INTRODUCTION
High-resolution and high-precision age models for proxy records are required to understand the teleconnections and feedbacks in the global climate system, including abrupt forcings such as volcanic eruptions (Lavigne et al. Reference Lavigne, Degeai, Komorowski, Guillet, Robert, Lahitte, Oppenheimer, Stoffel, Vidal, Surono, Pratomo, Wassmer, Hajdas, Hadmoko and de Belizal2013) and changes in sea ice cover (Denton et al. Reference Denton, Alley, Comer and Broecker2005; Li Reference Li2005). Without good chronologies, each climate record stands alone, and analyses of regional and global patterns are limited. High-precision correlations are especially important in assessing the realistic nature of climate model predictions, because models are much better at producing patterns in space than time series at individual locations. However, paleoclimatic reconstructions are typically time series measured at a single location, and are relatively isolated when age models are of low resolution. This weakness is only highlighted by the increasing number of very high-resolution proxy reconstructions, such as those from scanning XRF and paleomagnetic U-channel analysis (Grimm et al. Reference Grimm, Donovan and Brown2011; Vigliotti et al. Reference Vigliotti, Channell and Stockhecke2014; Noble et al. Reference Noble, Ball, Zimmerman, Maloney, Smith, Kent, Adams, Karlin and Driscoll2016). Great progress in regional and global reconstructions has been made in the last several decades using highly precise uranium-series age models on speleothems (Wang et al. Reference Wang, Cheng, Edwards, An, Wu, Shen and Dorale2001, Reference Wang, Cheng, Edwards, Kong, Shao, Chen, Wu, Jiang, Wang and An2008; Zhang et al. Reference Zhang, Cheng, Edwards, Chen, Wang, Yang, Liu, Tan, Wang, Liu, An, Dai, Zhou, Zhang, Jia, Jin and Johnson2008; Cheng et al. Reference Cheng, Edwards, Sinha, Spotl, Yi, Chen, Kelly, Kathayat, Wang, Li, Kong, Wang, Ning and Zhang2016), and tree-ring records provide (sub-) annual-precision records over the last 1000 to 2000 yr (e.g., Cook et al. Reference Cook, Seager, Cane and Stahle2007). However, for regions with few suitable speleothems, such as the western United States (Asmerom et al. Reference Asmerom, Polyak and Burns2010; Lachniet et al. Reference Lachniet, Denniston, Asmerom and Polyak2014), a well-dated network of terrestrial reconstructions may be lacking for periods predating tree-ring records.
High-precision, accurate chronologies are especially elusive in terrestrial climate records from archives such as lakes and meadows. Many lakes of interest in the western United States and other arid regions are climatically sensitive because they are hydrologically closed, but have little terrestrial vegetation to produce macrofossils or charcoal, and scant run-off to carry these materials into the lake (Benson et al. Reference Benson, Burdett, Kashgarian, Lund, Phillips and Rye1996, Reference Benson, Liddicoat, Smoot, Sarna-Wojcicki, Negrini and Lund2003; Placzek et al. Reference Placzek, Quade and Patchett2006; Kirby et al. Reference Kirby, Lund, Patterson, Anderson, Bird, Ivanovici, Monarrez and Nielsen2010). Other lakes are excellent recorders of glacial activity, a powerful indicator of seasonal temperature, but sit at elevations or latitudes where vegetation is sparse, and thus also lack macrofossils (Clark and Gillespie Reference Clark and Gillespie1997; Crann et al. Reference Crann, Patterson, Macumber, Galloway, Roe, Blaauw, Swindles and Falck2015; Xu et al. Reference Xu, Jiang, Yu, Yang and Chen2015). Although bulk sediment is sometimes used in the absence of other materials, such sampling homogenizes all the carbon-bearing materials in a lake, including dissolved, particulate, organic, inorganic, recycled, and freshly eroded carbon, and thus the resultant radiocarbon (14C) dates are of unknown accuracy. On the other hand, pollen is produced abundantly by most terrestrial plants and is spread widely by wind transport.
Pollen has long been used as a proxy indicator of vegetation change and, since the development of accelerator mass spectrometry (AMS) for 14C in the late 1980s, a number of attempts have been made to develop methods to isolate pollen for 14C dating (Brown et al. Reference Brown, Nelson, Mathewes, Vogel and Southon1989, Reference Brown, Farwell, Grootes and Schmidt1992; Long et al. Reference Long, Davis and de Lanois1992; Mensing and Southon Reference Mensing and Southon1999; Neulieb et al. Reference Neulieb, Levac, Southon, Lewis, Pendea and Chmura2013). None of the methods tried have been widely adopted, because either specialized equipment or extensive time was required. Most pollen dates reported in the literature are therefore still a combination of pollen and micro-charcoal, plant detritus, and amorphous organic material. The pollen is produced by vegetation with carbon directly from the atmosphere and (presumably) quickly deposited in the place where it is preserved, while the other materials all bear carbon of some indeterminate age (Piotrowska et al. Reference Piotrowska, Bluszcz, Demske, Granoszewski and Heumann2004; Munroe Reference Munroe2007; Li et al. Reference Li, Wang, Morrill, Anderson, Li, Zhang and Zhou2012; Lozano-García et al. Reference Lozano-García, Ortega, Roy, Beramendi-Orosco and Caballero2015).
Recent application of flow cytometry, a technique developed to sort biological cells and organisms, demonstrates the potential for efficient, complete purification of pollen from sediments (Byrne et al. Reference Byrne, Park, Ingram and Hung2003; Tennant et al. Reference Tennant, Jones, Brock, Cook, Turney, Love and Lee2013). This approach is potentially very powerful for improving terrestrial chronologies, but its application to real sediment sequences raises a number of questions: How short and variable is the time from production of pollen to deposition in the sediment? What is the potential for contributions from pre-aged pollen on the landscape, or erosion of ancient pollen in surface deposits? Can different sources of pollen be discriminated by sieving or by sorting parameters? What is the most efficient protocol for concentrating the pollen before cytometric sorting, and cleaning afterward? What kinds of variations in protocol may be required for different kinds of sediments?
As a preliminary exploration of these and other questions, we have applied cytometric sorting to pollen for AMS dating in a short sequence of sediment from Mono Lake, California. Located on the eastern flank of the Sierra Nevada (38.0°N, 119.0°W), on the western edge of the arid Great Basin region (Figure 1), Mono Lake is a climatically sensitive lake with potential for rich proxy information. A well-dated decadal-resolution absolute lake level curve for the last 2000 yr was developed by Stine (Reference Stine1987, Reference Stine1990a), taking advantage of the artificial reduction in lake level during the 1980s. Sediment cores from the lake can help to fill gaps in the curve, as well as provide a lake history before 2000 yr BP, if sufficient chronologies can be developed.
Location and digital elevation map of the Mono Lake basin in the eastern Sierra Nevada, on the western edge of the Great Basin. More than 75% of the modern inflow to the basin comes from the main Sierran streams, fed by rain and snowmelt from the mountains. Eruptions of the Mono-Inyo Craters to the south of the lake are the source of the North Mono-Inyo and South Mono tephras. Deep-lake silts of the Pleistocene cover much of the lake plain and late Pleistocene shorelines appear on the map as bathtub rings around the eastern edges of the basin on the DEM. In the modern climate little to no surface run-off drains the north, east, and south parts of the basin, but run-off from these areas may have been greater in the past (Zimmerman et al. Reference Zimmerman, Hemming, Hemming, Tomascak and Pearl2011). The BINGO-MONO10-4A-1N core site is marked by the black star, at 2.8 m water depth (1942.3 m above sea level [m asl]); maximum lake depth is ~1897 m asl, off the southern shore of Paoha Island (Scholl et al. Reference Scholl, Von Huene, St.-Amand and Ridlon1967).
One such core, BINGO-MONO10-4A-1N, records the last>10,000 yr in multiple proxies (Zimmerman et al. in revision) but over that period has an average uncertainty of±325 yr (95% confidence interval) in the age model (Figure 2). The age model is based on 14C dating of macrofossils and the identification of two tephras, the North Mono-Inyo and South Mono tephras. These tephras are comprehensively described from exposures around the basin (Sieh and Bursik Reference Sieh and Bursik1986; Bursik et al. Reference Bursik, Sieh and Meltzner2014) and their ages determined by 14C dating of leaves and twigs from bushes killed by the eruptions (Bursik and Sieh Reference Bursik and Sieh2013). The end of the North Mono-Inyo sequence is further constrained by a dendro-chronological age on the last phase of the eruption, of late summer AD 1350 (Millar et al. Reference Millar, King, Westfall, Alden and Delany2006). Pollen separated from the sediment between these tephras should therefore date to between 1345±55 and 615 +30/–15 cal BP, unless it has been stored on the landscape or altered by chemical preparation.
Bayesian age model for full BINGO-MONO10-4A-1N core; 95% confidence envelope is as narrow as 59 yr near dates, but expands to as much as 1480 yr in intervals between dates (arrows). Sub-plots at top are diagnostics of the Bacon age model and show (left) the distribution of the Markov Chain Monte Carlo (MCMC) iterations produced by Bacon, with a lack of structure indicating a stationary distribution; (center) the prior (solid gray line) and posterior (light gray histogram) distributions of the accumulation rates; and (right) prior and posterior distributions of the memory, a measure of the variability of the accumulation rate between neighboring depths in the core.
METHODS
Twelve samples were collected from between the South Mono and North Mono tephras in core BINGO-MONO10-4A-1N (hereafter: BINGO/10-4A), which is archived at the National Lacustrine Core Repository at the University of Minnesota (LacCore). Previous work (Davis Reference Davis1999) suggests that pollen concentrations in this interval at Mono Lake may be as high as 50,000 grains per cm3 of sediment. To balance stratigraphic resolution with sufficient pollen for dating, samples were 1 cm in the depth direction, and approximately 2 cm3 in volume. Pollen was concentrated at LacCore using the standard protocol for pollen intended for 14C dating (Figure 3).
Flow diagram of pollen concentration method used at the National Lacustrine Core Repository (LacCore) for pollen intended for 14C dating.
This protocol differs from the standard LacCore pollen prep for palynological studies in the following ways: baked glassware is used in place of organic carbon-based utensils (e.g., wooden stirrers, plastic tubes); samples are sieved to isolate the 20–118 micron fraction, rather than the 7–160 micron fraction; new heavy-liquid solution (LST – lithium heteropolytungstate) is used instead of hydrofluoric acid to remove silicates; resistant organic matter is removed with bleach rather than acetolysis; and no ethanol or tertiary butyl alcohol are used. The goal of this 14C-dating-optimized protocol is to isolate the sample from any other carbon-bearing materials, to prevent contamination of the carbon isotope signature of the pollen. Other 14C pollen preparation methods include use of hydrofluoric acid (Brown et al. Reference Brown, Nelson, Mathewes, Vogel and Southon1989, Reference Brown, Farwell, Grootes and Schmidt1992) but this is avoided in the LacCore protocol due to the use of prebaked glassware. The 12 samples were divided into two batches for the pollen concentration procedure, batch 1036 (8 samples) and batch 1037 (4 samples). All preparation steps were the same, except that batch 1036 was boiled in 0.4 N bleach for 3 min after nitric acid treatment, while batch 1037 was not, causing the 1036 samples to be white in color, while the 1037 samples were tan.
After concentration, the samples were sent to the Indiana University-Bloomington Flow Cytometry Core Facility (FCCF) for sorting by flow cytometry. Because the prepared pollen is in the 20–118 micron size fraction, the COPAS (Complex Object Parametric Analyzer and Sorter) Select machine was used. This sorter is specially designed to sort objects between 20 and 400 microns by optical density, size, and fluorescence. All samples were sorted using ultrapure water, and a minimum goal of 10,000 grains per sample was set, with 20,000–30,000 as an ideal, to strike a balance between sorting time and sufficient carbon for AMS analysis. Previous work suggests that this is enough pollen from western U.S. Holocene lake sequences to yield 80–150 micrograms of carbon, a small but feasible amount for a 14C analysis at CAMS.
The FCCF is a central facility and receives many biological samples from a wide variety of projects, potentially including materials exposed to tracer 14C. Tracer solutions contain tens to hundreds of times the natural level of 14C and are extremely powerful tools in biological experiments, but even a small amount presents a severe contamination risk to natural-level samples. To protect the sorted samples and the CAMS lab from potential tracer contamination, the COPAS sorter was swiped to test for tracer-level 14C just before each sorting session. Swipes were sent ahead to CAMS and measured separately; no tracer-contaminated swipes have been detected thus far.
14C analyses were done at the Center for AMS at Lawrence Livermore National Laboratory. Because of the extensive prior preparation, at CAMS the samples were only rinsed three times with high-purity MQ water, centrifuging between rinse steps, and combusted and graphitized according to standard procedures (Vogel et al. Reference Vogel, Southon, Nelson and Brown1984). AMS analysis followed routine protocols, and all dates are reported in conventional 14C yr with 1-sigma uncertainty (Stuiver and Polach Reference Stuiver and Polach1977). Dates for the tephras follow the weighted-mean 14C ages reported by Bursik and Sieh (Reference Bursik and Sieh2013); 555±40 14C BP and 1440±40 14C BP, respectively. Individual 14C dates (for both pollen and tephras) were calibrated using the IntCal13 calibration curve in the Calib 6.0 program (Stuiver and Reimer Reference Stuiver and Reimer1993; Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Freidrich, Grootes, Guilderson, Haflidason, Hajdas, Hatte, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013), and are reported as the median probability age with 2-sigma uncertainties (Table 1).
Pollen concentrations, radiocarbon dates, and calibrated ages.
*median probability as calibrated with Calib 6.1/IntCal09
calibrated ages and 2-sigma uncertainty rounded to the nearest 5-year
t.s.= too small; insufficient CO2 to make graphite
Bayesian age models for the core interval studied were developed in the Bacon program (Blaauw and Christen Reference Blaauw and Christen2011). The sediment shows undisturbed bedding and clear but minor variation of sediment type, and so we used the prior assumptions of superposition and increasing age with depth. Because we have 8 dates over only 30 cm, we allowed Bacon to use a 1-cm thickness, allowing the maximum flexibility; all other settings were the Bacon defaults. The assumption of superposition allowed Bacon to eliminate the tails of the 14C probability density functions where they result in age reversals, reducing the possible age range for any depth interval.
RESULTS
Although the bulk samples were all ~2 cm3, they yielded highly variable amounts of pollen (Table 1). Previous pollen work shows that pine dominates the Holocene assemblage at Mono Lake (average 50–70% of total; Davis Reference Davis1999), and this is reflected in smear slides of the concentrates. The smear slides also showed that the 20–118 micron fraction of the samples contained variable amounts of charcoal, silicate minerals, volcanic glass, and unidentifiable organic matter (Figure 4). In a conventional pollen concentrate, the silicates and glass are unimportant to the 14C date because they do not contain carbon, but the charcoal and organic matter would contribute carbon of an unknown age to the 14C measurement, potentially changing the age significantly. The bleach step did not appear to cause any difference in the behavior of batches 1036 and 1037; no offset in the 14C ages was observed, nor a systematic difference in initial purity or sorting time, indicating that the bleach step did not improve the purity of the sample in this case.
Photomicrographs of sample 1037-1 (43–44 cm): (a) after chemical pretreatment at LacCore, pollen is abundant, but charcoal and other organic material are also abundant; (b) after sorting by flow cytometry, sample is nearly pure pine pollen.
The sorting times varied from 2 to 8 hr because of differences in the size of the samples and the fraction of the material that was pollen (Table 1). Three samples yielded <1500 grains, and were not submitted for 14C dating. Two samples had extremely high pollen concentrations, with pollen grains representing 18–20% of the grains sorted, and yielded 30,000 and 54,000 grains in a few hours of sorting. The other seven samples produced 9000–17,000 grains each, with pollen concentrations of 1–2%. Sample 1037-1 had a particularly high pollen concentration, and was sorted to 54,000 grains, in order to split the sample into two aliquots for replicate 14C dates.
The 14C ages of the pollen samples are generally in stratigraphic order within their calibrated ranges (Figure 5). However, the pollen separates were all relatively small, yielding between 30 and 110 micrograms of carbon. One sample (1037-2) did not graphitize properly, and is not discussed further. Samples of less than ~300 micrograms of carbon require a background correction scaled to the sample size (Zimmerman et al. Reference Zimmerman, Steponaitis, Hemming and Zermeño2012) and so the analytical uncertainties were larger than desired, between±40 and±170 yr, 1-sigma (Table 1). Unfortunately, the combination of large analytical uncertainty and the shape of the calibration curve in this interval led to large uncertainties in the calibrated ages. Most of the calibrated ages had precisions of±150–250 yr (2-sigma), but the smallest sample had a calendar-calibrated precision of±445 yr.
Probability density functions (PDFs) of calibrated ages of pollen dates and North Mono-Inyo (NMI) and South Mono (SM) tephras, from Calib 6.0, using IntCal13 (Stuiver and Reimer Reference Stuiver and Reimer1993; Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Freidrich, Grootes, Guilderson, Haflidason, Hajdas, Hatte, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013). Note that the vertical axis is only to indicate the depth of each sample, and is not a linear scale. The amount of carbon in each pollen AMS sample is indicated to the right; samples close to 0.1 mg mostly reflect the complexity of the calibration curve (e.g., 32.5 cm), while PDFs of dates with <0.05 mg are smeared by the large AMS uncertainty (e.g., 52.0 cm).
The ages of the North Mono-Inyo and South Mono eruptions provide a basis for judging the accuracy of the pollen ages. The bottommost sample, just above the South Mono tephra, has a 14C age of 1920 (± 110) 14C yr, 480 yr older than the age of the tephra (1440±40 14C yr BP). The topmost pollen sample, just below the North Mono tephra, has a 14C age of 960 (± 100) 14C yr BP, 405 yr older than the tephra (555±40 14C yr BP).
A Bacon model for this core interval using only the tephras (without the pollen dates) (Table 2 Model 1, Figure 6A) results in a mean 95% confidence interval of 227 yr, with a minimum close to the younger tephra, and a maximum range near the middle of the sequence. A Bacon age model based only on the pollen (Table 2 Model 2; Figure 6A) is overall more precise, because the larger number of dates constrains the envelope of uncertainty better, but it is less accurate. Subtracting 400 14C yr from the age of sample 1036-1 stratigraphically below the North Mono-Inyo tephra makes it slightly older than that tephra, as it should be. Adopting that offset to adjust the other pollen dates yields a model with a mean 95% confidence interval of 158 yr, more precise than the tephras alone (Table 2 Model 3, Figure 6B).
Bacon age models for datasets discussed here: A: tephras only (left curve), and pollen only (right curve); B: tephras+pollen-400 14C yr; C: tephras+pollen-400 14C yr, with hypothetical 35-yr uncertainties.
Mean, minimum, and maximum widths in years, of 95% confidence interval for Bacon models.
DISCUSSION
The value of pollen dating in building lacustrine age models is two-fold: first, to provide dates at a desired depth in a sediment sequence, for example, across a significant change in a proxy record; and second, to provide the most precise dates possible, by adjusting the sample size. Both of these are rarely possible in macrofossil-poor lakes, where age models rely on dating the available macrofossils, wherever they are found, and whatever size they are. We have demonstrated that it is possible to acquire an age at any depth that contains sufficient pollen, but two issues need to be addressed for future application of the technique: first, the small size of the samples, which resulted in large uncertainties on the 14C ages; and second, the causes of the significantly older age of the pollen samples compared to the well-dated tephras. Although several potential improvements to the sample preparation methods may help to overcome these issues, geological processes are likely to be more significant, and more complex to address.
One problem that should be relatively simple to rectify in future pollen-dating studies is the size of the samples. In principle, the amount of carbon in a 14C sample may be increased by combining samples from adjacent depths. Although this reduces the resolution of the age model, 14C measurements with smaller analytical uncertainties can be calibrated with greater precision, leading to a better overall precision. The decision to combine adjacent samples relies on a reasonable ability to predict the size of the AMS sample from the sorted pollen separate, which in turn requires improvement in our handling procedures.
For example, in this study samples 1036-4 and 1036-5 had similar pollen counts (Table 1), but 1036-4 gave 0.11 mg of carbon and had an analytical uncertainty of±40 yr, while 1036-5 yielded 0.02 mg of carbon, resulting in an analytical uncertainty of±170 yr. Given the difficulty of removing the rinse water from the centrifuge tube while leaving the pollen behind, even after centrifuging, we suspect that the difference in weight between these two samples is because some pollen was lost from 1036-5 during the three rinse steps. Reducing or eliminating post-cytometry, pre-AMS cleaning steps, and development of efficient techniques for any required handling, should address this issue. Further, the color difference between the bleached and unbleached samples made handling of the latter significantly easier, suggesting that elimination of the pre-cytometry bleach step is desirable for samples where it is not required. Ideally, samples would be examined microscopically on arrival at the AMS sample lab, and the decision to combine samples or not made based on expected carbon content, sedimentation rate and depth between samples, required precision of the age model, and other project-specific considerations.
An informative exercise is to consider the precision of the age model if the samples had all contained >0.3 mg C. In that case, the AMS analytical uncertainties would likely have been the canonical 35 yr, and the simple calibrated age ranges would have been much smaller (Figure 7). For example, the worst-case sample in this experiment, 1036-5, had 0.02 mg C and a 14C uncertainty of±170 yr (1-sigma), which translates into a calibrated uncertainty of±445 yr (2-sigma). If this had been a 0.3 mg sample with the same central age (1890 14C yr BP) and a 35-yr analytical uncertainty, the calibrated 2-sigma uncertainty would be reduced to±95 yr.
Probability density functions (PDFs) of pollen dates with hypothetical 35-yr uncertainties, from Calib 6.0, using IntCal13 (Stuiver and Reimer Reference Stuiver and Reimer1993; Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Freidrich, Grootes, Guilderson, Haflidason, Hajdas, Hatte, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013). Note that vertical axis is only to indicate the depth of each sample, and is not a linear scale.
The reduction in calibrated range is primarily a function of the actual sample size, where the smallest sample (0.02 mg) had the largest reduction, and the calibrated uncertainty of the three largest samples (0.11–0.07 mg C) was reduced by only 0–20 yr. The irregular shape of the calibration curve plays a secondary effect in this sample set: two samples of 0.05 and 0.03 mg had nearly the same reduction in uncertainty, because the larger sample sits in a valley in the curve, while the smaller is on a little peak. Used in the same Bayesian model described above, this hypothetical “reduced uncertainty” dataset (with the 400-yr offset of the pollen dates as in Model 3) results in Model 4 (Table 2, Figure 6C), which has only slightly better precision than Model 3, using the tephras and the pollen with the real uncertainties (Table 2, Figure 6B). Either would permit correlation of paleoclimatic proxy features on the sub-centennial to multi-decadal scale.
A second effect that may be improved with examination of the chemistry protocols is the “old” age of the pollen relative to the tephras. Although the potential for contamination by various carbon-bearing materials has been recognized since the work of Brown et al. (Reference Brown, Nelson, Mathewes, Vogel and Southon1989) and carbon-based reagents were carefully avoided in this study, the addition of petroleum-based (14C dead) carbon from laboratory reagents may not be the only source of alteration during the chemical preparation. Byrne et al. (Reference Byrne, Park, Ingram and Hung2003) found that chemical pretreatment with a variety of non-carbon bearing reagents (trisodium phosphate, hydrochloric acid, potassium hydroxide, etc.) caused the δ13C of pine pollen samples to shift as much as 4 per mil lighter than untreated pollen. Although they could not rule out the possibility that the shift was caused by the physical removal of isotopically heavier components of the pollen, they suggested that the carbon-isotope composition of the sporopollenin phase itself may have been altered by the use of strong acids. Additional systematic testing of these variants in conjunction with flow cytometry purification and 14C dating is in order. Regardless, a shift of a few per mil in the isotopic composition of the pollen cannot account for the 400-yr offset of the pollen ages obtained in this study.
The geological setting and processes of a lake basin influence the accumulation of pollen in the lake’s sediments and the resulting pollen 14C age, and are likely more significant and more difficult to control than effects of sample preparation. For example, the shift of 400 yr observed in sample 1036-1 requires a contribution of only 15% of the carbon from pollen 4000 14C yr old, and only 6% from pollen >30,000 14C yr old. Both of these are likely ages of old pollen delivered to modern Mono Lake sediments, because higher lake levels during those periods deposited sediment at elevations that are now above lake level and may be actively eroded.
Further, the date at 34.5 cm depth is >250 yr older than the date below it, constituting a serious reversal in age at this scale (Table 1, Figure 6). If the dates were on macrofossil material, it might be assumed that the bit of charcoal or plant material had been stored and re-transported, and it would be ignored as an outlier. It is somewhat more difficult to explain a purified pollen age in this way. There was nothing different about this sample in the smear-slide evaluation of the concentrated separates before sorting, and the sorting time, grains sorted, and final pollen content were not strikingly different than the sample above. The two samples below (36.5 cm and 38.5 cm) did not contain enough pollen for dating, and the 34.5 cm sample occurs at the time of the pluvial between the mega-droughts of the Medieval Climate Anomaly, when Mono Lake (and other western Great Basin lakes) rose briefly and then fell (Stine Reference Stine1990b, Reference Stine1994; Adams Reference Adams2003). It seems likely that the change in lake level may have reactivated older deposits around the lake, which are common in the basin.
This effect is especially likely where old deposits on the surface are common, as is the case in the western Great Basin. Mono Lake, like many lakes in the region, was much larger during glacial periods, and the regression at the beginning of the Holocene exposed Pleistocene lake sediments all around the lake. These sediments are eroded into the lake by the streams, as well as being lifted from the surface by winds, depending on vegetation cover and wind speed/direction. Both fluvial and aeolian transport are likely to be variable in time as well as space, increasing, for example, during a dry period when vegetation dies, or a drop in lake level induces down-cutting by the streams. As suggested by the above example, old pollen may also be introduced by increased run-off and erosion during wet periods. The amount and age of reworked pollen contributing to a lake sequence will depend on basin geomorphology, patterns of surface run-off, the geometry and location of older deposits, and whether streams are aggrading, down-cutting, or in approximate equilibrium at any particular time.
Interestingly, Mensing and Southon (Reference Mensing and Southon1999) also found an offset of 400 yr between pollen and non-pollen dates in the sediments of Lake Moran, in the western Sierra Nevada. The 38–74 micron and >74 micron fractions of the sediment just above the Mazama tephra were consistent with the best age for that eruption, (~6640 14C BP) while pollen from the same depth interval was older by 400 yr (~7040 14C BP). Their pollen samples were purified by mouth pipetting, and so (similarly to our results) the older age cannot be attributed to non-pollen contaminants. Instead, Mensing and Southon (Reference Mensing and Southon1999) suggest the possibility that soil infiltration rates may have been altered by the tephra, inducing down-cutting and erosion of older pollen.
The potential for older reworked pollen to be included in the sorted pollen fraction makes microscopic examination of the pollen separates critical. If the older pollen is physically or chemically degraded and can be identified visually after sorting, the sample may be discarded as undateable if the fraction is high enough, or flagged as potentially problematic and the date treated with caution. In locations where older periods were characterized by significantly different vegetation (e.g., forest vs. grassland, oak vs. spruce), it may be possible to identify the presence of reworked pollen even where it isn’t visibly degraded. The possibility that sorting parameters used by the flow cytometry system might be able to distinguish pollen of different types and stages of degradation also remains to be explored.
Whatever the cause of the age offset, in this case the tephra dates can be used to do an approximate correction of the 14C dates, similar to the comparison of macrofossil-bulk sediment pairs in lakes where macrofossils are rare. The premise of the approach is to test the similarity of the offset at different depths in the sediment sequence, in hopes of finding a consistent offset through time that can be reliably used to correct bulk-sediment dates that do not have a macrofossil pair. Here we have corrected the pollen ages based on the well-constrained age of the North Mono-Inyo tephra, subtracting 400 yr from each 14C age before calibration. This results in the oldest date, just above the South Mono tephra, being 80 yr older than the mean 14C age of that tephra, well within the 2-sigma uncertainty of both dates.
As with bulk-macrofossil pairs, this correction method is not recommended without careful attention to the specific case at hand, however. If the cause of the offset is the chemical preparation, it should be consistent between samples prepared in the same way, and avoidable with appropriate chemistry. However, in the case of a contribution of old pollen from the environment, the correction will certainly vary between lakes, probably even in similar settings (e.g., western Great Basin lakes), as well as through time in a particular lake. Further, the contribution of old pollen may potentially change pollen 14C ages by hundreds or thousands of years, depending on the balance of production and deposition of contemporary pollen with erosion and deposition of old pollen stored on the landscape.
IMPLICATIONS AND FUTURE DIRECTIONS
Sorting by flow cytometry may potentially be a breakthrough in building pollen-based age models for lake records, but additional testing of preparation protocols is required. Several tasks will help to address the problems and uncertainties we observed here:
1. The techniques that have been used to prepare the raw mud for sorting, including sieving, heavy-liquid separation, and chemical concentration, must be compared to determine their value for flow cytometry of pollen for AMS. The protocol using the fewest preparation steps to yield the cleanest final sample is the ideal, but the variety of techniques needed for common kinds of mud (and pollen) must be systematically tested. An additional consideration is the time necessary to sort sufficient pollen grains from a sample for a reliable AMS analysis, as sorting costs are generally charged per hour.
2. The potential range of sorting parameters must be explored, to address the possibilities of separating pollen of different types, as well as the potential to eliminate badly degraded pollen from a sample during sorting.
3. How much pollen is in the sediment and what type (i.e., how big), as well as what other material remains after sieving and concentration, will determine the amount of sediment needed and the time required for sorting (and thus the cost). These set practical limits on the precision of the age models that can be produced by this approach, and suggest that different approaches may be appropriate for initial age models on long sediment sequences, versus high-resolution age control for time periods of particular interest.
4. The possible addition of modern and 14C-dead carbon during sample preparation for AMS measurement is routinely monitored for other sample types (e.g., wood, charcoal, carbonates) by preparing materials of known age and similar matrix alongside samples. To test for addition of old carbon or modern carbon during the preparation of pollen, modern and 14C-dead pollen standard materials must be identified and characterized. These should then be prepared alongside samples using the same physical and chemical protocols for both.
Once protocols are developed to reliably produce robust 14C dates, a number of geological questions will need to be studied and addressed on a case-by-case basis. What kinds of pollen (size, amount of carbon, plant life cycle, dispersion method) are most desirable for 14C dating? Once the pollen has been released from its plant of origin, how long can it sit on the landscape before being sufficiently degraded to be unrecognizable to the flow cytometer? How does the potential residence time vary with climatic and geomorphic characteristics of a lake basin? In a specific lake or climatic setting, what changes in landscape dynamics or climate conditions produce the most reliable, or most erroneous 14C ages? Are there particular kinds of lakes or climatic settings where pollen dating always, or never, produces reliable age models? What are the limits to the resolution achievable by the technique in different settings? With the proper attention to these dynamics of the specific basin being studied, AMS dating of flow cytometry-sorted pollen may be the key to high-precision, high-accuracy correlation of proxy records of past terrestrial climate.
ACKNOWLEDGMENTS
This article is dedicated in memory of R. Byrne, who introduced the idea of flow cytometry as a means of separating pollen for AMS dating, and encouraged us to explore the power and applications of the technique. Many thanks to to L. Reidy, M. Rosen, I. Hendy, L. Heusser, S. Mensing, C. Scholz, and I. Tunno for discussions about pollen and AMS dating of lake sequences; S. Stine for discussions of Mono Lake sediment dynamics; and A. Myrbo for many discussions about Bacon. S. Hemming is gratefully acknowledged for her continuing support of Mono Lake work and the lead author. This work was supported by LLNL LDRD grants 14-LW-091 and 17-ERD-052 to S.Z. and performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344; this is LLNL-JRNL-725175.
