Hunted gazelles evidence cooling, but not drying, during the Younger Dryas in the southern Levant

Significance

The Terminal Pleistocene Younger Dryas (YD) event is frequently described as a return to glacial conditions. In the southern Levant it has featured prominently in explanations for the transition to agriculture—one of the most significant transformations in human history. This study provides rare local measures of the YD by deriving gazelle isotopic values from archaeological deposits formed by Natufian hunters just prior to and during the YD. The results provide evidence for cooling, but not drying during the YD and help reconcile contradicting climatic reconstructions in the southern Levant. We suggest that cooler conditions likely instigated the establishment of settlements in the Jordan Valley where warmer, more stable conditions enabled higher cereal biomass productivity and ultimately, the transition to agriculture.

Abstract

The climatic downturn known globally as the Younger Dryas (YD; ∼12,900–11,500 BP) has frequently been cited as a prime mover of agricultural origins and has thus inspired enthusiastic debate over its local impact. This study presents seasonal climatic data from the southern Levant obtained from the sequential sampling of gazelle tooth carbonates from the Early and Late Natufian archaeological sites of Hayonim and Hilazon Tachtit Caves (western Galilee, Israel). Our results challenge the entrenched model that assumes that warm temperatures and high precipitation are synonymous with climatic amelioration and cold and wet conditions are combined in climatic downturns. Enamel carbon isotope values from teeth of human-hunted gazelle dating before and during the YD provide a proxy measure for water availability during plant growth. They reveal that although the YD was cooler, it was not drier than the preceding Bølling–Allerød. In addition, the magnitude of the seasonal curve constructed from oxygen isotopes is significantly dampened during the YD, indicating that cooling was most pronounced in the growing season. Cool temperatures likely affected the productivity of staple wild cereal resources. We hypothesize that human groups responded by shifting settlement strategies—increasing population mobility and perhaps moving to the warmer Jordan Valley where wild cereals were more productive and stable.

This study measures the δ¹⁸O and δ¹³C isotopic values of gazelle tooth carbonates to address the controversy regarding water availability during the Younger Dryas (YD) climatic event in the eastern Mediterranean region. The YD was an abrupt, millennial-scale cooling event centered on the Northern Hemisphere between ∼12,900 and 11,500 BP (1). The weakening or possibly even the complete shutdown of the North Atlantic meridional overturning circulation caused dramatic cooling that reversed the post Last Glacial Maximum warming trend (2). Although the cooling phenomenon associated with the YD is clearly manifested in northern latitudes, contradictory data regarding the impact of this event in the eastern Mediterranean region is far from resolved (Fig. 1). Vastly disparate interpretations of the environmental and climatic impact of the YD on the southern Levant have been proposed with some arguing for xeric and others for opposing mesic conditions (3–10).

Fig. 1.

Global and regional paleoclimate records for the past 20,000 y. (A) The chronology of EN, LN, and Prepottery Neolithic A (PPNA) cultures in the southern Levant based on calibrated ¹⁴C dates (14); individual calibrated (OxCal13) radiocarbon dates of HLT (red symbols ± 1σ) and HC EN (blue symbols ± 1σ). (B) Calculated Northern Hemisphere insulation at 60°N (22). (C) Composite Greenland ice core oxygen isotope (δ¹⁸O) values provide a proxy for paleotemperature (21). (D) DSB lake level reconstruction (24). (E) Eastern Mediterranean reconstruction of SST using alkenone unsaturation ratio ($U_{\mathit{37}}^{k\mathrm{’}}$) and Globigerinoides ruber δ¹⁸O values (23). (F) Soreq Cave speleothem record (3). S2 is a sapropel formation event in the Eastern Mediterranean Sea; YD is the Younger Dryas; and H1 is a Heinrich Event.

In the Mediterranean region of the Levant, the YD has been cited as a primary trigger for the transition to agriculture, a fundamental shift in human economic and social strategies that completely reconfigured human societies at the beginning of the Holocene (11–13). Clarifying regional impacts of the YD are essential for understanding the conditions that preceded this fundamental cultural change. In the Mediterranean Levant, the YD coincides with the Late Natufian (LN) cultural phase when the last hunter-gatherers occupied the region (11, 14).

In contrast to the Bølling–Allerød that was associated with unprecedented human sedentism in comparison with earlier periods, thinner site deposits and less investment in permanent architectural features reflect a reduction in site occupation intensity and an increase in population mobility coincident with the YD (12, 13, 15). The leading argument for the abandonment of a more or less sedentary lifestyle is an imbalance between human populations and their food resources related to a reduction in biomass production brought on by cool and dry conditions attributed to the YD (16, 17). Bar-Yosef and Belfer-Cohen (13) argue that experimentation with plant cultivation, triggered by the need to increase food production in response to declining resource abundance, ultimately lead to the beginning of agriculture at the onset of the Holocene. Other researchers recognize the importance of climatic change (18–20).

Multiple proxies have been used to reconstruct the climatic conditions of the YD. Oxygen isotope data from the Greenland ice core indicate clear evidence for a dramatic drop in temperature across the Northern Hemisphere (21) (Fig. 1C) when solar radiation approached its peak (22) (Fig. 1B). A local drop in sea surface temperature (SST) is also evident in the eastern Mediterranean Sea based on alkenone unsaturation ratios ($U_{\mathit{37}}^{k\mathit{’}}$) and positive shifts in the δ¹⁸O values of foraminifera (23) (Fig. 1E) in deep sea cores. Despite general agreement on regional cooling during the YD, multiple proxies from the eastern Mediterranean conflict over the degree of aridity. In the Dead Sea Basin (DSB) for example, the shorelines of paleolake Lisan increase in elevation (Fig. 1D), suggesting higher water availability in the Mediterranean catchment area (9, 24). Based on a correlation between modern mean annual precipitation and the δ¹⁸O values of meteoric water, Bar-Matthews et al. (3) interpreted a positive shift in cave speleothem δ¹⁸O values as evidence for dry conditions during the YD [see also Orland et al. (25) for a modern cave carbonate analog]. However, following comparisons of the δ¹⁸O data from deep sea cores and Lake Lisan carbonates, others interpret the speleothem δ¹⁸O data as the result of a source effect, rather than a drop in precipitation (26, 27). Additional controversy revolves around the relative availability of arid-adapted plant families during the YD, with Rossignol-Strick (8) arguing for an increase in the arid-adapted plants recorded in deep sea cores, a shift that was not observed in lacustrine pollen diagrams (4). Gvirtzman and Wieder (5) argue that the YD was more arid based on their evaluation of coastal aeolian loess deposits. In addition, a study of Negev sand grains lead Roskin et al. (10) to argue that exceptionally windy rather than arid conditions are responsible for the formation of aeolian sand dunes.

The current study contributes to this debate by introducing δ¹⁸O and δ¹³C values of gazelle tooth carbonates to reconstruct climate conditions in the southern Levant during the Natufian cultural period, just before and during the YD. Seasonal carbon and oxygen isotope data recovered from gazelle teeth directly deposited by humans at archaeological sites are used to investigate the impact of climatic change at both local and regional scales. Reconstructing climate directly from archaeological material circumvents problems that arise when regional climatic datasets are applied to local archaeological sites dated using different techniques and/or situated in different environmental contexts. In this study, climatic conditions surrounding sampled archaeological sites are reconstructed based on δ¹³C and δ¹⁸O values from gazelle carbonates. We also aim to resolve sharply contrasting interpretations regarding regional patterns of aridity for this period. Because the primary controversy concerns water availability, which is also the major limitation on modern day biomass productivity in the eastern Mediterranean (28), we focus especially on this marker. Finally we discuss potential implications of the results on Natufian settlement and subsistence on the threshold of the transition to agriculture.

This study compares seasonal climatic conditions during the Early and Late Natufian (EN and LN, respectively) at the sites of Hayonim Cave (EN and LN) and Hilazon Tachtit (LN), Galilee, Israel (29–31). The δ¹⁸O values are influenced by a number of climatic factors, and δ¹³C provides an independent proxy for water availability (32–34), and is measured here to constrain the interpretation of δ¹⁸O values. Carbon isotope values measured in gazelle teeth reflect water availability during the growth of C₃ vegetation consumed by gazelle and thus provide an indirect measure of how wet local conditions were when the C₃ plants were consumed (34). The site-based δ¹⁸O and δ¹³C values provided by gazelle tooth enamel carbonate are then directly compared with regional paleoclimate proxies to evaluate contrasting interpretations of climatic data.

Mountain gazelles (Gazella gazella) are the most ubiquitous hunted species in Natufian sites (15). Gazelle are ideal for paleoenvironmental and paleoclimatic reconstructions because the carbon isotope values (δ¹³C) of their body tissues reflect the carbon values of their plant diet. The carbon values mirror the amount of water available during the growth of C₃ plants in the diet and the proportion of C₄ vegetation consumed in steppic environments (34). Because tooth enamel forms incrementally, δ¹³C values measured sequentially in tooth enamel carbonate provide valuable seasonal climatic and environmental information. Oxygen isotope values (δ¹⁸O) measured in gazelle teeth provide complementary information about the isotopic composition of the consumed meteoric water, and the animal’s physiological condition, which changes under water stress (35, 36).

Hayonim Cave [HC; WGS84 32898277N 352689671W, 176 m above modern sea level (amsl)] and Hilazon Tachtit Cave (HTC; WGS84 32923617N 35216871W, 216 m amsl) are located about 10 km apart on the western slopes of the Mediterranean Hills, at the boundary between the Lower and Upper Galilee. They are situated at similar elevations, enjoy equal mean annual rainfall (∼700 mm/y) and are surrounded by a mix of Mediterranean maquis and batha vegetation communities (Supporting Information; Fig. S1). Although the Natufian deposits at HC cover both the Early (15,000–13,600 Cal BP) and Late phases (13,640–11,540 Cal BP; ref. 30), the radiocarbon dates and cultural remains from HTC indicate a more limited occupation in the LN phase (12,744–12,674 OxCal13 BP) (14). Because the sites are environmentally similar, the isotopic samples can be combined to compare climatic and environmental conditions across the EN and LN boundary.

Fig. S1.

Regional and local maps of Hayonim and Hilazon Tachtit Caves. White contour lines and black numerals represent modern isohyets (1961–1990).

Results

General Trends between Early and Late Natufian Phases.

The complete dataset of tooth enamel carbonate including unidentified gazelle molar fragments and lower third molars (M₃s) is detailed in Table S1. Before specific patterns are identified, it is important to note that mean δ¹⁸O and δ¹³C values from LN HC and LN HTC are statistically identical (two-tailed t test assuming unequal variance, δ¹⁸O P = 0.771; δ¹³C P = 0.987; Fig. 2). This result affirms that differences between the EN layer at HC and the LN layers at HC and HTC are the product of temporal changes in regional climate rather than site-specific microhabitat conditions or a change in gazelle hunting territories.

Table S1.

Gazelle tooth enamel carbonate isotope data

Sample	Phase	n	δ13C (‰)				δ18O (‰)
			$\overline{x}$	${\mathrm{\delta }}^{\mathrm{13}}{\text{C}}_{\text{min}}$	${\mathrm{\delta }}^{\mathrm{13}}{\text{C}}_{\text{max}}$	Δ13C	$\overline{x}$	${\mathrm{\delta }}^{\mathrm{18}}{\text{O}}_{\text{min}}$	${\mathrm{\delta }}^{\mathrm{18}}{\text{O}}_{\text{max}}$	Δ18O
1	EN HC	12	−12.46	−12.94	−12.05	0.90	−5.38	−6.47	−3.88	2.59
2	EN HC	14	−12.19	−12.64	−11.67	0.97	−5.36	−5.74	−4.81	0.94
3	EN HC	12	−10.75	−11.36	−10.30	1.07	−3.09	−4.05	−2.35	1.69
4	EN HC	8	−11.13	−11.72	−10.93	0.79	−3.91	−5.11	−1.36	3.74
6	EN HC	8	−11.15	−11.33	−10.84	0.50	−0.52	−0.86	0.07	0.94
8	EN HC	8	−12.41	−12.45	−12.29	0.16	−2.34	−2.92	−1.23	1.69
9	EN HC	4	−10.89	−11.29	−10.32	0.97	−2.21	−3.42	−0.97	2.45
11	EN HC	5	−12.00	−12.41	−11.60	0.81	−2.32	−2.98	−1.63	1.35
13	EN HC	7	−11.80	−12.65	−10.94	1.72	−3.16	−3.95	−2.25	1.71
33	EN HC	10	−11.90	−12.72	−10.95	1.77	−0.93	−2.26	−0.07	2.19
34	EN HC	12	−11.81	−12.34	−11.40	0.94	−1.78	−2.80	−0.64	2.16
35	EN HC	12	−11.37	−11.79	−10.99	0.80	−1.92	−2.76	−1.34	1.42
36	EN HC	3	−13.28	−13.51	−13.15	0.35	−1.50	−1.53	−1.46	0.07
37	EN HC	10	−12.02	−13.41	−10.88	2.52	−0.64	−2.20	0.44	2.64
38	EN HC	10	−12.20	−12.76	−11.44	1.32	−1.20	−1.99	−0.65	1.34
5	LN HC	6	−11.16	−11.60	−10.86	0.73	−2.82	−3.22	−2.32	0.90
7	LN HC	4	−11.44	−11.64	−11.16	0.48	−0.90	−1.88	−0.35	1.54
10	LN HC	8	−12.71	−13.50	−11.62	1.88	−2.90	−4.39	−1.71	2.68
12	LN HC	8	−12.06	−12.82	−11.71	1.11	−0.79	−2.27	0.70	2.97
14	LN HC	9	−13.08	−13.46	−12.56	0.89	−2.79	−3.77	−2.08	1.69
15	LN HC	8	−12.46	−13.45	−11.76	1.69	−1.72	−3.03	−0.60	2.43
17	LN HC	4	−14.25	−14.84	−13.59	1.25	−1.45	−2.34	−0.62	1.72
19	LN HC	2	−11.63				0.73
20	LN HC	1	−11.32				1.12
22	LN HTC	16	−12.38	−13.25	−11.85	1.39	−3.10	−4.58	−0.27	4.31
24	LN HTC	9	−12.12	−12.83	−11.40	1.43	−0.95	−2.22	0.54	2.91
25	LN HTC	9	−11.57	−12.03	−11.03	0.99	−1.87	−2.39	−1.51	0.88
26	LN HTC	11	−13.53	−14.27	−12.78	1.49	−4.64	−5.61	−3.55	2.06
27	LN HTC	3	−13.12	−13.47	−12.94	0.53	−3.49	−4.18	−2.62	1.56
28	LN HTC	6	−11.96	−12.74	−11.59	1.15	−0.20	−1.24	0.60	1.84
29	LN HTC	1	−12.47				0.24
30	LN HTC	5	−13.02	−13.59	−12.23	1.36	−1.73	−3.10	−0.55	2.56
31	LN HTC	3	−12.35	−12.85	−12.07	0.78	−1.72	−2.38	−1.28	1.10
32	LN HTC	5	−12.35	−12.75	−11.82	0.93	0.02	−0.60	0.82	1.42
39	LN HTC	15	−11.50	−12.37	−10.33	2.04	−0.78	−1.85	0.11	1.95
40	LN HTC	5	−11.73	−12.19	−11.43	0.76	−0.41	−0.57	−0.24	0.33
41	LN HTC	15	−12.02	−12.88	−11.15	1.72	1.67	0.51	2.91	2.40
42	LN HTC	12	−12.05	−12.61	−11.52	1.09	−0.30	−2.61	1.01	3.62
43	LN HTC	10	−11.67	−12.14	−11.25	0.89	−0.67	−1.26	−0.20	1.06
44	LN HTC	12	−12.08	−12.51	−11.57	0.94	−0.75	−1.63	0.31	1.93
45	LN HTC	15	−11.96	−12.60	−11.20	1.40	−0.57	−1.16	−0.16	1.00
18	HC	6	−12.13	−12.60	−11.50	1.10	−2.78	−3.32	−2.25	1.07
21	HC	2	−12.13				−3.25

Fig. 2.

Bivariate plot of all gazelle carbon and oxygen isotope enamel molar data (Table 1). Blue symbols and shaded spheres (±1σ): EN HC; green symbols: LN HC; red symbols and shaded spheres (±1σ) LN HTC; squares represent mean values; triangles represent the average minimal values measured in each tooth; diamonds represent the average maximal value.

The mean δ¹⁸O values measured in gazelle teeth from EN HC are significantly lower (negative) than those from LN HC and HTC (Table 1; Fig. 2). A similar relationship exists between the average most negative (minimum) and positive (maximum) measurements from each tooth cusp (Table 1; Fig. 2). Assuming that the lowest and highest δ¹⁸O values measured in each tooth represent the wet and dry seasons respectively, the seasonal similarity between EN and LN samples indicates stability in Mediterranean climates over time. A more detailed examination of the seasonal data are provided for the M₃s below.

Table 1.

Results summary

Site/period	EN HC	LN HC	LN HTC
${\delta }^{13}{\mathrm{C}}_{\text{mean}}$	−11.8 ± 0.7*	−12.2 ± 1.0	−12.2 ± 0.5*
${\delta }^{13}{\mathrm{C}}_{\text{min}}$	−12.4 ± 0.5†	−13.0 ± 1.2	−12.8 ± 0.6†
${\delta }^{13}{\mathrm{C}}_{\max }$	−11.3 ± 0.8	−11.9 ± 0.9	−11.7 ± 0.6
${\delta }^{18}{\mathrm{O}}_{\text{mean}}$	−2.4 ± 1.5‡,§	−1.3 ± 1.5‡	−1.1 ± 1.5§
${\delta }^{18}{\mathrm{O}}_{\text{min}}$	−3.3 ± 1.6¶	−3.0 ± 0.9#	−2.1 ± 1.50¶,#
${\delta }^{18}{\mathrm{O}}_{\max }$	−1.5 ± 1.4||	−1.0 ± 1.1	−0.3 ± 1.5||

One-tail t tests assuming unequal variance were performed on site mean, minimum, and maximum δ¹³C and δ¹⁸O values for EN HC–LN HC, EN HC–LN HTC, and LN HC–LN HTC pairs. Symbols mark the pairs with statistically significant results: *^{,†,§,¶,#,||}P < 0.05; ^‡P < 0.01. Means are displayed with ±1σ (‰). For complete results, see Table S2.

The mean δ¹³C values measured at LN HC and HTC are slightly negative [−0.42 – (−0.40) ‰] and significantly different from those measured at EN HC (P < 0.05; Table 1; Fig. 2). Because the ¹³C in C₃ vegetation becomes enriched when annual precipitation drops (33), this result suggests similar albeit slightly wetter growth conditions during the LN. The magnitude of the seasonal difference indicated by the most negative (presumed wet season) and least negative (presumed dry season) δ¹³C values is the same in the EN and LN periods (Table 1; Fig. 2).

Seasonal Isotope Patterns in M₃ Teeth.

Plots of sequential δ¹³C and δ¹⁸O values from individual M₃s from EN HC and LN HTC provide seasonal data beginning at the time of tooth formation. The M₃ begins forming at parturition in the late spring (April–June; wet season), and continues forming until the following dry summer and fall seasons (July–November) (37, 38). The data presented in this section has been wiggle-matched to fit consensus regression lines (see Fig. S2 for unmatched data). The isotopic expression of the current Mediterranean climate is clearly observable in the combined δ¹³C and δ¹⁸O patterns of the Natufian gazelle. A similar difference between the wet and dry season is also recorded in the teeth of modern gazelle (34). Both EN HC and LN HTC δ¹³C data form uniform and overlapping linear regression lines when plotted against time. Early in the tooth’s formation, the values are consistently more negative than those measured later in the process (Fig. 3 B, D, F). The δ¹⁸O patterns are similar to those for the gazelle molar fragment data (Fig. 2) showing more negative intercepts in the EN δ¹⁸O value second-order polynomial regression curve compared with the LN (Fig. 3E; this difference is maintained when outlying specimen 44 is removed from the regression; Fig. S3). Differences in the seasonal pattern however, are visible in the amplitude of the evaporative dry season indicated by the curvature of wiggle-matched consensus regression lines. Although the EN HC sample shows a pronounced wet–dry–wet season shift, the hyperbolic curvature is shallower in the LN HTC data, implying weaker evaporative conditions during the dry season (Fig. 3 A, C, E).

Fig. S2.

Sequential sampling results of individual M₃ teeth from Early Natufian Hayonim Cave and Late Natufian Hilazon Tachtit Cave that were not corrected for minor differences in fawn birth timing.

Fig. 3.

Sequential sampling results of individual M₃ teeth from EN HC (blue-shaded plot lines) and Late LN HTC (red-shaded plot lines). The symbols in A (δ¹⁸O) correspond to those in B (δ¹³C) EN HC; and those from C (δ¹⁸O) correspond with D (δ¹³C) LN HTC; combined early and late Natufian sequential δ¹⁸O (E) and δ¹³C (F) data.

Fig. S3.

Late Natufian Hilazon Tachtit sequential M₃ teeth δ¹⁸O value curvature. Upper shows the second-order polynomial equation when sample 41 is excluded from the analysis. Lower shows data as it appears in the main text with outlying sample 41 highlighted.

Discussion

The results presented in this study clarify some ambiguities regarding the climatic effect of the YD on Natufian populations inhabiting the Mediterranean zone in the southern Levant. The interpretation of high-resolution cave speleothem oxygen isotope data (3) requires modern day analogs (25) and supporting oxygen isotope-based paleoclimate proxies (27, 39). The current study uses oxygen isotope values obtained through the sequential sampling of gazelle tooth enamel as a proxy for the seasonal composition of environmental water; this is made possible through comparison with data on water availability obtained from complementary carbon isotope data. A significant advantage of this proxy is that it derives directly from cultural remains, enabling direct association of human behavioral adaptations with climatic and environmental conditions.

Our results reveal no evidence for the aridification that has typically been associated with the YD. To the contrary, the carbon isotope values of the large dataset of tooth enamel cusps show that the water available to C₃ plants was similar to or even greater in the YD than the EN phase, which falls within the Bølling–Allerød period (Fig. 2; Table 1). Because the carbon data show that this change cannot represent a decrease in precipitation, the positive shift in oxygen isotope values from the EN to the LN (${}^{\mathit{18}}\mathit{\Delta }_{EN-LN}$ = 1.32–1.14‰) must reflect a change in the oxygen isotopic composition of Mediterranean seawater. This change would result from a drop in the temperature of the seawater—the source of Eastern Mediterranean precipitation (Fig. 1; refs. 23, 39). This result joins a growing body of evidence pointing to wet conditions during the YD including an abrupt rise in the Dead Sea Basin shore lakes (Fig. 1D; ref. 9) and the formation of a dark manganese and barium oxide desert varnish crust on the paleo-lakeshore pebbles (6). These results also help to resolve some discrepancies in recent reconstructions of the YD. For example, the increase in the presence of the arid-adapted plant genus Artemisia during the YD was originally interpreted as a marker of aridification (8). Given that vegetation was not water stressed during the YD, this is likely the result of the expansion of aeolian sand and loess sediments to which Artemisia is adapted (5, 10, 40) and thus a product of edaphic rather than climatic conditions.

The scale of seasonal change represented in the carbon and oxygen isotopes values from EN and LN gazelle carbonates reflects the typical modern Mediterranean climate pattern (wet and cool winter, hot and dry summer). Gazelles shift from protein-rich green herbaceous vegetation during the wet season to evergreen ligneous vegetation during the dry season (38). This dietary shift is manifested in the gradual increase in δ¹³C values from early to late tooth formation (Fig. 4D; ref. 34). This period corresponds to a 6-mo period (37, 38) between the late spring, when most modern fawns are born, and the late fall, when the early rains begin but fresh green herbaceous vegetation has not yet sprouted. The oxygen isotope pattern fully agrees with the carbon pattern—negative oxygen values rise across the dry summer season and decline when temperature drops in the fall (Fig. 3 A and C). This seasonally cyclical transition from cold and rainy to hot and dry is measured locally in the δ¹⁸O values of modern water vapor collected from Israel’s seashore (Weizmann Institute) (41). Water vapor is an ideal analog for the oxygen isotope values of gazelle teeth because δ¹⁸O values in plants and animals are determined by evaporation and transpiration during the dry season when measureable rainfall is absent. The consensus curve formed between δ¹⁸O values and the distance from the tooth’s apex in M₃s dated before the YD (EN HC) mirrors the curve of modern water vapor, and the curve of gazelle teeth deposited during the YD (LN HTC) is shallower (Fig. 4). Comparisons of the gazelle trend lines show that the smallest difference between the EN and LN occur in the dry season (June–September), and the largest differences occur in the wet season (December–April; Figs. 3 and 4). Although wet season differences reflect a change in water source oxygen isotope values, the dry season data indicate that evaporation was less pronounced during the YD than the EN. The typical Mediterranean pattern of wet winters and dry summers is not altered during the YD, but a drop in temperature would have decreased the evaporative regime. This would have increased the water available for plants even if rainfall remained constant in the wet season. The decrease in the magnitude of difference in δ¹⁸O values between the wet and dry seasons during the YD is supported by high-resolution isotope and fluorescence patterns measured in this time period at Soreq Cave (speleothem N2; ref. 7).

Fig. 4.

Plot of seasonal near-surface water vapor δ¹⁸O values (gray symbols; each symbol type relates to a specific year) measured between 1998 and 2006 on Israel’s seashore overlain by gazelle M₃ tooth enamel regression lines. Pre-YD (EN HC): blue dashed line, YD (LN HTC): red line. Data from ref. 41.

If annual temperature dropped significantly during the YD and mean annual precipitation remained relatively constant, this alone could explain the modest negative shift in δ¹³C values between the EN to LN phases (ref. 34; Fig. 2). Soil and air temperature during the YD are unknown, but sea surface temperature proxies (TEX₈₆, ${\text{U}}_{37}^{\text{k}\mathrm{’}}$) indicate a drop of about 2–3 °C between the Bølling–Allerød and the YD (39). It is likely that seasonal and annual terrestrial temperature amplitudes would have been pronounced during a period of Mediterranean surface water cooling. Further work on carbonate temperature proxies (i.e., Δ₄₇) in cave speleothems might enable better reconstruction of terrestrial temperatures during the YD (42). Clarifying the speed of the onset of this climatic event is also important because it influences the ability of the local biome to adjust to changing climate conditions. Current data suggest that the cooling associated with the YD took place on a decadal scale (ca. 20 y) (43) and thus would have had a more profound effect on local biota and humans.

Other factors that could influence our carbon isotope data and mask increased aridity during the LN are a change in the relative balance of C₃ and C₄ plants in gazelle diets or a change in human hunting or gazelle foraging territories. In the Mediterranean Levant C₄ grasses are sporadic and available as a green food source only in summer when herbaceous C₃ vegetation is dry. Modern gazelles inhabiting these habitats in the Levant typically consume small quantities of C₄ grasses in the dry season (38). If conditions were much drier in the YD, then C₄ grasses may have been more abundant in Mediterranean habitats and potentially more important in gazelle diets, at least in the dry season. In this case, we would expect higher δ¹³C values and a greater difference between dry and wet season values in LN than EN gazelles (34). Our data do not meet either of these expectations (Fig. 3; Table 1). If conditions were drier in the YD, humans or gazelles also could have adapted by foraging at higher elevations where rainfall was high enough to offset the drying impact of the YD. If gazelle foraged at higher elevations, their δ¹⁸O values would be more negative (44); and again, our data contradict this expectation. Finally, the transport of near-complete gazelle carcasses (45) and tight clustering of both our within-group δ¹³C and δ¹⁸O values (Fig. 4; Table 1) suggest that humans foraged relatively close to home in both the EN and LN.

Our results challenge the entrenched model that warm temperatures and high precipitation are synonymous with climatic amelioration and cold and wet conditions are combined in climatic downturns. Our results indicate that precipitation remained constant across the Natufian, and temperatures became colder in the Mediterranean zone. The relationship between this climatic and cultural change is complex, partially due to spatial variation in the scale of climatic change given the diverse topographies and environments of the southern Levant (i.e., climatic change may be more pronounced in the Mediterranean zone than the Jordan Valley) and partially due to the low resolution of Natufian chronologies, in particular the transition from the EN to the LN. The onset of the YD has traditionally been associated with the beginning of the LN, however absolute dates, especially those from the Mediterranean zone suggest that the LN may have begun up to 600 y before the onset of the YD (14, 20). Thus, many of the cultural markers of the LN, such as changes in lithic technology and burial practices likely emerged before the YD. The archaeological material in this study derives from well-dated LN contexts (30, 31), and thus the scale of the YD’s impact can be detected in human subsistence and settlement strategies (15). Consequently, given evidence for a drop in temperature during the YD, we must now ask how this event might have affected the economy of highly sedentary Natufian societies?

Environmentally speaking, the answer probably lies in the ability of wild cereal species to maintain biomass productivity in a given stand at lower ambient temperatures. The discovery of sickle blades, ground stones, abraded tooth wear patterns, and increased abundance of grain phytoliths all point to the importance of grains in Natufian diets. Furthermore, grinding stones, and bedrock mortars are more common at LN than EN sites (11, 46, 47). Use wear analysis showed that these tools were used for grinding cereals and legumes typically associated with increasing dependence on cereals (46, 47).

The impact of lower growing season temperatures on cereal productivity is not well studied in semiarid Mediterranean environments because water availability rather than temperature is the primary factor determining crop biomass productivity today (48). Studies conducted on commercial and wild cold-adapted cereals reveal species-specific optimal temperature requirements for germination and growth (49). In a regionally relevant study from Israel the seeds of various wild barley (Hordeum spontaneum) populations were translocated into new environments (50). Seeds originating from warm climates showed high mortality when planted in a cool mountain plot where early growth temperatures range below freezing. Also, without exception all barley seeds sown in the cool location showed lower biomass productivity (<50% of optimal growth rate) after 2 mo of growth, although final reproductive size was not compromised (50). These results suggest the kind of impact that the YD could have had on wild grain growth. The rapid onset of the YD, may have caused an initial drop in grain production by killing seeds in winter frost events, and more importantly, delaying the germination and maturation of cereals and thus affecting the timing of the harvest. Although rapid natural selection for cold-adapted cereal variants likely took place, a delay in the germination and maturation of cereals would have been detrimental for human populations reliant on these seeds.

EN populations sustained their sedentary lifestyle by increasing environmental carrying capacity through the exploitation of abundant low-ranked foods such as cereal grains and small game animals, but this would have made them more sensitive to subtle compromises in food availability, especially of the low-ranked wild grasses that enabled sedentism in the first place. Thus, the YD may have disrupted the sedentary lifestyle of Natufian populations in the Mediterranean zone, which fits archaeological data suggesting that they adapted to new conditions by increasing population mobility (12, 15), and seeking more reliable grain stands in more climatically stable environments (e.g., the Jordan Valley; refs. 11, 51). Increased population mobility is evidenced by a significant decline in the scale and intensity of construction of architectural features and thickness of site deposits in EN versus LN sites (12). Some EN sites take on new primary roles as burial locales in the LN—at HC LN activity centered around burials interred toward the back of the cave. Only two insubstantial walls were constructed in the LN compared with several round structures, some with built hearths, slab-lined pits, and pavements in the EN (30). HTC also served primarily as a human burial site (29). Additionally, faunal indices of slow- and fast-moving small game reveal a significant drop in site occupation intensity during the LN in the Mediterranean zone (15). In contrast, the LN site of Nahal Ein Gev II indicates that the same is not true in the Jordan Valley. Excavations have revealed a substantial village with dense deposits and diverse architecture including the foundations of large structures, pits, and burials that reflect intensive occupation (52). Recent survey and testing at the site of Huzuq Musa farther south in the Jordan Valley also hint at large architectural features and dense archaeological deposits, although the site was only excavated to a limited extent (51).

Further work on the impact of cooling on wild cereal growth and productivity in Mediterranean environments is necessary to provide higher-resolution data on the potential affects of the YD on resource availability. Future work will also establish the degree of reliance of Jordan Valley residents on cereals during the LN.

Materials and Methods

Gazelle Tooth Sampling and Processing.

Two groups of gazelle molars were available for study (n = 44 teeth) from clear stratigraphic contexts at HC and HTC (Table S2). The first group includes lower third molars (M₃) from EN and LN contexts (n = 12). M₃s were chosen for analysis because unlike the M₁ and M₂ they are easy to identify to element when recovered in isolation (37). The seasonal signal created by sequentially sampling the M₃ can be directly compared to detect seasonal differences between the EN and LN phases. The second group includes molar cusp fragments from secure EN and LN contexts that were not sufficiently diagnostic to be assigned to a specific molar (n = 32). This sample was predominantly used to provide more robust climate data and to improve the confidence levels of the mean δ¹³C and δ¹⁸O values.

Table S2.

Provenience data for the gazelle molars available for this study

Number	Site	Year	Square		Subsquare	Depth below datum	Locus	Fauna Cat. No.	Phase
1	HC	1978	L	30	c	245–250		15903	EN
2	HC		H	25	a+d	266–270		16066–16067	EN
3	HC	1978	K	30	a	246–250	7	15908+15909	EN
4	HC	1969	I	25	c	238–248		16095	EN
5	HC	1975	K	26	c	184–189		15882	LN
6	HC	1978	K	30	a	235–240		16098+16099	EN
7	HC		J	26		190–195		16068	LN
8	HC	1977	M	28	b	204–210		15880	EN
9	HC	1978	J	30	b	245–250		15901	EN
10	HC	1979	O	24	a	180		15886	LN
11	HC	1977	M	27	a+c	208–217		15906	EN
12	HC	1979	O	27	d	145–150		16087	LN
13	HC	1978	I	29	d	240–245		16090	EN
14	HC	1978	Q	27	a+d	165		15899	LN
15	HC	1979	M	23	b	175–180		15891+15892	LN
16	HC	1977	M	24	d	173–178		15895	LN
17	HC	1975	L	25+26	c+a	188–195		15893	LN
18	HC		I	30	b	240		16074
19	HC	1979	O	30	d	145–150		15913	LN
20	HC	1975	K	26	d	193–202		15887	LN
21	HC	1975	M	28	b	154–158		779
22	HTC	2001	M	13	a	318–325	1	3014	LN
24	HTC	2001	J	14	d	310–315	2	3025	LN
25	HTC	2001	L	14	d	334	1	3021	LN
26	HTC	2001	M	13	b	335–340		3022	LN
27	HTC	2006	M	13	c	365–370	1		LN
28	HTC	2001	J	14	d	310–315		3026	LN
29	HTC	2005	M	14	a	357–361	1	5316	LN
30	HTC	2001	L	14	b	332	1	3018	LN
31	HTC	2005	J	14+15	c+d, b		2	4558	LN
32	HTC	2005	M	14	d	307–328	1		LN
33	HC	1998	O	24	c	222	8		EN
34	HC	1999	O	32	a	195–200		3605	EN
35	HC	1998	N	24	b	222–225		2519	EN
36	HC	1994	N	24	d	215	8	107	EN
37	HC	1994	P	32	a	176		2	EN
38	HC	1977	M	28	a	198–203			EN
39	HTC	2001	M	14	c	335–340		3039	LN
40	HTC	2008	K	14	a	373	2	5784	LN
41	HTC	2006	M	14	a	360–365	1	5302	LN
42	HTC	2000	K	12	d	320–325		1674	LN
43	HTC	2008	J	14	b	355	2	5519	LN
44	HTC	2001	K	14	c	315–320		3040	LN
45	HTC	2005	K	14	a	345	2	4365	LN

The external surfaces of the teeth were cleaned by manually removing adhering debris and sediment and were then mounted on a carrying glass. The teeth were microsampled using a micromilling system (New Wave Research). After mounting, the sampling area was mechanically cleansed using the micromill drill by removing 5 µm of enamel from the tooth’s surface. Sampling took place sequentially on a single labial tooth cusp starting from the tooth cervix and advancing toward that tooth enamel apex in 1-mm intervals. Each sample was drilled to an average depth of 450 µm. This strategy ensured comparability among specimens and allowed the seasonal signal from multiple teeth from a single cultural context to be aligned so that statistically significant results could be produced. The drill head was cleaned with 0.5 N hydrochloric acid and then dried with 100% ethyl alcohol after each sequential sample was extracted to eliminate potential contamination.

The enamel powder (∼3 mg per sample) was treated to remove organic remains and unbound exogenous carbonate following Balasse et al.’s (53) protocol. The samples were first treated with 2 mL of 2% NaOCl (sodium peroxide) in reaction tubes to remove organic remains. The samples were agitated and then left to react with the NaOCl solution for 24 h and then washed and agitated 3 times with Milli-Q H₂O (MΩ.cm 18.2 Milli-Q; Millipore). Later the samples were reacted with highly diluted 0.1 N pH 3 CH₃COOH (acetic acid). The acid reaction was limited to 4 h, and was terminated by repeated washing and centrifuging until the solution reached neutrality (pH 7). Water was decanted from the reaction tubes after the final centrifuging cycle and left in the desiccating oven to dry at 50 °C overnight.

The birth of modern mountain gazelle can vary within the spring season (March–May) in accordance with peak biomass productivity (38). Thus, the timing of initial tooth formation should not be fully synchronized among individuals. Wiggle-matching over minute distances (<3 mm) was used to align the seasonal patterns in individual teeth when both the carbon and oxygen isotope data from a tooth lined up with the average trend lines.

Stable Isotope Analysis.

The samples were analyzed at the stable isotope facility of the Department of Écologie et Gestion de la Biodiversité, at the Muséum National d'Histoire Naturelle, Paris, France. Approximately 600 µg of treated enamel powder was reacted for 240 s with 100% phosphoric acid at 70 °C in individual vessels in a Kiel IV carbonate device (Thermo Scientific). The evolving CO₂ was transferred by vacuum to a Delta V Advantage isotopic ratio mass spectrometer (Thermo Scientific) where m/z 45, 46, 47 were converted to isotopic ratios using the following equation:

$$\delta \hspace{0.17em}sample\hspace{0.17em}(‰)=\left(\frac{R\hspace{0.17em}sample-R\hspace{0.17em}standard}{R\hspace{0.17em}standard}\right)-1\times 1,000,$$

where R is the ratio of ¹³C/¹²C to ¹⁸O/¹⁶O normalized against the Vienna Pee Dee Belemnite (VPDB) standard.

The analytical precision of δ¹³C and δ¹⁸O values determined by repeated measurement of in house carbonate standard MarberLM (n = 53; ${\mathrm{\delta }}^{13}{\text{C}}_{\text{VPDB}}$ = 2.130‰ and ${\mathrm{\delta }}^{18}{\text{O}}_{\text{VPDB}}$ = −1.830‰) were 0.026‰ (1σ) and 0.044‰ (1σ) respectively. The MarberLM in house standard was calibrated against IAEA NBS 19 (NIST).