RSS 1.0Earth Science and Engineering Program
For more information visit: https://pse.kaust.edu.sa/study/academic-programs/earth-science-and-engineering/Pages/home.aspx
Recent Submissions
-
Robust data driven discovery of a seismic wave equation(arXiv, 2023-09-24) [Preprint]Despite the fact that our physical observations can often be described by derived physical laws, such as the wave equation, in many cases, we observe data that do not match the laws or have not been described physically yet. Therefore recently, a branch of machine learning has been devoted to the discovery of physical laws from data. We test such discovery algorithms, with our own flavor of implementation D-WE, in discovering the wave equation from the observed spatial-temporal wavefields. D-WE first pretrains a neural network (NN) in a supervised fashion to establish the mapping between the spatial-temporal locations (x,y,z,t) and the observation displacement wavefield function u(x,y,z,t). The trained NN serves to generate meta-data and provide the time and spatial derivatives of the wavefield (e.g., u_tt and u_xx) by automatic differentiation. Then, a preliminary library of potential terms for the wave equation is optimized from an overcomplete library by using a genetic algorithm. We, then, use a physics-informed information criterion to evaluate the precision and parsimony of potential equations in the preliminary library and determine the best structure of the wave equation. Finally, we train the "physics-informed" neural network to identify the corresponding coefficients of each functional term. Examples in discovering the 2D acoustic wave equation validate the feasibility and effectiveness of D-WE. We also verify the robustness of this method by testing it on noisy and sparsely acquired wavefield data.
-
Transport Properties of Oil-Co2 Mixtures in Nanopores: Physics and Machine Learning Models(Elsevier BV, 2023-09-22) [Preprint]Fundamental understanding and quantitative models of the transport properties of oil-CO2 mixtures in nanopores are indispensable for physics-based models of CO2-enhanced oil recovery in unconventional oil reservoirs. This study determines the Maxwell-Stefan (M-S) diffusivities of CO2-decane (1: CO2; 2: decane /C10) mixtures in calcite nanopore with compositions relevant to CO2 Huff-n-Puff by molecular dynamics (MD) simulations. In the compositional space explored, D12 characterizing CO2-C10 interactions is relatively insensitive to composition, in contrast to that of bulk mixtures with similar compositions. D1,s characterizing CO2-wall interactions increases sharply with CO2 loading in the nanopore. In contrast, D2,s characterizing C10-wall interactions shows a nonmonotonic dependence on C10 loading. In addition, surprisingly, D2,s is negative, opposite to the expectations for dense fluid mixtures or pure decane confined in nanopores. These features of the M-S diffusivities can ultimately be traced to the fact that CO2 molecules adsorb far more strongly on pore walls than the C10 molecules, which leads to significantly heterogeneous distribution of CO2 and C10 in the nanopore and a low mobility of the adsorbed CO2 molecules. As MD simulations are computationally expensive, a non-parametric machine learning technique called the multitask Gaussian process regression method, is used to build a surrogate model to predict M-S diffusivities based on limited MD data. The surrogate model performs well in the compositional space it was trained with a relative root mean square error less than 10%.
-
Oil Spill Risk Analysis For The NEOM Shoreline(2023-09-21) [Preprint]A risk analysis is conducted considering an array of release sources located around the NEOM shoreline. The sources are selected close to the coast and in neighboring regions of high marine traffic. The evolution of oil spills released by these sources is simulated using the MOHID model, driven by validated, high-resolution met-ocean fields of the Red Sea. For each source, simulations are conducted over a 4-week period, starting from first, tenth and twentieth days of each month, covering five consecutive years. A total of 48 simulations are thus conducted for each source location, adequately reflecting the variability of met-ocean conditions in the region. The risk associated with each source is described in terms of amount of oil beached, and by the elapsed time required for the spilled oil to reach the NEOM coast, extending from the Gulf of Aqaba in the North to Duba in the South. To further characterize the impact of individual sources, a finer analysis is performed by segmenting the NEOM shoreline, based on important coastal development and installation sites. For each subregion, source and release event considered, a histogram of the amount of volume beached is generated, also classifying individual events in terms of the corresponding arrival times. In addition, for each subregion considered, an inverse analysis is conducted to identify regions of dependence of the cumulative risk, estimated using the collection of all sources and events considered. The transport of oil around the NEOM shorelines is promoted by chaotic circulations and northwest winds in summer, and a dominant cyclonic eddy in winter. Hence, spills originating from release sources located close to the NEOM shorelines are characterized by large monthly variations in arrival times, ranging from less than a week to more than two weeks. Similarly, large variations in the volume fraction of beached oil, ranging from less then 50\% to more than 80\% are reported. The results of this study provide key information regarding the location of dominant oil spill risk sources, the severity of the potential release events, as well as the time frames within which mitigation actions may need to deployed.
-
Source-encoded waveform inversion in the Northern Hemisphere(Geophysical Journal International, Oxford University Press (OUP), 2023-09-21) [Article]We use source-encoded waveform inversion to image Earth’s Northern Hemisphere. The encoding method is based on measurements of Laplace coefficients of stationary wavefields. By assigning to each event a unique frequency, we compute Fréchet derivatives for all events simultaneously based on one ‘super’ forward and one ‘super’ adjoint simulation for a small fraction of the computational cost of classical waveform inversion with the same dataset. No cross-talk noise is introduced in the process, and the method does not require all events to be recorded by all stations. Starting from global model GLAD_M25, we performed 100 conjugate gradient iterations using a dataset consisting of 786 earthquakes recorded by 9,846 stations. Synthetic inversion tests show that we achieve good convergence based on this dataset, and we see a consistent misfit reduction during the inversion. The new model, named SE100, has much higher spatial resolution than GLAD_M25, revealing details of the Yellowstone and Iceland hotspots, subduction beneath the Western United States, and the upper mantle structure beneath the Arctic Ocean.
-
Spatiotemporal variability of droughts over the Arabian Peninsula and associated mechanisms(Research Square Platform LLC, 2023-09-19) [Preprint]This study examines the spatiotemporal variability of drought and associated physical processes over the Arabian Peninsula (AP). For this purpose, we computed the standardized precipitation evapotranspiration index (SPEI) for the period 1951–2020 using the Climate Research Unit and ERA5 Reanalysis datasets. By applying rotated empirical orthogonal function analysis on the SPEI data, we identified four homogeneous and coherent drought regions. In comparison with the southern region, the droughts in the northern homogeneous regions were more significantly correlated. All four sub-regions of the AP exhibit a significant drying trend (p < 0.01) with an abrupt acceleration in drought frequency and intensity over the last two decades. The increase in droughts is associated with the reduction of synoptic activity and an increase in the high pressure over the AP. Seasonally, potential evapotranspiration is the dominant driver of summer droughts in the AP, whereas both precipitation and temperature are important for driving winter droughts. The summer droughts, mainly over the northern AP, are due to the occurrence of an anomalous equivalent barotropic high associated with anomalous dry and hot conditions. However, anomalous dry conditions in winter are a result of an anomalous paucity of winter storms caused by the weakening of the sub-tropical jets.
-
Effects of multi-observations uncertainty and models similarity on climate change projections(npj Climate and Atmospheric Science, Springer Science and Business Media LLC, 2023-09-16) [Article]Climate change projections (CCPs) are based on the multimodel means of individual climate model simulations that are assumed to be independent. However, model similarity leads to projections biased toward the largest set of similar models and intermodel uncertainty underestimation. We assessed the influences of similarities in CMIP6 through CMIP3 CCPs. We ascertained model similarity from shared physics/dynamics and initial conditions by comparing simulated spatial temperature and precipitation with the corresponding observed patterns and accounting for intermodel spread relative to the observational uncertainty, which is also critical. After accounting for similarity, the information from 57 CMIP6, 47 CMIP5, and 24 CMIP3 models can be explained by just 11 independent models without significant differences in globally averaged climate change statistics. On average, independent models indicate a lower global-mean temperature rise of 0.25 °C (~0.5 °C–1 °C in some regions) relative to all models by the end of the 21st century under CMIP6’s highest emission scenario.
-
Downscaling using CDAnet under Observational and Model Noises: The Rayleigh-Benard Paradigm(2023-09-13) [Preprint]Efficient downscaling of large ensembles of coarse-scale information is crucial in several applications, such as oceanic and atmospheric modeling. The determining form map is a theoretical lifting function from the low-resolution solution trajectories of an infinite-dimensional dissipative dynamical system to their corresponding fine-scale counterparts. Recently, Hammoud et al. (2022b) introduced CDAnet a physics-informed deep neural network as a surrogate of the determining form map for efficient downscaling. CDAnet was demonstrated to efficiently downscale noise-free coarse-scale data in a deterministic setting. Herein, the performance of well-trained CDAnet models is analyzed in a stochastic setting involving (i) observational noise, (ii) model noise, and (iii) a combination of observational and model noises. The analysis is performed employing the Rayleigh-Bénard convection paradigm, under three training conditions, namely, training with perfect, noisy, or downscaled data. The effects of observational and model noise on the CDAnet downscaled solutions are analyzed. Furthermore, the effects of the Rayleigh number and the spatial and temporal resolutions of the coarse-scale information on the downscaled fields are examined. The results suggest that the expected l 2 -error of CDAnet behaves quadratically in terms of the standard deviations of the observational and model noises. The results also suggest that CDAnet responds to uncertainties similar to CDA with an additional error overhead due to CDAnet being a surrogate model of the determining form map.
-
Introduction to the Special Section for the Centennial of the Great 1923 Kanto, Japan, Earthquake(Bulletin of the Seismological Society of America, Seismological Society of America (SSA), 2023-09-13) [Article]
-
Insights from very Large Ensemble Data Assimilation Experiments with a High Resolution General circulation model of the Red Sea(Authorea, Inc., 2023-09-11) [Preprint]Ensemble Kalman Filters (EnKFs), which assimilate observations based on statistics derived from samples of ocean states called ensemble, have become the norm for ocean data assimilation (DA) and forecasting. These schemes are commonly implemented with inflation and localization techniques to increase their ensemble spread and to filter out spurious long-range correlations resulting from the limited-size ensembles imposed by computational burden constraints. Such ad hoc methods were found not necessary in ensemble DA experiments with simplified ocean/atmospheric models and large ensembles. Here, we conduct a series of 1-year-long ensemble experiments with a fully realistic EnKF-DA system in the Red Sea using tens-to-thousands of ensemble members. The system assimilates satellite and in-situ observations and accounts for model uncertainties by integrating a 4km-resolution ocean model with ECMWF atmospheric ensemble fields, perturbed internal physics and initial conditions for forecasting. Our results indicate that accounting for model uncertainties is more beneficial than simply increasing the ensemble size, with the improvements due to large ensemble leveling off at about 250 members. Besides, and in contrast to what is commonly observed with simplified models, the investigated ensemble DA system still required localization even when implemented with thousands of members. These findings are explained by (i) amplified spurious long-range correlations produced by the low-rank nature of the ECMWF atmospheric forcing ensemble, and (ii) non-Gaussianity generated by the perturbed internal physical parameterization schemes. Large ensemble forcing fields and non-Gaussian DA methods might be needed to take full benefits from large ensembles in ocean DA.
-
Doublet Huff and Puff (Dhp): A New Technology Towards Optimum Scco2 Sequestration with Stable Geothermal Recovery(Elsevier BV, 2023-09-11) [Preprint]The heat energy extracted from the geothermal reservoirs is clean and plays an important role in decarbonizing the energy sector. Carbon emissions increase in day-to-day operations due to the utilization of hydrocarbons, which contributes to global warming. Therefore, it is crucial to capture and storeSCCO2 while developing clean energy technologies. In this study, we develop a new technology that stores SCCO2 while extracting clean energy from geothermal reservoirs. Our goal is to achieve sustainable thermal recovery and maximize SCCO2 storage capabilities in geothermal reservoirs. To investigate the thermal recovery and SCCO2 storage behavior of geothermal reservoirs, we use a thermo-hydro (two-phase) mechanical (THM) model. This technology is adapted from conventional Huff-Puff technology used in the hydrocarbon industry and applied using well pairs with different injection and perforation operating cycles. We also compared the numerical results with the CO2 plume geothermal (CPG) models. The levelized cost of energy (LCOE) analysis is conducted and compared with the CPG models. The numerical results show that the reduction in production temperature is less than 10 % of the original temperature (base case), the injected SCCO2 accumulates at the top of the reservoir, and the cold front progresses in the vicinity of wells over time. We also investigate the sensitivity of the rock and operating parameters on the heat power and the amount of SCCO2 stored. The implementation of DHP technology is more economical than CPG in geothermal reservoirs (LCOEDHP
-
Scaling the “Memory Wall” for Multi-Dimensional Seismic Processing with Algebraic Compression on Cerebras CS-2 Systems(ACM/IEEE, 2023-09-11) [Conference Paper]We exploit the high memory bandwidth of AIcustomized Cerebras CS-2 systems for seismic processing. By leveraging low-rank matrix approximation, we fit memoryhungry seismic applications onto memory-austere SRAM waferscale hardware, thus addressing a challenge arising in many wave-equation-based algorithms that rely on Multi-Dimensional Convolution (MDC) operators. Exploiting sparsity inherent in seismic data in the frequency domain, we implement embarrassingly parallel tile low-rank matrix-vector multiplications (TLRMVM), which account for most of the elapsed time in MDC operations, to successfully solve the Multi-Dimensional Deconvolution (MDD) inverse problem. By reducing memory footprint along with arithmetic complexity, we fit a standard seismic benchmark dataset into the small local memories of Cerebras processing elements. Deploying TLR-MVM execution onto 48 CS-2 systems in support of MDD gives a sustained memory bandwidth of 92.58PB/s on 35, 784, 000 processing elements, a significant milestone that highlights the capabilities of AIcustomized architectures to enable a new generation of seismic algorithms that will empower multiple technologies of our lowcarbon future.
-
Transport mechanisms of nocturnal surface ozone over Riyadh, Kingdom of Saudi Arabia(Atmospheric Environment, Elsevier BV, 2023-09-10) [Article]This study investigated ozone (O3) transport mechanisms and their impact on observed nocturnal surface O3 (NSO) enhancements over Riyadh, Kingdom of Saudi Arabia (KSA). We used O3 measurements for the summer of 2012 to examine the NSO enhancements across the city. We found that out of the 88 days for which observations were available, NSO enhancements could be observed on 38 days. The average difference in the NSO concentration between NSO-enhanced days and nonenhanced days was ∼15 ppb, which was statistically significant (p < 0.05). Our analysis of the regional surface emission sources, potential vorticity (PV), and back trajectories using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model indicated that out of 38 days, the NSO enhancements on 18 days (47%), 14 days (37%), 4 days (11%), and 2 days (5%) are due to local transport, long-range transport from highly polluted regions and from stratospheric intrusion regions located at a considerable distance from Riyadh, and a combination of long-range transport from stratospheric intrusion and highly polluted regions, respectively. We then investigated the relevance of each transport mechanism for the increased NSO concentrations during five protracted episodes, each marked by three consecutive days of NSO enhancement. Our results demonstrated that the local transport, long-range transport of O3-rich air masses from highly polluted regions to the east and northeast of the KSA, and local descent over Riyadh increased the average NSO concentration by up to 43%. Moreover, we found that long-range transport from stratospheric intrusion regions north of the KSA increased the average NSO concentration by up to 46%. The combination of long-range transport and stratospheric intrusion increased the average NSO concentration by ∼180%. These increases were usually associated with strong northwest regional winds flowing over highly polluted and major stratospheric intrusion regions.
-
Seismic image of the mantle transition zone beneath northeastern china: evidence for stagnant pacific subducting slab, lithospheric delamination, and mantle upwelling(Geophysical Journal International, Oxford University Press (OUP), 2023-09-01) [Article]We provide a comprehensive image of the mantle transition zone (MTZ) beneath northeastern China by performing Variable Bin Radius Stacking of receiver functions. A massive seismic dataset consisting of over 133 000 receiver functions recorded by 1 208 broadband stations is processed. Our results reveal fine-scale topography on the 410- and 660-km discontinuities defining the upper and lower bounds of the MTZ, lateral variations in the MTZ thickness, and slab interfaces within the MTZ. In particular, unambiguous images of the slab interfaces provide direct evidence for the presence of the stagnant Pacific subducting slab below the eastern portion of the study area. A widespread deepening of the 410-km discontinuity is consistent with a hot and wet low-velocity upper mantle resulting from dehydration of the stagnant slab. Prominent depressions are evident in the depth to the 660-km discontinuity, with a thickened MTZ associated with the cold stagnating slab. Localized uplifts of the 660-km discontinuity are possibly caused by partial melt under the slab. These features attest to the influence of the Pacific plate on the MTZ. Additionally, a pronounced upwarp on the 660-km interface with a thin MTZ agrees with a previously hypothesized mantle upwelling through a slab window, possibly triggered by the sinking of the stagnant slab. Moreover, the western part of the study region is characterized by alternating ups and downs of the 410-km interface, while the topography of the 660-km discontinuity is relatively flat. We propose the western region is dominated by foundering of delaminated lithospheric blocks that induced upward mantle return flows upon entrance into the MTZ.
-
Molecular insights into fluid-solid interfacial tensions in water + gas + solid systems at various temperatures and pressures.(The Journal of chemical physics, AIP Publishing, 2023-09-01) [Article]The fluid-solid interfacial tension is of great importance to many applications including the geological storage of greenhouse gases and enhancing the recovery of geo-resources, but it is rarely studied. Extensive molecular dynamics simulations are conducted to calculate fluid-solid interfacial properties in H2O + gas (H2, N2, CH4, and CO2) + rigid solid three-phase systems at various temperatures (298-403 K), pressures (0-100 MPa), and wettabilities (hydrophilic, neutral, and hydrophobic). Our results on the H2O + solid system show that vapor-solid interfacial tension should not be ignored in cases where the fluid-solid interaction energy is strong or the contact angle is close to 90°. As the temperature rises, the magnitude of H2O's liquid-solid interfacial tension declines because the oscillation of the interfacial density/pressure profile weakens at high temperatures. However, the magnitude of H2O vapor-solid interfacial tension is enhanced with temperature due to the stronger adsorption of H2O. Moreover, the H2O-solid interfacial tension in H2O + gas (H2 or N2) + solid systems is weakly dependent on pressure, while the pressure effects on H2O-solid interfacial tensions in systems with CH4 or CO2 are significant. We show that the assumption of pressure independent H2O-solid interfacial tensions should be cautiously applied to Neumann's method for systems containing non-hydrophilic surfaces with strong gas-solid interaction. Meanwhile, the magnitude of gas-solid interfacial tension increases with pressure and gas-solid interaction. High temperatures generally decrease the magnitude of gas-solid interfacial tensions. Further, we found that the increment of contact angle due to the presence of gases follows this order: H2 < N2 < CH4 < CO2.
-
A conjugate fault revealed by the destructive Mw 5.6 (November 21, 2022) Cianjur earthquake, West Java, Indonesia(Journal of Asian Earth Sciences, Elsevier BV, 2023-08-31) [Article]On 21 November 2022, a destructive earthquake (Mw 5.6) struck Cianjur, West Java, Indonesia, resulting in at least 321 deaths, damage to 47,000 buildings, and economic losses of up to 7.7 trillion Indonesian Rupiahs (∼US $546 million). Prior to this earthquake, the fault on which slip occurred had not been mapped, thus making further analysis crucial for assessing future seismic hazard in the region. We constructed a detailed earthquake catalogue, which spanned the period from 10 days before to 48 days after the mainshock, using waveform migration and stacking, followed by relative relocation using a double-difference method. Source mechanisms for selected aftershocks were estimated using waveform inversion. Our results show three clear foreshocks preceding the mainshock, while the aftershocks reveal the presence of a conjugate fault pair trending NNW-SSE with a length of ∼8 km and WSW-ENE with a length of ∼5 km. Directivity analysis highlights bilateral rupture of the main shock toward N20°E and N200°E, although based on the focal mechanism solutions, it is likely that there was some slip on the conjugate fault. Analysis of the Coulomb stress change induced by the mainshock shows that areas to the NNW and WSW experienced an increase in stress, consistent with the observed aftershock pattern. The nearby fault to the south (the Rajamandala Fault) experienced an increase in stress, which likely elevates the risk of it rupturing in the future.
-
Integrating U-nets into a Multi-scale Full-waveform Inversion for Salt Body Building(IEEE Transactions on Geoscience and Remote Sensing, Institute of Electrical and Electronics Engineers (IEEE), 2023-08-31) [Article]In salt provinces, full-waveform inversion (FWI) is most likely to fail when starting with a poor initial model that lacks the salt information. Conventionally, salt bodies are included in the FWI starting model by interpreting the salt boundaries from seismic images, which is time-consuming and prone to error. Studies show that FWI can improve the interpreted salt provided that the data have long offsets, and low frequencies, which is not always the case. Thus, we develop an approach to invert for the salt body starting from a poor initial model, limited data offsets, and the absence of low frequencies. We leverage deep learning to apply multi-stage flooding and unflooding of the velocity model. Specifically, we apply a multi-scale FWI using three frequency bandwidths.We apply a network after each frequency scale. After the first two bandwidths, the networks are trained to flood the salt, while the network after the last frequency bandwidth is trained to unflood it. We follow the unflooding step, with a final FWI. We verify the method on the synthetic BP 2004 salt model benchmark. We only use the synthetic data of short offsets up to 6 km and remove frequencies below 3 Hz. We also apply the method to real vintage data acquired in the Gulf of Mexico region. The real data lack frequencies below 6 Hz and the streamer length is only 4.8 km. With these limitations, we manage to recover the salt body and verify the result by using them to image the data and analyze the resulting angle gathers.
-
Adsorption of Gases on Fullerene-like X12Y12 (X = Be, Mg, Ca, B, Al, Ga, C; Y = C, Si, N, P, O) Nanocages(Energy & Fuels, American Chemical Society (ACS), 2023-08-30) [Article]Density functional theory calculations are carried out to investigate the adsorption behaviors of CO2, NO, CO, and NH3 on 12 fullerene-like X12Y12 (B12N12, Al12N12, Ga12N12, B12P12, Al12P12, Ga12P12, Be12O12, Mg12O12, Ca12O12, C12Si12, C12N12, and C24) nanocages. The molecular electrostatic potential (MESP) analysis suggests that, for example, for the B12N12, Al12N12, and Ga12N12 nanocages, the electron-rich regions are centered on the N atoms. The deepest MESP minimum (Vmin) values suggest that replacement of C atoms in C24 by XY units increases the electron-rich nature of the nanocage. Generally, CO2 is found to be physisorbed, while NH3 is chemisorbed on the X12Y12 nanocages. NO is found to be strongly adsorbed on the B12P12, Be12O12, Ca12O12, and C24 nanocages, and CO is strongly adsorbed on the B12N12, B12P12, Be12O12, and C24 nanocages. An important result is that the Vmin values of the X12Y12 nanocages are linearly proportional to their CO2 or NO adsorption energies. The quantum theory of atoms in molecules analysis suggests strong covalent interactions in the CO2/Ca12O12, NO/Ca12O12, NO/C24, CO/C24, and NH3/C24 systems.
-
Adsorption of hazardous gases on Cyclo[18]carbon and its analogues(Journal of Molecular Liquids, Elsevier BV, 2023-08-30) [Article]Density functional theory calculations were performed to study the adsorption behavior of hazardous gases (NH3, CO2, CO, and NO) on twelve adsorbents (cyclo[18]carbon (C18) and its analogues Al9N9, B9N9, C6B6N6, C12B3N3, C14B2N2, d-C14B2N2, C16BN, C17Si; cyclo[12]carbon (C12) and its analogues Al6N6 and B6N6). The molecular electrostatic potential (MESP) maps of C18 revealed its polyynic structure with alternating electron-rich and electron-deficient regions. The most negative-valued MESP point (Vmin) indicates that the replacement of carbon atoms of C18/C12 by BN/AlN/Si units increases the electron-rich environment in the molecules. In general, NH3 and CO2 are found to be physisorbed on C18/C12 and its analogues. An important result is that the Vmin of C18/C12 and its analogues is linearly correlated with the NH3 and CO2 adsorption energies. The quantum theory of atoms in molecules (QTAIM) results indicate that the interactions of NH3 and CO2 with C18/C12 and its analogues are non-covalent in nature. In general, CO and NO are found to be chemisorbed on C18/C12 and its analogues. In contrast to the cases of NH3 and CO2 adsorption, the Vmin of C18/C12 and its analogues is generally inversely related to the CO and NO adsorption energies. The QTAIM results indicate the strong covalent character of the bonding for CO/C18, CO/C6B6N6, CO/d-C14B2N2, CO/C12, NO/C18, NO/C17Si, and NO/C12 systems. The adsorption process significantly influenced the chemical potential and/or hardness, especially for CO– and NO-adsorbed C18/C12 and its analogues.
-
Self-Supervised Pre-training Vision Transformer with Masked Autoencoders for Building Subsurface Model(IEEE Transactions on Geoscience and Remote Sensing, Institute of Electrical and Electronics Engineers (IEEE), 2023-08-28) [Article]Building subsurface models is a very important but challenging task in hydrocarbon exploration and development. The subsurface elastic properties are usually sourced from seismic data and well logs. Thus, we design a deep learning (DL) framework using Vision Transformer (ViT) as the backbone architecture to build the subsurface model using well log information as we apply full waveform inversion (FWI) on the seismic data. However, training a ViT network from scratch with limited well log data can be difficult to achieve good generalization. To overcome this, we implement an efficient self-supervised pre-training process using a masked autoencoder (MAE) architecture to learn important feature representations in seismic volumes. The seismic volumes required by the pre-training are randomly extracted from a seismic inversion, such as an FWI result. We can also incorporate reverse time migration (RTM) image into the seismic volumes to provide additional structure information. The pre-training task of MAE is to reconstruct the original image from the masked image with a masking ratio of 75%. This pre-training task enables the network to learn the high-level latent representations. After the pre-training process, we then fine-tune the ViT network to build the optimal mapping relationship between 2D seismic volumes and 1D well segments. Once the fine-tuning process is finished, we apply the trained ViT network to the whole seismic inversion domain to predict the subsurface model. At last, we use one synthetic data set and two field data sets to test the performance of the proposed method. The test results demonstrate that the proposed method effectively integrates seismic and well information to improve the resolution and accuracy of the velocity model.
-
Inshimtu – A Lightweight In Situ Visualization “Shim”(Springer Nature Switzerland, 2023-08-25) [Book Chapter, Conference Paper]In situ visualization and analysis is a valuable yet under utilized commodity for the simulation community. There is hesitance or even resistance to adopting new methodologies due to the uncertainties that in situ holds for new users. There is a perceived implementation cost, maintenance cost, risk to simulation fault tolerance, potential lack of scalability, a new resource cost for running in situ processes, and more. The list of reasons why in situ is overlooked is long. We are attempting to break down this barrier by introducing Inshimtu. Inshimtu is an in situ “shim” library that enables users to try in situ before they buy into a full implementation. It does this by working with existing simulation output files, requiring no changes to simulation code. The core visualization component of Inshimtu is ParaView Catalyst, allowing it to take advantage of both interactive and non-interactive visualization pipelines that scale. We envision Inshimtu as stepping stone to show users the value of in situ and motivate them to move to one of the many existing fully-featured in situ libraries available in the community. We demonstrate the functionality of Inshimtu with a scientific workflow on the Shaheen II supercomputer.
