Supporting datasets for estimating reef counts in the Coral Sea and northern Australia (NESP MaC 3.17, AIMS)

Full description

This collection contains supporting datasets created to assist in estimating the number of reefs in the Coral Sea and northern Australia. The 'Analysis-regions' shapefile provides standardised spatial boundaries defining reporting regions across northern Australia and adjacent waters (Great Barrier Reef, Torres Strait, Coral Sea, and northern/northwestern coastal areas) used to aggregate and compare reef count statistics. The 'Boundary-regions' shapefile identifies zones where the two sets of independent reef mapping produced in the North and West Australian Tropical Reef Features dataset (https://doi.org/10.26274/XJ4V-2739) can be reliably compared. This is to quantify the distribution of boundary errors that occur from repeat mapping of reefs. The 'Grazing-halos' dataset contains measurements of grazing halo extents around 100 randomly selected lagoonal patch reefs in the Coral Sea. These three datasets support a Monte Carlo simulation approach to estimating reef counts across northern Australia while explicitly accounting for uncertainties in reef boundary mapping and counting criteria. The Boundary-regions dataset enables comparison between multiple independent reef mappings, providing the statistical foundation for simulating boundary variations (via log-normal buffering of the EdgeAcc_m attribute). The Grazing-halo measurements quantify the typical extent of reef influence around Coral Sea lagoonal patch reefs, establishing empirical constraints on the clustering distance threshold (50-150 m) used to determine when neighbouring reef patches should be treated as a single ecological unit. The Analysis-regions shapefile provides management-relevant spatial reporting units (Coral Sea, Great Barrier Reef Marine Park, Torres Strait, and northern/northwestern Australia) that allow reef count distributions from the Monte Carlo simulation to be aggregated and compared across jurisdictional boundaries. Methods Analysis-regions Shapefile (data/v1/in/analysis-regions/analysis-regions.shp): The Analysis-regions shapefile was created by overlaying and dissolving multiple authoritative spatial datasets including the Commonwealth Marine Regions (DCCEEW, 2023b), Great Barrier Reef Marine Park boundaries (GBRMPA), Coral Sea Marine Park boundaries, and Natural Resource Management (NRM) regions for Torres Strait (DCCEEW, 2023a). The process began by using the outer boundary of the Australian Tropical Reef Features study area and performing a vector union with Commonwealth Marine Regions to create initial subdivisions. Manual digitisation in QGIS was then used to refine boundaries, with particular attention to: (1) dividing the Great Barrier Reef region using the GBRMP outer boundary and NRM Torres Strait boundary as guides; (2) delineating Torres Strait as a bioregion including both Australian and Papua New Guinea waters where reefs form a connected network; (3) separating nearshore (state waters) from offshore (Commonwealth waters) subdivisions in the North and North-west regions; and (4) simplifying offshore island cutouts (such as Rowley Shoals and Scott Reef) by dissolving them into their enclosing marine regions. The resulting shapefile contains 18 regions with attributes for Country, region, subregion, and numeric ID. For the reef counting analysis, only Australian regions (Country = 'Australia') excluding South-east are used. Full processing details and validation steps are documented in the associated GitHub repository. For the Torres Strait region we include the Papua New Guinea portion of the Torres Strait. In this way we ignore the EEZ divide between Australia and PNG and instead consider the connected reef system as one region. The analysis regions dataset is intended to provide regions for analysing the count of reefs within Australia and connected regions, such as Torres Strait. It does not represent the boundaries of external nations accurately. While the reef mapping does include some reefs in neighbouring countries it cannot be used to estimate the number of reefs in these countries as the mapping is incomplete in these regions. In the north west edge of the continental shelf there are the Sahul Banks that include many submerged coral reef banks. These banks are within the 1972 Australia - Indonesian Seabed Boundary on the Australian side, but now the Australian EEZ excludes these reefs. More research is needed to determine the sovereignty and interests over these coral reefs banks in order to determine how they should be included in any counts. Boundary-regions Shapefile (data/v1/in/boundary-regions/Boundary-comp-regions_v0-4-to-v0-1-EL.shp): The Boundary-regions shapefile was created to identify areas where two independent mappings of northwestern Australian reefs could be validly compared to assess boundary accuracy. The primary reef mapping (NW-Aus-Features_v0-4) was compared with a rough reef mask dataset (AU_AIMS_NESP-MaC-3-17_Rough-reef-shallow-mask_87hr.shp) that was created independently for shallow water habitat mapping purposes. Because the rough reef mask included both reef boundaries and shallow nearshore sediment areas, direct comparison was only valid in specific regions. Forty-four regions were manually digitised in QGIS to identify locations where: (1) the rough mask corresponded only to reef boundaries with no shallow sediment contamination, (2) both datasets showed comparable boundary detail and quality, and (3) comparison regions provided spatial distribution across the study area to reduce spatial biases. These regions were intentionally distributed broadly across northwestern Australia. In these boundary regions the primary reef boundaries and rough reef mask boundaries can be directly compared, allowing the distribution of differences between repeat mappings to be analysed. Grazing-halos Shapefile (data/v1/in/grazing-halos/CS-Lagoonal-Reef-Grazing-Halos.shp): The Grazing-halos shapefile contains measurements from 100 randomly selected lagoonal patch reefs in the Coral Sea. Reef selection used the Australian Tropical Reef Features dataset (Lawrey & Bycroft, 2025), filtering for Coral Sea features (Dataset='CS Features') with classification RB_Type_L3='Oceanic Lagoon Patch Coral Reef'. From 2,857 available reefs, 100 were randomly sampled using a fixed random seed (42) for reproducibility. Point features were generated at reef centroids using representative points guaranteed to fall within reef boundaries. Grazing halo measurements were performed manually in QGIS using two complementary Sentinel 2 image composites, R1 and R2 composites from Lawrey & Hammerton, 2022, to minimize image noise effects. For each reef, distances were measured in four cardinal directions (north, south, east, west) from the estimated hard substrate reef edge to the outer edge of the grazing halo or vegetation ring. Halo type was classified as: Light (cleared grazing halo), Dark (vegetation halo with no clear zone), None (no surrounding vegetation), or NULL (insufficient image clarity). Measurement challenges included depth effects (reefs at 40-60 m depth), where deeper reef substrates can appear similar in brightness to shallower sand, making boundaries difficult to distinguish. To address this, multiple image views were compared, and where available (North Flinders Reefs), 30 m resolution bathymetry data (Beaman, 2017) was used to identify substrate edges based on slope transitions. Given expected measurement uncertainty of 20-40 m, the four directional measurements were averaged, This reduced reduce noise in the grazing halo width estimates, but resulted in potentially underestimating the width distribution of asymmetric halos. Results showed 15% of reefs were unmeasurable due to image quality, 17% had no surrounding vegetation, 12% displayed vegetation halos (median 70 m, N=12), and 56% had measurable grazing halos (median 28 m, 90th percentile 80 m, N=56). These measurements informed the proximity threshold range (50-150 m) used in the Monte Carlo reef counting analysis. Full measurement protocols and validation approaches are documented in the GitHub repository (https://github.com/open-AIMS/AU_AIMS_Reef-count/). False positive reefs and Potential Reefs (data/v1/false-positive-reefs, data/v1/potential-reefs): As part of the Coral Sea mapping we identified potential reefs. These are locations where reefs might exist based on the information available (satellite and bathymetry), but the evidence was insufficient for them to be included as a mapped reef feature. These typically correspond to deep locations where little is known about these structures making it difficult to correctly interpret the imagery. Identifying how many deeper unmapped reefs might exist is important for determining the most likely reef count as each new advance in imagery and information has increased the number of known reefs. The potential reefs tries to compensate for the likely number of reefs that might be discovered in the future. The potential reefs in the Coral Sea were mapped as part of the Coral Sea mapping project. In the reef counting analysis we use the Coral Sea potential reefs dataset (https://doi.org/10.26274/PGJP-8462). As part of the Monte Carlo reef count analysis we need to assign each `FeatConf` classification a probability value. It is however difficult to quantify the uncertainty (false positives) without a significantly better information source than was used to map the reefs. We therefore estimate the magnitude of probabilities by looking back at historic mapping. The Torres Strait reefs were mapped by the same team, using similar methods and quality control in 2016. More recent satellite imagery allows us to determine if any of the reefs might be false positives and how many reefs were missed in the mapping. This can then provide an indication to the magnitude for the likely false positive and false negative error rate in other regions using a similar mapping strategy. For this we used the Sentinel 2 images (Hammerton & Lawrey, 2024a) and (Hammerton & Lawrey, 2024b) to review the existing Torres Strait reef mapping. We set the reef layer to only show the outlines. These outlines were set to low contrast, so that the focus was on the imagery. The Torres Strait region was progressively reviewed comparing the old reef boundaries to the new satellite imagery. Where a reef boundary was mapped, but there was no evidence in either composite satellite images then the reef was marked as a potential false positive (`false-positive-reefs/-false-postive-reefs.shp`). Where a reef was missed from the mapping it was saved in `potential-reefs/AU_Potential-reefs.shp`). This is more a record of missed reefs than potential reefs, as most of these locations now have enough evidence that they should be included in the reef mapping. Note: The Potential Reefs used in the Monte Carlo analysis were only from the preexisting Coral Sea and not the AU_Potential-reefs.shp. The AU_Potential-reefs.shp was only used for determining the likely false negative rate in Torres Strait as a way of quantifying the likely false positive and negative rate of the newer mapping in the Coral Sea and northern Australia. In future versions of this dataset the existing potential reefs in Torres Strait will be added to the Aus-Trop-Reef-Features dataset and removed from the AU_Potential-reefs dataset. Data Dictionary: analysis-regions.shp: - id: Feature count from 1 - 18 - subregion: Division between nearshore and offshore areas, for example: Temperate East (Offshore), Temperate East (Nearshore). - Country: Australia, Indonesia, New Caledonia, Papua New Guinea - region: Torres Strait, Timor East / Indonesia, Temperate East, South-west, Papua New Guinea, Oceanic, North-west, North, New Caledonia, Indonesia, GBR, Coral Sea CS-Lagoonal-Reef-Grazing-Halos.shp: The shapefile contains the following attributes: - ReefID: Unique reef identifier linking back to the source reef dataset - GrazHaloNm: Grazing halo distance North (metres) - manually measured - GrazHaloSm Grazing halo distance South (metres) - manually measured - GrazHaloEm Grazing halo distance East (metres) - manually measured - GrazHaloWm: Grazing halo distance West (metres) - manually measured - HaloType: Light (grazing halo), Dark (vegetation halo), None (no surrounding vegetation to measure halo), NULL (imagery is too unclear to take a measurement) Change Log: In this section we will note any changes made as part of revisions to this dataset. 2026-02-20 (v1) - Initial draft of this analysis focused on the Coral Sea, with a comparison with the rest of Australia. In this version the Potential Reefs and False positive reefs only covered Torres Strait.

Lineage

Maintenance and Update Frequency: notPlanned

Notes

Credit
This dataset collection was developed using funding from the Australian Government's National Environmental Science Program, by Parks Australia (Commonwealth of Australia) and the Australian Institute of Marine Science.