Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Contents lists available at ScienceDirect

Estuarine, Coastal and Shelf Science

journal homepage: www.elsevier.com/locate/ecss

Hydrodynamic modelling of tidal-fluvial flows in a large river estuary T

a,b,∗ a b c,d e e
S.D. Sandbach , A.P. Nicholas , P.J. Ashworth , J.L. Best , C.E. Keevil , D.R. Parsons ,
E.W. Prokockid, C.J. Simpsonb
a
Department of Geography, University of Exeter, Exeter, EX4 4QJ, UK
b
Division of Geography and Geology, School of Environment and Technology, University of Brighton, Brighton, Sussex, BN2 4GJ, UK
c
Departments of Geology, Geography and Geographic Information Science, Mechanical Science and Engineering and Ven Te Chow Hydrosystems Laboratory, University of
Illinois at Urbana-Champaign, 1301 W. Green St, Champaign IL, 61801, USA
d
Department of Geology, University of Illinois at Urbana-Champaign, 1301 W. Green St, Urbana, IL, 61801, USA
e
Department of Geography, Environment and Earth Sciences, University of Hull, Hull, HU6 7RX, UK

A R T I C LE I N FO A B S T R A C T

Keywords: The transition between riverine and estuarine environments is characterised by a change from unidirectional to
Sediment transport bidirectional flows, in a region referred to herein as the Tidally-Influenced Fluvial Zone (TIFZ). In order to
Tidal constituent improve our understanding of the hydrodynamics and morphodynamics of this zone, we present a combined
Residual flow field and numerical modelling study of the Columbia River Estuary (CRE), USA, tidally-influenced fluvial zone.
Tidal-fluvial interactions
The CRE is large measuring 40 km in length and between 5 and 10 km wide. A shallow water model (Delft3D)
Columbia river estuary
Model parameterisation
was applied in both 2D and 3D configurations and model sensitivity to the key process parameterizations was
investigated. Our results indicate that a 2D model constrained within the estuary can sufficiently reproduce
depth-averaged flow within the TIFZ of a stratified estuary.
Model results highlight the interactions between tidal-, fluvial- and topographic-forcing that result in depth
dependent tidal rectification, and thus zones of residual sediment transport that: i) may be flood-directed along
shallow channel margins and in the lee of bars, and simultaneously ii) is ebb-directed within deeper channel
thalwegs. This condition is enhanced at lower discharges, but increased fluvial discharge reduces the number
and size of regions with net flood-directed sediment transport and flow. These sediment transport patterns
provide a mechanism to extend the bar/island topography downstream, and generate flood-directed, ebb-di-
rected, and symmetrical bedforms, all within the same channel. Analysis of the model data reveals flood-directed
sediment transport is due to both tidal variability and mean flow. These results highlight the need to include the
mean flow component (M0) when considering the long-term morphodynamic evolution in a TIFZ.

1. Introduction estuary, and the mixed tidal-fluvial regime (downstream boundary of

freshwater tidal river reach; middle TIFZ), where tidal-flows are
Tidally-influenced fluvial zones (TIFZs) represent some of the most strongly impacted by varying river discharges, bed geometry/bathy-
complex environments on Earth (Dalrymple and Choi, 2007), and ex- metry, and water density (Jay and Smith, 1988). These factors drive
tend from the mouth of an estuary, or delta, through the freshwater asymmetric hydrodynamic flows with either flood- or ebb- dominance
tidal river reach, whose landward limit terminates at the most upstream (Dyer 1997, chapter 2), and associated net sediment transport (Guo
point of observable variations in water surface elevations caused by et al., 2014), and are referred to as tidal rectification (cf. Huthnance
tidal-flows (Hoitink and Jay 2016). TIFZs are commonly sub-divided 1973; Bowers and Al-Barakati 1997).
into three general hydraulic zones defined by the mean ratio of tidal vs The hydrodynamics and morphodynamics in tidally–dominated es-
fluvial energy input, and include (see Jay et al., 1990; Jablonski et al., tuaries have been well studied (cf. Guo et al., 2014; Zhou et al., 2014;
2016): i) tidally-dominated lower river (i.e., estuary), ii) tidally-domi- Chu et al., 2015; Guo et al., 2016), with net sediment transport related
nated, fluvially-influenced, or mixed tidal-fluvial regime, and iii) flu- to: i) tidal velocity asymmetries (Van Maren et al., 2004; Díez-Minguito
vially-dominated, tidally-influenced regime. Thus, it is at the boundary et al., 2012; Guo et al., 2014); ii) the generation of the M4 overtide
between the tidally-dominated lower TIFZ, or brackish water upper (Friedrichs and Aubrey, 1988); and iii) the triad interaction M2-O1-K1

∗
Corresponding author. Met Office, FitzRoy Road, Exeter, Devon, EX1 3PB, United Kingdom.
E-mail address: steven.sandbach@metoffice.gov.uk (S.D. Sandbach).

https://doi.org/10.1016/j.ecss.2018.06.023
Received 13 November 2017; Received in revised form 20 June 2018; Accepted 25 June 2018
Available online 27 June 2018
0272-7714/ © 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/BY/4.0/).
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Fig. 1. Aerial photography of the lower Columbia River Estuary showing (A) modelling domain location and boundaries and, (B) ADCP cross-section locations. Model
grid resolution has been reduced by a factor of 8 for clarity. Aerial photographs (2009) provided courtesy of the United States Department of Agriculture National
Agriculture Imagery Program.

(Hoitink et al., 2003). The focus of recent work on the middle to upper to apply a 2D model to the TIFZ of the CRE. Herein, we test this by
TIFZ includes suspended sediment dynamics (Achete et al., 2016), in- applying both 2D and 3D simulations with the model boundary located
teractions between mean and tidal flow components (Van Maren et al., both inside and outside the estuary.
2004; Chu et al., 2015), net water transport (Alebregtse and de Swart, Despite this recent interest, little is known about TIFZ channel scale
2016), and determination of the relative importance of non-stationary hydrodynamics and morphodynamic processes that determine how
river vs stationary tidal energy signals in governing streamwise fluc- channels and sand bars evolve, which ultimately are one of the im-
tuations in the water surface elevation (Matte et al., 2013; Guo et al., portant sedimentological building blocks of estuaries and deltas
2015; Jay et al., 2015), and/or flow velocity magnitudes and orienta- (Dalrymple and Choi, 2007; Leuven et al., 2016). Therefore, to in-
tions (Losada et al., 2017). vestigate TIFZ channel to barform evolution, a combined field and
A number of process-based hydraulic models have been applied to numerical modelling study was conducted on the meso-tidal lower
study the Columbia River Estuary (CRE) (Kärnä and Baptista, 2016), Columbia River (LCR), WA/OR, USA (Prokocki et al., 2015).
including the Regional Ocean Modelling System (Liu et al., 2009), EL- Thus, in this paper three principal questions are addressed:
CIRC (Zhang et al., 2004), SELFE (Zhang and Baptista, 2008) and
Delft3D (Elias et al., 2012). The predominant focus of these model 1. Can a 2D model be used to simulate accurately the LCR TIFZ (i.e.,
applications has been on plume dynamics and estuary mouth processes. estuary) where density stratification is present?
Because of this focus, the impact of stratification was important and 2. What are the dominant characteristics of LCR TIFZ channel scale
thus these models were all deployed in 3D mode. hydrodynamics and morphodynamics, and how do these processes
Modelling studies of other estuaries include 3D (e.g. Burla et al., affect bedform to barform evolution?
2010) and 2D (e.g. Bolle et al., 2010; Prario et al., 2011) representa- 3. How do fluctuations between low- and high- river flow impact LCR
tions. In most cases, the modelling domains include the estuary mouth TIFZ bedform to barform evolution?
and so apply a coastal boundary condition. Notable exceptions are
Prario et al., (2011) and Matte et al., (2017a, 2017b), who constrained To address these points, a modelling study of the LCR TIFZ in-
their models to within the estuary. This suggests that it may be possible cluding different levels of process representation to evaluate the

177
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

impacts of waves, 3D/2D, boundary location, and bedform roughness 6% of the profile were also removed in processing, to remove the effect
and turbulent viscosity, is developed herein. Details of the modelling of contamination by side-lobe interference. Due to the large length
setup are provided in section 3. The model that performed best was scales and short time-scales for survey data collection, it was not pos-
assessed by calibration, using measured water level data, and validation sible to take repeat measurements for stable flow conditions. Therefore,
with other hydrodynamic data (sections 4.1 and 4.2). Results from the in this paper, velocity data are depth-averaged to remove non-sta-
best performing model were then used to investigate channel-scale tionary turbulent signals.
hydrodynamics and morphodynamics within the LCR TIFZ (section The average boat velocity during the surveys was 1.09 m s−1, pro-
4.3). Due to the large variability in river-discharge, the effect of flow viding high-resolution (∼6.3 measurements m−2) velocity data along
discharge on model calibration/validation and channel scale process each cross-section. Seven transects were surveyed during the high river-
was also analysed. discharge period (field seasons 1 and 3). A subset of these transects
were re-surveyed during the low river-discharge period (field-season 2).
2. Columbia River estuary The locations of all the surveyed transects are illustrated in Fig. 1. The
transects surveyed at low river flow are marked as dashed lines. The
The Columbia River is the largest river entering the Pacific Ocean data were processed using the VMT Matlab toolbox (Parsons et al.,
from North America, with an average river-discharge of ∼7300 m3 s−1 2013). Raw and processed data files are available at: http://fluvialtidal.
(Jay et al., 2010). The discharge is seasonally variable with a typical com.
minimum autumn-early spring flow of ∼3000 m3 s−1 and a spring The field survey collection periods ranged from 4 to 8 h per de-
freshet between 10,000 and 15,000 m3 s−1. The tide is mixed diurnal ployment with an average of 5 h depending on the prevailing condi-
and semidiurnal with a tidal range of 3.6–4 m (Fain et al., 2001). Within tions. Each transect was typically surveyed at least five times. The ty-
the Columbia River littoral cell, the wave climate is seasonally-variable pical time for each survey was around 30 min but smaller channels
and high-energy with significant wave heights Hs = 3 m and peak could be crossed in 10 min while others took approximately an hour to
period Tp = 12 s in the winter season (Gelfenbaum and Kaminsky, survey. Processing of ADCP data using a time-interpolation method
2010). A deep thalweg (navigation channel), which is periodically produced no perceptible difference in the spatial distribution of velo-
dredged, meanders through the central portion of the estuary flanked to city.
the north and south by embayments (Fig. 1).
3. Modelling simulations
2.1. Data availability
3.1. Numerical model
Bathymetric and topographic measurements from various sources
and time periods are available offshore and within the estuary. The
Both 2D and 3D Delft3D-FLOW (Deltares, 2011) models were setup
National Oceanic and Atmospheric Administration (NOAA) collected
to simulate the Columbia River TIFZ. The horizontal momentum
Multi-Beam Echo Sounding (MBES) data predominantly within the
equations include forcing associated with: i) Coriolis force (model la-
Navigation Channel stretching from the mouth to river-mile 18 (2007),
titude ϕ = 46.1881°); ii) horizontal and vertical kinematic eddy visc-
river-mile 18 to 25 (2008), river-mile 25 to 28.5 (2011) and from river-
osity νv and νH; iii) water-surface and bed shear stresses; and iv) pres-
mile 28.5 to Beaver Army Terminal (2009). Areas outside the
sure gradient forces expressed through the application of a hydrostatic
Navigation Channel were surveyed using Single-Beam Echo Sounding
pressure assumption.
(SBES) by the United States Army Corps of Engineers (USACE) in 2003
Free surface and bedform shear stresses are included using the
and 2009. Shallow areas inaccessible to the USACE vessels were sur- → →
veyed using SBES in 2009 and 2010 by the Lower Columbia River quadratic friction law: τb = ρref Cd Uref Uref , where τb is the shear stress,
→
Estuary Partnership (LCREP). Exposed topography within the estuary is the reference density, Cd is a drag coefficient and Uref is the reference
was measured by Light Detection and Ranging (LiDAR) in 2005 by the velocity. At the free-surface, the air density and wind velocity are used
Puget Sound LiDAR Consortium (PSLC). Just outside the estuary mouth, in this formulation and Cd varying between 0.0025 and 0.0289 (Elias
NOAA (2007) and USACE (2009) conducted MBES bathymetric sur- et al., 2012). For bed roughness effects, a slightly modified version of
veys. For the remainder of the coastal zone, 20 km north and south and drag coefficient is used; the Chezy roughness parameter: C = g / Cd .
40 km west of the mouth, data collected by NOAA using a fathometer Also, in this case the water density and the near-bed (3D model) or
→
during the 1920s and 1950s are also available. depth-averaged (2D model) velocities are used (Uref ) in the shear stress
Gauging stations maintained by NOAA and CMOP (Coastal Margin parameterisation.
Observation & Prediction) are present throughout the system, collecting The viscosity associated with vertical shear is:
both hydrological and biological data. Hydrological data monitored in νv = νmol + max(ν3D, νvback ) , where νmol is the molecular fluid viscosity,
the estuary include water surface elevation, salinity, temperature, wave ν3D is the 3D eddy-viscosity and νvback is the vertical background eddy-
climate and discharge. The data used in this paper were obtained from viscosity. In the horizontal, the viscosity is: νH = νv + νHback + νSGS ;
CMOP (http://www.stccmop.org), NOAA (http://www.noaa.gov) and where νHback is the user specified horizontal background eddy-viscosity
USGS (http://waterdata.usgs.gov). The locations of the gauging stations and νSGS is the horizontal Sub-Grid-Scale (SGS) eddy-viscosity. The SGS
are shown in Fig. 1A. terms are computed using the Uittenbogaard HLES (Horizontal Large
Eddy Simulation) model (Deltares, 2011, pp. 532–537). A similar re-
2.2. Flow surveys lationship holds for the diffusive terms (Dv and DH) (Deltares, 2011, pp.
203). In the present paper, the background viscosity and diffusion are
Flow data were obtained using a Teledyne RD Instruments 1200 kHz defined following Elias et al., (2012) whereby νHback = DHback
ADCP, deployed from a small inflatable boat, which was set to de- = 1 m2 s−1 and νvback = Dvback = 10−6 m2 s−1. The 3D viscosity and
termine three-dimensional flow in a vertical column at 0.25 m interval diffusion (ν3D and D3D ) are computed using the k-ε turbulence model
bins at a rate of ∼1 Hz. The ADCP was deployed in moving-boat mode (Deltares, 2011, pp. 231–233). The cyclic advection scheme is used for
with vessel position and velocity correction provided via Real Time both momentum and scalar transport.
Kinematic GPS. The distance from the water surface to the first ADCP In some of the simulations (Table 1), the hydrodynamic model was
measurement bin was 0.91 m, which accounted for submergence of the coupled with Delft3D-WAVE (SWAN version 40.72ABCDE). The model
ADCP below the water-line and the blanking distance that removes the parameterisation of Elias et al., (2012) is used herein, including an
effect of acoustic ringing (Szupiany et al., 2007). Data from the bottom overlapping WAVE grid.

178
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Table 1 3.2.3. Parameters

Modelled time periods and configurations assessed including a summary of the A series of model simulations were conducted to calibrate and assess
optimal roughness parameters. model sensitivity. This assessment focused on: bed roughness; waves;
Modelling periods and mean fluvial discharge model dimensionality; horizontal viscosity; and downstream boundary
location.
Field season Mean fluvial discharge (m3 s−1) Model period The roughness values were specified spatially variable with: uni-
form roughness in the lower estuary and coastal domain (Csea); uniform
1 15,700 27/05/2011–27/06/2011
2 4200 15/09/2011–15/10/2011 roughness in the river (Criv); and a linear variation in the intervening
3 11,500 29/05/2012–29/06/2012 region (see Fig. 1A for transition points). This follows similar ap-
proaches by others (Baptista et al., 2005; Sassi et al., 2011; Elias et al.,
Configurations 2012; Van et al., 2016). The range of values assessed was based on the
values used in other studies of the CRE (Baptista et al., 2005; Elias et al.,
Case 2D/3D fx2 (m2s-1) Waves Csea (m1/2 s-1) Criv (m1/2 s-1)
2012).
1 3D 1.0 – 70 60 A series of simulations were conducted using the baseline para-
2 3D 1.0 Included 70 60 meters described in section 3.2 for FLOW and FLOW-WAVE config-
3 2D 1.0 – 60 60
urations. A further three model configurations were considered for the
4 2D 1.0 Included 60 60
5 2D 0.1 – 60 60
2D estuary-boundary simulations where the horizontal background
6 2D 10.0 – 60 65* viscosity is varied (0.1, 10 m2 s−1 and HLES). For each of these con-
7 2D HLES – 60 65 figurations, one-month long simulations were run (Table 1).

3.2. Model application 4. Results

3.2.1. Domains 4.1. Model calibration and sensitivity

Two domains were created (see Fig. 1A). The first covered an area
similar to the modelling domain used by Elias et al., (2012), and in- The Mean Tidal Amplitude difference MTAd = (MTAm - MTAo) and
cluded the Columbia River (∼30 km long and ∼3 km wide), Estuary Mean Water Level difference MWLd = (MWLm - MWLo) are used to
(∼40 km long and ∼5–10 km wide) and coastal zone (∼40 km per- calibrate the models. The subscript 'm' and 'o' indicate model predic-
pendicular and parallel to the shore). The second domain was a reduced tions and observations. Figure 2 shows filled contour plots of MWLd
version of the first, where the downstream boundary was moved from averaged across all available gauges for both field seasons two and
outside to inside the estuary (Fig. 1A). three. The dotted line corresponds to the zero-contour for MTAd
A variable resolution grid was used ranging from ∼80 m in the (=MTAd0). The optimal configuration is obtained where the zero-
estuary and up to 1000 m in the coastal region. In the 3D model si- contour for both MWLd (=MWLd0) and MTAd intersect. The 3D model
mulations, the resolution was reduced by a factor of two to facilitate over-predicts the MWL and so there is no zero-contour (Fig. 2A and B).
acceptable model run-times. The vertical resolution was greatest near The results indicate that the model is more sensitive to tidal amplitude.
the bed and free surface (4% of the depth) and lowest in the middle For both sets of the 3D simulations, Criv = 60 m½ s−1 and Csea = 70 m½
(17% of the depth), varying smoothly in-between over 10 layers. For s−1 provide optimal results (values are larger than Elias et al., 2012).
the remainder of this paper, these domains are referred to as the sea- The optimal roughness values are summarised in Table 1.
boundary and estuary-boundary domains respectively. The 3D sea-boundary model simulations (Fig. 2A and B) show that
2D and 3D models were run in the estuary- and sea-boundary do- the wave model increases the mean water level by around 0.025 m but
mains respectively. We initially tested a 2D model in the sea- boundary has a negligible effect on tidal amplitude. In the 2D estuary-boundary
domain but the performance was poor. We chose not to implement a 3D simulations (Fig. 2C and 2D) the wave model has a negligible effect but
model in the estuary-boundary domain because it was not possible to viscosity can lead to a significant shift in the tidal water levels (Fig. 2C
obtain accurate salinity boundary conditions. and F).
An inversion of the roughness sensitivity of the MWL metric to the
sea roughness parameter is evident for the 3D model but not the 2D
3.2.2. Boundary conditions model (not shown). The mean high and low water levels (Fig. 3A–D)
The models were forced using discharge and water level boundary show that the model is sensitive to the applied Csea at high water during
conditions. Water level data from Hammond tide gauge was used for low river-discharge, but insensitive during high-river-discharge (Fig. 3A
the estuary-boundary simulations. An astronomical water level was and B). The reverse is true for the mean low water level (Fig. 3C and D).
specified at the western boundary of the domain in the sea-boundary The response of the model at high and low water levels is related to
simulations. The amplitude and phase of the nine components used flood and ebb processes, and in particular to the work done by external
were obtained from Elias et al., (2012). forces. Considering only the bed shear forcing (F ∝ U 2) , the work done
The Local Mean Sea Level (LMSL) was specified using a conversion W (=FUdt) may be computed and partitioned into ebb (e) and flood (f)
from NAVD88 to LMSL provided by NOAA's VDatum software (vdatum. periods:
noaa.gov). The water level boundary was supplemented with two zero-
wd = (We − Wf )/(We + Wf ), (1)
normal water level gradients (Neumann) at the north and south
boundaries (Fig. 1A). Time-series discharge observations obtained from There is an increase in wd in the upstream direction, as well as a
the Beaver Army Terminal gauging station were used for the river clear difference in the work done during the periods of high and low
boundary. This data includes the effects of both river- and tidal- forcing. river-discharge. During low river flow, the work done in the flood
The wind climate was specified using measured data from NOAA period dominates in the lower to middle part of the estuary, corre-
(NDBC buoy 46,029; see Fig. 1A for location) and the same para- sponding to the increased sensitivity to roughness at high water levels
meterisation as Elias et al., (2012). (flooding process). During high river flow, the ebb work-done dom-
The salinity at the river and the sea boundaries during inflow were inates for most of the estuary, corresponding to the increased sensitivity
set equal to 0 and 33, respectively (Elias et al., 2012). A Thatcher- to roughness at low water levels (ebb process).
Harleman time-lag (Deltares, 2011) of 120 min was applied to provide A direct comparison between 2D estuary-boundary and 3D sea-
smoothing. boundary results (Fig. 4) is presented relative to the Hammond water

179
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Fig. 2. Filled contour plots of MWLd averaged across all available gauges and river flow periods (field-seasons 2 and 3). Red dotted line is the zero-contour for MTAd.
(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

level data. These results demonstrate that the 2D estuary-boundary using Reduced Major Axis (RMA) linear regression (compare with the
model performs best and this is most noticeable at the Skamokawa tide 1:1 line of equality). In Fig. 5E, the RMA gradient for each transect and
gauge. model configuration is shown. The average of the absolute difference
between 1 and the computed gradient ( 1 − m ) across all transects is
shown in Fig. 5F. The degree of scatter is quantified here using the
4.2. Model validation correlation coefficient r (Fig. 5G). The mean differences between a
perfect correlation coefficient of 1 and the actual correlation coeffi-
The salinity data were assessed using the Root Mean Square Error cients (1 − r ) are presented in Fig. 5H. These results show that the 2D
(RMSE), and the model skill using the Index of Agreement (Willmott, estuary-boundary model with uniform viscosity νHback = 1 m2 s−1 per-
1981). The results from this analysis are not presented here but they forms best.
demonstrate: i) neither model configuration (FLOW or FLOW-WAVE)
systematically out-performed the other; ii) the model skill is higher in
4.3. Hydrodynamics and morphodynamics in the TIFZ
the lower part of the estuary, a result also observed by Elias et al.,
(2012); and iii) the present model performs to a similar level of accu-
The 2D estuary-boundary model results (from configuration 3 in
racy as other modelling studies of the CRE (Elias et al., 2012; Kärnä
Table 1) were used to investigate the channel scale hydrodynamics and
et al., 2015; Kärnä and Baptista, 2016).
morphodynamics in the TIFZ of the CRE. This is the best performing
Although it is possible to obtain the non-stationary amplitude and
configuration based on our analysis. To begin, small regions (1.1 and
phase (Matte et al., 2013), herein we use the t_tide Matlab toolbox
1.2, Fig. 1) are used to investigate temporal flow patterns, total sedi-
(Pawlowicz et al., 2002) since the analysis window is one month. The
ment transport and the contribution of tidal and fluvial forcing. These
2D and 3D configurations (case 1 and 3: Table 1) and observations were
results are then used to generalise the results in the upper part of the
analysed (Table 2).
CRE (region 2, Fig. 1).
Both model configurations reproduce the tidal amplitudes to a si-
milar level of accuracy: the average and maximum amplitude errors
being 0.01 m and 0.05 m respectively. The 2D estuary-boundary model 4.3.1. Temporal flow patterns in region 1.1
performs slightly better at reproducing the tidal phase with an average/ The flow patterns during a high discharge period (field season 3,
maximum phase error of around 3/5° compared to 7/12° for the 3D sea- Table 1) where the flow changes from ebb to flood in region 1.1
boundary configuration. The results also show that the amplitudes are (Fig. 1B) are presented in Fig. 6. This region includes the Grassy Island
damped by increasing river discharge for M2, S2 and O1 but amplified ADCP transect (Fig. 1B). Measured and modelled data from three time
for K1. The components that are affected most are K1 and S2. periods are shown in Fig. 6G–I.
We have also computed the model skill and absolute error (Table 3). In Fig. 6A, the velocities are ebb-directed and approximately at
The results show that the RMSE is relatively large for the 3D case. This maximum magnitude. During the next 2 h, the velocity magnitude is
is surprising given the amplitude and phase differences between these reduced (Fig. 6B), and 1 h later (Fig. 6C) the flow remains pre-
configurations is small. This is due to poor representation of the MWL dominately ebb-directed, although some of the flow has reversed and a
as discussed in section 4.1 and demonstrated in Fig. 2A and B. recirculation zone develops.
In Fig. 5A–D, a direct comparison between the modelled and mea- In the following hour, nearly all the flow has reversed (Fig. 6D). In
sured velocity magnitude is given for four ADCP transects with error the Prairie Channel (Fig. 1B), the flow is flood-directed in the shallow
bars for variability. The degree of bias is illustrated using the line fitted areas but remains ebb-directed in the deeper part of the channel. This

180
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Fig. 3. (A–D) Mean high and low water levels predicted by the 3D FLOW-WAVE model. (E) Proportion of the total work done during flood and ebb calculated using
Equation (1). Vertical dash-line denotes Hammond tide gauge. Positive wd indicates ebb is dominant, and negative wd indicates flood is dominant (Fig. 3E).

181
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Fig. 4. Difference between predicted and measured mean water level.

bidirectional flow pattern is exhibited in both the measured and mod- Table 3
elled data (Fig. 6G) and the recirculation zone is still evident (Fig. 6D). Index of agreement and root mean square error computed for the low (field
The recirculation is driven by opposing ebb-directed flow from the season 2) and high (field season 3) discharge periods.
Prairie Channel and flood-directed flow from downstream. Thirty Gauge IOA RMSE (m)
minutes later (Fig. 6E), the flow in this area is entirely ebb-directed,
with the velocity magnitude in the shallow areas being much greater field-season 2 field-season 3 field-season 2 field-season 3
than in the deeper part of the channel. The flow profiles illustrate this
2D 3D 2D 3D 2D 3D 2D 3D
also (Fig. 6H). Finally, 1 h later (Fig. 6F and I), the flow velocities in the
deeper part of the channel are now greater than in the shallow areas. Ham – 0.979 – 0.979 – 0.22 – 0.23
The flow in this region is affected by both tidal and fluvial flows, as Ast 0.998 0.985 0.998 0.979 0.07 0.20 0.07 0.24
well as a clear influence of bed topography. Analysis of the results Ska 0.998 0.964 0.998 0.982 0.06 0.28 0.06 0.18

showed that the flow duration is flood-biased in some areas of this

region, typically along the margins of the channel and in the lee of
directed flow: d = U ·Us / U ·Us , where U and Us are the 2-dimensional
Grassy Island (with respect to the ebb direction). The same general flow
horizontal velocity vectors from the unsteady and the steady model
patterns are observed during the low discharge period although the
configurations respectively.
magnitudes of the ebb flow velocities and duration are reduced.
The direction parameter was then used to scale the velocity mag-
nitude to separate ebb- and flood- directed flow velocities and time-
4.3.2. Sediment transport in region 1.2 averaged across a lunar cycle (denoted using angled braces). The results
The influence of tidal and fluvial forcing, combined with bathy- of the analysis (Fig. 7A and B; vector length does not represent mag-
metric variability, may lead to mean flood-directed flow. This is de- nitude), show that regions of ebb-directed net flow exist with velocities
monstrated for a slightly larger area (region 1.2, Fig. 1B) than outlined greater than 0.04 m s−1 (Fig. 7A). Zones of flood-directed net flow are
in the previous section. To compute the mean flow direction, the model observed in the lee of Grassy Island and along the channel margin
was run in a steady configuration using mean water level height and (Fig. 7A). Increasing the river-discharge reduces the extent of these
discharge at the model boundaries. This was used to define a directional zones (Fig. 7B) and increases maximum tidal-averaged velocity (from
parameter d (= ± 1) to denote ebb- (d = +1) or flood- (d = −1)

Table 2
Modelled and measured tidal amplitude and phase for the low (field-season 2, roman) and high (field-season 3, italics) discharge periods.
Gauge Amplitude (m)

M2 K1 S2 O1

Station Obs 2D 3D Obs 2D 3D Obs 2D 3D Obs 2D 3D

Ham 0.95 – 0.95 0.31 – 0.29 0.30 – 0.32 0.25 – 0.24
0.91 – 0.95 0.49 – 0.49 0.17 – 0.20 0.24 – 0.23
Ast 1.01 0.97 1.02 0.30 0.29 0.30 0.30 0.30 0.32 0.23 0.23 0.24
0.96 0.91 0.97 0.47 0.46 0.46 0.16 0.16 0.18 0.22 0.21 0.21
Ska 0.88 0.89 0.92 0.26 0.24 0.24 0.24 0.27 0.29 0.18 0.19 0.19
0.79 0.79 0.79 0.37 0.38 0.36 0.13 0.13 0.13 0.15 0.16 0.15

Gauge Phase (Degrees)

M2 K1 S2 O1

Station Obs 2D 3D Obs 2D 3D Obs 2D 3D Obs 2D 3D

Ham 75 – 69 192 – 186 27 – 21 112 – 104
234 – 228 126 – 120 41 – 31 346 – 339
Ast 87 89 81 198 201 192 41 44 34 115 120 108
247 250 243 133 137 130 61 63 49 357 1 356
Ska 114 112 106 220 217 209 72 69 61 134 135 125
279 280 276 153 158 154 93 92 80 34 32 34

182
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Fig. 5. (A–D) Comparison between measured ADCP and predicted model (2D FLOW νHback = 1 m2 s−1) velocities for four transects. Error bars represent one standard
deviation from the within model-cell variability. (E & F) Gradient m obtained from RMA least-squares regression, (G & H) correlation coefficient.

0.18 m s−1 to 0.29 m s−1). contours presented, except for some areas close to the channel margin.
These results indicate that there is potential for net sediment During the low discharge period (Fig. 7C), sediment transport remains
transport in this region. In principle, the sediment transport rate may be principally ebb-directed. In the lee of Grassy Island, and in some mar-
computed directly using a sediment transport scheme. However, ap- gins of the Prairie Channel, there are regions of flood-directed sediment
plication of these schemes requires careful site-specific calibration. For transport with magnitudes greater than 5% of the local maximum.
simplicity, the Engelund and Hansen (1967) transport scheme is used These regions of flood-directed transport have a larger spatial extent
herein. A proxy for sediment transport is denoted with a lowercase s (= than the tidal-averaged velocity.
U5). For the remainder of this paper, this is referred to as the sediment
transport capacity. A similar approach has been adopted by others (e.g. 4.3.3. Tidal analysis of sediment transport in region 1.2
Hoitink et al., 2003; Van Maren et al., 2004). To understand the principal factors driving these flow and sediment
The sediment transport capacity was time-averaged over a lunar transport patterns, the flood- and ebb- directed velocity magnitude
tide. The results are presented (Fig. 7C and D) as a fraction of local were further analysed. The full signal is split into mean (U0) and tidally-
maximum tidal-averaged sediment transport capacity smax evaluated variable (UT) velocities: U = U0 + UT . To do this: the tidally-variable
within region 1.2 (vectors denote the time-averaged direction) from the part was computed using the t_tide Matlab toolbox (Pawlowicz et al.,
low discharge period. 2002) and the mean component obtained from the difference between
During the high discharge period (Fig. 7D), sediment transport is the total and the tidally-variable velocity (U0 = U − UT ). Using these
nearly entirely ebb-directed with only small regions of flood-directed reconstructed velocities, the mean (s0) and tidally-variable (sT) sedi-
transport, in particular in the lee of Grassy Island. In the Prairie ment transport capacities were computed.
Channel, the ebb-directed transport is greater than the range of The sediment transport capacities show that the tidally-variable part

183
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Fig. 6. Series of model and ADCP data showing the flow within a channel and across the Grassy Island transect (Fig. 1B) on the 18th June 2012 PST in region 1.1
(Fig. 1B).

(sT) is significantly larger than the mean part (s0) that is typically ebb- is presented for the upper to mid part of the estuary (region 2 – Fig. 1).
directed. In Fig. 7E and F, the tidally-variable sediment transport ca- Similar patterns are observed to those discussed previously. To compare
pacities are presented scaled by smax = 0.0576 m5 s−5. The tidally- these two ratios directly, the difference (RsT - Rs0) is calculated and
variable part (sT) is typically flood-directed with similar spatial dis- presented in Fig. 8C and D.
tribution for both high and low discharge periods. The largest sT values Analysis of the data showed that the dominant constitutes are O1
are aligned along the Prairie channel (denoted with dashed arrow) and M2 for the low and high discharge periods respectively. The second
where water depths are largest in region 1.2. largest constitute is K1 for both discharge periods. Hoitink et al., (2003)
To compare the total and reconstructed sediment transport capa- showed that the linear interaction between these components (K1, M2
cities, we define the following ratios: and O1) results in tidal asymmetry. They formulated analytical ex-
pressions to compute sediment transport (velocity raised to 3rd or 5th
RsT = sT / s ,
power). These formulations show that the long-term average sediment
Rs0 = s0/ s . (2)
transport depends only on the amplitude and phase of these constitutes.
Negative values indicate flood-directed sediment transport. To test these formulations in the CRE, we computed the sediment
The spatial distribution of RsT are similar for both flow periods transport (5th power) using the obtained tidal components and
(Fig. 7G and H). The sediment transport is flood-directed and large Hoitink's formulation. The sediment transport patterns are very dif-
compared to the total transport: i) along the north margin of the Prairie ferent to the sediment transport patterns computed using the time series
Channel and ii) in the layer between the Prairie Channel and the lee of data (not shown). In the CRE, the fluvial discharge is not negligible and
Grassy Island. thus the sediment transport capacity was also computed using the mean
Along the north margin of the Prairie Channel (i), the tidally-vari- flow (M0) combined with the same constitutes as Hoitink et al., (2003)
able sediment transport appears to be the principal mechanism that (K1, M2, O1, M4). The resultant sediment transport patterns (not shown)
results in flood-directed sediment transport. are very similar to the results obtained using the original time series
Across the layer between the Prairie Channel and the lee of Grassy (Fig. 8E and F). We also tested these results with the Van Maren et al.,
Island (ii), the total sediment transport capacity transitions from ebb- (2004) formulation and similarly found that the sediment transport
directed (within the Prairie Channel) to flood-directed (in the lee of (this time a 3rd power) computed using time series data and re-
Grassy Island) sediment transport. The large RsT magnitude in this re- constructed from components (M0, K1, M2, and O1) were very similar.
gion is a result of relatively small total sediment transport due to
competing tidal and mean flow. In the lee of the bar, the tidally-variable 4.3.5. Net sediment transport in region 2
and mean flow sediment transport capacities are flood-directed (not Calculated sediment transport capacities are presented In Fig. 8E
shown) and so the total sediment transport capacity is flood-directed. and F. The bed topography in this region is presented in Fig. 8G for
reference. For the area presented, smax = 0.5065 and 2.6324 m5 s−5 for
4.3.4. Tidal analysis of sediment transport in region 2 the low and high discharge periods respectively. The maximum sedi-
In Fig. 8A and B, the tidally-variable sediment transport ratio (RsT) ment transport capacities were obtained in the navigation channel in

184
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Fig. 7. Tidal averaged flow in region 1.2 defined in Fig. 1. (A & B) velocity and (C & D) total sediment transport, (E & F) tidally-variable sediment transport, (G & H)
tidally-variable sediment transport ratio. (< s > max = 0.0576 m5 s−5 local maximum during low river-discharge).

the upper part of the estuary. Finally, the ratio of sediment transport from the high and low river
In both the north and south embayments, most locations exhibit flow periods was computed (Fig. 8H). Sediment transport ratios greater
ebb-directed sediment transport around 0.05smax or less. In the re- than (less than) one indicate areas of increased (reduced) sediment
maining locations, small magnitude flood-directed and larger magni- mobility. The results show that this ratio is greater than one for most
tude ebb-directed sediment transport is observed. The larger magnitude (92%) of the area presented. The ratio is spatially-variable with the
ebb-directed transport is concentrated in the deeper scoured areas. largest values (∼5) being located near the head of the estuary, and
During low river flow periods (Fig. 8E), large regions of flood-directed reducing along the axis of the estuary seaward (∼2.5).
sediment transport are observed in the lee of the bars and islands.
Flood-directed sediment transport along the margins of the channels is
also evident.

185
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

Fig. 8. Tidal average sediment transport in area 2. (A & B) tidally-variable sediment transport ratio, (C & D) difference between sediment transport ratios, (E & F)
total sediment transport (< s > max = 0.5065 m5 s5 from the low discharge period), (G) bed level relative to NAVD88 and (H) ratio of the sediment transport from
high and low discharge periods.

5. Discussion rather than being explicitly resolved.

From analysis of salinity and water level data (not shown), we found
The model calibration results presented herein highlight an inver- that the 3D sea-boundary model performed to a similar level of accu-
sion of the 3D model roughness sensitivity to water level variations due racy compared to other models (Elias et al., 2012; Kärnä et al., 2015;
to a change in river-discharge. This was attributed to a change in the Kärnä and Baptita 2016) of the CRE. Our results also demonstrated that
amount of work done during the two discharge periods. However, the the 2D estuary-boundary configuration performed better than the 3D
2D model did not exhibit this behaviour because the bed shear stress is sea-boundary configuration in terms of calibration and validation sta-
related to the bulk momentum of the entire column in the 2D model tistics. This was based on analysis of observed water-level and flow

186
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

data. This may be due to modelling the more complex system and in- Furthermore, direct comparison between these parameters shows that
cluding coastal and mouth processes, and demonstrates the difficulties the tidally-variable ratio is dominant - in particular during the low
associated with modelling the CRE. One particular difficulty is related discharge period and along the channel margins.
to the specification of the seaward boundary. In the approach adopted Although Rs0 is generally small compared to RsT, the mean compo-
herein and by Elias et al., (2012), the sea boundary height is specified nent is crucial in determining the resultant sediment transport patterns.
using stationary mean water level. The associated error in the estima- This was observed in tests where the resolved constituents without the
tion of this height is 0.226 m (VDatum: http://vdatum.noaa.gov/docs/ mean flow (M0) were used to reconstruct the sediment-transport (not
est_uncertainties.html). In the 2D estuary-boundary simulations, the shown here). Furthermore, given K1, M2 and O1 are the dominant
water levels are specified using measured data, and the model boundary constituents in the CRE and have previously been linked to tidal
conditions are therefore more constrained. Our results suggest that, asymmetry (Hoitink et al., 2003), we also tested the following combi-
although stratification and vertical process representation are im- nation: K1, M2, O1, M4 plus M0. The spatial sediment transport patterns
portant, similar if not better model predictions can be obtained using a were computed by reconstructing the flow from the given constitutes.
2D model constrained within the estuary. The reconstructed sediment transport patterns are visually similar to
The differences in performance may also be due to running the 3D the total sediment transport, and demonstrates the importance of the
model at lower resolution. As a test, we ran additional simulations at mean flow in determining the sediment transport patterns in the CRE.
the lower resolution for the 2D case. Reducing the resolution results in This was also demonstrated by Chu et al., (2015) who found that the
slightly poorer calibration and validation statistics although the dif- mean flow can modify the temporal, and therefore long-term, transport
ferences are small (not shown). We therefore expect that the 3D model patterns.
calibration and validation statistics to be only marginally better if run The impact of river-discharge on the sediment transport patterns in
at higher resolution. the CRE is dominated by three factors. Firstly, due to topographic for-
The temporal and spatial flow data show that larger peak velocities cing, river-discharge is preferentially routed into deeper parts of the
are associated with the deeper areas of the channels, likely due to to- channels. This results in ebb-dominance in the deeper channels. Flood-
pographic forcing. The associated increase in momentum results in a dominance is still possible, and was demonstrated in the shallow and
greater lag in response to water-level change compared to compara- sheltered areas, but increasing river-discharge limits the spatial extent
tively slower flow in the shallower and more sheltered areas. This was of the flood-directed areas. Secondly, although the magnitude of the
demonstrated in an area of the south embayment (region 1.1). This mean flow can be small, its effect on determining the sediment trans-
results in net ebb-directed flow in the deeper channels and flood-di- port patterns is significant. Finally, the effect of increasing the dis-
rected flow into the shallower and more sheltered areas (region 1.2). charge leads to increased sediment transport.
Similar spatial distributions were obtained for net-sediment transport
capacities (region 1.2). Increasing the river-discharge results in a re- 6. Conclusions
duction in the spatial extent of the flood-directed flow and sediment
transport. Delft3D was applied to simulate the hydrodynamics of the Columbia
Although the sediment transport is principally ebb-directed, there River Estuary, with testing of two modelling configurations and para-
are areas of net flood-directed sediment transport, which has two im- meter sensitivity. The models were calibrated using water level data
plications for sedimentological processes. Firstly, in the lee of islands and validated using salinity and flow data. Overall, the model performs
and bars, flood-directed flow provides a source of sediment that could very well at reproducing the measured water levels and flow patterns.
subsequently be deposited, and a mechanism to extend bar/island to- Our results show that a 2D estuary-boundary model can be used to
pography downstream. Secondly, flood-directed flow may produce provide a good representation of planimetric hydrodynamics within a
flood-directed bedforms. If these bedforms are preserved in bar/island highly energetic and stratified estuary.
stratigraphy, they could provide a signature that may help interpret the A detailed description of the flow within an area of the south em-
origin and evolution of depositional forms in the TIFZ. However, their bayment illustrates the complex interactions between hydrodynamic
preservation potential is also linked to aggradation rate, and thus in forcing of the tidal flood with the momentum of the ebb-directed flow
regions of large amplitude gross-, but low amplitude net-, sediment during the turn of the tide. Flow reversals occur first in areas sheltered
transport, the bedforms produced are likely to be more symmetrical. by bar/island topography and along the margins of the channels,
Such a situation is observed at the margins of Prairie Channel and in the thereby producing planimetric bidirectional flows and recirculation
lee of Grassy Island during high river-discharge. At the same time, large zones.
ebb-directed sediment transport will likely produce ebb-directed bed- These interactions result in net sediment transport that is principally
forms. This yields the potential for spatial variations in the degree of ebb-directed. However, within the channels and sheltered areas, flood-
bedform symmetry and therefore the potential difficulty in associating directed flows and sediment transport are evident. The results indicate
certain bedform assemblages with a downstream position in the fluvial- that this is due to topographic forcing that focuses most of the river-
tidal zone (Venditti et al., 2012). discharge into deeper parts of the channels. Increasing the river-dis-
To better understand the principal driving mechanisms for flood- charge results in a reduction of the extent of the flood-directed flow and
and ebb-directed sediment transport, the tidal signal was split into sediment transport. Further analysis of the results showed that: i) flood-
tidally-variable and mean parts. To compare the influence of each, two directed sediment transport along the channel margins is due to tidal
parameters were used: RsT and Rs0 (Equation (2)). The tidally-variable variability, and ii) flood-directed sediment transport in the lee of bars is
part (RsT) is principally flood-directed, large compared to Rs0 and spa- due to both tidal variability and mean flow.
tially variable. In particular, large negative (flood-directed) RsT is ob-
served in two locations: i) along the margin of the channels where tidal Acknowledgements
variability is the main mechanism for flood-directed sediment transport
(Rs0 is typically ebb-directed); and ii) in areas where the principal net We are grateful to CMOP, USGS, NOAA, USACE, LCREP and PSLC
sediment direction changes (in the lee of bar topography), this high- (all fully referenced within the paper) for supplying bathymetric and
lighting areas where the mean part is also negative. The converse is true hydrological data critical to the model application. Calculations for this
for the mean part Rs0 (not shown). In region 2, Rs0 is fairly uniform and study were performed using the University of Exeter Supercomputer.
generally small compared to < s > during the low discharge period. This work was supported by the Natural Environment Research Council
However, during the high discharge period, Rs0 exhibits greater varia- (NERC) on a split grant award [grant numbers: NE/H007954/1, NE/
bility increasing in magnitude towards the head of the estuary. H006524/1, NE/H007261/1 and NE/H00582X/1]. Fieldwork was

187
S.D. Sandbach et al. Estuarine, Coastal and Shelf Science 212 (2018) 176–188

made possible by access to the extensive facilities at the Clatsop in and Calcification of Shallow, Stratified Estuaries, Physical Processes in Estuaries.
Community College Marine and Environmental Research and Training Springer Verlag, Berlin, pp. 21–41.
Jay, D.A., Giese, B.S., Sherwood, C.R., 1990. Energetics and sedimentary processes in the
Station; we thank in particular António Baptista, Michael Wilkin and Columbia River estuary. Prog. Oceanogr. 25 (1–4), 157–174. https://doi.org/10.
Katie Rathmell for their help in servicing equipment, providing access 1016/0079-6611(90)90006-N.
Jay, D.A., Zaron, E.D., Pan, J., 2010. Initial expansion of the Columbia River tidal plume:
to machine workshops, training in marine health and safety and pro- theory and remote sensing observations. J. Geophys. Res. 115https://doi.org/10.
vision of a Humber Class RIB. Rob Strick and Greg Sambrook Smith 1029/2008JC004996. C00B15.
helped obtain some of the field data. Finally, we thank Pat Killion for Jay, D.A., Leffler, K., Diefenderfer, H.L., Borde, A.B., 2015. Tidal-fluvial and estuarine
processes in the lower Columbia River. I: along-channel water level variations, Pacific
his navigation skills on the Tansy Point and essential local expertise in Ocean to bonneville dam. Estuar. Coast 38 (2), 415–433. https://doi.org/10.1007/
the ever-changing conditions in the Columbia River Estuary. s12237-014-9819-0.
Kärnä, T., Baptista, A.M., 2016. Evaluation of a long-term hindcast simulation for the
Columbia River estuary. Ocean Model. 99, 1–14. https://doi.org/10.1016/j.ocemod.
References 2015.12.007.
Kärnä, T., Baptista, A.M., Lopez, J.E., Turner, P.J., McNeil, C., Sanford, T.B., 2015.
Numerical modeling of circulation in high-energy estuaries: a Columbia River estuary
Achete, F.M., van der Wegen, M., Roelvink, D., Jaffe, B., 2016. Suspended sediment dy-
benchmark. Ocean Model. 88, 54–71. https://doi.org/10.1016/j.ocemod.2015.01.
namics in a tidal channel network under peak river flow. Ocean Dynam. 66 (5),
001.
703–718. https://doi.org/10.1007/s10236-016-0944-0.
Leuven, J.R.F.W., Kleinhans, M.G., Weisscher, S.A.H., van der Vegt, M., 2016. Tidal sand
Alebregtse, N.C., de Swart, H.E., 2016. Effect of river discharge and geometry on tides and
bar dimensions and shapes in estuaries. Earth Sci. Rev. 161, 204–223. https://doi.
net water transport in an estuarine network, an idealized model applied to the
org/10.1016/j.earscirev.2016.08.004.
Yangtze Estuary. Continent. Shelf Res. 123, 29–49. https://doi.org/10.1016/j.csr.
Liu, Y., MacCready, P., Hickey, B.M., Dever, E.P., Kosro, P.M., Banas, N.S., 2009.
2016.03.028.
Evaluation of a coastal ocean circulation model for the Columbia River plume in
Baptista, A.M.Y., Zhang, A., Chawla, M., Zulauf, C., Seaton, E.P., Myers, J.K., Wilkin, M.,
summer 2004. J. Geophys. Res. C Oceans 114https://doi.org/10.1029/
Burla, M., Turner, P.J., 2005. A cross-scale model for 3D baroclinic circulation in
2008JC004929. C00B04.
estuary-plume-shelf systems: II. Application to the Columbia River. Continent. Shelf
Losada, M.A., Díez-Minguito, M., Reyes-Merlo, M.Á., 2017. Tidal-fluvial interaction in the
Res. 25 (7–8), 935–972. https://doi.org/10.1016/j.csr.2004.12.003.
Guadalquivir River Estuary: spatial and frequency-dependent response of currents
Bolle, A., Wang, Z.B., Amos, C., De Ronde, J., 2010. The influence of changes in tidal
and water levels. J. Geophys. Res.: Oceans 122 (2), 847–865. https://doi.org/10.
asymmetry on residual sediment transport in the Western Scheldt. Continent. Shelf
1002/2016JC011984. .
Res. 30 (8), 871–882. https://doi.org/10.1016/j.csr.2010.03.001.
Matte, P., Jay, D.A., Zaron, E.D., 2013. Adaptation of classical tidal harmonic analysis to
Bowers, D.G., Al-Barakati, A., 1997. Tidal rectification on drying estuarine sandbanks.
nonstationary tides, with application to river tides. J. Atmos. Ocean. Technol. 30 (3),
Estuaries 20 (3), 559–568. https://doi.org/10.2307/1352614.
569–589. https://doi.org/10.1002/2014JC009791.
Burla, M., Baptista, A.M., Casillas, E., Williams, J.G., Marsh, D.M., 2010. The influence of
Matte, P., Secretan, Y., Morin, J., 2017a. Hydrodynamic modeling of the st. Lawrence
the Columbia River plume on the survival of steelhead (Oncorhynchus mykiss) and
fluvial estuary. I: model setup, calibration, and validation. J. Waterw. Port, Coast.
Chinook salmon (Oncorhynchus tshawytscha): a numerical exploration. Can. J. Fish.
Ocean Eng. 143 (5). https://doi.org/10.1061/(ASCE)WW.1943-5460.0000397.
Aquat. Sci. 67 (10), 1671–1684. https://doi.org/10.1139/F10-083.
04017010.
Chu, A., Wang, Z.B., de Vriend, H.J., 2015. Analysis on residual coarse sediment transport
Matte, P., Secretan, Y., Morin, J., 2017b. Hydrodynamic modeling of the st. Lawrence
in estuaries. Estuar. Coast Shelf Sci. 163, 194–205. https://doi.org/10.1016/j.ecss.
fluvial estuary. II: reproduction of spatial and temporal patterns. J. Waterw. Port,
2015.06.003.
Coast. Ocean Eng. 143 (5). https://doi.org/10.1061/(ASCE)WW.1943-5460.
Dalrymple, R.W., Choi, K., 2007. Morphologic and facies trends through the fluvial-
0000394. 04017011.
marine transition in tide-dominated depositional systems: a schematic framework for
Parsons, D.R., Jackson, P.R., Czuba, J.A., Engel, F.L., Rhoads, B.L., Oberg, K.A., Best, J.L.,
environmental and sequence-stratigraphic interpretation. Earth Sci. Rev. 81 (3–4),
Mueller, D.S., Johnson, K.K., Riley, J.D., 2013. Velocity Mapping Toolbox (VMT): a
135–174. https://doi.org/10.1016/j.earscirev.2006.10.002.
processing and visualization suite for moving-vessel ADCP measurements. Earth Surf.
Deltares, 2011. Delft3D-FLOW: Simulation of Multi-dimensional Hydro- Dynamic Flows
Process. Landforms 38, 1244–1260. https://doi.org/10.1002/esp.3367.
and Transport Phenomena, Including Sediments—User Manual, Version: 3.15,
Pawlowicz, R., Beardsley, B., Lentz, S., 2002. Classical tidal harmonic analysis including
Revision 14499, Report, Delft, Netherlands.
error estimates in MATLAB using T-TIDE. Comput. Geosci. 28 (8), 929–937. https://
Díez-Minguito, M., Baquerizo, A., Ortega-Sánchez, M., Navarro, G., Losada, M.A., 2012.
doi.org/10.1016/S0098-3004(02)00013-4.
Tide transformation in the Guadalquivir estuary (SW Spain) and process-based zo-
Prario, B.E., Dragani, W., Mediavilla, D.G., D'Onofrio, E., 2011. Hydrodynamic numerical
nation. J. Geophys. Res. 117https://doi.org/10.1029/2011JC007344. C03019.
simulation at the mouths of the Parana and Uruguay rivers and the upper Rio de la
Dyer, K.R., 1997. Estuaries: a Physical Introduction, second ed. John Wiley and Sons,
Plata estuary: a realistic boundary condition. Appl. Math. Model. 35 (11),
Chichester, UK.
5265–5275. https://doi.org/10.1016/j.apm.2011.04.013.
Elias, E.P.L., Gelfenbaum, G., van der Westhuysen, A.J., 2012. Validation of a coupled
Prokocki, E.W., Best, J.L., Ashworth, P., Parsons, D.R., Sambrook Smith, G.H., Nicholas,
wave-flow model in a high-energy setting: the mouth of the Columbia River. J.
A.P., Simpson, C.J., Wang, H., Sandbach, S., Keevil, C., 2015. Mid to late Holocene
Geophys. Res.: Oceans 117 (C9). https://doi.org/10.1029/2012JC008105. C09011.
geomorphological and sedimentological evolution of the fluvial-tidal zone: lower
Engelund, F., Hansen, E., 1967. A Monograph on Sediment Transport in Alluvial Streams.
Columbia River, WA/OR, USA. In: Ashworth, P., Best, J.L., Parsons, D.R. (Eds.),
Teknisk Forlag, Copenhagen.
Fluvial-tidal Sedimentology, Developments in Sedimentology Series. Elsevier,
Fain, A.M.V., Jay, D.A., Wilson, D.J., Orton, P.M., Baptista, A.M., 2001. Seasonal and tidal
Amsterdam, pp. 193–226 ISBN 9780444635297.
monthly patterns of particulate matter dynamics in the Columbia River estuary.
Sassi, M.G., Hoitink, A.J.F., de Brye, B., Vermeulen, B., Deleersnijder, E., 2011. Tidal
Estuar. Coast 24 (5), 770–786. https://doi.org/10.2307/1352884.
impact on the division of river discharge over distributary channels in the Mahakam
Friedrichs, C.T., Aubrey, D.G., 1988. Non-linear tidal distortion in shallow well-mixed
Delta. Ocean Dynam. 61 (12), 2211–2228. https://doi.org/10.1007/s10236-011-
estuaries: a synthesis. Estuar. Coast Shelf Sci. 27, 521–545. https://doi.org/10.1016/
0473-9.
0272-7714(88)90082-0.
Szupiany, R.N., Amsler, M.L., Best, J.L., Parsons, D.R., 2007. Comparison of fixed- and
Gelfenbaum, G., Kaminsky, G.M., 2010. Large-scale coastal change in the Columbia River
moving-vessel flow measurements with an aDp in a large river. J. Hydraul. Eng. 133
littoral cell: an overview. Mar. Geol. 273, 1–10. https://doi.org/10.1016/j.margeo.
(12), 1299–1309. https://doi.org/10.1061/(ASCE)0733-9429(2007)133:12(1299).
2010.02.007.
Van, C.P., de Brye, B., Deleersnijder, E., Hoitink, A.J.F., Sassi, M.G., Spinewine, B.,
Guo, L., van der Wegen, M., Roelvink, J.A., He, Q., 2014. The role of river flow and tidal
Hidayat, H., Soares-Frazão, S., 2016. Simulations of the flow in the Mahakam riv-
asymmetry on 1-D estuarine morphodynamics. J. Geophys. Res.: Earth Surf. 119,
er–lake–delta system, Indonesia. Environ. Fluid Mech. 16 (3), 603–633. https://doi.
2315–2334. https://doi.org/10.1002/2014JF003110.
org/10.1007/s10652-016-9445-4.
Guo, L., van der Wegen, M., Jay, D.A., Matte, P., Wang, Z.B., Roelvink, D., He, Q., 2015.
Van Maren, D.S., Hoekstra, P., Hoitink, A.J.F., 2004. Tidal flow asymmetry in the diurnal
River-tide dynamics: exploration of nonstationary and nonlinear tidal behavior in the
regime: bed-load transport and morphologic changes around the Red River Delta.
Yangtze River estuary. J. Geophys. Res.Earth Surf. 120 (5), 3499–3521. https://doi.
Ocean Dynam. 54 (3–4), 424–434. https://doi.org/10.1007/s10236-003-0085-0.
org/10.1002/2014JC010491.
Venditti, J.G., Nelson, P.A., Minear, J.T., Wooster, J.K., Dietrich, W.E., 2012. Alternate
Guo, L., van der Wegen, M., Wang, Z.B., Roelvink, D., He, Q., 2016. Exploring the impacts
bar response to sediment supply termination. J. Geophys. Res.Earth Surf. 117https://
of multiple tidal constituents and varying river flow on long-term, large-scale es-
doi.org/10.1029/2011JF002254. F02039.
tuarine morphodynamics by means of a 1-D model. J. Geophys. Res. 121, 1000–1022.
Willmott, C.J., 1981. On the validation of models. Phys. Geogr. 2, 184–194. https://doi.
https://doi.org/10.1002/2016JF003821.
org/10.1080/02723646.1981.10642213.
Hoitink, A.J.F., Jay, D.A., 2016. Tidal river dynamics: implications for deltas. Rev.
Zhang, Y., Baptista, A.M., 2008. SELFE: a semi-implicit Eulerian-Lagrangian finite- ele-
Geophys. 54 (1), 240–272. https://doi.org/10.1002/2015RG000507.
ment model for cross-scale ocean circulation. Ocean Model. 21 (3–4), 71–96. https://
Hoitink, A.J.F., Hoekstra, P., van Maren, D.S., 2003. Flow asymmetry associated with
doi.org/10.1016/j.ocemod.2007.11.005.
astronomical tides: implications for the residual transport of sediment. J. Geophys.
Zhang, Y., Baptista, A.M., Myers, E.P., 2004. A cross-scale model for 3D baroclinic cir-
Res. C Oceans 108 (C10). https://doi.org/10.1029/2002JC001539.
culation in estuary–plume–shelf systems: I. Formulation and skill assessment.
Huthnance, J.M., 1973. Tidal current asymmetries over the norfolk sandbanks. Estuar.
Continent. Shelf Res. 24, 2187–2214. https://doi.org/10.1016/j.csr.2004.07.021.
Coast Mar. Sci. 1 (1), 89–99. https://doi.org/10.1016/0302-3524(73)90061-3.
Zhou, Z., Coco, G., Jimenez, M., Olabarrieta, M., van der Wegen, M., Townend, I., 2014.
Jablonski, B.V., Dalrymple, R.W., Marzo, M., 2016. Recognition of strong seasonality and
Morphodynamics of river-influenced back-barrier tidal basins: the role of landscape
climatic cyclicity in an ancient, fluvially dominated, tidally influenced point bar:
and hydrodynamic settings. Water Resour. Res. 50, 9514–9535. https://doi.org/10.
middle McMurray Formation, Lower Steepbank River, north-eastern Alberta, Canada.
1002/2014WR015891.
Sedimentology 63, 552–585. https://doi.org/10.1111/sed.12228.
Jay, D.A., Smith, B.S., 1988. In: Dronkers, J., van Leussen, W. (Eds.), Residual Circulation

188