthy in particular is the age of KNM-ER 3733, reassessed as 1.65 ± 0.05 Ma.
Sedimentation began in the Omo-Turkana Basin about 4.24 Ma ago, as shown by measurements previously made at Kanapoi and at Lothagam, SW of Lake Turkana ( Fig. 3 and McDougall & Brown 2008). In the present study, we have dated the Naibar Tuff at 4.023 ± 0.038 Ma, consistent with its stratigraphic position just below the Topernawi and Moiti tuffs in the Koobi Fora region. Similarly, the ages on alkali feldspar from pumice clasts in Tuff B-delta in the Shungura Formation (Fig. 3)  Plagioclase crystals from pumice clasts in Tuff C-gamma yielded an imprecise arithmetic mean age of 3.07 ± 0.38 Ma, or a weighted mean age of 2.99 ± 0.09 Ma. These results are consistent within error with previous higher precision data, notably the age of the underlying Tuff B-10 at 2.97 ± 0.01 Ma (McDougall & Brown 2008).
Our new results for the Burgi Tuff gave a pooled age of 2.642 ± 0.042 Ma, consistent with the age measured on other pumice clasts of Burgi Tuff from elsewhere in the Koobi Fora region of 2.622 ± 0.027 Ma (McDougall & Brown 2008). The age of 2.271 ± 0.041 Ma for Tuff G at the base of Member G of the Shungura Formation (Fig. 3) is consistent with its position between Tuff F (2.324 ± 0.020 Ma, McDougall & Brown 2008), and Tuff G-3 (2.188 ± 0.036 Ma; McDougall & Brown 2008).
Dating of the Orange Tuff in the Koobi Fora Formation is particularly helpful, as previously this part of the section had no direct age control other than the age of the Malbe Tuff below (1.843 ± 0.023 Ma) and the Morutot Tuff above (1.607 ± 0.019 Ma) (see . The pooled age of 1.760 ± 0.026 Ma determined here is consistent with the stratigraphic order and considerably improves age control on this part of the section.
Analysis of the frequency of tuffs in the sequences with time indicates a pronounced peak of activity at c. 1.5 Ma, with other less pronounced peaks of explosive volcanic activity also identified at 2.3, 3.0 and 3.4 Ma.