Pluri-Gaussian rapid updating of geological domains
Sultan Abulkhair, Peter Dowd, Chaoshui Xu · 2025
Read Pluri-Gaussian rapid updating of geological domains by Sultan Abulkhair, Peter Dowd, Chaoshui Xu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-01575.txt.
Introduction
Sultan Abulkhair, Peter Dowd, and Chaoshui Xu present a 2025 method for updating geological domain maps in near real time as new mining data arrive. Resource models are often refreshed for continuous values like ore grade, but the domains themselves, the categories of rock, were left static. Domain maps are a core part of estimating a resource, so the gap matters.
Their method borrows from pluri-Gaussian simulation, which turns rock categories into smooth Gaussian fields and back again. A Gibbs sampler does the conversion, an ensemble Kalman filter folds in fresh sensor and drilling data, and a rotation step keeps several fields consistent. Tests used a synthetic deposit with known truth and a real case with production drilling samples. Grades are then updated inside the new boundaries.
Catalog record
| CREATOR | Sultan Abulkhair, Peter Dowd, Chaoshui Xu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-01575
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Pluri-Gaussian rapid updating of geological domains Sultan Abulkhair, Peter Dowd, Chaoshui Xu Abstract Over the past decade, the rapid updating of resource knowledge and the integration of real-time sensor information have garnered attention in both industry and academia. However, most studies on rapid resource model updating have focused on continuous variables, such as grade variables and coal quality parameters. Geological domain modelling is an essential component of resource estimation, which is why it is crucial to extend data assimilation techniques to enable the rapid updating of categorical variables. In this paper, a methodology inspired by pluri-Gaussian simulation is proposed for near-real-time updating of geological domains, followed by the updating of grade variables within these domain boundaries. The proposed algorithm consists of a Gibbs sampler for converting geological domains into Gaussian random fields, an ensemble Kalman filter with multiple data assimilations (EnKF-MDA) for rapid updating, and rotation based iterative Gaussianisation (RBIG) for multi-Gaussian transformation. We demonstrate the algorithm using a synthetic case study with observations sampled from the ground truth, as well as a real case study that uses production drilling samples for joint updating of geological domains and grade variables. Both case studies are based on real data from an iron oxide-copper-gold deposit in South Australia. This approach enhances resource knowledge by incorporating both categorical and continuous variables, leading to improved reproduction of domain geometries, closer matches between predictions and observations, and more geologically realistic resource models. 2025 \nomail \nomail keywords Ensemble Kalman filter with multiple data assimilations; Pluri-Gaussian simulation; Rotation-based iterative Gaussianisation; Geostatistics; Geological modelling; Resource knowledge 1 Introduction Resource modelling is a fundamental aspect of mining as it describes the location, quality, and quantity of mineral resources, which are essential for mine planning and optimisation ( Rossi and Deutsch 2014 ) . An important part of resource modelling is segmenting the ore body into spatial domains that have distinct geological and statistical characteristics. One way to account for these different domains and their grade distributions is through cascade modelling. In this method, the domains are modelled first, followed by estimating or simulating the grades within each domain ( Dowd 1986 ; Jones et al 2013 ; Madani and Abulkhair 2020 ; Abulkhair and Madani 2022 ; Zhexenbayeva et al 2024 ) . Conventional deterministic methods, such as wireframing, can produce accurate models when the drill hole spacing is dense and the geological domains are not overly complex ( Dowd 1986 ; Srivastava 2005 ) . However, in practice, mining deposits often involve multiple interconnected domains, and the available data may not be dense enough to create precise geological models. In contrast, stochastic geostatistical methods can generate multiple equally probable realisations of these domains while also quantifying uncertainty, particularly around the domain boundaries. Techniques such as sequential indicator simulation (SIS) ( Journel 1983 ; Deutsch 2006 ) and pluri-Gaussian simulation (PGS) ( Dowd et al 2003 ; Emery 2007 ; Armstrong et al 2011 ) are commonly used for this purpose. PGS is considered a superior method compared to SIS due to its capability to incorporate contact relationships among multiple categories. This is achieved by representing geological domains as two or more Gaussian random fields (GRFs), using techniques such as Gibbs sampling to enforce these contact relationships ( Emery 2007 ) . Madani and Emery 2015 introduced an improved hierarchical approach that can use more than two GRFs and local proportions governed by an interpreted geological model. Additionally, truncating non-stationary GRFs enhances the effectiveness of PGS in complex and heterogeneous environments, such as porphyry copper deposits ( Madani and Emery 2017 ) . Another alternative is multiple-point statistics (MPS) methods, which offer a better reproduction of connectivity, especially for heterogeneous and long-range structures ( Mariethoz and Caers 2014 ) . For instance, the direct sampling (DeeSse) algorithm applies the distance function to directly scan the training image and obtain spatial variability information ( Mariethoz et al 2010 ) . Aside from its applications in hydrogeology and petroleum facies modelling, DeeSse has been applied to model complex domains such as barren dykes ( Talebi et al 2019 ) , coal seam layers ( Abulkhair and Madani 2022 ) and gold veins ( Zhexenbayeva et al 2024 ) . The latter study showed considerably more accurate modelling of thin, long-range veins compared to PGS. However, the crucial downside of MPS remains its high simulation runtime and difficulty in deriving an appropriate training image. Nevertheless, regardless of the choice of modelling methodology, geological uncertainty remains a significant challenge, particularly around domain boundaries. Any of the aforementioned methods can only ensure precisely accurate results at the drill hole locations. As mining operations go deeper and grades decrease, it becomes increasingly important to ensure accurate resource modelling. One effective way to achieve this is by integrating real-time sensor observations available during mining operations at various stages of the mining value chain. For example, the ensemble Kalman filter (EnKF) has been extensively applied in the rapid updating of resource and grade control models over the past decade ( Benndorf 2015 ; Benndorf 2020 ; Wambeke and Benndorf 2017 ; Prior et al 2021b ; Prior et al 2021a ; Talesh Hosseini et al 2023 ) . Alternatively, methods, such as residual kriging to update GRF realisations ( Chevalier et al 2015 ) , actor-critic reinforcement learning ( Kumar and Dimitrakopoulos 2022 ) , and EnKF with multiple data assimilations (EnKF-MDA) ( Abulkhair et al 2025 ) have also been proposed for the same task. For instance, EnKF-MDA is based on updating the same data multiple times with an inflated measurement error ( Emerick and Reynolds 2012 ; Emerick and Reynolds 2013 ) . It offers a significantly closer match between model-based predictions and observations, especially in highly heterogeneous cases. Most studies on rapid resource model updating, however, have focused on continuous variables (e.g., grades, coal quality parameters, and compositional data). The main challenge with categorical variables (e.g., lithology, alterations, and facies) is how complicated it is to transform them into a format suitable for data assimilation with EnKF. Due to the Gaussianity assumption of EnKF, geological domains need to be parameterised into a continuous multivariate Gaussian space. For instance, Talesh Hosseini et al 2023 proposed a discrete wavelet transform to turn indicator variables into frequency coefficients and a subsequent EnKF rapid updating. However, it should be noted that the results validating this approach only involved two domains, meaning that its applicability to cases with complex contact relationships between multiple domains is unclear. Similarly, several methods have been proposed for the history matching of facies models, including PGS ( Liu and Oliver 2005 ) , principal component analysis (PCA) ( Sarma et al 2008 ) , level set functions ( Chang et al 2010 ) , and deep generative models ( Canchumuni et al 2019 ) . This study continues the work of Abulkhair et al 2025 , who proposed an algorithm that combines EnKF-MDA and multi-Gaussian transformation to rapidly update multivariate resource models. EnKF-MDA minimised the deviations between predictions and observations compared to conventional EnKF, while rotation based iterative Gaussianisation ( Laparra et al 2011 ; Abulkhair et al 2023 ; Cook et al 2023 ) helped to take into account complex multivariate relationships. In