Petrophysical characterisation of the Neoproterozoic and Cambrian successions in the Officer Basin

Keywords: Cambrian, characterisation, Exploring for the Future, laboratory measurement, log interpretation, Neoproterozoic, neural network, Officer Basin, permeability, porosity, reservoir, seal.

The Neoproterozoic–Paleozoic Officer Basin is a remnant of the Centralian Superbasin. It spans 525 000 km² across southeastern Western Australia and western South Australia (Fig. 1), with its present-day margins bounded by the Yilgarn Craton to the west, Musgrave Block to the north and Gawler Craton to the southeast (Morton and Drexel 1997; Grey et al. 2005). The Officer Basin remains a frontier basin for energy exploration, with significant uncertainty due to a lack of available data (Hashimoto et al. 2018). The stratigraphic architecture of the Officer Basin is poorly constrained, particularly due to limited biostratigraphic control in both the western and eastern Officer Basin (Morton and Drexel 1997; Apak and Moors 2001).

Fig. 1.  Major tectonic sub-divisions (modified from Gravestock 1997 and Carlsen et al. 2003), wells and hydrocarbon shows in the Officer Basin, superimposed on OZ Seebase 2021 (Geognostics 2021). MSL, mean sea level; GSWA, Geological Survey of Western Australia.

As part of Geoscience Australia’s Exploring for the Future (EFTF) program, a new chemostratigraphic framework was developed utilising elemental data obtained using Inductively Coupled Plasma-Optical Emission Spectrometry and Inductively Coupled Plasma-Mass Spectrometry testing by Chemostrat Australia on 1245 cuttings and 241 core samples from 10 Officer Basin wells (Munday et al. 2021). The five wells from the Western Australia area include Dragoon 1, Hussar 1, Geological Survey of Western Australia (GSWA) Vines 1, Lungkarta 1/ST1, and the five wells from the South Australia region include Birksgate 1, Giles 1, Kutjara 1, Mulyawara 1 and Trainor Echo 1 (Fig. 1). Key elements and ratios, which typically have the best correlation potential over a greater distance, were selected to subdivide the strata and highlight changes in sediment provenance, as well as to identify climatic and organic matter changes.

In Munday et al. (2021), five chemostratigraphic mega-sequences (MS1–MS5) were defined, predominantly in the Neoproterozoic and Cambrian successions of the Officer Basin. MS1–MS3 broadly correspond to the Neoproterozoic Centralian Supersequence (CS1–CS3) as defined by Bradshaw et al. (2021). MS4 corresponds to the Cambrian CS4 (Bradshaw et al. 2021) and includes most of the Marla Group (Khider et al. 2021; Munday et al. 2021). Each MS has a different number of sequences, dependent upon the variability in their chemical composition. These MSs, sequences and their corresponding stratigraphic units are shown in Fig. 2, incorporating the lithostratigraphy after Apak and Moors (2001), Apak and Carlsen (2003), Haines et al. (2008) and Khider et al. (2021) as follows:

1. MS1 has 11 sequences (MS1-S0–MS1-S10), corresponding to the Pindyin Sandstone and the Alinya, Browne, Hussar and Kanpa formations.

2. MS2 has three sequences (MS2-S1–MS2-S3), corresponding to the Pirrilyunka Formation in GSWA Vines 1.

3. MS3 has seven sequences (MS3-S0–MS3-S6), corresponding to the Wahlgu and Meramangye formations, Tarlina Sandstone, Murnaroo Formation, Dey Dey Mudstone, Karlaya Limestone, and Tanana, Munyarai and Narana formations.

4. MS4 has five sequences (MS4-S1–MS4-S5), corresponding to the Relief Sandstone, Observatory Hill Formation, Arcoeillinna Sandstone, and Apamurra and Lungkarta formations.

5. MS5 corresponds to the upper part of the Cambrian Trainor Hill Sandstone (Marla Group) in Trainor Echo 1 and the Table Hill Volcanics in Yowalga 3.

Fig. 2.  Stratigraphic columns of the Neoproterozoic–Cambrian successions in the Officer Basin after Khider et al. (2021), Haines et al. (2008), Bradshaw et al. (2021) and Hashimoto et al. (2018), integrating the chemostratigraphic mega-sequences and sequences of 10 Officer Basin wells (Munday et al. 2021). Column 1: Age (Ma); Column 2: period; Column 3: Centralian supersequence (CS) defined by Bradshaw et al. (2021); Columns 4 and 5: mega-sequence (MS) and sequence defined in Munday et al. (2021); Column 6: major formations in the Western Australia region (Western Officer); Column 7: main formations in the South Australia region (Eastern Officer). The legends indicate those sequences or formations with short age intervals.

The stratigraphic columns of the Neoproterozoic–Cambrian successions have been updated after Khider et al. (2021), Haines et al. (2008), Bradshaw et al. (2021) and Hashimoto et al. (2018) and integrate the new chemostratigraphic MSs and sequences of the 10 Officer Basin wells (Munday et al. 2021).

In the current study, petrophysical properties are analysed through laboratory testing, as well as analysis and interpretation of well logging data covering the Neoproterozoic and Cambrian successions from MS1 to MS4 in six selected Officer Basin wells: Birksgate 1 (Henderson and Tauer 1967), Giles 1 (Dunster 1986), GSWA Vines 1 (Apak et al. 2001), Munta 1 (Cucuzza 1987), Yowalga 3 (Shell Development (Australia) Pty Ltd 1981) and Hussar 1 (Karajas and Yaylor 1983). Some of these wells feature gas indications in mud-logging data and host other minor hydrocarbon indications (Fig. 1). For example, in Birksgate 1, the laboratory measured water saturation from cores range from 25 to 77% in MS4, and yellow spots of crude oil were observed on cores (Henderson and Tauer 1967). Petrophysical characterisations are reported following the chemostratigraphic MSs and sequences interpreted by Munday et al. (2021), which are placed within the stratigraphic framework of Khider et al. (2021).

Apak and Moors (2001) presented potential conventional reservoirs and their sedimentary and diagenetic genesis of the Neoproterozoic−Paleozoic successions in the Lennis area of the Officer Basin, based on laboratory measured porosity and permeability, and log-derived porosity. As part of the EFTF program, 41 legacy core samples were collected from 5 (Table 1) of the 6 wells analysed for testing of petrophysical properties.

Table 1.  Number of samples for petrophysical testing from five wells in the Officer Basin.

The samples listed in Table 1 represent a range of lithotypes and were analysed and tested in CSIRO’s laboratory to derive a variety of petrophysical properties of conventional, tight and shale reservoir rocks. The analyses and tests included:

1. Core imaging (scanning and digital photography) through medical X-ray Computerised Tomography to identify general texture and internal microstructure and to derive dry bulk density.

2. Gas porosity and permeability testing on dry core plugs using a CoreTest Inc. automated nitrogen Permeameter–Porosimeter for samples with permeabilities greater than 1 µD.

3. Unsteady state pulse decay testing using a CoreTest Inc. NDL-605 apparatus for ultra-low permeability samples (11 samples with permeabilities of less than 1 µD).

4. Grain density measured from powders (ground from dry core plugs) with a helium gas displacement pycnometer (Accupyc II 1340, Micrometritics).

5. Mercury injection capillary pressure (MICP) analysis of select samples.

The average laboratory measured total porosity of conventional, tight and shale reservoir rocks is 17.79, 7.09 and 5.38%, respectively, and geometric mean permeability is 5.63, 0.00135, and 0.00048 mD, respectively (Bailey et al. 2021). The average measured gas (nitrogen) porosity of tight and shale reservoir rocks is 2.26 and 1.07%, respectively. Table 2 presents the average petrophysical parameters of various rock types in the Neoproterozoic and Cambrian successions. Further details, and all results, from the testing program are presented in Bailey et al. (2021). Fig. 3 presents the relationships between air-permeability, corrected permeability after removing slippage effect (Klinkenberg correction) and correction from gas permeability to nano-scale permeability testing results.

Table 2.  Average petrophysical properties of the various rock types in the Officer Basin.

Fig. 3.  Relationships among various permeability based on laboratory measurements of 41 samples, including relationships between logarithmic air permeability (log KG) and Klinkenberg corrected permeability (log KC) of (a) conventional reservoir rocks, (b) tight reservoir rocks, (c) shales, and (d) between Klinkenberg corrected permeability and nano-scale permeability (log KN) of tight reservoir rocks and shales.

Alongside newly acquired laboratory measurements, previously conducted laboratory measurements of petrophysical properties were collated. These were mainly sourced from the well completion reports of the six selected wells (Fig. 1). Laboratory measurements sourced from these well completion reports primarily include grain density, porosity and air permeability, with porosity having been tested mainly using helium gas expansion methods (Ahmed 2002; Yu 2013). In this study, the relationships summarised in Fig. 3 were used to correct the gas permeability for conventional, tight and shale reservoir rocks. The newly acquired and previously measured laboratory data provide a combined dataset of 193 effective porosity and 177 permeability measurements, which were used as training targets for neural network interpretations.

Legacy well logging data were obtained from the Western Australia Petroleum and Information Management System (WAPIMS 2021) for the wells in the western Officer Basin and from the South Australian Resources Information Gateway (PEPS 2021; SARIG 2021) for the wells in the eastern Officer Basin. Well logging data were acquired by Schlumberger in Yowalga 3; Reeves/BPB in Hussar 1, Giles 1, and GSWA Vines 1; Welex in Birksgate 1; and Geoscience Associates (Australia) in Munta 1.

Birksgate 1 has gamma ray, compressional wave slowness, neutron and resistivity logs, but no bulk density log. Bulk density for this well was approximated from compressional wave slowness using Brocher’s polynomial model (Gardner et al. 1974; Brocher 2005). The remaining five wells have gamma ray, compressional wave slowness, neutron porosity and bulk density logs. Selecting a method for interpreting wireline logs depends on data availability and quality. Other well data, such as formation temperature, drillstem testing, organic and inorganic geochemical data, and mud properties, were collected from the relevant well completion reports held by WAPIMS (2021) and the South Australian Resource Information Gateway (SARIG 2021). In this study, well log data interpretation is performed by combining conventional and neural network interpretations.

After data collection and pre-processing, conventional interpretation was conducted in Schlumberger’s Techlog^TM platform. The volume fraction of shale (Vshale) was interpreted from gamma ray logs using the Larionov model (Schlumberger 1991) and neutron-density crossplots. Effective and total porosity were computed from compressional wave slowness in Birksgate 1 and from compressional wave slowness and neutron-density crossplots in the remaining five wells (Schlumberger 1991, 2003; Ellis and Singer 2007; Wang et al. 2021). The interpretation intervals were defined in reference to the chemostratigraphic sequences for the Neoproterozoic−Cambrian sedimentary succession defined in Munday et al. (2021).

Unsupervised machine learning algorithms, such as the K-means and Gaussian mixture model, SOM and many others, have been used for lithofacies identification (Balashov et al. 2019; Saputelli et al. 2019; Celma et al. 2020). As a neural network tool, SOM is used to classify multivariate data samples into petrophysical groups or classes using unsupervised learning. Compared to classic clustering algorithms, SOM provides the management of transitions between classes, which is important in analysing geological questions. In this study, SOM clustering was performed to define 10 petrophysical groups or classes using the Ipsom module in Schlumberger’s Techlog^TM platform, and utilising gamma ray, compressional wave slowness, bulk density, neutron porosity, interpreted Vshale, effective and total porosity. During the SOM modelling, hierarchical clustering with Ward’s method was used to cluster the input parameters. The petrophysical classes are then sorted by effective porosity and ordered from minimum to maximum porosity.

The outputs from SOM clustering include class indexes (1–10), and probability profiles so that the data samples belong to different classes (Columns 10–11, Figs 4, 5). The SOM classes present the clustered petrophysical groups, and the class index indicates the groups with different petrophysical property ranges. The cumulated class profiles provide another dataset for lithostratigraphic correlation between wells. Figs 4, 5 present the input logs and interpretation results from conventional methods, SOM modelling and neural network interpretation results. Neural network interpretation will be discussed in the next section.

Fig. 4.  Well profiles for GSWA Vines 1. Column 1: measured depth (MD, m); Columns 2 and 3: mega-sequences and sequences defined by Munday et al. (2021); Column 4: lithostratigraphic units (Zone_WA) defined by Haines et al. (2008) (LF, Lungkarta Formation; and VF, Vines Formation); Column 5: lithological descriptions from composite log; Column 6: gamma ray (GR, gAPI) and compressional wave slowness (DTC, µs/ft) logs; Column 7: interpreted Vshale (VSH, m³/m³), effective porosity (PHIEF, m³/m³) and total porosity (PHITF, m³/m³) from conventional interpretations; Column 8: neural network interpreted and laboratory measured porosity (PORNN and POR, m³/m³); Column 9: neural network interpreted and laboratory measured permeability (PERMNN and PERM, mD); Column 10: rock class from SOM clustering (Class); Column 11: cumulative probability curves of classes from SOM clustering (ClassCum).

Fig. 5.  Well profiles for Giles 1. Column 1: measured depth (MD, m); Columns 2 and 3: mega-sequences and sequences defined by Munday et al. (2021); Column 4: lithostratigraphic units (Zone_EFTF, Khider et al. 2021) (KL, Karlaya Limestone; TF, Tanana Formation; AF, Alinya Formation; and PS, Pindyin Sandstone); Column 5: lithological descriptions from composite log; Column 6: gamma ray (GR, gAPI) and compressional wave slowness (DTC, µs/ft) logs; Column 7: interpreted Vshale (VSH, m³/m³), effective porosity (PHIE, m³/m³) and total porosity (PHIT, m³/m³) from conventional interpretations; Column 8: neural network interpreted and laboratory measured porosity (PORNN and POR, m³/m³); Column 9: neural network interpreted and laboratory measured permeability (PERMNN and PERM, mD); Column 10: rock class from SOM clustering (Class); Column 11: cumulative probability curves of classes from SOM clustering (ClassCum).

Permeability is a key parameter for reservoir characterisation, as it is critical for analysis of fluid flow and production performance. Empirical, linear, simple non-linear equations or regression models have been proposed for permeability interpretation from well logs (Schlumberger 1991; Asquith and Krygowski 2004). However, it is difficult to estimate permeability from multiple input logs because of the complex relations between permeability and various logs, especially where the lithological types are complex, as is the case for the Neoproterozoic–Cambrian succession in the Officer Basin. The logarithmic permeability derived from gamma ray, compressional wave slowness, bulk density, neutron porosity, interpreted Vshale, effective and total porosity, and petrophysical class index using multivariate linear regression has the correlation coefficient (R²) of 0.3645 with the laboratory measurements. An ANN approach was used to learn the underlying correlations between the input parameters and laboratory measurements and to improve the approximation of permeability in this study.

ANNs have been used to solve a wide variety of tasks (Hopfield 1985; Hect-Nielsen 1990; Kalogirou 2000). In the petroleum domain, ANNs have been applied in estimating hard data, such as laboratory measurements, from seismic attributes, well logs and conceptual geological descriptions (Wong et al. 1995; Wang et al. 1999, 2021; Huang et al. 2001; Yan et al. 2014; Ouadfeul and Aliouane 2015; Julian 2016; Saputelli et al. 2019). Fig. 6 presents an example of an architecture of a neural network with input, output and three hidden layers. Each layer has different numbers of neurons.

Fig. 6.  Neural network architecture with one input, one output and three hidden layers.

The input layer imports the input parameters, the input of each neuron in hidden and output layers is a linear combination of neuron outputs in the former layer, and the output neuron gives the output parameter. Activation functions, such as rectified linear unit (ReLU), sigmoid and hyperbolic tangent functions, are used to transform input to output at each hidden neuron. A loss (error) function is constructed by comparing to the targets (laboratory measurements). During the neural learning process, the weights connecting neurons are updated using batch gradient descent or stochastic gradient descent algorithms. As a popularly used optimiser, adaptive momentum estimation (Adam) enables us to compute adaptive learning rates and avoid oscillations during optimisation by accelerating the descent in the right direction (Kingma and Ba 2015; Selvam 2018; Brownlee 2019; Stewart 2019). A variety of hyperparameters need to be configured for neural network learning, such as the number of hidden layers and neurons, as well as the selection of an activation function (Avila and Hauck 2017; Brownlee 2019). Sensitivity analysis is performed for optimisation purposes using the root mean square error, the mean absolute error and the correlation coefficient (R²) between approximations and targets.

In this study, the input parameters include a series of selected well logs (gamma ray, compressive wave slowness, bulk density and neutron porosity), interpreted Vshale, effective and total porosity using conventional methods, and SOM clustered petrophysical class index. Table 3 presents the R² between the input parameters and laboratory measured porosity and permeability in the six selected wells. The lithology logs, interpreted effective and total porosity, and petrophysical class index have higher R² with the laboratory measured porosity than the remaining input parameters. The bulk density, interpreted Vshale and porosity, and petrophysical class index are better correlated with the logarithmic permeability than other input parameters.

Table 3.  Correlation coefficients (R²) between the input parameters and laboratory measured porosity and permeability in the six selected wells.

A total of 193 and 177 training patterns were constructed for porosity and permeability interpretations, respectively. One single output parameter is either porosity or permeability (Fig. 6). The input logs were standardised among the six wells using the medians of logs in the halite and shale intervals within the same sequence. All the input parameters were normalised to be within the range of [−1, 1] before neural network learning.

The optimal neural network architecture for each porosity and permeability estimations has three hidden layers with 20, 200 and 20 hidden neurons based on sensitivity analysis. ReLU was selected as the activation function, and the Adam algorithm was used during the neural learning processes for both porosity and permeability estimations. The R² between the laboratory measured and neural network interpreted effective porosity and permeability (Columns 8–9, Figs 4, 5) are 0.7912 and 0.7167, respectively. The neural network interpreted porosity is slightly better correlated with target porosity than the porosity from conventional interpretations in Table 3. Compared to the permeability estimation results from multivariate linear regression, the neural network approach is capable of handling the complexity inside petrophysical data and of deriving the approximations with a significantly higher R² with targets. The interpreted porosity and permeability are summarised for different chemostratigraphic sequences and various reservoir types in the following section.

A variety of potential reservoirs and seals have been previously identified in the Neoproterozoic and Cambrian successions of the Officer Basin (Walter and Gorter 1994; Zang 1995; Morton and Drexel 1997; Apak and Moors 2001; Apak and Carlsen 2003; Tingate and McKirdy 2003; Samson 2018). This study focuses on the descriptions of Vshale, porosity and permeability distributions for reservoirs and seals of the six selected Officer Basin wells (Fig. 1). The sedimentary rocks are investigated by integrating laboratory measurements and well log interpretations for conventional, tight and shale reservoir rocks. The three reservoir types are defined using the following cutoffs on the volume fraction of shale (Vshale, Column 7, Figs 4, 5), neural network interpreted porosity (PORNN, Column 8, Figs 4, 5) and permeability (PERMNN, Column 9, Figs 4, 5):

1. Conventional reservoir rock: Vshale <0.5; porosity ≥0.1 and permeability ≥0.1 mD

2. Tight reservoir rock: Vshale <0.6; permeability <0.1 mD

3. Shale: Vshale ≥0.6.

Figs 7–9 present the distributions of Vshale, effective porosity and permeability of various presented sequences from MS1 to MS4 in the six selected wells in this study. The three potential reservoir types are present in both the western and eastern Officer Basin, apparent where multiple peaks appear in the distribution histograms (Figs 7–9). The distribution of petrophysical properties implies the potential for forming different types of reservoirs.

Fig. 7.  Distributions of interpreted Vshale, porosity and permeability in different sequences in Hussar 1 and Yowalga 3.

Fig. 8.  Distributions of interpreted Vshale, effective porosity and permeability in different sequences in GSWA Vines 1 and Giles 1.

Fig. 9.  Distributions of interpreted Vshale, effective porosity and permeability in different sequences in Birksgate 1 and Munta 1.

Table 4 lists the ratios of conventional and tight reservoirs (reservoir thickness over gross thickness), average Vshale and effective porosity, and geometric mean permeability of each MS (MS1–MS4). Table 5 presents the average Vshale and porosity, and geometric mean of permeability of potential conventional and tight reservoirs in each MS in the six selected wells. Besides the conventional and tight reservoirs, the laboratory measured total porosity of the Cambrian shales range from 8.27 to 12.86% with an average of 5.38%, whereas the total porosity ranges from 1.76 to 6.33% with an average of 4.13% in the Neoproterozoic shales (Bailey et al. 2021). The further study on reservoir potential includes the evaluations on, but not limited to, the organic matter richness of source rocks and thermal maturity, hydrocarbon migration, brittleness of shales and geomechanical properties.

Table 4.  Ratios of conventional and tight reservoirs (RCRG and RTRG), average interpreted Vshale, porosity and geometric mean of permeability of each mega-sequence (MS) in the six selected Officer Basin wells.

Table 5.  Average interpreted porosity and geometric mean of permeability of conventional and tight reservoirs in each mega-sequence (MS) in the six selected Officer Basin wells.

MS1 was deposited in a large thermal sag basin initiated in the early Neoproterozoic (late Tonian), and the sedimentary build-up was controlled by northwest–southeast extension (Morton and Drexel 1997; Apak and Moors 2001; Apak and Carlsen 2003; Tingate and McKirdy 2003; Boult and Rankin 2004; Bradshaw et al. 2021; Khider et al. 2021). The sequences include MS1-S2–MS1-S9 in Yowalga 3, MS1-S5–MS1-S10 in Hussar 1, and MS1-S0 and MS1-S2 in Giles 1. The sedimentary rocks of MS1 have been correlated to the Pindyin Sandstone, and Alinya, Browne, Hussar and Kanpa formations (Fig. 2) that comprise clastic, carbonate and evaporites inferred to have been deposited in a sabkha environment.

MS1 is interpreted to be a predominantly tight reservoir in Yowalga 3 and Giles 1, with a small portion (less than 20%) interpreted as A conventional reservoir in Hussar 1. The range in the average porosity and geometric mean permeability of the interpreted tight reservoirs are 5.7−6.39% and 0.00087−0.01307 mD, respectively, in the different MS1 sequences. The interpreted conventional reservoirs in the study wells show a range in average porosity and geometric mean permeability of 12.34−16.02% and 3.851−29.391 mD, respectively, in the MS1 sequences.

MS2 was identified only within GSWA Vines 1 by Munday et al. (2021) and corresponds to the Pirrilyungka Formation in Haines et al. (2008) (Fig. 2). MS2 is characterised by dolomitic clastics (massive sandstone, diamictite and shaly diamictite) and deposition ranging from mass flow in deep marine environments; to glacigene, shallow marine; to continental, pebbly outwash deposits in the GSWA Vines 1 area. MS2 has three sequences (MS2-S1–MS2-S3, Munday et al. 2021). A broad correlation can be seen within the geochemistry and the sedimentological subdivisions defined by Haines et al. (2008). MS2 mainly hosts tight reservoirs, with an average porosity and geometric mean permeability of 4.77% and 0.00139 mD, respectively.

MS3 broadly corresponds to the lithostratigraphy from the Tarlina Sandstone to the Tanana Formation, although this does vary depending on the lithostratigraphic nomenclature used within each well (Fig. 2). This MS is recognised predominantly within the eastern wells (Munday et al. 2021), including MS3-S1–MS3-S6 in Giles 1, MS3-S5 in Birksgate 1, and MS3-S4–MS3-S6 in Munta 1. This MS mainly consists of fluvial-deltaic-lacustrine massive sandstone with thin silty mudstone, fluvial-deltaic-marine and fluvial-tidal clastics, subtidal shelf limestone/mudstone, intertidal to upper subtidal calcareous siltstone with minor sandstone, and debris flow to tidal flat clastics (Zang 1995; Morton and Drexel 1997; Apak and Moors 2001; Tingate and McKirdy 2003).

Tentative sequences are interpreted within the western Officer Basin wells, such as MS3-S1–MS3-S2 in Hussar 1 and MS3-S0 in GSWA Vines 1. The sequence MS3-S0 corresponds to the Wahlgu Formation, which is characterised by glacially related diamictite, conglomerate, mudstone and sandstone (Apak and Moors 2001; Haines et al. 2008)

MS3 mainly hosts tight reservoirs with a range in average porosity and geometric mean permeability of 5.58–8.05% and 0.00088−0.01102 mD, respectively, in different sequences. The conventional reservoirs identified in GSWA Vines 1 and Hussar 1 correspond to the Wahlgu Formation, where they have a range in average porosity and geometric mean permeability of 19.72–21.74% and 1.1655–15.6723 mD, respectively, in these two wells.

MS4 was predominantly identified within the eastern study wells in Munday et al. (2021). This includes MS4-S1 and MS4-S4 in Giles 1, MS4-S1–MS4-S5 in Birksgate 1, MS4-S1 in GSWA Vines 1, MS4-S3 and MS4-S4 in Hussar 1, and MS4-S2–MS4-S5 in Munta 1. MS4 was deposited with highly variable accommodation in the Officer Basin (Morton and Drexel 1997; Tingate and McKirdy 2003).

In the eastern Officer Basin wells, MS4-S1 in Birksgate 1 and Giles 1 corresponds to the Relief Sandstone (Fig. 2), which was deposited in aeolian, fluvial and tidal environments where the facies change is related to cycles of relative sea-level change. The subsequent deposits mainly include micaceous siltstone and claystone with minor very fine-grained sandstone, limestone and dolomite that formed in a distal alluvial fan, fluvial, ephemeral lake to alkaline playa lake environment. The very fine- to medium-grained feldspathic sandstone formed in a fluvial-lacustrine environment, whereas calcareous and dolomitic fine- to very fine-grained sandstone and siltstone were deposited in a proximal alluvial fan environment (Morton and Drexel 1997). In GSWA Vines 1, MS4-S1 comprises the aeolian sandstone equivalent to the Lungkarta Formation (Haines et al. 2008) and hosts a massive, highly porous sandstone reservoir.

This MS hosts significant sandstone units interpreted to be potential conventional reservoirs. The range in the average porosity and geometric mean permeability of these reservoirs are 18.4−26.38% and 8.5–152.613 mD, respectively, in the different MS4 sequences. Interpreted tight reservoirs display a range in their average porosity and geometric mean permeability of 7–10.4% and 0.00089–0.00758 mD, respectively, in the MS4 sequences.

Within the Officer Basin, most of the formations contain lithologies that would make effective seals at all scales. All of the Neoproterozoic sedimentary rocks contain both carbonate and shale facies that are suitable as seals, and the best shale seals were deposited on flooding surfaces and form the bases of the Kanpa, Hussar, and Steptoe formations (Apak and Moors 2001; Raven and Dewhurst 2005). In this study, the laboratory measured gas (nitrogen) porosity ranges from 0.8 to 6.27% with an average of 2.9% in the Cambrian shales, whereas the porosity ranges from 0.15 to 2.85% with an average of 0.61% in the Neoproterozoic shales. Four shale samples with very low measured gas (nitrogen) porosities between 0.2 and 0.8% were selected for analysis via MICP testing. These samples are from the wells Giles 1 (992.25 mRT, MS3-S2), Birksgate 1 (972 mRT, MS4-S2), Yowalga 3 (2385.8 mRT, MS1-S5) and GSWA Vines 1 (985.52 mRT, MS2-S3). Table 6 presents the densities of fresh water (H₂O), methane (CH₄) and carbon dioxide (CO₂) under in situ pressures and temperatures (P–T). Table 7 presents the calculated gas (pure CH⁴ and CO₂) column heights under in situ P–T conditions calculated from MICP analysis. The derivation of gas column height from capillary threshold pressure is given in Ahmed (2002).

Table 6.  In situ P–T conditions estimated for four wells taken from sample depths, assuming hydrostatic pressure and 35°C/km geothermal gradient (plus 25°C average surface temperature).

Table 7.  Air-mercury capillary threshold pressures (Air-Hg Pth) for four samples from the given wells and calculated gas (pure methane, CH₄) and CO₂ column heights for in situ P–T conditions.

Besides the shales, salt penetrates from the lower part of the Browne Formation through to the overlying section. Salt provides an effective barrier to petroleum migrating from structural deeps (Apak and Moors 2001; Apak and Carlsen 2003) and, as well as forming good seals in petroleum systems, interbedded and dome salts (halite) have the potential for the future underground storage of hydrogen (Crotogino et al. 2018).

Petrophysical properties of the Neoproterozoic–Cambrian succession in the six selected wells from the Officer Basin have been derived through laboratory testing and well log interpretation. This was achieved through conventional and neural network interpretations and produced the continuous log values for porosity, permeability and Vshale. Integration of well logs, log interpretations, lithological descriptions and laboratory testing results, has determined the potential for conventional, tight and shale hydrocarbon reservoirs within the Neoproterozoic–Cambrian successions (MS1–MS4) in the Officer Basin as follows:

1. MS1–MS3, which correspond to the Centralian supersequences 1, 2 and 3, have mainly tight reservoirs with the average porosity ranging from 4.77 to 6.39% and geometric mean permeability ranging from 0.00087 to 0.01307 mD in the different sequences.

2. In MS4, the range in average porosity and geometric mean permeability of conventional reservoirs is 14.54–26.38% and 0.341–103.68 mD, respectively, in the different sequences. The range in the average porosity and geometric mean permeability of tight reservoirs are 5.58–8.6% and 0.00088–0.01102 mD, respectively, in the different sequences.

3. The ANN is capable of handling the complexity of petrophysical data, and the R² between the neural network interpreted permeability and laboratory measurements is 0.7167.

Interpretation of very low permeability shales via MICP analysis demonstrates high sealing capacities, with derived potential gas column heights of 520–1250 m for pure CH₄ and 390–930 m for CO₂. Known salt units are likely to act as high-quality seals. This study demonstrates the presence of key petroleum system elements (reservoir and seal) in the Officer Basin of South Australia and Western Australia and highlights the value of advanced neural network methods for filling significant data gaps.

The laboratory testing data supporting this study are available at Geoscience Australia’s eCat system (https://ecat.ga.gov.au/geonetwork/srv/eng/catalog.search#/metadata/145834). The well logging data and well completion reports are available at the South Australian Resource Information Gateway (SARIG) (https://minerals.sarig.sa.gov.au/QuickSearch.aspx), Petroleum exploration and production system (PEPS) – South Australia (https://peps.sa.gov.au/home), and the Western Australia Petroleum and Information Management System (WAPIMS) (https://wapims.dmp.wa.gov.au/WAPIMS/Search/Wells).

All authors confirm there are no conflicts of interest.

This study was funded by Geoscience Australia’s Exploring for the Future program. No funding from external organisations was received for this research.