Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
TU3.T10.2.pdf
1. Normal Modes Analysis
of RiverSonde Data in a
Tidal Channel
Calvin C. Teague, Donald E. Barrick
CODAR Ocean Sensors, Ltd.
Mountain View, CA
David Honegger
Oregon State University
Corvallis, OR
2. RiverSonde Description
• UHF (435 MHz, 70-cm wavelength) radar
• Bragg scattering from 35-cm wavelength waves
• Based on SeaSonde hardware
• 1 W transmit power
• MUSIC direction finding using 3-yagi antenna array
• 5–15 m range bins, 1° angle bins
• Typically installed on a river bank
5. Newport Experiment
• CODAR student grant program
• Installed in September 2010 at Newport, Oregon
• Channel connecting Yaquina Bay to Pacific Ocean
• Tidal flow between parallel jetties
8. Velocity Vector Estimation
• From a single site, find along- and cross-channel
components from least-squares fit to radials
• If 2 sites are available, find full vectors by combining
radial measurements from both sites
• From a single site, use Normal Modes Analysis of radial
measurements to infer full vectors
• For arbitrary boundary, numerical solution required
• For rectangular boundary, closed-form solution possible
• Useful for dynamic flow conditions like tidal reversals
9. Normal Modes Analysis
• Assume water incompressible
• Express horizontal flow in terms of velocity potentials and stream func-
tions
• Boundary conditions
– Zero normal flow at banks
– No impedance to tangential flow at banks
– Periodic boundary at ends of analysis region
−
→
• Horizontal surface velocity vector U
−
→
U = ∇ × [z (−Ψ) + ∇ × (ˆΦ)]
ˆ z
where z is the vertical unit vector, Ψ is the stream function, and Φ is
ˆ
the velocity potential
• Allow up to 20 modes across river, only 2 along river
• Closed-form solutions in terms of sines and cosines
10. Homogeneous Equations
• Stream function satisfying Dirichlet condition at bank
∇2ψn + νnψn = 0, where ψn|Γ = 0
D , v D = −∂ψn , ∂ψn
un n
∂y ∂x
where ψn is the n-th eigenfunction of the stream function Ψ, νn is the
corresponding n-th eigenvalue and uD and vn are the velocity compo-
n
D
nents in the x and y directions, respectively.
• Velocity function satisfying Neumann condition at bank
∂φn
∇2φn + µnφn = 0, where (ˆ · ∇φn) =
λ =0
Γ ∂λ Γ
N , v N ] = ∂φn , ∂φn
[un n
∂x ∂y
where φn is the n-th eigenfunction of the velocity potential Φ, µn is the
corresponding n-th eigenvalue and ˆ is the direction perpendicular to
λ
the boundary.
15. Mode Coefficients
Determination
• Evaluate model in terms of unknown mode coefficients
at each point where radar data are available
• At each point, equate sum of radial components of
model to radial radar measurement
• Repeat over all available radar measurements
• Solve overdetermined set of equations for mode
coefficients (~5000 equations in ~50 unknowns) using
least-squares
• Allow up to 20 modes across river for along-river
component (mmax), only 2 along river for both along- and
cross-river components (jmax)
16. Streamline Examples
NWPT_2010_12_07_0600 NWPT_2010_12_07_0945
2.0 2.0
300 300
1.5 1.5
250 250
y m
y m
200 200
1.0 ms 1.0 ms
150 150
100 0.5 100 0.5
50 50
200 100 0 100 200 200 100 0 100 200
0.0 0.0
x m x m
NWPT_2010_12_07_1300 NWPT_2010_12_07_1530
2.0 2.0
300 300
1.5 1.5
250 250
y m
y m
200 200
1.0 ms 1.0 ms
150 150
100 0.5 100 0.5
50 50
200 100 0 100 200 200 100 0 100 200
0.0 0.0
x m x m
17. Mode Limits
NWPT_2010_12_07_1530
2.0
300
1.5
250
u: jmax = 1, mmax = 5
y m
200
1.0 ms
v: jmax = 0, mmax = 2 150
100 0.5
50
200 100 0 100 200
0.0
x m
NWPT_2010_12_07_1530
2.0
300
1.5
250
u: jmax = 1, mmax = 20
y m
200
1.0 ms
v: jmax = 0, mmax = 2 150
100 0.5
50
200 100 0 100 200
0.0
x m
18. Lagrangian Particle Trajectories
• Compute velocity vectors at 5-minute intervals
• Seed study area with 100 particles randomly placed
every 2 minutes
• Integrate particle velocity in 10-second steps
• Display 10 locations of particles with lighter color for
older positions
• Movie covers 2.5 hours around a tidal reversal
20. Summary
• For an arbitrary boundary, Normal Modes solution must be
found numerically
• For the special case of a rectangular boundary, with no normal
flow across banks and periodic continuation at open
boundaries, a closed-form solution can be found as a series of
products of sines and cosines
• Least-squares fit of radial components of Normal Modes to
radar radial velocity vectors gives coefficients
• Lagrangian visualization of particle trajectories may be useful in
dynamic conditions like tidal reversals
• Future studies
• Compare this 2D fitting to 1D radial data with ADCP or other
in-situ measurements, especially during flow reversals
• Determine how many modes are meaningful