Search

In the underground mining process, the evolution of fissures, fractures, and breakages in overlying rock strata can lead to water inrush and many other serious risks, including rock bursts, roof collapses, and ground subsidence. Since directly observing strata behavior is challenging, downscaled similar material simulations are often relied upon as a proven effective method for strata pattern analysis. Traditionally, mere visual inspection or stress and displacement monitoring of the simulation have been employed for risk analysis and often cause imprecise and subjective interpretations. With recent advancements in deep learning image processing models, which enabled quantitative studies on variation dynamics of image sequences, this study proposes a novel method for analyzing strata behavior patterns by applying the spatiotemporal data feature extracting capability of the convolutional long short-term memory (ConvLSTM) model on a carefully conducted similar material experiment. The approach involves reconstructing the preprocessed image sequence data using ConvLSTM, thereby predicting and enhancing visual inspection by dynamically highlighting major strata fracture and breakage evolvements. Critical periods of rapid strata behavior are also reliably identified by thresholding the reconstruction errors. In addition, high-risk areas in each critical period are delineated using image differentiation and the fast nonlocal means (NLM) denoising algorithm. While other applicable models, including ConvLSTM-autoencoder (ConvLSTM-AE) and temporal convolutional network (TCN)-AE, also perform well, the predictive ConvLSTM model stands out with its superior reconstruction visual distinctiveness and model stability in the analysis. With provided valuable spatiotemporal information for understanding mining-induced strata behavior patterns, the obtained promising results offer precise dynamic guidance on predicting the formation of water inflow channels.

Microseismic monitoring during mining operations generates spatiotemporal data that could indicate strata fractures and deformations leading to water inrush anomalies. However, current water inrush prediction methods face challenges from the data nonstationarity and multidimensionality, resulting in low prediction precision and effectiveness. This study proposes an innovative data-driven approach for predicting mining water inrush using field 3-D microseismic monitoring data. The approach couples machine learning and deep learning models to analyze microseismic events, preprocessed using the density-based spatial clustering of applications with noise (DBSCAN) and the random sample consensus (RANSAC) algorithms for both data denoising and water inrush risk locating. Weighting periods are analyzed in periodic variations of event attributes using the fast Fourier transform (FFT), continuous wavelet transform (CWT), empirical mode decomposition (EMD), and seasonal and trend decomposition using loess (STL) methods. Anomalies are detected using the long short-time memory (LSTM) + absolute error (AE), isolation forest (iForest), and LSTM + iForest models. The study is conducted using a microseismic dataset acquired during intermittent water inflow anomalies in the Xingdong coal mine in China. The approach accurately predicts a major water inrush incident hours prior to its occurrence merging detected anomalies with the obtained weighting periods, which are also used for model calibration. Future studies could focus on the performance evaluation and calibration of the deep learning models using microseismic datasets from different mining operations, and expanding the approach’s scope by incorporating other geophysical exploration technologies like the electrical methods to further study the presence and movement of water in mines for improving mining safety.

Frost heave failure often occurs after the completion of tunnels in seasonal frozen areas. The safe operation of tunnels is impacted by freeze–thaw-damaged surrounding rocks. Therefore, the deterioration mechanism of surrounding rocks during a freeze–thaw cycle in a seasonally frozen area is of great significance for tunnel stability analysis. This study investigates the physical and mechanical characteristics of Houwai tunnel limestone samples after freeze–thaw cycles. Testing was conducted using freeze–thaw cycle tests, which yielded variations in the mass, longitudinal wave velocity, and uniaxial compressive strength of rock samples subjected to freeze–thaw cycles. A rock damage model of the Houwai tunnel was established based on the existing rock damage model combined with the parameters before and after freezing and thawing of the Houwai tunnel limestone after seasonal freeze in the Jilin Province. The relationship between the total damage variable and the strain of limestone was obtained under the freeze–thaw cycle, and the damage deterioration mechanism of the limestone was examined after the freeze–thaw cycle. The mechanical characterization of the deterioration damage of the surrounding rock predicted using the deterioration damage model is of great significance for the determination of freeze–thaw damage and prevention and control of freezing damage in tunnel engineering in seasonally frozen areas. The study on the damage characteristics of limestone in the study area can lay a foundation for the freeze–thaw failure mechanism of tunnels in seasonal frozen area.

Asymmetric loads often occur at tunnel entrances or exits during shallow buried tunnel construction. Due to asymmetrical loads, structures are subjected to complex forces, and the design process of these tunnels is different from that of regular tunnels. Therefore, it is necessary to discriminate asymmetric tunnels quantitatively. Currently, determining whether a tunnel is subject to asymmetrical loads is based on the empirical maximum burial depths and surface slope angles listed in the Chinese Code for the Design of Railway Tunnel. However, in real geological conditions, the maximum burial depths and surface slope angles of the tunnel are not necessarily completely consistent with the Chinese Code; in other words, the continuous discrimination of asymmetric tunnels is not realized. In this paper, the surrounding rock stress ratio of the arch shoulder was treated as index to define asymmetrical loading tunnels and a new approach to achieve the continuous discrimination of asymmetric tunnels was proposed. Compared to the empirical values in the Chinese Code, the results show that the asymmetric stress ratio tends to decrease with the decrease in the surrounding rock grade. Regardless of whether the grade is III, IVRock, IVSoilor V, the asymmetric stress ratios of the surrounding rock tend to converge and stabilize with increasing burial depths and decreasing dip angles, and the sensitivity of the slope angle to the stress ratio is bounded by 20°. The asymmetric stress ratio of the surrounding rock arch shoulder can be used as an index to ascertain whether a tunnel is subjected to asymmetric loads. When the surrounding rock grades are III and IVRock, the values calculated by the proposed approach are lower than the empirical values, and the approach is more conservative than the empirical method. When the surrounding rock grades are IVSoiland V, the values calculated by the proposed approach are in good agreement with the empirical values. The approach of continuous discrimination of asymmetric tunnels proposed in this paper is reliable. The findings of this study can help for better understanding of the identification of asymmetrical loading tunnels, the design of support structures and the safety risk assessment of tunnels.

CO2dissolution in water at different temperature and pressure conditions is of essential interest for various environmental, geochemical, and thermodynamic related problems. The topic is of special interest in studies of CO2geological sequestration in brine-bearing aquifers. In this Article, four powerful machine learning (ML) techniques—Radial Basis Function Neural Network (RBFNN), Multilayer Perceptron (MLP), Least-Squares Support Vector Machine (LSSVM), and Gene Expression Programming (GEP)—are implemented to develop economical, rapid, and reliable models to predict the solubility of CO2in water. To expand the prediction capability of the ML approaches, their control parameters are optimized by various techniques. To this end, four back-propagation algorithms are applied in the MLP learning phase, while Particle Swarm Optimization (PSO), Differential Evolution (DE), Genetic Algorithm (GA), and Firefly Algorithm (FFA) are used to optimize the RBFNN and LSSVM control parameters. A wide-ranged database including temperature and pressure as inputs and CO2solubility in pure water as output is utilized to develop the models, which are then compared with each other and also against existing models. The results demonstrate that the prediction performance of the proposed models is quite satisfactory. In addition, the comparison results reveal that the LSSVM-FFA is the best paradigm to estimate the solubility of CO2in pure water, as it outperforms the other proposed ML techniques as well as prior models. The overall RMSE and R2values for LSSVM-FFA are 0.3261 and 0.9930, respectively.

The natural gas compressibility factor indicates the compression and expansion characteristics of natural gas under different conditions. In this study, a simple second-order polynomial method based on the group method of data handling (GMDH) is presented to determine this critical parameter for different natural gases at different conditions, using corresponding state principles. The accuracy of the proposed method is evaluated through graphical and statistical analyses. The method shows promising results considering the accurate estimation of natural gas compressibility. The evaluation reports 2.88% of average absolute relative error, a regression coefficient of 0.92, and a root means square error of 0.03. Furthermore, the equations of state (EOSs) and correlations are used for comparative analysis of the performance. The precision of the results demonstrates the model’s superiority over all other correlations and EOSs. The proposed model can be used in simulators to estimate natural gas compressibility accurately with a simple mathematical equation. This model outperforms all previously published correlations and EOSs in terms of accuracy and simplicity.

Facilitated transport membranes for post-combustion carbon capture are one of the technologies to achieve efficient and large-scale capture. The central principle is to utilize the affinity of CO2for the carrier to achieve efficient separation and to break the Robson upper bound. This paper reviews the progress of facilitated transport membranes research regarding polymer materials, principles, and problems faced at this stage. Firstly, we briefly introduce the transport mechanism of the facilitated transport membranes. Then the research progress of several major polymers used for facilitated transport membranes for CO2/N2separation was presented in the past five years. Additionally, we analyze the primary challenges of facilitated transport membranes, including the influence of water, the effect of temperature, the saturation effect of the carrier, and the process configuration. Finally, we also delve into the challenges and competitiveness of facilitated transport membranes.

Rivers and their hyporheic zones play an important role in nutrient cycling. The fate of dissolved inorganic nitrogen is governed by reactions that occur in the water column and streambed sediments. Sediments are heterogeneous both in term of physical (e.g., hydraulic conductivity) and chemical (e.g., organic carbon content) properties, which influence water residence times and biogeochemical reactions. Yet few modeling studies have explored the effects of both physical and chemical heterogeneity on nutrient transport in the hyporheic zone. In this study, we simulated hyporheic exchange in physically and chemically heterogeneous sediments with binary distributions of sand and silt in a low‐gradient meandering river. We analyzed the impact of different silt/sand patterns on dissolved organic carbon, oxygen, nitrate, and ammonium. Our results show that streambeds with a higher volume proportion of silt exhibit lower hyporheic exchange rates but more efficient nitrate removal along flow paths compared to predominantly sandy streambeds. The implication is that hyporheic zones with a mixture of inorganic sands and organic silts have a high capacity to remove nitrate, despite their moderate permeabilities. With the advent of intensive cultivation and livestock farms, nutrient contamination has increased in streams. Microbial communities in streambeds reduce nitrate, effectively removing it from streams. However, quantifying this removal is difficult because it is influenced by many spatially variable factors, among which sediment type plays an important role. Streambeds can contain both sandy and silty sediments distributed in complex patterns. The present study employs computer simulations to analyze how nitrate is removed in a mixture of different streambed sediments. Role of physical and chemical heterogeneity of streambed sediments on denitrification is numerically investigatedSilt sediments promote denitrification since they experience longer residence times and are source of dissolved organic carbonStreambeds with mixture of sand and silt have high capacity to remove nitrate despite moderate permeabilities

We investigate the coupled dynamics of fluid mixing and viscously unstable flow under both miscible (single‐phase) and partially miscible (two‐phase) conditions, and in both homogeneous and heterogeneous porous media. Higher‐order finite element methods and fine grids are used to resolve the small‐scale onset of fingering and tip splitting. An equation of state determines the thermodynamic phase behavior and Fickian diffusion. We compute global quantitative measures of the spreading and mixing of a diluting slug to elucidate key differences between miscible and partially miscible systems. Hydrodynamicinstabilities are the main driver for mixing in miscible flow. In partially miscible flow, however, we find that relative permeabilities spread the two‐phase zone. Within this mixing zone dissolution and evaporation drive mixing thermodynamicallywhile reducing mobility contrasts and thus fingering instabilities. The different mixing dynamics in systems involving multiple phases with mutual solubilities have important implications in hydrogeology and energy applications, such as geological carbon sequestration and gas transport in hydrocarbon reservoirs. When fluids propagate through subsurface porous media, flow instabilities can be either detrimental or advantageous depending on the application. The interplay between viscous instabilities and fluid mixing has been studied extensively for single‐phase two‐component mixtures, but not for applications that involve two phases (e.g., oil and gas) that exchange multiple species. In this first study of mixing in partially miscible two‐phase flow, we find that the dynamics are profoundly different. In single‐phase flow, viscous fingering increases the interface between two fluids (of the same phase) and diffusion across that interface drives the mixing. In two‐phase flow, mass exchange between the phases creates a two‐phase mixing zone. In this zone, two phases coexist that have different velocities, which lead to an expansion of the mixing zone. Diffusion occurs both within each phase and between the phases, and the interphase mass exchange significantly enhances mixing. Because this thermodynamic mixing is extremely efficient, viscosity contrasts are quickly homogenized, and hydrodynamic instabilities (fingering) are less pronounced. In fluid mechanical terms, dissipation drives mixing in single phase, while production may dominate in two‐phase flow. These findings have broad implications for the prediction of instabilities and mixing behavior in subsurface applications that involve multiple partially miscible fluid phases. The mixing dynamics of two partially miscible fluids are revealed for the first timeMixing is governed by interphase multispecies exchange interacting with flow fingering and channelingExpansion of a two‐phase mixing zone reduces global variance budget in an interplay with dissipation

Reliable characterization of subsurface structures is essential for earth sciences and related applications. Data assimilation‐based identification frameworks can reasonably estimate subsurface structures using available lithological (e.g., borehole core, well log) and dynamic (e.g., hydraulic head, solute concentration) observations. However, a reasonable selection of the observation type and frequency is essential for accurate structure identification. To achieve this, we extended a recently developed stage‐wise stochastic deep learning inversion framework by coupling it with non‐isothermal flow and transport simulations. With the extended framework, the worth of three common observations (hydraulic head, concentration, and temperature) are compared under different observation noise and frequency. The framework combines the emerging deep‐learning (DL)‐based framework with the traditional stochastic approaches. This combination makes it possible to simultaneously compare the ability of these two methods to assimilate observation data. Our results show that including at least one type of dynamic observation strongly improves subsurface structure identifiability and reduces the uncertainty. However, the DL‐based framework is able to identify subsurface structures more accurately than stochastic identification methods under the same scenarios. Assimilation of certain types of dynamic observations could reduce the prediction error for related dynamic responses, but not necessarily for other uncorrelated dynamic responses. Observation data worth is affected by the observation noise and frequency. High observation noise increases the uncertainty of the prediction and reduces the estimation accuracy. However, the higher observation frequency can significantly improve the temporal dynamic information of observations. This information can compensate for negative impacts of high observation noise. Obtaining realistic subsurface structures is crucial for earth sciences and other related applications, including groundwater management, geological carbon storage, radioactive waste disposal, and geothermal applications. This study extended a recently developed deep‐learning‐based identification framework to consider hydraulic head, concentration, and temperature observations. In addition, this framework combines the traditional method with recently popular deep learning framework. Through this framework, we study the effects of different observation types, noise, and frequency on subsurface structure identification, so as to guide the reasonable use of observations in practice to improve the accuracy of subsurface structure identification. Our results show that by applying some of the latest deep learning technologies, the framework can obtain more accurate subsurface structures. Through this original research, we recommend preferentially using observations that can directly inform the process of interest. When collecting more data, it is more important to focus on collecting higher quality data. Effects of the observation type, noise, and frequency are studied for subsurface structure identificationObservation data worth is influenced by observation noise and observation frequencyDynamic observations that can directly inform processes of interest should be preferentially assimilated Effects of the observation type, noise, and frequency are studied for subsurface structure identification Observation data worth is influenced by observation noise and observation frequency Dynamic observations that can directly inform processes of interest should be preferentially assimilated

Groundwater monitoring networks are direct sources of information for revealing subsurface system dynamic processes. However, designing such networks is difficult due to uncertainties in the spatial heterogeneity of aquifer parameters such as permeability (k). This study combines deep learning and information theory with an optimization framework to address network design problems in heterogeneous aquifer systems. The framework first employs a generative adversarial network to parameterize heterogeneous kdistribution using a low‐dimensional latent representation. Then, surrogate models are developed based on the deep neural networks to perform uncertainty quantification of pressure heads and solute concentrations at locations of pre‐designed candidate monitoring stations. The monitoring stations are then ranked using the greedy search algorithm based on the maximum information minimum redundancy (MIMR) criterion. In order to depict the importance of each candidate monitoring location, the hotspot maps of the selection probability (Ps) are derived from MIMR repetition results. Comprehensive monitoring networks derived from the hotspot maps are then conducted as the final monitoring stations to improve monitoring information compared to MIMR results. Additionally, nine entropy quantization strategies are compared to evaluate their effects on monitoring network optimization results. Results indicate that caution should be taken when selecting entropy quantization strategies to achieve the accuracy required for model calibration and to improve the efficiency of monitoring optimization. Considering high‐dimensional uncertainties associated with aquifer parameters, the developed framework can provide important insights for monitoring network designs in various earth observational projects. Groundwater is widely distributed throughout the crust and lithosphere playing a critical role in geological evolution processes and environmental problems. This study presents an integrated framework for designing groundwater monitoring networks in heterogeneous aquifer systems. The proposed model takes advantage of the strength of deep‐learning algorithms and deals with high‐dimensional and highly non‐linear problems of entropy‐based methods for designing the monitoring networks. The framework can efficiently provide monitor locations for complex aquifers. The monitoring data collected by the designed network can significantly reduce uncertainties associated with high‐dimensional parameters and contain the least amount of redundant information. Further, the proposed method is significant in practice as credible groundwater forecasting requires model calibrations that use monitoring data. Effective designed monitoring networks are critical for reducing model uncertainties and properly reflecting subsurface dynamic processesDeep learning and maximum information minimum redundancy criterion were combined to design optimal monitoring networks for heterogeneous aquifersThe designed monitoring networks were verified to be influenced by different entropy quantification strategies Effective designed monitoring networks are critical for reducing model uncertainties and properly reflecting subsurface dynamic processes Deep learning and maximum information minimum redundancy criterion were combined to design optimal monitoring networks for heterogeneous aquifers The designed monitoring networks were verified to be influenced by different entropy quantification strategies

Heterogeneity in hydraulic conductivity (K) and channel morphology both control surface water‐groundwater exchange (hyporheic exchange), which influences stream ecosystem processes and biogeochemical cycles. Here we show that heterogeneity in Kis the dominant control on exchange rates, residence times, and patterns in hyporheic zones with abrupt lithologic contrasts. We simulated hyporheic exchange in a representative low‐gradient stream with 300 different bimodal Kfields composed of sand and silt. Simulations span five sets of sand‐silt ratios and two sets of low and high Kcontrasts (1 and 3 orders of magnitude). Heterogeneity increases interfacial flux by an order of magnitude relative to homogeneous cases, drastically changes the shape of residence time distributions, and decreases median residence times. The positioning of highly permeable sand bodies controls patterns of interfacial flux and flow paths. These results are remarkably different from previous studies of smooth, continuous Kfields that indicate only moderate effects on hyporheic exchange. Our results also show that hyporheic residence times are least predictable when sand body connectivity is low. As sand body connectivity increases, the expected residence time distribution (ensemble average for a given sand‐silt ratio) remains approximately constant, but the uncertainty around the expectation decreases. Including strong heterogeneity in hyporheic models is imperative for understanding hyporheic fluxes and solute transport. In streams with strongly heterogeneous sediments, characterizing lithologic structure is more critical for predicting hyporheic exchange metrics than characterizing channel morphology. Abrupt contrasts in permeability are a dominant control on hyporheic exchangePermeability heterogeneity radically changes hyporheic residence time behaviorAs sand connectivity increases, the uncertainty in the residence times decreases

This technical note presents a useful methodology for studying how the variance of hydraulic and/or reactive attributes of an aquifer are linked to the multiscaled and hierarchical sedimentary architecture of the aquifer. A new recursive equation is derived which quantitatively describes how the variance is related to sedimentary facies defined at all scales across an entire stratal hierarchy. As compared to prior published equations that emphasize differences in means among facies populations withina hierarchical level, it emphasizes differences acrosslevels. Because of the hierarchical relationships among the terms of the equation, we find it to be useful for conducting a holistic analysis of the relative contributions to the variance arising from all facies types defined across all scales. The methodology is demonstrated using appropriate field data, and is shown to be useful in defining parsimonious classification systems. Aquifer permeability and sorption are linked to sedimentary architectureA holistic, multiscale, and hierarchical analysis of variance is presented

A large proportion (∼80%) of the Indian subcontinent's precipitation comes from the Indian summer monsoon (ISM), which influences one‐fifth of the world's population. A long‐term reliable proxy for ISM is fundamental to understanding previous global climate change. We establish a mass‐wasting‐inferred proxy to examine the paleo‐hydrogeology (river undercutting history) of the Southeast Tibetan Plateau and reconstruct the dramatic variability of ISM intensity (precipitation) in the past 130,000 years. Our data suggest that mass‐wasting events, which provides us sufficient samples for paleoclimate research, are prone to dramatic climate changes, especially extreme climate environments. The Southeast Tibetan Plateau was subjected to at least four distinct ISM intensity phases in the past 130,000 years. We conclude that ISM intensity has a cyclicity featured by the Earth's orbit with obliquity, and that ISM intensity transition lags the global ice volume (sea level) change by 8–15 kyrs. The role of Indian summer monsoon (ISM) in the global climatic conditions has attracted considerable research attention. We reported a new long‐term proxy collected in Tibetan Plateau mass‐wasting deposits for ISM reconstruction, which is significant for coupling of tectonic evolution of Asia and global climate change in the context of global warming. This study has revealed the mutuality of paleo‐mass‐wasting events and climate changes, which confirms the relationship between surge of mass‐wasting events and global climate change (warming and increasing precipitation). ISM variability derived in our work for the last 130,000 years deduces new views of cyclicity and lag characteristics of ISM intensity. The correlation between the occurrence of mass‐wasting deposits and the dramatic global climate changes is revealedThe variability of Indian summer monsoon (ISM) is reconstructed based on our proposed mass‐wasting‐inferred proxyNew views of cyclicity and lag characteristics of ISM intensity are deduced based on the variability of ISM intensity The correlation between the occurrence of mass‐wasting deposits and the dramatic global climate changes is revealed The variability of Indian summer monsoon (ISM) is reconstructed based on our proposed mass‐wasting‐inferred proxy New views of cyclicity and lag characteristics of ISM intensity are deduced based on the variability of ISM intensity

The stochastic models and deep‐learning models are the two most commonly used methods for subsurface sedimentary structures identification. The results from the stochastic models typically involve uncertainty due to their nature. For the deep‐learning models, sufficient structure samples are necessary for training, but they are practically difficult to obtain. This study develops an inversion framework to combine the strength of these two models to overcome the limitations. The stochastic model is first adopted to generate the structure samples required by the deep‐learning models by integrating available observations. Then the trained deep‐learning model is utilized to reduce the uncertainty of the structures generated by the stochastic models. This integrated framework can successfully estimate the structures using available observations. Importantly, no additional structure training samples are required in the identification process. To summarize, the combination of the stochastic and the deep‐learning models shows great advantages in identifying subsurface sedimentary structures. This study develops an integrated inversion framework that uses sparse data obtained from different geophysical measurements to identify subsurface sedimentary structures. By combining the advantages of physically‐based geological models and advanced deep‐learning techniques, the developed framework can identify complex subsurface sedimentary structures in a more practical way than is currently possible. The advantage of this method is that the identification process relies only on sparse measurement data and does not need additional structure samples for deep‐learning model training. It has significant practical value because most deep‐learning‐based methods require a large number of training samples, which are practically difficult to obtain. Therefore, the developed framework has great advantages in practical subsurface sedimentary structures identification applications. A stage‐wise inversion framework is developed by combining the transition probability approach and the deep‐learning methodsThe developed framework can identify subsurface sedimentary structures using sparse geological data and flow and transport responsesThe estimated structures can honor the available flow and transport responses and capture the preferential flow pathways A stage‐wise inversion framework is developed by combining the transition probability approach and the deep‐learning methods The developed framework can identify subsurface sedimentary structures using sparse geological data and flow and transport responses The estimated structures can honor the available flow and transport responses and capture the preferential flow pathways

We investigate mechanisms that enhance lateral methane (CH4) plume migration in shallow aquifers that exhibit complex and multiscale sedimentary architecture. We show how heterogeneity in capillary pressure characteristics related to the sedimentary architecture causes gaseous CH4to spread over larger areas by retarding, deviating, or blocking upward buoyancy‐driven CH4migration. Simplifying or ignoring capillary pressure heterogeneity thus leads to overestimation of leaked CH4to the atmosphere, and underestimation of mobile gaseous CH4in aquifer. We show, both qualitatively and quantitatively, that meter‐scale sedimentary stratification contributes more to CH4plume migration than the millimeter‐ and centimeter‐scale strata comprising them. Results indicate that the extent of gaseous CH4leakage, and its associated impacts on groundwater quality and global warming, cannot be accurately assessed unless the sedimentary architecture and resulting heterogeneity in capillary pressure are represented. Increasing levels of methane (CH4) in shallow aquifers from both natural and anthropogenic sources have become a serious concern worldwide. Leaked CH4not only contaminates drinking water resources, resulting in long‐term adverse health effects, but may also enter the atmosphere, accelerating global climate change. To assess the potential risks associated with CH4leakage and design effective mitigation strategies, it is crucial to understand how geological features at different spatial scales affect the behavior of CH4in groundwater. Field experiments have shown that horizontal migration of CH4is beyond that expected due to groundwater flow alone, but the mechanism driving these phenomena is unknown. In this work, we show that this unexpected behavior may be caused by geological heterogeneity and the resulting heterogeneity in capillary pressure, which is the pressure difference at the interface between gaseous CH4and groundwater. Capillary heterogeneity in relatively homogenous and shallow aquifers places leading‐order controls on lateral CH4plume spreadingNeglecting capillary heterogeneity leads to overestimation of atmospheric CH4and underestimation of mobile gaseous CH4in aquiferMeter‐scale sedimentary stratification contributes more to CH4plume migration than millimeter‐ and centimeter‐scale features Capillary heterogeneity in relatively homogenous and shallow aquifers places leading‐order controls on lateral CH4plume spreading Neglecting capillary heterogeneity leads to overestimation of atmospheric CH4and underestimation of mobile gaseous CH4in aquifer Meter‐scale sedimentary stratification contributes more to CH4plume migration than millimeter‐ and centimeter‐scale features

Nitrous oxide (N2O) is a potent ozone‐depleting greenhouse gas produced by incomplete denitrification. Recent works on riverine N2O emissions focus mainly on contributions from in‐channel, benthic, and fluvial hyporheic environments under assumptions of steady‐state conditions and homogeneous sediment hydraulic conductivity (K). However, riparian floodplains are also a potentially important N2O source characterized by complex sediment heterogeneity and dynamic surface and groundwater interactions. We use numerical flow and reactive transport models to investigate the influence of complex sedimentary architecture and high‐flow events (e.g., storms) on N2O production. We interpret the correlation between flow and solute fields with the flow topological Okubo‐Weiss metric (OW) and the scalar dissipation rate weighted by soil organic matter (OM) fraction and soil saturation. We model a heterogeneous riparian floodplain based on field observations from the Theis Environmental Monitoring and Modeling Site, Ohio, USA. N2O production is greatest within intermediate‐Ksediments (e.g., sands) where denitrification rates are highest, and emissions increase by more than an order of magnitude during storms. Sensitivity analysis reveals that the denitrification rate is most influential for N2O flux, accounting for nearly 46% of the variance in production rates. Denitrification rates adapt to spatial changes in the flow topology (measured by OW) related to sediment heterogeneity and are strongly influenced by subsurface mixing dynamics. Mixing is greatest in shear strain‐dominated regions, while vorticity promotes OM dissolution and prolongs residence times. Accurate lithologic representation is imperative for characterizing subsurface N2O production dynamics, especially given growing concern regarding climate change driven hydrologic changes within watersheds worldwide. Subsurface N2O production increases in response to storms, but sediment heterogeneity controls the location and magnitude of emissionsAccurate representation of heterogeneity is important for characterizing subsurface nitrogen transformations and N2O production dynamicsHeterogeneity introduces shear into the groundwater flow topology, enhancing mixing and increasing the spatial extent of denitrification Subsurface N2O production increases in response to storms, but sediment heterogeneity controls the location and magnitude of emissions Accurate representation of heterogeneity is important for characterizing subsurface nitrogen transformations and N2O production dynamics Heterogeneity introduces shear into the groundwater flow topology, enhancing mixing and increasing the spatial extent of denitrification

Tides in coastal rivers drive river‐groundwater (hyporheic) exchange and provide opportunities for nitrate removal that may improve coastal water quality. Silt and sand layers in coastal floodplain sediments can alter the flow and transformation of nitrate. Our goal was to understand how sediment heterogeneity influences nitrogen dynamics near tidal rivers. Numerical simulations show that oxic, variably saturated sand layers and anoxic, organic‐rich silt layers are sites of nitrification and denitrification, respectively. The exchange of river water and nitrate through heterogeneous sediments increases with sand fraction, as sand lenses become longer and more connected. The amount of nitrate removed from river water also increases but represents a smaller portion of total nitrate exchange through the hyporheic zone, causing removal efficiency to decline. Our results suggest that accurate characterization of aquifer heterogeneity leads to an improved understanding of sites of nutrient transformation within floodplain sediments. Excess nitrate can degrade coastal water quality, but microbial reactions reduce its concentration within the bed and banks of tidal rivers, where surface water and groundwater mix. Spatial arrangements of different sediments (sand and mud) affect the mixing of river water and groundwater and thus affect nitrate removal. Here, we use computer models to simulate nitrate transformation along a tidal river with different amounts of coarse and fine sediments. Coarse sediments promote groundwater flow and nitrate production, while fine sediments promote nitrate removal. The amount of nitrate removed from river water is greater in aquifers with coarser sediments, but the removal efficiency decreases. Removal also varies with the spatial distribution of sand and mud in sediments but to a lesser degree. Computer models of nitrate transport should consider the distribution of different sediment types. Tidal aquifers are a net sink of nitrate; removal magnitude varies with permeability structureSediment organic matter is a local source of dissolved organic carbon driving small‐scale patterns of nitrate transformationNitrification dominates in sand layers, where oxygen and nutrients are mobile; denitrification dominates in organic‐rich, silt layers

In coastal rivers, tides facilitate surface water‐groundwater exchange and strongly coupled nitrification‐denitrification near the fluctuating water table. We used numerical fluid flow and reactive transport models to explore hydrogeologic and biogeochemical controls on nitrogen transport along an idealized tidal freshwater zone based on field observations from White Clay Creek, Delaware, USA. The capacity of the riparian aquifer to remove nitrate depends largely on nitrate transport rates, which initially increase with increasing tidal range but then decline as sediments become muddier and permeability decreases. Over the entire model reach, local nitrification provides a similar amount of nitrate as surface and groundwater contributions combined. More than half (~66%) of nitrate removed via denitrification is produced in situ, while the vast majority of remaining nitrate removed comes from groundwater sources. In contrast, average nitrate removal from surface water due to tidal pumping amounts to only ~1% of the average daily in‐channel riverine nitrate load or 1.77 kg of nitrate along the reach each day. As a result, tidal bank storage zones may not be major sinks for nitrate in coastal rivers but can act as effective sinks for groundwater nitrate. By extension, tidal bank storage zones provide a critical ecosystem service, reducing contributions of groundwater nitrate, which is often derived from septic tanks and fertilizers, to coastal rivers. Nitrate is one of the most common pollutants and can degrade coastal water quality. In the beds and banks of coastal rivers, tides increase the mixing of river and groundwater, which creates opportunities for nitrate production by nitrification and removal by denitrification. It is unknown which process prevails along tidal rivers and how these processes vary in strength as sediments become muddier and tides become larger near the coast. In this study, we used computer models to estimate rates of nitrate production and removal and where they are greatest along tidal rivers. We found that nitrate is ultimately removed from the aquifer, but that the majority of nitrate removed is from nitrification near the water table and groundwater sources. Although aquifers along tidal rivers are only marginally effective at removing nitrate from river water, their ability to remove groundwater nitrate is vital for maintaining coastal water quality. Surface water‐groundwater exchange decreases moving downstream, as sediment permeability decreases, despite increasing tidal rangeThe majority of denitrification removes nitrate produced in the riparian zone, with groundwater supplying the remaining nitrate removedTidal bank storage zones along rivers can be an effective sink for groundwater nitrate, mitigating nitrate export to coasts