Oceans Policy Development Csiro Contract March 1998

DRAFT

OCEANS POLICY DEVELOPMENT CSIRO CONTRACT MARCH 1998

LARGE MARINE DOMAINS OF AUSTRALIA’S EEZ

Vincent Lyne0, Peter Last0, Roger Scott0, Jeff Dunn0, David Peters1, Trevor Ward0

0CSIRO Division of Marine Research

1Department of Environment & Land Management, Tasmania

EXPLANATORY NOTES TO MAP SERIES

Methods

Scope and Philosophy of Current Project

The IMCRA-derived provincial regionalisation (CSIRO 1996) of the shelf region (coast to the shelf-break at the 200m isobath) forms the basis of the current Large Marine Domain (LMD) regions on the shelf.

The prime focus of the current effort is a derivation of the offshore (offshore of the 200m isobath) LMDs and their integration with those on the shelf.

We make no attempt in this limited project to derive any separate regionalisation of the slope region. It suffices to note here that preliminary inspection of extensive research expedition data from the slopes of Western Australia and New South Wales indicate intricate depthwise structuring that may not be related to surface offshore water properties or benthic substrate on the shelf. Thus a prime limitation of the current regionalisation is an assumed continuity of the LMD regions across the slope.

Ideally the philosophy of the current offshore regionalisation should follow that for the shelf (CSIRO 1996). That philosophy relied on a hierarchical framework with regionalisations at the provincial level derived from distributional data for endemic fish species. Whilst datasets on commercial fish species and extensive research expedition data do exist for the offshore region, the only currently available national dataset is that of chlorophyl derived from the NOAA CZCS satellite flown from the late 70 to 1986. Given that numerous oceanographic and geological datasets do exist, an alternate strategy for the offshore regionalisation was required. The current IMCRA (Version 3.2) uses a physical oceanographic regionalisation for the water column from the surface to 50m for the so-called pelagic regionalisation offshore of the shelf-break. For the demersal regionalisation the AGSO-derived slope analysis was used. The specification for this project was for one offshore regionalisation that integrated the water column and demersal regions which was also linked to the regions derived for the shelf.

The achievable strategy, given limitations of time and resources, was a multivariate analysis of the existing oceanographic, geological and chlorophyl datasets for the offshore region which was then integrated with the demersal regionalisation for the shelf.

Regionalisation Strategy

Key tasks in implementing the multivariate analysis comprised:

1.Collation and inspection of datasets. Criteria for selecting appropriate datasets included:

Relevance to regionalisation

Spatial extent and resolution

Ease of processing

IMCRA-derived datasets were used as far as possible. When this was not practical alternate datasets were sought and analysed.

2.Subsetting of suitable datasets onto the analysis grid. The grid extent was:

Latitude:-50oS to 0o

Longitude100oE to 180oE

Resolution0.5 o

Where the resolution of the data was finer than the grid an average was computed over the grid cell. In the case of the oceanographic data, an optimal interpolation procedure was used that averaged over a set number of the closest neighbours (CSIRO, 1996).

3. Gridded datasets were compiled into one column-oriented file that became the primary input for the multivariate analysis.

Datasets

Datasets collated and examined comprised:

Dataset / Type / Source / Comment / Analysed
Etopo5 / 0.5o grid / US/CSIRO / Too coarse, inaccurate / No
SIO Bathymetry / 2 min. grid / Scripps Inst. / OK / Yes
SIO Bathy Variance / 0.5o grid / derived CSIRO / Yes
SIO Gravity / 2 min. grid / Scripps Inst. / SIO Bathymetry preferred / No
Temp,Salinity 0-50m / 0.5o grid / CSIRO / IMCRA, refined 1998 / Yes
N,PO4,O2,Si 150m / 0.5o grid / CSIRO / IMCRA, refined 1998 / Yes
Seasonal T,S 0-50m / 0.5o grid / CSIRO / IMCRA, refined 1998 / Yes
Plate Age / 0.1 o grid / Dietmar Mueller / Large undefined age areas / Yes
CAMRIS Sediments / Polygons / ERIN / No attribute, no projection / No
AusSeabed / Points / Chris Jenkins / Coarse offshore / No
CZCS Chl. / 0.5o grid / CSIRO / Cloud problems S.West / Yes
Seasonal Chl / 0.5o grid / CSIRO / Cloud problems S.West / Yes
AGSO Bathymetry / 3sec. arc / ERIN / Could not import to A.Info / No

Since the datasets were compiled, problems with the AGSO bathymetry are in the process of being resolved. The Jenkins sediment datasets appears to be reasonably comprehensive for the shelf region but it is too sparse in the offshore area. The point data were interpolated but yielded what appear to be artificial patterns that would have unduly influenced the analysis. It was thus not included. The SIO Gravity dataset would have been a useful additional dataset but given the limited time available we had to restrict the number of large datasets that we analysed and unfortunately it was dropped in favour for the SIO bathymetry data.

A simple average was used to reduce high resolution datasets down to the half-degree resolution grid used for the analysis. In the case of the oceanographic datasets, optimal interpolation procedures were used to interpolate the point data onto the grid. Temporal and spatial variability were examined through the seasonal variability maps for temperature, salinity and chlorophyl, and through the bathymetry variation computed from the SIO bathymetry. For the seasonal maps a difference of winter from summer was used. For the bathymetry variation, maps were computed of averages taken over a 30min. grid and a 10min grid. The difference between these two was used to examine bathymetry variation.

Temperature and salinity were computed for the layer from 0-50m. Nutrients in the surface layers are generally stripped in offshore waters around Australia (possible exceptions being limited upwelling layers and waters in the Indo-Pacific region). To obtain a more meaningful signal a cut through the thermocline layer at 125-150m was taken for nitrate (N), Oxygen (O), Phosphate (P) and Silicate (Si).

Analyses

A number of analyses involving principal components, multivariate clustering and classifications were conducted. We report here the two main analyses:

• Principal component analysis of all the project data followed by a clustering and classification analysis of the first two principal components.
• Principal component analysis of the three key benthic datasets (Plate Age, Topography and Topographic Variation) followed by analysis of the first principal component.

The first analysis, which was dominated by the oceanographic variables, provided the large-scale general pattern expected of spatial structuring in the water column properties - albeit modified by the benthic variables. The second analysis was designed to reveal details of the seafloor structure. Thus these two analyses provide between them the major spatial structures in water column and seafloor to be expected from the variables analysed.

PCA Analysis of all variables

The first principal components analysis used all 13 variables:

1NitrateMean from 125-150m

2OxygenMean from 125-150m

3PhosphateMean from 125-150m

4SilicateMean from 125-150m

5TemperatureMean from 1-50m

6SalinityMean from 1-50m

7TemperatureSummer - Winter difference, 1-50m

8SalinitySummer - Winter difference, 1-50m

9Plate Age

10ChlorophylMean surface concentration from CZCS satellite

11ChlorophylJanuary - June surface concentration from CZCS satellite

12Topography

13TopographyVariation from large scale mean minus small scale mean

The first two PCAs, from the full dataset, were passed through an unconstrained clustering algorithm carried out on a reduced subset of the PCA vectors (every 10th value) as the full dataset of over 10,000 points taxed the limited memory of the computer on which the statistical analyses was conducted. This analysis suggested that 4 classes would adequately describe the parameter space spanned by the first two PCAs, with two additional classes for outlying points. Plots of this classification were used to classify (manually) the full dataset. In contrast, no classification was necessary for the benthic PCA analysis as the pattern emerging from the first PCA provided clear guidance on the spatial structuring.

A gradient analysis of the first 5 PCA was conducted using the so-called Ecotone Analysis methodology developed by Peters (1990) (see figures in appendix). This analysis was visualised by displaying the results of the gradients draped over the topography. Surface lighting from the north-west was used to simulate a 3-D effect, this unfortunately tends to obscure features in the south-east which are behind hilly regions (eg SE Tasmania).

Interpretation

PCA-based classification of all variables

The manual classification of the first two PCAs (see figures appended as part of this report) show firstly that the not all points were classified - most noticable in the two classes intersecting southern Australia. The large area of unclassified points in the south-western part of the figure is due to cloud contamination in the CZCS images.

With the exception of the North Western part of Australia, 4 classes describe the bulk of the spatial structuring. These being:

1. In the north-west area encompassing Java, Timor and Sulawesi a band representing Indo-Pacific structure. Around Australia, this extends from the eastern Arafura Sea to the southern end of the Sahul Shelf.

1. A band encompassing the Solomon Islands, the Solomon Rise and the Melanesian Basin which are not part of the Australian EEZ. The same band class appears as a north-west offshoot extending from the Exmouth Plateau.

1. Broad zonal bands on the lower half of western Australia leaking past into the Great Australian Bight. The same band-class extending from the north-eastern tip of Australia down past Sydney.

1. A zonal band of subtropical convergence structure with northern extents intersecting the lower half of NSW, and another possible intersection just east of the South Australian gulfs.

1. A scattered outlying class in the south notably around the south Is of New Zealand and to the south of Tasmania.

Benthic PCA

The benthic PCA class (see appended figures) shows structures dominated by the Plate Age with modifications due to topography in areas of uniform Plate Age. Principal features to note are:

1. A dominant southern structure extending from just south of Australia flowing around Tasmania and up the east coast to about Ulladulla before splitting off to an eastern extent that terminates at the western edge of the Lord Howe Rise and it’s connection with New Zealand. There appears to be a faint indication of possibly disparity in the structures on the eastern and western slopes of Tasmania. The south Tasman Rise stands out as a unique structure embedded within the main band.

1. On the eastern end of the main southern structure, a band confined to the west of the Lord Howe seamount chain extends up to the southern edge of the Great Barrier Reef.

1. Between the southern band and the slope off the Great Australian Bight a band sweeps to the west and around the western edge of the Naturaliste Plateau off south west of WA.

1. To the north of the Naturaliste Plateau, another less defined band (another dominant element from the Plate Age data) extends up to the Java Trench.

The gradient/ecotone analysis confirms a number of the boundaries evident in the PCA classification and the benthic PCA (SA gulf, south-east NSW. South-west WA, north-west Cape) but in addition suggests boundaries running to the north-west from about Geraldton and a series of striations running off the north-west coast of WA.

Large Marine Domain Boundaries

The 3 key information sets used in defining the Large Marine Domains (LMD) were:

1. The Demersal regionalisation from IMCRA 3.2

1. The classification of the PCA analyses, and gradient analysis, involving all the compiled data

1. The spatial pattern in the first PCA of the benthic data

The information sets were used to alter and verify the boundaries in the unofficial LMD map produced by CSIRO/Uni. of Tasmania for the MS&T working group in December 1997.

The main changes made were as follows:

1. Addition of an area enclosing the south Tasman Rise labelled as a “Sub-Antarctic” region (but with an alternate suggestion of a label as a “Southern Ocean” region). We did not investigate the link between this new region and the existing Macquarie Island region but note in passing that faunal elements on the south Tasman Rise have recently been shown to be of Sub-Antarctic origin (Last per. comm.). Clearly a more focussed effort is required to resolve this in the future.

2. Alteration of the boundary Eden/Cape Howe boundary off the NSW/Victoria border so that its eastern edge is more closely aligned with the clearly defined benthic structure seen in the benthic PCA analysis.

3. Alteration of the southern WA boundary so that its intersection with the WA coast was shifted to near Lancelin, about the mid-point of the South West biotone, and its offshore extension following the structure seen in the benthic PCA analysis.

4. The North West Cape boundary was well supported by the structure seen in the all-data classification and was not altered.

5. The north-western boundary was shifted to intersect the coast at about Cape Londonderry (west of Joseph Bonaparte Gulf) and extending offshore in sympathy with the alignment of structures seen in both the all-data classification and the benthic PCA.

6. The Cape York boundary was well supported and left unaltered.

7. The Sandy Cape boundary was well supported by the benthic PCA (with no suggestions from the all-data classification).

8. The offshore domains were left unaltered but we note here that the Lord Howe and Norfolk domains may require further investigation as to their similarity.

CAVEATS

1. This is a very rough investigation of the LMD structure of the Australian EEZ designed to provide no more than an impression of what the structure might look like. We make no pretence that this is a scientifically rigorous derivation because it isn’t.

1. The only biological dataset used in the analysis was chlorophyl which by itself is no better at discriminating biogeographic structuring than other water column properties. Extensive research information does currently exist (both biological and geological) which would help substantially refine the analysis, and more importantly provide the necessary ecological knowledge required to manage the LMDs.

1. A conceptual hierarchical framework is required for the LMDs which follows the framework envisaged for IMCRA and which accounts for the offshore areas. In the current analysis there is a mixture of scales which have been blended (for lack of other data); for instance the topographic variation map picks out such structures as individual seamounts which are more appropriately analysed as part of “habitat” units rather than as a basis for a provincial structure.

1. The slope region has been almost entirely glossed over in current analyses. There is very good existing biological information which can be analysed in an RAP (Rapid Assessment Procedure) manner to provide information on the structuring on the slope.

1. The Plate Age data dominates many structures seen in the benthic PCA. The validity and accuracy of this data needs careful assessment as it appears to be a critical information set for identifying offshore provinces.

1. The half-degree resolution grid was much too coarse in resolving a number of boundaries, particularly their interaction with the shelf. In future an integrative analysis (of shelf, slope and offshore) is suggested where spatial resolution is tied to the intrinsic scales of interest. Thus a polygonal analysis method (finite elements) is advocated so that a more robust integration of the boundaries can be achieved.

1. We made no attempt at resolving biotones and core provinces in the offshore region (as was done for the IMCRA shelf region by CSIRO) as this requires a careful analysis of the faunal elements of the offshore region (eg using the extensive sharks and rays information set).

1. Given the half-degree resolution grid, boundaries are no more accurate than this resolution.

1. We did not examine the Jenkins shelf sediment data. This dataset may be important in providing part of the basis for a comprehensive habitat scale information set for the shelf region (along with such datasets as the Kirkman seagrass and subtrate-type data). The AGSO 30 second arc bathymetry would likewise be useful in resolving the fine-scale habitat structure in the offshore region. Here again a much finer resolution analysis is required.

1. Time constraints prevented a full classification of the data and we manually compiled the final boundaries of the Large Marine Domains from the three analyses. A more comprehensive and robust classification analysis can be carried out given more time.. In particular the rich structure off the north-western part of WA needs careful resolution.

1. The claimable seabed areas were not included in this analysis because of a lack of available data. Their boundaries are shown on the map for consistency.

References

CSIRO (1996) Interim marine bioregionalisation for Australia: Towards a national system of marine protected areas. Report to Department of the Environment, Sport and Territories. Ocean Rescue 2000 report series on a National Representative System of Marine Protected Areas.