On 5 December 1997 at 23:26:51 local time (11:26:51 UTC), a large earthquake (Mw 7.7–7.9; we use 7.8) shook the region of the Kronotsky Peninsula, Kamchatka, Russia (Figs. 1, 2; Gordeev et al., 1998). The earthquake was characterized by a typical foreshock–mainshock–aftershock sequence (Gusev et al., 1998; Fedotov et al., 1998; Balakina, 2000; Zobin and Levina, 2001; Kuzin et al., 2007; Slavina et al., 2007). Most studies of the earthquake calculate a moment magnitude of 7.8 for the energy released in the first 60–80 s of the main rupture (e.g., Zobin and Levina, 2001). Gusev and Shumilina (2004), in reassessing many Kamchatka earthquakes, assign Mw 7.9 to Kronotsky 1997. In addition to the mainshock, and using GPS measurements, Gordeev et al. (2001) calculate Mw 7.7 for deformation in the pre-seismic half-month, and approximately Mw 7.9 for post-seismic deformation; Bürgmann et al. (2001) calculate Mw 7.7 of (post-seismic) aseismic energy release in the 2 months following the mainshock, also based on GPS data.

The locations of the mainshock and of any slip concentration for this earthquake have not been well resolved, and with one early exception (Sohn, 1998), locators have not used tsunami data. Based on seismic data, the locations of foreshocks and the mainshock/epicenter (Fig. 2) are in the northern part of the interpreted rupture area. A number of analytical locations of the mainshock lie under the NE Kronotsky Peninsula (Fig. 2; Table S2). Some analyses interpret the rupture to have propagated NE to SW (Petukhin et al., 1998), deepening toward the SW. Gusev (2004) maps the entire aftershock zone as part of the 1997 event (Fig. 1). On the other hand, the linear zone of aftershocks in the SW (Fig. 2) has been interpreted to be a separate stress zone (Kuzin et al., 2007) potentially along a separate transverse fault (Slavina et al., 2007). In an analysis focused on GPS data, Bürgmann et al. (2001) place the majority of the primary rupture energy in the southern half of the aftershock zone.

The most complete contemporary record of the 1997 Kronotsky tsunami is from far-field tide gages. Both proximal tide gages, in Ust' Kamchatsk and in Nikol'skoe (Bering Island) (Fig. 1), were not functioning when the tsunami arrived. The Petropavlovsk-Kamchatsky gage is very protected and shows a wave train with an amplitude of about 0.01 m (Zayakin and Pinegina, 1998). The tide gage at Nikol'skoe resumed recording after the first 10 h of the tsunami, with a few centimeters of amplitude remaining (Zayakin and Pinegina, 1998). The far-field tsunami had tide-gage amplitudes in Alaska/Aleutians and Hawaii in line with other tsunamis traveling to Hawaii from the Russian Far East (Table S3; Fig. S4). The tsunami was recorded on at least 12 tide gages, with the highest amplitude (half of wave height) of 0.3 m at Kahului, Maui, Hawaii (NCEI online database). Deep-water pressure sensors deployed at that time in the North Pacific were all in tsunami shadows for this tsunami source, and in all cases, the modeled and measured tsunami was within the noise level of the buoys (Bourgeois and Titov, 2001; no event page at http://nctr.pmel.noaa.gov/database_devel.html).

A truncated post-earthquake and tsunami survey by helicopter took place on 9 December 1997 (Leonov, 1998; Zayakin and Pinegina, 1998). The survey reached as far north as Kronotsky Cape on the Kronotsky Peninsula (Fig. 1) and found that the tsunami had not exceeded the unvegetated sandy beach. At this time, the beach was covered with a thin layer of ice and snow, which in places had been coated by the tsunami with a thin sand layer and elsewhere had been broken up by the tsunami (Fig. 3). The team did not have surveying equipment and estimated runup to be no more than 3 m (T. Pinegina notes), and the published report gave a maximum of 1–1.5 m. The turnaround point in the survey was dictated by fuel and available daylight.

On 5 December 1997, two rangers were in a cabin near Big Chazhma River (Fig. 1); one of them was interviewed (in Petropavlovsk-Kamchatsky) by T. Pinegina on 19 April 2001. They felt the earthquake that night, and the next day, as was routine, they went via snowmobile to survey the northern coastal part of Kronotsky Nature Reserve, in the Little Chazhma River area. At the mouth of the Big Chazhma River, they saw jumbled ice and seaweed on the snow; a cabin on the south bank of the Little Chazhma River was partly wetted, and there was seaweed on the snow. Normally the rangers crossed the river near this cabin, but the river was a jumble of ice and they had to go some distance upstream in order to cross (on ice). On the other side, they could not continue north because there was water in the low spot between beach and hill (see Fig. 4, our profile 100).

Based on results of the post-tsunami survey (reported to Sohn by V. Gusiakov), Sohn (1998) analyzed the tsunami with regard to its earthquake source and concluded that the main rupture must have lain largely under land, in order to explain the low runup accompanying a moment magnitude the author calculated as Mw 7.7.

The Kamchatka Peninsula has a short but rich historic record of large earthquakes and attendant tsunamis; herein we discuss only 20th century tsunamis originating in or having been recorded in the field region of Kamchatsky Bay (Table 1). In addition to locally originated tsunamis, Kamchatka is vulnerable to tsunamis from Chile, less so from Peru, and minimally from Japan, Alaska/Aleutians, and Central America, due to directivity (e.g., see Table S1). Based on scant records (Table 1), the 1960 Chile tsunami likely reached elevations of 3–5 m above sea level along Kamchatsky Bay (Fig. 1), on the order of 2 times higher than the 1952 southern Kamchatka tsunami in this bay (Table 1)

The largest documented local tsunamis from earthquakes near Kronotsky Peninsula (Fig. 1; Table 1) are two from 1923, both having local as well as far-field records (Table S4); both may have affected south-central Kamchatsky Bay. There was also a 24 February 1923 Mw 7.6 earthquake in this area (Fig. 1; Gusev, 2004); however, it has no historical tsunami record in the near or farfield. The Mw 8.0 1917 earthquake along the Steller fracture zone (Fig. S1) also did not produce a recorded tsunami. The 3 February 1923 Kronotsky Bay earthquake (Mw 8.5) was located south of Kronotsky Cape (Fig. 1), and its tsunami was large (6–8 m) in Kronotsky Bay (Table 1), decreasing northward; a sled team in the area during and after the earthquake reported a coastal ice rampart being pushed about 3 km upstream on the (Big?) Chazhma River, north of Kronotsky Cape. The 13 April 1923 north Kamchatsky Bay earthquake (Mw 7.3 in NCEI catalogue; 14 April local time) generated a very high tsunami in north to north-central Kamchatsky Bay (Tables 1, S1), with large (largest; naibolshii) effects south to Cape Shubert in south-central Kamchatsky Bay (Fig. 1; Troshin and Diaghilev, 1926). Based on tsunami amplitudes, Gusev and Shumilina (2004) suggested this April 1923 earthquake had a moment magnitude of 8.2 (Table S1, Fig. S2). In sum, the February and April tsunami runup was large south and north (respectively) of our field area, decreasing toward that field area.

The record of earthquakes and tsunamis on Kamchatka prior to the 20th century is spotty but improving (Zayakin and Luchinina, 1987; Godzikovskaya, 2010). Earthquakes on 17 May 1841 and 17 October 1737 originated in the region of the 1952 south Kamchatka great earthquake, so they likely did not have significant effects in (southern) Kamchatsky Bay (see Table 1, 1952 runup). Other tsunamis that may have affected southern Kamchatsky Bay are an autumn 1849 tsunamigenic earthquake in the vicinity of the Komandorsky Islands (Godzikovskaya, 2010) and a 1791 event with an intriguing account of having affected the mouth of the Kamchatka River (Ust' Kamchatsk), reported to reach 7 km upstream (Zayakin and Luchinina, 1987).

The southern field site (Fig. 1) which we call “Chazhma” (Fig. 4) is a narrow strip (∼ 400 m wide or less) of Holocene accumulative coastline along a rugged coast just north of the Kronotsky Peninsula. The two profiles near river mouths (Chazhma 210 and Chazhma 130; Fig. 4) maintain lower elevations (< 4 m) over much of their distance, though both reach elevations of more than 6 m above sea level. The other five profiles rise, typically in sharp steps indicative of Holocene uplift events (as in Pinegina et al., 2013), reaching typical maximum levels of 8–10 m (Fig. 4). Net uplift on these profiles is consistent with longer-term uplift of Pleistocene terraces on the Kronotsky Peninsula (Melekestsev et al., 1974).

The northern field site which we call “Storozh”, extending north to the Bystraya River (Figs. 1, 4), is a broader strip (typically 600 m wide) of Holocene accumulative coastal plain associated with active and drowned river mouths. Two of these profiles (140, 001; Fig. 4) drop in elevation behind one or more beach ridges. The other seven profiles are typified by a series of beach ridges, of which the seaward ridges are higher, reaching typically 6–7 m, with an average elevation of the profile in the range of 4–6 m (Fig. 4). Such profiles indicate minor subsidence or no vertical change in the late Holocene.

Sediment runup data (Table 3, Fig. 6) indicate that in southern to central Kamchatsky Bay, the 1997 Kronotsky tsunami ran up as much as 9.5 m, averaging 6.1 m, with moderate inundation distances of 100–300 m. The general pattern over about 100 km of coastline, including post-tsunami survey observations on Kronotsky Peninsula itself, is relatively smooth, and we also expect based on the pattern that there was run up north of our northernmost profile (Fig. 6), but north-central Kamchatsky Bay comprises sea cliffs, not coastal plain. The maximum elevation reached by the tsunami deposit is higher on southern (Chazhma) profiles. However, lower runup numbers on northern profiles may be an artefact of their lower elevations (Fig. 4); inundation distances are greater on these profiles (Table 3). On some profiles the 1997 deposit is absent.

Sediment runup and inundation data for “1923” indicate that this tsunami (or tsunamis) was larger than 1997 in the region of our profiles. The deposit we interpret as from “1923” is usually thicker and more extensive, and never less extensive, than the deposit from 1997 (see Figs. 5, 7, 8, 9). The “1923” deposit is present on all measured profiles whereas the 1997 deposit is missing on six (Table 3, Fig. 6). Only on profiles where the sediment limit was not found (e.g., 100), or where profiles dropped to low elevations at their landward extent (001, 180, 160, 140, 100, 130, 210) were “1923” deposits at similar or lower elevations than 1997; in many of these cases (001, 180, 160, 130), inundation distances for “1923” were longer. Even in the few cases where our field locations did not distinguish 1997 from “1923” by sediment runup or inundation (e.g., Storozh 140, Fig. 9), the “1923” deposit was coarser and/or thicker than 1997.

Between “1923” and 1997 deposits on a few profiles (Table 3), there is a thin, patchy and less extensive deposit which we attribute to the 1960 Chile tsunami (e.g., Fig. 4, right). We favor 1960 Chile over 1952 Kamchatka for two reasons. First, the 1960 tsunami was larger than 1952 in the Kamchatsky Bay region (Table 1); the more locally generated 1952 tsunami dies off in amplitude along the strike of the rupture (MacInnes et al., 2010), whereas the Chilean tsunami on Kamchatka is little affected by latitude (Zayakin and Luchinina, 1987). Second, supporting the 1960 interpretation, in one excavation on profile 001, this intermediate tsunami deposit lies above the Bezymianny 1955 tephra layer (Fig. 7).

In many excavations (e.g., profile 100 in Fig. 4, profile 110 in Fig. 8), there is a tsunami deposit within a few centimeters of the base of KS1907 and which is comparable to 1997 and “1923” in thickness and extent. Although pre-1907 sedimentation rates are difficult to determine this tsunami deposit must fall within the historical period, which extends back to 1737. However, the more complete historical records are from southern Kamchatka, and records from the second half of the 19th century are particularly spotty (Gusev and Shumilina, 2004). Thus there is no known historical event we can assign to this deposit; optically stimulated luminescence (OSL) dating might help in interpreting this deposit.

Our observations are consistent with 1997 being a seismogenic tsunami source with significant rupture energy expended in the northern portion of the zone of aftershocks. The extensive and relatively smooth distribution of runup (Table 3; Fig. 6) and the ratio of maximum runup to distance over which the tsunami had significant runup (on the order of 10-5) indicate that this tsunami was typical of a seismogenic source rather than a landslide source (see Okal and Synolakis, 2004). The far-field tide-gage records (e.g., Hilo, Table 1) are also indicative of a broad rather than a point source. Given that the post-tsunami survey reported runup that did not exceed the beach on the Kronotsky Peninsula and that the deposits we mapped north of the peninsula are from the 1997 tsunami, any source model must explain the low (“water”) runup on Kronotsky Peninsula and relatively high (“sediment”) runup north of this peninsula (Fig 6). Source–region models by Bürgmann et al. (2001) and Llenos and McGuire (2007), for example, do not include the northern aftershock area, and such models have been used to interpret Kamchatka subduction-zone behavior (e.g., Song and Simons, 2003; Bürgmann et al., 2005; Llenos and McGuire, 2007; Bassett and Watts, 2015). On the other hand, source regions by Gusev et al. (1998), Gusev (2004), and Levina et al. (2013) tend to include the entire aftershock zone, overlapping February 1923 in the south but also filling the gap between February 1923 and April 1923 (Fig. S1), which might not be consistent with the tsunami data. Slavina et al. (2007) interpret the southwestern aftershock activity (Fig. 2) to be on a separate, transverse fault, and Kuzin et al. (2007) interpret the SW portion of the (extended) aftershock region to be a separate stress zone, interpretations more consistent with tsunami data. Zobin and Levina (2001) favor most mainshock energy being generated in the middle zone defined by fewer aftershocks (see Fig. 2), but this region is in shallower water, less conducive to tsunami genesis. A recently published finite-fault model resolves to most slip being under the Kronotsky Peninsula, with most energy release focused in the north (Hayes, 2017; https://earthquake.usgs.gov/earthquakes/eventpage/usp0008btk#finite-fault). As with the Sohn (1998) analysis, the Hayes (2017) model cannot explain the 1997 tsunami runup because the rupture is mostly under the Kronotsky Peninsula. Shifting this pattern of deformation eastward could resolve the discrepancy.

There are reasons to favor either or both the 3 February 1923 and 13 April 1923 Kamchatka tsunamis as the generator(s) of the deposit above KS1907 that we identify as “1923” (e.g., Figs. 7, 8, 9). Given what is known (Table 1), south-central Kamchatsky Bay is the place most likely to have comparable runups from each. Both tsunamis have a record in Hilo, but one is runup and the other tide-gage amplitude. There is no case on Kamchatka of a pair of similarly measured records from the same locality with which to compare the two tsunamis, with the exception of observations that the April tsunami generated more damage at the Tsutsumi fish plant southeast of Ust' Kamchatsk (Table S4). The 3 February tsunami was larger in most catalogued locations (Table S4) but apparently smaller than April 1923 in north Kamchatsky Bay. The two 1923 tsunamis both occurred while the ground would have been snow covered so that following snowmelt, it would be nearly impossible to distinguish two different deposits. The source regions of the two 1923 Kamchatka tsunamis have been mapped (Fig. 1), but they are not easy to constrain in detail other than saying that the February earthquake was south of Kronotsky Peninsula and the April earthquake north of it (Fig. 1). The February earthquake has been catalogued as Mw 8.3–8.5 (ISC event 911271; NCEI) and the April earthquake as Mw 7.1–7.3 (ISC event 911331; NCEI), but the local and far-field tsunami runup for April 1923 suggests it may have been significantly larger (Gusev and Shumilina, 2004); based on its tsunami, Gusev suggests Mw 8.2 for the April earthquake. A moment magnitude around 7.8–8.0 for the April earthquake would be more consistent with its tide-gage amplitude in Hilo (Fig. S2).

For the record and analysis of tsunami deposits below KS1907, for each excavation we count the number of deposits between marker tephra and determine the approximate elevation above sea level and distance from shore of the excavation locale in that time (tephra) interval (Fig. S5; see Figs. 7, 8, 9 and their captions for more details on our interpretations). For some layers, an excavation may be their limit and for others not (e.g., Fig. 9). We do not attempt to correlate sand layers from excavation to excavation (or profile to profile), though there are cases where it is possible; the problem with distinguishing February 1923 from April 1923 deposits illustrates the potential for mis-correlation. The reasons that not all deposits are present in all excavations range from preservation to separation – for example, excavations near the coast will commonly contain amalgamated sand layers (e.g., Bourgeois et al., 2006). For each profile, we count the maximum number of tsunami deposits between tephra, which is our indication of how many tsunami events have occurred

In order to summarize paleotsunami sizes, we determine sediment runup – or the highest point seaward, whichever is higher – and sediment inundation for tsunami deposits on each profile. For each tephra interval along each profile, there will be deposits at maximum distances and maximum elevations; the two measures are treated separately because tsunami deposits are not correlated (in fact, high runup is associated with shorter, steeper profiles and long inundation with low-relief profiles). For example, for the historical deposits, two points are plotted (Fig. 10) – their point of maximum inundation and their point of maximum runup, which are usually on separate profiles.

A few of the paleo-events are comparable to Chile 1960 (Fig. 10), but most are likely from locally generated tsunamis because Chile 1960 was an outsized event, and its deposit is not well represented on the profiles. The 1997 tsunami has dimensions similar to the majority of paleotsunamis as represented by sediment runup on the order of 5–7 m (Fig. 10). The “1923” deposit (unknown if related to February, April, or both) is a “typical largest” event (Fig. 10). Recall that in these field sites there are few excavations at elevations of 10 m or more (Fig. S6), and that these higher elevations are on uplifted profiles, so in this situation we cannot have a record of older paleotsunamis reaching such elevations, simply as an artefact of the profile history (Fig. S5). This issue is present also for paleo-inundation on prograding profiles, but is not such a strong artefact in our data set. Overall, the number of deposits tends to decrease away from the coast and at higher elevations (density of points in Fig. 10), although there is a lot of scatter in the data, likely due to preservation and identification differences (e.g., Fig. 9).

To determine tsunami recurrence according to size, we consider all tsunami deposits above KS1 (∼ AD 300) at elevations greater than 5 m (Fig. 11). We only use excavations now at or reconstructed to be more than 5 m above sea level or landward of a beach ridge (reconstructed to be) higher than 5 m to be more confident we are analyzing tsunami deposits, not those of storms or floods, and to eliminate most non-local tsunamis. We did not use intermediate Shiveluch tephra layers between KS1907 and KS1 (Table 2) because their presence is not consistent enough to break down recurrence statistics, and the time intervals are short relative to the number of events, so statistical analysis cannot be supported. The grand total of the maximum number of events (per each interval) is 18 deposits, including the historical cases. For each event, we determine a maximum sediment runup, that is, if there are four deposits between two marker tephra on a given profile, we determine the four highest points those deposits reach; e.g., two may reach 8.3 m and the other two only 7.2 m (all four are considered to have reached 7.2 m). We use reconstructed distances and elevations for each time interval below KS1907. The maximum elevation is either sediment runup ( h) or maximum elevation before sediment runup (H; as in Fig. 5), whichever is higher. Independent of the determined maximum elevation, we determine a maximum sediment inundation for each deposit in each tephra interval.

All 18 deposits represent large tsunamis, reaching minimum elevations of 5 m (smaller not considered) and inland distances of 100 m, with each factor having a recurrence interval of about 100 years (Fig. 11). Note again that runup and inundation are not paired; high runup commonly occurs on shorter, steeper profiles and long inundation on lower profiles. Tsunamis reaching an elevation of at least 7 m have a recurrence of ∼ 200 years (Fig. 11). The largest reconstructed tsunamis as recorded by tsunami deposits have runup of 10 m or more and occur on average every 425 years. Tsunamis with inundation of 600 m or more occur on average every ∼ 570 years.

This work adds to the tsunami catalogue for 1997 Kronotsky and 1960 Chile, but not February or April 1923 Kamchatka events because we cannot differentiate between the (two) 1923 deposits. The near-field nature of the 1997 Kronotsky tsunami is significantly revised by our report of coastal profiles north of the Kronotsky Peninsula, adding substantial data to its catalogue. The 1997 tsunami reached runup heights of more than 9 m, averaging 6 m over about 60 km of coastline. As would be expected, tsunami heights (as indicated by deposits) and inundation distances are influenced by coastal topography, with higher runups on steep profiles and longer inundation on lower-relief profiles. Data catalogues do not commonly provide topographic profiles, yet this information can be critical to understanding a tsunami and potentially its generating source.

Based on deposits from 15 profiles and more than 100 excavations, we conclude that in southern to central Kamchatsky Bay, the 1923 tsunami (February or April, indeterminate) was larger than the December 1997 Kronotsky tsunami, but the summary and tabulated data (Fig. 6, Table 3) are tricky to interpret, with sediment inundation (L) being more indicative of tsunami size than runup (h) or highest point seaward of runup (H; e.g., see Fig. 5 illustration). On the basis of the total number of profiles exhibiting a deposit, “1923” is more extensive, but its average sediment runup (h) value is lower because the farthest point it reached on a number of profiles is actually lower than the closer-to-shore points for 1997. Moreover, even though “1923” exceeded more of the high beach ridges seaward of the (sediment) runup point (H), the average of those is almost the same as for 1997 (Table 3). Thus the most telling measurements distinguishing 1997 from “1923” are sediment inundation distances, with the average for “1923” almost twice that for 1997.

The 1952 tsunami deposit in southern Kamchatka (and the northern Kuril Islands; MacInnes et al., 2010) reaches greater heights and inundation distances along its earthquake rupture zone than any of the historical tsunami deposits along the northern part of the Kamchatka subduction zone (this study; also Pinegina, 2014). While this observation is not surprising given that 1952 was Mw 9.0 and the historical events to the north no larger than about Mw 8.5, this leaves us with the question: can (does) the northern part of the subduction zone produce Mw 9 events, or does Kronotsky Cape represent a locked or continuously slipping zone that keeps ruptures shorter, as in 1923? For that, we must turn to the prehistoric record.

The sediment runup and inundation data reported here require a reevaluation of rupture source models for the 1997 Kronotsky earthquake; we favor slip focused within the northern half of the aftershock zone shown in Fig. 2 (also see Fig. S9). Models which place most rupture energy to the south of or under the Kronotsky Peninsula (Fig. S9; e.g., Bürgmann et al., 2001, 2005; Llenos and McGuire, 2007; Bassett and Watts, 2015; Hayes, 2017) are not consistent with the tsunami data. The tsunami, rather than being unusually small for its generating earthquake's moment magnitude (Sohn, 1998), produced runup averaging 6 m over about 60 km of coastline, and 30 cm amplitude on the Hilo tide gage, requiring a “normal” offshore, subduction-zone rupture. Moreover, some significant portion of that rupture must be under substantial water depth to produce the indicated tsunami in the bay north of Kronotsky Cape, while not generating as much runup on the Cape, or to its south. While part of the rupture could well have been under the Kronotsky Peninsula and the relatively shallow region directly offshore, deformation in deeper water east and north of the peninsula is needed.

We conclude that a rupture consistent with the mainshock and aftershock locations from Kamchatka's network are more reasonable than more westerly locations, e.g., in the ISC catalogue (Fig. 2, Table S2). This issue is illustrated by the Hayes (2017) inversion, which takes the National Earthquake Information Center (NEIC) hypocentral location (Table S2) to start and, while this inversion results in most slip to the north (Fig. S9), locates that slip under the peninsula, where it cannot generate a tsunami. If this inversion were located based on the Kamchatka network's mapped mainshock, it might explain the 1997 tsunami.

The northern part of the Kamchatka subduction zone ruptured in two large tsunamigenic events in February 1923 and April 1923 (Fig. 1), and our study indicates that a substantial portion of the energy released by the 1997 Kronotsky earthquake was generated in a seismic gap between those earthquakes (and a large 24 February 1923 aftershock; Fig. 1), as originally recognized by Fedotov et al. (1998) and predicted by the authors' earlier work. The Kronotsky Peninsula lies landward of the (subducting) Emperor Seamount chain, which has been postulated to generate a locked or slowly slipping zone on the KSZ, a zone characterized by a relatively strong positive gravity anomaly (e.g., Bürgmann et al., 2005; Llenos and McGuire, 2007; Bassett and Watts, 2015; Fig. S9). The behavior of the subduction zone off/under Kronotsky Peninsula may have well kept the northern Kamchatka subduction zone from generating the 1952-scale (Mw 9) Kamchatka earthquakes, but the 1997 tsunami is evidence that this segment does rupture.

The area of the southern to central Kamchatsky Bay contains a relatively short but well-preserved record of paleotsunami deposits which can be calibrated with the historical record. Combined with the record in northern Kamchatsky Bay (Pinegina et al., 2012; the north-central bay is characterized by cliffs), the pattern of runup and inundation in the prehistoric record for the last 1700 years does not diverge from the 20th century record. Compared with southern Kamchatka, the region where Mw 9-scale events occurred in 1952 and 1737, the northern subduction zone has generated smaller and less extensive tsunamis, in agreement with analyses of Bürgmann et al. (2005) for the modern and Pinegina (2014) for the prehistoric record.

A robust, 1700-year-long record may be sufficient to generate a probabilistic hazard analysis that can be used for both local and far-field hazard studies, and not only for tsunami recurrence statistics but also for recurrence statistics that include tsunami size. Reconstructing paleo-runup and paleo-inundation requires, and is thus limited by, accurate reconstructions of past shoreline locations and past (relative) sea levels. Coastlines with well-established marker tephra can enable such reconstructions, as shown by this study.

As are seismologists, paleoseismologists are cautioned to qualify our generalizations by the lessons of the 11 March 2011 Tohoku earthquake and tsunami. Characterizing subduction-zone behavior and quantifying its hazards are goals which we will only ever accomplish imperfectly.