Tongchao Nan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Tongchao Nan has a total of 67 co-authors. Here are the top co-authors by number of shared publications:
- Chunhui Lu
The National Key Laboratory of Water Disaster Prevention Hohai University Nanjing China
7 shared publications Jichun Wu
7 shared publications- Jichun Wu
Key Laboratory of Surficial Geochemistry of Ministry of Education School of Earth Sciences and Engineering Nanjing University Nanjing China
7 shared publications - Jiangjiang Zhang
State Key Laboratory of Water Disaster Prevention Hohai University Nanjing China
6 shared publications Yifan Xie
5 shared publications- Jina Yin
State Key Laboratory of Water Disaster Prevention Hohai University Nanjing China
4 shared publications
Latest Publications
- Open Access
- 10 November 2026
- Published by American Geophysical Union (AGU) in Water Resources Research
- Vol. 62 (11)
Nitrate Transport, Denitrification, and Discharge in Tide‐Controlled Fractured Coastal Aquifers
Land‐derived nitrate degrades coastal groundwater quality, with denitrification in intertidal aquifers serving as a critical barrier to its seaward flux. While geologic heterogeneity has been recognized as a key factor affecting nitrate transport, denitrification, and discharge in coastal aquifers, few studies have investigated how fractures regulate nitrate fate, particularly the main fractures dominating water flow. This study first integrated discrete fracture matrix model (DFM)‐based sandbox experiments and numerical simulations to analyze land‐sourced conservative solute transport in homogeneous, horizontally, and vertically fractured intertidal aquifers. The validated model was then scaled up to assess nitrate dynamics under point and non‐point source scenarios, with sensitivity analyses further exploring how fracture geometry, aquifer hydraulic properties, and boundary conditions affect non‐point source nitrate fate in horizontally fractured aquifers. Results show that main fractures act as preferential pathways directing nitrate into intertidal seawater‐freshwater mixing zones, inducing denitrification hotspots at the fracture‐upper saline plume (USP) interface. Under point source conditions, horizontal and vertical fractures emphasize corresponding expansion of interfacial hotspots. In the non‐point source scenario, broader nitrate inputs promote denitrification near high and low tidal levels, weakening the prominence of fracture‐induced hotspots. Sensitivity analyses highlight that fracture's proximity to the USP or enhanced reactant convergence capacity can expand interfacial hotspots, facilitating nitrate removal and lowering discharge concentrations. In addition, the downstream N 2 plume's extent and concentration are governed by the size and activity of fracture‐induced hotspots, as well as the fracture's freshwater‐zone length.
- 10 November 2026
- Published by American Geophysical Union (AGU) in Water Resources Research
- Vol. 62 (11)
Open AccessNitrate Transport, Denitrification, and Discharge in Tide‐Controlled Fractured Coastal Aquifers
Land‐derived nitrate degrades coastal groundwater quality, with denitrification in intertidal aquifers serving as a critical barrier to its seaward flux. While geologic heterogeneity has been recognized as a key factor affecting nitrate transport, denitrification, and discharge in coastal aquifers, few studies have investigated how fractures regulate nitrate fate, particularly the main fractures dominating water flow. This study first integrated discrete fracture matrix model (DFM)‐based sandbox experiments and numerical simulations to analyze land‐sourced conservative solute transport in homogeneous, horizontally, and vertically fractured intertidal aquifers. The validated model was then scaled up to assess nitrate dynamics under point and non‐point source scenarios, with sensitivity analyses further exploring how fracture geometry, aquifer hydraulic properties, and boundary conditions affect non‐point source nitrate fate in horizontally fractured aquifers. Results show that main fractures act as preferential pathways directing nitrate into intertidal seawater‐freshwater mixing zones, inducing denitrification hotspots at the fracture‐upper saline plume (USP) interface. Under point source conditions, horizontal and vertical fractures emphasize corresponding expansion of interfacial hotspots. In the non‐point source scenario, broader nitrate inputs promote denitrification near high and low tidal levels, weakening the prominence of fracture‐induced hotspots. Sensitivity analyses highlight that fracture's proximity to the USP or enhanced reactant convergence capacity can expand interfacial hotspots, facilitating nitrate removal and lowering discharge concentrations. In addition, the downstream N 2 plume's extent and concentration are governed by the size and activity of fracture‐induced hotspots, as well as the fracture's freshwater‐zone length.
- 1 November 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (11)
Spectral-structure modulation of non-Darcy flow and heat transfer in Weierstrass–Mandelbrot rough fractures
Fluid flow and fluid–solid heat transfer in rough fractures depend on geometric organization that is not fully specified by scalar roughness descriptors. This computational study uses COMSOL Multiphysics to compare four finite-band Weierstrass–Mandelbrot fractures with matched vertical mechanical aperture, nominal joint roughness coefficient (JRC), and finite-scale structure–function dimension DSF = 1.200. A full-band reference (WM-F) is compared with long-, middle-, and short-wave-enriched geometries (WM-L, WM-M, and WM-S), retaining every frequency order. Steady incompressible flow and conjugate heat transfer are simulated for Reynolds numbers (Re) from 5 to 600. At JRC = 17, the fitted 10% inertial-contribution threshold occurs at Re ≈ 64 for WM-S and Re ≈ 659 for WM-L; the latter is an extrapolation beyond the simulated interval. WM-S develops the earliest and largest retained recirculation region. At Re = 600, its bulk-temperature-based overall heat-transfer coefficient h is 22.9% higher than that of WM-F, accompanied by a 45.3% higher pressure gradient. WM-L and WM-M have coefficients 2.5% and 1.8% above the reference while requiring pressure gradients 16.6% and 15.1% lower, respectively. These comparisons establish nonuniqueness of transport response within the matched-descriptor set. Spectral information may provide additional descriptive information relevant to transport, but its effect is not isolated from correlated differences in maximum slope, turning density, and other geometric outcomes.
- 1 November 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (11)
Spectral-structure modulation of non-Darcy flow and heat transfer in Weierstrass–Mandelbrot rough fractures
Fluid flow and fluid–solid heat transfer in rough fractures depend on geometric organization that is not fully specified by scalar roughness descriptors. This computational study uses COMSOL Multiphysics to compare four finite-band Weierstrass–Mandelbrot fractures with matched vertical mechanical aperture, nominal joint roughness coefficient (JRC), and finite-scale structure–function dimension DSF = 1.200. A full-band reference (WM-F) is compared with long-, middle-, and short-wave-enriched geometries (WM-L, WM-M, and WM-S), retaining every frequency order. Steady incompressible flow and conjugate heat transfer are simulated for Reynolds numbers (Re) from 5 to 600. At JRC = 17, the fitted 10% inertial-contribution threshold occurs at Re ≈ 64 for WM-S and Re ≈ 659 for WM-L; the latter is an extrapolation beyond the simulated interval. WM-S develops the earliest and largest retained recirculation region. At Re = 600, its bulk-temperature-based overall heat-transfer coefficient h is 22.9% higher than that of WM-F, accompanied by a 45.3% higher pressure gradient. WM-L and WM-M have coefficients 2.5% and 1.8% above the reference while requiring pressure gradients 16.6% and 15.1% lower, respectively. These comparisons establish nonuniqueness of transport response within the matched-descriptor set. Spectral information may provide additional descriptive information relevant to transport, but its effect is not isolated from correlated differences in maximum slope, turning density, and other geometric outcomes.
- Open Access
- 22 October 2026
- Published by Elsevier in SSRN Electronic Journal
A graph-based recurrent surrogate for coupled pressure-temperature dynamics in fractured geothermal reservoirs
Reliable prediction of pressure propagation and thermal breakthrough in fractured geothermal reservoirs requires explicit representation of fracture connectivity, but repeated high-fidelity simulations are costly. We developed a coupled recurrent conditional generative adversarial network (CRcGAN) for joint prediction of pressure and temperature fields. Matrix and fracture cells and their embedded-discrete-fracture-model (EDFM) connections are represented as a graph. A gated graph neural network encodes this topology, and an autoregressive pressure–temperature cascade exchanges information between the hydraulic and thermal branches. Conditional adversarial training complements supervised losses and emphasizes fracture nodes and nodes linked by non-neighboring connections. The model was trained and evaluated on 10,000 EDFM simulations spanning stochastic fracture networks, varying initial conditions, and 20 years of operation; 1,000 simulations were withheld for testing. Across the test set, mean absolute errors were 0.23 bar for pressure and 0.47 K for temperature. Ablation experiments showed that coupled recurrence, hydraulic conditioning, adversarial training, and regional weighting improved prediction accuracy. For 1,000 test samples, inference required 860 s, compared with 59,316 s for EDFM simulations in MRST, corresponding to a 69-fold speed-up. CRcGAN can therefore reproduce hydrothermal responses efficiently within the sampled parameter range and may support computationally intensive tasks such as scenario screening and fracture-network inversion.
- 22 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA graph-based recurrent surrogate for coupled pressure-temperature dynamics in fractured geothermal reservoirs
Reliable prediction of pressure propagation and thermal breakthrough in fractured geothermal reservoirs requires explicit representation of fracture connectivity, but repeated high-fidelity simulations are costly. We developed a coupled recurrent conditional generative adversarial network (CRcGAN) for joint prediction of pressure and temperature fields. Matrix and fracture cells and their embedded-discrete-fracture-model (EDFM) connections are represented as a graph. A gated graph neural network encodes this topology, and an autoregressive pressure–temperature cascade exchanges information between the hydraulic and thermal branches. Conditional adversarial training complements supervised losses and emphasizes fracture nodes and nodes linked by non-neighboring connections. The model was trained and evaluated on 10,000 EDFM simulations spanning stochastic fracture networks, varying initial conditions, and 20 years of operation; 1,000 simulations were withheld for testing. Across the test set, mean absolute errors were 0.23 bar for pressure and 0.47 K for temperature. Ablation experiments showed that coupled recurrence, hydraulic conditioning, adversarial training, and regional weighting improved prediction accuracy. For 1,000 test samples, inference required 860 s, compared with 59,316 s for EDFM simulations in MRST, corresponding to a 69-fold speed-up. CRcGAN can therefore reproduce hydrothermal responses efficiently within the sampled parameter range and may support computationally intensive tasks such as scenario screening and fracture-network inversion.
- Top Cited PapersOpen Access
- 27 February 2026
- Published by American Geophysical Union (AGU) in Water Resources Research
- Vol. 62 (3)
Improving Heterogeneous Aquifer Characterization Using a Deep Learning‐Based Ensemble Smoother With a Hybrid Prior Strategy
Accurate characterization of aquifer hydraulic properties is crucial for groundwater resource management and pollution control. Most traditional inverse methods for aquifer characterization rely on unbiased prior assumptions to make robust estimation; however, real‐world applications frequently involve biased priors that can significantly compromise characterization accuracy. To mitigate this issue, this study introduces a hybrid prior strategy that combines multiple plausible prior assumptions to form a comprehensive prior. This strategy is implemented within a deep learning‐based ensemble smoother, termed , enabling robust estimation of unknown, heterogeneous aquifer hydraulic properties without relying on unbiased prior assumptions. We systematically evaluate the performance of under unbiased, biased, and hybrid priors across hydraulic conductivity fields characterized by log‐Gaussian, channelized, and three‐facies distributions. Using the Kalman‐based ensemble smoother () as a benchmark, results demonstrate that the hybrid prior strategy significantly enhances the robustness of aquifer characterization under biased prior assumptions, with consistently outperforming in terms of both accuracy and stability. Notably, maintains high effectiveness even when the hybrid prior ensemble excludes unbiased samples, highlighting its adaptability and reliability in complex groundwater systems.
- 27 February 2026
- Published by American Geophysical Union (AGU) in Water Resources Research
- Vol. 62 (3)
Top Cited PapersOpen AccessImproving Heterogeneous Aquifer Characterization Using a Deep Learning‐Based Ensemble Smoother With a Hybrid Prior Strategy
Accurate characterization of aquifer hydraulic properties is crucial for groundwater resource management and pollution control. Most traditional inverse methods for aquifer characterization rely on unbiased prior assumptions to make robust estimation; however, real‐world applications frequently involve biased priors that can significantly compromise characterization accuracy. To mitigate this issue, this study introduces a hybrid prior strategy that combines multiple plausible prior assumptions to form a comprehensive prior. This strategy is implemented within a deep learning‐based ensemble smoother, termed , enabling robust estimation of unknown, heterogeneous aquifer hydraulic properties without relying on unbiased prior assumptions. We systematically evaluate the performance of under unbiased, biased, and hybrid priors across hydraulic conductivity fields characterized by log‐Gaussian, channelized, and three‐facies distributions. Using the Kalman‐based ensemble smoother () as a benchmark, results demonstrate that the hybrid prior strategy significantly enhances the robustness of aquifer characterization under biased prior assumptions, with consistently outperforming in terms of both accuracy and stability. Notably, maintains high effectiveness even when the hybrid prior ensemble excludes unbiased samples, highlighting its adaptability and reliability in complex groundwater systems.
- 30 January 2026
- Published by Elsevier in Journal of Environmental Management
- Vol. 400, 128766
Spatiotemporal prediction and attribution of groundwater storage anomaly using enhanced hybrid deep learning modeling with uncertainty quantification
- 30 January 2026
- Published by Elsevier in Journal of Environmental Management
- Vol. 400, 128766
Spatiotemporal prediction and attribution of groundwater storage anomaly using enhanced hybrid deep learning modeling with uncertainty quantification
- 17 September 2025
- Published by Elsevier in Advances in Water Resources
- Vol. 206
Modeling hydro-thermal processes in fractured geothermal reservoirs using embedded discrete fracture model (EDFM) and MRST
- 17 September 2025
- Published by Elsevier in Advances in Water Resources
- Vol. 206
Modeling hydro-thermal processes in fractured geothermal reservoirs using embedded discrete fracture model (EDFM) and MRST
- Open Access
- 24 June 2025
- Published by American Geophysical Union (AGU) in Water Resources Research
- Vol. 61 (6)
Enhancing Aquifer Characterization With Position‐Encoded Hyperparameters: A Novel ES‐SIFG Approach
To accurately predict groundwater flow and solute transport, it is essential to precisely characterize the highly heterogeneous aquifer conditions. Ensemble smoother with multiple data assimilation (ESMDA), though widely applied to identify aquifer properties and spatial features, encounters severe problems in practice due its fundamental assumptions of linearity and Gaussianity. To tackle this challenge, we first use a spatially‐informed field generator (SIFG) to hyperparameterize the conductivity field and encode position information into the hyperparameters, and then combine it with ensemble smoother to form a new inversion framework called ensemble smoother with SIFG (ES‐SIFG). In ES‐SIFG, followed by utilizing an ensemble smoother to update the hyperparameters rather than the aquifer parameters. The main innovation of ES‐SIFG is integrating positional information into hyperparameters, enabling the use of distance‐based covariance localization (CL) and significantly reducing the number of model simulations. The proposed method has been tested on parameter identification problems in 2‐D and 3‐D non‐Gaussian aquifers and compared to ESMDA with normal score transformation. Results indicate that ES‐SIFG outperforms ESMDA and is capable of accurately identifying non‐Gaussian aquifer parameters and reconstructing contaminant release history, particularly in resolving equifinality and preserving prior geological structure. Furthermore, SIFG allows usage of CL between hyperparameters and observations, which ensures the stable convergence of data assimilation processes even with very small ensemble sizes.
- 24 June 2025
- Published by American Geophysical Union (AGU) in Water Resources Research
- Vol. 61 (6)
Open AccessEnhancing Aquifer Characterization With Position‐Encoded Hyperparameters: A Novel ES‐SIFG Approach
To accurately predict groundwater flow and solute transport, it is essential to precisely characterize the highly heterogeneous aquifer conditions. Ensemble smoother with multiple data assimilation (ESMDA), though widely applied to identify aquifer properties and spatial features, encounters severe problems in practice due its fundamental assumptions of linearity and Gaussianity. To tackle this challenge, we first use a spatially‐informed field generator (SIFG) to hyperparameterize the conductivity field and encode position information into the hyperparameters, and then combine it with ensemble smoother to form a new inversion framework called ensemble smoother with SIFG (ES‐SIFG). In ES‐SIFG, followed by utilizing an ensemble smoother to update the hyperparameters rather than the aquifer parameters. The main innovation of ES‐SIFG is integrating positional information into hyperparameters, enabling the use of distance‐based covariance localization (CL) and significantly reducing the number of model simulations. The proposed method has been tested on parameter identification problems in 2‐D and 3‐D non‐Gaussian aquifers and compared to ESMDA with normal score transformation. Results indicate that ES‐SIFG outperforms ESMDA and is capable of accurately identifying non‐Gaussian aquifer parameters and reconstructing contaminant release history, particularly in resolving equifinality and preserving prior geological structure. Furthermore, SIFG allows usage of CL between hyperparameters and observations, which ensures the stable convergence of data assimilation processes even with very small ensemble sizes.
- Open Access
- 28 April 2025
- Published by Wiley
Improving Heterogeneous Aquifer Characterization Using A Deep Learning-Based Ensemble Smoother With A Hybrid Prior Strategy
Accurate characterization of aquifer hydraulic properties is crucial for effective groundwater resource management and pollution control. However, aquifer heterogeneity introduces high-dimensional and non-Gaussian complexities that pose significant challenges to conventional data assimilation methods. To address these limitations, we have proposed a deep learning-based ensemble smoother, $\text{ES}_{\text{DL}}$, which constructs a non-linear mapping from the innovation vector to the update vector, thereby enabling automatic detection and handling of non-Gaussian patterns in the data. While many existing studies assume unbiased prior knowledge of aquifer properties, real-world applications frequently involve biased priors that can significantly compromise characterization accuracy. To mitigate this issue, this study introduces a hybrid prior strategy that combines multiple types of prior information, reducing reliance on any single, potentially biased, prior assumption. This strategy enables $\text{ES}_{\text{DL}}$ to learn innovation–update relationships from an ensemble of diverse prior assumptions, enhancing its robustness in complex scenarios where unbiased priors are unavailable. We systematically evaluate the performance of $\text{ES}_{\text{DL}}$ under unbiased, biased, and hybrid priors across hydraulic conductivity fields characterized by log-Gaussian, channelized, and three-facies distributions, using the Kalman-based ensemble smoother ($\text{ES}_{\text{K}}$) as a benchmark. Results demonstrate that the hybrid prior strategy significantly enhances the robustness of aquifer characterization under biased prior assumptions, with $\text{ES}_{\text{DL}}$ consistently outperforming $\text{ES}_{\text{K}}$ in terms of both accuracy and stability. Notably, $\text{ES}_{\text{DL}}$ maintains high effectiveness even when the hybrid prior ensemble excludes unbiased samples, highlighting its adaptability and reliability in complex groundwater systems.- 28 April 2025
- Published by Wiley
Open AccessImproving Heterogeneous Aquifer Characterization Using A Deep Learning-Based Ensemble Smoother With A Hybrid Prior Strategy
Accurate characterization of aquifer hydraulic properties is crucial for effective groundwater resource management and pollution control. However, aquifer heterogeneity introduces high-dimensional and non-Gaussian complexities that pose significant challenges to conventional data assimilation methods. To address these limitations, we have proposed a deep learning-based ensemble smoother, $\text{ES}_{\text{DL}}$, which constructs a non-linear mapping from the innovation vector to the update vector, thereby enabling automatic detection and handling of non-Gaussian patterns in the data. While many existing studies assume unbiased prior knowledge of aquifer properties, real-world applications frequently involve biased priors that can significantly compromise characterization accuracy. To mitigate this issue, this study introduces a hybrid prior strategy that combines multiple types of prior information, reducing reliance on any single, potentially biased, prior assumption. This strategy enables $\text{ES}_{\text{DL}}$ to learn innovation–update relationships from an ensemble of diverse prior assumptions, enhancing its robustness in complex scenarios where unbiased priors are unavailable. We systematically evaluate the performance of $\text{ES}_{\text{DL}}$ under unbiased, biased, and hybrid priors across hydraulic conductivity fields characterized by log-Gaussian, channelized, and three-facies distributions, using the Kalman-based ensemble smoother ($\text{ES}_{\text{K}}$) as a benchmark. Results demonstrate that the hybrid prior strategy significantly enhances the robustness of aquifer characterization under biased prior assumptions, with $\text{ES}_{\text{DL}}$ consistently outperforming $\text{ES}_{\text{K}}$ in terms of both accuracy and stability. Notably, $\text{ES}_{\text{DL}}$ maintains high effectiveness even when the hybrid prior ensemble excludes unbiased samples, highlighting its adaptability and reliability in complex groundwater systems. - 19 February 2025
- Published by Elsevier in Advances in Water Resources
- Vol. 199
Effects of tides and subsurface dams on the land-sourced contaminant transport: Laboratory and numerical investigation
- 19 February 2025
- Published by Elsevier in Advances in Water Resources
- Vol. 199
Effects of tides and subsurface dams on the land-sourced contaminant transport: Laboratory and numerical investigation
- Open Access
- 8 March 2024
- Published by Copernicus GmbH
Effective Characterization of Fractured Media with A Novel Deep Learning-Based Data Assimilation Approach
In various research fields such as hydrogeology, environmental science and energy engineering, geological formations with fractures are frequently encountered. Accurately characterizing these fractured media is of paramount importance when it comes to tasks that demand precise predictions of liquid flow and the transport of solutes and energy within them. Since directly measuring fractured media poses inherent challenges, data assimilation (DA) techniques are typically employed to derive inverse estimates of media properties using observed state variables like hydraulic head, concentration, and temperature. Nonetheless, the considerable difficulties arising from the strong heterogeneity and non-Gaussian nature of fractured media have diminished the effectiveness of existing DA methods. In this study, we formulate a novel DA approach known as PEDL (Parameter Estimator with Deep Learning) that harnesses the capabilities of DL to capture non-linear relationships and extract non-Gaussian features. To evaluate PEDL’s performance, we conduct two numerical case studies of increasing complexity. Our results unequivocally demonstrate that PEDL outperforms three popular DA methods: Ensemble Smoother with Multiple Data Assimilation (ES-MDA), Iterative Local Updating Ensemble Smoother (ILUES), and ES with DL-based Update (ES-DL). Sensitivity analysis confirms PEDL’s validity and adaptability across various ensemble sizes and DL model architectures. Furthermore, even in scenarios where a mismatch exists between the reference model and the forecast model’s structure, PEDL adeptly identifies the primary characteristics of actual fractures.
- 8 March 2024
- Published by Copernicus GmbH
Open AccessEffective Characterization of Fractured Media with A Novel Deep Learning-Based Data Assimilation Approach
In various research fields such as hydrogeology, environmental science and energy engineering, geological formations with fractures are frequently encountered. Accurately characterizing these fractured media is of paramount importance when it comes to tasks that demand precise predictions of liquid flow and the transport of solutes and energy within them. Since directly measuring fractured media poses inherent challenges, data assimilation (DA) techniques are typically employed to derive inverse estimates of media properties using observed state variables like hydraulic head, concentration, and temperature. Nonetheless, the considerable difficulties arising from the strong heterogeneity and non-Gaussian nature of fractured media have diminished the effectiveness of existing DA methods. In this study, we formulate a novel DA approach known as PEDL (Parameter Estimator with Deep Learning) that harnesses the capabilities of DL to capture non-linear relationships and extract non-Gaussian features. To evaluate PEDL’s performance, we conduct two numerical case studies of increasing complexity. Our results unequivocally demonstrate that PEDL outperforms three popular DA methods: Ensemble Smoother with Multiple Data Assimilation (ES-MDA), Iterative Local Updating Ensemble Smoother (ILUES), and ES with DL-based Update (ES-DL). Sensitivity analysis confirms PEDL’s validity and adaptability across various ensemble sizes and DL model architectures. Furthermore, even in scenarios where a mismatch exists between the reference model and the forecast model’s structure, PEDL adeptly identifies the primary characteristics of actual fractures.