Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Ehsan Farahbakhsh has a total of 69 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    How subduction evolution drives sediment-hosted mineralisation along craton edges

    by Hojat Shirmard, Ben Mather, Ehsan Farahbakhsh, Craig O’NEill, R. Dietmar Müller

    Sediment-hosted mineral deposits cluster near craton edges, yet the geodynamic factors influencing this concentration remain poorly understood. To investigate the tectonic and geodynamic controls on craton-edge mineralisation, we integrate a 1.8 Ga global plate motion model with craton-edge mapping from full-waveform seismic tomography, geodynamic modelling, and a global mineral deposit database. Here we show that mineralised craton edges consistently cluster 800–1800 km from subduction zones at the time of formation, with some as far away as ~3000 km—a spatial pattern distinct from random craton-edge locations. Geodynamic models show that subduction generates broad mantle return-flow cells that focus lithospheric stress, strain, and weakening at comparable distances from the trench. We propose that this mechanically driven weakening promotes rifting, enhances permeability, and facilitates the infiltration of slab-derived volatiles, thereby preconditioning the lithosphere for metallogenesis. These results identify subduction–craton coupling as a first-order control on sediment-hosted mineral systems across multiple supercontinent cycles and provide a predictive geodynamic framework for understanding the distribution of continental resources through deep time.Subduction near cratons fosters conditions that enrich the mantle and promote mineral formation along craton edges.

  • Open Access

    DEEP-SEAM: an explainable semi-supervised deep learning framework for mineral prospectivity mapping

    by Zijing Luo, Ehsan Farahbakhsh, Stephen Hore, R. Dietmar Müller

    The global transition to clean energy is sharply increasing demand for rare earth elements (REEs), yet discovery rates are declining, especially in areas concealed by younger cover. Deep learning (DL) offers new opportunities for mineral prospectivity mapping (MPM), but its application is challenged by sparse labelled mineral occurrences, strong class imbalance, and limited model transparency. To address these issues, we present DEEP-SEAM, an explainable semi-supervised DL framework that integrates geological, geophysical, geochemical, remote sensing, and topographic datasets to predict REE prospectivity in the northern Curnamona Province, South Australia. The framework employs the Deviation Network (DevNet), a semi-supervised anomaly detection model that learns from a small number of known REE occurrences together with abundant unlabelled samples. DEEP-SEAM produces highly accurate predictions: the top 2 % of the mapped area contains 86 % of known REE deposits, and nearly all known occurrences fall within the highest-prospectivity zones. These areas show strong spatial association with felsic granites, major faults, and Mesoproterozoic metasedimentary sequences – features consistent with established REE mineral system models. To improve interpretability, we apply SHapley Additive exPlanations (SHAP), which highlight radiometric signatures, magnetic pseudo-gravity attributes, hydrothermal alteration indicators, and key geochemical principal components as the most influential predictors. These insights align with independent geological evidence, strengthening confidence in the predictive outcomes. DEEP-SEAM provides a transparent, scalable, and data-efficient approach for delineating REE prospectivity in complex and partially covered terranes, offering a valuable tool for reducing exploration risk and guiding future targeting efforts.


  • Open Access

    Spatiotemporal controls on porphyry mineralisation from global plate reconstructions and machine learning

    by Ehsan Farahbakhsh, Brent I. A. McInnes, Fabian Kohlmann, Maria Seton, Adriana Dutkiewicz, R. Dietmar Müller

    Porphyry mineral systems formed in volcanic arcs supply most of the world’s copper and constitute major sources of gold and molybdenum, making them central to both resource security and exploration strategy. Despite decades of study, exploration targeting at regional to global scales remains challenging because porphyry formation and preservation are governed by complex, long-lived tectonic and surface processes. In particular, uncertainties persist regarding how evolving subduction regimes, properties of the subducting slab, and post-formational erosion interact to control where porphyry deposits ultimately occur and are preserved in the geologic record.Current prospectivity models typically focus on present-day tectonic configurations or isolated geological proxies, leaving a critical gap in our ability to explicitly link porphyry mineralisation to the full spatio-temporal evolution of subduction systems and landscape processes over hundreds of millions of years. As a result, the relative importance of slab-derived inputs, overriding plate structure, and preservation potential remains poorly constrained at a global scale.Here, we address this gap by developing a global, data-driven prospectivity model for porphyry mineralisation using a machine learning framework trained on a worldwide database of known porphyry occurrences. The model integrates time-dependent features derived from global plate tectonic reconstructions from 540 Ma to the present, including ocean basin evolution, alongside erosion–deposition outputs from a landscape evolution model to explicitly quantify preservation likelihood and the probability of preserved mineralisation.Our results highlight several strong controls on porphyry occurrence. The most influential parameters are crustal thickness of the overriding plate, total sediment thickness, carbonate accumulation on the subducting plate, and distance from the trench. Porphyry deposits are most likely to occur 200–400 km from the trench, preferentially where overriding crust exceeds ~36 km in thickness and the downgoing slab carries more than ~200 m of pelagic sediments. While trench distance likely reflects slab dip and the degree of mantle wedge hydration, our findings suggest that mechanical coupling between plates plays a key role in modulating volatile transfer and deformation. Notably, carbonate thickness emerges as a positive predictor of mineralisation probability, implying that carbon-rich slabs may enhance volatile fluxes into the mantle wedge—an interpretation that challenges the traditional emphasis on silicate-dominated sediment inputs alone.We propose that thickened continental crust promotes magma differentiation and volatile retention, fostering the generation of fertile, chalcophile-metal–rich magmas, while sediment- and carbonate-rich slabs deliver abundant water- and carbon-bearing fluids that metasomatise the mantle wedge and redistribute metals such as Cu and Au. Together, these results demonstrate that porphyry systems are products of tightly coupled tectonic, geochemical, and surface processes operating over geological timescales. This integrated framework provides a new basis for global-scale exploration targeting and underscores the importance of considering both mineral system fertility and long-term preservation in assessing porphyry prospectivity.

  • Open Access

    A subduction distance control on craton-margin metallogenesis

    by Hojat Shirmard, Ben Mather, Ehsan Farahbakhsh, Craig O'NEill, R Dietmar Muller

    Craton margins represent long-lived lithospheric weak zones that host a disproportionate share of the world’s sediment-hosted Pb–Zn and Cu resources, yet the geodynamic conditions that distinguish fertile from barren margins remain poorly constrained. Here, we test the hypothesis that subduction dynamics exert a first-order control on craton-margin metallogenesis by integrating a 1.8-billion-year global plate motion model, machine learning–derived craton boundary mapping from full-waveform seismic tomography, a global database of age-coded sediment-hosted deposits, and numerical geodynamic simulations.Spatiotemporal analysis reveals that mineralised craton margins systematically cluster within 2000 km from active subduction trenches at the time of deposit formation—a spatial signal absent from 90,000 randomly generated craton-margin locations propagated through deep time. More than 90% of the total contained metal endowment of the analysed deposits lies within this 2000 km threshold, demonstrating that trench proximity is a robust discriminator between fertile and barren craton edges. This relationship is consistent across three supercontinent cycles and multiple deposit types, including Pb–Zn clastic-dominated, Mississippi Valley–type, and sediment-hosted Cu systems. Kinematic analysis further shows that deposits formed preferentially during episodes of moderate trench retreat, indicating that the overriding plate migrates oceanward and repositions cratons over previously subducted domains.Numerical geodynamic models reproduce a comparable spatial scale, with subduction-driven mantle return flow generating lithospheric strain-rate maxima at craton margins approximately 2000 km from trenches. These results indicate that subduction transmits stresses thousands of kilometres into the overriding plate, localising deformation at craton edges while preserving craton interiors. There is a slight offset between observed deposit clustering and modelled strain peaks, likely reflecting inherited lithospheric heterogeneity and three-dimensional mantle flow effects not captured by simplified two-dimensional models. Strain localisation enhances permeability, facilitating long-term metasomatic enrichment of the subcontinental lithospheric mantle by slab-derived fluids.Subduction-derived volatiles and ligands—including halogens, carbon, and reduced sulphur—play a critical role in metallogenic fertility by increasing the capacity of basinal brines to dissolve, transport, and precipitate metals. Episodes of trench retreat position cratons over previously enriched mantle domains, promoting the ascent of metal-bearing fluids and partial melts into sedimentary basins and triggering short-lived mineralisation events. This mechanism provides a physical explanation for the temporal clustering of giant sediment-hosted deposits during supercontinent breakup phases, when rollback-driven extension and back-arc processes were widespread.Together, the convergence of global reconstructions, statistical analyses, and geodynamic modelling demonstrates that subduction is a fundamental driver of craton-margin metallogenesis. By quantifying a predictive distance window linking subduction, mantle flow, and lithospheric weakening, this study provides a physically grounded framework for mineral exploration and reveals how deep-Earth dynamics regulate the long-term distribution of critical metal resources throughout Earth history.

  • Open Access

    Deforming-Plate Reconstructions Reveal Tectonic Regime Controls on Porphyry Copper Formation in the Arabian-Eurasian Tethyan Belt

    by Elnaz Heidari, Sabin Zahirovic, Ehsan Farahbakhsh, Brent I. A. McInnes, Fabian Kohlmann, Maria Seton, R. Dietmar Müller

    The western–central Tethyan belt, capturing the Arabian-Eurasian convergence zone, preserves a complex record of subduction, rifting, and continental collision, yet much of this complexity is simplified or omitted in existing global plate reconstructions. This limits our ability to robustly link tectonic evolution with the timing and setting of magmatic–hydrothermal systems such as porphyry copper deposits. Here, we present a new plate reconstruction model of the Arabian-Eurasian convergence system since the latest Jurassic that incorporates both rigid and deforming plates. Our approach re-evaluates the boundaries between continental and oceanic crust, pre-collision margin geometries, and plate affinities by integrating regional geological, geophysical, and geochronological constraints. This integration results in a revised kinematic history that incorporates previously unresolved plate-boundary configurations and deformation patterns into the reconstruction. The model spans 130 Ma to the present, beginning with the initial Iran–Eurasia collision. A key revision occurs in the eastern and northeastern domain, where the Sistan–Sabzevar system is reconstructed as an oceanic basin that opened through rifting between the Lut block and Eurasia, followed by ridge spreading and the development of double-sided subduction systems, culminating in the initial Iran–Eurasia collision at ~54 Ma. West and southwest of the Lut block, Arabia–Iran convergence involves a Cretaceous phase of rifting along the northern Arabian margin, followed by ophiolite obduction and sustained subduction south of the Lut block, before evolving into a sequence of diachronous continental collision stages, beginning with an initial soft collision at ~42 Ma and followed by three hard collision phases at ~35, ~26, and ~16 Ma. Along the southern margin, the Makran region is reconstructed as a long-lived subduction system linked to a migrating island-arc complex that remains active to the present day. Together, these revisions produce a kinematically consistent reconstruction that better reconciles basin development and regional structural shortening. Building on this revised rigid framework, we implement deforming plate networks to represent the distributed strain associated with rifting and continental collision. Time-dependent stretching factors derived from these meshes allow us to distinguish intervals dominated by extension from those dominated by shortening. Comparison with the spatio-temporal distribution of porphyry Cu mineralisation shows that ~62% of deposits formed during extensional or transtensional regimes, whereas ~38% are associated with compressional or transpressional conditions. These results demonstrate that porphyry systems develop across a broader range of tectonic states than commonly assumed. By explicitly linking plate-boundary evolution with time-dependent lithospheric deformation, this reconstruction provides a more realistic framework for evaluating the tectonic controls on magmatic–hydrothermal systems in convergent margins.

  • Open Access

    Reply on RC2

    by Zijing Luo, Ehsan Farahbakhsh, Stephen Hore, R. Dietmar Müller

    The clean energy transition demands a significant increase in exploration for critical minerals, particularly rare earth elements (REEs), beyond the well-explored surface deposits. Discovery rates have been declining for decades, escalating the need for new exploration methodologies. Deep learning (DL) utilizes multi-layer neural networks to automatically model high-level abstractions of data, extracting information relevant to the target task, thereby positioning itself as a potentially powerful tool for mineral prediction. But the non-linear and highly heterogeneous characteristics of complex exploration data sets paired with sparse, imbalanced training data and a lack of interpretability represent significant challenges. In this study, we have developed DEEP-SEAM v1.0, a novel explainable semi-supervised DL framework for prospectivity mapping of REE mineralisation in Northern Curnamona Province, South Australia. This framework proposes a comprehensive data preprocessing pipeline and introduces a semi-supervised anomaly detection DL model, termed the Deviation Network (DevNet). DevNet leverages a limited number of positive samples alongside a large number of unlabelled samples to effectively establish the mapping between multi-source exploration data and mineralisation probability. The results indicate that prospective mineralisation areas exhibit a strong spatial coupling with known REE deposits, with high-probability mineralisation areas primarily concentrated in faulted regions, felsic granites, and Mesoproterozoic strata. To address concerns about the poor interpretability of DL models, we incorporate a post-hoc model interpretation technique known as the SHapley Additive exPlanations method. The method facilitates an improved understanding of the decision-making mechanisms and logic underlying DevNet. By comparing DEEP-SEAM’s decisions with our understanding of mineral systems, we not only enhance the model’s transparency and interpretability but also strengthen the reliability and credibility of the predicted prospective mineralisation areas.

  • Open Access

    Reply on RC1

    by Zijing Luo, Ehsan Farahbakhsh, Stephen Hore, R. Dietmar Müller

    The clean energy transition demands a significant increase in exploration for critical minerals, particularly rare earth elements (REEs), beyond the well-explored surface deposits. Discovery rates have been declining for decades, escalating the need for new exploration methodologies. Deep learning (DL) utilizes multi-layer neural networks to automatically model high-level abstractions of data, extracting information relevant to the target task, thereby positioning itself as a potentially powerful tool for mineral prediction. But the non-linear and highly heterogeneous characteristics of complex exploration data sets paired with sparse, imbalanced training data and a lack of interpretability represent significant challenges. In this study, we have developed DEEP-SEAM v1.0, a novel explainable semi-supervised DL framework for prospectivity mapping of REE mineralisation in Northern Curnamona Province, South Australia. This framework proposes a comprehensive data preprocessing pipeline and introduces a semi-supervised anomaly detection DL model, termed the Deviation Network (DevNet). DevNet leverages a limited number of positive samples alongside a large number of unlabelled samples to effectively establish the mapping between multi-source exploration data and mineralisation probability. The results indicate that prospective mineralisation areas exhibit a strong spatial coupling with known REE deposits, with high-probability mineralisation areas primarily concentrated in faulted regions, felsic granites, and Mesoproterozoic strata. To address concerns about the poor interpretability of DL models, we incorporate a post-hoc model interpretation technique known as the SHapley Additive exPlanations method. The method facilitates an improved understanding of the decision-making mechanisms and logic underlying DevNet. By comparing DEEP-SEAM’s decisions with our understanding of mineral systems, we not only enhance the model’s transparency and interpretability but also strengthen the reliability and credibility of the predicted prospective mineralisation areas.

  • Open Access

    Predicting the preservation of buried ore deposits using deep-time landscape evolution modeling

    by Addison Tu, Sabin Zahirovic, Sara Polanco, Samuel C. Boone, Matt Boyd, Claire Mallard, Pedro Restrepo, Youseph Ibrahim, Luke Mahoney, Tristan Salles, et al.

    Porphyry copper discoveries are declining despite rising demand to meet net-zero targets, highlighting the need for innovative exploration strategies. While many advances have focused on ore formation at depth, a major challenge remains in understanding how erosion and uplift over millions of years affect deposit preservation. These postmineralization processes determine whether porphyry systems are exposed, buried, or eroded entirely. We present a physically based landscape evolution model that incorporates spatially variable erodibility, dynamic uplift histories, climate and sea level change, and evolving topography over geological timescales. This richer input data, combined with tighter calibration, enables quantification of preservation potential and marks a step beyond prior conceptual and time-static models. We apply the model to New Guinea’s geologically complex mountains and integrate it with machine learning–derived ore formation probabilities. The combined model predicts known porphyry endowment, identifies new targets, and constrains preservation likelihood, validating this open-source method as a flexible and affordable exploration tool in dynamic tectonic settings.


  • Open Access

    Reply on CEC1

    by Zijing Luo, Ehsan Farahbakhsh, Stephen Hore, R. Dietmar Müller

    The clean energy transition demands a significant increase in exploration for critical minerals, particularly rare earth elements (REEs), beyond the well-explored surface deposits. Discovery rates have been declining for decades, escalating the need for new exploration methodologies. Deep learning (DL) utilizes multi-layer neural networks to automatically model high-level abstractions of data, extracting information relevant to the target task, thereby positioning itself as a potentially powerful tool for mineral prediction. But the non-linear and highly heterogeneous characteristics of complex exploration data sets paired with sparse, imbalanced training data and a lack of interpretability represent significant challenges. In this study, we have developed DEEP-SEAM v1.0, a novel explainable semi-supervised DL framework for prospectivity mapping of REE mineralisation in Northern Curnamona Province, South Australia. This framework proposes a comprehensive data preprocessing pipeline and introduces a semi-supervised anomaly detection DL model, termed the Deviation Network (DevNet). DevNet leverages a limited number of positive samples alongside a large number of unlabelled samples to effectively establish the mapping between multi-source exploration data and mineralisation probability. The results indicate that prospective mineralisation areas exhibit a strong spatial coupling with known REE deposits, with high-probability mineralisation areas primarily concentrated in faulted regions, felsic granites, and Mesoproterozoic strata. To address concerns about the poor interpretability of DL models, we incorporate a post-hoc model interpretation technique known as the SHapley Additive exPlanations method. The method facilitates an improved understanding of the decision-making mechanisms and logic underlying DevNet. By comparing DEEP-SEAM’s decisions with our understanding of mineral systems, we not only enhance the model’s transparency and interpretability but also strengthen the reliability and credibility of the predicted prospective mineralisation areas.

  • Open Access

    Comment on egusphere-2025-3283

    by Zijing Luo, Ehsan Farahbakhsh, Stephen Hore, R. Dietmar Müller

    The clean energy transition demands a significant increase in exploration for critical minerals, particularly rare earth elements (REEs), beyond the well-explored surface deposits. Discovery rates have been declining for decades, escalating the need for new exploration methodologies. Deep learning (DL) utilizes multi-layer neural networks to automatically model high-level abstractions of data, extracting information relevant to the target task, thereby positioning itself as a potentially powerful tool for mineral prediction. But the non-linear and highly heterogeneous characteristics of complex exploration data sets paired with sparse, imbalanced training data and a lack of interpretability represent significant challenges. In this study, we have developed DEEP-SEAM v1.0, a novel explainable semi-supervised DL framework for prospectivity mapping of REE mineralisation in Northern Curnamona Province, South Australia. This framework proposes a comprehensive data preprocessing pipeline and introduces a semi-supervised anomaly detection DL model, termed the Deviation Network (DevNet). DevNet leverages a limited number of positive samples alongside a large number of unlabelled samples to effectively establish the mapping between multi-source exploration data and mineralisation probability. The results indicate that prospective mineralisation areas exhibit a strong spatial coupling with known REE deposits, with high-probability mineralisation areas primarily concentrated in faulted regions, felsic granites, and Mesoproterozoic strata. To address concerns about the poor interpretability of DL models, we incorporate a post-hoc model interpretation technique known as the SHapley Additive exPlanations method. The method facilitates an improved understanding of the decision-making mechanisms and logic underlying DevNet. By comparing DEEP-SEAM’s decisions with our understanding of mineral systems, we not only enhance the model’s transparency and interpretability but also strengthen the reliability and credibility of the predicted prospective mineralisation areas.