The Geostatistical Analyst extension of ArcGIS (ESRI® ArcGIS 10.5) was used to explore, model and generate surfaces for selected sediment geochemistry data. For interpolating the element Aluminium data points, an Empirical Bayesian Kriging model was applied to derive the predicted and the predicted standard error maps. Predicted errors for the model were: Regression function: ‎0.25 * x + 15552.69; Prediction Errors: Mean, 104.186; Root-Mean-Square, 14546.21; Mean Standardised, -0.0182; Root-Mean-Square Standardised, 1.072; and Average Standard Error, 14839.97.

The habitat mapping program was made possible through offset funds provided by INPEX-operated Ichthys LNG Project to Northern Territory Government Department of Environment and Natural Resources, and co-investment from Geoscience Australia and Australian Institute of Marine Science. The intent of this four year (2014-2018) program is to improve knowledge of the marine environments in the Darwin and Bynoe Harbour regions by collating and collecting baseline data that enable the creation of thematic habitat maps that underpin marine resource management decisions. The specific objectives of the survey were to: 1. Obtain high resolution geophysical (bathymetry) data for outer Darwin Harbour, including Shoal Bay; 2. Characterise substrates (acoustic backscatter properties, grainsize, sediment chemistry) for outer Darwin Harbour, including Shoal Bay; and 3. Collect tidal data for the survey area. Data acquired during the survey included: multibeam sonar bathymetry and acoustic backscatter; physical samples of seabed sediments, underwater photography and video of grab sample locations and oceanographic information including tidal data and sound velocity profiles.

A number of surveys were conducted (GA0351/SOL6187; GA4452/SOL6432, GA0358 and GA039) between 2014 and 2018 to sediment samples. The sediment dataset comprises total sediment metabolism, carbonate and element concentrations and C and N isotopes measurements. Detailed accounts of the surveys and methods are provided in Post Survey Reports ( https://pid.geoscience.gov.au/dataset/ga/89978; https://pid.geoscience.gov.au/dataset/ga/100945; and https://pid.geoscience.gov.au/dataset/ga/126219). Bottom sediments were collected using a Smith McIntyre Grab. The surface sediments (0-2 cm) within each grab were spooned into falcon vials and the pore waters were removed by centrifugation. Pore waters were removed within 20 minutes of collection, and salinity, temperature and pH measurements were taken. The pore waters were then filtered (0.45 µm) into 3 ml gas-tight vials (that were pre-charged with 0.025 HgCl2). The procedure was repeated on pore waters from a second bulk sample that was incubated for ~24 hrs at sea surface temperatures. The samples were then frozen for transport to the laboratories at Geoscience Australia (GA) where they were: (1) subsampled for specific surface area analysis; and (2) freeze-dried and ground in a tungsten carbide mill. The dried residue was divided into fractions for: (i) major and trace element analysis; (ii) bulk carbonate analyses; and (ii) total organic carbon and total nitrogen concentrations and isotopic ratios (after de-carbonation). Dissolved inorganic carbon (DIC) concentrations were determined on pore water samples using a DIC analyser and infrared-based CO2 detector (Geoscience Australia). CO2 production rates were calculated by concentration differences over the incubation period, after correction for CaCO3 fluxes. The average %RSD of the precisions and accuracies of the dissolved inorganic carbon measurements were 0.2. The accuracy of the wet/dry weight used in the calculations were better than 1%. Bulk carbonate was determined on dried ground sediment using the carbonate bomb method. The average precision (%RSD +/- s.d.) for 11 samples run in duplicate was 1.6 +/- 1.7%. Specific surface areas were determined using a 5-point Brunauer-Emmett-Teller (BET) adsorption isotherm on a Quantachrome NOVA 2200e analyser, with nitrogen used as the adsorbate. The samples were first cleaned of organic matter by slow heating for 12 hours to 350oC. Major and trace elements were determined in GA laboratories using X-Ray fluorescence (XRF; Phillips PW204 4kW sequential spectrometer) and ICP-MS (AGILENT 7500ce). Two Certified Reference Materials (CRMs) called CH-1 (Marine Sediment; Institute of Rock and Mineral Analysis, Beijing) and WG-1 (Woodstock Basalt; Australian National University, Canberra) were run in triplicate to calculate accuracy. Ten to twelve samples were also analysed in duplicate to measure precision. The accuracy and precision information is presented on element by element basis in accompanying spreadsheet. De-carbonated powders were sent to Environmental Isotopes Pty Ltd (Sydney) for isotopic analysis by mass spectrometry. Samples were back-corrected to account for the carbonate removal, using carbonate concentrations derived from the bomb method (this dataset). Error estimates for the C and N isotope values are ±0.15.

The Geostatistical Analyst extension of ArcGIS (ESRI® ArcGIS 10.5) was used to explore, model and generate surfaces for selected sediment geochemistry data. Deterministic methods (e.g. Inverse Distance Weighting (IDW)) and probabilistic methods (e.g. Kriging (ordinary (OK), Simple (SK), Universal (UK)) and Empirical Bayesian Kriging (EBK) were applied to interpolate geochemistry data across the study area. Where appropriate, outliers were removed from the analyses and data were normalised using either Log, Box-Cox, Empirical or Log Empirical transformations, depending of the interpolation method used. Tweaking of method properties (e.g. neighbourhood type, sector type, angle, trend removal, kernel function, anisotropy, sample subset size, number of simulations) for each of the spatial interpolation methods, a number of surfaces were derived with their cross-validation statistics. The best model was selected in terms of root-mean-square (RMS), mean standard error (MSE), mean standardised close to zero, the root-mean-square Standardised were close to one and the smallest RMS-MSE. For the best model, predicted surfaces and the predicted standard error maps were derived.

Variables used for the spatial interpolation were surrogates important for benthos: Chorin index, Total Reactive Chlorin, Total benthic CO2 flux, Total O2 uptake, 13C isotope, Carbon:Nitrogen ration, PCA Factor 1 and 2 and Total organic carbon.

Statistical analyses were undertaken on the sediment geochemical data to supplement the spatial interpretations using the Statistica V13.3 package. The analyses included Principle Components Analysis (PCA), Students t-tests and and Product-Moment Correlations. The PCA was undertaken on the element dataset after Centred Log Ratio transformation (see Grunsky et al. 2014).