boreas

Geophysical evidence for Holocene lake-level change in southern California (Dry Lake) BROXTON W. BIRD, MATTHEW E. KIRBY, IAN M. HOWAT AND SLAWEK TULACZYK

BOREAS

Bird, B. W., Kirby, M. E., Howat, I. M. & Tulaczyk, S. 2010 (January): Geophysical evidence for Holocene lakelevel change in southern California (Dry Lake). Boreas, Vol. 39, pp. 131–144. 10.1111/j.1502-3885.2009.00114.x. ISSN 0300-9483. Ground penetrating radar (GPR) data are used in combination with previously published sediment cores to develop a Holocene history of basin sedimentation in a small, alpine lake in southern California (Dry Lake). The GPR data identify three depositional sequences spanning the past 9000 calendar years before present (cal. yr BP). Sequence I represents the first phase of an early Holocene highstand. A regression between o8320 and 48120 cal. yr BP separates Sequence I from Sequence II, perhaps associated with the 8200 cal. yr BP cold event. Sequence II represents the second phase of the early-to-mid Holocene highstand. Sequence IIIa represents a permanent shift to predominantly low lake stands beginning 5550 cal. yr BP. This mid-Holocene shift was accompanied by a dramatic decrease in sedimentation rate as well as a contraction of the basin’s area of sedimentation. By 1860 cal. yr BP (Sequence IIIb), the lake was restricted to the modern, central basin. Taken together, the GPR and core data indicate a wet early Holocene followed by a long-term Holocene drying trend. The similarity in ages of the early Holocene highstand across the greater southern California region suggests a common external forcing – perhaps modulation of early Holocene storm activity by insolation. However, regional lake level records are less congruous following the initial early Holocene highstand, which may indicate a change in the spatial domain of climate forcing(s) throughout the Holocene in western North America. Broxton W. Bird, University of Pittsburgh, Department of Geology and Planetary Science, Pittsburgh, PA 15260, USA; Matthew E. Kirby (corresponding author: [email protected]), California State University, Fullerton, Department of Geological Sciences, Fullerton, CA 92834, USA; Ian M. Howat, The Ohio State University, School of Earth Sciences & Byrd Polar Research Center, Columbus, OH 43210, USA; Slawek Tulaczyk, University of California, Santa Cruz, Department of Earth and Planetary Sciences, Santa Cruz, CA 95064, USA; received 29th January 2009, accepted 16th June 2009.

Ground penetrating radar (GPR) has proved a useful tool for understanding the stratigraphy of depositional basins (James 1995; Brian & Frederick 1997; Schwamborn et al. 2002; Shuman & Donnelly 2006; Wang & Horwitz 2007). In remote locations where access is limited, challenging, or where the use of traditional seismic equipment is not feasible, portable GPR systems provide an alternative method to investigate large-scale, subsurface basin stratal geometries. As well, portable GPR units offer the ability to image subsurface structures at greater resolution than most traditional seismic equipment, which can improve correlations between sediment cores and geophysical data. Following initial work on Dry Lake by Filippelli & Souch (1999) and Filippelli et al. (2000), Bird (2005) and Bird & Kirby (2006) used a combination of sedimentological analyses to interpret an 8.4 m, 9000 cal. yr BP sediment core from Dry Lake’s depocenter. Their results suggest that the early Holocene was wetter than today, followed by a long-term Holocene drying. Furthermore, Bird & Kirby (2006) proposed that the early-to-mid Holocene glacial advance identified by Owen et al. (2003) may represent an abrupt cold interval associated with the 8200 cal. yr BP cold event (Barber et al. 1999). In this article, we use GPR data to test these results by examining basin-wide sedimentation dynamics over the same interval of time as that represented by the sedi-

ment cores. We note that, generally, GPR data should be collected prior to sediment core collection; however, as in our research, the availability of GPR equipment reversed the recommended practice of collecting the GPR data before the sediment core.

Study area Dry Lake is a moraine dammed, alpine ‘intermittent’ lake (2763 m a.s.l.) located at the headwaters of the Santa Ana River in the San Gorgonio Wilderness of southern California (Fig. 1). Dry Lake’s drainage basin is small, about 3.7 km2, and underlain to the north by biotite gneiss and schist, and to the south by quartz monzinite (Morton et al. 1980). There are three inlets to the lake; however, these inlets are active today only during times of spring melt, storm events, and extremely wet years (Fig. 1). Dry Lake has one outlet at the northwest corner of the lake basin where a moraine impinges on the adjacent valley wall (Fig. 1). At present, the outlet is stranded above average lake level. While detailed lake level data are not available for Dry Lake, the lake has overflowed after exceptionally wet winters, such as 1992, 1995, and 2005 (http://www. sgwa.org/trails.htm). During average to dry years, spring snowmelt creates marsh-like conditions in the

DOI 10.1111/j.1502-3885.2009.00114.x r 2009 The Authors, Journal compilation r 2009 The Boreas Collegium

132

Broxton W. Bird et al.

BOREAS

W 116 50’ 00”

W 116 49’ 00”

C 2681

2743

Dry Lake (2763 m)

metamorphic N 34 07’ 00”

15-16 kyr advance

granite

2865

9-5

A

2987

ky

ra

dv

an

ce

Pacific Ocean

3108

NA

Atlantic Ocean

Study Area

CA

3230 3352

500 m Elevation in m a.s.l.

35° N

N

115° W

AZ

B

Dry Lake

Los Angeles

Gulf of California

Pacific Ocean

Fig. 1. Regional map with study site. Small filled triangle on Dry Lake is core DLPC03-1 and 2; small unfilled triangle is approximate location of Filippelli et al. (2000) core. Relevant proposed glacial advances by Owen et al. (2003) are shown.

lake’s central basin that can persist into the spring and early summer. Instrumental meteorological data are lacking from Dry Lake; therefore, we use precipitation and temperature data from a nearby weather station at Big Bear Lake and Big Bear Dam (about 17 km NNW) to infer recent climate trends at Dry Lake. Characteristic of the coastal southwest, Dry Lake experiences peak precipitation between November and April. Precipitation falls mainly as snow at the lake, averaging 70–100 cm yr 1 water equivalent (Minnich 1984). Strong southwesterly winds during winter storms strip snow from southerly stoss slopes and preferentially redistribute it on northeasterly leeward slopes, resulting in significantly greater than average snow accumulation on these aspects (Minnich 1984). Once transferred to the leeward slopes, the succeeding runoff is within the drainage basin of Dry Lake. Therefore, winter precipitation and its subsequent runoff is the dominant water source to the modern lake. Summer precipitation is extremely limited on average, totaling only 7.5% of the annual average in the San

Bernardino Mountains. The small rise in precipitation during July–September reflects the limited contribution of the North American Monsoon, local convective storms, and occasional land falling cyclones to the hydrologic budget of the San Bernardino Mountains (Tubbs 1972).

Methods Core sedimentology and age control Two cores comprised of 27 (304 cm) segments and totaling 8.4 m (DLPC03-1 and DLPC03-2) were collected from the depositional center of Dry Lake using an Eijkelkamp hand-coring device in the Fall 2003 (Fig. 1). Despite excellent overall recovery, some gaps exist in the sediment record. We attribute these gaps to the coring process, but because of the high-density age control (20 AMS 14C dates) these gaps are accounted for in the age model and represent real gaps in sediment. Because the lakebed was exposed when the cores

Holocene lake-level change in southern California

BOREAS

133

1

200

8 2 7

Northing (m)

150 100 MHz 50 MHz

3

100 6

50 4

0 Fig. 2. Study site with core locations, GPR lines, and relevant geomorphic features.

N34.1194; W116.8288

0

were collected, the deepest part of the lake basin was determined visually. The GPR data discussed in this article confirm that the core location was the lake’s depocenter throughout the interval discussed (past 9000 cal. yr BP). Following collection, the cores were transported to the California State University, Fullerton (CSUF) Paleoclimatology Laboratory, where they were split, described, photographed, and subsampled for sediment analyses. Subsampling consumed one half of the split core. The second half is archived at CSUF in cold storage. For comparison only, a sediment core from Filippelli et al. (2000) is also used based on its description in their article. This core was collected in the lake’s shallow, littoral zone approximately 150 m southeast of the depocenter near an ephemeral inlet (Fig. 2). Because of its location, the Filippelli et al. (2000) core is lower resolution and terminates in coarse alluvium. Nonetheless, it provides a tie-point for ground-truthing the GPR data across the basin and correlation to DLPC03-1 and -2. Percent total organic matter and total carbonate (LOI 550 and 9501C, respectively) were determined at 1 cm contiguous intervals using the loss-on-ignition method (Dean 1974). Each sample was baked in a muffle oven at 5501C and 9501C, respectively, for 2 h, subtracting the residual weight from the original weight to determine the percent weight loss following the method by Dean (1974). LOI 9501C is not discussed further, as all values were 3%, or below, and therefore may represent interstitial water loss from the clay minerals instead of true carbonate loss, thus precluding useful interpretation (Dean 1974). Grain-size distribution was measured at 5 cm intervals (1 cm intervals over transitions and intervals of

5

50

100 Easting (m)

150

200

‘interest’) using a 2004 Malvern Mastersizer 2000 laser diffraction particle analyzer coupled to a Hyrdo 2000G. Approximately 0.1–0.5 cm3 of sediment was boiled in DI water and pretreated with Z30 mL of 30% H2O2 to remove organics. Organic material larger than 2000 mm was removed by wet-sieving prior to the addition of hydrogen peroxide. At the beginning of each measurement day, a tuff standard (TS2) with a known distribution between 1.0 and 16.0 mm (avg. 4.53 mm0.07; n = 2998) was measured twice (or more as necessary to establish analytical stability) and compared to past measurements to assess the equipment’s accuracy, repeatability, and stability. Thereafter, TS2 was run every 10 samples and once at the end of the day’s analyses for a final assessment. TS2 results were compared to values obtained by measuring known Malvern standards as an additional measure of stability. The measurement principle used was the Mie Scattering principle. All data are reported as volume percent and divided into sand (74.00–2000 mm), silt (3.90–73.99 mm), clay (0.02–3.89 mm), and mode. Microfossil counts of charcoal were taken at 1 to 5 cm intervals using a binocular microscope after being dried, weighed, and wet-sieved at 125 mm. Counts of split samples were corrected for the number of splits and corrected to the amount of charcoal per 5 g of dry sediment (Mullins 1998). An age model was constructed for the composite core – DLPC03-1 and DLPC03-2 – using 20 of the 27 AMS 14 C dates (Fig. 3); all ages were calibrated using the online Calib program (Stuiver & Reimer 1993). Seven of the dates showed age reversals (Bird & Kirby 2006). These samples were eliminated because they were collected from rapidly deposited layers (RDLs), which are


134

BOREAS

considered instantaneous events that contain reworked organics (Bird & Kirby 2006). RDLs were identified visually as well as by magnetic susceptibility, loss-onignition 5501C, and grain-size data (Bird 2005; Bird & Kirby 2006). The remaining 20 dates were plotted versus depth, assuming linear sedimentation between points (Fig. 3). To account for RDLs in the age model, the basal contact was assigned the age generated from initial linear regressions between AMS 14C dates. The top contact age was assigned the basal contact age plus one year to model an instantaneous event (Bird & Kirby 2006). Subsequent equations generated from linear regressions between dates were used to convert depths to ages.

wavelength, at regular intervals of one-half a wavelength; a radar measurement is taken at each interval until the signal fades. The rate at which the two-way travel time between the antennas increases with increasing distance between the antennas gives the average velocity of the radar signal through the sediments. This method yielded a sediment velocity of 0.12 m/ns, which is within the range of typical values for these sediments; it is the value we use to estimate absolute depths from the radar signal travel times. Due to the possible uncertainty of this measurement, however, reflector positions are also presented in units of two-way travel time. Sequence stratigraphy for the basin profiles was developed using standard techniques and terminology as outlined by Mitchum et al. (1977) and Scholz (2001), specifically, for lakes.

GPR acquisition Six intersecting GPR lines were collected across Dry Lake using a commercially built 1 kV profiling GPR system (Sensors & Software). All six lines were acquired using 100 MHz signals; four additional lines were also obtained with 50 MHz. Two of the lines directly intersect sediment core locations (Filippelli et al. 2000; Bird & Kirby 2006). In order to convert two-way travel times of radar reflections to absolute depths below the surface, the speed at which radar signals travel through the sediments (the sediment velocity) must be determined. The average sediment velocity at Dry Lake was measured through a Common Mid-Point (CMP) survey completed in the center of the basin (Fisher et al. 1992; Greaves et al. 1996). In the CMP survey, the transmitter and receiver antennas are moved progressively apart, starting from an initial separation of one

0

2000

4000

6000

8000

0 Sequence IIIb

10 000 0

1860 cal. yr BP

Sequence IIIa

Depth (cm)

200

5550 cal. yr BP

Sequence II

400

200

400

bottom reflector in Sequence II: ~ 8120 cal. yr BP

600

600 Early Holocene Regression: 8200 cal. yr BP cold event? top reflector in Sequence I: ~ 8320 cal. yr BP

800

Sequence I

0

800

2000 4000 6000 8000 10 000 Age (cal. yr BP from Bird & Kirby 2006)

Fig. 3. Core DLPC03-1 and -2 age in cal. yr BP versus depth (cm). GPR sequences are labeled on graph.

Results Core DLPC03-1 and DLPC03-2 sedimentology Fig. 4A–E shows the percentage total organic matter (LOI 5501C), percentage sand, percentage silt and clay, charcoal count data, and a visual description for core DLPC03-1 and 2. The sediment data reveal a large range of variability and average values throughout the core. Total organic matter is high and variable between the core bottom and 700 cm, low and variable between 700 and 250 cm, low and stable between 250 and 100 cm, and increasing with moderate variability between 100 and the core top (Fig. 4A). Sand is highly variable throughout the core (Fig. 4B). There is a general increasing trend with large variability from the core bottom to 550 cm, two large sand layers between 550 and 420 cm, and a general decreasing trend with moderate variability from 420 cm to the core top. As noted on the stratigraphic column, there are several prominent sand layers throughout the core. Clay is highly variable with a general decreasing trend from the core bottom to 550 cm; from 550 cm to the core top, clay is less variable and generally increasing to the core top (Fig. 4C). Like the sand and clay, silt is also quite variable (Fig. 4C). Silt shows a general decreasing trend from high values at the core bottom to low values by 550 cm; from 550 to 420 cm there are two pronounced low silt layers, which correspond to the high sand layers (Fig. 4C). From 420 cm through to the core top, silt averages about 60% and displays moderate variability. Charcoal is high and variable from the core bottom to 390 cm (Fig. 4D). From 390 cm to the core top, charcoal is relatively low with the exception of the spike at 40 cm and the upper 15 cm. The complex stratigraphy of core DLPC03-1 and 2 is shown by Fig. 4E. The most distinct feature, besides the sundry sand layers, is the color change at 275 cm. This sediment color change is characterized by a shift from grey to dark grey to black silt and clay interbedded with sand layers and charcoal


BOREAS

A

B

C

D

135

E

Fig. 4. DLPC03-1 and -2 sediment data with GPR sequences.

and organic-rich layers to a brown to dark brown silt and clay with fewer sand layers and a decreased occurrence of charcoal fragments. Another change in sedimentology occurs at 180 cm, which is characterized by a substantial decrease in the occurrence of sand layers. Excepting brief periods of increased sedimentation between 4500 and 4000 cal. yr BP, and 1500 and 1400 cal. yr BP, average sedimentation rates are much higher from 9000 to 5550 cal. yr BP (0.13 cm yr 1) than from 5550 to modern cal. yr BP (0.04 cm yr 1) (Fig. 3). GPR data Three sequences were identified using the 100 MHz GPR data based on bounding surfaces as defined by their unconformities or correlative conformities (Mitchum et al. 1977; Scholz 2001). Line 8-2 was used as the master line because of its basin edge location and clear depiction of several diagnostic reflection terminations and their stratal geometries (Fig. 2). It is our rationale that the basin edge location of line 8-2, abutted against a distinct bedrock promontory, increased the erosional susceptibility of the sediments as lake level rose and fell. Lines 1–6, 8-5, and 7-4 taper more gradually away from the central basin and they do not impinge on bedrock or moraine promontories. Moreover, lines 1–6, 8-5, and 7-4 taper into the modern inlets, which reduces the quality of the GPR data because of the complex depositional geometries associated with fluvial-deltaic environments (Fig. 2). As a result, these lines do not preserve the diagnostic reflection terminations and their stratal geometries at the same quality as

in line 8-2. Nonetheless, we were able to tie the four lines together using line 8-2 as our master profile. The three sequences (I, II, IIIab), from oldest to youngest, are color-coded to highlight major features. The presence of conductive pore water within the central basin reduced signal quality and transmission. Even so, we were able to correlate all three sequences across the basin for most lines. Sequence I (green) represents the bottommost sequence captured by the core sediments (Fig. 5). However, it is clear from the 50 MHz line 1–6 as well as the 100 MHz lines 7-4 and 5–8 that there exists basin fill in excess of possibly 200–300 ns (c. 18 m) below the modern lake sediment surface (Figs 6, 7), although the quality of data diminishes rapidly with depth in the central basin (Fig. 6). Using the 100 MHz data, Sequence I is characterized by classic divergent basin fill with even to hummocky reflection patterns (Figs 5–7). Accordingly, the reflections (i.e. sediments) tend to increase in thickness towards the central basin and away from the basin’s edge. Divergent patterns are common in basins where sedimentation is focused into a deeper, profundal environment. Because of the reduction in data quality with depth, the base of Sequence I is undefined. The top of Sequence I is characterized by the truncation of reflectors (green) against an overlying erosional surface (blue) (Figs 5,7). Erosional truncation indicates a drop in base level, which allows erosion of the ‘exposed’ or near exposed surface through various processes such as wave action, subaerial weathering and erosion, and/or fluvial processes. The age (8320 cal. yr BP) of the uppermost, truncated reflector

136


BOREAS

DL GPR Line 8-2: 100 MHz

0

GPR Depositional Sequences

Two- Way Travel Time (ns)

Sequence I: early Holocene high stand phase 1 50

Sequence II: early-to-mid Holocene high stand phase 2

100

Sequence IIIa: mid-to-late Holocene low stands Sequence IIIb: mid-to-late Holocene low stands

150 Station 8 0

10

Station 2 20

30

40

50

60

70

80

90

cal. yr BP (x1000) 0

2

4

6

8

10 0

0

GPR/Core-Age Relationships 0.60 m: ~ 1860 cal. yr BP

100 2.35 m: ~ 5550 cal. yr BP

300

50

400 500 100

Depth (m)

Two- Way Travel Time (ns)

200

600

5.70 m: ~ 8120 cal yr. BP

700

6.70 m: ~ 8320 cal. yr BP

800 900

150 Station 8 0

10

Station 2 20

30

40 50 Distance (m)

60

70

80

8.40 m: ~ 9000 cal. yr BP core basal age

line 1-6

90 intersection

Fig. 5. GPR 100 MHz line 8-2. Age-depth (cm) graph for core DLPC03-1 and 2 is overlain on the bottom image based on the CMP determined velocity.

in Sequence III (green) is based on correlating the inferred depth from the GPR profile using 0.12 ns/m to the calibrated AMS 14C age-depth relationship for core DLPC03-1 and 2 (Fig. 5). Sequence II (blue) represents the sediment unit between 8120 and 5550 cal. yr BP based on GPR inferred depth estimates and its comparison to the core DLPC03-1 and -2 calibrated AMS 14C age-depth model (Fig. 5). Above the truncated reflectors of Sequence III (green) is an erosional surface, which delimits the lower boundary of Sequence II (8120 cal. yr BP) (Figs 5–7). Onlapping and offlapping stratal terminations throughout Sequence II attest to its dynamic sedimentological history (Figs 5–7), although the broad extent of the Sequence II reflectors indicates a large area of basin fill during its depositional history (Figs

5–7). Onlapping reflectors indicate a rise in base level and the subsequent migration of sedimentation towards the rising base-level surface. Offlapping reflectors represent a decrease in base level wherein sedimentation follows the base-level surface. Internally, Sequence II is characterized by moderately defined divergent basin fill with evidence for sediment thickening towards the central basin. The quality of the data diminishes towards the edges of lines 7-4 and 5-8 because of their relationship to inlet sources (Fig. 7). The upper boundary of Sequence II delimits the reflector on which Sequence IIIa (red) offlaps (Figs 5–7). Sequence III (red (a) and yellow (b)) represents the uppermost sediment unit between 5550 and 1860 cal. yr BP (Sequence IIIa) and 1860 cal. yr BP and the modern sediment surface (Sequence IIIb); these ages are


BOREAS

Two-Way Travel Time (ns)

0

137

DL GPR Line 1-6: 100 MHz

50 100 150 200 0

20

40

60

80

100

120

140

160

180

200

0

20

40

60

80

100

120

140

160

180

200

0

20

40

60

80

100

120

140

160

180

200


0 50 100 150 200


0

100

200

300

major sediment source


0

Fig. 6. GPR 50 MHz line 1–6 and GPR 100 MHz line 1–6.

100

200

300

0

20

based on the GPR inferred depth estimates and its comparison to the core DLPC03-1 and -2 calibrated AMS 14C age-depth model (Fig. 5). There is some evidence for both offlapping and onlapping relationships between the base of Sequence IIIa (red) and the top of Sequence II (blue), which indicates a complex depositional history. There is no clear evidence for erosional truncation nor is there evidence for onlapping beyond the edges of Sequence II. Sequence IIIab are characterized by laterally continuous, fairly even, basin fill. Sequence IIIb (yellow) represents the uppermost reflector in the modern basin environment; it also represents a contracted basin with a minimal surface area of sedi-

40

60

80

100

120

140

160

180

200

Distance (m)

mentation. Sequence IIIb (yellow) reflectors generally offlap Sequence IIIa (red) reflectors. Combined GPR and core sediment data Using the inferred GPR depth model (i.e. CMP determined sediment velocity of 0.12 m/ns), the GPR sequences were overlain on the core sediment data with known depths to assess the significance of the sequences (Fig. 4). In other words, we test if the independently defined GPR sequences correspond to significant sedimentological differences in core DLPC03-1 and 2. It is important to note that there is no a priori reason for


138

BOREAS

DL GPR Line 7- 4: 100 MHz Two-Way Travel Time (ns)

0 50 100 150 200 Station 7

Station 4

0

50

100

150

200




0 50 100 150 200 Station 7

Station 4

0


0

50

100

150

200

DL GPR Line 5–8: 100 MHz

50 100 150 200 Station 5 0

20

Station 8 40

60

80

100

120

140

160

180

200

major sediment source Two-Way Travel Time (ns)

0 50 100 150 200 Station 5 0

20

Station 8 40

60

80

100

120

140

160

Distance (m)

any meaningful correlation between the GPR sequences and the core sedimentology, particularly because they are independently acquired data. One might expect, however, over the 9000 cal. yr BP history of Dry Lake that the sequences as defined by GPR correspond to statistically significant differences in core sedimentology. This is a reasonable assumption, since it is the changes in the sediment’s physical properties that produce the reflectors. To address this issue, Student T- and F-tests for unpaired data with unequal variance were determined to assess the significance of the means and the variance between adjacent sequences and their corresponding sediment units. The results indicate that

180

200

Fig. 7. GPR 100 MHZ line 7-4 and GPR 100 MHz line 5–8.

all of the GPR sequences as defined by their respective sediment units show statistically different means and variances for at least one of the five sedimentological measurements; several sequences show multiple differences (Table 1). This analysis provides reason to conclude that the independently defined GPR sequences correspond to sedimentologically distinct units. For example, Sequences II and IIIa show statistically significant differences in charcoal content, percentage clay, percentage silt, and percentage sand (Table 1). Overlain on the 50 MHz image (Fig. 6) is the percentage sand for core DLPC03-1 and 2. Although not a perfect match, there is an interesting relationship between changes in


BOREAS

139

Table 1. GPR sequence sediment data, T-test and F-test probabilities. T-test (means)

Seq IIIb vs. Seq IIIa

%TOM Charcoal counts %Clay %Silt %Sand

o0.0001 0.15 0.47 0.12 0.15

F-test (variance) %TOM Charcoal counts %Clay %Silt %Sand

Seq IIIb vs. Seq IIIa o0.0001 o0.0001 0.23 0.07 0.09

sand content and the occurrence of the GPR reflectors. From this statistical analysis and the visual sand-reflector comparison, it is concluded that the independently defined GPR sequences record real, measurable differences in depositional environments through time.

Discussion Previous sedimentological results from Dry Lake suggest that the early Holocene was wetter than today followed by a long-term Holocene drying trend (Bird 2005). Additionally, Bird & Kirby (2006) propose that the earlyto-mid Holocene glacial advance identified by Owen et al. (2003) may represent an abrupt cold interval associated with the 8200 cal. yr BP cold event (Barber et al. 1999). The GPR data presented in this article provide additional lines of evidence to test these statements by examining basin-wide sedimentation dynamics over the same time interval as that in the sediment cores. If the early Holocene was wetter than today, the GPR data should reveal a large early Holocene basin with a substantially larger surface area of active sediment deposition and preservation than occurs today in the contracted basin. The GPR data should also reveal a decrease in basin extent during the mid-Holocene (5550 cal. yr BP) that corresponds to a decrease in sedimentation rate as reflected in the sediment core age model (Fig. 3). Finally, if the early-to-mid Holocene glacial advance identified by Owen et al. (2003) is an abrupt cold interval associated with the 8200 cal. yr BP cold event (Barber et al. 1999), the GPR data should reveal evidence for rapid lake-level change 8200 cal. yr BP, if this event produced a change in effective moisture across southern California as proposed by Bird & Kirby (2006). In the succeeding sections, we assess each of the statements above (i.e. a large early Holocene lake; a mid-Holocene transition to a smaller lake; and, lake level change 8200 cal. yr BP) and forward our final interpretations based on the combination of sediment core and GPR data.

Seq IIIa vs. Seq II 0.06 0.006 0.0005 0.03 0.01 Seq IIIa vs. Seq II o0.0001 o0.0001 0.0004 0.0004 0.0003

Seq II vs. Seq I o0.0001 0.03 0.83 0.0002 0.003 Seq II vs. Seq I o0.0001 o0.0001 0.20 0.66 0.54

Evidence for a large, early Holocene lake and a mid-Holocene climate shift The most striking feature revealed by the GPR data is the substantial decrease in basin extent and surface area of active sediment deposition and sediment preservation from the early Holocene through to the modern (Figs 5–7). Sequences I and II occupy the entire basin; yet, Sequences IIIab are limited to only the modern central basin (Figs 5–7). These data indicate a substantial change in the basin’s extent of sediment deposition and sediment preservation over the past 9000 cal. yr BP. An important question to address, however, is whether this decrease in basin extent is due to external forcings (e.g. climate) or internal forcings (e.g. accommodation space). It is not unusual for small basins to diminish in size through time in response to sedimentation and its forced reduction in accommodation space (Leeder et al. 1998; Slagle et al. 2006). Generally, this decrease in accommodation space is a response to the interaction of a rising sediment base and a breached, or near breached, raised outlet, which acts to limit the lake’s total volume and, therefore, its capacity to preserve deposited sediment. Dry Lake’s current outlet, however, is at least 3–4 m above the lake’s deepest point, which should allow for a similar amount of sediment to accumulate before the lake is limited by accommodation space. Furthermore, when filled to overflowing, such as in 2005, the entire basin, which includes the spatial extent of Sequences I and II, is occupied by water, not just the modern depocenter as defined by Sequences IIIab (Figs 1, 5–7). Therefore, if the lake were permanent today, sediment deposition should occur across the entire basin, not just within the region of Sequences IIIab. We argue, therefore, that accommodation space is not the reason for the observed decrease in basin extent in the mid-Holocene, but that external climate controls are driving lake-level changes and the extent of sediment deposition in the basin. We suggest that the reduction in basin extent reflects an overall decrease in the duration and occurrence of


lake highstands as a result of changes in the ratio of precipitation to evaporation. With time, less water on an annual basis will generate an intermittent or ephemeral lake, particularly in an arid environment such as the San Bernardino Mountains, where lakes depend almost entirely on spring snowmelt. As the lake changes from perennial to ephemeral, the total surface area of sediment deposition and sediment preservation will decrease, which favors the erosion or non-deposition of sediment, especially in the littoral zones. Additionally, less total runoff will reduce the flux of allochthonous sediment to the lake basin. For an alpine lake such as Dry Lake, allochthonous sediment is the predominant source of sediment to the basin, so any decrease in its flux will alter the dynamics of sediment accumulation in the lake (Bird 2005). The result over centuries to millennia is a total decrease in the basin’s extent of sediment deposition and sediment preservation. The GPR data distinctly illustrate this decrease in basin extent through a contraction of the uppermost reflectors (Sequences IIIab) and their offlapping stratal geometries, which reflect an encroachment of littoral zone into the lake’s central basin (Figs 5–7). The GPR evidence for a contracting basin through the Holocene is supported by sedimentological and age data in the Dry Lake sediment cores. A sediment color change at 275 cm (6400 cal. yr BP) from a grey to dark grey to black silt and clay to a brown to dark brown silt and clay is very close – considering the possible error involved with the GPR inferred depth estimates – to the sequence boundary between Sequences II and IIIa (5550 cal. yr BP) (Fig. 4). This color change is interpreted as a switch from a perennial deep lake, under dominantly reducing conditions, to an intermittent or ephemeral lake where occasional, or sustained, subaerial exposure favored oxidation (Fig. 4). If it is assumed that this color change represents the transformation from Sequence II to Sequence IIIa, then it is also associated with statistically significant changes in charcoal content (decreases from Sequence II to IIIa), percentage clay (increase), percentage silt (increase), and percentage sand (decrease) (Fig. 4, Table 1). Following the interpretation by Bird & Kirby (2006), sand in Dry Lake is an indicator of runoff (i.e. storminess). Less sand in Sequence IIIa supports the interpretation that Sequence IIIa (and b) represents a transition to a smaller lake forced by less precipitation, less frequent storms, and a decreased capacity for transporting sand into the central lake basin. Less charcoal also supports the interpretation for a change to a more arid climate – less precipitation equals less biomass equals less available biomass for burning. Alternatively, less charcoal may represent less runoff (i.e. drier climate) and transport of charcoal into the lake basin. In either scenario, a drier climate explains the decrease in charcoal. In addition to the sedimentological changes above, all of the sediment cores from Dry Lake (Filippelli et al.

BOREAS

Relative lake level 0

high lake level

low lake level Sequence IIIb

1860

Sequence IIIa Approximate age (cal. yr BP)

140

5550

regression? tra

ns

Sequence II

gr

es

sio

n

early 8000’s

regression?

Sequence I

*Regression is coeval/ caused(?) by 8.2 kyr cold event and dates >8120 and