North and West Australian Tropical Reef Features - Boundaries of coral reefs, rocky reefs and sand banks (NESP-MaC 3.17, AIMS, Aerial Architecture)

This dataset contains a shapefile of over 10,000 boundaries of shallow tropical reefs of Australia including coral reefs, rocky reefs, and sand banks. This dataset is intended to assist with environmental accounting, marine management, marine research planning and environmental impact assessments.

All features were digitised manually from satellite imagery that was optimised for image clarity of marine habitats (Hammerton & Lawrey, 2024a; Hammerton & Lawrey, 2024b). This dataset covers from Houtman Abrolhos Islands in Western Australia, across Northern Territory, through to Western Cape York in Queensland. This dataset also covers offshore coral reefs including Lord Howe Island, Norfolk Island, Christmas Island and Cocos (Keeling) Island. This study area is intended to complement the existing mapping of the Great Barrier Reef, Torres Strait and Coral Sea Marine Park. As this dataset is mapped from satellite imagery it only includes reefs that are shallower than approximately 10 - 45 m in depth. This means that deeper hard substrate areas are not included.

This dataset focuses on mapping the outer boundary of marine features that are raised off the sea floor including rocky reefs, coral reefs and sand banks. The primary goal of this dataset is to indicate areas of likely high biological activity. Coral reefs and rocky reefs are typically high structural complexity habitats that are home to a disproportionately high number of species compared to the surrounding sea floor. While sand banks are generally abiotic and thus do not indicate areas of high biodiversity, they do provide an indication of current flow and thus connectivity. It is for this reason that large sand banks are also included in this mapping.

The classification of the mapped feature boundaries is primarily determined from the evidence provided from remote sensing rather than direct field observations. The mapping team learned how to read the satellite imagery by comparing field observations to the view of these locations in the satellite imagery. This training process used a subset of the available field observations. The remaining field observations were withheld for the purpose of validation.

This dataset was inspired by the GBRMPA GBR Features dataset, which mapped the outer extent of reef features across the GBR, at the scale of named reefs. This is in contrast to habitat mapping where reefs are subdivided into many small parts to differentiate between, for example, sandy, rubble, rocky and coral areas.

Ecological and geological mapping of reef boundaries:

While the original intent of the dataset was to map areas of high biological activity we found that the reefs in the Kimberley region posed a challenge. These reefs have grown out laterally from the land to the point where they join up neighbouring islands. The outer edge of these reefs have active coral growth, while the old reef flat portion is dominated by sand. Blue holes (large deep holes formed from the dissolution of the carbonate framework from rain during low sea level periods from glacial periods) on these reef flats indicate that they are carbonate in origin. Only mapping the active outer edge of these fringing coral reefs significantly underestimates the geological size of these reefs. We therefore included a category, 'Coral Reef Flat', to record the inner portion of these reefs that are largely devoid of coral and largely covered in thin layer of sand. The outer edges were mapped as 'Coral Reef'. Combining these two feature types provides the outer extent of the geological reef structure, while focusing on the 'Coral Reef' feature allows a focus on areas with high biological activity. This two part mapping was applied where there was clear evidence of the coral reef foundation that had a clear active growth region and a significant portion covered in sand (> ~150 m).

This dataset also maps rocky reefs due to their importance as a habitat. The 3D structure of rocky reefs provides protection for fish and an attachment substrate for macroalgae, further adding to the habitat. Rocky reefs are divided into 'Rocky Reef' and 'Low Relief Rocky Reef' categories. 'Low Relief Rocky Reef' corresponds to flat rock pavement that provides less structural complexity and thus is likely to be a less diverse habitat.

Feature classifications:

The reefs were classified against a new classification scheme, the Reef Boundary Type, developed for this dataset. It is intended to represent the diversity of reef types across tropical reefs of Australia, while supporting crosswalks to other habitat classification schemes such as the Natural Values Common Language (NVCL) used by Parks Australia, and the Seamap Australia Classification scheme.

To better represent the diversity of reef types across northern Australia, coral and rocky reefs were divided into classifications that capture regionally specific reef types, including 'High Intertidal Coral Reefs' that occur only in macrotidal environments, 'Limestone Reefs' that are a form of rocky reef where the rock is formed from ancient coral reefs, and types to distinguish between lagoonal and rim reefs on coral atolls. Each of these classifications aims to better represent the unique characteristics of each region.

The Reef Boundary Type scheme is hierarchical in design with the RB_Type_L3 (Reef Boundary Type Level 3) attribute providing the most detailed indication of the reef type, Level 2 grouping these classifications down to coral reefs, rocky reef, sand banks and islands, and Level 1 indicating if the feature is a reef, sediment or island. The classification scheme indicates the relative location of the feature with respect to land ('attachment' attribute) and a depth category. The attachment attribute has categories of 'fringing reef' (attached to islands or the mainland), 'isolated reef' (free-standing reefs), and 'oceanic' (indicating that the reef is off the edge of the continental shelf). The depth category (land, intertidal, very shallow, shallow and deep) that indicates the shallowest portion of the feature. The oceanic classification is used to assist with the crosswalk to the NVCL and the fringing or isolated classification is used to indicate whether the reef is associated with an island. Fringing reefs typically have a tapered asymmetrical structure with an outer reef slope that transitions to a shallow sediment dominated intertidal habitat. Isolated reefs are typically more symmetrical with a reef slope completely around the outer perimeter of the feature.

This dataset mapped three primary high level feature types: coral reefs, rocky reefs and sand banks. Since there is no well established definition of reefs for mapping purposes we provide the core definition that we used in this mapping:

Coral Reef Definition: The kernel of a coral reef is an area with some exposed hard substrate, formed by the modern accretion of hard coral and other calcifying organisms. On the continental shelf it rises above the surrounding sea floor and has outward growth due to accretion, typically resulting in rounded growth forms. On atoll platforms it corresponds to areas of exposed hard substrate, not covered in sediment. Coral reefs are not limited by depth, but by whether there is sufficient benthic light to support growth of some hard coral species. Coral reefs include drowned reefs (typically deeper than 25 m) where reef growth is halted, but significant coral colonies can still be found.

To ensure the mapped features correspond to the scale of whole reefs, exposed hard substrate areas (reef patches) are clustered by distance, including sand between these patches. A small sand buffer on the outer perimeter is included to capture sandy areas that are highly utilized by reef-associated organisms. The sand buffer size is informed by the size of grazing halos typical for the region being mapped. This is can be as much as 50 m in lagoonal reefs in the Coral Sea (not part of this dataset), but is typically less than 20 m.

In the Kimberley and in western parts of Northern Territory the coral reefs predominantly grow laterally due to the high tidal range. Many of the reefs are old, as evidenced by the presence of blue holes in the reef tops. These reefs have extensive reef flats that appear to mostly be covered in a thin layer of sediment. These reefs are typically split into the active growing region (RB_Type_L3 = 'Coral Reef') and the reef flat (RB_Type_L3='Coral Reef Flat'). This division is only done when there is a clear separation of zones on the reef. In version v0-4 of the dataset the mapping of the 'Coral Reef Flat' areas is incomplete (estimated 50 - 80% complete) as the primary focus was on mapping the highest biodiversity areas, the 'Coral Reef' features.

Coral reefs are distinguished from rocky reefs by evidence of sufficient modern accretion from the active growth of coral and other calcifying organisms. Many rocky reefs will have coral colonies growing on them but the colony density and environmental conditions lead to no carbonate reef matrix accretion. In some places we find coral reefs growing over the top of part of a rocky reef; areas where the rocky base is fully covered by a coral reef carbonate layer. In this dataset these two sections (rocky reef and the coral reef over part of the top) are mapped as separate touching reefs. In some areas, particularly the Pilbara in WA, many of the modern reefs are growing on ancient limestone reefs that are more than 10 million years old. These limestone reefs are considered as a subset of rocky reefs, even though they are carbonate in origin, because of the age of these foundations. In many cases we find modern coral reefs growing over the top of part or all of a limestone reef. Ideally these should be mapped as separate touching rocky and coral reefs. In practice distinguishing limestone and coral reefs from satellite imagery is very error prone and while we attempted to map these separately it is likely that the accuracy is not much better than 50%.

Rocky Reefs Definition:

A rocky reef corresponds to any intertidal or subtidal marine area where there are exposed (not covered in sand) rock. These generally correspond to rock that is of terrestrial in origin, but also includes limestone rocky deposits where there is no evidence of modern active accretion by hard coral. Rocky reefs are subdivided into categories based on their likely structural complexity and relief, whether they were part of an ancient coastline and if they are limestone. The boundary of rocky features were mapped using similar clustering rules as the coral reefs. This ensures a consistent mapping of features to a scale of ~1:100k.

Methods:

The classification of the mapped feature boundaries is primarily determined from the evidence provided from remote sensing rather than direct field observations. The project team prepared for the mapping by comparing the satellite imagery to known field observations from drone and underwater footage, as well as a set of pre-existing benthic surveys (Lawrey & Bycroft, 2024). This virtual fieldwork helped to identified the range of different reef types and how they could be distinguished from remote sensing. During the mapping process additional research was conducted on the reefs of the Kimberley and the Pilbara region, where the reef structures are significantly different from other regions in Australia. The use of additional field observations was paused as soon as the team was reasonably confident in the identification of the reefs in a given area. This was to ensure that some data was not used during the mapping, allowing it to be used during validation. Any field observations that were studied by the team during the manual mapping process could not be used for validation, as this could lead to an over-estimate of dataset accuracy.

Satellite imagery The marine features boundaries were primarily digitised from cloud-free satellite composite imagery from Sentinel 2 MSI (10 m resolution) . Two primary image composites were used: - All-tide composite (Hammerton, & Lawrey, 2024a): 10 m resolution Sentinel 2 composite imagery covering the whole study area. This imagery was formed from up to 200 images per Sentinel 2 scene smoothing out turbidity plumes and minimising image noise. - Low-tide composite (Hammerton, & Lawrey, 2024b): 10 m resolution Sentinel 2 composite imagery covering the inshore areas. This was produced by combining the ten lowest tide images. This allowed a clear view of shallow features in highly turbid areas. This composite also included near-infrared and infrared, which allows for viewing of the 3D structure of very shallow areas, helping to identify high-intertidal reefs.

Format of the data: - AU_NESP-MaC-3-17_AIMS_NW-Aus-Features.shp (Shapefile, 12 MB, XX features)

Dataset Reproduction: The reef mapping dataset was developed through five distinct stages, progressively integrating remote sensing data, bathymetric information, marine charts, and in-situ observations. These are provided as part of this dataset because they are used to help characterise various types of errors in the mapping, and are used as inputs to other dataset.

Stage 1 (v0-1_dual-maps) involved two independent reef mappings using Sentinel 2 satellite imagery: a detailed manual digitisation by Rachel Bycroft (Reef Boundaries RB.shp), classifying coral and rocky reefs at 1:150k-250k scale (with a target spatial error of 75-150 m), and a rapid, semi-automated mapping by Eric Lawrey (Rough Reef Mask, AU_AIMS_NESP-MaC-3-17_Rough-reef-shallow-mask_87hr.shp) that included intertidal areas but without benthic classification (max boundary error 250 m). The Rough Reef Mask was used as an input into the development of the shallow sediment area mapping used as inputs into the habitat mapping completed by UQ. The Reef Boundaries mapped by Rachel formed the basis of the draft of the North and West Australian Tropical Reef Features. These two datasets represent two independent mappings of the reef features. The Rough Reef Mask includes intertidal and shallow sediment areas and thus is not directly comparable to Reef Boundaries RB.shp in all areas. It is however useful for determining the number of reefs that were missed by Rachel as part of the initial manual mapping. It is also useful for studying the consistency and distribution of boundary errors.

Stage 2 (v0-2_merge-maps) merged these two independent mappings of the reef datasets (AU_AIMS_NESP-MaC-3-17_Rough-reef-shallow-mask_87hr.shp and 'Reef Boundaries RB.shp'), incorporating reefs missed in the detailed mapping and addressing high-priority issues from the independent review, thus creating a unified remote-sensing-based dataset. Combining these two mappings help to ensure that the detection rate of difficult to see reefs was better than any one person.

Stage 3 (v0-3_review). In this stage Eric Lawrey reviewed the output from v0-2_merge-maps, creating v0-2_merge-maps/Issues-v0-2. This review involved visually inspecting every reef mapped and comparing it against the dataset feature type definitions and digitisation rules. The goal was to ensure that the dataset is consistent in quality as possible. Adjustments were recorded in the Issues-v0-2 point dataset, and assigned a priority for adjustment. The main adjustments identified were trimming of sediment areas out of rocky reef features and the identification of many small fringing rocky reefs. More than 1000 adjustments were identified. Rachel applied corrections for approximately 50% of the priority 1 adjustments. Not all corrections were made due to limited time. Eric Lawrey then took over edits to the dataset in preparation for the dataset being provided for the project habitat mapping being undertaken by UQ. These adjustments included: - Trimming and review of rocky reef features, particularly around Mornington Island - Redo of the reefs surrounding Montebello/Barrow Islands based on monitoring site data from Bancroft, 2009. This changed these large reef areas from its original classification of rocky reef to Fringing Coral Reef and Shallow Reef Flat, to better capture that they these areas have a carbonate base and that there are high levels of hard and soft corals.

The stage 3 version of the NW-Aus-Features incorporates priority 1 corrections to the dataset.

Stage 4 (v0-4_merge-bathy) In this stage additional reefs were identified by reviewing bathymetric datasets and marine charts from the Australian Hydrographic Office (AHO). Only bathymetric features confirmed visually with satellite imagery were added; others were recorded separately for reference. Known reefs from external sources (AHO charts, ReefKIM, Big Bank Shoals, AIMS surveys) were noted, and depth data from marine charts were integrated to support depth classification.

The dataset corresponding to each stage of the process are available as different versions in the data download. This is provided because some of the analysis scripts compare the release dataset (v1_qaqc) with the original independent maps (v0-1_dual-maps) to determine how many false positives and negatives were mapped by each of the reef cartographers. Having all the stages available also allows tracking of how many additional reefs were mapped from the incorporation of bathymetry datasets.

Data dictionary:

Additional details are provided in the Reef Boundary Type Classification Scheme v0-4 document for RB_Type_L3, RB_Type_L2, RB_Type_L1, Attachment, DepthCat, DepthCatSr, FeatConf, TypeConf, EdgeSrc, and EdgeAcc_m.

• RB_Type_L3 (String): Reef Boundary Type Classification Level 3 - Most detailed classification applied to features. See Reef Boundary Type Classification Scheme v0-4 for details.
• Attachment (String): Attachment classification of the feature to islands and mainland. Values: 'Fringing', 'Isolated', 'Land' and 'Oceanic'
• DepthCat (String): Depth of the top 90th percentile of the feature. 'Land': Islands, 'Surface': Floating man made structures, 'Intertidal', 'Very Shallow': Shallower than -2.5 m LAT, 'Shallow', -30 m MSL to -2.5 m LAT, 'Deep', < -30 m MSL.
• DepthCatSr (String): Source of information used to estimate the depth category.
• FeatConf (String): Confidence that the mapped feature is a feature of interest and not an artefact.
• TypeConf (String): Confidence that the RB_Type_L3 classification is correct.
• EdgeSrc (String): Source of information used to primarily digitise the feature boundary.
• EdgeAcc_m (Integer): Estimated error in the digitisation of the feature boundary in metres from the true boundary for 90th percentile of digitised points.
• RB_Type_L2 (String): Level 2 classification of the Reef Boundary Type Classification. This groups all the different coral reef types to a 'Coral Reef' class, and the different rocky reef types to a 'Rocky Reef' class.
• RB_Type_L1 (String): Level 1 classification of the Reef Boundary Type Classification. The lowest detailed level of classification. It primarily distinguished between 'Reefs', all features with hard substrates, 'Sediment', soft sediment areas, and 'Land'.
• NvclEco: (String) Natural Values Common Language classification Ecosystem classification. This is a classification scheme used by Park Australia. • Oceanic shallow coral reefs: Coral reefs occurring seaward of the continental shelf break in depths shallower than 30 m. • Oceanic mesophotic coral reefs: Coral reefs occurring seaward of the continental shelf break in in the mesophotic zone: a reduced light zone between 30 m and the maximum depth at which there is sufficient penetration of sunlight to support photosynthesis. The maximum depth is variable dependent upon water clarity and may extend to 150 m in the clearest of waters however, as a national average it is nominally defined as 70 m. • Shallow coral reefs: Coral reefs shallower than 30 m • Shallow rocky reefs: Rocky reefs shallow than 30 m • Mesophotic coral reefs: Coral reefs deeper than 30 m
• NvclEcoCom (String): Natural Values Common Language classification Ecosystem Complex classification. • Oceanic coral reefs: Coral reefs occurring seaward of the continental shelf break.
• INUNDTN (String): [Queensland Intertidal and Subtidal ecosystem Classification]: Inundation
• SMB_CMP (String): [Queensland Intertidal and Subtidal ecosystem Classification]: Structural macrobiota composition
• AS_TidalZ (String): [Seamap Australia Classification]: Aquatic Setting Tidal Zone (Intertidal or Subtidal)
• AS_BDepth (String): [Seamap Australia Classification]: Aquatic Setting Benthic Depth (Littoral, Infralittoral)
• AS_System (String): [Seamap Australia Classification]: Aquatic Setting (Marine)
• AS_SubSys (String): [Seamap Australia Classification]: Aquatic Setting (Nearshore, Offshore) Note: We have only set this with the crosswalk and so is not assigned based on the position of the feature relative to the 30 m contour that normally separates this classification. Some features are marked as 'Nearshore;Offshore' to indicate that it might be in either category.
• BC_Level1 (String): [Seamap Australia Classification]: Biotic Component (Biota Present, Biota Absent)
• BC_Level2 (String): [Seamap Australia Classification]: Biotic Component (Invertebrates, Microbes, Vegetation)
• BC_Level3 (String): [Seamap Australia Classification]: Biotic Component (Non-Molluscan Filer Feeders, Stromatolite, Macrophytes)
• BC_Level4 (String): [Seamap Australia Classification]: Biotic Component (Coral Biota, Macroalgae, Seagrass)
• SO_Level1 (String): [Seamap Australia Classification]: Substratum Origin (Anthropogenic Origin, Biogentic Origin, Geologic Origin)
• SO_Level2 (String): [Seamap Australia Classification]: Substratum Origin (Algae, Carbonate, Terrigenous)
• SO_Level3 (String): [Seamap Australia Classification]: Substratum Origin (Halimeda, Coral, Limestone)
• SC_Level1 (String): [Seamap Australia Classification]: Stratum Component (Hard Substrata, Soft Substrata)
• Area_km2 (Decimal number, length: 10, precision: 6) Area of the feature in km2. Calculated with project projection of EPSG:3112 and the QGIS field calculator expression: round($area / 1000000,6)

Location of the data: This dataset is filed in the eAtlas enduring data repository at: data\custodian\2023-2026-NESP-MaC-3\3.17_Northern-Aus-reef-mapping\data\AU_NESP-MaC-3-17_AIMS_NW-Aus-Features

References: Bancroft, K.P. (2009). Establishing long-term coral community monitoring sites in the Montebello/Barrow Islands marine protected areas: Site descriptions and summary analysis of baseline data collected in December 2006. Marine Science Program Data Report MSPDR9. June 2011. Marine Science Program, Department of Environment and Conservation, Perth, Western Australia, 91p https://library.dbca.wa.gov.au/static/Journals/080598/080598-09.pdf