https://arxiv.org/api/nh1y5o34BNOkyUINGZoG5uiXoyI 2026-09-10T23:33:39Z 8935 90 15 http://arxiv.org/abs/2605.07060v3 2026-08-13T05:34:20Z

Physics-informed neural networks (PINNs) provide a mesh-free framework for solving PDE-constrained inverse problems, but their extension to Bayesian inversion still faces a fundamental difficulty: prior distributions are typically defined in the weight space of neural networks, whereas physically meaningful prior assumptions are more naturally expressed in function space. In this study, we introduce a unified framework, termed functional-prior-based approaches to Bayesian PDE-constrained inversion using physics-informed neural networks (fpBPINN), to incorporate functional priors into Bayesian PINN-based inversion. We consider two complementary approaches. The first is a functional-prior-informed Bayesian PINN (FPI-BPINN), in which a neural network weight prior is learned to be consistent with a prescribed functional prior, and Bayesian inference is subsequently performed in weight space. The second is function-space particle-based variational inference for PINNs (fParVI-PINN), which performs Bayesian estimation using ParVI directly in function space. We also show that random Fourier features (RFF) play an important role in representing Gaussian functional priors with neural networks and in improving posterior approximation. We applied the proposed approaches to one-dimensional seismic traveltime tomography and two-dimensional Darcy-flow permeability inversion. These numerical experiments showed that both approaches accurately estimated posterior distributions, highlighting the significance of introducing physically interpretable functional priors into Bayesian PINN-based inverse problems. We also identified the contrasting advantages of FPI-BPINN and fParVI-PINN, namely flexibility and accuracy, respectively.

2026-05-08T00:13:13Z Ryoichiro Agata Tomohisa Okazaki http://arxiv.org/abs/2608.10287v1 2026-08-10T22:44:52Z

Distributed Acoustic Sensing (DAS) measures dynamic strain along a fiber-optic cable, offering a robust, densely-sampled alternative to traditional seismic sensors. To ensure good ground-to-cable coupling, cables are typically buried in a shallow trench. Unburied surface deployments are attractive for rapid-response terrestrial applications as well as extraterrestrial missions, such as on the Moon, where burial is impractical. However, unburied DAS often suffers from severely degraded signal quality, due to poor strain transfer from ground to cable. The physical mechanism responsible remains unknown. Here, we identify bending stress relief as a mechanism that can explain this loss: suspended cable segments accommodate ground strain by bending rather than by stretching or compressing, reducing the measurable axial strain that reaches the fiber. We develop the first analytical and numerical model of unburied DAS coupling, representing the draped cable as a series of suspended segments between discrete ground contact points, to explain and quantify the bending stress relief mechanism. Our analysis reveals a dimensionless parameter, Theta, set by the ratio of the cable's initial gravity-induced sag to its radius, which governs the strain transfer efficiency. Once a segment's sag exceeds a quarter of the cable's radius, ground displacement starts to be absorbed by bending rather than being transferred as measurable axial strain. This framework predicts how mechanical properties, cable dimensions, pretension, and gravity affect strain transfer efficiency and provides quantitative guidelines for optimizing cable design and deployment strategies on both Earth and the Moon.

2026-08-10T22:44:52Z 18 pages, 3 figures Simone Probst Marc Serra Garcia Johan O. A. Robertsson Carly M. Donahue http://arxiv.org/abs/2608.10233v1 2026-08-10T21:10:51Z

Groundwater variability in Ghana remains poorly characterized due to limited long-term in-situ observations. This study investigates groundwater storage anomalies using GRACE-derived data from 2004-2024 combined with statistical analysis and unsupervised machine learning. Groundwater anomalies were standardized using Z-scores, while an ensemble-based Isolation Forest framework was applied for anomaly detection. The results revealed substantial temporal variability, with persistent groundwater deficits during 2004-2009 followed by increasing positive anomalies after 2018. A total of 12 anomalous months were identified, comprising 5 deficit and 7 surplus events, with the strongest anomalies associated with groundwater deficits. Spatial analysis showed more frequent deficit anomalies in northern Ghana and stronger surplus occurrence in southern regions. Comparison with statistical thresholds further indicated that the machine learning framework captured additional subtle deviations beyond conventional threshold-based methods. Overall, the integration of GRACE observations with unsupervised anomaly detection provides a practical framework for groundwater monitoring in data-scarce environments.

2026-08-10T21:10:51Z 8 pages, 7 figures, accepted at ICECCME 2026 George Yamoah Afrifa Theophilus Ansah-Narh Marcellin Atemkeng http://arxiv.org/abs/2608.10192v1 2026-08-10T20:10:25Z

Gravity driven sediment flows are responsible for a major portion of sediment redistribution within oceans, reservoirs, and lakes, with important implications in coastal erosion, siltation, carbon burial, and contaminant migration in aquatic systems. Despite the ubiquity of this phenomenon, current mechanistic understanding of sediment gravity flows (SGFs) remains limited. This knowledge gap is particularly acute in the case of cohesive, fine-grained sediments (i.e., muds) due to the complex properties of the clay matrix, including low permeability, viscoplastic rheology, and flocculation. In this work, we develop a computational fluid dynamics model that accurately predicts key features of cohesive, clay-rich SGFs based on independent measurements of the relation between sediment solid fraction and rheological yield stress. In particular, the model captures the four primary flow regimes (low density turbidity currents, high density turbidity currents, mudflows, and mudslides) observed in lock-exchange experiments with slurries containing smectite or kaolinite clay. The model is validated through comparison with previous experimental observations of sediment flow morphology, speed, and runout distance. Overall, we demonstrate the ability to predict the influence of intrinsic (particle size, grain density, and rheology) and extrinsic sediment properties (sediment topography and solid fraction) in the development of self-sustaining cohesive SGFs.

2026-08-10T20:10:25Z Mitchell D. Jans Cyprien Soulaine Judy Q. Yang Ian C. Bourg http://arxiv.org/abs/2604.23199v2 2026-08-09T15:44:52Z

We investigate how models of fluid properties and boundary conditions influence predictions of convective mixing in confined porous media, with relevance to subsurface carbon dioxide storage. Using high-resolution simulations at high Rayleigh-Darcy numbers (O($10^4$)), we analyze miscible fluids with linear, nonlinear, and non-monotonic density-concentration relationships under fixed- and free-interface in 2D and 3D. We show that, across all cases, mixing is quantified by the mean scalar dissipation, providing a unifying framework for convective-diffusive interactions. The density-concentration relationship affects mixing via the effective density contrast driving convection and the position of the maximum density. Free interfaces enhance early-time mixing through deformation, while long-term behavior depends on fluid properties and dimensionality. We demonstrate that simplified modeling assumptions (e.g., monotonic density laws or 2D flow) can lead to deviations in predicted mixing rates of up to O(10-100)%. These results offer guidance for model selection and improving predictions of convective mixing in geophysical systems.

2026-04-25T08:12:05Z Marco De Paoli Sergio Pirozzoli http://arxiv.org/abs/2609.05456v1 2026-08-07T19:19:01Z

In this study, we investigated the slow dynamic behavior of concrete with varying compressive strengths ($f'_c = 32\sim56$ MPa) and alkali-silica reaction (ASR) damage. Concrete prisms were conditioned by compressive loading and recovery was monitored by coda wave interferometry (CWI). A self-referencing temperature correction technique was used to minimize the effect of ambient temperature changes, demonstrating the importance of careful temperature control and correction. For intact specimens, both the recovery rate, $m_v$, and velocity drop magnitude, $|c|$, generally increased with compressive strength. Recovery times were calculated from the fit parameters, and higher strength specimens were found to recover faster (9.9 h for $f'_c = 32$ MPa, 6.6 h for $f'_c = 56$ MPa). When the same stress was applied, ASR-damaged specimens had increased softening and faster recovery rates, but longer recovery times (up to 458 h). These findings demonstrate the potential of slow dynamics in concrete characterization, through its ability to evaluate the strength of intact samples and its sensitivity to microcracking in damaged samples.

2026-08-07T19:19:01Z This manuscript is under review at the Journal of Nondestructive Evaluation Clayton Malone Jinying Zhu http://arxiv.org/abs/2509.14162v3 2026-08-07T14:27:52Z

Flood risk is correlated in space and time, challenging insurance systems that rely on diversification across assets. Financial instruments governing flood coverage are typically structured as 1 to 5-year contracts, exposing portfolios to climate-driven risk at interannual-to-decadal scales. Yet existing tools address climate risk either through seasonal forecasts extending only months or multidecadal projections misaligned with fiscal horizons, leaving a critical gap in actionable flood risk simulation. We introduce a multisite flood simulation framework combining attention-based analog retrieval with stochastic generation of multivariate flood frequency, intensity, and duration sequences. Applied to over 100 sites in the Mississippi River Basin, the model produces spatiotemporally coherent flood portfolios conditioned on interannual climate variability. Explainable AI attribution paired with wavelet analysis links simulated clustering to large-scale climate drivers, yielding physically interpretable flood clusters for portfolio-scale loss simulation. The framework provides plausible, out-of-sample flood risk catalogs for interannual-to-decadal insurance risk assessment and financial planning.

2025-09-17T16:46:56Z Nayak, A., Gentine, P., and Lall, U. (2026). "Attention-Based Stochastic Simulation of Climate-Informed Spatiotemporal Flood Risk for Multisite Insurance Portfolios." Geophysical Research Letters, 53 (15) Adam Nayak Pierre Gentine Upmanu Lall 10.1029/2026GL123701 http://arxiv.org/abs/2608.06543v1 2026-08-06T19:48:35Z

The IceCube Upgrade is a densely instrumented central region of the IceCube Neutrino Observatory, deployed during the 2025-26 polar season. It will reduce the detector's energy threshold and improve overall reconstruction capabilities for multi-GeV atmospheric neutrinos, which in turn enhance their sensitivity to Earth matter effects as they traverse through the deep Earth. In this study, we describe the potential of the IceCube Upgrade to observe Earth matter effects on atmospheric neutrinos and estimate the detector's sensitivity to probe key features of the Preliminary Reference Earth Model by utilizing these observations. We highlight the IceCube Upgrade's capability to estimate the mass of the Earth and verify the non-homogeneous distribution of matter density within the Earth. We also estimate the IceCube Upgrade sensitivity to measure the correlated densities of the Earth layers while incorporating constraints from the mass and moment of inertia of the Earth. Neutrino-based results would be independent and complementary to the seismic and gravitational measurements.

2026-08-06T19:48:35Z 20 pages, 13 figures, 2 tables, and 1 appendix The IceCube Collaboration R. Abbasi M. Ackermann J. Adams S. K. Agarwalla J. A. Aguilar M. Ahlers J. M. Alameddine S. Ali N. M. Amin K. Andeen C. Arg{ü}elles S. Athanasiadou S. N. Axani R. Babu X. Bai A. Balagopal V. S. W. Barwick V. Basu R. Bay J. J. Beatty J. Becker Tjus P. Behrens J. Beise C. Bellenghi S. Benkel S. BenZvi D. Berley E. Bernardini D. Z. Besson E. Blaufuss L. Bloom S. Blot F. Bontempo J. Y. Book Motzkin C. Boscolo Meneguolo S. B{ö}ser O. Botner J. B{ö}ttcher J. Braun B. Brinson Z. Brisson-Tsavoussis L. Brusa R. T. Burley D. Butterfield K. Carloni J. Carpio S. Chattopadhyay N. Chau Y. C. Chen Z. Chen D. Chirkin S. Choi A. Chubarov B. A. Clark D. A. Coloma Borja A. Connolly J. M. Conrad D. F. Cowen C. De Clercq J. J. DeLaunay D. Delgado T. Delmeulle S. Deng P. Desiati K. D. de Vries G. de Wasseige T. DeYoung J. C. D{\'ı}az-V{é}lez S. DiKerby T. Ding M. Dittmer A. Domi L. Draper L. Dueser D. Durnford K. Dutta M. A. DuVernois T. Ehrhardt L. Eidenschink A. Eimer C. Eldridge P. Eller E. Ellinger D. Els{ä}sser R. Engel H. Erpenbeck W. Esmail S. Eulig J. Evans P. A. Evenson K. L. Fan K. Fang K. Farrag A. Fattorini A. R. Fazely A. Fedynitch N. Feigl C. Finley D. Fox A. Franckowiak S. Fukami P. F{ü}rst J. Gallagher E. Ganster A. Garcia M. Garcia E. Genton L. Gerhardt A. Ghadimi C. Glaser T. Gl{ü}senkamp J. G. Gonzalez S. Goswami A. Granados D. Grant S. J. Gray S. Griffin S. Griswold K. M. Groth D. Guevel C. G{ü}nther P. Gutjahr C. Ha A. Hallgren L. Halve F. Halzen L. Hamacher M. Handt K. Hanson J. Hardin A. A. Harnisch P. Hatch A. Haungs J. H{ä}ußler K. Helbing J. Hellrung B. Henke L. Hennig F. Henningsen L. Heuermann R. Hewett N. Heyer S. Hickford A. Hidvegi C. Hill G. C. Hill R. Hmaid K. D. Hoffman A. Hollnagel D. Hooper S. Hori K. Hoshina M. Hostert W. Hou M. Hrywniak T. Huber K. Hultqvist K. Hymon A. Ishihara W. Iwakiri M. Jacquart S. Jain O. Janik M. Jansson M. Jin N. Kamp D. Kang W. Kang A. Kappes L. Kardum T. Karg A. Karle A. Katil M. Kauer J. L. Kelley M. Khanal A. Khatee Zathul A. Kheirandish T. Kim H. Kimku F. Kirchner J. Kiryluk C. Klein Y. Kobayashi S. Koch A. Kochocki R. Koirala H. Kolanoski T. Kontrimas L. K{ö}pke C. Kopper D. J. Koskinen P. Koundal M. Kowalski T. Kozynets A. Kravka N. Krieger J. Krishnamoorthi T. Krishnan K. Kruiswijk E. Krupczak A. Kumar E. Kun N. Kurahashi C. Lagunas Gualda L. Lallement Arnaud M. J. Larson F. Lauber J. P. Lazar K. Leonard DeHolton A. Leszczy{ń}ska C. Li J. Liao C. Lin Q. R. Liu Y. T. Liu M. Liubarska C. Love L. Lu F. Lucarelli W. Luszczak Y. Lyu M. Macdonald E. Magnus Y. Makino E. Manao S. Mancina A. Mand I. C. Mari{ş} S. Marka Z. Marka L. Marten I. Martinez-Soler R. Maruyama J. Mauro F. Mayhew F. McNally K. Meagher A. Medina M. Meier Y. Merckx L. Merten J. Mitchell L. Molchany S. Mondal T. Montaruli R. W. Moore Y. Morii A. Mosbrugger D. Mousadi E. Moyaux T. Mukherjee M. Nakos U. Naumann R. Neshat L. Neste M. Neumann H. Niederhausen M. U. Nisa K. Noda A. Noell A. Novikov A. Obertacke V. O'Dell A. Olivas R. Orsoe J. Osborn E. O'Sullivan B. Owens V. Palusova H. Pandya A. Parenti C. Parisel N. Park V. Parrish E. N. Paudel L. Paul T. Pernice T. C. Petersen J. Peterson S. Pick M. Plum A. Pont{é}n V. Poojyam B. Pries R. Procter-Murphy G. T. Przybylski L. Pyras C. Raab J. Rack-Helleis N. Rad M. Ravn K. Rawlins Z. Rechav A. Rehman I. Reistroffer E. Resconi C. D. Rho W. Rhode L. Ricca B. Riedel A. Rifaie E. J. Roberts S. Rodan M. Rongen A. Rosted C. Rott T. Ruhe L. Ruohan D. Ryckbosch J. Saffer D. Salazar-Gallegos P. Sampathkumar A. Sandrock G. Sanger-Johnson M. Santander S. Sarkar M. Scarnera M. Schaufel H. Schieler S. Schindler L. Schlickmann B. Schl{ü}ter F. Schl{ü}ter N. Schmeisser T. Schmidt F. Schmitt A. Scholz F. G. Schr{ö}der S. Schwirn S. Sclafani D. Seckel L. Seen M. Seikh S. Seunarine P. A. Sevle Myhr R. Shah S. Shah S. Shefali N. Shimizu M. Shin B. Skrzypek R. Snihur J. Soedingrekso D. Soldin P. Soldin G. Sommani D. Song C. Spannfellner G. M. Spiczak C. Spiering J. Stachurska M. Stamatikos T. Stanev T. Stezelberger T. St{ü}rwald T. Stuttard G. W. Sullivan I. Taboada S. Ter-Antonyan A. Terliuk A. Thakuri M. Thiesmeyer W. G. Thompson J. Thwaites W. Tian S. Tilav K. Tollefson J. A. Torres S. Toscano D. Tosi A. K. Upadhyay K. Upshaw A. Vaidyanathan N. Valtonen-Mattila J. Valverde J. Vandenbroucke T. Van Eeden N. van Eijndhoven L. Van Rootselaar J. van Santen J. Vara F. Varsi M. Velazquez M. Venugopal M. Vereecken S. Vergara Carrasco S. Verpoest D. Veske A. Vijai J. Villarreal C. Walck A. Wang E. H. S. Warrick C. Weaver A. Weindl J. Weldert A. Y. Wen C. Wendt J. Werthebach M. Weyrauch N. Whitehorn C. H. Wiebusch D. R. Williams L. Witthaus J. Woodward G. Wrede X. W. Xu J. P. Yanez Y. Yao E. Yildizci S. Yoshida F. Yu S. Yu T. Yuan S. Yun-C{á}rcamo A. Zander Jurowitzki A. Zegarelli S. Zhang Z. Zhang P. Zhelnin P. Zilberman C. Zilleruelo Ca{ñ}as http://arxiv.org/abs/2608.06086v1 2026-08-06T14:28:00Z

We demonstrate, by studying precession-driven flows, that inertial wave hydrodynamic turbulence can drive a robust magnetic dynamo action. Motivated by the stronger damping of large-scale geostrophic vortices in rapidly rotating planetary and stellar interiors, we introduce a controlled damping of the vortices, which usually accompany inertial wave turbulence and feed on wave energy. It is shown that even a small vortex damping results in a significant increase of the growth rate of the dynamo due to inertial waves in the kinematic regime, allowing it to persist for magnetic Prandtl numbers as low as $Pm \sim 10^{-3}$ and Poincaré numbers $Po\sim 0.025$. These critical values of $Po$ and $Pm$ for the dynamo onset decrease with increasing Reynolds number. The onset and growth of the dynamo appear to correlate with the coherent fluctuations of kinetic helicity. Spectral analysis shows that magnetic energy growth is primarily due to inertial-wave-induced induction over a broad range of scales. These results establish inertial waves as an efficient mechanism for magnetic field amplification in rapidly rotating low-$Pm$ flows relevant to planetary and stellar interiors.

2026-08-06T14:28:00Z 9 pages (including End Matter), 8 figures, submitted to Physical Review Letters A. Mishra G. Mamatsashvili M. Le Bars A. J. Barker F. Stefani http://arxiv.org/abs/2608.04901v2 2026-08-06T13:18:14Z

We present a phase-space formulation of elastic full-waveform inversion (FWI) based on local-angle-domain (LAD) representations of seismic scattering. Instead of minimizing acquisition-domain waveform misfits, the method minimizes image-domain inconsistencies in a phase-space domain parameterized by position, propagation direction, scattering angle, azimuth, and time. The formulation is motivated by the observation that velocity and impedance perturbations occupy largely distinct regions of phase space. For anisotropic elastic media, we derive a generalized objective function together with the corresponding adjoint-state gradient and Hessian operators. Representing LAD images through directional and opening wavenumbers provides a unified description of propagation and scattering effects. The approach is independent of the forward-modeling engine and can be implemented with finite-difference, finite-element, spectral-element, ray-based, or beam-based methods. Under Born linearization, ray/beam formulations allow explicit Hessian computation. Owing to the near-orthogonality of directional and opening-angle sensitivities, the Hessian is more diagonally dominant than in conventional FWI, improving parameter separation, reducing velocity-reflectivity crosstalk, and enhancing inversion stability. By organizing seismic information in a physically meaningful phase-space representation, LAD-FWI supports adaptive, resolution-driven inversion and may reduce cycle skipping through attributes that vary more smoothly than waveform residuals. The framework naturally integrates directional focusing, reflector-consistency analysis, anisotropic parameter estimation, converted-wave imaging, and diffraction-based characterization, providing a unified workflow for elastic FWI, migration-velocity analysis, impedance inversion, and high-resolution subsurface imaging.

2026-08-05T14:26:55Z 80 pages, 24 figures Zvi Koren http://arxiv.org/abs/2609.05455v1 2026-08-06T12:18:22Z

Erosion-induced landslide mobility is a prominent phenomenon in nature. However, this topic remains poorly understood, even though erosion greatly alters nearly every aspect of landslide mechanics, dynamics, run-out, mobility, deposition structure and the enormous destructive power carried by such catastrophic events. Here, I present a simple, elegant, and exact analytical model describing the kinematics of erosion-induced landslide mobility in terms of a newly constructed mobility controller containing essential forcing mechanisms in erosion. In contrast to the classical perspective, I formally prove that, irrespective of how large the erosion rate is, the erosion rate alone is not sufficient to determine enhanced or reduced mobility; mobility is also greatly regulated by the volume bulking rate, erosion drag or thrust, and flow depth. With this mechanical mobility controller, it is now possible to explain precisely when the mobility of an erodible landslide is enhanced or reduced. I demonstrate that, a mechanically valid, physically founded erosion model must be able to explain both the enhanced and reduced mobility of erosive landslides.

2026-08-06T12:18:22Z Shiva P. Pudasaini http://arxiv.org/abs/2608.05763v1 2026-08-06T08:51:43Z

Full waveform inversion (FWI) can recover high-resolution subsurface velocity models. Conventional waveform-difference objectives, however, are vulnerable to cycle skipping when the starting model is inaccurate. We introduce an FWI objective that compares features produced from modeled and observed seismic traces by SeisLM, a pretrained seismic foundation model. The SeisLM encoder remains frozen during inversion, and the feature discrepancy is differentiated with respect to the modeled traces to construct an adjoint source compatible with the standard adjoint-state framework. We also test a scheduled hybrid loss that combines the SeisLM feature loss with the conventional $L_2$ objective. Time-shift diagnostics show that the loss computed from features produced by the pretrained encoder has a broader and smoother basin around the correct alignment than either the waveform $L_2$ objective or the feature loss obtained from an encoder with the same architecture and randomly initialized parameters. In the Marmousi experiment, the SeisLM and hybrid objectives produce similar improvements during early-stage inversion and provide useful models for subsequent reflection-based $L_2$ refinement. In the 2D Overthrust experiment, which begins from a laterally invariant linear-gradient model, the SeisLM feature-loss workflow outperforms the conventional and hybrid workflows, indicating that introducing the $L_2$ contribution too early can reintroduce cycle-skipping sensitivity. In the 3D Overthrust experiment, conventional $L_2$ inversion stalls near the initial linear gradient, whereas the SeisLM feature loss guides the inversion toward a background model from which $L_2$ refinement recovers the principal structures. These results support using features produced by pretrained seismic networks to define early-stage FWI objectives rather than complete replacements for waveform-domain misfits.

2026-08-06T08:51:43Z Mustafa Alfarhan Matteo Ravasi Fuqiang Chen George Turkiyyah David Keyes http://arxiv.org/abs/2608.05593v1 2026-08-06T04:34:45Z

Alternate bars are ubiquitous bedforms in alluvial rivers, and excessive bar-height growth can increase flood risk and disrupt ecosystems. Predicting bar-height variability requires quantifying the dependence of bar growth rate on water discharge, yet this has rarely been achieved because conventional flume measurements interrupt the flow and cannot continuously capture bar evolution. Here we conducted laboratory experiments under four discharges and four channel slopes (16 conditions) and continuously mapped bed topography using Stream Tomography, a nonintrusive, flow-through measurement technique. Bar height exhibits sigmoidal growth and is well described by the Landau equation, allowing robust estimation of the growth rate and equilibrium height. The inferred dimensionless growth rate tends to increase as discharge decreases, but its sensitivity is weaker than that predicted by stability theory for a single alternate-bar mode. This discrepancy becomes more pronounced at lower discharge, indicating limitations of linearization and the influence of interactions among multiple bar modes. Furthermore, numerical flow simulations over the measured bed topography reveal that, during bar development, lower-discharge conditions produce a relatively larger bar height-to-depth ratio, making the flow more prone to deflection. In such cases, sediment transport is restricted downstream of depositional areas, thereby limiting bar migration. Consequently, flow and sediment transport become concentrated in the scour zones, enhancing local scour and promoting the growth of bar height. These results provide benchmark constraints for bar stability theories and help improve predictions of how bar height responds to changes in discharge.

2026-08-06T04:34:45Z 38 pages, 13 figures Shohei Seki Daichi Moteki Hiroyasu Yasuda http://arxiv.org/abs/2603.00393v2 2026-08-05T20:56:29Z

Inverse problems constrained by partial differential equations are often ill-conditioned due to noisy, incomplete data or inherent non-uniqueness. A prominent example is full waveform inversion (FWI), which estimates Earth's subsurface properties by fitting seismic measurements subject to the wave equation, where ill-conditioning stems from noisy, band-limited, finite-aperture measurements and complex geological structures. A Bayesian framework describes the solution more comprehensively: instead of a single estimate, a posterior distribution of plausible solutions characterizes the non-uniqueness and can be sampled to quantify uncertainty. However, no clear procedure exists for translating hard physical constraints, such as the wave equation, into priors amenable to existing sampling techniques. We address this by sampling the posterior in the dual space via an augmented Lagrangian formulation, which converts hard constraints into penalties suited to sampling algorithms while enforcing them progressively through multiplier updates, so they are satisfied in the limit. We integrate the alternating direction method of multipliers (ADMM) with Stein variational gradient descent (SVGD), a particle-based sampler: the constraint is relaxed at each iteration and the multiplier updates progressively enforce its satisfaction. This enables posterior sampling under hard constraints while inheriting the favorable conditioning of dual-space solvers, where partial constraint relaxation permits productive updates even when the current model is far from the true solution. We validate the method on a stylized Rosenbrock conditional inference problem and on frequency-domain FWI for a Gaussian anomaly model and the Marmousi II benchmark, demonstrating physically consistent uncertainty estimates and posterior contraction with increasing data coverage.

2026-02-28T00:34:31Z https://github.com/luqigroup/SVGDADMMSampler.jl Ali Siahkoohi Kamal Aghazade Ali Gholami http://arxiv.org/abs/2604.11488v2 2026-08-05T17:31:34Z

Observations show the multidimensional dynamics of meltwater and distribution of ice layers in the firn on the Greenland Ice Sheet. However, state-of-the-art large-scale models for firn hydrology are essentially one-dimensional, limiting their ability to explain observed datasets and failing to reduce uncertainty in surface mass balance and sea-level rise estimates. Here, we present a large-scale, multidimensional, multiphase, and thermomechanical model to simulate firn hydrology. The model is highly efficient due to a novel algorithm in which an extra equation for pressure is solved only in saturated regions. Furthermore, the model can apply spatially heterogeneous boundary conditions to the unsaturated-saturated domain and allows for the dynamic formation of fully impermeable ice layers. The numerical results show excellent comparisons against analytic solutions to one- and two-dimensional problems that involve coupled unsaturated-saturated flows, thermodynamics, and phase change. We further apply the model to investigate field data from southwest Greenland and find that lateral heterogeneities strongly influence the depth of melt percolation and ice layer formation. Improved understanding of these local, multidimensional processes will provide physics-based constraints on firn densification, reduce uncertainty in converting altimetric elevation change to mass change, and improve estimates of freshwater fluxes to the ocean under a warming climate.

2026-04-13T13:54:22Z 41 pages, 12 figures, 1 table Mohammad Afzal Shadab Surendra Adhikari C. Max Stevens Asa K. Rennermalm Jing Xiao Marc A. Hesse Reed M. Maxwell