Numerical Model Output Research Articles

Prediction of Sea Surface Current Around the Korean Peninsula Using Artificial Neural Networks

AbstractThe prediction of sea surface currents is essential for various marine activities such as disaster monitoring, fishing industries, and search and rescue operations. Continuous improvements in numerical models have made it possible to predict more realistic oceans using data assimilation and fine spatial resolution. However, such complicated ocean models require high computational power and time, making them difficult to use for near‐real time forecasting. To compensate for these high computational costs, novel approaches with efficient computational costs combined with numerical model outputs need to be developed. Artificial neural networks could be one of the solutions as they require low computational power for prediction, owing to pre‐trained networks. Here, we present a prediction framework applicable to surface current prediction in the seas around the Korean Peninsula using three‐dimensional (3‐D) convolutional neural networks. The network is based on a 3‐D U‐shaped network structure and is modified to predict sea surface currents using oceanic and atmospheric variables. The forecast procedure is optimized to minimize the error of the next day's sea surface current field with a spatial resolution of 1/24° and its recursively predicting structure allows more days to be predicted. The network performance is evaluated by changing the input days and variables to find the optimal surface‐current‐prediction artificial neural network model. The optimized model shows a root mean squared error of 2.3 cm/s for the first forecast day, demonstrating its strong potential for practical use in the near future.

Extreme low‐level wind jets in Storm Ciarán

AbstractSatellite imagery of Storm Ciarán as it approached the UK early on 1 November 2023 showed characteristic banding that fuelled media reports that Ciarán could contain a sting jet, with enhanced likelihood of damaging surface winds and gusts. Here operational numerical weather prediction model output is used with trajectory analysis to reveal the existence of at least two wind speed maxima attributed to sting jets. The small‐scale sting jets are located beneath, and likely affected by, far broader overrunning strong wind regions attributed to a deeply descending dry intrusion, thus highlighting the likely interaction of these airstreams at low levels.

Understanding Coherent Turbulence and the Roll‐Cell Transition With Lagrangian Coherent Structures and Frame‐Indifferent Fluxes

AbstractWe present the first analysis of frame‐indifferent (objective) fluxes and material vortices in Large Eddy Simulations of atmospheric boundary layer turbulence. We extract rotating fluid features that maintain structural coherence over time for near‐neutral, transitional, and convective boundary layers. In contrast to traditional analysis of coherent structures in turbulent boundary layers, we provide the first identification of vortex boundaries that are mathematically defined to behave as tracer transport barriers. Furthermore, these vortices are indifferent to the choice of observer reference frame and can be identified without user‐dependent velocity field decompositions. We find a strong agreement between the geometric qualities of the coherent structures we extract using our new method and classical descriptions of horizontal rolls and convective cells arising from decades of observational studies. We also quantify trends in individual vortex contributions to turbulent and advective fluxes of heat under varying atmospheric stability. Using recently developed tools from the theory of transport barrier fields, we compare diffusive momentum and heat barrier fields with the presence of rolls and cells, and determine a strong connection between heat and momentum orthogonality with the physical drivers of roll‐cell transformation. This newly employed frame‐indifferent characterization of coherent turbulent structures can be directly applied to numerical model output, and thus provides a new Lagrangian approach to understand complex scale‐dependent processes and their associated dynamics.

An accurate and efficient forecast framework for fine PM2.5 maps using spatiotemporal recurrent neural networks

Commonly used numerical prediction models for PM2.5 maps suffer from low accuracy and high computation cost, which cannot meet the requirements for fine-scale air pollution control. In this study, we propose a framework based on the spatiotemporal recurrent neural network (PredRNN) to efficiently generate accurate 3-h and 6-km PM2.5 maps with a lead time of 5 days. In this framework, two PredRNN networks are initially utilized to forecast PM2.5 concentration at ground monitoring sites and the spatial distribution of aerosol optical depth (AOD) by assimilating the output of numerical prediction model. Subsequently, the 3-h and 6-km PM2.5 forecasted maps with a lead time of 5 days can be inferred by establishing the regression links between the forecasted results of PM2.5 concentration at ground sites and AOD maps. We evaluate the proposed framework in the Beijing-Tianjin-Hebei urban agglomeration region during 2017–2020. Compared with the numerical prediction products of the Copernicus Atmosphere Monitoring Service, the proposed framework achieves higher accuracy, with R2 of 0.83 at the forecast base time and 0.70 at the fifth day. The spatial information richness is also enhanced by approximately 15.67% according to the information entropy metrics. Notably, the proposed framework only requires 1 min for forecasting 5-days PM2.5 maps. These results demonstrate that our framework can efficiently generate accurate and fine PM2.5 maps with a lead time of 5 days.

A new airborne system for simultaneous high-resolution ocean vector current and wind mapping: first demonstration of the SeaSTAR mission concept in the macrotidal Iroise Sea

Abstract. Coastal seas, shelf seas and marginal ice zones are dominated by small-scale ocean surface dynamic processes that play a vital role in the transport and exchange of climate-relevant properties such as carbon, heat, water and nutrients between land, ocean, ice and atmosphere. Mounting evidence indicates that ocean scales below 10 km have far-ranging impacts on air–sea interactions, lateral ocean dispersion, vertical stratification, ocean carbon cycling and marine productivity – governing exchanges across key interfaces of the Earth system, the global ocean, and atmosphere circulation and climate. Yet, these processes remain poorly observed at the fine spatial and temporal scales necessary to resolve them. The Ocean Surface Current Airborne Radar (OSCAR) is a new airborne instrument with the capacity to inform these questions by mapping vectorial fields of total ocean surface currents and winds at high resolution over a wide swath. Developed for the European Space Agency (ESA), OSCAR is the airborne demonstrator of the satellite mission concept SeaSTAR, which aims to map total surface current and ocean wind vectors with unprecedented accuracy, spatial resolution and temporal revisit across all coastal seas, shelf seas and marginal ice zones. Like SeaSTAR, OSCAR is an active microwave synthetic aperture radar along-track interferometer (SAR-ATI) with optimal three-azimuth sensing enabled by unique highly squinted beams. In May 2022, OSCAR was flown over the Iroise Sea, France, in its first scientific campaign as part of the ESA-funded SEASTARex project. The campaign successfully demonstrated the capabilities of OSCAR to produce high-resolution 2D images of total surface current vectors and near-surface ocean vector winds, simultaneously, in a highly dynamic, macrotidal coastal environment. OSCAR current and wind vectors show excellent agreement with ground-based X-band-radar-derived surface currents, numerical model outputs and NovaSAR-1 satellite SAR imagery, with root mean square differences from the X-band radar better than 0.2 m s−1 for currents at 200 m resolution. These results are the first demonstration of simultaneous retrieval of total current and wind vectors from a high-squint three-look SAR-ATI instrument and the first geophysical validation of the OSCAR and SeaSTAR observing principle. OSCAR presents a remarkable new ocean observing capability to support the study of small-scale ocean dynamics and air–sea interactions across the Earth's coastal, shelf and polar seas.

Fusion of Lagrangian drifter data and numerical model outputs for improved assessment of turbulent dispersion

Abstract. Transport and dispersion processes in the ocean are crucial, as they determine the lifetime and fate of biological and chemical quantities drifting with ocean currents. Due to the complexity of the coastal ocean environment, numerical circulation models have difficulties to accurately simulate highly turbulent flows and dispersion processes, especially in highly energetic tidal basins such as the eastern English Channel. A method of improving the results of coastal circulation modeling and tracer dispersion in the Dover Strait is proposed. Surface current velocities derived from Lagrangian drifter measurements in November 2020 and May 2021 were optimally interpolated in time and space to constrain a high-resolution coastal circulation MARS model, with careful attention given to selecting ensemble members composing the model covariance matrix. The space–time velocity covariances derived from model simulations were utilized by the optimal interpolation algorithm to determine the most likely evolution of the velocity field under constraints provided by Lagrangian observations and their error statistics. The accuracy of the velocity field reconstruction was evaluated at each time step. The results of the fusion of model outputs with surface drifter velocity measurements show a significant improvement (by ∼ 50 %) of the model capability to simulate the drift of passive tracers in the Dover Strait. Optimized velocity fields were used to quantify the absolute dispersion in the study area. The implications of these results are important, as they can be used to improve existing decision-making support tool or design new tools for monitoring the transport and dispersion in a coastal ocean environment.

Internal-tide vertical structure and steric sea surface height signature south of New Caledonia revealed by glider observations

Abstract. In this study, we exploit autonomous underwater glider data to infer internal-tide dynamics south of New Caledonia, an internal-tide-generation hot spot in the southwestern tropical Pacific. By fitting a sinusoidal function to vertical displacements at each depth using a least-squares method, we simultaneously estimate diurnal and semidiurnal tides. Our analysis reveals regions of enhanced tidal activity, strongly dominated by the semidiurnal tide. To validate our findings, we compare the glider observations to a regional numerical simulation that includes tidal forcing. This comparison assesses the simulation's realism in representing tidal dynamics and evaluates the glider's ability to infer internal-tide signals and their signature in sea surface height (SSH). The glider observations and a pseudo glider, simulated using hourly numerical model output with identical sampling, exhibit similar amplitude and phase characteristics along the glider track. Existing discrepancies are in large part explained by tidal incoherence induced by eddy–internal-tide interactions. We infer the semidiurnal internal-tide signature in steric SSH by the integration of vertical displacements. Within the upper 1000 m, the pseudo glider captures roughly 78 % of the steric SSH total variance explained by the full water column signal. This value increases to over 90 % when projecting the pseudo glider's vertical displacements onto climatological baroclinic modes and extrapolating to full depth. Notably, the steric SSH from glider observations aligns closely with empirical estimates derived from satellite altimetry, highlighting the internal tide's predominant coherent nature during the glider's sampling.

Computationally Efficient Emulation of Spheroidal Elastic Deformation Sources Using Machine Learning Models: A Gaussian‐Process‐Based Approach

AbstractElastic continuum mechanical models are widely used to compute deformations due to pressure changes in buried cavities, such as magma reservoirs. In general, analytical models are fast but can be inaccurate as they do not correctly satisfy boundary conditions for many geometries, while numerical models are slow and may require specialized expertise and software. To overcome these limitations, we trained supervised machine learning emulators (model surrogates) based on parallel partial Gaussian processes which predict the output of a finite element numerical model with high fidelity but >1,000× greater computational efficiency. The emulators are based on generalized nondimensional forms of governing equations for finite non‐dipping spheroidal cavities in elastic halfspaces. Either cavity volume change or uniform pressure change boundary conditions can be specified, and the models predict both surface displacements and cavity (pore) compressibility. Because of their computational efficiency, using the emulators as numerical model surrogates can greatly accelerate data inversion algorithms such as those employing Bayesian Markov chain Monte Carlo sampling. The emulators also permit a comprehensive evaluation of how displacements and cavity compressibility vary with geometry and material properties, revealing the limitations of analytical models. Our open‐source emulator code can be utilized without finite element software, is suitable for a wide range of cavity geometries and depths, includes an estimate of uncertainties associated with emulation, and can be used to train new emulators for different source geometries.

Efficient and effective calibration of numerical model outputs using hierarchical dynamic models

Efficient and effective calibration of numerical model outputs using hierarchical dynamic models

Enhancing tropical cyclone intensity forecasting with explainable deep learning integrating satellite observations and numerical model outputs

Enhancing tropical cyclone intensity forecasting with explainable deep learning integrating satellite observations and numerical model outputs

A Convolutional Neural Network for Tropical Cyclone Wind Structure Identification in Kilometer-Scale Forecasts

Abstract Detection and tracking of tropical cyclones (TCs) in numerical weather prediction model outputs is essential for many applications, such as forecast guidance and real-time monitoring of events. While this task has been automated in the 1990s with heuristic models, relying on a set of empirical rules and thresholds, the recent success of machine learning methods to detect objects in images opens new perspectives. This paper introduces and evaluates the capacity of a convolutional neural network based on the U-Net architecture to detect the TC wind structure, including maximum wind speed area and hurricane-force wind speed area, in the outputs of the convective-scale AROME model. A dataset of 400 AROME forecasts over the West Indies domain has been entirely hand-labeled by experts, following a rigorous process to reduce heterogeneities. The U-Net performs well on a wide variety of TC intensities and shapes, with an average intersection-over-union metric of around 0.8. Its performances, however, strongly depend on the TC strength, and the detection of weak cyclones is more challenging since their structure is less well defined. The U-Net also significantly outperforms an operational heuristic detection model, with a significant gain for weak TCs, while running much faster. In the last part, the capacity of the U-Net to generalize on slightly different data is demonstrated in the context of a domain change and a resolution increase. In both cases, the pretrained U-Net achieves similar performances as the original dataset.

Artificial neural networks ensemble methodology to predict significant wave height

The forecast of wave variables are important for several applications that depend on a better description of the ocean state. Due to the chaotic behaviour of the differential equations which model this problem, a well know strategy to overcome the difficulties is basically to run several simulations, by for instance, varying the initial condition, and averaging the result of each of these, creating an ensemble. Moreover, in the last few years, considering the amount of available data and the computational power increase, machine learning algorithms have been applied as surrogate to traditional numerical models, yielding comparative or better results. In this work, we present a methodology to create an ensemble of different artificial neural networks architectures, namely, MLP, RNN, LSTM, CNN and a hybrid CNN–LSTM, which aims to predict significant wave height on six different locations in the Brazilian coast. The networks are trained using NOAA’s numerical reforecast data and target the residual between observational data and the numerical model output. A new strategy to create the training and target datasets is demonstrated. Results show that our framework is capable of producing high efficient forecast, with an average accuracy of 80%, that can achieve up to 88% in the best case scenario, which means 5% reduction in error metrics if compared to NOAA’s numerical model, and a increasingly reduction of computational cost.

Machine Learning‐Based Wave Model With High Spatial Resolution in Chesapeake Bay

AbstractA high‐resolution wave model is crucial for accurate modeling of sediment and organic material transports, but its computational costs hinder direct coupling to an ecosystem model. We developed a machine learning model using long short‐term memory to simulate large‐scale, high‐resolution waves. Trained with numerical wave model (NWM) outputs and wind data from nine locations, our model successfully replicates NWM results for daily mean significant wave height and period in Chesapeake Bay with identical spatial resolution. Compared to the NWM, the data‐driven model has root‐mean‐square errors below 6 cm for daily mean significant wave height and 1 s for the wave period in the bay. It demonstrates excellent model skills and can accurately forecast daily mean significant wave height and period at NOAA wave stations comparable to NWMs. Using minimal wind data and having a short runtime, our data‐driven model shows promise as an alternative for wave forecasting and coupling with sediment and ecological models.

Slow slip event displacement on 2018 offshore Boso Peninsula detected by Sentinel-1 InSAR time-series analysis with numerical weather model assistance

SUMMARY The south-eastern offshore of the Boso Peninsula in Japan periodically experiences short-term slow slip events (SSEs) every few years. On 2018 June, an SSE occurred with the maximum surface horizontal displacement reaching up to 4.7 cm by according to the operational global navigation satellite system (GNSS) network. This study performed a time-series analysis of interferometric synthetic aperture radar (InSAR) with Sentinel-1 SAR images to investigate detailed spatial pattern of surface displacements caused by the SSE. With the assistance of an atmospheric delay correction with a regional numerical weather model output, the InSAR time-series analysis successfully captured displacement signals in three paths, whose maximum amplitudes in line-of-sight directions were 1.46, 1.86 and −0.80 cm. A checkerboard test revealed that the resolution of the slip inversion was higher when InSAR was used than that using GNSS, especially in and around the inland. The slip inversion with the actual displacement data derived from the InSAR time-series analysis was performed with the L-curve optimization, showing that the estimated slip area was concentrated offshore south-eastward from the Boso Peninsula with the maximum slip of 5 cm and the estimated moment magnitude of 6.4. As similar to previous SSEs in the Boso Peninsula, a seismic swarm simultaneously occurred in the downdip area adjacent to the estimated slip with the SSE occurrence, suggesting a different friction characteristics between them. This study demonstrates usefulnesses of the InSAR observation for capturing detailed spatial characteristics of small-displacement events like SSEs and of the hybrid use of the externally derived delay correction with the time-series analysis to improve the displacement detection accuracy.

Oil Production Rate Forecasting by SA-LSTM Model in Tight Reservoirs

Abstract The accurate forecasting of oil field production rate is a crucial indicator for each oil field’s successful development, but due to the complicated reservoir conditions and unknown underground environment, the high accuracy of production rate forecasting is a popular challenge. To find a low time consumption and high accuracy method for forecasting production rate, the current paper proposes a hybrid model, Simulated Annealing Long Short-Term Memory network (SA-LSTM), based on the daily oil production rate of tight reservoirs with the in situ data of injection and production rates in fractures. Furthermore, forecasting results are compared with the numerical simulation model output. The LSTM can effectively learn time-sequence problems, while SA can optimize the hyperparameters (learning rate, batch size, and decay rate) in LSTM to achieve higher accuracy. By conducting the optimized hyperparameters into the LSTM model, the daily oil production rate can be forecasted well. After training and predicting on existing production data, three different methods were used to forecast daily oil production for the next 300 days. The results were then validated using numerical simulations to compare the forecasting of LSTM and SA-LSTM. The results show that SA-LSTM can more efficiently and accurately predict daily oil production. The fitting accuracies of the three methods are as follows: numerical reservoir simulation (96.2%), LSTM (98.1%), and SA-LSTM (98.7%). The effectiveness of SA-LSTM in production rate is particularly outstanding. Using the same SA-LSTM model, we input the daily oil production data of twenty oil wells in the same block and make production prediction, and the effect is remarkable.

Site-scale groundwater pollution risk assessment using surrogate models and statistical analysis

Petroleum pollution in soil and groundwater has emerged as a significant environmental concern worldwide. As a sustainable and cost-effective in-situ remediation technique, Monitored Natural Attenuation (MNA) exhibits significant promise in addressing sites contaminated by petrochemicals. This study specifically targets a typical petrochemical-contaminated site in northern China and employs GMS software to establish a comprehensive physical model. The model relies on time-series monitoring data of phenol concentrations spanning from 2018 to 2020, effectively simulating both the leakage and natural attenuation of phenol. Within this study, the adsorption coefficient and maximum adsorption capacity emerge as the foremost influential factors shaping the outcomes of the model. Given the inherent heterogeneity of the site and the variability of hydrochemical conditions, parameters such as dispersion, porosity, and adsorption coefficient exhibit significant uncertainties. Consequently, relying on traditional deterministic models to predict the feasibility of MNA technology is not reliable. Therefore, this study employs machine learning (ML) methods to construct stochastic parameter models based on physical processes. The Random Forest Regression (RFR) algorithm, after trained, demonstrates strong alignment with numerical model output, exhibiting an average Nash-Sutcliffe Efficiency (NSE) >0.96. Using a stochastic approach, RFR iteratively computes phenol concentration across 6000 sets of parameters. Applying probability statistics, the model shows a notable reduction in the likelihood of phenol concentrations exceeding a threshold, dropping from 64.0% to 15.7% before and after natural attenuation. In parameter uncertainty, the stochastic model emphasizes natural attenuation's efficacy in mitigating phenol pollution risk (porosity being the most influential factor). This case study proposed a novel method to quickly assess the pollution risks at petrochemical sites under the influence of the uncertainty of pollutant transport and reaction parameters. The results can provide a reference for the pollution risk assessment at petrochemical sites, especially in sites with high stratigraphic heterogeneity or insufficient transport parameter data.

Equatorward shift of ENSO-related subtropical jet anomalies in recent decades

Using observations, NCEP-NCAR and ERA5 reanalysis datasets, numerical experiments and CMIP6 model outputs, this study investigates long-term changes in the meridional position of subtropical jet anomalies in association with El Niño-Southern Oscillation (ENSO). It is found that the subtropical jet anomalies associated with ENSO index (i.e., Niño3.4 index in this study) over both Northern Hemisphere (NH) and Southern Hemisphere (SH) significantly (at the 99% confidence level) shift equatorward since the 1960s, at the speeds of 1.28 and 1.00° per decade, respectively. In recent six decades, the subtropical jet anomalies have moved equatorward by 4.34 and 3.19° over NH and SH, respectively. Our analysis indicates that the tropical positive sea surface temperature (SST) anomalies associated with Niño3.4 (corresponding to its warm phase) during 1990–2020 (P2) are weaker than those during 1950–1980 (P1). The weak, positive SST anomalies induce tropical weak, positive surface latent heat flux and precipitation anomalies, which heat tropical troposphere and produce weak, positive geopotential height anomaly with a narrow meridional width over the tropics during P2. According to geostrophic wind relations, anomalous positive zonal winds occur over the subtropics in the upper troposphere, meaning strengthened subtropical westerly jets over NH and SH. Hence, the relatively weak, narrow positive geopotential height anomaly during P2 favors the equatorward shift of Niño3.4-related subtropical jet anomalies via geostrophic wind relations. The results from numerical experiments and CMIP6 model outputs also indicate that changes in tropical SST anomalies associated with Niño3.4 contribute to this equatorward shift, supporting the results from reanalysis datasets. These results suggest that the intensity of tropical SST anomalies linked to ENSO should be well resolved to better simulate and predict subtropical jet variations.

Machine learning for predicting offshore vertical wind profiles

The accurate characterization of the vertical wind profile over the sea that covers the rotor swept area of modern wind turbines is a key challenge for wind energy yield calculations. Since offshore wind measurements are scarce, early-phase projects tend to use numerical model outputs before planning a dedicated measurement campaign. This study aims to develop and validate a machine-learning model that can assimilate wind parameters measured at the first level of the meteorological masts as input and provide a wind speed profile covering the rotor swept area of modern turbines that is more accurate than numerical weather prediction models. The methodology is based on a random forest model implemented in the python package Scikit-Learn. Three offshore sites in the North Sea have been selected for this study, namely FINO3, IJmuiden and Nordsee Ost’s (NSO) met mast, which are 100 to 350 km apart. Each site has an instrumented 100-m meteorological mast along with a Doppler wind lidar measuring up to 300 m. The baseline selected for comparison was the Weather Research and Forecast (WRF) model. To assess the accuracy of the random forest model two models were trained at IJmuiden and FINO3 and tested at all three locations. Hence, we examine the model performance at the site of training, the so-called same site approach, as well as new sites, the so-called round robin approach. The same site approach quantifies the model consistency, while the round robin shows the degree of spatial robustness. The results show that the model is consistent, where the model trained and tested at FINO3 showed a mean absolute error (MAE) reduction of 68% compared to WRF. This model is also robust, when applied at IJmuiden, 350 km away with a MAE of 1.2 m/s, 8% improved compared to WRF outputs. This study therefore shows the potential to implement machine-learning methods in the prediction of vertical wind speed profiles over the sea. One potential application for the presented methodology is the extension of wind profiles measured by floating lidar systems to higher heights, where current wind lidar products have low availability and higher associated uncertainties, when measuring at a higher height.

Single and multi-objective optimization of PVT performance using response surface methodology with CuO nanofluid application

The nanofluid application in photovoltaic thermal (PVT) systems is widely known as an effective solution for improving output performance. However, studies exploring the optimum operation point of PVT systems with nanofluid application are scant in the current literature. The novelty of the study lies in exploring the optimum operating conditions of each PVT output using single-objective optimization and the overall performance optimization using multi-objective study. The conditions were explored with the application of copper oxide nanofluid as a coolant in PVT systems. The investigation was conducted using a numerical equation-based model developed in TRNSYS simulation environment. An experimental test was conducted to validate the outputs of numerical model. After validating the numerical model, TRNSYS model was investigated for 0.10%-0.50% volume concentration and 60–120 kg/h mass flow. The output data was then imported into Design-Expert software for performance optimization. Central composite design was selected as it is the most suitable fractional factorial design option in response surface methodology. In single-objective investigation, the optimization conditions for PVT outputs are identified as follows, the optimum electrical efficiency output was 14.64% at 90 kg/h fluid flow and 0.10% concentration, the thermal efficiency output was 34.94% at 120 kg/h fluid flow and 0.50% concentration, the overall energy efficiency output was 49.28% at 88 kg/h fluid flow and 0.10% concentration, the overall exergy efficiency output was 16.45% at 60 kg/h fluid flow and 0.50% concentration. The multi-objective study indicated that the collective performance of PVT outputs is optimum at 91 kg/h fluid flow and 0.10% concentration.

Bayesian ensemble machine learning-assisted deterministic and stochastic groundwater DNAPL source inversion with a homotopy-based progressive search mechanism

A hyperheuristic homotopy algorithm (HH-HA) and a homotopy-based swarm intelligence algorithm for parallel stochastic search are proposed to improve the search ergodicity and stability of deterministic and stochastic inversion approaches, respectively, for the identification of dense nonaqueous phase liquid (DNAPL) source characteristics and contaminant transport parameters in groundwater. Meanwhile, a Bayesian ensemble machine learning (BEML) method for numerical simulation surrogate modeling is proposed to enhance the learning and generalization capability, while significantly reducing the computational cost of repetitive CPU-demanding likelihood evaluations in inversion iterations. Two cases are presented for the verification and application of the proposed inverse modeling methods. The results show that the BEML surrogate model is powerful in approximating the numerical model outputs, reducing the mean relative errors of the contaminant concentration and pressure head predictions to 1.18% and 1.29%, respectively. The homotopy-based progressive search mechanism is highly effective in achieving more accurate identification values for deterministic inversion and a more reliable posterior distribution for stochastic inversion. The HH-HA shows stable performance, approaching the global optimum in wide areas and reducing the mean relative error of identification from 11.95% to 4.11%. The homotopy-based parallel stochastic search lines adequately cover the search space tracking the change of homotopy parameter, and the mean relative error of point estimation is reduced to 2.03%. Though the deterministic inversion approach can provide intuitive identification values, the reliability of the inversion outputs cannot be quantified. In contrast, the proposed stochastic inversion system is of great significance for providing decision makers with comprehensive reference information.