Soil carbon-13 isoscape for continental Australia

Full description

This collection comprises a soil δ¹³C isoscape for continental Australia and the underlying soil stable isotope observations used in its development.

Specifically, the collection comprises:

• Soil δ¹³C README, observations and metadata (.xlsx): An Excel workbook (soil_d13C_observations_metadata.xlsx) containing 364 sample measurements of soil organic carbon δ¹³C (‰), sample coordinates, sample source, and other associated metadata. The workbook includes separate worksheets containing a README, data dictionary (Fields), site and sample metadata (Samples), and soil δ¹³C measurements (Measurements).

• Soil δ¹³C observations, flattened dataset (.csv): A tab-delimited flat file (soil_d13C_observations_flatfile.csv) containing 364 sample measurements of soil organic carbon δ¹³C (‰), sample coordinates, sample source, and associated metadata. Data has been combined in this file into a single table to facilitate import and analysis in statistical software.

• Continental soil δ¹³C isoscape (GeoTIFF): A 90 m resolution raster (EPSG:3577) providing spatial predictions of soil organic carbon-13 isotope values (δ¹³C) across continental Australia (australia_soil_d13C_isoscape_90m_v1.tif).

• Prediction uncertainty layer (GeoTIFF): A 90 m resolution raster (EPSG:3577) providing a spatially explicit uncertainty surface derived from Quantile Regression Forest modelling, expressed as the width of the 95% prediction interval (australia_soil_d13C_uncertainty_90m_v1.tif).

This collection provides the first nationally consistent soil δ¹³C isoscape for Australia and can be used to support research into carbon cycling, ecosystem function, vegetation dynamics, stable isotope ecology, land-use change, and environmental modelling. The dataset also provides a baseline resource for future investigations of spatial and temporal variation in soil carbon isotope composition.

A manuscript describing the sampling design, laboratory methods, modelling framework and validation procedures in detail, "Continental-scale controls on soil δ¹³C reveal the importance of vegetation, microbial communities, and climate" (Munroe and Welti), is currently under peer review.

Lineage: The data in this collection were generated as part of a study to develop the first continental-scale soil δ¹³C isoscape for Australia. Soil samples were sourced from two nationally significant soil monitoring programs: the Terrestrial Ecosystem Research Network (TERN) and the Soil Carbon Research Program (SCaRP). Samples were collected between 2009 and 2019 from natural and managed ecosystems spanning the major climatic and biogeographic regions of Australia.

For TERN sites, soil samples were collected from permanent ecosystem monitoring plots following established AusPlots field protocols (White et al. 2012). A spatially and environmentally representative subset of sites was selected for stable isotope analysis using biodiversity-based optimisation methods (Martín-Forés et al. 2021, Munroe et al. 2021). Additional samples were sourced from the SCaRP soil archive to improve representation of agricultural and production landscapes (Baldock et al. 2013, Karunaratne et al. 2025). Exact geographic coordinates are not provided for all SCaRP sampling locations. These samples were collected from privately owned land under agreements with landholders that restricted the public release of site coordinates. Consequently, location information for some SCaRP samples have been withheld to protect landholder confidentiality.

Soil samples were prepared and analysed for stable carbon isotope composition (δ¹³C) using isotope ratio mass spectrometry at the CSIRO Analytical Services Unit (ASU) in Adelaide, South Australia. A diffuse reflectance mid-infrared spectroscopy with multivariate partial least-squares regression analysis (MIR/PLSR) based model was used to screen all TERN samples for carbonate concentrations. Samples containing carbonate contamination were identified and excluded from model development. Stable carbon isotope values are reported relative to the Vienna Pee Dee Belemnite (VPDB) standard and expressed in per mil (‰).

Environmental predictor variables describing climate, vegetation, soil properties and fungal and microbial diversity were assembled from nationally consistent spatial datasets and resampled to a common 90 m spatial resolution. A Random Forest modelling framework incorporating spatial cross-validation was used to predict soil δ¹³C values across continental Australia and identify the dominant environmental drivers of spatial variation. Model uncertainty was quantified using Quantile Regression Forests, and uncertainty is provided as the width of the 95% prediction interval.