A document by Freya M E Muir. Click on the document to view its contents.

A document by Oleksiy Dudnik. Click on the document to view its contents.

Segmentation of soil constituents in X-ray CT imagery is critical for advancing our understanding of soil structure dynamics, but challenges arise due to the overlapping X-ray attenuation. In this study, we explore the potential of nnUNet, a deep learning-based semantic segmentation model, for applications in soil science. We evaluated nnUNet on three challenging datasets: (1) heterogeneous soil structure with numerous material classes sharing gray value ranges, (2) fine roots in noisy images, and (3) permafrost core with gradual gray value transitions between sediment types. The performance of nnUNet was compared to other reference methods, namely ilastik, Rootine v.2 and manual thresholding. For Dataset 1, the segmentation accuracy surpassed that of ilastik for the particulate organic matter class. Dataset-specific challenges, such as annotating fine structures (Dataset 2) or poorly defined boundaries between material classes (Dataset 3), impacted the performance of nnUNet, compared to more specialized methods. The mean validation Dice scores were 0.71, 0.90, 0.92, for Dataset 1, 2 and 3, respectively, which suggested overall good model performance. Our study underscores that deep learning models like nnUNet perform well for the segmentation of complex soil structures and could assist the development of a generalized segmentation model, thereby fostering standardization in soil structure analysis.

Climate change reduces ocean oxygen levels, posing a serious threat to marine ecosystems and their benefits to society. State-of-the-art Earth System Models (ESMs) project an intensification of global oxygen loss in the future, but poorly constrain its patterns and magnitude, with contradictory oxygen gain or loss projected in tropical oceans. We introduce an oxygen water mass framework– grouping waters with similar oxygen concentrations from lowest to highest levels– and separate oxygen changes into two components: the transformation of oxygen in water masses by biological, chemical or physical processes along their pathways in ’ventilation-space’, and the redistribution of these water masses in ’geographic-space’. The redistribution of water masses explains the large projection uncertainties in the tropics. ESMs with more realistic representations of water masses provide tighter constraints on future redistribution than less skilled ESMs, leading to over a third more of tropical area exhibiting consistent oxygen projections (58% vs 22%), and a 30% reduction in model spread for tropical oxygen projections. These higher-skilled ESMs also project weaker global deoxygenation than less skilled models (median of -6.5 vs -9.5 Pmol O2 per °C of surface warming) controlled by an increase in global water residence times, and they project a stronger increase in oxygen minimum zone ventilation by ocean mixing. These tighter constraints on future oxygen changes are critical to anticipate and mitigate impacts for ecosystems, and inform management and conservation strategies of marine resources.

An idealized aquaplanet moist general circulation model (GCM) with realistic clear-sky radiative transfer and no convective parameterization possesses multiple climate equilibria. When forced symmetrically about the equator, the intertropical convergence zone (ITCZ) can migrate to an off-equatorial equilibrium position, given a slab ocean boundary condition with sufficient heat capacity, as a result of a destabilizing radiation-circulation coupling. Here, we present GCM simulations with this model with imposed hemispheric asymmetry in the boundary conditions. The simulations with imposed hemispheric asymmetry are used to determine whether the ITCZ always resides in the higher-energy hemisphere or whether multiple equilibria persist. We find that multiple equilibria can persist for hemispherically asymmetric boundary conditions of non-trivial magnitude. Starting from a control simulation that exhibits the ITCZ destabilization, energetic arguments for the response to asymmetric forcing provide guidance about the forcing magnitude required to shift the ITCZ into the opposite hemisphere.

A document by Takashi Miyamoto. Click on the document to view its contents.

NASA’s Orbiting Carbon Observatory-2 (OCO-2) has the goal of accurately measuring column-averaged dry-air mole fractions of carbon dioxide (XCO2). In order to fit the measured radiances, many parameters besides CO2 are included in the optimal estimation state vector, including atmospheric water vapor and temperature. The current operational XCO2 retrieval algorithm (V11) solves for a multiplicative scaling factor on an a priori water vapor profile and an additive offset on an a priori temperature profile. However, simulations have indicated that water vapor and temperature each have 1.5-3 degrees of freedom in the vertical column. This means that the retrieval is limited in its ability to fit the true profiles of temperature and water vapor. Here, we use singular value decomposition (SVD) to determine the three most explanatory profile ”shapes” of water vapor and temperature error, then retrieve a single scaling factor applied to each shape. We assess retrieval errors by comparing to the Total Carbon Column Observing Network (TCCON) and a collection of CO2 models. We find that after applying quality filtering using Data Ordering Genetic Optimization (DOGO) and a custom bias correction, the scatter of the XCO2 error versus TCCON is reduced from 1.02 to 1.01 ppm (2.3% reduction in variance) for land glint observations, 1.04 to 0.96 ppm (14.5% reduction in variance) for land nadir observations, and 0.68 to 0.66 ppm (4.7% reduction in variance) for ocean glint observations. We also see a small improvement in the agreement between OCO-2 XCO2 and CO2 models over oceans and the Amazon.

Large-Scale Travelling Ionospheric Disturbances (LSTIDs) are wave-like ionospheric fluctuations, often triggered by geomagnetic storms, which play a critical role in Space Weather dynamics. In this work, we present a machine learning model able to forecast the occurrence of LSTIDs over the European continent up to three hours in advance. The model, based on CatBoost-a gradient boosting framework-and trained on a humanvalidated LSTID catalogue, uses a diverse set of physical drivers, ranging from geomagnetic indices, and solar wind and activity data, to ionosonde measurements. It results in three distinct operating modes, tailored for scenarios with varying relative costs of false positives and false negatives. In high-risk settings it is crucial to enable researchers and decisionmakers to understand and trust the predictions made by the model. In our case, explainability is ensured through the SHAP formalism, a game-theoretic approach to explaining model output. The validation phase-involving a global evaluation and interpretation step, followed by an event-level validation against independent detection methods-highlights the model's predictive robustness, suggesting its interesting potential for real-time Space Weather forecasting. Depending on the operating mode, we report an improvement ranging from +72% to +93% over the performance of a rule-based benchmark. Our study concludes with a comprehensive analysis of future research directions and actions to be taken towards full operability. We discuss probabilistic forecasting approaches from a cost-sensitive learning perspective, along with performance-centric model monitoring. Finally, through the lens of the conformal prediction framework, we comment on uncertainty quantification for end-user risk management and mitigation.

A document by Mario Arroyo-Solórzano. Click on the document to view its contents.

Marginal seas significantly impact the global carbon cycle. However, current knowledge on the role of marginal seas is limited, and only a few in situ datasets on air-sea CO2 exchange are available. This study presents the first direct measurements of air-sea CO2 flux carried out in the central Mediterranean region. Measurements used in this study were made at the Lampedusa Oceanographic Observatory (35.49°N, 12.47°E) and cover the period from December 2021 to June 2023. The daily air-sea flux is calculated based on high-resolution measurements of sea temperature, salinity, CO2 partial pressure in the ocean and in the atmosphere, and wind, using a wind-dependent parameterization for gas transfer velocity. The data show that the Central Mediterranean currently acts as a CO2 sink, with both absorption and emission phases over the course of a year. However, large differences exist between the two autumn-winter seasons comprised in the dataset, with a 30% lower air-sea flux during early 2023 than in early 2022. The possible impact on the CO2 flux of the intense and long-lasting marine heatwave which took place in the period from May 2022 to April 2023 has been investigated by analyzing the dependency of the calculated air-sea CO2 flux on SST and temperature. The reduced incidence of intense wind episodes during the heatwave period appears to be the main driver for the reduced air-sea CO2 exchange taking place during winter 2023, while the higher temperatures appear to produce a small effect on the calculated flux.

AI-based model emulators have emerged as a pragmatic strategy for calibrating Earth System models or their components (e.g., land, atmosphere, ocean), circumventing the previously insurmountable hurdle of the process-heavy models’ computational expense. Such emulators require large, spatially diverse datasets for training, however, which – in the land/hydrology context – contrasts with parameter estimation approaches that have traditionally emphasized optimizing model performance for individual basins, followed by similarity-based transfer schemes for parameter regionalization. Compared to calibrating basins individually, direct land/hydrology process model calibration approaches typically perform worse when trained jointly to large collections of basins. Building on insights from large-sample deep learning hydrologic modeling, this study introduces a Large-Sample Emulator (LSE) approach that unifies and streamlines process model parameter calibration and regionalization. Tested across 627 basins in the continental United States using the Community Terrestrial Systems Model (CTSM), the LSE approach consistently improves runoff predictions in all basins, outperforming the Single-Site Emulator (SSE) in both single-objective and multi-objective calibration tasks. Moreover, LSE-based regionalization in unseen basins, evaluated through spatial cross-validation, achieves better results than the default parameters in most cases. This LSE framework offers a promising strategy for effective large-domain process-based model calibration and regionalization.

Mode waters are critical for ocean ventilation and carbon sequestration. Using observations, we trace their subduction pathways and biogeochemical evolution. Solving modified mixing equations that account for respiration reveals that less than 50% of the oxygen changes along mode waters ventilation pathways are due to respiration within the water mass, the rest being due to mixing with oxygen-poorer surrounding waters. Consequently, measured changes in oxygen or Apparent Oxygen Utilization overestimate respiration by a factor of up to two, as do derived biogeochemical quantities such as carbon export. Measured nitrate changes either overestimate or underestimate remineralization depending on surrounding concentrations. Mean true respiration rates in mode waters range from -0.1 to -0.4 µmolkg−1yr−1. Southern Ocean mode waters export significantly more carbon (0.16 PgC) compared to other water masses (0.002-0.060 PgC), while subtropical mode waters exhibit the highest export rates, 1.5–9.5 mgCm-3yr-1, compared to less than 2.5 mgCm-3yr-1 in subpolar mode waters.

The future response of the Antarctic Slope Current (ASC) to projected transient changes in winds, radiative forcing, and meltwater input is examined in a high-resolution eddy-resolving ocean-sea ice model. Meltwater changes are found to dominate the almost-circumpolar strengthening of the ASC. The strength of the ASC increases non-linearly despite the linear meltwater perturbation applied, with a 9% increase in strength from 2000-2025 contrasting a 71% increase from 2025-2050. This non-linear increase is, however, not seen in the coastal currents on the continental shelf. The non-linear growth of ASC strength is attributed to meltwater accumulating on the continental shelf and poleward shifting warm waters north of the ASC, both triggered by reduced dense shelf water formation, and causing an accelerating ASC through thermal wind balance. Despite the strengthening ASC, any vulnerabilities in the current system could bring poleward shifted warmer waters onto the shelf, with implications for Antarctic ice shelf melt.

Terrestrial processes control land-to-atmosphere fluxes of water, energy, and carbon and are also influenced by climate. Feedbacks in the land-atmosphere coupled system can therefore potentially modulate changes in land surface water fluxes. Prior work has focused on one specific aspect of land-atmosphere coupling modulated by soil moisture, while largely ignoring other ways in which coupling between the land and atmosphere could alter climate. We use a novel experimental design of paired perturbed parameter ensembles in an Earth system model to isolate the extent to which land-atmosphere feedbacks modulate the global water cycle by perturbing land parameters spanning a wide range of terrestrial processes which affect evapotranspiration. We find support for the traditional soil moisture-precipitation coupling in some dry or transitional dry-to-wet regions where evapotranspiration is soil-moisture limited. However, we also find that land-atmosphere feedbacks have a dampening effect on evapotranspiration in wet regions. The effect in wet regions is more widespread and more robust than the impact in dry regions which is consistent with first principles but surprisingly has not been quantified in prior work. By adopting a more holistic definition of land-atmosphere coupling, our analysis provides insights into where and how land-atmosphere feedbacks modulate terrestrial processes, posing a challenge to the widespread practice of developing and evaluating land models in an uncoupled configuration and then deploying them to understand and predict terrestrial processes in a coupled context.

A document by Muhammad Umar Akbar. Click on the document to view its contents.

Flow instabilities and viscous fingering limit access to available subsurface pore space for carbon storage. Foams provide a pathway to mitigate these shortcomings and enhance storage capacity in subsurface systems by shifting the flow dynamics toward a stable flow, improving residual gas trapping, and reducing gravity segregation. Foam systems require careful consideration of various factors including surfactant formulation and  effective concentration to maximize foam stability. This work aims to optimize foam stability and carbon storage potential in the context of CCUS, by testing CO2-foams with different surfactant formulations and concentrations using a microfluidic platform. The microfluidic devices contain surrogate porous media that are representative of Berea sandstone and a Indiana limestone. The mediums are designed using X-ray tomography and porosity data from the core samples where pore connections are informed by throat size distribution data. The devices are fabricated utilizing an in-house photolithography method. Surfactants solutions are made using brine and three zwitterionic surfactant formulations with varying concentrations, ranging from below to above the critical micellar concentration (CMC). The solutions, along with CO2, are injected into the medium using a Surfactant Alternating Gas (SAG) scheme. Data is collected in the form of segmented high-resolution images of the medium during and after the injection process. Foam stability is evaluated based on changes in foam texture over time and bubble-size populations. The potential for carbon storage is evaluated based on residual fraction of trapped gas in the pore structure. Optimal surfactant formulation and concentration are identified for achieving a stable CO2 foam, a higher CO2residual saturation in the pore space for the lowest total cost.

Tree height estimation is crucial for accurate biomass estimation and forest monitoring. Synthetic Aperture Radar (SAR) offers two products useful for tree height estimation: SLC (Single Look Complex) images and derived tomographic cubes. In this work, we present machine learning methods to predict forest height in two ways, directly from SLC images and from tomographic cubes, providing valuable context to the upcoming European Space Agency’s Biomass Satellite mission.

Belharra is a wave energy hotspot located on the French Basque coast. Here, unusually large waves occasionally break over a 15-meter deep seamount, producing world-class conditions for big wave surfing. This does not occur often, even though the region frequently encounters highly energetic swells, especially during winter. This work attempts to better understand the underlying processes leading to extreme waves at this location. A combination of spectral and phase-resolving wave models enables analysis of how the incident swell conditions and tide level affect the wave field at both regional and local scales. We carry out a detailed wave-by-wave investigation of the local conditions for a wide range of plausible scenarios. Qualitative comparisons are drawn with data from the trajectory of a surfer riding waves at Belharra during a big swell event. Regional and local features in the bathymetry appear to govern the concentration of swell energy at this spot. Long-period swells from directions around 300° - 315° generate the most energetic local conditions. Lower tide levels lead to greater breaking intensities, albeit with weaker wave amplifications.