Cosmogenic nuclides in sediments accumulate throughout the sedimentary cycle as near-surface material is exposed during weathering and exposure of the source rock, transport in a specific drainage system, and to a much lesser degree following burial at some intermediate or final destination. Unlike the more commonly used radioactive cosmogenic nuclides, which may decay substantially or even completely over multiple sedimentary cycles, 21Ne is stable. This means that the concentration of 21Ne measured in sediments may have accumulated over several cycles of exposure and deposition. For example, after sediments reach the depositional basin, they can be re-exhumed and once again exposed and transported in a new sedimentary cycle. Therefore, the concentration of cosmogenic 21Ne measured in sediment represents the total exposure during previous and current sedimentary cycles, unless the sediment is exposed during transport to temperatures exceeding the geological closure temperature of Ne in quartz (90–100 ∘C; Shuster and Farley, 2005). The loss of Ne due to diffusion could occur either during burial at depths of ∼2–3 km given a geothermal gradient of 30–50 ∘C km-1 or if rock reaches high enough temperatures for an extended time, which has been recorded in hot desert environments (e.g., McFadden et al., 2005).

We collected and analyzed 10 samples in total, 8 Hazeva formation samples, and 2 in situ Jordanian cherts. The Hazeva samples include three samples of quartz sand (MHS1, MHS3, and MHS5), and five individual chert pebbles (MHC2, MHC23, MHC5a MHC2b, and MHC6) were obtained from two Miocene Hazeva exposures (Fig. 2b and c; Table 1). At both sites, samples were collected from deeply shielded locations to minimize the effects of post-burial production (see Sect. 5.1 for further discussion). The quartz sand and the chert pebbles were both transported by the Miocene Hazeva system and share a similar exposure history. However, the quartz sand was exposed in previous sedimentary cycles throughout the Mesozoic and Paleozoic, where it accumulated cosmogenic 21Ne. In contrast, the chert was deposited in the Eocene and then exposed, transported, and buried during the Miocene (Avni et al., 2012). Therefore, while the cosmogenic 21Ne measured in the quartz sand represents multiple sedimentary cycles, the cosmogenic 21Ne measured in the chert pebbles represents erosion and transport during a single sedimentary cycle in the Miocene Hazeva River. Additionally, two individual samples of in situ chert nodules (EJC3 and EJC5) were collected from exposed bedrock outcrops of the Eocene source rock in central Jordan (Fig. 2a). Unlike the Miocene samples, which were exposed during at least one full sedimentary cycle, the Jordanian chert nodules accumulated cosmogenic nuclides only during exhumation to the currently exposed surface. Therefore, the cosmogenic nuclide concentrations measured in the Jordanian cherts represent averaged rates of erosion over the last ∼105 yr.

Chert pebbles (ranging 4–14 cm, b axis) were crushed, and both chert and sand samples were sieved to 250–850 µm. Chert and quartz samples were processed to separate clean SiO2 at The Fredy & Nadine Herrmann Institute of Earth Sciences cosmogenic isotope laboratory, The Hebrew University of Jerusalem, following standard procedures (Hetzel et al., 2002; Kohl and Nishiizumi, 1992). The samples were first leached in a HCl/HNO3 mixture (3:1) at a temperature of 150 ∘C for 1.5 h dissolving carbonates and iron oxides. This procedure was followed by Franz magnetic separation to remove magnetic grains, including quartz grains that contain inclusions of magnetic material. Samples were then leached three times in a 1 % HF/HNO3 mixture for 7, 12, and 24 h at 70 ∘C, removing the outer rims of the quartz grains. Aliquots of all 10 etched samples were then analyzed for Ne isotopes at the Berkeley Geochronology Center. Chert samples were washed with isopropanol to remove fine chert particles attached to the chert grains. Aliquots from samples MCH5A and EJC5 were crushed to compare the degassing results with the uncrushed aliquots. Ca. 70 mg from the chert samples and ca. 150 mg from the quartz samples were encapsulated in a tantalum packet and heated under vacuum using a diode laser micro-furnace at 2–4 heating steps between 450 and 1250 ∘C for 15 min at each temperature step. Ne isotope measurements used the BGC “Ohio” system and the procedure described in Balco et al. (2019). Amounts of 20–30 g of leached and clean quartz from three quartz samples and three chert samples were processed to separate Be and Al oxides following Kohl and Nishiizumi (1992) and Bierman and Caffee (2001). These were then analyzed for 10Be/9Be and 26Al/27Al at the Center for Accelerator Mass Spectrometry, Lawrence Livermore National Laboratory, and calibrated against house standards and blanks.

Exposure and burial times as well as erosion rates were calculated based on Balco (2007) and scaled using time-independent scaling (Stone, 2000) and production mechanisms based on Balco et al. (2008), giving sea-level high-latitude production rates of 4.96 atoms per gram SiO2 per year (denoted as atoms (g SiO2)-1 yr-1 hereafter) for 10Be, 30.6 atoms (g SiO2)-1 yr-1 for 26Al (Balco et al. (2008), and 18.1 atoms (g SiO2)-1 yr-1 (Borchers et al., 2016; Luna et al., 2018).

For the chert samples, <2 % of the total 21Ne and no more than 1 % of the total 20Ne measured were released above 950 ∘C (see Tables S1–S4 in the Supplement). Therefore, subsequent analyses were performed at 450, 700, and 950 ∘C heating steps for chert samples and 950 and 1250 ∘C heating steps for quartz samples (Table 1). Of the total 21Ne measured, >85 % was released at the low-temperature steps, below the 950 ∘C step in the chert samples, and below the 1250 ∘C step in the quartz samples (see Tables S1–S4). Also, low-temperature 21Ne/20Ne and 22Ne/20Ne ratios fall on the spallation line, within analytical uncertainty. Therefore, we conclude that excess 21Ne relative to an atmospheric isotopic 21Ne/20Ne ratio of 0.002959 (21Neex=21Ne/20Nemeasured-21Ne/20Neair) in the low-temperature steps is a good representation for cosmogenic 21Ne (21Necos; see Figs. S8–S12). While most samples show some increase in the low-temperature 21Neex, sample MHC2 shows no enrichment in 21Ne/20Ne ratio and very little enrichment in 22Ne/20Ne ratio compared to atmospheric composition in the low-temperature steps. In the 950 ∘C step, there is enrichment compared to atmospheric values. However, as only ∼12 % of the total 21Ne was released in the 950 ∘C step, determining the concentration of cosmogenic 21Ne in sample MHC2 is beyond analytical abilities. Therefore, this sample was not considered in further calculations, discussion, or interpretations. It is important to note that even with cosmogenic isotopic values of 21Ne/20Ne and 22Ne/20Ne ratios at the low-temperature steps, distinguishing the cosmogenic component of 21Neex from the nucleogenic component, produced by the decay of U and Th within the crystal lattice, is not trivial. Nonetheless, as all chert samples (Eocene chert nodules and Miocene chert pebbles) share the same lithology, any differences in the 21Neex concentrations must be due to the cosmogenic component.

The chert pebbles and quartz sands sampled at both Miocene Hazeva sites show variable concentrations of 21Necos ranging between 0.00±1.88×106 and 8.89±1.83×106 atoms (g SiO2)-1 (Fig. 3). At both Miocene Hazeva sites, the cosmogenic 21Ne concentrations measured in chert pebbles are similar or lower compared to sand samples. These measured concentrations agree with our understanding that the sand samples contain quartz grains that originated from various sandy units that were deposited throughout the Phanerozoic and could have undergone several sedimentary cycles before they were exhumed and transported by the Miocene fluvial system. The sand samples could also have higher concentrations of nucleogenic 21Ne as the source rock for this sand is >800 Ma (Kolodner et al., 2009). Conversely, the Hazeva chert samples are derived from a relatively young Eocene source rock and were exposed during one sedimentary cycle in the Miocene. Both samples of Jordanian chert nodules collected from in situ Eocene outcrops show similar cosmogenic 21Ne concentrations, higher compared to the Miocene Hazeva chert pebbles (Fig. 3).

Diffusion kinetics of Ne in quartz have been examined experimentally and theoretically (Shuster and Farley, 2005; Tremblay et al., 2014), but they have yet to be tested on chert samples, where the diffusion length-scale is not straightforward. While diffusion kinetics in chert are likely to be similar to quartz, more work is needed to determine that with certainty. Nevertheless, diffusion is not likely to have been significant over a ∼20 Myr time span in the measured Miocene chert samples. While temperatures in exposed cherts in the Levant region can reach 60–70 ∘C during midday in the summertime due to solar heating, it is unlikely that samples that were transported fluvially were exposed continuously at the surface. The examined chert samples did not exhibit any visible cracking or fractures commonly identified with thermal stresses, leading us to believe that temperatures were not high enough to cause significant diffusion of Ne out of the chert samples.

10Be and 26Al concentrations were measured in three Miocene sand samples (MHS1, MHS3, and MHS5), the two Eocene chert nodules (EJC3 and EJC5), and two chert pebbles (MHC5b and MHC6). 10Be results for sample MHC5b and 26Al results for sample MHS1 are not available (Table 1). Miocene sand and chert samples show 10Be and 26Al concentrations that are low and consistent with extended periods of burial (≤0.39±0.03×105 atoms (g SiO2)-1 for 10Be and ≤4.33±0.55×105 atoms (g SiO2)-1 for 26Al). Currently eroding Eocene nodules show higher concentrations of 10Be and 26Al, with sample EJC3 showing a 26Al/10Be ratio that is consistent with production at the surface (6.75; Balco et al., 2008), and sample EJC5 showing a lower 26Al/10Be ratio, suggesting a more complicated exposure history (see Discussion section).

When examining concentrations of cosmogenic nuclides in sediments that have been buried for extended periods, post-burial production needs to be considered. At or near the surface, spallation interactions are the main pathway for in situ production of cosmogenic nuclides, accounting for >95 % of 26Al, 10Be, and 21Ne (Dunai, 2010). However, the relative contribution of production by muon interactions increases with burial depth. While production rates are relatively low, they can be significant when integrated over long periods, especially for stable nuclides. The post-burial component does not represent surface processes, and therefore it is crucial to account for its contribution to the measured cosmogenic component. For radioactive cosmogenic nuclides, such as 10Be and 26Al, their initial concentrations (acquired during exposure) decrease post burial due to radioactive decay, with 26Al decreasing faster than 10Be according to their corresponding half-lives (e.g., Balco and Rovey, 2008; Granger, 2006; Granger and Muzikar, 2001; Lal, 1991).

We calculated the expected concentrations of cosmogenic 26Al, 10Be, and 21Ne in sediments over a burial period of 18 Myr, the likely age of the fluvial system stabilization (Bar and Zilberman, 2016). We then compared these calculated concentrations to the measured concentrations of 26Al, 10Be, and 21Necos in Miocene chert and sand samples (Fig. 4). Both 10Be and 26Al measurements are only available for two buried sand samples, one buried chert pebble, and two in situ chert nodules (Table 1). The measured 10Be and 26Al concentrations have reached an equilibrium that is consistent with an extended period of burial at depths between 20 and 120 m (given that overburden consists of clastic sediments with a density of ∼2 g cm-3). The discrepancy between the current burial depth, only tens of meters below the surface, and the deduced burial depth is likely the result of surface erosion that occurred during the last ∼2 Myr (Matmon and Zilberman, 2017, and references therein). Additionally, the relatively large uncertainty on muogenic production rates could account for some of this discrepancy (Balco, 2017; Balco et al., 2019). Our calculations show that the cosmogenic 21Ne produced post burial over 18 Myr at depths between 20 and 120 m is lower than the 21Neex measured in the presented samples (including their uncertainties). The maximal calculated post-burial cosmogenic 21Ne concentration accounts for ∼1.3×106 atoms (g SiO2)-1, which is lower than the analytical uncertainty for all measured Miocene samples except for MHC2, where no cosmogenic 21Ne was measured. However, sample MHC2 is not considered in the interpretations of the results. Therefore, we consider post-burial cosmogenic 21Ne production to be insignificant for the presented Miocene exposure times.

Exposure times at the surface calculated from cosmogenic 21Ne concentrations measured in in situ chert nodules from the central Jordanian Plateau (EJC3 and EJC5) range between a minimum of 193 kyr and a maximum of 454 kyr (correlating to cosmogenic 21Ne concentrations of 8.08±1.48×106 and 12.10±2.43×106 atoms (g SiO2)-1).

In comparison to the Jordanian samples, quantifying exposure times during the Miocene using cosmogenic 21Ne concentrations is not trivial, most notably due to the challenge in evaluating the local cosmogenic production rates. The production rate of cosmogenic nuclides increases with altitude as the air pressure and shielding effect of the atmosphere decreases (Stone, 2000). While the latitude of the Arabian Peninsula during the early Miocene was similar to today (Meulenkamp and Sissingh, 2003, and references therein), accounting for the elevation of the Miocene samples during the production of cosmogenic 21Ne raises two difficulties. First, it is not possible to determine with certainty the elevation of the central Jordanian Plateau during the Miocene. It is clear that from the Late Cretaceous up until the late Eocene, the Arabian Peninsula was mostly submerged below sea level and that during the Oligocene it was uplifted to a sufficient elevation to allow for significant surface erosion (Garfunkel, 1988). During the early Miocene, broad valleys (500–1000 m wide and ∼100 m deep) incised the regional truncation surface that developed in the region, where the Hazeva formation was later deposited (Avni et al., 2012). This timeline of events leads us to believe that significant surface uplift occurred prior to the initiation of the Miocene Hazeva fluvial system at ∼18 Ma. Nevertheless, this stratigraphic evidence is insufficient to determine whether the Arabian Peninsula reached its current elevation during the early-to-mid Miocene or whether additional uplift occurred over the past 20 Myr. Studies that focus on exhumation along the eastern flank of the Dead Sea rift do not provide clear evidence to constrain the timing of surface uplift. Surface uplift histories based on thermodynamic cooling ages (Feinstein et al., 2013), and river profiles (Wilson et al., 2014), conclude that during the last ∼30 Myr the western half of the Arabian Peninsula was uplifted to its current elevation (Feinstein et al., 2013; Wilson et al., 2014). However, in a recent work, Morag et al. (2019) present thermochronologic constraints using apatite (U-Th)/He and fission-track data from a transect across the eastern flank of the Suez Rift in SW Sinai. The researchers suggest that uplift and exhumation along the western side of the Suez Rift flank slowed substantially post ∼18 Ma. This decline reflects a decrease in uplift, which could indicate that the Jordanian Plateau has reached close to its current elevation (∼1000 m) when the Hazeva River was active. One more approach to evaluate the paleo-elevation of the central Jordanian Plateau is to calculate this elevation given a known distance between the source point and the base level and an evaluated slope. The Hazeva fluvial system drained westward to the Mediterranean at an elevation of ∼0 m a.s.l., and over a distance of ∼200 km from the Mediterranean coast to the location of exposed chert nodules. Given a moderate stream gradient of ∼0.5 %, the elevation of the central Jordanian Plateau is ∼1 km a.s.l. Given the different types of evidence reported, it is reasonable to presume that the western flank of the Arabian Peninsula reached its current elevation (∼1 km) during the early-to-mid Miocene. The use of a single elevation to calculate paleo-production rates introduces a second difficulty, as it does not account for spatial variations in elevation due to catchment topography. Without any tangible information about the size and steepness of the catchment area of the Hazeva River, we are unable to correct for different elevations and production rates throughout the basin. These uncertainties in paleo-production rates, due to assumptions in catchment paleo-elevation, result in longer calculated exposure times. Accounting for uncertainties described above, we assume an elevation range of 500–1000 m a.s.l. and latitude of 20–30∘ for the calculated Miocene exposure times.

The calculated exposure times of sediments in the Miocene Hazeva fluvial system are variable and range between a minimum of 0-0+59–0-0+86 kyr measured in chert pebble sample MHC5b and a maximum of 278±63–408±63 kyr measured in quartz sand sample MHS5 (Table 2). Comparing the two silicate members, concentrations (and exposure times) of the sand samples are overlapping or higher than the chert samples (Fig. 3). This observation agrees with our understanding that the cosmogenic 21Ne measured in the Miocene chert pebbles represents the total time of exposure during exhumation from bedrock coupled with transport in the Hazeva River. At the same time, the sand samples have undergone previous sedimentary cycles and contain inherited cosmogenic 21Ne. Therefore, sand samples cannot be used to calculate the time that sediments were exposed during transport in the Hazeva fluvial system or to infer erosion rates. Unlike the sand samples that have feasibly undergone previous exhumation, erosion, and deposition, the Miocene chert samples have not undergone previous sedimentary cycles. Hence, all cosmogenic 21Ne measured was produced during erosion and transport in the Hazeva River and rates of surface processes during the Miocene can be evaluated using the Miocene chert samples.

The cosmogenic 21Ne exposure times calculated from the Jordanian chert samples range from 269±63 to 378±76 kyr. Exposure times that were calculated from 10Be and 26Al concentration measured in sample EJC5 overlap within uncertainty with 21Ne calculated exposure values (Table 2). In contrast, exposure times calculated from 10Be and 26Al concentrations measured in sample EJC3 are much shorter ∼13–16 kyr, an order of magnitude difference. While we cannot explain this discrepancy, we believe that the representative results are longer exposure times. Firstly, the 21Ne calculated exposure time in sample EJC3 agrees with the 21Ne, 26Al, and 10Be calculated exposure times for sample EJC5. Secondly, the timescales of exposure times measured in cherts in eroding surfaces at the hyperarid Negev desert are similar and range from ∼2×105 to ∼2×106 yr (Boroda et al., 2014; Fruchter et al., 2011; Matmon et al., 2009). We conclude that exposure times in modern central Jordanian Plateau chert nodules range ∼300–400 kyr. It is important to note that the calculated exposure times in the Jordanian cherts represent only exposure at the surface, and they do not include exposure during transport, in contrast to the Miocene chert pebbles.

When examining ancient exposure times, we must first consider the timescales over which cosmogenic nuclides are averaged. The question arises as to whether the reported exposure times accurately represent the environmental conditions of a certain period (e.g., the early-to-mid Miocene) or if the calculated times are the result of episodic oscillation or catastrophic geomorphic events. For currently exposed in situ samples, the modern exposure times are relatively long, integrating hundreds of thousands of years, over which such oscillations or rare catastrophic events would be averaged. As for the Miocene exposure times, samples were collected from two separate sites and different depths, so it is unlikely that they all represent the exception. We, therefore, consider the range of times obtained from Miocene samples to be a good representation of Miocene surface processes.

The calculated exposure times of the Jordanian chert nodules are equivalent to erosion rates of ∼4–12 mm kyr-1 (Table 2), consistent with other rates measured in the region (Matmon and Zilberman, 2017, and references therein). Calculation of paleo-erosion rates is not as straightforward, as Miocene cherts were sampled post-deposition and represent exposure both during erosion from bedrock and transport in the Hazeva River. However, Miocene exposure times are either shorter or overlap within uncertainty with times of in situ Jordanian chert. Thus, actual bedrock erosion rates during the Miocene must have been faster than the prevailing rates mentioned above.

While we cannot determine how much faster paleo-erosion rates were during the Miocene, any increase in erosion rates in a hyperarid desert must be the consequence of different environmental conditions that prevailed in the region at that time. An increase in rates of erosion is most commonly attributed to perturbations in fluvial basins in response to tectonic uplift and/or warmer or wetter climatic conditions (e.g., DiBiase and Whipple, 2011; Romans et al., 2016; Schaller and Ehlers, 2006; Val et al., 2016; Willenbring et al., 2013). For example, increased precipitation brings about higher river discharge and enhancement of the stream power available for bedrock erosion and sediment transport. Erosion rates in fluvial systems also respond to tectonically induced changes in base level that increase slope steepness and instability, resulting in higher stream power and more sediment readily available for transport. Here we examine evidence from previous studies of the climatic and tectonic conditions that prevailed in the region during the Miocene, capable of forcing the deduced increase in erosion rates.

Many works which quantify the rates and timing of surface uplift related to the rifting of the Red Sea are confined to the edges of the Arabian plate and do not give good constrains for intercontinental uplift (Morag et al., 2019; Omar et al., 1989; Omar and Steckler, 1995). These studies used thermochronometric methods and focused on the uplifted flanks of the Suez Rift along which the Precambrian basement of the Arabian-Nubian Shield is exposed. Constraining uplift of the Arabian Plateau is more challenging as the exposed strata are composed mostly of carbonate rocks, which are not suitable for this type of method. While some studies point to a decrease in exhumation rates during the mid-Miocene (∼18 Myr; Morag et al., 2019), surface uplift and topographic changes could still drive large-scale landscape response, manifesting as increased erosion rates and the establishment of the Hazeva fluvial system.

In addition to tectonic forcing, there is ample evidence for a warmer and wetter climate in the region during the Miocene. Locally, the appearance of mammals in the Negev, along with arboreal and grassy vegetation during the early-to-mid Miocene, supports a humid environment (Goldsmith et al., 1988; Horowitz, 2002; Tchernov et al., 1987). A tropical-to-subtropical climate prevailed in the eastern Arabian Peninsula, as indicated by fossilized mangrove roots (Whybrow and McClure, 1980). Locally, Kolodny et al. (2009), interpreted the 18O in lacustrine limestone from the lower part of the Hazeva unit to be deposited by 18O-depleted paleo-meteoric water. They proposed that the presence of a warm ocean to the southeast of the region during the late Oligocene–early Miocene resulted in tropical cyclones being more prevalent and increasing rainfall in the region.

Together, the above observations suggest climatic conditions that could promote erosion rates that are faster than observed rates in hyperarid conditions and that support the existence of a large and maintained fluvial system, such as the Hazeva River, during the Miocene.