Paleomagnetic secular variation for a 21,000-year sediment sequence from Cascade Lake, north-central Brooks Range, Arctic Alaska

Abstract. Two > 5-m-long sediment cores from Cascade Lake (68.38° N, 154.60° W), Arctic Alaska, were analyzed to quantify their paleomagnetic properties over the past 21,000 years. Alternating-field demagnetization of the natural remanent magnetization, anhysteretic remanent magnetization, isothermal remanent magnetization, and hysteresis experiments reveal a strong, well-defined characteristic remanent magnetization carried by a low coercivity magnetic component that increases up core. Maximum angular deviation values average < 2°, and average inclination values are within 4° of the geocentric axial dipole prediction. Radiometric ages based on 210Pb and 14C were used to correlate the major inclination features of the resulting paleomagnetic secular variation (PSV) record with those of other regional PSV records, including two geomagnetic field models and the longer series from Burial Lake, located 200 km to the west. Following around 6 ka (cal BP), the ages of PSV fluctuations in Cascade Lake begin to diverge from those of the regional records, reaching a maximum offset of about 2000 years at around 4 ka. Several correlated cryptotephra ages from this section (reported in a companion paper by Davies et al., this volume) support the regional PSV-based chronology and indicate that some of the 14C ages at Cascade Lake are variably too old.



Introduction 20
In paleoclimate and paleo-environmental studies of lake sediments, a firm chronological framework is needed to place observations and corresponding interpretations into regional and global contexts. For Holocene and late Pleistocene lacustrine studies, 14 C dating is the most common technique for developing such frameworks. However, dating Quaternary lake sediments in the Arctic using 14 C can be problematic due to scarcity of terrestrial macrofossils, and the inheritance of "old carbon" eroded from the surrounding landscape (Gaglioti et al., 2014). Another challenge is the possibility of a hard-water effect, which 25 involves the incorporation of radioactively inert bicarbonate ions into aquatic organisms (e.g., Rea and Colman, 1995;Moore et al., 1998). Researchers in the Arctic and elsewhere have therefore turned to supplementing existing 14 C chronologies with additional methods, including paleomagnetic data (e.g., Lisé-Pronovost et al., 2009;Barletta et al., 2010;Ólafsdóttir et al., 2013) and tephrochronology (Lowe et al., 2015;Davis et al., this issue).
The latest geomagnetic field models provide estimates of site-specific inclination, declination, and field intensity for the past 30 10 kyr (CALS10k.1b; Korte et al., 2011) and 9 kyr (pfm9k.1b; Nilsson et al., 2014). Although these models are not perfectly constrained for the Alaskan Arctic and they are inherently smoothed both spatially and temporally, their major directional and intensity features can be used for age control by correlation with site-level records (e.g., Barletta et al., 2010). In addition to correlations afforded by field models, individual site-level paleomagnetic records supported by well-constrained geochronological control serve as references for correlated ages based on wiggle matching of prominent fluctuations. 35 We present a u-channel and hysteresis-based analysis of Cascade Lake, Alaska, paleomagnetic properties and demonstrate that the sediments are suitable for recording directional and relative paleo intensity (RPI) variability since around 21 ka. We then present a PSV-derived age model for Cascade Lake sediments, constructed from correlated age control points using both field https://doi.org/10.5194/gchron-2021-19 Preprint. Discussion started: 16 June 2021 c Author(s) 2021. CC BY 4.0 License. models and an inclination record from Burial Lake, Alaska (Dorfman, 2013). This age model, together with new correlated ages based on cryptotephra identifications (Davies et al., this volume), indicates that our PSV-derived age model provides 40 improved age control over the 14 C data alone.

Cascade Lake
Cascade Lake (68.38° N, 154.60° W; 990 m asl; Fig. 1) is located on the northern front of the north-central Brooks Range, slightly east and outside of the main Kurupa River valley. The lake has an area of ~ 1 km 2 and reaches a maximum depth of ~ 45 40 m. According to surficial mapping by Hamilton (1980), the Cascade Lake basin was likely glaciated during the Itkillik I (early Wisconsin) but not during the Itkillik II glaciation (late Wisconsin) (Hamilton, 1982); however, Itkillik II ice likely entered the drainage of Cascade Lake and may have delivered meltwater to the lake at that time. Cascade Lake presently has no significant inflow, and the only small outlet flows westward to Kurupa Lake (~ 920 m asl). The small catchment size (~ 10 km2) and lack of major inflow likely results in lake deposits dominated by remobilized hillslope sediment derived from till, in 50 conjunction with variable fluxes of eolian and biogenic sediment. Figure 1: Cascade Lake bathymetry with core sites (2 = CASC-2, 4 = CASC-4). Inset shows the location of Cascade Lake in the northcentral Brooks Range, Alaska, along with the locations of previously published paleomagnetic secular variation and relative paleointensity 55 records used for comparison with this study: Burial Lake (BUR, Dorfman, 2013); Grandfather Lake (GFL, Geiss and Banerjee, 2003); Gulf of Alaska (85JC, Walczak et al., 2017); Chukchi Sea (8JPC, Lisé-Pronovost et al., 2009); Beaufort Sea (803, Barletta et al., 2008). https://doi.org/10.5194/gchron-2021-19 Preprint. Discussion started: 16 June 2021 c Author(s) 2021. CC BY 4.0 License.

Geomagnetic Setting
The relative proximity of Cascade Lake to the magnetic north pole may make the deposits especially sensitive to geomagnetic 60 field variations (Cox, 1970); however, this concept is incompletely tested. Experiments by Aurnou et al. (2003) suggest that fluid movement of "polar vortices" in this region of the outer core may be manifested as geomagnetic field behavior within the "tangent cylinder" (poleward of ~69.5° N and 69.5° S) that is unique to variability outside of the "tangent cylinder" (St-Onge and Stoner, 2011). These potential behaviors of the geomagnetic field are not well understood, however, and recent observations from Svalbard (Ólafsdóttir et al., 2019) suggest mid-latitude geomagnetic dynamics, at least for that part of the 65 polar regions are more likely to apply. Cascade Lake also lies at the edge of the "North American Flux Lobe", one of several regions of concentrated magnetic flux observed from the historical time averaged field (e.g., Bloxham and Gubbins, 1987;Bloxham et al., 1989). Recent research suggests that these flux patches may display oscillatory behavior (Gallet et al., 2009;Stoner et al., 2013), and could be related to heterogeneities in the lower mantle (e.g., Amit et al., 2010;Korte and Holme, 2010) that, along with the possible impact of the "tangent cylinder", could result in complex field behaviors in these locations. 70

Core recovery and initial analyses
Cascade Lake sediment cores were collected in 2013 using a single-drive (7.7 cm inner-diameter tubing), percussion, piston coring system from the frozen lake surface. Core CASC-4A (5.53 m long) was collected near the depocenter (33.8 m deep) of the primary basin, and CASC-2D (5.41 m long) is from the shallower (18.9 m deep) sub-basin (Fig. 1). An Aquatic Instruments 75 Universal corer was used to recover the surface sediment at each site. Cores were shipped to the National Lacustrine Core Facility, University of Minnesota, where whole cores were scanned with a Geotek MSCL-S fitted with a gamma-ray density sensor to measure wet bulk density (WBD) at 0.5 cm resolution. Cores were then split, described, and imaged using a Geotek Corescan-III line scanner. Magnetic susceptibility and color spectrophotometry were measured using a Bartington point sensor (MS2E) and a Konika Minolta color spectrophotometer, respectively, at 0.5 cm intervals. Additional data available from site 80 CASC-4 include biogenic silica measurements at 5-10 cm intervals and loss-on-ignition (LOI) to determine organic matter (OM) content at 2-5 cm intervals. The analytical methods and further interpretation of these results are provided in Steen (2016).

Magnetic analyses
Sediment cores were sampled for paleomagnetic analysis with rigid plastic 2 x 2 cm u-channels (up to 1 m long), which were 85 taken as close as possible to the center axis of the core, unless either side of a core section appeared less disturbed by minor coring deformation. Low-field magnetic susceptibility (kLF) was measured at 1 cm intervals using an automated u-channel magnetic susceptibility track fitted with a 35 mm Bartington MS loop MS2C metering coil at the Paleo-and-Environmental Magnetism Laboratory (Pmag Lab), Oregon State University. Remanence measurements on u-channels were acquired using a 2G Enterprises model 755-1.65UC cryogenic superconducting rock magnetometer equipped with in-line alternating field (AF) 90 demagnetization coils for automated AF demagnetization and measurement and an in-line direct current (DC) coil for anhysteretic remanent magnetization (ARM). An off-line pulse magnetizer was used for imparting an isothermal remanent magnetization (IRM) to the u-channels. Magnetic data processing and visualization was completed using a modified version of the UPmag MATLAB software program (Xuan and Channell, 2009).
The natural remanent magnetization (NRM) of u-channels was studied through progressive AF demagnetization of the X, Y, 95 and Z axes and measurement using 16 steps from 10 to 100 mT. Anhysteretic remanent magnetization (ARM) acquisition was performed by imparting a small (0.05 mT) DC field during AF demagnetization on the Z axis, at steps identical to the NRM.
The ARM acquired at peak AF of 100 mT was also AF demagnetized at the same steps as the NRM. Isothermal remanent https://doi.org/10.5194/gchron-2021-19 Preprint. Discussion started: 16 June 2021 c Author(s) 2021. CC BY 4.0 License. magnetizations (IRMs) were imparted with the off-line pulse magnetizer using fields of 0.1, 0.3, and 1.0 T, with the latter nominally representing a saturation isothermal remanent magnetization (SIRM). IRMs imparted at 0.1 T were not 100 demagnetized, and 0.3 T IRMs were AF demagnetized at same 16 steps as the NRM. The SIRMs were demagnetized at three steps (30, 60, and 100 mT). Due to the response function associated with the pick-up coils of the magnetometer, each measurement integrates across a ~7 cm interval (Oda and Xuan, 2014). Therefore, the first and last 5 cm from each u-channel segment are not considered. Additionally, NRM, ARM, ARM acquisition, and IRM were not measured for two short segments (CASC-4A-2a and CASC-2D-2a,13 and 14 cm, respectively). 105 Because the cores were not azimuthally oriented upon retrieval, declination data were rotated such that each u-channel section has a mean declination of 0°. This rotation correction is unlikely to be a perfectly accurate representation of declination, but serves as an objective transformation to meaningfully visualize the declination data. Present declination is ~ 16.5° E at Cascade Lake, which could be indicative of the magnitude of the expected offsets.
Hysteresis experiments on subsampled bulk material were carried out using a Princeton Applied Research Vibrating Sample 110 Magnetometer (VSM), with the purpose of quantifying changes in magnetic grain size and magnetic mineralogy at the Institute of Rock Magnetism, University of Minnesota. Saturation magnetization (Ms), saturation remanence (Mrs), coercivity (Hc) and coercivity of remanence Hcr-dot calculated from the hysteresis loops themselves (after Fabian and von Dobeneck, 1997) were derived from hysteresis loops measured to a 1000 mT saturating field and corrected for paramagnetic contributions using the high field slope above 800 mT. 115

210 Pb and 14 C dating
Core CASC-4B was collected from within 10 m of CASC-4A and the two were correlated by using visual stratigraphy. Core CASC-4A was sampled for 210 Pb analysis at 0.5 cm intervals over the uppermost 10 cm. 210 Pb content was estimated by direct measurements of 210 Po activity. 226 Ra activity was also analyzed at 10.0-9.5, 6.5-6.0, and 1.5-1.0 cm below lake floor (blf) to determine levels of unsupported 210 Pb present in the sediment. A constant-rate-of-supply (CRS) age model (Appleby and 120 Oldfield, 1978) was applied to allow for changing initial concentrations of unsupported 210 Pb with changing sedimentation rates.
The radiometric-based age model of Cascade Lake was constrained by 11 accelerator mass spectrometry (AMS) 14 C dates.
Dated material consisted of various terrestrial plant macrofossils, insect parts, resting eggs, and aquatic vegetation, which were analyzed at the Keck AMS Facility at University of California, Irvine. 14 C dates were calibrated with CALIB 7.1 (Stuiver et 125 al., 2005) using the IntCal13.14C Northern Hemisphere terrestrial calibration curve (Reimer et al., 2013). AMS 14 C dates obtained from core CASC-4B were combined with CASC-4A dates through visual stratigraphic matching to construct a composite CASC-4 age model. Calibrated 14 C dates are reported as the median of the probability distributions, with 1σ error ranges. CLAM software for classical age modeling (Blaauw, 2010) was used to fit the calibrated 14 C age distributions with a smooth spline with a smoothing parameter of 0.8. The 95% confidence interval from the distribution of 1000 iterations is used 130 as the uncertainty in the age-depth model.

General lithologic and magnetic stratigraphy
The sediments recovered in cores from both subbasins of Cascade Lake are undeformed by the coring procedure. The sequence can be separated into three distinct lithologic units based on visual stratigraphy, wet bulk density, organic-matter content, and 135 variations in magnetic parameters ( L1 (562-393 cm). Lithologic unit 1 (L1) is a pebbly, brown diamicton with a matrix of clay and silt. It contains glacially striated pebbles 1-3 cm in diameter. This section was not sampled using u-channels. There is no clear evidence of an 140 unconformity or depositional hiatus between the diamicton (L1) and the overlying lacustrine unit (L2) in CASC-4A or CASC-2D.
L2 (393-355 cm). Unit 2 (L2) is composed of beige-brown, diffusely to sharply laminated silt and clay that transitions gradationally to grayish-brown silt and clay near the top of the unit. WBD values (1.5-2.0 g cm -3 ) are significantly lower than in unit L1, and OM steadily increases upward through the unit. Unit L2 exhibits a spike in magnetic parameters relative to unit 145 L3, with kLF exhibiting a low of 4.50 x 10 -5 at 382 cm to a high of 1.06 x 10 -4 SI at 361 cm, ARM acquisition at 100 mT AF exhibits a low of 7.89 x 10 -3 at 381, to a high 2.95 x 10 -2 A m -1 at 363 cm), and IRM0.3T at 0 mT AF exhibits a low of 4.81 x 10 -1 at 380, to a high 2.03 A m -1 at 355 cm) and SIRM at 0 mT AF exhibits a low of 8.97 x 10 -1 at 380, to a high 2.35 A m -1 at 355 cm. (Fig. 2). A ~5-cm-thick interval of olive-green to gray, faintly laminated sediment near the top of unit L2 in CASC-2D is visually similar to unit L3. This distinctly colored interval corresponds with high intensities in all parameters that are 150 slightly offset in depth suggesting a complex magnetic mineralogic suite that hint at the authogenic creation of greigite, but without further analysis are beyond the scope of this manuscript to identify.
L3 (355-0 cm). Unit 3 consists of irregular, millimeter-to centimeter-scale bands of olive-green, greenish-gray, and brown silt and clay. Unit L3 is similar in thickness in the two cores (311 to 0 cm in CASC-2D). Magnetic properties are in general more variable in the lower part of the unit, becoming more consistent up core. kLF is, however, somewhat different and similarly 155 variable throughout, decreasing from highs around 5.5 x 10 -5 SI at the base of the unit to around 3 x 10 -5 SI at the top. ARMaq at 100 mT are generally lower 1.8 x 10 -2 and variable (1σ = 4.2 x 10 -3 A m -1 ) below 220 cm and higher ~2.8 x 10 -2 and variable (1σ = 4 x 10 -3 A m -1 ) above, with a high value of 3.82 x 10 -2 A m -1 at 157 cm. IRM0.3T at 0 mT AF are less variable with mean values of 0.56 A m -1 and 1σ of 0.11 A m -1 . SIRM at 0 mT AF are slightly higher with a mean 0.7 A m -1 and 1σ of 0.12 A m -1 .
Core CASC-2D displays a dark, highly magnetic layer (IRM0.3T = 9.26 A m -1 ; 307 cm blf) that forms a sharp contact with 160 laminated sediments below, but grades diffusely into laminated sediments above. Immediately up-core of this interval, IRM0.3T does not cleanly decrease in intensity with higher AF suggesting that the SQUID electronics are unable to keep up with the change in counts resulting in artifacts within the subsequent magnetic measurements. This is a common problem associated with overly strong samples that persists even after several AF demagnetization steps. Therefore, IRM0.3T data from ~300-251 cm blf in CASC-2D are disregarded (shown shaded in Fig. 2), and SIRM for the u-channel section (387-251 cm blf) was not 165 measured.

S-ratios and kARM/kLF
Cascade Lake S-ratios display an overall increasing trend up-core (Fig. 2). CASC-4A contains an S-ratio peak at ~ 356 cm, at 175 the transition from L2 to L3. This peak is not visible in CASC-2D, due to complications in measuring IRM at 0.3 T and 1.0 T in the u-channel section containing the L2-L3 transition. S-ratios vary between 0.40-0.92 in Cascade Lake sediment, with an increasing up-core trend indicating an increasing proportion of low-coercivity magnetic material, an interpretation supported by the up-core increase in ARM (Fig. 2). https://doi.org/10.5194/gchron-2021-19 Preprint. Discussion started: 16 June 2021 c Author(s) 2021. CC BY 4.0 License.
Cascade Lake kARM/kLF and kARM/SIRM (Steen, 2016) and ratios also show an increasing trend, and reach a clear minimum in 180 L2 (Fig. 2). Assuming these are mostly driven by the ferrimagnetic assemblage, these ratios suggest a coarser magnetic grain size in L2 compared to L3, and generally decreasing up-core magnetic grain sizes in L3 consistent with the progressive addition of a separate fine-grained ferrimagnetic component to the magnetic assemblage.

Hysteresis, magnetic grain size and mineralogy
Hysteresis experiments were performed on ten representative samples from the three units in core CASC-4A. After slope 185 correction for paramagnetic contributions, data from the three units exhibit distinct hysteresis loops and positions on conventional plots (cf., Day et al., 1977;Wang and Van der Voo, 2004) (Fig. 3). However, the grain-size boundaries depicted are only a useful approximation as the proportion of high-coercivity materials, especially in L3 and possibly L2, are not directly applicable to a Day plot. Unit L1 hysteresis loops are characterized by high coercivities and a subtle "wasp-waisted" shape (Roberts et al., 1995;Tauxe et al., 1996); the data plot to the far right on a Day diagram, suggesting a significant 190 antiferromagnetic (e.g., hematite, goethite) contribution, with a fine ferrimagnetic component (Roberts et al., 2018). Unit L2 hysteresis loops are also "wasp-waisted", with an apparent antiferromagnetic and perhaps coarser ferrimagnetic contributions (Roberts et al., 2018). Unit L3 data are consistent with fine ferrimagnetic material, approaching single-domain threshold on a Day diagram on top of substantial concentration. Samples from L1 and L3 plot within the pseudo-single domain (PSD) range (~0.1-10 µm for Ti-poor magnetite) on a typical Day plot (Day et al., 1977), while the L2 sample falls just outside of this boundary (Fig. 3b). High Mr/Ms values for L3 samples ( Fig. 3a,b,e) suggest the presence of very fine PSD, or even SD, magnetite within this unit (Roberts et al., 1995). Hysteresis measurements indicate that L2 (Fig. 3d) may incorporate significant amounts of both ferrimagnetic and antiferromagnetic 205 material, while L1 (Fig. 3c) is likely dominated by high-coercivity material (e.g., hematite) and a smaller proportion of PSD magnetite (Roberts et al., 1995).
The kARM vs. kLF calibration of King et al. (1982) also suggests extremely fined grain magnetic grain sizes of < 0.1 µm for L3, and somewhat coarser L2 magnetic grain sizes of < 5 µm (Fig. 4). While this relationship likely accurately represents the relative magnetic grain size relationship between L2 and L3, these plots should not be interpreted as absolute grain size 210 estimates. However, due to the increased influence of high coercivity material (hematite) on the SIRM and hysteresis data, this relationship (kARM/kLF) may be a more realistic indicator of relative magnetite grain size at Cascade Lake.  King et al. (1982) for samples of dispersed (~ 1% by volume) equidimensional magnetite grains. Dashed lines are empirically derived characteristic grain sizes. L2 (orange squares) and L3 (gray circles) are plotted separately to discern lithologically-separate differences in magnetic grain size. L1 (glacial till) is not included because that interval was not sampled with u-channels.

Characteristic remanent magnetization
Cascade Lake sediments carry a generally strong, stable, and well-defined NRM signal (Fig. 5). Component inclination and 220 declination were calculated using Principal Component Analysis (PCA; Kirschvink, 1980) from the UPmag software (Xuan and Channell, 2009) over the 20-70 mT (11 steps) AF demagnetization range after removal of a low coercivity viscous remanent magnetization intermittently observed throughout the u-channels (Fig. 6) (80.5°) and CASC-2D (74.8°) are close to, and for much of each record, they vary around the predicted geocentric axial dipole (GAD) inclination for Cascade Lake (78.8°), indicating that these sediments are reliable recorders of geomagnetic direction.

235
(GAD) inclination for Cascade Lake (78.8°). Yellow bars indicate intervals with MAD values greater than 5°. Because core sections are not azimuthally oriented, and because each section is oriented differently, the declination record of each u-channel section has been rotated to a mean of 0°.

Normalized remanence (relative paleointensity)
To assess whether these sediments record variations in geomagnetic field strength, we explore three commonly used normalization approaches over the stable ChRM demagnetization range (NRM20-70 mT). First, NRM at the 30, 45, and 60 mT 245 demagnetization steps was normalized using kLF (NRM/kLF). Second, the slope of the NRM at demagnetization ranges 20-50, 20-70, and 40-70 mT was normalized using the slope of the ARM at the same demagnetization steps (NRMX-YmT/ARMX-YmT).

Composite paleomagnetic record 260
Due to the missing upper sediment, shallower coring depth, increased lithologic variability and higher MAD values in CASC-2D, we focus instead on CASC-4A as the main paleomagnetic record from Cascade Lake. However, CASC-2D contains a thicker version of unit L2. To create a more detailed composite sequence, the data from unit L2 in core CASC-2D were spliced onto the entirety of the paleomagnetic data from core CASC-4A by matching their inclination records. The data from 473-343 cm blf from CASC-2D were used to extend the paleomagnetic record downward to a composite depth of 520 cm blf, 265 specifically by setting 389 cm depth in CASC-4A equal to 343 cm depth in CASC-D2. This match is supported by the transition from units L2 to L3 in each core, which agrees with the inclination match. The match also suggests that the CASC-2D paleomagnetic record is missing > 40 cm of sediment at the top of the sequence when compared to CASC-4A. The resulting composite sequence (CASC-4A-2D) therefore comprises the quality and completeness of CASC-4A for the upper 389 cm, plus the data from the expanded L2 section in CASC-2D over the lower 131 cm. This composite L2-L3 sequence is underlain 270 by glacial diamicton (L1) in both cores, which was not analyzed for continuous magnetic properties.

Radiometric age model
Based on CRS modeling of 210 Pb data (Table 1) (Appleby and Oldfield, 1978), the uppermost 3.6 cm of sediment from site 285 CASC-4 was likely deposited within the last ~140 years (Fig. 8). Uncertainty in the CRS age model was estimated by firstorder propagation of counting error after Binford (1990), yielding an age range of 150-135 years (95% C.I.) at 3.6 cm blf. This depth is interpreted as the limit of meaningful age estimation by 210 Pb dating, and suggests sedimentation rates of 24-27 cm kyr -1 in the uppermost portion of CASC-4A.
The 210 Pb age profile was combined with 14 C dates (Table 2) to generate a provisional age-depth model for site CASC-4 ( Fig.  290   8). Two 14 C dates (UCIAMS 134422, 128096) were rejected as outliers; both are anomalously old relative to the down-core trend of the surrounding 14 C dates. On the basis of this age model, the average sedimentation rate is 23.7 cm kyr -1 , as evaluated to the deepest 14 C age at 351 cm blf. The extrapolated age of the unit L2-L3 boundary at 355 cm is estimated at 15.2 ka (15.6-14.1 ka, 95% C.I.). The average 95% confidence interval over the entire radiometric age model is ± 345 years.   Table 1). The constant-rate-of-supply (CRS) model was used for 210 Pb age-depth modeling (after Appleby and Oldfield, 1978), and error (2σ) at the bottom of each sampled interval was calculated using a first-order propagation of error 300 technique (Binford, 1990). (B) 14 C dates (Table 2) were fit with a smooth spline with a "smoothing parameter" of 0.8 using CLAM software (Blaauw, 2010). Rejected dates are shown in red. Dotted lines define the 95% confidence interval for the CLAM age model.  C dates from surface core CASC-4B. All other dates are from CASC-4A.

Magnetic assemblage 305
Hysteresis experiments, remanence measurements and ratios (S-ratios, kARM/kLF) document an up-core transition (L1 to L3) from a magnetic assemblage dominated by high coercivity, likely antiferromagnetic (e.g., hematite) minerals as demonstrated by S-ratios, to one where very fine-grained ferrimagnetic (e.g., (titano)magnetite) components are a much more important part of the assemblage, although high-coercivity components are still present. This transition reflects the deglacial and environmental evolution of the catchment, with unit L1 comprising minerogenic sediment glacially sourced from local 310 bedrock. In contrast, unit L3 is magnetically dominated by low-coercivity magnetic material (e.g., (titano)magnetite), but also incorporates a subordinate amount of high-coercivity material.

Development of PSV age model
Assuming the inclination changes recorded in Cascade Lake sediment accurately reflect the timing and variations of Earth's magnetic field, comparisons with other inclination records can be used for stratigraphic correlations, and for geochronological 315 purposes when well-dated records are available. Due to the possibility of relatively localized geomagnetic field behavior, we focus our comparisons with existing regional records mainly on Burial Lake (Dorfman, 2013), the one closest to our study site in the north-central Brooks Range. We visually matched prominent features (tie points) in the Cascade Lake inclination record to geomagnetic field models and the Burial Lake paleomagnetic record. The field models used for the inclination comparisons are CALS10k.1b (Korte et al., 2011) and pfm9k.1b (Nilsson et al., 2014); both provide inclination, declination, and field 320 intensity predictions for any given site coordinates over the specified time interval. The CALS10k.1b model incorporates archeomagnetic data from the GEOMAGIA50 database (Donadini et al., 2006;Korhonen et al., 2008) as well as 75 selected sedimentary paleomagnetic records from the SED12k data compilation (Donadini et al., 2009;Korte et al., 2011). The pfm9k.1b model uses a dataset similar to that of Korte et al. (2011), but excludes sedimentary sequences dated by using 14 C on bulk sediment, archeomagnetic data with large temporal uncertainties, and some records with paleomagnetic behavior 325 incompatible with the majority of others. Both field models have an estimated temporal resolution of around ± 500 years, based on sediment age uncertainties and bootstrap sampling (Korte et al., 2011) as well as comparisons of model power spectra at different coordinates (Nilsson et al., 2014). The Burial Lake 14 C age model is likely more reliable than the Cascade Lake 14 C age model because Burial Lake 14 C samples only include terrestrial macrofossils (decreasing the likelihood of a hard-water or reworked old carbon effects) and because the sedimentation rate at Burial Lake is rather linear over the past ~ 17 kyr 330 (Dorfman, 2013;Dorfman et al., 2015;Finkenbinder et al., 2015). We adopted the published 95% confidence intervals from the published 14 C age model for the age uncertainty of the age-control points. https://doi.org/10.5194/gchron-2021-19 Preprint. Discussion started: 16 June 2021 c Author(s) 2021. CC BY 4.0 License.
We developed two alternative PSV inclination-matched age models (Steen, 2016), but focus here on the more likely version (PSV-1). It was constructed from 14 tie points selected by visual wiggle-matching of the prominent features in the Cascade Lake inclination record with CALS10k.1b (9 tie points), pfm9k.1b (8 tie points), and the Burial Lake records (9 tie points) 335 ( Fig. 9; Table 3). Not all tie points are present in all records due to the limited age of the CALS10k.1b and pfm9k.1b models (10 ka and 9 ka, respectively) and limited prominent features in the upper part of the Burial Lake record. The depths of all tie points and their age uncertainties were fit with a smooth spline using CLAM software to generate the Cascade Lake PSV-1 age model (Fig. 10). Where the ages of tie points from Burial Lake are somewhat older than those from the field models, between about 270 and 170 cm, the PSV-1 age modeled more closely follows the field models because it is based on a higher 340 density of control points.
The PSV-1 age model and the radiometric age model overlap within their uncertainties from the base of the radiometrically dated section, at around 365 cm, up to around 160 cm (between ~ 16 and 6 ka) (Fig. 10). The PSV-1 age model can be extended beyond the radiometric age model to ~ 21 ka using the tie points between the Cascade Lake inclination record (CASC-4A-2D depth scale) correlated with the similar record at Burial Lake (Dorfman 2013), although the radiocarbon age-depth model for 345 this portion of the Burial Lake sediment core is not well constrained. Above around 160 cm depth (~ 6 ka), the ages of fluctuations in Cascade Lake begin to diverge from those of the regional records, reaching a maximum offset of around 2000 y at around 4 ka, and then decreasing towards the present.  1  60  2272  486  2  80  2753  302  3  155  4808  529  4  177  7276  138  5  189  --6  203  --7  228  9878  407  8  246  --9  284  11937  495  10  357  15455  705  11  419  17056  386  12  427  17414  551  13  454  18130  785  14 509 20518 1131 * Tie points shown in Figure 9. # One half of 1 sigma range. § Assumed age uncertainty of ± 500 years.  : Cascade Lake PSV-1 inclination age control points using the CALS10k.1b (Korte et al., 2011) and pfm9k.1b (Nilsson et al., 2014) field models and the Burial Lake inclination record (Dorfman, 2013). Tie-point ages are listed in Table 3.

Age model discrepancies
The disagreement between the PSV-1 and the radiometric age models may arise for multiple reasons. First, the three 14 C dates between 140 and 30 cm might be too old. Because the surrounding bedrock includes limestone and dolomite of the 360 Pennsylvanian-Mississippian Lisburne Group (Mull et al., 1994), a hard-water effect is likely operating within the lake, resulting in artificially old 14 C dates. Hard-water effects are known from many lakes; for example, they are responsible for age offsets of several hundred years in the Great Lakes (e.g., Rea and Colman, 1995;Moore et al., 1998). Due to a lack of sufficient terrestrial macrofossil material, all dated 14 C samples from Cascade Lake contained some component of aquatic organic material ( Table 2) that likely incorporated bicarbonate with no radiocarbon. Alternatively, 14 C dates may be too old due to 365 terrestrial "old carbon" being washed into the lake from organic matter stored within the watershed. Gaglioti et al. (2014) demonstrate that a significant amount of ancient organic carbon was incorporated into sediments of another lake (Lake of the Pleistocene) in the region during the late glacial and Holocene, especially during intervals when permafrost thaw thickened the active layer.
Some of the age offset might be explained by a post-depositional remanent magnetization (pDRM) lock-in effect, where the 370 acquisition of the magnetization occurs below the sediment water interface (Irving and Major, 1964). PSV records from organic-rich varved lake sediments in Sweden matched reference curves best when accounting for lock-in depths of 21-34 cm (Mellström et al., 2015;Snowball et al., 2013). Using differences between floating varve chronologies and 14 C wiggle-matched chronologies, Mellström et al. (2015) suggested that complete lock-in may not be achieved until 50-160 cm at Gyltigesjön and 30-80 cm at Kälksjön (west-central Sweden) in sediment with high organic-matter content and low bulk density. If a 375 similarly pronounced pDRM lock-in depth was a feature of the Cascade Lake sediment, then the radiometric age model could be closer to correct. Importantly, however, new cryptotephra analyses (Davies et al., this issue) from the same core analyzed for magnetic properties (CASC-4A) identified tephras whose ages support the PSV-1 age model as well as our conclusion that the pDRM lock-in depth is not significant in Cascade Lake (Fig. 10). Specifically, tephras in the upper 1.1 m of sediment have been geochemically 380 correlated with eruptions of Opala (~1400 cal BP), Mt. Churchill (~1700 cal BP), Ruppert tephra (~2450 cal BP) and Aniakchak (~3500 cal BP). These correlated tephra ages provide strong evidence for a 14 C age offset due to a hard-water effect or remobilization of old carbon in the watershed after around 6 ka, with this effect reaching a peak around 4 ka. Because the 14 C samples contain a mixture of aquatic and terrestrial materials (Table 2), which was necessary to obtain sufficient mass for conventional AMS 14 C techniques, it is not possible to investigate the extent to which the old carbon signal is carried by aquatic 385 versus the terrestrial components.

Comparison with regional records of geomagnetic field variability
Inclination, declination and RPI variability for the past ~ 21 kyr at Cascade Lake can be compared with other regional Alaskan, western North American, and Arctic records during the late Pleistocene and Holocene (Fig. 11). Previous published records include Burial Lake (Dorfman, 2013), core 85JC, Gulf of Alaska (Walczak et al., 2017) core 8JPC, Chukchi Sea (Lisé-390 Pronovost et al., 2009, Grandfather Lake, southwest Alaska (Geiss and Banerjee, 2003), and core 803, Beaufort Sea (Barletta et al., 2008). Comparisons with the lower part of the sequence (unit L1 and L2, > 16 ka) should be treated with caution as magnetic mineralogy with low S-ratios, implying a large percentage of hematite, while lower ARM and kARM/k suggest little magnetite and what is there is likely to be coarser grained (Fig 3 and 4), along with higher MAD values (Fig. 5) all imply a lithology less suited for geomagnetic reconstruction. in those intervals are not optimal for recording the geomagnetic field. 395 Similar inclination features can be identified in individual records, including core 85JC and Grandfather Lake (not incorporated in CALS10k.1b or pfm9k.1b) (Fig. 11a). Notably, the timing of low inclination at Cascade Lake around 2.8 ka is similar to https://doi.org/10.5194/gchron-2021-19 Preprint. Discussion started: 16 June 2021 c Author(s) 2021. CC BY 4.0 License. that at Grandfather Lake (~ 2.4 ka), core 803 (~ 2.5 ka), and core 8JPC (~ 2.9 ka). This pervasive feature may be related to a significant eastward swing in declination that is observed in northern Europe at ~ 2.8-2.5 ka, described as the "f-event" (e.g., Turner and Thompson, 1981;Snowball and Sandgren, 2002;2004). There is evidence that centennial-and millennial-scale 400 PSV and intensity features in North America and Europe are often out of phase during the Holocene , possibly resulting from an oscillating "eccentric dipole" field behavior (e.g., Gallet et al., 2009), with alternating periods when geomagnetic flux patches centered over either North America or Europe are dominant. Therefore, an "archeomagnetic jerk" such as the "f-event" (Gallet et al., 2009), identified as an eastward declination swing in North Atlantic PSV records, may be recorded as an inclination low in parts of North America because of growth of flux over European shifted toward an intensity 405 high Walczak et al., 2017). Though the timing of this North American inclination low is broadly contemporaneous with the European "f-event", a deeper understanding of spatial geomagnetic field behavior and better chronological control is required to confirm that these events are correlative.
Additionally, rapid, high-amplitude inclination and declination changes occurring around 18-17 ka in Cascade Lake are not associated with shifts in sediment type in unit L1, and they are replicated in the Burial Lake and 85JC records (Fig. 11a, b). 410 The occurrence of a possible geomagnetic excursion has been documented at ~ 17 ka builds upon previous knowledge of low inclination around ~ 18-17 ka in the western U.S. (Rieck et al., 1992;Liddicoat and Coe, 2011;Turrin et al., 2013).
Transformation of Cascade Lake component directions to virtual geomagnetic poles (VGPs) yields a minimum VGP latitude of 51.2° N, which is a 38.8° displacement from the geographic north pole. High-amplitude inclination shifts at this time are contemporaneous with low relatively paleointensity estimates (Fig. 11) and elevated MAD values (Fig. 5) that might be 415 expected during excursions, which usually exhibit low geomagnetic field intensity (e.g., Laj and Channell, 2007;Channell, 2014). More documentation of geomagnetic field behavior during this interval, especially at high latitudes, is necessary to positively correlate the ~ 17 ka excursion documented elsewhere with rapid inclination shifts in the Cascade Lake, Burial Lake, and 85JC records.

Relative paleointensity 420
The mechanism by which the past intensity of the geomagnetic field is preserved in sediments is not completely understood (Roberts et al., 2013). Therefore, to infer RPI, a series of quality assessment requirements must be met. A comprehensive review of RPI studies (Tauxe, 1993, and references therein) outlines some prerequisites for developing meaningful paleointensity estimates: (1) the dominant magnetic mineral should be magnetite in the grain-size range of 1-15 µm; (2) the ChRM of the magnetite should display a stable, well-defined vector component; (3) depositional remanent magnetization 425 (DRM) must accurately record the geomagnetic field, and there should be minimal discrepancy between inclination of the ChRM and that of the expected GAD for the site latitude; (4) down-core concentrations of magnetic minerals should not vary by more than an order of magnitude; (5) the ChRM should be normalized by multiple methods (e.g., kLF, IRM, ARM) with similar results; and (6) RPI should not correlate strongly with rock magnetic parameters of sediment (e.g., ARM, IRM, susceptibility, coercivity). Stoner and St-Onge (2007) add to these criteria, recommending that: (1) all samples should be 430 subjected to alternating field (AF) demagnetization to reveal component magnetization and high-and low-coercivity components; and (2) ChRM MAD values should be no more than 5° for RPI studies.
Unit L3 appears to meet most of the criteria for RPI studies, as evidenced by inclination relative to expected GAD inclination (Fig. 5), remarkably low MAD values (Fig. 5), and stable demagnetization behavior (Fig. 6). Magnetic grain-size estimation ( Fig. 3 and 4) suggests fine PSD magnetite, while SD magnetite can be a very efficient paleomagnetic recorder, even when it 435 is of biogenic origin (Roberts et al., 2011). https://doi.org/10.5194/gchron-2021-19 Preprint. Discussion started: 16 June 2021 c Author(s) 2021. CC BY 4.0 License. Figure 11: Cascade Lake paleomagnetic record compared with regional and global records. (A) Paleomagnetic inclination relative to geocentric axial dipole (GAD) predictions. (B) Paleomagnetic declination. (C) NRM/IRM compared with regional and global geomagnetic 440 intensity estimates and regional relative paleo intensity records. Included are Burial Lake, Brooks Range, Alaska (Dorfman, 2013), core 85JC, Gulf of Alaska (Walczak et al., 2017), Grandfather Lake, southwest Alaska (Geiss and Banerjee, 2003), core 803, Beaufort Sea (Barletta et al., 2008), and core 8JPC, Chukchi Sea (Lisé-Pronovost et al., 2009). Paleointensity estimates are provided for the CALS10k.1b (Korte et al., 2011) and pfm9k.1b (Nilsson et al., 2014)  We propose that, based on high R-values, minimal resemblance to the normalizer, and general agreement with various normalization methods, NRM20-70mT/IRM20-70mT is the best selection for an RPI proxy for the Cascade Lake record. Due to the rapid lithologic change and significantly lower normalized remanence estimates in L2, we suggest that RPI estimates for this interval are unreliable, as lithologic variability has possibly not been adequately removed.
RPI estimates for Cascade Lake (NRM20-70mT/IRM20-70mT) for the past ~10 kyr are similar to core 803 NRM/ARM and with 450 field strength estimates from CALS10k.1b and pfm9k.1b over the last few thousand years (Fig. 11c). Specifically, RPI peaks at ~ 4.8, 2.3, and 0.9 ka are persistent regional features, though CALS10k.1b and pfm9k.1b display an earlier mid-Holocene RPI peak at 4.4-4.2 ka. But the general shapes of the curves are quite different, especially prior to 4 ka, with higher values than the spherical harmonic models predict. This is partially supported by 8JPC that also shows generally higher intensity during this time interval. Estimates of Cascade Lake RPI earlier than 10 ka should be treated with caution until additional late 455 Pleistocene and Holocene paleointensity records become available from this region.

Conclusions
Paleomagnetic data from Cascade Lake sediment add to our knowledge of geomagnetic field variability in the Arctic back to 21 ka. RPI reconstructions over the past 15 ka, especially the NRM20-70mT/IRM20-70mT proxy, are promising and compare well with regional records. However, similarities with ARM concentrations suggest that the record should be treated with caution 460 because the implied higher-than-expected intensities during the mid-Holocene, if supported through replication, would have significant geomagnetic implications.
The paleomagnetic data also extend and improve the radiometric-based age model from Cascade Lake. Prominent features (tie points) in the Cascade Lake paleomagnetic inclination record can be wiggle-matched with similar features in two geomagnetic field models (Korte et al., 2011;Nilsson et al., 2014) and in the inclination record from Burial Lake (Dorfman, 2013). Tie 465 points can be correlated between Cascade and Burial Lakes to extend the age model at Cascade Lake below the lowest 14 C age at 15 ka, back to 21 ka. Tie-point ages diverge from the radiometric-based ages in the upper 1.6 m of the sediment, by up to about 2000 years at around 4 ka. The recent discovery of four identifiable late Holocene cryptotephra in this section of the core (Davis et al., this issue) supports the PSV chronology and suggests that hard water or aged organic material is a likely explanation for the age offset. In a companion study, Davies et al. (this issue) integrate the PSV age model with corroborating 470 evidence based on 14 C, 210 Pb, and cryptotephra, in a Bayesian framework to generate the best estimate for the age-depth relation. This firm geochronological footing increases the value of the Cascade Lake sedimentary sequence as an archive for future paleoenvironmental and paleoclimatologic studies.

Data availability
The down-core data for the multi-parameter properties analyzed on the sediment cores from Cascade Lake, 475 including all data used to plot the figures in this article, are in an Excel workbook file as one of the resources available with this online publication.
Author contribution DPS conducted the analyses, wrote the initial draft and curated the data. JSS validated the interpretations and provided training and laboratory resources. JPB co-led and co-funded the project. DSK conceptualized, co-led and co-480 funded the study. All co-authors provided critical reviews and revisions.
Competing interests The authors declare that they have no conflict of interest. data discussions.
Financial support This research was supported by the National Science Foundation grants #1107662 (DSK) and #1107854 490 (JPB), and awards from Pioneer Natural Resources and the Geological Society of America (DPS).