Environmental covariates stack for biodiversity and ecosystem modelling at 90m and 250m for continental Australia

Full description

These stacks of 125 environmental covariates are a subset of those compiled by TERN, with the addition of new covariates developed by TERN, Geoscience Australia and CSIRO. These data were pre-processed and standardised for use in the Habitat Condition Assessment System version 3, at 250m and 90m grid resolutions. The covariates are equally applicable to biodiversity and ecosystem modelling where the conceptual framework requires that the signature of anthropogenic land use be minimised. That is, the environmental covariates represent, as far as possible, the equilibrium state of the environment prior to industrialisation of the landscape following European colonisation of the Australian continent (c.1750). While the temporal window associated with source data used in developing many of these covariates varies, each covariate is intended to represent the long-term (multi-decadal) equilibrium state, approximating pre-1750. The careful selection of covariates that meet this conceptual framework requirement of 'non-anthropogenic' distinguishes this covariate stack from other compilations.

The 250 m dataset is published in a geographic reference system (0.0025 degrees of latitude and longitude) in GDA94 (EPSG: 4283). The 90 m dataset is published in a projected reference system (Australian Albers in GDA94; EPSG: 3577). Each dataset is published as a cloud-optimised GeoTIFF (COG) in its original units and as the z-score standardised version, for the convenience of different applications. The z-score parameters for each covariate are included in the collection.

Dataset names do not change, only the folder in which they are contained distinguishes the resolution and whether the data are standardised or not.

Each dataset is also published with the expanded coastline and hole-filled, as described in Liu et al. (2025) to ensure all continental data pixels have a value. This ensures flexibility for other uses in making decisions about where to set the coastline. The original data compiled by TERN (Searle et al., 2022) shows where data holes were filled, as this varied between datasets. The input and output mask datasets used in developing the 250m and 90m versions of HCAS are therefore also supplied to facilitate use in different applications.

All compiled covariates are potentially relevant in a range of ecosystem and biodiversity modelling applications, including HCAS, and represent a subset of the hundreds of covariates in the original TERN compilation (Searle et al., 2022). Descriptions and citations associated with each dataset were reviewed, checked against sources, and updated for accuracy. The updated descriptions and source citations are provided for the 125 datasets included in this collection.

For method details, see Liu et al. (2025): Liu N, Williams KJ, Searle R, Wilford J, Botha E, Johnson S, Read A, Valavi R, Lehmann E, Giljohann KM and Joehnk K (2025) Environmental covariates for predicting the reference state of ecosystems in the Habitat Condition Assessment System (HCAS). Technical Report EP2025-2352. CSIRO, Canberra, Australia.
Lineage: This compiled set of 125 environmental covariates at two output resolutions (90m and 250m) derives from a range of sources, including new developments described in the report by Liu et al. (2025): Liu N, Williams KJ, Searle R, Wilford J, Botha EJ, Johnson S, Read A, Valavi R, Lehmann E, Giljohann KM and Joehnk K (2025) Environmental covariates for predicting the reference state of ecosystems in the Habitat Condition Assessment System (HCAS). Technical Report EP2025-2352. CSIRO, Canberra, Australia.

The hundreds of options in the 90m environmental covariates stack compiled by TERN for digital soil modelling stack were reviewed to determine which variables minimised the anthropogenic signature of land use. The soil attributes modelled by TERN typically include remote sensing variables as biotic predictors to represent the range of ecosystems around Australia that influence soil formation. The use of remote sensing results in an anthropogenic signature of land use change. Therefore, TERN remodelled a subset of soil attributes using only environmental covariates (Searle 2023). This resulted in a new suite of eleven soil attributes described as ‘HCAS-optimised’. TERN further aggregated these attributes across the multiple depth predictions into topsoil and subsoil surfaces, and artificially filled data holes and slightly expanded coastlines, resulting in 22 covariates.

Geoscience Australia developed predictions of rock and regolith chemistry such as conductivity and mineral oxides using predictors derived from the TERN 90m stack of environmental covariates plus additional custom lithology covariates. These also included remote sensing, such as the barest earth Landsat products, among other predictor variables. Therefore, Geoscience Australia remodelled the two depth-conductivity estimates and ten mineral oxides using environmental predictors that minimised the anthropogenic signature of land use, to enable those outputs to be used in HCAS.

The analysis extent for HCAS included the full range of intertidal and coastline dynamics between 1987 to 2016, buffered by a further 3km. The purpose of this spatial extent is to minimise edge effects associated with year-to-year variability in land extent due to changing coastlines. All environmental covariates were expanded to include this extent using nearest neighbour.

Additional environmental covariates were developed to improve characterisation of wetlands, surface water, snow and coastal dynamics relevant to improvements being sought for the Habitat Condition Assessment System (HCAS). This resulted in an additional 13 covariates. Details are provided in the technical report Liu et al. (2025): Liu N, Williams KJ, Searle R, Wilford J, Botha E, Johnson S, Read A, Valavi R, Lehmann E, Giljohann KM and Joehnk K (2025) Environmental covariates for predicting the reference state of ecosystems in the Habitat Condition Assessment System (HCAS). Technical Report EP2025-2352. CSIRO, Canberra, Australia.