DEA Coastlines

Geoscience Australia Landsat Coastlines Collection 3

1.0.0-beta flat-square
Digital Earth Australia
Geoscience Australia Landsat Collection 3
Resource type
Date modified


Australia has a highly dynamic coastline of over 30,000 km, with over 85% of its population living within 50 km of the coast. This coastline is subject to a wide range of pressures, including extreme weather and climate, sea level rise and human development. Understanding how the coastline responds to these pressures is crucial to managing this region, from social, environmental and economic perspectives. 

What this product offers

Digital Earth Australia Coastlines is a continental dataset that includes annual shorelines and rates of coastal change along the entire Australian coastline from 1988 to the present. 

The product combines satellite data from Geoscience Australia's Digital Earth Australia program with tidal modelling to map the typical location of the coastline at mean sea level for each year. The product enables trends of coastal erosion and growth to be examined annually at both a local and continental scale, and for patterns of coastal change to be mapped historically and updated regularly as data continues to be acquired. This allows current rates of coastal change to be compared with that observed in previous years or decades. 

The ability to map shoreline positions for each year provides valuable insights into whether changes to our coastline are the result of particular events or actions, or a process of more gradual change over time. This information can enable scientists, managers and policy makers to assess impacts from the range of drivers impacting our coastlines and potentially assist planning and forecasting for future scenarios. 


  • Monitoring and mapping rates of coastal erosion along the Australian coastline 
  • Prioritise and evaluate the impacts of local and regional coastal management based on historical coastline change 
  • Modelling how coastlines respond to drivers of change, including extreme weather events, sea level rise or human development 
  • Supporting geomorphological studies of how and why coastlines have changed across time 

Related products


Bishop-Taylor, R., Sagar, S., Lymburner, L., Alam, I., & Sixsmith, J. (2019). Sub-pixel waterline extraction: Characterising accuracy and sensitivity to indices and spectra. Remote Sensing, 11(24), 2984. Available:

Data access

Link to data WFS service (annual coastlines data)
WFS service (rates of change statistics data)
Link to maps Interactive DEA Coastlines product on DEA Maps
Web services WMS (layer name: dea:DEACoastLines)
CMI RESTful node ID 581
Security classification Unclassified
Update frequency Annually
Product life span -

Access notes

DEA Maps

To explore DEA Coastlines on the interactive DEA Maps platform, visit the link below:

To add DEA Coastlines to DEA Maps manually:

  1. Open DEA Maps.
  2. Select Add data on the top-left.
  3. Select Coastal > Digital Earth Australia Coastlines > Digital Earth Australia Coastlines (beta)
  4. Click the blue Add to the map button on top-right.

By default, the map will show hotspots of coastal change at continental scale. Red dots represent retreating coastlines (e.g. erosion), while blue dots indicate seaward growth. The larger the dots and the brighter the colour, the more coastal change that is occurring at the location. 

More detailed rates of change will be displayed as you zoom in. To view a time series chart of how an area of coastline has changed over time, click on any labelled point (press "Expand" on the pop-up for more detail):

DEA Maps zoom example

Zoom in further to view individual annual coastlines:

DEA Maps coastlines example

Note: To view a DEA Coastlines layer that is not currently visible (e.g. rates of change statistics at full zoom), each layer can be added to the map individually from the Coastal > Digital Earth Australia Coastlines > Supplementary data directory.

Loading DEA Coastlines data from the Web Feature Service (WFS) using Python

DEA Coastlines data can be loaded directly in a Python script or Jupyter Notebook using the DEA Coastlines Web Feature Service (WFS) and geopandas:

import geopandas as gpd

# Specify bounding box
ymax, xmin = -33.65, 115.28
ymin, xmax = -33.66, 115.30

# Set up WFS requests for annual coastlines & rates of change statistics
deacl_coastlines_wfs = f'' \
                       f'service=WFS&version=1.1.0&request=GetFeature' \
                       f'&typeName=dea:coastlines&maxFeatures=1000' \
                       f'&bbox={ymin},{xmin},{ymax},{xmax},' \
deacl_statistics_wfs = f'' \
                       f'service=WFS&version=1.1.0&request=GetFeature' \
                       f'&typeName=dea:coastlines_statistics&maxFeatures=1000' \
                       f'&bbox={ymin},{xmin},{ymax},{xmax},' \

# Load DEA Coastlines data from WFS using geopandas
deacl_coastlines_gdf = gpd.read_file(deacl_coastlines_wfs)
deacl_statistics_gdf = gpd.read_file(deacl_statistics_wfs)

# Ensure CRSs are set correctly = 'EPSG:3577' = 'EPSG:3577'

Loading DEA Coastlines data from the Web Feature Service (WFS) using R

DEA Coastlines data can be loaded directly into R using the DEA Coastlines Web Feature Service (WFS) and the sf package:


# Specify bounding box
xmin = 115.28
xmax = 115.30
ymin = -33.66
ymax = -33.65

# Read in DEA Coastlines annual coastline data, using `glue` to insert our bounding 
# box into the string, and `sf` to  load the spatial data from the Web Feature Service 
# and set the Coordinate Reference System to Australian Albers (EPSG:3577)
deacl_coastlines = "{ymin},{xmin},{ymax},{xmax},urn:ogc:def:crs:EPSG:4326" %>% 
  glue::glue() %>%
  sf::read_sf() %>% 

# Read in DEA Coastlines rates of change statistics data
deacl_statistics = "{ymin},{xmin},{ymax},{xmax},urn:ogc:def:crs:EPSG:4326" %>% 
  glue::glue() %>%
  sf::read_sf() %>% 

Technical information

The DEA Coastlines product contains three layers: 

Annual coastlines

Annual coastline vectors from 1988 to 2019 that represent the median or ‘typical’ position of the coastline at approximately mean sea level tide (0 m AHD) for each year (Figure 1). 

  • Semi-transparent coastlines have low certainty due to either few non-cloudy satellite observations, poor tidal modelling performance, or aerosol issues (see Caveats and limitations)
DEA CoastLines coastline layer

Figure 1: Annual coastlines from DEA Coastlines visualised on the interactive DEA Coastlines web map

Rates of change statistics

A point dataset providing robust rates of coastal change statistics for every 30 m along Australia’s non-rocky (clastic) coastlines (Figure 2). The most recent 2019 coastline is used as a baseline for measuring rates of change.

DEA CoastLines rates of change statistics layer

Figure 2: Rates of change points from DEA Coastlines visualised on the interactive DEA Coastlines web map

On the interactive DEA Coastlines web map, points are shown for locations with statistically significant rates of change only (p-value < 0.01, see sig_time below). Each point shows annual rates of change (in metres per year; see rate_time below), and an estimate of uncertainty in brackets (95% confidence interval; see se_time). For example, there is a 95% chance that a point with a label -10.0 m (±1.0 m) has an erosion rate between -9.0 and -11.0 metres per year.

The rates of change statistics dataset contains the following attribute columns that can be accessed by clicking on labelled points in the web map: 

Annual coastline distances
  • dist_1990, dist_1991 etc: Annual coastline distances (in metres) relative to the 2019 baseline coastline. Negative values indicate that an annual coastline was located inland of the 2019 baseline coastline. By definition, the dist_2019 baseline column will always have a distance of 0 m.
Rates of change statistics
  • rate_time: Annual rates of change (in metres per year) calculated by linearly regressing annual coastline distances against time (excluding outliers). Negative values indicate retreat and positive values indicate growth. 
  • sig_time: Significance (p-value) of the linear relationship between annual coastline distances and time. Small values (e.g. p-value < 0.01 or 0.05) may indicate a coastline is undergoing consistent coastal change through time. 
  • se_time: Standard error (in metres) of the linear relationship between annual coastline distances and time. This can be used to generate confidence intervals around the rate of change given by rate_time (e.g. 95% confidence interval = rate_time * 1.96)
  • outl_time: Individual annual coastlines are noisy estimators of coastline position that can be influenced by environmental conditions (e.g. clouds, breaking waves, sea spray) or modelling issues (e.g. poor tidal modelling results or limited clear satellite observations). To obtain reliable rates of change, outlying years are excluded using a robust Median Absolute Deviation outlier detection algorithm, and recorded in this column. 
Climate driver statistics
  • rate_soi: The slope of any linear relationship between annual coastline distances and the Southern Oscillation Index or SOI (in metres change per unit of SOI). Negative values indicate that coastlines have historically retreated during La Niña years. Please note: this comparison was made against SOI data from NOAA which does not apply a standard scaling factor to SOI values unlike Australia's BOM. Divide rate_soi and se_soi values by 10 to get rates of coastal change per increase in BOM-scaled SOI values.
  • sig_soi: Significance (p-value) of the linear relationship between annual coastline distances and SOI. 
  • se_soi: Standard error (in metres) of the linear relationship between annual coastline distances and SOI.
  • outl_soi: A list of any years excluded from the SOI regression by the robust outlier detection algorithm.
Other derived statistics
  • retreat, growth: True/False columns indicating whether a shoreline was retreating (i.e. moving inland) or growing (i.e. moving seaward) based on the rate_time column.
  • sce: Shoreline Change Envelope (SCE). A measure of the maximum change or variability across all annual coastlines, calculated by computing the maximum distance between any two annual coastlines (excluding outliers).
  • nsm: Net Shoreline Movement (NSM). The distance between the oldest (1988) and most recent (2019) annual coastlines (excluding outliers). Negative values indicate the shoreline retreated between the oldest and most recent coastline; positive values indicate growth.
  • max_year, min_year: The year that annual coastlines were at their maximum (i.e. located furthest towards the ocean) and their minimum (i.e. located furthest inland) respectively (excluding outliers).
  • breaks: An experimental list of any years identified as non-linear breakpoints in the time series. This can be useful for verifying that a significant trend is indeed linear, or identifying areas of rapid non-linear change (e.g. associated with coastal development or management).

Coastal change hotspots

A point layer giving the average rate of change (in metres per year) for significant statistics points within a moving 5 km window along the coastline (Figure 3). This is useful for visualising regional or continental-scale patterns of coastal change. 

DEA CoastLines summary layer

Figure 3: Coastal change hotspots from DEA Coastlines visualised on the interactive DEA Coastlines web map

Accuracy and limitations

Precision and accuracy

An extensive validation against high resolution coastal monitoring validation datasets was conducted to evaluate the precision and accuracy of the DEA Coastlines product. Validations were performed using existing validation dataset beach profile lines where possible. For each profile line, we measured distances to the intersection between the profile line and each annual DEA Coastline, and compared this to the location where the profile line first intersected with 0 m AHD elevation (i.e. mean sea level). The following datasets were included in the initial round of validation:

  • Coastal Movement Lines derived from photogrammetry for south-western WA (WA Department of Transport)
  • ETA Hydrographic Survey data for south-eastern Queensland (City of Gold Coast)
  • NSW Beach Profile Database derived from photogrammetry and LiDAR for the entire NSW coast (NSW Office of Environmment and Heritage; NSW Water Research Laboratory)
  • Narrabeen-Collaroy Beach Survey Program (Turner et al. 2016)

Precision was considered to be the most critical metric for evaluating the product, as high precision would infer that DEA Coastlines can be used to reliably compare trends in coastal erosion or retreat through time, even if overall accuracy was reduced by a consistent inland or sea-ward offset from true coastline positions (Bishop-Taylor et al. 2019b). The precision of DEA Coastlines was evaluated by comparing correlations between DEA Coastlines and validation dataset along-profile distances, and by computing the standard deviation of differences between the two datasets.

Standard deviation precision ranged from 6.1 m (Narrabeen-Collaroy Beach Survey Program) to 14.2 m (City of Gold Coast), while DEA Coastlines were highly correlated with all validation datasets (ρ = 0.96–0.99; see Figure 4). This sub-pixel precision exceeds the 30 m resolution of the input data (see Bishop-Taylor et al. 2019b), and indicates that DEA Coastlines can closely reproduce true coastline positions and shape both spatially and across time (see Figure 5). 

DEA CoastLines validation results

Figure 4: DEA Coastlines validation results compared to a) Narrabeen-Colloroy Beach Survey Program, b) WA DoT Coastal Movements, c) NSW Beach Profile Database, and d) City of Gold Coast hydrographic survey data
Comparison of DEA CoastLines and WA DoT Coastal Movements validation data at Rockingham, WA

Figure 5: Comparison of DEA Coastlines and WA DoT Coastal Movements validation data at Rockingham, WA

Absolute accuracy was evaluated by calculating two common metrics: Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). Absolute accuracy measured by MAE varied from 5.8 m to 23.7 m, and from 7.3 m to 27.2 m when measured by RMSE. These results were influenced by landward biases in modelled shoreline positions that varied by study area (i.e. shorelines mapped further inland than their true positions). These biases are likely to be caused by differences in spectral characteristics of the underlying substrate, and were an expected consequence of the consistent water index threshold used to facilitate waterline extraction at continental scale (Sagar et al. 2017, Bishop-Taylor et al 2019a, 2019b).

For example, in south Western Australia accuracy was high (MAE of 8.6 m, RMSE of 11.9 m) with only a small 5.0 m landward bias in shoreline positions. However, City of Gold Coast results showed the lowest overall accuracy (MAE of 23.7 m, RMSE of 27.2 m) due to a large 23.2 m landward bias in DEA Coastlines shoreline positions. When this bias was corrected, DEA Coastlines showed a strong ability to reproduce temporal patterns of coastal change over time, with bias-corrected accuracies of 10.2 m MAE and 14.2 m RMSE respectively; see Figure 6:

Bias-corrected results for City of Gold Coast

Figure 6: Bias-corrected validation results for DEA Coastlines compared to City of Gold Coast hydrographic survey data
Limitations of validation approach
  • Annual DEA Coastlines represent typical or median coastline positions at approximately Mean Sea Level throughout an entire year. To compare annual DEA Coastlines to validation datasets, multiple validation observations in a year were combined into a single measurement using a median calculation. In the case where only a single validation observation was taken for a year, this single observation may not be reflective of typical shoreline conditions across the entire year period. Because of this, validation results are expected to be more reliable for validation datasets with multiple observations per year.
  • The current validation approach was biased strongly towards Australia's south-western, southern and south-eastern coastlines due to the availability of historical coastal monitoring data. This bias prevented us from including more complex intertidal environments in our validation, which is likely to have inflated the accuracy of our results due to issues outlined in Caveats and limitations below.

Caveats and limitations

  • Rates of change statistics may be inaccurate or invalid for some complex mouthbars, or other coastal environments undergoing rapid non-linear change through time. In these regions, it is advisable to visually assess the underlying annual coastline data when interpreting rates of change to ensure these values are fit-for-purpose. Regions significantly affected by this issue include:
    • Cambridge Gulf, Western Australia
    • Joseph Bonaparte Gulf, Western Australia/Northern Territory
  • Annual coastlines may be less accurate in regions with complex tidal dynamics or large tidal ranges, and low-lying intertidal flats where small tidal modelling errors can lead to large horizontal offsets in coastline positions. Annual coastline accuracy in intertidal environments may also be reduced by the influence of wet muddy substrate or intertidal vegetation, which can make it difficult to extract a single unambiguous coastline (Bishop-Taylor et al. 2019a, 2019b). It is anticipated that future versions of this product will show improved results due to integrating more advanced methods for waterline detection in intertidal regions, and through improvements in tidal modelling methods. Regions significantly affected by intertidal issues include:
    • The Pilbara coast, Western Australia from Onslow to Pardoo
    • The Mackay region, Queensland from Proserpine to Broad Sound
    • The upper Spencer Gulf, South Australia from Port Broughton to Port Augusta
    • Western Port Bay, Victoria from Tooradin to Pioneer Bay
    • Hunter Island Group, Tasmania from Woolnorth to Perkins Island
    • Moreton Bay, Queensland from Sandstone Bay to Wellington Point
  • Coastlines may be noisier and more difficult to interpret in regions with low availability of satellite observations caused by persistent cloud cover. In these regions it can be difficult to obtain the minimum number of clear satellite observations required to generate clean, noise-free annual coastlines. Affected regions include:
    • South-western Tasmania from Macquarie Heads to Southport
  • In some urban locations, the spectra of bright white buildings located near the coastline may be inadvertently confused with water, causing a land-ward offset from true coastline positions. 
  • Some areas of extremely dark and persistent shadows (e.g. steep coastal cliffs across southern Australia) may be inadvertently mapped as water, resulting in a landward offset from true coastline positions. 
  • 1991 and 1992 coastlines are currently affected by aerosol-related issues caused by the 1991 Mount Pinatubo eruption. These coastlines should be interpreted with care, particularly across northern Australia. 


The following software was used to generate this product: 

Relevant websites


Bishop-Taylor, R., Sagar, S., Lymburner, L., & Beaman, R. J. (2019a). Between the tides: Modelling the elevation of Australia's exposed intertidal zone at continental scale. Estuarine, Coastal and Shelf Science, 223, 115-128. Available:

Bishop-Taylor, R., Sagar, S., Lymburner, L., Alam, I., & Sixsmith, J. (2019b). Sub-pixel waterline extraction: Characterising accuracy and sensitivity to indices and spectra. Remote Sensing, 11(24), 2984. Available:

Sagar, S., Roberts, D., Bala, B., & Lymburner, L. (2017). Extracting the intertidal extent and topography of the Australian coastline from a 28 year time series of Landsat observations. Remote Sensing of Environment, 195, 153-169. Available:

Turner, I. L., Harley, M. D., Short, A. D., Simmons, J. A., Bracs, M. A., Phillips, M. S., & Splinter, K. D. (2016). A multi-decade dataset of monthly beach profile surveys and inshore wave forcing at Narrabeen, Australia. Scientific data3(1), 1-13.



Data sources

Processing steps

  1. Load stack of all available Landsat 5, 7 and 8 satellite imagery for a location
  2. Convert satellite observations to a remote sensing water index (MNDWI)
  3. For each satellite image, model ocean tides into a 2 x 2 km grid based on exact time of image acquisition
  4. Interpolate tide heights into spatial extent of image stack
  5. Mask out high and low tide pixels by removing all observations acquired outside of 50 percent of the observed tidal range centered over mean sea level
  6. Combine tidally-masked data into annual median composites from 1988 to the present representing the coastline at approximately mean sea level
  7. Apply morphological extraction algorithms to mask annual median composite rasters to a valid coastal region
  8. Extract waterline vectors using subpixel waterline extraction
  9. Compute rates of coastal change at every 30 m along Australia's non-rocky coastlines using linear regression

Major algorithms

Schema / spatial extent

Digital Earth Australia CoastLines extent

Update frequency Annually
Temporal extent
Min. longitude -4846590.00
Max. longitude -1887450.00
Min. latitude -1015650.00
Max. latitude 2121650.00
Coordinate reference system Australian Albers / GDA94 (EPSG: 3577)
Cell size X 30.00
Cell size Y 30.00

Example images

DEA CoastLines at Busselton, WA
DEA CoastLines at Cape Peron, Rockingham, WA
DEA CoastLines at Port Geographe, WA
DEA CoastLines at Point Stuart, NT
DEA CoastLines rates of change statistics at Perth, WA
DEA CoastLines rates of change statistics at Adelaide, SA
DEA CoastLines rates of change statistics at Long Bay, SA
DEA CoastLines at Corner Inlet, Vic
DEA CoastLines at Dutton Way, Vic
DEA CoastLines at Queenscliff, Vic
DEA CoastLines rates of change statistics at Old Bar, NSW
DEA CoastLines at Tweed Heads, NSW and QLD
DEA CoastLines at Jumpinpin Channel, QLD
DEA CoastLines at Barubbra Island, QLD


Commonwealth of Australia (Geoscience Australia)

Principal contributors

Robbi Bishop-Taylor, Rachel Nanson, Stephen Sagar, Leo Lymburner

Subject matter experts

Robbi Bishop-Taylor


CC BY Attribution 4.0 International License

Rights statement

© Commonwealth of Australia (Geoscience Australia) 2019. Creative Commons Attribution 4.0 International License.


The authors would like to thank the following organisations for providing validation data and valuable feedback on preliminary versions of this product: 

  • Centre for Integrative Ecology, Deakin University 
  • City of Gold Coast, Queensland
  • Department of Transport, Western Australia 
  • Griffith Centre for Coastal Management, Griffith University 
  • Sunshine Coast Council, Queensland 
  • TASMARC Project, Antarctic Climate & Ecosystems Cooperative Research Centre 
  • Water Research Laboratory, University of New South Wales 
  • Colin Woodroffe, University of Wollongong

This research was undertaken with the assistance of resources from the National Computational Infrastructure High Performance Data (HPD) platform, which is supported by the Australian Government.