Mesoscale oceanographic data analysis and data assimilative modelling with application to Western Australian fisheries

Project Number:



CSIRO Oceans and Atmosphere Hobart

Principal Investigator:

David Griffin

Project Status:


FRDC Expenditure:





Understanding the influence of environmental effects on recruitment is an important aspect of fisheries stock assessment (and hence sustainable management) to help interpret whether fluctuations in recruitment are due to environmental effects or to the impact of fishing on the spawning stock. In many cases, the effect of breeding stock cannot be detected unless environmental effects have been taken into account. At present the measure of variation in the strength of the Leeuwin Current along the WA coast from Shark Bay to Esperance is based on monthly and annual mean sea levels at Fremantle, which are highly correlated with sea levels at other locations and have few missing values. The impacts of this and other environmental factors such as westerly winds are presently being examined in isolation rather than obtaining an overall measure of the impact on the circulation. This project will integrate these separate oceanic and atmospheric processes into circulation models and thus result in improved estimates of ocean currents, water temperature and salinity for regions adjacent to important fisheries, providing a better understanding of the effects of environmental variability on recruitment as well as better prediction of recruitment levels. The current assessment of the impact of the breeding stock on the level of recruitment in the western rock lobster assumes that larvae hatched from all areas contribute to the settlement at each location. Although extensive mixing of larvae during the larval phase occurs, it is possible that larvae hatched from certain areas may contribute proportionally more to the puerulus settlement in different regions. This project will provide an improved understanding of the relative importance of the breeding stocks from different regions, enabling better assessment of the impact of the breeding stock and an improved stock-recruitment relationship which is fundamental for the proper management of the fishery. There is therefore a need for a physical oceanographic analysis tool that can be used to test theories on the influence of favourable/unfavourable larval advection, or temperature, or larval mortality. A well understood physical/biological interaction would enable efficient ocean and fisheries observation and monitoring programs to be established to maximize the skill of larval survival predictions.


1. To develop algorithms for operational estimation of near-surface currents and temperatures off south-western Australia based on satellite altimetry and thermometry.

2. To develop and test a three-dimensional data-assimilating model of ocean dynamics off Western Australia, to be run in hindcast mode, archiving data for the last ten years.

3. To run tracking scenarios for rock lobster larvae to describe larval behaviour under different environmental conditions (extended to other larvae as time permits).

4. To provide advice on management issues that may be addressed by improved ocean understanding, such as the spawning locations of successful larvae, and correlations between larval success and ocean conditions.

Mesoscale oceanographic data analysis and data assimilative modelling with application to Western Australian fisheries

Final Report
ISBN:1 876996 01 3
Author(s):David Griffin
Date Published:September 2001

The more that is understood about the factors controlling the abundance of an exploited fish stock, the more optimally it can be harvested for sustainable yield and profit. It has been known for some time that catches of western rock lobster are closely related to the number of larvae surviving their year at sea and settling as puerulus on the coast, and that the variation in settlement, in turn can be statistically predicted using several types of ocean data. What is not understood is why sea level, for example, should be a predictor of larval survival. This needs to be understood so that insight into reasons why the correlation might break down (as it did in 1998) can be gained, and so that a better predictor can be found. The value of a reliable indicator of the environmental influence on larval survival is that, for example, a year of very poor larval settlement can then be correctly attributed to either over-fishing of the breeding stock, or poor survival rates at sea.

This project addressed the question of why larval survival rates are so variable by taking a process-based, modelling approach, as distinct from the statistical, or correlative approaches taken to date. The modelling approach has only recently become feasible, for two reasons. One is that the computing demands are high, but the more significant recent advance is the advent of satellite techniques for mapping near-surface ocean currents. We used these maps to calculate where winds and ocean currents from 1993 to 1998 carried numbers of individual western rock lobster larvae, from hatching to far offshore, then back to the coast. 

Crucial to the success of this project was that relatively much has been learnt, from decades of sampling from ships, about the behaviour of larval western rock lobsters in the deep ocean. In particular, we needed to know the details of when larvae rise to the surface and when they descend to depth, and how this varies with larval age, time of day, phase of moon, etc. We also needed to know what triggers a late-stage larva to metamorphose to the non-feeding, fast-swimming puerulus stage that is found settling on inshore reefs. This trigger, however, is unknown, so our model simply assumes that all larvae at least 270 days old make this transition if they find themselves over the continental slope near new moon, a behaviour that is consistent with available observations.

In the first phase of the project, many types of ocean data were assembled, and two techniques developed for making accurate maps of the ocean currents. The first technique was relatively straightforward: observations of sea level height and surface temperature were used to estimate the surface currents directly via approximations of the physical equations. The second technique is called data-assimilative modeling. The satellite data were ‘assimilated’ into a hydrodynamic model.

Our data-assimilating model of the ocean currents of Western Australia is the first of its kind in Australia. However, it took somewhat longer to complete than originally planned, and we still consider it to be a work in progress. Work on this model was de-prioritised in favour of using the maps of ocean near-surface currents diagnosed directly from the satellite data, which turned out, fortunately, to be more accurate than hoped for. The accuracy of these ‘altimetric’ current maps was assessed by comparing them with the velocities of satellite-tracked drifting buoys, and animations of ocean thermal imagery. 

The current maps and (also only recently available digitally) daily wind maps were then combined with information on larval behaviour and many simulations were performed of the fate of six year-classes of lobster larvae. The simulations confirm the importance of the summer southerly winds in transporting larvae ‘upstream’ against the mean onshore and southward flows that exist just below the surface, and which help return larvae to the coast. The simulations also highlight the role of energetic eddies, which mix the larvae at velocities much in excess of the larger-scale flows.

It is the vigorous mixing by eddies in our simulations that produce the result that the location of hatching of larvae quickly becomes immaterial to its destiny. To test this hypothesis was one of our primary objectives because of its relevance to the potential benefit of preferentially protecting certain spawning regions.

With regard to explaining the observed correlation of sea level with larval settlement,  our simulations confirm that sea level does serve as an indicator of both the strength of the Leeuwin Current, and the intensity of eddies associated with it, but does not support the hypothesis that the direct (transporting) influence of the currents on the larvae is responsible for the large (ie five-fold) changes observed in how many return to the coast.

So the mystery of why larval settlement correlates with sea level remains, although we now have a clearer picture of how any return at all. 

We concluded our project with a very preliminary study of the potential importance, to larval survival, of the inter-annual variability of food availability. For this we used the recently launched SeaWiFS satellite that senses ocean colour, from which near-surface chlorophyll abundance can be estimated. These data show that there was less phytoplankton in the water in the summer of 1997-98 than in the next two summers, perhaps explaining why the settlement in 1998 was very low, even though the prediction based on sea level was for average settlement. 

The next step to take is to include temperature- and prey field-dependent larval growth and mortality in an advection model such as the one developed here. In addition to explaining the inter-annual variability of settlement, the inclusion of growth and mortality in the model could also change our finding about the importance of hatching location, because of the regional differences that exist in primary production.

To complete this project, we have produced an educational CD-ROM with all the data assembled, along with results of the larval transport simulations, presented in the form of movies that can be viewed on any computer. The CD can be browsed at

Keywords: Western Rock Lobster, larval advection, ocean currents, altimetry, data assimilation.