Distal tephra beds provide important records of pyroclastic volcanism that enhance our overall understanding of eruptive frequencies, magnitudes, compositions, and hazards. Some beds also serve as widespread chronostratigraphic markers. Lacustrine sediments near Summer Lake, Oregon (United States), record numerous eruptions of Cascade arc sources over a period exceeding 2.5 m.y. Late Pleistocene sediments exposed in outcrop have yielded 88 visible tephra beds, including many beds not previously documented. Of these beds, 44 are characterized by rhyolitic glass, 40 contain predominantly basaltic or intermediate glass, and 4 are strongly heterogeneous in composition. Only 23 have been correlated to deposits outside of the Summer Lake basin. The remaining 65 beds provide a record of Cascade arc volcanism that is as yet unique to Summer Lake. Age-depth relations are well constrained for the upper 6 m of section, but are less certain in the lower 12.4 m. Tephra correlations and an overall age model suggest the following: bed B1 originates from an eruption of Mount Mazama (Crater Lake) ca. 20 ka. Beds I and W likely originate from eruptions of Mount St. Helens ca. 80 and 190 ka. A 7-cm-thick tephra bed correlated to Shevlin Park Tuff probably dates to ca. 198 ka. Tephra correlated to the Antelope Well tuff from Medicine Lake volcano dates to ca. 215 ka. Bed NN, at the base of the section, has an estimated age of at least 240–250 ka and probably originated from Newberry Volcano. Overall, this record significantly refines the Pleistocene tephrostratigraphic framework for western North America.
Much of what we know about the Cascade volcanic arc (western North America) has been determined through studies of lava flows and proximal tephra deposits (e.g., Bacon and Lanphere, 2006; Hildreth, 2007). Because unconsolidated pyroclastic materials on volcanic slopes are readily eroded, and because of the significant potential for burial by subsequent eruptions, proximal pyroclastic records are often incomplete (e.g., the set C tephras of Mount St. Helens; Mullineaux, 1996). Distal tephra deposits, especially those from favorable depositional settings like lakes, offer the potential to fill in many of the gaps in the proximal record, thereby enhancing our understanding of eruptive frequencies, magnitudes, compositions, and related hazards. Distal tephra beds also provide information on particle size and thickness distributions that enhances our understanding of ash-dispersal processes.
Tephra beds are also widely used as chronostratigraphic markers, providing independent age control for a large array of interdisciplinary studies ranging from archaeology to paleoseismology to surficial processes (e.g., Mehringer and Foit, 1990; Langridge, 1998; Hermanns et al., 2000). Tephra correlation also offers the potential to link together glacial, lacustrine, marine, and terrestrial records over long distances (e.g., Sarna-Wojcicki et al., 1991; Zdanowicz et al., 1999; Negrini, 2002). Furthermore, because tephra is dispersed rapidly across large areas, it can be used to test the synchroneity of environmental and climate changes with a precision often unmatched by radiometric dating techniques.
Summer Lake, which occupies a portion of the northwestern subbasin of pluvial Lake Chewaucan in south-central Oregon (Figs. 1 and 2), is a key reference locality with numerous tephra beds found together in a single stratigraphic context. Much work has also been done studying the paleoclimate, paleomagnetic, and paleoseismic records preserved in ∼18.4 m of outcrop and multiple cores (e.g., Negrini et al., 1988, 1994, 2000; Langridge, 1998; Cohen et al., 2000; Sarna-Wojcicki et al., 2001; Zic et al., 2002). The earliest work on the tephrostratigraphy is that of Allison (1945), who described the top few meters of section from two locations along the Ana River (Fig. 2). Starting from the bottom of this interval, Allison divided the stratigraphy into 19 distinct units, 6 of which (2, 4, 6, 8, 12, 18) are tephra layers. Conrad (1953) revisited the same localities and described many more beds, 42 in total. From sites C, E, and F (Fig. 2), Davis (1985) documented 54 beds, including the 6 numbered beds of Allison (1945) and 48 beds that he designated A–NN. Davis (1985) also obtained major element glass compositions by electron probe microanalysis (EPMA) for 33 of the 54 total beds and used these data to correlate several of the beds to locations outside of the basin and to identify some of the source volcanoes. Additional tephra beds from two cores, the Wetland Levee (WL) and Bed and Breakfast (B&B) cores (Fig. 2), and an age-depth model for the combined core and outcrop sequence were added later (Negrini et al., 2000). Four additional cores also exist, including two taken adjacent to outcrop site C (31 and 64 m) (Erbes, 1996; Sarna-Wojcicki et al., 2001) and two from the Summer Lake Playa site G (SPG) site (33 and 37 m) (Negrini, 2001) (Fig. 2). The total thickness of basin-fill sediment is thought to be ∼1.5 km (Travis, 1977), so the potential exists to also obtain much older records.
In a composite measured section on the Ana River, Lake County, Oregon, reproduced in Erbes (1996) and Negrini et al. (2001), J.O. Davis suggested the existence of additional tephra beds in the C, E, and F outcrop sections. Erbes (1996) and Langridge (1998) also suggested the presence of additional tephra beds based on study of site C outcrop and core, the WL and B&B cores, and additional outcrop locations.
Our study, which focuses primarily on the outcrop tephrostratigraphy in the Ana River canyon, was initiated on the basis of: (1) the presence of many tephra beds of undetermined glass composition, (2) the suggested presence of additional tephra beds, (3) the apparent lack of suitable correlatives to large pyroclastic deposits such as the Shevlin Park Tuff, and (4) uncertainties in the stratigraphic relations between outcrop locations. To this end, the C, E, and F outcrop sites were reexcavated (Figs. 3–5) and sampled in detail for tephra (Table 1 and Supplemental Tables S1–S31).
More than 200 tephra samples from the 3 outcrop locations have been analyzed for their glass compositions by EPMA (Table 2; Supplemental Table S4, see footnote 1). We chose to focus on glass compositions because these provide one of the most consistent and distinguishing characteristics for the identification and correlation of tephra beds (Sarna-Wojcicki et al., 1991). Whereas bulk tephra compositions typically vary with distance due to differential settling of glass and crystals, glass compositions are often uniform over long distances (Sarna-Wojcicki et al., 1991). In situations where major element glass compositions of two or more tephra beds overlap, additional information from mineralogy, age, stratigraphic relations, and/or the trace element composition of the glass are often used to distinguish and identify them (Sarna-Wojcicki, 2000). Note also that glass compositions may differ substantially from bulk compositions when crystallinity is high, the glass often containing a greater proportion of silica than the bulk tephra. For example, the latest Pleistocene tephras from Glacier Peak volcano have dacitic bulk compositions (Gardner et al., 1998), but contain homogeneous, high-silica rhyolite glass (Kuehn et al., 2009). Thus, glass and bulk compositions often are not directly comparable.
Most analyzed samples were mounted in epoxy as bulk tephra, polished, and then carbon coated. Wavelength-dispersive analyses of the site C and core samples were conducted primarily at the GeoAnalytical Laboratory of Washington State University (WSU) using a Cameca Camebax microprobe (15 keV accelerating voltage; 8–10 μm beam diameter, and 12 nA current). Similar analyses of the site E and F samples were conducted primarily at the University of Alberta (UA) using a JEOL 8900 microprobe (15 keV accelerating voltage; 10 μm beam diameter, and 6 nA current). For the four most abundant elements, Na, K, Al, and Si, both instruments were calibrated using the same natural obsidian standard, UA5831 (also known as CCNM 211 at WSU), which originates from Lipari Island, Italy. The remaining elements were calibrated on differing glass and mineral standards. As both of these long-established analytical procedures were designed primarily for rhyolitic glasses, P2O5, which is typically near detection limits in rhyolitic glasses, is not included in the data set.
At the University of Alberta, UA5831 obsidian and Old Crow tephra (UT1434) were periodically reanalyzed as secondary standards (typically 5–10 points each at the beginning and end of each day and again between about every 4 unknowns) to monitor calibration quality and detect any instrument drift. To maximize consistency between analyses obtained on different days, minor corrections were applied based on the secondary standard values obtained for UA5831. At WSU, the obsidian standard was analyzed as an unknown at the beginning and end of each day to assess calibration quality and detect drift, and the instrument was recalibrated as needed.
All results in Tables 2 and S4 (see footnote 1) are reported as oxide values normalized to 100%, volatile free (Froggatt, 1992). Most individual analyses with especially low totals (below ∼90%–92%) and obvious crystalline contaminants have been removed from the data set. Some of the mafic samples analyzed are very microcryst rich, and scatter in analytical data for these samples suggests some residual crystalline contamination (Table S4 [see footnote 1]). For the glass standard and the most homogeneous tephra samples (i.e., those with the smallest standard deviations), the observed variability (Tables 2 and S4 [see footnote 1]) largely reflects the precision of the analytical method.
Some modest differences are apparent between the data sets produced in the two laboratories. For mafic tephras, FeO values tend to be ∼0.5–1 wt% higher in the UA data (Table 2), probably due to the use of different Fe standards. There also are some differences in Na2O and SiO2 values that vary from sample to sample, and these are likely due to variations in Na migration resulting from the different beam currents.
To compare compositional values of different samples and evaluate potential correlations, two approaches were used: (1) the similarity coefficient of Borchardt et al. (1971, 1972), and (2) bivariate plots combining both unknowns and reference data. The similarity coefficient (SC), a simple ratio between concentrations, was employed in the same manner used in Kuehn and Foit (2006). This includes differential weighting of the various oxides to reflect differences in analytical precision. Higher SC values indicate more similar compositions, and identical values result in an SC of 1.0. Repeated analysis of the same homogeneous tephra on the same instrument typically results in similarity coefficients of 0.97–0.99 between sample means, thus SCs >0.97 are considered good evidence for correlation. Comparison of less homogeneous samples and comparison of analyses of the same tephra performed at different laboratories commonly results in lower similarity coefficients. An SC of 0.92 typically is considered the lowest acceptable value for correlation (Froggatt, 1992), provided that other significant and compelling evidence is available. For compositionally heterogeneous samples with well-defined end members, it is possible to use the SC to compare the end-member compositions separately. Similarly, the individual populations in bimodal and polymodal samples may also be treated separately.
For heterogeneous samples, bivariate plots are generally superior to the SC for evaluating potential correlations because plots allow full examination of multiple populations, compositional ranges, covariation trends, and the relative frequencies of different glass compositions. Even in more homogeneous samples, bivariate plots sometimes reveal subtle variations that can be important for testing correlations. Plots are typically less useful when comparing to published data, because the individual data points used to calculate the mean values are often not reported. In this study, bivariate plots were used extensively to test correlations between the outcrop localities. Where the necessary reference samples or reference data sets were available, bivariate plots were also used to evaluate distal correlations.
TEPHRA BEDS AND STRATIGRAPHY
At least 88 visible tephra beds are preserved in 18.4 m of lacustrine sediments at the C, E, and F outcrop locations along the Ana River (Fig. 6; Table 1). Most consist of particles of silt to fine or medium sand size, but several contain coarser grains, including three beds that contain small pumice lapilli. Table 1 summarizes the thickness, particle size, and outcrop color for each tephra bed, along with the silica content of the glass. Silica ranges are provided for samples in which observed variability noticeably exceeds analytical error. For the remaining samples, observed variability is largely a function of analytical error, so approximate mean values are provided instead. Photographs of most of the numerous tephra beds are available in Figures S1–S38 in the Supplemental Figure File2. Names for previously known tephra beds and designations for new beds follow the scheme of Allison (1945) and Davis (1985). The six tephra layers designated only by number were named by Allison (1945) in his description of the top few meters of section. Davis (1985) included these designations in his nomenclature and named additional tephra from the top of the section down using the alphabet. After the first 26 he used double letters (e.g., tephra Z is followed by AA). When new tephra beds were found between previously identified beds, he added sequential numbers after the name of the upper tephra (e.g., T1 is the name of the tephra found between tephras T and U). Note that several tephra beds described here have more familiar names based on correlatives studied elsewhere. For examples, tephra 12 is correlative to the Mount St. Helens Cy tephra, also known farther south in the Great Basin as the Marble Bluff bed (Davis, 1985).
The youngest recognized tephra at Summer Lake, the Mazama tephra from Crater Lake, is present as pale yellow to brown lapilli and ash in surficial dune sediments that unconformably overlie the lacustrine deposits (Davis, 1985). The youngest water-lain tephra beds (D–A) and the youngest lacustrine sediments found along the Ana River are known only from the top of site E. Here, the sequence is inclined slightly to the west, and it is possible that even younger beds may be present to the west of the canyon exposures beneath surficial sediments. The oldest deposits exposed in canyon outcrop are found at the bottom of site C and adjacent areas. Here, tephra bed NN is present several centimeters below the surface of the Ana River (Figs. 6 and S4 [see footnote 2]; Table S1 [see footnote 1]).
Stratigraphic relations are not entirely certain for a small number of beds. Stratigraphic relations between JJ0.6 and JJ1, for example, are not known directly as they have not been found together (although based on their relative positions between beds JJ and KK as shown in Fig. 6, it is suspected that JJ0.6 is younger than JJ1). Similarly, the age of EE3 in relation to EE1 and EE2 is unclear, as EE3 was found only at site E, and EE1 and EE2 were observed only at site C.
Several tephra beds have conspicuous internal layering, including beds 2, GG, and JJ (Table 1). The bottom of tephra 2 typically consists of two 8–10-mm-thick upward-fining layers (Fig. S30 [see footnote 2]). These are overlain by multiple thin layers that consist of tephra mixed with a lesser amount of silt; this in turn is capped by a lighter colored, 2–3-mm-thick tephra layer. A possible explanation for this pattern is two pulses of the same eruption followed by a short time for reworking and redeposition followed by an additional eruption of the same magma. Tephra GG typically has a 1 cm white base that is overlain by alternating lighter and darker gray layers. At site C, the base of tephra JJ is a predominantly white, 1-mm-thick layer consisting of silt- to very coarse sand–sized pumiceous ash. Above this is a 2-mm-thick layer that contains a mixture of dark gray, medium- to coarse sand–sized grains and white, lower density grains of coarse to very coarse sand size. Above this is a 1-cm-thick layer of orange to dark gray color that consists largely of the darker tephra. The uppermost layer consists of 2 mm of silt-sized gray ash.
The host sediments for the Summer Lake tephra beds typically also contain significant tephra glass. This likely results from a combination of reworking together with the significant input of tephra to the basin from the Cascade volcanic arc, some of which is too disseminated to form visible beds. Pure samples were difficult to obtain for some of the thinner beds, and the resulting EPMA data sets contain a chaotic mixture of glass compositions (Table S4 [see footnote 1]). These mixtures are likely to represent a combination of primary deposition (the visible tephra bed) and reworked tephra (from the host sediments). In many cases, samples that appear to contain a significant reworked tephra component also contain some shards with elevated K2O and low Na2O values (Table S4 [see footnote 1]), features that are common in glass that has undergone alkali exchange (Shane, 2000). This altered glass may possibly derive from Miocene and Pliocene tuffs and tuffaceous sediments (Walker, 1963; Travis, 1977) exposed in the Winter Ridge escarpment that bounds the western side of the basin (Fig. 2).
Although the lake sediments consist predominantly of silt and clay, significant sand is present. This consists largely of ostracode valves, and several sand beds composed almost entirely of ostracodes are present (Tables S1–S3 [see footnote 1]). Also present in the sequence are several continuous or semicontinuous carbonate beds (tufas) and several beds with hard or soft carbonate nodules. A few of the more prominent tufas and sand beds are useful for correlation (Fig. 6).
Previous work (Davis, 1985; Davis, inErbes, 1996, and inNegrini et al., 2001; Negrini and Davis, 1992; Erbes, 1996; Langridge, 1998; Negrini et al., 2000) documented two significant unconformities in outcrop and core and suggested several additional minor unconformities or hiatuses. As these sometimes appear to vary laterally in their character (i.e., between disconformity and angular unconformity), we describe them in nonspecific terms. At site C, the main unconformity is marked by several centimeters of sand over a tufa bed and is located below tephra beds L and L1 at a depth of 5.83 m (Figs. 3, 6, and S1 [see footnote 2]; Table S1 [see footnote 1]). At site F, the same unconformity is marked by as much as 22 cm of cross-bedded sand on top of a thin layer of ooids and a discontinuous tufa located at a depth of ∼4.8 m below tephra bed L (Figs. 5, 6, and S1 [see footnote 2]; Table S3 [see footnote 1]). At site E, thick sand and tufa layers are absent, and the location of the equivalent unconformity is assigned to an abrupt transition at 6.83 m from finer silt and clay to coarser silt that is overlain by a thin ostracode sand (Figs. 4 and S1 [see footnote 2]; Table S2 [see footnote 1]). The same unconformity is also known from the WL core and is probably related to a lowstand during marine oxygen isotope stage 5 (Cohen et al., 2000; Negrini et al., 2000).
Davis (1985) described a second major unconformity at site F that is marked in our excavation by conspicuous cross-bedded sands and tufa below tephra bed T1 at a depth of 7.06 m (Figs. 5, 6, and S2 [see footnote 2]; Table S3 [see footnote 1]). Davis (1985) interpreted the five tephra beds that he found below this unconformity to be older than tephra NN, which is found at the base of site C, and so gave them designations SS–OO. Based on the field character of these beds in our deeper excavation and the tephra glass compositions, we reinterpret the beds below Davis's lower unconformity as correlative to tephra layers found at sites C and E (e.g., beds OO and W are equivalent) (Fig. 6). This removes most of the time gap associated with Davis's earlier interpretation.
A prominent thick sandy bed, but no tufa, is also present beneath bed T1 at site E at a depth of 9.54 m in a stratigraphic position equivalent to Davis's (1985) lower site F unconformity (Figs. 6 and S2 [see footnote 2]; Table S2 [see footnote 1]). At site C an apparently correlative sandy interval is present beneath bed T1 at 8.99 m, but is much less prominent (Figs. 6 and S2 [see footnote 2]; Table S1 [see footnote 1]). At site C, a distinct repeating paleomagnetic waveform continues across this interval and further argues against a significant gap in time (Negrini et al., 1994).
An additional unconformity was observed higher in the section at site E, where disrupted beds K and L are truncated by sandy sediments and a tufa at a depth of 6.45 m (Figs. 6 and S3 [see footnote 2]; Table S2 [see footnote 1]). Equivalent sand and tufa beds are also present 10–15 cm below bed J3 at sites C and F (at total depths of 5.06 and 4.05 m, respectively), so the same unconformity might exist there as well (Fig. 6; Tables S1 and S3 [see footnote 1]). A few centimeters higher and just below bed J3 is an abrupt transition from silty to sandy sediments that may mark a second unconformity in this interval (Figs. 6 and S3 [see footnote 2]; Tables S1–S3 [see footnote 1]). Davis (inErbes, 1996) and Erbes (1996) suggested a possible unconformity and soil horizon associated with sand and tufa beds ∼0.6 m below tephra bed LL1 at site C (Fig. 6; Table S1 [see footnote 1]).
Variations in lake level (Negrini et al. 2000; Cohen et al., 2000) are clearly apparent in the field. Deeper water intervals are associated with finer sediments that are often finely laminated and tend to develop a blocky texture. These deeper water sediments on occasion contain isolated pebbles that may represent dropstones, suggesting that deeper water intervals are times associated with colder climates. Shallower intervals are associated with lower clay contents and higher sand contents, and may be associated with tufas, unconformities, and sand beds containing ostracodes coated with carbonate and/or iron oxides.
CORRELATIONS AND AGES
Age control for the Summer Lake area sediments is derived from the correlation of tephra beds and paleomagnetic variations to other dated records plus a limited number of radiometric dates on the Summer Lake deposits (Table 2). On this basis, age-depth models have been constructed for the upper sediments (Negrini and Davis, 1992; Zic et al., 2002) and the entire outcrop sequence (Negrini et al., 2000). For the uppermost ∼5 m, the chronology is well constrained, but there is considerable uncertainty in the ages of the middle to lower portions of the outcrop sequence.
The uppermost correlated tephra is the Mazama tephra from Crater Lake, Oregon. Mazama tephra is present in surficial sediments that unconformably overlie the lake beds (Davis, 1985), and thus provides a minimum age for the youngest lacustrine sediments. Mazama tephra has a weighted mean radiocarbon age of 6730 ± 40 14C yr B.P. (7470–7620 calendar yr B.P.) (Hallett et al., 1997), and Zdanowicz et al. (1999) reported an essentially equivalent age of 7627 ± 150 cal yr B.P. based on identification of Mazama tephra in the GISP2 (Greenland Ice Sheet Project) core.
Other previously correlated units in the upper portion of the section include bed D (Mount St. Helens Mp); bed 18 (Trego Hot Springs tephra); bed E1 (Tulelake T2438); bed F (Wono tephra); bed G (9715K, found at but not originating from Newberry Volcano); bed 12 (Mount St. Helens Cy); 8, 6, and 4 (beds of the Pumice Castle set); and bed 2 (Ice Quarry tephra) (Davis, 1985; Rieck et al., 1992; Negrini et al., 2000; Kuehn and Foit, 2006) (Fig. 6; Tables 2, S2, and S3 [see footnote 1]). All of these beds except E1 and G have associated dates by 14C, thermoluminescence (TL), or K-Ar methods (Figs. 6–8; Table 3). The Trego Hot Springs and Pumice Castle beds originate from Crater Lake (Davis, 1985). The Ice Quarry tephra originates from Newberry Volcano (Kuehn, 2002; Kuehn and Foit, 2006). Beds G and F (Wono tephra) correlate to coarser deposits in the vicinity of Newberry Volcano to the north (Kuehn, 2002, Kuehn and Foit, 2006) (Fig. 1; Table 2).
The glass in bed B1 is similar in composition (SCs of 0.97) to proximal Llao Rock tephra fall deposits at Crater Lake (Table 2) that preceded the climactic Mazama tephra by 100–200 yr (Bacon and Lanphere, 2006). B1, however, is too old to directly correlate with the Llao Rock deposit, but the strong compositional similarity may be interpreted to suggest Mount Mazama (Crater Lake) as the likely source.
Bed I, newly analyzed herein, is very similar in glass composition (SCs to 0.99) to the much younger bed 12 (Mount St. Helens Cy) found 2–3 m above it (Table 2). It is also similar in composition to a series of tephra beds thought to be of Mount St. Helens origin that are known from Washington and Idaho and that date to ca. 70–120 ka (Busacca et al., 1992; Berger and Busacca, 1995; Bouchard et al., 1998; Whitlock et al., 2000) (Table 2). Based on glass compositions and estimated ages, the most likely potential correlatives to bed I are Carp Ash-8 and Carp Ash-9 from Carp Lake (Whitlock et al., 2000) (Fig. 1; Table 2) (SCs to 0.97 and 0.95, respectively) and the EMSH ash of similar age from sites KP-1 and WA-5 (Busacca et al., 1992) (Fig. 1; Table 2) (SCs of 0.94–0.95 to sample KP-1D). The EMSH ash, which is found in loess in eastern Washington State, is associated with a TL date of 83 ± 8 ka (Berger and Busacca, 1995) (Table 3), consistent with the position of bed I in the Summer Lake age-depth model in Negrini et al. (2000).
The next lower dated layer is bed N. Bed N and the much thinner bed M just above it were indicated by Davis (1985) to be above the major unconformity at sites C and F. Berger (1991) obtained a TL date of 102.3 ± 11 ka on bed N (Table 3), and this age and the stratigraphic context indicated by Davis (1985) were used in Negrini et al. (2000) to construct an age-depth model. However, new excavations at sites C and F failed to locate the M and N pair in Davis's (1985) indicated stratigraphic context. Neither are they noted at site C in Davis's composite section (inErbes, 1996; Negrini et al., 2001) or in Erbes (1996). A pair of beds matching Davis's (1985) description of M and N was found in the site E excavation, however, and it is from this locality that Berger (1991) obtained his dated sample.
A similar pair of beds (0.6–0.8 and 6 cm thick) was also sampled during reexamination of the WL core. These are equivalent to WL-37–1 and WL-37–2 in Negrini et al. (2000). They have the same thicknesses (Table S2 [see footnote 1]), glass composition (Table 2), and relative stratigraphic position as M and N at site E. They also were interpreted in Negrini et al. (2000) as being below the major unconformity that occurs below bed L in the WL core, below bed L at sites C and F, and probably also below bed L at site E (Figs. 6 and S1 [see footnote 2]; Tables S1–S3 [see footnote 1]). Reinterpreting beds M and N as being below the unconformity requires revision of the published age-depth model (discussed further herein), and also permits the correlation of bed N1 with bed N. Bed N1, located at site F, has the same glass composition (Table 2), the same thickness, and the same particle size range as bed N (Fig. 9; Tables S2 and S3 [see footnote 1]). Bed N1 also has the same stratigraphic relations with the unconformity above and beds O, P, and Q below (Fig. 6). Beds M and N are also similar in glass composition to Newberry tephra 9912D and thus are probably correlative (Kuehn and Foit, 2006) (Table 2).
A further implication of correlating N and N1 and reinterpreting the stratigraphic position of the associated unconformity is that it is now clear that beds OO–SS (described from beneath the unconformity at site F by Davis, 1985) need not be older than the base of the site C exposure. Of these beds, Davis (1985) reported a glass composition only for OO. Notably, this composition is very similar to that of bed W (Table 2), suggesting that W and OO are the same bed. In Kuehn and Foit (2006), W was identified at site C as a new bed W1, but the stratigraphic position relative to carbonate beds above it at site C and the relations shown in the composite measured section of Davis (inErbes, 1996; Negrini et al., 2001) suggest that this is, in fact, bed W of Davis (1985).
Beds Q, S, T, T1, and V are compositionally similar to tephra beds at Newberry Volcano and are likely correlative (Kuehn and Foit, 2006). The glass composition of bed Q is similar to that of Newberry tephra 9920C; S, T, and T1 are very similar in composition to Newberry tephra beds 984F and 984G5; and V is similar to Newberry tephras 978D and 0004F, which are tephra fall and pyroclastic flow deposits, respectively (Kuehn, 2002; Kuehn and Foit, 2006) (Table 2). Bed V has also been previously correlated to tephra bed T1193 found at 32.28 m in the Tulelake core (Rieck et al., 1992; Herrero-Bervera et al., 1994; Negrini et al., 1994) (Table 2). Consistent stratigraphic relations provide further supporting evidence for the Summer Lake–Newberry correlations. Beds Q, S, T, T1, and V at Summer Lake occur in the same stratigraphic order as the beds at Newberry Volcano that are interpreted as correlative (Kuehn, 2002).
Like bed I, bed W is similar in glass composition to bed 12 (Davis, 1985; Kuehn and Foit, 2006) (SCs of 0.97–0.98) and to the ca. 70–120 ka Mount St. Helens set C–like beds discussed herein (Table 2) (e.g., SCs of 0.96–0.97 to Carp Ash 8). The position of bed W at Summer Lake suggests that it is significantly older than the inferred ages of these beds and therefore not directly correlative. The strong compositional similarity together with the age relations suggests that bed W represents an earlier eruption of Mount St. Helens.
Glass in bed AA1 is very heterogeneous and is somewhat bimodal (Fig. 7A). The two end members average ∼56 and 69 wt% SiO2, and the overall average is ∼62–63 wt% SiO2 (Table 1). The reported mafic end-member composition was calculated using data points with SiO2 <57.3 wt%. The silicic end-member composition was calculated using data points with SiO2 >67.4 wt% and with CaO and FeO <2.9 and <4.3 wt%, respectively (Table S4 [see footnote 1]).
Glass in bed AA1 is strikingly similar to that in proximal Shevlin Park Tuff, a largely bimodal andesitic ash-flow tuff found primarily to the west of Bend, Oregon, and erupted from a vent located east of the Three Sisters (Conrey et al., 2002; Sherrod et al., 2004). Lanphere et al. (1999) reported 40Ar/39Ar ages from plagioclase of 260 ± 15 ka (plateau) and 255 ± 74 ka (isochron) for Shevlin Park Tuff (Table 3). Bed AA1 glass compositions are an excellent match to the compositional range, trend, and largely bimodal frequency distribution of Shevlin Park Tuff (Fig. 7A; Tables 2 and S4 [see footnote 1]). Similarity coefficients to proximal reference material are also good, with SCs as much as 0.97 for the silicic end member, to 0.97 for the mafic end member, and to 0.94 for the bulk composition. Most of the analyses of proximal Shevlin Park Tuff glass presented here are WSU data provided by R.M. Conrey (2002, personal commun.). When it was observed that this initial data set did not fully replicate the most mafic shards in bed AA1, two of the most mafic Shevlin Park clasts identified by X-ray fluorescence were obtained from R.M. Conrey and analyzed at the UA (Table S4 [see footnote 1]). The additional analyses from these clasts significantly improved the match between the Shevlin and AA1 data sets apparent on bivariate plots and illustrate the importance of obtaining sufficiently representative reference material.
Shevlin Park Tuff was previously correlated with Summer Lake beds JJ (Gardner et al., 1992; Gardner and Negrini, 2001) and NN (Conrey et al., 2001). Bivariate plots (Figs. 7A, 7B) clearly distinguish Shevlin and JJ. Both the end-member compositions and frequency distributions are substantially different. Bed NN glass is relatively homogeneous in composition, unlike that in Shevlin Park Tuff (Table 2).
Beds DD, EE, GG, and II were previously correlated to a series of beds preserved in lacustrine sediments at Pringle Falls, Oregon (Herrero-Bervera et al., 1994; Negrini et al., 1994) (Figs. 1 and 6). Glass in Summer Lake beds DD, EE, GG, and II are very similar in composition to Pringle Falls beds K, H, D, and S, respectively (Table 2). This set of beds also occurs in the same stratigraphic sequence at both locations.
Pringle Falls D (and thus Summer Lake GG) was also previously correlated to tephra PI-OR at Paoha Island in Mono Lake (Herrero-Bervera et al., 1994) (Fig. 1). In addition, bed GG is very similar in glass composition to Newberry tephra 9917C (Table 2), and therefore may have originated from Newberry Volcano (Kuehn and Foit, 2006). New bed EE2 is similar in glass composition to Pringle Falls bed E (SC 0.98; Table 2) and is likely correlative. Pringle Falls E was previously correlated to Paoha Island and Benton Crossing in California (Herrero-Bervera et al., 1994; Liddicoat et al., 1998, 1999) (Fig. 1). Whitlock et al. (2000) also noted a strong similarity between Pringle Falls E and Carp Lake Ash-14 (Table 2), but preferred an age model that precludes correlation.
Bed GG and its associated magnetic excursion provide the most precise age control point in the lower portion of the Summer Lake outcrop sequence. Herrero-Bervera et al. (1994) reported an 40Ar/39Ar plateau age on plagioclase of 227 ± 8 ka from the correlated Pringle Falls D bed. Singer et al. (2008) recalibrated this result to 198 ± 59 ka and 221 ± 10 ka for the isochron and plateau, respectively (Table 3). Singer et al. (2008) also dated an additional sample of Pringle Falls D plagioclase by 40Ar/39Ar and reported an age of 211 ± 6.4 ka (mean of 16 isochrons). GG and Pringle Falls D are also both closely associated with the Pringle Falls geomagnetic excursion (Herrero-Bervera et al., 1994; Negrini et al., 1994). It was proposed (Negrini et al., 2000; Negrini, 2002), based on the similarity of subsequent paleomagnetic secular variations, that the Pringle Falls excursion is correlative with a ca. 190 ka excursion found in marine sediment cores (e.g., Henyey et al., 1995). However, the balance of radiometric dating evidence suggests that, instead, the Pringle Falls excursion is an older excursion (ca. 220 ka) that has been dated at other locations around the world both with the 40Ar/39Ar method and marine oxygen isotope stratigraphy (e.g., Channell, 2006). The corroborating 40Ar/39Ar ages include an isochron age of 218 ± 7 ka on groundmass (weighted mean of six samples) from the Albuquerque Volcanoes, New Mexico (Singer et al., 2008), and an 40Ar/39Ar age of 227 ± 8 ka on the Mamaku Ignimbrite, New Zealand (Houghton et al., 1995; McWilliams, 2001). Berger (2001) reported a significantly younger TL age of 142 ± 33 ka for Pringle Falls bed D.
Initial EPMA analyses of bed JJ1 and proximal reference material for Bend Pumice analyzed on the same instrument on the same day produced essentially identical mean values, suggesting a potential correlation (see WSU results for samples C62 and 96–19, the first two entries associated with bed JJ1 in Table 2). Because of the greater standard deviations in the JJ1 data and dates indicating a much older age for Bend Pumice (Lanphere et al., 1999), additional analyses were obtained to further evaluate the possible correlation. The resulting much larger data set on a total of six samples (UA results in Table 2), analyzed consecutively on the same day on the same instrument and including additional reference material, instead revealed significant differences that preclude correlation. Bivariate plots clearly show that glass in the JJ1 sample from site C (C61) is bimodal (Fig. 7C; Table 2). The homogeneous Bend Pumice reference data plots between the two modes (Fig. 7C). The stratigraphically equivalent bed in the SPG-A core yielded only the higher silica mode (Table 2).
Bed KK has been previously correlated to the heterogeneous Antelope Well tuff (also known as Medicine Lake andesite tuff), which erupted from Medicine Lake volcano (Herrero-Bervera et al., 1994; Negrini et al., 1994). Donnelly-Nolan et al. (2008) regarded Antelope Well tuff as the single most important stratigraphic unit at Medicine Lake, as it forms the only widespread marker bed present at the volcano. To further test the correlation with bed KK, an Antelope Well reference sample, 194-M(2)A, was obtained from E. Wan (2007, personal commun.) at the Menlo Park campus of the U.S. Geological Survey and analyzed at WSU by F. Foit (2007, personal commun.). Additional analyses were obtained at the UA. Similar mean values (SC as high as 0.96) are generally supportive of the correlation between KK and Antelope Well tuff (Table 2), but the very similar distributions of glass compositions apparent on bivariate plots (Fig. 7E) are even more compelling and provide strong evidence for correlation. The plots demonstrate that KK and Antelope Well tuff have essentially identical major populations (range, trend, and frequency distribution), and both have similar trace populations of higher silica glass.
Bed KK was previously correlated with tephra in cores from Tulelake and Walker Lake (Herrero-Bervera et al., 1994; Negrini et al., 1994) (Fig. 1; Table 2). Negrini et al. (1994) and Sarna-Wojcicki et al. (2001) noted similarities between bed KK and the Wadsworth tephra bed of Davis (1978) (Table 2), but this latter correlation remains uncertain. Whitlock et al. (2000) also noted a similarity to Carp Ash-15 (Table 2), but preferred an age model that precludes correlation.
Several dates are available for bed KK and potentially correlative deposits. Berger (1991) reported a TL date of 201 ± 27 ka on glass in sample SML-21a from Summer Lake site C and a 201 ± 45 ka TL date from glass in sample JOD88B from the possibly correlative Wadsworth bed. Herrero-Bervera et al. (1994) reported 40Ar/39Ar plateau and isochron ages of 171 ± 43 and 149 ± 110 ka, respectively, for plagioclase obtained from pumice in proximal Antelope Well tuff. All of the ages above overlap at 1σ error. Older dates associated with the overlying bed GG (ca. 211 ka) favor a bed KK age toward the higher end of the error range associated with the dates discussed above. In contrast, Donnelly-Nolan et al. (2008) concluded that the age of the Antelope Well tuff is ca. 180 ka on the basis of a 40Ar/39Ar age, stratigraphic constraints from other (unspecified) dated units, and evidence that the tuff erupted through an ice cap on the volcano (Donnelly-Nolan and Nolan, 1986). One possible solution is an eruption age of 225–230 ka. This is consistent with most of the age estimates and with the constraint provided by the ice cap interpretation, because this age range is within a several-thousand-year-long glacial interval indicated by the marine oxygen isotope curve (Martinson et al., 1987). Although this older age is beyond the 1σ error range of the 171 ± 43 ka plateau date on the Antelope Well tuff, it is still within the 95% confidence interval. Alternatively, the glacial history of the Medicine Lake volcano as suggested by Donnelly-Nolan and Nolan (1986) may not be representative of time-averaged global ice conditions, a supposition that is consistent with the relatively low magnetic susceptibility of the Summer Lake sediments containing bed KK. Such magnetic properties appear to be indicators of shallow lakes during interglacial times (Negrini et al., 2000). A final alternative is that the correlation between bed KK and Antelope Well tuff is incorrect. This, however, would require the existence of an additional major eruption of broadly similar age, with essentially identical major element glass geochemistry, and no known proximal deposits. It would also require the Antelope Well tuff to have not left any recognizable bed at Summer Lake.
Bed LL was dated by TL. Sample SML-5 from bed LL at site C yielded a TL data of 160 ± 35 ka (Berger. 1991). Berger (2001) considered this to be a minimum age due to an unusual dose-response curve produced by the sample.
Bed NN, the lowermost bed in the outcrop sequence, is very similar in glass composition to homogeneous Newberry tephra units 9881C and Qdt-Qto (Table 2) and thus is probably correlative. Units Qdt and Qto are Newberry ash-flow tuffs that were mapped separately by MacLeod et al. (1995) and were later correlated on the basis of similar whole pumice compositions by J. Donnelly-Nolan (cited in Jensen et al., 2009; see also Kuehn and Foit, 2006), and on the basis of glass compositions (Kuehn, 2002). Newberry tephra 9881C is indistinguishable in glass composition from the Qdt-Qto flow and is likely the tephra fall equivalent (Kuehn and Foit, 2006). We consider earlier correlations of bed NN with Shevlin Park Tuff (Conrey et al., 2001) and with a preceding tephra fall preserved at Columbia Canal (Columbia Canyon) (Sarna-Wojcicki et al., 2001) to be less likely. Shevlin Park Tuff is strongly heterogeneous and is an excellent match to bed AA1. Columbia Canal glass is close to NN on most elements, but the reported abundance of K2O is significantly lower (Table 2). Some constraint on the age of bed NN is provided by a 40Ar/39Ar age of ca. 300 ka on plagioclase from Qdt and Qto (Donnelly-Nolan et al., 2004).
Although a large number of mafic tephra beds is preserved at Summer Lake, some of them containing sand-sized grains, none of them has been correlated to specific source vents outside of the basin. In the Fort Rock–Christmas Lake basin, ∼40–60 km to the north, are numerous maars, tuff cones, and cinder cones that are regarded as Pliocene–Pleistocene in age (Heiken, 1971). Although these are the closest potential sources, their ages are poorly constrained, and it is uncertain how many of them might be late Pleistocene in age. In contrast, late Pleistocene mafic vents are well known in the Cascade arc (e.g., Bacon and Lanphere, 2006; Hildreth, 2007). The closest Cascade vents are ∼90–120 km to the west of Summer Lake (Fig. 1), and are located generally up wind. Cascade arc sources are thus likely to be major contributors of mafic tephra to the Summer Lake basin.
REVISED AGE-DEPTH MODEL
A revised age-depth model is presented in Figure 10 and Table 4. Above the major unconformity (∼6.5 m), the model is largely unchanged from that of Negrini et al. (2000) and Zic et al. (2002). In this part of the section, previously discussed glass compositions and the age model suggest that bed B1 originates from an eruption of Mount Mazama (Crater Lake) ca. 20 ka (Table 4). Bacon and Lanphere (2006) reported proximal eruptive activity from this time in the form of several rhyodacite domes with 40Ar/39Ar isochron ages of 24 ± 3 and 18 ± 4 ka. The existence of this corresponding activity at Crater Lake suggests that both the correlation and the inferred ages of the uppermost sediments are accurate. The 83 ± 8 ka age of the EMSH ash from Mount St. Helens (Berger and Busacca, 1995), a deposit that is potentially correlative with bed I, is also consistent with the earlier chronologies suggested for the Summer Lake sediments above the major unconformity.
We propose changes to the previously published age model, below the major unconformity, based primarily on our new interpretation that tephra N and N1 are the same bed, on a new correlation of bed AA1 with the Shevlin Park Tuff, and on new dates (Singer et al., 2008) associated with correlatives of bed GG and the Pringle Falls geomagnetic excursion.
We have shifted the preferred chronology below the unconformity toward older ages to better honor the numerous and convergent radiometric ages now associated with the correlative tephra of bed GG and the Pringle Falls geomagnetic excursion. The new chronology also honors the radiometric ages of the Shevlin Park Tuff and Antelope Well tuff, though at the extremities of their 1σ uncertainty estimates. (See the related discussion herein on the Antelope Well–KK correlation.) The resulting chronology is a plausible, albeit poorly constrained, age model for the lower two-thirds of site C using the limits of the single standard deviation error bars for the 40Ar/39Ar ages on beds GG (Pringle Falls D) and KK (Antelope Well tuff) as the key control points. The resulting model has an average sedimentation rate of 16 cm/ka and suggests approximate ages of 190 ka for W, 196 ka for Shevlin Park Tuff, 215 ka for Antelope Well tuff, and 218 ka for bed LL (Fig. 8; Table 3). These age estimates are compatible with the isochron age for Shevlin Park Tuff (Lanphere et al., 1999), the TL date on bed R (Berger, 1991), and the interpretation of the TL date on bed LL as a minimum age (Berger, 2001). The model does not, however, fit the TL date on bed N. The N date is significantly younger than that for bed R below it, and lacking any sedimentological evidence for an unconformity between these beds, we suggest that the date on N should be interpreted as a minimum age.
Projecting the model to the base of the section suggests an age of 240–250 ka for bed NN. Considering the significant uncertainties inherent in this model, the suggestion by Davis (inErbes, 1996; Negrini et al., 2001) and Erbes (1996) of a possible unconformity below bed LL, and the 300 ka approximate age of tephra correlated to bed NN, the base of the sequence could be older.
REPEATED ERUPTIONS FROM INDIVIDUAL SOURCES AND PETROLOGIC INTERPRETATIONS FOR SELECTED BEDS
At Summer Lake, there are many examples of tephra beds that are very similar to others above or below them in the sequence. In many cases, these geochemical similarities can be interpreted to suggest repeated eruptions from the same volcano. Using the overall age-depth model, the relative timing of these events can be considered. In addition, a number of deposits show patterns of heterogeneity that suggest the involvement of zoned magma chambers and/or multiple magmas. Many examples of both are outlined here, largely in order from older to younger beds.
The compositional heterogeneity observed in bed KK (Fig. 7E) could potentially be explained by either zoned or multiple magmas. Although it is difficult to determine which from glass compositions of individual shards alone, the nearly equal frequency distribution along the main trend and smaller compositional range (e.g., compared to AA1 and JJ) could be construed as evidence for a zoned chamber. Two diffuse beds, JJ0.6 and JJ0.4, ∼40 and 60 cm above KK, contain glass that spans the same compositional range as bed KK. We interpret these beds to be the result of redeposition rather than additional eruptions because erosion and redeposition of the thick underlying KK bed are thought to be more likely than the occurrence of two additional eruptions with exactly the same pattern of heterogeneity.
Glass in bed JJ is strongly heterogeneous with two relatively abundant end-member compositions and fewer data points between them (Fig. 7B). On a plot of MgO versus SiO2 (right side plot of Fig. 7B), the major mafic component plots off of the trend formed by the silicic end member and intermediate data points. This pattern could be modeled using a system with three end-member magmas. An alternate model consists of a homogeneous mafic magma and a zoned intermediate to rhyolitic magma. Three additional beds have glass compositions that are similar to the silicic end member of JJ. These include II1, located ∼4 cm above JJ, and JJ0.2, located ∼6–8 cm below JJ. II1 and JJ0.2 and glass compositions are indistinguishable from the silicic end member of JJ (Fig. 7B). Assuming a constant sedimentation rate, the age model suggests that bed II1 may be ∼200 yr younger than JJ and that JJ0.2 may be ∼400 yr older. The identical compositions and short time interval are compatible with origin of II1, JJ, and JJ0.2 from three eruptions of the same magma chamber, only one of which also involved the eruption of mafic magma. Bimodal bed JJ1, located ∼33 cm below JJ and estimated to be older by ∼3 k.y., contains glass compositions that plot together with all 3 of the aforementioned beds (Figs. 7B and 7D). The apparent absence of the lower silica component in the stratigraphically equivalent sample from the SPG-A core (Table 2) suggests that the two modes may represent two different eruptions. The small compositional difference between modes could have originated by tapping different portions of a variably evolved magma chamber.
On most bivariate plots, bed AA1 and proximal Shevlin Park Tuff appear to have two prominent end members, a more tightly clustered silicic end member and a somewhat broader mafic end member with a lower abundance of data points between end members (Fig. 7A). The relative lack of intermediate data points suggests magma mixing. On some plots (e.g., K2O versus MgO, not shown) there appear to be two mafic components, thus suggesting the involvement of three magmas in the eruption. Conrey et al. (2001) came to a similar conclusion based on a more comprehensive set of petrologic data.
Glasses in beds HH and II are very similar, being distinguished only by small offsets in SiO2, Al2O3, and CaO. On this basis, it is possible that these derive from the same source. Their relative depths indicate an interval of ∼250 yr between them (Table 4).
The glass compositions of beds S, T, and T1 are very similar in composition (Table 2). Thus, these beds could have been erupted from the same magma system. Bed S and T glasses are indistinguishable. The somewhat older T1 is shifted slightly toward higher SiO2 and lower CaO and FeO [Table 2; cf. sample E52 (T) with E53 (T1) and C35 (T) with C36 (T1)]. The spacing between beds S and T suggests ∼500 yr between them. T and T1 are closer together and probably differ in age by ∼200 yr. The silicic glass in bed T0.5 is indistinguishable from T1, but it is unclear whether this bed represents an additional event or simply redeposited tephra T1.
Several pairs of compositionally indistinguishable beds are above S and T, and each of these may also represent a pair of closely spaced eruptions from the same source magma. Included here are beds P and Q, which are probably separated by ∼120 yr. Compositionally indistinguishable beds M and N are separated by ∼3 cm of sediment, which corresponds to ∼200 yr. Glass in bed H0.2 is indistinguishable from that in bed 8 found 30–35 cm below it. The age model suggests that they differ in age by ∼6 k.y. (Table 4).
Bed F, correlated to Wono tephra, is heterogeneous in composition with a 3–4 wt% range in silica (Tables 1, 2, and S4 [see footnote 1]). F and Wono lack clearly defined end members; rather, there is a continuous trend with a somewhat higher frequency of data points closer to the less silicic end (Kuehn, 2002). Bed F1, which is ∼1–4 cm lower and is perhaps 100–200 yr older, is compositionally similar to F and Wono tephra. Bed G, which is ∼3 cm below F1 and is perhaps 400 yr older, is relatively homogeneous in composition, but it plots on the trend line defined by F and Wono tephra compositions. A few Wono clasts found at Newbery Volcano also contain a small fraction of glass with the same composition as tephra G (Kuehn, 2002). These geochemical relationships imply that G, F1, and F (Wono) derive from the same zoned magma system. Although the exact source these tephras is unknown, Wono tephra has a northward-coarsening particle size distribution (Davis, 1985; Kuehn, 2002) and occurs on the northwest flank of Newberry Volcano as a 63-cm-thick deposit containing pumice as much as 4 cm in diameter. This distribution pattern suggests an eruptive source somewhere in the general vicinity of the Three Sisters. Beds B and A also contain glass that is similar in composition to bed F, and these may record additional activity at the same source that produced Wono tephra. Bed B is located 1.4 m above bed F and is ∼10 k.y. younger. Bed A is 10 cm above bed B and is ∼1.6 k.y. younger than B. Bed A contains multiple populations of glass, and it is possible that it contains reworked glass from bed B.
Collectively, the Summer Lake outcrop and core records represent one of the most complete and detailed tephrostratigraphic records of Pleistocene volcanism in western North America. As such, this sequence serves as a key reference locality that puts together in a single stratigraphic context multiple beds that are known from many different locations. Thus it serves a critical role in refining the Pleistocene tephrostratigraphic framework for western North America. The expanded catalog of tephra beds provided herein also offers, for example, the potential to tie together paleoclimate records of the Great Basin more precisely than previously.
At least 88 visible tephra beds, 25 of them newly described, are preserved in 18.4 m of lacustrine sediments exposed in outcrop sections along the Ana River. Thicknesses of these beds range from <1 mm to >10 cm. Half of the tephra beds, 44, are characterized by rhyolitic glass, 40 beds contain predominantly basaltic or intermediate glass, and 4 are characterized by wide ranges in glass composition; 23 beds are known from localities outside of the basin, and 20 of these have been connected to specific source volcanoes, including Mount St. Helens, Newberry Volcano, Crater Lake, and Medicine Lake volcano. The remaining 65 beds are as yet known only at Summer Lake and therefore provide a unique record of Cascade arc volcanism. The large number of mafic tephra beds present at Summer Lake, close to half of the total, indicates that mafic eruptions contribute significantly to the distal ash fall produced by the Cascade arc. Some of these mafic eruptions were also sufficiently powerful to disperse sand-sized tephra ∼100 km downwind. Evidence of repeated eruptions of individual sources, many of which appear to be spaced a few hundred years apart, provides useful information about frequencies of pyroclastic eruptions in the Cascade arc.
Two of the compositionally heterogeneous tephras present in the sequence have been correlated to major ash-flow tuff deposits found outside of the basin (AA1 to Shevlin Park Tuff and KK to Antelope Well tuff). By obtaining well-representative proximal reference samples, geochemically analyzing them using the same procedures applied to the Summer Lake tephras, and carefully comparing the data using bivariate plots, robust correlations have been established. Rather than providing a hindrance, the observed heterogeneity in these deposits significantly strengthens the correlations, as it is unlikely that additional eruptions would so closely replicate all of the patterns observed in the data, including the means, compositional ranges, covariation trends, and frequency distributions of different glass compositions.
Our refined stratigraphy includes a single major unconformity associated with marine oxygen isotope stage 5e, and there are probably several minor unconformities present as well. From 0 to ∼6 m depth, above the major unconformity, the age-depth model is relatively well constrained. Age-depth relations for the rest of the section have significant uncertainties, and additional work is critically needed to provide better age control for the lower two-thirds of the sequence. It is possible, however, to construct a reasonable, but still significantly uncertain, age model using the available dates. Our revised age model increases the ages for the lower two-thirds of the sequence by ∼20 k.y., and suggests approximate ages of 190 ka for bed W, 198 ka for bed AA1 (Shevlin Park Tuff), and 215 ka for bed KK (Antelope Well tuff), although possibly conflicting age control for the Antelope Well tuff adds uncertainty to our revision. The estimated age of bed NN at the base of the section is 240–250 ka, but given the limited age control the actual age could be significantly greater.
Tephra studies at the University of Alberta were supported by a Natural Sciences and Engineering Research Council Discovery grant and an Alberta Ingenuity New Faculty Award to Duane Froese. The School of Earth and Environmental Sciences at Washington State University (WSU) provided instrument time for electron probe microanalysis (EPMA) of the site C and Summer Lake Playa site G (SPG) core samples. The Department of Chemistry at California State University, Bakersfield, provided cold room storage for the SPG, WL (Wetland Levee), and B&B (Bed and Breakfast) cores. Elmira Wan supplied a reference sample of the Antelope Well tuff, and Franklin Foit provided an analysis of the sample by EPMA at WSU; Foit also provided EPMA data for proximal Llao Rock tephra. Richard Conrey provided EPMA data for bed JJ and proximal Shevlin Park Tuff. Marty St. Louis of the Summer Lake Wildlife Refuge arranged for the permits necessary to conduct field studies along the Ana River. Bill Cannon of the U.S. Bureau of Land Management, Lakeview District, conducted the archaeological examination of the field site needed for the permit process. Careful and comprehensive comments by two anonymous reviewers led to important improvements in the manuscript.