3D parallel surface-borehole TEM forward modeling with multiple meshes

As an efficient geophysical exploration method, the time-domain airborne electromagnetic (AEM) data often show sign reversal in late-time channels due to induced polarization (IP) effect. The traditional imaging and inversion methods without considering the IP effect cannot recover the true electrical structure of the earth, so it is necessary to develop 3D EM forward modeling and inversion techniques with IP effect. In this paper, we propose a 3D forward modeling method for time-domain AEM with IP effect based on unstructured finite-element (FE) method. To describe the IP effect of a medium, we introduce the Cole-Cole model and transform it into fractional derivative form using frequency-time conversion. Then, we discretize it using the difference discretization format of Caputo fractional derivative. Finally, we use the vector FE method based on unstructured tetrahedral mesh and unconditionally stable second-order backward Euler’s scheme to discretize Maxwell’s equations in space and time. In this way, we can solve the 3D forward modeling problem for AEM with IP effect in time domain. We verify the accuracy of our algorithm by comparing it with the 1D semi-analytical solution for a half-space model, and then calculate EM responses for typical abnormal models and analyze the characteristics of IP effect.

Objective: Due to the recent development of multiple parallel programming tools with varying features, it is difficult to choose the best tool according to the needs of the user. Methods: This problem is addressed by making a comparative analysis study of different features like license type, source code availability, targeted platforms and languages supportedby these diverse tools. There are different parallel programming languages that support the present multi-core architecture like Message Passing Interface (MPI) and Open Multi-Processing (OpenMP). These are widely used to provide different performance characteristics of parallelism in different test cases. The new architecture and the complexity strengthens the need to monitor and analyze the performance of the various OpenMP kernels and constructs on multi-core processors. Findings: There are many papers that have been published in the past but non of them focuses on a comparative study among the performance analysis tools (PATs) that we mostly opt for. This paper intends to analyze the parallel computing ability of OpenMP and MPI, besides helping the user to understand which tool suites his task the best. Improvement: This study shows that MPI offers the best performance characteristics in the field of shared memory programming whereas OpenMP is a better choice because of the global style of the resulting program. It also provides a roadmap to select the best tool when designing a parallel programming system.

Many high-performance distributed memory applications rely on point-to-point messaging using the Message Passing Interface (MPI). Due to the latency of the network, and other costs, this communication can limit the scalability of an application when run on high node counts of distributed memory supercomputers. Communication costs are further increased on modern multi-and many-core architectures, when using more than one MPI process per node, as each process sends and receives messages independently, inducing multiple latencies and contention for resources. In this paper, we use shared memory constructs available in the MPI 3.0 standard to implement an aggregated communication method to minimize the number of inter-node messages to reduce these costs. We compare the performance of this Minimal Aggregated SHared Memory (MASHM) messaging to the standard point-to-point implementation on large-scale supercomputers, where we see that MASHM leads to enhanced strong scalability of a weighted Jacobi relaxation. For this application, we also see that the use of shared memory parallelism through MASHM and MPI 3.0 can be more efficient than using Open Multi-Processing (OpenMP). We then present a model for the communication costs of MASHM which shows that this method achieves its goal of reducing latency costs while also reducing bandwidth costs. Finally, we present MASHM as an open source library to facilitate the integration of this efficient communication method into existing distributed memory applications.

We have developed a technique to calculate lateral density distribution of the sedimentary basin basement by combining linear gravity–density inversion and 2D forward modeling. The procedure requires gravity anomaly data, depth-to-basement data and density data for the sediments (density–depth distribution). Gravity effect of density variations in the basement was extracted from the total gravity anomaly by removing the joint effect of the sediments with vertical density variations and homogeneous basement of average density contrast (calculated by 2D modeling). Gravity effect of the sediments was calculated using depth-to-basement data and density–depth function, based on borehole logging data. Bouguer slab formula was used in order to simplify basement model and calculate the operator for the linear inversion of gravity effect of the basement into lateral density distribution. The results were tested by forward modeling, and differences between observed and modeled gravity data were used for the next iteration of the inversion and correction of density values in each point along the profile. The procedure was tested using synthetic models and field example, and both results were satisfying, especially taking into account the simplicity of the inversion technique. The main problem was the effect of abrupt changes in the basement topography on density distribution, but it was downsized by filtering. Basement density maps were compiled based upon the density distribution along the profiles.

Inverse and 3D forward gravity modelling for the estimation of the crustal thickness of Egypt

Structural and stratigraphic basin modelling has been used to examine quantitatively the Cretaceous and Tertiary post-rift subsidence history of the northern North Sea Basin following Late Jurassic extension. 2D forward and reverse, syn-rift and post-rift modelling has been used to determine the magnitude and timing of departures from the McKenzie post-rift thermal subsidence trend for regional stratigraphic profiles in the Outer Moray Firth, the South Viking Graben and the North Viking Graben. Observed stratigraphy has been reverse post-rift modelled using flexural backstripping combined with decompaction and reverse thermal subsidence calculations, and forward modelled through the syn-rift and post-rift period using the flexural cantilever model of continental rift basin formation. Long-term global eustasy has been included in the analysis. Only reliable palaeobathymetric markers, such as erosion surfaces and coals, have been used to constrain subsidence history. Both forward and reverse 2D modelling show that a Palaeocene regional uplift followed by a rapid Eocene subsidence must be superimposed on McKenzie post-rift Cretaceous and Tertiary basement subsidence in order to generate the stratigraphy observed. The magnitude of Palaeocene uplift deduced from modelling was of the order of 375 m in the Outer Moray Firth (central North Sea) increasing to 525 m in the North Viking Graben (northern North Sea). Rapid Early Eocene subsidence was of the order of 160–310 m. The most likely mechanism for regional Palaeocene uplift and rapid Eocene subsidence in the northern North Sea was the Palaeocene development of the Iceland plume situated at a distance of 700–900 km from the study area in the Early Tertiary. Forward and reverse modelling predict axial bathymetries in the graben of up to 1300 m for the Early Cretaceous and 900 m for the Upper Cretaceous. The large palaeobathymetries at the end of the Cretaceous provided much of the accommodation space for Tertiary sedimentation.

A hybrid parallel scheme for the ensemble square root filter (EnSRF) suitable for parallel assimilation of multiscale observations, including those from dense observational networks such as those of radar, is developed based on the domain decomposition strategy. The scheme handles internode communication through a message passing interface (MPI) and the communication within shared-memory nodes via Open Multiprocessing (OpenMP) threads. It also supports pure MPI and pure OpenMP modes. The parallel framework can accommodate high-volume remote-sensed radar (or satellite) observations as well as conventional observations that usually have larger covariance localization radii. The performance of the parallel algorithm has been tested with simulated and real radar data. The parallel program shows good scalability in pure MPI and hybrid MPI–OpenMP modes, while pure OpenMP runs exhibit limited scalability on a symmetric shared-memory system. It is found that in MPI mode, better parallel performance is achieved with domain decomposition configurations in which the leading dimension of the state variable arrays is larger, because this configuration allows for more efficient memory access. Given a fixed amount of computing resources, the hybrid parallel mode is preferred to pure MPI mode on supercomputers with nodes containing shared-memory cores. The overall performance is also affected by factors such as the cache size, memory bandwidth, and the networking topology. Tests with a real data case with a large number of radars confirm that the parallel data assimilation can be done on a multicore supercomputer with a significant speedup compared to the serial data assimilation algorithm.

The results of the 3D finite-difference forward modeling of magnetotelluric data acquired around the Serra da Cangalha impact crater located northeastern Brazil, South America are presented. The 13 km impact structure has been investigated because of its explicit 3D shape. The central part of the model covers an area of approximately 169 Km2. The model of the study area, constructed using a 3D mesh of 22 x 19 x 25 (29325 cells) was based on the a priori information of the geology of the region, and the results of the 2D inversion results. Eleven frequencies were calculated for the 3-D model (0.01, 0.03, 0.1, 0.3, 1, 3, 10, 30, 100, 300, 1000 Hz). The resistivity structure obtained suggests perturbations of the upper crust due meteorite impact; the base of the uplifted structure extending to 3.0 km depth was well delineated. The observed 3D effects are explained as a shallow resistive ring that represents a local 3D body embedded in a regional layered earth. The results confirm the importance of 3D forward modeling in impact crater studies. The three-dimensional model results indicate that a valuable picture of the regional electrical resistivity can be obtained from 3-D interpretation of the determinant impedance tensors using polar diagrams. The study has helped in providing an insight into the post-impact electrical signature of the upper crust.

This project was designed as a broad foundational study of spectral induced polarization (SIP) for characterization of contaminated sites. It encompassed laboratory studies of the effects of chemistry on induced polarization, development of 3D forward modeling and inversion codes, and investigations of inductive and capacitive coupling problems. In the laboratory part of the project a physico-chemical model developed in this project was used to invert laboratory IP spectra for the grain size and the effective grain size distribution of the sedimentary rocks as well as the formation factor, porosity, specific surface area, and the apparent fractal dimension. Furthermore, it was established that the IP response changed with the solution chemistry, the concentration of a given solution chemistry, valence of the constituent ions, and ionic radius. In the field part of the project, a 3D complex forward and inverse model was developed. It was used to process data acquired at two frequencies (1/16 Hz and 1/ 4Hz) in a cross-borehole configuration at the A-14 outfall area of the Savannah River Site (SRS) during March 2003 and June 2004. The chosen SRS site was contaminated with Tetrachloroethylene (TCE) and Trichloroethylene (PCE) that were disposed in this area for several decades till the 1980s. The imaginary conductivity produced from the inverted 2003 data correlated very well with the log10 (PCE) concentration derived from point sampling at 1 ft spacing in five ground-truth boreholes drilled after the data acquisition. The equivalent result for the 2004 data revealed that there were significant contaminant movements during the period March 2003 and June 2004, probably related to ground-truth activities and nearby remediation activities. Therefore SIP was successfully used to develop conceptual models of volume distributions of PCE/TCE contamination. In addition, the project developed non-polarizing electrodes that can be deployed in boreholes for years. A total of 28 of these electrodes were deployed at the SRS site in September of 2002. The project found that (1) currently available field instrumentation need to be faster by an order of magnitude for full SIP to be engaged for broadband characterization in the field, (2) some aspects of the capacitive coupling problem in borehole geometries can be solved by use of a high impedance receiver, (3) a careful investigation of ways to adequately compare inversion results to ground-truth data is warranted, (4) more laboratory studies should be directed to understand the influence of micro-organisms and long residence time of contaminants (aging) on spectral IP properties.

Identifying the lithospheric structure of a Precambrian orogen using magnetotellurics: The Capricorn Orogen, Western Australia

We present a novel methodology for integration of high-angle/horizontal (HA/HZ) well data into 3D geomodels as a natural extension to well placement workflows. Log interpretation, typically done in 2D cross-sections, is based on 1-D automated inversion, yielding near-wellbore reservoir structure and properties. 2D cross-section, updated by inversion and further refined using 2D/3D modeling, is subsequently retrofitted into the geomodel so as to minimally perturb the original topology. Changes in positions/dips/azimuths of boundaries and faults, cell properties, and further local grid refinements are applied automatically. The updated geomodel honors high-resolution logs and low-resolution seismic/nearby wells data. We demonstrate this on a typical real-time well placement scenario. Electromagnetic log interpretation codes are integrated as a high performance computing (HPC) Web service into a geosteering/model update workflow. As the initial model, we use 3D geomodel constructed from seismic/vertical well data. A "curtain" cross-section is extracted, edited based on 1D inversion, and further refined to match HA/HZ logs through 2D and 3D forward modeling, by changing properties/dips/layer thicknesses /fault positions, while preserving the original topology. Then, updated node coordinates and cell properties of the affected pillar grid region are calculated to optimally retrofit the changed 2D cross-section into the grid. These changes are then automatically applied to the 3D model. For quality control, we recompute the 2D cross-section from the refined geomodel. Ultimately, we arrive at the geomodel that honors both seismic and resistivity well-log data. The combination, in a single workflow, of physics-based log modeling codes, Services-Oriented Architecture, HPC framework, and the solver to optimally retrofit 2D cross-sections into 3D models, creates a qualitatively new opportunity for well placement engineers. This integrated workflow (1) maximizes the value of deep directional resistivity well-logs and real-time well placement interpretation by incorporating them into the source of data for building geomodels; (2) radically speeds up the model refinement loop by automatically calculating and applying the modifications to 3D reservoir model; (3) enables geoscientists to directly refine geomodels while geosteering. The latter has not been a standard practice, hindered by challenges of scale difference between geomodels and well-logs and lack of availability of efficient modeling codes.

Fast Fourier Transform (FFT) is an important part of many applications, such as in wireless communication based on OFDM (Orthogonal Frequency Division Multiplexing). With Cloud Radio Access Networks, implementing FFTs on multiprocessor clusters is a challenging task. For instance, supporting the Long Term Evolution (LTE) protocol requires processing 100 independent FFTs (with sizes ranging from 128 to 2048 points) in 66.7 μs. In this work, seven native FFT candidate implementations are compared. The considered implementation environments are: OpenMP (Open Multi-Processing) on 1 core, MPI (Message Passing Interface) on 1 core, 2 cores, and 3 cores, Hybrid OpenMP+MPI on 1 core and 3 cores, and MPI on an heterogeneous platform composed of Xeon-Phi and 3 cores. The reported experimental results show that the latter method meets the latency requirements of LTE. It is shown that the OpenMP and MPI paradigms running only on MICs (Many Integrated Cores) cannot benefit fully from the computing capability of many-core architectures. The heterogeneous combination of Xeon+MICs provides a better performance.

SUMMARY While the forefront of AEM research is focusing on the challenges of 3D modelling, the wide AEM community still rely on less sophisticated computational techniques for their calculations. Inversion of large time domain AEM surveys still prove a computational challenge within a 1D formulation, and require much more computational resources than can be delivered by an office workstation. Emerging Monte-Carlo based 1D Bayesian inversion schemes provide another example of applications that are currently limited by the 1D forward modelling rate. In this abstract we describe our research in modifying the AarhusInv AEM inversion code to utilize next generation massively parallel co-processors. While our results are early and based on very little optimization, we still achieve comparable levels of performance (>80%) from a single co-processor and a 48 cpu core server. We estimate that performance on the co-processor can be speeded up by approximately another 4x with a limited amount of code restructuring/rewriting.

Hybrid Message Passing Interface (MPI) and Open Multi-Processing (OpenMP) parallel programs are pivotal for scalability and efficiency in high-performance computing (HPC), especially as systems approach exa-scale operations. These programs leverage distributed and shared memory systems, making them indispensable for scientific simulations, data analysis, and numerical computations. However, synchronization defects such as deadlocks and race conditions pose significant challenges to reliability and performance, often eluding traditional static and dynamic analysis tools due to the complexity of hybrid systems. This study introduces a deep learning-based models for automated defect prediction in hybrid MPI and OpenMP programs. Using a balanced dataset of 1,500 C++ files, three neural architectures—Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and a hybrid CNN-LSTM model—were evaluated. To preprocess the code, Abstract Syntax Tree (AST)-based token extraction was employed, capturing structural and semantic relationships critical for detecting defects. Token extraction methods included detailed C++ syntax analysis via the Clang library and a custom regular expression approach. The results reveal that Clang-token-based representation provided the most effective input for defect prediction, enabling CNN models to achieve an accuracy of 97%. By integrating AST-based structural insights with the predictive power of deep learning, this research offers a scalable solution for enhancing the reliability of hybrid parallel programs. The findings establish a foundation for future advancements in automated defect detection, addressing the pressing needs of high-performance computing systems.

The demand of processing speed in modern digital signal processing is huge and requires more computational power than a single processor can provide. To address these demands, distributed and parallel processing in embedded systems is becoming a requirement. So it is essential to do research on algorithm and the parallel programming in the parallel embedded system. MPI (message passing interface) is one of the most important parallel programming standards. Porting MPI to the embedded system will enhance application portability and make parallel programming much easier. This paper presents a novel design and implementation of MPI based on link DMA for DSP-based systems. Through computation the value of and PingPong, the results show that the implementation is suitable for parallel programming, especially suitable for large volume of data.