The Role of Holocene Relative Sea-Level Change in Preserving Records of Subduction Zone Earthquakes
- 1.5k Downloads
Eustasy and glacio- and hydro-isostatic adjustment are the main drivers of regional variability of Holocene relative sea-level (RSL) records. These regional variations in Holocene RSL influence the preservation of coastal wetland stratigraphic records of prehistoric earthquakes along subduction zone coasts. The length and completeness of prehistoric earthquake records is intrinsically linked to the accommodation space provided by gradually rising (<3 mm/year) Holocene RSL. In near-field regions that were located beneath northern hemisphere ice sheets (e.g., western Vancouver Island), RSL fall from a mid-Holocene highstand has limited prehistoric earthquake records to the last 1 ka. In intermediate field regions (e.g., southern Washington and central Oregon), gradual RSL rise over the last ∼7 ka has preserved widespread records of prehistoric earthquakes. In far-field regions (e.g., Sumatra, Chile, and Japan), fragmentary stratigraphic evidence of prehistoric earthquakes has been preserved only during periods of gradual RSL rise prior to a mid-Holocene highstand, or during the last 1–3 ka, when RSL was within 2 m of modern sea level, and thus within the tidal frame.
KeywordsRelative sea level Glacio-isostatic adjustment Prehistoric earthquakes Accommodation space Coastal wetland stratigraphy Subduction zone
Holocene relative sea-level (RSL) change is the net effect of eustatic factors (land ice mass change and ocean thermal expansion), regional glacio- and hydro-isostatic adjustment (GIA), ocean dynamics, tectonic uplift or subsidence, and other local factors that produce complex patterns of RSL rise or fall over space and time [1, 2, 3]. These complex regional patterns of RSL change have influenced the deposition and preservation of sediment representing coastal wetland environments (e.g., tidal marshes and mangroves) that preserve stratigraphic records of Holocene RSL change [4, 5, 6]. The formation and preservation of coastal wetland stratigraphic sequences are strongly dependent on the accommodation space created by RSL change [7, 8, 9]. The most spatially and temporally complete coastal wetland stratigraphic archives are found along coastlines with gradually rising RSL during the Holocene . In contrast, coastlines that have experienced an RSL fall  from a mid-Holocene highstand have fragmented and spatially limited coastal wetland stratigraphic records.
Holocene coastal wetland stratigraphy can preserve evidence of RSL change related to prehistoric earthquakes and tsunamis [8, 12, 13, 14, 15]. The societal and scientific value of prehistoric records of the largest, most infrequent subduction zone earthquakes and their tsunamis was underscored by the devastating impacts of the 2004 Indian Ocean and 2011 Tohoku events [16, 17]. If seismic and tsunami hazard assessments in these regions had fully considered coastal geologic evidence for past subduction zone earthquakes and tsunamis, losses might have been substantially reduced [17, 18]. One of the most widely applied methods of reconstructing subduction zone earthquake histories on millennial timescales employs coastal wetland stratigraphic sequences to identify sudden changes in RSL during coseismic vertical deformation of the coast [12, 13, 19, 20, 21]. However, the time spans and completeness of prehistoric earthquake records varies widely among subduction zone coastlines.
In this review, we discuss the factors that drive patterns in Holocene regional RSL change along subduction zone coasts and how these patterns influence the formation of coastal wetland stratigraphic sequences likely to preserve evidence of past subduction zone earthquakes. We contrast the relatively complete Holocene stratigraphic archives of prehistoric earthquakes along the Cascadia subduction zone, where RSL has been gradually rising during the Holocene, with incomplete and fragmentary records from the subduction zone coasts of Sumatra, Chile, and Japan, where late Holocene (last 4 ka) sea-level fall has limited the preservation of prehistoric earthquake evidence. Considering stratigraphic records of coseismic vertical deformation in the context of Holocene RSL change reveals the most likely time periods and locations for the preservation of millennial-scale records of past earthquakes along various subduction zone coasts.
Spatial Variability of Holocene Relative Sea-Level Change
The spatial expression of RSL change during the Holocene varies among regions once covered by the northern hemisphere ice sheets (Fig. 1a; near-field), located at the periphery of these ice sheets (Fig. 1b; intermediate-field), and regions distant from these major glaciation centers (Fig. 1c–e; far-field) [1, 11, 28].
Near-field regions are strongly influenced by local ice (un)loading, which produces vastly contrasting patterns in RSL change during the Holocene. Near the center of former northern hemisphere ice sheets (e.g., Hudson Bay), the rate of glacio-isostatic uplift exceeded the rate of eustatic sea-level rise during the Holocene . But at the margins of the ice sheet (e.g., western Vancouver Island Canada), the rate of eustatic sea-level rise outpaced glacio-isostatic uplift until ∼7 ka, after which glacio-isostatic uplift became the dominant control on RSL, resulting in a mid-Holocene highstand .
RSL change in intermediate-field regions is influenced by the formation and collapse of the proglacial forebulge . Ice loading during the last glacial maximum (∼26 ka; LGM) caused the migration of mantle material away from ice load centers, resulting in uplift of a forebulge at the periphery of ice sheets [31, 32]. Progressive melting of the ice led to the collapse of this forebulge (glacio-isostatic subsidence) as mantle material returned to the former load centers . In intermediate-field regions (e.g., US Atlantic and Pacific coasts), isostatic and eustatic effects worked in tandem to produce rapid RSL rise up to ∼7 ka BP . After ∼7 ka, the eustatic input diminished and continuing glacio-isostatic subsidence became the predominant control on RSL rise [3, 28].
In far-field regions distant from northern hemisphere ice sheets (e.g., Atlantic and Pacific coasts of South America), eustatic, hydro-isostatic, and glacio-isostatic processes are dominant [35, 36]. In these regions, the RSL pattern is characterized by a rise to a mid-Holocene sea-level high- or still-stand between ∼7 and 5 ka [37, 38]. The fall in RSL to present was produced when the rates of eustatic sea-level rise was exceeded by the combined effects of hydro-isostatic loading of the continental shelf (continental levering) and water migrating away from far-field equatorial ocean basins in order to fill space vacated by collapsing forebulges (equatorial ocean siphoning) [1, 2].
The Preservation of Stratigraphic Evidence of Earthquakes During Long-Term RSL Change
Prior knowledge of the Holocene RSL history of a subduction zone is the first step in deciding which sites to target to reconstruct earthquake histories from coastal wetland stratigraphy. Increasing tidal inundation during gradual (<3 mm/year) RSL rise leads to increased sediment deposition and the growth and aggradation of wetland vegetation . However, if RSL rises too fast (>10 mm/year), organic matter contributions from coastal vegetation are reduced, accelerating erosion and replacing coastal wetlands with subtidal environments [9, 40, 41]. RSL fall starves coasts of sediment and limits coastal wetland formation to protected estuaries, barrier lagoons, and deltas where sediment supply is relatively high .
At coastlines that experience long-term RSL fall, similar coseismic subsidence or uplift results in minimal burial of wetland soil O horizons and tidal flat mud because of limited accommodation space and low sedimentation rates [52, 53, 54, 55, 56]. Coastal wetland stratigraphic sequences that form under RSL fall are thin (Fig. 2b) and discontinuous with many unconformities. As the coast emerges out of the tidal frame during RSL fall, coastal wetlands are progressively stranded  and the upper sediment of emerged wetland stratigraphy is eventually bioturbated, oxidized, and incorporated into upland soils (Fig. 3b). Emergence also leads to sediment erosion that further obscures stratigraphic contacts [8, 53, 56]. Evidence of prehistoric earthquakes in such emerged wetland sequences is typically difficult to correlate among cores or sites .
Regional RSL and Earthquake History
In order to demonstrate the influence of differing rates of Holocene RSL change on the length and completeness of prehistoric earthquake records, we highlight RSL changes and earthquake histories at near-, intermediate-, and far-field regions at four subduction zones. Our near-field (western Vancouver Island) and intermediate-field (southern Washington, and central Oregon) regions are located along the Cascadia subduction zone. Cascadia’s RSL history has been reconstructed using sea-level index points that record the position of RSL over time . Each index point contains information about its (a) geographic location; (b) calibrated radiocarbon age and error (2σ); and (c) elevation of former sea level and its vertical error (2σ) . Sea level index points for Cascadia are listed in a comprehensive RSL database for the Pacific coast of central North America .
To estimate rates of RSL change for the Cascadia subduction zone, we use a spatio-temporal empirical hierarchical statistical model, as in Kopp et al. . The spatio-temporal model fits all index points simultaneously, not on a site-by-site basis. Details of the model are described in Online Resource 1.
Examples of subduction zones in far-field regions include the Sunda (northern and western Sumatra), Chile (central and southern Chile), and Japan (Sendai Plain, Tokyo metropolitan area, and eastern Hokkaido) subduction zones. Due to the absence of RSL databases in our far-field regions, we relied on glacio-isostatic adjustment model predictions to characterize RSL change over the Holocene [33, 61]. For each far-field region, we generated sea-level predictions using two GIA models—the ICE-6G_C (VM6) model  and the ICE model .
The differences in the RSL predictions during the Holocene from these two models is driven predominately by differences in the deglaciation histories of the Antarctic Ice sheet, with a larger total ice-volume equivalent sea-level contribution (26 m versus 13.6 m)  and melting continuing until 1 ka (compared to 4 ka) in the Bradley et al.  model (Online Resource 2). For the ICE model, we generated predictions for a suite of earth models to look at the impact of changes in lithosphere thickness and upper and lower mantle viscosity on RSL predictions . Rates of past RSL change were derived from the GIA predictions by calculating the numerical derivative of each curve over 1-ka time intervals for an upper and lower range of rate values.
Within regions near-, intermediate-, and far-field from northern hemisphere ice sheets, we selected the longest records of prehistoric earthquakes preserved in coastal wetland stratigraphy to compare to the RSL histories and models. The selected studies used a standard set of criteria  to support their interpretations of coseismic and interseismic origin for the changes in coastal wetland stratigraphy. The key criteria used were the lateral extent of sharp stratigraphic contacts; the suddenness and amount of coseismic vertical deformation; the synchroneity of coseismic vertical deformation among regional sites; and the coincidence of tsunami deposits with sudden changes in stratigraphy. To plot prehistoric earthquake histories against RSL histories and models, we used calibrated radiocarbon age ranges (2σ) for earthquakes reported in the original publications (Online Resource 3).
Case Study: Cascadia Subduction Zone
Because the Cascadia subduction zone (Fig. 1a, b) has not experienced a great earthquake since the M8.8–9.2 Cascadia earthquake in AD 1700 [63, 64], we must rely on paleoenvironmental reconstructions of changes in RSL to reconstruct the patterns, timing, and magnitudes of past earthquakes. Fortunately, the creation of sediment accommodation space by gradual Holocene RSL rise [7, 59, 65] along Cascadia’s coasts has produced unusually complete stratigraphic archives of coseismic subsidence spanning thousands of years [12, 20, 45, 50, 51, 65, 66, 67]. There are, however, significant differences in the lengths of prehistoric earthquake records at sites formerly covered by LGM ice sheets, compared with those located south of the ice margin (Fig. 1a, b).
Near-Field Region Sites at Cascadia
Intermediate-Field Region Sites at Cascadia
In southern Washington, the oldest coastal wetland deposits are dated to ∼5.3 ka . RSL rise at rates of <3 mm/year thereafter allowed sediment and wetland vegetation to aggrade  and stratigraphic evidence of coseismic subsidence to be preserved [49, 77]. Up to ten instances of coseismic subsidence over the last ∼5.3 ka are recorded in the coastal stratigraphy at John’s River and Willapa Bay (Figs. 1b and 5e).
Preservation of Earthquake Stratigraphy in Far-Field Regions
Subduction zone coastlines in far-field regions that have recorded an RSL highstand during the Holocene, such as the Sunda, Chile, and Japan subduction zones, preserve only geographically limited, discontinuous stratigraphic records of coseismic subsidence or uplift (Figs. 1c–e, 7a–f, and 8a–h) [8, 14, 15, 21, 53, 54, 55, 56, 80, 81]. The RSL history at far-field sites has generally restricted stratigraphic records of coseismic subsidence or uplift to brief time windows of preservation during (1) periods of slow (<1.5 mm/year) RSL rise prior to the mid-Holocene highstand (6–3 ka), which created accommodation space, or (2) the last 1–3 ka when RSL was within ∼2 m of modern sea level, and therefore stratigraphic sequences have remained within the tidal frame (Great Diurnal Range is between 1 and 2 m at all our far-field sites) .
Sunda Subduction Zone
Chile Subduction Zone
In southern Chile, coastal stratigraphic records of past earthquakes are limited to the last 2 ka (Fig. 8a) [53, 87, 88]. At the Maullin estuary, Cisternas et al.  documented a ∼2-ka-long record of seven buried soil contacts marking repeated coseismic subsidence (Figs. 1d and 8e). At Chiloe Island, Garrett et al.  found a ∼1-ka-long record of coseismic subsidence that correlated with the four youngest subsidence events at Maullin (Fig. 1d). The stratigraphic sections at both sites in southern Chile are <1.5 m thick and preserved within the ∼2-m modern tidal range. However, Garrett et al.  noted the occurrence of high marsh soils at elevations below modern mean sea level, implying RSL rise during the last 1 ka. This is in contrast to the late Holocene emergence noted in studies to the north [14, 53, 55, 87, 88, 89] and in our GIA predictions. Garrett et al.  suggest that net tectonic subsidence of Chiloe Island, or glacio-isostatic subsidence related to the Patagonian Ice Sheet and not considered in current GIA models, may have caused RSL rise during the last 1 ka.
Japan Subduction Zone
At the Sendai Plain, the Tokyo metropolitan area, and eastern Hokkaido, Japan, GIA models predict high rates (>5 mm/year) of RSL rise from ∼12 to ∼7 ka, followed by slowing rates of rise (<1 mm/year) to an RSL highstand of ∼3 m between 6 and 4 ka, and then gradual (<1 mm/year) RSL fall until present (Fig. 8b–d). The prograding beach ridge sequences that formed during late Holocene RSL fall in the Sendai Plain contain limited evidence of coseismic vertical deformation (Fig. 1e and 8b) . Sawai et al.  documented evidence of coseismic subsidence associated with two tsunami sand beds (dated to ∼1.5 and ∼1 ka) preserved within the stratigraphy of a sheltered lowland in Odaka (Fig. 8f). Further south, in the Tokyo metropolitan area, stratigraphic records of coseismic vertical deformation are limited to the last ∼1 ka (Fig. 8c). At the Miura Peninsula, Shimazaki et al.  found evidence of three instances of coseismic uplift that abruptly changed a shallow subtidal bay environment into a tidal flat (Fig. 8g). The three uplift events occurred during three historical Kanto earthquakes in AD 1923, 1703, and 1060–1400. Similar to southern Chile, the stratigraphic sections in the Sendai Plain and the Tokyo Metropolitan area are within the modern tidal range (∼1.5 m).
Stratigraphy beneath multiple coastal wetlands fringing estuaries in eastern Hokkaido, northern Japan, contains evidence of up to six postseismic uplift events in the last 3 ka (Figs. 1e and 8h [15, 81, 91, 92]. Each uplift event is marked by a change between a tidal flat mud and a freshwater marsh, signaling at least 1 m of slow, postseismic uplift following earthquakes larger than any in the region’s written history [91, 93]. The tide gauge records in eastern Hokkaido, northern Japan, display submergence of 8–10 mm/year over the last ∼100 years . This pronounced subsidence over the twentieth century has been ascribed to interseismic strain accumulation that has not been reversed by recent earthquakes . Infrequent postseismic uplift may help reconcile eastern Hokkaido’s twentieth century submergence with its stratigraphic evidence of long-term RSL fall during the Holocene [81, 93].
At subduction zone sites in near-field regions (i.e., western Vancouver Island), RSL fall over the last ∼6 ka has limited stratigraphic evidence of past earthquakes to the last 1 ka.
At subduction zone sites in intermediate-field regions (i.e., southern Washington and central Oregon), gradual RSL rise over the last 7–5 ka has produced widespread coastal wetland stratigraphy that contains the longest (>5 ka) documented stratigraphic records of repeated coseismic subsidence. Subtle differences in the onset of gradual (<3 mm/year) RSL rise influence the length of stratigraphic records of coseismic subsidence in central Oregon (∼7 ka) and southern Washington (∼5 ka).
At subduction zone sites in far-field regions such as Sumatra and central Chile, gradual (<1.5 mm/year) RSL rise leading up to the mid-Holocene highstand (6–3 ka) provided the accommodation space necessary for coastal wetlands to form and for prehistoric earthquake evidence to be preserved. In southern Chile and Japan, prehistoric earthquake records are limited to the last 1–3 ka, when RSL was within 2 m of modern sea level, and thus within the tidal frame, preserving geographically limited coastal wetlands despite slowly falling RSL. Subtle differences in RSL influence the length of these far-field records.
This work was supported by funding from National Science Foundation awards to TD (EAR-439021), SEE (EAR-1419844), BPH (EAR-1357756, 1419824, 0809392), and REK (ARC-1203415). TD would like to thank Antonio Osa for his unwavering support and encouragement (AMQPLV). We thank Alan Nelson and Harvey Kelsey for constructive reviews that improved the manuscript. This paper is a contribution to IGCP project 639 and PALSEA2. MV contributes to the the A*MIDEX project (n° ANR-11-IDEX-0001-02). On behalf of all authors, the corresponding author states that there is no conflict of interest.
- 5.Horton BP, Engelhart SE, Kemp AC, Sawai Y. 14.25 Microfossils in tidal settings as indicators of sea-level change, paleoearthquakes, tsunamis, and tropical cyclones A2—Shroder, John F. In: Treatise on geomorphology [Internet]. San Diego: Academic Press; 2013. p. 292–314. Available from: http://www.sciencedirect.com/science/article/pii/B9780123747396003948.
- 8.Dura T, Rubin CM, Kelsey HM, Horton BP, Hawkes A, Vane CH, et al. Stratigraphic record of Holocene coseismic subsidence, Padang, West Sumatra. J Geophys Res Solid Earth. 2011; 116(B11).Google Scholar
- 10.Kemp AC, Horton BP, Engelhart SE. SEA-LEVELS, LATE QUATERNARY | Late quaternary relative sea-level changes at mid-latitudes A2—Elias, Scott A. In: Mock CJ, editor. Encyclopedia of quaternary science (second edition) [Internet]. Amsterdam: Elsevier; 2013. p. 489–94. Available from: http://www.sciencedirect.com/science/article/pii/B9780444536433001400
- 15.Kelsey H, Satake K, Sawai Y, Sherrod B, Shimokawa K, Shishikura M. Recurrence of postseismic coastal uplift, Kuril subduction zone, Japan. Geophys Res Lett. 2006;33(13).Google Scholar
- 17.Sawai Y, Namegaya Y, Okamura Y, Satake K, Shishikura M. Challenges of anticipating the 2011 Tohoku earthquake and tsunami using coastal geology. Geophys Res Lett. 2012; 39(21).Google Scholar
- 37.Pluet J, Pirazzoli PA. World atlas of Holocene sea-level changes. Vol. 58. Elsevier; 1991.Google Scholar
- 41.Kirwan ML, Guntenspergen GR, D’Alpaos A, Morris JT, Mudd SM, Temmerman S. Limits on the adaptability of coastal marshes to rising sea level. Geophys Res Lett. 2010;37(23).Google Scholar
- 42.Pratolongo PD, Kirby JR, Plater A, Brinson MM. Temperate coastal wetlands: morphology, sediment processes, and plant communities. Coast Wetl Integr Ecosyst Approach 2009; 975.Google Scholar
- 43.Atwater BF, Hemphill-Haley E. Recurrence intervals for great earthquakes of the past 3,500 years at northeastern Willapa Bay, Washington. USGPO; Information Services [distributor]; 1997.Google Scholar
- 52.Nelson AR. SEA-LEVELS, LATE QUATERNARY | Tectonics and relative sea-level change A2—Elias, Scott A. In: Mock CJ, editor. Encyclopedia of quaternary science (second edition) [Internet]. Amsterdam: Elsevier; 2013. p. 503–19. Available from: http://www.sciencedirect.com/science/article/pii/B9780444536433001412
- 57.Shennan I, Long AJ, Horton BP. Handbook of sea-level research. John Wiley & Sons; 2015.Google Scholar
- 59.Engelhart SE, Vacchi M, Horton BP, Nelson AR, Kopp RE. A sea-level database for the Pacific coast of central North America. Megathrust Earth Sea-Level Chang Tribute George Plafker. 2015;113:78–92.Google Scholar
- 63.Heaton TH, Kanamori H. Seismic potential associated with subduction in the northwestern United States. Bull Seismol Soc Am. 1984;74(3):933–41.Google Scholar
- 64.McCrory PA, Blair JL, Waldhauser F, Oppenheimer DH. Juan de Fuca slab geometry and its relation to Wadati-Benioff zone seismicity. J Geophys Res Solid Earth. 2012; 117(B9).Google Scholar
- 65.Hutchinson I. Holocene sea level change in the Pacific Northwest: a catalogue of radiocarbon dates and an atlas of regional sea-level curves. Institute of Quaternary Research, Simon Fraser University; 1992.Google Scholar
- 68.Dallimore A, Enkin RJ, Pienitz R, Southon JR, Baker J, Wright CA, et al. Postglacial evolution of a Pacific coastal fjord in British Columbia, Canada: interactions of sea-level change, crustal response, and environmental fluctuations-results from MONA core MD02-2494. This article is one of a series of papers published in this special issue on the theme Polar Climate Stability Network. Can J Earth Sci. 2008;45(11):1345–62.CrossRefGoogle Scholar
- 74.Darienzo ME, Peterson CD, Clough C. Stratigraphic evidence for great subduction-zone earthquakes at four estuaries in northern Oregon, USA. J Coast Res. 1994;850–76.Google Scholar
- 76.Peterson CD. Holocene sedimentary framework of Grays Harbor basin, Washington. 1992.Google Scholar
- 77.Atwater BF, Tuttle MP, Schweig ES, Rubin CM, Yamaguchi DK, Hemphill-Haley E. Earthquake recurrence inferred from paleoseismology. Quat Period U S. 2003; 331–50.Google Scholar
- 82.United Kingdom Hydrographic Office. Admiralty Tide Tables: Pacific Ocean. Vol. 4. United Kingdom Hydrographic Office; 2016.Google Scholar
- 85.Briggs RW, Sieh K, Amidon WH, Galetzka J, Prayudi D, Suprihanto I, et al. Persistent elastic behavior above a megathrust rupture patch: Nias island, West Sumatra. J Geophys Res Solid Earth. 2008; 113(B12).Google Scholar
- 86.Meltzner AJ, Sieh K, Chiang H-W, Shen C-C, Suwargadi BW, Natawidjaja DH, et al. Coral evidence for earthquake recurrence and an AD 1390–1455 cluster at the south end of the 2004 Aceh–Andaman rupture. J Geophys Res Solid Earth. 2010; 115(B10).Google Scholar
- 88.Bartsch-Winkler S, Schmoll HR. Evidence for Late Holocene relative sea-level fall from reconnaissance stratigraphical studies in an area of earthquake—Southern Chile. Tecton Controls Signat Sediment Successions Spec Publ 20 IAS. 2009;40:91.Google Scholar
- 89.Hong I, Dura T, Ely LL, Horton BP, Nelson AR, Cisternas M, et al. A 600-year-long stratigraphic record of tsunamis in south-central Chile. Holocene. 2016; 0959683616646191.Google Scholar
- 90.Shimazaki K, Kim HY, Chiba T, Satake K. Geological evidence of recurrent great Kanto earthquakes at the Miura Peninsula, Japan. J Geophys Res Solid Earth. 2011; 116(B12).Google Scholar
- 95.Dyke AS. An outline of North American deglaciation with emphasis on central and northern Canada. Quat Glaciations: Extent Chronol. 2004;2:373–424.Google Scholar