Sediment-filled karst depressions and riyad – key archaeological environments of south Qatar

Systematic archaeological exploration of southern Qatar started in the 1950s. However, detailed local and regional data on climatic fluctuations and landscape changes during the Holocene, pivotal for understanding and reconstructing human–environment interactions, are still lacking. This contribution provides an overview on the variability of geomorphic environments of southern Qatar with a focus on depression landforms, which reveal a rich archaeological heritage ranging from Palaeolithic(?) and Early Neolithic times to the Modern era. Based on a detailed geomorphic mapping campaign, sediment cores and optically stimulated luminescence data, the dynamics of riyad (singular rawdha; shallow, small-scale, sediment-filled karst depressions clustering in the central southern peninsula) and the larger-scale Asaila depression near the western coast are studied in order to put archaeological discoveries into a wider environmental context. Geomorphic mapping of the Asaila basin shows a much greater geomorphic variability than documented in literature so far with relict signs of surface runoff. An 8 m long sediment core taken in the sabkha-type sand flats of the western basin reveals a continuous dominance of aeolian morphodynamics during the early to mid-Holocene. Mounds preserved by evaporite horizons representing capillarites originally grown in the vadose zone are a clear sign of groundwater-level drop after the sea-level highstand ca. 6000–4500 years ago. Deflation followed the lowering of the Stokes surface, leaving mounds where the relict capillarites were able to fixate and preserve the palaeo-surface. Abundant archaeological evidence of Early and Middle Neolithic occupation – the latter with a clear focus inside the central Asaila basin – indicate more favourable Published by Copernicus Publications on behalf of the Deutsche Quartärvereinigung (DEUQUA) e.V. 216 M. Engel et al.: Archaeological environments of south Qatar living conditions than today. In contrast, the sediment record of the investigated riyad in the south is very shallow, younger and controlled by surface discharge, deflation and the constantly diminishing barchan dune cover in Qatar over the Middle and Late Holocene. The young age of the infill (ca. 1500 to 2000 years) explains the absence of findings older than the Late Islamic period. Indicators of current net deflation may relate to a decrease in surface runoff and sediment supply only in recent decades to centuries. In the future, geophysical prospection of the riyad may help to locate thicker sedimentary archives and the analysis of grain size distribution, micromorphology, phytoliths or even pollen spectra may enhance our understanding of the interplay of regional environmental changes and cultural history. Kurzfassung: Die systematische archäologische Erkundung des Südens der katarischen Halbinsel begann in den 1950er Jahren. Lokale und regionale Daten zu den die regionale Kulturentwicklung begleitenden klimatischen Schwankungen und Landschaftsveränderungen im Holozän fehlen allerdings weitgehend. Dieser Beitrag liefert einen Überblick über die geomorphologische Variabilität des südlichen Teils von Katar mit einem Schwerpunkt auf karstbürtigen Hohlformen, deren verhältnismäßig reiches archäologisches Erbe bis in das Paläolithikum(?)/Frühe Neolithikum zurückreicht. Basierend auf einer detaillierten geomorphologischen Kartierkampagne, Sedimentkernen und Altersdatierung mit Hilfe optisch stimulierter Lumineszenz erfolgt eine Rekonstruktion der Sedimentationsgeschichte der räumlich bedeutenden Riyad (Singular: Rawdha; flache, kleinskalige, sedimentgefüllte Karsthohlformen) und des größeren Beckens von Asaila in der Nähe der Westküste, um die während der vergangenen Dekaden getätigten archäologischen Funde und Erkenntnisse in einen erweiterten Umweltkontext einzuordnen. Die geomorphologische Kartierung des Asaila-Beckens zeigt eine weit größere Variabilität an Oberflächenformen als bisher in der Literatur dokumentiert, darunter auch deutliche Indikatoren für reliktischen Oberflächenabfluss. Während der 8 m lange Sedimentkern aus dem sabkhaähnlichen Bereich des westlichen Beckens eine kontinuierliche Dominanz der äolischen Morphodynamik über das frühe bis mittlere Holozän andeutet, sind die von kleinen hügelartigen Formen dominierten Bereiche durch reliktische, einst in der vadosen Zone gewachsene kapillare Evaporithorizonte fixiert worden. Dies ist ein deutliches Zeichen für vorherrschende Deflation seit dem Hochstand des Meeresspiegels (und des Grundwasserspiegels) vor ca. 6000–4500 Jahren. Zahlreiche frühund mittelneolithische Streufunde – letztere mit einem klaren Schwerpunkt im zentralen Asaila-Becken – weisen auf günstigere Lebensbedingungen zu dieser Phase hin. Die sedimentäre Füllung der untersuchten Riyad ist deutlich flacher und jünger und wird durch Oberflächenabfluss, Deflation und die sukzessiv abnehmende Barchan-Dünenbedeckung in Katar während des Mittelund Spätholozäns determiniert. Das junge Alter der Füllung (ca. 1500–2000 Jahre) erklärt das Fehlen archäologischer Funde, die älter als spätislamisch datieren. Indikatoren aktuell dominierender Deflation (lineare Korrasionsrillen, MikroYardangs) sind mögliche Hinweise auf einen Rückgang von Oberflächenabfluss und Sedimenteintrag innerhalb der letzten Jahrzehnte bis wenigen Jahrhunderte. Eine geophysikalische Prospektion der Riyad, die die Lokalisierung möglichst mächtiger Sedimentpakete erlauben würde, in Kombination mit der Analyse von Korngrößenverteilung, Mikromorphologie, Phytolithen oder gar Pollenspektren, könnte in Zukunft Teil einer Strategie sein, hier detailliertere Daten zur Paläo-Umweltdynamik der zentralen Bereiche Katars zu generieren.


Introduction
Pioneering archaeological surveys and excavations in Qatar started more than 60 years ago (e.g. Glob, 1958;Kapel, 1967;de Cardi, 1978;Tixier, 1980;Inizan, 1988). During the last decade, these were supplemented by intensified interdisciplinary research on all phases of the peninsula's cultural heritage (e.g. Al-Naimi et al., 2011;Rees et al., 2011;Eichmann et al., 2014;Gerber et al., 2014;Drechsler, 2014;Drechsler et al., 2013Drechsler et al., , 2016Izquierdo Zamora et al., 2015;McPhillips et al., 2015). While evidence for the Palaeolithic remains under debate (Glob, 1958;Kapel, 1967;Al-Naimi et al., 2010;Scott-Jackson et al., 2015;Drechsler, 2014;Drechsler et al., 2016), Early and Middle Neolithic flint scatters along with burial cairns at a number of locations testify the occurrence of mobile to semi-stationary groups at that time . The Early Neolithic mainly comprises Qatar-B sites sensu Kapel (1967) and In-izan (1988), tentatively dated to the 7th millennium BCE. The Middle Neolithic is often represented by surface finds of tile knives, scrapers, bifacial arrowheads and, in some cases, Ubaid-style pottery (Tixier, 1980;Inizan, 1988;Drechsler, 2014), which implies links with other Ubaid-related sites of the southern Arabian Gulf coast dated to the 6th-5th millennium BCE (Oates, 1978;Uerpmann and Uerpmann, 1996;Kainert and Drechsler, 2014). The earliest known settlement of Qatar was identified in Wadi Debayān at the northwest coast, dating back 7500 years Tetlow et al., 2013). Of similar age are the fishermen's huts near Shagra at the east coast, which were associated with the Ubaid period by Inizan (1988). The decline of the Ubaid culture marked the onset of a period represented by very few cultural remains in Qatar, coinciding with the "Dark Millennium" as defined for the southeastern gulf shores (Uerpmann, 2003;Preston et al., 2012;Muhesen and Al-Naimi, 2014). The Bronze Age is only sparsely represented in Qatar, most prominently in the form of pottery and a purple-dye industry at Al-Khor north of Doha (Edens, 1999;Carter and Killick, 2010). Such finds are nearly absent in the southern part of the country . Burial cairns are distributed over the entire peninsula and mostly from the Iron Age (ca. 300 BCE-300 CE) (Bibby, 1965;Buckley, 1973;Konishi et al., 1988), even though some of them have recently been dated to as old as the Ubaid period Izquierdo Zamora et al., 2015). No evidence exists for Iron Age settlements, but a coexistence of nomadic pastoralism and sedentary lifestyles has been postulated (Muhasen and Al-Naimi, 2014). Permanent settlements emerged around the 8th century CE (Abbasid period), in particular along the northwestern coast (Guérin and Al-Naimi, 2009;Macumber, 2015). The nomadic and semi-nomadic Bedouin culture, however, coexisted and persisted well into the 20th century CE, focussing on the shallow karst depressions of the Qatar peninsula as campsites (McPhillips et al., 2015).
Underpinning research on the concomitant environmental changes is clearly in its infancy, consisting mainly of the outcome of the QNHER (Qatar National Historic Environment Record) project covering the northern part of the peninsula (e.g. Cuttler et al., 2011;Macumber, 2011Macumber, , 2015Tetlow et al., 2013). Isolated contributions to the framework of earlier archaeological missions mainly focus on Late Quaternary coastal changes and are rather preliminary (Vita-Finzi, 1978;Perthuisot, 1980). Thus, when the South Qatar Survey Project (SQSP) started in late 2012 as a joint operation between Qatar Museums and the Orient Department of the German Archaeological Institute, little well-established information existed regarding dynamics of vegetation, water availability and landforms. In the course of the archaeological survey, shallow, sediment-filled karst depressions referred to as riyad (singular rawdha or rawdah) (Batanouny, 1981;Babikir, 1986;Sadiq and Nasir, 2002;Al-Yousef, 2003;Macumber, 2011Macumber, , 2015 turned out to be focal points of historic occupation, in contrast with the barren surrounding hamada (Eichmann et al., 2014;Gerber et al., 2014;Pfeiffer, 2015). A second priority of the SQSP was assigned to the large depression of Asaila in centralwestern Qatar (locally referred to as Jaow Al Bahath), which, compared to the rest of Qatar, preserves a high concentration of Early and Middle Neolithic single flint artefacts and flint artefact scatters, along with numerous findings of the pre-Islamic, Islamic and Modern periods (Kapel, 1967;Inizan, 1988;Pelegrin and Inizan, 2013;Drechsler, 2014;Drechsler et al., 2016).
This paper provides a general overview of the variability of geomorphic environments of southern Qatar (defined here as the part of the peninsula south of the Dukhan road; Fig. 1) over Holocene timescales. Based on a detailed geomorphic mapping campaign, sediment cores and optically stimulated luminescence (OSL) data, the dynamics of riyad and the basins of Asaila and adjacent Jaow Aqeeq (Fig. 1) are evaluated in order to put the Early Neolithic to Islamic archaeological discoveries of the SQSP (Drechsler, 2014;Eichmann et al., 2014;Gerber et al., 2014;Pfeiffer, 2015) into a wider environmental context.

Geology and tectonic setting
The Qatar peninsula protrudes from the Arabian Peninsula into the Arabian Gulf. It represents an anticlinal structure of uplifted Palaeocene to Middle Eocene limestones, dolomites, marls, chalks and shales, intercalated with evaporites ( Fig. 1). Lower and Middle Eocene carbonates comprise 80 % (Rus Formation ∼ 10 %; Dammam Formation ∼ 70 %) of Qatar's surface (Al-Yousef, 2003;Al-Saad, 2005). Miocene units are mainly found in southern Qatar, represented by the Dam and Hofuf formations. The Dam Formation consists of a sequence of shallow marine limestone, gypsum, dolomite and mud, whereas the Hofuf Formation also contains continental conglomerates with a matrix of aeolian sand and gypsum (Cavelier, 1970;Al-Yousef, 2003).
The topography is dominated by the central N-S-trending Qatar Anticline, which has been driven by tectonic uplift since the Palaeogene and created smaller anticlinal and dome structures. The Dukhan Anticline (NNW-SSE) in the west is another major structural feature, bearing the largest oil reservoirs of the area, creating a steeper surface relief than the Qatar Anticline and causing slight tilting of the entire limestone sequence (Al-Yousef, 2003).

Hamadas
Notable topographic elevations occur only in the south, reaching 103 m a.s.l. (above mean sea level) at the highest point. The majority of the peninsula is flat and the most widespread landform is hamadas (Fig. S8 in the Supple-  Qatar, 1980), overlying an ASTER global digital elevation model, which is a product of the US Ministry of Trade and Economy (METI) and the National Aeronautics and Space Administration (NASA). The main study area of the Asaila basin as well as coring sites in the Jaow Aqeeq basin further north (ASA-C1) and the southern riyad are depicted (QAT 41, QAT 66). Tectonic features are adapted from Al-Yousef (2003). The distribution of freshwater conditions in groundwater, which, depending on different sources, may have an upper limit of total dissolved solids of 3000 ppm (e.g. Heberger and Donnelly, 2015), are based on data of the Qatar Department of Environment, cited in Macumber (2012). The overview map indicates the location of the main map and palaeoclimate records referred to in the text. ment), locally referred to as hazm (very gently sloping) or mistah (entirely flat), i.e. stone pavements covering most of the Dammam limestone province ( Fig. 1) (Perthuisot, 1980;Batanouny, 1981). These plains are covered by mostly angular, in situ limestone gravel (see Benazzouz, 2004;Goudie, 2004b). Vegetation cover of the hamadas is very scarce; only isolated xerophile shrubs and trees such as Tetraena qatarense, Acacia tortilis or Lycium shawii are found between the surface stones (Batanouny, 1981).

Rocky ridges
Rocky ridges of southern Qatar relate to the N-S-trending anticlines, in particular to the Miocene Dam Formation along the southwest coast between Dukhan and south of Umm Bab (Fig. 1). They form mesas and buttes, controlled by varying resistivity of carbonate strata and reg surfaces, i.e. pavements of smaller, par-autochthonous clasts (see Benazzouz, 2004). Patchy sand accumulations in depressions of the rocky ridges are the only sites where sparse vegetation is found today, consisting mainly of Panicum turgidum and Zygophyllum qatarense (Fig. S14) (Batanouny, 1981).

Karst depressions
The term rawdha (Arabic "garden") refers to the fine-grained infill of shallow, round, slightly elongated or more irregular (e.g. when coalesced) inland depressions of 100 m up to a few kilometres in diameter. Riyad result from solution and collapse of the Eocene Rus and Dammam gypsum and limestone and are most abundant in the central part of the peninsula (Fig. 1). Their formation is related to the presence and orientation of anticlinal joints and fractures and, as hypothesized by Sadiq and Nasir (2002), to intensified karstification during the wetter Middle Pleistocene. Riyad sediments usually consist of light brown silty fine sand provided by sheet floods; carbonate and salt contents are low (Babikir, 1986). Usually they are topped by thin drapes of coarse aeolian sand or even nebkhas -mounds of aeolian sediment trapped and fixated by shrubs (Goudie, 2004a) -of up to 2 m height (Perthuisot, 1980;Macumber, 2011;Engel and Brückner, 2014). Most riyad are currently subject to deflation as indicated by micro-yardangs and linear corrasion features at the surface Figs. S32, S33). The riyad of Qatar's interior provide evidence for occupation indicated by a considerable amount of pottery of different wares mostly dating into the (Late) Islamic and Modern periods. Stone structures of temporary character reflecting the presence of Bedouins are abundant (Eichmann et al., 2014;. The widest and deepest depressions of southern Qatar, significantly exceeding the dimensions of riyad, are the Asaila basin ("Acila depression" in Inizan, 1988;Macumber, 2012;Pelegrin and Inizan, 2013; local name is Jaow al Bahath), separated from the Dukhan continental sabkha by a massive limestone ridge, and the Jaow Aqeeq basin (Fig. 1).
Both depressions result from collapse and solution over major faults and joint-flow drainage estimated to be active at least since the Miocene. Sadiq and Nasir (2002) suggest a genetic sequence reaching from deepening and widening cylindrical karst pits, which coalesce subterraneously (compound karst pits) and develop larger bottle-and bowl-shape karst pits through collapse processes. These pits gradually fill up with predominantly aeolian deposits to form such mature depressions or basins.

Wadis
Wadis are most prominent in the southwest of Qatar, where they accumulate some silt and clay and are intercalated by gravel sheets, which result from episodic rainfall events. The wadis show a relatively dense vegetation cover characterized by Pennisetum divisum, Acacia ehrenbergiana, L. shawii and, where aeolian dynamics increase, Leptadenia pyrotechnica (Batanouny, 1981).

Coastal sabkhas
The low-lying coastlines of Qatar support the formation of coastal sabkhas, i.e. saline flats in intertidal position. Coastal sabkhas are extensive in the southeast, around Khor al-Udaid, and are characterized by temporary flooding, a water table close to the surface and the precipitation of evaporites within the sediment column and on the surface. A large continental sabkha without surface connection to the sea is formed in the synclinal depression east of the Dukhan Ridge. Its lowest point is 6 m below sea level (Al-Yousef, 2003). The coastal sabkhas are Holocene features mostly resulting from coastal progradation along the entire Qatari coast following the mid-Holocene sea-level highstand (e.g. Billeaud et al., 2014;Strohmenger and Jameson, 2018). Coastal sabkhas with fine-grained soil may host halophile vegetation such as Arthrocnemum glaucum, Juncus rigidus or Aeluropus lagopoides. Where they merge into tidal flats not bordered by beach ridges, mangroves of Avicennia may establish (Batanouny, 1981).

Barchan dunes
In addition to the sand deposits found along the rocky ridges of southwestern Qatar, aeolian processes formed barchan dune fields in the southeast. The availability of quartz sand as source material and the regional Shamal wind system approaching from NW to NNW are the defining factors (Embabi and Ashour, 1993;Rao et al., 2001;Al Senafi and Anis, 2015). Once having crossed and covered the peninsula from NNW to SSE, the sediment source area towards the inner gulf became cut off due to the Holocene marine transgression into the Arabian Gulf, just before the mid-Holocene sealevel highstand. The southeastern dune population migrating with a speed of several metres per year represents a relict landform constantly diminishing in size as the dunes "calve" into the Arabian Gulf . While the barchan dunes themselves are free of vegetation, Cyperus conglomeratus may establish along their margins (Batanouny, 1981).

Present and former climate
The present climate of Qatar is arid, though the relative humidity may rise up to 90 %. Annual rainfall amounts to 50-80 mm and mainly occurs during winter and spring. However, the spatio-temporal pattern of rainfall is highly irregular (Embabi and Ashour, 1993). The NW-to-NNW Shamal winds are active mostly during early June to mid-July and November to March (Rao et al., 2001;Al-Yousef, 2003;Al Senafi and Anis, 2015), and they drive aeolian morphodynamics throughout the peninsula (Embabi and Ashour, 1993;.
Climate as the dominant factor shaping the physical landscape, controlling water availability and influencing occupation patterns of Qatar, has varied in the past. During glacialinterglacial cycles, and even on millennial timescales over the Holocene, geological records from different parts of the Arabian Peninsula indicate considerable fluctuations (e.g. Fleitmann et al., 2007;Engel et al., 2012Engel et al., , 2017Preston et al., 2012;Dinies et al., 2015Dinies et al., , 2016Guagnin et al., 2016;Parker et al., 2016, Breeze et al., 2017Parton et al., 2018). The closest palaeoclimate record on the Arabian Peninsula is located at Ras al-Khaimah, UAE (Preston et al., 2012;Parker et al., 2016), where lake deposits reflecting a rainfall surplus date between 9.0-8.3 and 3.0 ka cal BP. The only reference presenting local Late Quaternary rainfall variability of Qatar describes humid conditions during the Last Glacial Maximum 20 kyr ago and increasing aridity towards the present with a short humid deviation during the mid-Holocene. However, this curve has to be considered with caution since no specific data source is provided (Perthuisot, 1980). As recharge rates are low, permanent freshwater bodies are absent. Easily accessible, potable groundwater aquifers representing the main limiting factor of ancient settlement activity are rare. Most freshwater aquifers occur in the north, while in the south they are very local and isolated (Fig. 1). Groundwater tables at the coast tend to incline towards the sea level. According to Macumber (2011Macumber ( , 2015, the Asaila depression is one of the few sites with relatively easy access to shallow fresh to slightly brackish groundwater in the south of Qatar.

Methods
This study is based on work carried out at the Asaila basin, the basin of Jaow Aqeeq and two specific riyad in the southern part of the peninsula (Fig. 1). At Asaila, we conducted a detailed geomorphic mapping campaign, carried out a magnetometer prospection and took an 8 m long sediment core (QAT 63). At Jaow Aqeeq, a 3 m long sediment core (ASA-C1) was taken in order to compare sedimentation patterns inside both basins with different hydrological conditions. The sediment infill of two riyad was investigated based on a trench (QAT 41) and a short sediment core (QAT 66).

Geomorphic mapping
To develop our understanding of past and present geomorphic processes shaping the Asaila depression, a detailed geomorphic mapping campaign was initiated to identify environmental controls of landscape dynamics and explain the spatial distribution of archaeological surface findings. Pre-existing mapping resources included a multispectral IKONOS mosaic from 2004 (resolution ∼ 1 m pixel −1 ), the geological map of Qatar (State of Qatar, 1980), the topographic map of Qatar (State of Qatar, 1971), a predecessor geomorphic map of Inizan (1988) and local geological maps of Al-Yousef (2003), which were used for an initial classification of landforms as polygons in ArcMap v10.3.1. A total of 2 weeks of field mapping in March 2016 included photographic documentation, verification and refinement of the existing documentation. Site location and topographic profiles were realized using a TOPCON Hiper Pro differential global navigation satellite system (DGNSS) with a lateral and vertical error of ±2 cm, a manual GPS and a laser rangefinder (TruPulse 200, Laser Technology Inc.). The resulting map describes (i) geomorphic surface units based on the rock type and type of sediment deposit, (ii) morphodynamics and morphogenesis (aeolian, fluvial, etc.), and (iii) morphometry (slopes, microrelief). Vegetation within the depression was recorded and determined using Batanouny (1981) and Norton et al. (2009).

Magnetometer prospection
To verify the existence of former surface runoff patterns into the Asaila basin (see geomorphic map in Inizan, 1988) and to localize relict fluvial landforms, a magnetometer prospection was carried out in the area of the Acila 36 excavation (Inizan, 1988;Pelegrin and Inizan, 2013) in the northern part of the basin (Fig. 2). In order to reach the highest possible sensitivity, to realize a time-efficient prospection and to receive additional information on the enrichment of magnetic minerals in lateral sediment layers, we used the Cs total field magnetometer (Scintrex SM4G-Special) using the "duo-sensor" configuration (see examples in Fassbinder, 2015in Fassbinder, , 2017. Five adjacent grids of 40 m × 40 m were prospected. All details regarding the prospection and data analysis are presented in Supplement Sect. S1.

Sampling and survey of sediment archives
To study potential changes in depositional environments, which would have had essential implications for the interpretation of the survey findings, a vibracore (QAT 63) was taken in the western part of the Asaila basin using an Atlas Copco Cobra mk1 coring device and open steel probes of 6 and 5 cm  Table S2) and the magnetogram (Fig. 4). in diameter. A second shorter core (QAT 63 OSL) was taken at the same site using closed, opaque PVC liners in order to retrieve sample material for optically stimulated luminescence (OSL) dating . For comparison, we extended the analysis of sedimentary archives to the adjacent basin of Jaow Aqeeq, located seaward of the rocky ridge separating the Asaila depression from the Dukhan sabkha. Sediment core (ASA-C1) was taken in the northeastern part of the basin of Jaow Aqeeq (Fig. 1). In order to investigate the formation and dynamics of the key archaeological environments of the riyad, many were visited and described during the field surveys between 2012 and 2016, while two of them were studied in more detail using a sediment core (QAT 66) and a trench (QAT 41).

Grain size analysis
Due to their overall coarse texture, samples of QAT 63 were analysed for grain size and grain shape by applying dynamic image analysis (Retsch Camsizer P4, particle size range of 30-30 000 µm). Samples from Jaow Aqeeq (ASA-C1) and the riyad (QAT 41, QAT 66) were measured using a laser particle analyser (Beckman Coulter LS 13320; particle size range of 0.04-2000 µm) because these had a significant silt component. For analysis of the latter, the pre-treatment procedure includes air-drying of sample material, careful hand pestling and dry-sieving < 1 mm. H 2 O 2 (15 %) was added to a < 1 mm aliquot of 0.2-1.0 g in order to remove organic carbon. The sample was washed using a centrifuge until a pH value of 6-8 was reached to avoid neutralization after the addition of 0.5N Na 4 P 2 O 7 (55.7 g L −1 ) for aggregate dispersion. Calculation of statistical parameters was performed using the GRADISTAT statistic package (Blott and Pye, 2001).

X-ray diffraction analysis
Selected surface samples from the Asaila depression were subjected to X-ray diffraction (XRD) analysis using a Siemens D5000 powder diffractometer (Cu tube) to determine if the landforms are actively formed or of relict nature (active formation may be indicated by the presence of easily soluble evaporites). Samples were measured over a range of 5-75 • (2θ ) with a step size of 0.05 • and a time of 4 s per step (aperture slit = 0.5). The diffractograms were analysed by employing the EVA software package and the ICDD (International Centre for Diffraction Data) database.

OSL dating
Due to a lack of material suitable for radiocarbon dating, age estimates for selected units in the Asaila depression and the riyad are derived from OSL dating of sand-sized quartz grains (see Sect. S2 and Figs. S1-S4 for details).

Landforms of the Asaila basin
The Asaila basin has an extent of ca. 12 km 2 . It is mostly closed, apart from its western margin, where it is connected with the southwestern extension of the Dukhan sabkha through a series of ridges and small depressions. Detailed field mapping revealed a surprisingly wide range of landform units ( Fig. 2; see Sect. S4 for detailed explanations), some of which clearly relate to the former presence of surface water and were probably formed by surface processes during the Holocene.
The basin is framed by the flint-bearing limestone plateaus of the Dammam Formation, consisting of two clear plateau levels, i.e. the higher limestone plateau (ca. 13-18 m Qatar National Datum, QVD), including the rugged higher limestone plateau along its basinward transition (ca. 8-15 m QVD) and the lower limestone plateau (ca. 2-8 m QVD). The higher limestone plateau forms a massive NE-SW-trending structural landform and separates the Asaila basin from the adjacent Dukhan sabkha. Another unit of the higher limestone plateau bounds the basin in the northeast. The transition from the lower limestone plateau to the floor of the Asaila basin occurs quite abruptly in the southwest, south and east, in the form of a vertical step (Fig. 3a). In the north and northwest, the transition is relatively smooth with broad units of terraced slope debris and sand-and-gravel sheets (Figs. 2,3a). In the northernmost part of the basin, where the lower limestone plateau merges into the gravel sheet, the excavation Acila 36 of Inizan (1988) and Pelegrin and Inizan (2013) is located. At this type site for the Middle Neolithic Qatar-B industries, a new magnetometer prospection was carried out (Fig. S13). Despite poor magnetic susceptibility of the loose sediment cover due to low heavy mineral and iron oxide contents and internal magnetization contrasts of only ±0.8 nT, an unequivocal 10-15 m wide channel structure covered by sand and gravel and running towards the basin can be inferred (Fig. 4). The finding corroborates the assumption of Inizan (1988) of a major inactive surface water trajectory into the basin in this area (Fig. 2). Other geomorphic features reflecting more recent fluvial activity include the following: i. The first feature is a broad, vegetation-free, linear surface depression of 10-20 cm depth (Fig. 5) with a weak, reddish-brown surface crust of gypsum, quartz sand and calcite (Fig. S6, Table S2) as well as high amounts of clay and silt (Fig. S31). It extends from the central northern margin and bifurcates towards the centre of the basin (Fig. 2).
ii. The second feature is a wadi channel at the eastern end of the Asaila basin, characterized by retrogressive erosion (Figs. S29, S30). The wadi is filled with slope debris and experiences aeolian overprinting in the form of sand ramps (Figs. S12, S29a). However, horizontally bedded platy clasts in a poorly sorted, consolidated sandy matrix and an outwash channel exposing barren bedrock point to at least episodic surface water flows in recent times (Figs. S29a, S30a).
The most extensive landform unit inside the Asaila basin is the hummocky sand flats (Fig. 2), which mostly comprise nebkha fields originating from Zygophyllum qatarense shrubs. Some of the nebkhas seem inactive and are protected by thin evaporitic crusts made of gypsum, calcite and minor amounts of sylvine, mixed with quartz sand (samples ASA 1 and 2; Fig. S5, Table S2). In some areas, even halite was present (ASA 3; Fig. S5, Table S2).
Some areas in the central southern, northern and eastern parts are densely covered by characteristic mounds significantly higher than the nebkhas. They are fixated and protected by evaporitic horizons of massive to porous gypsum and minor amounts of calcite and sylvine (samples ASA 5, 6, 8-10; Figs. 3b, S6, S7, Table S2). These mounds vary from perfectly round, up to 1 m high, to narrow, cross-cutting ridges, several tens of metres long (Fig. 3b). Furthermore, large parts inside the basin are covered by vegetation-free sabkha-type sand flats (Fig. 3a, b) with gypsum-and halitecontaining surface crusts, in some parts even polygonal structures (Figs. S19, S20).

Sedimentary infill of the Asaila depression
Sediment core QAT 63 was taken in the low-lying sabkhatype sand flats of the northwestern Asaila basin (Figs. 2,  S21). It reached a depth of 8 m and consists of moderately sorted to moderately well-sorted fine to very fine gravelly medium sand (sensu Blott and Pye, 2001) (Fig. 6). Despite the slight variation in fine, medium and coarse sand, no facies changes occur. Grain size distributions are generally symmetrical to slightly coarsely skewed, apart from samples at 0.58-0.50 m below surface (b.s.) and ca. 1.55-1.40 m b.s., which are very coarsely skewed and poorly sorted. These samples, along with the surface sample, also have the lowest value for grain sphericity and width to length ratio of the entire sequence. A trench next to QAT 63 revealed a mas-sive subsurface gypsum crust at 0.60-0.50 m b.s., while the overlying sands are clearly cross-bedded (Fig. 6b).
Three samples were taken for OSL dating at 0.40 (small trench), 1.90-1.60 and 2.90-2.60 m b.s. (both from core QAT 63 OSL). They reveal coarse-grain (150-200 µm) quartz ages of 5800 ± 300, 6400 ± 300 − −7300 ± 300, and 7500 ± 300 years, respectively (Fig. 6, Table S1). While for the uppermost sample right above the gypsum crust the measured water content of 7 % was used for age calculation, two ages were generated for the sample 1.90-160 m b.s., which is probably located in the vertical fluctuation range of the local groundwater table. The younger age implies a water content of 10 % as measured in the laboratory, the older one assumes water saturation. While fluctuating water contents over time introduce dating uncertainties, the overdispersion of the rather symmetric equivalent dose distributions of 8 %-15 % suggests that incomplete signal resetting is not an issue (Fig. S4).

The depression of Jaow Aqeeq
Jaow Aqeeq is another larger-scale topographic depression north of Asaila, located close to the coastal sabkha of Dukhan and seaward of the SW-NE-trending limestone ridge separating Asaila from the Dukhan sabkha (Figs. 1, S32). Both depressions share strong geomorphic similarities, even though traces of surface runoff are more prevalent in Jaow Aqeeq compared to Asaila. Outside the depression along its eastern margin, a wide range of archaeological surface finds include fireplaces, clusters of pottery sherds and lithic artefacts (flakes, blade tools, arrowheads), cairns, shells, coins, metal pieces, bulbs of flint, remains of a provisional mosque, and several modern structures. Some of the lithic artefacts may correspond to the Early Neolithic Qatar-B industry, but the context is largely undated. Multiphase utilization of the site is evident. Distinct areas of domestic use and associated lithic workshops (later overprinted by cairns) as well as mosques and modern structures have been mapped (Fig. 7b). The record inside the basin is very poor so far, only consisting of modern garbage (Schönicke et al., 2016). The surface of the inner depression reveals higher moisture and higher abundance of halite-and gypsum-dominated crusts as compared with the Asaila basin.
In order to improve our understanding of the differences between Asaila and nearby Jaow Aqeeq and to evaluate sediment infill, hydrological conditions and palaeoenvironmental changes, sediment core ASA-C1 was taken inside the Jaow Aqeeq depression close to a temporary shallow standing waterbody (Figs. 7, S32). At the surface, the buckled gypsum crust, several centimetres thick and in some places thinly covered by white halite crystals, appeared very similar to sabkhas along the coast of southwest Qatar (e.g. Dukhan sabkha; Al-Yousef, 2003). The groundwater table was located at 35 cm b.s.  Inizan (1988), located at the basinward end of the lower limestone plateau (see Fig. 1 for overview). The area, where several scatters of both worked and unworked flint (green dots on a) were found during the survey of Drechsler et al. (2016), some also associated with ceramics (white dot), slopes towards the southwest and merges into the sand and gravel sheets and the sabkha-type sand flats of the basin. The magnetogram shows an inactive, subsurface channel structure running towards the basin (white lines).  (Fig. 7). The sand mostly consists of well-rounded quartz grains and minor appearances of feldspar, gypsum, reworked Dammam limestone and some dark heavy minerals. The sections with the highest amounts of silt and clay were checked for microfossil indicators for lacustrine environments, but only very isolated, highly abraded skeletal fragments were found. In most cases, the occasional gravel component coincides with large gypsum crystals, which occur in large numbers at several horizons, as either hardgrounds or gypsum mush (ca. 2.30, ca. 1.55, 1.00-0.90 m b.s., upper 20 cm). The sections 1.80-1.53 and 0.73-0.49 m b.s. appear well-stratified with al-ternating coarser and finer textures. Most sections are moderately sorted to moderately well-sorted, except those containing a more significant silt/clay or gypsum component.

The southern riyad
Two riyad were investigated (Fig. 8). At QAT 41 (site HAR 5183/QNHER 5183 in Gerber et al., 2013), a considerable surface relief has formed through wind sculpturing. The loose coarse sand on top is distributed in patches, partly developing wind ripples (Fig. 9b). While smaller limestone pebbles at the surface may result from sheet flood input into the endorheic depression of ca. 0.065 km 2 , the small boulders, mostly found along the rawdha margins, were brought in by humans to stabilize their tents. Further archaeological findings comprise food remains (shells) . A pit dug in the central part revealed only 80 cm of sedimentary infill. Overlying the Dammam limestone is a thin unconsolidated layer of limestone pebbles, up to 3 cm in diameter, in a fine sandy matrix. The section 0.75-0.48 m b.s. shows consolidated, greyish brown medium sand with a minor silt fraction as well as fine to coarse sand (Fig. 10). Furthermore, CaCO 3 concretions as pseudomorphs along former root channels were documented. The following layer (0.48-0.42 m b.s.) has a similar sand matrix but contains a greater amount of precipitated CaCO 3 . The upper unit (0.42-0.00 m b.s.) consists of weakly consolidated silty fine sand with some carbonate concretions and root remains (Fig. 9b).
The two thicker units below and above the CaCO 3 horizon reveal OSL dates of 1200 ± 100 (0.60 m b.s.) and 710 ± 30 (0.30 m b.s.) years, respectively (Table S1). Since scatter and shape of equivalent dose distributions suggest relatively complete signal resetting prior to deposition (Fig. S4) and both samples are situated clearly above the groundwater level, the OSL ages provide robust estimates of the time of sediment deposition. Vertical sediment mixing through haloturbation potentially biasing the equivalent dose distributions can be excluded due to negligible salt contents (Babikir, 1986). Sediment core QAT 66 was taken at the centre of a second, much larger rawdha (ca. 0.74 km 2 ) with a complex system of wadis entering from all sides (Fig. 8). QAT 66 hit bedrock at 0.79 m b.s., exposing an overlying sequence of consolidated, well-sorted, greyish light brown silt with a very small finesand component. Downcore changes are negligible. The deposit is compact and entirely dry. It forms an even surface with very sparse and low herbaceous vegetation grazed by goats and sheep. Shrubs trap aeolian sand and form nebkha mounds of > 1 m in elevation (Fig. 9a).  QAT 41 represents a trench in rawdha HAR 5183 (Gerber et al., 2013) and QAT 66 a sediment core in a rawdha, which was not part of the archaeological survey (Fig. 9).  (Figs. 1, 8); the location of grain size samples is indicated on the core (Fig. 10). (b) Profile QAT 41 exposing the sediment infill of a smaller, more elliptic rawdha (HAR 5183), with brief stratigraphic log, OSL ages (Table S1) and grain size samples (Fig. 10).

The geomorphic system
Geomorphic mapping based on field surveys and satellite imagery reveals the polygenetic nature of landforms in the Asaila basin, driven by tectonic, aeolian and fluvial processes as well as subsurface hydrology. The macroscale relief is determined by the roughly N-S-trending anticline or syncline structures, in particular the Al Huriyeh syncline (Fig. 1). The basin's origin may be controlled by a NNE-SSW-striking fault (Inizan, 1988) initiating long-term, subsurface limestone dissolution and collapse since the Miocene (Sadiq and Nasir, 2002) and eventually creating a large-scale morphological depression at the surface. Definite field evidence for such a fault, however, is missing, and other sources emphasize the lack of surface expressions of major faulting on the Qatar peninsula (e.g. Cavalier, 1970;Sadiq and Nasir, 2002).
Surface water exists but is very short-lived and only occurs episodically during strong rainfall. Geomorphic evidence was mapped in the eastern part, in the form of fluvial bedforms in a narrow, shallow channel and a collapsed (sub)surface drainage system, i.e. a potential karst spring. The wadi channel mapped in the central northern part of the basin with its reddish-brown surface appears inactive; halite as a potential sign of recent flooding is absent (sample ASA 7). However, it unequivocally represents a significant pathway of surface runoff into the basin. Another major path-way of surface inflow exists in the northernmost part of the basin, where the higher limestone plateau is dissected in an ENE-WSW orientation (Fig. 2). The small valley entering at the northernmost extension of the basin hosts a subsurface channel morphology (Fig. 4), testifying to the relict nature of more significant surface runoff. There is, however, no evidence for the timing of this increased runoff.

Prevalence of aeolian processes
Grain size measures of sediment core QAT 63 reflect a persisting aeolian environment. OSL data indicate that the sequence captures at least the entire Holocene. Both datasets in combination with present-day sabkha-type surfaces, nebkha formation, sand ripples, streamlined gypsum mounds showing clear signs of corrasion and sand ramps reflecting the main Shamal corridor show that aeolian processes have dominated the Asaila landscape at least since the arrival of humans in the Neolithic. QAT 63 rejects any type of lacustrine environment inside the Asaila basin during that time, which has been speculated by Macumber (2012).

Fluctuations of the groundwater table and capillary evaporite formation
The three horizons of poorly sorted sediments, as well as a lower width to length ratio and sphericity of grains in the uppermost 1.5 m of QAT 63 are related to the ongoing precipitation of mostly gypsum crystals from ascending groundwater in the vadose zone. These processes are similar to Figure 10. Grain size distribution of samples from the riyad (Fig. 9) and, for comparison, a typical barchan dune of south Qatar (QAT 67; curves almost entirely overlap).
those in continental sabkhas, where brines are significantly more highly concentrated (Sonnenfeld and Perthuisot, 1989;Yechieli and Wood, 2002;Ginau et al., 2012). Upward capillary movement from a shallow groundwater table (ca. 1.5 m deep in the Asaila basin today) leads to a halite-dominated and carbonate-containing crust at the surface, underlain by one or several layers of gypsum, also referred to as capillarites (Sonnenfeld and Perthuisot, 1989). This is also reflected by surface XRD samples from the hummocky sand flats, which, similar to the sabkha-type sand flats, represent active equilibrium surfaces of deflation and aeolian deposition. Crust formation is to a large extent driven by capillary rise and the position of the groundwater table. Sample ASA 3 still contains halite, whereas samples ASA 1 and 2 do not contain halite and show how easily salts are deflated and dissolved at the surface (Sonnenfeld and Perthuisot, 1989). Even though Asaila is located at the southwestern fringe of the freshwater lens of Qatar (Fig. 1), subsurface gypsum dissolution in the Rus and underlying Umm er Radhuma Formations results in elevated levels of salinity and brackish to saline groundwater in the southern peninsula (Lloyd et al., 1987;Macumber, 2011Macumber, , 2015. Asaila has relatively moderate salinity levels of around 3000 ppm (Macumber, 2012), which is still sufficient to precipitate capillary evaporites in the vadose zone and, in the eastern sabkha-type sand flats, thin buckled gypsum crusts at the surface. However, comparison with the sediment infill and the thick, halitedominated surface crust at Jaow Aqeeq (ASA-C1) shows how a higher groundwater table and significantly higher groundwater salinity of about 8000 ppm (Macumber, 2012) may lead to more intense precipitation of evaporites in interstitial pore waters and the establishment of inland sabkha conditions. While Jaow Aqeeq seems to directly correspond with the Dukhan coastal sabkha in terms of hydrogeological exchange, the higher limestone plateau separating the Asaila depression from the southeasternmost extension of the Dukhan sabkha is also reflected by the strong SE-NW gradient on the groundwater salinity map in Macumber (2012).

The role of relative sea-level changes in landscape formation
The rate of aeolian sand sedimentation of 1-2 mm yr −1 in the upper 3 m of QAT 63 averaged over the time between ca. 7500 and 5800 years ago as inferred from OSL data is driven by two factors: (i) greater sand availability further north (upwind), as large parts of the Qatar peninsula were then still covered by dune fields , and, (ii) at that time, relative sea level, which is coupled with groundwater levels in low-lying areas near the Qatari coast (Macumber, 2011), rose by several metres (Lambeck, 1996), reaching a highstand around 6000 years ago (Perthuisot, 1977;Engel and Brückner, 2014;Parker et al., 2018;Strohmenger and Jameson, 2018;Rivers et al., 2020). A rising groundwater table and capillary fringe stabilize newly deposited sand in dry climates and lead to net accumulation. Such a raised equilibrium surface close to the shallow groundwater table is referred to as the Stokes surface, below which deflation does not occur due to cohesion of the sediment provided by capillary moisture and initial cementation (Fryberger et al., 1988). It seems that after the sea-level highstand phase (+2-3 m, ca. 6000-4500 years ago; Vita-Finzi, 1978;Engel and Brückner, 2014;Parker et al., 2018;Strohmenger and Jameson, 2018;Rivers et al., 2020), deflation began due to the lowering of the groundwater and the capillary fringe along with sea level, thereby affecting the lowering Stokes surface. The characteristic mounds covering specific areas of the Asaila basin (Figs. 2, 3a) are mainly preserved by porous, platy or needle-shaped gypsum (surface samples ASA 5, 6, 8-10). We assume that these gypcretes are relict horizons of interstitial gypsum in the vadose zone at the time of the sea-level and groundwater-level highstand. Predominant deflation afterwards in the era of sinking groundwater levels led to the removal of the halite-carbonate surface crust (see Sonnenfeld and Perthuisot, 1989), the aeolian sand below and parts of the more massive gypsum crusts. Where the gypsum withstood erosion, the crust protected the underlying deposits from denudation resulting in the formation of mounds.

Origin of the linear mounds
The origin of the linear mounds (Figs. 2, 3b) in the central southern part of the basin, some of which form irregular grids, remains enigmatic. In some parts, they resemble inverted canals as known from other arid environments, e.g. historical Lower Mesopotamia  or southern Peru (Beresford-Jones et al., 2009), but no other indication of an anthropogenic origin was found. Alternatively, the network of linear mounds could follow the pattern of subsurface small-scale joints, which provide better conditions for the capillary rise of water to form stable gypsum crusts. Further in-depth investigations are necessary to shed light on the origin of the linear shapes.

Landscape dynamics and the archaeological record
The limited availability of potable water has always determined ancient settlement patterns in Qatar, explaining why most archaeological sites are located on the northern peninsula. The generally poor groundwater quality in the southern part Macumber, 2015) meant that Neolithic occupation concentrated mostly around the Asaila basin, where access is fairly easy and the salinity is moderate . Only in recent times have further potentially Palaeolithic and Neolithic sites been discovered closer to the coast (Drechsler, 2014). Several Early Neolithic flint knapping workshops were identified along the margins of the Asaila basin based on the spectrum of single diagnostic Qatar-B artefacts or artefact scatters (dated to ca. 7500-6500 BCE). They existed mostly in direct proximity to outcrops of flint raw material (Pfeiffer, 2015;Drechsler et al., 2016), comprising regular blades from bidirectional naviform cores (Inizan, 1988;Pelegrin and Inizan, 2013). In contrast, Middle Neolithic artefacts, i.e. unifacial and bifacial points, scrapers, and bifacially chipped winged and tanged arrowheads following the "Arabian bifacial lithic tradition" sensu Edens (1982Edens ( , 1988) (ca. 6500-4500 BCE), show the greatest concentrations inside the basin (Fig. 11). They occur either as scatters of cores, flakes and flint tools in combination with burnt limestone and ashy sediment in the centre of the basin, indicating both in situ flint knapping and domestic activities, or as single tool findings at the lower western margins, where they point to incidental tool usage and discard . Successful refitting of single pieces of the same flint artefacts within a radius of only a few metres inside one of the survey units inside the basin shows that relocation of artefacts is negligible (Schönicke et al., 2016). There appears to have been quite a substantial occupation in the Middle Neolithic with a focus in the centre of the basin, which, at that time, might have provided a higher potential for grazing of both domesticated and wild animals .
Even though no local proxy record of Holocene climatic changes exists in Qatar, the high-resolution δ 18 O curve from a speleothem at Hoti Cave, Oman (Fleitmann et al., 2007;Fig. 12b), and the palaeo-lacustrine record from Awafi near Ras al-Khaimah, UAE (Parker et al., 2016;Fig. 12d), may provide important references. Both sites received a moisture surplus during Early Holocene to mid-Holocene sum-mer seasons from the Indian Summer Monsoon (Fleitmann et al., 2007) and even more so from the intensified East African Summer Monsoon (EASM) penetrating into the Arabian Peninsula. The role of the mid-level westerlies for the Early Holocene Humid Period (EHHP) in the southeastern Arabian Peninsula, however, remains unclear (Parker et al., 2016). Recent climate model simulations by Jennings et al. (2015) and Guagnin et al. (2016) suggest that Qatar might have had an increase in rainfall following an intensification of the African-Asian monsoonal systems, although far less than the sites of Awafi, Hoti Cave, or another site in south Oman, i.e. Qunf Cave (Fig. 12c). The onset of lacustrine conditions as reflected by the Ti and magnetic susceptibility records from Awafi (Parker et al., 2016;Fig. 12d) overlaps well with the intensification of Middle Neolithic occupation at Asaila Fig. 12g), which at its later stage also benefitted from the high groundwater table connected to the sea-level highstand (Lambeck, 1996;Parker et al., 2018;Strohmenger and Jameson, 2018;Fig. 12e). The Hoti Cave record indicates an even earlier onset of the EHHP, which may reflect more favourable environmental conditions and incipient human occupation at Asaila already in the Early Neolithic (Fig. 12b, g). The same applies to the northern Arabian palaeolake record from Tayma (Saudi Arabia), which reflects only a very short EHHP and appears offset, probably due to a complex interplay of a range of global and regional atmospheric moisture sources, e.g. the EASM, Mediterranean winter rains and winter-spring tropical plumes Enzel et al., 2015;Neugebauer et al., 2018;Parton et al., 2018). Macumber (2018) associates the phase of a wetter climate around the 6th and 5th millennia BCE with the formation of a massive subterraneous freshwater lens over large parts of Qatar. The subsequent absence of archaeological traces at Asaila -and a reduction of sites in the entire country (Muhesen and Al-Naimi, 2014) as well as across the wider gulf region (Uerpmann, 2003) -is associated with aridification on the eastern and southeastern Arabian Peninsula, sea-level drop and associated groundwater level fall (at least after the highstand plateau ended around 2500 BCE), and the onset of a predominant deflation regime inside the basin. After a hiatus, Islamic (610-1972 CE) to Modern (post 1972 CE) pottery and remains of campsites were found (Figs. 11,12g), along with cairns on the higher plateau Drechsler et al., 2016), which are difficult to date Izquierdo Zamora et al., 2015).

Rawdha formation
The uniform sedimentary infill of south Qatar's shallow karst depressions (riyad) appears rather young. The one which was dated in this study (QAT 41) has accumulated its silts and sands over the last 2000-1500 years as determined by OSL data. The low ages of the sediment infill explain the absence of older archaeological findings, which, in the riyad surveyed Figure 11. Spatial distribution of archaeological surface findings (basemap from IKONOS 2004) inside and around the Asaila basin. Selected squares (indicated by mapped surface findings) of 500 m × 500 m were surveyed by systematic back-and-forth walking in order to guarantee a 100 % coverage . Geomorphic units of the innermost Asaila basin are shown for orientation (see details in Fig. 2). "Prehistoric" refers to flint artefacts with unspecific character, which, in theory, may date from any period between the Palaeolithic and historical times  by the SQSP, mainly consist of Late Islamic to Modern period pottery wares, modern trash, Chinese porcelain, coins or temporary mosque structures, all dating to the 18th century CE or younger (Eichmann et al., 2014;Schönicke et al., 2016).
Sedimentation inside the riyad is induced by surface runoff events as indicated by small wadi channels and runnels directed towards some of the larger landforms (Macumber, 2015; Fig. 8). The broad grain size distribution, in particular from the distal record inside the large rawdha (QAT 66), relates to colluvial processes, even though a contribution of aeolian dust as indicated by geochemical data from riyad surfaces of Qatar's interior presented by Yigiterhan et al. (2018) cannot be excluded. As small, endorheic basins, the riyad represent pivotal sites of meteoric groundwater recharge (Macumber, 2011(Macumber, , 2015 and groundwater access through wells. They become flooded during rainfall events, which explains the distribution of remains of temporary camps only along their margins Schönicke et al., 2016). Accordingly, the silt component of their infill (Figs. 9, 10) settled out of suspension. Increased soil moisture, strong evaporation and significant sediment contributions of the local limestone-dominated hamada lead to carbonate crust formation in the vadose zone, as observed in QAT 41, and to incipient cementation. The finer grain size spectrum of QAT 66 (Fig. 10) compared to QAT 41 corresponds with its distal location in the centre of a very large rawdha.
Thus, net accumulation inside the riyad may represent a proxy for surface runoff, which is controlled by rainfall events and, as in our case, by land cover. In the Early Holocene and mid-Holocene, much of Qatar -most probably including the southern riyad -was covered and filled by mobile sands. Only after the southeastward migrating barchan dunes, which lost their source due to the transgression of the gulf (Embabi and Ashour, 1993;, had left the riyad zone, could fine-grained colluvial deposits accumulate during surface runoff events. The rawdha HAR 5183 (QAT 41) is located ca. 21 km upwind of the migrating end of the barchan dunes west of Qatar's inland sea (Khor al-Udaid) (Fig. 1) according to the mean Shamal azimuth of 332 • deviation from the north (Embabi and Ashour, 1993). This azimuth seems to have remained stable during the Holocene based on the similar orientation of drowned barchanoid dunes inside the Gulf of Salwa (Al-Hinai et al., 1987). Considering a migration rate of ca. 8-10 m yr −1 measured for medium-sized barchan dunes in Qatar over the time span of several decades , the karst depression became dune-free around ca. 2600-2100 years ago, or even later, if higher migration rates of smaller-sized barchan Figure 12. Synopsis of regional palaeoclimate and sea-level data in combination with accumulation-deflation phases in Asaila and the southern riyad as well as the chronological density of surface archaeological findings in Asaila. (a) TOC (total organic carbon) record and relative abundance of Poaceae pollen (grasses) (Dinies et al., 2016) combined with phases of the highest lake shorelines  considering revised chronology in Dinies et al., 2015) and maximum grassland expansion (Dinies et al., 2015) from the palaeo-lake and wetlands in the sabkha of Tayma, northern Arabia. Oxygen isotope records of stalagmites from (b) Hoti Cave, northern Oman, and (c) Qunf Cave, southern Oman (Fleitmann et al., 2007) (VPDB is relative to the Vienna Pee Dee Belemnite standard). (d) Ti flux and magnetic susceptibility from the lacustrine record of Awafi, UAE, where low values indicate landscape (dune) stability and higher moisture availability (Parker et al., 2016). (e) Envelope curve of relative sea-level stands from Dosariyah, Saudi Arabia, based on 14 C-dated sea-level index points considering vertical and lateral error margins . It is shown in combination with the global eustatic sea-level function and a regional modelled sea-level prediction for the northern head of the Arabian Gulf (Lambeck, 1996), the peak of which is tentatively shifted to account for 14 C calibration (MSL is mean sea level). (f) Sedimentation rates at Asaila and the southern riyad as inferred from OSL data of this study (Figs. 6, 9b, Table S1), and the inferred shift to a deflation regime. Barchan dune cover at rawdha QAT 41 was estimated based on migration rates inferred by . Furthermore, the number of sites identified in and around the Asaila basin per year of the time span of each period is shown for the different historical and prehistoric periods (EN: Early Neolithic; MN; Middle Neolithic; LN: Late Neolithic; C: Chalcolithic; U: Uruk; D: Dilmun; BA: Bronze Age; IA: Iron Age; H: Hellenistic; S: Seleucid; PIs: pre-Islamic; Is: Islamic) as a proxy of the intensity of human occupation . The grey vertical bar crossing all proxy data curves emphasizes the Middle Neolithic, which is prominently represented by artefacts inside the Asaila basin (Fig. 11). dunes are taken into account. This coincides with the start of the runoff-related silt and sand accumulation of the rawdha at some point before 1500 years ago, as inferred from OSL data. The archaeological record indicates significant human usage of the riyad during the last 300 years, when their infill already resembled today's situation. The only potential climate signal deduced from the surveyed riyad so far may be a shift to more pronounced aridity in very recent times indicated by ubiquitous surface deflation patterns and small, decimetre-scale yardangs (Figs. S33, S34). This process may have supported a concentration of young archaeological findings at the surface. Taking these observations into account, it can be concluded that the further north a rawdha is located, the longer it has been dune-free, and more time may be represented by its silty sediment record.

Conclusions
Among the diverse arid landform units of south Qatar, the Asaila basin as well as the numerous riyad of the central peninsula reveal the richest archaeological record. The former has a density of 108 surface findings per square kilometre extrapolated from the three surveyed squares inside the basin (Fig. 11). In contrast to previous maps of the Asaila basin, which only differentiate between "sabkha" and "aeolian deposits" (Al-Yousef, 2003), we demonstrate greater geomorphic variability and additional signs of (relict) surface runoff. The areas of mounds of different shapes -for the first time systematically investigated in this study -are important indicators for groundwater (and, thus, sea level) control of accumulation and deflation inside the basin. While the 8 m long sediment core reveals a continuous dominance of aeolian sedimentation over the Early Holocene to mid-Holocene, the mounds, cemented by capillary evaporites originally grown in the vadose zone, are a clear sign of deflation after the mid-Holocene sea-level (and groundwaterlevel) highstand. Abundant archaeological evidence of Early and Middle Neolithic occupation -the latter with a clear focus inside the Asaila basin -indicate more favourable living conditions. Whether these imply denser vegetation and better access to groundwater and are linked to the EHHP inferred from other records on the southeastern Arabian Peninsula (e.g. Fleitmann et al., 2007;Preston et al., 2012;Parker et al., 2016) still awaits verification. The existence of more favourable conditions inside the Asaila basin is particularly evident when compared with the adjacent basin of Jaow Aqeeq, which shows how a higher groundwater table and higher groundwater salinity are reflected by higher amounts of capillary gypsum and thicker surface evaporite crusts.
In contrast, the sediment records of the riyad in southern Qatar are very shallow, younger (only ca. 1500 years in the case of QAT 41) and apparently controlled by surface runoff, deflation and the constantly diminishing barchan dune cover over the Middle and Late Holocene (see . The young age of the infill explains the exclusive presence of young artefacts, mainly covering the Late Islamic to Modern periods. In combination with the indicators of current deflation, it may relate to a decrease in rainfall and surface runoff in recent decades to centuries. Whether this shift is related to inactivity and burying of runoff channels identified at the northern margin of the Asaila (Figs. 4, 5) is a matter of future investigations. It remains to be stated that the Late Quaternary environmental changes on the peninsula of Qatar are still largely unknown due to a lack of suitable geological archives. This report adds to the scarce information available, e.g. from Wadi Debayān (e.g. Tetlow et al., 2013), northeast Qatar, or the Ras Abrouq peninsula north of Dukhan, where archaeological layers were encountered down to a depth of 2.6 m (Smith, 1978;Vita-Finzi, 1978), even though chronological resolution and climatic significance still need to be improved. In the future, geophysical prospection of the riyad further north, e.g. using the combined approach of electrical resistivity tomography and seismic refraction tomography as applied for dolines on Crete by Siart et al. (2010), may help to locate thicker rawdha infill with more detailed palaeoenvironmental information through a higher-resolution age model in combination with the analysis of grain size distribution, micromorphology, phytoliths or even pollen spectra. Data availability. OSL and XRD data are provided in the Supplement file. All grain-size-related data will be provided by the corresponding author upon request.
Author contributions. The study was conceived by ME and HB. The archaeological survey was led by CG, PD, KP and RE. PD provided data on the spatial distribution of archaeological findings in Asaila. Geomorphic mapping was performed by SR, ME, and AP. ME, HB, KP, AP, DW and SR took sediment cores. ME, HB and KP logged and sampled the rawdha trenches. JWB conducted the magnetometer prospection and data processing. DB measured luminescence signals and calculated ages. SO conducted XRD measurements and interpreted XRD spectra. ME and DW carried out sedimentological analyses. ME wrote the paper. All authors read, commented on and approved the paper.
Competing interests. The authors declare that they have no conflict of interest.
Special issue statement. This article is part of the special issue "Geoarchaeology and past human-environment interactions". It is not associated with a conference.