Landslides are a significant global geological hazard, with adverse and far for human life, the economy and the natural environment on an annual basis worldwide. Accurately estimating the spatial and temporal distribution of landslide probability is crucial for reducing these losses. Existing landslide warning systems may fail to consider the selection of non-landslide samples and the dynamic process of landslides, potentially compromising the accuracy of landslide warning systems. This study explores the impact of different selections of non-landslide samples and satellite rainfall datasets on the early warning model for landslides in the Guangdong-Hong Kong-Macao Greater Bay Area (GBA). Through Pearson correlation analysis, critical factors associated with landslide occurrences were identified, including elevation, slope, aspect, distance to roads and rivers, soil type, plan curvature, profile curvature, Topographic Wetness Index (TWI), and Normalized Difference Vegetation Index (NDVI). In this study, a semi-supervised random forest (SSRF) model incorporating frequency ratios (FR) to evaluate landslide susceptibility in the GBA. The susceptibility and rainfall threshold model were subsequently combined into a dynamic landslide hazard warning system through a matrix approach. The findings revealed that the maximum area under the curve (AUC) value for a landslide to non-landslide ratio of 1:4 is 0.973. The very high susceptibility zone is typically located between 125 and 250 meters away from roads, which is a common characteristic of the highly urbanized areas. Moreover, the validation phase yielded successful predictions for 67 out of 96 landslide events, providing effective early warning and a reference point for disaster mitigation and prevention.

Accurate prediction and monitoring of vadose zone soil water content (VZSWC) is important yet challenging for sustainable water resources management. While machine learning algorithms have shown promise, given limited and noisy subsurface data, they often struggle with overfitting, poor generalization, and physical inconsistency. In this study, we explore the effectiveness of physics-informed neural networks (PINNs) in addressing these challenges by presenting PINN-SM, a physics-informed neural network model tailored for predicting physically consistent VZSWC profiles by incorporating Richardson’s equation as the governing partial differential equation. Our model extends a traditional PINN architecture through inclusion of process-informed input variables and an attention mechanism that captures the time lag in the response of VZSWC to rainfall events. Using data from a monitoring station in Austin, Texas with soil moisture sensors at different depths, PINN-SM demonstrated superior predictive performance compared to the traditional PINN model, reducing RMSE by 80% across all depths. Furthermore, PINN-SM outperformed conventional artificial neural networks (ANNs), achieving approximately 25% lower RMSE across all depths. This performance advantage was particularly pronounced in peak event prediction, where PINN-SM achieved around 50% lower RMSE across all depths. Results indicate that compared to conventional ANNs, PINN-SM can more effectively capture underlying patterns while avoiding overfitting and excel at extreme events’ predictions. By integrating physical constraints with meteorological inputs and modeling temporal dependencies through attention mechanisms, PINN-SM represents a significant advancement in VZSWC prediction, with potential applications in improving subsurface hydrological monitoring and satellite-based forecasting systems.

Documenting and characterising past submarine landslides is fundamental to understanding their distribution and frequency through time, and critical to assessing the associated hazard. The widespread availability of marine geophysical data at the active Hikurangi subduction margin, east of Aotearoa New Zealand, provides an excellent basis to map regional trends in landslide occurrence. We present a database that documents mass transport deposits (MTDs) in 30 marine geophysical surveys, encompassing ~45,400 line-km of 2D seismic profiles. We map and characterise 737 MTDs, showing variations in size, location and style of failure, which we attribute to changes in geomorphic setting from north to south. MTDs in the northern Hikurangi margin, characterised by a high taper wedge and seamount subduction, show a broad range in size, with the highest proportion of MTDs displaying blocky or intact internal architecture. The central margin, characterised by lower wedge taper, hosts the most MTDs (51%), albeit with the thinnest (on average) and clustering within interridge basins. The southern Hikurangi margin hosts widespread submarine canyons and the largest (on average) MTDs, based on area and thickness. We demonstrate the importance of seismic archives in providing new insights into MTD preservation and discuss the bias between seafloor geomorphology and subseafloor seismic data in quantifying MTD occurrence. Our findings support the interrogation of the varied and complex causes of submarine landslides along active margins generally, as well as regions prone to cascading geohazards and landslide-induced tsunami.

Visible/infrared imagery from passive satellites is commonly relied upon to study low cloud microphysics over ocean, including for the Southern Ocean (SO), but relatively little validation has been undertaken for the SO. In this article, we compare low-cloud effective radius (re), cloud droplet number concentration (Nd) and cloud liquid water path (LWP) retrievals from the NASA Moderate Imaging Spectroradiometer (MODIS) with surface measurements collected during the Macquarie Island Cloud & Radiation Experiment (MICRE). MODIS 3.7-μm band retrievals show little bias and moderately good correlation relative to MICRE retrievals for liquid-phase low clouds when restricted to Solar Zenith Angles < 65o on spatial scales of 50 to 100 km. However, the low overall bias in 3.7-μm band effective radius (re 3.7) retrievals partly results from cancellation of errors: re 3.7 is overestimated for non-to-lightly precipitating clouds and underestimated for heavier drizzling clouds by ~1 to 1.5 μm. In contrast, 1.6-μm, and 2.1-μm band re retrievals are biased high. Nd may likewise be slightly under- or overestimated depending on the concentration, but there is insufficient data to provide confidence in this result. Interestingly, a composite of MODIS retrievals from 2002-2020 shows a distinct region of enhanced cloud cover and Nd (and lower re) in the wake of Macquarie Island associated with orographic cloud formation. MODIS retrievals of aerosol optical depth (AOD) and Angstrom Exponent (AE) upwind and downwind of the island do not differ significantly. Comparison with MICRE measurements suggests that MODIS Collection-6 AOD retrievals are reasonable, while AE is problematically large.

The Geospace Environment Modeling (GEM) program regularly issues ”community challenges” in which researchers examine a particular space physics phenomenon or geomagnetic activity event, often running numerical models to assess dominant processes and understand the timing and relationship of observed signatures. The GEM Methods and Validation (M&V) Resource Group helps those GEM leaders running challenges to maximize participation and optimize scientific return from the significant time investment of these endeavors. This article gives a brief history of GEM community challenges and details those best practices that lead to an inclusive and valuable experience.

Remote Sensing-derived Flood Inundation Maps (RS-FIM) are an attractive and commonly used source of evaluation benchmarks. Errors in model-predicted FIM (M-FIM) evaluation results due to biases in RS-FIM benchmarking are quantified by introducing a high-confidence benchmark FIM, which was manually delineated from ultra-resolution imagery, as Ground Truth. The evaluation results show considerable differences in M-FIM accuracy assessment when using lower-quality benchmarks. A RS-FIM enhancement (gap-filling) procedure is presented and its effect on FIM evaluation results is analyzed. The results show that the enhancement is insufficient for significantly improving the robustness of the evaluation. The impact of including/excluding Permanent Water Bodies (PWB) on FIM evaluation results is analyzed. The results show that including PWB in FIM evaluation can significantly inflate the model accuracy. A novel evaluation strategy is proposed and analyzed. The proposed evaluation strategy is based on excluding low-confidence grid cells and PWB from the M-FIM evaluation analysis. Low-confidence grid cells are those that were estimated to be flooded by the gap-filling procedure, but were not classified as such by the remote sensing analysis. The results show that the proposed evaluation strategy can dramatically improve the robustness of the evaluation, except when a considerable number of false positives exist in the RS-FIM. The analyses showcase the many challenges in FIM evaluation. We provide an in-depth discussion about the need for standards, user-centric evaluation, the use of secondary sources, and qualitative evaluation.

Climate change has intensified the frequency and severity of droughts, significantly impacting water resources, agriculture, and ecosystems. Traditional drought indicators typically focus on recent conditions rather than future projections, and conventional forecasting methods often struggle to capture the complex, non-linear relationships between long-term climate variables and droughts. This project aims to fill this gap by developing a machine-learning model to project drought conditions in Iowa, specifically focusing on the U.S. Drought Monitor categories. The developed model, a long short-term memory, was rigorously validated to ensure reliability and accuracy. With a Root Mean Squared Error of 0.19 and an R² Score of 91%, the model achieved a high level of accuracy, making it effective in guiding conservation practices and enabling timely interventions. The model was trained on historical data from 2012 to 2019 and thoroughly evaluated using out-of-sample data from 2002 to 2011. It exhibited strong performance in the projection of drought conditions across Iowa’s Hydrologic Unit Code 08 watersheds. The model predicted drought conditions for 2030–2050 using three climate models: MPI-ESM1-2-HR, BCC-CSM2-MR, and CNRM-ESM2-1. Results indicate that droughts in the coming decades will become more intense, prolonged, and frequent, with projections suggesting intensities up to twice as severe and durations and frequencies in northwestern regions up to nine times higher than historical records. Moreover, this research developed an interactive application for visualizing future drought conditions in Iowa. This tool aids users in making informed water management decisions by providing stakeholders with detailed visualizations and technical information.

Total Electron Content (TEC) is central to characterizing ionospheric response to solar and geomagnetic activity. Variations in TEC structures over time provide insight into underlying physical processes and inform monitoring of space weather events, which pose a risk to navigation and communication systems. JPL processed GNSS observations over 20 years provide a series of 15-minute Global Ionospheric Maps (GIMs) of spatial resolution 1x1 degree longitude/latitude. We translate these into geomagnetic coordinates centered about the sub-solar point and we isolate the top 1% of TEC values in each map to define High Density Regions (HDRs) of TEC. Image processing tools are used to develop an algorithm that detects and tracks these to compile a set of contiguous, uniquely labelled space-time TEC HDRs. We find that HDRs naturally divide into two populations by peak area, separated by a size of 8.0x106 km2, which is around the continental scale. These populations are studied for different storm conditions - quiet (Kp<4), moderate (4 ≤ Kp < 7) and extreme (Kp ≥ 7): small HDRs form primarily around four magnetic latitude bands and move roughly parallel to lines of constant magnetic latitude towards later MLT. Large HDRs form around the same latitude bands but follow more complex paths. The statistical nature of these results could be used in predictive ionospheric models and identify reproducible trends on these spatial/temporal scales.

Conventional Probable Maximum Precipitation (PMP) methods face several limitations, including the lack of statistical uncertainty characterization, subjectivity in storm maximization, and the assumption of a stationary climate. To address these limitations, we propose a nonstationary PMP approach that combines the novel stochastic spatiotemporal rainfall generation model StormLab with a nonstationary Generalized Extreme Value (GEV) model. We applied the new approach to the Upper Red River Basin in the south-central United States. StormLab provided 10,000 years of high-resolution (6-hour, 0.03°) precipitation fields from 1901 to 2100, based on 50 ensembles of a global climate model (GCM). A nonstationary GEV model was fitted to the simulated precipitation annual maxima, providing PMP estimates under different climate periods with an associated annual exceedance probability (AEP). The simulated precipitation was then integrated with a hydrologic model to generate annual peak discharge in major tributaries and to estimate the probable maximum flood (PMF). Our approach produces PMP estimates for areas ranging from 10-20,000 mi2 and durations from 6 to 360 hours. Results show a 16-25% increase in PMP with an AEP of 10-4 from 2020 to 2100 at different spatial and temporal scales. Higher increases of 35% and 37% are projected in PMF with the same AEP in two major tributaries. The PMP and PMF results were further compared with previous PMP/PMF estimates. This study demonstrates the value of utilizing stochastic rainfall models and GCM large ensembles to inform PMP and PMF analysis in a changing climate.

EISCAT3D, which is under its final stage of construction, will be the first incoherent scatter radar (ISR) system to provide the three-dimensional ion velocity across hundreds of kms in vertical and horizontal directions. This presents a tremendous opportunity to study the three-dimensional nature of ionospheric electrodynamics. Here we present a data-driven regional model of the electric field based on the EISCAT3D observations, where the measured F-region ion velocity data are fitted to a regional electric potential produced by a grid of spherical elementary systems. Performance of the model is demonstrated using simulated ionospheric parameters obtained from the three-dimensional GEMINI model. To simulate realistic radar measurement of the ion velocity, error estimates obtained from the e3doubt package are added to the ground truth GEMINI data. Our model can be used either with multistatic or monostatic measurements of the ion velocity, and it can also integrate ion velocity data from other platforms, such as satellite sensors, into existing ISR measurements. The model captures the ground truth electric field including its complex spatial structure with average percentile differences of about 1.5%. Most accurate results are achieved with the multistatic data, but the general spatial structure of the electric field can be captured also with monostatic data, if optimal beam patterns and regularization are used. The modeling method is also applied using real monostatic line-of-sight ion velocity data measured by the Poker Flat ISR. The modeled electric field shows reasonably well-behaved variations in latitude and longitude within the radar’s field of view.

This study presents a comprehensive analysis of the 1.5 years’ worth of OGO 6 BIMS data, focusing on the density profiles of ionospheric oxygen and nitrogen ions across various environmental parameters, including solar activity, geomagnetic activity, seasons, latitude, altitude, and magnetic local time. While the abundances of both O+ and N+ decrease monotonically with increasing altitude, the two species follow significantly different power laws, such that nN+ decreases at a slower rate (1/x) than nO+ (1/x2). This leads to a linear increase in the nN+/nO+ ratio with altitude. The average nN+/nO+ ratio reaches its maximum in the midnight sector, whereas the average densities of both oxygen and nitrogen ions peak on the day side. The peak in oxygen ion density leads that of nitrogen ion by 30º in longitude. These differences are most likely due to the different spatio-temporal scales for production and loss for those two ion species. The nN+/nO+ ratio displays a strong seasonal dependence, reaching its highest values during the summer season.

Conventional tsunami simulations rely on accurate bathymetric data, posing challenges in regions lacking such information. We introduce a novel approach using ambient noise interferometry to derive empirical Green’s functions of infragravity waves (IGWs) from noise correlation functions (NCFs) extracted from a 10-year DART dataset in the Pacific Ocean. Our analysis reveals pronounced propagating behavior in NCFs, indicative of wave dispersion relationships. Long-period NCFs align with shallow-water wave dynamics, making them suitable for tsunami simulations. By eliminating the need for precise bathymetry, our method offers a practical solution for data-sparse regions. A case study of an Alaska tsunami demonstrates our NCFs effectively fit observed pressure data, outperforming conventional COMCOT simulations. The fidelity of our results underscores the potential of ambient noise interferometry-derived NCFs to enhance tsunami predictions, even in complex environments. Our findings advance tsunami research and have significant implications for disaster preparedness and mitigation.

We explore the skill of the fully initialized decadal forecasting system EC-Earth3 in re-forecasting annual frequencies of recurrent, regional-scale atmospheric circulation types in the extra-tropics. To this end, 6-hourly sea-level pressure data from 10 model integrations run within the dcppA-hindcast experiment are transformed into discrete time series of categorical circulation types following the Jenkinson-Collison approach. The annual occurrence frequencies of these types are verified against those obtained from a reference reanalysis (ERA5) to estimate the hindcast skill, which is then compared with the skill of a 10-member ensemble run under historical conditions to estimate the added value of initialization. While the results for the Northern Hemisphere are heterogeneous, the initialized experiments provide systematic skill and added value in the Southern Hemisphere during austral summer, up to the second forecast year. Skill correlates with the Ratio of Predictable Components and the westerly wind conditions are particularly prone to the signal-to-noise paradox.

Future projections and past reconstructions of Antarctic Ice Sheet stability and sea-level rise depend on knowledge of continental shelf bathymetry, which controls water circulation under floating ice and interactions between the ice shelf and seafloor. We present a bathymetry model of the Venable Ice Shelf in the Bellingshausen Sea sector, from an inversion of airborne gravity data. The new model reveals troughs up to ~1.6 km deeper than previously mapped, providing pathways for warm Circumpolar Deep Water to access the grounding line. A bathymetric high beneath the western Venable Ice Shelf is identified as a former pinning point that has been intermittently grounded since ~1935 and allows us to interpret crevasse patterns in satellite imagery as evidence of mid-20th century ice-shelf thinning, extending the ice-shelf thickness record beyond the satellite era.

Phytoplankton carbon biomass represents a critical indicator for marine ecosystem function. Here, we examine changes in surface PHYC (PHYCOS) within a suite of Earth System Models participating in the Coupled Model Intercomparison Project, Phase 6 (CMIP6), comparing an extreme climate change case with a preindustrial control simulation. While most models produce mutually consistent spatiotemporal patterns of PHYCOS, they diverge under climate change. We train machine learning emulators on the model output and feed these emulators different combinations of preindustrial and climate-altered inputs. Feeding climate-altered environmental predictors to the emulator trained with preindustrial inputs allows us to explain the majority of change away from the Arctic, suggesting that climate change largely shifts the location and timing of particular conditions. Maintaining constant inputs while varying the emulator, however, shows the impact of shifts in apparent relationships between environmental drivers and phytoplankton biomass, particularly in the Arctic. Low-latitude divergence in the relative biomass change is driven by differences in the response to nutrients, with some models showing influence of temperature. Differences in high-latitude relative biomass changes are driven by strong intermodel differences in the response to light, mixed layer depth, and nutrients. Relative to observations, models show much stronger sensitivity to light and macronutrients, suggesting that they may overrespond to climate change. However, this may be mitigated by unrealistically weak responses to iron.

Tropical cloud fractions from combined CALIPSO and CloudSat observations are classified based on a K-means cluster analysis. These satellite instruments together retrieve detailed vertical structure of clouds, allowing for an objective decomposition of tropical clouds into distinct regimes. The identified cloud regimes are overall robust except that an increasing number of redundant high-cloud clusters are produced as the specified number of clusters is increased. This redundancy arises because geometrically thin clouds with slightly different altitudes can be mathematically distant from one another in the clustering procedure despite their morphological similarity, suggesting that an excessive number of clusters only generate unrealistic subclasses of high clouds. Objective evaluation metrics indicates that the most appropriate number of clusters is four, where the clusters may be labeled as near clear-sky, low cloud, high cloud, and deep convective regimes. A five-cluster classification is also deemed as reasonable, where a fifth regime of congestus clouds joins the other four regimes practically equivalent to the four-cluster solution. This study suggests that determining the optimal clustering solutions requires a careful interpretation from meteorological perspectives beyond a mere statistical assessment of clustering performance.

This study examines West Africa’s climate vulnerability under stratospheric aerosol injection (SAI) using two climate models: UKESM1 and CESM2. We analyzed temperature and precipitation patterns between 2050-2069 compared to 2015-2034 under two scenarios: SSP2-4.5 and ARISE-SAI-1.5. Our methodology involved evaluating temperature and precipitation anomalies, using Signal-to-Noise Ratio (SNR) analysis to assess climate signal reliability, and analyzing precipitation extremes through Cumulative Distribution Function (CDF) and Probability Density Function (PDF). Under SSP2-4.5, UKESM1 showed temperature increases of 1.8{degree sign}C while CESM2 showed increases of 1.0-1.2{degree sign}C. However, under ARISE-SAI-1.5, UKESM1 showed cooling (-0.3{degree sign}C below reference period) while CESM2 maintained slight warming (0-0.3{degree sign}C above reference). SNR analysis revealed that SAI significantly dampened the warming signal in UKESM1, making precipitation and temperature trends less detectable. CDF and PDF analyses showed that while SAI may reduce warming, it increases precipitation variability and uncertainty. These results emphasize the importance of multi-model comparisons when assessing geoengineering impacts, as models can produce varying results based on their sensitivity to radiative forcing and other factors.

Stable isotopes of hydrogen and oxygen in precipitation provide insights into water cycle dynamics, yet characterizing the meteorological processes driving isotopic variability remains challenging. We introduce a method to rasterize air mass trajectory data from HYSPLIT for isotopic analysis and assess the impact of trajectory initiation height on HYSPLIT’s representation of local precipitation and temperature. Incorporating HYSPLIT rasters into spatial analysis improved deuterium composition isoscape accuracy for precipitation compared with a model relying solely on surface predictors (i.e., temperature, elevation). The resulting isoscape patterns reflected strong, temperature-dependent seasonality, aligning well with previous studies. Trajectory cluster analysis further elucidated the links between meteorological patterns and observed seasonality in isotope compositions. We also analyzed daily precipitation at two sites (n=204, n=138) over ~2 years using Random Forest and Multiple Linear Regression. Our analysis identified a significant amount effect and evidence for sub-cloud evaporation. However, additional factors-such as cloud-top temperature, reflectivity, and convective versus stratiform fractions-are likely needed to explain residual daily-scale variance. Finally, we determined that an initiation height of 1000 m was sufficient for our study domain. Using the local condensation level as the initiation height provided no substantial improvement in correlations between modeled and observed precipitation or temperature.