On 26 January at 21:00 PST (Pacific standard time), 1700 CE, a large earthquake occurred along the Cascadia Subduction Zone, the interface between the Pacific and North American plates along the coasts of California, Oregon, Washington, and British Columbia (Satake et al., 2003). The earthquake created a tsunami with 10–12 m run-up heights that struck the Pacific Northwest and propagated across the Pacific to Japan (Atwater, 1992; Satake et al., 2003; Goldfinger et al., 2003). It is estimated to have most likely been a moment magnitude (M_w) 9.0, with between 13 and 21 m of coseismic slip on an offshore fault 1100 km long (Satake et al., 2003; Witter et al., 2011). The 1700 earthquake was preceded by an earthquake in ∼960 CE (740-year interval) and another in ∼750 CE (210-year interval), with three additional subduction events before these that make up a recent cluster of six megathrust events over the past 1500 years (Atwater et al., 2003). During the 1700 Cascadia earthquake, ground motion and peak ground acceleration (PGA) are modeled from ∼0.5 to 1.2 g along the Oregon coast (WDNR, 2012). Thus the shaking during this event would have been violent and widespread.

As a result of subsidence, some coastal forests dropped below sea level and were flooded. Boles and root masses of these trees still remain and can be found from northern Oregon to southern Washington. Radiocarbon dating of this wood showed the earthquake occurred around 1700. However, aligning growth patterns of adjacent living trees with those of the flooded, dead trees consistently showed that the last year of growth was 1699, indicating the earthquake occurred between October 1699 and April 1700 (Yamaguchi et al., 1997). The exact origen time of the earthquake was estimated by calculating the travel time for an unexplained tsunami that struck Japan on 26 January 1700 (Satake et al., 2003; Atwater et al., 2006). Surviving trees also recorded the earthquake's date by anomalous changes in ring width or wood anatomy (Atwater and Yamaguchi, 1991; Jacoby et al., 1997). This tree-ring and dating evidence for coastal disturbance is indeed compelling; however, the evidence was derived from trees along just 100 km of coastal southern Washington and northern Oregon or ∼5 % of the coastline expected to be affected by a Cascadia megathrust earthquake.

Additional methodologies have been employed to assess the coastal-wide impacts of the 1700 earthquake. For example, a coastal-wide inventory of liquefaction features associated with the 1700 earthquake found no features along the Oregon coast despite numerous exposures of clean sand deposits that must be susceptible to liquefaction, even at low levels of seismic shaking (Obermeier and Dickenson, 2000). The locations for these field studies in Oregon were also sites where evidence for great Holocene subduction earthquakes (in the form of crustal subsidence) have been identified (Nelson et al., 1995). The only liquefaction features identified to date (and thus direct evidence of seismic shaking) were found along the Columbia River 35–50 km east of the coast, and these indicate moderate shaking intensity of 0.2–0.35 g (Obermeier and Dickenson, 2000).

As a first step in estimating tree disturbance in South Beach, we produced a model of the tsunami inundation level and expected flow speed for the 1700 earthquake based on estimates of size, location, displacement, and coastal subsidence (Fig. 2a–c; Witter et al., 2011). Thus, the modeled run-up height of the 1700 tsunami can be used as a basis to investigate possible impacts along the coast and estuaries of South Beach. Figure 2a and b show the model results of tsunami inundation level and flow speed for South Beach assuming the “L”- or large-sized earthquake (M_w=9.0) for the 1700 CE event (Wei, 2017). The L model assumes a finite-fault source with maximum vertical coseismic displacement of 15.2 m and subsidence of ∼1.03 m at South Beach (Fig. 2c; Witter et al., 2011). The Witter et al. (2011) coseismic subsidence estimate differs slightly from the Satake et al. (2003) estimate of nearly 1 m at South Beach because it is based on coseismic slip from turbidite records (Goldfinger, 2011) and includes a rupture model with slip partition into a splay fault in the accretionary wedge. The earthquake source duration was not taken into account in the model.

Figure 2(a, b) Model of the maximum tsunami inundation level and the maximum flow speed for South Beach assuming the L- or large-sized earthquake (M_w=9.0) for the 1700 CE event (Wei, 2017). The L model assumes a finite-area fault source with maximum coseismic displacement of 15.2 m and subsidence of ∼1.03 m at South Beach. Contour levels are shown. (c) The L model assumes a finite-area fault source with maximum coseismic displacement of 15.2 m and subsidence of ∼1.03 m at South Beach (Witter et al., 2011).

Two models were used to compute the tsunami inundation levels and flow speed (Wei, 2017) based on four-level one-way nested model grids at the spatial resolutions of 1 arcmin (∼1.8 km), 12 arcsec (∼360 m), 2 arcsec (∼60 m), and $\mathrm{1}/\mathrm{6}$ arcsec (∼5 m). The tsunami simulation model MOST (Method of Splitting Tsunami) (Titov and Gonzalez, 1997) used in this study is based on the shallow-water wave equations and uses the estimates of coseismic slip to account for deep-water wave generation and propagation. The present MOST code version utilizes the graphics processing unit (GPU) technology that has led to significant reduction of computational time. The MOST model then provides the boundary conditions computed from the level-1 grid (Fig. 2a) for a Boussinesq model (Zhou et al., 2011), which takes into account wave dispersion when computing the nearshore wave-propagation field and onshore tsunami inundation in level-2, level-3, and level-4 grids (Fig. 2b shows the coverage of level-4 grid). The digital elevation model (DEM) and bathymetric grid of Newport and South Beach in level 4 were used in the tsunami inundation models. The elevation grid is derived from the digital elevation model provided by the Oregon Department of Geology and Mineral Industries (DOGAMI). This dataset contains lidar data based on DOGAMI lidar data quadrangles for Toledo south, Newport north, and Newport south. The horizontal datum of the DEM is WGS 84. The vertical datum is NAVD 1988, and it is then converted to mean higher high water (MHHW) level, which is the vertical datum in our tsunami inundation models. MHHW is 2.317 m above the NAVD 1988 and 1.185 m above the actual mean sea level (MSL) at Newport according to the datum information at the National Ocean Service (NOS) tide gauge at South Beach. Typically, when performing hazard assessments, mean high water (MHW) or MHHW is assumed over the entire duration of a tsunami (Wei, 2017), and using MHHW as the vertical datum usually gives a more conservative estimate of the tsunami impact. In the present study, we prefer to use MHHW, instead of the actual tidal level, as our model reference level due to (1) the uncertainty of the time of the event, which is based on estimates from Japanese records (Satake et al., 1996) and could vary over a window of 1–2 h; (2) the uncertainty of the earthquake and tsunami source; and (3) the uncertainty in the amount of sea level change, which is >0.5 m over the past 300 years based on a rate of 1.77 mm annual increase. The impact of these uncertainties on the model could overshadow the difference between MHHW and the actual tidal level, and it adds an additional level of uncertainty to the model results. A Manning's coefficient of friction of 0.03 is uniformly applied for both the land and ocean components of the tsunami propagation model. It is an average Manning's coefficient that Chow (1959) proposed for coastal and riverine areas (0.025–0.033) and for land surface (0.03–0.04). It is worth nothing that this Manning's coefficient has been widely used in MOST-based tsunami model forecast methodology and hazard assessments (Tang et al., 2009; Wei et al., 2008, 2013; Titov et al., 2016; Zhou et al., 2011). To more realistically estimate the tsunami impact produced by the 1700 event, we removed the two jetties at the entrance of Yaquina Bay from the model DEMs, which leads to greater tsunami inundation levels and impact at South Beach. The tsunami model results discussed hereafter are based on the revised DEMs without the jetties.

Figure 3(a) Model of maximum tsunami inundation depth at South Beach for the L-sized earthquake 1700 CE event; (b) zoom-in view of the tsunami inundation depth at Mike Miller State Park. Gray dotted circles show location of trees used in this study on the north side of the stand. Red numbers are the tree locations whose growth chronologies are shown in Fig. 4a and b. White numbers shows trees that were cored, but chronologies do not include the years before 1700 CE. Colors on map show inundation depth from the model, implying 0–10 m of inundation depth at the Mike Miller Park Douglas-fir stand. Green areas are high ground locations that show no inundation. The yellow line shows, for comparison, the model of maximum run-up height for the M_w=9.2 “XXLarge” earthquake scenario (Priest et al., 2013). Base maps made using © Google Earth 2016.

The tsunami inundation model presented here indicates that the L earthquake, with the coseismic subsidence taken into account, would produce a tsunami that could inundate South Beach to run-up heights up to 17 m (Fig. 2a) and inundation depths up to 16 m (Fig. 3a and b). The height of the water level at the western section of Mike Miller Park is generally between 12 and 15 m and reduces to between 9 and 12 m on the eastern side. It is important to note that “tsunami water level” is a term used to describe the elevation reached by seawater measured relative to a stated datum (MHHW herein). In contrast, “inundation depth” refers to the local water depth or height of the tsunami above the ground after taking into account the coseismic subsidence at a specific location, as shown in Fig. 2c. However, there is significant variation in the topography of South Beach, and several areas are predicted to experience a range of inundation depths much less than the 16 m maximum. For example, the model shows the amount of inundation decreases eastward of the beach, and the location of the old-growth Douglas-fir stand at Mike Miller State Park in South Beach may be subjected to a range of inundation depth from negligible to as much as 10 m (Fig. 3b). Moreover, the South Beach stand would likely have been subjected to flow velocities between 2 and 10 m s⁻¹ (Fig. 2b). These velocities are lower than most of the westward portions of the South Beach Peninsula because the stand is located on topography that can have up to 10 m higher elevation than most of the westward terrain. Nevertheless, it would seem these tsunami current velocities would be high enough to cause significant damage to the South Beach trees through the large mass and momentum of this volume of seawater that would be observable in the tree growth.

Lastly, it is worth noting that the L earthquake tsunami model presented here also involves the activation of splay faults in the overriding plate above the subduction zone. Motion on these splay faults introduce a larger coseismic subsidence along the coastline and therefore represent a more extreme inundation scenario for the 1700 CE event than previous models. Based on the turbidite records reported by Goldfinger et al. (2011), the L and larger earthquake scenarios occurred four times in the past 10 000 years and thus are referred to as a 2500-year event, although the general earthquake size class and associated time interval for an L event is estimated to be 800 years by Witter et al. (2011).

In an attempt to further quantify the widespread effects of the 1700 CE earthquake, we obtained tree-ring records from a stand of old-growth Douglas-fir trees (Pseudotsuga menziesii) that pre-date 1700 (Fig. 4a and b). The stand is located in an Oregon State Park in South Beach, Oregon, roughly 600 m east of Highway 101 (Fig. 3a and b). Old growth trees of 300+ years of age are rare along the Oregon coast; thus this stand of trees within the inundation zone presented a unique opportunity to search for direct physical evidence of the impact of a Cascadia Subduction Zone earthquake and tsunami inundation in a populated area where tsunami models indicate significant inundation levels and run-up heights.

Figure 4(a) Tree-ring growth records of old-growth Douglas-fir trees (Pseudotsuga menziesii) located in Mike Miller State Park, South Beach, Oregon (see Fig. 1). The vertical axis shows ring growth (in cm), and the time range covers several decades before and after 1700 CE. The color of each growth record relates to alpha-numeric labels of individual trees shown in the legend, with the location of trees shown in Fig. 3b. The designation “A/B” represents two cores from the same tree. The black arrow marks the 1700 CE date, and red arrows highlight the 1691, 1738, and 1745 CE large growth reductions that may have been caused by significant climatological events. Panel (b) shows detailed growth record of trees in (a) 4 years before and after 1700.

Download

To ground truth the model of the 1700 CE tsunami, we collected tree cores at breast height from 37 dominant or codominant old-growth trees at the South Beach site using a 0.8 m increment borer. One to two cores were collected from each tree, after which cores were mounted, sanded with increasingly fine lapping film, and cross-dated (Phipps, 1985). Each core was then measured using a Velmex TA tree-ring measuring device to the nearest 0.001 mm (Velmex, Inc. Bloomfield, NY). Cross dating was then statistically verified using the program COFECHA, which is designed to find errors in chronologies (Holmes, 1983). A master growth-increment chronology was then developed by detrending each measurement time series using negative exponential or regression functions to retain as much low-frequency variability as possible, as well as a second chronology developed using a 50-year 50 % frequency-cutoff cubic spline to highlight interdecadal to interannual growth variability. All chronology construction was performed using the program ARSTAN, a tree-ring standardization program based on detrending and autoregressive time series modeling (Cook and Krusic, 2005). Figure 3a and b show the location of the Douglas-fir trees sampled for this study in relation to the modeled tsunami run-up heights for South Beach. Of all the trees sampled, a total of 12 cores from 8 trees pre-dated 1700 (Fig. 4b).

As noted, the tsunami inundation model presented here (Fig. 2a and b) indicates the L earthquake would produce a tsunami that could inundate the lowlands of South Beach to inundation depths up to 18 m. However, the Douglas-fir old-growth stand that is the subject of this study lies on and along the western edge of two parallel north–south striking topographic highs (likely paleo-dune ridge lines). The tsunami model presented here indicates that while many of the trees in this area may have experienced as much as ∼10 m of inundation depth, several trees are also on high ground and may have experienced much less, or even zero, inundation.

Tree-ring data detrended using negative exponential functions did not reveal major stand-wide releases or suppressions around 1700 (data not shown) nor did data detrended using the 50-year spline functions. Detailed examination of the growth-ring samples indicates that although individual cores have below-average growth, and one experiences what could be interpreted as a post-1700 growth release, variability around 1700 is not necessarily exceptional in the longer-term context of the ∼310-year history of the dataset (Fig. 4a and b). Indeed, there are several other growth reductions in the record that are the same magnitude or larger than the disturbance at 1700 CE. Most notably, there are large suppressions observed beginning around 1691 and again in 1739 and 1745 (arrows on Fig. 4a). The 1739 reduction has been observed in other old-growth stand chronologies throughout Cascadia and may be due to a significant climatological event such as a drought (Carroll and Jules, 2005; Carroll et al., 2014).

5.1 Control sites in western Oregon Cascades and Coast Range

To better detect unusual growth anomalies around 1700, we compared the South Beach Douglas-fir tree-ring data to two other Douglas-fir datasets from the Oregon Coast Range and one from the western Cascade Mountains, all of which would have experienced similar climate conditions but not tsunami inundation. The first of these was an old-growth stand on Marys Peak (∼46 km east of South Beach) in the central coast range. The second is Browder Creek located ∼160 km east in the western Cascade Mountains (Black et al., 2015; Fig. 5a and b). The third was a chronology generated from samples of dead trees in lakes in the western and central Oregon Coast Range. These lakes had formed when landslides impounded streams, and the preserved drowned trees were then used to establish the date of lake formation, and thus the landslide event (Struble et al., 2020). Eight lakes had a combined number of 15 trees (and 31 sets of measurements) that pre-dated 1700 and were used to generate a “control” chronology. As with Mike Miller trees, all measurement time series in the control datasets were detrended with 50-year splines. The Oregon Coast Range sites range in elevation from 137 m (Hamar Lake) to 380 m (Klickitat Lake), with the exception of Marys Peak, which is 900 m. The lone western Cascade site, Browder Creek, is 1108 m. All sites are at low enough elevation that they are most limited by summer (July–September) drought as opposed to higher-elevation sites that are most sensitive to temperature. Relationships with drought are somewhat stronger at Marys Peak and Browder Creek, but this is likely due to their more inland locations.

Figure 5(a) Difference in growth chronologies between Mike Miller and reference sites at Marys Peak, Browder Creek, and Oregon Coast Range lake sites. The 1700 CE chronology indicated by the red arrow, and the significant growth differences between coast and inland sites in 1739–1741 and 1745–1748 are highlighted by red and black lines, respectively. Panel (b) shows location map with Marys Peak and lake sites relative to South Beach. The number of cores available at each site during the 1700 time period are South Beach (14), Marys Peak (28), Oregon Coast Range lakes (31), and Browder Creek (30). The Oregon Coast Range lakes include Beaver Dam Lake, Carlton Lake, Esmond Lake, Hamar Lake, Kauppi Lake, Klickitat Lake, Soup Lake, and Wasson Lake. Map created using digital elevation data points complied by the National Center for Environmental Information (http://ncei.noaa.gov, last access: September 2020).

When compared to the control sites, South Beach tree growth is significantly lower than the lakes, Marys Peak, or Browder Creek in 1700 (t test of p<0.001) after detrending all series with the 50-year 50% frequency-cutoff cubic splines. Moreover, the lake trees and Marys Peak trees do not significantly differ in growth in 1700, suggesting the South Beach Douglas-fir growth is unusually low across sites with the most similar climatic histories and sensitivities.

Analysis of tree-ring data presented here indicates there is a reduction in Douglas-fir tree growth at a site associated with the 1700 Cascadia Subduction Zone earthquake and tsunami in South Beach, Oregon, relative to other inland sites in the Oregon Coast Range. The growth reduction is not outside the range of variability, as illustrated by other much more severe reductions over the 310-year South Beach chronology (Fig. 5a and b). However, there is at least a subtle growth reduction that deviates from other nearby locations. Although beyond the scope of this study, further chemical analysis is needed to show that the tree rings collected at South Beach exhibit higher δ¹³C or water-soluble ion levels to establish that the trees here were immersed in seawater.

The tsunami inundation model presented here, which assumes the L- or large-sized earthquake (M_w=9.0) for the 1700 event (Wei, 2017), shows the resulting tsunami would have inundated the South Beach Douglas-fir stand. The growth reduction in Douglas-fir at South Beach is less pronounced than those observed following the 2011 Japanese tsunami or the post-1700 growth suppressions observed by Jacoby et al. (1997) along the Columbia River ∼220 km to the north of our study site. While modest, spanning no more than 2 years relative to the control sites, and not absolute conclusive evidence, the growth reduction is at least consistent with such an event and with both the magnitude and multi-year time period of growth reductions observed by Jacoby et al. (1997) along the Columbia River ∼220 km to the north.

Speculatively, the observed growth reductions could possibly represent multiple large earthquakes (magnitude 8+) over that time period rather than one great magnitude 9 earthquake at 1700. Although we cannot rule out multiple earthquakes using our tree-ring growth data, the record of the orphan tsunami in Japan is evidence that only one large earthquake occurred. Moreover, the tsunami inundation model presented here would not necessarily need to change in the multiple earthquake scenario since zero inundation at the old growth stand in South Beach is one possible model result.

Although tree rings and various geologic lines of evidence have been useful in establishing the date of the last Cascadia earthquake, there is still some question as to the degree of peak ground motion associated with the event. As previously discussed, a coastal-wide inventory of liquefaction features associated with the 1700 CSZ event indicate only moderate levels of ground shaking in coastal Oregon and Washington (Obermeier and Dickenson, 2000). Multiple, smaller magnitude megathrust earthquakes before, during, and after 1700 would be one possible explanation for a lower than expected peak ground motion and fewer observed liquefaction features throughout the region. Mapping seismically induced landslides in the Oregon Coast Range is potentially another means to assess levels and distribution of seismic shaking impacts from the 1700 CSZ event, given large-magnitude earthquakes in mountainous regions around the world typically trigger thousands of landslides, and slope failures constitute a significant proportion of the damage associated with these events (Stuble et al., 2020). Recent studies have demonstrated the utility of dendrochronology to date the ages of landslides in these settings (Stuble et al., 2020). However, despite the Oregon Coast Range exhibiting thousands of landslides, none have been conclusively associated with the 1700 subduction earthquake, and despite proximity to the megathrust rupture, most deep-seated landslides in the Oregon Coast Range were triggered by rainfall (Perkins et al., 2018; LaHusen et al., 2020; Struble et al., 2020). Thus a continued search for physical evidence of tsunami inundation, earthquake shaking, and coseismic landslides is needed to refine expectations of the inundation, as well as intensity and distribution of ground shaking during future Cascadia megathrust earthquakes.

The MOST model was created by Vasily Titov (Titov and Synolakis, 1995, 1998). The MOST codes used in the present study have been re-written utilizing the graphics processing unit (GPU) technology to boost the computational speed. The Boussinesq model codes were created by Hongqiang Zhou (Zhou and Teng, 2010), previously a research scientist at the University of Washington and now a data scientist at TISTA Science and Technology Corporation.

The Mike Miller, Cape Perpetua, and Marys Peak tree-ring data were added to the NOAA National Centers for Environmental Information International Tree-Ring Databank, https://www.ncdc.noaa.gov/data-access/paleoclimatology-data/datasets/tree-ring (last access: January 2021) (NOAA, 2021).

RPD prepared the manuscript with contributions from all co-authors. BAB and RPD collected the tree-ring samples, performed growth disturbance analysis, and wrote the manuscript. YW developed the tsunami model code, performed the simulations, and wrote the tsunami section of the manuscript. SGM drafted initial maps used in the manuscript.

The authors declare that they have no conflict of interest.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This project was supported by the National Oceanic and Atmospheric Administration's Pacific Marine Environmental Laboratory.

This paper was edited by Maria Ana Baptista and reviewed by Jean Roger and two anonymous referees.