3.2 Data processing
Velocity data retrieved from the AD2CPs were converted into ENU system
(i.e., \(v_{E}\), \(v_{N}\),\(v_{U}\) for East-, North-, and Up-ward
velocity, respectively), whereas ADCP data were already recorded in ENU
format. Raw ENU data were then imported to Matlab (version 2019b) for
initial quality control to remove noise generated by the interference of
water bubbles, large suspended particles, echo intensity, and other
disturbance factors (Lan et al., 2019). The procedure used for quality
control is a modified version of Guerra and Thomson (2017)’s algorithm.
For the data acquired using AD2CPs, values of echo intensity ≥ 25 dB and
correlation magnitude ≥ 30% are used as threshold limits for
high-quality data; for the data acquired by ADCP, values of echo
intensity ≥ 30 dB and correlation magnitude ≥ 50% are instead used as
thresholds. Velocity data were then despiked using the Phase-Space
Thresholding Method (Goring & Nikora, 2002) and eventually averaged
over the time length of individual bursts (i.e., 5 minutes). Moreover,
ADCP data were also averaged vertically over 4 successive bins to allow
for a more direct comparison with the AD2CP data. Overall, a total of 21
bins of velocity data were obtained for each instrument. For each bin,
the horizontal velocity \(\overrightarrow{v}\) was calculated as the
vector sum of the eastward (\(\overrightarrow{v_{E}}\)) and northward
(\(\overrightarrow{v_{N}}\)) velocity components (see Figure 5a) whereas
depth-averaged velocities (DAVs ) were computed for each measuring
station as the average value of the whole \(\overrightarrow{v}\)profile. Measured values of upward velocities
(\(\overrightarrow{v_{U}}\)) were instead maintained unaltered. Water
depth (\(Y\)) data were also obtained from pressure sensors integrated
within the instruments. Based on the surveyed topographic profile of
each cross-section (Figure 4b), we were able to identify the water depth\(Y_{B}\) corresponding to bankfull conditions. This allowed us to
differentiate velocity data recorded for water levels higher and lower
than the bankfull threshold (i.e., above- and below-bankfull water
stages). Stage-velocity diagrams were also obtained based on binary
plots of water levels (\(Y\)) and depth-averaged values of flow velocity\((DAVs\)) at each monitoring station (Figure 6 a-d). BothDAVs and \(Y\) were also put in relation to the rates
water-level change \(\dot{Y}\) = dY/dt (Figure 6 e-p).
Tidal asymmetries were investigated based on two distinct metrics,
concerning the asymmetry in flood vs. ebb peak tidal velocities and
flood vs. ebb durations, respectively (Figure 5 e-g). Since the flow
velocity is a function of the water depth, the peak tidal velocity index
(\(\rho_{v}\)), which is the ratio between the flood and ebb peak of\(\left|\overrightarrow{v}\right|\) (Friedrichs & Aubrey, 1988; Guo
et al., 2019), was calculated at different water depths (i.e., at
different positions along the water column) at 20 cm intervals. To
further differentiate between flow dynamics within the channel and
outside of it, distinct calculations of \(\rho_{v}\) were performed
averaging results by considering only velocity values measured at water
depths (\(Y\)) smaller and larger than the bankfull depth (\(Y_{B}\)),
respectively (Figure 5e,f). In contrast, asymmetries in tidal duration
(\(\rho_{d}\)) were computed as the ratio between the duration of the
falling and rising limb of the tidal wave (Friedrichs & Aubrey, 1988;
Guo et al., 2019). Both \(\rho_{d}\) and \(\rho_{v}\) provide a
straightforward tool to differentiating flood-dominated
(\(\rho\)>1) and ebb-dominated (\(\rho\)<1) tidal
flows.
To simplify the interpretation of velocity-data time series and filter
out outliers, data were phase averaged and subdivided into two distinct
groups based on the values of the high-tide water depth (\(Y_{H}\))
observed during each individual tidal cycle. Specifically, tidal cycles
for which \(Y_{H}\)>3.7 m (i.e., the sixth to thirteenth
tidal cycle in Figure 5) were classified as “high-amplitude tides”
(HAT ), whereas all the other tidal cycles were considered “low
amplitude tides” (LAT ) (Tu et al., 2019; Voulgaris & Meyers,
2004; Wang et al., 2013). For each tidal cycle, the instant
corresponding to \(Y_{H}\) was assigned the time value of \(t\)=0. Then,
data collected six hours before and after \(Y_{H}\) were
ensemble-averaged at five-minute intervals (Figure 7a,b)
Finally, in order to better investigate flow structures and unravel
possible secondary (i.e., cross-sectional) circulations, velocity data
were reprojected into two different components, namely, the primary
(i.e., streamwise) velocity \(V_{P}\), corresponding to the main
direction of in-channel tidal flows, and the secondary (i.e.,
cross-sectional) velocity \(V_{S}\), oriented orthogonally to \(V_{P}\)(Bever & MacWilliams, 2016; Finotello, Ghinassi, et al., 2020; Lane et
al., 2000). In order to define the directions of \(V_{P}\) and\(V_{S}\), previous studies have typically taken advantage of
reprojection techniques based on flow data recorded along the entire
channel cross-section by ADCP instruments mounted on moving vessels
(Finotello, Ghinassi, et al., 2020; Lane et al., 2000; Parsons et al.,
2013). These techniques cannot however be applied to our data, since our
instruments were operated in stationary mode, and significant
differences appear when observing flow velocities at above- and
below-bankfull stages. Thus, we assumed that the direction of \(V_{P}\)corresponds to the direction of the maximum horizontal velocity
(\(\overrightarrow{v_{\max}}\)) observed at the bottom vertical layer
(i.e., \(Y\) = 0.2 m in Figure 7c~f). Such a definition
is based on the observation that the orientation of\(\overrightarrow{v_{\max}}\) at the channel bottom is unequivocally
defined and remains consistent during both the ebb and flood phases (see
Figure 4 and Figure 7c~f). Once \(V_{P}\) is defined,
the orientation of secondary velocity (\(V_{S}\)) is immediately derived
as the direction perpendicular to \(V_{P}\). Details regarding the
determination of \(V_{P}\) and \(V_{S}\) at different measuring stations
can be seen in Figure 8,9,10 and 11. Close-up views of \(V_{S}\) vectors
for below-bankfull stages only are also shown in Supplementary Figure
S1.
4 Results