Transport and recruitment of silver-lip pearl oyster larvae on Australia's North West shelf.

ABSTRACT Silver-lip pearl oyster (Pinctada maxima) spat surveyed in the Eighty Mile Beach section of the North West Shelf have been used in conjunction with outputs from a particle dispersion model to identify likely spawning grounds. The dispersion model consisted of a 3-dimensional regional circulation model in which large numbers of individual particles were tracked over the period 1994 to 1999. From the settlement areas defined by the spat data, larvae were tracked back in time over their estimated pelagic phase of 24-31 days within the main spawning period of mid October to late December. The reverse calculation was also undertaken looking at larval dispersion from known broodstock populations. Results demonstrate that large tidal currents in the region move larvae back and forth across the shelf, whereas lower frequency currents influence their net transport. Whereas some model larvae traveled more than 60 km, most were transported less than 30 km. The model results suggest that spawning in the Eighty Mile Beach region is concentrated around the recently surveyed broodstock distribution between 8 and 15 m depth, with potential smaller contributions from the northeast. These spawning events are likely to lead to successful recruitment locally and alongshore to the southwest. They also feed larvae into neighboring shallow coastal environments (through tidal oscillations) and deeper waters to the west (-20 m). However, spat abundances seem to be low in these areas, suggesting that recruitment is strongly limited by habitat availability and possibly high mortality rates in shallow water. High local abundances of broodstock and spat observed occasionally in deeper water (-30 m) seem to be supported by intermittent larval transport from inshore populations. However, spawning in this area seems to contribute little to recruitment in the inshore populations.

KEY WORDS: Pinctada maxima, oysters, larvae, larval transport, recruitment

INTRODUCTION

The pearling industry on Australia's North West Shelf (NWS) relies substantially on shell caught from the inshore waters of Eighty Mile Beach (Fig. 1). The most sought after species is the silver-lip pearl oyster, Pinctada maxima (Jameson) that produces the silver-white South Sea pearl. Oyster beds on the NWS provide "wild shell," which is collected by divers, before being seeded and placed back on the seabed in framed net enclosures or "panels." After recovering for a few months, they are transported to farms where the pearls can develop. These activities are managed as a gauntlet fishery, where oysters are caught at the optimum size for seeding (120-160 mm) leaving the larger mother-of-pearl (MOP, >175 mm) as breeding stock.

[FIGURE 1 OMITTED]

Recently surveyed MOP distributions show increasing abundance with water depth, whereas the abundance of recruits (120-160 mm) tends to decrease (Hart & Friedman 2004). This decoupling of broodstock and recruitment might be explained by net onshore transport of larvae spawned in deeper waters being offset by high rates of fishing and natural mortality in shallow waters. This is the basis for the long-standing belief within the pearling industry that deep "unfished" stocks could be broodstock source for the commercially fished inshore stock. However, prior to this study, such a hypothesis had not been tested.

It is clearly critical to the management of the fishery to understand the P. maxima reproductive cycle (Rose et al. 1990) and establish the geographical extent of the spawning stock supporting the fishery. The objective of this study is to use a particle dispersion model in combination with results from pearl oyster surveys to investigate larval transport and settlement. Specifically, potential settlement distributions have been estimated based on surveyed MOP spawning populations and compared with results from spat abundance surveys. The reverse calculation has also been made, estimating potential spawning distributions based on spat abundance data and comparing these with the MOP distribution data. This has allowed us to test the hypothesis that deep-water MOP stocks support inshore recruitment, and population connectivity relationships more generally, within P. maxima stocks.

MATERIALS AND METHODS

Larval dispersion patterns were modeled by tracking particles transported by ocean currents estimated from a circulation model for the North West Shelf. Particle trajectories were used to estimate both potential settlement sites and potential spawning sites. Spawning distributions were based on a MOP survey conducted in September 2001, whereas settlement distributions were based on 3 years (2001 to 2003) of spat surveys. These surveys were focused off Eighty Mile Beach, where most of the fishery is concentrated.

Circulation Model

The circulation model was developed as part of the North West Shelf Joint Environmental Management Study (NWSJEMS 2002, Condie et al. 2003) and was based on code referred to as MECO (Model for Estuaries and Coastal Oceans). MECO is a generalpurpose finite-difference hydrodynamic model applicable to scales ranging from estuaries to ocean basins. It uses a curvilinear orthogonal grid in the horizontal and fixed "z" coordinates in the vertical. A comprehensive description of the underlying theory is provided in the MECO Scientific Manual (Herzfeld et al. 2002). MECO has found previous applications in systems such as the Port Phillip Bay (Walker 1999), Bass Strait, the Great Australian Bight and Southeastern Australia (Bruce et al. 2001) and the Gulf of Carpentaria (Condie et al. 1999).

MECO was implemented for the NWS on a rotated latitude-longitude grid with horizontal resolution of approximately 10 km (Condie et al. 2005a). This grid extended from Cape Cuvier in the southwest to the Bonaparte Archipelago in the northeast and well beyond the shelf break (Fig. 1). The vertical resolution expanded from 3 m near the surface to a maximum of 200 m at its maximum depth of 1200 m. Truncating the depth at this level had little effect on the circulation, but significantly improved computational times. The bathymetry was prescribed by spatially averaging a 30 sec (0.9 km) product provided by Geosciences Australia onto the model grid.

Inputs required by the model included forcing caused by wind, atmospheric pressure gradients, and open boundary conditions such as temperature, salinity, and sea level. Wind fields were taken from the NCEP-NCAR 40-y reanalysis dataset (Kalnay et al. 1996). These fields had a 12 hourly time-step and a spatial resolution of 1.8[degrees], which were linearly interpolated onto the model time-step and model grid. The interpolated product generally showed good agreement with locally measured winds at subdiurnal frequencies (Fig. 2a). However, smaller scale processes, such as the daily sea breeze and occasional tropical cyclones, were less adequately resolved. Winds during the major spawning period of mid October to December were predominantly from the west.

[FIGURE 2 OMITTED]

Temperature and salinity fields around the lateral boundaries of the model were interpolated from a global circulation model known as the Australian Community Ocean Model or ACOM (Schiller et al. 2000). In the absence of reliable surface fluxes, interior temperatures and salinities were modified by relaxing them towards ACOM values with a 10-day relaxation timescale. Sea levels on the boundaries were also taken from ACOM output, with the addition of tidal constituents derived from tide gauge observations around the Cape Cuvier and Bonaparte Archipelago areas in combination with global tidal model estimates along the offshore boundaries (Eanes & Bettadpur 1995).

Particle Dispersion and Tracking

The modeled currents provided an indication of the instantaneous movements of larvae in the water column. However, additional information was required to estimate advection and dispersion patterns. Individual-based particle-tracking techniques were adopted for this purpose. In the absence of any detailed information on larval swimming behavior, all particles were assumed to be nonmotile and neutrally buoyant. These assumptions are relatively easy to justify in the extremely energetic environment of Eighty Mile Beach, where tidal currents and turbulent mixing velocities would normally be expected to far exceed realistic larval swimming speeds (e.g., Condie 1999).

A large number (-[10.sup.5]) of neutrally buoyant particles were initially seeded randomly through the water column across the model domain, with highest concentrations on the inner shelf. The circulation and particle movement calculations were then conducted simultaneously, with particle positions being updated every l0 min by the interpolated model current velocities. A random walk component was also added to the trajectory to represent the dispersive influence of turbulent motions not resolved by the circulation model. Each particle was individually tracked and its location recorded every 3 hours.

Particles followed complex paths that were sensitive to their initial location. Individual trajectories therefore provided a limited view of likely dispersion patterns associated with large numbers of larvae. A statistical description of the dispersion results was therefore developed from large numbers of trajectories (Condie et al. 2005b). From the spawning areas defined by the MOP data, larvae were tracked over their estimated pelagic phase of 24-31 days within the main spawning period of mid October to late December (Rose et al. 1990). More specifically, particles within the spawning areas (defined later) at midnight on October 15 were tracked for 24 days, after which their locations were recorded every 3 hours over the following 7 days. The process was then repeated starting 3 h later (i.e., 3:00 AM on October 15) and then at all subsequent 3 hourly steps until the end of the 31-day pelagic phase corresponded to December 31. To ensure that the results were not unduly dominated by those particles remaining within spawning areas for extended periods, each particle was only permitted to initiate a new trajectory 7 days after it had initiated the last one.

The combination of all recorded locations provided the final estimate of potential settlement distribution. This differs from an estimate of actual settlement, which would require additional factors such as habitat distribution, habitat preference, and larval mortality to be taken into account. The approach taken to estimate spawning areas defined by the spat data was the same as that described above, except that trajectories were tracked back in time. Results were obtained for all modeled years (1994 to 1999), which were further amalgamated into an average distribution over the entire 5-year modeling period.

Biological Surveys

A pearl oyster stock survey conducted in September 2001 revealed that distributions are largely limited by habitat availability (Hart & Friedman 2004). This finding, combined with the longevity of the large broodstock (or MOP) (Rose et al. 1990), suggests that distributions are likely to be relatively stable and hence that the 2001 MOP distributions represented an acceptable proxy for the modeling period (1994-1999). A measure of the relative abundance of MOP was based on the average number of shell collected for each hour the vessel drifted within the target populations (Hart & Friedman 2004).

Spat surveys (measuring 0+ and 1+ age classes) were undertaken over the 3-year period from 2001 to 2003 using the method described in (Hart & Joll 2005). Because only partial coverage was provided by each survey, the three were combined here into a single estimate of settlement distribution. In brief, the spat methodology involved examining adult shell collected by commercial divers, and counting any attached spat (Hart & Joll 2005). These counts were assumed to be representative of relative spat densities across all suitable habitats available in the area. The average number of spat per shell within a 10-nm grid cell therefore provided a measure of spat relative abundance. Information on the spat year classes were not required for the current analysis, however were useful for evaluating year-to-year trends in settlement.

RESULTS

Modeled Currents and Sea Levels

Instantaneous current patterns on the NWS were dominated by strong tides, with speeds approaching 2 m [s.sup.-1] during the spring tide (Fig. 3a). In the main fishing areas around Eighty Mile Beach, the tidal movements were predominantly in the cross-shore direction and diminished with distance offshore. The model currents have been shown to compare favorably with observations around the Dampier region to the west (Condie et al. 2005a). However, no direct current observations have been taken in the Eighty Mile Beach region (Godfrey & Mansbridge 2000). The best indication of model performance in the local region was therefore provided through comparisons of sea level variability.

[FIGURE 3 OMITTED]

Sea level was strongly dominated by the very large tidal signal, with ranges up to 10 m in King Sound. The model generally reproduced the tidal sea-level signal quite accurately, with correlation coefficients increasing from r = 0.87 at Broome to r = 0.96 at Port Hedland (Condie et al. 2005a). However, it is the lower frequency component that largely controls net transport over the pelagic larval phase (24-31 days). Because subtidal currents are generally highly correlated with subtidal sea-level differences, model performance in the Eighty Mile Beach region was examined in terms of sea-level differences between Broome and Port Hedland. This analysis revealed a strong coastally trapped wave signal in the model and observations, with a period of approximately 14 days in both instances. However, it is the variability over the typical larval phase that is of most immediate interest. For example, comparisons of the running mean of the sea-level difference over 28 days produced a correlation of r = 0.89 (Fig. 2b).

Monthly averaged surface currents during the period October to December (main spawning period) were typically northeastward at a few centimeters per second (Fig. 3b). However, close to Eighty Mile Beach the surface flow was directed offshore, suggesting a net upwelling over the monthly timescale. These currents patterns were a response to the prevailing westerly winds and are consistent with the few satellite tracked ocean drifters observed in the area at this time of the year (Cresswell et al. 1993). The depth-averaged currents were weaker and tended to form a clockwise gyre off Eighty Mile Beach (Fig. 3c). The resulting flow close to the coast was predominantly westward, counter to the prevailing winds. This monthly averaged pattern persisted throughout the spawning period (Fig. 3d).

Estimates of Settlement Distribution

The surveyed MOP distribution (Fig. 4a) showed low to moderate relative abundances (0-60 [h.sup.-1]) around the 10 m depth contour, with a tendency towards higher values at the northern extremity of the sampling area (60-120 [h.sup.-1]). However, the largest abundances (110-160 [h.sup.-1]) were found in a patch further offshore around the 30-m depth contour. Potential larval settlement distributions were estimated by assuming that MOP survey abundances were proportional to the spawning population (although results were qualitatively similar when all cells with surveyed MOP were given equal weighting).

[FIGURE 4 OMITTED]

Results averaged across the model runs (1994 to 1999) indicated two main potential settlement areas (Fig. 4b). The larger was centered just inshore of the 10 m depth contour (121[degrees]24'E, 19[degrees]6'S), with the smaller centered just inshore of the 20 m depth contour (121[degrees]7'E, 19[degrees]11'S). Because larvae could potentially settle at any time during the tidal cycle, the size of these areas was largely determined by the dimensions of the tidal ellipse. The relative settlement rate in the two areas varied from year to year and in 1994 the model results showed more settlement offshore than onshore (not shown). However, the predicted settlement areas were not associated with the two distinct spawning areas. Rather, most MOP regions contributed to both settlement areas (C to F in Fig. 4a and Fig. 5a) reflecting temporal and spatial variability in flow patterns within the 2.5-month spawning period. The main exception to these trends was the surveyed area near the 30 m depth contour, whose modeled larvae tended to move into deeper water ([greater than or equal to] 40 m) to the southwest without contributing to either of the main settlement areas (A in Fig. 4a and Fig. 5a).

[FIGURE 5 OMITTED]

The model predicted settlement region covers nearly all the areas surveyed for spat, with predicted high settling rates corresponding with the main survey area (cf. Fig. 4b and 4c). However, spat surveyed further offshore and to the north (Fig. 4c) corresponded to low settlement rates and may have originated from MOP outside the MOP survey area. There were also large areas of predicted settlement, both near-shore and offshore, that were not in the area of the spat surveys. Whereas spat may have existed in some of these areas, the limited availability of suitable habitat would likely have resulted in low abundances.

Estimates of Spawning Distribution

The surveyed spat distribution, as described by Hart and Joll (2005), tended to follow the 10-m depth contour along Eighty Mile Beach at low to medium relative abundances (0-20 per thousand shells) with a bias towards deeper water to the north (Fig. 4c). There were a few high abundance of cells (20-30 per thousand shells) in this zone, with more in habitats further offshore around the 15-m and 30-m depth contours (Fig. 4c). Potential spawning distributions were estimated by assuming that the spat survey abundances were proportional to the settling population (although results were qualitatively very similar when all cells with surveyed spat were given equal weighting).

Estimated spawning areas associated with the surveyed spat distributions were centered just inshore of the 10-m depth contour (121[degrees]24'E, 19[degrees]6'S), but relatively high values extended offshore almost to the 20-m depth contour and onshore almost to the coast (Fig. 4d). The size of these areas was again a function of the tidal ellipse and there was very limited interannual variability (not shown). Spat counted in the main survey area along the 10-m depth contour were mainly spawned in this area (C, D and E in Fig. 4c and 5b), although cell E also received model larvae from further to the northeast. The model indicated that spat surveyed further south (B) and north (F) would have come from spawning in deeper water, whereas those offshore around the 30-m depth contour (A) would have come from marginally shallower depths to the northeast (Fig. 4c and 5b).

Apart from extending into very shallow water, the spawning area estimated from the surveyed spat (Fig. 4d) was consistent with the surveyed MOP distribution (Fig. 4a). The main exception was the offshore MOP (A in Fig. 4a), which the model suggests spawn into deeper water outside of the spat survey region (Fig. 5a).

DISCUSSION

The overall picture emerging from the model results is that spawning in the Eighty Mile Beach region is concentrated around the surveyed MOP distribution between 8- and 15-m depth, with potential smaller contributions from further northeast. These spawning events are likely to lead to successful recruitment locally and alongshore to the southwest so that the main pearl oyster producing populations are likely to be self seeding. These spawning events also feed larvae into neighboring shallow coastal environments (through tidal oscillations) and deeper waters to the west (~20 m). However, spat abundances appear to be lower in these areas, suggesting that recruitment is either limited by habitat availability and possibly increased mortality in shallow water, or else spat larvae preferentially select other settlement habitats (as opposed to live adult shell). High local abundances of MOP and spat observed in deeper water (~30 m) appear to be supported by intermittent larval transport from inshore populations and may be further assisted by high quality habitat and low mortality rates. However, spawning in this area seems to contribute little to recruitment in the inshore populations.

The validity of the settlement and spawning estimates is dependent on a number of factors. The first is the accuracy of the predicted circulation. Whereas the model has been shown to perform relatively well in the Dampier region (Condie et al. 2005a), the data available for validation near the fishing grounds was limited to sea-level information. The second factor is the comprehensiveness of the MOP and spat surveys. For example, some habitats may have been under sampled and others may still be undiscovered.

A third source of model uncertainty is associated with larval behavior. For example, larvae remaining close to the surface would tend to follow the wind dominated circulation to the northeast (Fig. 3b), rather than the depth averaged circulation (Fig. 3c). This seems unlikely in this region given that tidal motions drive energetic vertical mixing out to at least the 30-m depth contour and that the MOP and spat surveys suggest transport to the southwest. A potentially more significant behavioral characteristic would be any ability of larvae to select settlement sites. During the week long period in which settlement can occur (Rose et al. 1990), tidal motions carry individual larvae back and forward up to 20 km in the cross-shore direction. If larvae encountered preferred habitats over depths from around 8-15 m deep during this period, then the estimated settlement distribution based on the observed MOP (Fig. 4b) would be very similar to the observed spat distribution (Fig. 4c).

[FIGURES 3-4 OMITTED]

The results of the model suggest that the long-standing hypothesis within the pearling industry, namely that deeper "unfished" stocks are a broodstock source for the commercially fished inshore stock, is not likely to be true. The inshore stocks appear to be self-sustaining, and may even be providing larvae to deeper stocks in irregular recruitment events. These findings have important management implications, particularly if fishing for MOP were to recommence at some time in the future. However, current management and fishing strategies seem to protect the critical broodstock that support the long-term sustainability of the existing "gauntlet" fishery.

ACKNOWLEDGMENTS

The authors thank the Pearling industry, particularly Paspaley Pearls and Broome Pearls Pty Ltd for allowing the use of their vessels and pearl divers to conduct the spat sampling and research surveys. This work was supported by grants from the Pearling Industry of Western Australia, the Western Australian Fisheries Department and the Fisheries Research and Development Corporation (FRDC).

LITERATURE CITED

Bruce, B. D., S. A. Condie & C. A. Sutton. 2001. Larval distribution of blue grenadier (Macruronus novaezelandiae Hector) in south-eastern Australia: further evidence for a second spawning area. Mar. Freshwat. Res 52:603-610.

Condie, S. A. 1999. Settling regimes for non-motile particles in stratified waters. Deep-Sea Res. 1. 46:681-699.

Condie, S. A., J. Andrewartha, J. V. Mansbridge & J. Waring. 2005a. Modeling circulation and connectivity on Australia's North West Shelf: North West Shelf Joint Environmental Management Study Technical Report, CSIRO Marine Research, 61 pp.

Condie, S., C. Fandry, D. McDonald, J. Parslow & K. Sainsbury. 2003. Linking ocean models to coastal management on Australia's North West Shelf. EOS, Transactions, American Geophysical Union 84:49-53.

Condie, S. A., N. R. Loneragan & D. J. Die. 1999. Modeling the recruitment of tiger prawns (Penaeus esculentus and P. semisulcatus) to nursery grounds in the Gulf of Carpentaria, northern Australia: implications for assessing stock-recruitment relationships. Mar. Ecol. Prog. Ser. 178:55-68.

Condie, S. A., J. Waring, J. V. Mansbridge & M. L. Cahill. 2005b. Marine Connectivity patterns around the Australian continent. Environ. Mod. Software 20:1149-1157.

Cresswell, G., A. Frische, J. Peterson & D. Quadfasel. 1993. Circulation in the Timor Sea. J. Geophys. Res. 98:14379-14389.

Eanes, R. & S. Bettadpur. 1995. The CSR 3.0 global ocean tide model. Center for Space Research, Technical Memorandum, CST-TM-95-06.

Godfrey, J. S. & J. V. Mansbridge. 2000. Ekman transports, tidal mixing, and the control of temperature structure in Australia's northwest waters. J. Geophys. Res. 105:24021-24044.

Hart, A. M. & L. JolI. 2005. Growth, mortality, recruitment, and sex-ratio in wild stocks of silverlipped pearl oyster Pinctada maxima (Jameson) (Mollusca: Pteriidae) in Western Australia. J. Shellfish Res. (this volume).

Hart, A. M. & K. J. Friedman. 2004. Mother of pearl shell (Pinctada maxima): stock evaluation for management and future harvesting in Western Australia, FRDC Project 1998/153, Fisheries Research Contract Report No. 10, Department of Fisheries, Western Australia. 84 pp.

Herzfeld, M., J. Waring, J. Parslow, N. Margvelashvili, P. Sakov & J. Andrewartha. 2002. MECO, Model of estuarine and coastal oceans, V4.0, scientific manual. CSIRO Marine Research.

Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. Iredell, S. Saha, G. White, J. Woollen, Y. Zhu, M. Chelliah, W. Ebisuzaki, W. Higgins, J. Janowiak, K. C. Mo, C. Ropelewski, J. Wang, A. Leetmaa, R. Reynolds, R. Jeune & D. Joseph. 1996. The NCEP/NCAR 40-year reanalysis project. Bull. Amer. Meteor. Soc. 77: 437-471.

NWSJEMS. 2002. North West Shelf joint environmental management study interim report. Available at: http://www.marine.csiro.au/ nwsjems/reports/index.btml. 140 pp.

Rose, R. A., R. E. Dybdabl & S. Harders. 1990. Reproductive cycle of the Western Australian silverlip pearl oyster, Pinctada maxima (Jameson) (Mollusca: Pteriidae). J. Shellfish Res. 9:261-272.

Schiller, A., J. S. Godfrey, P. C. Mcintosh, G. Meyers & R. Fiedler. 2000. Interannual dynamics and thermodynamics of the Indo-Pacific Oceans. J. Phys. Oceanogr. 30:987-1012.

Walker, S. J. 1999. Coupled hydrodynamic and transport models of Port Phillip Bay, a semi-enclosed bay in south-eastern Australia. Mar. Freshwat. Res. 50:469-481.

SCOTT A. CONDIE, (1) * JIM V. MANSBRIDGE, (1) ANTHONY M. HART (2) AND JOHN R. ANDREWARTHA (1)

(1) CSIRO Marine and Atmospheric Research, GPO Box 1538, Hobart, Tasmania 7004, Australia; (2) Western Australian Fisheries and Maine Research Laboratories, PO Box 20, North Beach, Western Australia 6020, Australia

* Corresponding author: E-mail: [e-mail]