Record of paleo water divide locations reveals intermittent divide migration and links to paleoclimate proxies

Significance

Drainage divide migration reshapes drainage basins, redistributing water, erosion, and mass of sediments and rocks, as well as ecological niches. Existing approaches struggle to detect variations in migration rates on 10⁴ to 10⁵ y timescales, hindering our ability to resolve transience in migration rates and their link to environmental forces acting on the landscape. Combining field observations, alluvial terrace dating, and numerical simulations, we infer divide migration dynamics in the Negev Desert, Israel, since ~230 ka. We find that episodes of accelerated migration coincide with regional climate fluctuations indicated by regional paleoclimate proxies. These results reveal a previously undetected resolution of drainage divide dynamics and suggest potential associations with climate fluctuations, enhancing our understanding of Earth’s topographic, hydrologic, and ecological evolution.

Abstract

Drainage divide migration alters the geometry of drainage basins, influencing the distribution of water, erosion, sediments, and ecosystems across Earth’s surface. The rate of divide migration is governed by differences in erosion rates across the divide and is thus sensitive to spatiotemporal variations in tectonics and climate. However, established approaches for quantifying divide migration rates offer only indirect evidence for the motion of the divide and provide only migration rate averages. Consequently, transience in divide migration cannot be resolved, hindering the ability to explore environmental changes that drive the dynamics of such potential transience. Here, we study a set of datable terraces identified as markers of paleo-divide locations, which provide direct evidence for the paleo motion of the divide. The location and age of the terraces reveal intermittent divide migration at timescales of 10⁴ to 10⁵ y, with phases of rapid migration—at rates more than twice the average—which coincide with documented regional paleoclimate fluctuations. These findings highlight the intermittent nature of divide migration dynamics over geomorphic timescales and its potential sensitivity to climate changes, underscoring the impact of such changes on the planform evolution of drainage basins.

Drainage divides define the geometry and delineate the area of drainage basins by setting the boundaries between adjacent drainages. As early as the 19th century, Gilbert (ref. 1) posited that asymmetric slopes across a divide and the associated erosion rate gradients are expected to cause divide migration, which, in turn, changes the geometry of the basins over time (2–6). Such changes could result in reorganization of the drainage pattern over multiple scales (e.g., ref. 7), with impacts on landscape functionality, including changes to groundwater and surface water hydrology (e.g., ref. 8), alterations to ecological niches and biodiversity (9, 10), and rerouting of sediment systems with consequences to the interpretation of offshore and fluvial stratigraphic sections (e.g., ref. 11). Therefore, the direction, rate, and driving mechanisms of divide migration are progressively recognized as fundamental factors controlling landscape evolution in the broader sense (8, 12–14).

The migration of divides is commonly attributed to climatic, tectonic, and lithologic perturbations over varying timescales (e.g., refs. 15 and 16), which alter the spatial distribution of erosion rates and potentially create gradients in erosion rates across divides that result in divide migration (10, 17, 18). Additionally, the specifics of these perturbations can alter prevailing erosional processes, facilitating migration of divides that were previously static. Numerical and experimental studies have demonstrated that the peak rates of divide migration in response to such perturbations are within the range of 10² to 10³ m my⁻¹ (13, 19–21).

Identifying and measuring divide migration in the field is challenging, because datable geomorphic markers that could potentially preserve the migration record are commonly erased by the erosional nature of the migration process. Therefore, studies targeting migration of divides in natural settings, utilized morphometric proxies for divide stability, including relief, slope, and χ gradients across the divide (19), where χ is a parameter that quantifies the expected steady-state elevation along the channels (3). However, while these proxies predict the sense of divide migration, they cannot determine the migration rate without additional data (19, 22).

A common approach for field-based assessments of migration rates combines basin-averaged erosion rates, inferred from cosmogenic nuclides (CRN) concentrations in fluvial sediments, and slope measurements across the divide (22–24). However, this approach has several limitations: First, CRN-based erosion rate estimates represent temporal averages (25), which means they cannot detect short-term variations in erosion rates during the averaging period. Therefore, when the rates are averaged over short periods, they may not be representative of the long-term basin erosion. Second, in many cases, migrating divides bound catchments that are steep and landslide-dominated, potentially leading to biases in the inferences of the CRN-based erosion rates (26) and resulting in large uncertainties in the computed migration rates (27). Third, this method relies on a simple geometric model (23) that may not necessarily align with the local field conditions. The alignment between the model and field may be further complicated by the model reliance on the scale and technique used for slope measurements (24), along with the possibility of slope variations over time.

To date, the limitations of the aforementioned approaches, along with the scarcity of field sites capable of capturing temporal changes in erosion rates across the divide, have resulted in a lack of direct evidence for the actual motion of divides, as well as indication for temporal variations in this motion. This gap further hampers our ability to discern migration rate variations in response to environmental changes such as climate fluctuations. Here, we leverage a unique field setting in which geomorphic markers of paleo-divide locations are preserved in the form of alluvial terraces. Absolute and relative dating of the abandonment ages of these terraces, using soil analysis and luminescence techniques, provide direct evidence for the motion of the divide. These ages resolve changes in the rate of divide migration over timescales of 10⁴ to 10⁵ y, revealing potential links between drainage divide dynamics and regional climatic changes.

Conceptual Model

Here, the dynamics and rate of divide migration are explored in a drainage reversal setting (Fig. 1). The exploration is guided by a conceptual model relating divide migration to the formation of alluvial terraces along reversed drainages. Drainage reversal occurs when a channel reverses its flow direction while exploiting its antecedent valley (6), and is commonly identified by barbed tributaries and a wind gap (28), i.e., a valley-confined section of the drainage divide that is located between the reversed drainage and the beheaded valley that lost its headwater area due to the reversal (Fig. 1). Reversals are common near cliffs, where highland channels that originally drained away from the cliff are redirected toward it (6, 27, 29). Harel et al. (6), demonstrated that in many cases, reversals toward cliffs are related to localized wind gap migration away from the cliff along antecedent valleys. This migration is commonly facilitated by a higher erodibility of the valley bed compared to the more resistant interfluves due to, for example, the presence of erodible sediments filling the valley (e.g., Fig. 1).

Fig. 1.

Cartoons of the conceptual model. (A) At time T0, prior to drainage reversal, the wind gap coincides with the cliff line and is, by definition, hydrologically disconnected from fluvial/colluvial sediment supply. In this example, the valley is filled with alluvial-colluvial sediments which are more erodible than the rocky interfluves. (B) Upon wind gap migration at time T1, the drainage area and the discharge of the reversed drainage increase, leading to incision of the reversed main stem and entrenchment of side tributaries into the valley bed, forming terraces close to the wind gap. (C and D) Further wind gap migration at time T2 forms newer terraces near the wind gap, and therefore the abandonment ages of the terraces increase away from the wind gap toward the cliff. In the case of continuous migration (C), terrace ages (represented by grayscale coloring) are expected to evenly distribute along the reversed channel, while in episodic migration scenarios (D), groups of terraces with similar ages form during rapid migration phases. While this model is applicable across a wide range of fluvial valley scales (6, 30), the depiction of colluvial fans and their activity is based on observations specific to the field area in this study.

Wind gap migration in a reversal setting progressively shrinks the area of the beheaded valley near the wind gap, while in the reversed valley, across the wind gap, the migration advances incision, eventually forming datable terraces indicative of past wind gap location (Fig. 1 A and B). Wind gap emplacement within the antecedent valley halts fluvial deposition around the wind gap (Fig. 1A) as it is, by definition, hydrologically disconnected from sediment sources (31). Incision of side tributaries near the wind gap further isolates the surface around it. Upon further wind gap migration, the entrenchment of side tributaries that bound the new wind gap location form a new terrace, while continued incision of existing tributaries maintains the isolation of the older terraces (Fig. 1B). According to this model, the abandonment age (i.e., timing of the last fluvial/colluvial deposition) of the terraces and wind gap surface corresponds to the time when the wind gap had migrated to or beyond these surfaces (Fig. 1). Therefore, the abandonment age is expected to increase from the wind gap toward the cliff. In locations where such terraces are well preserved, gradual or stepped changes in terrace ages may indicate whether the divide migration was continuous (Fig. 1C) or episodic (Fig. 1D). Consequently, dating these terraces offers a unique opportunity to investigate the dynamics of divide migration at a temporal resolution that was previously unexplored.

Using Reg Soils for Terrace Dating

A well-established approach for dating the abandonment age of alluvial terraces in hyperarid environments involves investigation of the reg soil (gypsic haplosalid aridisols) that develops on the terraces and exhibits a characteristic temporal evolution (32, 33). Reg soil formation begins after the terrace is abandoned and fluvial activity ceases and continues as long as the terrace surface remains. As the soil forms, eolian material (mainly dust) accumulates on the surface and infiltrates the fluvial sediments, leading to the development of two distinct soil horizons: a thin vesicular horizon (Av) that develops underneath the clasts and becomes enriched in silt and clay with time (34) and a deeper, gravel-poor reddish horizon (B). This B horizon overlays a gravelly horizon (C) that represents the original fluvial deposits of the terrace. As eolian material accumulates, salts from dust and rainwater precipitate and dissipate into the lower B horizon and C horizon, contributing to gravel shattering within the soil (35). In well-developed reg soils, the salts can form diagnostic horizons indurated by gypsum and halite (33). The reg soil horizons typically reddens with age due to enrichment in iron oxide (36). Concurrent with the soil development, surficial clasts undergo physical weathering that homogenizes and decreases their size, eventually forming a characteristic “desert pavement” layer on the terrace surface (37). In addition to using the time-dependent properties of reg soils for the relative dating of abandoned terraces (32), various studies have leveraged the reg soil unique structure to determine the abandonment age using absolute dating methods, including luminescence and CRN-based techniques (34, 38, 39). These methods assume that the eolian-transported grains in the Av and B horizons accumulate over time, and in the absence of bioturbation, their burial age increases with depth. Importantly, assuming persistent dust flux, the age of eolian-transported grains at the upper part of the C horizon records the transition from an active fluvial surface to reg soil formation, close in time to the terrace abandonment (34, 39, 40). In this study, we employ an integrated approach of relative and absolute dating of reg soils that develop over the terraces formed during drainage reversal. Using the conceptual model presented above (Fig. 1), we link the abandonment age of the terraces to wind gap migration, thus inferring paleo-divide locations and migration rate changes over time.

Geomorphic Setting

Drainage Reversals in the Southeastern Negev Desert.

This study focuses on abandoned terraces identified along reversed ephemeral channels in the Shaharut plateau (5, 6), located in the southeastern Negev Desert, Israel (Fig. 2). The plateau consists of Cretaceous subhorizontal carbonate-dominated strata. To the east, the plateau is bounded by a 400 to 600 m high escarpment that marks the western boundary of the Arava, a rift-shaped valley that hosts the plate boundary (Fig. 2A) between the Arabia plate and Sinai subplate (41). In the southern portion of the study area, the drainage divide between the Shaharut plateau and the Arava valley mostly aligns with the edge of the escarpment cliff, separating low-relief antecedent channels that flow west along wide (typically ~50 to 150 m) and deep (typically ~50 to 80 m) valleys, and east-flowing steep channels that drain the escarpment toward the Arava Valley (Fig. 2B).

Fig. 2.

The study area. (A) Orientation map depicting the plate boundary between the Sinai subplate and Arabia plate (red line). The red star marks the location of the study area. (B) Shaded-relief map based on the 11.6 m px⁻¹ resolution TanDEM-X DEM (42), illustrating the wind gaps and the associated drainage reversals toward the Arava Valley, reflected by local shifts of the main drainage divide from the cliff. Flow lines are color-coded by χ value (calculation details in SI Appendix, section S1). χ gradients across the main divide indicate nonequilibrium drainage basin geometry where a westward divide migration is expected. (C) Field photo of the complementary site GAY showing the sampling location at the wind gap between the reversed and the beheaded channels. A person icon near the sampling location is for scale. (D) Shaded elevation map based on a 0.5 m px⁻¹ resolution Lidar-based DEM from the Geological Survey of Israel, emphasizing the wind gaps and barbed tributaries that indicate drainage reversal in the complementary sites GAY and AYS. The camera symbol marks the location and orientation from which the photos of (C) and (E) were taken, and black polygons approximate the area captured by the photos. (E) A panorama of the complementary site AYS shows the wind gap sampling location. Note the person at the bottom right for scale.

A distinct asymmetry in topographic gradients and χ values across the divide (6) suggests that the divide is unstable and is expected to migrate westward to achieve greater stability (Fig. 2B). This westward migration aligns with tens of drainage reversals with variable migration distances (6, 43), shifting drainage area eastward toward the Arava (5, 31, 44, 45).

The study’s field sites are located along and near the Shaharut Plateau drainage divide (Fig. 2 B–E), where several wind gaps separate reversed valleys from beheaded valleys. The main site in the study, where we explore the divide migration dynamics, referred to as Maale-Yitro-North (MYN), includes a set of well-preserved terraces along a reversed channel incised into the valley fill. The study is further supported by analyzing data from two additional wind gaps at complementary field sites (Fig. 2 C–E), Grofit-Aya (GAY) and Aya-South (AYS).

Field sites.

Fig. 3 illustrates the reversal system in the main site (MYN). West of the wind gap, the beheaded valley is drained by a low-relief channel, joined by distributaries that drain local colluvial fans originating from the valley hillslopes (e.g., the colluvial fans in Fig. 1). The distributaries shallowly (<~1 m) incise into patches of slightly elevated fan surfaces within the valley bed (Fig. 3A). The desert pavement on these surfaces is darker than the gravel in the active channel, indicating the surfaces are transiently hydrologically disconnected. East of the wind gap, the reversed channel incises into erodible alluvial-colluvial sediments that fill the valley bed (Fig. 3B). The channel is joined by a barbed tributary from the north (Fig. 3A), and several tens of meters east of the confluence with this tributary, bedrock is exposed at the channel bed. Farther downstream, ~500 m eastward from the wind gap, the channel crosses the escarpment, forming a steep waterfall.

Fig. 3.

Overview of the main site (MYN, location in Fig. 2B). (A) Abandoned terraces and locations of soil sampling profiles on a shaded elevation map, based on a 0.5 m px⁻¹ resolution DEM from the Survey of Israel. Profiles B1 and B2, situated west of the wind gap, were dug into transiently hydrologically disconnected fan surface patches along the valley bed (SI Appendix, Fig. S1). Cross-section A–A′, B–B′, and C–C′, respectively, are shown in panels B and C. (B) Topographic cross-section A–A′, based on a 0.09 m px⁻¹ resolution DEM (SI Appendix, section S2 and ref. 46, and a schematic subsurface contact between the bedrock and the alluvial-colluvial valley fill. The cross-section reveals a slope decrease from the steep hillslopes toward the flatter terrace that is bounded by a steep riser resulting from the incision of the reversed channel. (C) Location and elevation of the soil sampling profiles, projected along the topographic transect B–B′ that crosses the edge of the terraces. Transect C–C′ depicts an elevation profile along the beheaded channel, the wind gap, and the reversed channel. Accounting for the offset caused by a normal fault\slump, the terrace elevations exhibit a slight west grading toward the wind gap (the semitransparent line represents reconstructed profile), indicating that before the reversal, they were part of the antecedent west-flowing valley bed (SI Appendix, section S3 and ref. 6).

Along the reversed valley, side tributaries of the reversed channel incise into the valley fill and dissect it into terraces. We focus on the terraces along the north bank that are relatively well preserved, as they are distinctly separated from the channel by steep risers (Fig. 3B) and are covered by well-developed desert pavement, pointing to the existence of mature reg soil on the terrace. Well-developed desert pavement and the lack of active channels on the wind gap location indicate that it is hydrologically disconnected as well (31).

A local fault or slump offset eastern terraces from terraces closer to the wind gap by ~5 m (SI Appendix, Fig. S2). Geomorphic evidence suggests that the offset occurred after the formation of the terraces and was not involved in their formation (SI Appendix, section S3). After correcting the offset, the terraces grade west, toward the wind gap (Fig. 3C), indicating that before the reversal, they were part of a continuous surface belonging to the antecedent, west-grading valley bed (6).

In the complementary sites (GAY and AYS), we focus solely on the wind gap area. Given that the wind gaps and terraces exhibit the common property of being hydrologically disconnected, we refer to them collectively as “terraces” and employ the same techniques to determine their abandonment ages.

Based on the conceptual model (Fig. 1), we explored the terraces in all study sites (MYN, AYS, and GAY) and determined their abandonment ages, based on reg soil relative and absolute dating approaches as explained above (and in Materials and Methods), aiming at i) Testing the conceptual model, ii) Constraining temporal changes in wind gap migration rates by reconstructing the paleo-locations of the wind gap through time (e.g., Fig. 1 C and D), and iii) Exploring potential drivers for wind gap migration dynamics in this region.

Results

Reg Soil Chronosequence.

The properties of the reg soil profiles are presented in SI Appendix, section S4 and Table S1. The profiles generally show the distinct stratigraphy of reg soils, including (from top to bottom): Desert pavement on the surface (SI Appendix, Table S2), a vascularly horizon (Av), a reddened gravel-poor horizon (B), and a hardened, clast-rich horizon (C). All soils contain varying amounts of gypsum and halite. At the main site (MYN), profile R4 deviates from the common reg stratigraphy, likely due to erosion of the upper part of the soil (SI Appendix, section S3 and Fig. S2http://www.pnas.org/lookup/doi/10.1073/pnas.2408426122#supplementary-materials).

The time-dependent properties of the soils that were measured for relative dating of the terraces (Materials and Methods) are presented in Fig. 4 and described in detail in SI Appendix, section S4. At the main site (MYN), these properties, including the soil color (SI Appendix, Table S1) clearly show a trend of eastward increased maturity of the reg soil and thus an eastward increase in terrace abandonment age, in agreement with the conceptual model proposed above (Fig. 1).

Fig. 4.

Time-dependent parameters of reg soils in the study area, supporting a trend of soil maturation from west to east in the main site MYN, and indicating relatively young soils at the wind gap profiles on all sites. (A) Percentage of surface area covered by desert pavement clasts (SI Appendix, Table S2). (B) Volumetric percentage of silt and clay in the Av horizons, demonstrating a higher content in the eastern profiles. (C and D) Maximal concentrations of gypsum and halite (C and D, respectively), represented by the SO₄⁻² and Cl⁻ concentrations extracted from the soil. Profiles R1–R4 in the main site (MYN) are distinctly richer in gypsum relative to the other profiles. Note the logarithmic y axis in panel D, emphasizing the extreme maximal halite concentrations in R1, R2, and B1 in the main site (MYN).

Aside from the desert pavement, the reg soil time-dependent properties in the wind gap profiles of the complementary sites (GAY and AYS) are generally comparable to the properties of the profiles located closer to the wind gap in the main site (Fig. 4). Observed variations among the main and complementary sites may be attributed to the sensitivity of the time-dependent parameters to local conditions, including hydrology, geology, and dust flux.

Luminescence Properties of OSL and pIR-IR₂₅₀ Ages.

Absolute dating (Materials and Methods) is based on optically stimulated luminescence (OSL). In cases where the OSL signal was saturated, indicating the sample age is likely older than the obtained luminescence age we used the post-infrared infrared stimulated luminescence signal (47) measured at 250 °C (pIR-IR₂₅₀), which extends the time-range of luminescence dating (48). The results of the OSL and pIR-IR₂₅₀ dating are presented in SI Appendix, Table S3 and Fig. S3. The reliability of these ages is supported by the well-behaved properties of the samples (SI Appendix, section S5, including Abanico plots in SI Appendix, Fig. S7) and by the consistent increase in age with sample depth (within uncertainty) across all profiles, in line with the theory of reg soil development from bottom to top by the accumulation of eolian-transported grains under the surficial clasts (49).

For older samples that were saturated or close to saturation (OSL signal >120 Gy) and were measured using both OSL and pIR-IR₂₅₀, the pIR-IR₂₅₀ ages are consistently older by 23 to 73 ka than the OSL ages (Table 1). To compare the ages obtained from the OSL and pIR-IR₂₅₀ methods, we selected the lowest sample from profile B1 and dated it using both techniques. This young and unsaturated sample was chosen due to the high reliability of its OSL age (19 ± 1 ka), indicated by low signal overdispersion (OD = 9%, Table 1 and SI Appendix, section S5), which reflects sample homogeneity and suggests the sample is well bleached. The pIR-IR₂₅₀ age of this sample is 32 ± 2 ka, 13 ± 3 ka older than the OSL age, likely due to the hard-to-bleach residual signal associated with the pIR-IR₂₅₀ method (48). As this residual value is not expected to increase with age in the older pIR-IR ages (50), we suggest that 13 ± 3 ka is a representative residual value of the >180 ka old terraces, which is approximately equivalent to the age uncertainty of the old terraces. Consequently, we prioritize the pIR-IR₂₅₀ ages for the older samples, as they are considered more accurate than the underestimated OSL ages.

Table 1.

Luminescence samples, including field and laboratory data. A full list of the analytical data is found in SI Appendix, Table S3.

Site name	Profile	Horizon	Lab. Code	Depth (cm)*	Mineral	Signal	Dose rate (Gy/a)	OD (%)†	N	De (Gy)	Age (Ka)‡
MYN	R1	By1	AYS-13a	3 to 8	QZ	OSL	1.42 ± 0.04	73	17/18	25.8 ± 3.3	18.2 ± 2.4
		By2	AYS-14	12 to 27	QZ	OSL	0.79 ± 0.02	33	10/10	122 ± 13	154 ± 17
					KF	pIR	1.20 ± 0.07	23	7/8	272 ± 13	227 ± 17
	R2	By1	AYS-15	7 to 12	QZ	OSL	1.00 ± 0.03	66	19/19	77 ± 12	77 ± 12
		By2	AYS-16	23 to 35	QZ	OSL	0.80 ± 0.03	16	18/18	166 ± 7	206 ± 11
					KF	pIR	1.21 ± 0.07	10	8/8	277 ± 11	229 ± 16
	R3	By1	AYS-17	17 to 22	QZ	OSL	1.00 ± 0.03	29	14/14	147 ± 12	147 ± 13
					KF	pIR	1.43 ± 0.07	6	8/8	277 ± 8	194 ± 11
		By2	AYS-18	27 to 42	QZ	OSL	0.89 ± 0.03	19	13/13	154 ± 9	173 ± 11
					KF	pIR	1.31 ± 0.07	6	8/8	296 ± 8	226 ± 14
	R4	2Cy2	AYS-25	13 to 16	QZ	OSL	0.93 ± 0.03	20	19/19	117 ± 6	126 ± 7
	R5	Cy1	AYS-19	13 to 20	QZ	OSL	1.13 ± 0.03	24	18/19	21.1 ± 0.8	18.8 ± 0.9
		Cy1	AYS-20	27 to 30	QZ	OSL	0.86 ± 0.03	17	17/18	22.9 ± 0.8	26.6 ± 1.2
	WG1	By2	AYS-6	11 ± 2	QZ	OSL	1.29 ± 0.04	20	18/19	15.5 ± 0.6	12.3 ± 0.6
		By3	AYS-5	23 ± 2	QZ	OSL	1.64 ± 0.05	16	19/19	45.7 ± 1.9	27.9 ± 1.5
		C	AYS-4	33 ± 2	QZ	OSL	1.52 ± 0.05	29	19/19	41.8 ± 2.9	27.5 ± 2.1
	WG2	By3	AYS-3	15 ± 2	QZ	OSL	1.07 ± 0.03	22	19/19	21.5 ± 1.1	20.1 ± 1.2
		Cyz	AYS-2	30 ± 2	QZ	OSL	0.97 ± 0.03	27	18/19	32.6 ± 1.5	33.6 ± 1.9
	B1	By	AYS-23	5 to 14	QZ	OSL	0.95 ± 0.03	25	17/19	16.4 ± 0.8	17.2 ± 1.0
		Cy	AYS-24	15 to 25	QZ	OSL	0.73 ± 0.02	21	18/18	13.3 ± 0.7	18.1 ± 1.0
	B2	By1	AYS-21	8 to 12	QZ	OSL	1.32 ± 0.04	15	19/19	19 ± 6	14.8 ± 0.7
		Cy1	AYS-22	27 to 32	QZ	OSL	0.97 ± 0.03	9	19/19	18.0 ± 0.5	18.5 ± 0.7
					KF	pIR	1.40 ± 0.07	14	8/8	42.2 ± 2.1	31.6 ± 2.3
GAY	GAY	B2	AYS-7	23 ± 2	QZ	OSL	1.59 ± 0.05	25	19/19	18.4 ± 1.1	11.6 ± 0.8
		Bym	AYS-8	45 ± 2	QZ	OSL	1.42 ± 0.05	23	19/19	26.0 ± 1.4	18.3 ± 1.2
AYS	AYS	B1	AYS-1	23 ± 2	QZ	OSL	1.07 ± 0.03	15	19/19	25.1 ± 1.0	23.4 ± 1.2

Notes: QZ = quartz; KF = alkali feldspar; pIR = pIRIR₂₅₀, N = number of aliquots used for calculation out of those measured (e.g., SI Appendix, Fig. S6); OD = over dispersion, the scatter beyond that expected from instrumental noise.

^*The ±2 cm range for the depth in some of the samples is estimated.

^†Low OD [<30% (51)] values reflect homogenous and well-bleached samples.

^‡Ages in regular font are saturated for OSL signal and therefore are considered underestimated.

A full list of the analytical data is found in SI Appendix, Table S3.

Luminescence Abandonment Ages.

Following Porat (ref. 39), we determine the abandonment ages of the terraces as the luminescence age of the samples taken from the upper portion of the C horizon, while trying to avoid young grains diffused into it from the B horizon, whose effect on luminescence-based age can be particularly meaningful in old terraces. The C horizon is identified by the increase of clasts and hardness relative to the overlying gravel-poor B horizon (SI Appendix, section S4). The considerations for assigning the abandonment age of each terrace are provided in SI Appendix, section S6. In agreement with the eastward increase soil maturity trend, the abandonment ages of the terraces at the main site (MYN, Fig. 5A) show a consistent increase in age from the wind gap toward the cliff. The terraces can be divided to three groups and a single terrace based on their age, presented below from west (youngest) to east (oldest):

Fig. 5.

Terrace abandonment ages in the study area and their interpretation for divide paleo-location and migration rate. (A–C) Aerial images of the study sites showing the locations of sampling points, mapped terraces, and their corresponding abandonment ages (location map in Fig. 2B). Age labels are color-coded from old (white) to young (black). Labels with dashed frame lines indicate minimum or maximum ages. At the main site MYN (A), the ages in the beheaded valley reflect transient abandonment of colluvial fan patches, whereas the ages at and east of the wind gap represent irreversible abandonment, marking the timing of wind gap migration across each terrace (Discussion). Accordingly, clusters of these ages mark rapid migration phases within narrow time windows, consistent with the episodic migration model (Fig. 1D). Minimum abandonment ages from the wind gaps at the complementary sites AYS (B) and GAY (C) agree with the age range of the wind gap group at the main site (MYN). (D) Divide migration scenarios at the main site (MYN). Shaded pink areas mark possible locations of the wind gap over space and time, as constrained by the terrace ages (horizontal bars, accounting for age uncertainty) and terrace extents (vertical bars). The solid blue line represents a migration scenario with minimal migration rate during the rapid migration phases (SI Appendix, section S7), followed by slow migration periods. The dashed red line depicts a scenario with the maximal migration rate (instantaneous) during the rapid migration phases, followed by extended stall periods. This latter scenario includes an optional phase of rapid migration related to the maximum abandonment age of R4 (<126 ± 7 ka. Abandonment ages of terraces B1 and B2 in the main site (MYN), and of the wind gaps in the complementary sites (GAY and AYS) are not shown here, as they are not part of the divide migration sequence in the main site (MYN).

i) The beheaded group includes the westernmost profiles B1 and B2 dated to 19 ± 1 and 18 ± 1 ka (OSL), respectively. These profiles are located on small colluvial fans west of the wind gap and serve as control profiles. Their ages likely reflect the time since these fan portions were fluvially active, before they were transiently disconnected from their feeding tributaries by avulsions and/or incision (see Discussion).

ii) The wind gap group includes terraces WG1, WG2, and R5, dated to 28 ± 2, 34 ± 2, and 27 ± 2 ka (OSL), respectively.

iii) Terrace R4, dated to 126 ± 7 ka (OSL) was sampled from the uppermost gravelly fluvial horizon (C), which, unusually, is not overlain by a gravel poor horizon (B), possibly due to soil erosion. As the transition between the gravel poor (B) and fluvial (C) horizons is not exposed in this terrace, the dated horizon does not necessarily reflect the latest deposition before abandonment. Consequently, its age is considered as the maximum abandonment age of the terrace.

iv) The eastern group includes the easternmost terraces R1, R2, and R3, dated to 227 ± 17, 229 ± 16, and 212 ± 29 ka (pIR-IR₂₅₀), respectively. The age and age uncertainty of R3 reflects the range of two luminescence samples taken from the C horizon (SI Appendix, section S6).

The conceptual model that links the terrace abandonment ages to the wind gap migration (Fig. 1) allows calculating the long-term average wind gap migration rate in the main site (MYN). Specifically, the abandonment age of the easternmost surface R1 reflects the time required for the wind gap to migrate from its ancient location east of R1 to its present position (Fig. 5A), revealing an average rate of ~1,143 ± 86 m my⁻¹ since 227 ± 17 ka (calculation details in SI Appendix, section S7). Temporal changes in the migration rate within this average, based on the abandonment ages of the other terraces, are discussed in the following section.

The abandonment ages of the wind gap at the complementary sites AYS and GAY are minimum ages of 23 ± 1 ka (OSL) and 18 ± 1 ka (OSL), respectively (Fig. 5 B and C and SI Appendix, section S6). These minimum ages suggest that the actual abandonment ages in the complementary sites are likely older and may align with the age range of the wind gap group in the main site MYN.

Discussion

Relating Abandonment Ages to Pattern and Rates of Wind Gap Migration.

The abandonment ages of the control profiles at the beheaded side of the main site (B1 and B2 in MYN), located over colluvial fan surfaces (Fig. 5A), indicate a reg soil that began to develop at ~20 to 17 ka (including age uncertainties). Here, at the beheaded side of the wind gap, the shallow incision of the distributary system into the fan surfaces indicates that the fan is still active. Consequently, an additional event that could potentially bury or erode the reg soil, for example a local avulsion, is likely in the future. Therefore, the ages on the beheaded side of the wind gap likely represent transient abandonment that is part of local fan dynamics.

Conversely, the terraces on and east of the wind gap (WG1, WG2, and R1 to R5 in MYN) are completely isolated from the active drainage by the incision of the main stem of the reversed channel and its tributaries (Fig. 3). Therefore, at these terraces, the abandonment ages represent irreversible terrace abandonment. The formation of such terraces necessitates westward wind gap migration to, or past, the terrace location, as illustrated by the conceptual model in Fig. 1. Therefore, we suggest that the abandonment age of the terraces on and east of the wind gap at the main site (MYN), represents the time when the wind gap first arrived at, or west of, the terrace location.

Similar to the episodic divide migration scenario in the conceptual model (Fig. 1D), reflected by groups of terraces with similar ages along the reversed channel, the terrace groups with similar age in the main site (MYN) suggest phases of rapid wind gap migration past this group. The timing of these migration phases can be constrained by the abandonment ages of the terraces in the group (Fig. 5D). Considering the ages of the two main terrace groups at and east of the divide, we suggest that two primary phases of rapid wind gap migration occurred at 245 to 183 ka, and 36 to 26 ka, as represented by the age ranges of the eastern terrace group and the wind gap terrace group (including age uncertainties), respectively. Between these phases, the wind gap likely stalled or migrated slowly (dashed red and solid blue lines in Fig. 5D, respectively). During these stall/slow migration periods, the wind gap was located adjacent to or at the west-most terrace of the terrace group, similar to its present-day configuration where the wind gap is located at the west-most surface of the wind gap group (Figs. 1 and 5).

Terrace R4, located at the main site (MYN) between the eastern and the wind gap terrace groups, has a maximum abandonment age of 126 ± 7 ka (OSL). This age does not align with either terrace group and could represent a possible rapid phase of divide migration across the terrace, occurring at or after 126 ± 7 ka (dashed red line in Fig. 5D). Alternatively, it may reflect a single timestamp between the old and recent rapid migration phases, during a prolonged period of continuous slow divide migration (solid blue line in Fig. 5D).

Based on the abandonment ages and spatial extent of the terrace groups, we estimate the minimal migration rates during the rapid migration phases (solid blue line in Fig. 5D) to be ~2,662 and ~2,200 m my⁻¹ for the old and recent rapid migration phases, respectively (SI Appendix, section S7). These rates are more than two times the average migration rate of 1,143 ± 86 m my⁻¹ (see results) and are more than five times higher than the maximal migration rate during the slow migration period, ~442 m my⁻¹ (solid blue line in Fig. 5D).

Importantly, these rapid migration rates represent minimum estimates, as they are based on a scenario that includes a period of slow, but not negligible, migration rate between 183 and 36 ka. An alternative scenario, involving very rapid episodes of divide migration followed by prolonged stalls (red dashed line in Fig. 5D), would result in significantly higher migration rates, with differences between slow and rapid phases potentially spanning several orders of magnitude.

These results establish distinct phases of rapid drainage divide migration, interspersed with periods of significantly slower migration or even stalling. This intermittent pattern of divide migration highlights that even when average migration rates are quantified, they may not fully represent the long-term dynamics of the reorganizing drainage basin when they average over short timescales (10² to 10⁴ y). Conversely, averages calculated over longer timescales could encompass long intervals of slow migration rates interspersed with relatively short phases where instantaneous rates are meaningfully higher.

Associations between Rapid Migration Phases to Regional Paleoclimate.

The minimum abandonment ages of the complementary sites wind gaps (23 ± 1 ka at AYS and 18 ± 1 ka at GAY), are in overall agreement with the abandonment age of the wind gap terrace group at the main site (MYN, 36 to 26 ka), likely implying that the latest migration phase was part of a broader regional pattern of wind gap migration. This phase likely began after 36 ka and ended around 22 ka, and no later than 17 ka (considering the minimum age uncertainties of the wind gaps at the complementary sites AYS and GAY, respectively). This migration phase was followed by wind gap stalling, or a slow rate of migration, that continues to the present. Accordingly, we propose that the driving force behind this recent migration phase, and likely older phases as well, is a regional factor. Given that tectonic activity is less likely to be the driving force in this study area (SI Appendix, section S8), we investigate the link between rapid migration phases (shaded gray areas in Fig. 6) and climatic fluctuations by exploring correlations between these phases and paleoclimate proxies from the southern Negev and Arava region.

Fig. 6.

Correlations between inferred rapid migration phases and paleoclimate proxies, and modeled wind gap migration through time in the MYN site. (A) The location and the abandonment age of terraces, shown by error bars representing their age uncertainty and their spatial extent. East of the wind gap, groups of similar abandonment ages are interpreted as associated with phases of rapid wind gap migration (shaded areas). Lighter shading around 126 ka represents an optional rapid migration phase associated with the maximum abandonment age of R4. The semitransparent curves represent simulated wind gap migration under different climate scenarios (Model details in Materials and Methods and SI Appendix, section S9). The dashed gray curve shows simulation results for constant climatic conditions, represented with time-invariant erosional efficiency, where migration rate fluctuations primarily occur through the effect of tributary capture due to divide migration (i.e., ref. 30). The solid pink curve shows simulation results for time-variable climate conditions represented with erosional efficiency that varies through time, assuming shifts of erosional efficiency that correspond to the onset of regional climate changes (Materials and Methods). (B) Regional paleoclimate proxies from the central and southern Negev area (Discussion). Horizontal lines with short vertical lines at their edges represent the extent of time periods explored by each study. Top to bottom: increased groundwater recharge in the Negev <38 ka (blue, ref. 52), periods of weathering and sediment production leading to terrace formation and aggradation in Wadi Yael (black, ref. 53), slackwater deposits indicating increased storm intensity and frequency in Wadi Netafim (red, ref. 54), abandonment ages of terraces in Wadi Shehoret (orange, ref. 55, based on refs. 39 and 40) and Wadi Evrona (purple, ref. 34), travertine deposits in the Negev and Arava (cyan, ref. 56), and periods of speleothem growth in the southern and central Negev, constraining the two Negev Humid Periods (green, ref. 57). Wadis Shehoret, Evrona, Netafim, and Yael are within a radius of <30 km from the study area. Values of δ¹⁸O in benthic foraminifera from deep-sea cores over time (58) are shown for reference to global climate (semitransparent curve, note the flipped y axis). This global climate proxy does not fully align with the regional paleoclimate proxies and the coinciding phases of rapid wind gap migration. Vertical dashed lines represent boundaries between even and odd Marine Isotope Stages (MIS), associated with global glacial and interglacial periods, respectively (59).

Although the climate in the southern Negev has been generally hyperarid for at least ~1.5 Ma (60), several observations suggest that this period was interspersed with phases of reduced aridity (61). Specifically, within a radius of < 30 km from the study area, several local paleoclimate proxies point to a short-term climatic change between ~35 and 20 ka, though not necessarily throughout this entire period (Fig. 6B). This climate change was suggested to include an increased frequency of extreme storms and floods, leading to intensified weathering and sediment production (34, 40, 53–55), concurrent with increased groundwater recharge across the Negev, dated to <38 ka (52). These observations align with the most recent phase of wind gap migration in the study area, represented by the abandonment ages of the wind gap group at the main site, MYN (constraining the last migration phase to be between 36 and 26 ka), and the wind gap ages at the complementary sites AYS and GAY (indicating that the last migration phase occurred before 23 ± 1 ka and 18 ± 1 ka, respectively). This late phase of Pleistocene wind gap migration is, to date, the youngest documented planform drainage reorganization in the southern Negev highlands.

The older phase of divide migration in the study area (245 and 183 ka, Fig. 5A) correlates with one of the “Negev humid periods” (NHP) in the central and southern Negev (Fig. 6B), that was dated to 220 to 190 ka (56). Negev humid periods are characterized by the growth of speleothems in the central and southern Negev (57), indicating higher precipitation rates relative to those of the present hyperarid climate. These periods generally correlate with the development of travertines throughout the Arava Valley (56). Unlike the relatively localized and short-lived phase of ~35 to 20 ka, the Negev humid periods were suggested to reflect a climatic change with greater spatial and temporal extent. The last two periods, occurring around 220 to 190 ka (NHP-2) and 142 to 109 ka (NHP-1), generally coincide with interglacial periods (61) and were linked to southern climatic systems of intense precipitation, which occurred when prevailing northwestern climate systems weakened (61). The timing of the Negev humid period 2, between 220 to 190 ka, is within the time frame of the oldest phase of divide migration in this study (245 to 183 ka, including age uncertainties), represented by the terrace ages of the eastern group at the main site (MYN).

In this context, we note the correlation between the luminescence age sampled at terrace R4 (126 ± 7 ka) and the peak of Negev humid period 1, at around ~125 ka (61). This correlation can support the interpretation of the abandonment age of terrace R4 as a phase of rapid divide migration shortly after 126 ± 7 ka (dashed red line in Fig. 5D), similar to the other two terrace groups, rather than the scenario of abandonment as part of continuous slow migration (solid blue line in Fig. 5D).

Overall, the two phases of rapid divide migration (245 to 183 and 36 to 26 ka) inferred at the main site (MYN) correlate with various paleoclimate proxies from both the immediate vicinity of the study area and the broader Negev region. These correlations, including the potential contemporaneity between Negev humid period 1 and the age of terrace R4, and the evidence against tectonic control as the main driver of regional divide migration (SI Appendix, section S8), suggest that climate change is most likely the primary driver of wind gap migration in this study area.

Climate Change Dominates Over Drainage Topology in Accounting for Wind Gap Migration in Numerical Simulations.

While phases of rapid divide migration can be driven by external factors, such as the climate changes discussed above, they may also result from internal basin dynamics associated with the geometry of the main channel and its tributaries. This geometry could play a critical role because when the migrating divide crosses a tributary confluence, this tributary is captured by the reversed channel, increasing the discharge, stream power, and erosion rates of the reversed, expanding basin, potentially accelerating the divide migration (30). To explore whether the observed wind gap migration pattern is consistent with the effect of side tributary capture or if it necessitates external drivers such as climate change, we applied a landscape evolution model that simulates two scenarios of wind gap migration in the main site, MYN (Materials and Methods and SI Appendix, section S9). The first scenario is with uniform climate, where migration is affected only by the drainage geometry, and changes in discharge are primarily influenced by the wind gap crossing tributary confluences, followed by tributary capture (30). The second scenario involves intermittent climate changes, represented by shifts between periods of high and low erosional efficiency.

Fig. 6A shows the distribution of the abandonment ages in the main site (MYN), and the location of the wind gap through time for the two simulated scenarios of divide migration in the main site (MYN): uniform climate (dashed gray curve) and time-dependent climate (solid pink curve). The results show that the uniform climate scenario cannot reproduce the terrace abandonment age pattern in the main site (MYN), suggesting that the drainage topology alone is insufficient to explain the recorded wind gap migration dynamics. Conversely, the time-dependent climate scenario reproduces the observed pattern of abandonment ages.

Mechanistic Links between Climate Change and Divide Migration.

Contrary to previous studies that associated divide migration with spatial gradients in uplift and/or climate across the divide (3, 15, 18), the small scale and the spatially uniform climatic conditions of the field sites explored here (Figs. 1 and 2) suggest that the association we show between divide migration and climate fluctuations likely reflects temporal changes in a spatially uniform climate. Although only a few studies have explored the influence of such changes on the dynamics of divide migration (62–64), we note that geomorphologic theories suggest links between temporal changes in spatially uniform climate and dynamic of divide migration. For example, a commonly used geometric model for divide migration suggests that the migration rate is proportional to the difference in erosion rate across the divide (i.e., refs. 13 and 23). These erosion rates often scale with the discharge of channels (65), and hence with the magnitude and frequency of rainfall events. Therefore, for an initially asymmetric discharge, increasing the discharge (i.e., due to an increase in spatially uniform rainfall), can amplify the difference in erosion rates across the divide, expediting its migration rate. Temporal changes in spatially uniform climate can also affect the dynamics of divide migration through nonlinear relations between erosion and discharge (65, 66), as well as across-divide differences in hydrology, base level, or dominant erosional processes (10, 64, 67, 68).

Changes in spatially uniform climate can also trigger divide migration through their effect on sediment flux and deposition. For example, climatic periods that are characterized by high sediment flux from hillslopes can promote sediment deposition and aggradation at the valley bottom, primarily in the beheaded side of the wind gap where wind gap migration reduces discharge and associated sediment transport capacity of the main channel (5, 45). The resulting aggradation of this main channel elevates the local base level for tributaries and distributaries, which, along with localized deposition of hillslope sediments, increase the likelihood for damming and modifying existing flow pathways. These changes increase the probability of discharge diversion toward the reversed channel. Such discharge diversion can occur via avulsions (30), as exampled by the avulsed tributary near the wind gap that was diverted toward the reversed channel in Fig. 1 C and D. Markers of such avulsions, indicating a discharge diversion, are presently observed near the wind gap in the main site (5). This discharge diversion enhances incision rate in the reversed channel, leading to further wind gap migration, aggradation of the beheaded channel, and so on, generating a migration-aggradation-diversion-incision feedback that can explain the episodes of the rapid divide migration inferred for the study area.

Conclusions

This work leverages a unique field site in the Negev Desert, Israel, to address fundamental knowledge gaps regarding the dynamics of drainage reorganization, primarily whether drainage divide migration is continuous or episodic, and what may drive episodes of rapid migration. Utilizing datable terraces along reversed channels that mark the paleo-locations of a divide, we show concrete evidence for divide motion with an average divide migration rate of ~1,143 ± 86 m my⁻¹ over the past ~230 ka. The spatial pattern of terraces age in the study area reveals that this timeframe encompasses periods of divide stalling or slow migration, interspersed with phases of rapid migration that are more than double the average rate. These phases, occurring at ~36 to 26 ka and ~245 to 183 ka (and possibly another phase ~<126 ka), correlate with various evidence for climate change in the southern Negev during these periods, potentially linking the intermittent nature of divide migration to climate changes. This linkage is further supported by numerical simulations that rule out the possibility that these episodes are caused by changes in discharge due to capture of side tributaries. The high temporal resolution of divide migration dynamics revealed in this study uncovers previously undetected patterns of intermittency over timescales of 10⁴ to 10⁵ y, potentially linking these dynamics to climate change.

Materials and Methods

To determine the abandonment age of the terraces we used relative and absolute dating approaches, relying on the theory of reg soil development in hyperarid areas that is provided in Using Reg Soils for Terrace Dating. The relative dating is based on soil chronosequence, utilizing time-dependent properties of the soil, and the absolute dating is based on luminescence techniques.

Reg Soil Chronosequence.

To use the time-dependent properties of the reg soil for relative dating at the main site (MYN), we first mapped the terrace boundaries using a high-resolution DEM (9 cm px⁻¹), an orthophoto (1.7 cm px⁻¹), and field observations (SI Appendix, section S2 and Table S4). We then located the flattest and least disturbed area on each terrace and dug a soil pit in it. Overall, we excavated nine pits (Fig. 3 and http://www.pnas.org/lookup/doi/10.1073/pnas.2408426122#supplementary-materialsSI Appendix, Figs. S1 and S3 and Table S1): five east of the wind gap on terraces along the reversed channel (R1 to R5), two on the wind gap (WG1 and WG2), and two west of the wind gap on fan surfaces along the beheaded channel (B1 and B2) that were sampled as control profiles. At the complementary sites (GAY and AYS), we dug a single pit at the center of each wind gap (Fig. 2 C–E). The pits were excavated to a depth of 35 to 80 cm and reached into the gravel-rich horizon (C). An exception is the pit at the wind gap of the complementary site GAY, where manual digging equipment was insufficient to reach beyond the gravel-poor (B) horizon (SI Appendix, Table S1). The soil profiles were described and sampled for analyses (69). The following properties were measured: i) Percentage of the surface area covered by clasts and the clasts size, based on field visual estimates, ii) Texture and color of the soil horizons (SI Appendix, section S10), and iii) Mass concentrations of gypsum and halite in all profile horizons (SI Appendix, section S11).

Luminescence Dating of Reg Soils.

To constrain the absolute abandonment age of the terraces, we used luminescence dating, which is based on measuring signals that accumulate in the mineral lattice (typically quartz or feldspar) due to natural radioactivity following bleach (reset) after sunlight exposure (70). The dating targets were grains dispersed few cm below the boundary of the gravel-poor horizon (B) and the gravel-rich horizon (C), whose last exposure to sunlight is proposed to correspond closely to the latest fluvial/colluvial deposition and the terrace abandonment (34, 39, 40). The samples were collected in dark conditions with a hammer and a chisel, sometimes with the assistance of a cordless drill. In some cases, due to challenging field conditions or unclear horizon boundaries, the sample was taken from deeper than the target horizon (<10 cm). All samples were dated with quartz grains using the optically stimulated luminescence (OSL) signal, which is widely used for dating terrace abandonment ages (34, 38–40, 53, 55, 60). As the OSL signal is limited in its dating range, samples with equivalent dose (De) >120 Gy were defined as samples with saturated OSL signal, giving minimum ages. These samples were also dated with K-feldspar, using the post-infrared infrared stimulated luminescence signal (47) measured at 250 °C (pIR-IR₂₅₀), which extends the time range of luminescence dating (48). SI Appendix, section S12 and Table S5http://www.pnas.org/lookup/doi/10.1073/pnas.2408426122#supplementary-materials provide information about these signal properties, and describe the sample preparation and measurement processes, which followed Faershtein et al. ((48).

To compare the OSL and pIR-IR₂₅₀ methods and to estimate how well the pIR-IR₂₅₀ signal is bleached under natural environmental conditions in our sites, we also applied the pIR-IR₂₅₀ dating to a young sample with a reliable quartz OSL age (the lowest sample of profile B1), whose properties indicate it is well bleached (Table 1 and SI Appendix, Fig. S7). To confirm that the age of the soil horizons increases down-section and to identify potential unconformities in the profile, in most profiles we also sampled above the gravel-rich horizon (C), resulting in a total of 1 to 3 samples per profile (Table 1 and SI Appendix, Fig. S3).

Divide Migration Numerical Simulations.

A one-dimensional landscape evolution model was applied to explore whether the observed wind gap migration pattern in the main site (MYN) could stem from climate fluctuations or reflect discharge changes resulting from capture of side tributaries as a wind gap migrates along the reversed valley (30). A full description of the model setting, parameters, and sensitivity analysis is provided in SI Appendix, section S9. Briefly, the model solves for the temporal changes in erosion rate and elevation across the wind gap, and the consequent wind gap migration, combining detachment-limited fluvial incision and linear diffusive sediment transport. The model’s initial conditions are based on the lithologic, topographic, and drainage network topology of the main site (MYN), including the distribution of drainage area along the model domain (i.e., ref. 30). Simulations start with a wind gap situated just east of the easternmost terrace R1 (dated to 227 ± 17 ka), approximating its location before the oldest phase of rapid divide migration (SI Appendix, Fig. S9) The impact of changes in climate on phases of wind gap migration is explored by varying a model parameter that describes the climate-dependent erosional efficiency (65).

Two scenarios of uniform and changing climate were explored, both calibrated (by adjusting the erosional efficiency parameter) such that the wind gap approaches its current location after 220 ky. The first scenario of uniform climate assumes a time-invariant erosional efficiency. Here, changes in erosion rate across the wind gap that control its instantaneous migration rate result only from the drainage topology and depend on the position of tributary confluences relative to the migrating wind gap (30). In the second scenario, the erosional efficiency changes through time, reflecting climate variations, and includes shifts between periods of high and low erosional efficiency, whose timing is set by periods of regional climate change (Discussion and Fig. 6). The beginning of the high erosional efficiency periods corresponds to the onset of two most recent the Negev humid periods at 220 to 190 ka (NHP-2) and 142 to 109 ka (NHP-1) (57), and the short-term climatic change period of 35 to 20 ka (53). The duration of the high erosional efficiency periods is adjusted to fit the abandonment ages of each terrace group in the main site MYN.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

DEM dataset have been deposited in DEM of Maale Itro North (MYN) site, Israel (https://doi.org/10.5281/zenodo.10775106) (46).