Modelling and Remote Sensing in Support of Olympic Sailing

Dr Roger Proctor
ProudmanOceanographic Laboratory

Dr Timothy J Smyth
Plymouth Marine Laboratory

Hamish Willcox
Royal Yachting Association

Colin Bell
ProudmanOceanographic Laboratory

Abstract

Computer models and remote sensing were combined to provide inputs into tactical decision making at the Olympic Games in September 2000 in Sydney, Australia. The complexity of the oceanographic conditions within and outwith Sydney Harbour necessitated different approaches to current prediction. Tidal currents within the harbour were predicted using numerical models. Currents offshore were predicted using a combination of sea surface temperatures derived from satellite imagery and simple regression formulae relating wind patterns to current speed and direction. These data were condensed down to supplement information on wind conditions to produce sets of simple sailing instructions, the ‘call book’, which varied according to the conditions, the area and the race type.

INTRODUCTION

The Olympic Games in 2000 focussed a great deal of scientific effort on the locality of Sydney Harbour to forecast both the meteorological and oceanographic conditions leading up to and during the Games. We report here on the effort directed towards the Royal Yachting Association (RYA) preparations for Team GB. These began in earnest with the pre-Olympic regatta in September 1999 and intensified as September 2000 arrived. Our role in these preparations was not so much to dynamically predict the hour-by-hour changing currents but more to form an understanding of the prevailing current regimes and provide insight into changes which may occur during race periods, allowing sailing tactics to be modified accordingly.

Fig. 1: Sydney Harbour showing race areas A-F

Sydney Harbour is an interesting place to hold a sailing competition. The topographical complexity of the land and coastline around the harbour (Figure 1) means that winds and currents are significantly affected by the local topography. Within the harbour estuary currents are dominated by moderate tidal flows with headland-generated eddies superimposed. Outwith the harbour, the coastal waters experience a complex cocktail of oceanographic processes, including tidal, meteorological and wave-induced flows but dominated by the (usual) southwards flow of the East Australian Current. Race areas were set up both inside and outside the harbour (see Figure 1), and boat classes (eleven in all) moved from one course to another for different races, thus requiring detailed information on how different boats may be affected by the conditions encountered in each area.

Oceanographically, the starting point for the study was a literature survey and examination of existing charts.

Within the estuary the available tidal stream charts (produced by the Royal Australian Navy) show the basic features of a semi-diurnal (two tides per day) tidal regime with neap/spring flow into and out of the estuary and the many headlands altering this current flow with eddies. However the tidal diamond detail on the charts does not encompass all of the racing areas and so to provide a complete picture of tidal streams a tidal model of the estuary was constructed. We were not the only ones to reach this conclusion and several models of currents in Sydney Harbour were produced in the period prior to the Games (e.g. Das et al., 2000).

In recent years much interest has been applied to the ocean region adjacent to Sydney because of plans for offshore deepwater sewage disposal. Middleton et al. (1997) provide a brief but comprehensive review of the oceanography of the continental shelf in the Sydney region and indicate that surface currents are predominantly alongshore and can have components associated with 4 major forcing mechanisms:

•    the East Australian Current (EAC), which originates in the Coral Sea to the north and transports a narrow band of warm water down the east coast of Australia to the Tasman Sea. This is a strong and persistent feature with southward flows up to 1.5m/s (3 knots) in the shallow waters off Sydney. Occasionally this flow can separate from the coast north of Sydney (at Seal Rocks, 32^oN) and spawn large clock-wise rotating eddies (200km diameter) (see Figure 2 of Middleton et al. (1997)) which can significantly reduce, and even reverse, the flow of the EAC

•    large scale winds which tend to oscillate Northeast/Southwest with the eastward passage of meteorological low pressure systems over central Australia producing northeast/southwest currents of up to 0.75 m/s (1.5 knots)

•    coastal trapped waves, generated by strong easterly winds blowing over the shallow waters of Bass Strait between southern Australia and Tasmania, which propagate northwards and pass Sydney 2 days after their generation. These currents can reach speeds of 0.6 m/s (1.2 knots) and oscillate with the same 5-15 day period of their generating winds

•    tides, which can have speeds reaching 0.6 m/s (1.2 knots) on spring tides and show most variation at the mouth of the estuary.

Thus the observed currents are composed of 3 forcing functions producing currents of similar magnitude but with different periods of motion superimposed on a quasi-steady feature with a dominant current. The challenge, therefore, was to separate these components and predict when the dominant southwards flow of the EAC may be changed.

The overall objective of these investigations was to contribute to a ‘call book’ of situations for each race area which both the racing coaches and the sailors themselves could study to adopt the best tactic for a given race. The call book would combine information of the meteorological conditions (constructed by the RYA meteorological consultant David Houghton and colleagues) and the currents.

THE TIDAL MODEL

Tidal models are relatively simple to construct and extremely useful tools for providing information on tides. Coupled to a ‘user-friendly’ graphical interface they allow the full spectrum of tidal changes to be easily interpreted.

An existing POL (Proudman Oceanographic Laboratory) computer code, the same one routinely used for forecasting tides and storm surges around the UK coastline (see Flather, 2000), was adapted in a 2-stage development to Sydney Harbour. First, a regional model of eastern Australia and the Tasman Sea was constructed with a horizontal resolution of 5 minutes latitude by 5 minutes longitude (approximately 10 km) extending from 27.5^oS to 37.5^oS. Bathymetry at the same resolution was extracted from the global DBDB5 public domain dataset. Tidal data for open boundary conditions was extracted from the global tidal model solution (FES95.2) of Le Provost et al. (1998). The four dominant tidal constituents were available, two diurnal (O₁, K₁) and two semi-diurnal (M₂, S₂). The regional model was used to produce a set of tidal harmonic constituents (surface elevations and currents) around the boundary of the Sydney Harbour model (considered the edge of Figure 1). Bathymetry for the Sydney Harbour model (100m resolution) was constructed from available charts (personal communication, Pauline Weatherall, British Oceanographic Data Centre) and the model run to produce a database of tidal harmonics. Both the models were checked for accuracy against available tidal data, mostly from coastal tide gauge sites.

The first application of the Sydney Harbour model was to the pre-Olympic regatta of September 1999. The harmonic constants database from the Sydney model was input to POLPRED a Windows-based software package for displaying any tidal state from gridded model elevations and currents (www.pol.ac.uk/appl/polpred.html). Its accuracy was first gauged by tidal predictions for Fort Dennison (Figure 2), a small island close to Sydney Harbour Bridge (between race areas A and B on Figure 1) and subsequently by comparison with the tidal stream atlas and additional measurements by the RYA coaches. Sample tidal current streams from POLPRED, a zoom view of the race areas B and C, are shown in Figure 3. The ease with which current patterns in a particular area can be displayed, added to the ability to quickly step forwards or backwards in time at any chosen time interval provided the team with insights, into for example, the effects of a particular headland on currents in a race domain, which had previously been unavailable.

Fig. 2: Harmonic tidal predictions (––) and model predictions (o o),

Fort Dennison, September 1999

Fig. 3: POLPRED screen showing a snapshot of tidal current speed and direction

The value of POLPRED, running on a laptop, was also that it could be used to predict currents at later times during the 12 month run-up to the Olympics when small groups would visit the harbour for additional training sessions, and to predict tidal currents during the Games period. The tidal data was also reformatted for input to another software tool used by the RYA, Graeme Winn’s “Deckman for Windows” package (www.islandcomputers.co.uk/Deckman.html) which allows real-time wind forecasts to be combined with tidal diamond information to produce surface current forecasts at any location within the estuary.

 OFFSHORE CURRENTS

Although the Sydney model does not entirely cover the offshore race areas E and F it does provide tidal stream data in sufficient detail. The main foci for offshore current prediction were a) the state of the EAC and b) the possibility of coastal trapped waves, as wind forecasts (and thus wind-driven currents) could be obtained from the excellent service provided by the Australian Bureau of Meteorology and interpreted by David Houghton.

Fig. 4:  AHVRR sea surface temperature (SST) showing section

a)  Monitoring of the EAC was through satellite imagery provided by the NERC Remote Sensing Group at Plymouth Marine Laboratory (PML) (www.npm.ac.uk/rsdas). Level 1b AVHRR satellite data at 4km resolution (GAC) covering the eastern coastline of Australia was automatically downloaded from the Satellite Active Archive and processed. Sea Surface Temperature (SST) was calculated from the AVHRR data within 3 hours of data reception using automatic processing software developed at PML. To avoid contamination by clouds, which are masked out during the processing, running 3 day median SST composites were automatically produced every day; median composites were chosen over mean composites as they eliminate spurious points within the individual images which make up the composite. A typical image produced during the Olympics is shown in Figure 4. SST across a 100km (60 nautical mile) eastward transect north of Sydney Harbour at 33^o 46.2’N (marked on Figure 4) was automatically extracted in ASCII format from the data field and plotted to monitor the daily changes in near shore temperatures (Figure 5). However, the imagery itself could only be used qualitatively to observe gross SST features (e.g. the shape of the EAC) because the large scale winds blowing along the coast could also modify the near shore sea surface temperatures. As Middleton et al. (1997) point out, if the winds blow for sufficient time (2-4 days), a NE wind will cause upwelling at the coast bringing cooler water to the surface and causing strong horizontal stratification, whereas winds from the SW will result in downwelling and a homogenising of near shore sea surface temperatures. In addition, through the establishment of wind-driven sea surface slopes, a geostrophic current will be established in the wind direction, i.e. southwards for a NE wind and northwards for a SW wind. Thus knowledge of the predicted wind fields and their history was important for interpreting the composite satellite images, but the images themselves were important to provide wide area information on the behaviour of the EAC itself.

Fig. 5: Daily composite SST from the section shown in Figure 4

b)  Coastal trapped waves (CTWs) are generated by strong winds in Bass Strait and propagate equatorward reaching Sydney 2 days after generation and so they are, to some extent, predictable. These are not waves in the normal sense of wind waves since they have little observable effect on the sea surface; they may be thought of as oscillating currents with a period of 5-15 days and alongshore variability of several hundred kilometres. These waves have been observed and studied off the east coast of Australia for many years (e.g. Church et al., 1986). The task for us was to find a quick ‘on the spot’ method for predicting the likely occurrence of CTW currents off Sydney Harbour. Fortunately Griffin and Middleton (1991) had devised a scheme to predict these currents with some success. They used linear regression models to explain the time-filtered (band-passed) variance in the observed currents in terms of past alongshore wind stresses in up to three remote regions. From experiments carried out with different linear regression models we opted for one of the simplest i.e.

          v_Sydney(t) = 0.93t(y₁,t-24) + 0.52t(y₁,t-48) + e(t)

      where v_Sydney(t) represents the CTW alongshore current off Sydney, t is time in hours,t(y₁) denotes the alongshore wind stress at Montague Island, some 285 km to the south, an automatic recording station with data available over the internet from the Bureau of Meteorology, and e(t) represents the error. Griffin and Middleton state that Montague Island windstresses gave the best single station predictor of CTW currents off Sydney in summer. Thus using observed winds (converted to windstress) at Montague Island 24 and 48 hours previously, a CTW current off Sydney was predictable. This could be reduced further by the simplification that windspeeds >30knots would produce currents >0.4m/s.

Procedures were thus in place to provide current estimates for the EAC, CTWs, wind forcing and tides. Assuming a steady state for the first three currents, these estimates were simply added together and superimposed on the tides. The main intention being to provide advance warning of likely changes in the expected EAC for race courses E and F.

THE CALL BOOK

Each day the RYA collected a measurement of sea surface temperature and the current speed from off the mouth of Sydney Harbour. These were used for ground truthing the offshore prediction scheme (the results of which were emailed to the team base in Sydney) and for determining tactics for the day.

Fig. 6: Model tidal predictions, Fort Dennison, September 2000

Analysis of the tidal patterns within Sydney Harbour led us to some simplifications of the tidal state. Tidal heights for Fort Dennison for September were plotted (Figure 6) to explain the tidal current atlas for the training period (September 1-16) and for the Olympic racing period (September 17-29). This showed that:

a)  the pattern of high and low waters has a repeat pattern with the Olympic racing period having the same sequence of tides as the preceding training period;

b)  this pattern starts with successive tides having nearly the same tidal range, moves to a period where successive tides alternate with one tidal range significantly larger (or smaller) than the next, and finishes with successive tides becoming more equal.

For the training period it can be seen that on September 3 the first tide of the waking day is big and the next is small. But by September 10 the first tide is small and the second tide large. With ‘Unequal’ tides ebb and flood flow will be stronger with the larger range, with the ‘Equal’ tides ebb and flood currents are similar but stronger than the ‘Unequal – large’ currents. It was also noted that the length of time between successive low and high waters is different depending on the tidal range. Typically, the time between successive ‘equal’ highs and lows is 6 hours, between successive highs and lows of ‘unequal’ small tides is less than 6 hours, and between successive highs and lows for ‘unequal’ big tides is more than 6 hours. From September 17, when racing started, this cycle is repeated.

Thus, for simplicity, tidal streams representative of the two different tidal conditions were produced:

a)      for ‘equal’ successive tides (based around September 2)

b)      for ‘unequal’ successive tides (based around September 9)

‘Equal’ tide currents for days of equal tides and the ‘Unequal’ tide currents for days of unequal tides. These were identified as follows: ‘Equal’ 1-4, 14-18 and 27-30 September; ‘Unequal’ 5-13 and 19-26 September

To keep it all as simple as possible 2 sets of current charts were produced – one for courses B and C and another for D. For each set tidal current fields for three tidal cycles, for the ‘Equal’, ‘Unequal Small’ and ‘Unequal Large’ tidal range were produced. Each set showed currents at 2-hourly intervals from HW-6 hours to HW+6 hours (HW relative to Fort Dennison). An example for course D is shown in Figure 7.

Fig. 7: Tidal current vector output for course D

These currents were combined with detailed wind analyses to indicate what tactics to adopt with winds from different sectors and currents at different tidal states. Figure 8 shows one of the pages from the call book for course D under NE winds. Pages for wind direction from eight sectors (N, NW, E, SE, … etc.) were produced for each course and current state for each race determined from the tidal streams.

SUMMARY

Each day a briefing session was held first thing in the morning to discuss the latest information and to advise the sailors of their courses and what their tactics should be. However, once the sailors are competing they have to rely on their skill and experience to interpret what they have heard and what they see on the course.

Does all this preparation and support make a difference? The team returned the best haul of medals of any GB team ever, three gold and two silver medals. Clearly the sailors are of extremely high calibre but John Derbyshire, the Team Manager, considered that this support made a significant contribution to the success of the team.

Fig. 8: Sample Call Book page for course C

ACKNOWLEDGEMENTS

We are grateful to the Satellite Active Archive for access to the imagery used, and would like to thank Catherine Edwards for exploring the use of the regression models for CTW prediction and Robert Smith for redrawing the call book figure.

REFERENCES

Church, J.A., Freeland, H.J. & Smith, R.L. 1986. Coastal-trapped waves on the east Australian continental shelf, Part I. Propagation of modes. Journal of Physical Oceanography, 16, 1929-1943.

Das, P., Marchesiello, P. & Middleton, J.H. 2000. Numerical modelling of tide-induced resildual circulation in Sydney Harbour. Marine and Freshwater Research, 51(2), 97-112.

Flather, R.A. (2000). “Existing operational oceanography.” Coastal Engineering, 41(1-3): 13-40.

Griffin, D.A. & Middleton, J.H. 1991. Local and remote wind forcing of New South Wales inner shelf currents and sea level. Journal of Physical Oceanography, 21, 304-322.

Le Provost, C., Lyard, F., Molines, J.M. & Genco, M.L. 1998. A hydrodynamic ocean tide model improved by assimilating a satellite altimeter-derived data set. Journal of Geophysical Research, 103, 5513-5529.

Middleton, J.H., Cox, D. & Tate, P. 1997. The Oceanography of the Sydney Region. Marine Pollution Bulletin, 33 (7-12), 124-131.

Roger Proctor

is a physical oceanographer at the Proudman Oceanographic Laboratory and has been a Principal Scientific Officer (Band 4) since 1990. He obtained his PhD in Numerical Modelling at Liverpool University in 1981 and previously graduated (1976) with a degree in Mathematics from Teesside Polytechnic. He has provided oceanographic support to Team GBR at each Olympics since the 1988 Korean Olympic Games.

Tim Smyth

is a remote sensing scientist at the Plymouth Marine Laboratory and has been in post since 1997. He obtained his PhD in Radar Meteorology in 1997 having previously graduated (1994) with a degree in Physics and Meteorology at Reading University. He is a fellow of the Royal Meteorological Society.

Hamish Willcox

was the Team GBR 470 coach during the Sydney 2000 Olympics. He is currently the weather program manager for ‘One World America’s Cup’. Prior to this he played the same role for the America’s Cup Challenger ‘Prada’ from Italy. In the 1980s Hamish won 3 Olympic 470 class world championships and during the 90s spent 7 years as New Zealand National Yachting Coach. Hamish has attended every Olympic Games from 1984 to 2000 as either a sailing reserve or coach.

Colin Bell

is a software developer at the Proudman Oceanographic Laboratory where he was worked for nearly 10 years, having graduated from Leeds University with a degree in Computer Science. For the last two years, he has also been head of the Applications Group which deals with exploitation of the laboratory’s science.