Groundwater fluxes and interacting zones between groundwater and surface water are crucial for understanding the water dynamics of the critical zone. Groundwater within the critical zone plays a significant role in the ecosystem, biodiversity, and water supply. However, estimating these fluxes remains a key challenge because they are not directly measured in the field. Model calibration involves adjusting key parameters—such as saturated hydraulic conductivity and soil-water retention properties—using observed data like hydraulic head and river discharge, while initial and boundary conditions are prescribed to define the model l setup. That calibration is often done by comparing simulated soil water saturation and water table level to piezometers. Nevertheless, real flows occur in 3D in a complex medium containing heterogeneities with various lithologies, with different hydraulic parameters such as hydraulic conductivity and porosities.Geophysical methods such as electrical resistivity tomography (ERT), seismic refraction tomography (SRT), and multichannel analysis of surface wave (MASW), which are sensitive to lithology, content, and nature of fluid, represent helpful tools for hydrogeological modelling, both in terms of model parameterization and physical property characterization. ERT, which is particularly sensitive to lithology, allows us to identify and delineate heterogeneities, while seismic methods, which are sensitive to mechanical properties, will enable us to infer the water saturation and the piezometric surface in the near surface through the P-wave and S-wave velocities ratio (Poisson’s ratio, e.g. ) (Dangeard et al., 2021).We propose a workflow combining geophysics and geostatistics to reconstruct the heterogeneities and the piezometric surface in an alluvial plain context. We implemented the workflow in a 30 x 30 m area at the Avenelles site of the Orgeval Critical Zone Observatory (CZO), which is part of the French network of CZOs OZCAR. ERT, SRT, and MASW surveys were carried out along 7 profiles to obtain 2D sections of electrical resistivity, P and S wave velocities (6 profiles of 72 electrodes/geophones and one profile of 48 electrodes/geophones, with 0.40 m spacing leading to 12,708 apparent resistivity data, 33,888 first wave arrival picks, and 277 dispersion curves). Geophysics allows us to pass from punctual piezometer data to 2D vertical sections. However, carrying out 3D geophysical acquisition is cumbersome. To overcome these limitations, we then use geostatistics to get a distribution of our geophysical parameters in the 3D volume delineated by the geophysical survey. Once the 3D interpolation is done by kriging methods, we can retrieve a view of the heterogeneities distribution in the near surface as well as the water table position to inform hydrogeological inversion. Furthermore, with the addition of petrophysical relationships, it is possible to estimate saturation and porosity distribution for a future 3D hydrogeological physics-based model run to better characterise groundwater fluxes. Finally, all these workflows, including complementary methods, could be performed on different dates for time-lapse monitoring of the water table.

In a context where low-carbon transportation is increasingly essential, the diagnosis and maintenance of railway infrastructures have become critical challenges. Current assessment techniques still rely heavily on destructive testing of embankments, sublayers, and underlying soils. These structures are also exposed to more frequent and less predictable extreme climatic events, threatening their mechanical integrity and long-term stability. High-density, high-resolution geophysical methods therefore offer a compelling non-destructive alternative, particularly for characterizing and monitoring soil mechanical properties. Over the past decade, major advances have been made in seismic acquisition, processing, and interpretation. We present an overview of our recent contributions, primarily based on surface-wave methods, which require low-energy sources and are well adapted to railway environments. We developed high-yield acquisition strategies using landstreamers, combined with classic active and passive (train) sources. Stacking and interferometry-based approaches extract multimodal dispersion images, enabling detection of lateral variations within embankments and continuous site monitoring. Deep learning supports semi-automatic picking and interpretation, while Bayesian inversion and physics-aware AI models integrate geotechnical and hydrogeological data to rapidly infer petro- and hydro-facies and track water-table dynamics. We conclude with a roadmap for future developments, including the integration of distributed acoustic sensing.

Groundwater systems are components of the Critical Zone, in dynamic balance with the climate and human pressure. Water and heat fluxes control biogeochemical processes and regulate both resource potential and system vulnerability. Evaluating system responses to climate and land use changes requires an estimation of subsurface structure, flow dynamics, and the spatial distribution of recharge and discharge zones. Classical hydrological data, such as water levels and river discharge, are commonly coupled with unconventional methodologies, including heat tracing, electrical resistivity tomography, and surface wave analysis. Their coupling via time-lapse monitoring and inversion frameworks enables the characterization of thermal, electrical, and mechanical interactions, thus encouraging a transition from static structure description to transient representations of groundwater. We identify key challenges, including data limitations, model uncertainty propagation, and the integration of transient variables and parameters into inversion workflows. Petrophysical and geostatistical approaches help overcome these issues by coupling geophysical and groundwater models, quantifying spatio-temporal uncertainties, and addressing scale change. Finally, choices regarding experimental design, parameter reduction, dimensionality, model hypotheses, and inversion scale must be assessed to balance parsimony with model accuracy. These developments underscore the central role of hydrogeophysics in advancing Critical Zone science and sustainable groundwater management. Emphasizing transient processes is essential for capturing how subsurface systems evolve over time in response to environmental changes.

By 2050, climate changed-induced extreme heat waves and urban heat island phenomenon will lead to severe soil drought, which can affect trees health and sustainability of green spaces. By-product of the pyrolysis of biomass, biochar incorporation can favor soil water retention and can be seen as a potential tool to mitigate future soil drought. In order to test the ability of biochar incorporation into soil to improve water retention, experimentations are necessary to evaluate its efficiency. To do so, experimental tests involving monitoring would require destructive, costly and sparse spot measurements that are not representative of the heterogeneous nature of studied technosols. This work explores how geophysics can overcome such limitations, by enabling spatialized, non-invasive monitoring of soils at the scale of experimental plots. Control and biochar-amended plots (n > 3) were monitored over time with spectral induced polarization (SIP) and active seismic methods. Initial results suggest that SIP can distinguish the presence of biochar through marked contrasts in resistivity and phase, which evolve with time suggesting a potential monitoring of biochar aging (e.g. fragmentation, migration and oxidation). At the same time, seismic surface-wave velocity measurements show sensitivity to seasonal variations, as well as a quasi-systematic decrease in velocities in amended soils, which could reflect changes in porosity and/or water content. This approach serves as a proof of concept for highlighting the potential of geophysics as an in situ diagnostic tool, capable of monitoring the effect of biochar by providing reliable and integrative indicators of water content in heterogenous urban soils.

To effectively address engineering challenges and risks, it is crucial to characterize mechanical properties of near-surface environments. The Multichannel Analysis of Surface Waves (MASW) has proven to be a valuable active seismic imaging technique by providing near-surface shear (S)-wave velocities estimations. However, its application to urban areas requires further development. This study leverages well-constrained experimental sites to assess the viability of a passive-MASW technique, utilizing seismic waves induced by high-speed train traffic instead of conventional active sources. We suggest employing short 96-geophone uniform linear arrays to capture surface waves in a broad frequency band (10-200 Hz). Train passages are automatically detected and categorized regarding to the train travel direction. Seismic interferometry and phase-weighted stack techniques are applied to generate virtual shot-gathers that are transformed into high-resolution multi-modal dispersion images. Our results demonstrate a strong coherence between the picked dispersion curves from the passive-MASW approach and those obtained through traditional active MASW with a hammer source. We discuss the validity of higher modes and explore array density limits to ensure reliable results. Our findings highlight that seismic interferometry, coupled with a high phase-weighted stack power, effectively recovers energy at high frequencies, enhancing the characterization of multi-modal surface-wave dispersion associated with thin near-surface layers.

Soil nature and water saturation of the near-surface directly influence land use resilience and sustainability in urban areas facing intense climate forcing and human activity. Seismic methods are widely applied in this context for subsurface characterization and monitoring, but often fall short in delivering joint geological, hydrogeological, and geomechanical descriptions. We explore the effectiveness of a passive seismic approach, coupled with artificial intelligence (AI), to characterize geological structures and capture groundwater dynamics. We present a deterministic inversion technique powered by a language model, to translates seismic wave velocity measurements into petrophysical parameters in the form of textual descriptions. Results successfully delineate subsurface structures with their respective composition and mechanical characteristics, while accurately predicting daily water table levels. Validation demonstrates high accuracy, with a normalized root-mean-square error of 8%, while delivering fast insights into subsurface conditions. This underscores the potential of AI to enhance subsurface characterization across multiple scales.

Monitoring underground water reservoirs is challenging due to limited spatial and temporal observations. This study presents an innovative approach utilizing supervised deep learning (DL), specifically a multilayer perceptron (MLP), and continuous passive-Multichannel Analysis of Surface Waves (passive-MASW) for constructing 2D water table height maps. The study site, geologically well-constrained, features two 20-meter-deep piezometers and a permanent 2D geophone array capturing train-induced surface waves. For each point of the 2D array, dispersion curves (DCs), displaying Rayleigh-wave phase velocities (VR) across a frequency range of 5 to 50 Hz, have been computed each day between December 2022 and September 2023. In the present study, these DCs are sampled in wavelengths ranging from 4.5 to 10.5 m in order to focus the monitoring on the expected water table depths. All VR data around one of the two piezometers is used to train the MLP model. Water table heights are then predicted across the entire geophone array, generating daily 2D piezometric maps. Model's performance is tested through cross-validation and comparisons with water table data at the second piezometer. Model’s efficiency is quantified with the root-mean-square error (RMSE) and the coefficient of determination (R²). A R² is estimated above 80 % for data surrounding the training piezometer and above 55 % for data surrounding the test piezometer. Additionally, the RMSE is impressively low at 0.03 m at both piezometers. Results showcase the effectiveness of DL in generating predictions of water table heights from passive-MASW data. This research contributes to advancing our understanding of subsurface hydrological dynamics, providing a valuable tool for water resource management and environmental monitoring. The ability to predict 2D piezometric maps from a single piezometer is particularly noteworthy, offering a practical and efficient solution for monitoring water table variations across broader spatial extents.

Assessing Railway Earthworks (RE) requires non-destructive and time-efficient diagnostic tools. This study evaluates the relevance of shear-wave velocity (V_s) profiling using Multichannel Analysis of Surface Waves (MASW) for detecting Low Velocity Layers (LVLs) in disturbed RE zones. To enhance time-efficiency, a towed seismic setup (Landstreamer) was compared with a conventional one. Once qualified, the Landstreamer was deployed on the ballast for roll-along acquisition, showing greatly improved efficiency and good imaging capability. A probabilistic framework adopted in this study additionally enhances quantification of uncertainties and helps in interpretation of V_s models, facilitating reliable decision-making in infrastructure management.

Quantifying the water and heat fluxes at the interface between surface water (SW) and groundwater (GW) is a key issue for hydrogeologists to consider for safe yield and good water quality. However, such quantification with field measurements is not straightforward because the SW-GW changes depend on the boundary conditions and the spatial description of the hydrofacies, which aren't well known and are usually guessed by calibrating models using standard data like hydraulic heads and river discharge. We provide a methodology to build stronger constraints to the numerical simulation and the hydrodynamic and thermal parameter calibration, both in space and time, by using a multi-method approach. Our method, applied to the Orgeval Critical Zone Observatory (France), estimates both water flow and heat fluxes through the SW-GW interface using long-term hydrological data, time-lapse seismic data, and modeling tools. We show how a thorough interpretation of high-resolution geophysical images, combined with geotechnical data, provides a detailed distribution of hydrofacies, valuable prior information about the associated hydrodynamic property distribution. The temporal dynamic of the WT table can be captured with high-resolution time-lapse seismic acquisitions. Each seismic snapshot is then thoroughly inverted to image spatial WT variations. The long-term hydrogeological data (such as hydraulic head and temperature) and this prior geophysical information are then used to set the parameters for the hydrogeological modeling domain. The use of the WT geometry and temperature data improves the estimation of transient stream-aquifer exchanges.

Quantifying water and heat fluxes at the surface water (SW)–groundwater (GW) interface is crucial for ensuring sustainable water management and quality. However, direct field-based quantification remains challenging due to the dynamic nature of SW-GW interactions, which are influenced by poorly constrained boundary conditions and spatial hydrofacies distributions. Traditionally, these parameters are inferred through model calibration using conventional data, such as hydraulic heads and river discharge. Many regional studies have treated rivers as curvilinear GW divides, with flow either converging toward or diverging from the river center—an assumption rooted in Tóth’s theory, which correlates surface and subsurface drainage boundaries. However, this oversimplification fails to account for geological heterogeneity, river morphology, variable hydraulic conditions, and anthropogenic influences like withdrawals. While regional-scale studies commonly examine SW-GW exchanges, their coarse resolution limits the ability to resolve localized hydraulic gradients. Understanding flow dynamics in heterogeneous environments, such as alluvial plains, requires a more detailed, integrated approach. Here, we present a multi-method framework to strengthen numerical simulations and improve hydrodynamic and thermal parameter calibration in both space and time. Applied to the Orgeval Critical Zone Observatory (France), our approach estimates SW-GW fluxes using a combination of long-term hydrological data (10 years), time-lapse seismic imaging, and numerical modeling. We demonstrate how high-resolution geophysical imaging, combined with geotechnical data, enables a detailed characterization of hydrofacies and provides valuable prior constraints on hydrodynamic properties. Time-lapse seismic acquisitions offer a high-resolution view of groundwater table (WT) dynamics, with each seismic snapshot carefully inverted to capture spatial WT variations. By integrating these geophysical insights with long-term hydrogeological observations (hydraulic head and temperature), we refine parameterization within the hydrogeological modeling domain, leading to improved estimates of transient stream-aquifer exchanges. Finally, we outline future steps toward achieving a fully coupled hydrogeophysical model to further enhance SW-GW interaction predictions.

Soil moisture is an essential ecosystem resource and a major control of the Earth's hydrological cycle and energy balance, closely interacting with the climate system. However, investigating deep soil moisture dynamics at large scales presents significant challenges due to the sparse distribution and limited spatial representativeness of in situ monitoring networks, while various remote sensing methods mainly address surface soil moisture within the top few centimeters. This study illustrates how seismic waves can effectively detect variations in deep soil moisture. We examine continuous seismic data from 791 stations across South-Central Europe for the period 2016-2020. Our findings confirm a strong correlation between variations in seismic velocity and deep soil moisture content. Notably, the seismic observations pinpoint areas impacted by severe soil moisture deficits related to the 2016-2017 European drought event. The seismic method presented in this study offers new opportunities in addressing the observational gap of this critical environmental parameter.

The Orgeval Long-term Research Observatory, part of the French critical zone network (OZCAR RI), is a 104 km² agricultural catchment, located 70 km east of Paris, in France. The Orgeval catchment is representative of intensive agriculture (80 % of its total area), the main land use in the Seine river basin. For more than 50 years, both quantity and quality of water are monitored throughout the catchment, from sub-hourly to yearly time scales. This rich dataset allows improving the understanding of critical zone structure and reactivity, in a holistic and interdisciplinary approach. Specific basic and applied research topics relate to extreme hydrological events, agricultural tile drainage, land use planning, and more generally the evolution of agricultural activities facing climate change and urban growth. The Orgeval research observatory is a unique testbed to investigate the functioning and evolution of the Critical Zone. Multidisciplinary approaches are implemented thanks to collaborations between research institutes and universities, combining knowledge and methods from different disciplines, such as hydrology, ecology, biogeochemistry, geophysics and socioeconomics. Created in 1962, the observatory was initially devoted to study floods and weathering research questions. Since then, is evolved towards other societal and environmental topics. Initiated in 1975, lots of monitoring and research was dedicated to diffuse agricultural pollution, especially nitrates, which contributed to a better understanding of the interactions between agricultural activities and surface and groundwater quality. Since the 2000’s, research questions opened to pesticides and biology and biodiversity. The Orgeval observatory is also highly adapted to develop technological innovations, such as in situ biochemical monitoring. A multi-scale observation strategy is implemented in both space and time, ranging from local (with more than 80 monitoring sites) to regional scale, and from time-lapse campaigns to high-frequency measurements (from 1 Hz for geophysics to 1 h or 1 week for chemistry), most often with a long-term approach. The main measurements include: Water level and water discharge : at the outlet of each sub-catchment. groundwater level : in piezometers in the riverbanks and in the aquifers. Precipitation : in addition to Météo-France stations. Main weather variables : air temperature, humidity, and radiation. Soil moisture : from the soil surface to a depth of 1.5 meters. Water quality : dissolved organic and inorganic carbon, dissolved gases (O 2 , CO 2 , Rn), major and trace ions, nutrients, but also water, carbon or strontium stable isotopes. This includes the RiverLab prototype, installed in June 2015 at the outlet of Avenelles sub-catchment for high-frequency measurement (every 30 minutes) of the river's chemical composition. Organic and inorganic contaminants : pesticides but also metals and microplastics in surface and groundwater. Ecotoxicological and ecological indices : ecological assessment. Surface and groundwater temperature : using heat as a tracer of surface-groundwater exchanges. Hydrogeophysics : ERT, GPR, seismic, especially in the riparian areas. Borehole core samples and logging : lithofacies description and characterization. Water level and water discharge : at the outlet of each sub-catchment. groundwater level : in piezometers in the riverbanks and in the aquifers. Precipitation : in addition to Météo-France stations. Main weather variables : air temperature, humidity, and radiation. Soil moisture : from the soil surface to a depth of 1.5 meters. Water quality : dissolved organic and inorganic carbon, dissolved gases (O 2 , CO 2 , Rn), major and trace ions, nutrients, but also water, carbon or strontium stable isotopes. This includes the RiverLab prototype, installed in June 2015 at the outlet of Avenelles sub-catchment for high-frequency measurement (every 30 minutes) of the river's chemical composition. Organic and inorganic contaminants : pesticides but also metals and microplastics in surface and groundwater. Ecotoxicological and ecological indices : ecological assessment. Surface and groundwater temperature : using heat as a tracer of surface-groundwater exchanges. Hydrogeophysics : ERT, GPR, seismic, especially in the riparian areas. Borehole core samples and logging : lithofacies description and characterization. All these data support the development and use of numerical models, for scientific questions but also for environmental impact assessment and territorial management. This catchment is used to develop and/or validate numerical methodologies at the headwater catchment scale, in parallel to larger scale modeling (typically for the Seine river basin)

Railway Trackbed (RT), which collectively describes the subgrade structures that support rail tracks, is of great importance to the effective maintenance and rehabilitation of the rail network. Therefore, a comprehensive understanding of the mechanical condition of Railway Earthwork (RE) is necessary. The development of non-destructive and efficient methods for the characterization of REs is a priority. Previous studies have investigated the potential of surface waves for the characterization of RE. Preliminary results indicate that this approach is effective, particularly when using high yield acquisition setup such as landstreamer. However, these instruments generate an amount of data that necessitates optimized and automated processing. The potential of Deep Learning (DL) to automate the processing of surface wave data is being explored. In this study, the primary objective is to identify the energy maxima and propagation modes in dispersion images. A supervised convolutional neural network (CNN), designated as ‘U-Net’, was selected to perform segmentation tasks. This model, called ‘U2-pick’, integrates two U-net architectures: one for maxima identification and another for propagation mode identification. The training dataset was constructed using synthetic data that is representative of a French High-Speed-Line (HSL) RE structure. The preliminary outcomes on the synthetic datasets are encouraging, demonstrating accurate identification of energy maxima and mode classification. However, the predictions made on field datasets revealed that while the energy peaks were identified with a high degree of accuracy, the mode assignment proved to be less satisfactory, especially in the case of higher modes. Finally, a comparison of the accuracy and picking time was performed using more standard tools like maxima search, semi-automatic, and Machine Learning (ML) tools.

Monitoring groundwater tables (GWTs) remains challenging due to limited spatial and temporal observations. This study introduces an innovative approach combining an artificial neural network, specifically a multilayer perceptron (MLP), with continuous passive Multichannel Analysis of Surface Waves (passive-MASW) to construct GWT depth maps. The geologically well-constrained study site includes two piezometers and a permanent 2D geophone array recording train-induced surface waves. At each point of the array, dispersion curves (DCs), displaying Rayleigh-wave phase velocities VR $\left({V}_{R}\right)$ over a frequency range of 5-50 Hz, were measured daily from December 2022 to September 2023, and latter resampled over wavelengths from 4 to 15 m, to focus on the expected GWT depths (1-5 m). Nine months of daily VR ${V}_{R}$ data near one piezometer, spanning both low and high water periods, were used to train the MLP model. GWT depths were then estimated across the geophone array, producing daily GWT maps. The model's performance was evaluated by comparing inferred GWT depths with observed measurements at the second piezometer. Results show a coefficient of determination (R2) of 80% at the training piezometer and of 68% at the test piezometer, and a remarkably low root-mean-square error (RMSE) of 0.03 m at both locations. These findings highlight the potential of deep learning to estimate GWT maps from seismic data with spatially limited piezometric information, offering a practical and efficient solution for monitoring groundwater dynamics across large spatial extents.

Increasing anthropogenic and climate pressures on water resources and thermal energy call for a better understanding of the transient water storage and the water fluxes within the Critical Zone (CZ). Recharge, as the main water inflow feeding groundwater (GW), is critical for the proper management of GW systems. GW recharge is defined as the water percolating from the last unsaturated horizon down to the water table and is therefore broadly inaccessible to direct observations. Recharge is spatially heterogeneous and controlled by multiple factors such as porous media properties and hydrogeological conditions. Hydrogeophysics provide valuable approaches to determining hydraulic parameters in unconsolidated and unsaturated soils. In this domain, electromagnetic and electrical methods predominate due to their obvious dependence on water content. While crucial for water-related assessments, the transition to mechanical properties emphasizes the complementary role of seismic techniques. Specifically, seismic refraction tomography and surface-wave dispersion analysis stand out in delimiting boundaries between saturated and unsaturated zones. Recent studies underscore the synergy of employing both 2D electrical and seismic methods, showcasing their collective efficacy in identifying hydrofacieses and delineating the water table. However, these techniques fall short of providing a detailed saturation profile in the unsaturated zone. Recent studies suggest to employ the Van Genuchten model, coupled with a rock physics model that incorporates capillary suction effects, to determine the mechanical properties of the soil, accounting for both depth and saturation dependencies. This method enables the analytical 1D modeling of both P- and S-wave velocities in various hydrofacieses with various water table depths (in static conditions). Then by utilizing these velocity models, it is possible to calculate synthetic P-wave travel times (P-TT) and surface-wave dispersion (SWD) from an artificial seismic setup. This constitute a forward problem from saturation versus depth models towards seismic data. In this study, we propose to do the inverse problem, e.g. estimating the VG parameters (VG) from P-TT and SWD. We use the database provided by Carsell and Parrish to compute synthetic observations in wide a priori ranges. We propose the employment of a straightforward grid search and formulate the results in a Bayesian framework. Our results indicate that both SWD and P-TT are responsive to changes in water saturation, allowing for the retrieval of the VG parameters from observed data. Moreover, our study highlights that the sensitivity of geophysical data varies with soil composition, particularly underscoring the complexities of estimating VG parameters in soils with a high sand content.

Pressure (P) or shear (S)-wave velocity models of the near-surface can be simultaneously estimated along coincident arrays from P-wave refraction tomography and surface-wave (SW) dispersion inversion methods. Over the past decade, this approach has been integrated into the hydrogeophysics toolbox to image spatial variations of VP/VS (or Poisson) ratio, as its evolution is strongly associated with water content (or saturation) contrasts. The relevance of this method has been verified in various Critical Zone (CZ) observatories, each with distinct hydrogeological characteristics such as continuous multi-layered hydrosystems or fractured environments with strong discontinuities. It has also proven successful in other contexts and application scales, including a hydrothermal site or partially saturated glass beads in a laboratory experiment. However, we identified two major issues: (1) the combined use of P-wave traveltime tomography and SW dispersion inversion involves distinct characteristics of the wavefield and different assumptions about the medium, providing VP and VS models with different sensitivity, resolution, investigation depth, and posterior uncertainties; (2) the involved inversion processes use a small number of layers that cannot properly describe the continuous variations of subsurface hydrological properties. In particular, we noted that VP/VS (or Poisson) ratio was only consistent with strong saturation contrasts and often faced difficulties in retrieving water content variations in the unsaturated zone. This underscores the need to use petrophysical approaches to build alternative forward models and improve inversion processes. Adapted rock physics models have thus recently been developed to take capillary suction effects into account in the effective stress of the soil. In this study, we first present several datasets obtained from various contexts in which SW dispersion variations have been observed and related to changes in water content and/or water table depths. We then suggest using the previously cited rock physics models to simulate these data and show how it helps in understanding the involved hydrofacieses and processes. We finally address the relevance of surface-wave dispersion inversion approaches involving such forward models and discuss the possible use of additional attributes of the seismic wavefield to constrain interpretations.

Monitoring groundwater tables (GWTs) is challenging due to limited spatial and temporal observations. This study presents an innovative approach utilizing supervised deep learning, specifically a Multilayer Perceptron (MLP), and continuous passive-Multichannel Analysis of Surface Waves (passive-MASW) for constructing 2D GWT level maps. The study site, geologically well-constrained, features two 20-meter-deep piezometers and a permanent 2D geophone array capturing train-induced surface waves. For each point of the 2D array, dispersion curves (DCs), displaying Rayleigh-wave phase velocities (V_R) across a frequency range of 5 to 50 Hz, have been computed each day between December 2022 and September 2023. In the present study, these DCs are resampled in wavelengths ranging from 4 to 15~m in order to focus the monitoring on the expected GWT levels (between -1 and -5 m). Nine months of daily V_R data around one of the two piezometers is used to train the MLP model. GWT levels are then estimated across the entire geophone array, generating daily 2D GWT maps. Model’s performance is tested through cross-validation and comparisons with GWT level data at the second piezometer. Model’s efficiency is quantified with the root-mean-square error (RMSE) and the coefficient of determination (R²). The R² is estimated at 80% for data surrounding the training piezometer, and at 68% for data surrounding the test piezometer. Additionally, the RMSE is impressively low at 0.03 m at both piezometers. Results showcase the effectiveness of DL in estimating GWT level maps from passive-MASW data, offering a practical and efficient monitoring solution across broader spatial extents.

The dynamic cone penetrometer (DCP) provides local soil resistance information. The difference in the vertical and horizontal data resolution (centimetric vs. multi-metric) makes it difficult to spatialize the DCP data directly. This study uses a high-resolution Vs$V_s$ section, extracted by the seismic surface-wave method, as the auxiliary and physical constraint for mapping the DCP index (DCPI). Geostatistical formalism (kriging and cokriging) is used. The associated measurement error of the seismic surface-wave data is also included in the cokriging system, that is, the cokriging with variance of measurement error (CKVME). The proposed methods are validated for the first time on a test site designed and constructed for this study, with known geotechnical perspectives. Seismic and high-intensity DCP campaigns were performed on the test site. The results show that with decimating the number of DCP soundings, the kriging approach is no longer capable of estimating the lateral variation in the test site, and the root-mean-square error (RMSE) value of the kriging section is increased by 87%$87\%$. With the help of Vs$V_s$ sections constraining the lateral variability model, the RMSE values of the cokriging and the CKVME sections are increased by 25%$25\%$ and 17%$17\%$.

Effective structural assessment of urban infrastructure is essential for sustainable land use and resilience to climate change and natural hazards. Seismic wave methods are widely applied in these areas for subsurface characterization and monitoring, yet they often rely on time-consuming inversion techniques that fall short in delivering comprehensive geological, hydrogeological, and geomechanical descriptions. Here, we explore the effectiveness of a passive seismic approach coupled with artificial intelligence (AI) for monitoring geological structures and hydrogeological conditions in the context of sinkhole hazard assessment. We introduce a deterministic petrophysical inversion technique based on a language model that decodes seismic wave velocity measurements to infer soil petrophysical and mechanical parameters as textual descriptions. Results successfully delineate 3D subsurface structures with their respective soil nature and mechanical characteristics, while accurately predicting daily water table levels. Validation demonstrates high accuracy, with a normalized root mean square error of 8 while delivering broader insights into subsurface conditions 2,000 times faster. These findings underscore the potential of advanced AI techniques to significantly enhance subsurface characterization across diverse scales, supporting decision-making for natural hazard mitigation.