Relativistic electrons in the radiation belts can be transported as a result of wave-particle interactions (WPI) with ultra-low frequency (ULF) waves. Such WPI are often assumed to be diffusive, parametric models for the radial diffusion coefficient often being used to assess the rates of radial transport. However, these WPI transition from initially coherent interactions to the diffusive regime over a finite time, this time depending on the ULF wave power spectral density, and local resonance conditions. Further, in the real system on the timescales of a single storm, interactions with finite discrete modes may be more realistic. Here, we use a particle-tracing model to simulate the dynamics of outer radiation belt electrons in the presence of a finite number of discrete frequency modes. We characterize the point of the onset of diffusion as a transition from separate discrete interactions in terms of wave parameters by using the “two-thirds” overlap criterion (Lichtenberg & Lieberman, 1992), a comparison between the distance between, and the widths of, the electron’s primary resonant islands in phase space. Further, we find the particle decorrelation time in our model system with typical parameters to be on the timescale of hours, which only afterwards can the system be modeled by one-dimensional radial diffusion. Direct comparison of particle transport rates in our model with previous analytic diffusion coefficient formulations show good agreement at times beyond the decorrelation time. These results are critical for determining the time periods and conditions under which ULF wave radial diffusion theory can be applied.

Predicting the future contribution of the ice sheets to sea level rise over the next decades presents several challenges due to a poor understanding of critical boundary conditions, such as basal sliding. Traditional numerical models often rely on data assimilation methods to infer spatially variable friction coefficients by solving an inverse problem, given an empirical friction law. However, these approaches are not versatile, as they sometimes demand extensive code development efforts when integrating new physics into the model. Furthermore, this approach makes it difficult to handle sparse data effectively. To tackle these challenges, we propose a novel approach utilizing Physics-Informed Neural Networks (PINNs) to seamlessly integrate observational data and governing equations of ice flow into a unified loss function, facilitating the solution of both forward and inverse problems within the same framework. We illustrate the versatility of this approach by applying the framework to two-dimensional problems on the Helheim Glacier in southeast Greenland. By systematically concealing one variable (e.g. ice speed, ice thickness, etc.), we demonstrate the ability of PINNs to accurately reconstruct hidden information. Furthermore, we extend this application to address a challenging mixed inversion problem. We show how PINNs are capable of inferring the basal friction coefficient while simultaneously filling gaps in the sparsely observed ice thickness. This unified framework offers a promising avenue to enhance the predictive capabilities of ice sheet models, reducing uncertainties, and advancing our understanding of poorly constrained physical processes.

We report a citizen science-motivated study on the cause of an unusually bright red aurora as witnessed from Hokkaido, Japan during a magnetic storm on December 1, 2023. Such an intense red aurora event has occurred in the Halloween 2003 super storm, but the Dst index peak of this December 2023 storm was only -107 nT. In spite of the moderate storm amplitude, the extremely high solar wind density of >50 /cc caused the aurora oval extension to 53 MLAT (L=2.8). We discuss that the drift loss of the ring current particles across the small-size magnetopause is important, and Hokkaido was at the right position to see the direct effect of the large particle injection of the storm-time substorm.

The survival of salt marshes, especially facing future sea-level rise, requires the supply of sediment. Sediment can be supplied to salt marshes via two routes: through marsh creeks and over marsh edges. However, the conditions of tides and waves that facilitate sediment import through these two routes remain unclear. To better understand when and how sediment import towards salt marshes occurs, measurements spanning two months were conducted in Paulina Saltmarsh. The results show that the marsh creek and the marsh edge do not import sediment simultaneously. The marsh creek tends to import sediment during neap tides with waves. A small tidal range results in weaker flow during ebb tides, reducing the export of sediment. Strong waves, particularly during this period, enhance the sediment supply from mudflats to the marsh creek. Additionally, waves can directly affect sediment re-suspension in the marsh creek during spring tides when the water level is above the marsh canopy. The marsh edge benefits from contrasting tidal and wave conditions, with sediment imported during spring tides with weak waves. Waves during spring tides re-suspend sediment, impeding the sediment deposition, and thus leading to sediment export over the marsh edge. These results highlight the potential sediment transport routes to marshes under varying conditions, shedding light on their implications for the long-term survival of salt marshes.

We present an artificial neural network (ANN) model that reconstructs > 30 keV electron flux measurements near the geomagnetic equator from low-Earth-orbit (LEO) observations, exploiting the global coherent nature of the high-energy trapped electrons that constitute the radiation belts. To provide training data, we analyze magnetic conjunctions between one of National Oceanic and Atmospheric Administration’s (NOAA’s) Polar Orbiting Environmental Satellites (POES) and National Aeronautics and Space Administration’s (NASA’s) Van Allen Probes. These conjunctions occur when the satellites are connected along the same magnetic field line and allow for a direct comparison of satellites’ electron flux measurements for one integral energy channel, > 30 keV and over 64,000 such conjunctions have been identified. For each conjunction, we fit the equatorial pitch angle distribution (PAD) parameterized by the function JD = C·sinNα. The resulting conjunction dataset contains the POES electron flux measurements, L and MLT coordinates, geomagnetic activity AE index, and C and N coefficients from the PAD fit for each conjunction. We test combinations of input variables from the conjunction dataset and achieve the best model performance when we use all the input variables during training. We present our model’s prediction for the out-of-sample data that agrees well with observations, R2 > 0.80. We demonstrate the ability to nowcast and reconstruct equatorial electron flux measurements from LEO without the need for an in-situ equatorial satellite. The model can be expanded to include existing LEO data and has the potential to be used as a basis of future radiation-belt monitoring LEO constellations.

We propose the “endo-exo” conceptual framework to account for the varied and complex episodic landslide movements observed during progressive maturation until collapse/stabilization. This framework captures the interplay between exogenous stressors such as rainfall and endogenous damage/healing processes. The underlying physical picture involves cascades of local triggered mass movements due to fracturing and sliding. We predict four distinct types of episodic landslide dynamics (exogenous/endogenous-subcritical/critical), characterized by power-law relaxations with different exponents, all related to a single parameter ϑ. These predictions are tested on the dataset of the Preonzo landslide, which exhibited multi-year episodic movements prior to a final collapse. All episodic activities can be accounted for within this classification with ϑ≈0.45±0.1, providing strong support for our parsimonious theory. We further show that the final catastrophic failure of this landslide is clearly preceded by an increased frequency of large velocities corresponding to a transition to a supercritical regime with amplifying positive feedbacks.

This study explores the impact of deep ( >5 m) critical zone (CZ) architecture on vegetation distribution in a semi-arid snow-dominated climate. Utilizing seismic refraction surveys, we identified a significant correlation between saprolite thickness and LiDAR-derived canopy heights (R²=0.47). We argue that CZ structure, specifically shallow fractured bedrock under valley bottoms, redirects groundwater to locations where trees are established—suggesting they are located in specific locations with access to nutrients and water. This work provides a unique spatially exhaustive perspective and adds to growing evidence that in addition to other factors such as slope, aspect, and climate, deep CZ structure plays a vital role in ecosystem development and resilience.

The atmospheric CO2 growth rate is a fundamental measure of climate forcing. NOAA's growth rate estimates, derived from in situ observations at the marine boundary layer (MBL), serve as the benchmark in policy and science. However, NOAA's MBL-based method encounters challenges in accurately estimating the whole-atmosphere CO2 growth rate at sub-annual scales. We introduce the Growth Rate from Satellite Observations (GRESO) method as a complementary approach to estimate the whole-atmosphere CO2 growth rate utilizing satellite data. Satellite CO2 observations offer extensive atmospheric coverage that extends the capability of the current NOAA benchmark. We assess the sampling errors of the GRESO and NOAA methods using ten atmospheric transport model simulations. The simulations generate synthetic OCO-2 satellite and NOAA MBL data for calculating CO2 growth rates, which are compared against the global sum of carbon fluxes used as model inputs. We find good performance for the NOAA method (R = 0.93, RMSE = 0.12 ppm/year or 0.25 PgC/year). GRESO demonstrates lower sampling errors (R = 1.00; RMSE = 0.04 ppm/year or 0.09 PgC/year). Additionally, GRESO shows better performance at monthly scales than NOAA (R = 0.77 vs 0.47, respectively). Due to CO2's atmospheric longevity, the NOAA method accurately captures growth rates over five-year intervals. GRESO's robustness across partial coverage configurations (ocean or land data) shows that satellites can be promising tools for low-latency CO2 growth rate information, provided the systematic biases are minimized using in situ observations. Along with accurate and calibrated NOAA in situ data, satellite-derived growth rates can provide information about the global carbon cycle at sub-annual scales.

The amount and spatial distribution of surface displacement that occurs during an earthquake are critical information to our understanding of the earthquake source and rupture processes. However, the earthquake surface displacement generally occurs over wide regions, includes multiple components affecting the ground surface at different spatial scales, and is challenging to characterize. In this study, we assess the sensitivity of optical imagery and topography datasets of different resolutions to the earthquake surface displacement when using optical image cross-correlation (OIC) techniques. Results show that the average noise in the output displacement maps linearly increases with decreasing image resolution, leading to greater uncertainty in determining the geometry of the faults and the associated displacement. Fault displacements are, on average, under-estimated by a factor ~0.7-0.8 when using 10 m compared to 0.5 m resolution imagery. Our analysis suggests that an optical image resolution of ≤1 m is necessary to accurately capture the complexity of the ground displacement. We also demonstrate that sub-meter vertical accuracy of the digital surface/elevation model (DSM/DEM) is also required for accurate image orthorectification, and is better achieved using high-resolution stereo optical imagery than existing global baseline topography data. Together, these results highlight the measurement needs for improving the observation of earthquake surface displacement towards the development of future Earth surface topography and topography change observing systems.

Although supervised deep learning (DL) offers a potent solution for removing noise from seismic records, challenges are encountered owing to the scarcity of noise-free labels. The innovative Noise2Noise method eliminates the need for clean training targets and extends the applicability of deep learning to seismic data denoising. In this study, we introduce the Noise2Noise Enhancement (N2NE) framework, which improves upon the conventional noise reduction methods used in seismic processing. The applicability of this framework was quantitatively examined using actual field noise under two scenarios: with and without repeated shots. In scenarios with repeated shots, the N2NE framework enhances the conventional stacking method. In addition, the substack strategy, which employs smaller substacks for preliminary noise suppression before DL training, boosts noise suppression. In scenarios without repeated shots, the N2NE framework refines conventional denoising methods (F-X deconvolution) by utilizing information from the common-shot and receiver domains. The N2NE framework lays a foundation for future research on N2N-based seismic denoising methods and contributes to improving the quality of seismic records and the efficiency of data acquisition.

Enhanced geothermal systems (EGS) are promising for generating clean power by extracting heat energy from injection and extraction of water in geothermal reservoirs. The stimulation process involves hydroshearing which reactivates pre-existing cracks for creating permeability and meanwhile inducing microearthquakes. Locating these microearthquakes provide reliable feedback on the stimulation progress, but it poses a challenging nonlinear inverse problem. Current deep learning methods for locating earthquakes require extensive datasets for training, which is problematic as detected microearthquakes are often limited. To address the scarcity of training data, we propose a transfer learning workflow using probabilistic multilayer perceptron (PMLP) which predicts microearthquake locations from cross-correlation time lags in waveforms. Utilizing a 3D velocity model of Newberry site derived from ambient noise interferometry, we generate numerous synthetic microearthquakes and 3D acoustic waveforms for PMLP training. Accurate synthetic tests prompt us to apply the trained network to the 2012 and 2014 stimulation field waveforms. Predictions on the 2012 stimulation dataset show major microseismic activity at depths of 0.5–1.2 km, correlating with a known casing leakage scenario. In the 2014 dataset, the majority of predictions concentrate at 2.0–2.9 km depths, consistent with results obtained from conventional physics-based inversion, and align with the presence of natural fractures from 2.0–2.7 km. We validate our findings by comparing the synthetic and field picks, demonstrating a satisfactory match for the first arrivals. By combining the benefits of quick inference speeds and accurate location predictions, we demonstrate the feasibility of using transfer learning to locate microseismicity for EGS monitoring.

Thermal runaway is a ductile localization mechanism that has been linked to deep-focus earthquakes and pseudotachylyte formation. In this study, we investigate the dynamics of this process using one-dimensional, numerical models of simple shear deformation. The models employ a visco-elastic rheology where viscous creep is accommodated with a composite rheology encompassing diffusion and dislocation creep as well as low-temperature plasticity. To solve the nonlinear system of differential equations governing this rheology, we utilize the pseudo-transient iterative method in combination with a viscosity regularization to avoid resolution dependencies. To determine the impact of different model parameters on the occurrence of thermal runaway, we perform a parameter sensitivity study consisting of 6000 numerical experiments. We observe two distinct behaviors, namely a stable regime, characterized by transient shear zone formation accompanied by a moderate (100 - 300 Kelvin) temperature increase, and a thermal runaway regime, characterized by strong localization, rapid slip and a temperature surge of thousands of Kelvin. Nondimensional scaling analysis allows us to determine two dimensionless groups that predict model behavior. The ratio tr/td represents the competition between heat generation from stress relaxation and heat loss due to thermal diffusion while the ratio Uel/Uth compares the stored elastic energy to thermal energy in the system. Thermal runaway occurs if tr/td is small and Uel/Uth is large. Our results demonstrate that thermal runaway is a viable mechanism driving fast slip events that are in line with deep-focus earthquakes and pseudotachylyte formation at conditions resembling cores of subducting slabs.

We study the dynamic evolution of dayside magnetopause reconnection locations and their dependence on the interplanetary magnetic field (IMF) cone angle via 3-D global-scale hybrid simulations. Cases with finite IMF Bx and Bz but By=0 are investigated. It is shown that the dayside magnetopause reconnection is unsteady under quasi-steady solar wind conditions. The reconnection lines during the dynamic evolution are not always parallel to the equatorial plane even under purely southward IMF conditions. Magnetopause reconnection locations can be affected by the generation, coalescence, and transport of flux ropes (FRs), reconnection inside the FRs, and the magnetosheath flow. In the presence of an IMF component Bx, the magnetopause reconnection initially occurs in high-latitude regions downstream of the quasi-perpendicular bow shock, followed by the generation of multiple reconnection regions. In the later stages of the simulation, a dominant reconnection region is present in low-latitude regions, which can also affect reconnection in other regions. The global distribution of reconnection lines under a finite IMF Bx is found to not be limited to the region with maximum magnetic shear angle.

Seismicity at active volcanoes provides crucial constraints on the dynamics of magma systems and complex fault activation processes preceding and during an eruption. We characterize time-dependent spectral features of volcanic earthquakes at Axial Seamount with unsupervised machine learning methods, revealing mixed frequency signals that emerge from the continuous waveforms about 15 hours before eruption onset. The events migrate along pre-existing fissures, suggesting that they represent brittle crack opening driven by influx of magma or volatiles. These results demonstrate the power of novel machine learning algorithms to characterize subtle changes in magmatic processes associated with eruption preparation, offering new possibilities for forecasting Axial's anticipated next eruption. This novel method is generalizable and can be employed to identify similar precursory signals at other active volcanoes.

The ionospheric convection electric field is often assumed to be a potential field. This assumption is not always valid, especially when the ionosphere changes on short time scales $T \lesssim 5$~min. We present a technique for estimating the induction electric field using ground magnetometer measurements. The technique is demonstrated on real and simulated data for sudden increases in solar wind dynamic pressure of $\sim$1 and 10 nPa, respectively. For the real data, the ionospheric induction electric field is 0.15$\pm$0.015 mV/m, and the corresponding compressional flow is 2.5$\pm$0.3 m/s. For the simulated data, the induction electric field and compressional flow reach 3 mV/m and 50 m/s, respectively. The induction electric field can locally constitute tens of percent of the total electric field. Inclusion of the induction electric field increased the total Joule heating by 2.4\%. Locally the Joule heating changed by tens of percent. This corresponds to energy dissipation that is not accounted for in existing models.

The low frictional strength of smectite minerals such as montmorillonite is thought to play a crucial role in controlling the rheology and the stability of clay-rich faults. In this study, we perform coarse-grained molecular dynamics simulations on a model clay system, in which clay platelets are simplified as oblate ellipsoids interacting via Gay-Berne potential. We study the rheology and the structure development during shear in this model system, which is sheared at constant strain rates for 10 strains after compression and equilibrium. We find that the system exhibits velocity-strengthening behavior over a range of normal stresses from 1.68 to 56.18 MPa and a range of strain rates from 6.93´105 to 6.93´108 /s. The relationship between the shear stress and the strain rate follows the Herschel-Bulkley model. In general, shear is localized at lower strain rate and higher normal stress, whereas the homogeneous shear is realized at higher strain rates. The structure change by the shear is analyzed from various aspects: the volume fraction, the particle orientation, the velocity profile, and the parallel radial distribution function. We find that particle rearrangement and compaction dominate at the early stage of shear when the shear stress increases. Shear band starts to form at the later stage when the shear stress decreases and relaxes to a steady-state value. The structure development at low strain rates is similar to that in previous experimental observations. The stacking structure weakens during shear, and restores logarithmically with time in the rest period.

In geological characterization, the traditional methods that rely on the covariance matrix for continuous variable estimation often either neglect or oversimplify the challenge posed by subsurface non-stationarity. This study presents an innovative methodology using ancillary data such as geological insights and geophysical exploration to address this challenge directly, with the goal of accurately delineating the spatial distribution of subsurface petrophysical properties, especially, in large geological fields where non-stationarity is prevalent. This methodology is based on the geodesic distance on an embedded manifold and is complemented by the level-set curve as a key tool for relating the observed geological structures to intrinsic geological non-stationarity. During validation, parameters 𝜌 and 𝛽 were revealed to be the critical parameters that influenced the strength and dependence of the estimated spatial variables on secondary data, respectively. Comparative evaluations showed that our approach performed better than a traditional method (i.e., kriging), particularly, in accurately representing the complex and realistic subsurface structures. The proposed method offers improved accuracy, which is essential for high-stakes applications such as contaminant remediation and underground repository design. This study focused primarily on twodimensional models. There is a need for three-dimensional advancements and evaluations across diverse geological structures. Overall, this research presents novel strategies for estimating non-stationary geologic media, setting the stage for improved exploration of subsurface characterization in the future.

A document by Bhagyashree Waghule. Click on the document to view its contents.

The minerals carrying the magnetic remanence in geological samples are commonly a solid solution series of iron-titanium spinels known as titanomagnetites. Despite the range of compositions within this series, micromagnetic studies that characterize the magnetic domain structures present in these minerals have typically focused on magnetite. No studies systematically comparing the domain-states present in titanomagnetites have been undertaken since the discovery of the single vortex (SV) structure and the advent of modern micromagnetism. The magnetic properties of the titanomagnetite series are known to vary with composition, which may influence the domain states present in these minerals, and therefore the magnetic stability of the samples bearing them. We present results from micromagnetic simulations of titanomagnetite ellipsoids of varying shape and composition to find the size ranges of the single domain (SD) and SV structures. These size ranges overlap, allowing for regions where the SD and SV structures are both available. These regions are of interest as they may lead to magnetic instability and “pTRM tails’ in paleointensity experiments. We find that although this SD+SV zone occupies a narrow range of sizes for equidimensional magnetite, it is widest for intermediate (TM30-40) titanomagnetite compositions, and increases for both oblate and prolate particles, with some compositions and sizes having an SD+SV zone up to 100s of nm wide. Our results help to explain the prevalence of pTRM tail-like behavior in paleointensity experiments. They also highlight regions of particles with unusual domain states to target for further investigation into the definitive mechanism behind paleointensity failure.

This study presents new seismic imaging of the Andean subduction zone through P-wave hybrid finite-frequency and ray-theoretical tomography. We measured both differential and absolute traveltimes using broadband seismic waveforms from stations in an array of ocean-bottom seismographs near the Chile Triple Junction (CTJ) and stations within 30° from the array. These data were combined with the global traveltime dataset to obtain a global P-wave velocity structure with a focus on central to southern South America. The new tomographic image showed the Nazca slab geometry as a continuous fast anomaly, which is consistent with seismic activity and prior slab models. Furthermore, two notable structures were observed: a broad extension of the fast anomaly beneath the Nazca slab at 26–35° S and a slow anomaly east of the CTJ. The checkerboard resolution and recovery tests confirmed the reliability of these large-scale features. The fast anomaly, isolated from the Nazca slab, was interpreted as a relic Nazca slab segment based on its strong amplitude and spatial coincidence with the current Pampean and past Payenia flat slab segments. The slow anomaly near the CTJ was consistent with the previously inferred extent of the Patagonian slab window. Moreover, the active adakitic volcanoes are aligned with the southern edge of the anomaly, and the plateau basalts are located within the anomaly. Our model showed that the slow anomaly extended to a depth of up to 250 km, suggesting a depth limit that the asthenospheric window can influence.

A document by Danica Roth. Click on the document to view its contents.

This paper compares the ability of deep learning and entropic learning methods to predict the probability of the Niño3.4 index being above 0.4 (El Niño), below 0.4 (La Niña) or within both of these thresholds (neutral) at lead times of 3 up to 24 months. In particular, the performance, interpretability, and training cost of entropic learning methods, represented by the entropy-optimal Scalable Probabilistic Approximation (eSPA) algorithm, are compared with deep learning methods, represented by a Long Short-Term Memory (LSTM) classifier, trained on the same dataset. Using only data derived from observations over the period 1958-2018 and a corresponding surface-forced ocean model, the problem manifests as a canonical smalldata challenge. Relative to the LSTM model, eSPA exhibits substantially better out-of-sample performance in terms of area under the ROC curve (AUC) for all lead times at ⇠ 0.02% of the computational cost. Comparisons of AUC with other state-of-the-art deep learning models in the literature show that eSPA appears to also be more accurate than these models across all three classes. Composite images are generated for each of the cluster centroids from each trained eSPA model at each lead time. At shorter lead times, the composite images for the most significant clusters correspond to patterns representing mature or emerging/declining El Niño or La Niña states, while at longer lead times they correspond to precursor states consisting of extra-tropical anomalies. Finally, modifications to the baseline dataset are explored, showing that improvements can be made in the parsimony of the trained eSPA model without sacrificing predictive power.

Numerical modeling of the fabric of naturally deposited sands depends on being able to accurately reconstruct individual grains in a reasonable amount of time. To this end, X-Ray Computed Tomography (XRCT) is an excellent tool. However, while the currently available processing workflows, e.g. (Stamati, O., et al, 2020), have been successfully used to obtaining the shapes of clean, pluviated sands, the image reconstruction is a lot more challenging in resolving the individual grains and fabric of naturally deposited fine sands, such as shown in Figure 1. The focus of this study has been to develop a more robust that can rapidly reconstruct the grain avatars in sufficient detail.

Temperature is of primary importance for many physical properties in the Martian soil. We measured diurnal and annual soil (and surface) temperature variations using the NASA InSight Mars mission’s HP3 radiometer and thermal probe. At the depth of the probe of 0.5 - 36 cm, an average temperature of 217.5K was found varying by 5.3 - 6.7 K during a sol and by 13.2K during the seasons. The damping of the surface temperature variations in the soil were used to derive a thermal diffusivity of 2.30±0.03×10−8 m2/s for the depth range of the diurnal wave - thermal skin depth 2.5±0.04 cm - and 3.74±0.61×10−8 m2/s for that of the annual wave, with a thermal skin depth of 84±10 cm. The temperatures measured are conducive to the deliquesence of thin films of brines in the soil. These are of astrobiological interest and may explain the formation of the observed cemented duricrust.