Australia’s Commonwealth Marine Reserve (CMR) network covers 2.76 million km2 of continental shelf, slope, and abyssal habitat.
The shelf provides a mosaic of shallow water habitats (to about 200 metres deep) that support a diversity of marine species and species assemblages, but little is known about the extent of these habitats or the status and trends of their associated assemblages. One way to address this important knowledge gap is to map the seafloor using multi-beam sonar (MBS) but in these water depths this would be a slow process.
A survey vessel dedicated to mapping 24 hours a day, every day of the year, would take 3.5–17.5 years to cover the 306,627 km2 of CMR seafloor lying within the biologically diverse 40–200 m depth range. Even then, data would be available only for the seafloor habitats, not the resident species and assemblages. A more pragmatic approach is required to provide early and cost effective methods to gain insight about the extent of shelf habitats in CMRs and the status and trends of marine life.
This project successfully implemented a new sampling approach that enables managers to acquire accurate baseline measurement of environmental assets in the CMR network. Trends in these assets can also be collected in a cost-effective manner during the prolonged period required to complete mapping. Establishing the baseline status of the CMR network at an early stage, and tracking trends in key indicators, are essential for management agencies to get timely information on the performance of the CMR network.
IMAGE: (A) Multibeam sonar habitat mapping and sample locations (1─40) of the GRTS design drop camera survey of shelf habitats of the Multiple Use Zone of the Flinders CMR showing the cluster of mixed reef sample points discovered in the north-western corner of the reserve. MBS mapping of the drop camera locations (B), and of a continuous patch of seabed (C) shows the typical patchiness of the reef habitat. Image: GA/CSIRO/IMAS
Approach
The alternative approach to continuous MBS acquisition termed a GRTS (Generalised Random Tessellation Stratified) design works by collecting samples at small scales across broad areas. Sampling inference methods are then used to draw conclusions about the entire region of interest (such as a zone or reserve). This approach can complement continuous MBS acquisition and allow monitoring of status and trends to start during the long lead-in time required to acquire continuous MBS data.
Theme 1 of the Marine Biodiversity Hub trialled the GRTS approach in the Flinders CMR, and subsequently employed it in the Geographe CMR and Houtman-Abrolhos Key Ecological Features (KEFs). The technique is endorsed and used by the United States National Park Service, but has not previously been implemented in Australia nor adapted to the peculiarities of sampling from vessels in marine environments beyond the reach of divers.
Key advantages
The GRTS methodology ensures that sample sites in an area of interest, such as the shelf habitats of a CMR, are distributed in a spatially balanced way. This approach avoids problems experienced with other sampling designs such as judgemental sampling, simple random sampling, or the placement of samples on a regular grid. The flexibility of the GRTS approach also means that the survey design can accommodate unexpected loss of sampling effort due to poor weather, equipment breakdowns, or the vagaries of the funding cycles, without comprising statistical validity.
IMAGE: The GRTS design allows survey teams to preferentially sample areas of interest if the location is known in advance. This facility was used in surveys of the Flinders CMR, and to discover seagrass beds and reefs in the Geographe CMR. Images: CSIRO (left) and Curtin University.
Judgemental samples cannot be used to infer the status of regions that are not sampled. Simple random samples are undesirable because they tend to clump in space (see figure above). On average they are less likely than spatially balanced samples to detect features of interest, such as regions of high benthic biodiversity in the CMR network associated with distinct reef features that can be tens to thousands of metres apart. The clumping also acts to inflate estimates of indicator variance, (an essential factor in determining evidence for a trend). And while regular grid sampling can allow the mean value of an indicator to be assessed, it does not allow the variance of the indicator to be estimated.
New knowledge and opportunities
By applying spatially balanced designs, Marine Biodiversity Hub project teams discovered what appears to be a cluster of mixed reef habitats in the north-western corner of the Flinders CMR reserve (see figure at left), and seagrass beds and reefs in the Geographe CMR that are much more extensive than previously thought (see stories Surveys of Commonwealth marine reserves: Flinders and Surveys of Commonwealth marine reserves: Geographe. Furthermore, the GRTS design allows survey teams to preferentially sample areas of particular interest if the location of these areas is known in advance. This facility was put to good use in the survey within the Tasman Fracture CMR to target reef habitats of rock lobster. It ensured a good sample size to test the effect of the exclusion of fishing within part of the CMR on the size-frequency distribution of rock lobster, and ensured that the data collected are representative of the CMR as a whole.
Outputs and outcomes
Spatially balanced survey designs have been successfully trialled at several CMRs and KEFs, facilitating design and model-based inference of the extent and status of benthic habitats that would have not been identified using judgmental, simple random or regular survey designs.
IMAGE: A hypothetical surface showing the typical patchiness of a benthic biodiversity indicator in the South-east CMR network, together with the features of 40 Simple Random Samples (blue dots) and 40 spatially balanced GRTS samples (green dots). The clumping of the SRS samples is clearly evident, and in this instance the SRS samples fail to identify the biodiversity hotspot in the bottom right of the sample frame. Both designs capture the true mean value of the indicator but the confidence interval of the SRS estimator is 20% larger than that of the GRTS estimator because the variance of the indicator is higher. Image: CSIRO
contact
Keith Hayes
keith.hayes@csiro.au
(03) 6232 5260