Coral Sea Features - Dataset collection - Coral reefs, Cays, Oceanic reef atoll platforms, and Depth contours (AIMS)

Full description

This dataset consists of shapefiles for coral atoll platforms, coral reefs, coral cays, shallow depth contours (5 m and 20 m) and names for the Coral Sea region. The primary purpose of this dataset is to support management of the Coral Sea Marine Park, academic research and the creation of maps for the region. This dataset was created by manually digitising features from composite satellite imagery (Sentinel-2 and Landsat 8/9), cross-referenced against bathymetry datasets and marine charts.

Each shapefile layer includes attributes that record the level of confidence in the mapping of each features and their classification against the Natural Values Common Language. The dataset covers reefs both within and just outside the Australian Coral Sea Marine Park boundary.

Core Data Layers:

1. Atoll Platforms (CS_AIMS_Coral-Sea-Features_2025_Atoll-platforms.shp)
These are polygons representing the geological foundation of each atoll, mapped to a depth of 100 m. This depth represents the approximate maximum depth of photic coral growth in the Coral Sea. Much of the atoll platforms, except where there are coral reefs, are covered in soft sediment formed from calcifying algae. All areas shallower than 100 m were included in this layer regardless whether they had shallow reefs on them, were deep banks, or in some cases a large block of an atoll that has detached and settled next to the main atoll.

2. Reef and Cay Boundaries (CS_AIMS_Coral-Sea-Features_2025_Reefs-cays.shp)
This layer contains the boundaries of coral reefs and coral cays. In this dataset a coral reef a region of highly connected hard substrate patches, formed by hard corals and other calcifying organism, the sand between these patches and the surrounding sand that is highly ecologically connected to the reef, based on grazing halos. Coral reefs are not limited in depth, but by the presence of hard substrate and light availability.
In general, hard substrate patches closer than 100 m apart are joined together to form larger reef features. As a result, large atoll platforms, such as Lihou Reefs are represented as large, almost continuous reefs around the rim of the atoll, with hundreds of smaller reefs within the atoll lagoon, while atolls with no lagoon, such as Moore Reefs, there is a one-to-one match between the coral reef boundary and the atoll platform.
Coral cays are represented as a region where a cay has been observed over time, rather than a single time snapshot of their boundary. The cay boundary represents the outer hull of the movement of the cay over the study period from 2014 – 2023 (see Coral Cays Over Time). Cays are excluded from the reef polygons so that the reef and cay features do not overlap. Each cay has a Stability attribute to indicate the level of movement of the cay and whether it is always present.

3. Depth Contours: 5 m and 20 m (CS_AIMS_Coral-Sea-Features_2025_Depth5m_MSL.shp, CS_AIMS_Coral-Sea-Features_2025_Depth20m_MSL.shp)
These polygons approximate shallow depth zones derived from satellite-derived bathymetry. The 5 m contour delineates near-surface reef flats likely to experience strong wave action. The 20 m contour represents the shallow areas of the atolls that can be explored by dives and reef areas likely to have the highest coral diversity. Where shallow areas (20 m) do not overlap with reef boundaries this is because they correspond to sandy areas with no hard substrate.

4. Names (CS_AIMS_Coral-Sea-Features_2025_Names.shp)
This point layer records the names of places associated with the Coral Sea. These names are positioned based on their original sources and are not directly connected to mapped features. They are intended to assist with creating maps of the region. This includes official names (from AHO marine charts and the national gazetteer) and unofficial names sourced from literature.

Supplemental Extra Data Layers:

These layers are provided for completeness of the dataset.

1. Coral Cays Over Time (extra/Cays-over-time/CS_AIMS_Coral-Sea-Features_2025_Cays-over-time_2014-2023.shp)
This layer tracks the shape and position of individual coral cays between 2014 and 2023, highlighting the ephemeral or stable nature of each cay. The target was to map each cay once per year from Sentinel 2 imagery. Sentinel 2 imagery was reviewed from the first day of imagery available. The first available low cloud, low sun glint and low wave image was used. All the cays visible in the image were then mapped. The date was then jumped to the next year. The process was then repeated. In cases where insufficient suitable images were found the temporal spacing was shortened, down to as low as three months. If this still didn't result in a suitable set of images then Landsat 8 - 9 imagery was used starting from 2014. The union of all the observed cays was used to define the cay boundary in the Reef-cays shapefile. The cay region boundary was drawn manually from the polygons in this dataset. This dataset includes the dates and source imagery used to map each boundary. Note: The tidal levels were not reviewed or normalised when selecting the days to be digitised.

2. Depth class assessment (extra/Depth-class-assessment/CS-Features-Satellite-depth-class-vs-AHO.shp)
This shapefile provides a comparison between the depth classifications based on satellite imagery with depth measurements taken from AHO marine charts. This includes 880 reefs where there was a strong alignment between the reef boundary and the AHO marine charts. The marine chart depths were recorded for each reef, along with the depth classification determined from satellite imagery using the constrast enhanced red channel and blue channel. This dataset is analysed by script 07-sat-depth-class-vs-aho.py to determine the accuracy of the satellite image classifications.

2. Summary stats (extra/CS_AIMS_Coral-Sea-Features_2025_Summary-stats.xlsx)
This spreadsheet contains a copy of the attributes from the Atoll Platforms and Reefs and Cay Boundaries and includes a series of pivot tables to calculate summary statistics for the count and area of reefs of various categories.

3. Potential unmapped reefs (extra/Potential-reefs/CS_AIMS_Coral-Sea-Features_2025_Potential-unmapped-reefs.shp)
This layer corresponds to point locations of features that might correspond to deep coral reefs that were not included in the main reefs and cay boundaries dataset. This layer was created to assist in estimating the level of uncertainty in the mapping of deep patch reefs, to determine the number of reefs that are potentially missing from the reef dataset. Confidence that these points correspond to real reefs is low (maybe 5 – 30% are reefs) because it is highly likely that many of them correspond to visual artefacts, noise or algal patches that look similar to small deep reefs.

4. Position-mask (extra/Position-mask/CS_AIMS_Coral-Sea-Features_2025_Position-mask.shp)
This mask delimiting the lagoonal areas from the rim areas of the atolls. This was used
Reef Identifiers

5. Study Area (extra/Study-area/CS_AIMS_Coral-Sea-Features_2025_Study-area.geojson)
This layer represents the bounds of the region that was studied. It includes areas where reefs were searched for, but none found. This is used as the bounding polygon for the dataset metadata.

Reef Identifiers:

To assist in management and communication about specific features all reefs, cays and atoll platforms were allocated permanent identifiers based on an approach inspired by the reef ID system used on the GBR, but with the capacity to allocate identifiers to features globally. The identifier can be used to uniquely communicate about a reef, even when there is no existing name. The identifiers are allocated on a global 100x100 grid, where each grid cell is 3.6 degrees x 1.8 degrees in size, with 0, 0 corresponding to -180 degree longitude and -90 degrees latitude. All features within each grid cell are allocated a feature count, ordered by the sum of their latitude and longitude. Should new features need to be added in the future, all existing identifiers should remain unchanged.
The basic structure of the identifier is {letter to indicate the class of features}-{grid code: 2 digit longitude, 2 digit latitude }-{feature count within grid}, for example A-9140-03 for North Flinders Reefs and R-9140-078 for Main Cay Reef. All atolls start with the letter ‘A’, while reefs and cays start with ‘R’.
It should be noted that the identifiers are not hierarchical and so the identifiers of the reefs on an atoll are not directly linked. In most cases they are however likely to have the same grid code.


Methods Summary

Satellite imagery was the primary data source for reef mapping. Bathymetry and marine chart data was used for validation and calibration, but not for reef boundary mapping. Reef boundaries were digitised from cloud-free composite imagery from Sentinel 2 MSI (10 m) and Landsat 8-9 (30 m) spanning 2015–2021, optimised for benthic visibility. Images used in these composites were manually selected, filtering for low cloud cover, high water clarity, and low sun glint. Google Earth Engine was used for all processing. Full details of the image preparation are provided in Lawrey & Hammerton, 2022.

Atoll platform boundaries were mapped using multi-beam bathymetry data, supplemented by satellite imagery where high-resolution bathymetry was unavailable. The 100 m depth contour was used to define platform boundaries, approximating the maximum depth of photic coral growth. In areas lacking direct bathymetric data, boundaries were inferred from slope patterns observed in satellite imagery and adjacent contours. Boundaries were manually digitised, with accuracy estimated based on the most ambiguous perimeter sections.

Reef boundaries were mapped by identifying exposed hard substrate areas, formed by coral accretion, and clustering nearby reef patches based on distance. Typical clustering distances ranged from 50 to 250 m, with the largest distances applied to the sandy back-reef slopes of atoll rim reefs to prevent excessive fragmentation. Sand areas between reef patches were included within the reef boundary, ensuring ecological continuity. A small sandy buffer was added to account for grazing halos, with its size informed by typical halo dimensions in the region. This methodology ensures reef features contain at least 5% hard substrate while avoiding excessive fragmentation of large reef structures.

Reef features were classified based on their shallowest point as very shallow (<2.5 m LAT), shallow (2.5–25 m LAT), or deep (>25 m LAT) using Sentinel-2 spectral band visibility, calibrated against Australian Hydrographic Office (AHO) depth data. Red (B4) band visibility indicated very shallow features, green (B3) band visibility indicated shallow reefs, and features absent from both were classified as deep. Detection of the reef within each band required at least nine contiguous pixels brighter than the surrounding background noise. Classification thresholds were tuned against 880 reefs, achieving 87–88% accuracy (see CS-Features-Satellite-depth-class-vs-AHO.shp and 07-sat-depth-class-vs-aho.py). The depth threshold categories were chosen to align closely with the Natural Values Common Language (NVCL) framework.

Coral cays were mapped manually 7 - 10 times each across a time series from 2014 to 2023 using Sentinel-2 and Landsat-8 near-infrared bands. This time series was then used to determine the stability of the cays and the area where they are likely to exist. The outer hull of the mapped time series was used to produce the cay boundaries. No systematic tidal adjustments were used, as cay movement over time exceeded errors due to tidal variations.

The 5 m depth contour was derived using the Stumpf band-ratio method, leveraging logarithmic transformations of Sentinel-2 green (B3) and blue (B2) bands. Calibration was performed on the GBR outer reefs to optimize depth estimates while compensating for variations in benthic reflectance. The 20 m contour was mapped using a direct threshold on the green band (B3) due to its lower noise than the band-ratio method at this depth.

Official Names for the Coral Sea were sourced from the AHO ENC series and AHO chart series maps, as well as the ICSM Gazetteer of Australia 2018. Additional unofficial names were sourced from literature. The source of the names was recorded in the dataset. The position of the names were based on a combination of the positions in the name sources. Historic maps of the Coral Sea from TROVE were reviewed to establish and confirm that intended feature associated with each name.


Limitations / Errata

Coral reefs: All features in this dataset were mapped by visual interpretation of satellite imagery and limited bathymetry. Consequently, deeper areas could be misclassified or omitted, especially small reefs below 40 m.

Atoll platforms: The positional accuracy of the 100 m atoll platform boundary varies from 50 – 1000 m depending on the source used to digitise the edge. This error estimate is recorded for each feature.
This dataset should NOT BE USED for NAVIGATION purposes. There are likely to be small shallow reefs patches that represent a navigation hazard that are not represented.

Depth contours: The depth contours are a broad guide and should not be used for navigation as they are known to miss small navigation hazards.

Cay boundaries: While the mapped cay boundaries are intended to encompass the range of movement of the cays, it was only based on a relatively short history. It is therefore likely that the cays will eventually shift outside the mapped extents.
Cays over time: The accuracy of the perimeter of the mapped cays snapshots are estimated to be 20 – 50 m based on a combination of the limited imagery resolution, tide uncertainties and the uncertainty in determining areas exposed to air.

Depth classifications: Most depth classifications were determined by the visibility of the top of the reefs in the satellite image (see DepthCatSr, 'Sentinel 2 Red' and 'Sentinel 2 Green'). This process does not compensate for CDOM affecting depth visibility. It also does not compensate for tide levels, although since we used image composites made from multiple images each with different tides, the image composites are likely to be near the median tide. In v1-2 we identified that depth thresholds with the highest accuracy for 'Very shallow' and 'Deep' depth classifications are -2.4 m and -24 m (see 07-sat-depth-class-vs-aho.py), with an overall accuracy of 88.1%. Features with an available marine chart depths (DepthCatSr = 'AHO Marine Chart') were classified using thresholds of -2.5 m and -25 m. We retain these thresholds (-2.5 m and -25 m) in v1-2 to retain consistency between AHO marine chart assigned depth categories and satellite derived classifications. This resulting a slightly lower accuracy (87.6%) than optimum (88.1%) for the satellite image based classification.


Relevance and Discovery:

This dataset is ideal for regional-scale planning, habitat modelling, and conservation prioritization in the Coral Sea, particularly within the Coral Sea Marine Park. It provides an unprecedented level of detail for understanding reef geomorphology and potential mesophotic habitats, supporting both research questions (e.g., the role of lagoon patch reefs) and practical management tasks (e.g., zoning decisions, environmental impact assessments). The inclusion of dynamic cay boundaries offers insights into beach and turtle-nesting site stability.

This dataset contains mapped reefs for the following atoll platforms: Abington Reef, Ashmore Reef, Beva Reef, Boot Reef, Bougainville Reef, Calder Bank, Cato Bank, Chilcott Islet, Dart Reef, Diamond Islets, Diane Bank, Eastern Fields, Flora Reef, Frederick Reefs, Herald Cays, Herald Surprise, Homes Reefs, Kenn Reefs, Lihou Reefs, Magdelaine Cays, Malay Reef, Marion Reef, McDermott Bank, Mellish Reef, Moore Reefs, North Flinders Reefs, Osprey Reef, Portlock Reefs, Saumarez Reefs, Selfridge Bank, South Flinders Reefs, Vema Reef, Willis Islets and Wreck Reefs.

Format of the data:

- CS_AIMS_Coral-Sea-Features_2025_Reefs-cays.shp (Polygon shapefile, 6.57 MB, 3248 features)
- CS_AIMS_Coral-Sea-Features_2025_Atoll-platforms.shp (Polygon shapefile, 340 KB, 69 features)
- CS_AIMS_Coral-Sea-Features_2025_Depth5m_MSL.shp (Polygon shapefile, 3.3 MB, 5546 features)
- CS_AIMS_Coral-Sea-Features_2025_Depth20m_MSL.shp (Polygon shapefile, 3.4 MB, 1388 features)
- CS_AIMS_Coral-Sea-Features_2025_Names.shp (Point shapefile, 85 KB, 130 features)
- CS_AIMS_Coral-Sea-Features_2025_Cays-over-time_2014-2023.shp (Polygon shapefile, 348 KB, 663 features)
- CS_AIMS_Coral-Sea-Features_2025_Position-mask.shp (Polygon shapefile, 580 KB, 2 features)
- CS_AIMS_Coral-Sea-Features_2025_Potential-reefs.shp (Point shapefile, 12 KB, 265 features)
- CS_AIMS_Coral-Sea-Features_2025_Study-area.geojson (Polygon GeoJSON, 1 KB, 1 feature)
- CS_AIMS_Coral-Sea-Features_2025_Summary-stats.xlsx (Excel spreadsheet, 437 KB, 7 tabs)



Data dictionary:

The following data dictionary applies to Reefs-cays and Atoll-platforms:

- FeatConf (Feature Confidence): (String, 10 characters)
Likelihood that this feature is a real marine feature and not a misinterpretation of noise or an anomaly in the imagery, such as interpreting a cloud shallow or a turbid plume as a reef. This is estimated based on how clear the feature is in the available imagery. If the confidence was too low the feature was not digitised. The allocation of this classification is based on expert assessment based on the evidence available for the feature.

• Low – This also includes features that are very close to visible limit of detectability where the feature is just visible above the surrounding noise in the imagery. (Confidence is > 60%).

• Medium – Feature is visible in multiple images, but there is some uncertainty as to whether it might be a misinterpretation of the benthic substrate, such as sediment covered in organic material, rather than a hard substrate. (Confidence is > 85%).

• High – Feature is clearly visible with crisp boundary, good repeat evidence (Confidence is > 95%).


- TypeConf (Classification Type Confidence): (String, 10 characters)
Confidence that classification assigned to the feature is correct. While a feature might be repeatedly visible in the imagery, and so its Feature Confidence is high, we may have trouble assessing its classification and so assign Classification Type Confidence of low. Confidence is assigned based on expert assessment based on the evidence available. This will often be based on the clarity of the visual textures of the feature and how unusual the feature is.

• Low: Low confidence (< 60%). The classification is inferred largely from indirect evidence, such as the shape or contentual location. No or little textural information is available, or the feature is unusal in structure.

• Medium: Medium confidence (60% - 90%). Some of the visual characteristics are visible, but there might be regional differences, or mixed types within the region. Most features would tend to be in this category.

• High: High confidence (90% - 98%) in the type classification. The distinguishing visual characteristics are visible in the imagery and the feature is homogenous (i.e. predominately one type)

• Very high: High confidence (99%) in the type classification .


- EdgeSrc (Edge Image Sources): (String, 255 characters)
The source of the imagery / dataset used to digitise the feature or refine its boundary.
The following are a description of the imagery used for map the features:

• Sentinel 2 B02 composite (Lawrey & Hammerton, 2022): Sentinel 2 composite imagery (R1 collection). Blue channel contrast enhanced to show faint deep features.

• Landsat 8 B1 composite (Lawrey & Hammerton, 2022): Blue channel of Landsat composite imagery. This was used when Sentinel 2 was not available.

• Selfridge-bank Sentinel 3 (Lawrey & Hammerton, 2022): One remote bank (Selfrige Bank) was mapped from low resolution Sentinel 3 imagery because there was no Sentinel 2 or Landsat imagery.

• Sentinel 2 B02 Supplemental imagery 2022-05-14: In the northern Coral Sea near Torres Strait the composite imagery was not good enough to map the reefs and so an additional individual image was used.

• Sentinel 2 B02 composite R2 (Lawrey & Hammerton, 2022): The composite imagery from created from two sets of imagery grouped from the clearest imagery (R1) and the next best imagery (R2). In some local patches the R2 imagery is superior to the R1 imagery and so was the primary source of imagery used for the mapping.

• Hull of cay outline time series. Sentinel 2 - B8 (2016 to 2023, n=5), Landsat 8-9 (2014 to 2022, n=5): This indicates that the cay region was mapped from a time series of five Sentinel 2 images and five Landsat 8-9. The Sentinel 2 imagery was the Near infrared band (B8). The feature boundary was determined from the combined outer hull of the position and shape of the cay from the multiple satellite time steps.

• S2 TrueColour + cay time series. Sentinel 2 - B8 (2017 to 2023, n=8), Landsat 8 - B4 (2014, n=1): Similar to the 'Hull of cay output time series', but the true colour composite imagery was also referred to understand the likely distribution of the cay.

• Bathymetry - Visioning the Coral Sea 2020 DOI:10.26186/140048; Sentinel 2 B02 composite DOI:10.26274/NH77-ZW79: This is an EdgeSrc for the mapping of a Atoll platform. This reference indicates that the edge was mapped from both multi-beam (Visioning the Coral Sea) and Sentinel 2 blue channel imagery. The DOIs indicate the dataset origin.


- EdgeAcc_m (Edge accuracy): (Integer)
Positional accuracy of the feature boundary. This was estimated during digitisation based on the uncertainty in the visual edge boundary. The units of this attribute are in metres. In general smaller features tended to have lower edge errors and the highest errors occurred with low gradients in noisy environments. The mapped feature boundary with a buffer (inwards and outwards) equal to the EdgeAcc_m should contain the true edge of the feature for 95% of the perimeter.

• Atoll Platforms: The edge accuracy was estimated based on the source data used for the section being digitised. Edges estimated from multi-beam data were considered to have a spatial positioning error (typically assumed to be 20 m), combined with an error from the grid size (30 - 60 m). Where the estimated platform edge was outside the multi-beam data then a visual estimate of the reef edge slope was used to determine the edge. The error was estimated based on the slope of the reef edge and the visibility of the edge.

• Reefs: The edge accuracy was estimated from the uncertainty in the mapped boundary. Higher levels of uncertainty in the mapped feature boundaries were associated with: deep features at the edge of visibility, complex clustering of many small features, ambiguous substrate classification, gradual reef slopes near the limit of visibility, Coloured Dissolved Organic Matter (CDOM) in the water column causing variable depth of visibility leading to ambiguity in understanding the topography of the atoll platforms, and significant brightness shifts across the image due to banding in the Sentinel 2 from its push broom sensor.

• Cays: The mapped cay boundaries corresponds to the area where the cay is likely to be within, taking into account the movement of the cay. The EdgeAcc_m is an indication of the mapping accuracy and the movement of the cay within its mapped boundary. The EdgeAcc_m indicate the likely distance you might have to travel from the mapped cay region boundary to find the actual edge of the cay at any point in time. The error was estimated from largest distance between cay region edge (hull of all mapped reef boundaries over time) to the edge of the most distant snapshot of the cay boundary. Where the cay is ephemeral then the error corresponds to the distance between the centre and the furthest edge of the cay region.


- Stability: (String, 10 characters)
This is an estimate of the stability of the boundary of the feature. This attribute is most useful for understanding the stability of coral cays. This is estimated by looking at the change in the shape of the feature over time, typically assessed over a 8 to 10 year period. This assessment is very important for coral cays due to their high level of movement.

• Ephemeral - The feature is only present < 50 % of the time. For cays this means that there is an area above mean high water that periodically forms in the mapped cay region. In general, cays in this category they tend to be small and move around and typically cover less than 10% of the region that they appear in.

• Very low - The feature is only present 50 - <100% of the time. The feature moves around a lot within its region and typically only covers less than 20% of the region area.

• Low - The feature is always present but its size and outline change significantly over time. It's outline only minimally overlaps over time. The feature typically covers 20 - 50% of its mapped area.

• Medium - The feature is always present, but its outline changes by up to 30% of its width.

• High - The feature is always present, but its outline changes 10% - <30% of its width.

• Very high - This feature is always present. Its outline changes < 10% of its width.

• Fixed - This feature is permanent and its outline changes minimally, less than the measurement error. All hard reefs fall into this category.


- Stage: (Integer)
Stage of the project that the feature was added to the dataset. This was intended to help track how late in the project a feature was detected and mapped. Each stage involved a review, or the incorporation of an additional information source or technique.


- DepthCat: (String, 15 characters)
Depth of the shallowest portion of the feature, divided into categories.
• Very Shallow: Less than 2.5 m below Lowest Astronomical Tide (LAT)
• Shallow: Between 2.5 m and 25 m below LAT
• Deep: Greater than 25 m below LAT
• Land: Feature corresponds to land. This corresponds to cays.


- DepthCatSr: (String, 40 characters).
Source of information used to classify the depth of the top of the feature.
• Sentinel 2 Red: The `Very Shallow` classification was estimated based on whether the reef was visible in the red channel of the Sentinel 2 B4 cloud free composite imagery. The contrast was enhanced so that features barely visible above the noise floor could be detected. Features where there are were at least 9 pixels (3x3 pixels) brighter than the surrounding noise, located within the centroid of the reef boundary were classified as very shallow reefs. The TrueColour imagery was used to verify if small peaks were noise or real reef tops, i.e. does the small peak in the red channel spatially align with the reef visible in the green and blue channels. Features that were not very shallow or deep (see Sentinel 2 Green) were considered as shallow.

• Sentinel 2 Green: The `Shallow` classification was estimated based on whether the reef was visible in the green channel of the Sentinel 2 B3 cloud free composite imagery. The contrast was enhanced so that features barely visble above the noise floor could be detected. These were classified as `Shallow` reefs. Reefs with no areas visible in the green channel were classified as `Deep` (mesophotic) reefs. To improve the sensitivity of the classification the R2 imagery (Sentinel 2 B3 from DeepFalse imagery) was transparently overlaid on the R1 imagery (Lawrey 2022) with 50% transparency to visually average the imagery noise. This was to combine both image composites to further lower the image noise. To help distinguish between noise and a real shallow reef a positive detection required at least 9 pixels (3x3 pixels) to be brighter than their surrounding noise and located close to the centroid of the reef boundary. Where there was doubt the R1 and R2 imagery was individually viewed to check if the visual peak spatially correlated with the expected reef peak visible in the Blue channel, which can see much deeper.

• AHO Marine Chart: The reef top high was determined from the depths listed on the AHO Marine Charts and this depth was used to determine the classification. Depths < 2.5 were classified as `Very Shallow`; depths >= 2.5 and < 25 were classified as `Shallow`, and depths >= 25 classified as 'deep'. Only a subset of the reefs had AHO Marine Chart depth estimates available. The AHO Marine Chart depth estimates were only assigned to reefs where there was a very good alignment of the depth measure on the chart and the centroid of the mapped reef.

• Landsat 8 Blue: Landsat depth estimation was used when Sentinel 2 imagery was not available. This was only used on one feature.

• Landsat 8 Green: Landsat depth estimation was used when Sentinel 2 imagery was not available. This used the same approach as the 'Sentinel 2 Green' but with Landsat 8 composite imagery.

• Sentinel 3: Sentinel 3 imagery was only used on one remote feature where Landsat and Sentinel 2 imagery was not available. This depth estimate was only approximate.

- NvclEco: (String, 40 characters) 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.


- NvclEcoCom: (String, 40 characters)
Natural Values Common Language classification Ecosystem Complex classification.
• Oceanic coral reefs: Coral reefs occurring seaward of the continental shelf break.

- Country: (String, 30 characters)
Country that this feature predominantly resides in, as determined by the Australian EEZ.
•    Australia
•    PNG
•    New Caledonia

- 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)

- AvArea_km2: (Decimal number, length: 10, precision: 6)
Average area of the feature. For reefs this matches the Area_km2 attribute. This attribute is intended to indicate the typical area of each cay, when it is present. The less stable a cay the more is moves around leading to a cay region, the outer hull of the all the mapped movement of the cay over time, that is significantly larger than the cay appears at any one time. AvArea_km2 is intended to give an estimate of the typical area of the cay at any one time when it is present. For ephemeral and very low stability cays, where they are not always present, the average only includes time slices where the cay was present.

- ReefID: (String, 15 characters)
Permanent identifier assigned to the feature. This can be used in place of a name. Structure of the identifier 'R-{grid longitude}{grid latitude}-{feature count in grid cell}'.


- RB_Type_L3 (Reef Boundary Type Level 3 Classification): (String, 50 characters):
A indication of the type of the feature. This is the most detailed classification assigned to the feature. This classification scheme is intended to have a one-to-one mapping with attributes that can be assigned in the National Benthic Habitat Layer. This dataset uses a subset of the full classification scheme.

• Atoll Shallow Patch Coral Reef: Small isolated coral reef occurring in the lagoon of a coral atoll with highest point shallower than 30 m. They are characterised by their small size (< 1 km across) and isolation from neighbouring reefs (typically separated by > 500 m). They are also typically round and have very steep, near vertical, reef slopes. Most of these reefs have a significant grazing halo around them, leading to a pale ring around the central reef in satellite imagery.

• Atoll Deep Patch Coral Reef: Small isolated coral reef occurring in the lagoon of a coral atoll with highest point deeper than 30 m. These are generally the same as the Atoll Patch Coral Reef, but they do not reach close to the surface. The depth definition was chosen to align with the Natural Values Common Language Mesophotic Coral Reefs definition.

• Atoll Shallow Rim Coral Reef: Coral reef on the outer edge of a coral atoll where the structure of the reef is affected by the currents or waves on the edge. Generally within 2 km of the edge. Shallower than 5 m at the highest point. The outer edge of these reefs face the open ocean with a medium to high slope. This outer edge is exposed to oceanic waves, clearing much of the sediment from the reef slope. The back side of the reef faces the atoll lagoon and is typically gently sloping and very sandy.

• Atoll Shallow Flow Coral Reef: Coral reef on the outer edge of a coral atoll where the structure of the reef is affected by the currents or waves on the edge. Generally within 2 km of the edge. Between 5 - 30 m in depth at highest point. The shape of the hard substrate structures should be influenced by currents to be In some cases, these reefs completely surround the outer edge of the atoll forming a closed lagoon, such as Osprey Reef in Coral Sea.

• Atoll Deep Flow Coral Reef: Coral reef on the outer edge of a coral atoll where the structure of the reef is affected by the currents or waves on the edge. Generally within 2 km of the edge. Below 30 m at highest point.
These correspond to drowned low relief reefs that form on the edge of coral atolls. They tend to have a structure that is dominated by tidal flows, or by consistent oceanic currents. They often form between shallow Atoll Rim Coral Reefs where the water is forced in and out of the atoll lagoon. They typically have a striped pattern along the direction of the current flows, formed by groves in the coral.

• Atoll Shallow Platform Coral Reef: Isolated atoll coral reef with no lagoon that is shallower than 30 m at the highest point. These correspond to the coral reef portion of isolated atolls that are completed filled with no lagoon and are typically small (< 4 km across). The shape of the Atoll Platform Coral Reef will typically closely match that of the Atoll Platform it sits on.

• Atoll Deep Platform Coral Reef: Isolated atoll coral reef with no lagoon that is deeper than 30 m at the highest point.

• Vegetated Cay: A sandy island on top of a coral reef from accumulated calcium carbonate sand with permanent vegetation.

• Unvegetated Cay: A sandy island on top of a coral reef from accumulated calcium carbonate sand with no vegetation. Can be ephemeral. Boundary corresponds to region that the cay exists over in 5 - 10 years.

- RB_Type_L2 (Reef Boundary Type Level 2 Classification): (String, 50 characters):
Higher level feature classification of the feature type.
• Reef Island: A coral cay, an island formed on the top of a coral reef
• Coral Reef: This feature corresponds to a reef formed from hard coral and other calcifying organisms.

- RB_Type_L1 (Reef Boundary Type Level 1 Classification): (String, 50 characters):
• Land: An area that stay dry at mean high tide. It should be noted that for emphemeral cays, there will not always be land in the mapped cay boundary.
• Reef: A feature that is raised above the surrounding area that has exposed (no covered in sediment) hard substrate.

- Name: (String, 50 characters)
Name of the feature. Where available the Indigenous name should be given preference. If no Indigenous name is available, then use the European name. Where both names are available then both are listed. For example, 'Jeaka (Dungeness Reef)'. If only an unofficial name is available then include this name, but mark it with a star '*' at the end to indicate that it is unofficial. For example 'Juliette Cay*'. Names on the AHO Marine Charts are considered as official names.

- NameSrc (Name source): (String, 255 characters)
Reference to the source used for determining the name.
• Australian Hydrographic Office. (2021). AHO Chart Series chart service. Retrieved July 24, 2024, from
https://amsis-geoscience-au.hub.arcgis.com/datasets/geoscience-au::aho-chart-series/about
• Peter Sayre. (2019). Australia's Coral Sea Islands & Marine Park. Bianca Vessel Management Pty. Ltd. Publications.
• AHO ENC Simplified Series (Oct 4, 2022): Australian Hydrographic Office. (2021). AHO ENC Simplified Series service. Retrieved Nov 28, 2022, from
https://amsis-geoscience-au.hub.arcgis.com/datasets/geoscience-au::aho-enc-series/about
• AHO Chart Series (28 Nov 2022): Australian Hydrographic Office. (2021). AHO Chart Series chart service. Retrieved Nov 28, 2024, from
https://amsis-geoscience-au.hub.arcgis.com/datasets/geoscience-au::aho-chart-series/about
• Beaman (2012) 3DGBR Geomorphic Features: Beaman, Rob. (2012). Project 3DGBR: Great Barrier Reef and Coral Sea Geomorphic Features (JCU) [Dataset]. School of Earth and Environmental Sciences, James Cook University [producer]. eAtlas [distributor]. http://creativecommons.org/licenses/by/3.0/au/ [license]. Available from https://eatlas.org.au/data/uuid/25685ba5-6583-494f-974d-cce2f3429b78

- OtherNames: (String, 255 characters)
Other names that this feature is known by, including unofficial names that are sourced from literature.

- OtherNaSrc (Other Name Source): (String, 255 characters)
Reference to the source of the other names.

- Position(String, 15 characters)
Position of the feature position relative to the edge of the atoll platforms. The distance of 2 km was chosen based on reviewing the typical distance that the edge affects the reef structures and algal growth patterns on Tregrosse Reefs and the Magdeline Cays and Coringa Islet platforms.
• Atoll lagoon: Feature more than 2 km from the edge of the atoll platform. These features are less affected by waves and tidal currents.
• Atoll rim: Feature is less than 2 km from the edge of the atoll platform. These features are subjected to oceanic waves and in some places significant tidal flows.

- ReefIDNote: (String, 120 characters)
Record when and why a ReefID change was made. For example: '2026-02-23: Herald cays split into two reefs (South west cay and north west cay)'

- PrevReefID: (String, 12 characters)
Previous ReefID that this feature was known by. When a reef is split this field records any old IDs that this area corresponded to. Example: 'R-9140-550'.

The following data dictionary applies to Depth5m and Depth20m:

- FID: (Integer)
Row number of the feature

- IMG_SRC: (String)
Source of the imagery used to determine the depth. 'Sentinel 2': Composite imagery used to map the reefs (Lawrey & Hammerton, 2022). 'ArcGIS World': ArcGIS Online World Imagery map service (https://www.arcgis.com/home/item.html?id=10df2279f9684e4a9f6a7f08febac2a9) Accessed 30 Jan 2023.

- PROCESS: (String)
Method used to determine the depth contour. 'Auto SDB': Automated Satellite Derived Bathymetry. For the 5 m depth contour this was based on Strumpf, 2003 band ratios, and for the 20 m contour a threshold applied to the Sentinel 2 green channel. 'Manual': Small features that were removed by the spatial filtering applied to automatic SDB, were manually digitised from high resolution imagery, with a non-filtered estimate of the SDB to assist in depth estimation.


Change Log:

In this section we will note any changes made as part of revisions to this dataset. See https://github.com/eatlas/CS_AIMS_Coral-Sea-Features/blob/main/changelog.md for more details.
2026-02-23 (v1-2)
- Fixed classification for reef R-9140-452 from 'Atoll Deep Platform Coral Reef' to 'Atoll Deep Flow Coral Reef'.
- Reclassified reef R-9140-892 from 'Atoll Shallow Platform Coral Reef' to 'Atoll Deep Platform Coral Reef'. Review of imagery showed that is was likely deeper than 30 m.
- Split the main reef (R-9140-550) corresponding to Herald Cays into two reefs - South West Cay reef retained the original ReefID of R-9140-550 and the new smaller reef, North West Cay reef, was allocated a new ReefID of R-9140-1873.
- Improve description of the cays over time dataset. Added an analysis of the satellite derived depth classifications against the AHO marine charts. This includes adding data/v1-2/extra/Depth-class-assessment/CS-Features-Satellite-depth-class-vs-AHO.shp as the input data for the analysis and 07-sat-depth-class-vs-aho.py to the source code. Migrated the figured generated by 06-plot-cay-region-vs-cay-area.py into data/v1-2/figures/06 and added figures generated by 07-sat-depth-class-vs-aho.py to data/v1-2/figures/07.
- As part of finalising the reef count for the Coral Sea an additional review of the potential reefs dataset was done. This was to determine the size distribution to assign to these features. An additional 30 potential reefs were added across Diane Bank and Tregrosse Bank.
2025-03-25 (v1-1)
- Porpoise Cay (R-9337-031) was incorrectly as unvegetated in v1. Its RB_Type_L3 classification was set to 'Vegetated Cay' and the associated data/extra/CS_AIMS_Coral-Sea-Features_2025_Summary-stats.xlsx updated to match.
Added AvArea_km2 to Reefs-Cays to better represent the size of cays relative to their cay region. ReefID and Area_km2 were added to Cays-over-time.
2025-03-07 (v1)
- Initial publication of the dataset.

Lineage

Maintenance and Update Frequency: notPlanned

Notes

Credit
This dataset collection was developed using funding by Parks Australia (Commonwealth of Australia) and the Australian Institute of Marine Science.