Harvey M. Kelsey

Summary of Coastal Geologic Evidence for Past Great Earthquakes at the Cascadia Subduction Zone

Earthquakes in the past few thousand years have left signs of land-level change, tsunamis, and shaking along the Pacific coast at the Cascadia subduction zone. Sudden lowering of land accounts for many of the buried marsh and forest soils at estuaries between southern British Columbia and northern California. Sand layers on some of these soils imply that tsunamis were triggered by some of the events that lowered the land. Liquefaction features show that inland shaking accompanied sudden coastal subsidence at the Washington-Oregon border about 300 years ago. The combined evidence for subsidence, tsunamis, and shaking shows that earthquakes of magnitude 8 or larger have occurred on the boundary between the overriding North America plate and the downgoing Juan de Fuca and Gorda plates. Intervals between the earthquakes are poorly known because of uncertainties about the number and ages of the earthquakes. Current estimates for individual intervals at specific coastal sites range from a few centuries to about one thousand years.

Tsunami history of an Oregon coastal lake reveals a 4600 yr record of great earthquakes on the Cascadia subduction zone

Bradley Lake, on the southern Oregon coastal plain, records local tsunamis and seismic shaking on the Cascadia subduction zone over the last 7000 yr. Thirteen marine incursions delivered landward-thinning sheets of sand to the lake from nearshore, beach, and dune environments to the west. Following each incursion, a slug of marine water near the bottom of the freshwater lake instigated a few-year-to-several-decade period of a brackish (≤4‰ salinity) lake. Four additional disturbances without marine incursions destabilized sideslopes and bottom sediment, producing a suspension deposit that blanketed the lake bottom. Considering the magnitude and duration of the disturbances necessary to produce Bradley Lake’s marine incursions, a local tsunami generated by a great earthquake on the Cascadia subduction zone is the only accountable mechanism. Extreme ocean levels must have been at least 5–8 m above sea level, and the cumulative duration of each marine incursion must have been at least 10 min. Disturbances without marine incursions require seismic shaking as well. Over the 4600 yr period when Bradley Lake was an optimum tsunami recorder, tsunamis from Cascadia plate-boundary earthquakes came in clusters. Between 4600 and 2800 cal yr B.P., tsunamis occurred at the average frequency of ~3–4 every 1000 yr. Then, starting ~2800 cal yr B.P., there was a 930–1260 yr interval with no tsunamis. That gap was followed by a ~1000 yr period with 4 tsunamis. In the last millennium, a 670–750 yr gap preceded the A.D. 1700 earthquake and tsunami. The A.D. 1700 earthquake may be the fi rst of a new cluster of plate-boundary earthquakes and accompanying tsunamis. Local tsunamis entered Bradley Lake an average of every 390 yr, whereas the portion of the Cascadia plate boundary that underlies Bradley Lake ruptured in a great earthquake less frequently, about once every 500 yr. Therefore, the entire length of the subduction zone does not rupture in every earthquake, and Bradley Lake has recorded earthquakes caused by rupture along the entire length of the Cascadia plate boundary as well as earthquakes caused by rupture of shorter segments of the boundary. The tsunami record from Bradley Lake indicates that at times, most recently ~1700 yr B.P., overlapping or adjoining segments of the Cascadia plate boundary ruptured within decades of each other.

Great Cascadia earthquakes and tsunamis of the past 6700 years, Coquille River estuary, southern coastal Oregon

Cascadia subduction zone earthquakes dropped tidal marshes and low-lying forests to tidal flat elevations 12 times in the last 6700 cal yr B.P. at the Coquille River estuary in southwestern Oregon. The youngest buried soil, preserved in tidal marsh deposits near the estuary mouth, records the A.D. 1700 earthquake that ruptured the entire Cascadia margin. Eleven other buried marsh and upland soils found in tributary valleys of the estuary provide repeated evidence for rapid, lasting relative sea-level rise interpreted as coseismic subsidence. Additional stratigraphic criteria supporting a coseismic origin for soil burial include: lateral soil correlation over hundreds of meters, fossil diatom assemblages that indicate a maximum of 1.2‐3.0 m of submergence, and sand deposits overlying buried soils consistent with earthquakeinduced tsunamis that traveled 10 km up the estuary. Twelve earthquakes occurred in the last 6500‐6720 cal yr B.P., recurring on average every 570‐590 yr. Intervals between earthquakes varied from a few hundred years to over 1000. Comparisons of the Coquille record to earthquake histories from adjacent sites in Oregon, southwestern Washington, and northwestern California suggest that at least two earthquakes in the last 4000 yr did not rupture the entire length of the subduction zone. An earthquake 760‐1140 cal yr B.P. in southwestern Washington may have ruptured as far south as Coos Bay but probably stopped before it reached the Coquille estuary because no buried soil records the event, and tidal marsh conditions were set to record an earthquake. An earthquake limited to a southern segment of the Cascadia margin 1940‐2130 cal yr B.P. probably did not rupture north of the Coquille estuary. An analysis of relative sea-level histories from either side of the Coquille fault failed to find conclusive evidence for late Holocene vertical deformation. However, we cannot preclude recent upper-plate faulting. If the fault is active, as geomorphic features suggest, then constraints on the highest possible elevation of mean tide level allow a maximum vertical slip rate of 0.2‐ 0.4 mm/yr in the past 6200‐6310 cal yr B.P.

Strain Partitioning between structural domains in the forearc of the Hikurangi Subduction Zone, New Zealand

The Pacific plate obliquely converges with the Australian plate at latitude 39°50′S along the Hikurangi margin off the east coast of the North Island of New Zealand. An extensive and youthful subaerially exposed forearc on the east coast of the North Island in the Hawkes Bay area provides the opportunity to document contemporaneous forearc deformation in this obliquely convergent margin setting. Geologic mapping and analysis of strain at both mesoscale and megascale indicates that strain is partitioning into domains of extension, contraction, and strike-slip. The domains are elongate and trend parallel to the margin. Measurements of net shortening and transcurrent slip in the forearc show that the obliquely convergent motion is transferred across the plate interface. Deformation rates calculated for the past 1–2 m.y. for structures in all six forearc domains account for 50–70% of the margin-parallel motion required by Pacific-Australian plate convergence and about 6% of the plate motion perpendicular to the plate boundary. At the surface in the forearc, this obliquely convergent motion is manifest not by transpressional faults but rather by paired structural domains that consist of a strike-slip fault zone and an accompanying contractional fault-and-fold zone on the trenchward side. There are two such transcurrent faulting-and-contraction couplets, one where the backstop daylights at the arcward edge of the forearc and another couplet trenchward of a relatively undisturbed forearc basin. The small amount of shortening, relative to strike-slip, in the onshore part of the forearc suggests that shortening perpendicular to the plate boundary may be concentrated offshore and that most of the component of plate motion perpendicular to the plate boundary may be accommodated by slip along the subduction zone megathrust.

Plate-boundary earthquakes and tsunamis of the past 5500 yr, Sixes River estuary, southern Oregon

Eleven plate-boundary earthquakes over the past 5500 yr have left a stratigraphic signature in coastal wetland sediments at the lower Sixes River valley in south coastal Oregon. Within a 1.8 km2 abandoned meander valley, 10 buried wetland soils record gradual and abrupt relative sea-level changes back in time to ;6000 yr ago. An additional, youngest buried soil at the mouth of the Sixes River subsided during the A.D. 1700 Cascadia earthquake. Multiple lines of evidence indicate that tectonic subsidence caused soil burial, including permanent relative sea-level rise following burial, lateral continuity of buried soil horizons over hundreds of meters, diatom assemblages showing that sea level rose abruptly at least 0.5 m, and sand deposits on top of buried soils demonstrating coincidence of coseismic subsidence and tsunami inundation. For at least two of the buried soils, liquefaction of sediment accompanied subsidence. The 11 soil-burial events took place between 300 and ;5400 yr ago, yielding an average recurrence interval of plateboundary earthquakes of ;510 yr. Comparing paleoseismic sites in southern Washington and south coastal Oregon with the Sixes River site for the past 3500 yr indicates that the number and timing of recorded plate-boundary earthquakes are not the same at all sites. In particular, a Sixes earthquake at ;2000 yr ago lacks a likely correlative in southern Washington. Therefore, unlike the A.D. 1700 Cascadia earthquake, some Cascadia plate-boundary earthquakes do not rupture the entire subduction zone from southern Oregon to southern Washington. In the lower Sixes River valley, the upperplate Cape Blanco anticline deforms sediment of late Pleistocene and Holocene age directly above the subduction zone. Differential tectonic subsidence occurred during two of the plate-boundary earthquakes when a blind, upper-plate reverse fault, for which the Cape Blanco anticline is the surface fold, slipped coseismically with rupture of the plate boundary. During these two earthquakes, sites ;2 km from the anticline axis subsided ;0.5 m more than sites ;1 km from the axis.

Holocene fault scarps near Tacoma, Washington, USA

Airborne laser mapping confirms that Holocene active faults traverse the Puget Sound metropolitan area, northwestern continental United States. The mapping, which detects forest-floor relief of as little as 15 cm, reveals scarps along geophysical lineaments that separate areas of Holocene uplift and subsidence. Along one such line of scarps, we found that a fault warped the ground surface between A.D. 770 and 1160. This reverse fault, which projects through Tacoma, Washington, bounds the southern and western sides of the Seattle uplift. The northern flank of the Seattle uplift is bounded by a reverse fault beneath Seattle that broke in A.D. 900–930. Observations of tectonic scarps along the Tacoma fault demonstrate that active faulting with associated surface rupture and ground motions pose a significant hazard in the Puget Sound region.

Late Holocene earthquakes on the Toe Jam Hill fault, Seattle fault zone, Bainbridge Island, Washington

Five trenches across a Holocene fault scarp yield the first radiocarbon-measured earthquake recurrence intervals for a crustal fault in western Washington. The scarp, the first to be revealed by laser imagery, marks the Toe Jam Hill fault, a north-dipping backthrust to the Seattle fault. Folded and faulted strata, liquefaction features, and forest soil A horizons buried by hanging-wall-collapse colluvium record three, or possibly four, earthquakes between 2500 and 1000 yr ago. The most recent earthquake is probably the 1050–1020 cal. (calibrated) yr B.P. (A.D. 900–930) earthquake that raised marine terraces and triggered a tsunami in Puget Sound. Vertical deformation estimated from stratigraphic and surface offsets at trench sites suggests late Holocene earthquake magnitudes near M7, corresponding to surface ruptures >36 km long. Deformation features recording poorly understood latest Pleistocene earthquakes suggest that they were smaller than late Holocene earthquakes. Postglacial earthquake recurrence intervals based on 97 radiocarbon ages, most on detrital charcoal, range from ∼12,000 yr to as little as a century or less; corresponding fault-slip rates are 0.2 mm/yr for the past 16,000 yr and 2 mm/yr for the past 2500 yr. Because the Toe Jam Hill fault is a backthrust to the Seattle fault, it may not have ruptured during every earthquake on the Seattle fault. But the earthquake history of the Toe Jam Hill fault is at least a partial proxy for the history of the rest of the Seattle fault zone.

Topographic form of the Coast Ranges of the Cascadia Margin in relation to coastal uplift rates and plate subduction

The Coast Ranges of the Cascadia margin are overriding the subducted Juan de Fuca/Gorda plate. We investigate the extent to which the latitudinal trend in average topography of the Coast Ranges is a function of the latitudinal change in attributes related to the subduction process. These attributes include the variable age of the subducted slab that underlies the Coast Ranges and average vertical crustal velocities of the western margin of the Coast Ranges for two markedly different time periods, the last 45 years and the last 100 kyr. These vertical crustal velocities are computed from the resurveying of highway bench marks and from the present elevation of shore platforms that have been uplifted in the late Quaternary, respectively. Topography of the Coast Ranges is in part a function of the age and buoyancy of the underlying subducted plate. This is evident in the fact that the two highest topographic elements of the Coast Ranges, the Klamath Mountains and the Olympic Mountains, are underlain by youngest subducted oceanic crust. The subducted Blanco Fracture Zone in southernmost Oregon is responsible for an age discontinuity of subducted crust under the Klamath Mountains. The northern terminus of the topographically higher Klamaths is offset to the north relative to the position of the underlying Blanco Fracture Zone, the offset being in the direction of migration of the fracture zone, as dictated by relative plate motions. Vertical crustal velocities at the coast, derived from bench mark surveys, are as much as an order of magnitude greater than vertical crustal velocities derived from uplifted shore platforms. This uplift rate discrepancy indicates that strain is accumulating on the plate margin, to be released during the next interplate earthquake. In a latitudinal sense, average Coast Range topography is relatively high where bench mark-derived, short-term vertical crustal velocites are highest. Because the shore platform vertical crustal velocites reflect longer-term, permanent uplift, we infer that a small percentage of the interseismic strain that accumulates as rapid short-term uplift is not recovered by subduction earthquakes but rather contributes to rock uplift of the Coast Ranges. The conjecture that permanent rock uplift is related to interseismic uplift is consistent with the observation that those segments of the subduction zone subject to greater interseismic uplift rates are at approximately the same latitudes as those segments of the Coast Ranges that have higher magnitudes of rock uplift over the long term.

Structural evolution along the inner forearc of the obliquely convergent Hikurangi margin, New Zealand

The accretionary margin of the Hikurangi forearc on the southeast coast of the North Island of New Zealand is part of the leading edge of the Australian plate, which is overriding the obliquely converging Pacific plate. We investigate the last 10 m.y. of deformation history of the innermost (western) quarter of the total width of the forearc through analysis of the sedimentologic and structural evolution of the Eketahuna area on the east coast of the North Island. The Eketahuna area is ideal for such a study because emergence of the margin in the Quaternary has exposed a complete late Neogene rock record. This record has allowed us to chronicle the strain history. From 10 Ma to about 2.5 Ma this forearc region was the locus of subsidence and marine deposition. In the latest Pliocene this part of the margin began to shorten through folding and reverse faulting, bringing an end to basin filling. The period of shortening was brief, and by the late Pleistocene, reverse faulting had ceased and was immediately succeeded by dextral strike-slip faulting, in some cases along the same faults. Presently, the dominant strain regime in the inner quarter of the forearc is strike-slip faulting. This structural history illustrates that, over time, the pattern of strain partitioning has changed in the Hikurangi forearc. The switch from crustal shortening to dextral shear along the major faults in this area in the last 1 m.y. may be a response to more than 10° of clockwise rotation in the southern Hikurangi forearc in Pliocene and Pleistocene time. This rotation is a consequence of the fact that the accretionary margin is undergoing continuous deformation between the obliquely converging Australian and Pacific plates in this area at the southernmost end of the Kermadec-Hikurangi subduction system. The inboard portion of this young accretionary margin is exceptionally well exposed today, probably in part because of the late Neogene subduction of relatively thick, buoyant crust of the Hikurangi-Chatham plateau.

Formation of inner gorges

Abstract Under certain conditions, a distinctive type of landform, the inner gorge, evolves in drainage basins undergoing persistent base level lowering. Inner gorges are typical of basins in north coastal California, U.S.A., and one such basin was selected for detailed study. Based on analysis of along-channel distribution of rock types, streamside land-sliding, and stored sediment, as well as computation of stream power, the following set of conditions appears to accompany the formation of an inner gorge: 1. inner gorge parent materials are relatively competent, homogenous rock types; 2. in inner gorge reaches, there is an absence of any base level control that would prevent downcutting; 3. continued downcutting occurs due to continued base level fall, and this fall is most likely induced by tectonic uplift; 4. mass slope failures that form steep straight slopes (debris slides) are the main denudational hillslope process in the inner gorge; 5. sufficient stream power is available to transport landslide-derived material out of the inner gorge reach such that alluvial sediment does accumulate in the reach over time; and 6. the initial condition of the landscape is a low relief, erosionally mature land surface into which the inner gorge is cut. This initial condition implies that inner gorges develop in tectonically rejuvenated landscapes. The processes of denudation on inner gorge slopes, acting over a time scale of 105 years, result in a style of landform evolution that approximates parallel slope retreat. This type of slope retreat maintains the distinct inner gorge morphology during sustained downcutting. Though the analysis is confined to one basin in northern California, so as to provide a detailed framework to understand inner gorge landform, inner gorges appear to be commonplace in areas of moderate to high uplift where lithologic and tectonic conditions are appropriate for their formation. A global search for this landform, yet to be undertaken, is needed to confirm whether it indeed occurs where the above conditions are met.