Supplementary plots include a map showing a reconnaisance tsunami pit transect across the Ishinomaki plain (Fig. 1); a location map of hte Haiyuan Fault (Fig. 2); a location map of Cascadia core sites (Fig. 3); A plot showing correlation parameters between key Cascadia core sites (Fig. 4); a plot showing sensitivity to energy state model parameters (Fig. 5); a variation of the energy state model with four key core sites plotted individually (Fig. 6). Table 1 gives the paleoseismological sequence for the Haiyuan fault at Gaowanzi by 3D trenches; Table 2 summarizes the segment based rupture history for the Haiyuan Fault; Table 3 gives the locations for Cascadia cores used in this study; Table 4 gives radiocarbon ages and error ranges, and mass values for correlated turbidites; Table 5 gives a Pearson correlation matrix for mass vaues of correlated at the four Cascadia core sites in this study.
Figure S1. Tsunami pit transect across the Ishinomaki plain. To examine the robustness of the AD 869 Jogan tsunami inundation, we dug five shallow pits across the Ishinomaki Plain west of Ishinomaki. The area is a broad coastal plain with coast parallel beach ridges a few meters in height spaced appoximately .5-.8 km apart. White line indicates approximate coastline in in Jogan time. Red line indicates approximate limit of discernable deposition from the 2011 tsunami. Blue line indicates approximate limit of inundation from the 2011 tsunami based on salt intrusion damage to rice paddies. This area was investigated by Shishikura et al. (2007) who show Jogan tsunami deposit distribution on the Ishinomaki plain. Please see this reference for stratigraphic core descriptions and regional correlation of the Jogan deposit. Five pits are shown by orange markers. Pit one contained no Jogan tsunami as the shoreline was north of Pit 1 during Jogan time (Shishikura et al., 2007). Pits 2-5 all contained the Jogan tsunami, identifiable by its close association with an overlying tephra layer (Shishikura et al., 2007). White boxes indicate appoximate Jogan deposit thickness. At Pit 2, the Jogan deposit was 10-15 cm thick immediately below the ash layer, while the 2011 deposit was 1-2 cm thick. At Pits 3 and 4 the Jogan deposit exceeded 30 cm in thickness. At Pit 5, which is ~125m south of the base of the hills backing the coastal plain, the Jogan deposit was still 5-7 cm thick, suggesting the tsunami runup was terminated by the coastal hills. The inundation distance and thickess of the Jogan deposit along this this is significantly greater than the 2011 tsunami evidence after accounting for the change in coastline position. On the Sendai plain, Jogan deposit extent is similar to the 2011 deposit. Inset shows figure location and approximate epicentral location and source area of the 2011 Mw 9.0 earthquake.
Figure S2. Tectonic setting of the Haiyuan fault. (a) Major active faults and historical earthquakes (ML ≥4.5) in the region adjacent to the Haiyuan fault. Surface ruptures associated with the 1920 and 1927 earthquakes are shown in orange. Tianzhu seismic gap is highlighted in red (after Lasserre et al., 2001). (b) Surface traces of the Haiyuan fault system near Tianzhu and Songshan, superimposed on Shuttle Radar Topography Mission (srtm) Digital Elevation Models (dems). Elevation contour interval is 250 m. Figure from Liu-Zeng et al., 2007.
Figure S3. Cascadia margin turbidite canyons, channels and 1999-2002 core locations. Major canyon/channel systems are outlined in blue. Primary core sites shown with yellow rim, all other 1999-2002 cores are grey. White core numbers preceded with cruise number “M9907”, collected on the RV Melville, Yellow symbols are preceded with the cruise designation “RR0207”, collected on the R/V Revelle. Core EW9504-16PC shown in red. Earlier OSU cores shown in grey. “PC” = Piston Core; “BC” = Box Core; “KC = Kasten core; “GC” = Gravity core; “TC” = Trigger core. Trigger cores omitted for clarity. Inset of Effingham Inlet shows collection site of Pacific Geoscience Centre (PGC) collected piston cores.
Figure S4. A. Correlation of vertical series of coarse fraction pulses per turbidite for Juan de Fuca, Cascadia, Hydrate Ridge and Rogue cores. Table and chart show the number of fining upward coarse units per turbidite for events observed at all four sites. Correlation matrix shows Pearson correlation coefficient for the four series. Four line plots offset slightly for visibility. The number of coarse pulses per event remains quite constant among widely separated core sites. B. Correlation of turbidite mass for Cascadia, JDF channels, Hydrate Ridge and Rogue Apron. Mass (dimensionless here) is derived from gamma density traces (see Goldfinger et al. 2012 methods for details).
Figure S5. Energy State parameter sensitivity. This plot shows variation of the parameter space governing the energy state plot of Figure 4 in the main text. The first two panels (A and B) show variation of the starting point at T18, the first Holocene turbidite. Though nothing is known of this value, energy state of the Cascadia fault through time is not sensitive to this value. Variation of this value over a wide range shifts the entire plot on the vertical (energy state) scale, but does not affect the temporal pattern. The lower two panels (C and D) show variations of the scale factor parameter. This dimensionless scale factor is generated by the time series itself, and is that value required to maintain a no-net energy gain or loss profile through the Holocene time series. Variation of this plot requires a net increase (panel A), or a net decrease in energy state of the Cascadia fault, which we assume would not occur over the relatively short (10,000) year time scale of this study. Net changes, if they do occur, also do not affect the inference of variable energy gain/loss patterns of the fault over time. Error ranges are as described for Figure 4. As shown in Figure 4, predictable behavior does not necessarily occur at either energy maxima or minima, although a longer time series could reveal a pattern or tendency not observed in this time series.
Figure S6. Energy state model for four individual sites. Panels A-D compare primary core sites along the Cascadia margin. A: Juan de Fuca canyon/channel values from core M9907-12PC. This panel is and panel B are similar to Figure 4 in the main text, which average Juan de Fuca and Cascadia Channels. B: Cascadia channel values from core M9907-23PC. C: Hydrate Ridge west basin, values for core RR0207-56 PC. D: Averaged values for cores M9907-30 and 31PC from Rogue Apron. T4-T8 are further averaged between piston and trigger cores for the upper section where there is overlap of the turbidite section. Primary features of the energy series are represented at all four primary sites, including long gaps following T11 and T6, outsized T11 and T16 (though T16 is reduced in size at Rogue). Pattern of energy accumulation/dissipation between the JDF and Hydrate Ridge sites is striking given the separation of ~ 340 km between these sites. The Rogue Apron pattern is somewhat different, but has significant commonality with the other sites. Decline in energy state through T5-T1 is more evident at Rogue Apron, where a discrepancy between Bradley Lake tsunami heights and recurrence intervals was discussed by Goldfinger et al. [2010] and Witter et al. [2012].
Figure S7. Late Holocene earthquake time series expressed as energy gain and loss per event, NE Japan margin. Energy gain is proportional to recurrence between events in years. Energy loss is proportional to seismic moment, scaled to result in no net gain or loss of energy through the late Holocene. Mw estimated for historic Japanese events within the 3/11/11 M9.0 rupture area from various soures, and plotted in Figure 1. Jogan (869 A.D.) earthquake and two predecessors from Minoura et al. [2001] are assumed to be similar to the 2011 Mw 9.0 event. Plot shows long-term energy cycling of the megathrust, and complex behavior over time. First event assumed to occur 800 years prior to earliest event reported by Minoura et al., [2001]. Smaller events apparent in the latter half of the most recent supercycle are unknown in previous cycles, but are most likely present.
The Haiyuan fault zone can be divided into three segments: western, middle and eastern segments. Table S1 is based on a 3D trench at Gaowanzi, trench 13 in Zhang et al. (2005)’s Fig. 1. The 3D trench includes a trench parallel to the fault and 10 trenches across the fault with a distance of ~2m. All the trenches have a width of 1.8~2m and a depth of 2.5~3.5m. There is a lack of co-seismic horizontal displacement on the oldest events (E1 and E2 ) because of unclear marker in the trenches.
Table S2 is made from a book about the Haiyuan fault zone. This fault is a mainly strike-slip fault. Many trenches were excavated across the fault trace only to get the age of rupture event. Only the Gaowanzi trench gave the co-seismic horizontal displacement. The displacements in Table S2 are from this 3D trench, as shown in Table S1. The rupture length is compiled from all the trenches of the books and the Goawanzi 3D trench. The trenches on the eastern segment of Zhang et al.(2005) are not included.
Table S1. Paleoseismological sequence on the Haiyuan fault at Gaowanzi by 3D trenches (Yonkang et al., 1997)
Table S2. Displacement, rupture length of paleoearthquakes and historical earthquake along the Haiyuan fault (Yonkang et al., 1997; State Seismological Bureau, 1990).
Table S3. Core positions shown in Figure S3.
Table S4. Radiocarbon ages, interevent times and turbidite mass per event for 19 Cascadia margin turbidites correlated > 600 km along strike. Mass is arbitrarily scaled and dimensionless as shown.
Table S5. Pearson correlation matrix for mass per event for the four sites included in Fig. 4 in main text, supplementary Figures S5 and S6, and Supplementary Table S4.
Goldfinger, C., Witter, R., Priest, G.R., Wang, K., Zhang, Y. (2010), Cascadia Supercycles: Energy Management of the long Cascadia Earthquake Series, paper presented at the Seismological Society of America Annual Meeting, Portland, Oregon.
Goldfinger, C., Nelson, C.H., Morey, A., Johnson, J.E., Gutierrez-Pastor, J., Eriksson, A.T., Karabanov, E., Patton, J., Gracia, E., Enkin, R., Dallimore, A., Dunhill, G. & Vallier, T. (2012), Turbidite Event History: Methods and Implications for Holocene Paleoseismicity of the Cascadia Subduction Zone, USGS Professional Paper 1661-F, Reston, VA, U.S. Geological Survey, 184 p, 64 Figures.
Lasserre, C., B. Bukchin, P. Bernard, P. Tapponnier, Y. Gaudemer, A. Mostinsky, and D. Rong (2001). Source parameters and tectonic origin of the 1996 June 1 Tianzhu (Mw 5.2) and 1995 July 21 Yongden (Mw 5.6) earthquakes near the Haiyuan Fault (Gansu, China), Geophys. J. Int. 144, 206–220.
Liu-Zeng, J., Klinger, Y., Xu, X., Lasserre, C., Chen, G., Chen, W., Tapponnier, P., and Zhang, B., (2007), Millennial Recurrence of Large Earthquakes on the Haiyuan Fault near Songshan, Gansu Province, China: Bulletin of the Seismological Society of America, 97, 14-34.
Shishikura, M., et al. (2007), Age and distribution of tsunami deposit in the Ishinomaki plain, Northeast Japan, Annual Report on Active Fault and Paleoearthquake Researches, 31-46.
Institute of Geology State Seismological Bureau, and Seismolocial Bureau of Ningxia Hui Autonomous Region. The Haiyuan Active Fault Zone. Beijing: Seismological Press, 1990. 177-255 (in Chinese)
Witter, R. C., Zhang, Y., Wang, K., Goldfinger, C., Priest, G. R., and Allan, J. C., (2012), Coseismic slip on the southern Cascadia megathrust implied by tsunami deposits in an Oregon lake and earthquake-triggered marine turbidites, J. Geophys. Res., 117, B10, B10303.
Yongkang, R., Duan, R., Deng, Q., Jiao, D., and Wei Min, W., (1997), 3-D Trench Excavation and Paleoseismology at Gaowanzi of the Haiyuan Fault，Seismology and Earthquake，19, 297-107 (in Chinese).
Peizhen Zhang, P., Wei Min, W., Qidong Deng, Q, and Fengying Mao, M., (2005), Paleoearthquake rupture behavior and recurrence of great earthquakes along the Haiyuan fault, northwestern China, Science in China Series D: Earth Sciences, 2005, 48, 3, 364-375.
[ Back ]
