GA Bare Earth Covariates

Enhanced Barest Earth Landsat Imagery for Soil and Lithological Modelling

Remotely sensed datasets provide fundamental information for understanding the chemical, physical and temporal dynamics of the atmosphere, lithosphere, biosphere and hydrosphere. Satellite remote sensing has been used extensively in mapping the nature and characteristics of the terrestrial land surface, including vegetation, rock, soil and landforms, across global to local-district scales. With the exception of hyper-arid regions, mapping rock and soil from space has been problematic because of vegetation that either masks the underlying substrate or confuses the spectral signatures of geological materials (i.e. diagnostic mineral spectral features), making them difficult to resolve. As part of the Exploring for the Future program, a new barest earth Landsat mosaic of the Australian continent using time-series analysis significantly reduces the influence of vegetation and enhances mapping of soil and exposed rock from space (Roberts et al. 2019). Here, we provide a brief background on geological remote sensing and describe a suite of enhanced images using the barest earth Landsat mosaic for mapping surface mineralogy and geochemistry. These geological enhanced images provide improved inputs for predictive modelling of soil and rock properties over the Australian continent. In one case study, use of these products instead of existing Landsat TM band data to model chromium and sodium distribution using a random forest machine learning algorithm improved model performance by 28–46% (Wilford and Roberts, 2020).

Supported by TERN Landscapes, Geoscience Australia have just finalised the development of a range of new Landsat derived Bare Earth products. All together there are 23 raster at 25m resolution. These rasters have been resampled to fit the spatial support of the TERN Landscapes Digital Soils Modelling covariate stack (approx. 30m resolution) . These national extent rasters have also split into tiles to ease download sizes, as well as being incorporated into the Principal Components raster stack product.

These data sets as well as a range of other relevant DSM covariate datasets are available for download at the TERN Landscape File Download Site - https://drop.ternlandscapes.org.au/#/

More information can be found at;

Roberts D., Wilford J. & Ghattas O., 2019. Exposed soil and mineral map of the Australian continent revealing the land at its barest. Nature Communications 10:5297.

Wilford J and Roberts., 2020. Enhanced barest earth Landsat imagery for soil and lithological modelling, In: Czarnota K., et al. (eds.), Exploring for the Future: extended abstracts, Geoscience Australia, Canberra, 1–6.

Contact : ClientServices@ga.gov.au Geoscience Australia

The products include :

Individual bare earth bands

BLUE
GREEN
RED
NIR
SWIR1
SWIR2

Normalised ratios bands

((RED - BLUE) / (RED + BLUE)) == "-ND-RED-BLUE.tif"
((SWIR1 - NIR) / (SWIR1 + NIR)) == "-ND-SWIR1-NIR.tif"
((SWIR1 - SWIR2) / (SWIR1 + SWIR2)) == "-ND-SWIR1-SWIR2.tif"
((NIR - GREEN) / (NIR + GREEN)) == "-ND-NIR-GREEN.tif"
((SWIR1 - BLUE) / (SWIR1 + BLUE)) == "-ND-SWIR1-BLUE.tif"
((SWIR2 - NIR) / (SWIR2 + NIR)) == "-ND-SWIR2-NIR.tif"
((SWIR2 - RED) / (SWIR2 + RED)) =="-ND-SWIR2-RED.tif"
((RED - GREEN) / (RED + GREEN)) == "-ND-RED-GREEN.tif"
((SWIR2 - GREEN) / (SWIR2 + GREEN)) == "-ND-SWIR2-GREEN.tif"

PCA directed and selected

PC models are trained with ocean models removed.

These PC models are based previous work used to delineate iron oxides and clays (see Wilford and Roberts 2020 for more details and full references). Interpretation of these thematic products may differ in Australia due to differences in surface reflectance and that the eigenvectors are based on statistics for the entire continent.

Ferric PC2 of BLUE and RED (Chavez and Kwarteng 1989) == "-FERRIC-PC2.tif"
Ferric - PC4 of BLUE, RED, NIR, SWIR1 (Loughin, 1991) == "-FERRIC-PC4.tif"
Hydroxyl PC1 of SWIR1/SWIR2 and NIR/RED (Fraser and Green, 1987) == "-HYDROXYL-1-PC1.tif"
Hydroxyl PC2 of SWIR1/SWIR2 and NIR/RED (Fraser and Green, 1987) == HYDROXYL-1-PC2.tif"
Hydroxyl - PC2 of SWIR1, SWIR2 (Chavez and Kwarteng 1989) == "-HYDROXYL-2-PC2.tif"
Hydroxyl - PC3 of BLUE, NIR, SWIR1, SWIR2 (Loughin, 1991) == "-HYDROXYL-3-PC3.tif"
Hydroxyl - PC4 of BLUE, NIR, SWIR1, SWIR2 (Loughin, 1991) == "-HYDROXYL-3-PC4.tif"

Band additions

Carbonate/quartz (non-clays highly reflective) BLUE + SWIR2

These datasets have been split into the following regions (with overlap to insure no edge effects if later processed)

QLD

NSW_VIC_TAS

Page updated

Google Sites

Report abuse