Leonardo Azevedo

Critical raw material (CRM) exploration increasingly combines Earth Observation, geophysics, geochemistry, mineral prospectivity mapping, probabilistic modelling, and autonomous sensing. However, these capabilities are commonly developed or deployed as separate components, while the explicit coupling of observation, uncertainty-aware geological inference, and subsequent exploration decisions remains comparatively underdeveloped. This paper presents Closed-Loop Mineral Intelligence (CLMI) as a conceptual and methodological framework for organising these capabilities within an iterative Observe–Model–Decide process. The contribution does not propose a new stand-alone learning algorithm or claim an operationally validated exploration system. Instead, it formalises a reference architecture in which heterogeneous observations update probabilistic geological representations, uncertainty informs the prioritisation of subsequent observations, and expert oversight remains part of the decision loop. The framework is examined through an illustrative lithium-bearing pegmatite scenario that specifies the intended information flow, candidate data sources, modelling functions, decision points, and human intervention without reporting empirical performance. A validation roadmap is further defined to make the framework testable in future case studies, including requirements for benchmark baselines, uncertainty calibration, information-gain assessment, exploration-efficiency metrics, computational feasibility, and human–AI interaction. By distinguishing established technological components from the proposed system-level integration, the paper positions CLMI as a research framework and agenda for adaptive, uncertainty-aware mineral exploration rather than as a demonstrated replacement for current professional practice.

SUMMARY: Accurate source localization is critical for monitoring volcanic tremors, microseismic events and earthquakes. Compared to other source localization algorithms, waveform-based interferometric imaging exhibits superior robustness. It proves particularly effective when confronting weak signal phase signatures, noise interference and inaccurate velocity models. Despite shared methodological commonalities, high-resolution imaging and localization of volcanic tremors prove notably more challenging than other passive seismic tasks. This difficulty primarily arises from sparse monitoring networks, complex wavefield interactions with the subsurface and the indistinct phase signatures inherent in tremor signals. Consequently, waveform-based interferometric imaging tends to suffer from more severe blurring effects (i.e. poorer spatial resolution) when applied to volcanic tremor localization and compared against other application fields. To address these resolution limitations, we propose a maximum likelihood estimation-based iterative peeling interferometric imaging method. This approach synergizes the inherent noise robustness of interferometry with a rigorous iterative framework designed to deblur the source image. By constructing data-driven empirical wavelets and employing an adaptive whitening filter dynamically updated via the Sherman–Morrison formula, the method sequentially strips source responses from the residual wavefield. This adaptive procedure directly counters the energy smearing caused by sparse arrays, effectively suppressing sidelobe interference to achieve a highly focused source resolution. Synthetic tests based on the Katla scenario demonstrate that the proposed method yields significantly higher spatial resolution than existing techniques. Furthermore, it exhibits robust performance even under adverse array geometries, concurrent multisource interference and velocity model inaccuracies. Field application to the 2011 Katla data further validates this by recovering source locations consistent with known ice cauldrons.

Rising greenhouse gas emissions, particularly carbon dioxide (CO2) from human activities, are contributing to climate change. One way to reduce CO2 emissions to the atmosphere and to reduce the global warming rate is via the deployment of Carbon Capture, Utilization, and Storage (CCUS). CCUS is particularly vital for hard-to-abate industrial sectors.CO2 captured from combustion processes can be transported via pipelines, injected into the subsurface, and be permanently stored in geological formations. Although CO2 geo-storage is technically viable, high costs hinder its applicability and affect the scalability of the CCUS value chain. A primary contributor to these costs is the industry standard demanding high-purity CO2 streams. To evaluate and address this barrier, we present an integrated techno-economic model capable of assessing the full CCUS value chain, from industrial capture and pipeline transport to subsurface injection, under varying purity specifications for both Enhanced Oil Recovery (EOR) and dedicated geological storage scenarios. Applying this model to a typical example of a (hard-to-abate) industrial case study demonstrates that reducing CO2 purity from 100% to 80% yields a value-chain cost reduction of ~$16/tgas for EOR and ~$15/tgas for geostorage. These findings highlight that relaxing CO2 purity requirements is a practical step to improve economic feasibility without compromising the safety of the system. Hence, allowing decision-makers and industry stakeholders to accelerate commercial CCUS deployment.

Seismic oceanography re-uses legacy marine multichannel reflection data (MCS) giving it a new porpose for ocean knowledge enrichement. Targeting the water column, detailed MCS processing for high-resolution water column imaging facilitates new interpretations and ocean modelling technics. Seismic acoustic response is intrinsically conected to the water column propreties, such as temperature an salinity (e.g., Azevedo, L. et al., 2021), this brings the study of fine-scale ocean processes to another level giving the possibility of lateraly predicting their spatiotemporal continuity.To retrieve the water collumn reflections the MCS processing workflow requires specific steps that enhance the signal-to-noise ratio and preserve the true amplitude content (Duarte, et al., 2025) . Later, the resultant images can be interpreted within the context of expected and present ocean processes in the study region (Duarte, et al., 2024) and integrating measurements acquired in oceanographic campains.Since seismic oceanography signal depends on the water column propreties, seismic oceanography inversion enables the re-constrution of the ocean temperature and salinity distributions. However, seismic inversion is inherently non-linear, identical seismic responses can be originated from different combinations of temperature and salinity pairs. Consequently, the inversion problem is ill-posed and admits multiple solutions, requiring smart strategies to condition and solve the ambiguity (Azevedo, L. et al., 2021).Our results demonstrate a sucessful atempt on reconstructing the high-resolution temperature and salinty models for two Portuguese regions with stochastic seismic oceanography inversion. This aproach highlights how seismic oceanography can resolve fine-scale thermohaline structures, often neglected by conventional sampling technics and shades lighight into its spatial continuity while assessing the predictions of the uncertainty. The resulting models provide valuable insights confirming the potential of seismic data as a complement tool for oceanographic studies and encouragning its integration in further oceanographic studies.

Shelf-to-basin sediment transport plays a key role in the evolution of continental margins and the carbon cycle. Yet, the mechanisms linking oceanographic forcing to sediment resuspension and redistribution remain poorly quantified. Intermediate nepheloid layers (INLs) are widely recognized as substantial sediment conveyors in this context, but their formation mechanisms, morphologic imprint, and spatio-temporal evolution remain elusive in direct observations and process-based interpretation. Here, we present multidomain observations from the Eastern Levant Basin (ELB) that document the internal-wave-induced origin and seasonal recurrence of INLs. We integrate ultra-high-resolution (decimeter-scale) multichannel Seismic Oceanography (SO), high-resolution bathymetry, in situ conductivity-temperature-depth (CTD) profiles, oceanographic reanalysis products from the Copernicus Marine Service, and spaceborne Synthetic Aperture Radar (SAR) imagery, to link water-column stratification, internal wave activity, sediment resuspension, sediment transport, and their morphologic outcomes. These data sets span multiple years and capture the same processes during comparable seasonal stratification regimes, allowing assessment of process persistence rather than isolated events. Our results show that during periods of strong seasonal column stratification (i.e., Brunt-Vaisälä frequencies in the order of N ≈ 0.01 s⁻¹), shoaling internal waves promote sediment resuspension from the shelf edge and basinward transport. Seismic profiles reveal laterally continuous, gently inclined low-amplitude reflection packages with thicknesses of up to 10 meters that detach from the seabed and are interpreted as INLs flowing up to 10 km from the continental slope. This is confirmed by in-situ CTD measurements showing aligned water-column turbidity peaks up to 10 kilometers offshore the area of resuspension. Calculated internal wave (IW) beam angles relative to local slope show a clear correlation between transmissive (subcritical) zones, seafloor erosion, and locations of sediment detachment, while reflective (supercritical) areas show the appearance of sediment-wave patterns interpreted as upslope-migrating steps. Our results are consistent across different years and geophysical datasets. Co-located SAR imagery independently confirms the presence and orientation of internal wave packets during these periods.Together, these observations provide robust field-based evidence that internal wave-driven sediment mobilization is a seasonally recurrent process, governed by water-column stratification and seafloor criticality. These observations link sediment transport, oceanographic dynamics, and slope geomorphology in an increasingly stratified, warming ocean.

With a focus on geo-modeling applications for sustainable deep mineral exploration, we propose affordable, but still accurate subsurface modeling technique that can generates realistic 3-D geological models. Conventional geostatistical methods based on two-point statistics, often compromise their performance in deep and structurally complex geological settings mostly due to limitation in modelling complex spatial continuity patterns. On the other hand, deep generative modelling techniques, such as generative adversarial networks (GAN), allow to predict complex spatial patterns but have difficulties to create large-scales models in three-dimensions and be locally conditioned by observations.We introduce a deep generative framework that adapts conditional GANs with spatially adaptive normalization (cGAN–SPADE) for 3-D geological modeling under sparse and evolving data conditions to predict high resolution subsurface models with real-time data assimilation capabilities. The goal is to generate geo-models based on a priori geological information (i.e., expected geometries and probability maps) with real-time model update as new data are acquired during drilling.The cGAN-SAPDE is trained with samples based on prior geological knowledge and existing borehole experimental data. Training proceeds through a generator and discriminator scheme in which generator produces new models based on input training data while the discriminator output is the probability of input image being real based on the corresponding conditioning map.A conditioning map is introduced at each generator’s layer, where it modulates the intermediate activations using SPADE normalization. This mechanism injects spatially varying conditioning information into the network, enabling the generator to preserve structural coherence and fine-grained spatial details in the synthesized outputs.Experimental results on industry-standard challenging 3-D synthetic data sets show the ability of the network to predict high-resolution 3-D geological models that simultaneously match a priori information and direct measurements acquired in real-time scenario.This project has received funding from the European Union’s Horizon Europe Research and Innovation Program under the Grant Agreement No.101178775

CO2 geostorage is a well-established solution for long-term carbon storage towards carbon neutrality goals. Injected CO2 streams are often captured at fossil fuel plants and/or oil and gas fields and consequently contaminated with impurities. These impurities require an energy-intensive purification process, thereby increasing costs. This high cost is considered among the main barriers to effective CCS deployment. One way to reduce the costs is by reducing the purity of the injected CO2 stream without compromising the safety and performance of the operation. While some impurities enhance storage efficiency, others reduce it. To understand this effect, determining the CO2 stream purity requires the study of the impact of impurities on the CO2 plume spatiotemporal evolution. To assess these impacts, we propose herein a geo-modelling workflow using compositional fluid simulation in porous media to evaluate the effects of impurities (N2, O2, NO2, SO2, H2S, CO, and H2O). We use a simplified version of the Illinois Basin Decatur Project reservoir model for computational efficiency. Monte Carlo simulation is applied for sensitivity analysis of a wide range of impurities, allowing the identification of impurity mixtures from hard-to-abate process plants that have good injectivity and long-term CO2 storage. Our numerical examples show that non-condensable gases, such as N2, O2, and H2S, significantly lead to lower injectivity and high dispersion of the CO2 plume due to lower stream density. In contrast, H2O has a positive impact on injectivity and storage, while NO2, SO2, and CO have negligible effects and can be tolerated within specific concentration ranges. The study also highlights that some compositions of impurities have less significant impact on injection and on the spread of the gas plume.

A central aspect for understanding the CO ₂ geostorage safety in subsurface geological formations is the reliable prediction of the CO ₂ plume spatiotemporal evolution after carbon dioxide injection. This objective is often achieved through detailed numerical fluid flow simulations, and reliable predictions depend on the grid cell size of the reservoir grid and the fluid modeling approach parameterizations. We assess herein the combined impact of grid cell size, reservoir boundaries, and fluid modeling on the trade-offs between computational costs and the numerical simulation accuracy of CO ₂ storage. To reach these objectives, we use a high-resolution three-dimensional reservoir benchmark model from the Illinois Basin Decatur Project. First, we crop and upscale the original high-resolution model. Second, we apply different fluid flow modelling approximations. While model reduction and upscaling result in a decrease of the reservoir porosity and permeability variability, the use of an appropriate fluid flow model can maintain simulation precision under varying grid conditions, highlighting the importance of balancing grid size and resolution with computational demands for reliable CO ₂ storage predictions.