Introduction
The Pacific Northwest of North America hosts one of the world’s best geologic records of megathrust earthquakes. Still, there remain uncertainties about the frequency, magnitude, and characteristics of past ruptures on the Cascadia subduction margin. An ongoing question is whether prehistoric earthquakes inferred along the margin via correlation of deep-sea turbidites, subsided coastal-marsh and tsunami deposits, and lacustrine turbidites represent full-margin events or a series of closely spaced partial ruptures (Melgar, Reference Melgar2021; Staisch, Reference Staisch2024). As recently discussed by Staisch (Reference Staisch2024), this uncertainty stems largely from the limitations of radiocarbon dating. Challenges in dating past earthquakes from marine and terrestrial sedimentary archives include finding sufficient, well-placed organic remains, the need to account for the erosion of underlying sediment before event-layer emplacement, temporal uncertainties in the magnitude of the marine reservoir effect, accounting for in-built ages of organic material with pre-earthquake storage histories (e.g., in upland soils or lake-margin sediments) that are remobilized during the earthquake event, and the inherent uncertainty of radiocarbon dating itself. Although these challenges may be addressed by acquiring larger numbers of radiocarbon dates to bracket the ages of event deposits, dating delicate remains that might not survive long storage, improvements to reservoir corrections, and the application of Bayesian statistical techniques, none of these approaches can eliminate the limitations of radiocarbon-based age models, and independent estimates are highly desirable.
One approach to dating earthquake event deposits that does not rely exclusively on radiocarbon is to place them in the context of independently developed paleoclimate reconstructions. Tree-ring chronologies, for example, have been used to develop millennial-scale records of seasonal precipitation in western North America (e.g., Stahle et al., Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020). Similarly, lacustrine sediments have been used to reconstruct hydroclimatic effects on lake level and water-column chemistry (e.g., Steinman et al., Reference Steinman, Pompeani, Abbott, Ortiz, Stansell, Finkenbinder, Mihindukulasooriya and Hillman2016; Parish et al., Reference Parish, Wolf, Higuera and Shuman2022). The occurrence of past earthquakes relative to wet or dry periods identified in these reconstructions offers the potential to refine estimates of their timing. Here, we examine the sedimentary record of earthquakes and paleo-precipitation from Ozette Lake, a glacially formed lake located on the northwestern corner of the Olympic Peninsula in western Washington (Fig. 1).
Figure 1. Topographic and bathymetric map of Ozette Lake on the Olympic Peninsula of western Washington, USA, showing sediment coring locations (1–13). Bathymetry is expressed as shaded color and 20-m contours (after Dartnell et al., Reference Dartnell, Brothers, Ritchie, Sherrod, Currie, Dal Ferro and Powers2024). The inset map shows the location of the Ozette Lake (OL) relative to the Cascadia subduction zone and the location of Castor Lake (CL) used in the paleo-precipitation record recovered by Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012).
Ozette Lake depositional system
Ozette Lake is an oligotrophic to mesotrophic lake situated about 2 km inland from the Pacific Ocean (Fig. 1) and has a mean water surface elevation of 8 m above mean sea level. It has a surface area of 29.5 km2 and a catchment area of about 200 km2 that is underlain by Cenozoic sedimentary and volcanic rocks and Quaternary glacial sediments. The lake comprises northern, eastern, and western subbasins, and has an average and a maximum depth of 40 and 98 m, respectively.
The Cascadia trench, where the Juan de Fuca Plate subducts obliquely beneath the North American Plate at a rate of about 40 mm/yr (DeMets and Dixon, Reference DeMets and Dixon1999), lies 135 km west of Ozette Lake. The region has experienced great subduction earthquakes, the most recent being on January 26, 1700, CE (Atwater, Reference Atwater1987; Satake et al., Reference Satake, Shimazaki, Tsuji and Ueda1996; Yamaguchi et al., Reference Yamaguchi, Atwater, Bunker, Benson and Reid1997; Atwater et al., Reference Atwater, Musumi-Rokkaku, Satake, Tsuji, Ueda and Yamaguchi2016). Based on studies of subsided marshes, tsunami deposits, and deep-sea and lacustrine turbidites, subduction earthquakes are estimated to affect the region at intervals of 300–500 years (e.g., Atwater and Hemphill-Haley, Reference Atwater and Hemphill-Haley1997; Kelsey et al., Reference Kelsey, Nelson, Hemphill-Haley and Witter2005; Nelson et al., Reference Nelson, Kelsey and Witter2006; Goldfinger et al., Reference Goldfinger, Nelson, Morey, Johnson, Patton, Karabanov and Gutiérrez-Pastor2012, Reference Garrison-Laney, Miller, Haugerud and Kelsey2017; Brothers et al., Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024). Ozette Lake sediments record episodic disturbance events. Brothers et al. (Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024) traced 30–34 Holocene turbidites throughout the lake and presented evidence that they were emplaced during subduction earthquakes. Here, our focus is on the youngest four of these event layers.
Sediments in Ozette Lake also have the potential to record variations in precipitation on annual and longer timescales. Several tributaries enter the lake along its northern and eastern margins, including Umbrella Creek, Big River, Crooked Creek, and Siwash Creek (Haggerty et al., Reference Haggerty, Ritchie, Shellberg, Crewson and Jalonen2009; Brothers et al., Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024; Fig. 1). The Ozette River, the single outlet of the lake, flows 8.5 km to the Pacific Ocean from its northern end. Sediment supply to the lake has produced a thick post-glacial deposit, accumulating at a rate of 0.65–1.1 mm/yr since 13.5–14.0 ka B.P. (Brothers et al., Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024). Anthropogenic activities have increased these rates in the past century, particularly close to the larger tributaries. Commercial timber harvesting in the catchment began in the 1930s and was especially intensive in the 1950s–1980s, leading to pronounced delta progradation at the mouth of Umbrella Creek (Haggerty et al., Reference Haggerty, Ritchie, Shellberg, Crewson and Jalonen2009; Ritchie and Bourgeois, Reference Ritchie and Bourgeois2009). Lorenson et al. (Reference Lorenson, Brothers, Singleton, Derosier, Padgett, Sherrod, Kluesner and Swarzenski2024) used 210 Pb and 137Cs measurements from sediment cores collected from Ozette Lake in 2019 and 2021 to document linear sediment accumulation rates for the past ca. 100 years and since 1963 (the peak of above-ground testing of nuclear weapons in the northern hemisphere) of around 2.5 mm/yr at the lake’s northern end near Umbrella Creek, 1.7 mm/yr in the eastern subbasin near Tivoli island, and 1.0 mm/yr near the lake’s southwestern end.
Streamflow and turbidity on the northwest Olympic Peninsula are highly seasonal. The region, which experiences a temperate, oceanic climate (Köppen Zone Cfb) strongly influenced by Pacific Ocean atmosphere dynamics, is characterized by mild temperatures year-round and abundant precipitation. Warmer and dry conditions generally prevail between April and September when the strong North Pacific high-pressure system is positioned over the eastern Pacific. Clockwise flow around this system results in winds from the west and northwest, low humidity, and clear skies (Gavin and Brubaker, Reference Gavin and Brubaker2015). Cooler, wetter conditions occur from October to March in response to the intensification and southward movement of the Aleutian low-pressure system. Counterclockwise flow around the system steers moisture and winds inland from the south and southwest (Gavin and Brubaker, Reference Gavin and Brubaker2015). The bulk of precipitation, which averages around 260 cm/yr at Quillayute Airport, 12 km from Ozette Lake, falls during these months (Western Regional Climate Center, https.//wrcc.dri.edu/). The region is also affected by atmospheric rivers, long narrow pathways of water vapor transport that episodically cause extreme storms and floods, most commonly during autumn months in western Washington (Dettinger et al., Reference Dettinger, Ralph and Rutz2018).
On an interannual to multi-decadal basis, the El Niño-Southern Oscillation (ENSO) and related Pacific Decadal Oscillation (PDO) influence precipitation amounts in the Pacific Northwest by modifying the strengths and positions of the North Pacific High and Aleutian Low (Mantua and Hare, Reference Mantua and Hare2002; Verdon and Franks, Reference Verdon and Franks2006; Wise Reference Wise2010; Steinman et al., Reference Steinman, Abbott, Mann, Ortiz, Feng, Pompeani, Stansell, Anderson, Finney and Bird2014). During positive ENSO (El Niño) and PDO phases, the Aleutian Low strengthens and moves eastward, causing storm tracks to shift southward and leading to generally drier-than-average conditions in the region (McAfee and Wise, Reference McAfee and Wise2016). Negative ENSO (La Niña) and PDO phases are associated with the opposite conditions. Both ENSO- and PDO-driven variability in cool-season precipitation has been recognized in instrumental and longer-term proxy data from the region (d’Arrigo et al., Reference d’Arrigo, Villalba and Wiles2001; MacDonald and Case, Reference MacDonald and Case2005; Nelson et al., Reference Nelson, Abbott, Steinman, Polissar, Stansell, Ortiz, Rosenmeier, Finney and Riedel2011; Steinman et al., Reference Steinman, Abbott, Mann, Ortiz, Feng, Pompeani, Stansell, Anderson, Finney and Bird2014; Stahle et al., Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020; Shea et al., Reference Shea, Steinman, Brown and Schreiner2022). This variability is likely recorded in the sediments of Ozette Lake, where precipitation controls stream inflow and lake level (Haggerty et al., Reference Haggerty, Ritchie, Shellberg, Crewson and Jalonen2009; Slajus, Reference Slajus2020).
Methods
Core collection and processing
To explore the sedimentary record of earthquakes and hydroclimate, 7-cm-diameter gravity cores ranging 58–103 cm in length were recovered from 14 sites in Ozette Lake in June 2019 in partnership with the Washington Geological Survey (Fig. 1; Supplementary Table 1). All core sites, except for one, were in the deepest parts of the east and west subbasins, at water depths ranging from about 50–98 m. Duplicate cores were recovered at four sites (6, 7, 8, and 11; Fig. 1). Immediately following collection, the core surfaces were stabilized with Zorbitrol, a sodium polyacrylate gel. The cores were then sealed and shipped to the Continental Scientific Drilling Facility at the University of Minnesota, where they were scanned with a Geotek Standard Multi-Sensor Core Logger (Geotek Limited, Daventry, U.K.) to measure bulk sediment density, acoustic wave velocity, electrical resistivity, and loop-sensor magnetic susceptibility at 5-mm resolution. One of the cores (11B) was retained whole and shipped to the Institut National de la Recherche Scientifique, Centre – Eau Terre Environnement in Québec City, Canada, for analysis using a Siemens SOMATOM Definition A S+ 128 computerized tomography (CT) scanner at 0.25-mm increments. This core was chosen for intensive study because it was recovered from the lake’s western subbasin, which is separated from the lake’s major fluvial sources by an inter-basin ridge (Fig. 1). Sedimentation rates were expected to be moderate at this site and several earthquake-triggered layers were anticipated to be preserved in the meter below the lake bottom. The remaining cores were split lengthwise, and one half was photographed using a Geotek Geoscan-IV digital line scan camera mounted on a dedicated multi-sensor core logger–camera image scanner track. The core halves were then scanned at 0.5-cm increments on a Geotek XYZ Multi-Sensor Core Logger for magnetic susceptibility and visible-spectrum reflectance (VIS-RS) in the wavelength range of 360–740 nm using a Konica Minolta Spectrophotometer CM-2600d. After scanning, the cores were sampled at regular intervals and from layers of interest for radiocarbon and particle size analyses.
Organic carbon
Twelve samples from light- and dark-colored layers in the cores were analyzed for total organic carbon, carbon-to-nitrogen ratios, and carbon isotopic composition (ẟ13C). In preparation for these analyses, 0.5–1.0 g of sediment was dried and exposed to HCl vapor for 24 hours to remove carbonates. Following drying, sediment samples were measured for %TOC (total organic carbon) and %TN (total nitrogen) concentrations on a Thermo 1112 Flash elemental analyzer coupled in continuous flow to a Thermo Delta V+ isotope ratio mass spectrometer (IRMS). δ13C values (‰) were calculated relative to the VPDB (Vienna Peedee Belemnite) scale using International Atomic Energy Agency (IAEA) standards IAEA-C6, IAEA-C8, and IAEA-600 of known isotopic values as within-run standards. Replicate analysis of standard material gave a precision of ± 0.3‰.
VIS-RS data were also used to assess organic carbon content and composition in the core sediments. Following Rein and Sirocko (Reference Rein and Sirocko2002) and Meyer et al. (Reference Meyer, Van Daele, Fiers, Verleyen, De Batist and Verschuren2018), the relative magnitude of a trough between 660 and 670 nm in the reflectance curves measured at 0.5-cm increments down the cores was used to calculate the RABD660;670 parameter. RABD660;670 is a reliable proxy for the relative concentration of chlorophyll a (Chl a) and its derivative compounds in marine and lacustrine sediments (Zander et al., Reference Zander, Wienhues and Grosjean2022).
Radiocarbon analysis and age-depth modeling
Samples for radiocarbon analysis were selected from above and below turbidite layers and included fir needles, leaf fragments, and root fibers (Supplementary Table 2). The samples were rinsed with DI water and purified with 1N HCL and NaOH (the Acid-Base-Acid method; Olsson, Reference Olsson and Berglund1986) to remove contaminants they may have acquired during storage in soils or lake sediments. After purification and drying, the samples were submitted to the National Oceanographic Accelerator Mass Spectrometry Laboratory (NOSAMS) at the Woods Hole Oceanographic Institution. Results were calibrated using the IntCal20 curve (Reimer et al., Reference Reimer, Austin, Bard, Bayliss, Blackwell, Ramsey and Butzin2020).
An age-depth model was constructed using a Bayesian approach implemented through the R software package Bacon (Blaauw and Christen, Reference Blaauw and Christen2011). As discussed below, the Ozette Lake cores can be correlated based on event layers (turbidites) and a distinctive pattern of alternating lighter and darker layers in the intervening sediments (Fig. 2). Because all four turbidites were not sampled in every core, and because they were not underlain and overlain by sufficient, well-placed, datable material in a single core, we carefully correlated laminae between cores to project seven radiocarbon samples collected from multiple cores into stratigraphic positions in core 11B with a precision of a few mm. For the age model, the depths for these samples were corrected for the presence of the event layers, which represent instantaneous deposition. Two other dates were added to the “event-free” age model, including one for the core surface (collected in 2019 CE). An age of 250 cal yr BP was assumed for the uppermost event layer in the cores (EL1) based on correlation to the well-documented 1700 CE Cascadia subduction earthquake.
Figure 2. Cores collected from the eastern and western subbasins of Ozette Lake. Four event layers (EL1–EL4) are correlated between the cores. Note the consistent pattern of light and dark bedding in all the cores. The color of core photos was enhanced using Adobe Photoshop histogram equalization. Centimeter increments are indicated to the left of each core photograph.
Dynamic time warping
Dynamic time warping (DTW), an algorithm used to objectively align time series (Berndt and Clifford, Reference Berndt and Clifford1994; Ajayi et al., Reference Ajayi, Kump, Ridgwell, Kirtland Turner, Hay and Bralower2020), was applied to assess the correlation between sediment patterns in the cores and regional climate records. DTW accomplishes alignments by non-linearly stretching or compressing the depth (or time) axis of a “candidate” record (e.g., CT Intensity core 11B) to find the strongest statistical correlation to a “target” record (e.g., precipitation data and reconstructions). Our analysis used a MATLAB implementation of DTW developed by Hay et al. (Reference Hay, Creveling, Hagen, Maloof and Huybers2019), which selects the optimal warping path by minimizing the sum of the squared differences between the two records subject to an edge penalty that reflects the presumed extent to which the beginning and end times of the two records are the same and a diagonal penalty that controls the how much the candidate record can be warped. In our study, the code was modified to enforce a strong overlap between the youngest part of the records and the top of the core (edge=2) while assigning no penalty (edge=1) at the base of the core. The diagonal penalty (g) was tuned manually, being set just large enough to prevent the insertion of major time gaps and the collapse of major sedimentary cycles in the warped core.
To test the hypothesis that the light/high-density and dark/low-density layers reflect decadal-scale variations in precipitation and runoff, we compared the high-resolution, event-free CT intensity record from core 11B to the region’s instrumental and proxy records of cool-season precipitation. Before the application of DTW, the CT intensity data from core 11B were pre-processed to remove distinct event layers thicker than 1 cm and thought to be deposited at near-instantaneous time scales. Because the candidate (core) and target (climate) records represent different proxies, both records were linearly detrended and normalized by their standard deviations.
DTW optimizes the local correlation between the candidate and target time series. For two periodic time series the candidate record could align similarly well with more than one section of the target record. Because constraints are placed on the amount of warping, the result may therefore be sensitive to the starting age model. To investigate this sensitivity, 5000 age-depth models were simulated based on the allowable limits of the Bacon age model and then applied to the core 11B CT intensity. These simulated age-depth models were formed by seeding the ages of the mean Bacon model with Gaussian noise, where the minimum–maximum range in the Bacon age model was taken to represent plus or minus three standard deviations. These age-depth data were then fitted with a 5th-order polynomial to enforce a monotonic increase in age with depth. Applying DTW to the CT intensity with an ensemble of starting age-depth models yields a series of warped age-depth profiles that can be used to assess the uncertainty in the local warping path and the estimated age of event layers.
The instrumental record of hydroclimate for Clallam County, Washington, began around 1900 CE. Monthly precipitation data were accessed from the Western Climate Mapping Initiative website (Westmap; https://cefa.dri.edu/Westmap/Westmap_home.php; accessed May 14, 2024). The age model for the Ozette cores suggests that the instrumental period is represented by roughly the upper 15 cm in cores from the western (distal, relative to stream discharge to the lake) subbasin. Therefore, only this section of the 11B core was warped to the instrumental record.
Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) produced a 1500-year record of 20-year-averaged winter precipitation in the Pacific Northwest using oxygen isotope data from sediment cores recovered from Castor and Lime lakes, Washington. These two small lakes, located east of the Cascade Range in Okanogan and Pend Oreille counties, respectively (Fig. 1), are sensitive to climate variability due to their location in the Mediterranean-influenced, hot-summer humid continental (Dsa) Köppen climate regime. Cool-season rainfall is much less in this area than on the Olympic Peninsula. Still, it follows similar trends (Fig. 3). The Castor-Lime precipitation reconstruction relies on an age model based on 137Cs, tephrochronology, and radiocarbon dates.
Figure 3. Ten-year mean November–March precipitation for Clallam County, Washington, where Ozette Lake is located, and Okanogan County, Washington, where Castor Lake is located (Westmap; https://cefa.dri.edu/Westmap/Westmap_home.php).
The North American Seasonal Precipitation Atlas (Stahle et al., Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) is a web application that provides access to cool- and warm-season reconstructions of total precipitation based on tree ring chronologies. The reconstructions span the years 0–2016 CE and are available on a 0.5° × 0.5° latitude/longitude grid centered over North America. We downloaded the cool-season reconstruction for the Olympic Peninsula (47–48.5°N, 125.3–123.5°W). We fit a 10-year cubic spline to the time series data so that the resolution was comparable to that of the Ozette Lake sediment core dataset and the Castor-Lime lakes reconstruction.
Results
Core stratigraphy
The cores were examined for stratigraphic completeness. Some, but not all, of the cores recovered included an intact sediment–water interface, marked by a loosely packed, red-brown oxidized sediment layer, approximately 0.5 cm thick (e.g., cores 4A, 9A, 13A, Fig. 2). In the remaining cores, the absence of this oxidized layer indicates minor disturbance of the surficial sediments during core recovery, possibly by the bow wave created as the weighted gravity coring device was lowered (Dück et al., Reference Dück, Lorke, Jokiel and Gierse2019). Careful comparison of the laminae in the core 11B CT scan to high-resolution photos of those in neighboring cores (9A, 11A, and 13A) indicates about 1 cm of sediment is missing at the core top. A light-colored lamina about 1 cm below the top of the CT scan is correlative to one at 2 cm depth in the 13A core photo and is overlain by darker-colored, lower-density sediments that were not captured in the CT data.
Below the surficial oxidized layer, core sediments consist of laminated to thinly bedded, clayey silt, rich in diatoms and interpreted to record every-day or background conditions in the lake, interstratified with four, generally coarser, normally graded event layers (EL1–EL4) interpreted as turbidites (Fig. 2). The turbidites have sharp bases typically overlain by medium- to fine-grained sand enriched in plant debris, including spruce, hemlock, and fir needles. The turbidites fine upward to silt and have thin, clay-rich caps. In some cases, two superposed, fining-upward sequences are discernable within single turbidites (EL1 in core 4A, EL2 and EL3 in core 5A; Fig. 2). The turbidites vary laterally across the lake in particle size and thickness, ranging from 0.5 to > 60 cm thick. EL1 and EL2 are thickest in the eastern subbasin of the lake, particularly in the deepest depocenter (coring sites 1A and 3A; 97.5 and 89 m water depth, respectively). In the western subbasin, EL1, EL3, and EL4 are similarly thicker in the deep part of the depocenter than at shallower depths (e.g., at sites 9A, 10A, and 12A at 78–79 m water depth). EL2 is very thin in most cores recovered from the western subbasin (it is 1.5–3 cm thick in cores 9A and 8A, respectively), and it is difficult to discern in neighboring cores. Careful examination of the basal contacts of turbidites with underlying background sediments suggests minor, centimeter-scale scouring during emplacement of some turbidites in the eastern subbasin (e.g., EL1 in cores 3A, 4A, and 5A; Fig. 2), but no detectable scouring beneath the finer and thinner layers in the western subbasin.
Background sediments in the cores display alternating centimeter-scale, lighter and darker layers that can be correlated from core to core (Figs 2 and 4). Although the stratification is too variable in thickness to be considered varved, within some of the light and dark layers, thinner, regularly spaced laminae of a few mm thickness are apparent and may record annual deposition. Geophysical core logs indicate that the lighter-colored layers have higher magnetic susceptibility and are generally denser than the darker layers (Fig. 4). The CT scan of core 11B demonstrates that the lighter layers have higher CT intensity, reflecting their higher density. The ratio of reflectance at 590 and 690 nm, a proxy for the relative abundance of clay minerals such as biotite, illite, and chlorite (Trachsel et al., Reference Trachsel, Grosjean, Schnyder, Kamenik and Rein2010), is highest in the lighter color layers (Fig. 4). The lighter-colored, denser layers have lower organic carbon content and RABD660;670 values than the darker layers (Figs 4 and 5). The more positive carbon isotopic values and higher C:N ratios of organic matter in the light-colored layers compared to the dark layers are consistent with an increased contribution of terrestrial plant organic carbon (OC) relative to algal material, but the difference is not statistically significant.
Figure 4. Data from adjacent cores 12A and 11B. Event layers (turbidites; EL1–EL4) are highlighted using colors as in Figure 2. The correlation of darker-colored background sediments is shown with gray bars. Darker-color layers generally have relatively lower values of greyscale color, magnetic susceptibility, density, CT intensity and R590/R690, and generally higher RABD660;670 values, indicating that they are depleted in iron-bearing mineral particles and clays and enriched in organic carbon relative to the lighter layers. The core photo has been color-enhanced with Adobe Photoshop histogram equalization; dark blebs on the CT scan are bubbles from core degassing.
Figure 5. Organic carbon content and composition in lighter- versus darker-colored layers collected from Ozette Lake cores 11, 12, and 13.
Age-depth model
The radiocarbon-based Bacon age-depth model (Fig. 6) indicates sediment accumulation rates of around 0.9 mm/yr since the deposition of the uppermost turbidite, EL1, with an assumed age of 250 cal yr BP (January 26, 1700, CE), and slower, time-averaged sediment accumulation rates between about 0.5 and 0.6 mm/yr for the previous thousand years, consistent with the results of Brothers et al. (Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024). The radiocarbon-based age-depth model implies deposition of EL2 between 522 and 616 cal yr BP, EL3 between 846 and 1086 cal yr BP, and EL4 between 1130 and 1316 cal yr BP (Table 1).
Figure 6. Bacon age-depth model for core 11B. Green symbols denote dates for the core surface and the uppermost event layer (EL1), which is assumed to have been deposited during the 1700 CE Cascadia subduction zone earthquake. Blue symbols show the probability range (2σ) of calibrated radiocarbon dates from plant material collected above and below event layers EL1–EL4.
Table 1. Ozette Lake turbidites and their permissible correlation to northern Cascadia Subduction Zone paleoseismic records (from north to south)
1 Tanigawa et al. (Reference Tanigawa, Sawai, Bobrowsky, Huntley, Goff, Shinozaki and Ito2022); 2Enkin et al. (Reference Enkin, Dallimore, Baker, Southon and Ivanochko2013); 3Blais-Stevens et al. (Reference Blais-Stevens, Rogers and Clague2011); 4Garrison-Laney and Miller (Reference Garrison-Laney, Miller, Haugerud and Kelsey2017); 5Goldfinger et al. (Reference Goldfinger, Nelson, Morey, Johnson, Patton, Karabanov and Gutiérrez-Pastor2012); 6Atwater and Hemphill-Haley (Reference Atwater and Hemphill-Haley1997); Atwater and Griggs (Reference Atwater and Griggs2012); 7Witter (Reference Witter2008)
* When DTW warping allows for more than one possible age, the age most consistent with using the mean Bacon age-depth model for the candidate core is listed first (see Figs 13 and 14).
** Mean age not reported
*** EL2 is absent in core 11B but present in neighboring cores
Comparisons with climate records and proxies
Dynamic time warping indicates a strong cross-correlation (0.90) between core 11B CT intensity and November–March precipitation since 1900 CE (Figs 7 and 8), supporting the hypothesis that precipitation is a primary control on sediment density, color variations, and other properties of background sediment in Ozette Lake. The warped CT record indicates that sediment deposited in 1963 lies at 6–7 cm depth in 11B, at the sediment depth indicated by the 137Cs data of Lorenson et al. (Reference Lorenson, Brothers, Singleton, Derosier, Padgett, Sherrod, Kluesner and Swarzenski2024) for a core recovered at a nearby site, providing additional confidence that CT intensity accurately tracks precipitation history. Similarly, the base of the dynamically warped interval corresponds with a prominent light-colored interval that Lorenson et al. (Reference Lorenson, Brothers, Singleton, Derosier, Padgett, Sherrod, Kluesner and Swarzenski2024) interpreted to have accumulated around the turn of the twentieth century based on their 210Pb data.
Figure 7. CT intensity measured from core 11B plotted against the 10-yr mean November–March precipitation in Clallam County, Washington, since 1900 CE. Positive normalized values represent relatively more precipitation and negative values relatively less. Upper panel shows the CT record scaled to the mean radiocarbon-based Bacon age-depth model. Lower panel shows the CT “candidate” record after it is dynamically time warped to the “target” precipitation record.
Figure 8. Comparison of the age-depth relationship for the upper part of core 11B after dynamic time warping (red line) against 10-year average November-March precipitation in Clallam County, Washington to the ages indicated by the Bacon model. Dynamic time warping was applied using the mean Bacon age-depth model (green line) as the starting model for core 11B. The warped ages lie well within the 95% certainty of the Bacon radiocarbon-based model. The lower panel indicates that the shift of warped ages from the mean Bacon values is a maximum of 14.5 years.
When dynamic time warping is applied using longer-term climate reconstructions as the target time series, the cross-correlations are lower, 0.49 using the Castor-Lime precipitation reconstruction (Steinman et al., Reference Steinman, Abbott, Mann, Stansell and Finney2012; Fig. 9) and 0.53 using the tree-ring-based precipitation reconstruction (Stahle et al., Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020; Fig. 10). Still, even with the modest warping path recovered, alignment of the dominant features in both records is observed (Figs 9 and 10). The age-depth models produced by dynamically time warping the CT intensity (with the mean Bacon age-depth model applied initially) to the instrumental data and both reconstructions lie within or just on the edge of the 95% confidence intervals of the Bacon model (Figs 11 and 12).
Figure 9. Event-free computed tomography (CT) intensity measured from core 11B plotted against the 20-year average cool season, isotopically based precipitation reconstruction of Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012), where positive normalized values represent relatively more precipitation and negative values relatively less. Upper panel shows the CT record scaled to the mean radiocarbon-based Bacon age-depth model. Lower panel shows the CT “candidate” record after it is dynamically time warped to the “target” paleo-precipitation reconstruction. The positions of the four event layers (EL1–EL4) in the core are shown relative to both the un-warped and warped CT data.
Figure 10. Event-free computed tomography (CT) intensity measured from core 11B plotted against the 10-year average cool season, tree-ring-based precipitation reconstruction of Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020), where positive normalized values represent relatively more precipitation and negative values relatively less. Upper panel shows the CT record scaled to the mean radiocarbon-based Bacon age-depth model. Lower panel shows the CT “candidate” record after it is dynamically time warped to the “target” paleo-precipitation reconstruction. The positions of the four event layers (EL1–EL4) in the core are shown relative to both the un-warped and warped CT data.
Figure 11. Comparison of the age-depth relationship for core 11B after dynamic time warping against the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) cool-season precipitation reconstruction to the ages indicated by the Bacon model. Dynamic time warping was applied using the mean Bacon age-depth model (green line) as the starting model for core 11B. The warped ages lie well within or just outside of the 95% certainty of the Bacon radiocarbon-based model. The lower panel indicates that the shift of warped ages from the mean Bacon values is a maximum of 70 years.
Figure 12. Comparison of the age-depth relationship for core 11B after dynamic time warping against the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) cool-season precipitation reconstruction to the ages indicated by the Bacon model. Dynamic time warping was applied using the mean Bacon age-depth model (green line) as the starting model for core 11B. The warped ages lie well within or just outside of the 95% certainty of the Bacon radiocarbon-based model. The lower panel indicates that the shift of warped ages from the mean Bacon values is a maximum of 51.5 years.
When DTW is applied to a suite of allowable age-depth models for the CT intensity record, the uncertainty in warped event-layer ages can be understood and, in some instances, multiple local warping paths (or event ages) are found to be acceptable (Table 1). Using the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) and Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) records as targets, the age of EL1 is tightly constrained to 264 and 241 cal yr BP, respectively (Figs 13 and 14). EL2 can be warped to a position of either 502 or 590 cal yr BP within the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) record, the former being consistent with the result when the mean Bacon age model is applied to the candidate core. When the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) reconstruction is used as the target instead, all warping paths are consistent with an age of 542 cal yr BP for EL2 (Figs 13 and 14). EL3 is warped to an age of 1044 cal yr BP within the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) record. For the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) target, there is a group of paths consistent with an age of 943 cal yr BP, but the suite of DTW models also allows for a slightly younger age of 861 cal yr BP. Event ages for EL4 are the least well constrained (Table 1), with two allowable ages suggested when warping against either of the longer-term climate records. For the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) target, acceptable ages are found near 1290 and 1312 cal yr BP, and for the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) record, warping paths suggest an age of 1188 or 1243 cal yr BP. All DTW determined age estimates for the event layers are within the minimum to maximum range of the Bacon age model (Table 1).
Figure 13. Event layer (EL) age estimates based on a series of 5000 starting age-depth models for the core 11B computed tomography (CT) intensity warped against the target record of Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012). Age estimates for (a) EL1, (b) EL2, (c) EL3, and (d) EL4 in cal yr BP. All histograms are generated with a bin size of 0.5 yr, with the y-axis in counts. Red line indicates the median value for each cluster of allowable warped age estimates. Uncertainties in Table 1 represent the 95% range for each cluster of ages.
Figure 14. Event layer (EL) age estimates based on a series of 5000 starting age-depth models for the core 11B computed tomography (CT) intensity warped against the target record of Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020). Age estimates for (a) EL1, (b) EL2, (c) EL3, and (d) EL4 in cal yr BP. All histograms are generated with a bin size of 0.5 yr, with the y-axis in counts. The red line indicates the median value for each cluster of allowable warped age estimates. Uncertainties in Table 1 represent the 95% range for each cluster of ages.
Discussion
Interpretation of turbidites
The four normally graded event layers in the gravity cores share characteristics with earthquake-triggered turbidites identified from other lakes in seismically active regions. The turbidites fine rapidly upward from coarse basal sands, suggesting deposition from short-lived, energetic flows that evolved rapidly from earthquake-triggered slope failures (Vandekerkhove, Reference Vandekerkhove, Van Daele, Praet, Cnudde, Haeussler and De Batist2020; Brothers et al., Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024). Brothers et al. (Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024) used seismic reflection data to show that turbidites in the eastern subbasin of Ozette Lake, proximal to its significant fluvial sources, thin and fine away from deltas and other subaqueous slopes where mass-wasting deposits are present. In the lake’s western distal subbasin, where the present study is focused and where there are no fluvial sources, thinning of the turbidites away from the steep slopes provides especially important evidence of the role of slope failures in turbidite formation (Brothers et al., Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024). The superposed, fining-upward sequences topped by a single clay-rich cap in some turbidites further suggest derivation from multiple, coeval subaqueous slope failures associated with earthquake shaking (Schnellmann et al., Reference Schnellmann, Anselmetti, Giardini, McKenzie and Ward2002; Leithold et al., Reference Leithold, Wegmann, Bohnenstiehl, Joyner and Pollen2019). The age-depth model indicates ages for the turbidites that are consistent with their formation during four regionally identified subduction earthquakes (Table 1; Brothers et al., Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024). As discussed previously, the uppermost turbidite, EL1, is interpreted to have formed during the last, well-characterized, and dated CE 1700 event, which caused coseismic subsidence of the coastal margin of Washington and Oregon as recorded in preserved marsh soils (Atwater, Reference Atwater1987; Atwater et al., Reference Atwater, Musumi-Rokkaku, Satake, Tsuji, Ueda and Yamaguchi2016). No turbidites have been deposited in Ozette Lake in the subsequent 319-year record. This observation almost certainly rules out flood events as the triggering mechanism.
Background sediments: origin of light and dark layers
Based on average sediment accumulation rates of 0.5–0.9 mm/yr, the variations in the characteristics of background sediments in Ozette Lake are the product of decadal-scale environmental changes. The relatively high magnetic susceptibility and clay content of the light-colored layers and their relatively high density and low organic carbon content suggest a dominance of minerogenic particles, likely discharged to the lake during periods of high stream flow. Under such conditions, suspended fine-grained sediment would be introduced to the lake in surface- or mid-water plumes (hypo- and homopycnal flows) and then dispersed to the distal parts of the lake by wind-driven waves and currents (Wilhelm et al., Reference Wilhelm, Amann, Corella, Rapuc, Giguet-Covex, Merz and Støren2022). Discrete clay-rich laminae may record individual flooding events, while cm-scale light-colored beds are likely an amalgamation of many such events during multi-year periods when conditions were relatively wet.
The darker-colored layers are interpreted to have accumulated during drier periods when the relatively constant annual flux of biogenic debris (diatoms, plant detritus, etc.) to the lake bottom was less diluted with minerogenic particles. The darker layers likely owe their lower density to higher organic content and deposition as loose aggregates of biogenic and mineral particles (e.g., Droppo et al., Reference Droppo, Leppard, Flannigan, Liss, Evans, Wisniewski and Wisniewski1997; Hodder, Reference Hodder2009). The uniform composition of organic matter in the lighter and darker layers, as reflected by indistinguishable ẟ13C and C:N ratios, suggests its derivation from a similar mixture of terrestrial plant debris and lacustrine algae during periods of both higher and lower stream discharge into the lake. Based on these indicators of uniform composition, higher RABD660;670 values in the darker layers likely reflect total organic carbon content and are consistent with mineral dilution as a primary control on the differences between the light and dark layers.
Correlation of light and dark layers to historical and reconstructed hydroclimate
Using DTW to account for variations in rates of sedimentation accumulation and the uncertainty of the Bacon age-depth model, we first compared the CT data from core 11B to the instrumental precipitation record during the cool, wet season (November–March) in Clallam County, Washington. The age model for the Ozette cores suggests that the instrumental period, extending back to 1900 CE, is represented by roughly the upper 15 cm of distal cores 9A, 10A, 11A, 12A, and 13A (Fig. 2). Therefore, this portion of duplicate core 11B was selected for analysis. However, comparing these neighboring cores indicates that core 11B may be missing about 1 cm at its top, representing ca. 10 years of sediment accumulation (based on the estimated accumulation rates of about 0.9 mm/y).
The strong cross-correlation (0.90) between core 11B CT intensity and November–March precipitation (Fig. 7) supports the hypothesis that precipitation primarily controls sediment density, color variations, and other background sediment properties in Ozette Lake. Within the upper ∼15 cm of the Ozette Lake cores, two dark, low-density intervals record dry conditions between about 1920 and 1950, peaking in 1940, and between 1980 and the early 1990s. Three lighter-colored, higher-density intervals correspond with wet conditions in the early 1900s, from around 1950 to 1975 and from the early 1990s to the early 2000s. Near the base of the historical portion of cores 11B and 12A (Figs 2 and 4), three prominent mm-thick light-colored laminae may represent storms that produced three of the four highest average daily discharges measured in over a hundred years of record on the nearby Elwha River in 1897, 1900, and 1901 (https://nwis.waterdata.usgs.gov/nwis/peak?site_no=12045500&agency_cd=USGS&format=html, accessed May 14, 2024).
The explanation for relatively high sediment accumulation rates at the 11B site indicated by the warping path (near horizontal segments in Fig. 8) between about 1905–1910 and 1943–1947 is uncertain. The earlier episode could reflect the response of stream tributaries to the severe storms that occurred several years earlier. Routing of sediment mobilized on hillslopes and stream banks during those events may have led to relatively elevated sediment discharge to the lake in the succeeding years. The mid 1940s, in contrast, were very dry years. Higher linear accumulation rates at that time may reflect the loose packing of the low-density, organic-rich sediments deposited on the lake floor. Notably, an increase in sediment accumulation rates associated with intense timber harvesting and forest road construction in the Ozette Lake catchment between 1960 and 1981 is not apparent at the 11B site. However, it has been documented in the eastern parts of the lake proximal to the mouths of major streams (http://npshistory.com/publications/olym/lake-ozette-sediment.pdf, accessed May 14, 2024).
The CT record and the longer-term precipitation reconstructions track one another well with modest warping applied to the candidate core, particularly with the most recent half of the records (Fig. 12). The overall cross-correlations, however, are weaker than the instrumental record (Figs 9 and 10). The weaker correlation is unsurprising. While both the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) Castor-Lime lakes isotopically based and Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) tree-ring-based reconstructions reflect seasonally integrated precipitation, the Ozette sedimentary record very likely reflects individual storms that may occasionally have occurred during otherwise dry years. For the Castor-Lime Lake record, additional uncertainties may result from sparse age control, which for the past circa 1200 years is provided by 137Cs for the most recent part of the record, the Mount Saint Helens W tephra (471–468 cal yr BP; Yamaguchi, Reference Yamaguchi1985), and five radiocarbon dates (one from Castor Lake and four from Lime Lake; Steinman et al., Reference Steinman, Abbott, Mann, Stansell and Finney2012, Reference Steinman, Nelson, Abbott, Stansell, Finkenbinder and Finney2019) that were tuned to tree ring records (Nelson et al., Reference Nelson, Abbott, Steinman, Polissar, Stansell, Ortiz, Rosenmeier, Finney and Riedel2011; Lehmann et al., Reference Lehmann, Steinman, Finkenbinder and Abbott2021). Furthermore, Parrish et al. (Reference Parish, Wolf, Higuera and Shuman2022) called into question the interpretation of the Castor-Lime isotopic record as solely reflecting variations in cool-season precipitation and suggested that it might partially reflect variable evaporation over time.
On the other hand, age control for the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) dendrochronology-based reconstruction is excellent, but for the period older than 1400 CE (550 cal yr BP) it relies on a limited number of tree-ring chronologies, some up to 1000 km from the Olympic Peninsula. In addition, as explored by Wise (Reference Wise2021), tree-ring records in the Pacific Northwest may underestimate high-intensity precipitation events, including those associated with atmospheric rivers, because higher intensity storms may contribute more water to overland flow and less to the root zone soil moisture that is used by trees. Smith et al. (Reference Smith, Wegmann, Leithold and Bohnenstiehl2019) found that the Late Holocene sedimentary archive from Lake Quinault, the next large lake on the Olympic Peninsula south of Ozette, is dominated by riverine flood-event layers correlating to atmospheric rivers affecting the southwestern Olympic Peninsula. The sedimentary record of Ozette Lake likely also favors the preservation of atmospheric river-dominated flood-event layers. Moreover, the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) cool-season reconstruction for December through April may miss important portions of annual total precipitation occurring in autumn months, especially in November (Wise, Reference Wise2021). As such, the Ozette Lake sedimentary record may provide a detailed, long-term record of catchment-scale hydroclimate, which the existing regional proxy records do not fully reflect.
Both the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) Castor-Lime Lake isotopic and Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) dendro-based western Washington reconstructions provide information about important modes of ocean–atmospheric variability and their effects on precipitation in the Pacific Northwest, including the El Niño Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO) (Nelson et al., Reference Nelson, Abbott, Steinman, Polissar, Stansell, Ortiz, Rosenmeier, Finney and Riedel2011; Steinman et al., Reference Steinman, Abbott, Mann, Stansell and Finney2012, Reference Steinman, Abbott, Mann, Ortiz, Feng, Pompeani, Stansell, Anderson, Finney and Bird2014; Stahle et al., Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020). The results from Ozette Lake suggest that its sedimentary fill, which spans the entire Holocene (Brothers et al., Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024), can serve as an additional, very long, ocean-adjacent terrestrial record of these phenomena.
The Ozette Lake hydroclimate record may be especially important in evaluating the range of natural (pre-European settlement) limnological parameters such as tributary delivery of fine-grained sediments and lake surface stage, both of which are important factors in the viability of threatened Ozette Lake sockeye (Oncorhynchus nerka) that utilize the lake and its shoreline habitats for breeding and juvenile rearing (Good et al., Reference Good, Waples and Adams2005; Currens et al., Reference Currens, Fuerstenberg, Graeber, Rawson, Ruckelshaus, Sands and Scott2009; Haggerty et al., Reference Haggerty, Ritchie, Shellberg, Crewson and Jalonen2009). As suggested by previous studies (Steinman et al., Reference Steinman, Abbott, Mann, Ortiz, Feng, Pompeani, Stansell, Anderson, Finney and Bird2014; Shuman et al., Reference Shuman, Routson, McKay, Fritz, Kaufman, Kirby, Nolan, Pederson and St-Jacques2018; Mark et al., Reference Mark, Abbott, Steinman, Fernandez, Wise, Walsh and Whitlock2023), for example, particularly wet cool-season conditions in the Pacific Northwest during the Medieval Climate Anomaly (MCA, ca. 950–1250 CE) appear to have caused high stream discharge to the lake and are especially apparent when the CT record is warped to the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) precipitation reconstruction. In contrast, the dry cool-season conditions of the 1920s to 1950s seem to have been matched only in the years before the MCA.
The age of Ozette turbidites and comparison to other records of regional earthquake history
The results of DTW provide estimates of the age of the four turbidites in the Ozette Lake sediment cores based on the tuning of the CT record to the independently dated precipitation reconstructions of Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) and Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020; Table 1). The best-known earthquake along the Cascadia margin occurred in 1700 CE (250 cal yr BP). It was dated by comparison of tree rings in “ghost forests” that were killed by coseismic subsidence of marsh soils to nearby living “witness trees” (Yamaguchi et al., Reference Yamaguchi, Atwater, Bunker, Benson and Reid1997). Subsequently, the event was refined to precisely ca. 21:00 on January 26 (05:00 UTC, January 27) of that year based on written records of an orphan tsunami impacting Japan’s coastline (Satake et al., Reference Satake, Shimazaki, Tsuji and Ueda1996). Because this section of the radiocarbon calibration curve results in large uncertainties, most researchers have assumed that the most recent turbidites and tsunami layers at various sites represent the 250-cal-yr-BP event, as done here when constructing the Ozette Lake age-depth model. Brothers et al. (Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024) modeled the age of EL1 in Ozette Lake at between 292 and 523 cal yr BP (mean 397 cal yr BP) using OxCal Bayesian age modeling software (Bronk Ramsey, Reference Bronk Ramsey2009), suggesting possible in-built ages for their samples. Dynamic Time Warping of the 11B CT record to the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) reconstruction produces an age of 264 cal yr BP for the uppermost turbidite, while DTW using the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) precipitation reconstruction indicates an age of 241 cal yr BP. Both reconstructions indicate that the earthquake happened just after the peak of a very wet period.
Numerous attempts have been made to find accurate and precise dates for earlier earthquakes in Cascadia, primarily by identifying, carefully dating, and correlating tsunami deposits, subsided marsh soils, and offshore turbidites from multiple sites and applying Bayesian statistical techniques to combine them and derive probability distributions (e.g., Goldfinger et al., Reference Goldfinger, Nelson, Morey, Johnson, Patton, Karabanov and Gutiérrez-Pastor2012; Nelson et al., Reference Nelson, DuRoss, Witter, Kelsey, Engelhart, Mahan, Gray, Hawkes, Horton and Padgett2021). Recent studies indicate that the penultimate earthquake in northern Cascadia, first identified by Goldfinger et al. (Reference Goldfinger, Nelson, Morey, Johnson, Patton, Karabanov and Gutiérrez-Pastor2012) as having produced their second youngest deep sea turbidite (T2), may have been a partial rupture of the northern Cascadia margin (e.g., Goff et al., Reference Goff, Bobrowsky, Huntley, Sawai and Tanagawa2020, Tanigawa et al., Reference Tanigawa, Sawai, Bobrowsky, Huntley, Goff, Shinozaki and Ito2022). Dates for this event range from 384–573 cal yr BP based on the deep sea turbidite record (Goldfinger et al., Reference Goldfinger, Nelson, Morey, Johnson, Patton, Karabanov and Gutiérrez-Pastor2012) to 550–631 cal yr BP based on tsunami deposits preserved in Discovery Bay, Washington (Garrison-Laney and Miller, Reference Garrison-Laney, Miller, Haugerud and Kelsey2017), between 520 and 620 cal yr BP based on tsunami deposits from southwestern Vancouver Island (Tanigawa et al., Reference Tanigawa, Sawai, Bobrowsky, Huntley, Goff, Shinozaki and Ito2022), and between 565 and 668 cal BP for EL2 in Ozette Lake (Brothers et al., Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024).
The DTW results using the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) reconstruction and the mean Bacon age-depth model for core 11B yield an age estimate of 502 cal yr BP for EL2, consistent with the age determined by Goldfinger et al. (Reference Goldfinger, Nelson, Morey, Johnson, Patton, Karabanov and Gutiérrez-Pastor2012), but also allow for an age closer to 590 cal yr BP, more consistent with other estimates for the penultimate subduction earthquake in northern Cascadia. Tuning to the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) dataset yields an age estimate of 542 cal yr BP, slightly younger than the result of Brothers et al. (Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024). The DTW results for EL3 using the Steinman (2012) reconstruction suggests an age of 1044 cal yr BP, whereas using the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) reconstruction allows ages of either 943 or 861 cal yr BP. The younger age is more consistent with radiocarbon-based estimates of the next oldest earthquakes in some regional studies (the W event of Atwater and Hemphill-Haley, Reference Atwater and Hemphill-Haley1997; T3 of Goldfinger et al., Reference Goldfinger, Nelson, Morey, Johnson, Patton, Karabanov and Gutiérrez-Pastor2012); however, the older ages (943 and 1044 cal yr BP) are more consistent with those determined Brothers et al. (Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024) for EL3 in Ozette Lake (930–1055 cal yr BP). Finally, for EL4, DTW against the Steinman et al. (Reference Steinman, Abbott, Mann, Stansell and Finney2012) record indicated ages of 1290 or 1312 cal yr BP, or using the Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) data, ages of 1188 and 1190 cal yr BP. All but the oldest (1312 cal yr BP) of these ages overlap with at least one of the other regional estimates for the fourth oldest Cascadia subduction earthquake. This older age is, however, more consistent with the 1301–1405 cal yr BP age estimate for EL4 provided by Brothers et al. (Reference Brothers, Sherrod, Singleton, Padgett, Hill, Ritchie, Kluesner and Dartnell2024).
Conclusions
In this study, the centimeter-scale alternation of lighter-colored, higher-density layers with darker-colored, lower-density layers in Ozette Lake bottom sediments was hypothesized to record decadal scale wetter and drier periods in the lake’s Late Holocene history. To test this idea, we compared these patterns to historical instrumental records of cool-season precipitation in the region using the technique of dynamic time warping, which allows for modest variation of sediment accumulation rates over time. After a strong correlation was revealed, we compared sediments deeper in the lake’s stratigraphy to two independent cool-season precipitation reconstructions for the region, one derived from oxygen isotopes and one based on seasonal tree-ring records. Correlation of the sediment properties in the Ozette Lake sediments to these reconstructions suggests that its sedimentary archive likely holds clues to variations in seasonal precipitation across northwestern Washington over the past thousands of years, possibly since the retreat of the Cordilleran Ice Sheet, ca. 14 ka.
Importantly, correlation of the Ozette Lake sedimentary record to the independently dated cool-season precipitation reconstructions provides estimates of the timing of the past four Cascadia subduction earthquakes within the uncertainty of a radiocarbon-based age model. Although the reconstructions have some shortcomings, and the correlations permit several age estimates for each subduction zone earthquake event, they indicate dates that are more precise than possible using radiocarbon dating alone. The tree-ring-based reconstruction of Stahle et al. (Reference Stahle, Cook, Burnette, Torbenson, Howard, Griffin and Diaz2020) is an especially valuable chronometer because it is independent of radiocarbon dating. The approach here offers the possibility of more precise dating of earthquake events recorded in lakes along the Cascadia margin, addressing stubbornly intractable questions about their synchroneity and the behavior of megathrust earthquakes in the region.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/qua.2025.19.
Acknowledgments
This research was funded by a grant from the Faculty Research and Professional Development Program of NC State University (project 27581-42331). Research in Olympic National Park, Washington, USA, was facilitated by Dr. Steven Fradkin under US Department of the Interior Scientific Research and Collecting Permit OLYM-2019-SCI-0020. We are grateful to the talented personnel at the Continental Scientific Drilling Facility at the University of Minnesota, Minneapolis, Minnesota, USA, for their assistance with core processing and to Dr. Pierre Francus at Institut National de la Recherche Scientifique, Centre – Eau Terre Environnement in Québec City, Canada, for providing the CT scan. The hospitality of John and Mary Wegmann and Hannah Wiepke was invaluable for our study and is greatly appreciated. Reviews by Associate Editor Jaime Urrutia Fucugauchi, Byron Steinman, and an anonymous reviewer helped to strengthen the manuscript and are gratefully acknowledged.
Open access
