Late Holocene Geomorphology of the Columbia River Estuary , Oregon and Washington , USA

Abundant river sediment supply and an open-water central bay area characterize the geomorphology of the large Columbia River estuary (~ 100 km in length). Lateral floodplains and marsh islands do constrict the uppermost reaches of the estuary, but the central axes of the lower estuary are dominated by shallow sand shoals (0–4 m water depth). A total of 58 vibracores are used to document the grain size and age (0–2,500 CyrBP) of late Holocene deposits in the estuary. Sedimentation rates in stable floodplains (1.1 m ka) reflect rates of relative sea level rise (0.75 m ka). Sedimentation rates of muddy sand accretionary banks and prehistoric sand shoals (1.5–7 m ka) greatly exceed coeval rates of sea level rise, so they must represent short–term rates of vertical accretion resulting from channel lateral migration and associated cut and fill processes. The apparent paradox of unfilled accommodation space in the estuary is resolved by 1) winter wind–wave erosion of sand shoals to -3 m NAVD88 elevation and 2) asymmetric fluvial-tidal advection that results in net seaward transport of bed load in shallow tidal channels (> – 10 m NAVD88) and shallow subtidal shoals (> – 4 m NAVD88) during spring river flooding.


Introduction
The extensive Columbia River estuary (~ 100 km in length) provides a geomorphic paradox.The submerged incised river valley is abundantly supplied with river sediment (10-15 million tons per year) (Sherwood & Creager, 1990).The river mouth lacks a delta and the shallow estuary has not completely in-filled with tidal marshes or floodplains (Figure 1).The lack of a deep central bay in the Columbia River estuary is consistent with a high throughput of bed load (Gates, 1994;Baker et al., 2010) during late Holocene time.However, latest Holocene infilling of the central bay did not reach completion, as demonstrated by 1) the lack of substantial lateral floodplain development and 2) the presence of an open-water central bay that extends 30 km upriver of the tidal inlet.The open bay reaches of the estuary contrast with extensive flood plain constriction in the upriver sections of the lower Columbia River valley, which extends another 100 km upriver to the modern head of tide in the Cascade Range Gorge (Peterson et al., 2014).
The expanse of open water in the central bay of the Columbia River estuary (Figure 1) appears to imply the availability of accommodation space (Dalrymple et al., 1994) for new sediment accumulation in the estuary.These conditions, therefore, likely encouraged interpretations that the lower estuary serves as a net sink for littoral sand entering the river mouth in deep channels (Walter, 1980;Sherwood & Creager, 1990).However, long-term declines in sedimentation rates from mid-Holocene to late-Holocene time in the lower Columbia River valley (Peterson et al., 2013) demonstrate substantial bypassing of river sand (about 5 million cubic meters per year) to the Pacific Ocean.The Holocene depositional filling of the inner-shelf (Twichell et al., 2010) and the late Holocene progradation of barrier beach plains on either sides of the Columbia River mouth (Vanderburgh et al., 2010) are attributed to the substantial throughput of bed load from the Columbia River to the marine side (Peterson et al., 2010).The estuary (boxed, 100 km in length) occurs in the lower Columbia River valley, which serves as the border between Washington and Oregon States.The antecedent river valley is tidally influenced from the mouth at Astoria, OR to the present head of tide at Bonneville, WA, totaling a river distance of 225 km.
The goals of this paper are to reconcile the apparent opposing prehistoric conditions of 1) high supply rates of Columbia River sediment to the Columbia River estuary with 2) a general lack of intertidal marsh islands or laterally accreting supratidal floodplains in the central axes of the lower estuary (Figure 1).A total of 58 vibracores (4-6 m depth) are used to establish the deposit grain sizes and sedimentation rates in the estuary during latest Holocene time (0-2.5 ka).Early historic bathymetric surveys of the estuary are utilized together with wind-wave fetch to calculate bottom orbital velocities in the central axes of the middle and lower estuarine reaches.Early historic conditions of tidal, fluvial, and wind-wave forcing in the estuary are used to develop a conceptual model of bed load bypassing through the estuary to the Pacific Ocean in latest Holocene time.The unusual geomorphic development of the Columbia River estuary is relevant to regional sediment management in the bi-state region (Gelfenbaum & Kaminsky, 2010), and more broadly, to the modeling of bed load dynamics in shallow estuaries that are influenced by 1) high-rates of river sediment supply, 2) fluvial-tidal flow regimes, and 3) very-high wind-wave energies.

Background
As a major shipping route the lower Columbia River (Figure 1) has been systematically surveyed for bathymetry since at least 1839 (Hickson & Rudolf, 1951).Historic bathymetric change maps were compiled by CREDDP (1983) to establish potential erosion or accretion following 1) wide-scale impoundments of the Columbia River and its tributaries (1937-present), 2) long-term impacts of repeated jetty construction in the Columbia River mouth (1905-present), and 3) extensive deepening and channel bank stabilization of the Columbia River ship channel (1903-present) (Simenstad et al., 1990;Sherwood et al., 1990;Byrnes & Li, 1998).The CREDEP bathymetric change maps (CREDDP, 1983) were utilized in this research to select core site locations that were not directly impacted by channel dredging and/or stabilization.Most recently, light detection and ranging radar (LiDAR) has been used to map the intertidal deposit elevations in the Columbia River estuary (Pugetsoundlidar, 2013).The LiDAR data were used in this study to establish core site elevations relative to the NAVD88 elevation datum.
Bathymetric and hydrographic surveys near the mouth of the Columbia River (Hickson & Rudolf, 1951) demonstrate the presence of a shallow central bay since earliest-historic time in 1792 (Figure 2).North and south channels (~ 8-15 m water depth) dissected shallow sand shoals (< 5 m depth) that extended to within several kilometers of the tidal inlet.The main channel at the Columbia River mouth stabilized since the onset of jetty construction, beginning at the turn of the last century (circa 1902).During early historic time (1792-1885) the north and south tidal channels generally decreased in depth from the mouth (10-15 m depth) the south channel at the Port of Astoria (to 8-12 m depth), a distance of about 20 km from the mouth.The narrow south channel (~ 500 m width) was artificially deepened during the 1900s to enable the passage of large ships to upriver ports at the Port of Portland, OR (Figure 1). Figure 2. Columbia river mouth geomorphology  Representative channel depths (m) below sea level and extent of shallow sand shoals including inshore shoals (< 4 m depth) and offshore shoals (< 6, 8 or 9 m depth) are redrafted from Hickson and Rudolf (1951).Jetty construction to stabilize the mouth of the Columbia River at the south spit was initiated (bold line) at the turn of the last century (1902).
Early measurements of riverine-tidal discharge near the Columbia River mouth are presented in Table 1 for representative seasonal conditions of flood and ebb tides (Hickson & Rudolf, 1951).As expected, the early historic river discharges, which existed prior to the numerous impoundments of the Columbia River and its tributaries (Simonstead et al., 1990), exerted a significant asymmetry in tidal flux on a seasonal basis.Combined fluvial-ebb tide discharge exceeded flood tide exchange by 29.7-120.7 percent in the spring months of Columbia River flood.Under maximum river flood conditions, following late-spring snow melt runoff in the eastern tributaries, the fluvial-ebb tide discharge exceeded flood tide discharge by 356 percent.Salt water in the lower estuary generally extends 10-13 km upriver from the mouth in the north and south channels (Figure 2), but it is displaced towards the mouth during spring periods of river flood discharge (CREDDP, 1983).Early historic peak floods (> 35,000 m 3 s -1 discharge) have been reduced (< 20,000 m 3 s -1 discharge) and shifted into later summer periods by the numerous impoundments (> 130 in number) in the Columbia River tributaries.The impoundments of the Columbia River and its tributaries were constructed after 1937 (Sherwood et al., 1990).Fox et al. (1984) summarized modern geomorphologic features in the Columbia River estuary to include major and minor tidal channels, and intertidal shoals and marshes.Surface sediments in the Columbia River estuary have been analyzed for grain-size distribution (Sherwood & Creager, 1990).The subtidal flats and deeper channels are floored by sand.Mud in the modern intertidal deposits is restricted to the protected bays including Baker Bay, Youngs Bay, Grays Bay and Cathlamet Bay (Figure 1).Asymmetric bed forms were imaged on a seasonal basis by side-scan sonar in the estuary during both flood and ebb tidal cycles (Sherwood & Creager, 1990).A predominance of upriver flow orientations demonstrated landward directed salt wedge or gravity flow in the major channel thalwegs (15-20 m depth) in the lowest estuarine reaches.The landward oriented sand waves in the deepest channels (> 15 m water depth) extended 1-13 km upriver of the mouth during fall-winter, but only 1-5 km upriver of the mouth during spring river floods.Seaward-directed asymmetric bed forms dominate the middle reaches of the estuary (13-25 km upriver of the mouth) during all tidal stages throughout the year.Notes: The original data from the mouth of the Columbia River (1933), as provided in cubic feet per second (Hickson and Rudolf, 1951), are converted to cubic meters per second.The total number of tidal cycles averaged for each month in 1933 was April (total 8), May (total 13), and September (total 7).The maximum flood discharge during the May (1933) sampling period is shown under MAX Flood, as reported by Hickson and Rufolf (1951).Percent discharge difference is taken from the ratio of the discharge difference divided by the flood tide and normalized to percent.The maximum flood tide difference is taken from the ratio of maximum discharge difference divided by the May seasonal flood tide discharge and normalized to percent.
Sedimentation rates in the Columbia River estuary during latest Holocene time are established in this study from dated core intervals.Radiocarbon analysis of buried organic materials is supplemented by several methods of relative dating including 1) tephrochonology (Peterson et al., 2012a), 2) cosesimic subsidence (Peterson et al., 2012b), and 3) the first occurrence of introduced mollusks.Four tephra layers of importance to the late Holocene deposits in the estuary were produced by local eruptions from Mount St Helens (Figure 1) including the set-W at 0.5 ka, the set-P at 2.5-3.0 ka and the set-Y at 3.5-4.0ka, and from an unknown eruptive source (SH/A) at 1.3 ka (Peterson et al., 2012a).
Great earthquakes in the study region (Atwater, 1987;Peterson et al., 2000) have produced episodic events of abrupt submergence (0.5-2.0 m sea level rise) leading to abruptly buried wetland marshes in the Columbia River estuary and adjacent estuaries.Radiocarbon dating has been used to correlate the buried marsh horizons in the lower Columbia River estuary (Jurney, 2001) to regional coseismic susbsidence events (Peterson et al., 2012b).The regional subsidence events include the 1 st Sub (AD 1700), 2 nd Sub (~ 1.1 ka), 3 rd Sub (~ 1.3 ka), 4 th Sub (~ 1.7 ka) and 5 th Sub (~2.6 ka).Additional microfossil records of these subsidence events in Willapa Bay and Grays Harbor in southwest Washington (Figure 1) coast have been reported by Shennan et al. (1996) and Atwater and Hemphill-Haley (1997).

Methods
Continuous cores (7.5 cm diameter) were taken to 4-6 m depth by vibracoring in shallow subtidal shoals, tidal channel accretionary banks, and in fringing tidal marshes (Figure 3).Core sites (58 in number) were selected on the basis of approximately equal distances between stations in across-estuary transects, that were separated at several kilometer distances along the length of the estuary.Major and minor channels were avoided due to historic dredging of the channels for navigation purposes.Early historic bathymetric charts  of the lower reaches (Figure 2) were used to target pre-jetty shoals and muddy tidal flats for vibracoring near the river mouth.Natural intertidal islands and shallow subtidal sand shoals in the middle and upper reaches were selected for vibracoring, based on the earliest triangulated-survey bathymetric charts (1841) of the Columbia River estuary (Wilkes, 1841).Historic bathymetric change maps (CREDDP, 1983) were used to avoid any sites showing anomalous erosion or accretion, respectively, due to apparent channel dredging and dredge spoil disposal in the estuary.Vibracore sites were geo-referenced using GPS-12 channel WASS assisted real time differential processing (+/-5 m estimated potential error) (Table 2).Core site elevations were estimated on the basis of 1) LiDAR 1/9 th arc sec digital elevation models in supratidal and intertidal settings (Pugetsoundlidar, 2013) and 2) measured water depth relative to predicted tide level in subtidal settings.Core site elevations are reported to 0.1 m NAVD88, the reported accuracy of the 1/9 th arc sec LiDAR, but are assumed to include vertical errors of +/-0.25 m due to uncertainties of the horizontal positions (12 channel RT-GPS 2-5 m horizontal error).The regional datum (0 m NAVD88) is about one meter below mean tidal level (MTL).It is nearly equivalent to mean lower water (MLW) in the semidiurnal mesotidal estuary.
Radiocarbon dating was performed on peat and twig samples to establish basal deposit dates and potential maker bed events, including both tephra layers and buried wetland or subsidence contacts.The radiocarbon samples included 1) wood, as predominantly twigs and small detrital wood fragments mixed with sediments and 2) peat, as felted roots/leaves in peaty mud layers.Radiocarbon samples were submitted to Beta Analytic Inc., Florida, USA, for AMS radiocarbon dating and age calibration analysis.Some of the radiocarbon dates were previously reported to support the tephrochronology correlation analyses and great earthquake subsidence records in the estuary, as outlined above in Section 2. Target tephra layers in the estuary vibracores, including ash or concentrated lapilli, were confirmed by the presence of isotropic glass shards, as observed in picolyte™ mounting medium under petrographic microcroscopy (250 x).Coseismic subsidence contacts were identified by abrupt upcore decreases in the abundance of peat or densely rooted organic layers.Elev.
(m) estuary prior to 1937 (Counts, 1985(Counts, , 1986)).The small robust clams (1-3 cm diameter) live on the surface of freshwater sand and gravel deposits.Their presence in core samples indicates deposits of historic age. 210Pb dating of historic core sections in two core sites establish the approximate first occurrence ages of Corbicula shells in Grays Bay and Youngs Bay, which are lateral embayments in the estuary (Figure 1) (Petersen et al., 2004).The maximum depths of articulated Corbicula shells in the dated cores corresponded to 210 Pb decay curve ages of AD 1910-1920 in the Grays Bay and Youngs Bay tidal flats.

Results
Preliminary analyses of modern satellite imagery and deposit surface elevations of the Columbia River estuary (Figure 3; Table 1) indicate three intertidal and supratidal depositional environments including 1) supratidal floodplains (> 2.5 m elevation NAVD88), 2) intertidal marsh islands (~ 1.0-2.5 m elevation), and 3) intertidal sand shoals and channel margins or accretionary banks (1.0 to -1.0 m elevation).Supratidal floodplains dominate the upper reaches of the estuary.Intertidal marsh islands and sand shoals dominate the middle reaches or central bay region between Tongue Point and Aldrich Point.Intertidal sand shoals and subtidal channels (< -1.0 m elevation) characterize the central axes of the middle and lower reaches of the estuary.Core sites in the estuary are grouped by geographic areas including in down-river order from east to west 1) Clatskanie floodplains (CRCF), 2) Aldrich Point islands (CRAP), 3) Grays Bay (CRGB), Tongue Point Sands (CRTS), Cathlamet Bay (CRCB), Baker Bay (CRBB), Skipanon Sands (CRSK), and Youngs Bay (CRYB) (Figure 3).These geographic areas are briefly characterized by core site surface elevations, dominant sediment grain sizes, and deposit age ranges.Specific islands, sand shoals, and channels in the estuary are individually named elsewhere (USCGS, 1967(USCGS, , 1968)).

Aldrich Point Islands
The Aldrich Point islands are mid-channel vegetated islands (0.8-2.3 m elevation) that occur in the broadening river valley of the upper-middle reaches of the Columbia River estuary (Figure 3; Table 2).The single major river channel in the upper reaches of the Clatskanie flood plains diverges into 2-4 shallower subtidal channels (5-10 m depth) that surround the Aldrich Point islands.The multiple distributary channels and mid-channel islands in the Aldrich Point area resemble the bay head delta settings of Hill and Fitzgerald (1992) and Dalrymple et al. (1994).The Aldrich Point islands demonstrate shallow peaty mud or mud overbank deposits (0.5-2.0 m depth) above muddy sand accretionary bank deposits or channel sand deposits (Figure 5).The shallow overbank deposits are young, reaching only about 0.5 ka in age in sites CRAPI1, CRAP02, and CRAP03.The older lateral flood plain at BlindS reaches at least 1.3 ka in age based on correlated coseismic subsidence events (3 rd Sub at 1.3 ka) (Table 2).The muddy sand accretionary bank deposits (3-5 m thickness) in the Aldrich Point islands demonstrate relatively young ages (0.5-1.3 ka).The oldest channel sand deposit at about 6 m depth or -3.75 m elevation in CRHI1 yields a basal radiocarbon age of 1.4-1.7 ka.A shallower channel sand deposit at -1.75 m elevation in site CRAP01 is tentatively dated to 1.3 ka by a second thin tephra layer (SH/A at 1.3 ka).The Aldrich Point island deposits (0.5-1.3 ka) are substantially younger than the lateral flood plain deposits (2.5-3.0 ka) in the upper reaches of the Columbia River estuary (Table 3).

Cathlamet Bay and Grays Bay
The middle reaches of the Columbia River estuary include the two lateral embayments, Calthlamet and Grays Bay, and the intervening intertidal sand shoals (CRTS) in the central axis (Figure 3).Tidal marshes are developed in small tributary head deltas in Cathlamet Bay (JohnD) and Grays Bay (CRGB07) (Figures 6 and 7).Muddy sand deposits in the protected intertidal flats and channel accretionary banks of Cathlamet Bay (CRCB05, CRCB06, CRCB09, CRMI1) and Grays Bay (CRGB06) give way to intertidal sand flats and shallow sand channels (< 10 m depth) in the central axis sand shoals.Muddy sand deposits in channel accretionary banks at -4 to -5 m elevation reach 1.0-1.3ka in age in Cathlamet Bay (CRCB02 and CRCB05) and Grays Bay (CRGB06).Sedimentation rates in those channel accretionary bank settings reach 3-4 m ka -1 .The sedimentation rate for a muddy sand accretionary bank at CRMI (~ 0.5 to -1.5 m elevation) is estimated to be about 3.3 m ka -1 for the period from 528-659 to 989-1368.The accretionary bank sedimentation rates are much larger than measured rates of relative sea level rise (~ 0.75 m ka -1 ) for latest Holocene time in the study area (Peterson et al., 2010).The high sedimentation rates in the accretionary banks might reflect channel cut and fill deposition following channel lateral migration in middle reaches of the Columbia River estuary.Sedimentation rates of mud and peaty mud in overbank deposits from protected tidal marshes in Cathlamet Bay (JohnD) and Grays Bay (CRGB07) are much lower (1.6-1.8 m ka -1 ).Those overbank sedimentation rates in the protected tidal marshes more closely follow the relatively low-rates of latest-Holocene sea level rise in the estuary (Peterson et al., 2013).

Youngs Bay and Baker Bay
Youngs Bay and Baker Bay are shallow intertidal embayments that are located on either side of the central axis of the lower estuarine reaches, near the mouth of the Columbia River (Figure 3).Youngs Bay and Baker Bay are partially protected from high wave energy entering the mouth of the Columbia River by their leeward positions, respectively behind the south spit and Cape Disappointment.Tidal flats (-1 to 1 m elevation) and marshes (1 to 2 m elevation) have nearly infilled the tidal creek valleys that formed the two embayments (Figure 3; Table 2).Tidal marsh deposits in Youngs Bay (CG1/2 and L&C) record multiple paleotsunami sand sheets (2-4 in number) that are associated with coseismic subsidence events (Sub events # 1-6) at 0.3-2.6 ka (Table 3; Figure 8) (Jurney, 2001).High sedimentation rates (3-5 m ka -1 ) are found for Baker Bay core sites (CRBB06 and CRBB05) suggesting very rapid infilling of the embayment subtidal flats and/or tidal channel accretionary banks during the last 1,000 years (Figure 9).

Central Axis Sand Shoals
Muddy sand deposits in the lateral embayments of Cathlamet, Grays, Youngs and Baker Bays (Figure 3) transition to intertidal sand flats and shallow subtidal sand shoals in the central axis regions of the middle reaches (CRTS01, CRTS02, CRTS03, CRTS04, CRTS05, CRCB01, CRCB07) and lower reaches (CRSK03, CRSK02, CRSK04, CRB07) of the estuary (Figures 6,7,8,9).Very-high sedimentation rates are locally indicated for some of the central axis sand shoals, where whole valves or large shell fragments from the Corbicula mollusk were found to subsurface depths of more than one meter.Corbicula was introduced to the CRE in the early 1900s (Counts, 1985;1986).The first occurrences of Corbicula are locally found to maximum subsurface depths of 2.5 m in site CRSK02, 2.1 m depth in site CRGB04, 2.7 m depth in site CRTS01, and 2.9 m in site CRCB04 (Figures 6,8,9).
The historic ages and multi-meter depths of the Corbicula shells in some of the central axis sand shoals (Figure 3; Table 3) lead to very-high sedimentation rates of 2-3 cm yr -1 in historic time.The very-high sedimentation rates in some of the sand shoals greatly exceed rates of expected relative sea level rise (~ 0.1 cm yr -1 ) as projected into the last 100 year interval from the latest Holocene sea level curve (Peterson et al., 2010;Peterson et al., 2013).If such high sedimentation rates occurred widely in the central axis of the estuary during latest Holocene time (0 -2.5 ka) then the shallow subtidal shoals would have converted to intertidal marsh islands and/or to supratidal floodplains prior to historic time.Are such high-sedimentation rates in some of the central axis sand shoals due to 1) anomalous infilling of the lower estuarine reaches following jetty construction or channel dredge spoil disposal (CREDDP, 1983) or 2) natural channel migrations and corresponding cut-and-fill mechanisms in shifting sand shoals?Also, what processes maintained the sand shoals at low intertidal or shallow subtidal levels during historic time and possibly throughout latest Holocene time?These questions are addressed below in the discussion section.

Depositional Environments
Core logs from 58 sites in the Columbia River estuary indicate six depositional environments based on dominant sediment textures and cores site elevations (Figure 3).Peaty-mud overbank deposits occur in supratidal floodplains in the upper reaches (Figure 4) and in intertidal marsh islands in the middle reaches (Figure 5).Muddy sand deposits occur in protected lateral embayments and in intertidal accretionary banks in the middle reaches (Figures 5, 6, 7).Sand deposits occur in exposed sand shoals and in subtidal channel deposits in the central axes of the middle and lower reaches (Figures 6-9).Depositional sequences of 1) peaty mud overbank deposits, 2) muddy sand accretionary bank deposits, and 3) sand shoal or sand channel deposits differ substantially in their basal ages (Figures 10A and 11).For example, lateral floodplain progradations in the uppermost reaches of the estuary have constricted the main Columbia River channel(s) to the north of CLATB3 in transect A-A' since 2.5-3.5 ka and between sites CRCF02 and CRCF04, respectively, in transect B-B' since 2.5-3.5 ka and 2.8-2.9 ka.The relative stability of the deep main channel(s) in the uppermost reaches of the estuary contrast with the relative instability of shallow multiple channels in the middle and lower estuarine reaches.Three to four shallow channels separate intertidal marsh islands in transects C-C' and D-D' where muddy accretionary banks (CRAP07, CRAP01, CRMI1, CRGB06, CRCB02, CRCB05) date back to only about1.3ka at several meters depth subsurface.In the lower estuarine reaches multiple shallow channels (3-4 in number) dissect intertidal sand shoals and shallow subtidal sand shoals in the central axis of the CRE.Sand shoal deposits in transects E-E' and F-F', respectively date from 0.1 ka (CRTS01, CRSK02) to 1.0-1.3ka (CRCB02), at several meters depth subsurface.The sand shoal basal dates are on the average about five times younger than the floodplain basal dates from similar subsurface depths.

Geomorphic Controls on Depositional Environments
The older lateral floodplains, dating back to about 2.5 ka, are concentrated in the upper reaches of the Columbia River estuary, especially upriver of Aldrich Point (Figures 3,11).A narrowing of the Columbia River valley at Aldrich Point (4 km valley width) might have increased the floodwater elevations there, leading to the elevated floodplain surfaces (> 2.5 m elevation) in the uppermost reaches of the estuary.The substantial constriction of the Columbia River valley at Aldrich Point also eliminated the potential for westerly wind-wave erosion in the eastern uppermost reaches of the estuary.The younger intertidal marsh islands and shallow muddy sand flats, dating back to about 1.3 ka, are most prominent in the wider middle reaches (9-14 km in width) of the estuary.Cathlamet Bay is located upriver of Tongue Point on the south shore of the estuary.Grays Bay is located upriver of Point Ellice on the north shore of the estuary.River valley constrictions at Point Ellice and Tongue Point limited westerly wind-wave propagation into Cathlamet Bay and Grays Bays (Figure 10B).
Protection from wind-wave erosion permitted the development of the intertidal lateral marshes and marsh islands in the protected embayments in the lower and middle reaches of the Columbia River estuary.Cathlamet Bay is particularly well-sheltered from the strongest wind-wave forcing from the southwest (bearings 245-270°) during winter storms (Figure 10B).Only a narrow range of northwest wind-waves from the mouth (bearings 280-285°) can reach the interior of Cathlamet Bay.Grays Bay is sheltered from summer wind-wave forcing from the northwest (bearings 280-310°) and to a lesser extent from southwest wind-waves at bearings less than 240°.Erosional shorelines or gavel lag shorelines presently occur where wind-wave fetch is unobstructed, such as at Point Ellice and Tongue Point (Figure 10B).Intertidal sand shoals and intervening subtidal channels occur in the central axes of the middle and lower reaches.Sand shoals in the middle and lower reaches of the estuary are exposed to wind-waves from both the southwest quadrant (250-270°) and the northwest quadrant (285-310°).
Dated vibracore sequences from the Columbia River estuary do not show expected trends of decreasing sedimentation rates in sites that were most exposed to wind-wave erosion.The sand shoals in the central axes of the lower and middle reaches were exposed to the greatest range in wind-wave bearings and fetch distances in the estuary (Figure 10B).Some core sites in sand shoals in the central axes demonstrate the highest sedimentation rates in the estuary (average 2.5 m per century) (Table 4).These extreme sedimentation rates are recognized by the onset of the introduced Corbicula shells at several meters depth in continuous cores.Most core sites in the central axes sand shoals lack Corbicula shells at depth (Figures 5-9), suggesting that the extreme sedimentation rates are highly localized within the central axes areas.Sedimentation rates are also compiled for the marsh islands (average 3.3 m ka -1 ) and for the tidal flats (average 3.6 m ka -1 ) in the protected embayments.The modest sedimentation rates in the protected embayments are nearly an order of magnitude smaller than those for localized conditions in the central axes sand shoals.The lateral floodplains in the protected embayments and in the uppermost reaches up-river of Aldrich Point (Figure 4) show the lowest sedimentation rates in the estuary (average 1.1 m ka -1 ).Some mechanism(s) other than wind-wave erosion must control short-term sedimentation rates in the Columbia River estuary.quadrants.Monthly onshore wind-speeds (arrows in m/s) are from offshore buoy data (Byrnes & Li, 1998).Bathymetry is digitized from 1967 NOAA navigation charts (± 1 ft MLLW converted to m NAVD88) (USCGS, 1967;1968) that predate the most recent impacts from the Columbia River tributary impoundments and navigation channel deepening.

Sediment Level Curves
Deposit ages versus corresponding elevations (NAVD88) are compared for three sediment types or depositional environments in the Columbia River estuary (Figure 12).Peaty mud represents overbank deposition in floodplains and tidal marshes.Muddy sand represents deposition in intertidal accretionary banks and in protected shallow embayments.Sand represents deposition in intertidal sand shoals and subtidal channel deposits.The floodplain and tidal marsh deposits accumulated overbank mud at 1-2 meters above relative sea level for corresponding ages.Long-term sedimentation rates in the stable floodplains and tidal marshes are controlled by rates of relative sea level rise in the study region (~ 0.75 m ka -1 ).Rising sea level increases accommodation space in the upper-intertidal and supratidal settings, thereby controlling sedimentation rates of overbank mud deposition.Dated muddy sand and sand deposits generally occur well below sea levels of corresponding age during the last 1,500 years.Their sedimentation rates (1.9-4.2 m ka -1 ) (Table 4) greatly exceed the corresponding rate of sea level rise (0.75 m ka -1 ).The rates of sea level rise do not control short-term sedimentation rates in the sand and muddy sand depositional environments of the Columbia River estuary.Instead, the high-sedimentation rates represent short-term cut and fill processes resulting from channel lateral migrations.The very-high Figure 12.Sediment level and sea level curves Sediment level curves (deposit age versus elevation) for three depositional settings, as based on 1) peaty mud in supratidal floodplains and upper-interidal marshes (solid circles), 2) muddy sand in protected embayment tidal flats and channel accretionay banks (solid polygons) and 3) sand in channels and shallow sand shoals (solid squares) of the Columbia River estuary.A local relative sea level curve (solid stars) is based on salt marsh data from adjacent tidal bays, Grays Harbor and Willapa Bay (Figure 1) (Peterson et al., 2010).Columbia River estuary data are from Tables 2 and 3.

Early Historic Bathymetry
Intertidal islands, shallow subtidal shoals, and multiple major tidal channels characterized the early historic bathymetry (1841) of the lower and middle reaches of the Columbia River estuary (Wilkes, 1841).The intertidal islands and shallow subtidal shoals (> -4 m elevation) represented about 80-90 % of the mapped estuary surface area (Figure 13).The major tidal channels (2-3 in number within most cross-sections) ranged from -5 to -20 m in elevation and represented about 10-20 % of the mapped estuary surface area.The 1841 estuary bathymetry differs from the 1968 bathymetry (Figure 10B) in 1) two channels at the pre-jetty river mouth, 2) a wider north channel, and 3) a narrower (pre-dredged) middle channel.In other regards the two plan views of the estuary, separated by more than 100 years in time, show little difference in estuary geomorphology.
Bathymetric profiles (1)(2)(3)(4)(5)(6), which were taken along representative track lines in the 1841 bathymetric survey (Figure 13), are shown relative to the NAVD88 datum in Figure 14.The narrow tidal channels in the uppermost middle reaches at profiles 1 and 2 ranged from -5 to -15 m in elevation.The intertidal marsh islands in profiles 1 and 2 represented 50-75% of the cross-section distance.The shallow subtidal shoals (0 to -3 m elevation) represented about 75% of the cross-section distances in profiles 3 and 4 from the widest middle reaches of the estuary.No deep tidal channels (below -8 m elevation) crossed the middle reaches from west to east at profile 3.
The central axes sand shoals in the lower reaches at profile 5 ranged between 0 and -3 m elevation.The subtidal channels in the lower reaches widened to more than 50% of the cross-section distance in profile 6 at the river mouth.The subtidal sand shoals in the lower and middle reaches of the Columbia River estuary demonstrate a very narrow elevation range (0 to -3 m NAVD88 or 1-4 m below mean tidal level, as taken from lead line soundings (Wilkes, 1841).These relatively uniform elevations of the shallow subtidal sand and muddy sand shoals that existed in early historic time (Figure 14) are attributed to channel bank deposit scalping by wind-wave erosion, as discussed below.Small black dots represent soundings on track lines, tied to shoreline navigation points, as surveyed by timed triangulations (Wilkes, 1841).Dotted areas represent shallow shoals, islands, and tidal flats between major tidal channels.Part B: Bathymetry charts (1841) overlain by selected traverses (numbered 1-6) used for bathymetric profiles in Figure 14.Shallow sand shoals (≥ -4 m elev.inshore and ≥ -6 m elev.offshore) are shaded.Marsh islands (above mean sea level) are stippled.Modern geographic features (arrows) are named.

Wind-Wave Erosion of Shallow Subtidal Sand Shoals
The potential for high wind-wave energy in the Columbia River estuary is provided by coincident river valley enhancements of westerly wind flow and long fetch distances (10-30 km) in the middle and lower reaches (Figure 10B).Wind speed data from the meteorological station AST03 at Tongue Point, OR (Figure 3; NOAA, 2013) are analyzed for the year 2012 to help characterize the potential for wind-wave erosion in the estuary.Wind speeds above 6 m s -1 and 8 m s -1 , respectively, represent the upper 5 and 2.5 percent of the wind speed interval distribution.Similar wind speed ranges, means and standard deviations are found for the preceding years 2005-2011, as continuously recorded at AST03 (NOAA, 2013).Very-high wind speed periods (average period velocity ≥ 8.0 m s -1 ) with long durations (period duration ≥ 1.0 hr) are analyzed for date, duration, mean bearing, and mean velocity (Table 5).The longest duration periods (8-21.6 hr) of very-high wind speeds generally occur during the winter season (November to March).They generally blow out of the southwest (bearing 228±43° 1-σ) and typically reach period-averaged velocities of 8-10 m s -1 .1.6 228 heights of 0.3-0.6 m.The shallow depths of sand resuspension by wind-wave orbital motions are consistent with the elevations of the shallow subtidal sand shoals (0 to -3 m NAVD88) in the central axes of the middle and lower reaches (Figure 14).The broad bimodal distribution of high wind-speed bearings, ranging from 190° to 350° (Table 5), likely accounts for the infrequent wind-wave resuspension of muddy-sand in the shallow bottoms of the lateral embayments such as in Grays Bay, Youngs Bay, and Grays Bay (Figures 10B and 11).Notes: wind-wave height (m) and period (s) are predicted on the basis of wind speed (8 and 10 ms -1 ), fetch distance (km), and water depth (m), using SPMWave.htmlcalculator (USGS, 2013a).Wave development is not duration limited for long periods (multiple hours) of very-high wind velocity over fetch distances (10-30 km) in the Columbia River estuary (Figure 10b).Assumed water depths are based on bathymetric profiles over intertidal and shallow subtidal sand shoals (> -4 m elevation) surveyed in 1841 (Wilkes, 1841).

Conceptual Models of Sand Deposition, Re-Suspension, and Throughput in the Central Axes
The Columbia River estuary was supplied with abundant sand (> 5 million m 3 yr -1 ) during late Holocene time (Peterson et al., 2013), but the estuary failed to infill with upper-intertidal overbank deposits.Measured sedimentation rates in the central channel axes and lateral embayments (Table 4) were substantially greater than coeval rates of sea level rise (Figure 12).The high sedimentation rates resulted from tidal channel lateral migration and corresponding cut-and-fill deposition (Figure 15).Given the high-sedimentation rates in the estuary it would be expected that overbank deposition would have followed the channel cut and fill sequences.Instead, shallow sand shoals (0 to -4 m elevation) dominated the central axes in the middle and lower reaches (Figures 13 and 14) (Wilkes, 1841).The central axes sand deposits were eroded by wind-waves (Tables 5-7) that developed under conditions of long fetch distances (Figure 10B) and winter periods of westerly storm winds.The maximum storm wind velocities (8-10 m s -1 ) and long fetch distances (20-30 km) are estimated to have yielded near-bottom orbital velocities of greater than 0.2 m s -1 at water depths of 2-3 m.Superimposed on the 3.0 m tidal range in the estuary the maximum wind-wave erosion of the central axes sand shoals could have extended downward to 4 m below mean tidal level or -3 m NAVD88.
Wind-wave resuspensions of sand and muddy sand deposits in the exposed shallow shoals (Figures 10B and 11) were superimposed on tidal flood and ebb currents in the mesotidal estuary.Tidal currents are assumed to have drifted the wind-wave resuspended sediments to tidal channels, where combined riverine and tidal ebb flows then transported the sediments seaward (Figure 16).There is a potential seasonal offset between maximum wind-wave erosion in winter (Table 5) and maximum river discharge in spring (Table 1) in the Columbia River estuary.Sand that was stripped from the shallow shoals and delivered to deeper shoal margins and tidal channels in winter (November-March) was subsequently transported seaward by asymmetric riverine-tidal flow during prolonged spring floods (April-June) (Sherwood et al., 1990;Jay & Naik, 2011;Jay et al., 2011).The seaward transport of bed load occurred in shallow tidal channels (> -10 m NAVD88) and in the subtidal sand shoals (-1 to -4 m elevation) (Figure 13).These seaward-transport corridors are located well above the strongest gravity-driven flows in the partially stratified deep channels (-10 to -20 m elevation) near the tidal inlet (CREDDP, 1983;Jay & Smith, 1990).Small inflows of bed load in the bottoms of the deepest tidal inlet channels were outmatched by much greater outflows of bed load across the shallow tidal channel margins and subtidal shoals in the lower estuary reaches.4.0 0.28 Notes: wind-wave orbital velocity (m s -1 ) is estimated for predicted wave height (m), wave period (s), and water depth (Table 6), using RunWaveCalcs.htmlcalculator (USGS, 2013b).Representative wave parameters are from predicted wind speeds of 8.0 m s -1 (*) and 10 m s -1 ( # ).Orbital velocities > 0.2 m s -1 are above the threshold for sand erosion.2) and early historic bathymetric surveys (Figure 14).Maximum depth of wind-wave erosion (-3 m elevation NAVD88) is based on estimated maximums of wind-wave heights (Table 6) and corresponding near-bottom wave orbital velocities > 0.2 m s -1 at 0-4 m water depth or > -3 m NAVD88 (Table 7).(Sherwood and Creager, 1990).Net seaward flows dominate the middle estuary reaches (solid circles) and the shallow water depths (>-10 m NAVD88) in the lower reaches, particularly during spring river flooding (Table 1).The overall asymmetry of riverine-tidal advection in the shallow tidal channels and subtidal shoals transports sand through the central axes of the estuary to the Pacific Ocean.

Conclusions
Traditional models of incised valley accommodation space (Dalrymple et al., 1994;Heap & Nichol, 1997;Smith et al., 2009) do not explain the paradox of abundant bed load supply and an unfilled central bay in the large Columbia River estuary.The paradox in the Columbia River estuary is resolved by analyzing prehistoric core deposits, early historic bathymetry and riverine-tidal discharges, and modern wind-wave forcing conditions.Sediment grain size and deposit age relations are derived from dated vibracore sequences.These data provided sedimentation rates for representative depositional settings during the latest Holocene (0-2,500 14 CyrBP).Low sedimentation rates for peaty mud deposits in semi-stable floodplain and marsh island settings were controlled by low rates of net sea level rise in the estuary.In contrast, high sedimentation rates in muddy sand accretionary banks and in sandy channel deposits exceeded coeval rates of sea level rise.High sedimentation rates in the sand and muddy sand deposits in the central axes and in some lateral embayments of the estuary were related to relatively high rates of recent channel lateral migration.The oldest sand deposits that were preserved at several meters depth in the middle and lower estuary reaches extend back to only 1,500 yeas in age.The high sedimentation rates should have led to central bay infilling by overbank mud deposition in upper-intertidal marsh islands and/or lateral floodplains.However, near-bottom orbital velocities of the large wind-waves eroded the sand shoals to shallow subtidal depths during major winter storms.The seasonal erosion of channel accretionary banks by winter wind-waves prevented the central bay from infilling with overbank mud deposits.The remobilized bed load sediments were delivered to deeper shoal margins and tidal channels in winter months.During subsequent spring months of river flooding (Jay et al., 1990;Jay & Naik, 2011;Jay et al., 2011) the strong asymmetries in riverine-tidal discharges in the shallow shoals and tidal channels transported sand seaward to the Pacific Ocean.These unusual circumstances permitted the shallow Columbia River estuary to bypass bed load to the marine side (Gates, 1994;Baker et al., 2010, Peterson et al., 2013) and yet to remain unfilled by upper-intertidal overbank deposits in the central axes and in the larger lateral embayments for the last several thousands of years.

Figure 1 .
Figure 1.Location of the Columbia river estuary

Figure 3 .
Figure 3. Vibracore sites in Columbia River estuary Location of core sites (solid black circles) relative to major geographic features (named), intertidal shoals and floodplains (dashed lines) in the Columbia River estuary.Anemometer data is from the AST03 Met Station (solid square) at Tongue Point, 46.208 N 123.767W (NOAA, 2013).Background imagery is from Google Earth (2013).

Figure 4 .
Figure 4. Core logs from vibracores in Clatskanie flood plains

Figure 5 .
Figure 5. Core logs from vibracores in Aldrich Point islands

Figure 6 .
Figure 6.Core logs from vibracores in Cathlamet Bay

Figure 8 .
Figure 8. Core logs from vibracores in Baker Bay

Figure 10 .
Figure 10.Core site traverses and wind directions

Figure 13 .
Figure 13.Early bathymetric survey (1841) Part A: Earliest-historic complete bathymetry (1841) of the Columbia River estuary from the Wilkes Expedition.Small black dots represent soundings on track lines, tied to shoreline navigation points, as surveyed by timed triangulations(Wilkes, 1841).Dotted areas represent shallow shoals, islands, and tidal flats between major tidal channels.Part B: Bathymetry charts (1841) overlain by selected traverses (numbered 1-6) used for bathymetric profiles in Figure14.Shallow sand shoals (≥ -4 m elev.inshore and ≥ -6 m elev.offshore) are shaded.Marsh islands (above mean sea level) are stippled.Modern geographic features (arrows) are named.

Figure 14 .
Figure 14.Representative profiles of early bathymetry (1841) Selected bathymetric profiles from 1841 soundings in the Columbia River estuary.Data are converted to meters (m) NAVD88, assuming mean sea level is at 1.0 m NAVD88.

Figure 16 .
Figure 16.Hypothetical directions of annual net flow

Table 1 .
Early historic discharge measurements near the mouth of the CRE

Table 2 .
Core site positions and surface elevations in tidal flats and marshes

Table 3 .
Radiocarbon, tephra, subsidence, Pb 210 and Corbicula dates from late holocene tidal flats and marshes

Table 5 .
Storm wind-speeds in the Columbia River estuary for the year 2012

Table 6 .
Predicted wind-wave height and period

Table 7 .
Estimated near-bottom orbital velocities from predicted wind-waves and water depths