The paleohydrology of ephemeral stream systems is an important constraint on paleoclimatic conditions in arid environments, but remains difficult to measure quantitatively. For example, sedimentary records of the size and extent of pluvial lakes in the Mojave Desert (southwestern USA) have been used as a proxy for Quaternary climate variability. Although the delivery mechanisms of this additional water are still being debated, it is generally agreed that the discharge of the Mojave River, which supplied water for several Pleistocene pluvial lakes along its course, must have been significantly greater during lake highstands. We used the 10Be concentrations of 10 individual quartzite pebbles sourced from the San Bernardino Mountains and collected from a ca. 25 ka strath terrace of the Mojave River near Barstow, California, to test whether pebble ages record the timing of large paleodischarge of the Mojave River. Our exposure ages indicate that periods of discharge large enough to transport pebble-sized sediment occurred at least 4 times over the past ∼240 k.y.; individual pebble ages cluster into 4 groups with exposure ages of 24.82 ± 4.36 ka (n = 3), 55.79 ± 3.67 ka (n = 2), 99.14 ± 12.07 ka (n = 4) and 239.9 ± 52.16 ka (n = 1). These inferred large discharge events occurred during both glacial and interglacial conditions. We demonstrate that bedload materials provide information about the frequency and duration of transport events in river systems. This approach could be further improved with additional measurements of one or more cosmogenic nuclides coupled with models of river discharge and pebble transport.


Arid landscapes shaped by ephemeral stream flow are challenging to characterize and model because of the complex spatial and temporal changes in geomorphic processes and the incomplete records of climate factors that drive those processes. Innovative approaches such as modeling the residence time of sand in ephemeral streams based on luminescence ages (e.g., McGuire and Rhodes, 2015), using various types of tracers to track sand (e.g., Crickmore, 1967; Rathburn and Kennedy, 1978; Milan and Large, 2014) or coarser material (e.g., Church and Hassan, 1992; Lamarre et al., 2005) are improving our ability to characterize these types of geomorphic systems, as are recent studies of climatic drivers (e.g., Enzel et al., 2003, and references therein; Miller et al., 2010; Kirby et al., 2012; Lyle et al., 2012; Antinao and McDonald, 2013). However, placing constraints on the paleodischarge of fluvial systems remains a difficult problem.

Reconstructions of the paleohydrology of arid regions, particularly southwestern North America, are often based on geologic and geomorphic evidence of the size and extent of alluvial fans (e.g., Wells et al., 1987; Bull, 1991; Harvey and Wells, 1994; Harvey et al., 1999) or eolian deposits (e.g., Lancaster and Tchakerian, 2003), the stable isotopic records in speleothems (e.g., Winograd et al., 2006; Wagner et al., 2010), or lacustrine sedimentary records, particularly of pluvial lakes (e.g., Enzel and Wells, 1997; Enzel et al., 1992, 2003; Wells et al., 2003). A pluvial lake is a closed (endorheic) basin that fills with water during wetter climate periods. Observations of pluvial lake sedimentary records in the Mojave Desert, southern California, indicate that the highest lake levels, and therefore wettest climate conditions, were generally contemporaneous with glacial stages (e.g., Reheis and Redwine, 2008; Reheis et al., 2007, 2012).

The timing and surface elevations of pluvial lake stages within the Mojave River basin have been used to estimate the water balance in the Mojave River watershed and the discharge required to sustain one or more pluvial lakes along the Mojave River course (Enzel and Wells, 1997). Although the source of the discharge necessary to produce and maintain pluvial lakes along the Mojave River course remains a topic of considerable debate (e.g., Enzel et al., 2003, and references therein; Miller et al., 2010; Kirby et al., 2012; Lyle et al., 2012; Antinao and McDonald, 2013), estimates of the magnitude and timing of Mojave River discharge are based on spatially and temporally disparate alluvial, lacustrine, and marine sedimentary records. To our knowledge, no attempt has been made to test these hypotheses using a more direct measure of Mojave River paleodischarge.

We present 10 new cosmogenic 10Be exposure ages of individual quartzite pebbles collected from alluvium deposited on a strath terrace of the Mojave River near Barstow, California (Figs. 1 and 2). The age of alluvium on the strath is constrained by new optically stimulated (OSL) and infrared stimulated (IRSL) luminescence ages. We hypothesize that pebble exposure ages record episodic large discharge events in the Mojave River. The geology and geomorphology of the Mojave River watershed, and constraints on these characteristics provided by previous research in the region, provide a nearly ideal experimental setting to test this hypothesis.

1. The quartzite pebbles are sourced from metasedimentary and sedimentary rocks exposed primarily in the upper elevations of the San Bernardino Mountains (Fig. 1) (Dibblee, 1973; Sadler and Reeder, 1983; Morton and Miller, 2006). These rocks are dominantly conglomerates, and so the quartzite pebbles that are eroded out of hillslopes are presorted and rounded, minimizing the possible effects on 10Be concentrations of grain comminution or sorting during transport (e.g., Belmont et al., 2007; Carretier and Regard, 2011; Carretier et al., 2015, and references therein). The only other source of quartzite pebbles in the Mojave River watershed upstream of our sampling location is from Mojave River alluvium, either in the channel bed or preserved in Mojave River alluvial fill terraces (Dibblee, 1960a, 1960b, 1960c, 1960d, 1965, 1966, 1970, 1973; Dibblee and Bassett, 1966; B. Cox, 2013, personal commun.). Although there are other occurrences of quartzite in the region (Bowen and Ver Planck, 1965; Sadler and Reeder, 1983), their extent inside the Mojave River basin is not significant (Sadler and Reeder, 1983).

2. The discharge of the Mojave River is driven primarily by precipitation in the San Bernardino Mountains. The Mojave River headwaters in the San Bernardino Mountains, which are ∼5%–10% of the ∼9500 km2 Mojave River drainage area, receive >1000 mm/yr, whereas the remainder of the Mojave River drainage area receives between 125 and 150 mm/yr (as measured near Baker, California; Enzel and Wells, 1997; Reheis et al., 2012). Hydrological records and historical accounts indicate that the discharge of the Mojave River is strongly correlated with precipitation in the San Bernardino Mountains; nearly the entire mean annual discharge of ∼9.5 × 106 m3 is derived from that 5%–10% of the Mojave River drainage area in the San Bernardino Mountains (Enzel et al., 1992, 2003; Wells et al., 2003). Currently much of this discharge is lost to infiltration along the river bed, with only extreme floods reaching Afton Canyon, ∼65 km east of Barstow. However, late Pleistocene and Holocene sedimentary records in Mojave Desert pluvial lake basins indicate that the Mojave River has previously carried enough discharge for it to have episodically flowed as far as Death Valley during the Pleistocene (Wells et al., 2003).

3. The age of alluvium deposited on the strath is well known, which allows us to precisely and accurately account for postdepositional accumulation of 10Be, and isolate that fraction of the measured concentration that records the pebbles exposure time in the hillslope and channel network. In addition to our new luminescence ages, the sedimentary record of pluvial Lake Manix, the deposits of which the strath at our sampling site are cut into, has been examined for nearly a century (e.g., Buwalda, 1914; Blackwelder and Ellsworth, 1936; Jefferson, 1985, 2003; Steinmetz, 1988; Meek, 1990, Reheis and Redwine, 2008; Reheis et al., 2012). It indicates that the Mojave River entered the Cady subbasin of Lake Manix beginning ca. 500 ka (Cox et al., 2003), alternately terminating in the Cady, Troy, and Coyote subbasins until ca. 190 ka, when the lake expanded into the Afton subbasin (Reheis et al., 2007). There were three recognized distinct lacustrine phases between 190 ka and ca. 25 ka, when Afton Canyon was incised and Lake Manix drained into the Soda and Silver Lake basins to the east (Reheis and Redwine, 2008; Reheis et al., 2012), extending the Mojave River across the former floor of Lake Manix. Although Mojave River delta sediments extended far into Lake Manix after ca. 40 ka (Reheis et al., 2012), cut-and-fill fluvial deposits of the Mojave River inset into the delta did not reach beyond the upper delta until the lake failed and drained. Those deposits occur at a lower elevation than our sample location.

Because of these geologic constraints, pebble exposure ages should be representative of the total amount of time that each pebble has spent in the Mojave River hillslope and channel network, providing new insight into when, and how frequently, the Mojave River carried enough discharge to transport pebble-sized sediment. Our results show that (1) the Mojave River has undergone at least three periods of discharge large enough to transport pebble-sized sediment over the past ∼240 k.y.; (2) those periods of high Mojave River discharge do not correspond to glacial periods as defined by global temperature proxies; and (3) transport time from the Mojave River headwaters in the San Bernardino Mountains to the sample location, ∼160 km downstream, can be very short, likely <∼2 k.y.


Quartzite pebbles were collected from alluvium mantling a strath terrace of the Mojave River ∼35 km downstream of Barstow (34.952°N, 116.558°W; Figs. 1 and 2). The pebbles form a well-developed, moderately varnished pavement (Figs. 2B, 2C) supported by an ∼6-cm-thick vesicular A horizon overlying a weak to moderately developed cambic B horizon developed in fluvial pebbly sand. This soil formed in an ∼1-m-thick deposit inset ∼9 m into and overlying Lake Manix deposits. The deposit on the strath gently slopes north toward the Mojave River channel, and is cut on the north side by a 6 m lower strath with similar soils.



Luminescence from quartz and feldspar separates was used to provide age control on the alluvium deposited on the strath terrace from which pebbles were sampled by measuring OSL on quartz and IRSL on potassium feldspars in the same sample. The site (Fig. 2A) is ∼523 m above sea level, 20 m below a nearby highstand beach and inset more than 9 m into Lake Manix sediments. The alluvium deposited on the strath is 1.2 m thick, with soils developed in the upper 40 cm. The basal section, 80 cm thick, is composed of coarse arkosic sand with rounded pebbles of quartzite and angular clasts of local metavolcanic rock. It is concordantly overlain by medium to coarse arkosic sand with rounded quartzite pebbles. We sampled for luminescence near the top of the lower unit, below the base of soil development (depth of 45 cm). Samples were collected and processed following Mahan et al. (2007). Cosmic ray dose rate was estimated according to Prescott and Hutton (1994). Detailed methods for measuring the equivalent dose using single aliquot regeneration were presented in Mahan et al. (2007).

Cosmogenic 10Be

Pebbles (mean intermediate axis = 5.15 cm; Supplemental Table 11) were collected from the surface pavement at site MRT2 (Fig. 2). All had surface desert varnish and were on a vesicular A horizon, indicating long-term postdepositional aggradation of eolian dust and a near-surface origin (Wells et al., 1995; McFadden et al., 1987). Samples were prepared following standard methods (Supplemental File2). The 10Be/9Be ratios were measured at the Center for Accelerator Mass Spectrometry, Lawrence Livermore National Laboratory (Livermore, California), and normalized to the 07KNSTD standard series (Nishiizumi et al., 2007). Pebble exposure ages were calculated using the CRONUS online exposure age calculator (Balco et al., 2008) (Supplemental Table 23).


Luminescence Chronology

The quartz OSL age of the sand deposited on the strath surface is 25.4 ± 1.42 ka. The feldspar IRSL age is 24.3 ± 2.18 ka. The uncertainties are reported at 2σ (Table 1).

Pebble 10Be Exposure Ages

Individual pebbles have measured 10Be concentrations (Nmeas) between 1.62 ± 0.16 × 105 and 17.81 ± 1.92 × 105 atoms/g quartz at 2σ confidence (Table 2). These concentrations correspond to exposure ages between 20.71 ± 4.25 and 239.9 ± 52.16 ka (Table 2; Fig. 3A). Uncertainties on pebble ages are the 2σ external uncertainties related to reference 10Be production rate (Balco et al., 2008). A normal kernel density estimate (NKDE) of the 10 pebble ages (Fig. 3A) indicates three distinct multipebble clusters of ages. These clusters have mean ages and standard deviations of 24.82 ± 4.36 ka (n = 3), 55.79 ± 3.67 ka (n = 2), and 99.14 ± 12.07 ka (n = 4). There is a single pebble that defines the oldest limit of our data, 239.9 ± 52.16 ka.


The strath and overlying alluvial deposit that we sampled are inset into lacustrine sediments of Lake Manix, and represents the first known deposit of the Mojave River to this site. The quartzite pebbles in the alluvium overlying the strath have been transported to the sample location either directly from the pebble conglomerate source lithologies in the San Bernardino Mountains (Fig. 1) or have been recycled from Mojave River alluvium adjacent to the Mojave River channel, either from cut-and-fill terraces or from the Mojave River delta into Lake Manix. The luminescence dates of the alluvium deposited on the strath are indistinguishable from the age of the demise of Lake Manix (Fig. 3), suggesting that the river cut this channel in the lake floor in <1 k.y.

We interpret the 10Be data as exposure ages, i.e., the amount of time that the quartzite pebbles have spent near the surface during erosion from pebble conglomerates as prerounded, presorted pebbles from hillslopes in the San Bernardino Mountains, plus the transport time to where they were sampled (including any time in storage), plus the exposure time on the strath. It is very likely that each pebble has a complex exposure and burial history (e.g., Granger, 2006), including very different initial 10Be concentrations (inheritance; Ninh); there are myriad reasonable nuclide buildup and decay scenarios for each of those histories. However, because there is no unique 10Be production and decay path for any given measured 10Be concentration, we use the simple exposure ages.

Pebble 10Be Exposure Ages and the Problem of Inheritance

Our interpretation of the pebble 10Be concentrations as a proxy for time spent in transport and storage along the Mojave River assumes that the concentration of 10Be that would have accumulated in the pebbles during exposure and transport on hillslopes in the San Bernardino Mountains prior to them being shed into the Mojave River network (inheritance) is relatively small. Although this assumption may be valid, it is also possible that the 10Be inheritance could be both a significant component of our measured 10Be concentrations and highly variable from pebble to pebble. Whereas recent measurements of 10Be concentrations in modern fluvial sand collected from small San Bernardino Mountains watersheds occupy a narrow range between ∼10 × 104 and 20 × 104 atoms/g quartz (Binnie et al., 2007), low-temperature thermochronometric data from the San Bernardino Mountains indicate very low exhumation rates of a high-elevation, low-relief surface (<20 m/m.y.), and it is possible that quartzite pebbles, which are weathering out of Miocene sedimentary rocks, could persist on low-gradient hillslopes for long and highly variable amounts of time (Spotila et al., 1998, 2002), resulting in very high and variable 10Be inheritance in our pebbles.

We assess the robustness of our interpretation of pebble ages, namely that the different aged groups of pebbles represent distinct periods of time that pebbles could be transported by the Mojave River, by using some simplifying assumptions and the results of previous work to explore some possible values of 10Be inheritance in our pebbles.

We calculate inheritance, Ninh, in our youngest set of three pebbles using 
where Nmeas is the measured concentration of 10Be, Nterr is the concentration of 10Be accumulated during 24.4 ± 2.4 k.y. of exposure on the terrace (the mean and 2σ variance of the OSL and IRSL ages), and Ntrans is the concentration of 10Be accumulated during transport in the hillslope and fluvial network. We assume that Ntrans accumulated for a minimum period of time between 1 yr and 10.3 k.y., the least and greatest differences between the 2σ confidence intervals of the mean luminescence age of the terrace and the youngest pebble 10Be age. For Nterr we assume constant production rates of 5.97 and 0.215 10Be atoms/g quartz/yr by neutron spallation and muogenic processes, respectively. These production rates were calculated using the CRONUS calculator for the sampling location, at an elevation of 521 m and a material density of 2.44 g/cm3. For Ntrans we assume catchment mean production rates of 7.64 and 0.239 10Be atoms/g quartz/yr by neutron spallation and muogenic processes, respectively, also calculated using the CRONUS calculator, considering a catchment average elevation of 841 m. We do not account for either topographic or snow shielding. Both of these would decrease the 10Be production rate, resulting in higher modeled concentrations of Ninh. We also do not account for burial during transport, which would increase Ntrans and decrease modeled concentrations of Ninh.

Using these assumptions we calculate Nterr = 1.497 ± 0.146 × 105 atoms/g quartz and Ntrans between 8 ± 1 and 8.08 ± 1.04 × 104 atoms/g quartz. The resulting modeled values of Ninh from Equation 1 are between 0.63 ± 0.63 × 104 atoms/g quartz and 4.01 ± 1.015 × 104 atoms/g quartz. These are the middle of the range between the minimum and maximum modeled Ninh of each pebble, corresponding to the maximum and minimum assumed transport times, respectively. We compare these model inherited 10Be concentrations in our pebbles to three different end-member model sources of inherited 10Be.

1. We use the renormalized 10Be concentrations of sand-sized sediments from high-elevation, low-relief basins in the San Bernardino Mountains (Supplemental Table 2 [see footnote 3]; Binnie et al., 2007), where the pebble source areas are found, to account for inheritance. One could argue that sand is delivered to the channel network by different processes that operate at different rates from those that deliver pebbles, such as landslides and other processes that deliver material from greater depths beneath the surface (e.g., Niemi et al., 2005; Yanites et al., 2009; Puchol et al., 2014). Measurements of 10Be nuclide concentrations in different size fractions indicate that there is no consistent trend in concentration as a function of grain size (cf. Carretier et al., 2015). Previous work (Olivetti et al., 2012) indicated that coarser size fractions have up to two times lower 10Be concentrations, so we halve Binnie et al.’s (2007) nuclide concentrations to compare possible inheritance related to faster erosion rates to our model Ninh.

2. We use the long-term exhumation rates derived from low-temperature thermochronometric data of Spotila et al. (1998, 2002) to calculate surface 10Be concentrations in the low-relief watersheds of the upper San Bernardino Mountains and compare those to our modeled Ninh values. Even if pebbles form a long-lived lag on the surface, it is doubtful that that lag would persist at time scales longer than this exhumation rate, and so this will place a maximum bound on model Ninh values.

3. We calculate equivalent surface exposure ages of low-relief catchments in the upper San Bernardino Mountains and model the expected 10Be depth profile for that part of the landscape. We can transform the expected 10Be nuclide concentration at any given depth into the apparent age of a pebble from that depth to assess the possibility that all of our pebbles were derived from San Bernardino Mountains hillslopes at the same time but from different depths (e.g., a deep-seated landslide).

Modeled Pebble 10Be Inheritance Compared to Recent Catchment-Averaged 10Be Concentrations

We assume that any 10Be not accumulated during 24.4 ± 2.4 k.y. of exposure on the strath is due to nuclide inheritance and compare those concentrations to the renormalized 10Be concentrations of Binnie et al. (2007) (Supplemental Table 2 [see footnote 3]; Fig. 4). In general, the pebbles have greater inheritance than the measured 10Be concentrations of fluvial sands. Greater model 10Be inheritance of our pebbles can be interpreted in two different ways. The first is that the pebbles spent more variable, and longer, periods of time on hillslopes in the San Bernardino Mountains, where they accumulated larger concentrations of 10Be (∼4–8×) than sand collected by Binnie et al. (2007). This interpretation implies that pebble transport rates in the high-elevation, low-relief pebble source areas are low compared to sand, and that some pebbles may persist in this lag for much longer periods of time than others (e.g., Spotila et al., 1998, 2002). In this scenario the pebbles would have all been delivered to the Mojave River drainage network and deposited on the strath at about the same time, but had highly variable inheritance.

However, our modeled pebble Ninh values are also consistent with the interpretation that the four distinct sets of pebbles had similar inherited 10Be concentrations when they entered the Mojave River drainage network and spent different amounts of time in transport to where we sampled them. If we consider the youngest set of pebbles, modeled Ninh values are similar to half of the 10Be concentrations measured by Binnie et al. (2007) (Fig. 4). This indicates that the initial exposure duration of pebbles on high-elevation, low-relief hillslopes of the San Bernardino Mountains, and therefore hillslope denudation rates at the time that pebbles were delivered to the Mojave River network, were similar. This similarity indicates slightly faster hillslope denudation rates; recent denudation rates in the headwaters of the Mojave River, calculated using the Binnie et al. (2007) renormalized 10Be concentrations, have a mean of 96 ± 20 m/m.y., whereas denudation rates calculated using modeled Ninh values of the youngest set of Mojave River pebbles have a mean of 31 ± 14 m/m.y. This threefold difference could be attributed to greater water availability driving more rapid hillslope transport, or different transport processes, during the times that pebbles were shed into the Mojave River network. In this scenario, pebbles with similar ages would have similar inheritance and the difference in exposure ages between the sets would be the result of different durations of transport between the San Bernardino Mountains source area and where the pebbles were deposited on the strath.

Modeled Pebble 10Be Inheritance Compared to 10Be Concentrations Expected from Long-Term Exhumation Rates

Spotila et al. (1998, 2002) used low-temperature thermochronometric data from the San Bernardino Mountains to infer long-term (∼2.5 m.y.) exhumation rates of the mountain range; they inferred an average exhumation rate of 40 m/m.y. for the entire San Bernardino Mountains, but assumed that the exhumation rate of the high-elevation, low-relief headwaters of the Mojave River is lower, <20 m/m.y., based on the presence of Miocene-aged sedimentary cover, including the quartzite pebble conglomerates from which our pebbles are derived. Given this estimate of long-term exhumation, and using the largest modeled Ninh value of 16.155 ± 0.84 × 105 atoms/g produced in the upper 10 m of the surface, assuming a density of 2.44 g/cm3, an elevation of 2050 m, at rates of 16.73 and 0.343 atoms/g by spallation and muons, respectively, the sample is at secular equilibrium with respect to 10Be. This could be interpreted as one pebble with a very long exposure time (≥4 m.y.) in the quartzite cobble lag being mixed with a suite of much younger pebbles that were transported to and deposited on the strath at the same time. We, however, infer from this result that the modeled Ninh of the oldest individual pebble is not representative of the true 10Be inheritance. Although we cannot rule out values of Ninh in the older pebbles that are higher than the modeled Ninh of the youngest group, that would imply highly variable processes and process rates continuously delivering pebbles to the Mojave River network through time, it is just as likely that pebbles were delivered by similar processes acting at similar rates at distinct times. In this scenario, pebbles enter the Mojave River network with similar Ninh, but in discrete pulses, possibly driven by variability in regional climate that results in greater water availability to transport coarse sediment.

Modeled Pebble 10Be Inheritance Compared to 10Be Concentrations from an Expected Depth Profile

The combination of the long-term erosion rate estimated from low-temperature thermochronometry and our largest modeled value of Ninh indicates that the 10Be concentration of the apparent oldest pebble in our data set has reached secular equilibrium; this is something that has been documented only in the most arid landscapes (Klein et al., 1986; Brook et al., 1995; Ivy-Ochs et al., 1995; Bierman and Caffee, 2001; Nishiizumi et al., 2005). Given that recent denudation rates are significantly higher than the longer term thermochronologic estimate (Binnie et al., 2007; Spotila et al., 1998, 2002) (Supplemental Table 2 [see footnote 3]), and that part of the San Bernardino Mountains, though not the pebble source areas, were glaciated at least three times over the Quaternary (Owen et al., 2003), this situation seems unlikely. However, there is another possible way to explain the distribution of our modeled Ninh values.

Instead of pebbles with highly variable Ninh being sourced from the surface at the different times, they may have been sourced at the same time, but from different depths. Mass wasting processes (e.g., landslides, debris flows) can supply material from depth that would be relatively coarse, like our pebbles. This coarse sediment would have a lower 10Be concentration than material at the surface. The difference in depth-dependent concentrations of 10Be could be interpreted as differences in exposure age.

To explore this possibility, we calculated the 10Be concentration over a 10 m depth profile beneath a 1 Ma surface lowering at 20 m/m.y. using the same assumptions we used when considering long-term exhumation rates based on low-temperature thermochronometry. We selected 1 Ma as a seemingly reasonable compromise between the ca. 4 Ma age required for secular equilibrium (e.g., Nishiizumi et al., 2005; Granger, 2006) and the Holocene time scale represented by catchment-averaged erosion rates in the headwaters region of the Mojave River where the pebbles are sourced (Fig. 1) (Binnie et al., 2007). We also do not consider this scenario outside of the low relief, high-elevation area of the San Bernardino Mountains where the Mojave River headwaters are located, as it is this part of the landscape where pebbles are originally derived, and any pebbles from lower elevations within the watershed would overestimate 10Be concentrations due to nuclide production during fluvial transport.

The modeled depth profile is shown in Figure 5A. The 10Be concentrations decrease from a surface value of 69.18 ± 0.96 × 105 atoms/g (2σ uncertainty of 7%) to 1.09 ± 0.08 × 105 atoms/g. We use this depth profile to consider a scenario where an equal number of pebbles from 10 cm depth intervals down to a depth of 10 m are shed into the Mojave River at the same time by a landslide. This would be similar to a large block toppling from a cliff side of a canyon cut into the high-elevation, low-relief surface of the San Bernardino Mountains. The relative probability of these hypothetical pebble 10Be concentrations is shown in Figure 5C and compared to our modeled Ninh values, shown in Figure 5B. The model results demonstrate that it might be possible for some of the apparently older Mojave River pebbles to have entered the channel network with Ninh similar to our model values. However, our distribution of 10Be concentrations in our depth profile model is not consistent with the distribution of model Ninh of our pebbles.

Timing of Pebble Erosion and Large Mojave River Paleodischarge

Quartzite pebbles are shed into the Mojave River from a small portion of the high-elevation regions of the San Bernardino Mountains (Fig. 1; Dibblee, 1973; Sadler and Reeder, 1983; Morton and Miller, 2006) or recycled from Mojave River alluvial deposits. The Binnie et al. (2007) renormalized 10Be concentrations indicate that quartzite pebble source areas in the San Bernardino Mountains have equivalent exposure ages, i.e., the amount of time pebbles spend in the zone of 10Be production, between ∼4.4 k.y. and ∼15.4 k.y. (Supplemental Table 34). If we were to account for this inheritance in our strath pebble 10Be concentrations, the ages of the youngest pebbles would be much younger than the strath age from both our luminescence ages and the chronology of Lake Manix (Reheis et al., 2012). It is possible that the sampled pebbles entered the Mojave River drainage network by different, more rapid, geomorphic processes than the fine sand sampled by Binnie et al. (2007) (e.g., Niemi et al., 2005; Yanites et al., 2009; Kober et al., 2012). In order to not violate stratigraphic principles, the youngest pebbles must have transport times of ∼1 k.y. or less, which is a factor of ∼8 shorter than the mean equivalent exposure age of the catchments in the Mojave River basin sampled by Binnie et al. (2007). We interpret this to mean that the production and delivery of sediment to the Mojave River in the San Bernardino Mountains was more rapid during the periods of time represented by our pebble ages than it has been over the period of time represented by the Binnie et al. (2007) data.

We interpret the mean ages of the clusters, identified by the NKDE, as at least three periods of discharge in the Mojave River large enough to entrain new pebble-sized sediment supplied from hillslopes in quartzite source areas, as well as large enough to remobilize pebbles stored in alluvial deposits transported during earlier periods of high discharge. The precision of our individual pebble ages, and of the age clusters, does not allow us to draw any conclusions about the time scale of these transport periods below the time scale of the stated uncertainty, nor does it allow us to make any inferences about the specific moisture delivery mechanism, seasonality, or intensity of discharge-supplying precipitation. With those caveats in mind, in our inferred scenario the first set of pebbles would have entered the Mojave River drainage network ca. 240 ka, been transported some unknown distance downstream, and then deposited. The next event, at least as resolved by our data, occurred ca. 100 ka, during which new pebbles were shed into the Mojave River, transported some distance downstream, and mixed with pebbles from the ca. 240 ka events that were remobilized. This process would have been repeated ca. 54 ka and ca. 25 ka, at which point the pebbles were deposited on the strath.

The periods of time represented by our pebble ages may also represent times of rapid deposition in the delta within Lake Manix, which was later exhumed. Some of its sediment may have been redeposited on the strath as the Mojave River extended across the floor of the drained lake. Periods of deeper water and delta building are known, in a general way, for Lake Manix but are difficult to compare to pebble ages. The delta extended to within 500 m of our sample site starting ca. 50 ka but pebble-sized sediments are rare in the delta in this area. Coarser clasts are more common >10 km to the west, an area where delta sedimentation timing is poorly constrained but in all probability is older than 50 ka. It is possible to argue that the deeper lake episodes coincident with Marine Isotope Stages (MIS) 6 and 4 may have resulted from enhanced discharge, causing more gravel deposition in the delta. If so, burial and shielding could explain the age pulses that postdate those two events.

Alternatively, more rapid sediment production and delivery from hillslopes to the channel network may be indicative of cooler, wetter conditions during the times represented by our pebble ages, when weathering might have been more efficient at generating pebble supply and sustained fluvial discharge high enough to transport those pebbles (e.g., Miller et al., 2010; Ellwein et al., 2011). These high discharge proxy ages occur during both glacial and interglacial periods (Fig. 3).

That we observe age clusters of pebbles during interglacial periods is consistent with the interpretations of other data sets from the region. Based on sedimentologic and stable isotope data from a core of Lake Manix sedimentary deposits <1 km to the northeast of our sampling location, Reheis et al. (2012) inferred sustained shallow (<5 m) to moderate (5–8 m) depth lakes in the Manix basin during several interglacial periods between ca. 500 and 25 ka. In addition, a review of the geomorphic signals of Quaternary climate change in southwestern North America by Antinao and McDonald (2013) showed intermittent shallow Mojave Desert lakes during the latest Pleistocene–earliest Holocene. In both cases Antinao and McDonald (2013) attributed relatively high moisture during interglacial periods to winter Pacific Ocean sources. However, because of the resolution of our 10Be ages, we are unable to infer the water sources, seasonality, or durations of these wet periods, which could be anything from a single wet season or heavy snow year, to several years or decades of strong Northern Annular Mode resulting in increased moisture from atmospheric rivers (e.g., Reheis et al., 2012), or one or more strong El Niño–Southern Oscillation cycles (e.g., Antinao and McDonald, 2013). Regardless of the main climate drivers responsible for increased Mojave River discharge, our data indicate that it is likely not as simple as glacial = wet, interglacial = dry. This is illustrated by study of effective moisture recorded in soil proxies of U isotopes in opal. Maher et al. (2014) argued that more moisture was present in valley-bottom soils during MIS 3 and early MIS 2 but distinctly before the Last Glacial Maximum (LGM), and suggested that the LGM was cold and dry but was preceded by a wet period, explaining the diachronous behavior of lakes and glaciers. Our data are consistent with this interpretation.

Mojave River Paleodischarge

The quartzite pebbles deposited on the strath terrace east of Barstow are sourced exclusively from the low-relief, high-elevation region of the San Bernardino Mountains, perhaps cycled multiple times in Mojave River fluvial deposits. Although it might be possible for the observed distribution of pebble 10Be exposure ages to be the result of a single event, such as a deep-seated landslide, our analyses of different scenarios indicate that it is also likely that the pebble ages represent periods of time when Mojave River discharge was large enough to transport pebble-sized sediment in Mojave River bedload. We can use the sizes of our pebbles and some basic measurements of the geometry of the modern Mojave River channel to estimate the magnitude of those past discharge events.

Transport occurs when the driving forces of flow depth, h, and gradient, S, exceed the resisting forces of the weight and cohesion of sediment, and can be expressed as (Heede, 1976): 
where g is the gravitational constant, θ is Shields’ parameter, estimated as 0.045, D is the grain diameter, and ρw and ρs are the density of water (0.9997 g/cm3) and sediment, respectively. This relationship can be rearranged to give the minimum flow depth necessary to move a specified grain size at a given channel gradient, 

We determined D by measuring the principal axes of each of the pebbles. The density of each pebble was determined by measuring each pebble’s mass and volume. These characteristics, as well as roundness and sphericity, both determined in a computer-aided design program, are presented in Supplemental Table 1 (see footnote 1). The mean intermediate axis of our 10 pebbles is 5.15 cm. Using a mean pebble density of 2.44 g/cm3 and the mean slope of the Mojave River between the Mojave Dam and the sampling location determined from the 10-m-resolution National Elevation Dataset (NED; http://ned.usgs.gov/) (0.0024; Fig. 6A), the minimum flow depth required to entrain pebbles the size that we analyzed is 1.6 m.

We also drew 54 thalweg-perpendicular channel cross sections at regular intervals, using the same NED, between the Mojave Dam and the sampling site (Supplemental Table 45). Using these data we calculate a mean cross-sectional area of the Mojave River channel of 844 m. Although we recognize that the geometry of the Mojave River channel and the amount and caliber of its bedload may have been different in the past, using this cross-sectional area the minimum discharge required for the modern river to achieve a flow depth of 1.6 m and entrain pebbles of the size we analyzed is 1346 m3/s (flow depth × area).

The U.S. Geological Survey (USGS) has maintained a stream gauge at Barstow (10262500; 34°54′25″N, 117°01′19″W) since 1 October 1930. These flow data, to 12 March 2014, are presented in Figure 6B. There are several instances of flow into the current Mojave River terminus at Silver Lake Playa, some of which occurred prior to the USGS stream gauge record upstream at Barstow. These large discharge events have been attributed to large winter floods, tied directly to large amounts of precipitation or snow melt in the San Bernardino Mountains, and occurred in 1862, 1867, 1884, 1891, 1909, 1916, 1922, 1938, two in 1969, 1978, 1980, 1983, and 1993 (Enzel and Wells, 1997, USGS stream gauge data). The maximum measured discharge at Barstow over the gauging interval (1 October 1930 to present) was 512.53 m3/s on 3 March 1938. Our simple calculation indicates that a discharge at least ∼2.5 times the size of the largest flood on record would be required to move pebbles past Barstow. Average daily flow at Barstow for the largest floods on record (e.g., Enzel and Wells, 1997) is significantly lower than their corresponding maximum discharge. This indicates that pebbles moved from their source to the strath in either high-frequency, very short duration (minutes to hours?) events, or less frequent, longer duration (weeks to months?) events. The precision of our 10Be-derived ages precludes us from stating definitively what specific ranges of frequency-duration combinations were necessary to move pebbles the ∼160 km from the headwaters of the Mojave River to the strath where we collected them, and therefore prevents us from speculating about the specific type of climatic driver responsible for the minimum discharge needed to transport pebbles (e.g., Miller et al., 2010; Ellwein et al., 2011; Antinao and McDonald, 2013; Kirby et al., 2012). We can, however, use our pebble ages and the known age of alluvium deposited on the strath to estimate the amount of time necessary to transport pebbles from their source.

Duration of a Given Pebble Transport Event

The age of the strath from luminescence (24.35 ± 2.38 ka) and the age of the youngest quartzite pebble using 10Be (20.71 ± 4.26 ka), are indistinguishable within their respective uncertainties. Thus, the minimum amount of time needed to transport quartzite pebbles from their source in the San Bernardino Mountains to our sample location is ∼100 yr. Even if the younger 2σ bound of the OSL age of the alluvium on the strath (21.97 ka) and the older 2σ bound of the youngest individual pebble (24.97 ka) are considered, the difference is only 3 k.y. This is consistent with recently reported sand transport rates from the Mojave River. McGuire and Rhodes (2015) used the observed downstream increase in IRSL apparent ages, combined with measures of the degree of partial bleaching, to demonstrate cyclical bleaching and burial of grains as they are transported downriver. These observations, combined with a sediment transport model for sand between the Mojave River dam at Forks and Barstow, indicates fine sand transport times between 210 and 800 yr, consistent with our estimated minimum pebble transport time.

Additional Cosmogenic Nuclide Considerations

The fact that sampled pebbles group into four sets of ages indicates that at least some of the pebbles have been recycled several times. Between each erosion-transport event pebbles could have been exposed at the surface or been buried beneath alluvium. During this time, initial 10Be concentrations would have changed due to radioactive decay and secondary 10Be production, the rate of which depends in part on the depth of burial (e.g., Granger, 2006). It is interesting to note that the variance of ages between pebbles in a given group increases with increasing group mean age. This could be the result of a group of pebbles having similar initial 10Be concentrations when they enter the Mojave River, those concentrations diverging to greater and greater extents with more and more recycling events due to increasingly divergent histories of depth and duration of burial. More detailed discussion is beyond the scope of this work, but the data presented here may provide a foundation for further exploration through a combination of a larger data set, measurements of nuclide pairs, such as 26Al-10Be or in situ 14C-10Be, and numerical models of discharge and pebble exposure, transport, and burial.


We used the 10Be concentrations of 10 individual quartzite pebbles, collected from a ca. 24 ka strath terrace of the Mojave River near Barstow, California, to determine the timing of large paleodischarge of the Mojave River as it propagated to a downstream basin. Our exposure ages indicate that periods of discharge large enough to transport pebble-sized sediment have occurred at least 3, and possibly 4, times over the past ∼240 k.y. These inferred large discharge events occurred during both glacial and interglacial conditions, indicative of high runoff and Mojave River discharge available to transport pebble-sized sediment related to mechanisms other than simply glacial-interglacial cyclicity. This conclusion is consistent with recent work that presents evidence for a greater role of anomalous moisture transport. Although our data set is small, it is suggestive of what we might be able to learn about the frequency and duration of pebble transporting events through some combination of more measurements of one or more cosmogenic nuclides, and models of river discharge and pebble transport.

This work was conducted as part of the Mojave Neotectonics Project, supported by the U.S. Geological Survey National Cooperative Geologic Mapping Program. Accelerator mass spectrometer measurements of 10Be/9Be were completed at the Center for Accelerator Mass Spectrometry at Lawrence Livermore National Laboratory by Susan Zimmerman and Robert Finkel. This work benefitted from discussions with Greg Balco, Brett Cox, David Bedford, Yehouda Enzel, and Alan Hidy. Exploratory 10Be studies on Mojave River pebbles by Lewis Owen motivated our more detailed study. Constructive reviews by Stephen DeLong and three anonymous reviewers, as well as editorial guidance from Jose Hurtado and Shan de Silva, improved the manuscript.

1Supplemental Table 1. Physical characteristics of pebbles analyzed for 10BE. Please visit http://dx.doi.org/10.1130/GES01134.S1 or the full-text article on www.gsapubs.org to view Supplemental Table 1.
2Supplemental File. Pebble sampling and Cosmogenic nuclide sample preparation. Please visit http://dx.doi.org/10.1130/GES01134.S2 or the full-text article on www.gsapubs.org to view the Supplemental File.
3Supplemental Table 2. CRONUS Earth Calculator input parameters. Please visit http://dx.doi.org/10.1130/GES01134.S3 or the full-text article on www.gsapubs.org to view Supplemental Table 2.
4Supplemetal Table 3. Cosmogenic 10Be data of Binnie et al. (2007) re-normalized for comparison to Mojave River pebble data presented in Table 2. Please visit http://dx.doi.org/10.1130/GES01134.S4 or the full-text article on www.gsapubs.org to view Supplemental Table 3.
5Supplemental Table 4. Mojave River channel geometry between the dam at the forks and the location where pebbles were collected. Please visit http://dx.doi.org/10.1130/GES01134.S5 or the full-text article on www.gsapubs.org to view Supplemental Table 4.