On Venus, the coexistence of large volcanic highlands interpreted as the surface expression of long-lived mantle plumes, alongside coronae, smaller features thought to be caused by transient thermal diapirs, remains enigmatic. Using numerical models of mantle convection with sharp and broad mineral phase transitions, we show that both scales of upwellings are generated in a stagnant lid planet with an interior temperature 400 K warmer than Earth's. The smaller plumes originate from a ~600 km deep internal layer that exists as a consequence of the different sequence of mineral phase transitions that occurs in warmer mantles. These models further produce essential features of Venus' geodynamics, including apparent large scale downwellings, variable heat flow and plains dynamic topography.

Landslide risk is traditionally predicted by process-based models with detailed assessments or point-scale, attribute-based machine learning (ML) models with first- or second-order features, e.g., slope, as inputs. One could hypothesize that terrain patterns might contain useful information that could be extracted, via computer vision ML models, to elevate prediction performance beyond that achievable with low-order features. We put this hypothesis to the test in the state of Oregon, where a large landslide dataset is available. The image-processing convolutional neural networks (CNN2D) using 2D terrain data obtained either higher Precision or higher Recall than attribute-based random forest (RF1D) models, but could not improve both simultaneously. While CNN2D can be set up to identify more real events, it would then introduce more false positives, highlighting the challenge of generalizing landslide-prone terrain patterns and the potential omission of critical factors. However, ensembling CNN2D and RF1D produced overall better Precision and Recall, and this cross-model-type ensemble was better than other ways to ensemble, leveraging information content of fine-scale topography while suppressing its noise. These models further showed robust results in cross-regional validation. Our perturbation tests showed that 10m resolution (the smallest possible) produced the best model in a range of resolutions. Rainfall, land cover, soil moisture, and elevation were the most important predictors. Based on the results of the analysis, we generated landslide susceptibility maps, providing insights into spatial patterns of landslide risk.

The ocean plays a major role in the moderation of anthropogenically-induced climate change by absorbing roughly a quarter of anthropogenic CO2 (Cant). This absorption of Cant by the ocean leads to ocean acidification, threatening marine’s life. The North Atlantic Ocean encompasses the highest ocean storage capacity of Cant per unit area. The subpolar North Atlantic gyre is subject to a large seasonal to decadal variability that might impact Cant storage. To investigate Cant evolution over the 2011-2021 period and its relationship with ocean dynamics in this region, we use the Argo-O2 array combined with neural networks and a back-calculation method (φCTO method). We compute monthly time-series of Cant in the Labrador and Irminger Seas. We show that Cant concentrations in the first 2000 dbar of the Labrador and Irminger Seas are strongly affected by winter deep convection, especially between winter 2015 and winter 2018. The Cant inventories in the top 2000 dbar of the Labrador and Irminger Seas increase through time, at rates of 1.63±0.32% yr-1 and 1.49±0.30% yr-1, respectively. Our monthly Argo-based Cant estimates complement high-quality ship-based measurements acquired at a biennial or lower frequency. Additionally, this study shows that Cant concentrations and Cant inventories in deep convection areas may depend on the method employed to calculate Cant. As a consequence, we take over the model ensemble idea and propose to use several methods to compute Cant, which would give its methodological uncertainty.

Trans-dimensional Bayesian sampling has been applied to subsurface imaging and other inference problems across the Earth Sciences. A particular style of Markov chain Monte Carlo (McMC) method, known as reversible-jump has been used almost universally in such studies. This algorithm allows sampling across variably dimensioned model parameterizations. However, for practical reasons, it is limited to cases where the number of free parameters differ in a regular sequence between alternate models, usually by addition or subtraction of a single variable. Furthermore, jumps between model dimensions rely on bespoke mathematical transformations, which are bespoke to each class of application. As a result, implementations are dependent on the choice of model parameterization employed. A framework for Trans-conceptual Bayesian sampling, which is a generalization of trans-dimensional sampling, is presented. Trans-C Bayesian sampling allows exploration across a finite, but arbitrary, set of conceptual models, i.e. ones where the number of variables, the type of model basis function, nature of the forward problem, and assumptions on the measurement noise statistics, may all vary independently. The new framework avoids parameter transformations and thereby lends itself to development of automatic McMC algorithms, i.e. where the details of the sampler do not require knowledge of the parameterization. Algorithms implementing Bayesian conceptual model sampling are presented and illustrated with examples drawn from geophysics, using real and synthetic data. Comparison with reversible-jump illustrates that Trans-C sampling produces statistically identical results for situations where the former is applicable, but also allows sampling in situations where Trans-D would be impractical to implement.

A new closure, ROSSMIX 2.0, for the effect of meso-scale eddies in non-eddy-resolving ocean models is presented and evaluated. It combines aspects of several previous closures in a simplified approach: local linear stability analysis is used to predict the vertical and lateral shape of eddy correlations, while a wave energy equation co-integrated in the ocean model predicts their amplitudes. The new closure is implemented and evaluated with good success in an idealised channel model of vertical and lateral shear instability, and in a realistic quasi-global ocean model.

not-yet-known not-yet-known not-yet-known unknown This study investigates the F-region plasma density and irregularities response during the May-2024 geomagnetic storm using data from Swarm and GRACE-FO missions. While Swarm-B provided measurements at 11/23 LTs, data from GRACE-FO and Swarm-A were collected at 05/17 LTs and 07/19 LTs, respectively. Following the storm commencement on May 10, the equatorial ionization anomaly (EIA) was particularly strong (depleted) on the dayside (nightside). Additionally, an unexpected development of the EIA was observed in the early morning 05/07 LTs. Deviations during the storm’s initial and recovery phases indicate the strong influence of penetration and disturbance dynamo electric fields. Significant equatorial plasma density depletion (EPD) activity with apex altitudes reaching up to ~ 5000 km during pre-midnight on May 11 and ~ 3400 km in the early morning was also noticed. We suggest that the morning EPDs are locally generated rather than being remnants from the previous night.

Non-Newtonian fluid flows in porous media are critical in various subsurface and geotechnical engineering applications. However, accurately predicting such flows remains challenging due to complex fluid rheology and intricate pore structures. This study focuses on polymer fluids with shear-thinning rheology. To address the limitations of existing models, we derive a theoretical conductance model for polymer flow in a capillary tube, based on which a customized pore-network method is developed. Simulations reveals three distinct flow regimes, highlighting the impact of complex rheology on flow dynamics. Notably, flow heterogeneity amplifies as the shear-thinning feature directs more flows through wider pores, where the effective viscosity decreases more significantly compared to narrower ones. The conductance model also enables formulating a generalized Darcy's law for non-Newtonian fluids, validated through Pore-network modeling. The proposed framework is adaptable to a broad range of non-Newtonian fluids, offering valuable insights for scaling up to field-scale applications.

The lack of accurate information on the snow water equivalent (SWE) including its spatio-temporal variations in mountain catchments remains a key problem in snow hydrology and water resources management. This is partly because there is no sensor to measure SWE beyond local scale. At Mt. Zugspitze, Germany, a superconducting gravimeter senses the gravity effect of the seasonal snow, reflecting the temporal evolution of SWE in a few kilometer scale radius. We used this new observation to evaluate two configurations of the Alpine3D distributed snow model. In the default run, the model was forced with meteorological station data. In the second run, we applied precipitation correction based on an 8 m resolution snow depth image derived from satellite observations. The snow depth image strongly improved the simulation of the snowpack gravity effect during the melt season. This result suggests that satellite observations can enhance SWE analyses in mountains with limited infrastructure.

Surface information derived from exospheric measurements at planetary bodies complements surface mapping provided by dedicated imagers, offering critical insights into surface release processes, dynamics of various interactions within the planetary environment, erosion, space weathering, and planetary evolution. This study explores a method for deriving the elemental composition of Mercury’s regolith from in-situ measurements of its neutral exosphere using deep neural networks (DNNs). We present a supervised feed-forward DNN architecture—a network of fully-connected neural layers, the so-called multilayer perceptron (MLP). This network takes exospheric densities and proton precipitation fluxes, derived from a simulated orbital run through Mercury’s exosphere, as inputs and predicts the chemical elements of the surface regolith below. It serves as an estimator for the surface-exosphere interaction and the processes leading to exosphere formation, including micrometeoroid impact vaporization, ion sputtering, photon-stimulated desorption, and thermal desorption. Extensive training and testing campaigns demonstrate the MLP DNN’s ability to accurately predict and reconstruct surface composition maps from simulated exospheric measurements. These results not only affirm the algorithm’s robustness but also illuminate its extensive capabilities in handling complex data sets for the creation of estimators for modeled exospheric generation. Furthermore, the tests reveal substantial potential for further development, suggesting that this method could significantly enhance the analysis of complex surface-exosphere interactions and reduce uncertainties in planetary exospheres models. This work anticipates the analysis of data from the SERENA (Search for Exospheric Refilling and Emitted Natural Abundances) instrument package aboard the BepiColombo Mercury Planetary Orbiter, with its nominal phase starting in 2026.

Solar flares are among the most intense solar phenomena, and precise forecasting of solar flares holds significant importance in space weather research. The ability to predict solar flares is crucial for protecting the near-Earth space environment and preserving the integrity of our technological infrastructure from potential detrimental consequences. Over the last two decades, machine learning and deep learning models have made a significant impact on the prediction of solar flares, leveraging their capacity to learn from high-dimensional data spaces. However, the scarcity of high-quality data from the field of solar flare prediction becomes a daunting challenge for such tasks. One of the methods to tackle the scarcity of high-quality data is to generate synthetic samples (i.e., data augmentation). In this study, we aim to explore the role of data augmentation on time series-based flare prediction models, namely, deep learning-based methods. We utilize the latest time series-based benchmark dataset extracted from the vector magnetograms of Solar Dynamics Observatory’s Helioseismic and Magnetic Imager (SDO/HMI). Specifically, we use seven-time series data augmentation techniques to enrich our dataset and train three machine learning models for multivariate time series classification. To our knowledge, this is the first research effort that attempts to explore data augmentation’s impact on the solar flare prediction problem.

Bayesian inference methods are widespread in geophysics and over the decades have been extensively applied to inverse problems, where they are particularly represented through Monte Carlo methods. These are popular as they allow to easily quantify uncertainties and parameter trade-offs but on the downside usually require large computational effort. Here, we propose to use methods from generative deep learning, in particular Normalizing Flows, as global proposal distribution in a Markov chain Monte Carlo context. A simple synthetic example demonstrates that a Normalizing Flow proposal can provide samples with high efficiency and low inter-dependence even in higher dimensions. In a waveform inversion example, it is shown that Normalizing Flows can significantly improve efficiency in sampling complicated posterior distributions. This effect can be increased by adaptive sampling, i.e.\ further refining the Normalizing Flow on its own outputs. The methods presented in this study are thought to be the first application of deep learning-based proposal distributions in geophysics, and contribute to the development of deep-learning enhanced Monte Carlo methods for geophysical inverse problems.

We measured the acoustic properties of ice-bearing sand packs in the laboratory using a novel acoustic pulse tube within the frequency range of 1-20 kHz, similar to sonic well-logs. We analysed how wave velocity and attenuation (the inverse of quality factor) change with ice saturation during melting. We found strong frequency-dependent correlations for both acoustic parameters with ice saturation. For any frequency within the studied range, velocity decreases and attenuation increases as the ice melts. We used two-phase and three-phase rock physics models to assess our experimental results, and the comparison highlights the influence of ice formation location, sediment frame permeability, and gas content on both velocity and attenuation. Our results pave the way for monitoring ice saturation from sonic measurements as ice saturation has contrasting effects on velocity and attenuation and the effects vary with frequency. Overall, this research contributes to a better understanding of the acoustic response of ice-bearing sediments and provides valuable insights for various applications, including permafrost monitoring and gas hydrate dissociation studies.

The Extremely Low Frequency band (ELF: 0.03 – 1000 Hz) electromagnetic signals from thunderstorm lightning discharges can propagate around the globe in the Earth-ionosphere resonance cavity and thus be used for ionosphere monitoring. We use ELF observations of the World Wide Lightning Location Network (WWLLN) impulses to examine ELF propagation velocity and arrival azimuth under diurnal changes in two days of September 20th and 21st, 2023. Also temporary effects of solar flares’ ionizing fluxes are monitored, leading to increasing the ELF signal propagation speed modulated by the X-ray flux intensity. We present for the first time a simple method for automatic and large-scale analysis, utilizing data from two registration systems (ELF and WWLLN) and enabling easy evaluation of changes in wave propagation speed. The compared samples of WWLLN impulses generated in selected azimuth and distance sectors for Africa and America reveal varying effects of signal refraction, with increased azimuth changes for signals propagating across the ionospheric ionization gradients. The method has a potential to become a standard tool for the analysis and monitoring of the lower layers of the ionosphere.

Global, regional and local sea-level projections rely on complex process-based models of the climate-ocean-cryosphere system. While extrapolation of observational data has been examined on global and regional scales, this approach has not yet been used for the additional complexities of local coastal sea-level projections. Here, we evaluate the sea level trend and acceleration for a global network of tide-gauge observations over 1970-2023, which are then extrapolated to provide local projections up to 2050 and compared with the process-based projections from the IPCC Sixth Assessment Report (AR6). For 2050 relative to 2020, the observation-based projections agree with AR6 process-based projections within the 90% uncertainty range at the majority (99%) of 237 tide gauges. Thus, the observation-based projections provide complementary perspectives of near-term local sea-level changes, and this agreement provides increased confidence in the current understanding and projections of sea-level changes over coming decades.

Geologically Europe is an ensemble of Phanerozoic and Precambrian terranes. Our work aims to understand the structure of the former collision zones of these terranes by imaging seismic discontinuities in the upper mantle. We used an improved version of the S-receiver function method without deconvolution and obtained high quality profiles through the study area. Below the Bohemian Massif, the southeastern Carpathians and Pannonian Basin and below the northern Apennines we found evidence for inclined velocity discontinuities below the lithosphereasthenosphere boundary reaching down to about 180 km depth. This might be interpreted as fossil subduction or delamination in collision zones in the Variscan area or ongoing subduction in the Alpine area. However, in the Vrancea seismic zone in the southern Carpathians this newly observed discontinuity does not agree with the near vertical distribution of the intermediate depth earthquake sources. A model of the structure below the Pannonian Basin and the Vrancea seismic zone might need to contain both, subduction and delamination.

Wildfires have become a major ecological issue impacting humans and the environment alike. To prevent wildfires from starting in the first place, better spatial prediction tools are needed to assess wildfire risk between locations to allow better specialization of firefighting agencies in performing preventative measures. Current methods of predicting spatial wildfire risk rely on satellite imagery, which is significantly limited in terms of spatial and temporal resolution. The use of unmanned aerial vehicles (UAVs) for assessing wildfire risk has significant advantages over satellites due to their much higher spatial and temporal resolution. This project sought to develop a novel deep learning model to predict wildfire risk through aerial imaging specialized for deployment on UAVs. Convolutional neural network architectures were investigated due to their ability to extract implicit spatial features from data. Google Earth Engine was used to assemble a geospatial dataset containing patches of vegetation burnability indices each mapped to a wildfire risk index based on historical fire prevalence. The results for the model show extremely high performance across all metrics evaluated. From this proof of concept research, UAV-based prediction offers a promising prospective solution to the drawbacks of satellite imaging in remote sensing of wildfire risk.

Venus’s thick atmosphere rotates in the same direction as the solid body, but ~60 times faster. This atmospheric superrotation has produced dozens of windblown ejecta deposits (“parabolas”) on the surface of Venus. The formation and modification of parabolas is an interplay between impacts, aeolian erosion, and atmospheric dynamics. We conducted a survey to explore the nature of these sedimentary surface features and obtained three new results. Firstly, we observe trends in parabolas’ morphology that shed light on how they are deposited and eroded. Changes in the size and radar albedo of parabolas are likely linked to the height and density (respectively) of ejecta plumes at time of formation. Next, we discovered that parabolas show orientations inconsistent with present atmospheric dynamics, and thus may record a change in these dynamics or geologically recent true polar wander at a rate of ~1° Myr-1. Lastly, we observe a definitive example of a volcanically embayed parabola, which provides strong evidence for large-scale geologic activity in the recent past. These results provide important insights into the history and character of geologic processes on Venus that will guide observations by upcoming missions. If true polar wander is the cause of the anomalous parabola orientations, then atmospheric superrotation has probably persisted for at least the age of Venus’s surface. This mode of atmospheric circulation is unique to Venus in the inner Solar System, but is likely common to terrestrial exoplanets that are tidally locked in close orbits to their parent stars.

Water scarcity represents a critical global challenge, which is driven by diverse complex interactions between natural and anthropogenic factors. Long-term water scarcity often results in depletion of water resources in so-called water scarcity hotspots. To understand the interactions among social, ecological and hydrological components within water scarce systems at such hotspots, we applied causal discovery to observational time series of socio-economic, meteorological, and ecological variables. This resulted in a network representing the causal relations between these variables and Terrestrial Water Storage (TWS). Recognizing the limitations of causal discovery, we supplemented the network with expert knowledge. From this we derived Structural Causal Models (SCMs) that simulate the causal mechanisms influencing TWS trends at the water scarcity hotspots. The resulting SCMs have a variable performance with a median r^2 of 0.52 compared to TWS observations. The SCMs allowed us to estimate the impact of anthropogenic and natural changes on TWS variability at water scarcity hotspots. Our analysis identified population dynamics as the most significant cause of TWS change in hotspots. As such, this study demonstrates how causal discovery and SCMs can enhance modelling of human-water system dynamics affected by water scarcity, improving the understanding of these systems and potential impacts of future changes on water storage and availability. For future research, more detailed data on human-water use is needed to improve the robustness of these models. This is essential for developing effective water management strategies to mitigate water scarcity at hotspots.