River discharge is a prerequisite for an understanding of flood hazard and water resource management, yet we have poor knowledge of it, especially over remote basins. Previous studies have successfully used a classic hydraulic geometry, at-many-stations hydraulic geometry (AMHG), and Manning's equation to estimate the river discharge. Theoretical bases of these empirical methods were introduced by Leopold and Maddock (1953) and Manning (1889), and those have been long used in the field of hydrology, water resources, and geomorphology. However, the methods to estimate the river discharge from remotely sensed data essentially require bathymetric information of the river or are not applicable to braided rivers. Furthermore, the methods used in the previous studies adopted assumptions of river conditions to be steady and uniform. Consequently, those methods have limitations in estimating the river discharge in complex and unsteady flow in nature. In this study, we developed a novel approach to estimating river discharges by applying the weak learner method (here termed WLQ), which is one of the ensemble methods using multiple classifiers, to the remotely sensed measurements of water levels from Envisat altimetry, effective river widths from PALSAR images, and multi-temporal surface water slopes over a part of the mainstem Congo. Compared with the methods used in the previous studies, the root mean square error (RMSE) decreased from 5,089 m3s-1 to 3,701 m3s-1, and the relative RMSE (RRMSE) improved from 12% to 8%. It is expected that our method can provide improved estimates of river discharges in complex and unsteady flow conditions based on the data-driven prediction model by machine learning (i.e. WLQ), even when the bathymetric data is not available or in case of the braided rivers. Moreover, it is also expected that the WLQ can be applied to the measurements of river levels, slopes and widths from the future Surface Water Ocean Topography (SWOT) mission to be launched in 2021.

Solving the problems presented by climate change depends on geospatial technology in great measure. Through shifting pest and disease dynamics, increasing frequency of extreme weather events, and modifying growth circumstances, climate change aggravates problems with plant health. For example, remote sensing data and satellite-based climate models help one to forecast how variations in temperature and precipitation patterns can influence the distribution of plant diseases and pests. This predicting makes proactive adaption tactics possible (Beddington et al., 2012). Geospatial instruments also allow one to track crop performance under the influence of climateinduced stresses including heat waves and drenches. This tracking helps to create strong farming methods. Geospatial technology helps stakeholders to carry focused actions that preserve plant health and guarantee the stability of food systems in a changing environment by offering thorough understanding of how climate interacts with plant life. This capacity is particularly important in sensitive areas where smallholder farmers depend on agriculture for their food security and means of income.

Gullies are actively changing landforms on planet Mars. The prevailing hypothesis, supported by a suite of different studies, states that present-day activity in these gullies is caused by fluidised granular flows driven by the sublimation of seasonal CO2 ice. However, the long-term formation process of gully landscapes is a contentious issue as water-driven debris-flow processes could easily explain erosion. In contrast, we do not know if CO2-driven granular flows can cause a significant amount of erosion. In this study, we conducted flume experiments investigating the flow dynamics and erosion capacity of CO2-driven granular flows under different substrate and flow settings. Our experiments show that CO2-driven granular flows under Martian conditions are efficient erosive agents, which can erode and entrain large volumes of unconsolidated material in various environmental (i.e. substrate and flow) settings. In general, erosion and entrainment enhance the mobility of CO2-driven flows. However, the frost and thermal conditions of the slopes and the flow composition determine the erosion efficiency of these flows. Finally, based on terrestrial debris-flow erosion theory we estimate that collisional forces at the base of CO2-driven flows can cause erosion of the underlying bedrock or permafrost.

Maud Rise, is a seamount in the Southern Weddell Sea where open-ocean polynyas occasionally occur. Polynyas are large openings in sea ice that form in winter in association with deep ocean convection. In winter, freezing waters are found in the mixed layer and warmer deep waters lie below the pycnocline. Destratification of the water column enables the mixing of heat into the mixed layer and is required for the formation and sustenance of polynyas. Open-ocean polynyas are rare formations that can disrupt the conventional abyssal overturning circulation pathways and affect the heat and carbon budget of the region. Maud Rise lies in the path of the eastern limb of the Weddell Gyre, which is the main source of salt and heat in this region. We use satellite-based altimetry and ocean state estimates to study how variability in the eastern Weddell Gyre inflow impacted stratification around Maud Rise via the transport of salt anomalies. Salinity predominantly controls stratification in polar waters. We find that reduction in the eastern inflow induced sub-surface freshening in the years leading to the most recent polynya in 2017. Remotely advected sub-surface salinity anomalies control the region's stratification on interannual time scales, and thus influence how much heat and salt is stored below the pycnocline. The inflowing limb of the eastern Weddell Gyre responds to tropical Pacific and Southern Ocean climate variability via atmospheric teleconnections.

In this study, we developed and implemented a deep learning program to classify the suitability of regions for landing on the Moon's south pole, utilizing data from NASA's Lunar Reconnaissance Orbiter (LRO), launched in 2009. The LRO mission, which included the creation of a three-dimensional map of the Moon using the Lunar Orbiter Laser Altimeter (LOLA) instrument, and high-resolution images from the Lunar Reconnaissance Orbiter Camera (LROC), enabled detailed mapping of lunar topography, with a particular focus on craters in the south polar region. Images with a resolution of 1 - 20 meters per pixel were analyzed and characterized using a Convolutional Neural Network (CNN) to identify terrain features that could indicate safe landing sites. The binary analysis classified the areas as either suitable or unsuitable for landing, considering risks such as explosion or overturning due to the terrain. This approach enhances the identification of secure landing sites, supporting future lunar missions like Artemis 3.

Terrestrial ecosystems mitigate CO2 accumulation in the atmosphere by annually absorbing ~30% of anthropogenic emissions. The degrees to which CO2 and climate drive this absorption are uncertain, which presents a challenge for future planning around carbon mitigation scenarios. To reduce this knowledge gap, we use a Bayesian model–data integration framework (CARDAMOM) to build and analyze a global terrestrial biosphere reanalysis that optimally reconciles multiple lines of Earth Observations with mechanistic model processes. The Earth observations informing the model dynamics include satellite- and inventory-based constraints on distributions and change in terrestrial C storage (e.g., live biomass, soil organic C, and net biosphere exchange of CO2) and mechanisms of that change (e.g., photosynthesis, deforestation, water storage anomalies, or fire). We find that the impact of 2001–2021’s atmospheric CO2 increase on terrestrial C storage (+38 PgC) opposes and far outweighs the impact of climate trends over over the same period ($-$8.2 PgC). Globally, CO2-induced carbon gains occurred primarily in living biomass pools, while climate-induced losses occurred primarily in dead organic C pools, but relative gains and losses vary regionally. The fact that live biomass and dead organic C exhibit distinct and often opposing responses indicates that mechanistically resolving ecosystem function underlying the terrestrial C cycle’s emergent behavior is crucial for estimating the strength and resilience of the land C sink over the coming decades.

We develop a novel methodology for operating a Distributed Acoustic Sensing (DAS) technology at laboratory scale. Fiber Optic (FO) cables can be used for sensing seismic waves at unprecedented spatial resolution ($<1$ meter). Sensing seismic waves in laboratory experiments is fundamental for a better understanding of the phemomena reproduced at scale. However, installing seismic sensors in laboratory experiments could be difficult, due to the limited space available. Here, we installed a DAS interrogator connetected to a fiber optic (FO) cable for investigating seismic waves generated by seismic sources in a $200 \times 73$ centimeters sand-box, generally used for landslide analogue models. With such apparatus, we were able to sense seismic waves traveling in the sand, which could be potentially generated by, e.g., precursor phenomena of landslides. Our results demonstrate the capability of DAS technology in recording seismic waves at such laboratory scale. Our novel methodology, based on a probabilistic Bayesian approach, gives us fully consistent results in locating man-operated seismic events occurring in the sand-box. This study opens to the possibility of using DAS technology to monitor large-scale (1-10m) laboratory analogue experiments aimed to reproduce geophysical and tectonic processes.

Sulfates are abundant at the Martian surface, incorporating sulfur delivered by volcanic eruptions and degassing. The thickest sulfate accumulations are found in the interior layered deposits (ILDs) of Valles Marineris. The ILDs have been interpreted as either purely sedimentary deposits or weathered volcanic products. We performed nonlinear spectral unmixing of CRISM data located in the lower, massive member of the ILD pile in Ophir Chasma. The best spectral fit is obtained when primary igneous minerals (orthopyroxene, plagioclase) and coquimbite were present, alongside kieserite and szomolnokite, which were identified in previous ILD studies. Geomorphological evidence shows that the igneous minerals are from a source within the ILDs. Our findings suggest that the deposition of the lower member of the ILDs was influenced by Tharsis-related syntectonic and synvolcanic activity within the subsided Valles Marineris plateau, and by the redox state of an overlying Valles Marineris sea in a geologic context akin to volcanogenic massive sulfide deposition on Earth. Subsequent climate cooling, potentially related to Tharsis activity waning, would have led to gradual sea freezing, resulting in further alteration in acid-cold environment, and precipitation of the polyhydrated sulfates that compose the upper, layered ILD member. ILD erosion by glacial flows down to the currently exposed chasma floor level and aeolian erosion have shaped the current ILD morphology. In summary, the comprehensive ILD pile formed by in-situ weathering of volcanic products, redox state instabilities, and water level fluctuations in a warming and cooling Valles Marineris sea.

Foreshock, though well-documented phenomena preceding many large earthquakes, have limited forecasting utility due to their non-pervasive occurrence and non-distinctive characteristics. Using California as an example, we investigate how seismic monitoring capability, particularly the completeness magnitude (Mc), influences the inferred proportion of mainshocks with foreshocks (Pf). We test four foreshock identification methods, namely the fixed-window, nearest neighbor clustering, empirical statistical methods and the epidemic-type aftershock sequence (ETAS) model. The fixed-window method shows Pf decreasing with higher Mc due to misclassifying background events as foreshocks. In contrast, clustering and empirical statistical methods yield relatively stable Pf across different Mc values. The ETAS model finds more foreshocks in California are associated with aseismic driving processes at low Mc, while its capacity to distinguish between aseismically-driven and cascade-driven foreshocks diminishes at high Mc. These results show that improved seismic monitoring capability does not significantly increase Pf but is crucial for distinguishing processes driving foreshocks.

Tropical cyclone (TC) rapid intensification (RI) is a major source of uncertainty in TC prediction. Here we examine observed basin-specific relationships between RI and El Niño-Southern Oscillation (ENSO), where RI is defined as a TC strengthening by >=30 kt within 24 hr. During El Niño, there is a significant increase in RI in the eastern North Pacific, western North Pacific and South Pacific, with the opposite behavior in the North Atlantic. During La Niña, RI has an approximately opposite relationship in the Atlantic and Pacific. The relationship between ENSO and Indian Ocean RI is weak. These changes are closely linked to environmental conditions modulating RI, including mid-level moisture, vertical wind shear and sea surface temperatures. Because there is disagreement between the recently-observed La Nina-like trend and an El Niño-like trend simulated by climate models, anticipating future RI trends is conditional on improved model representation and physical understanding of ENSO.

We investigated tsunamis of an Mw 7.1 earthquake in the Hyuganada Sea in 2024, observed by a new dense and wide seafloor observation network, N-net, installed in the western Nankai Trough. A joint inversion of the offshore tsunami and onshore GNSS data revealed a maximum slip of 2.4 m with a significant slip near the centroid from the teleseismic analysis. The joint inversion provided reliable constraints for both up-dip and down-dip extents of the fault, while the inversions using either dataset showed limitations in the fault constraint. Comparisons with past earthquakes indicate the 2024 earthquake ruptured part of the asperity of the 1961 earthquake but not those of the 1996 earthquakes. Our fault modeling jointly using offshore and onshore data suggests the interplate seismic coupling ratio in this region is < 0.4, which was much smaller than those in the anticipated megathrust earthquake source region in the Nankai Trough.

We develop a novel three-dimensional viscoelastic modeling method using an implicit finite difference scheme, while the properties of sparse and block matrices are utilized to make the solution of the method possible. The advantage of this method is that the introduction of implicit format makes it alleviate the numerical dispersion problem with the same accuracy and has better stability. In order to verify the validity and accuracy of the method, this paper elaborates its basic principles and analyzes the results through a series of forward simulations of simple models. In addition, we also gives an outlook on the direction of the subsequent refinement of the 3D viscoelastic implicit finite difference method as well as the potential application prospects.

Recent advances in machine learning (ML) techniques show promise for estimating soil hydraulic properties from soil datasets. Pedo-transfer functions (PTFs) can facilitate the mapping of the complex relationship between soil properties and soil hydraulic properties, e.g., lateral hydraulic conductivity—a necessity for estimating lateral subsurface flow in distributed hydrological models. In wflow_sbm model, the horizontal-to-vertical saturated hydraulic conductivity ratio (fKh0) is a sensitive parameter, but no established PTF exists. Our objective is to investigate the potential of ML algorithms in estimating PTFs for fKh0 prediction. In this study, publicly available calibrated fKh0 (i.e., optimized) across Great Britain were utilized to train two machine learning algorithms: Random Forest (RF) and Boosted Regression Trees (BRT), employing SoilGrids dataset. Both algorithms effectively predicted fKh0 of 92 sub-basins (out of 115 sub-basins in the 25% test set), demonstrating a high correlation with the optimized values, with RF slightly outperforming BRT. As a next step, we compared wflow_sbm simulated discharge results using uncalibrated fKh0 (default value) and our predicted values. The predictions notably improved discharge simulations, with a median KGE increasing from 0.55 to 0.75. Subsequently, we generated two globally distributed fKh0 maps to investigate the transferability of the ML-based PTFs in the Loire basin, France. ML-based PTFs improved performance in 75% of sub-basins, with an average KGE increase of 0.06. Finally, we assessed the uncertainty in fKh0 predictions, confirming the robustness of the ML-based PTFs. Our study highlights the potential of ML methods for estimating soil hydraulic properties, aiding parameter estimation for distributed hydrological models.

Monin-Obukhov similarity theory (MOST) is widely used in numerical weather prediction to model surface fluxes of momentum, heat, and water vapor. However, MOST is based on assumptions of steady state and horizontally homogeneous turbulence that can lead to prediction errors in and around convective storms. To understand the nature of these errors, we used wind and eddy covariance flux measurements from the Atmospheric Radiation Measurement (ARM) Southern Great Plains Atmospheric Observatory to evaluate MOST predictions in fair-weather and convective storm environments, specifically those of mesoscale convective systems and air mass thunderstorms. Predicted wind profiles agreed well with observations in fair-weather cases, while in convective storm cases, MOST systematically overestimated shear in cold pools, after gust front passage. Surface layer stability was found to be important in assessing MOST within convective storm environments. The overestimation of wind shear in cold pools suggests the role of non-local fluxes in transferring momentum downward. We discuss reasons for discrepancy and agreement with past studies, and conclude with recommendations to improve prediction of surface winds and fluxes in convective storm simulations.

Model biases in tropical precipitation have persisted for nearly 30 years. These biases are highly sensitive to the region and/or season of interest, with commonalities in the east Pacific and Atlantic Ocean basins, possibly due to their similar observed climatological northern hemisphere intertropical convergence zone (ITCZ). However, the colloquial name ”double ITCZ bias” comes from a time and/or zonal mean, and many studies do not acknowledge that these model biases are most prominent during boreal winter and spring when models overestimate a southern hemisphere ITCZ, not a double ITCZ. In this study, we explore high-resolution characteristics of the ITCZ in observations, reanalyses, and 25 Coupled Model Intercomparison Project 6 (CMIP6) models over the east Pacific Ocean. We devise and apply an algorithm that determines a region’s dominant daily ITCZ configuration, its ”ITCZ state,” based on the daily mean precipitation field. The five ITCZ states include: northern hemisphere (nITCZ), southern hemisphere (sITCZ), double (dITCZ), equatorial (eITCZ), and absent (aITCZ). We find that nearly all CMIP6 models gravely underestimate dITCZs and nITCZs and overestimate sITCZs during January through May, in contrast with what ”double ITCZ bias” suggests. Surprisingly, all reanalyses also underestimate dITCZs and overestimate sITCZs. Errors in ITCZ state interannual variability are consistent with mean errors in reanalyses, while sITCZ interannual variability is far too low relative to the mean in most CMIP6 models. Lastly, all reanalyses and the seven CMIP6 models that can produce dITCZs overestimate precipitation rates in dITCZs, while nearly all CMIP6 models overestimate precipitation rates in sITCZs.

The eastern tropical Pacific (ETP) Ocean is projected to warm faster than the Atlantic or Indian Oceans in the 21st century, yet this prediction is highly uncertain due to model-observation discrepancies. The potential impacts of this uncertainty on regional terrestrial hydroclimates are largely unknown, which is problematic for climate risk assessments. To address this, we designed novel atmospheric model experiments simulating future global warming with and without enhanced ETP warming, superimposed upon an idealized El Niño-Southern Oscillation (ENSO) cycle. Our results show that enhanced ETP warming significantly influences future terrestrial hydroclimates in several regions across the Americas. In southern Mexico, Central America and the Amazon region, enhanced ETP warming exacerbates long term drought trends and extreme drought events, while the opposite is true in south-central South America. Along the west coast of the continental United States, the effects of enhanced ETP warming manifest as extreme precipitation anomalies. These findings illustrate how climate impact projections may be misrepresented in conventional multi-model analysis, which does reflect true uncertainty of the future tropical Pacific warming pattern.

The characterization of pore-throat structures in tight sandstones is crucial for understanding fluid flow in hydrocarbon reservoirs and groundwater systems. Both thin-section and Mercury Intrusion Capillary Pressure (MICP) offer insights rock petrophysical parameters. However, thin-section analysis is limited by its 2D nature and subjective interpretation, while MICP provides 3D pore-throat distributions, it lacks direct visualization of pore morphology. This study evaluates AI-assisted thin-section image analysis for pore-throat characterization by comparing its results to MICP-derived measurements. A machine learning-based workflow was developed using color thresholding, K-Means clustering, and medial axis transformation to segment pore structures in thin-section images. Throat width, porosity, and permeability were quantitatively assessed against MICP to determine the accuracy and reliability of the technique. The analysis of 26 sandstone samples outlined differences between the two methods. Thinsection analysis showed porosity values from 1.37% to 53.37%, with average pore-throat sizes between 5.63 µm and 30.09 µm, while permeability estimates ranged from 0.01 mD to 344.35 mD. Correlation analysis showed moderate agreement for throat size (r=0.62) and permeability (r=0.61), but weaker for porosity (r=0.32), highlighting the differences in how each method captures pore connectivity. Results demonstrate that the AI-assisted segmentation provides a scalable and reproducible approach but is constrained by thin-section imaging resolution. While MICP remains reliable for permeability evaluation, its comparison with AI-driven image analysis helps assess the reliability of the method. Future research should refine segmentation algorithms, incorporate pretrained data to validate AIderived pore-throat attributes for improved reservoir quality assessment and predictive modeling.

• Arctic phytoplankton net primary production can be predicted for multiple years, largely due to the Arctic shelves. • This predictability is driven by ocean temperatures, which are an important limit on phytoplankton growth and maintain predictability. • Phytoplankton net primary production predictability in the Arctic may increase in the near future (2030s) relative to the 2010s.