The ruins of the Medieval city of Dmanisi occupy the promontory at the confluence of the Mashavera and Pinezauri rivers in the lower Caucasus of southern Georgia (Fig. 1). A substantial settlement from the Bronze Age until the late Medieval Period, the entire surface of the promontory is covered by remains of masonry architecture, and a castle and fortress on the bedrock rise on the southern promontory (Fig. 1). Fortified by 80 m cliffs on two sides, and situated on the ancient silk road from the south, this was an ideal location for the ancient city. Remarkably, stratified Early Pleistocene deposits are preserved under the city ruins. Those deposits preserve thousands of animal fossils, lithic artifacts, and almost 80 fossils of early Homo erectus (Coil et al., 2020; Ferring et al., 2011; Gabunia and Vekua, 1995; Gabunia et al., 2000; Lordkipanidze et al., 2005, 2007, 2013; Margvelashvili et al., 2022; Mgeladze et al., 2011; Shelia et al., 2020; Vekua et al., 2002, 2005). Most of the recovered fossils and artifacts were from excavation Blocks 1 and 2 in the center of the Dmanisi promontory and were found in pipe and gully deposits of Stratum B1 (Fig. 1c).

From the first discoveries of Early Pleistocene fossils through the recovery of large samples of fossils in Blocks 1 and 2, it was clear that bone preservation in these deposits was excellent (Tappen et al., 2002, 2007; Tappen 2009). The association of many lithic artifacts with those bones, and also numerous carnivore remains and carnivore-modified bones gave rise to obvious questions about agents of bone accumulation; importantly, geologic, taphonomic, and spatial analyses virtually excluded any possible role of fluvial processes in bone accumulation (Coil et al., 2020; Lordkipanidze et al., 2007; Tappen 2009). Rather, burial of bones in pipe and gully deposits did point to the important role of rapid sedimentation and weak weathering regimes in bone preservation (Coil et al., 2020; Ferring et al., 2011). While excavation and recovery of many fossils and artifacts continued in Block 2, the surrounding areas of the promontory were unknown, save for shallow exposures in tests done in the early 1990s (Djaparidze et al., 1989).

As excavations in Block 2 continued, test excavations were conducted over several seasons at M5, located about 100 m upslope from the main excavations in Block 2 (Fig. 1). The 6 m thick M5 section has nine conformably superposed stratigraphic units (Ferring et al., 2011) including a slope facies of Stratum B1 that contrasts with the B1 pipe and gully deposits in Blocks 1 and 2. The striking differences in the geology of M5, as well as the first discovery of artifacts in Stratum A deposits there, showed that Dmanisi's geoarchaeological record was spatially and stratigraphically extensive. Many more geoarchaeological exposures over the promontory were clearly necessary, and a program of testing was implemented. The goals of those studies were to better document the stratigraphic and spatial extent of archaeological and paleontological materials, recover data on the sedimentary and geomorphic contexts across the promontory, and continue efforts to better define the relative age ranges of the serially deposited strata.

Here we present a summary of the geoarchaeological investigations conducted in concert with expanded block and test excavations over the past eight field seasons (Figure 1, Figure 2). These have provided substantial new data on the spatial extent, sedimentary environments, and formation contexts on the Dmanisi promontory. In addition, work at M5 includes dating the first sample of the Mashavera Basalt from the uppermost flows on the promontory surface, and petrographic study of the M5 section.

The archaeological site of Dmanisi (41° 20′14.09″ N; 44° 20′ 41.66″) is located about 65 km southwest of the capital city of Tbilisi in the Kvemo Kartli region of Georgia (Fig. 1). The site is situated at an average elevation of 910 m mean sea level (msl) on a promontory that is formed on two sides by the deeply entrenched Mashavera and Pinezauri rivers (Fig. 1). The southwestern extremity of this promontory connects with upland hills (up to ca. 1200 m msl) that have formed on faulted Upper Cretaceous tuffs and ignimbrites. Higher hills (to ca. 1500 m msl) to the south formed on diverse plutonic, volcanic, metamorphic, and sedimentary rocks (Supplementary Online Material [SOM] Fig. S1). The diverse lithology of those rocks is represented in the modern Pinezauri River gravels as well as among the diverse raw materials of Dmanisi's lithic artifact assemblages (Baena et al., 2010; Ferring et al., 2011; Gudjabidze, 2003; Mgeladze et al., 2011).

There is no evidence that either of the rivers flowed across the sloping surface formed by the Mashavera Basalt. Accordingly, the paleohydrology of the promontory was that of slope in a small catchment basin on which eolian sediments were serially deposited and variably affected by weathering and colluvial processes.

Since the last Mashavera lava flows, Dmanisi has been a promontory standing above the confluence of the two rivers. Today those gorges are deeply incised, with vertical walls exposing up to 80 m of Mashavera Basalt that filled the gorge and covered the Cretaceous rocks forming the pre-Mashavera bedrock spur at the rivers' confluence (SOM Figs. S2 and S3). Exposures of the contact of the Mashavera Basalt with underlying rocks and sediments show that at the time of the lava flows, the steep-walled river valleys were nearly in their present positions, and that the surrounding upland landforms were probably quite similar to those of today (SOM Fig. S2). The Mashavera Valley is narrow and deep for ca. 20 km downstream from the promontory, where it opens to gentler slopes and broad alluvial terraces. West of the promontory, the Mashavera Valley rises rather steeply to an open plateau with elevations of ca. 2000 m near Dmanisi Town and continues to rise toward the Djavakheti ridge about 80 km from the site (Nomade et al., 2016).

Since the discovery of Early Pleistocene fossils at Dmanisi in 1983, two different stratigraphic schemes have been used. The first stratigraphic scheme recognized six subhorizontal strata (labeled I–VI) exposed in the ‘Room 11’ paleontological excavations and the early excavations in Block 1 (Fig. 1; Djaparidze et al., 1989; Jöris, 2008). A different scheme identified two major stratigraphic units (A and B) based on the unconformity at the A–B contact observed in profiles in and around Block 1, as well as piping features that formed in A strata but were filled with B sediments (Gabunia et al., 2000). The M5 section resolved uncertainties created by the complex stratigraphy of Blocks 1 and 2 and resulted in a formal stratigraphy for the Dmanisi Formation, composed of Stratum A (A1–A4) and Stratum B (B1–B5; Ferring et al., 2011). Expanded excavations as well as a testing program over the promontory, described later, have all showed that the A1–B5 stratigraphy of M5 is applicable to at least 40,000 m2 of the promontory.

The presence of pipe features at Dmanisi was first recognized in Block 1, Room 11, and the Medieval cellar profile (D1B) north of Block I by Ferring and Swisher (Fig. 1; Gabunia et al., 2000) who noted their implications for both stratigraphy and formation processes. Subsequent exposure of pipes and related gully features in Block 2 showed that those deposits contained all of the hominin fossils and the vast majority of vertebrate fossils from that block (Coil et al., 2020; Lordkipanidze et al., 2007, 2013). Clearly, piping and related gullying processes were key to understanding not only the sedimentary and geomorphic development of Dmanisi during the formation of Stratum B1 but also had to be evaluated as a possible factor in the accumulation of so many well-preserved bones.

Pipes are tunnel-like features that form on hillslopes and in gullies through the combined action of hydraulic erosion and mass movement within the pipes (Bernatek-Jakiel et al., 2017; Bryan and Jones, 1997; Farifteh and Soeters, 1999; Gutierrez et al., 1997; Halliday, 2007; Jones et al., 1997; Pickford, 2018; Verachtert et al., 2010, 2013; Zhu, 1997, 2012). Bernatek-Jakiel and Poesen's (2018) comprehensive review shows that pipes have a truly global distribution, and exhibit remarkable variability with respect to climate, soil types, topographic settings as well as size, shape, and formation patterns. Pipe features at Dmanisi are all buried by Strata B2–B5, and are filled with B1 deposits. These ‘paleopipe’ features (see Saunders and Dawson, 1998), discussed later, include filled pipes, pipe breaches, and gullies formed on collapsed pipes.

Several attempts have been made since the early 2000s to date at the Dmanisi site (see SOM S1 and S2). All 40Ar/39Ar ages made directly on the Mashavera Basalt at Dmanisi and both the Mashavera and the Orzomani Basalts 14 km west of Dmanisi demonstrate that the Dmanisi formation is bracketed between 1.848 ± 0.008 Ma and 1.759 ± 0.005 Ma (2σ level; Gabunia et al., 2000; Messager et al., 2011; Nomade et al., 2016). The 40Ar/39Ar age of 1.81 ± 0.03 Ma and 1.81 ± 0.05 Ma obtained on stratum A1 ash by Garcia et al. (2010) and De Lumley and Lordkipanidze (2006), respectively, need to be taken with caution as they are based on very unprecise low K-bearing minerals (i.e., plagioclase) as well as glass shards that are known to be affected by 39Ar recoil during irradiation and argon lost during alteration which can bias 40Ar/39Ar toward older ages. Accordingly, those ages will not be used hereafter. Unspiked K/Ar and 40Ar/39Ar have been obtained at the LSCE Laboratory (France). Detailed methodological and analytical protocols for each method can be found in SOM S1.

All previous paleomagnetic studies were carried out in the thin lithological sections from Blocks 1 and 2, where piping and gulling is the rule rather than the exception (see SOM S2). This scenario was completely different when the M5 section was studied by Ferring et al. (2011). This vertically ordered section (free of gullies and pipes) was thicker (6 m) and displayed a clear paleomagnetism. Again, normal (Olduvai) and reverse (upper Matuyama) polarities were recorded in A and B strata, respectively.