Predicting continental distribution of soil mercury concentration in Australia

Professor Corey J. A. Bradshaw
Global Ecology | Partuyarta Ngadluku Wardli Kuu, Flinders University
e-mail

Team:

• Associate Professor Larissa Schneider, Australian National University
• Adjunct Professor Patrice de Caritat, Curtin University
• Professor Simon Haberle, Australian National University

Aims

• identify external and indirect determinants of mercury (Hg)
• understand environmental conditions that influence mercury retention and mobility
• predict continental distribution of soil mercury

Scripts

• HgGH.R: all required R code combined

Data

Sample point

• geochem.csv: geochemical data
• field.csv: sample point characteristics
• hgTSID.csv: re-analysed [Hg] estimates (ng/g)
• gs.csv: grain-size category percentages

Spatial

Most of these files are too large to store in this repository directly, so in most cases the links refer to the original repository URLs where you can download the datasets.

base

• aus.shp: Australia boundary shapefile (zipped)

land use

• NLUM_v7_250_ALUMV8_2020_21_alb.tif: land use of Australia 2010–11 to 2020–21 (download geotif raster from original site)

vegetation

• wwf_terr_ecos.shp: WWF ecoregions shapefile (download zipped file from original site)
• OzWALD.GPP.AnnualMeans.nc: vegetation carbon uptake (gross primary production) NetCDF (download from original site)
• OzWALD.LAI.AnnualMeans.nc: leaf area index NetCDF (download from original site)

geology & geochemistry

• GeologicUnitPolygons1M.shp: 1:1,000,000 geological unit polygon shapefile (download zipped file from original site)
• lithreclass.csv: reclassified lithology groups text file
• radmap_v4_2019_filtered_ML_KThU_RGB_24bit.tif: potassium:thorium:uranium geotif raster (download data package from original site)
• radmap_v4_2019_filtered_ML_ppmTh_32bitfloat_grid.tif: thorium ppm geotif raster (download data package from original site)
• radmap_v4_2019_filtered_ML_ppmU_32bitfloat_grid.tif: uranium ppm geotif raster (download data package from original site)
• radmap_v4_2019_filtered_ML_pctk_32bitfloat_grid.tif: % potassium geotif raster (download data package from original site)
• NTO_000_005_EV_N_P_AU_NAT_C_20231101.tif: soil nitrogen (0-5 cm) geotif raster (download from original site)
• PTO_000_005_EV_N_P_AU_NAT_C_20231101.tif: soil phosphorus (0-5 cm) geotif raster (download from original site)
• pHc_000_005_EV_N_P_AU_NAT_C_20140801.tif: soil pH (0-5 cm) geotif raster (download from original site)
• CLY_000_005_EV_N_P_AU_TRN_N_20210902.tif: % soil clay content geotif raster (download from original site)
• SLT_000_005_EV_N_P_AU_TRN_N_20210902.tif: % soil silt content geotif raster (download from original site)

water

• OzWALD.annual.Pg.AnnualSums.nc: annual precipitation NetCDF (download from original site)
• PrescottIndex_01_3s_lzw.tif: Prescott index geotif raster (download from original site)
• OzWALD.Ssoil.AnnualMeans.nc: soil water availability NetCDF (download from original site)

Predicted

• HgPredSpatRFlog10.nc: This is the NetCDF file for the output map of Australia-wide Hg concentration (log₁₀ values) predicted from the random forest model (resolution = 0.005° × 0.005° latitude/longitude ≈ 0.55 × 0.55 km ≈ 0.306 km²). Due to Github file-size constraints, we have broken the file into similar-sized chunks, and then compressed them using the fast, lossless compression algorithm zstd. First, decompress each .zst chunk using the following command in Terminal (or equivalent): zstd -d 'HgPredSpatRFlog10.nc_chunk_a*.zst', and then combine chunks a to i using the following Terminal (or equivalent) command: cat HgPredSpatRFlog10.nc_chunk_* > HgPredSpatRFlog10.nc. You can import the NetCDF file in R using the ncdf4 package and its function nc_open. This produces a ncdf4 object that can be converted to a SpatRaster object using the rast function in package terra.
• HgPredSpatRF.nc: This is the NetCDF file for the output map of Australia-wide Hg concentration (back-transformed to linear values) predicted from the random forest model (resolution = 0.005° × 0.005° latitude/longitude ≈ 0.55 × 0.55 km ≈ 0.306 km²). Due to Github file-size constraints, we have broken the file into similar-sized chunks, and then compressed them using the fast, lossless compression algorithm zstd. First, decompress each .zst chunk using the following command in Terminal (or equivalent): zstd -d 'HgPredSpatRF.nc_chunk_a*.zst', and then combine chunks a to i using the following Terminal (or equivalent) command: cat HgPredSpatRF.nc_chunk_* > HgPredSpatRF.nc. You can import the NetCDF file in R using the ncdf4 package and its function nc_open. This produces a ncdf4 object that can be converted to a SpatRaster object using the rast function in package terra.

required R libraries

• ade4, adegraphics, adespatial, boot, dismo, dplyr, fields, gbm, geodata, geosphere, ggplot2, kableExtra, ncdf4, patchwork, pdp, randomForestExplainer, ranger, rnaturalearthdata, sf, sp, spatialRF, spatstat, spdep, terra, tidyverse, usdm