Abstract

Approximately 2.75 million years ago, the Turkana Basin in Kenya experienced environmental changes, including increased aridity and environmental variability. Namorotukunan is a newly discovered archaeological site which provides a window into hominin behavioral adaptations. This site lies within the upper Tulu Bor and lower Burgi members of the Koobi Fora Formation (Marsabit District, Kenya), presently a poorly understood time interval due to large-scale erosional events. Moreover, this locale represents the earliest known evidence of Oldowan technology within the Koobi Fora Formation. Oldowan sites, older than 2.6 million years ago, are rare, and these typically represent insights from narrow windows of time. In contrast, Namorotukunan provides evidence of tool-making behaviors spanning hundreds of thousands of years, offering a unique temporal perspective on technological stability. The site comprises three distinct archaeological horizons spanning approximately 300,000 years (2.75 − 2.44 Ma). Our findings suggest continuity in tool-making practices over time, with evidence of systematic selection of rock types. Geological descriptions and chronological data, provide robust age control and contextualize the archaeological finds. We employ multiple paleoenvironmental proxies, to reconstruct past ecological conditions. Our study highlights the interplay between environmental shifts and technological innovations, shedding light on pivotal factors in the trajectory of human evolution.

Introduction

The remarkable ability of humans to inhabit nearly every terrestrial ecosystem is a result of the synergy between biological and technological evolution1. The long term significance of our evolutionary relationship with technology arises from the discovery of stone artifacts within Plio-Pleistocene sediments2,3. This study provides new evidence from the Turkana Basin for the relationship between climatic and environmental shifts and the development of stone tool technologies by hominins. The earliest known assemblages of Oldowan artifacts ( ~ 2.9-2.6 Ma) are confined to four localities in eastern Africa4,5 (Fig. 1). The earliest known localities provide insights into technological behaviors at single time horizons in the deep past. Despite advances in our understanding of early human technology, the specific mechanisms through which environmental changes influenced technological evolution in the earliest Oldowan remain poorly understood. Demonstrating how early hominins adapted their tool-making practices in response to changing environments would provide new insights into the evolutionary pressures that shaped these innovations. Here, we describe multiple assemblages of stone artifacts from well-constrained horizon, with age estimates at 2.75, 2.58, and 2.44 Ma, from the Koobi Fora Formation in the northeastern portion of the Turkana Basin (Paleontological Collection Area 40, archeological site Namorotukunan, National Museums of Kenya ID: FwJj 52). Detailed knowledge of environmental patterns within this region allows us to explore the interplay between periods of environmental change and the presence of an early Oldowan techno-complex that first evolved in the late Pliocene, and emphasizes temporal continuity throughout the Early Pleistocene, in eastern Africa.

Fig. 1: Map of Turkana Basin with the Namorotukunan Archeological Site and timeline of currently known events in the Plio-Pleistocene.

a Geographical context of the Koobi Fora Formation (red stripes), the paleontological collection area 40 (green square), and the location of the site of Namorotukunan (black dot); [map produced Natural Earth and NOAAA ETOPO 2022[95](https://www.nature.com/articles/s41467-025-64244-x#ref-CR95 “NOAA National Centers for Environmental Information. ETOPO 2022 15 Arc-second global relief model. https://doi.org/10.25921/fd45-gt74

(2022).“)]; b Stratigraphic context of the Koobi Fora Formation highlighting members and key volcanic ash marker levels, yellow bars refer to the age of archeological horizons (tephrostratigraphy after McDougall et al.96); c A chronology of key Plio-Pleistocene hominins from the East African Rift System (EARS)11,74,97,[98](https://www.nature.com/articles/s41467-025-64244-x#ref-CR98 “Villmoare, B. et al. New discoveries of Australopithecus and Homo from Ledi-Geraru, Ethiopia. Nature 1–7 https://doi.org/10.1038/s41586-025-09390-4

(2025).“) d A chronology and key localities associated with hominin lithic technology3,6,12 (images of Nyayanga provided by E. Finestone; images of Lomekwi and BD1 based on 3D models; artifact images are for representation and not to scale) and the investigations at Namorotukunan: red arrows represent the artifact levels in the archeological excavations (photos DRB), and colored circles (lettered A-G) represent geologic sections investigated to develop a synthetic stratigraphic column (presented in Figs. 2 and 3).

The earliest phases of tool manufacture, dating back to over 3.0 million years ago6 (although see Archer7), highlight percussive technology, which is ubiquitous in hominin records and shared with other primates8,9. Tool use, associated with extractive foraging, is a recurring trait in some extant primates10. The oldest systematic production of sharp-edged stone artifacts, known as the Oldowan, is found in the hominin behavioral record at eastern African sites: Ledi-Geraru and Gona in the Afar Basin (2.6 Ma), Ethiopia, and Nyayanga in western Kenya (2.6-2.9 Ma)3,11,12,13.

In this study, we describe the earliest known Oldowan technologies and their paleoenvironmental context in the Koobi Fora Formation (Fig. 1). The presence of butchery marks on bones within the Namorotukunan assemblage underscores the role of sharp-edged tools in the foraging behavior of these hominins and suggests that the development of Oldowan technology was associated with the exploitation of resources mediated by tool use14. Open habitats such as savannas and grasslands expand at the end of the Pliocene and may have facilitated an adaptive shift in hominins towards a regular exploitation of foods requiring the use of tools (e.g., USOs, bone marrow, meat)15,16. Here we examine three distinct temporal horizons containing evidence of sharp-edged tool technology spanning ~300,000 years. The consistent technological approaches across this interval suggest an enduring technological adaptation in the hominin lineage throughout the late Pliocene and earliest Pleistocene.

Results

Geological context

The Koobi Fora Formation (KF Fm.) refers to the sedimentary strata of the Omo Group on the eastern side of Lake Turkana and lies unconformably on Miocene and Pliocene volcanic rocks and associated sediments17,18. The Omo Group encompasses these and other Plio-Pleistocene lacustrine and fluvial deposits that fill a series of alternating half-grabens of the Omo-Turkana Basin, which in turn is part of the larger East African Rift System (EARS19,). Excellent age control throughout the Omo Group (4.02 to ~0.75 Ma) is provided by 40Ar/39Ar dates on volcanic ash layers, K/Ar ages on intercalated basalts and paleomagnetic stratigraphy20. The KF Fm. is divided into eight members (see Fig. 1), each delineated by a dated volcaniclastic tephra marker bed at the base of the member (except the basal Lonyumon Mbr19.). In Area 40, where Namorotukunan is located, sedimentary strata are estimated to range roughly between 4.3 to 1.6 Ma21,22, with a gap between 3.0 −2.5 Ma represented by the Burgi Unconformity21,23,24. Above this unconformity, a series of lake clays indicate a local change in tectono-sedimentary patterns and the Paleo-Lake Lorenyang transgression in the Turkana basin21,24. To the west of Namorotukunan, across the North Gele Fault19, a younger sedimentary succession is exposed that includes at least the Upper Burgi and KBS Members, as evidenced by the presence of the KBS Tuff within that sequence21.

Geological investigations focused on a 46-meter-thick stratigraphic interval examined in seven sections (Fig. 2). These sections are assigned to local geological units used for mapping around Namorotukunan (Supplementary Fig. 1). The lowermost unit in this framework is a tephra level (TB), identified as the Tulu Bor Tuff21 (Supplementary Table 1). The uppermost level is a sequence of lake clays associated with the hill locally known as Namorotukunan, from which the archeological site name is derived (L2b). This lithological succession comprises three distinctly different sediment packages separated by discrete boundaries (Supplementary Table 2).

Fig. 2: Simplified geology (upper part of the figure), paleoriver network, and the location of archeological sites.

The beds are slightly tilted towards NE, exposing older rocks towards the southwest and younger rocks in the Namorotukunan Hill in the north. Several linear horizons are characterized by coarse clastic sediments (e.g., cobbles, pebbles, and sands) with distinct bedding features (e.g., cross-through bedding) representing paleo-channels belonging to paleo-river systems whose flow direction is marked as light blue arrows. A series of geological sections (A-G) (lower part of the figure) were used to develop a synthetic log and the magnetostratigraphic context. The main lithologies include red paleosols (R1–R4), sands and poorly cemented sandstones (S1–S3), gray to gray-brown paleosols (P1–P6), conglomerates and gravels (C1–C7), lacustrine clays (L1–L2b), lacustrine sands with ripple marks (LS1), and diatomite-rich lacustrine clays (LD). These lithologies reflect a variety of depositional environments, from fluvial and alluvial systems to lacustrine settings. Green bars reflect straight line distances between stratigraphic sections. Image derived from UAV imagery collected on site.

The oldest sediment package is characterized by red silts, paleosols, and sands overlying the Tulu Bor Tuff (TB). It consists of poorly developed paleosols with root traces and carbonate-rich silts intercalated with sands and fine gravels. This package is interpreted as a fluvial floodplain (R1 to R4, Fig. 2 and Supplementary Fig. 2; Supplementary Table 2).

The second package is composed of gravels and paleosols, indicating paleochannels and swamp clays associated with a fluvial system (C1-2, C3, C6), characterized by alternating channel and overbank with intervals of paleosols, interpreted as belonging to a floodplain (S3 to C7). All three artifact-bearing horizons are situated within this sedimentary package. The lower part of this interval contains darker paleosols (P0-P3) rich in carbonate concretions and rhizomes, which alternate with sandy gravel deposits that preserve sand casts of plant roots and rhizomes (Supplementary Fig. 2). Brownish-red, poorly developed paleosols characterize the upper part (P4-P6). Some of the upper paleosol levels (e.g., P5) feature large irregular desiccation cracks (Supplementary Fig. 2), filled with well-cemented sands from the overlying bed (e.g., C6). Cut and fill sequences are limited to three local paleo-rivers corresponding to levels C1-2, C3, and C6. The uppermost cut (C6) is the most laterally extensive and shows characteristics of a meandering river. In some areas, this channel eroded underlying sediments down to the paleosol level P3. These cut and fill structures, representing larger paleoriver features, are highly localized and transition laterally into sandy-silty beds interpreted as overbank deposits.

The artifact-bearing horizons occur in the paleoriver system, where currents in the main channel of the ancient, braided river systems were robust enough to transport cobbles and pebble-sized clasts that served as the source materials for the stone tools. The paleoriver system included low-energy environments, where slack water and point-bar deposits as well as floodplain soil surfaces provided the geological context for the preservation of artifacts and faunal remains.

The uppermost and third sedimentary package lies above paleosol level P6 and consists of lake clays (L1 to L2b, see Fig. 2, and Supplementary Fig. 2), occasionally featuring sandy units (e.g., LS1) or diatom-rich sediments (e.g., LD) and a single level containing frequent catfish fossil remains (LS1-L2a transition). Previous investigations conducted on the slopes of Namorotukunan Hill (see Fig. 7C in Baldes et al.24) identified the Burgi Unconformity at the base of the L1 bed. However, no age constraints were provided to determine the extent of the time gap represented by this unconformity, leaving its duration, if any, unresolved.

Within the investigated Area 40 sedimentary sequence, four age markers were identified that allowed development of a comprehensive age model for the entire sequence (Fig. 3). The oldest of these markers is a tephra marker bed, which represents the base of the interval. This layer has been identified as Tulu Bor Tuff β21 and is dated by a 40Ar/39Ar feldspar age of 3.44 ± 0.02 million years22,25,26,27. Geochemical analyses (Supplementary Table 1) also confirm this as the Tulu Bor Tuff β.

Fig. 3: Age-depth models for Namorotukunan.

obtained by correlating the depth axis, represented by the polarity pattern obtained from Namorotukunan (vertical axis) with the age axis, represented by the Geomagnetic Polarity Timescale99 combined with additional geochronological data (horizontal axis), represented by paleomagnetic data represented by magnetic chrons, subchrons and excursions31 and lake phases described from the Turkana Basin28,47,100,101. Two scenarios are considered: a. scenario a assumes continuous sedimentation between gravel levels C7 and lake clays L1; b., represents the second scenario which places the Burgi Unconformity at the C7/L1 lithologic boundary, as previously suggested by Kidney21 and Baldes et al.24, providing slightly older estimates for sites NMT2 and NMT3. The scenario with the Burgi Unconformity provides two correlation options for the NZ2-RZ2 polarity interval, either with the Feni excursion (PL) or the Olduvai subchron (L). We favor the first option due to the absence of the KBS tuff in the lacustrine sediments corresponding with the NZ2 polarity zone. For lithological column and paleomagnetic polarity zone legend, see Fig. 2.

Paleomagnetic investigations of all the lithologies present, except for the unconsolidated coarse sands (R1-R4), revealed four distinct magnetic polarity intervals (Supplementary Fig. 3). The lowest part of the section, with a thickness of 16.5 meters, corresponds to a normal polarity zone (NZ1). Subsequently, there is a section ~20 meters thick (RZ1) that features reverse polarities. This is followed by a shorter normal polarity segment (NZ2), around 9 meters thick. The top of the Namorotukunan Hill, less than 1 meter thick, is characterized again by the beginning of another reversed polarity zone (RZ2). In most of Area 40 this unit has been removed by modern erosion.

The Tulu Bor Tuff forms the lower boundary of the sequence at 3.44 Ma, and the Paleo-Lake Lorenyang transgression28,29 marks its upper boundary at around 2.2 Ma. These two markers allow us to correlate the local magnetic polarity sequence observed at Namorotukunan with the established Geomagnetic Polarity Timescale (Supplementary Data 1, Supplementary Data 2 GPTS30). Specifically, NZ1 corresponds with the Gauss (C2An) Chron, RZ1, and RZ2 represent parts of the Matuyama (C2r) Chron, and NZ2 correlates with the Feni excursion (Formerly Réunion excursion)31 or, alternatively, with the Olduvai sub-chron. The Mammoth (C2An.2r) and Kaena (C2An.1r) reverse polarity subchrons from the lower part of the Gauss Chron were not found. These are either located in the sandy intervals S1 and S2, (incompatible with paleomagnetic analysis) or these sandy intervals are associated with disconformities. This limitation impacts our estimations of sedimentation rates in the lower interval but critically, does not significantly affect the age estimation of the artifact levels investigated here.

The studied interval includes the Upper Burgi Unconformity (UBU), a surface interpreted in parts of the Koobi Fora Formation as marking a depositional hiatus of several hundred thousand years24. However, the UBU is not clearly expressed in Area 40, where the transition from fluvial to lacustrine facies is conformable and no field evidence for an erosional surface is present. The overlying lake clays match Upper Burgi lithologies described by Baldes24 and the absence of the KBS Tuff in these well-preserved sediments—despite its occurrence ~600 m west of Namorotukunan across the North Gele Fault21—suggests that the Namorotukunan section does not reach the KBS Member. If a hiatus is present, it must be confined to a narrow interval within RZ1, implying a much shorter duration than has been proposed for central Koobi Fora. We thus consider two possibilities: Scenario A, in which sedimentation is continuous, and Scenario B, in which the UBU represents a brief interruption. Both scenarios securely place the archeological levels in the late Pliocene–early Pleistocene and do not alter our behavioral or chronological interpretations.

Considering the correlation of R1 with the Matuyama chron, we are presented with two correlation options for the NZ2 polarity interval: either aligning it with the Feni excursion or the Olduvai subchron. We favor the first option—the correlation with the Feni excursion—due to the absence of the KBS Tuff in the lacustrine sediments corresponding to the NZ2 polarity zone. The KBS Tuff is an important and geographically expansive stratigraphic marker that, if present, would support a correlation with the Olduvai subchron. KBS has been described ~600 west of Namorotukunan, across the North Gele fault21, overlaying a thick succesion of lake clays, inconsistent with the section at Namorotukunan. Its absence in Namorotukunan suggests that the NZ2–RZ2 interval more likely corresponds to the Feni excursion.

In addition to the level of the Tulu Bor Tuff β (3.44 ± 0.02 Ma), paleomagnetic data provide three new chronological points for developing the age model for the studied section and obtaining an age estimate for the artifact-bearing horizons at Namorotukunan (Fig. 3). These points are the Gauss-Matuyama reversal (NZ1/RZ1) at 2.610 million years30, the onset of the Feni subchron (RZ1/NZ2) at 2.137 million years, and the termination of the Feni subchron (NZ2/RZ2) at 2.116 million years32.

We present two alternative age models based on the different correlations described above. These age models reflect two possible scenarios: the first scenario (Fig. 3a) assumes continuous sedimentation between gravel levels C7 and lake clays L1, while the second scenario (Fig. 3b) incorporates the Burgi Unconformity at the C7/L1 lithologic boundary, as previously suggested by Kidney21 and Baldes et al.24. This second scenario results in slightly older age estimates for sites Namorotukunan-3 (NMT3) and Namorotukunan-2 (NMT2).

In this paper, we use the first scenario, which provides more conservative (younger) age estimates. It is important to note that these scenarios do not affect the age estimation of the oldest archeological levels presented in this paper.

Archeological findings

Archeological research was carried out within the paleontological collection Area 40 at the Namorotukunan site (36.329 E, 4.399 N). Three excavations (in a ~ 5000m2 area), named Namorotukunan-1 (NMT1), Namorotukunan-2 (NMT2), and Namorotukunan-3(NMT3), corresponding to discrete lithological units S3-C1 (for NMT1), C2-P2 (for NMT2), and C3 (for NMT3) yielded a total of 1290 artifacts: NMT1 (n = 198), NMT2 (n = 775), and NMT3 (n = 317), along with associated fossils (Supplementary Table 3). Artifact-bearing horizons are concentrated in sands and fine gravels (Fig. 4). Artifacts associated with the C3 horizon, (NMT3) appear to have undergone minor post-depositional disturbance as indicated by the absence of the smaller fraction of artifacts and linear orientation for this 2.44 Ma archeological horizon (NMT3; see Supplementary Figs. 4, 5, 6,7). Abrasion on artifacts is present in some of the artifacts, yet thin sharp edges are preserved in the assemblages (% of assemblage exhibiting abrasion ranges from 18-22%). Geochemical and mechanical damage to the surfaces of many of the fossil bone specimens prevents a detailed taphonomic analysis of bone modification in the faunal assemblage. However, some specimens within the 2.58 Ma assemblage include well-preserved surfaces that possess butchery marks indicating that hominins used tools to extract high-quality dietary resources from large mammalian remains (see Supplementary Fig. 8).

Fig. 4: Wall sections of three main stratigraphic units.

The three archeological horizons (NMT3 dated to 2.44 Ma; NMT2 attributed to 2.60 Ma; and NMT1 attributed to 2.75 Ma). These archeological horizons are correlated to the composite stratigraphy for the Area 40 region. Wall sections depict the distribution of artifacts in these stratigraphic units. The right diagrams reflect the proportion of artifacts in major typological categories102 and the major rock types present in these archeological assemblages (see Supplementary Data 3). A high prevalence of chalcedony is evident in all the archeological horizons. Yellow circles represent individual stone artifacts greater than 2 cm, white-blue stars represent faunal specimens recovered in situ during excavations (see Supplementary Fig. 6), and the red diamond represents the cut-mark bone described in Supplementary Fig. 8. Note that fauna was poorly preserved in the older and younger time horizons.

Detailed Descriptions Of The Archeological Finds

Artifacts from Namorotukunan exhibit many features that are diagnostic of anthropogenic conchoidal fracture (e.g., prominent bulbs of percussion, clear striking platforms, contiguous flake scars; although see Proffitt et al.9,33) and are dominated by smaller flakes and simple cores. High proportions of sharp-edged flakes and fragments, comprising 94.2 to 79.4% of the assemblages, indicate sharp edges were the likely focus of this technology. Early Oldowan assemblages tend to have flakes that are of similar dimension to their cores (as opposed to Acheulean sites where cores are much larger) and have fewer flake scars on cores (as compared to younger Oldowan assemblages3,12 (Fig. 5). Various technological attributes indicate that the Namorotukunan assemblages are more similar to other early Oldowan assemblages (e.g., Bokol Dora 1; Nyayanga) than assemblages younger than 1.8 Ma (Fig. 5; and Supplementary Fig. 10,11). Evidence of an understanding of fracture mechanics (e.g., relationship between platform angles and other technological features) is similar between hominins that produced the assemblages at Namorotukunan and older Oldowan assemblages (Supplementary Fig. 11). Cores from the Namorotukunan assemblages also show similar patterns to those seen at other early Oldowan assemblages ( > 2.3 Ma) where cores are rarely rotated during production (% Cores that are Unifacial; Fig. 5). Although percussive tools are infrequent in the Namorotukunan assemblage, the appearance of heavily battered artifacts ( < 1% of assemblage; NMT2) confirms the presence of percussive activities (Supplementary Fig. 12). Comparisons of raw material selectivity demonstrate fine grained chalcedony at proportions that are significantly greater than its presence in nearby conglomerates (Supplementary Fig. 13). We note that although hominins at Namorotukunan appear to be selective in their selection of specific rock types, there is little evidence of transport of stone. Gravels (braided river system deposits) that carried cobbles suitable for artifact manufacture were present in coarser grained sediments directly adjacent to all archeological sites. This level of selectivity is evident in all three assemblages and suggests a persistent preference for materials that fracture consistently (Supplementary Fig. 14).

Fig. 5: Early Oldowan technology.

A Principal components analysis of major technological features of several of the earliest stone tool industries currently known (Supplementary Data 6, 7). PCA scores were used to calculate a K-means cluster analysis. The shaded polygons (convex hulls) are the results of this K-means cluster analysis. In addition to early archeological assemblages, we include here assemblages made unintentionally by both Capuchins (SCNP33) and long-tailed macaques (Lobi Bay9). B The analysis of eigenvectors indicates that most Oldowan (and primate) assemblages are distinguished from Lomekwi by their smaller size and relative infrequency of cores. Most early Oldowan assemblages cluster together based on the relative size of flakes compared to cores. C Technological variables describing core forms analyzed from 3D models of cores (See “Methods”). Boxplots represent the interquartile range of data. Whiskers represent the upper and lower quartile.; In numerous variables the Namorotukunan assemblage is similar to early Oldowan ( > 2.0 Ma) assemblages but significantly different from the Lomekwi 3 assemblage as well as the later Oldowan and early Acheulean assemblages (Supplementary Figs. 9, 10, 11).

Paleoenvironmental proxies

The stratigraphic sequence in Area 40 of the KF Fm provides a framework for examining the paleoecological conditions from ~3.44 to ~2.0 million years ago. The existence of paleosols, along with accompanying sediments and fauna, facilitates a multiproxy reconstruction of the fluctuating environmental conditions within this period. Pedogenic carbonates, plant wax biomarkers, and phytoliths provide detailed information (Fig. 6, and Supplementary Fig 15 and Supplementary Data 4) on the physiognomic structure of the ecosystems across the late Pliocene and early Pleistocene in the region. The range of values from some paleoenvironmental proxies indicate the presence of diverse habitats within some temporal intervals (e.g., δ13C values of pedogenic carbonates −10 to −2 ‰). However, when fitting linear estimates through the data, there are coincident changes that suggest an increase in C4 vegetation and a decrease in water availability in the gravel and paleosol interval (2.7-2.2 Ma; Supplementary Fig. 16). This is supported by increases in the δ13C values of C31 and C33 n-alkanes that indicate an increase in C4 vegetation and likely, open habitats, in the gravel and paleosol interval.

Phytolith analysis indicates an overall decrease in grass phytoliths that characterize the gravel and paleosol interval (Fig. 6). High incidences of palm and sedge phytoliths before these red silts, paleosols, and sands interval, indicate the presence of substantial standing water in these habitats (3.4 – 2.6 Ma) that is notably absent in the younger sediments, the gravel and paleosol interval (2.6-2.2 Ma). Significant increases in microcharcoals, associated with landscape-scale wildfires, point to a more seasonally arid habitat beginning around the onset of the gravel and paleosol interval (2.7-2.6 Ma; Fig. 6). Seasonality probably played a key role, as shown by investigations of paleo-fire frequency on more recent timescales in Africa34. Phytoliths from the 2.7-2.2 Ma time interval (Fig. 6) show increased shrub-derived morphotypes and decreased grass phytoliths, further suggesting a shift towards more seasonally arid habitats. This is further supported by the presence of phytoliths of C4 Chloridoideae grasses suggesting xerophytic habitats consistent with arid conditions at this time.

Chemical index of alteration (CIA) and mean annual precipitation (MAP) was calculated based on bulk geochemistry of paleosol samples (see “Methods” and Supplementary Note 2). High CIA values indicate removal of labile cations (Ca2+, Na+ and K+) relative to Al3+ during the chemical weathering of silicate minerals under wet climatic conditions and low values indicate absence of the chemical weathering under dry climate35,36. CIA values (51 to 64) indicate incipient to moderately weathered paleosols. Low CIA values (<55) between 2.9 and 2.7 Ma indicates dry climate (Fig. 6). MAP calculated using geochemical climofunctions for paleosols varies between 231 and 855 mm/year, broadly classifying the climate as semi-arid to semi-humid. MAP follows a similar trend as of CIA and lowest MAP is recorded between 2.7 and 2.6 Ma indicating an increased aridity at this time interval. The time frame between 2.7 and 2.2 Ma is associated with the relatively low MAP estimates. This pattern is reversed at the return of lacustrine environments (i.e., Paleo-Lake Lorenyang) at the top of the sequence (2.2-2.0 Ma) (Fig. 6).

Environmental magnetic analysis of bulk sediments confirms the reliability of our paleomagnetic approach. The analysis indicates that magnetite is the primary iron oxide, carrying magnetic properties in nearly all samples. However, at the onset of the gravel and paleosol interval (Supplementary Fig. 17), we detected a notable exception: evidence of a second iron mineral, hematite.

There are relatively few identifiable faunal specimens from the entire sedimentary sequence in Area 40 (Supplementary Data [5](https://www.nature