Low-enthalpy geothermal doublets for direct-use heating are highly sensitive to subsurface heterogeneity and operational uncertainty. This study quantifies these uncertainties for the Delft campus geothermal system using an integrated workflow that couples geological modeling with GPU-accelerated, high-spatial-resolution reservoir simulation. Ensembles of three-dimensional regional-scale facies models of the Delft Sandstone Member, with and without conditioning to Geothermie Delft (GTD) well data, were generated using object-based modeling and sequential indicator simulation. Porosity was modeled by sequential Gaussian simulation, and permeability was derived from a nuclear magnetic resonance-based porosity–permeability correlation calibrated to GTD well data, yielding higher permeability than core-based correlations for porosity below 15%. Lorenz coefficients indicate strong variability in property distributions, resulting in a wide spread of production temperatures. In total, 2000 geological realizations were simulated over 50 years using the GPU-enabled open-source Delft Advanced Research Terra Simulator (open-DARTS) under maximum and 2025-demand production schemes. Conditioning to GTD wells adds considerable data worth by constraining uncertainty in production well bottom-hole temperature (BHT) and pressure (BHP), while keeping injection pressure within Dutch regulatory limits. SIS models exhibit greater temperature and pressure variability than OBM models due to lower sand-body continuity. Despite large 80% confidence intervals, P50 production temperatures remain comparable for conditioned models. Distance-based generalized sensitivity analysis identifies net-to-gross ratio and the porosity–permeability correlation as dominant controls on thermal response. The 2025-demand scheme delays cold-front propagation. Results demonstrate that ensemble-based, GPU-accelerated high-spatial-resolution simulations enable robust and efficient uncertainty quantification for direct-use geothermal systems, highlighting the importance of well conditioning and reservoir heterogeneity characterization in constraining thermal responses. ... Read more

Borehole images are commonly used to characterize fractures in the subsurface, but interpretations however might be prone to interpreter bias. In this study, we compare five independent interpretations of a single borehole image log from the Geneva Basin, and assess the agreement/disagreement among interpreters regarding fracture frequency and orientation. We then relate the agreement/disagreement with the geology encountered by the borehole and the technical quality of the imagelog. The results show that certain lithologies such as bioclastic limestone favour consistent frequency interpretations. In general, there is little agreement on fracture orientation, but the highest agreement is where the dip of bedding surfaces is low ( ... Read more

Naturally fractured reservoirs are essential for subsurface energy production and storage. However, the complexity and uncertainty inherent to fracture network properties make it difficult to characterise fluid flow within them. This study presents an unsupervised machine learning workflow that constrains uncertainty by establishing a systematic link between the pressure transient response observed at the well and the underlying fracture network properties. We generate a geologically consistent ensemble of 4,850 discrete fracture networks (DFNs) and simulate pressure transient responses for the same geometries under three matrix-fracture permeability configurations. For each dataset, we group pressure derivative responses into geologically interpretable flow behaviour clusters using Dynamic Time Warping (DTW) based K-medoids clustering. The resulting cluster medoids provide representative pressure derivative responses that summarise the dominant flow regime sequence within each class. The workflow consistently identifies four stable clusters across all datasets, each characterised by a distinct and repeatable sequence of diagnostic flow regimes consistent with a bounded range of fracture network properties. Feature importance ranking and SHAP values derived from a random forest classifier show that fracture intensity, wellbore fracture length, and backbone fracture fraction provide the strongest geological controls on cluster separation and hence on the emergent diagnostic signatures. Comparing clusters across datasets shows that 64.1% of DFNs retain their cluster membership, indicating that the clustering structure is primarily controlled by DFN geometry. However, for a fixed DFN, the pressure transient response varies across matrix-fracture permeability configurations, producing systematic shifts in derivative levels and in the dominance of specific flow regimes. The representative pressure derivative responses associated with each flow behaviour cluster are therefore not identical across datasets and must be interpreted within the matrix-fracture permeability configuration. Overall, the proposed framework provides a basis to constrain geological uncertainty and prioritise high-impact parameters for data acquisition in naturally fractured reservoirs, thereby improving reservoir characterisation and decision making in the early appraisal stage. ... Read more

Mitigating bias in geological interpretations in the domain of structural geology and more specifically in seismic images and outcrop data is a well-known challenge. So far, the same issue in borehole image interpretation has rarely been addressed, even though these data are crucial for our knowledge of the subsurface. The Lower Cretaceous carbonate rocks of the Geneva Basin, Switzerland is a naturally fractured geothermal reservoir which has been penetrated by the GEo-01 well. Borehole images of this well were interpreted independently by five interpreters offering a unique database to compare and contrast. In this study we analyzed the impact of lithology and borehole image quality/complexity on the frequency and orientation of fracture picks acquired by the five interpreters. For the fracture frequency, we demonstrate that in GEo-01, the lithology sets the baseline for the level of agreement between the different interpreters, while image quality/complexity plays only a moderate role. On the other hand, the variability between interpretations of the fracture orientation appears mainly to be affected by the dip of strata, where a trend is observed between intervals with low bedding dip angles and more consistent feature orientation picking. As a recommendation for future exploration in the Geneva Basin and beyond, we suggest guidelines based on the geological context of the well in order to increase the value of borehole image interpretations. ... Read more

Fluvial clastic sequence is the typical reservoir rock to develop the project of Direct Use Geothermal Systems (DUGS). However, limited and unknown subsurface data bring challenges to accurately characterize the geology of the fluvial depositional environment and to capture the details of the sand body distribution and connectivity. In this work, we utilize the process-based modeling (PBM) approach (Flumy) and object-based modeling approach (OBM) (Fluvsim) to develop a robust framework to evaluate the heterogeneity representation of the fluvial depositional environment. An ensemble of geological models with different global net-to-gross ratios (N/G) is created using PBM and OBM. The generated high-resolution models aim to cover a wide range of N/G from 30% to 80% which is typical for Dutch geothermal sediments. For the given models, a static analysis using the Lorenz coefficient shows a large range of variability in heterogeneity levels for both PBM and OBM models. This wide range of heterogeneity levels leads to a broad variation in thermal breakthrough time during 100 years of thermal production. We also find that the OBM and PBM converge to nearly the same distribution of predicted thermal breakthrough time when the overall N/G is 80%. ... Read more

Sediments deposited by rivers constitute a key source for the water, energy, and raw materials that fuel our everyday lives. Finding and managing these resources require to predict the spatial distribution of those fluvial deposits in the subsurface. This remains a major challenge due to the scarcity and poor quality of subsurface data, but also because traditional modeling approaches struggle to reproduce the continuity of those deposits while conditioning those data. To help tackle this challenge, we introduce FluvDepoSet, a large dataset of synthetic 3D models of fluvial deposits. Contrary to common practice, the 20,200 samples in FluvDepoSet were generated using a landscape evolution model, CHILD. CHILD simulates the processes driving landscape evolution, which leads to a high computational cost but ensures that the resulting 3D models are consistent with geological principles. Those processes include river lateral migration, aggradation, and incision with overbank deposition to simulate the evolution of a meandering river over tens of thousands of years. While all the samples share the same basic setting, seven parameters controlling those processes are randomly drawn from uniform distributions for each sample. This results in the buildup of different stratigraphies through time, with coarse sediments deposited in point bars along the river and fine sediments deposited in the floodplain. Those stratigraphies are then transferred to a regular structured grid, which is common to all samples and includes the fraction of coarse sediments and the deposition time. These properties are stored in HDF5 files, each file corresponding to one sample. Thanks to its large number of samples, FluvDepoSet is well-suited for sensitivity analyses to better understand the impact of fluvial deposits on applications – for instance through subsurface flow and transport simulations – and for machine learning to better support subsurface characterization and decision-making – for instance through the generation of more plausible conditional models of fluvial deposits. ... Read more

Low-enthalpy geothermal doublets for direct-use heating are highly sensitive to subsurface heterogeneity and operational uncertainty. This study quantifies these uncertainties for the Delft campus geothermal system using an integrated workflow that couples geological modeling with GPU-accelerated, high-spatial-resolution reservoir simulation. Ensembles of three-dimensional regional-scale facies models of the Delft Sandstone Member, with and without conditioning to Geothermie Delft (GTD) well data, were generated using object-based modeling and sequential indicator simulation. Porosity was modeled by sequential Gaussian simulation, and permeability was derived from a nuclear magnetic resonance–based porosity–permeability correlation calibrated to GTD well data, yielding higher permeability than core-based correlations for porosity below 15%. Lorenz coefficients indicate strong variability in property distributions, resulting in a wide spread of production temperatures. In total, 2000 geological realizations were simulated over 50 years using the GPU-enabled open-source Delft Advanced Research Terra Simulator (open-DARTS) under maximum and 2025-demand production schemes. Conditioning to GTD wells reduces uncertainty in production bottom-hole temperature (BHT) and pressure (BHP) while maintaining injection pressure below Dutch regulatory limits. SIS models exhibit greater temperature and pressure variability than OBM models due to lower sand-body continuity. Despite large 80% confidence intervals, P50 production temperatures remain comparable for conditioned models. Distance-based generalized sensitivity analysis identifies net-to-gross ratio and the porosity–permeability correlation as dominant controls on thermal response. The 2025-demand scheme delays cold-front propagation. Results demonstrate that ensemble-based, GPU-accelerated high-spatial-resolution simulations enable robust and efficient uncertainty quantification for direct-use geothermal systems, highlighting the importance of well conditioning and reservoir heterogeneity characterization in constraining thermal responses. ... Read more

Fractures are ubiquitous in geological formations and can often have an impact on subsurface applications such as geothermal energy, groundwater management or CO2 storage. Quantifying the relationship between the uncertainties inherent to fracture networks and the corresponding flow behaviour for these applications remains an open challenge. Simulation studies that are based on outcrop analogues of fracture networks have yielded many new insights about heat and mass transfer in fractured geological formations but are restricted to a limited number of fracture network realizations, simplified assumptions about fracture network properties or deterministic models, making it difficult to analyse a wide range of uncertainties. This study introduces a flexible workflow that generates ensembles of geologically plausible fracture networks that can be based on statistical data from outcrop analogues. The fracture networks are generated using a computationally efficient approach that combines mechanical and statistical methods. The ensembles are then seamlessly linked to multi-purpose flow and transport simulations where the fractures are represented explicitly in a porous and permeable rock matrix. This approach can enable new uncertainty quantification methods, supported by machine-learning-based emulators, to analyse how fracture network properties, such as fracture intensity, fracture aperture or fracture orientation, influence heat and mass transfer in fractured geological formations. The workflow is illustrated using two classic example applications pertinent to fracture network modelling – one based on outcrop data to assess thermal behaviour in geothermal systems, and one synthetic study to analyse the transition from matrix-dominated to fracture-dominated flow – and released as open-source code. ... Read more

This study investigates the use of machine learning (ML) models as ready-to-use surrogate models for the Random Finite Element Method (RFEM) in a 3D context. The training set comprises 4000 RFEM realizations, with random fields covering multiple spatial correlation lengths as input and their corresponding factors of safety (FoS) as the target. Our results show that a convolutional neural network performs best for predicting the FoS compared to two other ML models - support vector regression and random forest - combined with principal component analysis. The best-performing model in 3D can conduct a stochastic analysis of 4000 simulations within seconds, compared to 83 days for a standard RFEM reliability analysis, highlighting the efficiency of the approach. ... Read more

Characterising fractures in geothermal reservoirs is crucial for understanding heat and fluid flow, as fractures control reservoir permeability. Due to data scarcity, estimating fracture network properties remains uncertain. Dynamic data, such as well tests, provides indirect insights into subsurface properties and workflows have been developed to illustrate how uncertainty in fracture data affects flow behaviour. However, they use simplified, randomly generated fracture geometries limiting their applicability to real-world scenarios. This study presents a machine learning workflow for characterizing fractured reservoirs using transient data, focusing on geothermal reservoirs. A comprehensive dataset of 5000 geologically consistent Discrete Fracture Networks (DFNs) was generated using GeoDFN and directly linked to MRST for simulations. The workflow then applies a k-medoids clustering approach, using dynamic time warping (DTW) as a distance metric, to cluster pressure responses with similar transient behaviour. We identified 18 distinct pressure behaviour. Linking clusters to fracture properties reveals that fracture intensity, aperture, and length have the most significant impact on pressure behaviour, while fracture set type was found to be the least important factor. Future work will extend this workflow to temperature transient data and apply advanced machine learning techniques for both forward and inverse modelling of fractured geothermal reservoirs. ... Read more

Reducing the uncertainty of reservoir characterization requires to better identify the small-scale structures of the subsurface from the available data. Studying the seismic response of meter-scale, stratigraphic heterogeneities typically relies on the generation of reservoir models based on outcrop examples and their forward seismic modelling. To bridge geological information and seismic modelling, these methods allocate values of acoustic properties, such as mass-density and P-wave velocity, according to discretized properties like layer-type lithology or facies units. This strategy matches the current workflow in seismic data inversion in industry, where modelling workflows are based on lithofacies distributions. However, from stratigraphic modelling, we know that meter-scale heterogeneities occur within certain facies and lithologies. Here, we evaluate the difference on the seismic response between allocating acoustic properties in a grain size–based, semi-continuous manner versus discretized manners based on lithology and facies classifications. To do so, we generate a reference geological simulation that we populate with acoustic properties, mass-density and P-wave velocity, using three different strategies: (1) based on grain size distribution; (2) based on facies distribution; and (3) based on lithology. The method we propose includes the generation of realistic geological simulations based on stratigraphic modelling and the transformation of its output into acoustic properties, honouring the intra-lithology and intra-facies, small-scale structures. We, then, generate seismic data by applying a forward seismic modelling workflow. The synthetic data show that the grain size–based simulation allows the identification of small-scale, stratigraphic heterogeneities, such as beds with strong density and velocity contrasts. These stratigraphic structures are smoothened or may completely disappear in the facies and lithology discretized simulations and, therefore, are not (well) represented in the synthetic seismic data. Recognizing meter-scale, stratigraphic heterogeneities is relevant for the characterization of the fluid flow in the reservoir. However, current discrete and lithology-based strategies in seismic inversion are not able to resolve such heterogeneities because real subsurface properties are not discrete properties but continuous, unless there are stratigraphic discontinuities such as erosional surfaces or faults. This research works towards a better understanding of the relationship between changes in these continuous properties and the observed seismic data by introducing greater complexity into the discretized geological simulations. Here, we use synthetic seismic images with the goal of eventually aiding in fine-tuning seismic inversion methodologies applied to real seismic data. One pathway is to foster the development of inversion approaches that can leverage stratigraphic modelling to get stronger geological priors and replace the standard but inadequate multi-Gaussian prior. ... Read more

This research focuses on investigating the relative performance of a range of machine learning algorithms, namely the artificial neural network, support vector machine, Gaussian process regression, random forest, and XGBoost, for predicting the undrained shear strength from cone penetration test data. This is to assess how machine learning could help us lower the need for laboratory test data. The training dataset compiles 526 data from 12 regions and the testing dataset consists of 20 data from a polder located close to Leiden in the Netherlands. In addition, k-fold and group k-fold cross-validation strategies are both applied to validate the models. The poor performance of the models during group k-fold cross-validation suggests that, while machine learning techniques can perform well when site-specific data are included during training, they struggle to generalize without site-specific data. This highlights the difficulty of capturing soil heterogeneity and suggests that either machine learning methods should be trained on specific sites for which some data are already available, or much larger training datasets are needed. ... Read more

Many stratigraphic features occur at a scale that is at the edge or below vertical seismic resolution. Thus, they cannot be directly observed in the seismic data, while still having an important effect on the fluid flow within the system. The better understanding of these sub-seismic scale features or heterogeneities can help decrease subsurface uncertainty. Here we present a novel method that integrates forward stratigraphic modelling, petrophysics, and geophysics to decipher the seismic imprint of heterogeneities in wave-dominated, shallow marine environments. The proposed three-stepped method starts with defining geology-related input parameters for BarSim, a stratigraphic forward modelling software that produces models that include stratigraphic architecture, grain size distribution, and facies distribution. Then, the geological data is translated, cell by cell, into petrophysical data (density, Vp, and Vs) using emphirical relationships. Finally, the forward seismic modelling is performed by combining a finite difference approach strategy and angle-dependent full wavefield migration to retrieve the angle gathers This method also allows the generation of large amounts of field-independent data suitable for machine learning applications. ... Read more

Understanding the impact of fractures on fluid flow is fundamental for developing geoenergy reservoirs. Pressure transient analysis could play a key role for fracture characterization purposes if better links can be established between the pressure derivative responses (p′) and the fracture properties. However, pressure transient analysis is particularly challenging in the presence of fractures because they can manifest themselves in many different p′ curves. In this work, we aim to provide a proof-of-concept machine learning approach that allows us to effectively handle the diversity in fracture-related p′ curves by automatically classifying them and identifying the characteristic fracture patterns. We created a synthetic dataset from numerical simulation that comprised 2560 p′ curves that represent a wide range of fracture network properties. We developed an unsupervised machine learning approach that can distinguish the temporal variations in the p′ curves by combining dynamic time warping with k-medoids clustering. Our results suggest that the approach is effective at recognizing similar shapes in the p′ curves if the second pressure derivatives are used as the classification variable. Our analysis indicated that 12 clusters were appropriate to describe the full collection of p′ curves in this particular dataset. The classification exercise also allowed us to identify the key geological features that influence the p′ curves in this particular dataset, namely (1) the distance from the wellbore to the closest fracture(s), (2) the local/global fracture connectivity, and (3) the local/global fracture intensity. With additional training data to account for a broader range of fracture network properties, the proposed classification method could be expanded to other naturally fractured reservoirs and eventually serve as an interpretation framework for understanding how complex fracture network properties impact pressure transient behaviour. ... Read more