Mirak is located in a dry mudflat (53°25’53” E; 35°28’10” N; ca. 1035 masl) that extends from Semnan's piedmont along the southern flank of the Alborz Mountains to the northern edge of the Iranian Central Desert, also known as Dasht-e Kavir. This area belongs to an elongated Quaternary drainage system, which is characterized by alluvial fans and sedimentary pediments, with widespread saline playa-lake systems downstream.

Geomorphological fieldwork has revealed a pedo-sedimentary succession represented by a 9-m-thick alluvial-aeolian record made up of two main sequences (Sequence I: alluvial basal complex; Sequence II: windblown deposits) (see Fig. 4 for the synthetic stratigraphy and description). Sequence I can be divided into six lithostratigraphic units from the base upwards (Units 9-4; thickness: ca. 5 m). Horizontal bedding of silty clays (Units 8, 6 and 4) were observed in the stratigraphy. This stratigraphic column also includes at least two bodies of very fine sands (Units 7 and 5) and a deflated horizon (Unit 4a); these three units contain the three main archaeological assemblages (3 to 1 respectively). Sequence I accumulated on a pediplain that was flooded repeatedly as a consequence of shallow sheet flooding and channel overflows. Indeed, horizontal bedding prevails in the very fine sand units (Units 7 and 5), but current ripple marks and internal planar cross bedding occur as well (lithofacies Sr and Sp according to Miall 1996). The whole units are distinguished by intense post-depositional weathering, resulting both from the repetition of decreases in the level of the water table and increases in the rate of evaporation, as evidenced by Fe–Mn concretions, iron hydroxide precipitates, calcareous nodules, gypsum crystals, and desiccation cracks. These features are associated with the development of ‘Aridisol horizons’ (Bk, By) (USDA, Soil Survey Staff, 2014). The upper limit of this alluvial pedocomplex is marked by a major discontinuity, which corresponds to the contemporary deflation surface of Dasht-e Kavir. Sequence II of Mirak 8 mound can be divided into four lithostratigraphic units from the base upwards (Units 3–0; thickness: about 4 m). The first part of this sequence consists of calcareous aeolian sand deposits (Unit 3) reworked by a hydromorphic horizon (Unit 2) topward. The uppermost part of Sequence II is composed of another aeolian sand deposit (Units 1 and 0) affected by the development of an entisol (USDA).

Samples for luminescence dating were collected: six samples were taken from the northern section of Sequences I and II and two samples from the lower part of the eastern section of Sequence I. Preliminary dating results, including analytical data, are listed in Table 1 and plotted in their stratigraphic position in Fig. 4 (see also Supplementary Material for details on the implemented procedure). OSL and post-IR IRSL measurements were performed on multi-grain aliquots (see SM for methodological details). Overall, the OSL ages are coherent with the stratigraphy and the inter-aliquot dispersion in equivalent doses for each sample is very low; this allows us to consider the OSL ages to provide accurate, reliable ages. As a result, the ages range from 50 ± 3 to 28 ± 2 ka in the Sequence I that contains the three primary archaeological deposits, with a maximum age of 28 ± 2 ka for Layer 1, dates between 28 ± 2 and 38 ± 2 for Layer 2, dates between 47 ± 2 ka and 47 ± 4 ka for Layer 3 (samples MK 16/2 and MK16/3), which correspond to the Last Glacial cycle. The ages range from 1.2 ± 0.2 to 0.6 ± 0.1 ka in Sequence II.

Based on the available data, it seems that Sequence I at Mirak Mound 8 may have formed as a response to climatic changes during the first half of the Last Glacial (probably during MIS 3), which was more humid than the Upper Pleniglacial (MIS 2 age equivalent). The records of these transitional periods are less well understood in the context of the northern Iranian paleoclimatic and chronostratigraphic framework for the Last Interglacial/Glacial cycle (i.e. the Lake Urmia core: Djamali et al., 2008; loess records of North of Iran: Kehl et al., 2005 and Frechen et al., 2009). Human occupations at Mirak were contemporaneous with the aggradation of a covered pediment incised by anastomosed channels. This period was apparently characterized by rapid flooding events (i.e. sheet flood or crevasse splay deposits) that buried Paleolithic deposits. If the increase in rainfall events explains the accumulation of sediments deposited during flash floods, we do not exclude the influence of episodic uplift upstream, as pointed out by Kehl (2009). Nevertheless, no stratigraphic evidence of tectonic effect has been observed. Aggradation of the alluvial cover ended with a major aridification marked by an aridisol development, perhaps indicating the end of the Last Glacial period. The scenario of a long-term drying coincided with a sedimentary hiatus, between the units 4 and 3, which separated the Pleistocene desiccated flat from the Holocene sand dune (sequence II). The Holocene history of Mirak is defined by at least two phases of calcareous dust deposits interrupted by the temporary development of a perched water body on the first aeolian mound. The formation of this hydromorphic horizon (Unit 2) in arid environment implies a groundwater flux from the bedrock and its resurgence by a perched spring.

The study of the excavated sediments evidently indicates that, in contrast to the current arid desert situation, the northern part of the Iranian Central Desert, at least for the studied region, had a completely different climate during the end of the Pleistocene. During the first half of the Last Glacial Period (MIS 4/3) the climate of the region was cold and humid. This cold and humid period was characterized by a slow alluvial aggradation interrupted by rapid flooding events. Two of the archaeological layers at Mirak 8 are dated to the chronological interval between 50 ± 3 ka and 28 ± 2 ka, which corresponds to these humid climate conditions. Moreover, the different nature of the sediments of the archaeological layers in comparison to the upper and lower sterile sediments implies that, during each of these occupational events (which might have lasted hundreds of years), the site was situated on dry land, possibly surrounded by streams and marshes. Concerning the Holocene sequence (Sequence 2 in Mirak 8), the current study of Mirak 8 demonstrates that the mounds in Mirak cannot be seen as remnants of a more extensive deposit that has been eroded away, but correspond to Holocene nebkha or shadow dunes. In Mirak 8, it results in at least two phases of calcareous dust deposits interrupted by the temporary development of a perched water body on the first aeolian mound. The two OSL ages derived from the very top of the mound (MK15/1: 0.6 ± 0.1ka) and where the mound meets the flatland area around it (MK15/4: 1.2 ± 0.2 ka) indicate that the formation of the mound of Mirak 8 took approximately 800–1600 years. Therefore, the mounds of the site were formed at a relative rapid pace. As mentioned above, such hydromorphic horizon in arid environment implies a groundwater flux from the bedrock and its resurgence by a perched spring. In that general pattern of dune formation, the vegetation may have played an important role in the formation of the mound by fixing the calcareous sands and consequently the preservation of the underlying Pleistocene sequence, as originally proposed by Vahdati Nasab et al. (2013).

The lithic assemblage presented here consists of 2313 pieces from the easternmost excavation at Mirak 8. More than 85% of these artifacts are made of varieties of chert (i.e. chert, flint, and jasper). Other raw materials are tuff, sandstone, siltstone, and rarely, microconglomerate. Within this assemblage, 976 pieces are typologically indeterminable fragments or splinters, most of which have very small dimensions; their techno-typological identification is problematic, if not impossible (see Jayez and Vahdati Nasab, 2016) for the morphological characteristics of what is considered “indeterminable”). These pieces have been excluded from our analysis, which leaves 1337 chipped stones to present, including 212 chips (Table 2), which are flakes mostly with dimensions of less than 1 cm (see Shea, 2013: 32).

The three identified levels occurred in distinct stratigraphic units, although some vertical dispersion are observed within each level that may be related to some taphonomic agents (work in progress) (Fig. 5); however, the general state of preservation of the lithics is very good, which testifies to the fact that, if taphonomy has impacted the deposits, the impact has been limited. In that respect, Level 1 appears to occur in a deflated unit; this may explain the wider dispersion and lower density in objects than those observed for levels 2 and 3. In addition, few artefacts were located between these three main archaeological deposits, usually in whole roots or vertical brakes. Apart these very few pieces (around 30 among 2313 artefacts), the artefacts appear to depict three main archaeological deposits.

The L1 lithic assemblage appears to be an industry oriented toward a blade/bladelet production from at least carinated elements. No cores have been found and other tool types are underrepresented. Carinated burins are found exclusively in L1 (Fig. 6: 1–3), which are usually found in the so-called “Zagros Aurignacian” (see: Olszewski and Dibble, 1994, 2002) or “Baradostian” (see: Olszewski, 1993 and Solecki, 1958) techno-complexes. The characteristic artifact-types of this assemblage include blades, bladelets, and their associated fragments (Fig. 6:4), although Arjeneh (Font-Yves) points and lamelles Dufour as major components of both central Zagros and southern Zagros Upper Paleolithic lithic industries (see Conard and Ghasidian, 2011, Ghasidian et al., 2017, Hole, 1970, Hole and Flannery, 1967 and Otte et al., 2007) are not observed in the current assemblage of Mirak. Among the platform rejuvenation flakes in this assemblage is one large core tablet (Fig. 6: 5) that indicates the large dimensions of prismatic cores from which the blades were produced. Finally, Levallois–Mousterian tools are completely absent in this layer. As a whole, this layer shows Upper Paleolithic affinities. However, this assessment must be taken with caution; indeed, the density of artifacts is low (see Table 2) and the assemblage was preserved in a deflated horizon and thus may not be fully representative of the original one.

The L2 lithic assemblage is characterized by the mixed presence of Upper Paleolithic industries with characteristics of prismatic blade/bladelet production and Levallois–Mousterian flake-based production, the technological structure of which shows also a moderate percentage of cores and tools (Table 2). Few blades and bladelets are either plain or regularly retouched or notched (Table 4, Fig. 7: 4–5). Semi-parallel arrises (flake scar ridges) are present on the dorsal face of some blades (Fig. 7: 3). The Mousterian industry artifact types are dominated by the Levallois technique (for Mousterian artifact-types see, e.g., Geneste, 1985). They mostly indicate recurrent reduction strategy with radial/centripetal preparation (Fig. 7:6). Levallois flakes with facetted and chapeau de gendarme platforms are present as well as the byproducts of Levallois reduction strategy in the form of core preparation elements. Bladelets and some of the blades have been made via the prismatic reduction method. It is worth mentioning that Levallois blade cores are rare in the Mirak assemblage. Regarding the tools (Table 4), among the interesting specimens is one elongated convergent blade with facetted striking platform and inverse notching/retouching (Fig. 7: 1). As a whole, Layer 2 appears to present a mix of characteristics usually associated with both Upper and Middle Paleolithic industries.

In L3, cores and tools are well represented (twice as much as in upper layers). Most of the cores are Levallois cores (Fig. 8: 11). If the debitage composition of the layers is considered, L3 lithic assemblage, with less than 5% blade and bladelet products, indicates an industry based on flake production, a number of them being relatively elongated (Table 3). This assemblage includes Mousterian industry artifact types similar to L2 with the dominance of Levallois technique with mostly recurrent reduction strategy with radial/centripetal preparation. Regarding the tools (Table 4), Levallois–Mousterian tools such as various scrapers (Fig. 8: 8, 10) are present in high quantity in L3 as well as some points (Fig. 8: 2, 5–7); they are comparable to those found in the Zagros Mousterian (see Dibble, 1993, Dibble and Holdaway, 1993 and Lindly, 1997). Some tools have basal trimming, which might be attributed to the hafting methods (Fig. 8: 1, 8). There are also Levallois points with retouch on both faces and facetted striking platform (Fig. 7: 9). As a whole, L3 has clear Middle Paleolithic affinities.

However, the vertical distribution and field data (such as post-depositional reworking) shows that some pieces cannot be precisely allocated to Layers 2 or 3; among these pieces, one carinated scraper as well as a prismatic core (Fig. 8: 3), here vertical between L2 and L3 respectively, are usually considered as a typical implement of Upper Paleolithic industry of Baradostian/Zagros Aurignacian (see Olszewski and Dibble, 2006 and Otte et al., 2007).

In contrast to what has been claimed previously about the Middle Paleolithic affinities based on the study of the surface collection of the lithic industries in Mirak (Rezvani and Vahdati Nasab, 2010 and Vahdati Nasab et al., 2013), the data derived from the three seasons of excavations at Mirak 8 has clearly identified the presence of vestiges of at least three main archaeological layers.

The upper layer, i.e. L1, has clear Upper Paleolithic affinities, with some resemblance (carinated burins) with the so-called “Zagros Aurignacian” (Olszewski and Dibble, 1994, 2002) or “Baradostian” (Olszewski, 1993 and Solecki, 1958) techno-complexes. Yet Arjeneh (Font-Yves) points and lamelles Dufour as major components of both central Zagros and southern Zagros Upper Paleolithic lithic industry (see Conard and Ghasidian, 2011, Ghasidian et al., 2017, Hole, 1970, Hole and Flannery, 1967 and Otte et al., 2007) are not observed in the current assemblage of Mirak; if confirmed, Mirak would more closely resemble Garm Roud, which dates back to around 33 kya (Berillon et al., 2016, Berillon et al., 2007 and Chevrier et al., 2006). However, the low density of artifacts in L1 and its evident reworked nature does not allow a precise determination of this upper assemblage and needs caution in its interpretation. The Upper Paleolithic affinities are consistent with the OSL dating, which gives a maximum age of 28 ± 2 kya, and with the presence of a large Upper Paleolithic open-air site, i.e. Delazian, at a very close proximity (2 km northeast of Mirak, see Vahdati Nasab and Clark, 2014).

The lower layer, i.e. L3, has clear Middle Paleolithic affinities; with regards to Levallois–Mousterian tools (various scrapers and points); the assemblage is comparable to those of classical Zagros Mousterian sites (see Dibble, 1993 and Dibble and Holdaway, 1993; Lindly, 1997). Interestingly, this assemblage dates between 47 ± 2 kya and 47 ± 4 kya, which, although compatible with the general chronological framework of Middle Paleolithic assemblages in the Middle East, appears very recent compared to the Middle Paleolithic known in the Zagros, and earlier than the early Upper Paleolithic assemblages of the Zagros (Becerra-Valdivia et al., 2017).

The intermediate layer, i.e. L2, might be the most controversial one, representing a mixture of industries with both Middle and Upper Paleolithic affinities, in a chronological framework that dates it to between 28 ± 2 and 38 ± 2 kya. Based on this preliminary chronological framework, L2 appears possibly contemporary of the early Upper Paleolithic of the Zagros (Becerra-Valdivia et al., 2017), where facetted platforms or sidescrapers have been identified (e.g., Tsanova, 2013). However, more studies are ongoing, including the technotypology of the intermediate layer, as well as a refined chronology based on recent OSL samples that should allow us refining our understanding of the nature, the chronology and the context of this intermediate human occupations at Mirak. However, to our knowledge, despite some potential resemblances with the early Upper Paleolithic of the Zagros, such mixture of industries has no equivalent to date within the regional context of the Iranian Central Plateau, and together with its chronological framework points the complexity of the cultural evolution in a wide area between Zagros and Central Asia.

Finally, beside this general pattern, a carinated scraper and a prismatic core, which are typical of the Baradostian/Zagros Aurignacian (see Olszewski and Dibble, 2006 and Otte et al., 2007), have been collected between the lowest layers (L2 and L3) and could not be allocated precisely to one of these levels. Although, as mentioned above, this may be related to some post-depositional reworking based on field data and vertical distribution, this points to the originality of the lithic industries of Mirak, and one cannot exclude some potential affinities of the Mirak's lower lithic assemblages with early Upper Paleolithic or late Middle Paleolithic cultures of the Zagros. Further analysis of the lithic assemblage will address these potential affinities.