Loess chronology and composition as interpreted through the lens of fossiliferous lake sediments in central Illinois, USA

Located along the Illinois River valley in central Illinois, Lake Ben and Smith Lake (Fig. 1) ostensibly have silty lacustrine and paludal records that span the waning of the penultimate glaciation (late marine isotope stage [MIS] 6; ca. 135 ka; Curry and Baker, 2000; Curry et al., 2011) to the present. My primary objective in this research was to determine the timing and nature of waning and eventual cessation of loess deposition at the end of the Wisconsin Episode (last glaciation). Loess was deposited in the shallow closed lakes or wetlands, which acted as sediment traps. Ancient Lake Ben and modern Smith Lake are ~0.5 km and 1.0 km across, respectively, with adjacent slopes of <12%. Different basin morphologies and other factors such as distance from outwash terraces along the Illinois River valley resulted in different sediment accumulation rates (SARs) and sediment textures in the lake records.

Another purpose of this investigation was to characterize the grain-size distribution and composition (chemistry and mineralogy) of loess redeposited in the lakes. Characterization studies of loess (Smith, 1942; Willman and Frye, 1970; Grimley et al., 1998; Nash et al., 2018) have been limited by bioturbation and pedogenesis in the solum (Mason and Jacobs, 1998). Loess is the parent material of ~63% of soils in Illinois (Fehrenbacher et al., 1986). Understanding the mineralogy and geochemistry of Peoria Silt and its regional changes will be important for applied studies, such as establishing baseline concentrations of soil contaminants like lead (Pb). These parameters are also of scientific interest, such as in provenance studies (Grimley, 2000; Muhs et al., 2007).

Concentrations of elements such as Al are known to increase upward through B horizons of loessial sola, but the timing of their addition to the soil is not known, which hampers interpretation of their origin (Mason and Jacobs, 1998; Dreher and Follmer, 2004). Assessment of whether additions of elements like Pb and Al are natural or anthropogenic in Illinois agricultural soils may be determined in places where loess records are thickest with suitable dateable material. Lake basins are good targets for such studies, although most in Illinois have been drained for agriculture. The interpretive framework and results of laboratory data presented here will become richer from ongoing studies.

Many threads of loess investigation have their beginnings in the North American midcontinent, including Illinois. Smith (1942) provided the first rigorous evidence supporting the modern understanding of loess as an eolian deposit. He mapped the extent and thickness of late Pleistocene loess in Illinois and demonstrated trends in thinning and fining of silty loess downwind of outwash terrace sources. Most scholarly treatments of loess genesis in the midwestern United States have roots in his essential concepts (e.g., Grimley, 2000; Grimley et al., 1998; Muhs et al., 2013, 2018a; Nash et al., 2018). Continued advancements and interest in loess research are reflected in well-attended “Loess Fests” sponsored by the International Union for Quaternary Research's Loess and Pedostratigraphy Focus Group (Schaetzl et al., 2018).

Patterns of loess thinning and fining in central North America strongly suggest wind directions from generally west to east (Smith, 1942). In the 1980s, the first global circulation models indicated primary wind directions in the opposite direction, creating a point of contention between observations of field geologists and outputs of ever-improving computer models. Conroy et al. (2019) resolved the technical side of this puzzle by showing that the large size of computational cells used by the earlier models compromised resolution of a northern jet stream and its interaction with equatorial circulation. The new results are in general harmonious with field observations (Conroy et al., 2019).

One aspect of loess deposition that has evolved considerably since the days of Smith's early studies is the mechanisms that impact sediment sorting and accumulation rates. Conditions conducive to entrainment and accumulation of eolian deposits include a dry sediment source, a wind strong enough to loft particles into the atmosphere (or saltate them across the ground surface), and a stable place for the sediment to accumulate. Entrained and transported dust was composed of relatively fine particles (in Illinois, this is typically particles of very fine sand and finer), and the saltated materials were fine to medium sand particles. The conceptual end members of continental eolian deposits are loess (from atmospheric fallout) and dunes (from saltated particles).

The dust separates into a lower level of relatively coarser material and an upper level that may extend to the stratosphere (Pye, 1995). The lower part of the dust cloud is enriched in very fine sand and coarser silt, which characterize the texture of much sediment constituting the topsoil parent material in Illinois. The finer fraction may be deposited very far away from its source. A notable example is the dust originating from dry western Africa accreting in soils forming in the Caribbean (Muhs et al., 2007). Grain-size ratios are useful for assessing changes in the distance from the loess source (Grimley et al., 1998; Nash et al., 2018).

Deposits of silty and sandy outwash, deposited by seasonal glacial floods, were the source of proximal loess. Dust clouds formed during gusty windstorms from unvegetated terrace surfaces adjacent to glacial sluiceways. Although the processes that promote lofting of material are not contentious (Pye, 1995), the mechanisms involved with sediment accumulation are not as well understood. On a state or county scale, changes in loess thinning and fining are attributed to physics: the interaction of wind turbulence, particle size, and gravity. In some areas of Alaska, loess accumulates faster under a tree canopy than in grasslands, indicating a biological aspect of sediment trapping (Muhs et al., 2018a).

Sediment chemistry and mineralogy provide a framework for assessing the relative contributions of the major southern outlet glaciers of the Laurentide ice sheet: the Lake Michigan, Green Bay, Des Moines, Superior, and James lobes (Grimley, 2000; Curry and Grimley, 2006). The diversion of the Illinois River from its old course as part of the ancient Mississippi River to its modern course is a case in point (Curry, 1998). Prior to the diversion, outwash in the middle Illinois River valley included material derived from the Lake Michigan, Superior, and Green Bay lobes; after the diversion, outwash sediment originated solely from the Lake Michigan lobe, possibly with contributions from the Green Bay lobe via the Rock River. In places located throughout central Illinois, prediversion sediment contains relatively abundant magnetic minerals as measured by magnetic susceptibility (Grimley et al., 1998). The uppermost age of a wetland deposit buried by gray lacustrine sediment at Lomax, Illinois, suggested a maximum diversion age of 24,400 calibrated (cal) yr B.P. (Curry, 1998), an age later corroborated in other studies (e.g., Grimley et al., 1998; McKay et al., 2008; Nash et al., 2018; Curry et al., 2018b).

Profiles of element concentrations using energy-dispersive X-ray fluorescence (XRF) provide chemical spectra that invite interpretation of sediment provenance. An age model (discussed below) allows further interpretation of the nature of the change, be it abrupt or gradational. XRF scans are more commonly performed on sediment cores of lacustrine sediment, such as the cores from Lake Ben, rather than cores of loess successions. For studies of loess, XRF methods are also used, but on subsamples from core or outcrop that are typically fused with Li flux and analyzed by XRF dispersive methods.

Over the past 70 yr, approximations of the age of Peoria Silt have evolved with technological advancements in absolute dating methods. The first radiocarbon ages of loess were done on the aragonite comprising terrestrial snail shells; ~100 g aliquots were needed per analysis compared to the 0.03 g needed today for an age determination by accelerator mass spectrometry (AMS). These larger pre-AMS samples likely contained secondary carbonate, as their ages have proven to be “too young.” Pretreatment of gastropods shells done today includes: (1) omission of shell fragments festooned with sugary secondary carbonate, (2) breaking shells and picking only the outer whorls, and (3) ~15% reduction of sample mass through etching with weak acid. Further, studies have confirmed that certain species reliably provide ages without measurable contamination from radiocarbon-“dead” Paleozoic carbonate ingested by the organisms (Pigati et al., 2013, 2015). Shells of Succinea sp. are most commonly used for ages of Peoria Silt in Illinois (Nash et al., 2018).

Following these more rigorous pretreatment protocols, age models from two sites located ~40 km south and west of the study lakes indicated that loess began accumulating ca. 29 cal k.y. B.P., an age supported by earlier studies (Curry and Follmer, 1992; Grimley et al., 1998; Grimley, 2000; Curry et al., 2018b). Loess deposition rates slowed from ca. 23–21 cal k.y. B.P. (a period associated with development of the Jules Geosol), increased from 21–18 cal k.y. B.P., and ended between ca. 18 cal k.y. B.P. and 16 cal k.y. B.P. (Nash et al., 2018). The 23–21 cal k.y. B.P. age estimate for the Jules Geosol is ~3 k.y. older than previous estimates (Frye and Willman, 1973; Grimley et al., 1998).

Over the past 25 yr, many radiocarbon age estimates of Wisconsin Episode loess in Illinois have been determined by assay of soil organic matter (SOM) via pyrolysis-combustion (Wang et al., 2003a). A series of ages from Keller Farm, Illinois (Wang et al., 2000, 2003a, 2003b; Fig. 1), is cited as evidence that loess age and rates of deposition were different in Illinois than in neighboring states (Rousseau and Sima, 2014; Muhs et al., 2018a). Keller Farm SOM age records indicated that: (1) loess SARs were linear with an abrupt shift from rapid to less rapid accumulation at 20 cal k.y. B.P., (2) the last deposits of proximal loess were as young as the Younger Dryas (ca. 11 cal k.y. B.P.), and (3) long-term El Niño–Southern Oscillation cyclicity was consistent with ~400–600 yr periodicity of loess-paleosol couplets (Wang et al., 2000, 2003b). Subsamples of the same depth intervals yielded gastropod shells (identified by Jeff Pigati) with radiocarbon ages that progressively differed from SOM radiocarbon ages, with the uppermost interval yielding a difference of ~6.7 k.y. (Grimley et al., 2021). Age-depth profiles indicate snail and SOM ages converge at ca. 22.0–21.5 cal k.y. B.P. (Fig. 2). Age-depth profiles of modern soils in Illinois consisting of 1 m loess over loam diamicton (Wang et al., 2018) predict the observed age-depth pattern seen here; the upward-younging trends are caused by mixing of ancient carbon of the parent material with carbon in percolating organic acids from the modern soil.

Both lake basins examined in this study are located near the edge of a narrow east-to-west–trending erosional remnant of the Illinois Episode till plain upland (Fig. 3). The upland is bounded by steep slopes that on the south drop more than 27 m in ~1.5 km to a late Wisconsin terrace of likely lacustrine origin adjacent to the Holocene floodplain of Salt Creek (Fig. 4). Modern and ancient groundwater recharge to these lakes has been modest and likely absent during drought. The relatively great relief and well-drained setting likely partly explain why a long Sangamon Episode (last interglacial) lacustrine record is lacking at Lake Ben but is well preserved in basins to the south such as Pittsburg and Raymond Basins (Curry and Baker, 2000).

The lakes are located ~50 km southwest of the margin of the Last Glacial Maximum (LGM) terminal moraine (Shelbyville moraine; Fig. 3) and from 10 km to 45 km from the 35-m-wide Manito terrace, the largest LGM outwash plain along the Illinois River. Lake Ben and Smith Lake are topographically and hydrologically closed basins (Fig. 4); the primary nonorganic component of the lacustrine record is exogenic, with minor contributions from hillslopes. The Manito terrace was the likely primary source of proximal loess redeposited in the lakes. The lakes are located ~40 km northeast of Cottonwood School, a type section of the Peoria Silt (Willman and Frye, 1970). Outwash comprising the Manito terrace traveled at least 50 km from the LGM margin where it crossed the Illinois River valley in Peoria, Illinois (Fig. 3). Assuming wind direction from west to east (Smith, 1942), the next nearest loess source, the Mississippi River valley, is ~190 km west. Smith Lake is currently 7.5 ha with a topographic watershed covering ~45.9 ha. If the level rose to overflowing, the lake would cover ~25.2 ha (Fig. 4). The watershed of Lake Ben covers 172.7 ha; maximum paleolake area was ~38.5 ha. Publication dates of the earliest plat maps suggest approximately when Lake Ben was drained to make it agriculturally viable. Artificial drainage of Lake Ben occurred sometime after 1874 and likely before 1891, but possibly as late as 1933 (Fig. 5).

The genesis of the study basins was likely due to collapse of buried ice blocks after they were covered by glacial sediment shortly after retreat of Illinois Episode ice. The sediment cores show that the sublacustrine succession at Lake Ben is matrix-supported diamicton. The grayish brown color (10YR 5/2) and <2 mm matrix texture (loam) are consistent with the regional state map, which shows local deposits of the Vandalia Member of the Glasford Formation (Lineback, 1979). The age of this unit likely correlates with MIS 6, ca. 190–135 k.y. B.P. (Curry et al., 2011). From oldest to youngest, the successions include: (1) Illinois Episode diamicton (Vandalia Till Member of the Glasford Formation; Willman and Frye, 1970); (2) Illinois Episode lacustrine and colluvial deposits (Teneriffe Silt; Willman and Frye, 1970); (3) Illinois, Sangamon, and Wisconsin Episode peat (Berry Clay Formation and Roxana Silt; Hansel and Johnson, 1996; Curry et al., 1994); (4) Wisconsin Episode silty (lacustrine) sediment (Equality Formation) interpreted as mostly redeposited loess (Peoria Silt; the research emphasis of this paper); and (5) Hudson Episode organic-rich silty lacustrine and paludal sediment (Equality Formation, continued).

Initial interest in obtaining lake cores from this area was to examine fossil records of the last interglaciation (Sangamon Episode) like those in south-central Illinois (Curry and Baker, 2000). The first sediment cores were obtained in 1994 near the center of Smith Lake. Difficult sampling prompted the use of heavy-chain winches, which led to bent extension rods. Core recovery was poor, particularly the deepest material composed of compact peat. Radiocarbon ages of 26.3 ± 0.2 ¹⁴C yr B.P. and 32.1 ± 0.3 ¹⁴C yr B.P. were determined by conventional methods on samples from depths of 1108 cm (ISGS-2943) and 1172 cm (ISGS-2944; Curry et al., 2018b).

Interest in the Smith Lake record today focuses on the redeposited loess. Sediment cores were obtained using a Livingston piston corer aided initially by an electrically powered winch and later by a hand-cranked geared system. Core recovery was poor. Below about 1 m of soft Holocene sediment was ~2 m of hard, sticky fine silt. The soft upper meter of sediment was recovered with a Bolivia corer. Subsamples of all sediment cores yielded fossils suitable for radiocarbon dating (Table 1).

Significant core loss from the collective Smith Lake records prompted me to obtain core Mason 1a in 2018 with a hydraulic push probe from dry land situated in a nearby topographically closed basin, Lake Ben. The probe sampler easily pushed through the sticky fine silt and peat that impeded coring at Smith Lake. Most core loss zones were picked up by subsequent coring located 2 m and 4 m away from the master core (Mason 1b, 1c).

The 2018 sediment cores from Lake Ben were sampled at 1.52 m (5 ft) intervals except for initial drives of 1.20 m (4 ft). At the Limnological Research Center (LRC; University of Minnesota, Twin Cities), the halved acrylic-lined sediment cores were scanned for spectroscopy and magnetic susceptibility (MS), as well as photographically scanned at 600 dpi (Fig. 6). I later used these cores for all subsequent analyses discussed below.

With two early exceptions, all AMS samples were assayed by the Keck Carbon Cycling Institute at the University of California, Irvine. To obtain fossils for radiocarbon dating, I wet sieved field-moist 4-cm-long samples after soaking them in boiling water. Residual material was transferred to glass Petri dishes to dry. I photographed the dried material with a Nikon DS-Fi3 microscope camera mounted on a Nikon SMZ745 T trinocular microscope. AMS samples were given a standard acid-base-acid pretreatment. I explored age models by varying the thickness of analyzed sections and placement of lacuna using Bayesian modeling (Blaauw and Christen, 2011).

Scanning was done at 1 cm resolution on split field-moist cores using Northwestern University's Geotek multi-sensor core logger (MSCL-S) outfitted with an Olympus Delta energy-dispersive X-ray fluorescence (EDXRF) analyzer, a Konica Minolta CM-700d spectrophotometer, a Bartington MS2E magnetic susceptibility meter, and a 50 mm Canon camera. A dwell time of 30 s was utilized for the XRF. The error of the XRF measurements was 5%–10% and varied by element (Boyle, 2000). EDXRF measures concentrations of elements ranging in atomic weight from 23 amu to 238 amu (Na to U; Boyle, 2000). Na and Mg concentrations had errors >10%, and these data were excluded in my analysis. MS values reflect relative changes in the magnetic mineral content of the sediment (Grimley et al., 1998).

XRF results are dependent upon bulk density and grain size, with sandy materials being enriched in Si and Ca, and clayey materials enriched in elements such as Rb. To test the significance of grain-size effects, I analyzed 32 samples representative of the sediment succession using the ISGS Rigaku NEX CG EDXRF spectrometer under vacuum. Samples were dried at 105 °C overnight and sieved to <250 μm. Grain-size classes (<20 µm, 20–32 µm, 32–64 µm, and 64–125 µm) were separated via dry sonic sieving. Samples were mixed with a SPEX paraffin binder 3646 (9:1 ratio) in a Chemplex Industries, Inc., SpectroVial grinding vial (catalogue no. 1134) with two Chemplex Industries, Inc., ball pestles (catalogue no. 1211) using a SPEX mixer mill. Pellets were comprised of 4 g of mixture hydraulically pressed at 10 tons of pressure.

Geotek^© scanners at the LRC and at the laboratory of Professor Yarrow Axford at Northwestern University provided International Commission on Illumination (CIE) 1976 color space data (Robertson, 1977), in which L* indicates lightness, a* is the red/green coordinate, and b* is the yellow/blue coordinate. The LRC scanner also provided detailed MS measurements, and the Axford scanner provided EDXRF elemental-concentration data (discussed below). I did not have L*a*b* profile scans done on the Smith Lake cores.

I submitted dry core subsamples (<2 g) to the Prairie Research Institute Illinois State Water Survey–Illinois State Geological Survey Sediment Materials Laboratory for analysis by laser diffraction using a Malvern Mastersizer 3000 (Malvern Panalytical, 2010). Pretreatment included organic matter removal in 10% hydrogen peroxide at 60 °C for a minimum of 15 min.

Using the online program PAST (Hammer et al., 2001), I executed principal component analysis (PCA) and detrended correspondence analysis (DCA) of correlation matrices of the XRF and particle-size distribution data. I used variable loadings to assess data redundancy and determine the most effective elemental ratios to “fingerprint” sediment provenance. Eigenvector value profiles plotted with respect to time pinpointed when changes began and ended.

Fifteen samples of glycolated, oriented samples were analyzed at the Illinois State Geological Survey X-ray Diffraction (XRD)/X-ray Fluorescence (XRF) Materials Characterization Laboratory using a Scintag® XDS2000 spectrometer. I prepared oriented, glycolated slides following the standard procedure of Hughes et al. (1994). The step-scanned data were collected from 2° to 34° 2θ with a fixed rate of 1° per minute and step size of 0.02° 2θ.

I determined the relative proportions (by weight) of organic matter, carbonate minerals, and terrigenous sediment by using the standard loss-on-ignition (LOI) procedure of Dean (1974).

Over a nearly 20 yr span, 63 radiocarbon ages have been determined from fossils picked from the study sediment cores, including 35 from Lake Ben and 28 from Smith Lake (Table 1). Most sample ages were in chronological order with respect to depth. Three main fossil plant types used for radiocarbon assay include: sedge inflorescences and stalks, wood fragments, and fruits of Callitriche hermaphroditica (Figs. 7 and 8). The condition of all fossils was excellent, except for partial humification of wood and sedge fragments in the oldest samples. This humified material was responsible for coring refusal using the Livingston corers. The material was also difficult to disaggregate and process for various analyses because the material was visibly hydrophobic.

Callitriche hermaphroditica is rarely found in good enough condition to identify nor found in sufficient quantities for radiocarbon assay. Lake Ben sediment contained unusually well-preserved and abundant specimens, especially in the interval 25–22 cal k.y. B.P. (Figs. 7 and 8). The species is ecologically adept and may pollinate below or above water. In many sample intervals, it was the only fossil recognized. In selected intervals, radiocarbon assays of costratigraphic sedge and Callitriche fruit fossils allowed determination that the latter yielded dependable ages (Table 1).

The Lake Ben age model (Fig. 7) revealed maximum SAR varying from 0.25 cm/yr to 0.33 cm/yr during a 400 yr span (22.6–22.2 cal k.y. B.P.). SAR minima between 0.0125 cm/yr and 0.004 cm/yr occurred during an 8500 yr span (39.1–31.6 cal k.y. B.P.). In both cases, ~0.7 m of sediment accumulated. The upper part of the Lake Ben sediment core has 2σ errors less than 2000 yr, averaging 875 ± 240 yr B.P. (Fig. 9). The lower part (13.27–11.09 m deep, corresponding to 45.3–31.6 cal k.y. B.P.) has larger errors, averaging ~2500 ± 0.5 yr B.P.

Unlike Lake Ben, Smith Lake has age control in the upper meter of sediment (Fig. 10). The most prominent feature of the Smith Lake age model is the interval 250–193 cm depth (corresponding to 18.1–4.0 cal k.y. B.P.) with its large 2σ errors, which are as large as 10,200 yr. Below this zone of low SAR, the age error of the Smith Lake model is comparable to Lake Ben, averaging less than 1000 yr.

I noted four spectral zones (Fig. 11). Spectral zone A is characterized by spiky values, and high L* values represent rock particles. Removing these peaks from consideration, the values match what is seen by the eye: sediment that darkens upward. The upper contact with spectral zone B is abrupt; zone B is characterized by uniform light yellowish material (high L* and b* values), with little contribution from red hues (low a* values). The upper contact with spectral zone C is marked by increasing a* and b* values at a depth of 246 cm (17 cal k.y. B.P). Rising L* values in spectral zone C coincide with the gleysol; its truncation is evident at the C/D contact (Fig. 11).

I parsed data with unreasonably low or high values caused by sediment disturbance and partial desiccation near core tips. S and Fe had very spiky profiles, indicating postdepositional ion mobilization and remineralization. In the following discussions that characterize redeposited loess of the last glaciation, I discounted material above an angular unconformity at 120 cm depth caused by plowing as well as information below a depth of 1250 cm (42.0 cal k.y. B.P.).

Profiles of the selected data followed three patterns. XRF group A (Fig. 12A) includes high values of Ca, MS, and Zr. Values are greatest in a middle zone dating from 31.0–17.5 cal k.y. B.P. In the intervals above and below, their values dip to near or at background levels. XRF group B include elements that are near background from 42–31 cal k.y. B.P., have moderate values from 31.0–17.5 cal k.y. B.P., and have highest values in samples dating from 17.5–10.0 cal k.y. B.P. XRF group B includes the greatest number of elements exemplified by Al, Rb, K, Si, Ti, and Y (Figs. 12B and 12C). XRF group C (Fig. 12D) includes elements that have similar trends as group B but are marked by relatively higher variability (e.g., Pb, Cu, and Nb). Elements uniquely abundant in the lower part of the ¹⁴C-dated record are Cd and Ag, which comprise XRF group D (Fig. 12E). Sr provides a notable profile (Fig. 12F) that has the highest relative concentrations in the middle time period (31.0–17.5 cal k.y. B.P.) and somewhat lower values in the upper interval (17.5–10.0 cal k.y. B.P.) and a unique U-shaped trend from ca. 37.5–31.0 cal k.y. B.P., bottoming out ca. 33.4 cal k.y. B.P.

Principal component score profiles (Fig. 13) corroborate correlations suggested by visual inspection of variable profiles (Figs. 12A–12D). The primary feature of the axis 1 score profile (which explains 45% of data variability) is the upward decrease of Ca, Zr, and MS beginning at ca. 18.5 cal k.y. (XRF group A; Fig. 12A) at the expense of a larger group of elements exemplified by Al, Rb, and K (XRF group B; Figs. 12B and 12C) and Pb and Cu (XRF group C; Fig. 12D). Maximum axis 2 scores occur in sediment older than ca. 30.7 cal k.y. B.P., corresponding with Ag and Cd abundance profiles.

In Figure 14, I averaged chemical data in 500 yr bins starting with 10.0–9.5 cal k.y. B.P. (= 9.8) and ending with 42.0–41.5 cal k.y. B.P. (= 41.8). The PCA consists of 865 1-cm-thick intervals, and by parsing the data into 500 yr bins, the number of points was reduced by an order of magnitude (66). The youngest ages, from 17.8–9.8 cal k.y. B.P., all plot in quadrants III and IV and are enriched with XRF group B and C elements (Al, K, Rb, Pb, Cu, Nb, Y). Further parsing revealed that materials dating from ca. 16.3–9.8 cal k.y. B.P. are enriched with Si, Ti, K, and Al (in quadrant III) relative to the period 19.3–16.8 cal k.y. B.P., in which the materials have more Th, Zn, Ag, and Cd (in quadrants I and IV).

Provenance characterization may be explored in PCA by selecting variables that exhibit independent behavior expressed by right-angle relationships in the biplot. Variables in XRF group A (Ca, MS, and Zr) plotted in quadrant II at nearly right angles to those elements in XRF groups B and C. XRF group D, comprised of Cd and Ag, plotted independently in quadrant I. Elements that plot in similar multivariate space could be assessed further for redundancy.

One approach to assess sediment composition end-member mixing is to take the most positive and negative values of PCA axis 1 (Fig. 14). A stratigraphic plot of the ratio (Ca + MS)/(Al + Rb) revealed four zones spanning 40.5–10.0 cal k.y. B.P. (Fig. 15). Basal zone A features low values of all variables shown dating from 40.5–30.8 cal k.y. B.P., the interval with relatively abundant Cd and Ag. Zone B dates from 30.8–20.8 cal k.y. B.P. and has Ca and MS values greater than Al and Rb. Zone C is transitional and dates from 20.8–17.4 cal k.y. B.P. Upper zone D (17.4–10.0 cal yr B.P.) characteristics are marked by near-background values of MS and Ca and ever-increasing values of Al and Rb. The upper boundary of transitional zone C is less distinctive and could be placed at 20.8 cal k.y. B.P. (following the trends of the ratio) or as young as ca. 19.2 cal k.y. B.P. by emphasizing the abrupt drop in Ca concentration.

To assess the effects of particle size on XRF values, I prepared four grain-size subsamples from eight intervals. I did this test to determine if the rise in XRF group B elements dating from 17–14 cal k.y. B.P. was primarily due to changes in sediment provenance or to changes in grain size, with each grain-size class having a unique composition. A significant difference between the EDXRF data from pellets (Fig. 16) versus scans (Fig. 14) was the addition of Mg and Cl in the pellet analyses and exclusion of MS (magnetic susceptibility). The results demonstrated the unique elemental composition of the younger fine loess relative to the older coarse loess. Samples rich in Ca, Mg, Cl, and Mn (and with high MS values) dated at 21.2 cal k.y., 22.7 cal k.y., and 29.7 cal k.y.; samples poor in those elements, and rich in Al, Rb, and other elements, dated from 12.8 cal k.y. and 16.5 cal k.y.; and samples with intermediate compositions dated at 18.4 cal k.y. and 32.8 cal k.y. (Fig. 16).

The sediment succession in the Lake Ben cores is predominantly silt (Table 2), with sand-silt-clay percent means of 5.8–81.8–12.4 and standard deviations of 2.3–4.8–5.9, respectively, for 52 samples. The median and weighted mean grain size of all samples is medium silt (21.9 ± 5.6 µm and 28.0 ± 7.8 µm, corresponding to 5.5φ ± 0.2φ and 5.2φ ± 0.4φ, respectively); medium silt comprises the largest proportion (31.5% ± 5.2%) among the silt fractions.

A PCA of these data revealed that about half of the scores express inverse relationships between phi ranges 6–4 and 11–7 (15.6–62.5 µm and 0.5–7.8 µm, respectively; Fig. 17). Most subsamples with ages younger than ca. 18 cal k.y. B.P. have median grain size of between 10 µm and 20 µm and a fine tail that extends beyond 1 µm (e.g., curve 14.2 cal k.y. B.P. on Fig. 18).

Profiles of the ratio 6φ–4φ versus 11φ–7φ revealed many of the same trends exhibited by XRF, notably low values from ca. 17–10 cal k.y. B.P., and the long-lived saddle of low values dating from ca. 38–31 cal k.y. B.P. (Fig. 19). Peaks of coarser material that are supported by more than one value occurred at ca. 40.2 cal k.y. B.P., 30.5 cal k.y. B.P., 27.0 cal k.y. B.P., 25.0 cal k.y. B.P., and 23.0 cal k.y. B.P.

XRD analyses of oriented, glycolated slides returned diffractograms of variable quality (Fig. 20). Clay mineral zones (CMZs) A through D were identified based on primary mineralogy and trends and total counts-per-second of major clay mineral peaks (Table 3). Notably, CMZ C was rich in well-crystallized smectite (i.e., with especially sharp and well-formed peaks; Fig. 20B), although in terms of particle size, clay content was only ~14% greater than overlying and underlying zones. CMZ D included three samples within the plow zone.

Four zones were revealed in profiled LOI data from Lake Ben core 1a (Table 4; Fig. 21). LOI zone A was marked by low organic carbon (OC) content; it was beyond the limitations of radiocarbon dating, and its age is unknown. LOI zone B contained abundant organic matter, including two samples with OC values of ~40%, indicating the material is nearly entirely composed of organic matter. The mean and standard deviation of inorganic and organic carbon of LOI zone B was 0.7% ± 0.1% and 23.5% ± 10.0%, respectively (Table 4). Characteristic LOI zone C OC values were low, and inorganic carbon was relatively large. LOI zone D was marked by low inorganic carbon values.

Armed with a well-constrained age model, I aimed the focus of this analysis on material dating from 45.3 ± 1.5 cal k.y. B.P. to 14.7 ± 0.1 cal k.y. B.P., corresponding to ~10.5 m of sediment in core Mason 1a (Figs. 7 and 8; Table 1). I identified three lithofacies (A–C) with intervening transition zones based on age, fossil type, SAR, CIE (color) spectroscopy, XRF-based element concentration profiles, particle-size distribution, clay mineralogy determined by XRD, and organic and inorganic carbon contents determined by LOI (Fig. 15; Tables 5A and 5B). The entire succession is rich in silt (>80%; Table 2), reflecting the depositional history of loess redeposited during the last glaciation in shallow lacustrine and paludal environments.

Lithofacies A (older than 43.4–31.6 cal k.y. B.P.) is dark, fossiliferous, leached, OC-rich, silt loam with an earthy odor. I found this hydrophobic material difficult to prepare for slides for XRD data due to partial cementation with humic acids. This tough material stopped coring at Smith Lake. Matted sedge stalks comprised the bulk of several subsamples. The high organic matter content explains the relatively lower concentrations of elements detected by XRF scanning. A factor of interest is the low SAR values. From 39.1–35.9 cal k.y. B.P., SAR values are at their lowest in the entire record, less than 0.008 cm/yr. Values less than 0.009 cm/yr broaden the span of low SAR to 41.7–30.3 cal k.y. B.P. (Fig. 7).

The transition from lithofacies A to B is thin (19 cm) and chronologically abrupt (30.9–30.5 cal k.y. B.P., a span of 0.4 k.y.). It is markedly more abrupt than the upper transition from lithofacies B to C, which is 139 cm thick and took more time (3.3 k.y.). Seeds and fruits of aquatic vegetation are abundant in several horizons.

Other than continued dominance of silt, I found that lithofacies B (30.2–21.3 cal k.y. B.P.) has the opposite character of the overlying and underlying lithofacies. It is marked by high concentrations or values of Ca and MS, L*, inorganic carbon (as mostly CaCO₃), and SAR. Small and well-preserved fossils occur, including sedge stems and inflorescences in sediment dating from 30.2–27.0 cal k.y. B.P. and, notably, fruits of Callitriche hermaphroditica from 25.2–22.0 cal yr B.P. (Figs. 7 and 8). These macrofossils are fragile and typically recovered as damaged fragments; excellent preservation is due to the high SAR.

The upper transition zone from lithofacies B to C is marked by upward sediment darkening (= lower L* values, but not greater organic matter content as measure by LOI), high variability (related to laminations of dark and light sediment), and trending values of most elements. Although upward-diminishing Ca and MS values occur in the transition zone, Ca reaches near background values at 17.5 cal yr B.P. about 1 k.y. after MS values bottom out (Fig. 15). This phenomenon is likely due to changes to finer, more clay-rich material (that typically lacks magnetic minerals), but it may also be due to finer-grained magnetic minerals dissolving under reducing conditions under slowing SAR.

Lithofacies C occurs below the plow zone at a depth of 200–253 cm, corresponding to an age of ca. 14.7–18.0 cal k.y. B.P. The shallowest laboratory analyses began at 200 cm (= 14.7 ± 1.4 cal k.y. B.P.). Features marking lithofacies C include an increase in the concentrations of Al, K, Rb, Si, Ti, Pb, Cu, Nb, and V, and background or near-background values for Ca and MS (Fig. 15).

Major topics and issues rising from this study include (1) the kinship of the lake silts (recognized as Equality Formation by the Illinois State Geological Survey) with regionally recognized Peoria and Roxana Silts (loess), as indicated by similar particle-size distribution characteristics, mineralogy, and age; (2) proximal and distal loess facies differentiated by grain-size ratios and multivariate analysis of particle-size distribution data; (3) notably slow SAR from 39–31 cal k.y. B.P. during the Alton subepisode; (4) the abrupt transition in physical sediment character from the Alton to Michigan subepisodes; (5) elemental “fingerprints” and grain-size characteristics of loess versus generic dust, and sediment provenance; (6) diversion of the Mississippi River; (7) cessation of Peoria Silt deposition; and (8) correlation of element profile changes in modern loessial soils with the Lake Ben record. These discussions emphasize the importance of detailed site-specific studies for advocating regional correlation to similarly robust analyses.

The laminated silty sediments of lithofacies B and the transition zones above and below lithofacies B formally classify as the Equality Formation (Hansel and Johnson, 1996). Sediment age, texture, clay mineralogy, and chemical composition indicate that lithofacies B and allied transition zones of sediment core Mason 1a are waterlain facies of regionally widespread loess deposits known as Roxana and Peoria Silts (Willman and Frye, 1970; Hansel and Johnson, 1996). The correlation is supported by nearly identical lithology (particle-size distribution, mineralogy) and the timing of high SAR pulses of diachronous advances of the glacier margin during the Michigan and Alton subepisodes of the Wisconsin Episode (Hansel and Johnson, 1996; Curry et al., 2018b). Laminations and normally graded couplets indicate differential settling of very fine sand and silt grains through shallow water. Spring thaw likely resulted in rapid deposition of dust that fell earlier on the frozen lake surface, as well as on catchment slopes. Distal late glacial loess was deposited in the lakes with minimal bioturbation; the material also protected sedge inflorescences and stalks from oxidation, including immediate consumption by bacteria, mold, or animals. The sedges and other fossils also provide radiocarbon dateable material, which is lacking from contemporaneous distal loess deposited in adjacent terrestrial environments. Prairie soils not only store carbon but cycle carbon in part by pedogenic mixing of younger carbon (Wang et al., 2018). The lacustrine succession of redeposited loess was conformable and contained enough organic matter for radiocarbon age control.

Multivariate analyses of the grain-size data with respect to Wentworth's (1922) classification revealed characteristics of proximal and distal loess observed worldwide. The coarser fraction accumulates via rapid fallout within a few kilometers of the loess source, whereas the finer fraction is lofted higher into the atmosphere and deposited farther downwind (Pye and Zhou, 1989). Proximal and distal loess may be differentiated by grain-size ratios, specifically very fine sand-to-medium silt (6φ–4φ [15.6–62.5 µm]) versus fine silt to fine clay (11φ–7φ [0.5–7.8 µm]) (Fig. 19). Most loess deposits are a blend of these facies, resulting in a coarse mode with a skewed tail of fines (Pye, 1995; Sun et al., 2004; Prins et al., 2007).

Identification of colluviated sediment reworked into lakes versus sediment delivered directly to the lake through air-fall may be initially assessed by comparing grain-size characteristics. Detailed grain-size analyses conducted with the same method (laser diffraction) used in this study were done for samples from the New Cottonwood School (NCS) site located ~40 km south of the study lakes (Fig. 1) and the Thomas Quarry (TQ) site, ~30 km farther south (Nash et al., 2018). The bluff sites are within 0.5 km of the loess source, whereas Smith Lake and Lake Ben are 10–13 km distant. Regional profiles of mean grain size of Peoria Silt show that the more distal sites should have mean grain size ~20% smaller than sites adjacent to the loess source (Smith, 1942). For all data from lithofacies B–comparable material, this prediction holds for the data comparison between Lake Ben and Thomas Quarry, but not for New Cottonwood School (Table 6). While it is satisfying that Thomas Quarry and Lake Ben yield close comparisons, the data from New Cottonwood School suggest local differences in wind intensity, wind direction, or spatial variability of vegetation with differing sediment-trapping efficiency.

In southwestern Wisconsin, where Roxana Silt is in general <1.5 m thick, the age estimate for onset of Roxana Silt deposition was backcast to 55 cal k.y. B.P. (Leigh and Knox, 1993). This estimate is consistent with studies in the type region (McKay, 1979a), where the Roxana Silt is in places an order of magnitude thicker (e.g., 14.6 m at the Pleasant Grove type section; Willman and Frye, 1970; Fehrenbacher et al., 1986). A notable characteristic of the Roxana Silt is its abrupt thinning away from the source valley train in the Illinois River valley. My results agree with the regional isopach map of McKay (1979a); age-equivalent Roxana Silt in Lake Ben is ~400 cm thick (1490–1090 cm depth) compared to 800–1160 cm measured at the type Cottonwood School section (Willman and Frye, 1970; Fehrenbacher et al., 1986) and nearby New Cottonwood School section (Nash et al., 2018).

I was unable to improve upon previous age estimates of onset of Roxana Silt deposition at 55 k.y. B.P. (McKay, 1979a; Leigh and Knox, 1993). The ¹⁴C modeled part of Lake Ben's SAR record shows moderately high SAR of ~0.04 cm/yr at 45.3 cal k.y. B.P., fading to 0.02 cm/yr by 40.5 cal k.y. B.P. SAR values continue to fall, bottoming out at ~0.006 cm/yr from 39–34 cal k.y. B.P. (Fig. 7). Rates remain below 0.01 cm/yr until abruptly increasing to >0.05 cm/yr in the lower Peoria Silt. Evidence supporting slow SAR includes humic acid-cemented sedge stalks. When SAR rates ranging from 0.01 cm/yr to 0.04 cm/yr are used, backcasting the onset age of Roxana Silt gives ages from 61–49 k.y. B.P. An age of 55 k.y. B.P. is derived by assuming SAR of 0.015 cm/yr.

The mean SAR for the radiocarbon-dated part of the Athens subepisode at Lake Ben, 0.012 cm/yr, is intermediate between that for Cottonwood School (0.031–0.045 cm/yr) and southwestern Wisconsin (~0.006 cm/yr; Leigh and Knox, 1993). At least 70% of the Roxana Silt was (re)deposited in Lake Ben before ca. 40 cal k.y. B.P.; only 13.5% of the total thickness of the dated portion of the Roxana Silt was deposited during the Farmdale phase (roughly 34–29 cal k.y. B.P.; Curry et al., 2018b). The long period of very slow SAR (<0.02 cm/yr) from 40–32 cal k.y. B.P. has not been identified in the terrestrial record of thick Roxana Silt, but radiocarbon ages are few where the Roxana Silt is thicker than 2 m. Thick exposures of the Roxana Silt are rarely exposed and then rapidly slump and degrade, as at the Canteen section near Alton, Illinois (McKay, 1979b). Midsection beds there containing wood and peat yielded an age of ca. 43.6 cal k.y. B.P. (44.4–42.6 cal k.y. B.P.; McKay, 1979a), an age roughly corresponding with the onset of slow SAR at Lake Ben at 42 ± 2 cal k.y. B.P.

The transition from the Athens to Michigan subepisodes is manifested in core Mason 1a by an abrupt change in elemental composition between depths of 1095 cm and 1086 cm (= 30.9 ± 0.5 cal k.y. B.P. and 30.5 ± 0.3 cal k.y. B.P.; Fig. 22; see also Figs. 13 and 15). The age of onset of this major shift in composition, 30.9 ± 0.5 cal k.y. B.P., is consistent with the onset of the Farmdale phase in central Illinois at ca. 31.1 cal k.y. B.P. (Curry et al., 2018b) and other age estimates in Illinois (Willman and Frye, 1970; Hansel and Johnson, 1996; Grimley et al., 1998; Grimley, 2000; Curry et al., 2011). Few radiocarbon ages constrain the maximum age of the onset of the LGM in the region affected by the south-central outlet glaciers of the Laurentide ice sheet. Peak LGM conditions for the Lake Michigan and Huron-Erie lobes occurred initially at ca. 25–24 cal k.y. B.P. with a readvance at ca. 22–21 cal k.y. B.P. (Heath et al., 2018; Curry et al., 2018b; Loope et al., 2018). The age is somewhat older than the onset of deposition of tasmanatids (cysts of the Late Devonian green algae Tasmanites and Leiosphaeridia) in the Gulf of Mexico during the last glaciation (28.5 ± 1.2 cal k.y. B.P.; Kohl et al., 2020). Eroded by the Lake Michigan, Saginaw, and Huron-Erie lobes, the cysts were derived from Late Devonian black shales and deposited within distal outwash and farther downstream. The slight differences in age may be attributed in part to nuances of age models used by each study.

Changes in the elemental composition of loess and related sediment have been used to interpret provenance of the dust sourced from outwash from outlet glaciers of the Laurentide ice sheet (Grimley, 2000; Curry and Grimley, 2006; Muhs et al., 2018a). Important sources for the Illinois River valley outwash and dust include dolomite and shale rich in illite and chlorite from the Lake Michigan and Green Bay lobes (Grimley, 2000). Determination of the provenance of elements in dust sourced from outwash originating from lobes farther afield from Lake Ben is complicated by mixing of material with different sources and from differing pathways as well as from reworked material (Grimley, 2000). Published analyses of inductively coupled (argon) plasma–mass spectrometry (ICP-MS) analyses and EDXRF are available for select North American till, loess, and slackwater lake sediment (e.g., Curry and Grimley, 2006; Dixon-Warren and Stumpf, 2010; Muhs et al., 2018a). The XRF analyses of pressed pellets of various particle-size fractions (<20 µm, 20–32 µm, 32–64 µm, and 62–125 µm) show that the chemical composition is not significantly controlled by grain size (Fig. 16). Hence, I interpret that changes in the elemental profiles of XRF core scans reflect differing sediment sources.

The composition of outwash from lobes of the central outlet glaciers of the Laurentide ice sheet was approximated from analyses of slackwater lake sediment in Mississippi River tributaries near St. Louis, Missouri (Curry and Grimley, 2006). Mineral and chemical differences were determined among varicolored beds of Alton and Michigan subepisode age. Alton subepisode material included intercalated beds of reddish or gray silty clay interpreted as having respective provenance from the Superior lobe or lobes west of the Mississippi River valley (e.g., the Des Moines and James lobes). Our analyses determined that (1) kaolinite, redness (a*), K-feldspar, and Ca-plagioclase were diagnostic of the Superior lobe; (2) smectite and quartz were hallmarks of the western lobes; and (3) younger material with abundant dolomite, illite, and chlorite was derived from the Lake Michigan lobe.

To the mineral and spectroscopic analyses, I added elemental compositions from ICP-MS analyses (<4 µm fraction) of 14 additional samples to determine elemental fingerprints of sediment from the Des Moines, Superior, and Lake Michigan lobes (Fig. 23; Curry and Grimley, 2006). Analyses of source tills are not fully processed, but preliminary analysis and those of Thorleifson et al. (2007) for tills from Minnesota agree with the results discussed below. Samples rich in Ca, Sr, Mg, and, less intuitively, B are attributed to the dolomite-rich Lake Michigan lobe. The high concentration of carbonate means lower proportions of silicate minerals and, hence, relatively low concentrations of Si and Al. Reddish Superior lobe–sourced samples contain relatively high concentrations of K, Th, Ti, Cu, and other transition metals. The gray Des Moines lobe source contributed smectite-rich samples uniquely rich in trace elements Cd, Ag, Au, Hg, and Se. In tandem with the XRF scans, these ICP-MS results indicate the elements that are key to identifying provenance (Table 7). Key elements in the XRF data are not part of the ICP-MS package (Rb, for example), and vice versa (Mg and C).

The loess and lake provenance and chronology above provide a framework for interpreting results from sediment core Mason 1a. This record adds unparalleled age control from ca. 42–15 cal k.y. B.P. based on 34 AMS radiocarbon ages, a detailed record of waning loess deposition, and profiles of 24 elements (and MS) at 1 cm intervals via EDXRF scanning.

High Cd, Ag, Se, and Hg values in Lake Ben sediment dating from 37.5–31.0 cal k.y. B.P. indicate eolian contributions primarily from outwash from lobes located largely west of the Mississippi River (the Des Moines and James lobes). High concentrations of Cd were noted in the fine matrix of Des Moines lobe tills of western Minnesota (Thorleifson et al., 2007). Low values of these elements in younger sediment indicate little contribution of western lobe material to proximal and distal loess in Lake Ben (Fig. 24). This result is somewhat surprising because of the increase in smectite content (relative to illite) observed in sediment younger than 18 cal k.y. B.P., which has been attributed to glacially eroded Cretaceous shales of the western Great Plains. These results indicate the smectite was derived from far-traveled dust possibly from northern Africa (e.g., Claquin et al., 1999) and independent of regional glacial activity. Other possibilities include smectite sourced from weathered, poorly drained soil material (Hughes et al., 1993; Wilson, 1999) or bedrock in the unglaciated Great Plains (e.g., Muhs and Bettis, 2000; Muhs et al., 2018b).

The well-dated Ca and MS profiles (Figs. 12A and 15) are robust evidence of the Lake Michigan ice lobe from 31.3–17.9 cal k.y. B.P. (Table 8; Fig. 25). The slow SAR and correspondingly low Ca associated with the Alton subepisode (ca. 50.0–31.3 cal k.y. B.P.) may reflect partial to total leaching of carbonate. LOI analyses indicate modest amounts of carbonate minerals in this interval (Fig. 21). The low contents of calcite and dolomite in the XRD analyses of the <2 µm fraction (Table 3; Figs. 20C and 20D) indicate that most of the carbonate minerals occur as silt. Of course, Ca occurs in other minerals, but in lower proportions (e.g., Ca-plagioclase contains 14% Ca by weight compared to 40% of calcite).

From 31.0–17.9 cal k.y. B.P., Ca and MS account for at least half of normalized data variability (Fig. 13). MS has been attributed primarily to silt-sized ilmenite and hematite grains common to Superior and Lake Michigan lobe tills and outwash, especially the former (Grimley, 2000; Curry and Grimley, 2006). Large values of Ca reflect the interval's abundant carbonate mineral content, which is also reflected in high Sr, Mg, and inorganic carbon. Although Ca and MS are positively correlated, they typically do not occur in the same proportions (Fig. 15). About 85% of the time from ca. 25.5–17.4 cal k.y. B.P., Ca is proportionally more abundant than MS. This age range corresponds to the time when the Lake Michigan lobe was forming moraines in Illinois (Curry et al., 2018b) and likely entraining Paleozoic sedimentary rocks. From ca. 31–26 cal k.y. B.P., the relatively high values of MS (relative to Ca) are attributed to subglacial entrainment and fluvial erosion of Precambrian mafic and metasedimentary rocks rich in hematite, magnetite, and ilmenite north of the Illinois and Michigan Basins in Wisconsin, Michigan, and Ontario, Canada (Grimley, 2000). Hence, I attribute this change from MS-rich to Ca-rich silt to a general southward shift of the locus of erosion along the flow path of the Lake Michigan lobe.

The well-dated element scans and MS profiles allow comparison in the rate-of-change of contributions from Lake Michigan lobe sediment to the redeposited loess in Lake Ben. The onset of loess deposition ramped up from background Ca concentrations of ≤50 ppm to ~150 ppm in 700 yr (32.0–31.3 cal k.y. B.P.; Table 8) and then exploded to >400 ppm within 30 yr. The same magnitude of Ca decline took more than 2500 yr from 19.5–17.0 cal k.y. B.P.

Compositional changes in the latter part of the Michigan subepisode interval of the Lake Ben core suggest gradual waning and the eventual end of contributions from Lake Michigan lobe outwash and added importance of Ti and Cu, elements attributed uniquely to the Superior lobe (Fig. 24). Most elements became proportionally more abundant during this time due to little or no dilution by carbonate minerals from the Lake Michigan lobe. Al is an important component of both the western lobes and the Superior lobe (Fig. 23). Yet, elements unique to the western lobes (Cd, Ag, Se, and Hg) occur in very low proportions. The paucity of magnetic minerals in this younger interval is likely related to their relative abundance in the 6φ–4φ particle-size fraction versus the 11φ–7φ fraction, in addition to possible weathering under reducing conditions (Grimley, 2000). Such conditions likely explain low MS values in older intervals at 25.9–25.1 cal k.y. B.P. and >35.0–31.9 cal k.y. B.P. (Fig. 12A).

A notable shift in mineralogy and chemistry of LGM loess in central Illinois was caused by the Lake Michigan lobe diverting the Mississippi River at 24.4 ± 0.1 cal k.y. B.P (20.35 ± 0.85 ¹⁴C k.y. B.P.; Curry, 1998). Before the diversion, outwash originated from the Des Moines and other western lobes, and the Superior, Green Bay, and Lake Michigan lobes. After diversion, loess was derived solely from the Lake Michigan lobe. This shift is reflected in loess composition by a decrease in MS and increase in illite in the mineralogy of the <2 μm fraction (Grimley, 2000). In the Lake Ben sediment core, the shift is likely missing due to a 40 cm core loss zone at depths 733–691 cm (corresponding with ages 24.75–24.25 cal k.y. B.P.). MS values change from prediversion values of 50–80 m³/kg to postdiversion values ranging from 10 m³/kg to 30 m³/kg (Fig. 26). Moreover, the abundance profile of S increases by an order of magnitude from an average of ~400 ppm to greater than 4000 ppm (normalized values of 0.1–0.5 in Fig. 26). Postdiversion increases in S owe to coal-bearing strata in the underlying Illinois Basin.

The chronology, chemical composition, and sediment texture of loessial Peoria Silt as lacustrine Equality Formation in Lake Ben indicate deposition slowing and ceasing during an ~3.3 k.y. transition at 21.3–18.0 cal k.y. B.P. (Fig. 7). During this period, sediment changed from calcareous gray laminated silt to weakly calcareous, coarsely bioturbated, organic-rich fine silt capped by a light gray gleysol leached of carbonate. Modern agricultural activity truncated the gleysol (Fig. 27). Within the gleysol, the upward decrease in sedge fossils, increase in trace fossils (worm burrows), and angular blocky structure attest to changes from shallow lacustrine to paludal conditions, culminating in pedogenesis under poorly drained conditions. The SAR model indicates rapid decline from a maximum of 0.24 cm/yr to less than half that in ~200 yr between 22.2 cal k.y. B.P. and 22.1 cal k.y. B.P., reaching near background <0.03 cm/yr by 20.8 cal k.y. B.P. (Fig. 28). After a short uptick to 0.04–0.06 cm/yr from 19.5–17.9 cal k.y. B.P., background rates below 0.025 cm/yr resumed by 17.5 cal k.y. B.P. Two well-preserved sedge fossils preserved in 2-cm-thick, black organic-rich beds in duplicate cores of run 2 returned ages of 16.00 cal k.y. B.P. and 16.09 cal k.y. B.P. (Fig. 27), anchoring this part of the Bayesian age-depth relationship (Fig. 7). The organic-rich layers likely represent a brief pause in sedimentation that allowed plants to grow, but then a spurt of sedimentation buried the fossils against oxidation. Complementary interpretations of sedimentary and biological structures include upward particle-size fining, a decrease in the particle-size ratio of 6φ–4φ versus 10φ–7φ, and an increase in smectite/illite (Fig. 28).

Finer discrimination of arrival and cessation times of proximal and distal loess is indicated by DCA. The arrival of proximal loess between 30.9 cal k.y. B.P. and 30.7 cal k.y. B.P. is best expressed by (Ca + MS + Zr)/(Al + Zn + Cu) (Fig. 29). The cessation of proximal loess deposition, between 18.0 cal k.y. B.P. and 17.9 cal k.y. B.P, is better expressed by (Ca + MS + Zr)/(Al + K + Si), with trace amounts of distal loess persisting until ca. 16.2 cal k.y. B.P. (Fig. 29). Although the high smectite content indicates kinship with far-traveled western lobe loess, the “fingerprint” elements Cd and Ag are nearly absent (Fig. 24; Thorleifson et al., 2007). At present, we do not know the precise provenance of Cd and Ag nor the minerals with which they co-occur. However, this evidence suggests that much of the material incorporated into the white gleysol came from far-traveled dust not sourced from outwash of the Des Moines and James lobes.

The nature and age of proximal loess cessation are consistent with the building of the Tinley moraine, which retarded meltwater flow south from the Chicago outlet. Several minimum-limiting radiocarbon ages of fossil Dryas integrifolia rootlets, leaves, and stems indicate a minimum age of 17.6 cal k.y. B.P. for this moraine (Curry et al., 2018a, 2018b). Later drainage flowed along the south margin of Laurentide ice, debouching into the St. Lawrence and Mississippi Rivers. Proglacial Lake Chicago and large slackwater lakes in the middle Illinois River valley trapped silt-laden sediment (Curry et al., 2014). This setting starved the Manito terrace surface and Lake Ben of silt while Lake Chicago existed in the southern Lake Michigan basin. The chronology is consistent with the upper age limit of 17.6 cal k.y. B.P. from tasmanatids in the Mississippi River delta in the Gulf of Mexico (Kohl et al., 2020).

The ca. 18 cal k.y. B.P. age is associated with events indicative of global circulation. The same age is assigned for the youngest suite of LGM moraines in the Southern Alps of New Zealand. In the Northern Hemisphere, 18 cal k.y. B.P. is the age assigned to the end of significant meltwater discharge from western European glaciers to the Bay of Biscayne (Denton et al., 2022). The interhemispheric connection has been attributed to orbital forcing of Southern Hemisphere circulation, namely, latitudinal shifts of austral westerlies (Denton et al., 2021, 2022). In central North America, the 18 cal k.y. B.P. age attests to a significant, albeit not total, shift from south-central Laurentide ice sheet meltwater routes conveyed primarily by the Mississippi River to the St. Lawrence River—a change in riverine debouche points to the sea of ~17° latitude (28°N for the LGM Mississippi River and 45°N for the St. Lawrence River). Post–18 k.y. overflow across the Chicago outlet during the Glenwood and Calumet phases (ca. 17.6–15.0 cal k.y. B.P. and 14.0–12.6 cal k.y. B.P.; Curry et al., 2018a) was captured by slackwater lakes in the middle Illinois River valley, where the water likely warmed before either evaporating or overflowing to the Mississippi River and thence to the Gulf of Mexico. The cessation of proximal loess deposition in Lake Ben at 18.0–17.9 cal k.y. B.P. is compelling supporting evidence for interhemispheric synchronicity of cessation of glacier-supporting atmospheric and oceanic circulation and climatic patterns.

Several element profiles through the proximal, transitional, and distal loess facies are remarkably like those for the modern soil developed in the Peoria Silt at Cottonwood School (located in Fig. 1; Mason and Jacobs, 1998), particularly the upward increase in Al (Fig. 30), as well as the ratios of fine-to-coarse silt and clay percentages (Fig. 28). Mason and Jacobs (1998) noted that the timing of these changes in the modern soil was speculative due to bioturbation of the upper solum as well as the lack of reliable dateable material due to carbon cycling and bioturbation. Comparison of Al profiles from the Lake Ben sediment record suggested that ~85% of the soil at Cottonwood (from a depth of 25 cm to 172 cm) reflects transitional deposition of proximal to distal loess deposited from 18.0–13.3 cal k.y. B.P. I justify the one-to-one correlation of the Lake Ben and Cottonwood School profiles without using mass-balance techniques because both sites involve Peoria Silt proximal to the sediment source (albeit deposited in different settings). Corresponding to the lower 6 cm of plowing, the upper bounding age of ca. 13.3 cal k.y. B.P. is approximate and ~6 cm above the youngest radiocarbon age in the Lake Ben age model.

The fine texture and mineralogical characteristics in B horizons of surficial soils developed in loess in this region may thus be partly attributed to accreted eolian additions to the soil, and not entirely to in situ weathering. This observation may account for some differences noted between the last interglacial Sangamon Geosol and the modern soil, such as the latter's lack of randomly interstratified kaolinite-smectite, an index mineral found in Btg horizons of long-lived (~100 k.y.) interglacial soils (Hughes et al., 1993; Grimley et al., 2003). Instead, trends of element concentrations such as Ti and Cu and chronology are consistent with distal loess derived in part from the outwash of the Superior lobe and contributions from global dust. The latter may be represented by a rise in elements not unique to contributing lobes, such as the Lake Michigan lobe (MS and Ca) and western lobes (Cd and Ag). Of course, the modern soil also has evidence of physical and chemical weathering, as attested by pitted and etched grains of feldspar and hornblende (Cremeens et al., 1992). However, the similarity of Al concentration profiles in Peoria Silt at Cottonwood School and Lake Ben (given reasonable similarity in bulk density) indicates that much agricultural soil in Illinois was inherited from accreted distal eolian sediment.

Investigation of sediment cores sampled from two kettle lakes located along the middle Illinois River in the type region of the Peoria Silt revealed that eolian silt ceased accumulating as proximal loess by 18.0 cal k.y. B.P. and transitioned to distal loess by ca. 16.2 cal k.y. B.P. Deposition from ca. 16.2–14.7 cal k.y. B.P. is attributed to global dust that was relatively rich in fine silt, clay, and smectite.

A surprising result is the record of loess deposition during MIS 3 (57–29 cal k.y. B.P., equivalent to the Athens subepisode in Illinois). Lake Ben's record includes a long period (43–31 cal k.y. B.P.) of slow sediment accumulation, with modestly higher rates prior to and after that interval (Figs. 7 and 19). These results invite investigation of the rates of accumulation, age, and composition of thick deposits of the Roxana Silt along the lower Illinois River valley, especially with regard to evidence of glacial activity during that time in the upper Great Lakes region (e.g., Curry, 1989; Curry and Pavich, 1996; Ceperley et al., 2019; Kerr et al., 2021; Dalton et al.., 2022a, 2022b) and new evidence for an active Mackinaw lobe during this time in the northern part of the southern peninsula of Michigan (Schaetzl, 2023).

The changes in sediment character (mineralogy, chemistry, particle size, age) during slowing and eventual cessation of Peoria Silt deposition in Lake Ben are matched by adjacent terrestrial pedogenesis. Upticks in the concentration (and presumed mass balance) of Al in these matching profiles indicate that high levels or upward trends of other elements, such as Pb and Co (Dreher and Follmer, 2004), may be naturally occurring and not necessarily due to anthropogenic additions.

This paper was made possible through the assistance of numerous people over the past ~30 years. Donald Nannen graciously gave permission to core Smith Lake in 1995 and 2017, followed in 2019 by permission from his son Ben Nannen to core Lake Ben. The first Smith Lake cores were sampled through a joint effort with Eric Grimm, Russell Graham, and others with the Illinois State Museum, followed more recently by a herculean effort by Tom Lowell (University of Cincinnati) and his students. Many thanks to students of Jessica Conroy's paleolimnology class (University of Illinois Urbana-Champaign) who worked on class projects related to the Smith Lake cores. Most accelerator mass spectrometer samples were pretreated and assayed at the Keck Carbon Cycle Accelerator Mass Spectrometer Facility (University of California, Irvine). X-ray fluorescence, magnetic susceptibility, and spectroscopy core scans were accomplished at the Yarrow Axford laboratory (Northwestern University) by Peter Puleo; initial magnetic susceptibility/spectroscopy scans were done at the Limnological Research Center (University of Minnesota, Twin Cities). Identification of Callitriche hermaphroditica fruit/seeds was made by Richard V. Lansdown (Ardeola Environmental Services, Gloucestershire, UK). I am grateful to the peer review by Henry Loope and GSA Bulletin associate editor Richard Waitt, whose suggestions improved the manuscript.