5. North West Shelf
The North West Shelf (NWS) model domain extends 450 km along the coast and 425 km
offshore. Tropical Cyclone Bobby passed through this region at the end of February
1995, six weeks after a deployment of 16 current meter moorings and this event was
considered to be a stringent test of ROAM and its boundary conditions.
The NWS is a region where energetic internal tides are generated over the shelf and slope
regions (Holloway et al 2001). We do not attempt to model the internal tide as the length
scale required to adequately resolve internal waves would be prohibitive given that we
need an area large enough to adequately model the tropical cyclone. The internal tides are
evident in the available observations and have to be taken into account when making
comparisons with model results.
The data we have available for comparison with the model output have been the subject
of two papers and their results shall be referred to at times. Holloway et al (2001) show
that an energetic, shoreward propagating semi-diurnal internal tide is generated with
multiple generation sites coinciding with near-critical bottom slopes. Davidson &
Holloway (2003) model the tropical cyclone ocean response with the 3-D Princeton
Ocean Model.
5.1 Observations
5.1.1 Instrument locations
A detailed CTD survey was conducted in January 1995 during the mooring deployment.
Each CTD station (Fig 5.1.1) was occupied for 13 hours to measure one complete tidal
cycle with profiles taken every 30 mins except at the deep stations. We use data from
three of the moorings (indicated in pink) and the current meter depths are indicated in
Figure 5.1.2. The current meters recorded temperature and velocity with a 2 min interval
for the meters on M4 and a 5 min interval for M2. Filtered coastal tidegauge for the
locations indicated was also made available by the WA Department of Transport.
5.1.2 Internal tides
From the CTD section (Fig. 5.1.2) made up from vertical profiles averaged over 13 h it
would appear that there is a lens of water between 100 km and 150 km offshore where
the thermocline is more linear with depth. The lens however is an artefact of the presence
of the internal tide which is strongest in the region about M4. The timeseries of CTD
profiles taken at the M4 location in Figure 5.1.3 show that within a 13 h period water
from 110 m is raised to less than 50 m depth for a few hours and eventually returned to
its initial depth. A 5 day timeseries using M4 temperature records (Fig. 5.1.4) (and
spanning the time of the CTD profiling at that location) shows that the vertical
temperature structure switches rapidly over the tidal cycle between two very different
profiles and is not well represent by the mean profile. The internal tide is evident at both
the M2 and M4 locations throughout the deployment with the amplitude of the isothermal
excursions at M2 to be about half that observed at M4 (Holloway et al 2001).
The effect that the internal tides have on the mean temperature structure as seen in Figure
5.1.2 makes the comparison between temperature observations and model output
impossible even when the data has been filtered to remove tides.
5.2 Model initialisation and forcing
The model was initialised with a geographic-rectangular grid with 112 grid points
alongshore and 97 across-shelf giving an average grid spacing of 4 km and 4.4 km
respectively. The depth reaches 2000 m along most of the offshore boundary of the grid
and gets to a maximum of about 3000 m at the north-east corner. The model was run for
the month of February, with the cyclone entering the domain at about 23 Feb.
The model was nested in the two alternative and previously described products: OFAM
and SynTS. Both of these products provide initial values of temperature, salinity and
surface elevation to initialise the model as well as a timeseries of these variables at the
three ocean boundaries. Figure 5.2.1 shows the top 100 m of the vertical temperature
profile of both OFAM and SynTS at the location of the furthest CTD station as well as
the mean CTD profile (which showed very little isothermal fluctuations over the 13 h of
sampling). The SynTS profile is remarkably close to that observed including a close fit at
the thermocline despite the relatively course depth resolution. OFAM is cooler at the
surface and has a more gradual thermocline and at 100m is 2 degrees warmer than the
observed temperature.
A profile from the innermost CTD station is included in Fig. 5.2.1 showing the surface
temperature to be half a degree warmer than offshore, the thermocline to be more severe
and the bottom layer well-mixed. This inshore structure is not replicated by either SynTS
or OFAM which have little across-shelf structure.
The model is forced with winds and atmospheric pressure provided by RAMS which has
the tropical cyclone seeded in it. Except for the tidegauge heights neither OFAM nor
SynTS had any input data that provided them information about the presence of the
cyclone providing a good test of the ROAM boundary conditions. Figure 5.2.2 shows the
ROAM domain embedded in SynTS at the time that the cyclone crossed the middle of the
domain and the difference in the vertical temperature profiles at the boundaries.
5.3 Comparison of model with observations
5.3.1 Barotropic tides
The barotropic component of the velocity at each mooring is estimated by averaging the
observed velocity at each current meter. The equivalent model component compares well
for the duration of the model run before the cyclone indicating that the model is doing a
good job of replicating the barotropic tidal velocities in both phase and amplitude
(Figures 5.3.1 and 5.3.2).
The model response to the cyclone includes a strong barotropic velocity component
which is not evident in the observations and resulting in strong velocities at depth. This
appears to be a common problem to cyclone modelling and also occurred in the POM
model for the same cyclone (Davidson & Holloway, 2003).
5.3.2 Low-frequency response
Modelled timeseries of coastal height at the four tidegauges track the observations quite
well (Fig. 5.3.3). This is not a completely independent result as the same tidegauge
heights are input at the boundaries.
The timeseries of low-frequency passed filtered velocity (Figures 5.3.4 and 5.3.5) show
some discrepancies between observed and modelled. Some of the difference can be
accounted for by the inertial period barotropic response which can be seen at the deeper
current meters. Matching velocity amplitudes at each mooring is a quite high expectation
and is dependent on the track of the cyclone being accurate. Small changes in the cyclone
track of the order of a few kilometres can give very different velocity timeseries at a point
(Davidson and Holloway, 2003). The alongshore velocity timeseries (Fig 5.3.4) are
consistent with the modelled cyclone being slightly too close to the coast as the model
response at M6 and M4 is stronger than the observed response yet at M2 the model
response is weaker than that observed.
5.4 Comparison of model runs
Here we shall compare the automated ROAM run with a few other salient runs. The
automated run is has sponges on the boundaries and is run with the relatively robust
Mellor-Yamada 2.0 vertical mixing scheme. The auto run was also initialised with and
nested in SynTS. Given the nature of the forcing (i.e. a cyclone which crosses the
boundaries and crosses the domain) it is a significant result to find that removing the
sponge layers does not significantly change the model response.
References
Davidson F.J.M. and Holloway P.E. (2003). A study of tropical cyclone influence on the
generation of internal tides.
Holloway P.E., Chatwin P.G. and Craig P. (2001). Internal tide observations from the
Australian North West Shelf in summer 1995.