Geochronology of the pre-KBS Tuff sequence, Omo Group, Turkana Basin

The Omo–Turkana Basin of northern Kenya and southern Ethiopia developed in the northern Kenya Rift about 4.3 Ma ago in the Early Pliocene. Nearly 800 m of sediments, included within the Omo Group, crop out in the basin. Numerous rhyolitic tuffs in the sequence not only have facilitated secure correlations between the formations of the Omo Group but also have provided material for precise 40Ar/39Ar age measurements. Here we report ages on alkali feldspar crystals from pumice clasts within tuffs in the lower part of the Omo Group up to the level of the KBS Tuff, which has previously been dated at 1.87 ± 0.02 Ma. The results from 17 stratigraphic levels encompassing the 2.4 Ma time interval from the base of the group to the KBS Tuff provide a numerical time framework for the geological history of the lower part of the Pliocene sequence. The new ages, which have a precision of the order of 1%, are all consistent with the stratigraphic order, providing confidence that they accurately record the ages of the volcanic eruptions, with deposition of the tuffs and pumices occurring shortly thereafter.

In northern Kenya and southern Ethiopia the Omo-Turkana Basin, commonly referred to as the Turkana Basin, developed in the early Pliocene within the northern segment of the Kenya Rift ( Fig. 1). Up to 800 m of Pliocene-Pleistocene sediments, generally dipping less than 108W, were defined as the Omo Group by de Heinzelin (1983). Numerous vertebrate fossils, including hominins, have been recovered from the sequence and many archaeological sites are known. Dated minerals from rhyolitic tuffs in the succession are the basis for a numerical time scale for the basin.  presented 40 Ar/ 39 Ar age data for levels above the KBS Tuff (1.869 AE 0.021 Ma), only slightly older than the Pliocene-Pleistocene boundary (1.81 Ma; Gradstein et al. 2004). In this paper we provide comprehensive age data for the Pliocene lower part of the Omo Group. The resulting numerical time framework for this interval constrains the geological and palaeoclimatic history of the Turkana Basin, allows better age estimates for important vertebrate fossils (including hominins), and adds information bearing on the geomagnetic polarity time scale. As some tuffs within the Omo Group occur elsewhere in NE Africa, including in the Gulf of Aden and the Arabian Sea, the results are of wider application than the Omo-Turkana Basin itself.

Geological setting
The Omo-Turkana Basin is up to 500 km in length (northsouth) and as much as 100 km wide, with Lake Turkana occupying a significant fraction of the area (Fig. 1). Much current extension across the rift is associated with Lake Turkana, the focus of subsidence, where deposition is occurring at present (Dunkelman et al. 1989). The Omo River, draining the Ethiopian highlands, provides most of the detrital material to the northern half of the basin, and the Kerio and Turkwel rivers supply much sediment to the southern part of the basin. Seismic data show that the Plio-Pleistocene sequence beneath Lake Turkana is up to 4 km thick (Morley et al. 1999a, b).
The Omo Group consists of several formations defined in different parts of the basin. de Heinzelin (1983) mapped the Shungura Formation (766 m thick) in southern Ethiopia, where it dips c. 108W, and consists of the Basal Member, followed upward by Members A-L. Except for the Basal Member, a rhyolitic tuff lies at the base of each member. Thus, Tuff A is at the base of Member A, and successive units in the same member are labelled numerically as are the tuffs. In Kenya, east of Lake Turkana, the Koobi Fora Formation (560 m thick) as now defined (Brown & Feibel 1986) is divided into members, also with a tuff at the base of each, except for the basal Lonyumun Member, which lies on basaltic lavas. West of Lake Turkana a similar sequence of sediments is defined as the Nachukui Formation (730 m thick; Harris et al. 1988a, b), and it is also divided into members on the basis of rhyolitic tuffs. These formations consist dominantly of sands, silts and clays, deposited in a range of environments from fluvial, deltaic to lacustrine, with marked facies variation, both locally and regionally.
Some Omo Group strata, such as the Kanapoi and Nachukui formations, lie disconformably above Miocene volcanic rocks. Elsewhere Omo Group sediments were deposited on basalts of the Gombe Group (Watkins 1983;Haileab et al. 2004), but locally deposition began shortly before eruption of these basalts. Whole-rock K/Ar ages on Gombe Group basalts commonly yield ages of about 4 Ma. At Lothagam, the basal Nachukui Formation, represented by the Apak Member, disconformably overlies fluvial sediments of the Nawata Formation from which ages ranging from 7.4 AE 0.1 to 6.5 AE 0.1 Ma were reported previously (McDougall & Feibel 1999;Feibel 2003a).
Geochemical characterization of glass from the stratigraphically controlled tuffs was critical to working out the stratigraphy, and to demonstrating that the separate formations are all part of a single large depositional system, the Omo-Turkana Basin. Some tuffs are identifiable throughout the basin, whereas others are of much more restricted occurrence (Cerling et al. 1979;Brown et al. 2006). The source of many tuffs is probably within the region of Ethiopia drained by the Omo River, as evidenced by the presence of pumice clasts, which must have been transported by water. Volcanic ash blanketing the landscape around an explosive eruption vent would be rapidly washed into the drainage system and subsequently deposited, prior to the resumption of normal detrital sedimentation. Figure 2 shows schematically the stratigraphy of five constituent formations of the Omo Group. Important tuffs, either dated or successfully correlated between regions, are shown, but numerous other tuffs have been omitted for clarity.

Geochronological methods
Ages were measured by the 40 Ar/ 39 Ar technique mainly on single crystals of alkali feldspar separated from pumice clasts collected from tuffs in the Omo Group. However, where crystals were small (,0.5 mm), multiple crystals (2-6) were fused in a single analysis to have sufficient gas for precise measurement. Where glass from pumice clasts was chemically analysed, commonly the composition matched that of the enclosing tuff, demonstrating that the pumices are the products of the same eruption. Nevertheless, this was not always the case, so that the possibility of reworking must be considered. Each age provides a good estimate of the time elapsed since eruption, and we argue that deposition usually occurred soon after eruption, probably within tens to a few hundred years, so that the measured age closely approximates the actual depositional age.
The methods utilized in this work are closely comparable with those outlined by . Following fast neutron irradiation, single crystals were fused in an ultrahigh-vacuum system using an argon-ion continuous wave laser beam. The gases released were purified and the argon was analysed isotopically with a VG3600 mass spectrometer. Six to 22 feldspar crystals were analysed from each pumice clast, averaging about 10 per pumice. An arithmetic mean age and standard deviation of the population was calculated from results on each pumice, as the uncertainty of each analysis was comparable. We removed outliers by omitting any result more than 2 SD away from the mean, repeating the process until there were no further rejections. As the concordancy of results in a given dataset generally was high, the number of rejected analyses was low, averaging less than one in 15.
Results are summarized in Table 1, which lists the arithmetic mean age for feldspars from each sample after omission of outliers, together with the weighted mean age, and an isochron age derived from a 36 Ar/ 40 Ar v. 39 Ar/ 40 Ar correlation plot. The weighted mean age was calculated from the age results, with weighting by the inverse of the variance, disregarding the error in J, the irradiation parameter. The derived uncertainty was then combined quadratically with the estimated error in J to yield the standard error given in Table 1 and Figure 3. Throughout the text the arithmetic mean age is used with the uncertainty quoted being that of the standard deviation of the population. We believe that this conservative approach is more appropriate than using the weighted mean age and its standard error (see . It should be noted that the arithmetic mean age and weighted mean age in all cases agree to within the errors. Indeed, the average difference between the two mean ages for a given set of data averages less than 0.2%. It is also stressed that the probability plots are prepared using only the analytical uncertainty for each 40 Ar*/ 39 Ar K ratio calculated from the data, as listed in the Supplementary Publication (see below). Thus, the uncertainty of J for each irradiation is not included when the data are used to produce a probability plot. The isochron age is generally not used because many results plot close to the 39 Ar/ 40 Ar axis, yielding rather poor estimates of the trapped 40 Ar/ 36 Ar ratio. Nevertheless, there is generally very good agreement between ages derived using different approaches (Table 1). Three step heating experiments on bulk separates of feldspar were made, using a resistance furnace with excellent temperature control, again following the procedures of .
After Gradstein et al. (2004), the age of the Pliocene-Pleistocene boundary is taken as 1.81 Ma, and the Pliocene-Miocene boundary as 5.33 Ma. From McDougall et al. (1992), Cande & Kent (1995) and Lourens et al. (2004), the Gilbert-Gauss Chron boundary is estimated to lie between 3.57 and 3.596 Ma; the older boundary of the Mammoth Reverse Subchron, 3.29 to 3.33 Ma; the older and younger boundaries of the Kaena Reverse Subchron, 3.09-3.12 Ma and 3.02-3.04 Ma, respec-tively; the Gauss-Matuyama Chron boundary, 2.58-2.60 Ma; and the older boundary of the Olduvai Normal Subchron at 1.94 Ma. Table 1 are results of nearly 600 analyses in 57 groups on alkali feldspar crystals; analytical data and maps showing sample localities are available online at http:// www.geolsoc.org.uk/SUP18295. A hard copy can be obtained from the Society Library. In most cases, feldspars from more than one pumice clast from each tuff were analysed to check for consistency. Where the mean ages from the clasts agree well, an overall arithmetic mean age is calculated, as well as a weighted mean age (Table 1). Feldspar total fusion ages are shown against a schematic stratigraphy in Figure 2. Summary results of the step heating experiments are given in Table 2, with full analytical data given in the Supplementary Publication.

Summarized in
Here we discuss, from oldest to youngest, dating of 17 levels in the Omo Group that lie stratigraphically below the KBS Tuff. Analyses, including results from single crystals, are reported from 11 stratigraphic levels for the first time; four of these levels had not previously been dated, and two levels are shown to have yielded anomalous K/Ar ages. Ages given earlier for six stratigraphic levels from the lower part of the Omo Group have been adjusted upward by 0.7% to the reference age of 28.1 Ma used herein for the Fish Canyon Tuff sanidine fluence monitor for consistency with the results of . Schematic stratigraphic columns for the formations discussed in the text. Some of the named tuff beds are indicated together with their 40 Ar/ 39 Ar ages determined on alkali feldspars from pumice clasts from within the tuff or feldspar crystals collected directly from the tuff. Arithmetic mean ages are given, and the uncertainty quoted is the standard deviation of the population (see Table 1). Correlations of tuffs between sections are shown. The Shungura Formation composite section is after de Heinzelin (1983), the Koobi Fora Formation is essentially that provided by Brown & Feibel (1986), noting a significant hiatus, and that for the Nachukui Formation is derived from Harris et al. (1988a, b) with minor modification. The sequence at Lothagam is from information summarized by Feibel (2003a), and that at Kanapoi is after Leakey et al. (1998) andFeibel (2003b).  Feibel 1999). Although the peak in the probability plot of these results (Fig. 3) tails to both higher and lower ages, there are no outliers on the 2 SD criterion. Thus deposition began no later than c. 4.3 Ma ago, which is currently our best estimate for inception of the integrated basin. Higher in the sequence, above the Lothagam Basalt, lacustrine claystones of the Muruongori Member of the Nachukui Formation (Fig. 2) probably correlate with the Lonyumun Member of the Koobi Fora Formation (Brown & Feibel 1986;Feibel 2003a).

Kanapoi
The c. 60 m thick Kanapoi Formation ( Fig. 1; Leakey et al. 1998;Feibel 2003b), disconformably overlies Miocene basaltic volcanic rocks. Low in the formation two siltstones, stratigraphi-  Fig. 2), originally published by Leakey et al. (1995). The probability plot for the nine feldspar results from the lower pumiceous unit (95-37) is distinctly bimodal (Fig. 3), with the age peaks about 1.5% apart. If the three ages constituting the younger age peak were accepted as best representing the eruption age, and the others regarded as derived from a slightly earlier eruption, the age of 95-37 would decrease to 4.156 AE 0.020 Ma. Ages from the upper pumiceous siltstone provide a nearly ideal Gaussian probability plot (Fig. 3). Feldspars from four pumice clasts from the Kanapoi Tuff, which lies c. 15 m higher in the sequence, yielded a mean age of 4.108 AE 0.029 Ma (n ¼ 28) (Table 1, Fig. 2), updated from the results reported by . Again, the probability plot is almost ideally Gaussian (Fig. 3). The sequence is capped by the Kalokwanya Basalt, which has given K/Ar ages as old as 3.4 Ma , but its reversed magnetic polarity (Powers 1980) is inconsistent with the age. Feldspar ages from Kanapoi are consistent with the stratigraphic order, providing confidence that they closely reflect the eruption ages over an interval of about 0.1 Ma. We suggest that deposition occurred in each case shortly after eruption. The middle part of the Kanapoi sequence was deposited in a lacustrine environment (Feibel 2003b) in the same time interval that lacustrine sequences at Lothagam and Koobi Fora were deposited in the Lonyumun Lake Feibel 2003b). Compositionally, glass of the Kanapoi Tuff is distinct from that of other tuffs in the basin, and Feibel (1999) suggested that its source lies in the Kenya Rift to the south, rather than in Ethiopia. Following Namwamba (1993), he reported that the Kanapoi Tuff correlates with a tuff in the Baringo Basin. Leakey et al. (1995Leakey et al. ( , 1998) defined a new hominin species, Australopithecus anamensis, on the basis of fossils from the lower half of the sequence at Kanapoi. Most were found between the lower pumiceous siltstone and the Kanapoi Tuff, so that A. anamensis existed between 4.19 and 4.11 Ma ago. This hominin currently is among the earliest known that walked upright . Recently, White et al. (2006) assigned hominin fossils of similar age from Ethiopia to A. anamensis, providing support for the view of Leakey et al. (1995) that A. anamensis is ancestral to A. afarensis, having probably evolved from the older Ardipithecus ramidus.

Topernawi Tuff
Of 58 40 Ar/ 39 Ar single crystal age measurements on feldspars from seven pumice clasts from the Topernawi Tuff of the Nachukui Formation, none was rejected as a statistical outlier.
Mean ages for each pumice clast are remarkably concordant (Table 1), consistent with an origin from a single eruption. Pooled results give an overall mean age of 3.987 AE 0.025 Ma, superseding the mean K/Ar age of 3.81 AE 0.11 Ma calculated from data of Feibel et al. (1989), and indistinguishable from single crystal results given by Leakey et al. (2001). The probability plot has a single asymmetric peak that spreads to slightly younger ages (Fig. 3). Based on K/Ar age measurements the Topernawi Tuff was thought to be younger than the Moiti Tuff (Feibel et al. 1989). However, the mean 40 Ar/ 39 Ar age of the Topernawi Tuff is slightly older than that of the Moiti Tuff (see below), although their ages are essentially indistinguishable. This is consistent with the position of the Topernawi Tuff c. 5 m below the Moiti Tuff (Leakey et al. 2001).
Compositionally, glass from pumice 98-337A matches closely that of the glass from the enclosing Topernawi Tuff, whereas glasses from pumices 98-307A and B have significantly higher Al 2 O 3 and Fe 2 O 3 than the main compositional mode of the tuff. However, as ages on all analysed pumices agree so well, they probably result from essentially contemporaneous eruptions. This is supported by the close similarity of the average K/Ca ratios for the feldspars (Table 1).
A step heating experiment on a multigrained (15 mg) feldspar separate from pumice 93-1066 yielded a nearly flat age spectrum (Fig. 4) with a plateau age of 3.987 AE 0.015 Ma and an integrated total fusion age of 4.012 AE 0.056 Ma, agreeing with the plateau isochron age (Table 2). These ages agree closely with the single crystal age measurements. The age spectrum shows that since eruption there has been no significant reheating or disturbance of the feldspar K/Ar system.

Moiti Tuff
The Moiti Tuff marks the base of the Moiti Member and locally contains pumice clasts. In the upper part of the underlying Lonyumun Member, 5 m below the fluvially deposited type Moiti Tuff, an airfall tuff c. 1 m thick is compositionally indistinguishable from the fluvially deposited tuff. This suggests that lacustrine deposition continued until the time of eruption of the Moiti Tuff. As the Moiti Tuff is also found in deep-sea cores in the Gulf of Aden and the Somali Basin, more than 1600 km to the NE and 1200 km to the SE, respectively, it clearly resulted from a voluminous explosive eruption (Sarna-Wojcicki et al. 1985;Brown et al. 1992). McDougall (1985) regarded the mean K/Ar age of 4.10 AE 0.07 Ma on alkali feldspar from three pumices as a maximum because ages spread from 4.01 AE 0.04 to 4.15 AE 0.06 Ma, exceeding that expected for a single juvenile population. This raised the possibility of contamination with older feldspar. Although glass in the pumice clasts differs compositionally from that of the enclosing tuff, Brown et al. (1992) showed that the correlative tuff in the Gulf of Aden deepsea cores has some shards of similar composition, and concluded that the pumices and the tuff were derived from either the same eruption or a closely related one. Leakey et al. (1995Leakey et al. ( , 2001 reported single crystal 40 Ar/ 39 Ar age measurements on alkali feldspars separated from three pumice clasts from the Moiti Tuff, with an overall mean of 3.94 AE 0.03 Ma. After rejecting three outliers, and adjusting to the reference age of the fluence monitor used here, the arithmetic mean age is 3.970 AE 0.032 Ma (Table 1). The average age and K/Ca ratio for feldspars from each clast agree closely (Table 1), so a single eruptive event is being dated. The probability plot is roughly Gaussian with some tailing to younger ages (Fig. 3). The mean age for the Moiti Tuff feldspars is slightly younger than that of the underlying Topernawi Tuff in the Nachukui Formation, but the two pooled ages are indistinguishable.
Tuff VT-1 in the Wilti Dora area of the Middle Awash, Ethiopia, correlates geochemically with the Moiti Tuff (Hart et al. 1992;Brown et al. 1992   geochemically correlated the Lokochot Tuff of the Koobi Fora Formation (Fig. 2) with Tuff A, which lies c. 32 m above the base of the Shungura Formation (de Heinzelin 1983). On the basis of glass analyses, Brown & Fuller (2008) showed that the Lokochot (¼ A) Tuff also occurs in the Nkalabong Formation (Butzer 1971(Butzer , 1976 in the Kibish area (Fig. 1), and there it contains pumice clasts at a locality named Harpoon Hill. Thus, the Nkalabong Formation belongs to the Omo Group, and is at least partly equivalent to the Shungura Formation.

Lokochot Tuff
Mean ages on feldspars from three pumice clasts from Harpoon Hill range from 3.574 AE 0.024 to 3.618 AE 0.066 Ma, with an overall arithmetic mean age of 3.596 AE 0.045 Ma ( Table  1). The probability plot (Fig. 3) has a dominant peak but also some higher ages, especially two at c. 3.72 Ma from pumice 02-014. As neither age is an outlier on the 2ó criterion, they remain within the average. We view the mean calculated value of 3.596 Ma as the best estimate of the eruptive age of the Lokochot Tuff. This age agrees well with an orbitally tuned age of 3.57 AE 0.01 Ma (deMenocal & Brown 1999) estimated from Fig. 4. Age spectra for three alkali feldspar separates from pumice clasts within the identified tuff beds. Each box represents the age (AE 1 SD) measured in a single step of the step heating experiment. core at ODP Site 722 in the Arabian Sea. In addition, both Tuff A and the Lokochot Tuff have reverse polarity (Brown et al. 1978;Hillhouse et al. 1986) and lie stratigraphically just below the Gilbert-Gauss Chron boundary, which has an estimated age of 3.57 AE 0.05 Ma (McDougall et al. 1992).

Tulu Bor Tuff
The Tulu Bor Tuff is a pale grey tuff c. 5 m thick, commonly with cross-bedding, at the base of the Tulu Bor Member of the Koobi Fora Formation (Brown & Feibel 1986). It is a widely distributed tuff in the Omo Group , and correlates with Tuff B of the Shungura Formation, and U-10 of the Usno Formation .
Rounded pumice clasts (normally ,3 cm across) with small alkali feldspar phenocrysts occur sporadically in the Tulu Bor Tuff. To obtain sufficient material for analysis, feldspar crystals were combined from several pumice clasts. Two sets of analyses on composite feldspar separates from two lots of four small pumices from Area 207, Koobi Fora, yielded arithmetic mean ages of 3.469 AE 0.047 Ma (n ¼ 13) for 00-301A and 3.438 AE 0.023 Ma (n ¼ 17) for 00-301B (Table 1) with no outliers. Mean K/Ca ratios of feldspars from the two sets are in close agreement (Table 1), and the combined mean age from the two samples is 3.452 AE 0.038 Ma. The probability plot for 00-301B is nearly Gaussian, suggesting a population of juvenile crystals from a single eruption (Fig. 3). In contrast, the probability plot for sample 00-301A has multiple peaks indicative of feldspars of different ages, ranging from c. 3.54 to c. 3.38 Ma (Fig. 3). Feldspars from a composite sample of 14 small pumice clasts (00-294) from Area 261 provided a marginally older arithmetic mean age of 3.517 AE 0.048 Ma (n ¼ 17). The probability plot is bimodal (Fig. 3) with peaks at c. 3.51 and c. 3.57 Ma, suggestive of a mixed age population, but on the 2ó criterion no result can be rejected as an outlier. The mean age for this composite set of feldspars is slightly older than the age derived from the two sets of pumices in Area 207, further indicating that 00-294 was probably prepared from pumice clasts from two or more eruptions. The K/Ca average value for the feldspars from 00-294 is 2.0 AE 1.3, significantly different from the values of about nine on feldspars from the other two samples; this result indicates a different chemical composition and reinforces the view that different eruptions are being sampled.
As the depositional age can be no older than the age of the youngest pumices in the tuff, we suggest that the best estimate of its eruption age is 3.438 AE 0.023 Ma derived from sample 00-301B; this is a maximum age for tuff deposition. Even choosing the youngest ages from 00-301A and B the eruptive age can be no younger than 3.40 Ma, only c. 1% younger than the preferred estimate of age given above. The mean age of all three sets of feldspars is older at 3.475 AE 0.052 Ma, but the evidence is strong that material from slightly older eruptions is included, so that this overall mean is interpreted as an older limit for the main eruption.
The Tulu Bor Tuff is very widely distributed in NE Africa, as Brown (1982) demonstrated that the Sidi Hakoma Tuff in the Hadar Formation of Ethiopia is compositionally indistinguishable and thus very probably the same tuff. Glass shards of similar composition are present in core from DSDP Site 231 in the Gulf of Aden (Sarna-Wojcicki et al. 1985;Brown et al. 1992), and from ODP Sites 721 and 722 in the Arabian Sea (deMenocal & Brown 1999). Walter & Aronson (1993) confirmed the Tulu Bor Tuff-Sidi Hakoma Tuff correlation on the basis of further geochemical analyses. They also dated small (,0.4 mm) alkali feldspar crystals from the Sidi Hakoma Tuff by the 40 Ar/ 39 Ar technique. Four single crystals and four multiple (three) feldspar grains were analysed. Because ion beams were of small size, uncertainties for the analyses ranged from 1.4 to 10.7%. Walter & Aronson quoted a weighted mean age of 3.40 AE 0.03 Ma and an arithmetic mean age of 3.37 AE 0.11 Ma. These ages become 3.43 and 3.40 Ma, respectively, when adjusted to the same reference age as used in this study for the Fish Canyon Tuff sanidine fluence monitor. Thus, their age is indistinguishable from the more precise age derived in this study. The average K/Ca ratio for the feldspars from the Sidi Hakoma Tuff is 11.1 AE 2.6 (Walter & Aronson 1993), consistent with the values on feldspars from the Tulu Bor Tuff in Area 207, and further reinforces the correlation. An orbitally tuned age of 3.41 AE 0.01 Ma was derived by deMenocal & Brown (1999) for the Tulu Bor Tuff in the core from Site 722 in the Arabian Sea.
The Tulu Bor (¼ B) Tuff is normally magnetized (Brown et al. 1978;Hillhouse et al. 1986) and lies stratigraphically above the Gilbert Reverse to Gauss Normal Chron boundary. Our measured age of 3.44 Ma is consistent with its normal magnetic polarity.

Toroto Tuff
The Toroto Tuff, within the Tulu Bor Member of the Koobi Fora Formation, is exposed only in the southern part of the Koobi Fora region. It lies c. 5 m above the Tulu Bor Tuff and c. 10 m below the Allia Tuff Brown & Feibel 1986). The Toroto Tuff is grey, c. 2 m thick, and contains rounded pumice clasts up to 14 cm across. Anorthoclase crystals were separated from three pumice clasts, two from Area 204 and one from Area 207, several kilometres to the east. Single crystal 40 Ar/ 39 Ar ages from each pumice are concordant, as are average ages derived from each of the clasts, with no outliers (Table 1). The arithmetic mean age is 3.308 AE 0.022 Ma (n ¼ 24), and the probability plot is essentially Gaussian (Fig. 3). K/Ca values for the feldspars are rather scattered, but mean values for each pumice are indistinguishable from one another (Table 1). Concordance of the ages and the K/Ca values indicates that these pumices are probably products of a single eruption.
Earlier K/Ar measurements on multigrain feldspar separates from seven pumice clasts yielded a mean age of 3.32 AE 0.02 Ma (McDougall 1985).
Step heating experiments on two feldspar separates gave essentially flat age spectra with plateau ages of 3.32 AE 0.06 and 3.30 AE 0.06 Ma (McDougall 1985). Adjusting to the reference age for the fluence monitor used here we obtain plateau ages of 3.34 and 3.32 Ma for these samples. Our preferred best estimate of age is that derived from the single crystal analyses, 3.31 AE 0.02 Ma.
The Toroto Tuff has normal magnetic polarity (Hillhouse et al. 1986) and, together with the underlying Tulu Bor Tuff, is within the earliest subchron of the Gauss Normal Chron, just prior to the Mammoth Reverse Subchron, which has an estimated age of inception of 3.29-3.33 Ma (McDougall et al. 1992;Cande & Kent 1995;Lourens et al. 2004), so that the latter age estimate for the boundary seems marginally too old.

Ninikaa Tuff
The Ninikaa Tuff, defined by  with its type locality in Area 116, Koobi Fora, is also recognized in Areas 202 and 204 c. 55 m above the Tulu Bor Tuff (Brown & Feibel 1986) within the Tulu Bor Member, and c. 25 m below the base of the Burgi Member of the Koobi Fora Formation. In Area 116 the tuff is grey, c. 1.5 m thick, and has pumice clasts up to 20 cm across. Ages on feldspars from four pumices are all internally consistent, and average ages derived from each pumice also agree very well (Table 1). An overall arithmetic mean age of 3.066 AE 0.017 Ma (n ¼ 39) is accepted as a precise age for the eruption that produced the tuff. Despite the concordance of all the ages, average K/Ca ratios differ (Table 1). The probability plot is essentially Gaussian with a rather broad top (Fig. 3). A step heating experiment on a 15 mg feldspar separate from one pumice (93-1084A) yielded an almost ideal flat age spectrum with a plateau age of 3.059 AE 0.012 Ma, an integrated total fusion age of 3.076 AE 0.029 Ma (Fig. 4), and an isochron age of 3.040 AE 0.018 Ma (Table 2), in good agreement with the single crystal ages. McDougall (1985) also reported a nearly flat age spectrum for feldspar from a pumice from the Ninikaa Tuff with a plateau age of 3.01 AE 0.03 Ma and an incremental total fusion age of 3.02 AE 0.03 Ma. These become 3.03 and 3.04 Ma, respectively, when adjusted to the equivalent reference age utilized in this study for the fluence monitor. K/Ar ages on feldspars from nine pumice clasts from the Ninikaa Tuff in Area 116 averaged 3.11 AE 0.05 Ma (McDougall 1985). Thus, there is satisfactory agreement between earlier measured ages and the more precise ages presented here. Hillhouse et al. (1986) reported reverse polarity for the Ninikaa Tuff in Areas 202 and 204 but normal polarity at the type locality in Area 116. They suggested that the normal polarity might result from overprinting or because deposition occurred a little later owing to reworking. Our preferred eruption age of 3.066 AE 0.017 Ma fits within the Kaena Reverse Subchron in the Gauss Normal Chron.

Tuff B-10-1
Bed B-10-1, within Member B of the Shungura Formation (de Heinzelin 1983, p. 44), is a complex sandy deposit up to 5 m thick c. 70 m above the base of Member B and c. 20 m below the base of Member C. A minor altered tuff, c. 20 cm thick, at the base of the unit contains large (up to 4 mm), generally fresh, anorthoclase crystals. Single crystals (sample 03-039) gave ages ranging from 2.941 AE 0.020 to 3.086 AE 0.126 Ma. Four results were omitted as outliers, two of which had less than 20% radiogenic argon and large associated errors (4.1 and 5.3%). The final grouping of 11 out of 15 results yielded an arithmetic mean age of 2.965 AE 0.014 Ma (Table 1), with a nearly ideal probability plot (Fig. 3). Brown & Nash (1976) reported K/Ar ages on two feldspar separates from the tuff in B-10 that when recalculated to current decay constants (Steiger & Jäger 1977) are 3.01 AE 0.10 and 3.04 AE 0.10 Ma. Brown et al. (1985) measured K/Ar ages on two additional feldspar concentrates, obtaining 2.93 AE 0.03 and 2.98 AE 0.03 Ma. All these ages are essentially concordant and provide excellent control on the eruption age of this tuff. Brown et al. (1978) demonstrated that above unit B-3, this part of the section in the Shungura Formation has normal polarity, so B-10 lies in the Gauss Normal Chron above the younger boundary of the Kaena Reverse Subchron.

Burgi Tuff
The Burgi Tuff lies at the base of the Burgi Member of the Koobi Fora Formation, and is c. 1.6 m thick in its type locality in Area 207 at Koobi Fora, where it contains pumice clasts up to 20 cm across. There is a significant hiatus between the lower Burgi Member, exposed in the Allia Bay region, and the upper Burgi Member, which crops out farther north (Brown & Feibel 1986). The lower Burgi Member, 27 m thick in its type section, consists of several fluvial upward-fining cycles of sandstone grading to mudstone.
Feldspars from four pumices in the type area and from two additional pumices from a locality about 14 km to the NW were dated. Results from each of the pumices are essentially concordant (Table 1), with few outliers, yielding an overall mean age of 2.622 AE 0.027 Ma (n ¼ 77). The probability plot is roughly Gaussian with some skewing to older ages (Fig. 3). The similarity of the K/Ca ratios of the feldspars from each pumice supports derivation from a single eruption (Table 1). The new mean is slightly younger than the average K/Ar age of 2.68 AE 0.06 Ma reported by Feibel et al. (1989). The Burgi Tuff has normal polarity and Hillhouse et al. (1986) placed it within the Gauss Normal Chron c. 5 m below the boundary between the Gauss Normal and Matuyama Reverse Chrons. Harris et al. (1988a) named the Lokalalei Tuff, and used it to define the base of the Lokalalei Member of the Nachukui Formation. They correlated it with Tuff D of the Shungura Formation based upon similarity of glass compositions. A correlative tuff in Area 207 at Koobi Fora lies c. 20 m stratigraphically above the Burgi Tuff of the Koobi Fora Formation. At Lomekwi, west of Lake Turkana, a channel filled with Lokalalei Tuff contains abundant pumice clasts up to 15 cm across. Feldspars from four of these pumices and from a pumice in the correlative tuff in Area 207, Koobi Fora, were dated in three irradiations. Mean ages for the feldspars from each of the five pumices, listed in Table 1, are reasonably coherent. The calculated arithmetic mean age is 2.526 AE 0.025 Ma (n ¼ 35), with one analysis omitted as an outlier; the probability plot is broadly Gaussian (Fig. 3). A step heating experiment on a 20 mg sample of alkali feldspar from one pumice clast (93-1064A) yielded an essentially flat spectrum with a plateau age of 2.551 AE 0.011 Ma and an integrated total fusion age of 2.555 AE 0.024 Ma (Fig. 4), which agree with the plateau isochron age (Table 2). Previous K/Ar ages on bulk feldspar separates from six pumice clasts from Tuff D of the Shungura Formation averaged 2.52 AE 0.05 Ma . Thus, there is good agreement between K/Ar data from Tuff D of the Shungura Formation and the 40 Ar/ 39 Ar results from the Lokalalei Tuff in the Nachukui and Koobi Fora formations, with our preferred best age estimate being the pooled single crystal age of 2.526 AE 0.025 Ma.

Lokalalei (¼ D) Tuff
The K/Ca measurements on feldspars from pumice clasts in the Lokalalei Tuff have two distinct values. Three from Lomekwi range from 143 AE 94 to 285 AE 92 whereas a fourth is much lower at 20 AE 5, similar to that of 18 AE 13 in feldspars from pumice 83-284 from Area 207, Koobi Fora. As the ages are indistinguishable, the pumice clasts may reflect different phases of the same eruptive event. Brown & Nash (1976) previously reported a large range of K/Ca ratios in feldspars from Tuff D (¼ Lokalalei Tuff), consistent with the present findings.
Tuff D has reverse polarity and lies c. 10 m above the Gauss-Matuyama Chron boundary (Brown et al. 1978). Our preferred age for the Lokalalei (¼ D) Tuff of 2.526 AE 0.025 Ma is consistent with its reverse polarity.
Within or just a few metres below the Lokalalei Tuff at Lomekwi, a very robust hominin cranium, KNM-WT 17000, was described by Walker et al. (1986) and Leakey & Walker (1988), who assigned it to Australopithecus boisei. In both papers they recognized that Australopithecus aethiopicus might be an alternative name. Our best estimate of its age is 2.53 AE 0.05 Ma, indistinguishable from the age estimate of 2.50 AE 0.07 Ma given by Walker et al. (1986).

Tuff D-3-2
Tuff D-3-2, within Member D of the Shungura Formation, is a stratified yellowish tuff c. 4.5 m thick (de Heinzelin 1983). It contains loose crystals of anorthoclase, some of which were collected directly from the tuff. Eleven 40 Ar/ 39 Ar age measurements on sample 03-040 form a concordant dataset with a mean age of 2.443 AE 0.048 Ma (Table 1), and a somewhat asymmetric probability plot (Fig. 3). Member D is c. 37 m thick in its type section, and Tuff D-3-2 occurs c. 18 m above the base of the member (de Heinzelin 1983). As the Lokalalei (¼ D) Tuff has an age of 2.53 AE 0.03 Ma, the measured age of Tuff D-3-2 is consistent with its stratigraphic position and with its reverse polarity (Brown et al. 1978) within the Matuyama Reverse Chron. Our new age estimate supersedes the inconsistent K/Ar age results given by Brown & Lajoie (1971) and Brown et al. (1985).

Tuff F and Kalochoro Tuff
Tuff F, which defines the base of Member F of the Shungura Formation (de Heinzelin 1983), is commonly c. 4 m thick locally with pumice clasts whose glass composition matches that of the enclosing tuff. Feldspars from three pumice clasts from Tuff F were dated as single crystals using 10-16 crystals from each pumice (Table 1). Two older outliers from sample 01-088 were omitted, resulting in a mean age of 2.334 AE 0.029 Ma (n ¼ 8). The mean ages of feldspars from the two other pumice clasts agree to within the errors, so that the best estimate of age is given by the overall arithmetic mean of 2.324 AE 0.020 Ma (n ¼ 39; Table 1). The probability plot is only broadly Gaussian (Fig. 3), but has a clearly defined peak. Concordance of results from the three pumices gives confidence that they derive from a single volcanic episode, which is reinforced by the similarity of the K/Ca ratios averaging about 1280 (Table 1). Brown et al. (1985) and Feibel et al. (1989) reported conventional K/Ar age measurements on bulk feldspar separates from three pumice clasts from Tuff F of the Shungura Formation that yielded an overall average of 2.33 AE 0.02 Ma, so there is excellent agreement between all results. Tuff F is of reverse polarity (Brown et al. 1978), as expected at this age within the Matuyama Chron.
The Kalochoro Member of the Nachukui Formation has a thickness of 78 m in its type section (Harris et al. 1988a), and correlates with members F, G and the lowest part of Member H of the Shungura Formation. Harris et al. (1988a) correlated the Kalochoro Tuff with Tuff F of the Shungura Formation on the basis of glass composition. The Kalochoro Tuff is c. 1.5 m thick and contains pumice clasts up to c. 5 cm across. Feldspars from two pumices yielded concordant ages (Table 1), although three markedly older measurements in the range 2.83-4.35 Ma were omitted from results for 93-1059C-2. The average ages from the two pumices are concordant, and provide an overall mean of 2.331 AE 0.015 Ma (n ¼ 27; Table 1). The probability plot, however, shows two peaks less than 1% apart (Fig. 3). Nevertheless, the weighted mean age and arithmetic mean age are indistinguishable. The average age agrees well with that on Tuff F of the Shungura Formation, in keeping with the geochemical matching of the glass compositions of the two tuffs. However, we have kept results separate in Table 1 and Figure 2 because the average K/Ca value for the feldspars from the Kalochoro Tuff of c. 78 is very different from the overall average of c. 1280 found for feldspars from Tuff F, despite their indistinguishable ages. Glass compositions of pumices from the Kalochoro Tuff generally are distinct from those of the host tuff, in particular having markedly higher alumina content. We suggest that Tuff F and the Kalochoro Tuff were produced by the same explosive volcanic event, but that the pumice clasts in the two tuffs sample two different eruptions very closely spaced in time.

Tuff G-3
Tuff G-3, a sandy tuff c. 1 m thick, lies c. 20 m above the base of Member G (total thickness c. 200 m) of the Shungura Formation (de Heinzelin 1983, p. 93). Feldspars from two small pumice clasts from this tuff (00-297A, B) yielded average ages that agree well (Table 1), with an overall mean of 2.188 AE 0.036 Ma (n ¼ 23). The probability plot is roughly Gaussian, but has a secondary peak at a slightly younger age (Fig. 3). Tuff G-3 is of reverse polarity (Brown et al. 1978) and undoubtedly is within the Matuyama Reverse Chron. Submembers G-4 to G-8 are normally magnetized, and probably lie within the Reunion Normal Subchron (Brown et al. 1978;Kidane et al. 2006).
Given the age of 2.19 Ma determined here for Tuff G-3, and the age of 2.32 AE 0.02 Ma for Tuff F, it is likely that the K/Ar age of 2.33 Ma reported by Brown et al. (1985) for Tuff G is incorrect. We suggest that the pumices measured from Tuff G were reworked products of an earlier eruption. Assuming constant sediment accumulation between the base of Tuff F and Tuff G-3, Tuff G has an estimated age of 2.23 Ma.

Kangaki Tuff
The Kangaki Tuff lies in the upper part of the Kalochoro Member of the Nachukui Formation at Kangaki (Harris et al. 1988a, p. 10). It is a pale grey vitric tuff, c. 2.4 m thick, with pumice clasts up to c. 15 cm across that contain sparse alkali feldspar crystals. Mean ages on feldspars from four pumices, after omitting outliers, range from 2.034 AE 0.012 to 2.099 AE 0.018 Ma, a spread (3.1%) greater than expected, suggestive of a mixed population (Table 1). This is supported by the spread in average K/Ca values from 30 AE 16 to 72 AE 22 (Table  1). Recognizing the possibility of a mixed age population, we still regard the current best estimate as the overall mean age of 2.063 AE 0.032 Ma (n ¼ 48) because even the youngest mean age for pumice 98-328 is barely distinguishable. If results from pumice 00-296A, with the oldest average age, are removed, the mean age from the remaining three pumices only decreases to 2.052 AE 0.025 Ma (n ¼ 36). The probability plot has a dominant peak at c. 2.05 Ma with a secondary peak at c. 2.12 Ma (Fig. 3). Despite these problems, the eruption age is fairly well controlled at 2.06 AE 0.03 Ma, and provides useful constraints on the age of this part of the sequence. Importantly, the Kangaki Tuff lies stratigraphically a small distance below a prominent basinwide lacustrine interval. This lacustrine interval marks the base of the informal upper Burgi Member in the Koobi Fora Formation (Brown & Feibel 1986); in the Shungura Formation this interval begins in submember G-14. An age slightly younger than 2.06 Ma most probably pertains to both stratigraphic levels.

Age of hominins and archaeological sites
Estimates of age for several significant hominin fossils recovered from below the KBS Tuff were briefly discussed above where appropriate. However, age assignments for a few other important hominins are provided below. Leakey et al. (2001) named Kenyanthropus platyops, based on cranium KNM-WT 40000, from a mudstone in the Kataboi Member of the Nachukui Formation in the Lomekwi drainage. They showed that this bed lies c. 12 m above the Lokochot Tuff and 8 m below the Tulu Bor Tuff, so that linear stratigraphic scaling between the two tuffs yields an age of 3.50 Ma for the hominin, in agreement with the estimate given by Leakey et al. (2001). However, the uncertainty on this age is unlikely to exceed 30 ka, rather less than the error of 0.1 Ma given previously. Other hominin fossils from the same general area (Leakey et al. 2001) can also be dated a little more precisely, but require little change to the ages already assigned.  described several additional hominin fossils from the same general area in the Nachukui Formation. Some, from the Lomekwi Member from 5 to 17 m above the Tulu Bor Tuff, were assigned to Australopithecus afarensis. Ages estimated by  for these fossils of 3.26-3.35 Ma need to be increased to 3.31-3.40 Ma on the basis of the present age data.
Archaeological sites Lokalalei I (LA1) and 2C (LA2C), within the Kalochoro Member of the Nachukui Formation contain abundant Oldowan stone tools (Roche et al. 1999). A juvenile hominin molar (KNM-WT 42718) recovered from site LA1AE, near LA1, was assigned to Homo (Prat et al. 2005). These sites lie ,10 m above the Ekalalei Tuff, which is located above the level of the Kalochoro (¼ F) Tuff, the latter dated here at 2.33 AE 0.02 Ma. Roche et al. (1999) provided an estimated age of 2.34 Ma for these sites, slightly older than our estimate of 2.30 Ma.
Famed hominin cranium KNM-ER 1470 derives from 37 m below the KBS Tuff in the upper Burgi Member of the Koobi Fora Formation, and has always been regarded as little older than the KBS Tuff (Feibel et al. 1989). As noted above, the age of the Kangaki Tuff of 2.06 AE 0.03 Ma is only slightly greater than the that of the base of the upper Burgi Member, which lies 42 m below KNM-ER 1470. The older boundary of the Olduvai Normal Subchron lies c. 13 m below the KBS Tuff (Hillhouse et al. 1977). Thus KNM-ER 1470 must lie between 2.06 and 1.94 Ma, so that an age of 2.00 AE 0.05 Ma is a realistic estimate.

Chronological relations between Omo Group formations
The new ages also allow chronological relations between members of the principal formations of the Omo Group to be set out very clearly (Fig. 5). Where member boundaries are placed at correlative tuffs, they are of the same age, and in most instances that age has now been measured. Directly measured basal ages are absent only for the Okote Member of the Koobi Fora Formation, and Members E, G, H and J of the Shungura Formation, but bounding ages for these can be closely approximated between stratigraphic levels whose ages are well known. Fossil specimens are normally referred to members in each formation, so their ages are also well constrained. However, assignment to a member provides relatively coarse temporal resolution because in some cases members span an interval up to c. 0.8 Ma (e.g., the Tulu Bor Member of the Koobi Fora Formation). Further refinement requires stratigraphic scaling within the member or placement of the fossils with respect to other dated tuffs.

Climatic considerations
The role of climate in controlling or influencing the depositional environment within the Omo-Turkana Basin has long been recognized, and Brown et al. (2006) briefly summarized some effects in the upper part of the Omo Group. We only comment here on significant lacustrine deposition recorded in the sedimentary record of the Omo Group below the KBS Tuff. Soon after initiation of the basin c. 4.3 Ma ago, lacustrine sediments were deposited in the Lonyumun Member of both the Koobi Fora and Nachukui formations, and also in the Kanapoi Formation. At Lothagam, the Muruongori Member of the Nachukui Formation, consisting of lacustrine diatomites and claystones, was deposited after 4.2 Ma ago (McDougall & Feibel 1999), but prior to eruption of the Moiti Tuff, 3.97 Ma ago. The widespread development of lacustrine beds throughout the basin indicates that an areally extensive lake, called the Lonyumun Lake, was at a high level at this time. This extensive lake possibly resulted from a wetter climate over the period, rather than being a consequence of tectonic events. Similarly, widespread lake sediments in the upper Burgi Member of the Koobi Fora Formation are also well represented in the Nachukui and Shungura formations. Again, this may reflect an interval when a much wetter climate prevailed. As noted above, the age of this extensive lake is slightly younger than the eruption of the Kangaki Tuff at 2.06 Ma but older than the KBS Tuff (1.87 Ma). Other intervals in the lower Omo Group for which lacustrine deposition is inferred are less widespread than the two intervals discussed above. Trauth et al. (2005) showed that the lacustrine interval just older than the KBS Tuff is particularly well recorded in many rift basins in East Africa, emphasizing that higher rainfall must characterize this and other similar intervals.

Summary
Without exception measured ages presented in this study are consistent with the stratigraphic order. This is powerful evidence for the veracity of not only the physical measurements of age on the alkali feldspars but also the depositional time scale provided by the results. As discussed above, the ages are related to eruption events, but we argue that deposition occurred soon after, with the time between eruption and deposition being almost negligible compared with the ages.
The results given here cover the earliest Pliocene sedimentary strata known at present in the Turkana Basin at Lothagam with an age of 4.244 AE 0.042 Ma to the level of the KBS Tuff, well dated at 1.869 AE 0.021 Ma . In this interval of 2.4 Ma we have dated feldspars from 17 stratigraphic levels, nearly half of which had not been dated previously or which had anomalous ages. On average, we have measured ages at intervals of less than 150 ka so that by stratigraphic scaling it is usually possible to estimate ages for undated levels with uncertainties considerably less than 100 ka.
We wish to thank M. G. and R. E. Leakey for helping to facilitate this work over many years, and the National Museums of Kenya for their assistance. The governments of Kenya and Ethiopia are thanked for their co-operation. All age measurements were made in the Research School of Earth Sciences at the Australian National University; the support provided over an extended period is gratefully acknowledged. Special thanks go to J. Mya for mineral separation, and to R. Maier and X. Zhang for their very able technical assistance in the laboratory. Neutron irradiations were facilitated by grants from the Australian Institute of Nuclear Science and Engineering with the co-operation of the Australian Nuclear Science and Technology Organization, which managed and operated the HIFAR nuclear reactor. Most of the coordinates given are estimates made from Google Earth imagery using WGS84 as the datum. Direct Global Positioning System (GPS) observations have been made on the ground in a number of cases, generally with agreement between the two sets of measurements to within about 100 m (i.e. c. 0.0018). For samples from Koobi Fora, photocoordinates are also given, based upon aerial photographs taken in December 1970 by Hunting Surveys, London. Photograph number is given followed by coordinates in millimetres from left and top edge of the image, respectively; top edge is where the photograph details are shown.