The Morphometry of the Deep-Water Sinuous Mendocino Channel and the Immediate Environs, Northeastern Paciﬁc Ocean

: Mendocino Channel, a deep-water sinuous channel located along the base of Gorda Escarpment, was for the ﬁrst time completely mapped with a multibeam echosounder. This study uses newly acquired multibeam bathymetry and backscatter, together with supporting multichannel seismic and sediment core data to quantitatively describe the morphometry of the entire Mendocino Channel and to explore the age and possible causes that may have contributed to the formation and maintenance of the channel. The ﬁrst 42 km of the channel is a linear reach followed for the next 83.8 km by a sinuous reach. The sinuous reach has a sinuosity index of 1.66 before it changes back to a linear reach for the next 22.2 km. A second sinuous reach is 40.2 km long and the two reaches are separated by a crevasse splay and a large landslide that deﬂected the channel northwest towards Gorda Basin. Both sinuous reaches have oxbow bends, cut-off meanders, interior and exterior terraces and extensive levee systems. The lower sinuous reach becomes more linear for the next 22.2 km before the channel relief falls below the resolution of the data. Levees suddenly decrease in height above the channel ﬂoor mid-way along the lower linear reach close to where the channel makes a 90 ◦ turn to the southwest. The entire channel ﬂoor is smooth at the resolution of the data and only two large mounds and one large sediment pile were found on the channel ﬂoor. The bathymetry and acoustic backscatter, together with previously collected seismic data and box and piston cores provide details to suggest Mendocino Channel may be no older than early Quaternary. A combination of signiﬁcant and numerous earthquakes and wave-loading resuspension by storms are the most likely processes that generated turbidity currents that have formed and modiﬁed Mendocino Channel.


Introduction
Deep-water sinuous channels are conduits that transport sediments across continental margins and onto abyssal basins, often transporting sands, silts and clays hundreds of kilometers away from the margin. Almost all examples of modern deep-water sinuous channels are found on passive margins and are associated with submarine fans (e.g., [1][2][3][4], among many others). Mendocino Channel, off the U.S. northern California active continental margin (Figures 1 and 2), is one of the few examples (e.g., Reynisdjup Channel off Iceland and Hikurangi Channels off eastern New Zealand) of a modern deep-water sinuous channel that is not associated with a passive margin or a submarine fan although it also has been a conduit of coarse clastic sediments to a deep-sea basin [5]. During the 1984 mapping of the U.S. Exclusive Economic Zone off the western U.S. continental margin [6,7], a 42 km section of the channel was discovered on the abyssal seafloor at the base of the north side of the Gorda Escarpment segment of Mendocino Ridge, a ridge formed by the transform fault that strikes east from Gorda Ridge and beneath the North American continent. The short section of Mendocino Channel strikes east from Gorda Ridge and beneath the North American continent. The short section of Mendocino Channel was mapped using GLORIA, a long-range sidescan sonar ( Figure 3a) and was investigated on a subsequent cruise that collected widely spaced single-channel seismic-reflection profiles and several box cores within the channel and adjacent levee [5]. The GLORIA sidescan only provides images of 6.5-kHz backscatter but the images show the high-backscatter acoustic response of the channel floor. Unfortunately, the GLORIA system provided no measurements of bathymetry. The channel, called "Mendocino Channel" by Cacchione et al. [5], was traced on the GLORIA images from water depths of ~2450 m to ~2750 m. They suggested, based on NOAA bathymetric charts, that the channel is related to Mendocino and Mattole Canyons. Mendocino Channel is not in a submarine fan setting with associated sinuous distributary channels, such as are found on many submarine fans [1,2,8,[9][10][11][12][13][14][15][16][17][18][19][20] to cite just a few. Rather, Mendocino Channel is a single channel that trends across a debris apron at the base of the lower continental margin and continues across hemipelagic sediments deposited in a basin setting. The most similar channels to Mendocino Channel and its setting are Reynisdjup Channel off Iceland [21] and Hikurangi Channels off eastern New Zealand [22]. The entire eastern 850 km of Mendocino Ridge was mapped in 2009 with a multibeam echosounder (MBES) [23] and the new bathymetry includes the entire length of Mendocino Channel at a resolution of 40 m/pixel. The channel has many features that resemble those found on fan-related modern deep-water sinuous channels. In addition to resolving these common features, several enigmatic characteristics occur, such as; (1) the channel trends parallel to Gorda Escarpment but is perched several hundred meters above basin depths and descends to basin depths along its length; (2) an initial 48 km-long linear upper reach that makes an abrupt 90° bend towards the Gorda Escarpment but is immediately followed by (3) a 30 km-long section of sinuous channel that parallels the trend of the escarpment and not down slope to the NNW. This section is followed by (4) a 10.5 km reach with a broad flat region that (5) is abruptly diverted 20° and traverses 58.5 km away from the escarpment and out onto the basin. The term "reach" is used here senso lato and does not connote the strict subaerial hydraulic or geomorphic definition of the term. The purpose of this study is to use newly acquired multibeam bathymetry and backscatter, together with available supporting data to The entire eastern 850 km of Mendocino Ridge was mapped in 2009 with a multibeam echosounder (MBES) [23] and the new bathymetry includes the entire length of Mendocino Channel at a resolution of 40 m/pixel. The channel has many features that resemble those found on fan-related modern deep-water sinuous channels. In addition to resolving these common features, several enigmatic characteristics occur, such as; (1) the channel trends parallel to Gorda Escarpment but is perched several hundred meters above basin depths and descends to basin depths along its length; (2) an initial 48 km-long linear upper reach that makes an abrupt 90 • bend towards the Gorda Escarpment but is immediately followed by (3) a 30 km-long section of sinuous channel that parallels the trend of the escarpment and not down slope to the NNW. This section is followed by (4) a 10.5 km reach with a broad flat region that (5) is abruptly diverted 20 • and traverses 58.5 km away from the escarpment and out onto the basin. The term "reach" is used here senso lato and does not connote the strict subaerial hydraulic or geomorphic definition of the term. The purpose of this study is to use newly acquired multibeam bathymetry and backscatter, together with available supporting data to quantitatively describe the morphometry of the entire Mendocino Channel and to explore the age and possible causes that may have contributed to the formation and maintenance of the channel.
Geosciences 2017, 7, 124 3 of 28 quantitatively describe the morphometry of the entire Mendocino Channel and to explore the age and possible causes that may have contributed to the formation and maintenance of the channel. At the outset, it should be mentioned that Peakall et al. [24] challenged the analogy often made between subaerial and deep-water sinuous channels based on qualitative planform morphologies. They suggested differences in the fluid mechanics that acted within the two environments would have produced recognizable features. In particular, they stressed the importance of large-scale overbank flows that can produce recognizable morphologic features on submarine channel levees. Consequently, care has been taken to not over-interpret the processes that may have formed and modified Mendocino Channel using subaerial analogies. At the outset, it should be mentioned that Peakall et al. [24] challenged the analogy often made between subaerial and deep-water sinuous channels based on qualitative planform morphologies. They suggested differences in the fluid mechanics that acted within the two environments would have produced recognizable features. In particular, they stressed the importance of large-scale overbank flows that can produce recognizable morphologic features on submarine channel levees. Consequently, care has been taken to not over-interpret the processes that may have formed and modified Mendocino Channel using subaerial analogies.

General Setting
Mendocino Channel occurs at the base of Gorda Escarpment, the north-facing slope of the eastern-most section of Mendocino Ridge. The eastern end of the escarpment is the Mendocino Triple Junction, the point where the Gorda and Pacific Plates meet at the Mendocino Transform Fault. Gorda Escarpment is the eastern section of the transform fault that extends from Gorda Ridge to the west to North America on the east. Mendocino Channel extends westward from the continental margin for more than 148 km and out onto the southernmost Gorda Basin (Figure 2). Mendocino Triple Junction and Transform Fault are the locus of numerous earthquakes with magnitudes greater than Mw5 over the past 100 years whose affects have been felt over the entire region.
The regional bathymetry suggests the southern boundary of Eel Fan is just to the north of the new MBES bathymetry. Tréhu et al. [25] reported on a series of N-S multichannel seismic profiles that cross a section of Gorda Escarpment that includes Mendocino Channel. The profiles ( Figure 4) show that Mendocino Channel is perched against the north wall of Gorda Escarpment and has constructed a levee to the north, overlapped in places by landslide deposits and buried in other places by distal sediments of Eel Fan. The channel-levee complex is more than 250 m thick in places and was

General Setting
Mendocino Channel occurs at the base of Gorda Escarpment, the north-facing slope of the eastern-most section of Mendocino Ridge. The eastern end of the escarpment is the Mendocino Triple Junction, the point where the Gorda and Pacific Plates meet at the Mendocino Transform Fault. Gorda Escarpment is the eastern section of the transform fault that extends from Gorda Ridge to the west to North America on the east. Mendocino Channel extends westward from the continental margin for more than 148 km and out onto the southernmost Gorda Basin (Figure 2). Mendocino Triple Junction and Transform Fault are the locus of numerous earthquakes with magnitudes greater than M w 5 over the past 100 years whose affects have been felt over the entire region.
The regional bathymetry suggests the southern boundary of Eel Fan is just to the north of the new MBES bathymetry. Tréhu et al. [25] reported on a series of N-S multichannel seismic profiles that cross a section of Gorda Escarpment that includes Mendocino Channel. The profiles ( Figure 4) show that Mendocino Channel is perched against the north wall of Gorda Escarpment and has constructed a levee to the north, overlapped in places by landslide deposits and buried in other places by distal sediments of Eel Fan. The channel-levee complex is more than 250 m thick in places and was deposited above a thick deformed hemipelagic sequence that probably represent sediments of Eel Fan as well as landslide deposits.
Mendocino Channel evolves from two canyon heads, Mendocino and Mattole Canyons, that incise the northern California margin [5] (Figure 2). Mattole Canyon channel has been captured by Mendocino Canyon channel 22 km down-canyon and then evolves as a single channel that stretches for at least 148 km to the west-northwest. The channel trends parallel to Gorda Escarpment for 90.5 km before it has been diverted to the NW around a large landslide deposit that originated on the north flank of Gorda Escarpment. At this point, the channel trends NW away from Gorda Escarpment for at least another 48.5 km before the channel turns back on a more westerly trend for an additional 9.4 km. The MBES bathymetry no longer resolves the channel at 148 km down-channel, which is~114 km east of Gorda Ridge (Figure 1). The average slope of the channel is 0.48 • (8.4 m/km) from the capture point to the distal western-most point that is resolved by the multibeam bathymetry.
Geosciences 2017, 7, 124 5 of 28 deposited above a thick deformed hemipelagic sequence that probably represent sediments of Eel Fan as well as landslide deposits. Mendocino Channel evolves from two canyon heads, Mendocino and Mattole Canyons, that incise the northern California margin [5] (Figure 2). Mattole Canyon channel has been captured by Mendocino Canyon channel 22 km down-canyon and then evolves as a single channel that stretches for at least 148 km to the west-northwest. The channel trends parallel to Gorda Escarpment for 90.5 km before it has been diverted to the NW around a large landslide deposit that originated on the north flank of Gorda Escarpment. At this point, the channel trends NW away from Gorda Escarpment for at least another 48.5 km before the channel turns back on a more westerly trend for an additional 9.4 km. The MBES bathymetry no longer resolves the channel at 148 km down-channel, which is ~114 km east of Gorda Ridge (Figure 1). The average slope of the channel is 0.48° (8.4 m/km) from the capture point to the distal western-most point that is resolved by the multibeam bathymetry. Water depths of the floor of Mendocino Channel range from 1764 m at the capture point to 3025 m at the western-most extent. Landslides are prominent features on the continental margin adjacent to the eastern 25 km of Mendocino Channel as well as along the north wall of Gorda Escarpment. The California margin immediately north of the eastern reaches of the channel has one area of landslides that originated just north of Eel Canyon, another area of landslides just south of Eel Canyon, a landslide scar on the north levee of Mendocino Channel and a large landslide off the north wall of Gorda Escarpment (Figure 2). The landslide just south of Eel Canyon scattered debris at least 18 km away onto the basin floor. The landslide events are undoubtedly related to the high seismicity of the Mendocino Triple Junction and Transform Fault.

Multibeam Echosounder Data
The multibeam bathymetry and acoustic backscatter were collected with the NOAA Ship Okeanos Explorer equipped with a hull-mounted Kongsberg Maritime EM302 MBES system. This

Multibeam Echosounder Data
The multibeam bathymetry and acoustic backscatter were collected with the NOAA Ship Okeanos Explorer equipped with a hull-mounted Kongsberg Maritime EM302 MBES system. This MBES system transmits a 0.5 • wide fore-aft swath and forms up to 864 athwart-ship 1 • receive apertures over a maximum swath of 150 • . Individual soundings along track are spaced approximately every 20 m, regardless of survey speed.
An Applanix POS/MV 320 version 4 motion reference units (MRU) was used to correct for changes in instantaneous ship heave, pitch, roll and heading. The EM302 system can incorporate transmit beam steering up to ±10 • from vertical, and yaw and roll compensation up to ±10 • . The MRU was interfaced with a C&C Technologies C-Nav differential-aided GPS (DGPS) receiver that provides real-time correctors to the DGPS position fixes, providing spatial accuracies of <±0.5 m. All horizontal positions were georeferenced to the WGS84 ellipsoid and vertical referencing was to instantaneous sea level.
Water-column sound-speed profiles were calculated from casts of calibrated Sippican model Deep Blue expendable bathythermographs (XBTs) that measured to 760 m maximum water depth. XBT casts were routinely made every 6 h and between scheduled casts whenever the difference between measured sound speed at the transducers differed by more than 0.5 m/s from the sound speed calculated from the XBT value at the depth of the transducer.
The Kongsberg EM302 is capable of simultaneously collecting co-registered full time-series acoustic backscatter along with bathymetry. The backscatter data represent a time series of backscatter measurements across each individual beam footprint on the seafloor. If the received backscatter amplitudes are properly calibrated to the outgoing acoustic signal strength, receiver gains, spherical spreading, and attenuation, then the calibrated backscatter should provide clues about the composition of the surficial seafloor. A digital terrain model (DTM) was constructed from the MBES bathymetry into a non-projected geographic grid with a resolution of 40 m/pixel using a weighted moving average with a weight diameter of 3 that provides a minimum of 8 of the nearest soundings to influence each grid node of the DTM [26]. The multibeam DTM was combined with a bathymetry grid from the NOAA Coastal Relief Model (CRM) (http://www.ngdc.noaa.gov/mgg/coastal/crm.html) to provide a complete view of the eastern 50 km of the canyon-channel system. The CRM is a compilation of NOAA hydrographic surveys and university single-beam and multibeam bathymetry data gridded at 90 m/pixel. All map and perspective figures in this report were generated from this DTM. Similarly, the MBES backscatter data were gridded at 40 m/pixel and draped over the co-registered MBES bathymetry.

Channel Descriptions
Schumm and Brakenridge [27] subdivide fluvial pattern morphologies into (1) straight channels; (2) sinuous-thalweg straight channels; (3) meandering channels; (4) and braided channels. They cite experimental and field studies that have shown that critical thresholds in stream power, gradient and sediment load lead to changes in channel pattern. Although their conclusions may not be relevant to submarine channels because of differences in densities between air and river waters and between oceanic bottom water and turbidity currents, their insights suggest that those three parameters might be important to the formation of submarine channel planforms. Unfortunately, two of those three parameters (stream power and sediment load) are unknown in Mendocino Channel although channel gradient is easily measured from the multibeam bathymetry. Nevertheless, the spirit of their subdivision of channel patterns was used to segment Mendocino Channel.
Mendocino Channel was subdivided into 6 reaches based on the planform geometry. The complete channel includes the following segments: (1) Mendocino and Mattole Canyon channels; (2) USR, an upper straight reach; (3) USinR, an upper sinuous reach; (4) a crevasse splay; (5) LSinR, a lower sinuous reach; and (6) LSR, a lower straight reach. These segments follow in a progression from the continental shelf to the western extent of Mendocino Channel. Each reach has distinct characteristics, as discussed below, that made the subdivision obvious. Although the crevasse splay is not described as a reach, it was identified by planform geometry as such a unique feature that it merited its own subdivision.
Measurements were taken every 0.5 km along the thalweg of the entire length of Mendocino Channel with the junction of Mattole and Mendocino Canyon channels as 0 km. The measured channel parameters include (1) channel floor width; (2) channel top width; (3) channel vertical relief; (4) south-side levee height above the channel floor; (5) north-side levee height above the channel floor; (6) channel cross-sectional area calculated using the formula for a trapezoid; (7) water depth of channel floor and (8) perched height of channel floor above basin depth ( Figure 5). Of these measurements, relationships were found with levee heights, channel floor with channel top width, channel cross-section area, water depth of channel floor, perched height of channel floor, all with respect to down-channel distance. Other parameters typically measured in submarine channel meanders were measured (e.g., meander radius of curvature, meander wavelength, meander amplitude) but no statistically significant relationships were found. (4) south-side levee height above the channel floor; (5) north-side levee height above the channel floor; (6) channel cross-sectional area calculated using the formula for a trapezoid; (7) water depth of channel floor and (8) perched height of channel floor above basin depth ( Figure 5). Of these measurements, relationships were found with levee heights, channel floor with channel top width, channel cross-section area, water depth of channel floor, perched height of channel floor, all with respect to down-channel distance. Other parameters typically measured in submarine channel meanders were measured (e.g., meander radius of curvature, meander wavelength, meander amplitude) but no statistically significant relationships were found.

Mendocino and Mattole Canyon Channels
Mendocino Channel begins where Mendocino Canyon channel captured Mattole Canyon channel at the base of the northern California margin (white arrowhead on Figures 2 and 6). The upper reaches and junction of Mendocino and Mattole Canyons were first illustrated and discussed by H.W. Murray in the late 1930s (www.history.noaa.gov/stories_tales/mendocino.html). Murray's bathymetry was generated from a compilation of more than 80,000 laboriously collected lead-line soundings taken over a 3-year period and provides a remarkably good general overview of the two canyons, although Mendocino Channel was not discovered. Shepard and Dill [28] (their Figure 47) published a version of Murray's map in 1966. The combined MBES and CRM bathymetry provides for the first time an accurate and high-resolution view of the two canyons that shows their complexities. The head of Mendocino Canyon is at a water depth of 109 m ( Figure 6) with an overall gradient of 6.33 • . Mendocino Canyon channel has a rough and stepped subtle concave-down descent with 7 distinct step-downs that range from 30 to 172 m ( Figure 7). The widths of the broad U-shaped canyon valley progressively increase from 1 km at the proximal end to 2.47 km wide at the distal end just east of the capture point. A large 900 m wide, 4 km long terrace stands 200 m above the canyon channel floor on the gentle inside bend of the first bend of the valley (T a on Figure 6). About 2.5 km down-canyon, a two-stepped tilted terrace occurs on the inside of the next bend (T b and T c on Figure 6). The lower terrace (T c ) stands about 350 m above the canyon channel floor and the higher terrace (T b ) stands 730 m above the floor. A flat unpaired terrace (T d ) occurs across the channel from terrace T c on the south bank (T d on Figure 6) that stands 355 m above the Mendocino Canyon channel floor. The canyon valley makes a sharp 40 • left bend just down-canyon from the T c terrace and immediately before the capture of Mattole Canyon channel. The upper reaches of Mendocino Canyon channel have no apparent incised thalweg, although any thalweg with less than 0.5 m of relief would be below the vertical resolution of the MBES in these water depths.
The head of Mattole Canyon has a water depth of~74 m and begins as a gentle swale only 960 m from the present shoreline. The canyon is 26.8 km long with a gentle concave-up surface that has a two-step ( The width of the floor of Mattole Canyon channel varies from 515 m at the canyon head to 1155 m at the junction with Mendocino Canyon but the widest section of the canyon floor at 1.65 km occurs in a 3.8 km-long section area centered 12.8 km down canyon. Mattole Canyon has a three-step terrace just above the capture point with heights above the channel floor of 310, 510 and 660 m (T f , T g and T h on Figure 6). The three terraces T f , T g and T h are unpaired with terrace T d . Two scarps (L s on Figure 6) occur on the southern wall of Mattole Canyon and the canyon-channel floor adjacent to the scarps is much rougher than the canyon floor farther down.

Mendocino Channel Upper Straight Reach
The south side of first 12 km of the USR is constrained by the steep (>20°) wall of the Gorda Escarpment whereas the first 12 km of the north side of the USR is flanked by the steep wall of the continental margin and then landslide deposits shed off the California margin (Figures 2 and 8). The channel hugs the north wall of Gorda Escarpment along all but the last 12 km of its length. The upper straight reach extends west in a nearly straight strike for 41.6 km from the capture point of Mendocino and Mattole Canyon channels and ends at a sharp left-hand 90° bend to the south towards the wall of Gorda Escarpment ("c" on Figure 8). The channel floor is 1.1 km wide at the head of the reach and

Mendocino Channel Upper Straight Reach
The south side of first 12 km of the USR is constrained by the steep (>20°) wall of the Gorda Escarpment whereas the first 12 km of the north side of the USR is flanked by the steep wall of the continental margin and then landslide deposits shed off the California margin (Figures 2 and 8). The channel hugs the north wall of Gorda Escarpment along all but the last 12 km of its length. The upper straight reach extends west in a nearly straight strike for 41.6 km from the capture point of Mendocino and Mattole Canyon channels and ends at a sharp left-hand 90° bend to the south towards the wall of Gorda Escarpment ("c" on Figure 8). The channel floor is 1.1 km wide at the head of the reach and

Mendocino Channel Upper Straight Reach
The south side of first 12 km of the USR is constrained by the steep (>20 • ) wall of the Gorda Escarpment whereas the first 12 km of the north side of the USR is flanked by the steep wall of the continental margin and then landslide deposits shed off the California margin (Figures 2 and 8).  A side channel heads on the eastern summit region of Gorda Escarpment (Figures 8 and 9). The channel was first described by Tréhu et al. [25] and is clearly shown in their seismic data ( Figure 4, lines 3, 16 and 24). The new MBES bathymetry shows that the upper side channel does not head in a canyon or even on the margin, but simply appears as a broad shallow swale at the 750 m isobath on a broad high. The first 12.2 km of the upper side channel is relatively straight, but then the channel makes an abrupt series of three near 90° bends before the channel plunges down the north-facing Gorda Escarpment. The upper side channel appears to make yet another 90° bend on a bench about 300 m above the USR floor and traverses east along the bench for another 6.4 km before it enters the USR channel floor at 26.8 km down-channel. At this point, the upper side channel makes an abrupt 120° bend to trend down-channel (west) but no longer can be identified on the MBES bathymetry. The head of the USR is perched 300-360 m above the basin water depths and descends 200 m above the basin depths at the end of the reach (Figure 5g). Levees appear at 12 km down-channel on the north bank of the channel and levee (plus undifferentiated landslide and mass transport deposits) heights decrease from 350 m high at 12 km down-channel to 200 m high at 21 km down-channel. At this point, there is a marked decrease in both north-side and south-side levee heights where the channel has moved beyond the immediate margin and trends out of the zone of landslide and mass A side channel heads on the eastern summit region of Gorda Escarpment (Figures 8 and 9). The channel was first described by Tréhu et al. [25] and is clearly shown in their seismic data ( Figure 4, lines 3, 16 and 24). The new MBES bathymetry shows that the upper side channel does not head in a canyon or even on the margin, but simply appears as a broad shallow swale at the 750 m isobath on a broad high. The first 12.2 km of the upper side channel is relatively straight, but then the channel makes an abrupt series of three near 90 • bends before the channel plunges down the north-facing Gorda Escarpment. The upper side channel appears to make yet another 90 • bend on a bench about 300 m above the USR floor and traverses east along the bench for another 6. At this point, there is a marked decrease in both north-side and south-side levee heights where the channel has moved beyond the immediate margin and trends out of the zone of landslide and mass transport deposits and onto the proximal basin floor (black curly bracket on Figure 5a). The levees are never higher than~145 m throughout the remainder of the USR. There is a strong relationship (south levee R 2 = 0.853 and north levee R 2 = 0.830) of the levee heights with down-channel distance (Figure 5a). Also, there is a rough similarity in trends, but not in absolute values, of the width of the USR channel floor with the width of the USR channel top (Figure 5b). There is no statistical correlation (R 2

Mendocino Channel Upper Sinuous Reach
The upper sinuous reach (USinR) is a 37.5 km section that begins 42.5 km down-channel and ends at 78.5 km down-channel with an overall gradient of 0.33 • . The reach is perched 200 m above basin depths as it trends along the north wall of Gorda Escarpment and continues a westward descent to 140 m above basin depths at the western extent of the reach (Figure 5g). The north bank of USinR is bordered by the continuation of the extensive levee complex found along the upper straight reach but a sudden drop in levee height of 40 m down to basin depths occurs at 65 km down-channel (Figure 5a). The USinR is sinuous with 12 bends, one of which is an abandoned full 190 • cut-off meander, whereas other bends turn as little as 20 • Figure 11). Bend A has a flat, level bench (terrace or outer-bank bar [29]) 45 m above the channel floor compared to bend B directly across the channel that has a tilted bench only 3 m above the channel floor. Bend C has a flat, level terrace (or outer-bank bar?) 18 m above the channel floor. Cut-off meander D is a 190 • bend but a detailed analysis of the MBES bathymetry shows that a chute channel has broken through so that meander D is in fact a cut-off meander (Figure 11a,b). A relict cut-off channel and deep depression occurs on the south bank opposite bend A (Figure 11a,b). The depression is 30 m deep at its deepest and its floor is 15 m below the adjacent channel floor at bend A. The depression is separated from the channel floor at bend A by a steep (50 • ) 200 m wide, 23 m high sill, although the depression is inline with a broad gently curved abandoned channel that begins 18 m above the inner bend of the cut-off meander, and descends to depths 21 m deeper at the center of the depression. The depression and associated abandoned channel are erosional, not depositional, features.
A relict cut-off channel and deep depression occurs on the south bank opposite bend A (Figure  11a,b). The depression is 30 m deep at its deepest and its floor is 15 m below the adjacent channel floor at bend A. The depression is separated from the channel floor at bend A by a steep (50°) 200 m wide, 23 m high sill, although the depression is inline with a broad gently curved abandoned channel that begins 18 m above the inner bend of the cut-off meander, and descends to depths 21 m deeper at the center of the depression. The depression and associated abandoned channel are erosional, not depositional, features.

Mendocino Channel Crevasse Splay
A broad conspicuous 5 km-long flat area (4.5 km E-W × 1.7 km N-S) begins at 78.5 km downchannel just beyond the USinR. The area is bounded on the north by a discontinuous low bank. This section has the broadest channel width of the entire length of Mendocino Channel but it continues the perched descent to the west from 140 to 120 m above basin depths while it remains close to the north wall of Gorda Escarpment. The low, broad area resembles a crevasse splay where in the past the channel breached the north bank in several places (Figures 2 and 12). The north-south gradient between the banks is only ~0.02° whereas the east-west gradient is 0.33°, the latter gradient is typical of the USinR immediately up-channel. The north bank varies between 4 and 12 m high except for the suspected breach areas, which are less than 2 m high. High backscatter (−17-−12 dB) decreases in intensity to the north in the broad area of the crevasse splay ( Figure 12). Although the high backscatter might reflect debris shed off the escarpment, the location of the broad area of high backscatter is 6.5 km east of the eastern edge of the relief that reflects the large landslide deposit (see below) and nowhere else along the length of Mendocino Channel is high backscatter seen to extend to the north outside the channel.

Mendocino Channel Crevasse Splay
A broad conspicuous 5 km-long flat area (4.5 km E-W × 1.7 km N-S) begins at 78.5 km down-channel just beyond the USinR. The area is bounded on the north by a discontinuous low bank. This section has the broadest channel width of the entire length of Mendocino Channel but it continues the perched descent to the west from 140 to 120 m above basin depths while it remains close to the north wall of Gorda Escarpment. The low, broad area resembles a crevasse splay where in the past the channel breached the north bank in several places (Figures 2 and 12). The north-south gradient between the banks is only~0.02 • whereas the east-west gradient is 0.33 • , the latter gradient is typical of the USinR immediately up-channel. The north bank varies between 4 and 12 m high except for the suspected breach areas, which are less than 2 m high. High backscatter (−17-−12 dB) decreases in intensity to the north in the broad area of the crevasse splay ( Figure 12). Although the high backscatter might reflect debris shed off the escarpment, the location of the broad area of high backscatter is 6.5 km east of the eastern edge of the relief that reflects the large landslide deposit (see below) and nowhere else along the length of Mendocino Channel is high backscatter seen to extend to the north outside the channel.

Mendocino Channel Lower Sinuous Reach (Landslide and Beyond Landslide)
The lower sinuous reach (LSinR) includes a complex sinuous channel as well as a large landslide that shed debris off Gorda Escarpment and a channel blockage ( Figure 13). The LSinR begins 85.5 km down-channel at a water depth of 2759 m, just beyond the crevasse splay, with a series of tight meanders and ends at 125.7 km down-channel at a water depth of 2944 m. The sinuosity index for LSinR is 1.83. The reach has a gradient of 0.27° with very little resolved relief on the channel floor. However, there is a 5.8 km section of the channel that starts at 94.8 km down-channel (red dashed bracket on Figure 13) that is completely blocked by 3-5 m high rough relief that appears to be debris from the large landslide. The material that blocks the channel has backscatter values of −16-−18 dB, not much different than the backscatter values immediately up-channel in the reach (Figure 13b). The width of the channel floor continues to decrease and water depths continue to increase down-channel (Figure 5f). Levees occur on both sides of the LSinR along almost the entire length. However, the available seismic data do not image levees beyond the LSinR so the full extent of the levees is unknown. The backscatter values are high (−18 to −20 dB) at the beginning of the LSinR but the values decrease to levee values (−26-−30 dB) by 147.6 km down-channel.
As with the other up-channel reaches, the LSinR is perched above basin depths (Figure 5g). The LSinR continues the westward decline, from 120 m above basin depths at the beginning of the reach to less than 20 m at the end of the reach. At the end of the reach, the LSinR has essentially reached basin depths.
The initial west-trending LSinR channel abruptly changes course and trends northwest away from Gorda Escarpment at 89.9 km down-channel because of a large landslide deposit that originated on Gorda Escarpment (Figures 2 and 13). The top of the landslide deposit is 550 m high, stretches 21.7 km along the base against Gorda Escarpment and fans out 4.7 km away from Gorda Escarpment, giving it a minimum volume of ~25 km 3 of talus. There are no resolvable channel obstructions, and other than the area that blocks the channel, no high-backscatter debris within the LSinR that can be confidently attributed to the landslide. Also, there is no hint of a channel on the west side of the

Mendocino Channel Lower Sinuous Reach (Landslide and Beyond Landslide)
The lower sinuous reach (LSinR) includes a complex sinuous channel as well as a large landslide that shed debris off Gorda Escarpment and a channel blockage ( Figure 13). The LSinR begins 85.5 km down-channel at a water depth of 2759 m, just beyond the crevasse splay, with a series of tight meanders and ends at 125.7 km down-channel at a water depth of 2944 m. The sinuosity index for LSinR is 1.83. The reach has a gradient of 0.27 • with very little resolved relief on the channel floor. However, there is a 5.8 km section of the channel that starts at 94.8 km down-channel (red dashed bracket on Figure 13) that is completely blocked by 3-5 m high rough relief that appears to be debris from the large landslide. The material that blocks the channel has backscatter values of −16-−18 dB, not much different than the backscatter values immediately up-channel in the reach (Figure 13b). The width of the channel floor continues to decrease and water depths continue to increase down-channel (Figure 5f). Levees occur on both sides of the LSinR along almost the entire length. However, the available seismic data do not image levees beyond the LSinR so the full extent of the levees is unknown. The backscatter values are high (−18 to −20 dB) at the beginning of the LSinR but the values decrease to levee values (−26-−30 dB) by 147.6 km down-channel.
As with the other up-channel reaches, the LSinR is perched above basin depths (Figure 5g). The LSinR continues the westward decline, from 120 m above basin depths at the beginning of the reach to less than 20 m at the end of the reach. At the end of the reach, the LSinR has essentially reached basin depths.
The initial west-trending LSinR channel abruptly changes course and trends northwest away from Gorda Escarpment at 89.9 km down-channel because of a large landslide deposit that originated on Gorda Escarpment (Figures 2 and 13). The top of the landslide deposit is 550 m high, stretches 21.7 km along the base against Gorda Escarpment and fans out 4.7 km away from Gorda Escarpment, giving it a minimum volume of~25 km 3 of talus. There are no resolvable channel obstructions, and other than the area that blocks the channel, no high-backscatter debris within the LSinR that can be confidently attributed to the landslide. Also, there is no hint of a channel on the west side of the landslide deposits; consequently, it appears that the initiation of the landslide predates the formation of Mendocino Channel.  A series of outer-and inner-bend terraces (T1 through T10 on Figure 13) occurs throughout the LSinR, all at varying heights above the channel floor ( Table 1). Some of the terraces are flat in a channel-orthogonal direction whereas other terraces are tilted towards the channel ( Table 1). The terraces occur on the outside of channel bends (T1, T2, T3, T4, T5, T7, T8, T9, and T10) as well as along linear sections (T6 and T8). Paired terraces (similar water depths on adjacent inside and outside bend terraces) include T7 to T8 and T9 to T10 whereas all the other adjacent terraces are unpaired.
A subtle 2-3 m deep linear channel strikes NW from the LSinR (black hachured lines in Figure 13). The channel may be related to an older course of Mendocino Channel given that it is in line with the initial eastern straight section of the LSinR. The subtle channel's closest approach to the LSinR at its SE end is only 1.5 km but the floor of the subtle channel is 30 m below the level of the LSinR channel floor and the two are separated by a 15-20 m high ridge. The backscatter values of the linear channel are no different than the levee values in the surrounding area and the multichannel seismic profiles ( Figure 4) have no indication of this channel, both of which suggests the subtle channel has been buried by either LSinR levees or distal hemipelagic sediments of Gorda Basin.  A series of outer-and inner-bend terraces (T 1 through T 10 on Figure 13) occurs throughout the LSinR, all at varying heights above the channel floor ( Table 1). Some of the terraces are flat in a channel-orthogonal direction whereas other terraces are tilted towards the channel ( Table 1). The terraces occur on the outside of channel bends (T 1 , T 2 , T 3 , T 4 , T 5 , T 7 , T 8 , T 9 , and T 10 ) as well as along linear sections (T 6 and T 8 ). Paired terraces (similar water depths on adjacent inside and outside bend terraces) include T 7 to T 8 and T 9 to T 10 whereas all the other adjacent terraces are unpaired.
A subtle 2-3 m deep linear channel strikes NW from the LSinR (black hachured lines in Figure 13). The channel may be related to an older course of Mendocino Channel given that it is in line with the initial eastern straight section of the LSinR. The subtle channel's closest approach to the LSinR at its SE end is only 1.5 km but the floor of the subtle channel is 30 m below the level of the LSinR channel floor and the two are separated by a 15-20 m high ridge. The backscatter values of the linear channel are no different than the levee values in the surrounding area and the multichannel seismic profiles (Figure 4) have no indication of this channel, both of which suggests the subtle channel has been buried by either LSinR levees or distal hemipelagic sediments of Gorda Basin.

Mendocino Channel Lower Straight Reach (LSR)
The lower straight reach (LSR) begins at 125.7 km down-channel and continues for 23 km before the channel is no longer resolved in the MBES bathymetry or backscatter ( Figure 14). The reach has an overall gradient of 0.20 • and consists of several relatively gentle bends and five 90 • + bends. The LSR strikes west (~276 • ) for the first 5 km before it makes a broad 90 • bend to the north but only stays on that course for 1.1 km before it makes another 90 • bend to the west and runs for 2.9 km. At this point, the reach makes the third very broad 90 • bend to the north for 3 km before it makes a final 105 • turn to the west southwest (~261 • ) for the remainder of the resolved channel that ends at 148.7 km down channel. The last 105 • bend in the reach may reflect the influence of the southern extent of the distal hemipelagic sediments of Gorda Basin or Eel Fan as well as influence from subsurface structure (see basement beneath red arrow on line 9 of Figure 4) that may have steered the LSR away from its NW trend out towards Gorda Basin. The LSR differs from the other reaches in that no terraces are found on either side of any bends. Levee heights have considerable variation down-channel (Figure 5d). The north-side levee is only 4-10 m high and the south-side levee is 2-7 m high until an abrupt decrease in levee height of less than 2 m high at 132.5 km down-channel. The levees then increase in heights to 6-8 m at 136-138 km down-channel and then decrease in height until no bathymetric expression of a levee is found at 140 km down-channel (Figure 5d).
The perched channel height vs. distance down-channel is highly variable for the first 15 km but then at a water depth of~2990 m the channel has reached basin depths (Figure 5e).
A barely resolved channel trends to the southwest at the eastern beginning of the LSR with a 0.45 • gradient. The channel can be followed for 3.9 km and heads just after a sharp bend in the LSR, although it is roughly orthogonal to the LSR trend ( Figure 14). The incised depth of the channel ranges from 5 m at its head to less than 2 m at its western limit. The head of the side channel is 725 m WSW of the LSR and descends with a 0.8 • gradient from the top of the south wall of LSR. This abandoned channel has no acoustic backscatter signature and appears unrelated to a former course of Mendocino Channel prior to the large landslide because it is not in line with the LSR and is well north of the landslide deposits. It might be a breakout channel from an earlier stage of Mendocino Channel but presently there is a 24 m high south-side levee that separates the eastern-most abandoned channel from the nearest channel floor of the LSinR and the LSR.

Mendocino Channel Lower Straight Reach (LSR)
The lower straight reach (LSR) begins at 125.7 km down-channel and continues for 23 km before the channel is no longer resolved in the MBES bathymetry or backscatter (Figure 14). The reach has an overall gradient of 0.20° and consists of several relatively gentle bends and five 90° + bends. The LSR strikes west (~276°) for the first 5 km before it makes a broad 90° bend to the north but only stays on that course for 1.1 km before it makes another 90° bend to the west and runs for 2.9 km. At this point, the reach makes the third very broad 90° bend to the north for 3 km before it makes a final 105° turn to the west southwest (~261°) for the remainder of the resolved channel that ends at 148.7 km down channel. The last 105° bend in the reach may reflect the influence of the southern extent of the distal hemipelagic sediments of Gorda Basin or Eel Fan as well as influence from subsurface structure (see basement beneath red arrow on line 9 of Figure 4) that may have steered the LSR away from its NW trend out towards Gorda Basin. The LSR differs from the other reaches in that no terraces are found on either side of any bends. Levee heights have considerable variation down-channel ( Figure  5d). The north-side levee is only 4-10 m high and the south-side levee is 2-7 m high until an abrupt decrease in levee height of less than 2 m high at 132.5 km down-channel. The levees then increase in heights to 6-8 m at 136-138 km down-channel and then decrease in height until no bathymetric expression of a levee is found at 140 km down-channel (Figure 5d).
The perched channel height vs. distance down-channel is highly variable for the first 15 km but then at a water depth of ~2990 m the channel has reached basin depths (Figure 5e).
A barely resolved channel trends to the southwest at the eastern beginning of the LSR with a 0.45° gradient. The channel can be followed for 3.9 km and heads just after a sharp bend in the LSR, although it is roughly orthogonal to the LSR trend ( Figure 14). The incised depth of the channel ranges from 5 m at its head to less than 2 m at its western limit. The head of the side channel is 725 m WSW of the LSR and descends with a 0.8° gradient from the top of the south wall of LSR. This abandoned channel has no acoustic backscatter signature and appears unrelated to a former course of Mendocino Channel prior to the large landslide because it is not in line with the LSR and is well north of the landslide deposits. It might be a breakout channel from an earlier stage of Mendocino Channel but presently there is a 24 m high south-side levee that separates the eastern-most abandoned channel from the nearest channel floor of the LSinR and the LSR.

Discussion
The new multibeam bathymetry and backscatter data provide an unprecedented quantitative view of the morphology of the entire length of Mendocino Channel. The presence of Mendocino Channel, a sinuous channel on an active margin, begs an explanation for the age of the channel as well as what processes formed the channel. To investigate these questions, several lines of evidence were gathered from the literature and data archives that bear on explanations. The following discussion is focused on these questions.

The Age of Mendocino Channel
Twelve cores found in archives and the literature are either from within Mendocino Channel or from the northern levees, all located in the upper sinuous reach section ( Figure 15). The lithostratigraphies and 14 C dates of the cores show a complex pattern that is not straight forward to interpret. Three box cores (BX-1, BX-5 and BX-9) are from Cacchione et al. [5], two piston cores (M9907-47PC and M9907-51PC) are from Goldfinger et al. [30] and one piston core (Y74-1-08) is from www.ngdc.noaa.gov/mgg/curator/ data/melville/avon/avon09mv/051/m9907_051pc_handwritten_corelog.pdf. Each box core sampled the upper 40-50 cm of the channel floor and is composed of numerous thin sandy turbidites interbedded with hemipelagic mud [5]. The sandy turbidites are 2-12 cm thick and conventional 14 C dates from BX-1 range from 0.970 kaBP at 5-cm subbottom to 3.595 kaBP at 48-cm subbottom. None of the other box cores have been dated. BX-9 is the most up-channel box core and is located 33.1 km down-channel from the channel convergence point. BX-5 is located 11.4 km down-channel from BX-9 and BX-1 and is located 13.1 km down-channel from BX-5. BX-5 contains the thickest turbidites of the three box cores, more than likely because it was collected on a sharp bend in the reach whereas BX-1 and BX-9 were collected on straight sections of the USR.
The published locations of piston cores M9907-51PC and M9907-47PC are separated by less than 5 m so both piston cores were collected 6.7 km down-channel from BX-1. The top 50 cm of the two piston-core lithostratigraphies do not resemble one another at all and an AMS 14 C age at 72 cm subbottom from M9907-51PC is 160 yrBP [30]. No ages have been reported from M9907_47PC. Assuming the turbidites and sands represent events that lasted only days or at most months, then the turbidites can be subtracted from the lithostratigraphies to represent only the hemipelagic mud in the sections. This, of course, assumes that the top of each turbidite is defined by the sand unit and the overlying mud is solely of hemipelagic origin and not deposition from the tail of the turbidity current. Nevertheless, if the 45 cm of sandy turbidites from the top 72 cm of M9907-51PC (the depth of the youngest 14 C age) are subtracted from the section, then the linear sedimentation rate of the hemipelagic mud in this section of the core is 169 cm/kyr. Similarly, if the 14 C date of 820 yrBP at 400 cm depth in core M9907-51PC is used, and the turbidites in the top 400 cm are subtracted, the hemipelagic sedimentation rate is 241 cm/kyr; both sedimentation rates that appear very fast for hemipelagic mud. For comparison, the nearest Ocean Drilling Program (ODP) site (Leg 167 Site 1020) is in Gorda Basin 80 km northwest of Mendocino Channel and collected an upper sequence of hemipelagic mud and no sandy turbidites with a sedimentation rate of only 11.5 cm/kyr [31]. If the same assumption and calculation is performed on BX-1 (the closest box core to the two M9907-47PC and M9907-51PC piston cores), using the 3.595 ka 14 C age at 48 cm, then the linear sedimentation rate of the hemipelagic mud in BX-1 is 13.5 cm/kyr, a hemipelagic sedimentation rate that is comparable to that from the ODP site. Likewise, if the sandy turbidites are subtracted from the top 50 cm of M9907-47PC, and the 14 C age at 48 cm in BX-1 is used as the age at 48 cm of M9907-47PC, then the linear sedimentation rate of the top of hemipelagic mud in M9907-47PC is 8.1 cm/yr, a sedimentation rate not too different than that of BX-1 and the nearby ODP site. These order-of-magnitude differences in hemipelagic sedimentation rates in M9907-51PC versus M9907-47PC, BX-1 and the ODP core, and the observation that the 14 C age at 70 cm in M9907-51PC is considerably younger than the 14 C age at 50 cm in BX-1, makes piston core M9907-51PC suspect, either because the published AMS 14 C ages are grossly in error or the location of M9907-51PC is in error and the core is not from within Mendocino Channel.
Geosciences 2017, 7, 124 19 of 28 Figure 15. Location map of piston (circles) and box (squares) cores from within the channel and the north levee of Mendocino Channel. Numbers within the symbols indicate the core number in the core descriptions below the map. BX indicates a box core and the other cores are piston cores. Piston core descriptions were downloaded from www.ngdc.noaa.gov/mgg/curator/data/melville/avon/ avon09mv/047/. Descriptions of box cores are from [5]. AMS 14 C ages of box cores in italics from [5] and from piston cores from [30].
An undated piston core (Y74-1-08PC) whose location plots on the channel floor contains only two sandy turbidites in the top 50 cm; the top turbidite is 7 cm thick and the next older turbidite is 25 cm thick. When the turbidites are subtracted from the section and the above assumptions are applied, Figure 15. Location map of piston (circles) and box (squares) cores from within the channel and the north levee of Mendocino Channel. Numbers within the symbols indicate the core number in the core descriptions below the map. BX indicates a box core and the other cores are piston cores. Piston core descriptions were downloaded from www.ngdc.noaa.gov/mgg/curator/data/melville/ avon/avon09mv/047/. Descriptions of box cores are from [5]. AMS 14 C ages of box cores in italics from [5] and from piston cores from [30].
An undated piston core (Y74-1-08PC) whose location plots on the channel floor contains only two sandy turbidites in the top 50 cm; the top turbidite is 7 cm thick and the next older turbidite is 25 cm thick. When the turbidites are subtracted from the section and the above assumptions are applied, applying the 14 C age from BX-1 results in a linear sedimentation rate of 1.3 cm/kyr for the hemipelagic mud in this core, a sedimentation rate that seems too slow for hemipelagic sedimentation. This slow rate suggests the top of Y74-1-08PC is missing, a common occurrence in piston cores. The section below 50 cm subbottom is composed of 67 cm of sandy hemipelagic mud followed by a 119-cm thick sandy turbidite immediately overlying a 56-cm sandy turbidite. Piston core Y74-01-08PC was collected 33 km down-channel from BX-1, the farthest west box core from the channel floor and 91 km down-channel from the confluence point. Piston core Y74-1-08 is only 2.7 km away from the eastern edge of the large landslide from Gorda Escarpment, which may account for the thick sandy turbidite sequence from 50-300+ cm down core.
The lithostratigraphies of the 6 box cores (BX-2, BX-3, BX-4, BX-6, BX-7 and BX-8) collected from the north levee differ from that of the three box cores (BX-1, BX-5, BX-9) and the two piston cores (Y74-01-08PC and M9907-47PC) collected from the channel. The levee cores all have three or fewer thin sandy turbidites from 5-12-cm thick in the upper 50 cm that are interbedded with much thicker hemipelagic mud. Using the 14 C age of 3.595 kaBP from BX-1 at 49 cm on the levee box cores, subtracting the thin turbidites using the assumptions above, yields linear sedimentation rates that only range from 11.4-13.9 cm/kyr for the hemipelagic muds on the levee. The sedimentation rates of the hemipelagic muds from the levees, even though there are no age dates, are similar to the calculated sedimentation rates of the hemipelagic muds of BX-1 that has a 14 C date as well as to the sedimentation rate from the nearby ODP core.
The box and piston cores demonstrate that that channel has been a conduit for several sandy turbidity currents over at least the past 3.600+ kyrs but there is a conspicuous lack of high-amplitude reflectors in the multichannel seismic profiles that would indicate a significant amount of coarse sediment was deposited in the channel axis over time [4,16,32,33]. Although, the cores do not provide any evidence for how long the channel has existed, the single-channel seismic-reflection profiles from Cacchione et al. [5] and multichannel seismic-reflection profiles from Tréhu et al. [25] (Figure 4) show that the maximum thickness of the Mendocino Channel ranges from~200-~300 m, using the formula for the conversion of travel time to subbottom depth for the western U.S. margin [34]. If the assumption used above that the turbidites and sands represent events of only days or perhaps months duration, then subtracting the turbidites and sands, and using the occurrences of turbidites in the cores as representative of the turbidites in the seismic section, the hemipelagic sedimentation rate from BX-1 can be used as typical for the entire thickness of the Mendocino Channel (admittedly these assumptions could be challenged, but that is all the data available). If the hemipelagic sedimentation rate of 11.5 cm/kyr from ODP Leg 167 Site 2010 is used as the hemipelagic sedimentation rate for the channels in the multichannel seismic sections, then the initiation of the Mendocino Channel may be as old as 2.6-1.8 Ma (early Quaternary).

Potential Processes
Mendocino Channel is a rare example of a sinuous channel on an active margin and one that is not associated with a submarine fan. However, Mendocino Channel has many of the characteristics of sinuous channels found on passive margins that are associated with submarine fans. The potential process or processes that formed and maintained Mendocino Channel include (1) frequent seismicity; (2) periodic storm-wave loadings on the narrow shelf that were large enough to re-suspend and transport shelf sediments to the head of Mattole Canyon; (3) periodic low eustatic sea levels throughout the Quaternary that moved the shoreline to the heads of Mattole and Monterey Canyons; (4) tsunamis and (5) hyperpycnal plumes generated by periodic floods of the Mattole River (there is no river that directly feeds to Mendocino Canyon).

Seismicity
Many studies have suggested that nearshore and offshore seismic activity can trigger turbidity currents, (e.g., [30,35,36]; see Pope et al. [37] for a global summary). There is evidence that seismicity has had some impact on Mendocino Channel (e.g., [5,30]). The heads of Mendocino and Mattole Canyons are located at the Mendocino Triple Junction (red star on Figure 1) and the entire area is seismically active with great earthquakes that are suggested to have occurred on this margin with a recurrence of~500-530 yr and smaller earthquakes along the southern part of this margin with a higher frequency of recurrence [30,[38][39][40]. Although dating the numerous landslides on the California margin adjacent to Mendocino Channel has not yet been attempted, the box cores taken within Mendocino Channel contain a series of thin sandy turbidites interbedded with thicker hemipelagic muds [5]. The top sand in each box core is overlain by 3-5 cm of hemipelagic mud. Conventional 14 C ages of organic debris within the sands date the latest turbidite at 970 ± 80 yBP [5], but no turbidite was recovered that would correlate to the 1906 M w 7.8 San Francisco earthquake. However, the occurrence of numerous large landslides in the immediate area of Mendocino Channel (e.g., [25]), coupled with the local seismicity makes earthquakes prime candidates for the initiation of sediment failures that evolved into turbidity currents that coursed down the margin and onto the basin floor.

Wave-Loading Resuspension
It is widely recognized that some submarine canyons have continued to be conduits for shelf-sediment transport to the deep sea, even during eustatic high stands, e.g., [41]. Wave-loading by periodic storms has been suggested as a mechanism that can re-suspend shelf sediment and generate high concentrations of suspended sediment at canyon heads (e.g., [42] and references within; [43]). This process is more likely in regions with high rainfall with steep terrestrial slopes and extensive mass wasting that generates and transports large volumes of sediment to local coastal rivers. Such rivers would then transport much of the suspended and bed-load sediment to the shelf and form depocenters for eventual resuspension by storm waves (see Nittrouer et al. [44]). If a depocenter is close to a canyon head, then storm-generated re-suspended sediment would have the potential to be transported into and down the canyon as a gravity-driven turbidity current. The coastal mountains of northern California are steep and there is abundant rainfall (annually 1100-2500 mm/yr) and, in addition, major winter storms impact the coast (e.g., 3-10 major storms/yr for the past 50 yr). If major storms occurred during El Niño events, then the major storms would have been intensified [45]. Records of El Niño events have been dated as old as 130 ka [46], so wave-loading resuspension by major storms is a distinctly possible process that could have generated turbidity currents in Mendocino and Mattole Canyons.

Hyperpycnal Plumes
River floods with a high concentration of suspended sediment, with concentrations high enough to make the flood water denser than the sea water it flows into, could form a hyperpycnal plume. An important criterion for the generation of a hyperpycnal flow is the degree of estuarine mixing prior to the flow entering the ocean [47]. Hyperpycnal plumes are generated by typhoons or intense rainfall that couple high river discharge with large suspended-sediment load. The plumes have the potential to flow downslope by gravity and, if the density of the hyperpycnal flow remains denser than the bottom water, sediment-laden plumes have the potential to evolve into a density-driven turbidity current that can flow out onto a basin floor [43,[48][49][50][51]. Hyperpycnal plumes have been invoked for sediment waves off the California Margin just north of Mendocino Channel [52], in the Atlantic off Morocco [53], for deposits on Var Fan in the Mediterranean Sea [20,54], and off Taiwan [50,55,56]. This process has even been invoked as an alternate explanation for Humboldt Slide off northern California [57], although this interpretation is controversial (see discussion by Nittrouer et al. [51]). A dilute turbidity current was observed by Sumner and Paull [58] on an ROV dive in the head of Mendocino Canyon but only in water depths of less than 400 m. Whether this turbidity current reached the seafloor at water depths of 2000-3000 m is unknown, although Kao et al. [55] observed a hyperpycnal flow off Taiwan that reached water depths of 3000-3700 m. Regardless, this process would only be possible for the Mattole River and Canyon because the Mendocino Canyon has no adjacent river to feed sediment to the head of Mendocino Canyon.

Eustatic Sea Levels
Quaternary episodes of major regressions of eustatic sea levels located river mouths closer to adjacent canyon heads. Marine and ice-core studies have demonstrated that Quaternary eustatic sea levels dropped 120-130 m relative to today's sea level during maximum global glacial conditions ( [59]; see De Boer et al. [60] for discussion and references). The shoreline effect of the glacial eustatic low sea levels at the heads of Mendocino and Mattole Canyons is shown in Figure 16

Tsunamis
Tsunamis can have a profound effect on a continental shelf, especially on a very narrow shelf ( [61,62] and references within; [63][64][65]). The outgoing tsunami waves, after first eroding the shore, have the capacity to transport mud, sand and even large boulders to the ocean in the backflow. Once in the head of a canyon, gravity and the sediment-laden density of the backflows would transport the sediment down-canyon as gravity-driven sediment flows that potentially could reach a basin floor. Atwater and Hemphill-Haley [66] and Hemphill-Haley [67] reported tsunami deposits on coastal Washington State and Garrison-Laney et al. [68] reported on 5 tsunami deposits within the last 3500 years at the southern end of the Cascadia Subduction Zone, near the shoreline adjacent to

Tsunamis
Tsunamis can have a profound effect on a continental shelf, especially on a very narrow shelf ( [61,62] and references within; [63][64][65]). The outgoing tsunami waves, after first eroding the shore, have the capacity to transport mud, sand and even large boulders to the ocean in the backflow. Once in the head of a canyon, gravity and the sediment-laden density of the backflows would transport the sediment down-canyon as gravity-driven sediment flows that potentially could reach a basin floor. Atwater and Hemphill-Haley [66] and Hemphill-Haley [67] reported tsunami deposits on coastal Washington State and Garrison-Laney et al. [68] reported on 5 tsunami deposits within the last 3500 years at the southern end of the Cascadia Subduction Zone, near the shoreline adjacent to the heads of both Mendocino and Mattole Canyons. However, Weiss [69] modeled tsunami effects on non-cohesive sediments from the shelf and slope and calculated that even the 2004 Boxing Day tsunami in the Indian Ocean transported fine sand only 335 m offshore in a water depth of 985 m. This study suggests that if a nearby earthquake generated a submarine landslide that in turn initiated a tsunami, the resuspension and transport of fine sand and silt could have reached the heads of both Mendocino and Mattole Canyons.

Conclusions
The new multibeam bathymetry and backscatter data provide the first quantitative description of the entire Mendocino Channel. A more thorough understanding of the immediate area of Mendocino Channel was gained by combining the new analyses of the multibeam data with interpretations of published multichannel seismic and sediment cores from the channel and immediate vicinity. Several aspects of the entire Mendocino Channel system are intriguing. For instance, the concave-down profile of Mendocino Canyon channel suggests it is presently largely inactive and has accumulated sediment in the canyon-channel reaches. Conversely, the concave-up profile of Mattole Canyon channel suggests it is presently active and either has or is in the process of establishing an equilibrium profile. The similar water depths of the flat surfaces of terraces T g and T d on either side of Mattole Canyon, together with the sharp 80 m drop of the floor of Mattole Canyon at the capture point by Mendocino Canyon, is strong evidence that Mattole Canyon channel is older than Mendocino Canyon channel and that it was captured by Mendocino Canyon channel. The lack of obvious high-amplitude reflectors in the subsurface on multichannel seismic profiles ( Figure 4) suggests Mendocino Channel floor has not been the site of significant deposits of coarse sediments. However, the maximum~350 m thick north-side levees of the USR suggests a large amount of sediment has transited the channel over the life span of Mendocino Channel.
What changed the nature of the reaches from straight to sinuous and back to straight? Typically, major changes of planform characteristics of fluvial channels are the result of changes of sediment load, an increase in peak discharge or a change in channel gradient or some combination of the three [70]. The gradient of the USR decreases from 0.71 • at its beginning to 0.12 • at the beginning of the USinR. The gradient at the beginning of the LSinR is 0.18 • and varies down-channel less than ±0.2 • to the end of the reach. Is this small gradient enough to cause a change from linear to sinuous in channel character? Why did the levee heights suddenly decrease in height at two locations? The large north-side levee heights along the USR suddenly decrease from 200 m above the channel floor at 25 km down-channel to 100 m high, about the location where the channel emerges from the steep continental margin, but the change in channel character from the USR to the USinR occurs 41.8 km down-channel where the north and south levees remain less than 100 m above the channel floor. The sudden decreases in north levee heights might reflect the end of continental margin sediment mixed with the north levee sediments. The overall decrease in north and south levee heights in the USR is fairly smooth until the abrupt decrease in height at 22-25.5 km down-channel (Figure 5a). If this is the case, then the sudden decrease in north and south levee heights in the USinR at 62-65 km down-channel must reflect a reduction in the carrying capacity of the turbidity currents and a subtle decrease in channel gradient. The gradient of the channel floor of the USR just before it changes to the USinR is 0.71 • and it flattens to 0.12 • at the beginning of the USinR. The remainder of the USinR has a gradient of~0.33 • throughout the reach and both north and south levees suddenly end at 60-65 km down channel, 80+ km before the channel is no longer resolved in the bathymetry data. Nothing in the multichannel seismic subsurface data suggests any basement control for these changes.
It seems that a combination of significant and numerous earthquakes (seismicity) and wave-loading resuspension are the most likely processes that created the turbidity currents that formed and maintained Mendocino Channel. Eustatic lowered sea levels are possible contributors to the initial formation of the channel, but the lack of high-amplitude reflectors and the young ages for the turbidites in BX-1 suggest that eustatic sea levels are not the major forcing for activity within Mendocino Channel. Answers to the intriguing aspects and questions raised by the analyses presented here must await future dedicated seismic and sediment sampling cruises that can provide the necessary data to address these observations.

•
Newly acquired multibeam bathymetry and backscatter data, together with published multichannel seismic and sediment core data provides a quantitative geomorphometric analysis of the entire extent of Mendocino Channel and to explore the age and possible causes that may have contributed to the formation and maintenance of the channel. Changes from a linear reach to a sinuous reach and back to a linear reach are abrupt with 90 • bends.

•
There are several 180 • meanders and 2 cut-off meanders in the sinuous reaches. • Both inside-bend and outside-bend terraces are found on the sinuous reaches; some are paired terraces and some are unpaired.

•
Although landslides are evident along the north wall of Gorda Escarpment, there is very little evidence of landslide debris within Mendocino Channel. • Dated box and piston cores provide an estimate of the hemipelagic sedimentation rates that, when applied to the thickness of the channel-levee complex suggests the channel may be as old as 2.6-1.8 Ma (early Quaternary). • A combination of significant and numerous earthquakes (seismicity) and wave-loading resuspension are the most likely processes that created the turbidity currents that have flowed down Mendocino Channel.
after completion of each cruise for use by the scientific and industrial communities and the public. The data are available at http://ccom.unh.edu/data/mendocino-east-bathymetry as well as from https://maps.ngdc.noaa. gov/viewers/bathymetry/. Any use of the data should be referenced to doi:10.7289/V5B56GRV (bathymetry) and doi:10.7289/V56D5R03 (backscatter).

Conflicts of Interest:
The author declares no conflict of interest.