Gravity Data Research Articles

Density interface prediction using neural networks: Incorporating fluctuation features from observed gravity data

Predicting density interfaces in the subsurface using neural networks is a promising research area in deep learning-based gravity exploration. Constructing a comprehensive training dataset that captures the fluctuation characteristics of the density interface is vital for predictions. However, large datasets often introduce extraneous features and significantly increase the computational time. To address these challenges, we have developed a method that enhances the accuracy of the prediction by integrating fluctuation features of observed gravity data, specifically the z-component gravity data, into the training dataset. This is achieved by perturbing and deforming the approximate spectrum of the density interface derived from a physics-driven algorithm, thereby generating training samples that closely match the prediction target. Such an approach enables the neural network to concentrate on learning relevant fluctuation features associated with the target. Experimental results underscore the efficacy of our method in accurately discerning density interface features and its robustness against noise interference. Furthermore, our method maintains high prediction accuracy with smaller datasets, outperforming the conventional approaches. This advantage enhances the efficiency and practicality of neural networks in addressing density interface problems by reducing the computational costs for predictions with high accuracy.

Insights on the structural framework of the Egyptian Eastern Desert derived from edge detectors of gravity data

Advanced geophysical techniques are used on Bouguer gravity data to unravel the structural complexity of the Eastern Desert. Using methods such as tilt angle (TA), enhanced horizontal gradient amplitude (EHGA), lineament analysis, and 2Dcrustal modeling, we are able to differentiate between deep and shallow gravity sources and delineate the area’s geologic features. Using fast Fourier transform, positive and negative anomalies are identified. The positive anomalies are interpreted as uplifted block, anticline features, or intrusive occurrences of denser rocks, whereas the negative anomalies are interpreted as downfaulted blocks, providing crucial insights. Lineament analysis suggests that the predominant tectonic force that shaped the area is in the northwest direction. Integration of TA and EHGA reveals nuanced structural patterns. The constructed structural map indicates a network of uplifted and downfaulted blocks intersected by strike-slip faults with a northeast trend. The 2D crustal modeling illustrates the diverse composition of granitic and basaltic rocks, with a discernible thinning of the continental crust toward the Red Sea offshore area. This significantly advances the understanding of the Eastern Desert’s geologic evolution, offering valuable insights into the tectonic processes and crustal dynamics.

Inversion of the Deep Structure of the Northern Margin of the Ordos Block Using Teleseismic Receiver Functions and Gravity Data

Inversion of the Deep Structure of the Northern Margin of the Ordos Block Using Teleseismic Receiver Functions and Gravity Data

Gravity Predictions in Data-Missing Areas Using Machine Learning Methods

Gravity data, comprising a key foundational dataset, are crucial for various research, including land subsidence monitoring, geological exploration, and navigational positioning. However, the collection of gravity data in specific regions is difficult because of environmental, technical, and economic constraints, resulting in a non-uniform distribution of the observational data. Traditionally, interpolation methods such as Kriging have been widely used to deal with data gaps; however, their predictive accuracy in regions with sparse data still needs improvement. In recent years, the rapid development of artificial intelligence has opened up a new opportunity for data prediction. In this study, utilizing the EGM2008 satellite gravity model, we conducted a comprehensive analysis of three machine learning algorithms—random forest, support vector machine, and recurrent neural network—and compared their performances against the traditional Kriging interpolation method. The results indicate that machine learning methods exhibit a marked advantage in gravity data prediction, significantly enhancing the predictive accuracy.

Gravity data inversion for parameters assessment over geologically faulted structures—A hybrid particle swarm optimization and gravitational search algorithm technique

AbstractInterpreting gravity anomalies caused by fault formations is associated with hydrocarbon systems, mineralized areas and hazardous zones and is the main goal of this research. To achieve an effective and robust model over the geologically faulted structures from gravity anomalies, we present a nature‐inspired hybrid algorithm, which synergizes the physics of the particle swarm optimization and gravitational search algorithm with variable inertia weights. The basic principle of developed particle swarm optimization and gravitational search algorithm method is to synergistically use the exploratory strengths of gravitational search algorithm with the exploitation capacity of particle swarm optimization in order to optimize and enhance the effectiveness by both algorithms. The technique has been tested on synthetic gravity data with varying settings of noises over geologically faulted structure before being applied to field data taken from Ahiri‐Cherla and Aswaraopet master fault present in Pranhita–Godavari valley, India. The optimization process is further refined through normalized Gaussian probability density functions, confidence intervals, histograms and correlation matrices to quantify uncertainty, stability, sensitivity and resolution. When dealing with field data, the true model is never known; in these circumstances, the quality of the outcome can only be inferred from the uncertainty in the mean model. The research utilizes a 68.27% confidence intervals to identify a location where the probability density function is more dominant. This region is then used to evaluate the mean model, which is expected to be more appropriate and closer to the genuine model. Correlation matrices further provide a clear demonstration of the strong connection between layer parameters. The results suggest that particle swarm optimization and gravitational search algorithm is less affected by model parameters and yields geologically more consistent outcomes with little uncertainty in the model, aligning well with the available results. The analysed results show that the method we came up with works well and is stable when it comes to solving the two‐dimensional gravity inverse problem. Future research may involve extending the approach to three‐dimensional inversion problems, with potential improvements in computational efficiency and search accuracy for global optimization methods.

Mercury's Crustal Porosity as Constrained by the Planet's Bombardment History

AbstractKnowing the structure of the crust is critical to understanding a planet's geologic evolution. Crustal thickness inversions rely on bulk density estimates, which are primarily affected by porosity. Due to the absence of high‐resolution gravity data, Mercury's crustal porosity has remained unknown. Here, we use a model that was calibrated to the Moon to relate Mercury's impact crater population and long‐wavelength crustal porosity in the cratered terrains. Therein, porosity is created by large impacts and then decreased as the surface ages due to pore compaction by smaller impacts and overburden pressure. Our models fit independent gravity‐derived porosity estimates in the northern regions, where data is well resolved. Porosity in the cratered terrains is found to be 9%–18% with an average and standard deviation of 13% 2%, indicating lunar‐like crustal bulk densities of 2,565 70 kg from which updated crustal thickness maps are constructed.

Three-Dimensional Subsurface Model of Luk-Ulo Melange Complex, Karangsambung, Indonesia: Insights from Gravity Modeling

The Luk-Ulo Melange Complex (LMC) is characterized by a chaotic assemblage of mixed rocks with a block-in-matrix fabric. The exposed blocks consist of various scattered rock types, trending in an ENE-WSW direction. In the case of Mt. Parang, the origin of the diabase remains uncertain, with ongoing debate as to whether it is associated with in situ volcanic activity or represents an exotic block within the melange deposit. Subsurface data obtained through geophysical investigation can aid in modeling the geometry of intrusive bodies using inverse modeling techniques. In this study, we conducted a gravity survey and performed 3D inverse modeling to investigate the subsurface beneath Karangsambung. A total of 818 gravity data points and 28 rock density measurements were integrated with existing geological data to construct an a priori 3D geological model. To ensure the results align with geological concepts, the 3D inversion utilized a stochastic approach, allowing for the incorporation of multiple geological constraints over fifty million iterative procedures. Ultimately, the inversion successfully reduced the misfit between observed and calculated data from 2.71 to 0.55 mGal. Based on the inverted 3D model, the diabase rock in Mt. Parang is identified as having an intrusive origin. The intrusion model exhibited minimal changes in density, volume, and shape during the inversion process. Additionally, the model suggests the presence of a solidified magma reservoir at a depth of approximately 3 km, potentially related to Dakah volcanism. The inverted model also reveals the block-in-matrix structure of the Luk-Ulo Melange Complex in the northern area.

Precise Geoid Determination in the Eastern Swiss Alps Using Geodetic Astronomy and GNSS/Leveling Methods.

Astrogeodetic deflections of the vertical (DoVs) are close indicators of the slope of the geoid. Thus, DoVs observed along horizontal profiles may be integrated to create geoid undulation profiles. In this study, we collected DoV data in the Eastern Swiss Alps using a Swiss Digital Zenith Camera, the COmpact DIgital Astrometric Camera (CODIAC), and two total station-based QDaedalus systems. In the mountainous terrain of the Eastern Swiss Alps, the geoid profile was established at 15 benchmarks over a two-week period in June 2021. The elevation along the profile ranges from 1185 to 1800 m, with benchmark spacing ranging from 0.55 km to 2.10 km. The DoV, gravity, GNSS, and levelling measurements were conducted on these 15 benchmarks. The collected gravity data were primarily used for corrections of the DoV-based geoid profiles, accounting for variations in station height and the geoid-quasigeoid separation. The GNSS/levelling and DoV data were both used to compute geoid heights. These geoid heights are compared with the Swiss Geoid Model 2004 (CHGeo2004) and two global gravity field models (EGM2008 and XGM2019e). Our study demonstrates that absolute geoid heights derived from GNSS/leveling data achieve centimeter-level accuracy, underscoring the precision of this method. Comparisons with CHGeo2004 predictions reveal a strong correlation, closely aligning with both GNSS/leveling and DoV-derived results. Additionally, the differential geoid height analysis highlights localized variations in the geoid surface, further validating the robustness of CHGeo2004 in capturing fine-scale geoid heights. These findings confirm the reliability of both absolute and differential geoid height calculations for precise geoid modeling in complex mountainous terrains.

Assessing Groundwater Storage Change in the Great Artesian Basin Using GRACE and Groundwater Budgets

AbstractLarge, confined aquifer systems play a vital role in sustaining human settlements and industries in many regions. Understanding the sustainability of these water resources requires the evaluation of groundwater storage change. Direct in‐situ observation of groundwater storage is limited by the distribution and availability of groundwater level and aquifer storativity data. Here, we use and compare two auxiliary methods, applied at basin and sub‐basin scales, to assess groundwater storage changes in the Great Artesian Basin (GAB), one of the World's largest confined aquifer systems. The first, the groundwater budget, derives storage change as the residual of fluxes in and out of the GAB, assuming they are all accounted for and accurately estimated. The second uses time‐variable gravity data from GRACE satellites to estimate temporal changes in groundwater mass, assuming that all other components of the terrestrial water mass change detected by GRACE are correctly subtracted. Despite the depletion observed during the 20th century, groundwater storage is mostly stable during 2002–2022. An increase in storage is detected in the Surat sub‐basin, a major recharge area. This increase is attributed to an over‐representation of large recharge events during the study period and/or storage recovery following rehabilitation of free‐flowing bores. The approach consisting in disaggregating GRACE data assumes that water storage changes in confined aquifers is dominated by changes in the GAB, and as such, it may overestimate the increase in the GAB by incorrectly attributing the increase occurring in overlying aquifers to the GAB. In contrast, the recharge estimates used in the groundwater budgets do not account for flood recharge and might underestimate storage increase in the GAB.

Modification methods of the Stokes’ kernel for determining the (quasi-) geoid with the Remove-Compute-Restore technique

The geoid and quasi-geoid serve as the reference surfaces of the orthometric and normal height systems, respectively. In order to improve the accuracy of the (quasi-) geoid determined by the Stokes integral with use of the Remove-Compute-Restore (RCR) technique, various modification methods for the spherical Stokes’ kernels, including the spheroidal, cosine-, power-, and Molodensky-modified kernels, are studied in this paper. In addition to the traditional Molodensky-modified Stokes’ kernel, a more effective Molodensky-modified Stokes’ kernel is put forward. A general formula for spectral decomposition of the Stokes integral in the RCR mode is derived, followed by the spectral analysis to reveal the transfer principles of gravity data when using different Stokes’ kernels. The spheroidal and modified Stokes integrals can cause spectral leakage phenomenon, and a method to eliminate spectral leakage is presented based on spectral analysis. The research indicates the low truncation degree of the spheroidal Stokes’ kernel and the low modification degrees of the modified Stokes’ kernel affect the accuracy of the (quasi-) geoid significantly. Quantitative methods for estimating the empirical values of the parameters of the low-degree spheroidal and modified Stokes’ kernels are proposed and the effectiveness of the methods is validated through numerical tests.

Identification of Subsurface Geological Structures from Gravity Data by Modelling Techniques in Kalak-Bardarash Area, West of Erbil City Iraqi Kurdistan Region

A quantitative interpretation of gravity anomalies in the Kalak-Bardarash area, west of Erbil City in the Kurdistan Region of Iraq, was performed using two-dimensional analysis techniques. These techniques modeled geological structures identified from a previously created Bouguer anomaly map of the area. Five profiles were selected for the analysis, resulting in eight models constructed for both deep and shallow-seated rock units. The deep-seated rocks are represented by the basement rocks, with models constructed in the NNE-SSW, WNW-ESE, and NW-SE directions using the regional anomaly component. The depth of the basement rocks ranges from five to nine kilometers and is affected by numerous faults in the form of graben and step structures. The basement rock surface slopes towards the NE and ESE at a gradient of 110 m/km. The shallow-seated rocks are represented by the sedimentary cover, which includes Neogene, Paleogene, and Cretaceous-aged groups of formations. The residual component of the anomalies of the five profiles were interpreted. Three profiles, oriented NNE-SSW and nearly perpendicular to the main anticlinal structure, revealed that the Bardarash anticline is influenced by two faults striking both limbs and extending in the NNW-SSE direction, with an estimated throw of 200 meters on each side. The southern fault is believed to extend down to the basement rocks. The other two profiles, oriented NW-SE and nearly perpendicular to the Zab River, indicated the presence of two additional faults with a throw of approximately 600 meters in the deeper parts striking the basement rocks, dipping towards the southeast. The throw of these faults gradually decreases upward in the sedimentary cover, reaching about 200 meters. These faults directly influence the course of the Zab River, directing it in a NE-SW direction at this district.

Comprehensive gravity and magnetic data analysis for quantitative determining salt bodies in complex physical-geological environments

Comprehensive gravity and magnetic data analysis for quantitative determining salt bodies in complex physical-geological environments

The Evolution of Powell Basin (Antarctica)

Powell Basin is an ocean basin formed as a result of the Scotia Sea evolution. The existing tectonic models propose a variety of starting and ending ages for the spreading of the basin based on seafloor magnetic anomalies. Here, we use recent magnetic field data obtained from eight magnetic profiles in Powell Basin to provide insights into the oceanic spreading evolution. The differences found between the number of anomalies on both sides of the axis and the asymmetry in the spreading rates suggest different opening models for different parts of the basin. We propose a spreading model starting in the late Eocene (38.08 Ma) and ending in the early Miocene (21.8 Ma) for the northern part of Powell Basin. For the southern part, the opening started in the late Eocene (38.08 Ma) and ended in the middle Paleogene (25.2 Ma). The magnetic data have been combined with gravity and sediment thickness data to better constrain the age models. The gravity and sediment thickness information allow us to more accurately locate the position of the extinct spreading axis. Geothermal heat flow measurements are used to understand the relationship between the low amplitudes of the magnetic anomalies and the heat beneath them. Our proposed oceanic spreading models suggest that the initial incursions of the Pacific mantle outflow into the Powell Basin occurred in the Oligocene, and the initial incursions of oceanic currents from the Weddell Sea occurred in the Eocene.

Multiple Upward Residual Gravity and Magnetic Maps to Estimate the Deep Crustal Layers and Moho Discontinuity in Eastern Iraq

The upward continuation technique was applied to the gravity and magnetic data to obtain the residual anomalies maps in eastern Iraq. A new processing method is introduced to calculate the crustal layers depending on the relation between the total range of the residual anomalies of gravity, magnetic data and the upward continuation, with multi-elevations. The significant relationship shows slope variations, which may reflect the maximum contrast in the physical properties (density and/or susceptibility) and the intersection point representing the depth of layer contacts. This method was applied to the eastern central part of Iraq, adjacent to the Iranian borders. It is found that the average depth range to the Jurassic formations, top basement rocks, transition layer (Curie point), and Moho discontinuity are 3.1- 4.8 km, 8-12.9 km, 20.9-22.4 km and 43.6 km, respectively. The Sedimentary cover in the study area ranges from 8 to 13 km. The maximum depth to the Moho discontinuity obtained by gravity data in the study area is 43.6 km. The radially average power spectrum technique was applied in the study area to calculate the gravity depth and magnetic sources' depth. Three depth levels of sources from gravity and magnetic data were obtained. These depths are shallow sources, average basement rock and deep sources (Moho) at 3km, 13.2-12.1 km, and 32.2-34km, respectively. The obtained results concerning the main layers in the crust were compared between the two methods, and the new method shows more reliable results. The detected depth levels refer to top Jurassic, basement rocks, and Moho discontinuity. The new method is believed to be used to detect the crustal layers depth with good results.

Subsurface basement structures of the Usangu basin, East African rift system, with implications for basin structural configuration and hydrocarbon potential

The Usangu basin is a rift basin developed along the Eastern arm of the East African rift system trending in the NE-SW direction. Although the general structures of the basin have been well studied, the structural configuration of the basin and the spatial and depth variations of sediment thickness are still not well known. This study investigates the structures related basement configuration and the variation of sediment thickness within the basin using the Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM), aeromagnetic and gravity data. Results from DEM data indicate a few lineaments on the basin flanks representing the Usangu and Chimala border faults with no structures in the central part of the basin. The aeromagnetic and gravity data highlight three sets of normal and strike-slip faults, most of which trend NE-SW and others NNE-SSW, while a few trends WNW-ESE or NW-SE. Structures on the southwest of the basin reveal complex patterns attributed to the concentration of important tectonic and seismic activities in the study area. The Euler deconvolution and gravity models used to calculate the depth to the basement show that the basement is shallow in the north and south to southwest, and the basin deepens in the northeastern, northwestern and western parts. The findings also reveal that the basin has two grabens, troughs, depression and intrabasinal basement trending in the same direction as the basin configuration. The general thickness of sediments filling the basin ranges from 3 to 4.5 km, with the maximum accumulation of sediments reaching up to 4.8 km in the two depocenters at the south and southwest of the basin. The depth range of the sediments obtained implies that the basin has potential for hydrocarbon exploration with the possibility of natural gas occurrence.

Jurassic depocenters development, between Colorado basin, Chelforó sub-basin and Neuquén basin

Northern Patagonia, to the north of the Somún Cura Massif along the 39°S parallel, is characterized by the development of a series of Jurassic depocenters, initiating in the Andes to the west, across the Neuquén basin and the Pampean sector (Chelforó sub-basin) towards the Atlantic coast near the location of the Pedro Luro −1 well (onshore Colorado basin), and further east over the Argentine shelf, where multiple Jurassic depocenters form the Colorado, Rawson, and Valdés basins.Using marine and continental gravity data (Free air and Isostatic anomalies respectively), constrained by well and seismic data, an integrated interpretation workflow for the study area was carried out. The southern limit of the Neuquén Basin is controlled by the Huincul High, an E-W directed structure with a positive gravimetric anomaly, interpreted as a basement high limited by relative gravimetric lows to the north and south of the structure. The eastern sector includes a series of isolated depocenters, separated by basement highs (obliquely oriented) and limited to the south by the E-W anomaly of the Huincul High. These depocenters correspond to the Chelforó sub-basin, which was drilled in a recent exploration campaign obtaining Jurassic ages where seismic interpretation pointed out presumably Jurassic deposits. Between Valcheta and Sierra Grande, this positive anomaly turns NW and becomes the southern boundary of the Pedro Luro depocenter, partially drilled by the Pedro Luro- 1 well, with possibly Tithonian palynological records at total depth. Between the Pedro Luro depocenter and the Chelforó sub-basin some features consistent with basement highs with circular morphology have been identified on gravimetric data, located in the area limited between the Colorado and Río Negro rivers. Jurassic rift depocenters have also been interpreted further east, in the Colorado basin where depocenters are controlled by an E-W oriented basement high, and further east by NW trending structures near the position of the SDRs (seaward dipping reflectors on the continental-oceanic crust transition zone).The common evolution of these Jurassic depocenters is still poorly understood. They extend for more than 1500 km from the Andes to the Atlantic continental shelf and are controlled by structures located to the north of the Somún Cura Massif. The interpretation of seismic lines, and gravimetric data in areas without deep wells allow the identification of new exploration targets increasing the exploration potential of these basins.

A Comprehensive Geophysical Exploration of Sedimentary Exhalative Deposits: An Example from the Huaniushan Lead–Zinc–Silver Polymetallic Deposit in Gansu, China

The Huaniushan lead–zinc–silver deposit is a hydrothermal sedimentary exhalative deposit (SEDEX), and the mining area has complex geological conditions, with the main tectonic structure being the Huaheitan–Shuangfengshan Fault (F3), which controls the distribution of strata and magmatic rocks. Since the discovery of the Huaniushan lead–zinc–silver deposit, diverse interpretations of its genesis and metallogeny have been proposed, making it challenging to establish a definitive geological explanation. Moreover, using a single geophysical exploration method relies on limited rock physical parameters, making it difficult to effectively characterize underground structures. The combined use of multiple geophysical methods can effectively integrate the geophysical characteristics of different rock physical parameters, reducing the multiplicity and uncertainty of the inverse interpretation of geophysical data. The comprehensive interpretation of three-dimensional inversion based on various geophysical data, the construction of geological–geophysical models on geological grounds, the establishment of hidden ore exploration and positioning, and the rapid evaluation of geophysical technological systems are the current research trends in mineral exploration. In light of this, in this study, we conducted research on the three-dimensional inversion interpretation of gravity and magnetoelectric exploration data of the Huaniushan sedimentary exhalative lead–zinc–silver polymetallic deposit and constructed a three-dimensional geological–geophysical model of the study area based on the obtained three-dimensional physical structure of the underground density, magnetization intensity, resistivity, and polarizability of the study area, in combination with related geological and drilling hole data. Finally, we comprehensively interpreted the favorable mineralization sites in the study area.

Understanding the impact of deep structures on the hydrological setting of the Eastern Bahira Basin in Morocco using combined geophysical analysis of gravity, seismic, and electrical resistivity data

In arid regions like the Eastern Bahira Basin, communities mainly rely on groundwater for drinking and irrigation. However, efficiently managing these vital resources requires a deep understanding of the underlying aquifers’ structure and identifying the most suitable areas for exploitation. This presents a significant challenge for the success of water supply and irrigation programs in the Eastern Bahira Basin. This study is based on an integrative approach, combining Electrical Resistivity Tomography data, with the compilation and reinterpretation of pre-existing seismic, gravimetric and vertical electrical sounding data. This approach is based on compiling old gravimetric data and applying advanced processing techniques to determine the horizontal gradient maxima, which helps highlight the major structural alignments in the basin. Furthermore, the approach utilizes seismic data in order to enhance understanding of the deep structure of the basin, reinterpreting it in light of recent drilling data. The interpretation of the gravimetric and seismic data has also been validated by the results of vertical electrical soundings and electrical tomography that we recently acquired in the Eastern Bahira basin. The outcomes of this research provide new insights into the deep structure of the Eastern Bahira Basin and suggest the most promising hydrogeological prospects, thereby contributing to the success of the ongoing drinking water supply and irrigation program in the Eastern Bahira Basin.

Crustal melting and continent uplift by mafic underplating at convergent boundaries.

The thick crust of the southern Tibetan and central Andean plateaus includes high-conductivity, low-velocity zones ascribed to partial melt. The melt origin and effect on plateau uplift remain speculative, in particular if plateau uplift happens before continental collision. The East Anatolian Plateau (EAP) has experienced similar, more recent uplift but its structure is largely unknown. Here we present an 80 km deep geophysical model across EAP, constrained by seismic receiver functions integrated with interpretation of gravity data and seismic tomographic, magnetotelluric, geothermal, and geochemical models. The results indicate a 20 km thick lower crustal layer and a 10 km thick mid-crustal layer, which both contain pockets of partial melt. We explain plateau uplift by isostatic equilibration following magmatism associated with roll-back and break-off of the Neo-Tethys slab. Our results suggest that crustal thickening by felsic melt and mafic underplate are important for plateau uplift in the EAP, Andes and Tibet.

Imaging the innermost circumstellar environment of the red supergiant WOH G64 in the Large Magellanic Cloud

Significant mass loss in the red supergiant (RSG) phase has great influence on the evolution of massive stars and their final fate as supernovae. We present near-infrared interferometric imaging of the circumstellar environment of the dust-enshrouded RSG WOH G64 in the Large Magellanic Cloud. WOH G64 was observed with the GRAVITY instrument at ESO's Very Large Telescope Interferometer (VLTI) at 2.0--2.45 mu m . We succeeded in imaging the innermost circumstellar environment of WOH G64 -- the first interferometric imaging of an RSG outside the Milky Way. The reconstructed image reveals elongated compact emission with a semimajor and semiminor axis of sim 2 and sim 1.5 mas (sim 13 and 9 $R_ star ), respectively. The GRAVITY data show that the stellar flux contribution at 2.2 mu m \ at the time of our observations in 2020 is much lower than predicted by the optically and geometrically thick dust torus model based on the VLTI/MIDI data taken in 2005 and 2007. We found a significant change in the near-infrared spectrum of WOH G64: while the (spectro)photometric data taken at 1--2.5 mu m \ before 2003 show the spectrum of the central RSG with H$_2$O absorption, the spectra and prime photometric data taken after 2016 are characterized by a monotonically rising continuum with very weak signatures of H$_2$O . This spectral change likely took place between December 2009 and 2016. On the other hand, the mid-infrared spectrum obtained in 2022 with VLT/VISIR agrees well with the spectra obtained before 2007. The compact emission imaged with GRAVITY and the near-infrared spectral change suggest the formation of hot new dust close to the star, which gives rise to the monotonically rising near-infrared continuum and the high obscuration of the central star. The elongation of the emission may be due to the presence of a bipolar outflow or effects of an unseen companion.