>
Wave Height Measurements Using Acoustic Surface Tracking

Torstein Pedersen,
Nortek AS

Sven Nylund,
Nortek AS
Nortek AS
Industriveien 33
1337 Sandvika
Norway
inquiry@Nortek.no


André Dolle,
Thetis
Thetis
ZA Actisud
Le Beau Vézé
83320 Carqueiranne
France
inquiry@thetis.fr





Abstract
- Nortek has improved upon its AWAC, a
current and wave measurement sensor package, by
introducing a vertical, acoustic beam that detects the
surface. This added functionality allows for directly
measuring waves as opposed to inferring wave estimates
from wave energy spectra.

Traditionally, wave measurements from bottom-
mounted instruments, such as the combined pressure-
velocity (PUV) approach, are limited in their frequency
response. This is due to attenuation of the surface signal
with increasing depth. Recent advances employ the
alternative solution of measuring orbital velocities close
to the surface and incorporating the Maximum Likelihood
Method (MLM) estimate technique [1]. This improves the
accuracy at higher frequencies. However, for deployment
depths of 10 meters or deeper, these methods cannot
resolve waves periods that are 3 seconds or shorter.
Moreover, these bottom-mounted systems do not
measure the real surface time series, which makes it
difficult to calculate extreme value statistics.

The following paper provides an overview of the
process of (1) developing the surface track algorithms, (2)
comparing with a Datawell wave buoy off the coast of
Carqueiranne, France (3) and testing limiting conditions
such as breaking waves and greater depths (35 meters).


I. INTRODUCTION

Norteks AWAC (Acoustic Wave and Current, Fig. 1)
has traditionally measured both the pressure and orbital
velocities to estimate the wave frequency and
directional spectrum. Recently, we have modified the
firmware to allow us to detect the free surface using the
vertical beam. The modification eliminates the
constraint from the attenuation of wave properties with
depth. Therefore the AWAC is now capable of
measuring higher frequency waves in deeper water with
a greater degree of accuracy.

This approach of measuring waves is not necessarily
a new concept [2]. However it represents a
considerable step forward from existing bottom mounted
sensors now available, which generally rely just on the
pressure and velocity measurements.



Fig. 1 Deployed AWAC with four 1 MHz beams

The development and validation of the surface
tracking was performed over the course of three
separate experiments. The first was performed at
Drøbak, a site located in the fjord just south of Oslo.
The second experiment was performed in
Carqueiranne, France. Here we were able to directly
compare to a DataWell WaveRider buoy. Once we
established that the surface track measurements were
in good agreement with the wave buoy, we
implemented the surface track firmware in an AWAC
online in Hwa Lien, Taiwan. This last site demonstrated
that the AWAC is capable of measuring waves in depths
of 35 meters, with little compromise in data quality.

II. SYSTEM OVERVIEW

The AWAC is designed to measure both the current
profile and the wave directional spectrum using acoustic
Doppler technology. It can be used in stand-alone and
online mode. The target application is long term coastal
monitoring of waves and currents along the coast. The
wave measurement process employs a single cell per

1234 beam to minimize data volume and extend deployment
duration. Furthermore the cells are adaptively located
to ensure maximum signal strength.

The AWAC has four, 1 MHz transducers. One
center and the other three are equally spaced around it,
angled 25
° off the vertical axis. Beam width is 1.7° (3
dB point).

The instrument employs a fixed point DSP. Normal
memory size is 20-80 MB of flash, which provides
several months of current and wave data.

Other specifications:
Pressure sensor, 50 m range
Compass
Tilt sensor
Temperature sensor
1 Watt typical power consumption
9-16 Volts DC
1, 2, or 4 Hz Sampling
512, 1024, or 2048 samples per burst

III. PROCESSING

The approach used to detect the surface is relatively
simple. It can be broken down into the following
sequence of steps. (1) Transmit a pulse of a given
length; (2) Specify a receive window covering the range
of all possible wave heights; (3) Discretise the receive
window into multiple cells (~5 cm); (4) Apply a match
filter over series of cells to locate surface; (5) Use
quadratic interpolation to precisely estimate surface
location. An example of the amplitude time series for
the discretised signal is provided in Fig. 2.

Clearly we had to consider the prospect of false detects
and no detects. No detects were easily noted since
they did not exceed a specified threshold level for
detection. False detects on the other hand required
special determination. This began by identifying
samples that exceeded a specified bound relative to the
mean of the ensemble. This boundary was defined as
some multiple of the standard deviation of the
ensemble. This clean up step was iteratively performed
with increasingly tighter bounds to ensure all false
detects were removed. Finally, if the cumulative
number of false and no detects exceeded 10% of the
total number of samples in the ensemble, the ensemble
was considered corrupt and discarded.

Once the time series for the surface has been
established, we carry on with the traditional zero-
upcrossing method of estimating wave statistic.

Fig. 2 Example of a echo return from the surface.
The frequency limitation for the measurable waves
does not just lie with the Nyquist limit, but also with the
footprint created by vertical beam intersecting the
surface. Naturally, as the deployment depth increases,
the footprint increases. As a general rule, we follow a
Nyquist like reasoning; the frequency limit associated
with the footprint is when half the wavelength is on the
order of the diameter of the footprint.

IV. RESULTS

The organization of the results is presented in terms
of the objective of each experiment. Therefore the data
collected in Norway, France, and Taiwan is organized in
relation to the development of the surface tracking
algorithms and the subsequent validation.

A. Algorithm/Firmware Development

The first test was performed at Drøbak, a site local to
Nortek in Norway. The site offers the luxury of having
an AWAC online. This affords us the opportunity to
quickly test out new algorithms since we can both install
new firmware and upload collected data online. The
site was interesting in the sense that it is virtually
unexposed to the open ocean as it is still in the Oslo
Fjord (Fig. 3). This means that there are three possible
mechanisms for wave generation. These are (1) locally
wind generated waves, (2) transient waves from local
shipping traffic on way to Oslo, and perhaps if the
direction was right, (3) waves from open sea.

The AWAC is located on the sill of the fjord in 21
meters of water. Data was sampled at 4 Hz and
collected for over 17 minutes (1024 seconds). The
receive window was set at 8 meters in length and
subdivided into smaller bins so that there were 170
cells, each of which 4.7 cm long. We did not expect to
ever see any waves requiring such a large receive
window, however it provided ample opportunity for false
detects.

1235














Initial testing was quite encouraging since we
immediately noted higher frequency waves in the time
series and that transient events were regularly detected
since the location is exposed to considerable shipping
traffic. An example of this is presented in Fig. 4. Here
one can see that a passing ships wake. The
attenuated pressure signal is plotted as well.

Additionally, the locally generated wind waves are
clearly evident in the surface track but not in the
pressure signal. The spectrum of the surface track is
presented in the subsequent plot, demonstrating that
energy is detectable up to 1 Hz for the given setup.

The beam casts a footprint with a diameter of 0.62
meters on the surface. Therefore the limit associated
with the footprint is 1.1 Hz.




Fig. 3 AWAC test location noted by large
circle, Drøbak Norway.

Fig. 4 Surface Track (blue) and Pressue (Purple) time series indicating a passing ship. Bottom pane shows energy spectrum for the surface
track, note detectable energy up towards 1 Hz.

1236 Surprisingly, there were very few false or no detects
for the surface tracking. We attributed this to the fact
that the wave environment here is only exposed to small
waves and which are rarely breaking. Breaking waves
seem to be the most threatening to accurate surface
detection since there is greater possibility to falsely
detect the entrained bubbles below the waves. The
value of the tests in Drøbak was the realization of a
match filter and threshold level for which we had
confidence. We concluded that the next logical step
was to verify the accuracy of the measurements and
expose the method to both larger and