Holocene regional population dynamics and climatic trends in the Near East: A ﬁ rst comparison using archaeo-demographic proxies

This paper illustrates long-term trends in human population and climate from the Late Pleistocene to the Late Holocene (14,000 e 2500 cal. yr. BP) in order to assess to what degree climate change impacted human societies in the Near East. It draws on a large corpus of archaeo-demographic data, including anthropogenic radiocarbon dates (n ¼ 10,653) and archaeological site survey (n ¼ 22,533), and 16 hydro-climatic records from cave speleothems and lake sediments. Where possible, inferred population dynamics and climatic trends have been made spatially congruent, and their relationships have been statistically tested. Demographic proxies and palaeoclimatic records have been compared for the greater Near East as a whole and for seven major geo-cultural regions (Anatolia, Arabia, Cyprus, Iran, Levant, Mesopotamia, and South Caucasus). This approach allows us to identify regionalised patterns in population and climate trends. The results suggest a clear relationship between population and climate in the Late Pleistocene and Early Holocene (14,000 e 8326 cal. yr. BP) with population increasing in concomi-tance with wetter climatic conditions. During the Middle Holocene (8326-4200 cal. yr. BP) there is an increased regionalisation of demographic patterns, followed by marked interregional contrasts in the Late Holocene (4200-2500 cal. yr. BP). We identify a decoupling of demographic and climatic trends from the Middle Holocene onwards, and relate this to the existence of more complex societies. These were less vulnerable to gradual climatic shifts due to their logistical infrastructure, social organisation and technological capacity. We also assess the impact of ﬁ ve Rapid Climate Changes (RCC) which occurred during the study period on population levels. Although all ﬁ ve RCC (the so-called 10.2 k, 9.2 k, 8.2 k, 4.2 k, and 3.2 k cal. yr. BP events) are visible to some degree in our palaeoclimatic and demographic proxies, there are marked regional variations in magnitude and duration. © 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http:/


Introduction
Population growth occupies a central role in public debate due to its implications for subsistence strategies, environmental change, and migration, and its relationship with exogenous factors such as climate variations.In the archaeological and anthropological debate, population has been identified as a driver for cultural change (Naroll, 1956;Carneiro, 1962) and an explanation for variation in subsistence strategies (Boserup, 1965;Binford 1968;Shennan 2000;Peregrine 2004), social complexity (Johnson and Earle 2000;Feinman 2011), socioeconomic outputs (Bettencourt et al., 2007;Ortman et al., 2014;Altaweel and Palmisano 2019;Smith 2019) and intra-group conflicts (Goldstone 1993;Kohler et al., 2009;Turchin and Nefedov 2009).More recently, interest in human-environment interactions has prompted an increasing number of studies investigating the impact of human population on landscape, including to what degree population fluctuations were affected by climatic shifts (Langgut et al., 2016;Lawrence et al., 2016;Kaniewski and Van Campo 2017;Bevan et al., 2019;Roberts et al., 2019;Stephens et al., 2019).Building past human population models over the longue dur ee (and assessment of the causes of fluctuations) is critical for our understanding of cultural and environmental changes at a range of scales.
In this context, the Near East represents an excellent laboratory in which to investigate long-term relationships between demographic and climatic trends.This area is a mosaic of different cultural and environmental landscapes, which each experienced linked but different socio-ecological trajectories (Wilkinson 2003;Rosen 2007;Ur 2009Ur , 2015;;Roberts et al., 2011;Wilkinson et al., 2014;Izdbeski et al., 2016;Lawrence et al., 2017;Jones et al., 2019;Palmisano et al., 2019;Woodbridge et al., 2019).It also endured pronounced regional dry/wet episodes during the Late Pleistocene and Holocene which would have impacted on human behaviour (Burstyn et al., 2019;Jones et al., 2019).
The rise and fall of population and the emergence or collapse of complex societies has been attributed to shifts in climatic regimes.For example, the Younger Dryas stadial (~12,700e11,700 cal.yr.BP), a period of increased aridity and cooling, has been interpreted as a cause for a reduction in site occupation intensity (Belfer-Cohen and Bar-Yosef 2000; Guerrero et al., 2008;Goring-Morris and Belfer-Cohen 2010) and as a driver towards more productive techniques in plant cultivation as a response to declining natural sources (Cappers et al., 2002).A recent study by Roberts et al. (2018) showed that neolithisation and population increase occurred in the Fertile Crescent earlier than in the Anatolian plateau, and argued that this delay was caused by more favourable climatic conditions in the former area during the climatic deterioration of the Younger Dryas stadial (~12,700e11,700 cal.yr.BP).The uneven impact of these climatic shifts across the whole Near East could explain local cultural variations manifesting at different speeds and magnitudes.Some trends are also visible at pan-regional scales.For example, studies focused on Anatolia, Northern Mesopotamia and the Southern Levant have all suggested that from 4000/3500 cal.yr.BP onwards demographic trends become decoupled from those of climate, perhaps due to the technological advancement, organisational capacity and logistical infrastructure of more complex societies which makes populations more resilient to drought and food stress (Rosen 2007;Lawrence et al., 2016;Roberts et al., 2019).Alongside these longer-term trajectories, the Near East experienced several major rapid climate changes during the Holocene (RCC, the so-called 10.2 k, 9.2 k, 8.2 k, 4.2 k, and 3.2 k cal.yr.BP events).These are also considered to have affected demographic trends and triggered social responses, although there is much debate on the degree and nature of these.The 10.2 k cal.yr.BP event could be a factor in a cultural break visible across the Northern Levant and Upper Mesopotamia (Borrell et al., 2015).A consensus is beginning to emerge around the 9.2 k and 8.2 k cal.yr.BP events suggesting that they had a more limited impact on Near Eastern communities (Flohr et al., 2016;Allcock 2017), but the 4.2 k. cal.yr.BP event has prompted an animated debate, with some scholars arguing it was responsible for massive population decline and societal collapse (Weiss et al. 1993(Weiss et al. , 2017;;Kaniewski et al., 2018) and other preferring more nuanced views (Wilkinson 1997(Wilkinson , 2007;;Roberts et al., 2011;Greenberg 2017;Cookson et al., 2019).There is a broader agreement on the role of the 3.2 k cal.yr.BP event as a major driver of political and social collapse (Langgut et al., 2013;Cline, 2015;Izdebski et al., 2016;Kaniewski et al. 2015Kaniewski et al. , 2019)).
In this paper, we draw upon a large corpus of archaeological data (in the form of settlements from archaeological surveys and radiocarbon dates) and palaeoclimatic records across the whole Near East in order to provide the first systematic empirical analysis on a pan-continental scale in the very long run.Our assumption is that, all other factors remaining constant, increased aridity will negatively affect soil moisture, and therefore vegetation, food production and human population.However, this relationship is not straightforward, and is mitigated by human responses to change, including social and technological adaptation and population movement.The goal of the paper is to examine to what degree climate and population trends correlate with one another.In order to do this we compare trends visible in the summed probability distributions (SPD) of calibrated radiocarbon dates with several palaeoclimatic records.We also test the robustness of the SPD generated trends by comparing them with more traditional archaeological proxies for population (raw site count and estimated settled area) in three sub-regions.With some exceptions, studies concerning population patterns in the Near East have tended to focus on relatively small areas and short time periods.Here we include the entirety of the greater Near East in the radiocarbon dataset, from the Arabian Peninsula to the South Caucasus, including Anatolia, the Levant, Mesopotamia and Iran.Chronologically we range from the Late Pleistocene to the Late Holocene (ca.14,000e2500 cal.yr.BP).The three sub-regions used for comparison are determined by the availability of collated high-quality archaeological survey data.They include South-central Anatolia, Upper Mesopotamia and the Southern Levant.The fluctuations in demographic and climatic variables are quantitatively assessed to examine the correlation between the different demographic proxies, the correlations between climate change and population dynamics through time, and the impact of RCC events.

The study area
The Near East is a large area with a diversity of landscapes and climate regimes, including fertile drylands, alluvial plains, coastlands, high mountain chains and extensive deserts.Most of the area experiences warmer summers and colder winters, with rainfall occurring in the winter.Extremely dry conditions prevail in the southern part of the region, the Arabian Peninsula, which is mostly arid except in the highlands of Yemen and Oman where precipitation is increased by the Indian Monsoonal system (Enzel et al., 2015;Jones et al., 2019).The coasts of the Mediterranean and Black Seas, the Levant and Anatolia enjoy relatively abundant rainfall and rich agricultural lands (Fisher 2013).Across the entire region there are a series of marked average annual rainfall gradients, with values exceeding 2000 mm in the Caucasus and Zagros Mountains but less than 100 mm in the Arabian Peninsula and Southern Negev (Hemming et al., 2010;Lelieveld et al., 2012).
In order to deal with this variation, we have subdivided the Near East into seven major geo-cultural regions (Anatolia, Arabia, Cyprus, Iran, Levant, Mesopotamia, and South Caucasus; see Fig. 1).The regional classification is still very broad, and each area encompasses a range of landscapes.In delineating our study areas, we have attempted to balance regional scale and relative geographical and cultural homogeneity, with sufficient data coverage for archaeo-demographic proxies, particularly radiocarbon dates, to be effective.The geographical extent of our regions, from the Eastern Mediterranean to the Indus valley and from the Black Sea to the Arabian Sea, allows us to compare regionalised patterns of population and climate.While we have attempted to guarantee spatial and temporal congruence between the demographic and climatic records, this has not always been possible due to the distribution of paleoclimate records and archaeological data.Several of our climate records do not cover the entire chronological scope from the Late Pleistocene to the Late Holocene (14,000e2500 cal.yr.BP), but given the paucity of resources in the region we have included all those currently available in a quantified form.In this study, we refer to the triplex subdivision of the Holocene established by the International Commission on Stratigraphy (Walker et al., 2018): Early Holocene/Greenlandian (11,700e8326 cal.yr.Bp), Middle Holocene/Northgrippan (8326e4200 cal.yr.BP, and Late Holocene/ Meghalayan (4200 cal.yr.BP e 1950 AD).

Archaeo-demographic proxies
Population estimates build on the assumption that the density of the archaeological evidence found in a given study area is proportional to population (see Drennan et al., 2015 for a useful overview).The most popular proxies for inferring long-term changes in human populations in the Near East have been raw counts of archaeological sites and sums of estimated settlement sizes.These are commonly derived from archaeological field investigation such as landscape survey or site mapping (e.g.Sanders 1965;Adams 1965Adams , 1981;;Wright and Johnson 1975;Gophna and Portugali 1988;Finkelstein and Gophna 1993;Wilkinson 1999;Casana 2009;Ur 2013;Lawrence et al., 2016Lawrence et al., , 2017)).Over the past two decades, SPDs of archaeological radiocarbon dates have begun to be used to infer demographic trends in prehistory (see Rick 1987;Shennan and Edinborough 2007;Weninger et al., 2009;Shennan et al., 2013;Silva and Vander Linden, 2017;de Pablo et al., 2019).These approaches have had limited impact in the Near East, and where attempted have focused A. Palmisano, D. Lawrence, M.W. de Gruchy et al. Quaternary Science Reviews 252 (2021) 106739 on single regions or very specific time periods (cf.Borrell et al., 2015;Flohr et al., 2016;Palmisano et al., 2019;Woodbridge et al., 2019).Here we make use of a total of 10,653 radiocarbon dates from 993 sites collected from existing online digital archives (ArAGATS project: 2019; BANADORA: CDRC 2016; Borrell et al., 2015;CalPal: Weninger et al., 2018;CONTEXT: B€ ohner and Schyle 2006;EUROEVOL: Manning et al., 2016;Flohr et al., 2016 Reingruber and Thissen 2017), and electronic and print publications.To our knowledge, this is the largest collation of published radiocarbon dates for the Near East.We have not been able to include unpublished dates or those published in non-digital grey literature.Wherever possible the radiocarbon dates were crosschecked against different sources and georeferenced consistently (see Fig. 1).All the radiocarbon dates come from archaeological contexts and most of them are taken from samples of bone, charcoal, and seeds (see Table 1).Radiocarbon dates with poorly understood marine reservoir offsets (all shells), a standard error greater than 300 years, or that do not have anthropogenic causes (e.g.radiocarbon samples collected from environmental cores) have been excluded and are not part of the above total.The total number of dates exceeds the minimum sample size of 200e500 required to produce reliable SPDs of calibrated radiocarbon dates within a time interval of 8e10,000 years (Michczy nska and Pazdur, 2004;Michczy nska et al., 2007;Williams, 2012, 580e581).As a consequence, our dataset can be considered to be sufficiently large to overcome many of the potential sampling biases that might affect the patterns.The SPDs of calibrated radiocarbon dates are effective until 3000/2500 cal.yr.BP because after that archaeologists rely more on short-lived pottery types for dating archaeological layers.This period is also affected by the Hallstatt radiocarbon calibration plateau (ca.2750e2350 cal.yr.BP) which makes it difficult to obtain refined radiocarbon-based chronologies.
Both archaeological survey data (raw count, estimated settlement size) and radiocarbon dates are imperfect proxies and are subject to several issues that may affect the relationship between the archaeological data and past population levels.These include biases in research focus towards particular periods or regions, variations in the methods adopted and the intensity of the investigation, taphonomic loss and the visibility of diagnostic artefacts (Cherry 1983;Surovell et al., 2009;Contreras and Meadows 2014;Torfing 2015Torfing , 2016;;Becerra-Valdivia et al., 2020).The generation of SPD datasets requires excavation and sampling for radiocarbon dating.In areas of the world such as the Near East where modern states do not routinely integrate archaeological work into their planning and construction industries, the majority of archaeological excavation is driven by academic research projects and their attendant interests.As a result, such areas may be more affected by the concentration of research on particular topics, such as the emergence of farming, which in turn emphasises particular periods, than areas with a long tradition of commercial archaeology.This is less of a problem for archaeological surveys, which generally attempt to capture all sites within a landscape regardless of period (unfortunately often with the exception of very recent occupation).Unlike survey, SPDs do not capture variables such as site size or type.Expansion in the size of sites can lead to a counter-intuitive state of affairs in which the number of sites reduces but the population likely increased.This occurred in Southern Mesopotamia during the period of initial urban emergence from 6000 cal.yr.BP (Adams 1981), and in some areas of Upper Mesopotamia during the Early Bronze Age (Lawrence and Wilkinson 2015).However, SPDs of calibrated radiocarbon dates generally provide a better chronological resolution than artefactual data because the latter rely on the identification of changes in lithic and ceramic assemblages to define chronological periods and these changes may occur relatively slowly.Archaeological surveys which rely on surface collection cannot include stratigraphic information or data on proportions of pottery types, resulting in further chronological imprecision.
Despite these issues, several studies have shown broad agreement in the demographic trends produced by the SPD of radiocarbon dates and other archaeological indices (e.g.raw site count, estimated settlement size) (Tallavaara et al., 2010;French 2015;French and Collins 2015; Demjan and Dreslerov a2 0 1 6 ; Palmisano et al., 2017;Nielsen et al., 2019).The "Changing the Face of the Mediterranean" project, a Plymouth University-UCL collaboration, has confirmed that SPDs can be regarded as a robust proxy for modelling past human population in a region adjacent to our study area (see Bevan et al., 2019;Berger et al., 2019;Palmisano et al., 2019;Roberts et al., 2019;Stoddart et al., 2019;Weiberg et al., 2019).Here we assess the limitations of the different proxies through a cross-comparison between multiple archaeodemographic proxies.Archaeological site derived proxies (site count, settlement size) are compared to the SPD of calibrated radiocarbon dates for three sub-regions, South-central Anatolia, the Southern Levant and Upper Mesopotamia, to develop a better understanding of long-term population dynamics (see Table 1).The sub-regions were chosen because they have sufficient high quality systematic archaeological survey data to assess general trends.A fourth sub-region which has a similar archaeological survey dataset in our study region is Southern Mesopotamia, but here the paucity of excavation since the early 1990s means the number of published radiocarbon dates falls below the threshold for robust analysis.The majority of the cited papers where analogous work has been undertaken focus on relatively homogenous fertile lowland environments, and our study regions are similar.This means they are a poor test of the validity of the SPD data for dissimilar environments such as desert Arabia or mountainous Eastern Anatolia and the Caucasus, which also had rather different settlement histories.Further work is needed to bring together archaeological settlement datasets in these regions of sufficient size and scale to compare to the SPD record.However, SPDs are beginning to be used in some of these regions (Petraglia et al., 2020), and in the absence of appropriate settlement datasets we would argue that they can act as an empirically derived heuristic device for examining the causal relationship between climate change and population.This is not to claim that SPDs of radiocarbon dates are a more appropriate proxy than archaeological survey data for population.We use them here due to their wider spatial and temporal availability, and chronological precision.Ideally we would use both archaeo-demographic proxies throughout the whole study area and for the entire study period, but settlement datasets at these scales are not yet available.
The archaeological settlement data for South-central Anatolia and Upper Mesopotamia were collected and harmonised from reports and gazetteers of archaeological surveys of varying intensity (see the Appendix B for references).Settlement data were recorded as geo-referenced points (unprojected WGS84) per standardised cultural periods, which have been defined in absolute calendric years to provide maximum comparative potential among archaeological sites.Here we include those places identified as habitation sites or possible habitations and have removed sites such as cemeteries or mines where no evidence of settlement was recovered.The archaeological survey data cover a period spanning from 10,000e2500 cal.yr.BP as, unlike the radiocarbon dates, there are insufficient data for the earlier periods.We recorded a total of 1336 sites and 3543 occupation phases for south-central Anatolia and 509 sites divided into 1772 site-phases for Upper Mesopotamia (Table 1 and Fig. 1: a, c; see the Appendix B for the full list of archaeological surveys).A total of 20,688 sites and 66,183 occupation phases were collected for the Levant and then standardised from two extant online databases: 1) The Digital Archaeological Atlas of the Holy Land (Savage and Levy 2014), and 2) The West Bank and East Jerusalem Archaeological Database (Greenberg and Keinan 2009).One major caveat in this latter dataset is that the estimated size of settlements for each cultural period was not consistently available, and therefore we have only used the raw site count as a proxy for the population.

Palaeoclimatic records
For simplicity, in the present study we have used a palaeoclimatic proxy reflecting past hydro-climatic patterns and omitted those proxies indicating variability in temperature or pollen-based climate reconstructions.We have only included datasets where full results have been published and raw data are available.The palaeoclimatic dataset includes 16 stable oxygen isotope ratios (d 18 O) of speleothems and lake sediments collected from different regions (see Fig. 1 and Table 2) which provide relative precipitation levels.In the arid and semi-arid regions which make up most of the study area, we expect precipitation levels to correlate closely with vegetation and surface water availability (Jones et al., 2019).Although the spatial coverage of the available records is heterogeneous, with some areas such as Cyprus and the central part of the Fertile Crescent lacking natural palaeoclimatic archives, it is possible to identify regional climate variability across the whole Near East.A further issue to bear in mind is that most of our palaeoclimatic records (12 out of 16) are located at more than 650 m above sea level and reflect past climatic conditions in upland landscapes (see Table 2), while most of our archaeological data comes from lowland basins.
Stable oxygen isotopes from caves and lakes are used as hydrological indicators due to their rapid response to changes in water availability, and their fine-grained temporal resolution based on radiocarbon or uranium-series dating (Bar-Matthews et al., 2003;Fleitmann et al., 2007;Finn e et al., 2017).The higher and lower values of stable oxygen isotopes indicate respectively drier and wetter conditions.However, the interpretation of d 18 O values is not always straightforward as they can be affected by several environmental factors (e.g.vegetation, recharge conditions, open water evaporation) occurring in the original context of a given record.In addition, some authors have recently suggested that the d 18 O of the speleothem calcite represents the precipitation amount of the winter season rather than average annual conditions (Wassenburg et al., 2016;Deininger et al., 2017;Bini et al., 2019), and there is no unifying explanation for values across the Mediterranean basin (Moreno et al., 2014).A recent study by Baker et al. (2019) has highlighted that those sites from regions (eg.Northern Europe, northern America) with an annual mean temperature less than 10 C have an oxygen isotope composition strongly related to the isotopic composition of local rainfall.Instead, those regions, such as the Mediterranean basin and Near East, with an annual mean temperature between 10 and 16 C, can be more susceptible to moisture balance change due to the evaporative fractionation of stored karst water and selective recharge.Conversely, the d 18 O values of lacustrine carbonate are believed to correlate with hydroclimate conditions during summer (Leng and Marshall 2004;Bini et al., 2019) although more complex explanations have been proposed (Zielhofer et al., 2019).These palaeoclimatic records can be sensitive to air mass trajectories and to shifts in evaporative concentration.
It is not the main goal of this paper to describe in detail the uncertainties of each climate record, and several recent reviews provide excellent syntheses of the available data (Roberts et al., 2008;Burstyn et al., 2019;Finn e et al., 2019;Jones et al., 2019).We invite the reader to refer to these and the original publications for detailed descriptions of each record (see Table 2).All the palaeoclimatic records of this study were collected from existing online repositories and databases (NOAA; SISAL database: Atsawawaranunt et al., 2019).We selected those records which were freely accessible online, could provide reasonable spatial coverage and included multimillennial time duration across the Holocene.While some records have a fine and detailed chronology (for example, Jeita and Sofular caves) other ones have a mean sampling interval greater than 200 years (for example Jerusalem West cave and Lake Zeribar).

Methods
In this paper we draw on developed methods to infer past population dynamics from archaeological settlements data (Palmisano et al., 2017) and to test statistically the SPDs of calibrated radiocarbon dates (cf.Shennan et al., 2013;Timpson et al., 2014;Crema et al., 2016;Bevan et al., 2017).We provide here a general description of those methods, which are explained in detail in the original publications.All analyses and figures are reproducible thanks to the dissemination of the datasets and four scripts written in R statistical computing language (Appendix A).

Archaeology settlement data derived proxies
The archaeological settlement data (raw site count, aggregated estimated settlement size) have been binned into a series of 200year time slices starting at 10,000 cal.yr.BP (period t 1 : 10,000e9800 cal.yr.BP) and ending at 2600 cal.yr.BP (period t 37 : 2800-2600 cal.yr.BP).Because the length of cultural periods varies according to the precision of the dating of archaeological artefacts (e.g.pottery), with the earlier periods often spanning hundreds of years, we applied aoristic analysis to deal with the temporal uncertainty of occupation periods and generated aoristic weights (for a more detailed explanation of the methodology see Crema et al., 2010Crema et al., , 1118 e1121 ; Crema 2012,4 4 6 e448; Palmisano et al., 2017, 63e65).The method assumes that the total probability of an archaeological event (site occupation phase in our case) within a given time span is 1, which indicates an absolute certainty that the site was in use in that time span.If we then divide by the length of the site's chronological range, we can represent the probability of existence for each temporal block (implicitly adopting a default uniform assumption).For instance, using time-steps of 200 years, a Middle Bronze Age site-phase ranging from 2000 to 1600 BC has an aoristic weight of 0.50 for each time-step (2000e180 0, 180 0e1600).In addition, to mitigate the discrepancy between long individual phases within typo-chronological schemes and likely shorter term durations of site occupations, we applied Monte Carlo methods to generate randomised start occupation periods for those sites with low-resolution information, assuming a mean length of occupation of 200 years within each cultural period (cf.Crema 2012, 450e451;Orton et al., 2017,5 e6;Palmisano et al., 2017, 63e64).The 200-year duration was used to correspond to the median duration of the cultural periods (or site-phases) of the settlements employed in this study and to offer a clear contrast for those periods having large time spans (e.g.1000 years or more).The resulting probabilistic distributions (aoristic weights, randomised start dates) provide a robust dataset to compare with other archaeo-demographic proxies.

Creating SPD of calibrated radiocarbon dates
The SPD of calibrated radiocarbon dates is the result of counting up (summed in the manner of a histogram) the calibrated raw radiocarbon years of each organic sample, which are expressed in the form of probability statements with error ranges.The idea is that the more people living in a given region, the more garbage, the more organic materials, and the more radiocarbon collected, and dated (Rick 1987).
We used the R package rcarbon (version 1.4.1;see Bevan and Crema 2020 ; Crema and Bevan 2020) for generating and statistically assessing the SPDs of calibrated radiocarbon dates.We calibrated and summed individual radiocarbon dates within a broader temporal coverage than our study period, between 15,000 and 2000 cal.yr.BP, to avoid any edge effects.We reduced any bias of oversampling of specific chronological phases or events (for example, multiple dates being taken from a single floor layer) by aggregating uncalibrated radiocarbon dates within 50-year bins; dates with a gap greater than 50 years were assigned to a separate bin.We then calibrated the radiocarbon dates and pooled mean probabilities within the bins to have only one date distribution for each 50-year archaeological context (cf.Timpson et al., 2014).Hence, the probability distributions of 10,653 calibrated radiocarbon dates from 993 sites have been aggregated into 4873 (50year) site bins and finally summed to produce SPDs for the Near East as a whole and for the seven sub-regions (Anatolia, Arabia, Cyprus, Iran, Levant, Mesopotamia, and South Caucasus).We opted to sum unnormalised distributions of calibrated radiocarbon dates as some recent work (Weninger et al., 2015;Bevan et al., 2017) has suggested that normalised dates create artificial peaks in the resulting SPDs due to the steepening portions of the radiocarbon calibration curve (we have used the IntCal20 curve; Reimer et al., 2020).
We compared the observed SPD of calibrated radiocarbon dates with two theoretical null models of demographic change to statistically determine if the inferred population fluctuations indicate meaningful departures from what may be expected by mere chance.Put simply, the observed SPD were compared with conditional random sets of hypothetical dates produced according to two theoretical null models of population growth (exponential and logistic).First, we fitted both an exponential and a logistic model of demographic growth to the observed data, and then backcalibrated new random samples (equal to the number of sitebins) drawn from the fitted model (Bevan and Crema 2020: mod-elTest).These hypothetical samples were then calibrated and their probability distributions summed to generate an expected SPD of the fitted null model.We repeated the same process 1000 times to provide a test of global goodness-of-fit and a 95% confidence envelope (in grey in the relevant figures).Departures of the observed SPD (solid black line in the figures) above and below the confidence envelope indicate respectively periods of increase (in red in the relevant figures) or decrease (in blue in the relevant figures) greater than expected according to a null model of growth.Although both logistic and exponential models are theoretical abstractions not describing any empirically derived population growth, they are useful for assessing significant patterns in population fluctuations across time (Turchin 2001).The logistic model, based on the Malthusian assumption that population growth is limited by a maximum imposed by the carrying capacity of the environment, could fit a scenario characterised by a dramatic increase in population with the introduction of farming and a gradual demographic slowdown in the later periods as local resource limits are reached (Bevan et al., 2017;Palmisano et al., 2019).The exponential null model has the advantage of mimicking site loss through time and/ or expected multi-millennial demographic trends (Shennan et al., 2013).Finally, we assessed regional variation in population dynamics by performing a permutation test to assess to what degree the SPDs of each region depart from a null model representing the panregional demographic trends across the whole Near East (cf.Crema 2016; Bevan and Crema 2019: permTest).This technique shuffles the labels identifying from which region each bin comes and generates 1000 SPDs from which a 95% confidence envelope is derived.Deviations from the null model indicate periods in which the population increase (in red) or decline (in blue) of each region is greater than the pan-regional trend.

Demographic trends inferred from SPDs of radiocarbon dates
We generated a normalised (black line) and unnormalised (in grey) SPD of calibrated radiocarbon dates (Fig. 2a) for the period from 14,000e2500 cal.yr.BP.Despite our preference for summing unnormalised calibrated dates, which avoids artificial spikes in the resulting SPDs, the two curves appear very similar, as shown by a high Pearson correlation coefficient (r ¼ 0.91, p-value < 0.01).Given that most dates come from wood charcoal or unidentified material, we also produced an SPD of unnormalised radiocarbon dates from short-lived radiocarbon samples (e.g.bones, collagen, seeds, grains).The SPDs including all dates (Fig. 2a) and only short-lived dates (Fig. 2b) are highly correlated (r ¼ 0.95, p-value ¼ 0.01, Pearson) and the so-called old wood effect seems not to have significantly affected the resulting post-calibration probability densities.Fig. 2c and d shows the comparison of the SPD of unnormalised calibrated radiocarbon with logistic and exponential null models respectively.The observed SPDs (black solid line) show coincident positive (in red) and negative (in blue) deviations from the 95% confidence interval only from 5300 cal.yr.BP onwards (Fig. 2c and d).The inferred demographic trends show significant departures from the null model and on the basis of the global pvalue (<0.01) we can say that the population did not grow neither logistically nor exponentially from the Late Pleistocene to the Late Holocene (14,000e2500 cal.yr.BP).The SPD of calibrated radiocarbon dates suggests that population increased prominently at around 12,700e12,600 cal.yr.BP, remained steady during the Younger Dryas stadial (~12,500e11,700), and saw significant growth at the beginning of the Holocene (~11,600e11,300 cal.yr.BP).Continuous growth of population occurred until 8500 cal.yr.BP.In the Middle Holocene (between 8500 and 5500 cal.yr.BP), the population was quite stable and was characterised by alternating periods of moderate increase and strong decline.The population increased strongly between 5300 and 4200 cal.yr.BP and peaked during 3200e2900 cal.yr.BP.The final decline in SPD after 2800 cal.yr. is likely not reflective of past population, but is rather due to the fact that researchers rely less on radiocarbon dating given the problem represented by the Hallstatt plateau, as explained above, and greater refinement in ceramic typologies.
Fig. 3 shows the regionally subdivided SPD of unnormalised radiocarbon dates compared against the pan-regional trend (grey envelope) described above.We examined to what degree regional demographic fluctuations depart from the pan-regional trends via a permutation test (cf.Crema et al., 2016).This technique also deals with issues resulting from the different sizes of the samples, as the grey envelopes of pan-regional trends produced are larger in those regions with fewer radiocarbon dates, reflecting greater uncertainty.All seven regions show significant departures from the panregional trend over the long-term (p-value < 0.01).The population level was quite low across the whole Near East during the Epipaleolithic, with the Levant and Mesopotamia showing significant positive departures from the pan-regional trend (Fig. 3d and e).In contrast, during the Holocene we have greater regional variation.Anatolia and the South Caucasus show similar demographic trends with a marked increase of population between 8000 and 7000 cal.yr.BP and during the Bronze Age (5000-3000 cal.Yr.BP), punctuated by a general decline in the Chalcolithic.The South Caucasus, unlike Anatolia, shows a significant positive departure from the pan-regional trend during the Late Bronze and Iron Ages (3500-2500 cal.yr.BP; Fig. 3b).The Levant and Cyprus show similar population patterns across the Holocene with some striking differences in particular periods.For instance, in Cyprus population departs positively from the global pattern during the Aceramic Neolithic (11,000e9500 cal.yr.BP) and the Late Bronze Age (Fig. 3c), while in the Levant we can see that population levels off significantly between 9000 and 7200 cal.yr.BP and reaches a peak in the Early Iron Age (3200-2800 cal.yr.BP, Fig. 3d).Mesopotamia shows positive deviations during the Pre-Pottery Neolithic (PPN), Neolithic and Early Bronze Age, while population declines noticeably in the later periods (Iron Age), likely related to a combination of a lack of excavations, an increase in ceramic typology accuracy and the Hallstatt plateau in the radiocarbon curve discussed above (Fig. 3e).Iran shows alternating patterns of population booms and busts across the whole Holocene (Fig. 3f).Finally, the Arabian Peninsula shows a significant departure from the pan-regional trend during the Neolithic (7400-6400 cal.yr.BP) and a marked demographic drop in the Bronze Age (Fig. 3g).However, in this area the available radiocarbon dates are patchy and, therefore, the inferred population dynamics should be regarded cautiously.
Pairwise Pearson's correlations between all regional SPDs have been calculated in order to assess how demographic patterns differed among the regions in three discrete sub-periods: 1) Late Pleistocene and Early Holocene (14,000e8326 cal.yr.BP, Table 3); 2) Middle Holocene (8326e4200 cal.yr.BP, Table 4); 3) Late Holocene (4200e2500 cal.yr.BP, Table 5).The results show similar demographic trends across the Near East in the Late Pleistocene-Early Holocene except for the South Caucasus (Table 3), increased regionalisation during the Middle Holocene (Table 4), and marked interregional differences in the Late Holocene (Table 5; see Fig. 4a).

Assessing population dynamics from multiple proxies: a standalone evaluation
Here, we compare the demographic trends for three areas where we have both site survey data and radiocarbon dates: Anatolia, Southern Levant and Upper Mesopotamia (Fig. 5).The archaeological settlement data derived proxies (raw count, total aggregated estimated size, aoristic weight, randomised start occupation dates) and the SPD of calibrated radiocarbon dates have been normalised on a scale from 0 to 1 in order to make them comparable.All proxies have been binned into 200-year time slices to correspond to the median site-phase lengths.
In Anatolia, we can see a correspondence between the sites derived proxies and the SPD of radiocarbon dates (see Fig. 5a and Table 6).The correlation between the SPDs of radiocarbon dates and the other proxies is stronger between 10,000 and 4000 cal.yr.BP (Table 7).This is due to the fact that in the later periods the SPDs tend to underestimate the population as archaeologists rely more on ceramic-based typo-chronologies.However, the settlement data appear to be poorly correlated with the SPD of south-central Anatolia (Table 6).This issue is related to the fact that the SPD curve for central Anatolia is skewed by the disproportionately large number of dates from Çatalh€ oyük (282 out of 902), and by the fact that radiocarbon dates have not been published for several known Bronze and Iron Ages centres.In the Levant, the settlement derived proxies and the SPDs appear to be strongly correlated (Fig. 5b, Tables 8 and 9).The Early Bronze Age (5400e4000 BP) is overrepresented in the SPD, and there is a major spike in the early part of the Iron Age (3300e2500 BP).Both of these can be interpreted as a result of biases in research intensity.The EBA saw the emergence of complex polities and as a result has received significant scholarly attention, while the interests of Biblical Archaeology in the Southern Levant have resulted in a large number of projects focused on this period.In Upper Mesopotamia, the settlement derived proxies and the SPD of radiocarbon dates are quite well correlated during the period between 8400 and 4000 cal.yr.BP (Fig. 5c; Tables 10 and 11).Fluctuations in the SPD during the Late Chalcolithic (6500e5000 cal.yr.BP) are not reflected in either of the settlement record proxies and as in the Southern Levant there is a peak in the Early Bronze Age which may in part reflect biases in research agendas.The settlement data in this region is especially poor for the Early Holocene, when the SPD suggests very high levels of population.Here unusual research interest in early farming communities is likely inflating the SPD, while surveys reliant on ceramics as key indicators of archaeological sites struggle to identify pre-pottery occupation.Conversely, from 4000 cal.yr.BP onwards the SPD depict a drastic drop in the population, while surveys suggest a minor decline followed by a major increase.
While the initial decline during the Late Bronze Age may be genuine, during the Iron Age surveys have identified large numbers of small rural sites, often at new locations away from the long-term occupations on settlement mounds (Wilkinson 2003).Very few of these have been subject to excavation and are therefore not captured by the SPD.
To summarise, the overall agreement between the radiocarbon SPD dates and the settlement-based proxies corroborates the use of the former as a quite good indicator for inferring past human population dynamics.However, it is important to stress that these proxies should be used complementarily and contextually to provide a more complete picture.Broadly speaking, the curves inferred from survey data tend to minimise the population levels in the earlier periods while the SPDs of radiocarbon dates overestimate population in prehistoric periods.Survey projects tend to rely on pottery for dating, and therefore prehistoric sites with no pottery or very long lived types are underrepresented.Excavators at such sites tend to take more radiocarbon dates than those working in later periods.A clear caveat in the SPD for Anatolia and Upper Mesopotamia is that they fail to show the demographic booms inferred from increased site counts and total settled area related to the Hittite (ca.3600-3200 cal.yr.BP; Fig. 5a) and Neo-Assyrian (ca.2900-2600 cal.yr.BP) Empires (Fig. 5c).In the discussion section below, we interpret our wider results in the light of these limitations.

A reconstruction of climate during the Late Pleistocene and the Holocene
The available palaeoclimatic records indicate that after a long period of cooling in the Late Pleistocene (the Younger Dryas: ~12,700e11,700 cal.yr.BP), which reversed the gradual climatic warming of the Bølling-Allerød interstadial (~14,700e12,700 cal.yr.BP), the onset of the Holocene was characterised by a switch towards warmer and wetter climatic conditions (Grootes and Stuiver 1997;Roberts et al., 2008;Dean et al., 2015).The transition from the Late Pleistocene to the Early Holocene seems to have occurred abruptly, as suggested by some records from Anatolia (see Fig. 6: 3e4, 6), the Levant (Fig. 6:7e10), and the Iranian plateau (Fig. 6:12).
The d 18 O values of the Sofular and Karaca caves are in contrast with all the other records in this study, showing a gradual decrease of precipitation since the beginning of the Holocene (Fig. 6; 2).However, the records from these two caves seem to reflect change in the d 18 O values of the Black Sea, the main source of moisture for these sites, rather than variation in rainfall (Fleitmann et al., 2009;G€ oktürk et al., 2011).The Holocene was characterised by wetter conditions between 10,000 and 7000 cal.yr.BP with peaks in rainfall at ~8500 and 7500 cal.yr.BP (Fig. 6;B urstyn 2019; Jones et al., 2019).After 7000 cal.yr.BP there is a progressive shift towards more arid conditions in most of the records.Although the general trends described above are quite homogeneous across the whole Near East, the records from Hoti and Qunf caves, both located in Oman on the Eastern side of the Arabian peninsula (Fig. 6: 15e16), show a decrease in precipitation one millennium later, at ~6000 cal.yr.BP, due to the strength of the Indian Ocean monsoon (Neff et al., 2001;Fletimann et al., 2007).The Late Holocene was generally a dry period characterised by pronounced sub-centennials fluctuations of dry episodes (4300e3800 cal.yr.BP; 3200-2900 cal.yr.BP) punctuated by a period of increasing precipitation (~3800e3200 cal.yr.BP; Grant et al., 2012;Burstyn et al., 2019).At a shorter timescale, the Holocene was characterised by multicentennial and multi-decadal dry and wet episodes that caused abrupt changes in the longer millennial climatic trends.In this study, we will focus on five major RCCs (the 10.2 k, 9.2 k, 8.2 k, 4.2 k, and 3.2 k cal.yr.BP 'events') that have been the focus of academic debate addressing the human responses to climate change (see Table 12).The 10.2 k cal.yr.BP event took place between 10,200 and 9800 cal.tr.BP and is visible in some palaeoclimatic records from Anatolia (Fig. 6:1e2, 5e6) and the Levant (Fig. 6:7 e8; Table 12).The 9.2 k cal.yr.BP event seems to have a relatively small impact in the Near East as the only records showing a substantial decrease in precipitation are Lake G€ olhisar and Sofular cave in Anatolia (Fig. 6:1e2) and the Hoti and Qunf caves in Oman (Fig. 6:1 5 e16; Table 12; Fleitmann et al., 2015).The 8.2 k cal.yr.BP event took place between 8250 and 8000 cal.yr.BP and was characterised by a decrease in global temperatures and snow accumulation rate (Thomas et al., 2007;Cheng et al., 2009).This event is visible in most of the records used in this study, although Lake Hula in Northern Israel (Fig. 6:8) and Lake Mirabad in Iran (Fig. 6:13) saw an increase in rainfall (Table 12).The 4.2 cal.yr.BP was a cold and dry episode between 4300 and 3800 cal.yr.BP, which marked the transition to the Late Holocene and a shift towards more arid climatic conditions.Several scholars have argued that it caused considerable environmental stress affecting Near Eastern societies (Weiss 2016;Kaniewski et al., 2018;Bini et al., 2019).This abrupt climatic shift is evident in almost all records and seems to indicate a drying phase across the whole Near East (Fig. 6; Table 12).The final significant RCC occurred between 3200 and 2900 cal.yr.BP and is in line with the longer multi-millennial drying trend characterising the Late Holocene (Fig. 6; Kaniewski et al., 2015 and2019).

Comparing demographic and climatic trends over long and short timescales: a regional analysis
The SPDs of calibrated radiocarbon dates and the z-scores of those palaeoclimate records with a sufficiently fine resolution have been binned into 50-year time slices.The use of a 50-year time window is only possible for those palaeoclimate records having a mean sampling interval lower than 50-years across the whole chronological scope (Fig. 6:2,7,11,14e16).
The palaeoclimate records with a coarser chronological resolution and with dates greater than 50 years from one to another have been binned into 100-year (Fig. 6: 4), 300-year (Fig. 6:1,3,5,6,8) 400-year (Fig. 6:13), 500-year (Fig. 6:10,12) and 600-year (Fig. 6:9) time slices.The SPD were binned into coarser time slices when compared with these palaeoclimate records.For those palaeoclimatic records with a finer resolution and a long chronological span, we have calculated a 500-year time window Pearson correlation with ten 50-year bins in each time window (Fig. 6: 2,7, and 16).The advantage of this approach is to identify periods of correspondence and divergence between human population size and palaeoclimate records over shorter periods from the Late Pleistocene to the late Holocene (14,000e2500 cal.yr.BP).We then calculated Pearson correlations with maximum and minimum values ranging respectively between 1 and -1 in order to assess if the population dynamics are positively or negatively correlated with hydro-climatic trends.
In Anatolia, the long-term demographic and climatic trends show diverging patterns, as shown by strong negative Pearson correlation values ranging between À0.39 and À0.74, with the only   exception being Lake Van (r ¼ 0.58, p-value <0.01).This is not surprising given that the population constantly and incrementally increases from the onset of the Holocene, while most of the palaeoclimatic records show a slow and gradual multi-millennial trend towards more arid conditions (Fig. 7).However, on a shorttime scale, we can see that the demographic rise in the Epipalaeolithic and Early Neolithic coincided with a wetter climate at the beginning of the Holocene, as shown by the correlation value with the shorter duration dataset from Dim cave (r ¼ 0.67, p-value <0.01).The significant increase in the population occurring between 8500 and 8000 cal.yr.BP seems to correspond with wetter conditions detectable in some records (Fig. 7:1 , 5 e6), while the decline during the Chalcolithic shows a good correspondence with a climatic drying phase after 7500 cal.yr.BP visible in all records except Lake G€ olhisar.The population increased again between 5300 and 4400 cal.yr.BP despite a general arid trend and seems to have been negatively affected by the RCCs at 4.2 and 3.2 cal.yr.BP.Although to a minor degree for the total area of the sites, these two events are also present in the archaeological settlement data derived proxies (Fig. 5a).
Given the lack of available palaeoclimatic records in the South Caucasus, we compared the SPD against the nearest climatic records in our dataset, such as Karaca cave and Lake Van (Fig. 8).The lack of data does not allow for a direct comparison between demographic and climatic trends for the earlier periods.The increases in population during the Neolithic and the Early Bronze Age (EBA), interspersed by a substantial decline during the Chalcolithic, seem to correspond with two relatively wet periods separated by a drying climate between 7000 and 5600 cal.yr.BP.The population declined significantly in coincidence with the RCC at 4.2 cal.yr.BP and peaked dramatically during the Late Bronze age and Iron Age despite arid conditions.As for the South Caucasus, we compared the SPD of radiocarbon dates from Cyprus with the closest palaeoclimatic records on the Levantine (Jeita Cave and Lake Hula) and Anatolian (Dim cave) coasts.An increase in the population and wetter climatic conditions during the Early Holocene appears synchronous, as suggested by strong Pearson correlation values between the SPD with the climatic trends provided by Dim cave (r ¼ 0.67, p-value<0.01)and Jeita cave (Fig. 9).The 500-year moving window approach seems to identify a sort of sinusoidal cycle of alternating strong and negative correlations between demographic and climatic trends during the Middle Holocene.In the Late Holocene, the population grew substantially and peaked during the Late Bronze Age (3500-3000 cal.yr.BP) despite drier climatic conditions.However, the SPD curve and the climate patterns from Jeita cave are highly correlated between 3500 and 2500 cal.yr.BP (Fig. 9).
In the Levant, the population began to increase during the Epipalaeolithic, and a demographic bulge occurred in concordance with wetter climatic conditions between 10,500 and 9500 (Fig. 10).The initial demographic boom was followed by a decline in population during the Pottery Neolithic (ca.8500-6500 cal.yr.BP), perhaps due to an overall climatic drying trend.The population peaked during the Early Bronze Age despite a drying of the climate and then declined markedly between 4200 and 3200 cal.yr.BP, a period characterised by pronounced sub-centennial dry-wet episodes superimposed onto an overall millennial drying trend.However, the population reached a peak in the Iron Age during a period of increasingly arid climatic conditions (Fig. 10).
In Mesopotamia, the population saw a first substantial growth in coincidence with ameliorated climatic conditions after the onset of the Holocene and boomed significantly between 9300 and 8300, when a peak in rainfall was reached (Fig. 11).A climatic drying trend between 8000 and 5000 seems to correlate with a fall in the population.Population peaked during the Early Bronze Age despite an even more arid climate.While a drop in population at the end of the Early Bronze Age could be related to the 4.2 k cal.yr.BP event, and the following decline during the Late Bronze Age is also confirmed by several studies using archaeological survey data (e.g.Ur 2013;Lawrence et al., 2017), very low demographic levels in the Iron Age are unrealistic and most likely due to a lack of a systematic collection of radiocarbon samples for this period.In fact, as indicated by the northern Mesopotamia settlements data (Fig. 5c), population started increasing and boomed during the Assyrian Empire.
In Iran we have a gradual increase in the population at the same time as wetter climatic conditions with the onset of the Holocene (Fig. 12).The significant booms and busts of the population during the Pottery Neolithic and Chalcolithic seem to be correlated with the climatic fluctuations inferred from the Lakes Zeribar and Mirabad.The time-series of the SPD curve and these two palaeoclimatic records are slightly offset because Lake Zeribar and Lake Mirabad have a mean sample interval greater than 170 years and were dated with conventional radiocarbon rather than with AMS techniques, meaning they are likely to be less accurate.In the Early Bronze Age, the population peaked again despite a shift towards We do not have radiocarbon dates from the Arabian Peninsula for the earlier periods and, therefore, we cannot assess how the population responded to better climatic condition during the transition from the Late Pleistocene to the Early Holocene.We see that population grew substantially in the Neolithic, when the climate is relatively wet, and then declined in concomitance with a drying period from 6000 cal.yr.BP onwards (see Petraglia et al., 2020, Fig. 13).The population increased again in the Bronze Age and reached a peak in the Iron Age despite arid climate conditions.
The box-plots of all the Pearson correlation values show that population dynamics were positively correlated with climatic trends during the Late Pleistocene/Early Holocene (median ¼ 0.65) while they decouple during the Middle and Late Holocene (Fig. 4b).It is important to stress that the correlation values could be even more negative during the Late Holocene given that in some regions the SPD of radiocarbon dates tend to depict unrealistic low population levels in the later periods (from ~3500 cal.yr.BP) when climate trends are towards more arid conditions.In fact, the site-   A. Palmisano, D. Lawrence, M.W. de Gruchy et al. Quaternary Science Reviews 252 (2021) 106739 derived proxies from South-central Anatolia, Southern Levant and Upper Mesopotamia show higher population levels when compared to the SPDs of radiocarbon dates during the Late Holocene (Fig. 5).Thus, rather than a single step change between the Early Holocene and the Middle and Late Holocene, we would argue that from the second half of the Middle Holocene onwards we see an ongoing decrease in correlation between demographic and climate proxies.

Discussion
In this section, we illustrate more comprehensively where and when correlations between population dynamics and climatic trends occurred and describe how socio-ecological trajectories varied across the Near East from the Early to the Late Holocene.We are aware that the spatial and chronological scope of this work does not allow us to disentangle all the possible causal linkages between human-climate interactions, and caution is needed when constructing narratives of this kind.Our intention is to offer a first systematic assessment through the rigorous analysis of archaeological and climatic evidence.As stated above, our underlying assumption is that, all other factors remaining constant, increased aridity leads to population decline, and conversely that increased humidity supports population increase.Where these correlations are not present, or the magnitude of population change differs from that which we might expect given the magnitude of climate change, other factors may be involved.These include adaptive strategies such as social transformation and technological change, rapid declines in population and the complexity of social organisation, or migration.We should also remember that human groups are complex systems subject to endogenous change which may be unrelated to climate.We interpret our datasets within this framework.

Late Pleistocene and the Early Holocene (14,000e8326 cal. yr. BP)
The Late Pleistocene was characterised by the BøllingeAllerød interstadial, an abrupt warm and moist period (between 14,700 and 12,700) followed by a drying and cold climate during the Younger Dryas (ca.12,700e11,700).The climate instability of the Late Pleistocene ended with the onset of the Holocene (ca.11,700 cal.yr.BP), an overall thermally stable period characterised by higher average temperatures and wetter climatic conditions, which fostered radical environmental changes such as the growth of woodland vegetation, the retreat of permanent snowlines, and sea level rise (Lotter 2003;Roberts et al., 2008;Dean et al., 2015).Recent work has demonstrated that the domestication of cereal and legumes may have taken 2000e3000 years (Fuller et al., 2012;Asouti and Fuller 2012), and the climatic stability of the Holocene likely facilitated the establishment of farming.It is unlikely that such a long-term process could have occurred during the Late Pleistocene, a period characterised by pronounced climatic oscillations (Richerson et al., 2001).Put simply, the new multimillennial moist climatic regime of the Early Holocene could have provided favourable conditions for the full domestication of crops, increasing sedentism and population growth (Richerson et al., 2001;Weninger 2017;Shennan 2018,29e30, 37e38).
However, the Late Pleistocene-Early Holocene transition shows a more complex scenario with different regionalised climatic patterns.Several spatial climatic models in recent work by Roberts et al. (2018: Fig. 9) suggest that, despite a general drying phase across the Near East, favourable climatic conditions could have existed in the Fertile Crescent during the Younger Dryas (ca.12,700e11,700 cal.yr.BP).As a consequence, unlike other Near Eastern regions, the Fertile Crescent experienced a stable climatic regime and could have acted as a refugia for plant and animal resources and local human populations.This phenomenon may be visible in the SPDs of calibrated radiocarbon dates, where a rapid and continuous population growth occurred between 13,000 and 11,000 only in the Levant and Mesopotamia (Fig. 3: d-e; Figs.10e11 ).In the other regions, the population level was quite flat and in Anatolia lied below the confidence envelope of the panregional trend (Fig. 3: a).With the onset of the Holocene, ameliorated climatic conditions and the demic spread of farming favoured population growth across the whole Near East.In Anatolia, a first substantial growth occurred in correspondence with the earliest evidence of sedentism and farming in south-central Anatolia at around 10,300 cal.yr.BP, as shown by the earliest archaeological levels from Boncuklu (Baird et al., 2012) and As ¸ıklı H€ oyük (Stiner 2014).A more dramatic increase in the population occurred between 8500 and 8000 cal.yr.BP when the whole of Anatolia was fully neolithicised (Brami 2015, Fig. 7).Both the Levant and Mesopotamia seem to have a static or declining population immediately after 11,000 cal.yr.BP, with some regional differences.The Levant experienced further population growth between 10,500 and 9500, after which followed a gradual decline (Fig. 10).In Upper Mesopotamia, a drastic decline of the population between 10,200 and 9800 cal.yr.BP was likely due to the 10.2 cal.yr.BP event that caused rapid cooling and drying, negatively affecting plant and animal biomass and the local incipient agrarian communities (Borrell et al., 2015, Fig. 11).However, the population started increasing again after the cooling event and boomed during the Pottery Neolithic between 9300 and 8200 cal.yr.BP, when permanent villages started spreading into fertile plains and along major rivers (Bader 1993;Campbell 2012;42 1e422).This dramatic increase in population occurred simultaneously with the wettest climatic conditions across the whole Holocene (Fig. 11).A dramatic increase of population in Cyprus occurred with the onset of the Aceramic Neolithic at 11,000 cal.yr.BP and continued until 9500 cal.yr.BP (Fig. 9).This increase of population is perhaps related to several colonising events, where farmer-herder communities from the Levantine coast settled on the island (Peltenburg 2004;Simmons 2011).In Iran, a first increase in the population occurred during the Pre-Pottery Neolithic (PPN) with the earliest sedentary sites (e.g.Ganji Dareh, Asiab, Sarab) and in the first half of the ninth millennium BP when the communities on the Iranian plateau and highlands adopted a fully Neolithic lifestyle (Hole 1987;Zeder 2005;Helwing 2012, Fig. 12).
To summarise, after a climatic advantage to populations in regions such as the Levant and the Fertile Crescent during the Younger Dryas stadial, the onset of the Holocene brought an overall increase in temperature and rainfall across the whole Near East and facilitated the adoption of a sedentary lifestyle and demographic growth.The demic spread of the Neolithic package from the Levant at the beginning of the Holocene (~12,000 cal.yr.BP) to the South Caucasus by ~8000 cal.yr.BP (Fig. 3; cf.Bellwood 2013,129e135) seems to have prompted a chronologically variable first substantial increase in population across different regions.In particular, population started growing during the Aceramic Neolithic and more markedly during the Pottery Neolithic (Fig. 3).On a multi- millennial scale, the regional SPDs of calibrated radiocarbon dates show population growth in concomitance with increasing wet climatic conditions during the Early Holocene (see Figs. 7, 9e12).In addition, the regional demographic trends appear to be similar during the Late Pleistocene-Early Holocene (14,000e8326 cal.yr.BP; see Table 3 and Fig. 4a) and suggest a clear relationship between population and climate among early farming communities (see Fig. 4b).

Middle Holocene (8326e4200 cal. yr. BP)
The beginning of the Middle Holocene was marked by the socalled 8.2 k cal.yr.BP event, an abrupt cold and arid episode that is visible in climate records across the whole Near East between ~8250 and 8000 cal.yr.BP (see Table 12).However, the inferred demographic trends seem to indicate that this climatic shift did not severely impact Near Eastern communities, and there is no evidence of a widespread abandonment of settlements.The population levels were quite steady in all regions with a slight decrease in Mesopotamia and Levant (Fig. 3:d-e), while Iran experienced an increase in population (Fig. 3f).The decline in the Levant began at around 8600 cal.yr.BP onwards and is unlikely to be linked to the 8.2 cal.yr.BP event.Given the prevailing stable warm and wet climatic conditions, our data supports the hypothesis that the abandonment of the Late Pre-Pottery Neolithic B (PPNB) farming villages was endogenous and related to the overexploitation of local resources after the substantial population increase at the onset of the PPNB (Goring-Morris and Belfer-Cohen 2010; Finlayson 2013, 130).Alternatively, the pronounced sub-centennial rainfall fluctuations between moist and dry conditions occurring between 9500 and 8000 cal.yr.BP (visible in the higher resolution climate records from Jeita and Soreq caves; Fig. 10:7,10) could have affected the fragile socio-economic systems of the Levantine communities (Bar-Yosef 2002;Stein et al., 2010).However, the overall drop in rainfall between 8250 and 8000 cal.yr.BP was relatively low across the whole Near East, and it is possible that Neolithic societies could have overcome this change through adaptation strategies such as the diversification of subsistence practices and storage facilities (Flohr et al., 2016,35 e36).
After the 8.2 kya event, the Middle Holocene trend was for a gradual shift towards more arid climatic conditions, with a more marked drying after ~7000 cal.yr.BP (see Fig. 6).In this context, we  3), where we see a gradual demographic decline from 8000 cal.yr.BP onwards followed by a substantial demographic boom during the Late Chalcolithic and Early Bronze Age between ~5500 and 4200 cal.yr.BP.A decline in population in these regions seems to be synchronous with a multi-millennial drying phase that was characterised by severe and rapid aridification after 7000 cal.yr.BP (see Figs. 7 and 11 ).This population pattern, visible not only from the SPDs of radiocarbon dates but also from the time-series of archaeological survey data (as raw count, estimated size, and probabilistic weights; see Fig. 5aec; cf.Ur 2013; Alcock 2017; Lawrence et al., 2017), indicate that a substantial drying phase during the middle Chalcolithic could have triggered settlement abandonment and a relocation of agro-pastoral communities to areas with more stable and abundant resources, such as the Orontes river valley (Clarke 2016, 113).However, during the Late Chalcolithic and the Early Bronze Age (~5500 and 4300 cal.yr.BP.) the population grew dramatically in Anatolia and Mesopotamia despite a more arid climate (Clarke 2016, 114).In this period, the societies in Anatolia and Mesopotamia became more adaptive to climate change through technological advancements in metal manipulation, subsistence strategies and logistic infrastructure (Rosen 2007), as well as sophisticated social organisation.This period saw the emergence of large urban centres exceeding 100 ha and polities characterised by complex social hierarchies (cf.Allcock 2017; McMahon 2019).
Urban-based institutions may have facilitated resilience by integrating diverse economic spheres enabling risk pooling (Wilkinson et al., 2014).
Unlike Anatolia and Mesopotamia, in the Levant population started increasing from the mid-seventh millennium cal.yr.BP when cultural changes and successful adaptations culminated in more complex societies and villages exceeding 10 ha (see Figs. 5c and 10;Levy 1998;Rowan 2013).After a decline at 5600 cal.yr.BP, the population grew dramatically during the EBA I-II (ca.5300-4800 cal.yr.BP) despite drier climatic conditions cf. (Finkelstein 1994;Finkelstein and Gophna 1993;Palmisano et al., 2019).This may be linked to specialised agricultural production, especially olives, and more integrated regional economies (Badreshany et al., 2019).Overall population levels remained relatively high during the fifth-millennium cal.yr.BP despite a deterioration in climate, but fluctuations between 4700 and 4000 cal.yr.BP seem to be correlated strongly with the patterns drawn from the highresolution paleoclimate records from Jeita cave (Fig. 10).
In the South Caucasus, the population boomed with the onset of the Neolithic, when the first communities of farmers settled in the Kura-Araxes interfluve cultivating domesticated cereals and practising animal husbandry (Fig. 8;Berthon 2014;Sagona 2018, 124e125).Unfortunately, we do not have paleoclimate records from the South Caucasus but it seems that a decline in population between 7000 and 5500 cal.yr.BP occurred in concomitance of a drying phase as indicated by the hydro-climate records from Karaca cave and Lake Van (Fig. 8).However, a substantial demographic increase occured during the Bronze Age and peaked between 5000 and 4500 cal.yr.BP when the climate shifts abruptly towards more arid climatic conditions (Fig. 8).
We do not have palaeoclimate records from Cyprus and we rely on the closest available records from the Levant.The population level appears to be low at the beginning of the Middle Holocene with substantial growth in the late ceramic Neolithic between 6300 and 6000 cal.yr.BP when large villages with evidence of storage facilities and social stratification predominated (Peltenburg 1996, Fig. 9).However, a marked aridification occurred at around 6000 cal.yr.BP at the same time as a widespread abandonment of villages and a switch from agro-pastoralism to hunting, perhaps as a response to resource depletion and climatic stress (Knapp et al., 1994;Clarke 2016, 110).In the Late Chalcolithic and the EBA (5000e4000 cal.yr.BP), the population increased again as farming strategies became more intensive with greater evidence of increased livestock numbers, storage facilities and advancements in food-processing technology (Peltenburg 1998;Webb 2014).
In Iran, unlike the other regions, there was no gradual decline of the population with the shift towards more arid climatic conditions.It seems that during the Middle Holocene the population experienced booms and bust in coincidence with wet and dry climatic oscillations as shown by the palaeoclimatic records from the lakes Zeribar and Mirabad (see Fig. 12).As for the other regions, the more complex Bronze Age societies of Iran were more resilient to climatic variations and experienced a substantial demographic boom despite a rapid shift to a more arid drying phase between 5300 and 4500 cal.yr.BP (Fig. 12).
In the Arabian Peninsula, the "Holocene moist phase" between 8200 and 6200 cal.yr.BP could have favoured the rise of rich ecological niches that provided the basis for successful subsistence strategies and human settlement (Magee 2014,4 2 e45).In this context, an increase of population during the Neolithic could be related either to the migration of pastoralist groups looking for fertile grassland from the southern Levant to Arabia (cf.K€ ohler-Rollefson 1992,1 3 e14) or to the autochthonous development of the Neolithic package that allowed successful subsistence strategies triggering a substantial demographic growth (Cleuziou and Tosi 2007;Crassard 2009; see Fig. 13).After 6200 cal.yr.BP population dropped severely as a consequence of an abrupt drying phase that could have undermined the fragile eco-systems on which local Neolithic communities relied.Our data show an approximate correspondence with the so-called 'Dark Millennium', from 5900 to 5300 cal.yr.BP when arid conditions precluded mobile pastoral opportunities causing the abandonment of much of the peninsula (Petraglia et al., 2020).After a new demographic increase between 5600 and 5300 cal.yr.BP the population declined sharply during the Late Neolithic and the EBA at the same time as pronounced subcentennial oscillations between wet and dry episodes superimposed onto an overall more arid climate (Fig. 13).

Late Holocene (4200e2500 cal. yr. BP)
The Late Holocene was the most arid period across the chronological spectrum analysed in this study.The onset of this period was characterised by the so-called 4.2 k. cal.yr.BP event, an abrupt shift towards more arid climate conditions and dry and wet episodes between 4300 and 3800 cal.yr.BP.The magnitude, duration and impact of this event has been much debated, with significant variation in each visible across the Near East (Weiss et al., 1993;Riehl et al., 2014;Weiss 2016 and2017;Kaniewski et al., 2018;Cookson et al., 2019; see Fig. 6).The SPDs of calibrated radiocarbon date show a substantial drop in population during the 4.2 ka cal.yr.BP event in all the Near Eastern regions (Fig. 3) with the exception of Cyprus which experienced demographic growth (see Fig. 9).The patterns visible in the SPD are less evident in the archaeological survey data, although they are present in Anatolia and Mesopotamia (Fig. 5a and c).In the Levant, the settlement data do not show a sharp decline in population but rather continuity (Fig. 5b).It may be that the comparative precision of the SPD dataset allows us to identify short term events which are not resolvable in settlement pattern data based on ceramic typologies.For example, the chronological period known as the Late Third Millennium in Mesopotamia and EBA IV in the Levant spans 4500 and 4000 cal.yr.BP (cf.Regev et al., 2012;H€ oflmayer et al., 2014;Ur 2010), meaning settlements occupied before, during and after an event at 4.2 k. cal.yr.BP could be conflated.The widespread decline in population should also be considered in relation to the end of the Middle Holocene, when, despite a more arid climate, Bronze Age Near Eastern communities experienced a demographic boom.An explanation could be that the rapid expansion of new large sites established in climatically marginal zones and sustained by a population drawn from outlying communities could have exerted growing pressure on local resources (cf.Wilkinson et al., 2014;Lawrence and Wilkinson 2015).As a consequence, an abrupt deterioration of the climatic conditions of even a small magnitude could have had a great impact on the stability of communities and large urban centres reliant on intensive agricultural production to feed large populations (Weiss et al., 1993;Wilkinson 1997;Wilkinson et al., 2007;Ur 2010;Massa and S ¸aho glu 2015;Schwartz 2017;Altaweel and Palmisano 2019;Cookson et al., 2019).A decline in precipitation at the end of the fifth-millennium cal.yr.BP resulted in drought stress to crops and reduced yields that may have caused a shortage of food and then triggered a domino effect leading to famine, migration and societal conflict (Ur 2010: Riehl 2012;Riehl et al., 2014).However, mitigation and adaptative strategies such as the switch to more drought-tolerant crop species and from agricultural to pastoral resources avoided the total collapse of urban centres that, nevertheless, shrank in size (eg. Tell Brak, Tell Mozan;Riehl 2008;Krakauker 2010;Riehl et al., 2012;Wilkinson et al., 2014).It is worth noting that the population after the collapse was still well within the expected values for an exponential growth model, while the final part of the Middle Holocene represents a substantial positive deviation.We might then consider the boom period of urbanisation in the Early Bronze Age to be the outlier, with 'collapse' representing a return to more sustainable population levels.
After the decline in population visible across the Near East between ~4300 and 3800 cal.yr.BP (Fig. 2:c-d), the SPDs of calibrated radiocarbon dates indicate strikingly different regional patterns (Table 5; Fig. 4a).It seems that different geo-cultural areas experienced a variety of socio-ecological trajectories which are reflected in their demographic signal.However, this pattern should be considered cautiously given that in Anatolia and northern Mesopotamia we have detected no good agreement between SPDs and settlements derived proxies after 4000 cal.yr.BP (see Fig. 5:a and  c).The following period saw a slight increase in precipitation between ~3800 and 3200 cal.yr.BP within an overall drying phase (Grant et al., 2012;Burstyn et al., 2019).In this context, Near Eastern communities seem to flourish again, as witnessed by an increase of the population in several regions (Fig. 3).From 4000 cal.yr.BP onwards, the SPDs tend to underestimate real population levels as archaeologists increasingly rely on ceramic dating methods.The archaeological survey data for Anatolia indicate an increase in population during the Middle Bronze Age and a peak during the Late Bronze Age with the establishment of the Hittite Empire (cf.Allcock 2017;Woodbridge et al., 2019, Fig. 5a).The Levant and Mesopotamia were characterised by patterns of demographic boom in the MBA (~4000-3600 cal.yr.BP) and Iron Age (3200e2600 cal.yr.BP) punctuated by a decline in population in the LBA between ~3500 and 3200 cal.yr.BP (see Finkelstein 1996;Ur 2013;Lawrence et al., 2017;Palmisano et al., 2019, Fig. 5b and c and 10-11).The socalled 3.2 k cal.yr.BP event seems not to have impacted so severely on the demographic trends in the Near East between 3200 and 2900 cal.yr.BP.However, there were different regional responses.Anatolia was strongly impacted by the 3.2 k cal.yr.BP, which may have amplified the internal conflict and weakening of the Hittite Empire at the end of the LBA (Yakar 1993;Allcock 2017, 77;Kaniewski et al., 2015 and2019;Woodbridge et al., 2019, Figs. 4a and 6).In contrast, the Levant was characterised by increases in population during the IA I-II (~3100e2700 cal.yr.BP; Figs.4b and  9).This period was characterised by the end of the Egyptian and Hittite domination in the southern and northern Levant respectively, enabling the establishment of medium-sized regional kingdoms such as Israel, Judah, Moab, Edom, Ammon and Phoenician city-states in Lebanon (Palmisano et al., 2019).In Mesopotamia, the SPD curve likely overstates a real decline in the Late Bronze Age but fails to capture a demographic peak during the expansionist phase of the Assyrian Empire facilitated by the rapid growth of imperial capitals and a transformation of the rural landscape (Wilkinson et al., 2005;Lawrence et al., 2016;Düring 2018, Figs. 5c and 11).
Elsewhere, highly divergent trajectories also suggest political and technological changes seem to have had more of an impact than climate.In Iran, an interplay of different factors, including the end of the Shutrukid Dynasty due to the military defeat against the Babylonian king Nebuchadnezzer I and drier climatic conditions could have caused a demographic crisis and the subsequent abandonment of urban centres, political fragmentation and increasing pastoralism (Potts 1999,3 7 0 e305; Alvarez-Mon 2012, 754e755; Fig. 11).By contrast in the South Caucasus, the period between LBA and IA I (3500e2800 cal.yr.BP) was characterized by increased social complexity and the intensification of agriculture through the use of irrigation.This may be seen as a successful adaptative strategy to cope with the multimillennial drying phase of the Late Holocene (Sagona 2018,3 78e379;Fig. 8) and helped to produce a demographic boom.Late Bronze Age Cyprus was characterised by significant social changes culminating in the increase of new settlements and urban centres and an apparent societal upheaval resulting in the abandonment of several Middle Cypriot settlements (Knapp 1997;Steel 2013, 865e870).Although the 3.2 k cal.yr.BP event may be implicated here (Kaniewski et al., 2019), Cyprus does not seem to have been affected by an island-wide demographic crisis since some sites such as Enkomi and Kition continued, and Aegean migrants settled along the coastal regions where polities survived and new trading harbours such as Amathus and Kourion were established (Iacovou 2008, 640;Fig. 9).In the Arabian Peninsula, we can see a negative correlation between climate and population, with the latter increasing substantially despite a drying trend (Fig. 13).In fact, the deteriorating climatic conditions during the Late Holocene could have acted as a stimulus to innovation for the Bronze Age and Iron Age communities.Major shifts in settlement and social organisation were brought by the domestication of the dromedary camel, an excellent resource of meat and dairy products, and later the adoption of the qanat or falaj, a system of subterranean irrigation channels (Boucharlat and Lombard 1985;Magee 2014,197e222).

The decoupling of climatic and demographic trends and the impact of RCCs
The dramatic increase of population with the onset of the Holocene in the Early Neolithic is correlated positively with a trend towards wetter climatic conditions that could have favoured the diachronous demic diffusion of farming from the Levant (~11,700 cal.yr.BP) to the South Caucasus by 8000 cal.yr.BP.Unlike the Early Holocene, characterised by a growing population, in the Middle Holocene we see cycles of demographic booms and busts, some of which are synchronous with short periods of amelioration or deterioration of climatic conditions (Clarke et al., 2016).During the second half of the Middle Holocene (~5500 cal.yr.BP), and into the Later Holocene, we can identify a decoupling of climate and human population.The climatic and demographic trends, despite regional variations, are either negatively correlated or not correlated at all (Fig. 4b).We interpret this as due to technological advancement and the building up of infrastructure of more complex societies less vulnerable to a more arid climate (Lawrence et al., 2016;Roberts et al., 2019).
During the Early and Late Holocene RCC events seems to have exerted a minor impact on Near Eastern Communities.While the 10.2 k cal.yr.BP event seems to have determined a cultural break in the local communities in Upper Mesopotamia and northern Levant (Borrell et al., 2015), the 9.2 k and 8.2 k cal.yr.BP events do not appear to be linked to a decline population or a broad abandonment of settlements (Flohr et al., 2016).On the other hand, despite the decoupling between population and climate during the second half of the Middle Holocene, the 4.2 k cal.yr.BP event seems to have exerted a dramatic impact on Near Eastern communities and caused a widespread demographic crisis.In this instance the abandonment of settlements or their reduction in size is perhaps the result of the interplay between unfavourable climatic conditions and a system of large urban centres that grew rapidly during the Early Bronze Age and were reliant on fragile systems of food provision (Wilkinson et al., 2014;Lawrence and Wilkinson 2015).The collapse in population visible during this RCC event should therefore be viewed in the context of the preceding demographic boom.The impact of the 3.2 ka cal.yr.BP event was more regionalised and did not affect population patterns across the whole Near East in the same way as the 4.2 cal.yr.BP event.The Late Bronze Age collapse may have impacted specific polities, such as the Hittite and Egyptian empires, but social and technological capacity was retained, and new political entities, such as the emergent kingdoms of the Levant, were able to sustain high populations (Rosen 2007, 101e102).We might speculate that as societies become increasingly affluent and complex, straightforward demographic proxies are less useful in demarcating responses to RCC events.Such societies may be capable of maintaining food production at a level which prevents large scale site abandonment or mortality, and therefore visibility in the sorts of proxies we have used in this paper, while still experiencing dramatic shifts in material wealth and social and political organisation.

Conclusions
This paper has highlighted, for the first time, the correlations between demographic and climatic trends from the Late Pleistocene to the Late Holocene (14,000e2500 cal.yr.BP) across the Near East.To do so, we used most of the known and freely accessible palaeoclimatic records available for the area and three types of archaeo-demographic proxies (radiocarbon dates, settlement estimated size, raw site counts).We argue that population trends are best estimated using a multi-proxy approach in order to compare independent archaeological indices and statistically assess to what extent they correlate with one another.This approach allows us to explore the strengths and weaknesses of each line of evidence and to build more robust narratives, both in the very long run and in specific well-defined periods.Unfortunately, high quality integrated datasets of archaeological settlement data are not (yet) available across the whole Near East.However, the overall agreement between the SPDs of calibrated radiocarbon dates and the settlement-based proxies for three sub-regions (South-central Anatolia, Southern Levant, and Upper Mesopotamia) suggests SPDs can be a good indicator for reconstructing past population dynamics during particular periods.SPDs can be used up to ~2800 cal.yr.BP as a good indicator, while archaeological survey data can extend until more recent periods (e.g.Medieval).In some regions, such as Anatolia, Mesopotamia and Iran, SPDs tend to underestimate the population from around ~4000-3500 cal.yr.BP when compared with the settlement derived proxies.However, archaeological settlement derived proxies have the opposite problem of "dwarfing" population levels in the earlier periods, especially before the widespread use of pottery.Where they can be confidently interpreted, the chronological precision available through SPDs of radiocarbon dates mean they are a useful addition to settlement data for comparison with palaeoclimatic records, especially in relation to very short term 'events'.Given the patchiness of the available data, we are aware of the caveats in inferring population trends by using only SPDs of calibrated radiocarbon dates and our results should be considered as a preliminary assessment.
The long-term demographic trends show similar trajectories between regions in the Late Pleistocene-Early Holocene (14,000e8326 cal.yr.BP), increasing regionalisation of demographic patterns in the Middle Holocene (8326e4200 cal.yr.BP), and marked inter-regional contrasts during the Late Holocene (4200e2500 cal.yr.BP).Comparisons between palaeoclimatic and demographic proxies have revealed either no statistical correlation or an overall negative correlation between population fluctuations and hydro-climate patterns when taking into account the whole chronological scope between 14,000 and 2500 cal.yr.BP.In other words, human populations grew and declined independently from long-term climate trajectories, and possible correlations are only visible for shorter periods.Unfortunately, most of the available hydro-climate patterns available for the Near East have a mean sampling interval greater than 150 years and, therefore, only a few of them make it possible to assess the correlations between climate and population for short periods.In addition, the patchiness of the available hydro-climate records and their dispersed spatial distribution means coherent spatial and chronological comparisons with the archaeo-demographic proxies are not always possible.
Although we have highlighted general trends suggesting a good correlation between climate and population during the Early Holocene and a decoupling from the second half of the Middle Holocene, the Near East shows a wide spectrum of regional socioecological trajectories.For instance, the transition from the Late Pleistocene to the early Holocene was characterised by high regional variation, with the Levant and Mesopotamia potentially acting as refugia during the Younger Dryas stadial (~12,700e11,700 cal.yr.BP; Roberts et al., 2018), attracting population while other areas declined.This regionality is also visible in the RCCs occurring in the Early and Middle Holocene (10.2, 9.2 and 8.2 k cal.yr.BP events), all of which had a limited and differential impact across the Near East.Even the more widespread 4.2 k and 3.2 k cal.yr.BP events prompted different local responses.The 4.2 k cal.event appears to have had the greatest impact, but our results show that this must be taken in the context of a booming pre-event population which likely put pressure on resources and constrained the ability of communities to adapt.The 3.2 k. cal.event had perhaps the most regionalised impact of all.This might reflect the generally increasing robustness of food production systems in the face of climate fluctuations on the one hand, and the growing importance of social and especially political organisation as factors in resilience levels on the other.
Future research efforts will focus on the integration of the SPDs of calibrated radiocarbon dates with multiple archaeodemographic proxies (raw site count, aggregated estimated settlement size, number of burials) covering the whole Near East in order to deliver more robust results and reconstruct population dynamics over a chronological spectrum extending up to Medieval and post-Medieval times.Additional archaeological datasets would also enable a more detailed investigation of the individual regional trends identified here.
It is clear from the present work that while a wealth of archaeological data exist across the Near East, a higher number of well dated palaeoclimatic archives are needed to provide a more even spatial and chronological coverage, and to produce more accurate interpretations of past human-climate interactions.The subdivision of our seven geo-cultural regions is justified on the basis of traditional research trajectories, real differences in physical geography and by the need to guarantee sufficient data to produce meaningful results.However, an increase in the spatial resolution of available palaeoclimatic data would allow us to subdivide further, and potentially to assess relationships across a wider range of environments.This would allow to assess how social behaviours changed in different ecological niches.Furthermore, the integration of vegetation pollen-based reconstruction would allow us to assess not only correlations between climate and population but also the impact of climate and population change on vegetation and the wider landscape.

Fig. 1 .
Fig. 1.Study area and spatial distribution of radiocarbon dates.Settlements data from a) South-central Anatolia, b) southern Levant, and c) Upper Mesopotamia.

Fig. 2 .
Fig. 2. a) Summed Probability Distribution (SPD) of normalised (black line) and unnormalised (in grey) calibrated radiocarbon dates; the bar-code like strip represents the median values of multiple calibrated radiocarbon bins; b) unnormalised SPD of short-lived radiocarbon dates; c) unnormalised SPD (solid line) vs. a fitted logistic and d) exponential null model (95% confidence grey envelope).Blue and red vertical bands indicate respectively chronological ranges within the observed SPD deviates negatively and positively from the null model.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 3 .
Fig. 3. Regional summed probability distributions (SPDs) of calibrated radiocarbon dates for a) Anatolia, b) South Caucasus, c) Cyprus, d) Levant, e) Mesopotamia, f) Iran, and (g)Arabia compared with a 95% Monte Carlo envelope of the pan-regional model produced via permutation of sub-regional dates.
The high-resolution climate records such as Sofular Cave in Turkey, Jeita Cave in Lebanon, Kuna Ba cave in northern Iraq, and the Qunf cave in Oman indicate that this climatic event was characterised by pronounced multi-decadal fluctuations of wet and dry episodes(Fig.6: 2, 7, 14, and 16;Kuzucuo glu, 2009;Shah et al., 2013;Cheng et al., 2015.)

Fig. 4 .
Fig. 4. Box-plots showing the pairwise Pearson correlation values between all regional SPDs (a) and between population and climatic trends (b); n ¼ number of correlations performed.

Fig. 7 .
Fig. 7. SPD of unnormalised calibrated radiocarbon dates for Anatolia vs. a logistic null model (95% confidence grey envelope) compared with palaeoclimate records.Blue and red vertical bands indicate respectively chronological ranges within the observed SPD deviates negatively and positively from the null model.The green dots represent significant (pvalue < 0.05) positive or negative Pearson's correlations (r) values ranging from þ1toÀ1 by using a 500-year-time moving window.The orange rectangle represents two examples of 500-year-time moving window (between 9600 and 9100 and 4800-4300 cal.yr.BP) in which the SPD of calibrated radiocarbon dates is significantly positively correlated with the climatic trends from Lake Sofular.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 8 .
Fig. 8. SPD of unnormalised calibrated radiocarbon dates for South Caucasus vs. a logistic null model (95% confidence grey envelope) compared with palaeoclimate records Karaca Cave and Lake Van.Blue and red vertical bands indicate respectively chronological ranges within the observed SPD deviates negatively and positively from the null model.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 9 .
Fig. 9. SPD of unnormalised calibrated radiocarbon dates for Cyprus vs. a logistic null model (95% confidence grey envelope) compared with palaeoclimate records.Blue and red vertical bands indicate respectively chronological ranges within the observed SPD deviates negatively and positively from the null model.The red dots represent significant (pvalue < 0.05) positive or negative Pearson's correlations (r) values ranging from þ1toÀ1 by using a 500-year-time moving window.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 10 .
Fig. 10.SPD of unnormalised calibrated radiocarbon dates for the Levant vs. a logistic null model (95% confidence grey envelope) compared with palaeoclimate records.Blue and red vertical bands indicate respectively chronological ranges within the observed SPD deviates negatively and positively from the null model.The red dots represent significant (pvalue < 0.05) positive or negative Pearson's correlations (r) values ranging from þ1toÀ1 by using a 500-year-time moving window.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 11 .
Fig. 11.SPD of unnormalised calibrated radiocarbon dates for Mesopotamia vs. a logistic null model (95% confidence grey envelope) compared with palaeoclimate records.Blue and red vertical bands indicate respectively chronological ranges within the observed SPD deviates negatively and positively from the null model.The red dots represent significant (pvalue < 0.05) positive or negative Pearson's correlations (r) values ranging from þ1toÀ1 by using a 500-year-time moving window.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Table 2
List of palaoeclimate records.

Table 3
Pearson Correlation Coefficient r-value matrix in the Late Pleistocene and Early Holocene between the regional SPDs (14,000e8326 cal.yr.BP).Strong correlations are indicated by bold numbers.*p-value < 0.05 Pearson Correlation Coefficient r-value matrix in the Middle Holocene between the regional SPDs (8326-4200 cal.yr.BP).Strong correlations are indicated by bold numbers.

Table 5
Pearson Correlation Coefficient r-value matrix in the Late Holocene between the regional SPDs (4200-2500 cal.yr.BP).Strong correlations are indicated by bold numbers.

Table 6
Pearson Correlation Coefficient r-value matrix between the demographic proxies of Anatolia between 10,000 and 2600 cal.yr.BP).Strong correlations are indicated by bold numbers.*p-value < 0.05

Table 8
Pearson Correlation Coefficient r-value matrix between the demographic proxies of southern Levant between 10,000 and 2600 cal.yr.BP.Strong correlations are indicated by bold numbers.*p-value < 0.05 arid conditions and then declined around the time of the 4.2 k cal.yr.BP event.Again, just as in Mesopotamia, the lower level of SPD curve during the Late Bronze and Iron Ages are likely due to research biases and do not represent past realities.

Table 12
Table showing the occurrence of the RCC in the paleoclimate records used in the present study.