Introduction
Laetoli, most famous for its exquisitely preserved hominin footprints in tuffs from the Pliocene Laetolil Beds, preserves detailed paleoanthropological, paleoecological, sedimentological, and volcanic records stretching from > 4.4 Ma to the present (Leakey and Harris, Reference Leakey and Harris1987; Harrison, Reference Harrison2011a, Reference Harrison2011b; overview in Harrison et al., Reference Harrison, Kwekason, Su, Reynolds and Bobe2022). About 35 km southwest of Olduvai Gorge, Laetoli provides an earlier record of hominin activity, albeit with potential overlap to Olduvai in the youngest Laetoli strata. The potential for overlap has recently been expanded, given the discovery of older strata beneath surface exposures at Olduvai revealed through geological coring conducted by the Olduvai Gorge Coring Project (OGCP; Njau et al., Reference Njau, Toth, Schick, Stanistreet, McHenry and Stollhofen2021). Tephra preserved at Laetoli and Olduvai provide a record of volcanic activity from the nearby Ngorongoro Volcanic Highlands (Figure 1). These tephra provide an opportunity to establish a tephrostratigraphy linking basins and paleoanthropological sites.
Figure 1. Regional map, showing location of Laetoli relative to Olduvai Gorge and volcanoes of the Ngorongoro Volcanic Highlands.
The stratigraphy of the Laetoli area (Figure 2) includes the Lower Laetolil Beds (> 4.36–3.85 Ma), Upper Laetolil Beds (3.85–3.63 Ma), Lower Ndolanya Beds (3.58 Ma), Upper Ndolanya Beds (2.66 Ma), Naibadad Beds (2.155–2.057 Ma), Olpiro Beds (< 2.057 Ma), and Ngaloba Beds, as described by Hay (Reference Hay, Leakey and Harris1987) and as dated by Drake and Curtis (Reference Drake, Curtis, Leakey and Harris1987) and Deino (Reference Deino and Harrison2011). These strata are exposed by badlands-style erosion and correlated using sedimentology and tephrostratigraphy (e.g., Hay, Reference Hay, Leakey and Harris1987; Ditchfield and Harrison, Reference Ditchfield, Harrison and Harrison2011; Harrison and Kweka, Reference Harrison, Kweka and Harrison2011). The Laetolil Beds are the most extensively exposed and distributed (Figure 2), while the younger strata (Naibadad and Olpiro beds) are unevenly distributed, often filling erosional surfaces cut into the older strata.
Figure 2. Map of Laetoli localities (a) and overview of Laetoli area stratigraphy (b), indicating extent of Laetolil and Ndolanya beds exposures and the positions of sites sampled in the current study (after McHenry, Reference McHenry and Harrison2011).
Laetoli is renowned for discoveries within the Laetolil Beds, which host hominin trackways and have yielded a high diversity of fauna, including Australopithecus afarensis (Leakey, Reference Leakey, Leakey and Harris1987a, Reference Leakey, Leakey and Harrisb; Harrison, Reference Harrison and Harrison2011c, Reference Harrison and Harrisond; Masao et al., Reference Masao, Ichumbaki, Cherin, Barili, Boschian, Iurino, Menconero, Moggi-Cecchi and Manzi2016). The younger strata, however, also host an intriguing fossil and archaeological record (e.g., Ndessokia, Reference Ndessokia1990; Harrison, Reference Harrison and Harrison2011c; Harrison et al., Reference Harrison, Kwekason, Su, Reynolds and Bobe2022). The Naibadad and Olpiro beds yield only a sparse fauna, the latter of which is associated with Oldowan tools (Ndessokia, Reference Ndessokia1990). The Lower Ngaloba Beds, biostratigraphically correlated to ca. 1.7–1.2 Ma, have produced faunal remains in association with Acheulian tools (Harris and Harris, Reference Harris and Harris1981; Leakey, Reference Leakey, Leakey and Harris1987c; Harrison and Kweka, Reference Harrison, Kweka and Harrison2011; Arenson et al., Reference Arenson, Harrison, Sargis, Taboada and Gilbert2022). The Upper Ngolaba Beds, dated to > 200 ka (Manega, Reference Manega1993), have yielded a fauna associated with an archaic Homo sapiens cranium and Middle Stone Age lithic artefacts (Day et al., Reference Day, Leakey and Magori1980; Leakey, Reference Leakey, Leakey and Harris1987c; Masao and Kimambo, Reference Masao, Kimambo, Fortes-Lima, Mtetwa and Schlebusch2021). The current work improves and expands upon the prior dating and tuff compositional analyses for the deposits in the Laetoli region that are younger than the Ndolanya Beds, with a focus on the Naibadad Beds, which overlap in age and in general composition (rhyolitic) with Ngorongoro volcano and the newly described Ngorongoro Formation recovered from OGCP cores at Olduvai (Njau et al., Reference Njau, Toth, Schick, Stanistreet, McHenry and Stollhofen2021).
Background
Geologically noteworthy localities within the Laetoli area were first enumerated in Hay (Reference Hay, Leakey and Harris1987); those sites most relevant to the current study are highlighted in Figure 2. The Ndolanya Beds (Hay, Reference Hay, Leakey and Harris1987) unconformably overlie the Yellow Marker Tuff (YMT), which marks the top of the Upper Laetolil Beds. The Lower Ndolanya Beds consist largely of pedogenically altered tuffaceous claystone with some tephra, while the Upper Ndolanya Beds are typified by clay-rich aeolian tuffs with some calcrete (Ditchfield and Harrison, Reference Ditchfield, Harrison and Harrison2011).
The Ogol lavas locally overlie the Upper Ndolanya Beds and consist of phaneritic to porphyritic vogesite or alkaline basalt, with olivine and augite phenocrysts in a feldspar and augite matrix (Hay, Reference Hay, Leakey and Harris1987; Mollel et al., Reference Mollel, Swisher, Feigenson, Carr and Harrison2011; Zaitsev et al., Reference Zaitsev, Arzamastsev, Marks, Braunger, Wenzel, Spratt, Salge and Markl2021). They have been variously interpreted as locally sourced lavas (Hay, Reference Hay, Leakey and Harris1987; Zaitsev et al., Reference Zaitsev, Arzamastsev, Marks, Braunger, Wenzel, Spratt, Salge and Markl2021) or as the erosional remnants of lava flows originating from Lemagurut volcano (Mollel et al., Reference Mollel, Swisher, Feigenson, Carr and Harrison2011).
The Naibadad Beds are younger than the Ogol lavas, although they are not found in direct superposition, except possibly at Locality 10E where their relative stratigraphic position remains ambiguous. The Naibadad Beds are principally tuff (> 95%), with minor claystone and conglomerate. The Naibadad Beds tuffs are trachytic to rhyolitic, with a likely Ngorongoro volcanic source (Hay, Reference Hay, Leakey and Harris1987; McHenry, Reference McHenry and Harrison2011).
The Olpiro Beds are represented by relatively small, local exposures, identified by Hay (Reference Hay, Leakey and Harris1987) at Localities 1, 16, and 23 and in unpublished theses by Ndessokia (Reference Ndessokia1990) and Manega (Reference Manega1993) at a few additional sites (e.g., Locality 2S). The Olpiro Beds are dominated by brown claystone with minor conglomerate and sandstone, along with a tuff (the Olpiro Tuff). While Hay (Reference Hay, Leakey and Harris1987) proposed a correlation between the Olpiro Tuff and Olduvai Bed II based on composition, more recent dating attempts have placed its age at > 2.0 Ma (Deino, Reference Deino and Harrison2011), which is much too old to support such a correlation since the base of Bed II is ca. 1.8 Ma (Deino, Reference Deino2012). Mollel et al. (Reference Mollel, Swisher, Feigenson, Carr and Harrison2011) proposed a correlation between the Olpiro Tuff and the Coarse Feldspar Crystal Tuff (CFCT) in Olduvai Bed I, but while the two are likely similar in age, a more detailed analysis of the CFCT (McHenry et al., Reference McHenry, Stanistreet, Stollhofen, Njau, Toth and Schick2020) revealed a dominantly rhyolitic composition unlike the trachytic composition of the Olpiro Tuff.
Recently published ages and tephra compositions for the OGCP cores collected from the nearby Olduvai basin in 2014 (McHenry et al., Reference McHenry, Stanistreet, Stollhofen, Njau, Toth and Schick2020; Deino et al., Reference Deino, Heil, King, McHenry, Stanistreet, Stollhofen, Njau, Mwankunda, Schick and Toth2021), which include new stratigraphy below the level exposed in outcrop now referred to as the Ngorongoro Volcanic Formation (Stanistreet et al., Reference Stanistreet, Stollhofen, Deino, McHenry, Toth, Schick and Njau2020), provide an opportunity to correlate between Laetoli and the Olduvai basin. Tuffs and ignimbrites sampled in the lower part of Core 2A (below the Olduvai Bed I basalt) have trachytic to rhyolitic compositions and ages between 2.167 Ma and 2.003 Ma (Deino et al., Reference Deino, Heil, King, McHenry, Stanistreet, Stollhofen, Njau, Mwankunda, Schick and Toth2021), comparable to previous results for the Naibadad Beds tuffs (e.g., Deino, Reference Deino and Harrison2011). Core 3A provides a similar record of Ngorongoro volcanism, with tuffs below the level of the Olduvai Bed I basalt dating from 2.112 to 2.030 Ma (Deino et al., Reference Deino, Heil, King, McHenry, Stanistreet, Stollhofen, Njau, Mwankunda, Schick and Toth2021). The 2014 OGCP cores did not reach bedrock, and a seismic survey of the Olduvai basin (Lu et al., Reference Lu, Hanafy, Stanistreet, Njau, Schick, Toth, Stollhofen and Schuster2019) reveals that there is likely an additional 200+ meters of basin fill below the level cored, suggesting that older beds of Laetoli could have stratigraphic equivalents still buried deep within the Olduvai basin.
Field methods
Key Naibadad and Olpiro beds localities named and described by Hay (Reference Hay, Leakey and Harris1987), Manega (Reference Manega1993), Ndessokia (Reference Ndessokia1990), and McHenry (Reference McHenry and Harrison2011) were re-visited in 2016 for detailed stratigraphic measurement and tuff sampling (Figure 3). Tephra samples (LT16 prefixes, 35 samples total) and 40Ar/39Ar samples (L16 prefixes, 21 samples total) were collected in parallel by two of us (LM and AD, respectively).
Figure 3. Measured stratigraphic sections, showing positions of samples collected. Color coded by bed (based on tephra composition). Sections are organized by locality number. Clusters of nearby stratigraphic sections are grouped, with a solid horizontal line beneath. No datum is implied by the positions of the bases of the stratigraphic sections. Section descriptions focused on relative positions of tephra layers, thus not all intervening lithologies were noted for all sites.
Samples dated here are tuffs and lavas from Laetoli and nearby potential proximal tephra and source areas to the northeast (Naisuri Gorge and Lemagurut Volcano). All 40Ar/39Ar dating samples were collected by one of us (AD) during the 2016 field season. The dated sample set consists of 21 tuffs derived from the upper Ndolanya Beds, a tuff apparently related to the Ogol Lava in Naisuri Gorge, and the Naibadad Beds. In addition, we dated a sample of the Ogol Lava, and a lava exposed on the flanks of Lemagurut Volcano.
At Locality 2N (the northern end of Locality 2; see Figure 2A), exposures along the slope of the ridge to the north were scouted for claystone and darker-colored tephra potentially corresponding to the Olpiro Beds (from which two samples were collected), and lower-lying tuffaceous outcrops to the south of this ridge were measured and sampled for the Naibadad Beds (Figure 4, four samples collected). At Locality 2S (the southern end of Locality 2), we searched for Manega’s (Reference Manega1993) purported Olpiro Beds exposure and measured a series of detailed sections of Naibadad Beds tuffs (five samples collected), some of which preserve volcanic glass. Figure 3 shows the erosional relationships among the three measured sections from this site.
Figure 4. Locality 2N field photo showing the contact between the Ndolanya Beds and Naibadad Beds and positions of sampled tephra units.
At Locality 9 (Figure 2A) we collected tephra samples at sites 9W (six samples) and 9 (two new samples), the latter of which overlaps with the section previously analyzed in McHenry (Reference McHenry and Harrison2011). Tuff positions and thicknesses were measured from the top of the Yellow Marker Tuff (top of the Laetolil Beds) up through the Naibadad Beds (Figure 3). Section 9W, 700 meters to the west of Locality 9 along the same hillside, was measured and sampled from the base to the top of the exposure (∼13 meters, all within the Naibadad Beds). The units identified in this section are almost entirely tuff, with calcretes separating most units.
Two Naibadad Beds tuffs were sampled from within a continuous stratigraphic section at Locality 10E (Figure 2A). The Ogol lava was sampled from the top of nearby Norsigidok Hill.
Locality 16 serves as the type locality for the Olpiro Beds because the Olpiro Tuff is best preserved and most easily identified here near its base (Hay, Reference Hay, Leakey and Harris1987). We collected three samples of the Olpiro Tuff, one sample of Ndolanya tuff, and two samples of Naibadad Beds tuffs at this locality, although complex unconformable stratigraphy made it impossible to collect from a single continuous section.
Hay (Reference Hay, Leakey and Harris1987) also identified the Olpiro Beds at Localities 1 and 23, while Manega (Reference Manega1993) provisionally identified an exposure at Locality 2S. We visited these locations and collected samples from units matching the stratigraphic descriptions of the Olpiro Beds, when they could be located. Four samples from three distinct Naibadad tuffs and one sample of the Yellow Marker Tuff (topmost unit of the Laetolil Beds) were collected from Locality 23. One candidate Olpiro tuff was sampled from Locality 1.
We also visited and collected two samples at Naisuri Gorge, a steep valley incised through volcanic stratigraphy located to the east between Laetoli and Lemagurut, and from the summit of Lemagurut itself (Figure 1). These samples contribute to our understanding of the regional chronostratigraphy and potential source areas.
Laboratory methods
40Ar/39Ar dating methods
The 40Ar/39Ar sampling and dating methods used here are similar to those used previously to date Laetoli tuffs (Deino, Reference Deino and Harrison2011), with modifications primarily related to the availability and use of a newer generation of Noble gas mass spectrometers with improved resolution, sensitivity, and multi-collection capability, as described below. Additional details as to the use of this equipment are provided in Deino et al. (Reference Deino, Dommain, Keller, Potts, Behrensmeyer, Beverly and King2019a, Reference Deino, Sier, Garello, Keller, Kingston, Scott, Dupont-Nivet and Cohen2019b). We also described a new approach to age estimation of central tendencies of replicate analyses of individual phenocrysts from a tuff, based on Bayesian inference (Deino et al., Reference Deino, Sier, Garello, Keller, Kingston, Scott, Dupont-Nivet and Cohen2019b).
Samples were processed at the Berkeley Geochronology Center (BGC) in preparation for 40Ar/39Ar dating. The tuff processing methodology is similar to that described in Deino (Reference Deino and Harrison2011), using conventional separation techniques consisting of disaggregation using a ceramic mortar and pestle, sieving, dilute HF and distilled water rinses, and hand-picking under a binocular microscope to obtain a finished product of clean feldspar grains as free of inclusions as possible.
The two lava samples were processed to obtain purified groundmass separates by hand crushing in a metal mortar and pestle, sieving (250–350 microns for Ogol Lava sample L16/10, and 150–250 microns for Lemagurut sample L16/LEM-1), removal of phenocrysts with a hand magnet, a magnetic Frantz Isodynamic Separator, separation with heavy liquids, and hand-picking. This was followed by several rinses in distilled water in an ultrasonic bath. As an experiment to see if the radiogenic content and argon systematics could be improved by acid treatment, a split of the Ogol Lava sample (‘L16/10 HCl/HNO3ʹ) was treated with a sequence of mild acid washes (HCl 1N, HCl 6N, HNO3 1N, HNO3 6N) at 60°C in an ultrasonic bath prior to the final distilled water rinse to potentially remove alteration products.
The final mineral separates were irradiated in the cadmium-lined in-core irradiation tube (CLICIT) position of the Oregon State University TRIGA reactor in three batches (BGC irradiation 471 for 0.83 hour, and 474 and 478 for three hours each). All irradiations employed sanidine phenocrysts from the Alder Creek Rhyolite of California (orbitally referenced age = 1.1848 ± 0.0006 Ma; Niespolo et al., Reference Niespolo, Rutte, Deino and Renne2017) as the neutron fluence monitor mineral. Standards and unknowns were co-irradiated in a circular configuration in wells in an aluminum disk, with standards at the cardinal positions, and either two or three unknowns situated between standards. The appropriate neutron fluence factor (the ‘J’ parameter of 40Ar/39Ar dating calculations; McDougall and Harrison, Reference McDougall and Harrison1999) for the unknown positions was calculated from a planar fit of the standard calibrations, with 1σ errors derived by Monte Carlo simulation in the predicted J value ranging from 0.1–0.3%. Reactor-induced isotopic production ratios for these irradiations were: (36Ar/37Ar)Ca = 3.65 ± 0.02 × 10−4, (38Ar/37Ar)Ca = 1.96 ± 0.08 × 10−5, (39Ar/37Ar) Ca = 6.95 ± 0.09 × 10−4, (37Ar/39Ar) K = 3.24 ± 0.16 × 10−4, (38Ar/39Ar) K = 1.220 ± 0.003 × 10−2, (40Ar/39Ar) K = 3.5 ± 0.9 × 10−4. Atmospheric 40Ar/36Ar = 298.56 ± 0.31 (Lee et al., Reference Lee, Marti, Severinghaus, Kawamura, Yoo, Lee and Kim2006) and decay constants follow (Min et al., Reference Min, Mundil, Renne and Ludwig2000).
After a period of several months to permit radiological ‘cooling’ after irradiation, the materials were analyzed at the BGC using two techniques. K-feldspar phenocryst separates were analyzed grain-by-grain, and each grain was tested using incremental heating. Here, heating levels (laser output levels) were raised progressively from low temperature to fusion in a series of sequential independently measured steps (4–10 steps), termed the single-crystal incremental heating (SCIH) approach. To conserve analytical resources, grains yielding ages much older than expected in the first two steps were not analyzed further. The two lava samples (three separates) were analyzed in three small (10–20 mg) multi-grain aliquots per sample of groundmass using incremental heating (‘MGIH’). All argon measurements were carried out using an automated extraction line leading to a Nu Instruments 5-collector Noblesse mass spectrometer, using ion-counting electronics.
Completed incremental heating analyses were checked for apparent age plateaus, using a modified approach to that of Fleck et al. (Reference Fleck, Sutter and Elliot1977). Here, we searched for a set of contiguous steps encompassing the greatest percent of 39Ar release that exhibited an acceptable MSWD (‘mean square of weighted deviates,’ with a threshold probability > 95% that the observed scatter is caused by analytical error alone and that geological scatter is not demonstrated). A plateau must comprise at least 50% of the total 39Ar release and consist of at least three consecutive steps. We added here an additional threshold for inclusion of a plateau in further data analysis, in that the overall experiment must have consisted of more than three steps.
We further analyzed the argon systematics using ‘inverse isochron’ regressions (36Ar/40Ar vs. 39Ar/40Ar) calculated from the plateau steps (Tables S1 and S2). However, we present these results, in the form of 40Ar/36Ar ‘trapped’ composition and isochron age, for comparison only. In the more highly radiogenic data suites from older, relatively highly potassic samples such as these, we find that isochrons either reproduce the plateau ages within error, or introduce anomalous scatter to age populations; thus, the plateau age is taken as the reference age of all incremental heating analyses.
The age populations for a given sample were examined for outliers using a two-step approach. Age outliers may have either a geological origin (e.g., detrital contamination, alteration, or trapped excess 40Ar), or be due to random analytical occurrence. Typically, a geological outlier is older than the primary eruption age of a tephra. We identified distinctly older grains using a gap-finding technique (Deino et al., Reference Deino, Dommain, Keller, Potts, Behrensmeyer, Beverly and King2019a, Reference Deino, Sier, Garello, Keller, Kingston, Scott, Dupont-Nivet and Cohenb), with the sensitivity set to a gap score of seven. This procedure is sufficient to flag most old outliers easily identified by a human observer yet have little effect on the primary age distribution. Secondly, we identified outliers by examining the deviation of results from the median age and omitted those with a normalized median absolute deviation (‘nMAD’) from the median > 3.
Finally, we present two approaches to calculating the central tendency and error of the filtered sample age populations. The first calculates a conventional weighted mean, with the error as a modified standard error (multiplied by root MSWD where the MSWD > 1). The second uses a Bayesian parameter estimation approach (Keller, Reference Keller2018; Deino et al., Reference Deino, Sier, Garello, Keller, Kingston, Scott, Dupont-Nivet and Cohen2019b) that accommodates a tailing toward older ages commonly observed in East African phenocrystic feldspar phenocryst 40Ar/39Ar single-crystal data sets from tuffs (e.g., Deino et al., Reference Deino, Sier, Garello, Keller, Kingston, Scott, Dupont-Nivet and Cohen2019b). Such tailing is evident here in several samples (e.g., L16/2s-1 and L16/16-1). Our Bayesian prior constraint centers on the assertion that, although inherited, xenocrystic, or detrital grains may predate deposition, no 40Ar/39Ar ages may postdate deposition, except due to analytical uncertainty. Consequently, we expect a relative distribution of true single-crystal closure ages that is truncated at the time of deposition. The majority of ages should cluster at this limit within analytical uncertainty; however, the proposed distribution should permit a tail of older ages from, for example, incompletely degassed xenocrysts. We found that an exponential decay model meets this requirement, and closely approximates the conventional weighted mean when excess analytical scatter is absent.
Electron probe microanalysis methods
For electron probe microanalysis (EPMA), samples were gently crushed in a ceramic mortar and pestle, sieved to obtain the 60–100 mesh size fraction, sonicated in a 4% HF solution for one minute, and rinsed a minimum of three times in DI water (until the decanted solution was clear). Where identified, individual grains of feldspar, augite, hornblende, glass, oxide minerals (ilmenite and titanomagnetite), olivine, aenigmatite, and andradite (melanite) garnets were then hand-picked under a microscope and mounted in epoxy, polished, carbon coated, and analyzed by electron microprobe (Cameca SX-Five at University of Wisconsin-Madison). Phase identifications were confirmed using energy dispersive spectroscopy (EDS), and then quantitatively analyzed using phase-specific routines by wavelength dispersive spectroscopy (WDS) following the methods described in McHenry et al. (Reference McHenry, Njau, de la Torre and Pante2016). Check samples (including glass, feldspar, augite, hornblende, and titanomagnetite) were analyzed at the start of every run, with calibrations adjusted to ensure consistency between runs. Specific analytical details, including beam parameters, standards, and check sample compositions, are reported in McHenry et al. (Reference McHenry, Stanistreet, Stollhofen, Njau, Toth and Schick2020).
Correlation methods
Tuffs were characterized compositionally using two main approaches: (1) presence/absence of specific minerals and (2) EPMA elemental data of selected phases. The presence/absence and qualitative relative abundance of minerals were determined using optical microscopy during hand picking for EPMA analysis and confirmed by EPMA. The elemental compositions of the analyzed augite, titanomagnetite, feldspar, and (where available) glass and hornblende of various samples were compared visually using a series of bivariate and ternary plots, looking for compositional clusters. A correlation is possible where mineral assemblage (presence/absence of phases) and compositional clustering across multiple phases agree, although with closely spaced, compositionally similar tuffs, this does not always produce a unique “fingerprint” for an individual eruption.
Results
40Ar/39Ar dating results
The 1991 steps on 386 feldspar phenocrysts from 33 tuff samples were analyzed by the SCIH method (Table S1). Of these experiments, 40 were either truncated for the sake of efficiency after one or two initial low-power steps because they were much older than the expected eruption age or did not yield more than three steps prior to completion of the incremental heating release. The remaining experiments generally yielded plateaus (304 out of 346), 88% of these encompassing more than 90% of the total 39Ar released (Table S2). After outlier detection and elimination from the plateau sample age populations as described in Methods, about half the samples are represented by simple, pseudo-normal distributions, while the others exhibit positive skewness (tail toward older ages) or pronounced multi-modality. Weighted-means of the plateau age populations are provided in Table 1.
Table 1. Summary of dating results
* Previously published in Deino, Reference Deino and Harrison2011.
† SCIH = single-crystal incremental heating; MGIH = multi-grain incremental heating; SCTF = single-crystal total fusion (Deino, Reference Deino and Harrison2011).
‡ Mean square weighted deviation.
§ Median Ca/K value.
‖ Weighted-mean age ± 1 SD modified standard error (standard error multiplied by root MSWD when MSWD > 1).
¶ Bayesian model age using exponential prior distribution ± 1 SD standard error.
Preferred ages shown in bold font.
Weighted-mean ages are shown in bold-italic font.
More than half of the samples exhibit greater scatter than expected from analytical error alone (high MSWD with > 95% probability that analytical error is insufficient to explain the observed dispersion). To provide a consistent approach to estimating a ‘best’ eruption age for these samples, we resorted to a Bayesian approximation technique utilizing an exponential prior distribution following Keller (Reference Keller2018) and Deino et al. (Reference Deino, Sier, Garello, Keller, Kingston, Scott, Dupont-Nivet and Cohen2019b). The estimated ages and probability distributions are shown in Figure 5 and listed in Table 1. The Bayesian parameter estimation approach yields ages that are in all cases younger than the weighted-mean ages, by a median of ca. 7 ka (minimum 0.8 ka, maximum 99 ka). Figure 6, illustrating the age difference as a function of the MSWD probability, shows that as the probability that analytical scatter alone is responsible for the observed distribution approaches zero, the age difference increases. This reflects the inappropriateness of the weighted-mean calculation, which assumes normality, while the Bayesian estimation approach emphasizes the younger end of the distribution as representing the true eruption age.
Figure 5. (A–U) Feldspar phenocryst SCIH age-probability distributions showing the age-probability density spectra of each dated tuff sample. Each plot shows the age distribution before outlier deletion, after outlier deletion, and the estimated Bayesian exponential model posterior distribution (see legend). A rank-order plot of each analysis is shown in the upper part of each graph.
Figure 6. Mean square of weighted deviates (MSWD) probability versus age difference. Variation of the MSWD probability with the difference in age of a weighted-mean versus Bayesian exponential age estimation for individual sample age populations.
Eighty-nine steps on three aliquots each from three lava samples were analyzed by the MGIH method (Table S1). Incremental heating plateau and isochron data are provided in Table S2. Most MGIH aliquots yielded plateaus by the definition given above (5 of 9 experiments; Figure 7). The 40Ar/36Ar ‘trapped’ ratios derived from isochron analysis of the plateau steps are all within error of the anticipated atmospheric composition of 298.56 (Table S2; Lee et al., Reference Lee, Marti, Severinghaus, Kawamura, Yoo, Lee and Kim2006), suggesting that inherited radiogenic 40Ar is not a detectable problem with these samples. Age-probability spectra of the MGIH experiments are shown in Figure 8.
Figure 7. (A–I) Lava flow matrix groundmass using incremental heating (MGIH) incremental 39Ar release spectra showing 39Ar incremental release spectra for three aliquots each from three lava samples. ‘Int. Age’ at the bottom of each panel provides the integrated age from computational re-combination of all gas steps. Apparent-age plateaus are shown by the double-barbed arrows. All uncertainties are 1σ analytical error.
Figure 8. Groundmass using incremental heating (MGIH) age-probability spectra and related parameters. (A) Age-probability density spectra of each sample, with weighted-mean age and 1σ error (including error in J, the neutron fluence parameter in 40Ar/39Ar dating). (B) Rank order plot of the individual aliquots grouped by sample. Uncertainties in age are given at 1σ standard error. (C) Ca/K atomic ratio derived as a by-product of the 40Ar/39Ar analysis. (D) Percentage of radiogenic 40Ar released.
The two sample treatments on the Ogol Lava sample, one washed only in distilled water (L16/10-1) and the other treated with a series of warm HCl and HNO3 washes (L16/10-1 HCl/HNO3) as described in Methods, yielded distinctly different ages; acid washing increased the plateau ages slightly from a weighted mean of 2.308 ± 0.013 to 2.346 ± 0.008 Ma, a statistically significant difference at the 95% confidence interval. It is not readily apparent from the 39Ar release spectra in Figure 8 which treatment yielded the most well-behaved argon systematics. The untreated material maintained greater consistency over the entirety of the release experiments, as reflected in the percentage of radiogenic 40Ar, the Ca/K ratio, and the extent of the plateaus. However, the acid-treated material shows a sharper division in Ca/K behavior between what is likely matrix outgassing in the early phase of the experiment, and plagioclase/pyroxene outgassing in the latter third or so of the release. In the absence of a compelling reason to choose one sample treatment over another, we take the weighted-mean of the two treatments of 2.336 ± 0.007 Ma as the representative age of the Ogol Lava. The age of lava sample (L16/LEM-1) obtained from the summit of Lemagurut Volcano is 2.341 ± 0.007 Ma, on the basis of a single plateau result from the three MGIH experiments. The 40Ar/36Ar age results are summarized in Figure 9 and are described in detail in the discussion.
Figure 9. Summary age distribution of latest Pliocene–Pleistocene volcanism of the Laetoli area. These plots display all 40Ar/39Ar ages on upper Laetoli stratigraphic units, including three ages recalculated from Deino (Reference Deino and Harrison2011). (A) Age-probability spectra combining samples for each stratigraphic unit, with weighted-mean unit age. Note that stratigraphic units may be composed of multiple eruptions separated in time, which have not been individually characterized. (B) Rank order of each sample grouped by stratigraphic unit. (C) Ca/K ratio of each sample. The volcanic source for tephra between ca. 2.2 and 2.0 Ma is identified as Ngorongoro volcano, while tephra and flows at ca. 2.3 Ma are from Lemagurut volcano. The volcanic source of the upper Ndolanya Beds remains enigmatic.
Electron probe microanalysis (EPMA) results
The mineral assemblages for the tuffs analyzed in this study are reported in Table 2. Phenocryst and glass compositions for the tuffs analyzed in this study are reported in S3–S10.
Table 2. Sample IDs, bed designations, and mineral assemblages
Mineral assemblages based on EPMA results on thin sections and grain mounts
XX = abundant; X = common; + = trace
Olp = Olpiro; Ndo = Ndolanya; Naib = Naibadad; L, M, U = Lower, Middle, Upper.
Examining the presence/absence data, together with the EPMA bivariate plots, several key clusters were observed that can be used to discriminate individual (or groups of similar) tuffs. While all samples contained some specific phases (e.g., feldspar of some kind, and augite), few samples contained abundant hornblende (only LT16-16.01, 02, and 03), and andradite garnet varied considerably in abundance between samples (e.g., not present in most Locality-9 samples, except for the basal Naibadad tuff). Augite was ubiquitous in the tuffs sampled but differed in composition (Figure 10), with distinct clusters characterized by low Na/high Fe, low Na/low Fe/high Ti/high Mg, and high Na/high Fe, although some samples contained more than one of these compositional groups. Feldspars also showed distinct compositional clusters (Figure 11), with some plotting as high-K sanidine (e.g. LT16-16.04, 1.01, 2.05, 2.04, 23.01), some intermediate plagioclase (e.g. LT16-16.01, 16.02, 16.03), and most in the anorthoclase compositional range. For samples that preserved glass, rhyolites LT16-23.02, 23.04, and 9.09 had the highest Si and lowest Al (Figure 12), while samples LT16-16.01 and 16.02 had the highest Al and plotted in the trachyte range. The glasses of samples LT16-9.04 and 2.08 overlap in composition (but not with other samples), suggesting a possible correlation.
Figure 10. Electron probe microanalysis (EPMA) of augite compositions for samples collected in this study. (A) MgO versus FeO. This plot shows that Olpiro tuffs (green symbols) have higher Mg augite than most Ndolanya (blue symbols) or Naibabad tuffs (all other colors) and contain no high-FeO augite. (B) Na2O versus FeO. This plot shows a distinctive low-Na augite population found within all Naibadad Beds tuffs (and absent in Ndolanya and Olpiro Beds tuffs), and that Olpiro Beds augite are low in FeO. (C) TiO2 versus FeO, showing elevated Ti for the Olpiro Beds augite.
Figure 11. Electron probe microanalysis (EPMA) of feldspar compositions for samples collected in this study. (A) BaO versus K2O. This plot shows compositional differences among the Olpiro (green symbols), Ndolanya (blue symbols), and Naibadad Beds feldspar, mostly in their K2O content. (B) FeO versus K2O, showing similar separations. (C) Feldspar ternary diagram, showing sanidine composition for Ndolanya Beds feldspar, anorthoclase for Naibadad Beds, and plagioclase for Olpiro Beds.
Figure 12. Electron probe microanalysis (EPMA) of glass compositions (EPMA) for samples collected in this study that preserve glass. (A) SiO2 versus Al2O3. Olpiro glass (green symbols) has the highest Al, while Upper Naibadad samples LT16-23.02 and 23.04 have the highest SiO2. (B) FeO versus Al2O3. (C) Alkali versus silica diagram, showing trachytic to phonolitic composition for Olpiro Beds and trachytic to rhyolitic composition for the Naibadad Beds glass. IUGS and Le Maitre, Reference Le Maitre1989. IUGS = International Union of Geological Sciences. No Ndolanya Beds samples collected for this study preserved volcanic glass.
For samples of uncertain stratigraphic origin (e.g., prospective “Olpiro Tuff” samples from sites other than Locality 16), the assemblage was compared against previously reported compositions for different beds within the Laetoli area (Hay, Reference Hay, Leakey and Harris1987; McHenry, Reference McHenry and Harrison2011). While Naibadad Beds samples had mostly consistent mineral assemblages (rhyolitic or trachytic glass, anorthoclase feldspar, augite), most candidate “Olpiro Tuff” samples yielded compositions more similar to either Ndolanya or Naibadad beds tuffs, lacking the hornblende expected for the Olpiro Tuff.
Discussion
New chronostratigraphic groups for upper strata at Laetoli, based on 40Ar/39Ar dating results
Figure 9 and Table 1 summarize the overall age results for samples of the younger strata at Laetoli, from the Upper Ndolanya Beds through the Naibadad Beds, including three ages from Deino (Reference Deino and Harrison2011) recalculated using the Bayesian approach described here for consistency. The ages determined for the tuffs of the Naibadad Beds can be grouped into three main intervals, with ten samples falling into the “Lower” (2.189–2.154 Ma), four in the “Middle” (2.115–2.104 Ma), and four in the “Upper” Naibadad Beds (2.036–2.004 Ma). Within these groups, ages mostly overlap within error; however, each group may be composed of several distinct eruptions separated by a short interval. While some sampled sites contain only Lower Naibadad Beds units (e.g., Localities 2, 2S), others contain units in more than one age group (e.g., Locality 16, Lower and Middle; Locality 10E, Lower and Middle; Locality 23, Lower and Upper; and Localities 9 and 9W, Lower, Middle, and Upper).
Compositionally, the Ca/K ratios are notably uniform within each age group but differ significantly between several of the groups. This observation suggests that each group has a unique volcanic source, but that sources shifted and evolved through time. For example, the oldest group, the Upper Ndolanya Beds, exhibits unusually low Ca/K ratios typical of high-K anorthoclase. Based on ages, and mineral compositional data described below, the tuffs of the Naibadad Beds are attributed to eruptions from the Ngorongoro Crater volcanic system, but the source of the ca. 300-kyr-older Upper Ndolanya Beds remains unknown. Finally, the anorthoclase composition trends toward decreasing Ca/K ratio from the Lower Naibadad Beds through the Upper Naibadad Beds and provides an intriguing hint of long-term evolution of the Ngorongoro Crater volcanic system (from ca. 2.2 to 2 Ma), the presumed source of these eruptions.
Relative timing of the Ogol Lavas and Lemagurut volcano
The age of the Lemagurut summit lava sample (L16/LEM-1) is 2.341 ± 0.007 Ma, which is statistically indistinguishable from that of the Ogol Lava age (2.336 ± 0.007 Ma) at the 95% confidence interval. These are not the same eruptions based on lithologic characteristics (e.g., L16/LEM-1 has large phenocrysts of plagioclase, which the Ogol Lava lacks, and the latter is characterized by abundant phenocrysts of pyroxene and olivine in LEM-1). Nevertheless, the age correspondence of late-erupted events from Lemagurut with the Ogol Lava, separated by a distance of 23 km, suggests they are dynamically related, perhaps through a direct magmatic connection or indirectly via synchronized volcanism. The narrow age grouping of 2.34–2.33 Ma and low-K compositional data illustrated in Figure 9 suggest that the Olgol Lavas, Naisuri Gorge tuffs, and the Lemagurut summit lava were all generated from the same magmatic system, namely Lemagurut Volcano. Zaitsev et al. (Reference Zaitsev, Arzamastsev, Marks, Braunger, Wenzel, Spratt, Salge and Markl2021) explored the link between the Ogol lavas and Lemagurut using petrologic and isotopic data and favored a model involving multiple stages of magma mixing and mingling between the two, which potentially both derived from the same original basaltic melt mantle sources.
Major compositional trends
Augite, feldspar, glass, and titanomagnetite compositions for the tuffs analyzed in this study are plotted in Figures 10–13. While the tuffs of the Laetolil, Ndolanya, Naibadad, and Olpiro beds differ significantly from each other in terms of their mineral assemblages, compositions, and ages (see Hay, Reference Hay, Leakey and Harris1987; McHenry, Reference McHenry and Harrison2011), compositions within each of these stratigraphic units can be quite similar. Thus, composition permits identification of strata at the level of major stratigraphic units (i.e., attribution of a tuff to the Ndolanya Beds, Lower Naibadad Beds, or Olpiro Beds, but rarely any more stratigraphically specific than that), a powerful capability given the complex erosional, tectonic, and depositional history of the area, but the correlation of individual tuffs between sites is more problematic.
Based on the tephra mineral assemblages for the Laetoli area stratigraphic units documented by Hay (Reference Hay, Leakey and Harris1987) and McHenry (Reference McHenry and Harrison2011) (Figure 2), the Naibadad Beds tuffs are rhyolites with feldspar, augite, and titanomagnetite compositions that distinguish them from both the underlying Ndolanya and overlying Olpiro beds. Overall, their augite is higher in Fe than the Olpiro Tuff and lower in Na than the Ndolanya tuffs (Figure 10). A distinctive low-Na and intermediate- to high-Fe augite population occurs in all Naibadad Beds tuffs sampled, which is not observed in tuffs from the other beds. The Olpiro Tuff is limited to lower-Fe, higher-Ti augite. Ndolanya Beds augite trends towards higher Na and Fe but partially overlaps the Olpiro Tuff at the lower-Fe end of this trend. Many Naibadad Beds tuffs contain populations of higher-Na augite in addition to their distinctive low-Na augite. Naibadad Beds feldspar is dominated by anorthoclase (mostly high-K, Figure 11), unlike the more K-rich sanidine of the Ndolanya tuffs or the intermediate plagioclase of the Olpiro Tuff. The high-Ti and low-Al titanomagnetite compositions of the Naibadad Beds tuffs also separate them from both the Ndolanya Beds and Olpiro Tuff (Figure 13).
Figure 13. Electron probe microanalysis (EPMA) of titanomagnetite compositions. (A) MnO versus FeO. This plot shows the higher Fe and lower Mn composition of the Ndolanya Beds titanomagnetite (blue symbols) and reveals likely detrital contamination of Ndolanya Beds-derived titanomagnetite into specific Naibadad Beds tuffs (e.g., LT16-23.05). Olpiro tuffs (green symbols) are intermediate. (B) TiO2 versus FeO shows a similar separation between Naibadad and Ndolanya. (C) TiO2 versus Al2O3. Some of the Ndolanya Beds oxides trend towards magnetite, while Olpiro tuffs trend towards higher Ti and Al.
An additional key difference between the beds is the abundance of hornblende and plagioclase in the Olpiro Tuff (both are rare to absent in all other units). Andradite (melanite) garnet is abundant in most Ndolanya and Laetolil beds tuffs (although andradite from each of those beds is compositionally distinct; see McHenry, Reference McHenry and Harrison2011) but absent from the Naibadad and Olpiro beds tuffs (Hay, Reference Hay, Leakey and Harris1987), except for the basal Naibadad Beds tuff. While glass is not preserved in all samples, where present the Naibadad and Olpiro beds tuffs are distinct from each other, with the former characterized by rhyolitic or high-Si trachytic glass and the latter with trachytic to phonolitic glass with higher Al2O3 (Figure 12). No Ndolanya samples from the current study preserved glass, but Hay (Reference Hay, Leakey and Harris1987) reported a nepheline phonolite composition.
Distribution of Olpiro Tuff
Except for the type locality for the Olpiro Beds (Locality 16, based on Hay, Reference Hay, Leakey and Harris1987), we were unable to identify the Olpiro Tuff at any of the locations at which the unit was previously reported (Localities 1, 2N or 2S, or 23). Either no tuff was found in the currently accessible exposures (e.g., Locality 23), or sampled tuffs were mineralogically or chronologically dissimilar (e.g., Localities 1, 2N, and 2S) from the Olpiro Tuff described by Hay (Reference Hay, Leakey and Harris1987) and from our sampling of the tuff at Locality 16.
Hay (Reference Hay, Leakey and Harris1987) described an assemblage of augite, hornblende, and anorthoclase/plagioclase feldspar for the Olpiro Tuff, in which andradite (melanite) garnet is absent. Samples of the Olpiro Tuff from Locality 16 in the current study are consistent with this, with augite higher in Mg than in tuffs from the other beds at Laetoli, abundant hornblende, and mostly plagioclase feldspar. Where glass is present, it is trachytic. However, this assemblage is not observed in samples from the Olpiro Beds elsewhere. In the case of Localities 1 and 2N, candidate Olpiro Beds tuffs were compositionally more like the older Ndolanya Beds, with high-Na augite, andradite garnet, and lack of hornblende.
Possible explanations for this discrepancy
One possibly explanation for this discrepancy is that some deposits previously assigned to the Olpiro Beds could be Ndolanya Beds exposures. The complex depositional, tectonic, and erosional history of the Laetoli area can confuse stratigraphic relationships, and the Ndolanya and Olpiro beds have some similar properties (e.g., both are dark and clay-rich). This is the most likely explanation for Locality 1, where the dark tuff near the top of the “Olpiro” Beds (described in Hay, Reference Hay, Leakey and Harris1987) clearly matches the Ndolanya Beds in both mineral composition (LT16-1.01) and age (sample L16/1-1). The same is true for a candidate Olpiro Beds exposure at Locality 2N (tuff: LT16-2.05; age: L16/2-1sw), while a second horizon from the same exposure yielded a Naibadad Bed age (L16/2-1w).
A second reason could be confusion in identifying Olpiro Tuff versus upper Naibadad tuffs, without supporting mineral compositional data. This is a possible explanation for the Olpiro Beds at Locality 2S (proposed by Manega, Reference Manega1993). The tuff indicated in Manega’s (Reference Manega1993) stratigraphic section as the Olpiro Tuff is mineralogically consistent with the Naibadad rather than Olpiro beds (sample LT16-2.14: Fe-rich augite, anorthoclase rather than plagioclase, and no hornblende), and yields a Lower Naibadad Beds age (sample L16/2s-4, 2.189 ± 0.004 Ma).
It should be noted that neither of these explanations negates the existence of the Olpiro Beds, as previously described by Hay (Reference Hay, Leakey and Harris1987) and excavated by Ndessokia (Reference Ndessokia1990). The Olpiro Tuff, described here (and by Hay, Reference Hay, Leakey and Harris1987) for Locality 16, may not be present at all sites corresponding to the Olpiro Beds. It should also be noted that the published age of the Olpiro Tuff (Deino, Reference Deino and Harrison2011) places it at ca. 2.057 ± 0.015 Ma, which overlaps the ages obtained for the Upper Naibadad Beds in the current study.
Correlations within Laetoli
While most exposures of the Naibadad Beds contain one or two distinct tuffs, Locality 9/9W exposures include multiple tuffs separated by calcretes indicating gaps in time between deposition. Compositionally, the lowest tuff in the series (geochemical sample LT16-9.04, 40Ar/39Ar sample L16/9w-1) includes andradite (melanite) garnet, which is absent or rare in overlying tuffs. This tuff has trachytic glass, with an age of 2.175 ± 0.005 Ma. An andradite and trachytic glass-bearing tuff is also recognized in the lower Naibadad Beds at other sites in the Laetoli area, including Locality 2N (geochemical sample LT16-2.06, 40Ar/39Ar sample L16/2-2c: 2.164 ± 0.006 Ma).
Based on their 40Ar/39Ar ages, as outlined above, the Naibadad Beds tephra sampled for this study can be divided into three distinct age groups, within which samples mostly overlap within error. However, beyond the distinctive andradite-bearing mineral assemblage observed in the lower tuff described above, few systematic differences in mineral assemblage, mineral composition, or major element glass composition were observed among the different age groups. Some (but not all) of the youngest Naibadad Beds samples, especially LT16-23.02 (equivalent to Ar sample L16/23-2) and LT16-23.04 (top and bottom of the same unit at Locality 23), plot apart from the rest of the Naibadad Beds in FeO in feldspar (lower), Al2O3 in augite (lower), and SiO2 in glass (higher). Both of these samples also contain minor aenigmatite, a mineral absent from all other Laetoli-area tuffs analyzed but present in the Naabi Ignimbrite and related units in the OGCP cores at Olduvai. As noted above, however, we did observe a slight progressive decrease in Ca/K ratio of feldspars analyzed for 40Ar/39Ar dating across the three age groups (Figure 9).
Possible correlations to Olduvai
Similarities in composition between the Naibadad Beds and Ngorongoro-derived tuffs, and the lack of other local rhyolitic and trachytic volcanoes of similar age, support a Ngorongoro source for the Naibadad Beds tuffs and tuffaceous material (Deino, Reference Deino and Harrison2011; McHenry, Reference McHenry and Harrison2011). Ngorongoro-derived rhyolites and trachytes are also preserved at Olduvai, at the base of outcrop exposures in Lower Bed I in the western part of Olduvai Gorge (Naabi Ignimbrite, Coarse Feldspar Crystal Tuff (CFCT)), along the 1st fault at the far eastern end of Olduvai (Naabi, CFCT, and Orkeri units; McHenry et al., Reference McHenry, Mollel and Swisher2008), and within the OGCP cores (McHenry et al., Reference McHenry, Stanistreet, Stollhofen, Njau, Toth and Schick2020).
While overall similar in age, assemblage, and composition, no volcanic deposits from Olduvai outcrops or the OGCP cores appear to have exact correlatives to the Naibadad Beds tuffs analyzed in this study. The Naabi Ignimbrite, dated at Olduvai to 2.033 Ma (Deino et al., Reference Deino, Heil, King, McHenry, Stanistreet, Stollhofen, Njau, Mwankunda, Schick and Toth2021), has distinctively high-Fe anorthoclase feldspar and Fe-rich augite, along with minor aenigmatite. Similarly aged tuffs in the Naibadad Beds (e.g., samples LT16-23.02, 23.04, and 9.09) have different mineral compositions or assemblages. While minor aenigmatite was identified in both samples LT16-23.02 (Ar sample L16/23-2) and LT16-23.04, and their glass compositions are similar, their feldspar and augite compositions do not overlap with Naabi compositions from Olduvai (Figures 14 and 15). Upper Naibadad Beds sample LT16-9.09 is compositionally similar to the Orkeri Tuff (dated to 2.047 Ma; Deino et al., Reference Deino, Heil, King, McHenry, Stanistreet, Stollhofen, Njau, Mwankunda, Schick and Toth2021) in its overall assemblage, but its feldspar has higher K and lower Ba.
Figure 14. Electron probe microanalysis (EPMA) of glass and augite compositions, compared to Olduvai. Plots include data from three Upper Naibadad samples (orange and purple symbols) from the current study and published EPMA data for Olduvai outcrop tephra (black symbols; McHenry et al., Reference McHenry, Mollel and Swisher2008) and OGCP core samples from Cores 2A and 3A (blue symbols; McHenry et al., Reference McHenry, Stanistreet, Stollhofen, Njau, Toth and Schick2020). (A) FeO versus Al2O3 (glass). (B) SiO2 versus Al2O3 (glass). (C) FeO versus Al2O3 (augite, low-Al range only). Most glass shards from Naabi Ignimbrite-aged sample LT16-23.02 overlap with the Naabi ignimbrite (outcrop) and Naabi samples from Core 2A. However, as shown in C, the augites of this sample are compositionally unlike the Naabi.
Figure 15. Electron probe microanalysis (EPMA) of feldspar compositions compared to Olduvai. Plots include data from three Upper Naibadad samples (orange and purple symbols) from the current study and published EPMA data for Olduvai outcrop tephra (black symbols; McHenry et al., Reference McHenry, Mollel and Swisher2008, Reference McHenry, Stanistreet, Stollhofen, Njau, Toth and Schick2020) and OGCP core samples from Cores 2A and 3A (blue symbols; McHenry et al., Reference McHenry, Stanistreet, Stollhofen, Njau, Toth and Schick2020). (A) FeO versus K2O (feldspar). (B) BaO versus K2O (feldspar). (C) Feldspar ternary diagram. These plots demonstrate that none of the Naibadad Beds samples contains the distinctive high-Fe anorthoclase composition characteristic of the Naabi Ignimbrite in outcrop or its correlative units in the OGCP cores.
Also, many of these units at Olduvai were ignimbrite or volcaniclastic-flow deposits, presumably emplaced directly from Ngorongoro, uphill from Olduvai. At Laetoli, while some Naibadad Beds tuffs are coarse grained (coarse lapilli/pumice bearing), they are direct fallout or fluvially reworked deposits, rather than ignimbrites or flows.
Results from the OGCP cores show that Olduvai was a depocenter prior to and during emplacement of the Naibadad Beds. This is demonstrated by a prominent lacustrine interval at Olduvai situated between two pulses of Ngorongoro volcanic activity (Stanistreet et al., Reference Stanistreet, Stollhofen, Deino, McHenry, Toth, Schick and Njau2020). This activity falls within the timeframe of Naibadad Beds deposition, yet lacustrine strata are not observed at Laetoli. The Olduvai Basin is also at present more than 300 meters topographically lower than Laetoli and, given the absence of known major faults between the OGCP drill sites and Laetoli, this topographic relationship was likely the same during deposition of the Naibadad Beds. The lahars and ignimbrites deposited at Olduvai directly from Ngorongoro volcano would have been prevented from reaching the Laetoli area due to the shielding effect of Lemagurut volcano, which was fully formed by 2.3 Ma (Figures 9 and 16). These deposits would also largely have been excluded from reaching Laetoli from Olduvai by the requirement of traveling uphill (although not impossible, particularly for pyroclastic flows).
Figure 16. (A) Map showing likely flow directions for Ngorongoro tephra emplacement at Olduvai and Laetoli. Lemagurut and Sadiman volcanoes both pre-date the Naibadad Beds and would have provided an obstacle for the emplacement of Ngorongoro-derived flows from the direction of Olduvai. (B) Chronostratigraphy of the Laetoli area from Upper Ndolanya Beds through the Upper Naibadad Beds, based on the 40Ar/39Ar ages reported in this study, compared to the Ngorongoro Formation (Olduvai cores) (after Deino et al., Reference Deino, Heil, King, McHenry, Stanistreet, Stollhofen, Njau, Mwankunda, Schick and Toth2021). Bed duration encompasses the range of ages from the oldest- to youngest-dated sample attributed to a chronostratigraphic group in this study, including uncertainties. The Ngorongoro Formation at Olduvai (upper and lower pulses) overlap the age ranges of the Upper and Middle Naibadad Beds at Laetoli.
We conclude that the Naibadad Beds tuffs were likely emplaced as fallout directly from Ngorongoro volcano (sometimes locally reworked), rather than having arrived from the Olduvai Basin, and that the Laetoli area was a shielded reentrant in the broader basin paleotopography. Prevailing winds, both now and inferred for the Pleistocene (based on adhesion ripples in a ca. 1.8-Ma tuff at Olduvai; Hay, Reference Hay1976), are northeasterly to easterly, consistent with transport of volcanic plumes from Ngorongoro in the direction of both Olduvai and Laetoli.
The Naibadad Beds tephra could have instead been emplaced by a tephra cloud following a southern route, traveling from Ngorongoro volcano over the summits of Sadiman and Lemagurut volcanoes or from the southern rim of Ngorongoro volcano around the south sides of these volcanoes and into the Laetoli area. The Eyasi escarpment, which could potentially divert Ngorongoro-derived drainages to the south, most likely post-dates the Naibadad Beds since most of the surface extension of the Eyasi basin occurred within the last 1.5 Ma (Le Gall et al., Reference Le Gall, Nonnotte, Rolet, Benoit, Guillou, Mousseau-Nonnotte, Albaric and Déverchère2008; Fletcher et al., Reference Fletcher, Abdelsalam, Emishaw, Atekwana, Laó‐Dávila and Ismail2018).
Hay (Reference Hay, Leakey and Harris1987) established that the Naibadad Beds tuffs were deposited in channel fills cut deep into the Ndolanya and Laetolil beds in places, along trends consistent with modern drainages. This implies that the modern southwest-trending drainage was already established by ca. 2.2 Ma. In our model, tephra fallout from Ngorongoro deposited at Laetoli and on the slopes of neighboring volcanoes, and/or volcaniclastic flows exiting the southern rim of Ngorongoro traveling southwest around Sadiman and Lemagurut, were redeposited fluvially as the Naibadad Beds, filling channels previously incised into the Ndolanya and Laetolil beds.
The lack of tephra with the exact same compositions in the Olduvai cores suggests that any tephra fallout cloud was of limited geographical extent, supplying tephra to the south but not the west. This could explain why, while similar, the Ngorongoro-derived tuffs preserved in the Olduvai basin are not the exact same units deposited in the Naibadad Beds at Laetoli. Different eruptions might have followed different paths (west, into Olduvai, or south, into the headwaters of Laetoli drainages). This also explains why some of the Naibadad Beds tuffs are coarse, but do not show surge, flow, or lahar features, why Naibadad Beds tuff deposits are thicker farther away from Lemagurut, and why they often fill incised channels. This could also explain a greater degree of detrital contamination in the Naibadad Beds observed in the mineral assemblages, as they were reworked in drainages containing materials derived from Ndolanya and Laetolil beds.
Implications and conclusions
The Naibadad Beds overlap in time and composition with the Ngorongoro Formation volcaniclastics described in the OGCP cores. The lacustrine interval between the two pulses of Ngorongoro volcanism in OGCP Core 2A occurred during Naibadad times, perhaps during a lull in volcanic activity not preserved at Laetoli.
If Eyasi was not present, the Olduvai basin might have been the local depocenter for the Laetoli area and the most local source of non-spring freshwater—not just during Naibadad Beds times, but earlier. The OGCP cores preserve sediments older than the Naibadad (as old as 2.24 Ma; Deino et al., Reference Deino, Heil, King, McHenry, Stanistreet, Stollhofen, Njau, Mwankunda, Schick and Toth2021), and a seismic survey of the Olduvai basin reveals that the basin is substantially deeper and could contain a sedimentary record about twice as thick as was sampled in the 2014 cores (Lu et al., Reference Lu, Hanafy, Stanistreet, Njau, Schick, Toth, Stollhofen and Schuster2019). This older basin fill, not yet cored, could include Ndolanya or even Laetolil beds-aged sediments, as Laetolil Beds-equivalent materials are known to underlie the Olduvai Beds in the southwestern corner of the Olduvai basin (Hay, Reference Hay1976).
If the Naibadad Beds include eruptive events from Ngorongoro not recorded in the OGCP cores, then they will need to be considered to obtain the complete record of Ngorongoro explosive volcanism (and evolution of the Ngorongoro caldera system). The dates reported here can help constrain this.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/qua.2025.17
Acknowledgments
The Laetoli field project was funded by the National Geographic Society, Leakey Foundation, and National Science Foundation [BCS-0309513, BCS-1350023, BCS-1322017, and EAR-2020044]. Permits were provided by the Tanzanian Commission for Science and Technology (COSTECH) and the Tanzanian Antiquities Division in Dar es Salaam. This manuscript was greatly improved thanks to feedback from Associate Editor Curtis Marean, reviewer Craig Feibel, and an anonymous reviewer.
Open access
