This is Version 2 of the Australian Soil Sand Content product of the Soil and Landscape Grid of Australia.

It supersedes the Release 1 product that can be found at https://doi.org/10.4225/08/546F29646877E

The map gives a modelled estimate of the spatial distribution of sand in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: 20 um - 2 mm mass fraction of the < 2 mm soil material determined using the pipette method Units: %; Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: The approach, based on machine learning, predicts each soil texture fraction at 90 m grid cell resolution, at depths 0–5 cm, 5–15 cm, 15–30 cm, 30–60 cm, 60–100 cm and 100–200 cm. The approach accommodates uncertainty in converting field measurements to quantitative estimates of texture fractions. Existing methods of bootstrap resampling were exploited to predict uncertainties, which are expressed as 90% prediction intervals about the mean prediction at each grid cell. The models and the prediction uncertainties were assessed by an external validation dataset. Results were compared with Version 1 Soil and Landscape Grid of Australia (v1.SLGA) (Viscarra Rossel et al. 2015). All predictive and functional accuracy diagnostics demonstrate improvements compared with v1.SLGA. Improvements were noted for the sand and clay fraction mapping with average improvement of 3% and 2%, respectively, in the RMSE estimates. Marginal improvements were made for the silt fraction mapping, which was relatively difficult to predict. We also made comparisons with recently released World Soil Grid products (v2.WSG) and made similar conclusions.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This is Version 1 of the Australian pH (Water) product of the Soil and Landscape Grid of Australia.

The map gives a modelled estimate of the spatial distribution of soil pH (1:5 soil water solution) in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: pH of a 1:5 soil water solution Units: None; Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: A full description of the methods used to generate this product can be found at - https://aussoilsdsm.esoil.io/slga-version-2-products/soil-ph-15-water

We used a Random Forest model to fit the relationship between measurements and covariates. The Random Forest model uses the bootstrap resampling approach to iteratively develop the relationships between target variable and predictor variables.

Our modelling also included a repeated (n =50) bootstrap resampling approach but was different in that on each iteration the selected data which were also field data had to be converted to a ‘lab’ measurement. This ‘lab’ measurement was derived by drawing a value at random from the empirical distribution corresponding to the field measurement. In this way, we can incorporate into the modelling, the observed variability that is associated with field measurements, which also provides a seamless way to incorporate both data types.

The process of spatial modelling was relatively standard after the data integration step was done. Models were developed for each specified depth interval: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm, 100-200cm. Our investigations also revealed there was some benefit to modelling the Random Forest model residuals using variograms. Together models were evaluated using a data set of size 10000 sites, meaning that the number of cases to evaluate models differed with each depth interval as more cases are found at the surface and near surface and drop off with increasing soil depth. We used the prediction interval coverage probability to assess the veracity of the uncertainty quantifications.

Soil pH mapping was output to the ~90m grid resolution in accordance with SLGA specifications.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This is Version 2 of the Australian soil pH (CaCl2) product of the Soil and Landscape Grid of Australia.

The map gives a modelled estimate of the spatial distribution of the pH of soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

An additional measure of model reliability is through assessment of model extrapolation risk. This measure provides users a spatial depiction where model estimates are made within the domain of the observed data or not.

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: soil pH (CaCl2) Units: pH units; Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 24; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: Release 2 has come about via several mechanism and presents a completely different approach as to how release 1 was developed. Namely:

1. A huge expansion of the available library of data corresponding to each of the main soil state factors has been made possible (Searle et al. 2022). This is through acquisition of new data sets and improvement of others compared with those used for version 1.

2. The incorporation of soil pH data measured using field method (Raupach's Indicator test method) into the modelling system. An empirical transfer function was developed based on measurements with both lab and field observations (52629) to extend to measures where only field data was available. Combining lab and field measures required a special model fitting to account for differing magnitudes of error in the pH data. Lab data was assumed to be error free, however pH and estimated uncertainty could be estimated by the empirical transfer function, then incorporated into the spatial modelling system.

3. Adoption of machine learning to derive empirical relationships between target variable (soil pH) and various data related to the state factors that help determine and control soil variability across landscapes, here the Australian continent and very nearshore islands. While the adoption of ML is not an entirely new advancement, the coupling of it with additional data, and integration of it within a psedo-3D predictive framework permit an improved ability to spatially and vertically characterise soils than Version 1 did.

4. Together with a more powerful and streamlined predictive modelling approach, the quantification of uncertainties draws on the use of the UNEEC (Uncertainty Estimation based on Empirical Errors and Clustering; Shrestha and Solomatine 2006) approach instead of bootstrapping approach so that prediction interval bounds are more custom to the variations in state factor information. Bootstrapping tends to create uniform prediction interval ranges, whereas UNEEC can distinguish areas of relatively lower and higher uncertainties based on differences in soil and landscape characteristics. Therefore, for Version 2, the uncertainties are more custom and tightly defined to the environment they are quantified in.

5. An approach to understand and characterise issues of model extrapolation has been developed. This seeks to highlight areas where there is high confidence that models are going be unreliable, because these areas are outside the range of the underpinning data used in modelling. This issue is addressed via combination of data geometric and distance-based techniques.

The sequence of steps below were carried out to develop the Version 2 products

• Data extraction from SoilDataFederator

• Development of transfer function using data cases with corresponding field and lab information.

• Integration of lab and field data whereby estimates of pH 4B1 from field data are propagated from empirical distributions in order for uncertainty of data is sufficiently handled in later spatial modelling steps.

• Point data intersection with covariates.

• Creation of model and test data sets. Test cases were extracted from datasets for each depth interval.

• Ranger model hyperparameter value optimisation

• Variogram model fitting of ranger model residuals.

• Spatialisation of ranger models and residual kriging models

• Uncertainty analysis with UNEEC method including rudimentary optimisation of class number size.

• Spatialisation of model uncertainties.

• Model evaluations with both test data and against SLGA Version 1 products.

• Delivery of digital soil mapping outputs and computer code to repository.

This is Version 1 of the Australian Available Phosphorus product of the Soil and Landscape Grid of Australia.

The map gives a modelled estimate of the spatial distribution of available phosphorus in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: Available Phosphorus Units: mg/kg; Period (temporal coverage; approximately): 1970-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: This dataset models the spatial distribution of Available Phosphorus using a commonly measured analyte, bicarbonate - extractable phosphorus (Colwell P) (Method 9B1 & 9B2 - Rayment and Lyons 2010). It provides estimates of Colwell P across Australia for each Global Soil Map (GSM) depth interval at a 3 arcsecond resolution (80 - 100m pixel depending on where in Australia). The data is supplied as single band GeoTiff rasters and includes the 5th, 50th and 95th percentile predictions (Based on a 90% confidence interval) for each GSM depth.

Legacy Colwell P data currently stored in government agency soil databases in Australia that are from non-fertilised, non-cropped relatively undisturbed sites is being used to estimate AP. No new P data was collected for this project. Agency data was accessed using the newly developed Soil Data Federator Web API (Searle, pers.coms.). The Cowell P point data was combined with environmental covariates from the TERN national set to build a model of how Cowell P varies across Australia. Covariates were selected that best reflected the geography, geology, and climate of Australia. The model was built using the machine learning algorithm, Random Forests, which is commonly used in digital soil mapping in Australia.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This is version 1 of the Australian15 Bar Lower Limit Volumetric Water Content (L15) product of the Soil and Landscape Grid of Australia.

The map gives a modelled estimate of the spatial distribution of L15 soil hydraulic property in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: 15 Bar Lower Limit Volumetric Water Content Units: percent; Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: A full description of the methods used to generate this product can be found at - https://aussoilsdsm.esoil.io/slga-version-2-products/soil-hydraulic-properties

We employed standard Digital Soil Modelling (DSM) (McBratney et. al., 2002) methods utilising publicly available soil observation data and publicly available environmental covariate data in an environmental correlation approach using machine learning to map the soil properties of volumetric (mm/mm) Drained Upper Limit (DUL) and Soil Lower Limit (L15) across the entire continent at 6 standard depths at 90m pixel resolution.

We used pedotransfer functions for estimating Drained Upper Limit - 1/3 bar (DUL) and Lower Limit - 15 bar (L15) from readily available soil attribute data using data from the National Soil Site Collation (NSSC) (Searle, 2014). Soil property data was obtained using the TERN SoilDataFederator (SDF) (https://aussoilsdsm.esoil.io/site-data/soildatafederator).

The spatial modelling of DUL and L15 is done at six standard depth intervals conforming to the GlobalSoilMap Specifications. (GlobalSoilMap Science Committee, 2015) ie 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. To facilitate modelling at these standard depths the observed data set depths were harmonised to these depths using a mass preserving spline method as described Bishop (1999). A total of 20545 soil profiles were splined in this way and used as inputs to the spatial modelling.

We utilised the publicly available Terrestrial Ecosystem Research Network (TERN) raster covariate stack. It is comprised of 154 individual raster data layers. (https://esoil.io/TERNLandscapes/Public/Pages/COGs/90m_Covariates.html).

The covariate stack was used as the independent variable data for the predictions across all grid cells and at each depth.

The Cubist Machine Learning algorithm (Quinlan, 1992) consisting of 50 bootstrapped model realisations was used to predicted DUL and L15 values (mean of the boostrap realisations) and estimate upper and lower confidence intervals (5% and 95%)

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This is Version 2 of the Australian Soil Depth product of the Soil and Landscape Grid of Australia.

It supersedes the Release 1 product that can be found at https://doi.org/10.4225/08/546F540FE10AA

The map gives a modelled estimate of the spatial distribution of soil depth in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: Depth of soil profile (A & B horizons) Units: metres; Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: Rather than fitting a single model of soil thicknesses we went for a nuanced approach which entailed three separate models for:

Model 1. Predicting the occurrence of rock outcrops.

Model 2. Predicting the thickness of soils within the 0-2m range

Model 3. Predicting the occurrence of deep soils (soils greater than 2m thick)

Models 1 and 3 used the categorical model variant of the Ranger RF which was preceded by distinguishing; for Model 1, the observations that were deemed as rock outcrops from soils. And for Model 3, distinguishing soils that were less than 2m thick (and not rock outcrops) from soils greater than 2m thick. Ultimately both Models 1 and 3 were binary categorical models. 50 repeats of 5-fold CV (cross-validation) iterations of the Ranger RF model were run for each Model variant.

Model 2 used the regression form of the random forest model. After removing from the total data set the observations that were regarded as rock outcrops and soil greater than 2m, there were 111,302 observations available. Of these, 67,698 had explicitly defined soil thickness values. The remaining 43,604 were right-censored data and were treated as follows. For each repeated 5-fold iteration, prior to splitting the data in calibration and validation datasets, values from a beta function were drawn at random of length 43,604. This value (between 0 and 1) was multiplied by the censored value soil thickness and then added to this same value, creating a simulated pseudo-soil thickness. Once the simulated data were combined with actual soil thickness data, the values were square-root transformed to approximate a normal distribution. Ranger RF modelling proceeded after optimising the Hyperparameter settings as described above for the categorical modelling. Like the categorical modelling, 50 repeated 5-fold CV iterations were computed.

All three model approaches were integrated via a simple ‘if-then’ pixel-based procedure. At each pixel, if Model 1 indicated the presence of rock outcrops 45 times or more out of 50 (90% of resampling iterations), the estimated soil thickness was estimated as rock outcrop, or effectively 0cm. Similarly, for Model 3 which was the model based on prediction of deep soils (soils >2m deep). In no situations did we encounter both Models 1 and 3 predict in the positive on 90% or more occasions simultaneously. If Model 1 or 3 did not predict in the positive in 90% of iterations, the prediction outputs of Model 2 were used.

After model integration, we derived a set of soil thickness exceedance probability mapping outputs. These were derived simply by assessing the empirical probabilities (at each pixel) and then tallying the number of occasions the estimated soil depth exceeded given threshold depths of 10cm, 50cm, 100cm, and 150cm. This tallied number was divided by 50 to give an exceedance probability for each threshold depth.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This is Version 1 of the Australian Soil Cation Exchange Capacity product of the Soil and Landscape Grid of Australia.

The map gives a modelled estimate of the spatial distribution of cation exchange capacity in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: Cation Exchange Capacity Units: meq/100g; Period (temporal coverage; approximately): 1970-2022; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: Version 1 Soil and landscape Grid of Australia (Grundy et al. 2015), produced digital mapping of Effective Cation Exchange Capacity (ECEC) which is defined as the total amount of exchangeable bases which are mostly sodium, potassium, calcium and magnesium (collectively termed as bases) in non-acidic soils and bases plus aluminium and hydrogen in acidic soils.

This product, Soil and Landscape Grid National Soil Attribute Maps - Cation Exchange Capacity, described here entails the use of those data pertaining to those data with CEC measurement.

This dataset is made of soil measurements using the following methods as described in Rayment and Lyons (2010): method not recorded (1096), 15A1 (161), 15A2 (365), 15B1 (553), 15B2 (34), 15C1 (3229), 15D1 (265), 15E1 (28), 15K1 (376). The distribution of these sites, colour-coded by each method is shown on Figure 1.

To complement the CEC measurement data, we used data cases (12474) where there is a measured CEC together with soil texture and soil organic carbon co-located measurements. A machine learning pedotransfer function model with these data, together with spatial covariates was used to extend the geographic spread and density of CEC data in order to potentially improve digital soil mapping efforts.

Extensive data processing was involved post data extraction from the SoilDataFederator

Spatial modelling is underpinned by the Cubist (Quinlan 1993) machine learning algorithm.

The spatial modelling integrates both measurement CEC data and CEC data derived by pedotransfer function. The derived CEC have an associated uncertainty and this is incorporated into the spatial model via a simple monte-carlo approach.

The spatial model included a soil depth interval term in order to exploit covariance relationships of soil information within a soil profile. Thus modelling is considered a full soil profile predictive modelling framework.

Prediction uncertainties in this work were done using an approach based on local-errors and clustering (UNEEC) method developed by Shrestha and Solomatine (2006).

Soil maps of predictions and associated uncertainties (expressed as lower and upper prediction limits for 90% confidence) were generated for the following depth intervals: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm, 100-200cm.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/COGs

This is Version 2 of the Australian Available Volumetric Water Capacity (AWC) product of the Soil and Landscape Grid of Australia.

The map gives a modelled estimate of the spatial distribution of AWC soil hydraulic property in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: Available Volumetric Water Capacity Units: percent; Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: A full description of the methods used to generate this product can be found at - https://aussoilsdsm.esoil.io/slga-version-2-products/soil-hydraulic-properties

We employed standard Digital Soil Modelling (DSM) (McBratney et. al., 2002) methods utilising publicly available soil observation data and publicly available environmental covariate data in an environmental correlation approach using machine learning to map the soil properties of volumetric (mm/mm) Drained Upper Limit (DUL) and Soil Lower Limit (L15) across the entire continent at 6 standard depths at 90m pixel resolution.

We used pedotransfer functions for estimating Drained Upper Limit - 1/3 bar (DUL) and Lower Limit - 15 bar (L15) from readily available soil attribute data using data from the National Soil Site Collation (NSSC) (Searle, 2014). Soil property data was obtained using the TERN SoilDataFederator (SDF) (https://aussoilsdsm.esoil.io/site-data/soildatafederator).

The spatial modelling of DUL and L15 is done at six standard depth intervals conforming to the GlobalSoilMap Specifications. (GlobalSoilMap Science Committee, 2015) ie 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. To facilitate modelling at these standard depths the observed data set depths were harmonised to these depths using a mass preserving spline method as described Bishop (1999). A total of 20545 soil profiles were splined in this way and used as inputs to the spatial modelling.

We utilised the publicly available Terrestrial Ecosystem Research Network (TERN) raster covariate stack. It is comprised of 154 individual raster data layers. (https://esoil.io/TERNLandscapes/Public/Pages/COGs/90m_Covariates.html).

The covariate stack was used as the independent variable data for the predictions across all grid cells and at each depth.

Fifty bootstrapped model realisations using the Cubist machine learning algorithm (Quinlan, 1992) were generated and were used to predicted DUL and L15 values (mean of the bootstrap realisations) and estimate upper and lower confidence intervals (5% and 95%) across the entire continent

The Available Water Capacity values were calculated by subtracting the L15 values of each layer from the DUL values of each layer and the upper and lower confidence intervals were estimated by combining the variances of the upper and lower confidence intervals of L15 and DUL.

To estimate the Total Available Volumetric Water Capacity (mm) to 1 and 2 metres we summed all the AWC layer values converted to mm of water to the estimated soil depth (Australia-wide 3D digital soil property maps - Depth of Soil (3 arc second resolution) Version 2) or the designated depth of the product - which ever was shallowest.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This is Version 1 of the Australian Total Soil Nitrogen product of the Soil and Landscape Grid of Australia.

The map gives a modelled estimate of the spatial distribution of total nitrogen in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels). An additional measure of model reliability is through assessment of model extrapolation risk. This measure provides users a spatial depiction where model estimates are made within the domain of the observed data or not.

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: Total soil nitrogen; Units: % (percentage of fine soil mass); Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 24; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: A full description of the methods used to generate this product can be found at -https://aussoilsdsm.esoil.io/slga-version-2-products/total-soil-nitrogen

The first effort to derive national digital soil mapping of total soil nitrogen (expressed as a percentage of fine soil mass) is published and available on the CSIRO Data Access Portal among other places. The present work sort to update this mapping as part of ongoing efforts to expand and improve Australia’s national mapping and characterisation of its soil resources. Collectively these national soil mapping efforts constitute the Soil and Landscape Grid of Australia. The original work has been deemed as Version 1 (completed 2015), while the new work logically is Version 2 (completed 2023). This work has been made possible through support and funding from Australia’s National Collaborative Research Infrastructure Strategy (NCRIS) via the Terrestrial and Ecosystem Research Network

As with the first effort, digital soil mapping is the underpinning framework for the ultimate creation of soil maps in this instance.

As with the other more recent national digital soil mapping efforts, the SoilDataFederator (Searle 2020) has been instrumental in the dynamic collation of disparate soil observational datasets from across the country. These data have been sourced mainly from each State and Territory Government departments tasked with soil survey and collection. Plus there are other data contributions from Universities and to a lessor extent individual research groups. The SoilDataFederator also taps into the larger CSIRO developed Natsoil database (CSIRO 2020) which holds the data related to research projects and field stations that CSIRO has managed.

The improvement in digital soil mapping has come about via several mechanism.

1. A huge expansion of the available library of data corresponding to each of the main soil state factors has been made possible (Searle et al. 2022). This is through acquisition of new data sets and improvement of others compared with those used for version 1.

2. Adoption of machine learning to derive empirical relationships between target variable (total soil nitrogen content) and various data related to the state factors that help determine and control soil variability across landscapes, here the Australian continent and very nearshore islands. While the adoption of ML is not an entirely new advancement, the coupling of it with additional data, and integration of it within a psedo-3D predictive framework permit an improved ability to spatially and vertically characterise soils than Version 1 did.

3. Together with a more powerful and streamlined predictive modelling approach, the quantification of uncertainties draws on the use of the UNEEC (Uncertainty Estimation based on Empirical Errors and Clustering; Shrestha and Solomatine 2006) approach instead of bootstrapping approach so that prediction interval bounds are more custom to the variations in state factor information. Bootstrapping tends to create uniform prediction interval ranges, whereas UNEEC can distinguish areas of relatively lower and higher uncertainties based on differences in soil and landscape characteristics. Therefore, for Version 2, the uncertainties are more custom and tightly defined to the environment they are quantified in.

4. An approach to understand and characterise issues of model extrapolation has been developed. This seeks to highlight areas where there is high confidence that models are going be unreliable, because these areas are outside the range of the underpinning data used in modelling. This issue is addressed via combination of data geometric and distance-based techniques.

References:

CSIRO (2020): CSIRO National Soil Site Database. v9. CSIRO. Data Collection. https://doi.org/10.25919/c4br-0r30.

Searle, Ross (2020): TERN Soil Data Federator. v1. CSIRO. Software Collection. http://hdl.handle.net/102.100.100/480151?index=1

There are over 150 national GeoTIFF rasters representing the SCORPAN factors across climate, parent material, biology, relief, soil and location.

Covariate rasters (stored as Cloud Optimised GeoTIFF files) are available as full national mosaics. The mosaics are available at 90m (3 arcsec) resolution and at 30m (1 arcsec) resolution.

General information about this data set can be found at - https://aussoilsdsm.esoil.io/dsm-covariates/covariate-data

The covariates are also available as Principal Components (PCA). Details of the PCA can be found at - https://aussoilsdsm.esoil.io/dsm-covariates/covariate-data-pca

These DSM covariate datasets are available for download at the TERN Landscape File Download Site - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

A detailed metadata listing for the covariate data is available at - https://shiny.esoil.io/Apps/Covariates/

Lineage: The covariate rasters in this dataset were obtained from a broad range of original data sources. All of these datasets are publicly available. They original data sets were processed to all have the same spatial support for the 90m and 30m stacks respectively.

A range of processing tasks may have been applied to individual rasters such as coordinate projection, resampling and resolution adjustments. All layers in this data set are in Geographic coordinates (EPSG:4326)

Information about the original datasets can be found in the individual metadata records - https://shiny.esoil.io/Apps/Covariates/

This is Version 2 of the Australian Soil Clay Content product of the Soil and Landscape Grid of Australia.

It supersedes the Release 1 product that can be found at https://doi.org/10.4225/08/546EEE35164BF

The map gives a modelled estimate of the spatial distribution of clay in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: 2 μm mass fraction of the less than 2 mm soil material determined using the pipette method; Units: %; Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: The approach, based on machine learning, predicts each soil texture fraction at 90 m grid cell resolution, at depths 0–5 cm, 5–15 cm, 15–30 cm, 30–60 cm, 60–100 cm and 100–200 cm. The approach accommodates uncertainty in converting field measurements to quantitative estimates of texture fractions. Existing methods of bootstrap resampling were exploited to predict uncertainties, which are expressed as 90% prediction intervals about the mean prediction at each grid cell. The models and the prediction uncertainties were assessed by an external validation dataset. Results were compared with Version 1 Soil and Landscape Grid of Australia (v1.SLGA) (Viscarra Rossel et al. 2015). All predictive and functional accuracy diagnostics demonstrate improvements compared with v1.SLGA. Improvements were noted for the sand and clay fraction mapping with average improvement of 3% and 2%, respectively, in the RMSE estimates. Marginal improvements were made for the silt fraction mapping, which was relatively difficult to predict. We also made comparisons with recently released World Soil Grid products (v2.WSG) and made similar conclusions.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

We used Digital Soil Mapping (DSM) technologies combined with collations of observed soil colour data from TERN's Soil Data Federation System, to produce surface and subsoil maps of soil colour at a 90m resolution.

The map gives an estimate of the spatial distribution of RGB soil colour across Australia.

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Period (temporal coverage; approximately): 1950-2020; Spatial resolution: 3 arc seconds (approx 90m); Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: The map was produced as per methods described at - https://aussoilsdsm.esoil.io/slga-version-2-products/soil-colour

Soil colour is arguably one of the most obvious and easily observed soil morphological characteristics. Soil scientists use soil colour to differentiate genetic soil horizons as well as for the classification of soil types, e.g. The Australian Soil Classification.

In Australia, prior work of mapping the colour of Australian soils was performed by Viscarra Rossel et al. (2010), but was limited to just surface soils, output mapping to 5km spatial resolution, and only utilised a relatively small collection of vis-NIR spectra (from which colour was inferred) to develop spatial soil colour models.

From data discovery via the Australian Soil Data Federator, we were able to compile over 300 000 soil colour field observations (dry soil condition) collected across Australia. About 160 000 were for topsoils, while about 140 000 were for subsoils. Rather than exclusively using vis-NIR spectra, a logical line of investigation is to exploit the availability of a comparatively larger field observed dataset.

Colour Space Conversions

Field classification of soil colours are near exclusively recorded using the Munsell HVC (Hue, Value, Chroma) colour system. Munsell HVC soil colour descriptions are not conducive for quantitative studies (Robertson 1977). Using a lookup table, we performed a conversion from the Munsell HVC colour space to the CIELAB colour space. The CIELAB colour space can describe any uniform colour space by the three variables: L*, a*, and b*. Each variable represents the lightness of the colour (L* = 0 yields black and L* = 100 indicates diffuse white), its position between red/magenta and green (a*, negative values indicate green while positive values indicate magenta) and its position between yellow and blue (b*, negative values indicate blue and positive values indicate yellow).

Digital soil mapping

Random Forest machine learning was used to independently model L*, a*, and b* target variables as a function of a suite of available national extent environmental covariates. While we did investigate various options for combined target variable modelling given the covarying relationships of the colour variables, neither were able to match the prediction skill of the independently treated approach. The L* variable was modelled as a categorical variable, both a*, and b* were modelled as continuous variables. For both top- and subsoil models, a dataset (n=10000) was selected out of each of the available datasets prior to any modelling for the sole purpose of evaluating the goodness of fit of the fitted models, akin to an out-of-bag model evaluation.

After modelling, the combined L*, a*, and b* were post-processed to line up the nearest HVC colour space chip using Euclidean distance quantification.

For colour visualisation of the soil colour maps, predictions were transformed to the RGB colour space using the same lookup table as for the conversion form Munsell HVC to CIELAB.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This is Version1 of the Australian Soil Organic Carbon product of the Soil and Landscape Grid of Australia at 30m resolution.

The map gives a modelled estimate of the spatial distribution of total organic carbon in soils across Australia.

It supersedes the Release 1 product that can be found at https://doi.org/10.4225/08/547523BB0801A

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 1 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: Mass fraction of carbon by weight in the < 2 mm soil material as determined by dry combustion at 900 Celcius Units: %; Period (temporal coverage; approximately): 1970-2021; Spatial resolution: 1 arc seconds (approx 30m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: Data on total organic carbon (TOC) concentration (%) was extracted with the Soil Data Federator (https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederatorHelp.html) managed by CSIRO. The Soil Data Federator is a web API that compiles soil data from different institutions and government agencies throughout Australia. The laboratory methods for total organic carbon included in the study are 6A1, 6A1_UC, 6B2, 6B2b, 6B3, 6B3a. We selected TOC data from the period 1970-2020 to get a compromise between representativity of current TOC concentration and spatial coverage. The data was cleaned and processed to harmonize units, exclude duplicates and potentially wrong data entries (e.g. missing upper or lower horizon depths, extreme TOC values, unknown sampling date). Additional TOC measurements from the Biome of Australian Soil Environments (BASE) contextual data (Bisset et al., 2016) were also included in the analyses. TOC concentration for BASE samples was determined by the Walkley-Black method (method 6A1). Upper limits for TOC concentration by biome and land cover classes were set according to published literature, consistent datasets (Australian national Soil Carbon Research Program (SCaRP) and BASE, and data exploration to exclude unrealistic TOC values (e.g. maximum TOC = 30% in temperate forests, maximum TOC = 14% in temperate rainfed pasture). Since TOC concentration in Australian ecosystems has been underestimated by previous SOC maps, we did not set conservative TOC upper limits, knowing that machine learning model would likely underestimate high SOC values.

The equal-area quadratic spline function were fitted to the whole collection of pre-processed TOC data, and then values extracted for the 0-5 cm, 5-15 cm, 15-30 cm, 30-60 cm, 60-100 cm, and 100-200 cm depth intervals, following GlobalSoilMap specifications (Arrouays et al., 2014}.

Covariates: We collected a set of 57 spatially exhaustive environmental covariates covering Australia and representing proxies for factors influencing SOC formation and spatial distribution: soil properties, climate, organisms/vegetation, relief and parent material/age. The covariates were reprojected to WGS84 (EPSG:4326) projection and cropped to the same spatial extent. All covariates were resampled using billinear interpolation or aggregated to conform with a spatial resolution with grid cell of 30 m x 30 m.

Mapping: The spatial distribution of soil TOC concentration is driven by the combined influence of climate, vegetation, relief and parent materials. We thus modelled TOC concentration as a function of environmental covariates representing biotic and abiotic control of TOC. The measurement of SOC and their corresponding value of environmental covariate at same measurement locations were used to fit the mapping model.

Mapping is made with Quantile regression forest, which is similar to the popular random forest algorithm for mapping. Instead of obtaining a single statistic, that is the mean prediction from the decision trees in the random forest, we report all the target values of the leaf node of the decision trees. With QRF, the prediction is thus not a single value but a cumulative distribution of the TOC prediction at each location, which can be used to compute empirical quantile estimates.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This is Version 2 of the Australian Soil Silt Content product of the Soil and Landscape Grid of Australia.

It supersedes the Release 1 product that can be found at https://doi.org/10.4225/08/546F48D6A6D48

The map gives a modelled estimate of the spatial distribution of silt in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: 2-20 um mass fraction of the < 2 mm soil material determined using the pipette method Units: %; Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: The approach, based on machine learning, predicts each soil texture fraction at 90 m grid cell resolution, at depths 0–5 cm, 5–15 cm, 15–30 cm, 30–60 cm, 60–100 cm and 100–200 cm. The approach accommodates uncertainty in converting field measurements to quantitative estimates of texture fractions. Existing methods of bootstrap resampling were exploited to predict uncertainties, which are expressed as 90% prediction intervals about the mean prediction at each grid cell. The models and the prediction uncertainties were assessed by an external validation dataset. Results were compared with Version 1 Soil and Landscape Grid of Australia (v1.SLGA) (Viscarra Rossel et al. 2015). All predictive and functional accuracy diagnostics demonstrate improvements compared with v1.SLGA. Improvements were noted for the sand and clay fraction mapping with average improvement of 3% and 2%, respectively, in the RMSE estimates. Marginal improvements were made for the silt fraction mapping, which was relatively difficult to predict. We also made comparisons with recently released World Soil Grid products (v2.WSG) and made similar conclusions.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

The MCCN project is to deliver tools to assist the agricultural sector to understand crop-environment relationships, specifically by facilitating generation of data cubes for spatiotemporal data. This repository contains Jupyter notebooks to demonstrate the functionality of the MCCN data cube components.The dataset contains input files for the case study (source_data), RO-Crate metadata (ro-crate-metadata.json), results from the case study (results), and Jupyter Notebook (MCCN-CASE 3.ipynb)Research Activity Identifier (RAiD)RAiD: https://doi.org/10.26292/8679d473Case StudiesThis repository contains code and sample data for the following case studies. Note that the analyses here are to demonstrate the software and result should not be considered scientifically or statistically meaningful. No effort has been made to address bias in samples, and sample data may not be available at sufficient density to warrant analysis. All case studies end with generation of an RO-Crate data package including the source data, the notebook and generated outputs, including netcdf exports of the datacubes themselves.Case Study 3 - Select optimal survey localityGiven a set of existing survey locations across a variable landscape, determine the optimal site to add to increase the range of surveyed environments. This study demonstrates: 1) Loading heterogeneous data sources into a cube, and 2) Analysis and visualisation using numpy and matplotlib.Data SourcesThe primary goal for this case study is to demonstrate being able to import a set of environmental values for different sites and then use these to identify a subset that maximises spread across the various environmental dimensions.This is a simple implementation that uses four environmental attributes imported for all Australia (or a subset like NSW) at a moderate grid scale:Digital soil maps for key soil properties over New South Wales, version 2.0 - SEED - see https://esoil.io/TERNLandscapes/Public/Pages/SLGA/ProductDetails-SoilAttributes.htmlANUCLIM Annual Mean Rainfall raster layer - SEED - see https://datasets.seed.nsw.gov.au/dataset/anuclim-annual-mean-rainfall-raster-layerANUCLIM Annual Mean Temperature raster layer - SEED - see https://datasets.seed.nsw.gov.au/dataset/anuclim-annual-mean-temperature-raster-layerDependenciesThis notebook requires Python 3.10 or higherInstall relevant Python libraries with: pip install mccn-engine rocrateInstalling mccn-engine will install other dependenciesOverviewGenerate STAC metadata for layers from predefined configuratiionLoad data cube and exclude nodata valuesScale all variables to a 0.0-1.0 rangeSelect four layers for comparison (soil organic carbon 0-30 cm, soil pH 0-30 cm, mean annual rainfall, mean annual temperature)Select 10 random points within NSWGenerate 10 new layers representing standardised environmental distance between one of the selected points and all other points in NSWFor every point in NSW, find the lowest environmental distance to any of the selected pointsSelect the point in NSW that has the highest value for the lowest environmental distance to any selected point - this is the most different pointClean up and save results to RO-Crate

This is Version 1 of the Soil Coarse Fragments product of the Soil and Landscape Grid of Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. This product contains six digital soil attribute maps for each of three depth intervals, 0-5cm, 5-15cm, 15-30cm These depths are consistent with the specifications of the GlobalSoilMap.net project (http://www.globalsoilmap.net/). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

These maps are generated using Digital Soil Mapping methods

Attribute Definition: Soil Coarse Fragments Class Probabilities as defined in the Australian Soil and Land Survey Field Handbook Units: Probability of CF class occurring; Period (temporal coverage; approximately): 1950-2022; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Total size before compression: about 8GB; Total size after compression: about 4GB; Data license : Creative Commons Attribution 4.0 (CC BY); Format: Cloud Optimised GeoTIFF.

Lineage: Data on the abundance of coarse fragments (particles > 2 mm) and gravimetric content (% weight) were extracted with using the the Terrestrial Ecosystem Research Network (TERN) Soil Data Federator

(https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html)

managed by CSIRO (Searle et al., 2021). The Soil Data Federator is a web API that compiles soil data from different institutions and government agencies throughout Australia. The abundance (% volume) is assessed visually in the field as part of the soil profile description using standards described in the Australian Soil and Land Survey field Handbook (National Committee on Soils and Terrain , 2009). The abundance of rock fragments per soil horizon on the cut surface of the soil profile surface of the soil horizon occupied by coarse fragments was grouped into six categories: very few (0-2 %), few (2-10 %), common (10-20 %), many (20-50 %), abundant (50-90 %) and very abundant (> 90%). The gravimetric content (% mass) is measured in the laboratory as percent mass of coarse fragments (particles > 2 mm) from the whole soil. Here, we take the profile surface abundance of coarse fragments as a proxy for volumetric coarse fragments (CFVol). The data was cleaned and processed to exclude duplicates and wrong data entries (e.g., missing values). The observations of CFVol (%) were converted into GlobalSoilMap depth intervals with the slab function of the aqp R package (Beaudette et al., 2021), assigning the most probable class to each depth interval. The gravimetric coarse fragments were also standardized to the GlobalSoilMap depth intervals with equal-area quadratic splines (Bishop et al., 1999). Observations of gravimetric coarse fragment content (〖CF〗_Weight) were transformed into volumetric with the equation:

〖CF〗_Vol (%)=〖Vol〗_CF/〖Vol〗_WhSoil (〖Weig ht〗_CF / ρ_CF)/(〖Weight〗_WhSoil /〖 ρ〗_WhSoil )=(〖CF〗_Weight×ρ_WhSoil)/ρ_CF ,

Where where ρ_WhSoil is the bulk density prediction for bulk soil from SLGA (Viscarra Rossel et al., 2014), ρ_CF is assumed to be 2.65 g cm-3 (Hurlbut and Klein (1977) in Mckenzie et al. (2002) and 〖CF〗_Vol is the volumetric coarse fragment content (continuous),which was assigned to the corresponding class. This resulted in CFVol observations for 110,308 locations.

Mapping was produces using quantile regression forest fitted with the observed coarse fragments class data and a large set of environmental variables as predictors.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This is Version 2 of the Australian Soil Bulk Density - Whole Earth product of the Soil and Landscape Grid of Australia.

It supersedes the Release 1 product that can be found at https://doi.org/10.4225/08/546EE212B0048

The map gives a modelled estimate of the spatial distribution of Bulk Density in soils across Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (https://esoil.io/TERNLandscapes/Public/Pages/SLGA/Resources/GlobalSoilMap_specifications_december_2015_2.pdf). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

Detailed information about the Soil and Landscape Grid of Australia can be found at - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/index.html

Attribute Definition: Bulk Density of the whole soil (including coarse fragments) in mass per unit volume by a method equivalent to the core method; Units: g/cm3; Period (temporal coverage; approximately): 1950-2021; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Data license : Creative Commons Attribution 4.0 (CC BY); Target data standard: GlobalSoilMap specifications; Format: Cloud Optimised GeoTIFF; Lineage: An attempt was made to update digital soil mapping of whole soil bulk density for Australia. This was an update of first attempt by Viscarra Rossel et al. (2014). Based on model evaluations using a dataset not included in any modelling, the updated version (2nd Version) represents a demonstrable improvement on the 1st version.

Since the first version, more measured site data has been made available and retrievable via the Australian SoilDataFederator. In 2014 there were 3776 sites with measured whole soil bulk density. For the new update, 6116 sites had measured data. Because of usually strong empirical relationships between bulk density, soil texture and soil carbon, the use of pedotransfer functions (to predict bulk density from soil texture and soil carbon) was performed with the intention of increasing data density and spatial coverage of data that would ultimately improve digital soil mapping prediction skill. This added a further 15735 sites after building a spatial pedotransfer function using a dataset of 12308 cases (3939 sites with bulk density, soil carbon and soil texture data).

The basic steps of the work entailed.

Use soil data federator to get pertinent soils observation data

Develop spatial pedotransfer function prediction whole soil bulk density using soil carbon and texture data.

Compile measured and inferred whole soil bulk density data (86306 cases), then setting aside a dataset of 7500 cases for external model evaluation.

Predictive models using random forest algorithm with 78806 data cases fitted. To account for uncertainties in pedotransfer function inferred data, Monte Carlo simulations were performed from the pedotransfer function model. Simulation was repeated 100 times.

Predictive model uncertainties quantified using UNEEC approach (Uncertainty Estimation based on local errors and Clustering).

Quantification of model extension limits derived using hybrid method involving multivariate convex hull analysis and count of observations.

Digital soil maps with quantified uncertainties (5th and 95th prediction interval limits) and assessment of model extrapolation risk were produced at 90m resolution for the following depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm, 100-200cm.

All processing for the generation of these products was undertaken using the R programming language. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html

This dataset contains soil classification and property data for a number of different sites around the Council, Alaska vicinity. These data consist of jpg files containing imagery and data tables and are previewable via the link at the top of the page.

Comprehensive dataset of 1,609 Soil testing services in United States as of July, 2025. Includes verified contact information (email, phone), geocoded addresses, customer ratings, reviews, business categories, and operational details. Perfect for market research, lead generation, competitive analysis, and business intelligence. Download a complimentary sample to evaluate data quality and completeness.

This is Version 1 of the Soil Organic Carbon Fractions product of the Soil and Landscape Grid of Australia.

The Soil and Landscape Grid of Australia has produced a range of digital soil attribute products. This product contains six digital soil attribute maps for each of three depth intervals, 0-5cm, 5-15cm, 15-30cm These depths are consistent with the specifications of the GlobalSoilMap.net project (http://www.globalsoilmap.net/). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels).

These maps are generated using Digital Soil Mapping methods

Attribute Definition: Soil Organic Carbon Fractions :- mineral-associated organic carbon (MAOC), particulate organic carbon (POC) and pyrogenic organic carbon (PyOC) Units: Various; Period (temporal coverage; approximately): 1950-2022; Spatial resolution: 3 arc seconds (approx 90m); Total number of gridded maps for this attribute: 18; Number of pixels with coverage per layer: 2007M (49200 * 40800); Total size before compression: about 8GB; Total size after compression: about 4GB; Data license : Creative Commons Attribution 4.0 (CC BY); Format: Cloud Optimised GeoTIFF.

Lineage: Soil organic carbon (SOC) is the largest terrestrial carbon pool. SOC is composed of a continuum set of compounds with different chemical composition, origin and susceptibilities to decomposition, that are commonly separated into pools characterised by different responses to anthropogenic and environmental disturbance. Here we map the contribution of three SOC fractions to the total SOC content of Australia’s soils.

The three SOC fractions: mineral-associated organic carbon (MAOC), particulate organic carbon (POC) and pyrogenic organic carbon (PyOC), represent SOC composition with distinct turnover rates, chemistry, and pathway formation. Data for MAOC, POC, and PyOC were obtained with near- and mid-infrared spectral models calibrated with measured SOC fractions. We transformed the data using an isometric log-ratio transformation (ilr) to account for the closed compositional nature of SOC fractions. The resulting , back-transformed ilr components were mapped across Australia.

SOC fraction stocks for the 0-30 cm were derived with maps of total organic carbon concentration, bulk density, coarse fragments and soil thickness. Mapping was done by quantile regression forest fitted with the ilr transformed data and a large set of environmental variables as predictors.

The resulting maps along with the quantified uncertainty show the unique spatial pattern of SOC fractions in Australia. MAOC dominated the total SOC with an average of 59% ±17.5%, whereas 28% ± 17.5% was PyOC and 13% ± 11.1% was POC. The allocation of TOC into the MAOC fractions increased with depth. SOC vulnerability (i.e., POC/[MAOC + PyOC]) was greater in areas with Mediterranean and temperate climate. TOC and the distribution among fractions were the most influential variables on SOC fraction uncertainty. Further, the diversity of climatic and pedological conditions suggests that different mechanisms will control SOC stabilisation and dynamics across the continent, as shown by the model covariates importance metric. We estimated the total SOC stocks (0-30 cm) to be 12.7 Pg MAOC, 2 Pg POC and 5.1 Pg PyOC, which is consistent with previous estimates. The maps of SOC fractions and their stocks can be used for modelling SOC dynamics and forecasting changes in SOC stocks as response to land use change, management, and climate change.

Code - https://github.com/AusSoilsDSM/SLGA Observation data - https://esoil.io/TERNLandscapes/Public/Pages/SoilDataFederator/SoilDataFederator.html Covariate rasters - https://esoil.io/TERNLandscapes/Public/Pages/SLGA/GetData-COGSDataStore.html