You are on page 1of 20

ARTICLE IN PRESS

Computers & Geosciences 34 (2008) 1243–1262


www.elsevier.com/locate/cageo

Calculating wave-generated bottom orbital velocities from


surface-wave parameters$
Patricia L. Wiberga,, Christopher R. Sherwoodb
a
Department of Environmental Sciences, University of Virginia, Charlottesville, VA 22904-4123, USA
b
Coastal and Marine Geology, U.S. Geological Survey, Woods Hole, MA 02543-1598, USA
Received 12 October 2006

Abstract

Near-bed wave orbital velocities and shear stresses are important parameters in many sediment-transport and
hydrodynamic models of the coastal ocean, estuaries, and lakes. Simple methods for estimating bottom orbital velocities
from surface-wave statistics such as significant wave height and peak period often are inaccurate except in very shallow
water. This paper briefly reviews approaches for estimating wave-generated bottom orbital velocities from near-bed
velocity data, surface-wave spectra, and surface-wave parameters; MATLAB code for each approach is provided. Aspects
of this problem have been discussed elsewhere. We add to this work by providing a method for using a general form of the
parametric surface-wave spectrum to estimate bottom orbital velocity from significant wave height and peak period,
investigating effects of spectral shape on bottom orbital velocity, comparing methods for calculating bottom orbital
velocity against values determined from near-bed velocity measurements at two sites on the US east and west coasts, and
considering the optimal representation of bottom orbital velocity for calculations of near-bed processes. Bottom orbital
velocities calculated using near-bed velocity data, measured wave spectra, and parametric spectra for a site on the northern
California shelf and one in the mid-Atlantic Bight compare quite well and are relatively insensitive to spectral shape except
when bimodal waves are present with maximum energy at the higher-frequency peak. These conditions, which are most
likely to occur at times when bottom orbital velocities are small, can be identified with our method as cases where the
measured wave statistics are inconsistent with Donelan’s modified form of the Joint North Sea Wave Project (JONSWAP)
spectrum. We define the ‘‘effective’’ forcing for wave-driven, near-bed processes as the product of the magnitude of forcing
times its probability of occurrence, and conclude that different bottom orbital velocity statistics may be appropriate for
different problems.
r 2008 Elsevier Ltd. All rights reserved.

Keywords: Ocean waves; Bed shear stress; Continental shelf; Wave boundary layer; MATLAB

1. Introduction
$
Code available from server at http://www.iamg.org/CGEditor/ Wind-generated waves on water induce orbital
index.htm. motion within the water column to a depth roughly
Corresponding author. Tel.: +1 434 924 7546;
fax: +1 434 982 2137.
equal to half the wavelength of the surface waves.
E-mail addresses: pw3c@virginia.edu (P.L. Wiberg), When water depth is less than half the wavelength,
csherwood@usgs.gov (C.R. Sherwood). this wave-induced orbital motion extends to the

0098-3004/$ - see front matter r 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cageo.2008.02.010
ARTICLE IN PRESS
1244 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262

bed. Interactions between wave-generated water 2. Bottom orbital velocity from simple linear
motion and the bed influence surface waves through wave theory
frictional dissipation of wave energy, and influence
the bed through mobilization of bed sediment by Interactions between surface waves and the
wave-motion-induced bed shear stresses. Mobiliza- seabed are most conveniently expressed in terms of
tion of bed sediment can alter bottom roughness, the wave-induced orbital fluid motion close to the
e.g., through formation or elimination of small-scale bed. For small-amplitude, monochromatic waves,
ripples, which in turn modifies wave energy dissipa- linear wave theory predicts that the horizontal
tion and bottom drag for bottom boundary layer component of orbital velocity, u0, evaluated at
flows. Wave energy dissipation, bottom roughness, z ¼ h (the bottom) is
and sediment transport are important parameters
Hp
in many hydrodynamic and morphodynamic u0 ¼ cosðkx  otÞ (1)
T sinhðkhÞ
models for the coastal ocean, lakes, and estuaries.
Proper representations of these parameters depend where H is the wave height (twice the amplitude, a,
on accurate estimates of wave-induced flow near of sea-surface fluctuations), T is the wave period, h
the bed. is the water depth, k is the wavenumber (2p/L,
Theoretical relationships have been derived for where L is wavelength), o is the radian frequency
wave-generated, near-bed flow under low-ampli- (2p/T), x is the coordinate in the direction of wave
tude, monochromatic waves. These have been propagation, and t is the time; the vertical compo-
extended by Madsen (1994) to the more typical nent of orbital velocity approaches zero as z-h.
case when a spectrum of waves is present, provided u0 varies sinusoidally through a wave period, reach-
the spectrum is known. Problems arise, however, ing a maximum velocity, commonly termed the
when only summary statistical parameters like bottom orbital velocity ub, when j cosðkx  otÞj ¼ 1:
significant wave height and peak period are avail- Hp oa
able, rather than the full spectrum, as is frequently ub ¼ ¼ (2)
T sinhðkhÞ sinhðkhÞ
the case in archived wave data, model output, and
hindcasts. In these cases, orbital velocities can be Bottom orbital velocities are directly proportional
estimated by coupling significant wave height and to wave height and depend inversely on water depth
peak period with a generic spectral form to (Eq. (2)). The dependence of ub on wave period is
approximate the contribution of the full wave subtler, however, because period enters Eq. (2)
spectrum to flow at the bed (e.g., Soulsby, 1987). through k (¼ 4p2 =½gT 2 tanhðkhÞ, where g is the
In this paper, we review the methods for gravitational acceleration) as well as T; solving for k
calculating bottom orbital velocities from spectral requires an iterative solution (see Appendix A).
surface-wave data and from near-bed velocity Fig. 1 shows ub as a function of depth for periods
measurements. Then we consider methods for ranging from 4 to 14 s. The calculations were done
estimating orbital velocities using surface-wave assuming H ¼ 1 m (local wave height); for other
statistics (significant wave height and peak period) values of H, ub should be multiplied by H. Because
to parameterize a general spectral form. In applying the term T sinh(kh) decreases with increasing period,
this parametric method, we consider two common longer period waves generate larger bottom orbital
spectral forms that differ in shape. The accuracy of velocities. The increase in ub with increasing T
the parametric method is confirmed by comparing becomes more pronounced as water depth increases.
bottom orbital velocities calculated using this In shallow water, the differences are smaller because
method with values calculated from full spectra orbital velocities become
pffiffiffiffiffiffiffi
ffi dominantly a function of
and from direct measurements of near-bed velocity depth (ub ¼ a g=h when sinh(kh)Ekh).
for continental shelf sites on the US east and west Small-amplitude wave theory assumes the bed is
coasts. MATLAB codes for these methods are frictionless. Orbital velocities calculated using this
provided in appendices; test cases are provided as theory are in good agreement with observed oscilla-
Supplementary material. We conclude with a dis- tory flows under monochromatic waves except very
cussion of uncertainties in bottom orbital velocity close to the bed where friction is important. Within
calculations and of how to choose an appropriate the thin layer of frictional influence (the wave
bottom orbital velocity statistic for calculations of boundary layer), velocities increase with height
near-bed processes such as sediment transport. above the bottom from 0 to u0. Associated with
ARTICLE IN PRESS
P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262 1245

T=4s
squared (rms) or significant wave height and some
6 measure of average and/or peak (dominant) wave
8 period over a given sampling interval.
100 10 Significant wave height, Hs, has traditionally been
12 defined as the average height of the highest one-
ub (m/s)

14 third of waves during a sampling interval (Longuet-


Higgens, 1952; Dean and Dalrymple, 1991). Assum-
10-1 ing surface-wave heights generally follow a Rayleigh
distribution as Longuet-Higgens (1952) argued and
others have verified, this definition of significant
H = 1m wave height is equivalent to
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z ffi
10-2 pffiffiffiffiffiffi
100 101 102 H s ¼ 4 m0 ¼ 4 S Z ðf Þ df (4)
Water depth (m)

Fig. 1. Bottom orbital velocity, ub, as a function of water depth h where m0 is the variance of water surface elevation,
for periods ranging from 4 to 14 s. Calculations assume local Z, and SZ is the spectral density of surface elevation
wave height H ¼ 1 m. Assuming a deep-water wave height of 1 m
as a function of frequency, f ( ¼ 1/T) (e.g., Ochi,
and conserving energy flux as waves shoal would produce a
somewhat higher value of local H and ub. For values of wave 1998). Significant wave height defined in terms of m0
height other than 1 m, indicated values of ub should be multiplied is sometimes referred to as Hm0. The variance, m0,
by H. Breaks in curves represent depth-induced wave breaking: can be calculated either directly from a water-level
wave heights are limited to 0.78 h (Dean and Dalrymple, 1991). time series or from the integral of the surface
Curves end when h ¼ L/2. Longer period waves affect bed to
elevation spectrum.pAP discrete ffiapproximation to the
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
much greater depths than do those with short periods.
latter, i.e., H s ¼ 4 S Z;i Df i , where the summa-
tion is taken over all frequency bins in the spectrum,
these velocity gradients are shear stresses, which are is used by the US National Data Buoy Center
greatest at the seabed and decrease to 0 near the top (NDBC; www.ndbc.noaa.gov) to calculate signifi-
of the wave boundary layer (e.g., Jonsson and cant wave height from wave buoy measurements. pffiffiffi
Carlsen, 1967; Jensen et al., 1989). Maximum values For a Rayleigh distribution, H rms ¼ H s = 2.
of wave-generated bed shear stress, tbw, are com- There are several ways to define wave period from
monly parameterized in terms of ub and a friction time series or spectra of surface elevation. Zero-
factor, fw, as crossing period is defined as the interval between
rf w 2 consecutive times when surface elevation crosses
tbw ¼ u (3) zero going up or down. Average zero-crossing
2 b
periodpcan be estimated from wave spectra as
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where r is the fluid density (e.g., Jonsson, 1966; T z ¼ m0 =m2 , where m2 is the second moment of
R
Wiberg, 1995). ffi Wave shear velocity is defined as
pffiffiffiffiffiffiffiffiffiffiffi the wave height spectrum (m2 ¼ f 2 S Z ðf Þ df ); this
uw ¼ tbw =r. Although wave-generated shear ve- is the method NDBC uses to calculate average wave
locities and shear stresses vary throughout a wave period (Earle, 1996). Mean period is defined as
period in response to variations in u0, estimates of Tm ¼ m R 0/m1, where m1, the first moment of SZ is
wave-related turbulent mixing and sediment trans- (m1 ¼ fS Z ðf Þ df ). Peak or dominant period, Tp, is
port on time scales longer than a single wave period defined as the period corresponding to the fre-
are frequently based on tbw, the maximum stress quency band in the spectrum SZ with the most
during each wave cycle. energy. This is a less stable measure of wave period
than Tm or Tz, which are based on the integrated
3. Surface-wave statistics and spectral estimates wave field. Nevertheless, Tp is the most widely
of ub reported measure of spectral wave period and is
useful for locating the peak in the wave energy
Wind-generated waves are not monochromatic. spectrum.
Instead, a spectrum of wave frequencies and heights A spectrum of wave-induced orbital velocity at
are present. The characteristics of the surface-wave the bottom boundary can be calculated from a
field are often summarized in terms of root-mean- surface-wave spectrum SZ by applying Eq. (2) to
ARTICLE IN PRESS
1246 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262
P 2 P 1=2
each frequency band i of the wave spectrum: f bz ¼ f i Su;i Df i S u;i Df i . A MATLAB file
to calculate bottom orbital velocity and period
4p2 o2i
Su;i ¼ 2
SZ;i ¼ S Z;i (5) from a measured wave spectrum is provided in
T 2i sinh ðki hÞ sinh2 ðki hÞ Appendix B.
(e.g., Soulsby, 1987). From this, a representative
bottom orbital velocity, ubr, can be calculated by 4. Estimates of ub from near-bottom velocity
summing the contributions from each frequency measurements
component of Su,i,
X Like surface waves, measured near-bottom velo-
u2br ¼ 2 S u;i Df i (6) cities are also not monochromatic, but instead
i
R include motions that range from very low frequen-
in analogy to a2rms ¼ H 2rms =4 ¼ 2m0 ¼ 2 S Z df and cies (tides, wind-driven currents, shelf waves) to
recalling that ub is the amplitude of wave orbital very high frequencies (turbulence). Wave motions
velocity at the bed. Combining Eqs. (5) and (6) often dominate bottom velocity spectra over periods
yields ranging from a few seconds and to a few tens of
" #!1=2 seconds. Following Eq. (6), which relates bottom
pffiffiffi X 4p2 orbital velocity to the variance of wave-induced,
ubr ¼ 2 S Z;i Df i (7)
i T 2i sinh2 ðki hÞ near-bed velocity, we can estimate ubr from near-
bottom velocity measurements as
The bottom orbital velocity, ubr, given by Eq. (7), pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
which Madsen (1994) terms the representative ubr ¼ 2ðvarðu0 Þ þ varðv0 ÞÞ (10)
bottom orbital velocity, is the amplitude of a where u0 ¼ u  ū and v0 ¼ v  v̄ are components of
monochromatic wave orbital motion with the same instantaneous velocity with means (indicated by
variance as the full spectrum (Soulsby, 1987; overbars) removed.
Madsen, 1994); ubr is also equivalent to the rms Strictly speaking, Eq. (10) is only true for error-
value of bottom orbital velocity. Assuming a free measurements of inviscid flow, because instru-
Rayleigh distribution of bottom orbital velocity, ment noise, turbulence, and low-frequency motions
we can define significant bottom orbital velocity ubs, not associated with surface waves all contribute to
in analogy to significant wave height, as the variance in measurements. In our experience
" #!1=2 with acoustic Doppler velocimeters, these contribu-
pffiffiffi X 4p2
ubs ¼ 2ubr ¼ 2 S Z;i Df i tions are small (a few percent) when waves are
i T 2i sinh2 ðki hÞ present and the instruments are operating properly,
(8) but it is wise to investigate the velocity spectra to
verify this. We often use an alternative approach
The assumption that bottom orbital velocity follows for calculating ubr based on near-bottom velocity
a Rayleigh distribution and the best choice of spectra that closely parallels Madsen’s (1994)
bottom orbital velocity (ubr, ubs or some other formulae. This method estimates the spectra, Suu
statistic) for calculations of flow and sediment and Svv, for the u and v components of velocity
processes at the sediment-water interface are exam- using the Welch method (e.g., Press et al., 1992) and
ined in a later section. determines the representative bottom orbital velo-
Madsen (1994) defined the representative wave city ubr as
frequency, fr, corresponding to ubr as rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X
P ubr ¼ 2 S uv;i Df i (11)
f i S u;i Df i
i i
f br ¼ P (9)
Su;i Df i where Suv ¼ Suu+Svv is the combined horizontal
i
spectrum. As above,pffiffiwe ffi can define significant
We define the representative bottom wave period orbital velocity ubs ¼ 2ubr .
Tbr ¼ 1/fbr. This is consistent with the definition In practice, estimates from Eq. (11) match those
of wave period in terms of the first moment of from Eq. (10) to within a few percent. For example,
the wave spectrum (Tm above). Bottom frequency the mean ratio of ubr values from Eqs. (10) and (11)
or period can also be calculated based on the for 1403 hourly burst measurements in the Hudson
second moment, analogous to Tz above, giving Shelf Valley (discussed below) was 0.9999 and the
ARTICLE IN PRESS
P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262 1247

standard deviation was 0.022; 90% of the values fell wave motions while higher-frequency (short period)
between 0.965 and 1.033. The advantage of the wave motion may be completely attenuated before
spectral approach is that it can also be used to reaching the bottom. As a consequence, unless the
calculate representative near-bottom wave periods water is shallow with respect to the wavelength of
(Eq. (9)), to filter out spectral energy contributions the waves, no general, single measure of wave height
outside the frequency range for waves, and to and period used in Eq. (2) can capture the
estimate surface-wave heights as contribution of the full wave spectrum to bottom
0 11=2 orbital velocity.
A better solution to the problem of calculating
BX Suv;i Df i C
H s ¼ 4@  2 A (12) bottom orbital velocity when measurements of near-
2pf i coshðki zuv Þ
i
sinhðki hÞ bed velocity or surface-wave spectra are not
available is to approximate the surface-wave spec-
where zuv is the elevation (above the bottom) of the trum with a generic surface-wave spectrum using
velocity measurements. Noise at high wavenumbers known values of wave height and period, and then
can inflate the wave-height estimates, so only to use this estimated spectrum and Eq. (7) or (8) to
frequencies in a limited range are used (we typically calculate bottom orbital velocities; we will refer to
consider 0.02pfp0.2 Hz for water depths on the this as the parametric method. A number of general
order of 10 s of m; the range would shift to higher forms for wind-generated wave spectra have been
values in shallower water). This method will not proposed, including the Pierson and Moskowitz
resolve contributions to Hs from high frequencies (1964) spectrum, the Joint North Sea Wave Project
whose associated wave motions have attenuated
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi (JONSWAP) spectrum (Hasselmann et al., 1973),
before reaching the bottom (i.e., f 4 g=ð4phÞ). and modifications of each (e.g., Mitsuyasu et al.,
1980). The general form of these spectra can be
5. Calculating orbital velocity from Hs and Tp written as
!x
Direct measurements of wave spectra from ag2 f
S Z ðf Þ ¼
nearby surface-wave buoys are ideal for determining ð2pÞ4 f 5 f p
bottom orbital velocity time series for a site of 2 !4 3
interest when high-frequency near-bottom velocity f 5gexp½ðf f p Þ2 =ð2s2 f 2p Þ (13)
 exp4b
measurements are not available. However, the fp
number of such buoys and the length of their
records are limited. Wave models, such as Wave- where fp is peak frequency ( ¼ 1/Tp) and a, b, x, g,
Watch III (WW3; Tolman, 1999) and wave hind- and s are parameters that adjust the magnitude and
casts such as the Wave Information Study (WIS; shape of the spectrum. Depending on the formula-
frf.usace.army.mil/WIS) can provide wave data at a tion, the parameters are constant, functions of f/fp,
much larger number of locations, including specific or functions of external variables, typically dimen-
study sites, and for longer periods of time. Wave sionless fetch x̄ ¼ gx=U 2 (where x is fetch length
models like WW3 calculate wave spectra, but and U is mean wind speed 10 m above the water
archived, gridded model output often includes only surface) or dimensionless peak frequency f̄ p ¼
significant wave height, dominant period, and f p U=g (Table 1). Values of f̄ p ¼ 0:13 are consistent
dominant wave direction. The same set of para- with fully developed seas (Pierson and Moskowitz,
meters is available for the WIS hindcasts. 1964; Alves et al., 2003) while larger values of f̄ p
Estimates of bottom orbital velocity for wind- correspond to fetch-limited conditions (e.g., Massel,
generated waves made with Eq. (2) using H ¼ Hs or 1996); values of f̄ p significantly less than 0.13 are
Hrms and T ¼ Tp, Tm or Tz are generally less indicative of swell.
accurate than estimates made from full spectra. For the JONSWAP and Pierson–Moskowitz
Each combination of these produces a different spectra, x ¼ 0 so SZ(f)f5; other parameter values
value ub; H ¼ Hrms and T ¼ Tp are the best choices are given in Table 1. Battjes et al. (1987) showed
for estimating ubr (Soulsby, 1987). The increase in ub that Eq. (13) provided a statistically significantly
with increasing T (Fig. 1) means that low-frequency better fit to ocean-wave data when SZf4 (x ¼ 1 in
(long period) components of the wave spectrum Eq. (13)) rather than f5. Donelan et al. (1985)
contribute disproportionately to the bottom orbital proposed a modified form of the JONSWAP
ARTICLE IN PRESS
1248 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262

Table 1
Spectral parameter values for general spectral form given in Eq. (13), where f̄ p ¼ f p U=g and U is mean wind speed 10 m above water
surface

a b x g s

Pierson–Moskowitza 0.0081 5/4 0 1 –


JONSWAP constant gb 0:033f̄ p
0:67 5/4 0 3.3 0.7 for f/fpp1,0.9 for f/fp41
JONSWAP variable gc 0:033f̄ p
0:86 5/4 0 4:42f̄ p
0:43 0.7 for f/fpp1,0.9 for f/fp41
d 0:55 3
Donelan 0:0165f̄ p 1 1 6:5 þ 2:6 lnðf̄ p Þ for f̄ p X0:16 0:08 þ 0:0013f̄ p
1:7 for f̄ p o0:16

a
Pierson and Moskowitz (1964).
b
Hasselmann et al (1973).
c
Mitsuyasu et al. (1980).
d
Donelan et al. (1985).

spectrum using this scaling (Table 1). Fig. 2 ences in these two forms for fully developed wave
compares the shapes of the Donelan and JONS- conditions provide useful information about the
WAP spectra (Table 1) for fully developed sensitivity of orbital velocity calculations to the
(f̄ p ¼ 0:13; Fig. 2a) and fetch-limited (f̄ p ¼ 0:50; shape of the surface-wave spectrum.
Fig. 2b) wave conditions. Two versions of the Given Tp, U10 and discrete frequencies fi spanning
JONSWAP spectrum are considered: one with a an appropriate range of values (we let f/fp span
constant g ¼ 3.3 (Cg; Hasselmann et al., 1973) and the range 0.2–5, corresponding to T/Tp ¼ 5–0.2),
the other with a variable g (Vg; Mitsuyasu et al., Eq. (13) and Table 1 can be used to calculate SZ,i for
1980); g, termed the peak enhancement factor, is the any of the spectral forms included in Table 1; Hs can
ratio of the spectral density maxima of the then be found using Eq. (4), orbital velocities can be
JONSWAP and Pierson–Moskowitz spectra for calculated using Eq. (7) or (8), and bottom period,
the same wave conditions. All spectra in Fig. 2 Tbr can be calculated from Eq. (9). For the problem
have been normalized to a variance of 1 for we are interested in—calculating bottom orbital
comparison. velocity from surface-wave parameters—we know
The normalized JONSWAP-Cg spectrum is in- Tp and Hs, but not necessarily U10. As a result, we
dependent of f̄ p and serves as a useful point of can recast the spectrum given in Eq. (13) into one
reference. For the fetch-limited conditions, f̄ p ¼ that depends explicitly on Hs:
0:50 (Fig. 2b), the three spectra are very similar. For 4
!x
larger values of f̄ p , the peak of the Donelan and wH 2s f p f
SZ ðf Þ ¼
JONSWAP-Vg spectra increase relative to the 16 f 5 f p
JONSWAP-Cg spectrum, while they decrease for 2 !4 3
smaller values of f̄ p because of the dependence of g f 5gexp½ðf f p Þ2 =ð2s2 f 2p Þ
 exp4b (14)
on f̄ p . The differences among the spectra are fp
pronounced for fully developed wave conditions.
Fig. 2a shows that the Donelan spectrum is broader (Appendix C); similar forms have been presented
and has a much smaller peak (less than half the by Chakrabarti (1986) and Soulsby (1987). The
JONSWAP-Cg peak value); the JONSWAP-Vg parameter w, which rescales the spectrum given by
peak falls in between the other two. Of the three, Eq. (14) such that Eq. (4) is still valid, is a constant
the Donelan spectrum is most similar to the for the JONSWAP-Cg spectrum; for g ¼ 3.3,
Pierson–Moskowitz spectrum for fully developed w ¼ 3.28, as noted previously by Soulsby (1987).
seas (Massel, 1996; Alves et al., 2003). For the Donelan spectrum, w is a function of
In the comparisons that follow, we focus on the dimensionless peak frequency, f̄ p (Fig. 2c, Appendix
Donelan spectrum because it has been shown to C), because g and s depend on f̄ p . Values of f̄ p
have the form most consistent with observed wave consistent with w can be estimated from empirical
spectra, and the JONSWAP-Cg spectrum with expressions for a or dimensionless variance, e
g ¼ 3.3 because of its simplicity. The large differ- (Appendix C); g and s can be calculated using the
ARTICLE IN PRESS
P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262 1249

smaller values of f̄ p A MATLAB file to perform


4 Donelan these calculations is provided in Appendix D.
-
fp = 0.13 Bottom orbital velocities and periods calculated
JONSWAP-C γ
using the Donelan spectrum and JONSWAP
JONSWAP-V γ
S/m0

2 spectrum (g ¼ 3.3) are compared in Fig. 3 for a


range of peak periods and water depths. The
Donelan spectrum gives bottom orbital velocities
0
that are about 5–10% smaller than values obtained
0 1 2 3 using the JONSWAP spectrum except when peak
f/fp period becomes small enough that its associated
wave orbital pmotion
ffiffiffiffiffiffiffiffiffiffiffiffiffidoes not extend to the bottom
(ToT deep ¼ 4ph=g) and only the low-frequency
4
- part of the spectrum contributes to bottom orbital
fp = 0.50 velocity. Although the relative differences become
large for small periods, the corresponding bottom
S/m0

2 orbital velocities are quite small. Bottom wave


period is less sensitive than bottom orbital velocity
to spectral form (Fig. 3b).
0
0 1 2 3 6. Intercomparison of ub calculated from wave
f/fp spectra, wave statistics and near-bed velocity

6 To evaluate the various methods for calculating


bottom orbital velocity, particularly the parametric
4 method (based on Hs and Tp and a general spectral
form), we selected two sites for which simultaneous
χ

2
surface-wave buoy data, including wave spectra,
and near-bed velocity time series are available. One
site is at a depth of 60 m on the northern California
0
10-2 10-1 100 continental shelf near the Russian River, where
- long-period Pacific Ocean swell propagates across a
fp
narrow shelf. The other is a site in the Hudson Shelf
Fig. 2. Comparison of Donelan et al. (1985) spectrum and two Valley, in the apex of the mid-Atlantic Bight at a
JONSWAP spectra, one with constant g ¼ 3.3 (Cg , Hasselmann water depth of about 39 m. This site, like most of
et al., 1973) and one with variable g (Vg, Mitsuyasu et al., 1980), the US east coast, has a wide shelf and is often
for (a) fully developed and (b) fetch-limited wave conditions. dominated by locally generated waves caused by
Spectra are normalized to an area of 1; dimensionless peak
winds associated with subtropical depressions,
frequency f̄ p ¼ f p U=g. (c) Normalizing factor, w (Eq. (14),
Appendix C) as a function of f̄ p for three spectral forms. Larger frontal passages and occasional tropical storms.
values of f̄ p are typical of fetch-limited waves while values of Wave conditions have been recorded near the
0.13 are associated with fully developed waves; smaller values edge of the Russian River shelf by NDBC buoy
are indicative of swell. 46013 since 1981 (water depth ¼ 123 m). The
standard data reported for this and other NDBC
relationships in Table 1 once f̄ p is determined. stations include hourly significant wave height,
Several iterations through the relationships for w, f̄ p , average and dominant (peak) wave period, and
g and s yield a set of values consistent with the non-directional wave spectra; directional spectra are
Donelan spectrum and the specified values of Hs available for some sites for a more limited period of
and Tp. For swell-dominated waves (relatively small time (see www.ndbc.noaa.gov for a listing of wave
Hs and large Tp), f̄ p calculated using this approach buoy locations and available data). During the
can be quite low (50.13). Because w is constant winter of 1991, a Geoprobe bottom tripod deployed
(wffi2.0, Fig. 2c) and spectral shape becomes at a depth of 60 m, near the inner-mid-shelf
constant (not shown) for f̄ p o0:05, values of w, g transition on the Russian River shelf, measured
and s consistent with f̄ p ¼ 0:05 can be used for velocity at 4 levels in the bottom 1.2 m of the water
ARTICLE IN PRESS
1250 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262

1.3 1.3
h = 15m

ubr-Donelan / ubr-JONSWAP-Cγ

Tbr-Donelan / Tbr-JONSWAP-Cγ
1.2 h = 30m 1.2
h = 60m
1.1 1.1

1 1

0.9 0.9

0.8 0.8
5 10 15 20 5 10 15 20
Tp Tp

Fig. 3. Comparison of (a) bottom orbital velocities, ubr, and (b) bottom periods, Tbr, calculated with JONSWAP (Hasselmann et al., 1973)
and Donelan et al. (1985) spectra for a range of water depths and peak periods. Values of ubr and Tbr for a Donelan spectrum are shown
relative to values calculated using a standard JONSWAP spectrum with g ¼ 3.3. ffiffiffiffiffiffiffiffiffiffiffiffiffi of ubrDonelan/ubrJONSWAPCg increase when peak
pValues
period becomes smaller than threshold period for deep-water waves, T deep ¼ 4ph=g.

column (Wiberg et al., 1994). Velocities were for this 40-day period gives BSS ¼ 0.84 for ubrspec
measured once per second for 8.5 min every hour. compared to BSS ¼ 0.40 for ubrHrmsTp (Table 2).
Hourly representative bottom orbital velocity, The agreement is very good in this case even though
ubrmeas, was calculated using the 8.5-min records the buoy and the tripod are not located at the same
as indicated in Eq. (10). Buoy-recorded significant depth. Similarly good agreement has been observed
wave height, wave period and ubrmeas are shown for between NDBC buoy 46022 (located at a water
the period of the Geoprobe deployment in Fig. 4. depth of 509 m) and sites on the Eel shelf, northern
Estimates of bottom orbital velocity, ubrHrmsTp pffiffiffi , California (water depths of 55–70 m) and between
calculated using Eq. (2) with H ¼ Hrms (¼ H s = 2) NDBC buoy 46025 (at a depth of 860 m) and a 60-m
and T ¼ Tp are reasonably good (Fig. 4c, Table 2), site on the Palos Verdes shelf, southern California
except when Tpo8.8 s, the deep-water p wave thresh-
ffiffiffiffiffiffiffiffiffiffiffiffi
ffi (Wiberg et al., 2002). Because the California shelf is
old for a water depth of 60 m (T deep ¼ 4ph=g). On largely narrow and open, waves experience rela-
these occasions, ubrHrmsTp ¼ 0 whereas ubrmeas40 tively little modification as they traverse the outer
owing to contributions from the low-frequency and mid-shelf (depths 450 m). On broader shelves
portion of the wave spectrum characterized by wave such as those on the US east coast, or in shallower
periods greater than Tp. Mean Tp over the deploy- water, surface-wave buoy measurements closer to
ment was 11.6 s; mean Tz was 7.4 s. The combina- the site of interest are likely to be required for
tion of Hrms and Tp provided better agreement with comparable agreement.
ubrmeas than combinations that included Hs and Tz; Spectral estimates of representative bottom wave
Hs would be a better estimator of ubs. On only a few period, Tbrspec (Eq. (9)) overpredict Tbrmeas by an
occasions during the Geoprobe deployment was average of 0.8 s (Fig. 5a) but agree relatively well.
Tpo8.8 s, whereas Tzo8.8 s about half of the time, The skill (BSS) of the spectral prediction of bottom
resulting in frequent and sometimes extended period is 0.09, but increases to 0.48 if the mean
periods when estimates of ubr based on Tz are difference between the two time series is removed,
close to 0. indicating that the spectral calculations give the
Representative bottom orbital velocities, ubrspec, correct temporal variation in period.
calculated using Eq. (7) and NDBC wave spectral Representative bottom orbital velocities, ubrpar,
data from buoy 46013 for a depth of 60 m are in and bottom periods, Tbrpar, calculated for a depth
substantially better agreement with the Geoprobe of 60 m during the winter of 1991 using the
measurements (Fig. 5b) than are estimates based on parametric method with the Donelan et al. (1985)
Eq. (2). Using the Brier skill score (BSS, Brier, spectrum are shown in Fig. 6; we used Hs and Tp
1950) as a measure of the agreement between time series from NDBC buoy 46013 (Fig. 4) as
calculated and measured (ubrmeas) values of ubr input. Estimated values of dimensionless peak
ARTICLE IN PRESS
P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262 1251

Hs (m)
2

0
02/01 02/08 02/15 02/22 03/01 03/08
Tbr−meas
30 Tp
Tz
20
T (s)

10

0
02/01 02/08 02/15 02/22 03/01 03/08

ubr−meas
ubr−HrmsTp
0.6
ub (m/s)

0.4
0.2
0
02/01 02/08 02/15 02/22 03/01 03/08
1991

Fig. 4. Significant (a) wave height, Hs, and (b) average, Tz, and peak, Tp, period from NDBC buoy 46013, seaward of Russian River shelf,
northern California, during a period in winter of 1991 when a Geoprobe bottom tripod was deployed at a depth of 60 m. Dashed line in
(b) indicates minimum period needed for orbital motion to reach bed. Representative (b) bottom wave period, Tbrmeas, and (c) orbital
velocity, ubrmeas (Eq. (10)) calculated from Geoprobe near-bed velocity measurements are shown by red curves. (c) ubrmeas is compared
with orbital velocities calculated using Eq. (2) with Hrms and Tp (ubrHrmsTp).

Table 2
Brier skill scores (BSSa) comparing bottom orbital velocities calculated using measured spectra (ubr-spec) and parametric method (ubr-par)
for Donelan et al. (1985) and JONSWAP (Hasselmann et al., 1973) spectral forms (see Appendix D) with bottom orbital velocities
calculated from near-bed velocity measurements (ubr-meas) for a US west coast site (Russian River shelf) and east coast site (N.Y. Bight)

Representative orbital Near-bed velocity Measured wave Donelan spectrum JONSWAP Hrms and Tp using
velocities calculated data (Eq. (10)) spectra (Eq. (7)) ubr- (Eqs. (14) and (7), spectrum Eq. (2)
from: ubr-meas spec Table 1) ubr-par (Eqs. (14) and (7), ubr-HrmsTp
Table 1) g ¼ 3.3

Russian R. shelf, CA
Full record 1.00 0.835 (0.638)b 0.610 0.398
Convergent 1.00 0.838 0.696 0.668 0.454
conditionsc
Hudson Shelf Valley
Full record 1.00 0.875 (0.717)b 0.677 0.501
Convergent 1.00 0.890 0.768 0.731 0.558
conditionsc

BSS40 indicates that calculated values of ubr are a better predictor of measured values than is mean measured ubr; BSS ¼ 1 if agreement is
perfect. CalculatedP and measured ubr are P shown in Figs. 4–7.
a
BSS ¼ 1  ðubrmeas  ubrcalc Þ2 = ðubrmeas  ubrmeas Þ2 (Brier, 1950).
b
ubr-par in these cases was calculated using values of dimensionless peak frequency f̄ p obtained from a rather than dimensionless variance
e (see Appendix C).
c
These skill estimates were calculated just for times when Donelan parametric calculations converged (97% of record for Russian River
shelf and 89% of record for mid-Atlantic Bight). Times when calculations did not converge are indicated by black dots in Figs. 6 and 7.
ARTICLE IN PRESS
1252 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262

25 Tbr−meas
Tbr−spec
20

T (s)
15

10

5
02/01 02/08 02/15 02/22 03/01 03/08
0.8
ubr−meas

0.6 ubr−spec
ub (m/s)

0.4

0.2

0
02/01 02/08 02/15 02/22 03/01 03/08
1991

Fig. 5. (a) Comparison of Tbrmeas calculated from Geoprobe near-bed velocity measurements with bottom wave period, Tbrspec,
calculated for a depth of 60 m using Eq. (9) and spectra measured by using NDBC buoy 46013 on Russian River shelf, northern California,
during winter of 1991. (b) Comparison of representative bottom orbital velocity calculated from Geoprobe near-bed velocity
measurements, ubrmeas, with orbital velocity, ubrspec, calculated using Eq. (7) and spectral wave data.

25
Tbr−meas
20 Tbr−par
T (s)

15

10

5
02/01 02/08 02/15 02/22 03/01 03/08

0.8
ubr−meas
0.6 ubr−par
ub (m/s)

0.4

0.2

0
02/01 02/08 02/15 02/22 03/01 03/08
1991

Fig. 6. (a) Comparison of bottom wave period, Tbrpar, calculated from Donelan spectrum (Eqs. (14), Table 1), with Tbrmeas calculated
from Geoprobe near-bed velocity measurements. (b) Comparison of bottom orbital velocity, ubrmeas, calculated from Geoprobe near-bed
velocity measurements with orbital velocity calculated from Donelan parametric spectrum, ubrpar. Black dots correspond to times when
wave spectra were bimodal and calculations using a Donelan spectrum did not converge (see Appendix C), indicating that surface-wave
parameters were inconsistent with this unimodal spectral form.

frequency f̄ p for this record are small, consistent velocities calculated from the parametric spectra
with swell-dominated wave conditions; f̄ p 40:13 are similar to those calculated from measured
during only 6% of the record. Bottom orbital spectra and from Geoprobe near-bed velocity
ARTICLE IN PRESS
P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262 1253

measurements (BSS ¼ 0.70); the BSS for the Neither spectral form yields values of Tbrpar with
JONSWAP spectrum is only slightly smaller (Table 2). a skill score BSS40 relative to Tbrmeas. Factors
The black dots at the bottom of Figs. 6a and 6b that contribute to the discrepancy between Tbrpar
indicate times when the parametric method using and Tbrmeas include the blockiness of the buoy-
the Donelan spectrum with f̄ p calculated based on e reported values of Tp used in the parametric
(see Appendix C) did not converge, indicating that calculations and the presence of wave conditions
the wave conditions were not consistent with a during which Tp drops to values near the deep-water
simple, unimodal spectral form. During most of threshold (Tdeep ¼ 8.8 s; dashed line in Figs. 4b
these times, the measured spectra were bimodal, and 6a) but sufficient wave energy is present at
with the largest peak at a frequency higher than frequencies lower than the parametric spectra can
0.11 s1, the inverse of the threshold period for account for (though the calculations converge) to
deep-water waves (Tdeep ¼ 8.8 s). The other spectral produce higher values of Tbrmeas. These instances
forms (JONSWAP and the Donelan spectrum with are generally associated with low values of ubrmeas.
f̄ p calculated based on a (see Appendix C)) yield Wave conditions have been recorded in the mid-
values of ubrpar for these times, but they are Atlantic Bight at NDBC buoy 44025 since 1975
significantly smaller than ubrspec or ubrmeas. (40.241N, 73.171W, water depth 39 m). During the
because the disproportionately large contribution spring of 2006, a Sontek acoustic Doppler veloci-
of the lower-frequency part of the spectrum to ub meter (ADV) measured orbital velocities 0.6 mab at
and Tb is missing in parametric spectral estimates Hudson Shelf Valley site A (40.391N, 73.791W ), a
for these bimodal spectra. Thus, one advantage of 39-m site approximately 55 km west of buoy 44025
using the Donelan spectrum with f̄ p calculated (Fig. 7). Estimated values of f̄ p for this record are
based on e for the parametric method is that it larger than for the northern California shelf,
identifies wave parameters that are inconsistent with consistent with a greater proportion of locally-
a general unimodal spectral form. generated wave conditions; f̄ p 40:13 during 27% of

0.4 ubr−meas
ubr−spec
0.3
ubr−par
ubr (m/s)

0.2

0.1

0
04/13 04/18 04/23 04/28 05/03 05/08 05/13 05/18 05/23 05/28

25 Tbr−meas
Tbr−spec
20
Tbr−par
15
Tbr (s)

10

0
04/13 04/18 04/23 04/28 05/03 05/08 05/13 05/18 05/23 05/28
2006

Fig. 7. Comparison of bottom orbital velocities and periods measured in Hudson Shelf Valley with estimates from NDBC buoy 44025.
(a) Direct estimates of ubrmeas from acoustic Doppler velocimeter (ADV) measurements at 0.6 m above bottom (Eq. (10)) compared with
estimates of ubrspec from buoy spectra and ubrpar from buoy Hs and Tp and a Donelan spectrum. (b) Comparison of measured
representative wave period, Tbrmeas with estimates Tbrspec and Tbrpar based on measured or estimated spectra (Eq. (9)).
ARTICLE IN PRESS
1254 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262

the record. Once again, bottom orbital velocities However, it is common practice to use these metrics
calculated from the measured surface-wave spectra for all measured spectra even though the statistical
(ubrspec) agree very well with values calculated from relationships may no longer be exactly satisfied. The
near-bed velocity data (ubrmeas) (BSS ¼ 0.88; Table formulations presented for bottom orbital velocity
2). Bottom orbital velocities calculated using the assume the absence of currents or very-low
parametric method with the Donelan .spectrum, frequency oscillatory motion (e.g., internal waves).
ubrpar, are also in good agreement (BSS ¼ 0.77; The presence of a net flow will bias the orbital
Table 2). The Donelan parametric method doesn’t velocities in the direction of the net flow during a
converge for 11% of the measured wave conditions, sampling interval.
again indicating wave parameters inconsistent with Spectral estimation of bottom orbital velocity
the general, unimodal spectral form. There remain using surface-wave observations, whether they are
some low-energy conditions when ubrpar drops measured spectra or parametric spectra scaled to
toward zero whereas the values derived from near- measured Hs and Tp, assumes that the measure-
bed velocity and measured spectra remain some- ments represent the wave conditions at the site of
what higher (Fig. 7b). These are associated with interest. Given the relatively long distances between
time when Tp is near or below the threshold period surface-wave buoys in most areas, consideration
of 7.1 s for deep-water waves at a depth of 39 m, but should be given to whether wave dissipation,
there are minor peaks at longer periods. These low- refraction, shoaling, local wind forcing, or wave-
frequency peaks are reflected in the large values of current interactions would be likely to substantially
Tbrspec during these times whereas Tbrmeas and change the waves between the buoy locations and
Tbrpar tend to be lower (Fig. 7a). This contributes the sites of interest. Ideally, the distance between the
to the fact that neither Tbrspec nor Tbrpar is a good buoy and the study site should be short and the
predictor (BSS40) of Tbrmeas for this time series. bathymetry simple. Gridded Hs and Tp from wave
models or hindcasts should similarly be at a
7. Uncertainties in spectral estimates of bottom resolution small enough to resolve effects of
orbital velocity dissipation, refraction and shoaling on wave height.
One potential source of uncertainty in applica-
There are a number of assumptions involved in tions of the parametric method that appears not to
calculating bottom orbital velocities from measured have a large effect on calculated bottom orbital
or estimated surface-wave spectra that should be velocity is the relative magnitude of the spectral
kept in mind when applying these methods. First, peak. Despite the large difference in spectral
the relationship between surface-wave conditions maxima for low values of dimensionless peak
and orbital velocity (Eq. (2)) assumes that wave frequency, f̄ p (Fig. 2a), the Donelan et al. (1985)
amplitude is small relative to wavelength, in which and the JONSWAP (Hasselmann et al., 1973)
case the waves can be considered linear. As wave spectral forms give similar values for ubrpar at the
steepness increases, estimates of bottom orbital two sites considered in this study, even though
velocity based on Eq. (2) become increasingly f̄ p p0:13 during a large fraction of the records.
inaccurate. Long-period waves, which have the Fig. 3 suggests differences in bottom orbital velocity
greatest effect on bottom orbital velocity (Fig. 1), of 5–10% depending on the spectral form used in
have long wavelengths and generally can be applications of the parametric method.
considered low amplitude outside the nearshore Larger errors are likely to occur when this
zone. Steep waves in the open ocean are usually method is applied to wave conditions that are
short period waves generated by local winds that are inconsistent with the unimodal spectrum assumed in
unlikely to contribute much to bottom orbital all of these spectral forms. The largest discrepancies
velocity. As waves enter shallow water, their between ubrmeas or ubrspec and ubrpar are observed
steepness increases and they become nonlinear. As for cases when the measured spectra are bimodal,
a result, the spectral methods described here should with the larger peak at higher frequencies (Figs. 6
be used with caution in shallow water. and 7). In these cases, calculations of ubrpar based
The wave statistics used here and elsewhere, e.g., on the Donelan spectrum do not converge, because
the relationship between m0 (the variance of surface the measured wave parameters are inconsistent
elevation) and Hs, assume a narrow-banded spec- with a unimodal spectrum. Thus, this method is
trum, i.e., a spectrum with a dominant central peak. useful for identifying these cases. Fortunately,
ARTICLE IN PRESS
P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262 1255

bottom orbital velocities are usually small under occurrence, p, times magnitude of the forcing, x,
these conditions, and are unlikely to contribute the effective forcing can be defined as the value
much to near-bed dynamics. Bottom orbital velo- associated with the peak in the distribution of px.
cities will also be small for low sea-state conditions, Similar arguments are used to calculate the effective
which may also not be well described by a simple discharge that maximizes sediment flux in rivers
unimodal spectrum. Finally, wave spectral formula- (e.g., Wolman and Miller, 1960; Lenzi et al., 2006).
tions such as the JONSWAP and Donelan spectra The probability distribution function (PDF) for a
assume that waves are traveling in one dominant Rayleigh distribution is
direction. Wave fields with significant directional
x
spreading or intersecting wave trains will not be well pðxÞ ¼ expðx2 =2b2 Þ (15)
represented using these spectral forms. b2

8. Representation of bottom orbital velocities for where b is the parameter that determines the width
calculations of near-bed processes and center of the PDF. If the distributionphas ffiffiffiffiffiffiffi been
normalized such that xrms ¼ 1, then b ¼ 0:5 and
Madsen (1994) argues that because the wave- x ¼ 0.886. It follows that the peaks in the distribu-
induced bottom shear stress, tbw, is proportional to tions of pxn occur at x ¼ 1, 1.23, and 1.414 for
the square of bottom orbital-velocity amplitude, ub n ¼ 1, 2, and 3, respectively.
(Eq. (3)), the full spectrum of orbital motions are Assuming that wave-induced near-bottom velo-
best represented by ubr as defined in Eq. (7), because cities follow a Rayleigh distribution and processes
a sinusoidal motion with this amplitude has the are forced by u0 (the full time series of orbital
same variance as the near-bed wave-induced velo- velocity at the bed (Eq. (1)) rather than orbital
city spectrum. Researchers have used a wide range velocity amplitude, ub (Eq. (2))), the effective u0 is in
of statistics to characterize near-bottom orbital fact u0rms. If, however, a process is dominated by
motions for shear stress and sediment-transport u20 (as one might argue frictional forces are) or u30
calculations. Many have followed Madsen’s (1994) (for sediment transport), then the effective value
guidance and use ubr (e.g., Soulsby, 1997; Wright of u is correspondingly pffiffilarger.
ffi In particular, the
et al., 1999), but many others (e.g., van Rijn, 1993; peak of pu30 occurs at 2u0rms ¼ ubr . Assuming a
Sherwood et al., 1994; Wiberg et al., 2002; Li and Rayleigh distribution for u0 also allows us to infer
Amos, 2001) used ubs (larger than ubr by a factor of that u1=10 ¼ 1:8u0rms ¼ 1:27ubr (following the rela-
pffiffiffi
2) derived either from spectra (Eq. (8)) or from Hs tionships for Rayleigh distributions in Dean and
and Tp (Eq. (2)). Other measures have also been Dalrymple, 1991).
used, including ub1/10 (Wiberg et al., 1994); u1/10, Do wave-induced near-bottom motions (u0)
for ‘‘irregular’’ waves (Soulsby and Whitehouse, follow a Rayleigh distribution like surface waves?
2005; O’Donoghue et al., 2005); and the ‘‘highest We might expect they would, given the linear
wave velocity measuredyduring each hourly burst’’ relationship between surface and bottom wave
(Cacchione and Drake, 1990; caption for their conditions. Empirical support is provided by
Fig. 8, p. 752). measurements at the Hudson Shelf Valley site
The distinctions among these measures of bottom (Fig. 7). Fig. 8 shows a typical histogram and
orbital velocity are important because wave-dependent cumulative distribution of instantaneous, normal-
near-bed parameters such as bed shear stress ized wave-induced current speeds from a 17-min
and sediment-transport rates are nonlinear func- burst at 4 Hz of velocity measurements made at
tions of wave motion. Madsen’s (1994) argument is the Hudson Shelf Valley site A on April 10,
that, because stress goes as u2, the appropriate 2006. Wave-induced current speeds were deter-
orbital velocity is proportional to the velocity mined as
variance. Arguments for a more extreme statistic qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
for sediment-transport calculations are based on the jUj ¼ ðu  ūÞ2 þ ðv  v̄Þ2 (16)
reasoning that transport increases nonlinearly with
bed shear stress, but the influence of the most where u and v are instantaneous horizontal compo-
energetic wave motions is mitigated by their nents of current velocity and ū and v̄ are burst-
infrequent occurrence. If dominant conditions are averaged values. The measured distributions compare
defined as those that maximize probability of well with a theoretical Rayleigh PDF (Eq. (15)) and
ARTICLE IN PRESS
1256 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262

1
|U|rms = 0.07, |U| = 0.15,
ubr = 0.11
0.8
bhat = 0.71, SkB = 0.99,

Cumulative Probability
Probability 0.1 SkW = 1.00
0.6

0.4
0.05

0.2

0 0
0 1 2 3 0 1 2 3
Normalized speed |U| /|U|rms Normalized speed |U| / |U|rms

Fig. 8. Comparison of measured near-bottom velocities with Rayleigh distribution. (a) Histogram of current speeds (normalized by rms
speed) in a typical 17-min burst (4096 measurements at 4 Hz) measured 0.6 mab at 40-m depth in Hudson Shelf Valley and best-fit
Rayleigh distribution (dark line). (b) Cumulative probability distributions of measurements (dashed line) and best-fit Rayleigh distribution
(gray line).

cumulative distribution F (the integral of Eq. (15)) parameterizing sediment-transport calculations in


 terms of bottom orbital velocity in general, and ubs
x2
F ðx; bÞ ¼ 1  exp  2 (17) specifically, assuming transport is proportional to
2b u3b . Other processes, as well as sediment transport
fit to the data, where x ¼ jUj=jUjrms , the normalized under high-energy conditions, may more effectively
speed (Fig. 8). The results shown in Fig. 8 are typical integrate wave motion over the full wave cycle. This
of the majority of bursts in the 2-month record of would argue for basing pffiffiffi calculations on some
bottom velocity at this site. statistic of u0, such as 2u0rms ¼ ubr . Furthermore,
It is common to use bottom orbital velocity, ub, while we have characterized processes in terms
the amplitude of near-bed wave-induced velocity, of the most probable value of forcing, other
rather than u0, the full wave-induced velocity characterizations, e.g., in terms of the average
record, to parameterize wave-generated flow at the or expected value of forcing, might also be
bed in calculations of sediment-transport and other appropriate depending on the problem of interest.
wave-related near-bed processes. The PDF of the Given these considerations, there may be no single
amplitude of a narrow-band, Gaussian random ‘‘best’’ parameterization of near-bed wave-induced
process, such as surface waves, follows a Rayleigh velocity.
distribution (Longuet-Higgens, 1952). This should
extend to bottom orbital velocities pffiffias
ffi well. This is 9. Conclusions
the basis for assuming that ubs ¼ 2ubr . If we were
to argue that near-bed wave-driven processes are Bottom orbital velocities, a measure of the
forced by p ubffiffiffi (rather than u0), then the peak of pu3b variance or energy of wave-induced bottom orbital
occurs at 2ubr ¼ ubs and the peak of pu2b occurs at motion at the bed, can be calculated in several ways
1:23ubr , close to the value of u1/10. depending on the available data. Calculations of
The decision as to whether u0 or ub best represents bottom orbital velocity from near-bed velocity
near-bed wave processes may depend on the measurements and from wave spectral data mea-
problem and conditions of interest. For example, sured at nearby surface-wave buoys are in excellent
sediment is only mobilized when the velocity is large agreement at two sites on the US east and west
enough that the associated shear stress exceeds coasts with different wave climates. How close
the critical shear stress for initial motion. This is a buoy has to be to a site of interest to provide
most likely to be true when u ¼ ub, an argument for good estimates of ubr depends on the extent of
ARTICLE IN PRESS
P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262 1257

surface-wave transformation between the sites. In the 0538 (PLW) and by the Coastal and Marine
absence of near-bed velocity or spectral data, bottom Geology Program, US Geological Survey (CRS).
orbital velocity can be estimated using significant This paper has benefited from comments from Brad
wave height and peak period to parameterize a Butman, Jessica Lacy, Andrea Ogston, and two
general spectral form, thereby producing an approx- anonymous reviewers. Any use of trade, product, or
imate surface-wave height spectrum from which firm names is for descriptive purposes only and does
orbital velocity and bottom period can be calculated. not imply endorsement by the US Government.
Estimates made using the Donelan et al. (1985) and
JONSWAP (Hasselmann et al., 1973) spectral forms
Appendix A. Methods for calculating kh
agree well with bottom orbital velocities calculated
from velocity and spectral data at both sites
The wavenumber, k, required for orbital-velocity
considered here. Calculations of ubrpar using the
calculations using linear theory appears implicitly in
Donelan spectrum do not converge during wave
the dispersion equation
conditions characterized by a bimodal spectrum and
a dominant high-frequency peak, as these conditions o2 ¼ gk tanhðkhÞ (18)
are inconsistent with the general unimodal spectral
and there is no direct method for determining k
form assumed in these calculations. The parametric
exactly. Practical solutions include iterative methods
spectral method is particularly useful for calculating
and various direct formulae; several of these have been
spatial fields of bottom orbital velocity from gridded
evaluated recently by Soulsby (2006). Iterative solution
surface-wave significant wave height and peak
using the Newton–Raphson method (Press et al., 1992)
period, such as wave models and hindcasts produce.
is efficient and can produce results to user-specified
Calculations of wave-driven sediment-transport or
accuracy (within the numerical precision of the soft-
other near-bed wave processes will vary depending
ware and computer used). Soulsby (2006) provides a
on the choice of bottom wave velocity statistic, e.g.,
concise version of this method, along with accuracy
representative or significant bottom orbital velocity,
estimates that allow it to be used with a fixed number
used to parameterize the effective wave velocity. The
of iterations without the need to perform convergence
best choice may depend on the specific problem being
tests. Our MATLAB implementation of his method is
addressed, but this general issue is one that merits
listed in Appendix E. A popular direct method by
further consideration.
Hunt (1979) cited in Dean and Dalrymple (1991)
involves a high-order polynomial, and several methods
Acknowledgements found by Carvalho (personal communication, 2006)
using an automatic problem solver involve nested
The authors gratefully acknowledge support by approximations. We implemented in MATLAB two
ONR Grants N00014-01-1-0690 and N00014-04-1- versions of the Newton–Raphson method and six

Table A1
Comparison of computational speed and accuracy of several methods for determining kh (wavenumber  depth), a term in linear wave
dispersion equations

Newton–Raphsona Newton–Raphsonb Huntc Gilbertd Fentone Carvalho Carvalho Carvalho


(Eq. (1))f (Eq. (2))g (Eq. (16))h

Relative time 1 0.24 0.20 0.13 0.35 0.49 0.36 0.79


MAD (  106) 0 63  109 578 2900 7400 124 139 29

Relative time was computed using MATLAB implementations over a large practical range of depths and periods; MAD is mean absolute
deviation of results minus Newton–Raphson method results with specified accuracy better than 106.
a
Newton–Raphson iteration following Press et al., (1992) until convergence better than 106.
b
Concise Newton–Raphson method of Soulsby (2006) with three iterations.
c
Hunt fifth-order polynomial (Hunt, 1979, cited in Dean and Dalrymple, 1991).
d
G. Gilbert method cited in Soulsby (2006).
e
Fenton (1990).
f
Carvalho (personal communication, 2006), his Eq. (1).
g
Carvalho (personal communication, 2006), his Eq. (2).
h
Carvalho (personal communication, 2006), his Eq. (16) (average of his Eqs. (1) and (2)).
ARTICLE IN PRESS
1258 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262

direct methods, and compared their speed and accuracy [nt,nf] ¼ size(s);
over a practical range of water depths (1–340 m) and % T ¼ 1 ./f;
wave periods (2–30 s). The results, summarized in w ¼ 2*pi*f;
Table A1, are not surprising. All of the direct methods % Determine kh usingSoulsby (2006)
are faster than the Newton–Raphson method when method (seeAppendix E)
requiring that precision in kh be better than 106, but kh ¼ qkhfs(w,h);
their accuracy varies. The venerable fifth-order method w ¼ repmat(w,nt,1);
of Hunt (1979) offers a good combination of speed and fm ¼ repmat(f,nt,1);
accuracy. The new Carvalho formulae are the most kh ¼ repmat(kh,nt,1);
accurate direct methods but are slower and less if(exist(‘df’)),
accurate than Soulsby’s Newton–Raphson formulation
if(length(df) ¼ ¼ 1),
with three iterations. This method (Soulsby, 2006) is
df ¼ df*ones(nt,nf);
recommended for its speed, accuracy, and simplicity. In
elseif(length(df) ¼ ¼ nf);
this context, it is worth noting that the value of g itself,
df ¼ repmat(df,nt,1);
which is usually taken as constant (exactly 9.80665
end
m/s2), in reality changes as a function of latitude,
else
elevation, and local geology. Variation associated
fi ¼ f(1)-(f(2)-f(1));
with latitude and variations in centrifugal force can fe ¼ f(end)+(f(end)-f(end-1));
cause changes in g of 70.23% between the equator
dfb ¼ diff([fi f(:)’]);
and the poles.
dff ¼ diff([f(:)’ fe]);
df ¼ mean([dfb;dff]);
df ¼ repmat(df,nt,1);
Appendix B. MATLAB function for calculating end
bottom orbital velocity from measured spectra
Su ¼ ((w.^2)./((sinh(kh)).^2)).*s;
%Eq. 5
function ubr ¼ sqrt(2.*sum((Su.*df)’)’ ); %Eq. 6
[ubr,Tbr] ¼ ubspecdat(h,s,f,df) fr ¼ (sum((Su.*fm.*df)’)’) ./
% UBSPECDAT—Calculate ubr and Tbr (sum((Su.*df)’)’); %Eq. 9
from measured spectra Tbr ¼ 1./fr;
% [ubr,Tbr] ¼ ubspecdat(h,s,f,df) fz ¼ sqrt((sum((Su.*fm.^2.*df)’)’)
% ./ (sum((Su.*df)’)’));
% Input: Tbz ¼ 1./fz;
% h ¼ water depth (m)—scalar or col.
vector with length
% s(nf) or s(nt,nf) ¼ array of spec-
Appendix C. Rescaling the general wave spectrum
tral densities normalized so that
(Eq. (13)) in terms of Hs
% Hs ¼ 4*sum(s,2)*df
% f ¼ row vector with central fre-
Most wave spectral equations were developed to
quencies (Hz)
predict Hs for given wind conditions, U. If instead
% df ¼ (optional) scalar or row vec-
Hs is known but U is not, it is useful to rescale the
tor with freq. bandwidths (Hz)
general spectra equation (Eq. (13)) in terms of Hs
% Returns:
rather than a. This has been done for several specific
% ubr ¼ representative bottom orbi-
spectral forms by Chakrabarti (1986) and Soulsby
tal velocity (m/s)
(1987). To accomplish this for the general spectral
% Tbr ¼ representative bottom wave
equation, we first change the independent variable
period (s)
in Eq. (13) to f* ¼ f/fp
% The alternative bottom period,
Tbz, is also calculated (see text). A f nx  
SZ ðf n Þ ¼ 6 n5
exp bf n4
% Chris Sherwood, USGS fpf
% Last revised March 8, 2006 1Þ2 =ð2s2 Þ
gexp½ðf
n

g ¼ 9.81; (19)
ARTICLE IN PRESS
P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262 1259

where A ¼ ag2/(2p)4. If we define the integral given in Table 1, to solve for f̄ p ¼ f p U=g for a
R R
ðf 6p =AÞ f n S Z ðf n Þ df n ¼ 1=w, then ðwf 6p =AÞ f n S Z ðf n Þ specific spectral form, from which the consistent
df n ¼ 1; the value of w depends on b, g, and s, values of g and s can be determined. For example,
which are either constant or vary with f* or f̄ p ¼ for the Donelan spectrum, f̄ p ¼ ½m0 f 4p wð2pÞ4 =
f p U 10 =g depending on the formulation (Table 1). ð0:0165g2 Þ1=0:55 . Alternatively, Young (1992) com-
R R bined expressions for dimensionless variance e ¼
Recalling Eq. (4), m0 ¼ f S Z ðf Þ df ¼ f 2p f n S Z ðf n Þ
m0g2/U4 and dimensionless peak frequency f̄ p with
df n ¼ ðA=f 6p Þ=ðf 2p =wÞ giving A ¼ m0 f 4p w. Thus we Donelan et al.’s (1985) empirical expression for
3:3
can rewrite Eq. (13) as dimensionless variance  ¼ 6:365  106 f̄ p to get
2 4 6 2 1=0:7
!x !4 3 f̄ p ¼ ½m0 f p =ð6:365  10 g Þ , which is indepen-
m0 f 4p w f f dent of w. Several iterations through Eq. (20) with
SZ ðf Þ ¼ exp4b 5gexp½ðf f p Þ2 =ð2s2 f 2p Þ
f5 fp fp these expressions for f̄ p and the relationships for g and
s in Table 1 for a given Hs and Tp should converge to
(20)
a consistent set of spectral parameters in Eq. (20).
which is equivalent to Eq. (14). Values of w for the Lack of convergence suggests that the spectral form is
JONSWAP and Donelan spectra are plotted in not consistent with the specified values of Hs and Tp.
Fig. 2c. We can also use the two relationships for A MATLAB function that implements these calcula-
A above to solve for a and, using the relationships tions is provided in Appendix D.

Appendix D. MATLAB function for calculating bottom orbital velocity using a parametric spectrum

function [ubr,Tbr,iter] ¼ ubspecpar(hs,tp,h,specform)


% UBSPECPAR—Caclulate ubr and Tbr from hs and tp using parametric spectrum
% [ubr,Tbr,iter] ¼ ubspecform(hs,tp,h,specform)
%
% Input:
% hs—Significant wave height (m)
% tp—Peak period (s)
% h—Water depth (m)
% specform—spectral formulation to use
% specform ¼ ‘D’ for Donelan spectrum (default)
% specform ¼ ‘J’ for JONSWAP spectrum
% Returns:
% ubr ¼ representative bottom orbital velocity (m/s)
% Tbr ¼ representative bottom wave period (s)
% iter ¼ number of iterations if Donelan spectrum is chosen
% If iter45 the calculation did not converge
% iter ¼ 0 for the JONSWAP spectrum
%
% Patricia Wiberg, UVa
% Last modified 9 Mar 2007
if (exist(‘specform’,’var’)),
specform ¼ ‘D’;
end;
g ¼ 9.81; dffp ¼ 0.01;
ffp ¼ 0.2:dffp:5; %nb ¼ length(ffp);
nt ¼ length(tp);
iter ¼ zeros(nt,1);
ubr ¼ zeros(nt,1); Tbr ¼ ubr; Tbz ¼ Tbr;
for i ¼ 1:nt,
ARTICLE IN PRESS
1260 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262

m0 ¼ hs(i).^2./16;
fp ¼ 1./tp(i);
f ¼ ffp.*fp; df ¼ dffp.*fp;
kh ¼ qkhfs(2*pi*f,h); %(seeAppendix E)
if specform ¼ ¼ ‘D’,
xi ¼ 1; tol ¼ 1e-3; m0sfD ¼ 0.;
while abs((m0-m0sfD)/m0 4tol),
fpbar ¼ (m0*fp^4/(g^2*6.635e-6))^(1/.7);
gam ¼ 1.7;
if fpbar4 ¼ 0.05,
% alpha ¼ 0.0165.*fpbar.^0.55;
if fpbar40.159, gam ¼ 6.5+2.606*log(fpbar); end;
sig ¼ 0.08+0.0013.*(fpbar).^-3;
else,
% alpha ¼ 0.0165.*0.05.^0.55;
sig ¼ 0.08+0.0013.*(0.05).^-3;
end;
eterm ¼ -((ffp-1).^2)./(2.*sig.^2);
ee ¼ exp(eterm); t2 ¼ gam.^ee;
t1 ¼ -(ffp).^-4;
sffpn ¼ (ffp.^xi./ffp.^5).*exp(t1).*t2; %dimensionless spectrum
XD ¼ 1./sum(sffpn.*dffp); %chi
sf ¼ m0.*XD*fp.^4.*1./(f.^4*fp).*exp(t1).*t2; %Eq. 20
m0sfD ¼ sum(sf.*df);
iter(i) ¼ iter(i)+1;
if iter(i)45, break; end;
end;
%m1sfD ¼ sum(f.*sf.*df);
%m2sfD ¼ sum(f.^2.*sf.*df);
%fmD ¼ m1sfD./m0sfD; TmD ¼ 1./fmD;
%fzD ¼ sqrt(m2sfD./m0sfD); TzD ¼ 1./fzD;
su ¼ (2*pi.*f./sinh(kh)).^2.*sf; %Eq. 5
ubrD ¼ sqrt(2.*sum(su*df)); %Eq.6
frD ¼ sum(su.*f.*df)./sum(su.*df); %Eq. 9
fzD ¼ sqrt(sum(f.^2.*su.*df)./sum(su.*df));
TbrD ¼ 1./frD; TbzD ¼ 1./fzD;
ubr(i) ¼ ubrD; Tbr(i) ¼ TbrD;
end;
if specform ¼ ¼ ‘J’,
gam ¼ 3.3; xi ¼ 0; %XJ ¼ 3.283; %The value of gam can be changed
jn ¼ find(ffp41);
sig ¼ 0.07*ones(size(ffp)); sig(jn) ¼ 0.09;
eterm ¼ -((ffp-1).^2)./(2.*sig.^2);
ee ¼ exp(eterm); t2 ¼ gam.^ee;
t1 ¼ -1.25.*(ffp).^-4;
sffpn ¼ (ffp.^xi./ffp.^5).*exp(t1).*t2; %dimensionless spectrum
XJ ¼ 1./sum(sffpn.*dffp); %chi
sf ¼ m0.*fp.^4.*XJ./f.^5.*exp(t1).*t2; %Eq. 20
%m0sfJ ¼ sum(sf.*df);
%m1sfJ ¼ sum(f.*sf.*df);
%m2sfJ ¼ sum(f.^2.*sf.*df);
ARTICLE IN PRESS
P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262 1261

%fmJ ¼ m1sfJ./m0sfJ; TmJ ¼ 1./fmJ;


%fzJ ¼ sqrt(m2sfJ./m0sfJ); TzJ ¼ 1./fzJ;
su ¼ (2*pi.*f./sinh(kh)).^2.*sf; %Eq. 5
ubrJ ¼ sqrt(2.*sum(su*df)); %Eq. 6
frJ ¼ sum(f.*su.*df)./sum(su.*df); %Eq. 9
fzJ ¼ sqrt(sum(f.^2.*su.*df)./sum(su.*df));
TbrJ ¼ 1./frJ; TbzJ ¼ 1./fzJ;
ubr(i) ¼ ubrJ; Tbr(i) ¼ TbrJ;
end;
end;

Appendix E. MATLAB function for calculating kh (wavenumber * depth) for linear wave theory using the
Soulsby (2006) Newton–Raphson method

function kh ¼ qkhfs( w, h )
% QKHFS—Quick iterative calculation of kh in dispersion relationship
% kh ¼ qkhfs( w, h )
%
% Input:
% w ¼ Angular wave frequency ¼ 2*pi/T where T ¼ wave period [1/s]
% h ¼ Water depth [m]
% Returns:
% kh ¼ wavenumber * depth [ ]
%
% Either w or h can be a vector, but not both.
% Hard-wired for MKS units.
% Orbital velocities from kh are accurate to 3e-12 !
%
% RLSoulsby (2006)‘‘Simplified calculation of wave orbital velocities’’
% HR Wallingford Report TR 155, February 2006, Eqns. 12a–14
% csherwood@usgs.gov
% Sept 10, 2006
g ¼ 9.81;
x ¼ w.^2*h./g;
y ¼ sqrt(x) .* (xo1)+x.* (x4 ¼ 1);
%this appalling bit of code is about 25% faster than a for loop
t ¼ tanh( y );
y ¼ y-( (y.*t -x)./(t+y.*(1-t.^2)));
t ¼ tanh( y );
y ¼ y-( (y.*t -x)./(t+y.*(1-t.^2)));
t ¼ tanh( y );
y ¼ y-( (y.*t -x)./(t+y.*(1-t.^2)));
kh ¼ y;
return;

Appendix F. Supplementary materials

Supplementary data associated with this article can be found in the online version at doi:10.1016/
j.cageo.2008.02.010.
ARTICLE IN PRESS
1262 P.L. Wiberg, C.R. Sherwood / Computers & Geosciences 34 (2008) 1243–1262

References Massel, S.R., 1996. Ocean Surface Waves: Their Physics and
Prediction. World Scientific, Singapore, 401pp.
Alves, J.H.G.M., Banner, M.L., Young, I.R., 2003. Revisiting the Mitsuyasu, H., Tasai, F., Suhara, T., Mizuno, S., Ohkusu, M.,
Pierson–Moskowitz asymptotic limits for fully developed wind Honda, T., Rikishi, K., 1980. Observations of the power
waves. Journal of Physical Oceanography 33, 1301–1323. spectrum of ocean waves using a cloverleaf buoy. Journal of
Battjes, J.A., Zitman, T.Z., Holthuusen, L.H., 1987. A reanalysis Physical Oceanography 10, 286–296.
of the spectra observed in JONSWAP. Journal of Physical Ochi, M.K., 1998. Ocean Waves. Cambridge University Press,
Oceanography 17, 1288–1295. Cambridge, 319pp.
Brier, G.W., 1950. Verification of forecasts expressed in terms of O’Donoghue, T., Doucette, J.S., van der Weerf, J.J., Ribberink,
probabilities. Monthly Weather Review 78, 1–3. J.S., 2005. Flow tunnel measurements of full-scale ripples in
Cacchione, D.A., Drake, D.E., 1990. Shelf sediment transport: an oscillatory flow. In: Van Rijn, L.C., Soulsby, R.L., Hoekstra,
overview with applications to the northern California P., Davies, A.G. (Eds.), SANDPIT: Sand Transport and
continental Shelf. In: LeMehaute, B., Hanes, D.M. (Eds.), Morphology of Offshore Mining Pits. Aqua Publications, The
The Sea, vol. 9, Part B. Wiley, New York, pp. 729–773. Netherlands, pp. V1–V10.
Chakrabarti, S.K., 1986. Unification of two-parameter energy Pierson, W.J., Moskowitz, L., 1964. A proposed spectral form for
spectrum models. Journal of Waterway, Port, Coastal and fully developed wind seas based on the similarity theory of
Ocean Engineering 112, 173–177. S.A. Kitaigorodskii. Journal of Geophysical Research 69,
Dean, R.G., Dalrymple, R.A., 1991. Water Wave Mechanics for 5181–5190.
Engineers and Scientists. World Scientific, Singapore, 353pp. Press, W.H., Flannery, B.P., Teukolsky, S.A., Vetterling, W.T., 1992.
Donelan, M.A., Hamilton, J., Hui, W.H., 1985. Directional Numerical Recipes in Fortran; The Art of Scientific Computing,
spectra for wind-generated waves. Philosophical Transactions 2nd ed. Cambridge University Press, Cambridge, 994pp.
of the Royal Society of London, A 315L, 509–562. Sherwood, C.R., Butman, B., Cacchione, D.A., Drake, D.E.,
Earle, M.D., 1996. Nondirectional and directional wave data Gross, T.F., Sternberg, R.W., Wiberg, P.L., Williams III,
analysis procedures. National Data Buoy Center, Technical A.J., 1994. Sediment-transport events on the northern
Document 96–01, Stennis Space Center, Mississippi, 37pp. California continental shelf during the 1990–1991 STRESS
Fenton, J.D., 1990. On calculating the lengths of water waves. experiment. Continental Shelf Research 14, 1063–1099.
Coastal Engineering 14, 499–513. Soulsby, R.L., 1987. Calculating bottom orbital velocity beneath
Hasselmann, K., Barnett, T.P., Bouws, E., Carlson, H., Cart- waves. Coastal Engineering 11, 371–380.
wright, D.E., Enke, K., Ewing, J.A., Gienapp, H., Hassel- Soulsby, R.L., 1997. Dynamics of Marine Sands. Thomas
mann, D.E., Kruseman, P., Meerburg, A., Müller, P., Olbers, Telford, London, 264pp.
D.J., K. Richter, K., Sell, ., Walden, H., 1973. Measurements Soulsby, R.L., 2006. Simplified Calculation of Wave Orbital
of wind-wave growth and swell decay during the Joint North Velocities. TR-155. HR Wallingford Ltd., Wallingford,
Sea Wave Project (JONSWAP). Deutsche Hydrographisches 12pp.+figures.
Institut, Hamburg, Suppl. A 8 (12), 95pp. Soulsby, R.L., Whitehouse, R.J.S., 2005. Prediction of Ripple
Hunt, J.N., 1979. Direct solution of wave dispersion equation. Properties in Shelf Seas. Mark 2 Predictor for Time
Journal of the Waterway, Port, Coastal and Ocean Division, Evolution. TR-154. HR Wallingford Ltd., Wallingford, 99pp.
American Society of Civil Engineers 105 (WW4), 457–459. Tolman, H.L., 1999. User manual and system documentation of
Jensen, B.L., Sumer, B.M., Fredsøe, J., 1989. Turbulent WAVEWATCH-III version 1.18. National Oceanic and
oscillatory boundary layers at high Reynolds numbers. Atmospheric Administration, National Centers for Environ-
Journal of Fluid Mechanics 206, 265–297. mental Prediction, Ocean Modeling Branch Contribution
Jonsson, I.G., 1966. Wave boundary layers and friction factors. Number 166, Camp Spring, Maryland, 110pp.
In: Proceedings of the 10th International Conference on van Rijn, L.C., 1993. Principles of Sediment Transport in Rivers,
Coastal Engineering. American Society of Civil Engineers, Estuaries, and Coastal Seas. Aqua Publications, Amsterdam.
Tokyo, pp. 127–148. Wiberg, P.L., 1995. A theoretical investigation of boundary layer
Jonsson, I.G., Carlsen, N.A., 1967. Experimental and theoretical flow and bottom shear stress for smooth, transitional, and
investigations in an oscillatory turbulent boundary layer. rough flow under waves. Journal of Geophysical Research
Journal of Hydraulic Research 14, 45–60. 100, 22,667–22,679.
Lenzi, M.A., Mao, L., Comiti, F., 2006. Effective discharge for Wiberg, P.L., Drake, D.E., Cacchione, D.A., 1994. Sediment
sediment transport in a mountain river: computational resuspension and bed armoring during high bottom stress events
approaches and geomorphic effectiveness. Journal of Hydrol- on the northern California continental shelf: measurements and
ogy 326, 257–276. predictions. Continental Shelf Research 14, 1191–1219.
Li, M.Z., Amos, C.L., 2001. SEDTRANS96: the upgraded and Wiberg, P.L., Drake, D.E., Harris, C.K., Noble, M., 2002.
better calibrated sediment-transport model for continental Sediment transport on the Palos Verdes shelf over seasonal to
shelves. Computers & Geosciences 27, 619–645. decadal time scales. Continental Shelf Research 22, 987–1004.
Longuet-Higgens, M.S., 1952. On the statistical distribution of Wolman, M.G., Miller, J.P., 1960. Magnitude and frequency of
the heights of sea waves. Journal of Marine Research 11, geomorphic processes. Journal of Geology 68, 57–74.
245–266. Wright, L.D., Kim, S.-C., Friedrichs, C.T., 1999. Across-shelf
Madsen, O.S., 1994. Spectral wave-current bottom boundary variations in bed roughness, bed stress and sediment suspension
layer flows. In: Coastal Engineering 1994: Proceedings, 24th on the northern California shelf. Marine Geology 154, 99–115.
International Conference, Coastal Engineering Research Young, I.R., 1992. The determination of spectral parameters
Council. American Society of Civil Engineers, Kobe, Japan, from significant wave height and peak period. Ocean
pp. 384–398. Engineering 19, 497–508.

You might also like