Earth and Environmental Sciences

Operational crop mapping requires classifiers capable of robust generalization across years. While feature importance is routinely used for model optimization, its temporal stability has rarely been systematically investigated, creating a critical gap in deploying reliable monitoring systems. This study moves beyond identifying “most important” features to systematically evaluate and quantify their inter-annual stability for enabling automated classification. Using six agricultural years (2018, 2019, 2020, 2023, 2024 and 2025) of Sentinel-1 and Sentinel-2 data over Morocco, we extracted 156 multi-sensor features across 12 monthly composites and analyzed their importance stability through statistical metrics, clustering, and novel composite indices: the Reliability Index (RI) and Automatic Selection Score (AuSS). This framework automates feature selection by ranking features with RI and AuSS and then applying Pareto optimization to identify a minimal stable feature set—without requiring annual retraining or expert intervention. Our analysis confirms a fundamental tension: the most discriminative features (e.g., NDVI, VH, VV) are also the most volatile, while stable features (e.g., NDRE, MSI, NDMI) offer modest predictive power. Hierarchical clustering revealed four behavioral typologies (Dominant Stable, Performant Volatile, Stable Minor, and Noise), guiding strategic feature management. Crucially, a Pareto analysis demonstrated that a refined portfolio of 6 indices (VH, VV, NDVI, NDRE, GCVI, RVI) captures 57.2% of cumulative predictive importance, filtering out inter-annual noise while preserving discriminative signal. The Voting Ensemble leveraging this Stable Portfolio maintained consistent high accuracy (87.4% accuracy, 87.2% F1-score) with minimal performance degradation during temporal transfer, while models based on volatile top features exhibited significant drops. Entropy analysis confirmed that all features in the Stable Portfolio provide consistent informational certainty, indicating that stability-driven selection does not increase model uncertainty. We conclude that feature stability is not merely a diagnostic metric but a foundational criterion for operational design. We propose a practical, metrics-driven framework for constructing automated crop classification systems that are more resilient to inter-annual climate variability.

River deltas exhibit a wide range of morphologies shaped by the balance between fluvial, wave, and tidal processes, yet existing classification frameworks often rely on qualitative descriptions or data-intensive environmental metrics that are not consistently available at global scales. To address these limitations, we present DeltaLatent, a deep-learning workflow that classifies delta morphotypes using only satellite imagery and latent-space representations. Landsat 8 images of 222 deltas were standardized using Google Earth Engine, converted to grayscale, edge-enhanced with a Prewitt filter, and manually rotated for orientation consistency. A convolutional autoencoder was trained to compress each image into a 256-dimensional embedding that captures key morphological patterns. Representative deltas from each morphotype were used to compute centroid embeddings, and the L1 distances between all embeddings and these centroids were converted into similarity scores plotted in a ternary morphospace. The resulting diagram successfully reproduces well-established fluvial, wave, and tidal end-members, while also capturing transitional and mixed systems consistent with geomorphic interpretations in the literature. DeltaLatent demonstrates scale flexibility, enabling meaningful comparison across deltas of very different sizes. This morphology-based, data-driven framework provides a reproducible alternative to traditional process-based classifications and offers a foundation for global morphodynamic analysis, monitoring, and hazard assessment.

The surface radiation budget, controlled by the surface albedo, is a key driver of the Earth’s energy balance. Snow-covered surfaces, with their high reflectivity, scatter a large fraction of the incoming shortwave radiation back to the atmosphere and thereby modulate both local and global energy exchange. In complex topography, multiple and anisotropic forward scattering redistributes solar radiation within valleys and across slopes. To investigate terrain-scattered radiation, we combined drone-based albedo measurements with simulations using the surface radiation model Groundeye in the Meierhofer Tälli near Davos, Switzerland. Albedo was measured at different heights above ground, which serves as a proxy for different footprints for remote sensing products or different grid resolutions in models. Our objective is to understand the impact of the complex terrain on albedo, and specifically how multiple reflections and anisotropic scattering affect local reflectivity at different heights above the ground. We discuss our results primarily in the context of the representation of albedo in complex terrain and the potential biases in energy balance models and remote sensing retrievals at different resolutions. We show that 1. The ground albedo at the chosen locations in this valley varies between 0.54 and 0.98, depending on local time, slope and aspect, and on the surrounding topography, if multiple reflections are considered in our model, and 2. That the changes in albedo with height are location-specific and can increase or decrease depending on the footprint of the upward radiation flux. The albedo decreases if more shaded areas enter the footprint and increases if more sunlit surfaces become visible. However, changes of albedo with height are not only a function of simplified terrain parameters, such as the skyview factor. These results highlight the complexity inherent in interpreting albedo measurements across mountainous terrain. Such a complex change in albedo with elevation above ground is demonstrated in connection with topographic effects and has implications for improving the parametrisation of surface reflectivity in models and the derivation of surface properties from remote sensing observations.

Accurate and precise estimation of forest volume and its changes are fundamental for sustainable forest management, resource assessment, and long-term inventory monitoring. Multi-temporal remotely sensed data provide spatial auxiliary information that can substantially increase the precision of estimates of volume and volume change obtained from probability-based field samples. However, methodological differences between indirect and direct change estimation approaches may lead to differences in variance estimation and uncertainty propagation. This study, therefore, compares forest volume and forest volume change estimation under design-based inference by integrating probability-sampled National Forest Inventory (NFI) data with multi-temporal Sentinel-2 auxiliary variables at 10 m and 20 m spatial resolutions, using the random forest (RF) prediction algorithm and both indirect and direct estimation approaches. Forest volume means are estimated for two inventory years, 2013 and 2023. For the indirect approach, simple expansion and model-assisted regression estimators are formulated separately for each year, and change is estimated as the difference between time-specific estimates. For the direct approach, plot-level change is used as the response variable, and model-assisted regression estimators are constructed using changes in auxiliary variables, with residual-based variance estimators. Random forest models using Sentinel-2 auxiliary data explained approximately 51–62% of the variation in forest volume, with RMSE values ranging from about 79 to 97 m3/ha. The 10 m resolution data produced slightly more precise predictions than the 20 m data, though the gain was small relative to the greater processing effort required. Model-assisted estimators using both 10 m and 20 m Sentinel-2 data produced substantially smaller standard errors than simple expansion estimators, with relative efficiency analysis indicating an approximately fivefold gain in efficiency. The direct and indirect model-assisted approaches produced similar estimates of volume change, although the direct approach resulted in smaller standard errors. Overall, remotely sensed auxiliary data primarily improved the precision of forest volume change estimates but not the magnitude of the estimated change.

Abstract. Triple oxygen isotopes are powerful tracers of hydrological processes, yet their variability in atmospheric water vapor and the underlying controls remain poorly understood. We present a one-year record of triple oxygen and hydrogen isotopes of atmospheric water vapor (δ18OV, d-excessV, 17O-excessV) measured below, within and above a downy oak forest canopy at the AnaEE platform O3HP in the French Mediterranean. This vapor dataset is complemented by isotopic data of precipitation (δ18OP, d-excessP, 17O-excessP) and groundwater, as well as monthly observations of stomatal conductance and transpiration. Seasonal variations in 17O-excessV and d-excessV likely reflect changing evaporative conditions at oceanic moisture sources. d-excessP and 17O-excessP show a similar seasonal pattern, enhanced by summer rain re-evaporation. However, no clear isotopic differences were observed in vapor or precipitation derived from different oceanic source regions and weather regimes, likely due to frequent mixing of multiple moisture sources. Diurnal variations in 17O-excessV and d-excessV reflect a combination of vegetation-related processes, including local evapotranspiration. However, the impact of evapotranspiration was not evident at daily timescales. Although precipitation often deviates from isotopic equilibrium with near-surface atmospheric water vapor at the event scale due to incomplete equilibration and rain re-evaporation, equilibrium water vapor reliably approximates the near-surface isotopic composition of atmospheric water vapor at annual scale. Our results highlight the potential of 17O-excess for understanding water exchange between the land and the atmosphere, regardless of climatic and vegetation conditions. They enhance the mechanistic interpretation of precipitation isotopes, which is essential for reliable paleoclimate reconstructions.

Study region The Middle Benue Trough, Nasarawa State, Nigeria. Study focus Groundwater quality deterioration is an increasing environmental and public health concern in this region; however, studies that integrate entropy-based water quality assessment, interpretable machine-learning modelling, explainable artificial intelligence, and health risk evaluation within a unified framework are limited. This study integrates hydrogeochemical analysis, entropy-weighted water quality indexing (EWQI), interpretable stacked-ensemble machine learning, and health risk assessment to evaluate seasonal groundwater quality dynamics across five Local Government Areas (LGAs). Twenty-four parameters from 600 groundwater samples collected during dry and rainy seasons were analysed. Hydrochemical facies comprised Na–Cl, mixed Ca-Mg-Cl, and Ca-Mg-HCO 3 water types, reflecting the combined influence of meteoritic recharge, carbonate and silicate weathering, cation exchange, and localised anthropogenic inputs. EWQI results identified Awe and Doma as contamination hotspots, with groundwater quality generally deteriorating during the rainy season and requiring management interventions. New hydrological insights for the region The stacked-ensemble model outperformed individual machine-learning algorithms for EWQI regression and classification, achieving excellent predictive performance (R 2 = 0.935; RMSE = 9.420) and classification accuracy (0.858). SHAP interpretation identifies As, Pb, Cd, and Mn as the main drivers of groundwater deterioration. Positive SHAP–HQ relationships demonstrated that contaminants exerting the greatest influence on groundwater quality deteriorations also contributed weak-to-moderate influence to health risks. The integrated framework provides a valuable tool to identify key contaminants, sensitive locations, and priority intervention targets in hydrogeologically complex areas.

Decametric-resolution leaf area index (LAI) is an essential parameter for fine-scale crop growth monitoring and ecosystem modeling. Prior-guided approaches using existing hectometric-resolution LAI products have demonstrated potential in large-scale decametric-resolution LAI estimation. However, within such approaches, the impacts of algorithm selection and band combination on retrieval accuracy remain insufficiently quantified, and the lack of model interpretability limits methodological transferability. To address these challenges, a multi-source data integration (MSDI) framework is developed to systematically assess the sensitivity of prior-guided LAI estimation to retrieval algorithms and spectral bands using Sentinel-2 imagery. In addition, Shapley Additive Explanations (SHAP) is employed to quantify the contributions of individual bands and interpret model behavior. The MSDI LAI was evaluated using ground LAI measurements and compared with Simplified Level 2 Product Prototype Processor (SL2P)-derived LAI and MODIS LAI products. The results indicated that Support Vector Regression (SVR) achieved the best performance in LAI estimation among six machine learning algorithms, likely due to its robustness in modeling nonlinear relationships across different training samples. Band optimization further reduced estimation uncertainty by >24% and increased R2 by >44% for SVR-derived LAI estimates. Moreover, MSDI outperformed SL2P, especially at 20 m resolution, with Bias, RMSE, and R2 values of 0.26, 0.76, and 0.71, respectively. Meanwhile, MSDI LAI exhibited a similar spatial distribution to MODIS LAI while providing substantially enhanced spatial detail and accuracy. SHAP analysis revealed that red-edge (RE) and shortwave-infrared (SWIR) bands contributed the most to LAI prediction, consistent with their sensitivity to vegetation canopy biophysical properties. Overall, this study highlights the importance of retrieval strategy optimization and model interpretability for improving prior-guided decametric-resolution LAI estimation and offers practical guidance for generating consistent LAI estimations across various scales.

ABSTRACT Based on 30 CMIP6 global climate models and observational data from 2374 meteorological stations in China during 1961–2020, the primary objective of this study is to investigate the sources of uncertainty in model projections from both thermodynamic and dynamic perspectives. The projection of mid‐summer precipitation trend is slight increasing in the near‐term (2021–2040), maximum in the mid‐term (2041–2060) and slowing down in the long‐term (2081–2100) over northern China monsoon region (NCMR), and its mean increasing trend is projected to be 3% (5%) under different scenarios in the near‐term (mid‐term), and slightly decreases to 2% in the long‐term. Especially, it is sensitive to temperature change under shared Socioeconomic Pathway 1‐2.6 (SSP1‐2.6). Optimal model simulations show that the dynamic term dominates the increase in NCMR midsummer precipitation, while the thermodynamic term contributes negatively, offsetting some of the dynamic‐induced rise. Therefore, it is highly likely that the dynamic term dominates the future increase in precipitation over the NCMR and is also the primary factor contributing to discrepancies in results among different models.

Saline–alkaline land serves as a potential arable land reserve for augmenting agricultural productivity and safeguarding food security. However, long-term monitoring of saline–alkaline land conversion remains challenging because of vegetation recovery, surface changes, hydrological modification, and agricultural phenology. Compared with CCDC and LandTrendr, the proposed MK-based framework detects conversion occurrence and timing while reducing dependence on dense observations, parameter tuning, and annual classification. This study examines the spatiotemporal dynamics of saline–alkaline land converted into paddies in Da’an City, utilizing Landsat time-series data (2007–2021) from the Google Earth Engine (GEE) platform. The analysis employed Mann–Kendall (MK) trend and mutation tests to monitor conversion processes and analyze spatiotemporal dynamics. Point-biserial correlation analysis was applied to evaluate the sensitivity of various remote sensing indices in detecting land conversion. The top fifteen indices, including the Land Surface Water Index (LSWI), Salinity Index 4 (SI4), and Salinity Index 5 (SI5), demonstrated strong correlations (|r| = 0.788–0.885) and significant pre- and post-conversion spectral differences (p < 0.01). Validation via confusion matrix confirmed that the June SI5 index attained the highest detection accuracy (overall accuracy: 94.15%; Kappa coefficient: 0.86), supporting the MK trend test’s efficacy in monitoring conversion processes. The MK mutation test achieved 80.36% temporal accuracy in determining conversion timing. The spatiotemporal analyses identified heterogeneity in saline–alkaline land conversion patterns. Spatially, large contiguous paddy fields dominated the eastern region, whereas fragmented conversion characterized the west, with minimal activity in the central zone. Temporally, the conversion area expanded rapidly before 2015 and then gradually declined, reaching a cumulative converted area of 276.29 km2 by 2021. This study elucidates spatiotemporal conversion dynamics to guide sustainable land use.

Abstract Groundwater conservation policies are typically evaluated using historical observations. However, in coupled human–water systems, outcomes can vary substantially even under unchanged mean conditions of the climatic, economic, and land-use inputs due to nonlinear dynamics and spatiotemporal interactions. This phenomenon is referred to as internal variability. Here, we quantify how internal variability shapes inference about policy performance using the Sheridan-6 Local Enhanced Management Area (SD-6 LEMA) in western Kansas, a well-documented irrigator-led conservation program. We employ an agent-based human–hydrologic model to generate large ensembles through alternative temporal sequences of precipitation and crop prices and alternative spatial configurations of initial crop fields, while preserving fixed mean system states. These ensembles are explicitly contrasted with scenarios representing shifts in mean climate, economic, and land-use conditions. A three-way factorial analysis of variance shows that internal variability accounts for more than half of total variance in both policy-relevant indicators (groundwater withdrawal and saturated thickness) and behavioral outcomes (crop choice and farmer behavioral state transitions). The influence of internal variability is further amplified under higher irrigation intensity, where tighter constraints increase system sensitivity to spatiotemporal fluctuations. These findings do not challenge the effectiveness of the SD-6 LEMA policy itself, but instead demonstrate that policy performance inferred from observational data or a single historical realization can be misleading over short to medium timescales. We conclude that robust groundwater policy evaluation requires ensemble-based approaches that characterize outcome distributions under fixed institutional rules, with important implications for policy robustness and transferability in complex adaptive groundwater systems.

This study adopts MAX-DOAS technology and optimal estimation methods to analyze the vertical profiles of aerosols, HONO, and O3 in Kunming during winter 2024, thereby clarifying their vertical distribution characteristics and diurnal variation rules. Aerosols are predominantly concentrated in the near-surface layer, with concentrations decreasing gradually with increasing altitude; their concentrations rise in the morning, reach a peak at 10 a.m., and then decline, with the maximum extinction coefficient reaching 1.3 km−1. HONO concentrations decrease exponentially with height, being the highest around sunrise, the lowest at noon, and then gradually accumulating, with the maximum concentration exceeding 3.14 × 1010 molec·cm−3. O3 achieves its maximum concentration in the 2.6–4 km altitude range in the afternoon (the highest concentration > 1.16 × 1012 molec·cm−3), while near-surface O3 concentrations increase steadily from morning onward. Model analysis demonstrates that without considering HONO, OH radicals shift from net production to net loss in the morning, leading to reduced O3 production rates and a decrease of approximately 8 ppb in the maximum O3 concentration. These findings highlight the significant vertical and diurnal variations in the three pollutants, emphasize HONO’s crucial role in OH radical generation and O3 formation, provide valuable scientific support for air pollution control, and confirm HONO’s important impact on regional atmospheric chemistry and the distribution of key pollutants.

Low-slow-small (LSS) target recognition using multi-dimensional radar remains challenging due to weak signatures, similar kinematics, and overlapping short-term Doppler patterns. Digital-array radar provides continuous, complementary Doppler-spectrum and kinematic measurements; however, their heterogeneity in dimension, distribution, and physical meaning often makes direct fusion under-exploit discriminative complementarity and inadequately model temporal track evolution. To address this, we propose a Doppler-Kinematic Spatio-Temporal Graph Learning framework named Dual-Stream Spatio-Temporal Cross-Attention Graph Convolutional Network (DS-STCAGCN) for LSS target recognition using multi-dimensional radar observations. The method separately encodes Doppler-spectrum and kinematic features to preserve their modality-specific characteristics, fuses them through bidirectional cross-attention, captures long-range temporal dependencies via self-attention, and aggregates local frame-to-frame correlations through graph convolution on a time-ordered observation graph. On the public L-band digital-array dataset LSS-DAUR-1.0, DS-STCAGCN achieves 99.73% mean accuracy and maintains 98.64% at 5 dB signal-to-noise ratio (SNR). On the passive-radar dataset LSS-PR-1.0, it reaches 99.86% mean accuracy, demonstrating strong cross-modal generalization. This work provides an effective spatio-temporal modelling framework for multi-dimensional radar sensing and robust LSS target recognition.

The North Tropical Atlantic (NTA) is a leading mode of tropical interannual variability. NTA warming shifts the Atlantic intertropical convergence zone (ITCZ) northward, manifesting itself as enhanced rainfall in the Northern Hemisphere (NH) and reduced rainfall in the Southern Hemisphere (SH), thereby driving broad climatic impacts worldwide. However, the ability of climate models to simulate the NTA’s influence on the ITCZ remains unclear. Here, we evaluate the performance of 34 AMIP and coupled models from phase 6 of the Coupled Model Intercomparison Project (CMIP6). In AMIP, most models simulate eastward-displaced SH anomalies. For the NH anomalies, although the multi-model ensemble mean reasonably reproduces the observed location, individual models exhibit larger biases. In CMIP, model biases are larger than those in AMIP: 9 of 34 (26%) models fail to reproduce the northward ITCZ shift, and most of the remaining models simulate a southeastward displacement in both anomalies. In both atmospheric and coupled models, these location biases are closely related to those in rainfall climatology, whereas biases in simulating the NTA spatial pattern play a limited role in CMIP.

Assessing urban resilience requires accurate data on building age as a temporal proxy associated with structural vulnerability, yet persistent cloud cover and rapid development constrain data availability in tropical and subtropical regions. We propose a computationally efficient framework that prioritizes annual data integrity over monthly granularity to map building construction years. By combining annual cloud-free Landsat NDVI (Normalized Difference Vegetation Index) composites with open-source building footprints, the framework utilizes a Temporal Template Matching (TTM) algorithm to detect the distinct “vegetation-to-built” transition signal. Evaluated across two dynamic and heavily cloud-contaminated metropolitan areas—Shenzhen, China, and Hanoi, Vietnam—this approach achieves a high producer’s accuracy; furthermore, in Shenzhen, where a monthly comparative analysis was conducted, it outperforms a noise-sensitive monthly LandTrendr-based baseline by a factor of nearly 1.8. Our findings demonstrate that under persistent cloud contamination, a coarser but consistent annual composite provides a more reliable signal than finer-grained alternatives. This scalable methodology generates critical building-age datasets, offering foundational structural intelligence for potential inputs into seismic risk modeling and resilient urban planning in rapid-growth and resource-constrained regions.

Live fuel moisture (LFM) is a critical determinant of wildfire behavior, especially in Southern California chaparral, yet spatially continuous and near-real-time estimates remain limited by sparse field measurements, and spatial transferability challenges. In addition, the transition from MODIS to VIIRS requires evaluation of the continuity of long-term satellite-based LFM monitoring. This study developed a framework using 2003–2022 Globe-LFMC 2.0 dataset that first compared MODIS-based multiple linear regression and random forest models, then applied bias correction, and transferred the framework to VIIRS to assess cross-sensor continuity. This framework was further extended for near-real-time application by integrating analog-year phenology estimation and survival-based dry-down timing estimation. For MODIS, random forest achieved higher accuracy with the full dataset, but its performance declined substantially under leave-one-county-out spatial cross-validation and showed reduced ability to capture low and fire-disturbed LFM values. Multiple linear regression showed more stable performance between full-dataset evaluation (R 2 = 0.63, RMSE = 12.35%) and spatial cross-validation (R 2 = 0.58, RMSE = 12.43%), indicating greater spatial transferability. MODIS models provided higher predictive skill overall than VIIRS, whereas the M-band-only VIIRS configuration produced the highest VIIRS performance and reproduced major seasonal and spatial LFM patterns. Independent validation using 2023–2024 Fire Environment Mapping System (FEMS) dataset showed that LFM thresholds of 85–90% provided the most balanced classification of elevated-risk conditions for both MODIS and VIIRS. This study shows that an interpretable, phenology-informed satellite framework can support spatially transferable and temporally consistent LFM monitoring for chaparral fire risk assessment.

Abstract Lake water quality reflects interactions between in-lake processes and catchment-scale drivers, yet integrated datasets that combine lake chemistry with spatially explicit catchment information remain limited. Here, we present a harmonized, multiscale dataset linking water chemistry to delineated catchment characteristics for lakes included in two national Swedish monitoring programs: the Swedish Lake Survey (SLS) and the Swedish Trend Lakes Program. The SLS comprises more than 4,800 lakes sampled on a six-year rotating basis, while the Trend Lakes Program includes 110 lakes sampled annually. Together, the dataset encompasses over 5,230 lakes distributed across Sweden. For each lake, catchments were delineated using consistent geospatial procedures, and nationally available spatial datasets were used to derive topographic attributes, land use and land cover, climate variables, modeled runoff, atmospheric sulfur and nitrogen deposition, and vegetation dynamics. The dataset integrates spatially explicit catchment attributes with temporally resolved water chemistry data, enabling cross-scale analyses of lake-catchment interactions, long-term environmental change, and responses to atmospheric deposition and climate variability.

Study region Lake Balkhash, a large endorheic lake in southeastern Kazakhstan, Central Asia, with strong spatial variation in water clarity. Study focus Accurate lake water storage estimation in arid regions requires reliable bathymetry, but large optically complex lakes remain difficult to map because of limited in situ data and rapid signal attenuation. This study develops a bathymetric retrieval framework for Lake Balkhash by integrating ICESat-2 photon-counting LiDAR and multi-temporal Sentinel-2 imagery. An adaptive DBSCAN-based algorithm was used to extract lakebed photons from noisy photon clouds, and the resulting depth samples were combined with Sentinel-2 spectral features in a Random Forest model to produce a continuous high-resolution bathymetric map. New hydrological insight for the region The proposed fusion framework performed well in reconstructing lakebed topography under varying water-clarity conditions, achieving an RMSE of 1.046 m against independent in situ validation data. The reconstructed bathymetry reveals clear morphological differences between the two sub-basins, suggesting that the shallow western sub-basin is more sensitive to hydrological fluctuations than the deeper saline eastern sub-basin. By establishing level–area–volume relationships, we reconstructed a long-term record of lake water storage changes. This record indicates a renewed decline in water storage during 2019–2024. These findings demonstrate the limitations of traditional area-based monitoring and provide a volumetric baseline for water resource assessment in Central Asia.

Marine litter has been quantified in several deep-sea environments, including complex habitats such as submarine canyons that host Vulnerable Marine Ecosystems. However, a standardized method for its characterization and quantification is still lacking. Here, we propose a simple and replicable method based on repeated 100 m transects conducted using Remotely Operated Vehicle video surveys. The method was tested at two sites off the French coasts, in the Western Mediterranean Sea in 2023 and the North-East Atlantic in 2025. Mean litter densities reached 30,101 items km −2 in the Lacaze-Duthiers Canyon, Gulf of Lion, and 18,333 items km −2 in the Lampaul Canyon, Bay of Biscay. The type of litter differed according to the canyons and/or bathymetry. Fishing gear largely dominated the litter at the head of the Lacaze-Duthiers Canyon, whereas plastic bags and packaging constituted most of the litter in the Lampaul Canyon. Litter densities were particularly high in coral reefs that inhabit these canyons, lines and bags being entangled in coral branches. Our results demonstrate the efficiency, robustness, and reproducibility of this approach, as it generated comparable results across two distinct canyons. The method represents a practical and reliable tool for supporting the assessment of criterion D10C1 (seafloor litter) of Descriptor 10 (Marine Litter) of the Marine Strategy Framework Directive (MSFD). It contributes to improved monitoring of marine debris, beyond seafloor litter derived from trawling data, as well as its potential impacts on deep-sea habitats, and will support the implementation of appropriate management plans for the conservation of these vulnerable ecosystems.

Accurate high-resolution mapping of soil organic carbon (SOC) is essential for agricultural management and carbon pool assessment in arid lakeside oases, a fragile aquatic-terrestrial transition ecosystem. However, targeted high-precision SOC mapping for typical lakeside oases remains insufficient: existing models have poor adaptability to the highly fragmented oasis landscapes, and fine-resolution SOC spatial products for the representative Bosten Lake oasis are lacking. To address this inadequacy, we integrated Sentinel-2 imagery with topographic, bioclimatic, and spectral environmental covariates and developed four machine learning models (Random Forest, XGBoost, SVR with RBF kernel, Cubist) for SOC prediction, based on 153 topsoil samples (0–20 cm) collected via stratified random sampling in the study area. Model performance was validated through 5-fold cross-validation, the optimal model was selected for 10 m resolution SOC mapping, and dominant driving factors were identified via SHAP analysis. The results showed that SOC content in the study area ranged from 2.37 to 20.63 g·kg−1 (mean = 10.59 g·kg−1), with moderate spatial variability (CV = 34.86%). The Cubist model achieved the highest mapping accuracy (R2 = 0.8166, RMSE = 1.5812 g·kg−1, MAE = 0.9247 g·kg−1). The generated high-resolution SOC map clearly revealed a spatial pattern of high values in the eastern well-irrigated cropland and low values in bare and salinized areas at the oasis edge. The Bare Soil Index (BSI), surface roughness, and Normalized Difference Red Edge Index 1 (NDRE1) were the dominant factors controlling SOC spatial distribution. This study mitigates the inadequacy of high-precision SOC mapping in typical arid lakeside oases, and the proposed framework is readily applicable to other fragmented arid landscapes worldwide and provides reliable spatial data and a scalable technical framework for precision agriculture and sustainable land management in similar fragile ecosystems.

Novel entities are diverse and complex; hence, defining their Planetary Boundary through control variables, indicators, and acceptable limits is challenging. Here, we highlight shortcomings of the current quantitative boundary, which applies a purely precautionary interpretation of safety that defines the safe operating space as zero novel entities that are not fully characterized. In practice, the zero limit is unattainable and undesirable for human societies that rely on services provided by novel entities. Furthermore, it is not operational within the framework of other boundaries, which have risk-based thresholds. We therefore propose a reboot of the Planetary Boundary for Novel Entities and recommend governance and monitoring strategies based on: (a) clarification of the "safe operating space" for novel entities, distinguishing pure precaution ("safe from the unknown") from operational precaution that tolerates a workable, non-zero level of uncertainty; (b) complementary, multilayer control variables spanning pressures, states, and impacts; (c) extensions to existing monitoring systems to provide quantitative indicators of distribution, accumulation, exposure, and impacts for subsets of novel entities to communicate the status and trends of stress from novel entities on the Earth system; and (d) recognition of regulatory and policy-related success stories that illustrate comprehensive, protective, and operational governance of novel entities.

Abstract Green tides dominated by Ulva prolifera have become a recurrent ecological hazard in the Yellow Sea, which is typically associated with benthic oxygen depletion and localized nutrient regeneration in coastal waters. Existing numerical studies primarily focus on simulating bloom expansion and surface-drifting pathways during the growth stage, with limited attention to the pattern and timing of coastal stranding and offshore sinking during the decay phase. In this study, we developed a Lagrangian particle-tracking framework to explicitly simulate the fate of Ulva prolifera during its decay stage, incorporating parameterizations for beaching and density-driven sinking. The model was driven and validated using high-frequency satellite retrievals from GOCI-II observations, to ensure high temporal adequacy in reproducing the bloom’s evolution. The model adequately reproduced the dynamic transitions among floating, stranded, and sunk biomass fractions. Our results showed that stranded and sunk particles increased rapidly with approximately 90% of total biomass removed from the surface layer within 96 hours, indicating a massive downward flux of biomass to the ocean floor where it’s consumed by the benthos. Stranding occurred predominantly along the southern coast of Yantai City influenced by persistent onshore winds and regional circulation, whereas sinking was concentrated in offshore deeper waters and frontal zones. By explicitly resolving the relative contribution of biomass loss by stranding (~3%) and deposition (>90%), this study advances current drifting modeling toward comprehensive biomass budgeting. The proposed framework may provide essential scientific basis for evaluating coastal and benthic ecological impacts, as well as for improving operational forecasting and coastal management strategies.

The surface urban heat island (SUHI) temporal variations are still not well understood, compared to its spatial features, partly due to the low temporal resolution of satellite infrared data. The ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) provides Land Surface Temperature (LST) observations at a 70 m resolution with varying acquisition times and reduced revisit periods, thus offering new opportunities to study the SUHI dynamics. This study relies on ECOSTRESS data acquired over a five-year period (2018–2023) to explore both the day-to-day variability of the SUHI and its diurnal cycle during summer. We analyze these variations at the neighborhood scale based on the Local Climate Zones (LCZ) framework. The analysis reveals a typical LST contrast among LCZ classes, with a clear distinction between built types and natural types, but also within those two groups. The day-to-day variation of this pattern is assessed and it is found to be higher during daytime than at night. The drivers of these variations are shown to be distinct between day and night, with a daytime variability mostly driven by vegetation health and cover while, at night, the mean wind speed and the cloud cover are the most important drivers. Continuous diurnal evolution curves of the SUHI intensity are produced and enable the observation of distinct diurnal dynamics for different LCZs. Our approach, fully based on observational data and statistical modeling, contributes to fill the lack of observations of the SUHI dynamics and could be reproduced for inter-city comparison.

Abstract Understanding CO 2 nanobubble formation in water‐saturated sandstone is critical for understanding fluid behavior in CO 2 storage systems. Here, CO 2 exsolution from an aqueous phase in a sandstone was investigated using small‐angle neutron scattering at 50°C during cyclic depressurization from 12 to 0.7 MPa. Nanoscale heterogeneities consistent with CO 2 clusters and nanobubbles (5–200 nm) were resolved during pressure reduction. Although bulk phase diagrams predict exsolution at ∼8 MPa at 50°C, detectable exsolution emerged only at 2.4 MPa, indicating strong confinement and surface effects. A progressive loss of signatures associated with nanobubbles <∼15 nm suggests preferential disappearance of nanobubbles, consistent with curvature‐driven coarsening (e.g., Ostwald ripening). Repeated cycling revealed partial qualitative reversibility, implying nucleation and saturation‐history (hysteresis) effects relevant to operational pressure transients in CO 2 storage. Our findings improve assessment of CO 2 mobility, trapping, and leakage risk under dynamic pressure in CO 2 storage systems.

Abstract The K-alkaline, Quaternary Roman Magmatic Province (RMP) includes several active and quiescent caldera-forming volcanoes along the Italian Tyrrhenian margin, which have erupted tens of intermediate to large volume ignimbrites, poorly studied in terms of erupted volume and associated plumbing system, with few noticeable exceptions like the 40 ka Campanian Ignimbrite from Campi Flegrei. Here we reappraise the 433 ka Tufo Rosso a Scorie Nere sabatino (TRSNs) ignimbrite from the Bracciano caldera. Using new field data, we calculate the erupted DRE volume to be 67–170 km 3 , corresponding to a mass of 0.87–4.35 × 10 14 kg, which classifies the eruption as VEI 7 and magnitude 7, and the second largest identified within the RMP, after 181–265 km 3 DRE of the Campanian Ignimbrite. This new volume estimate is one order of magnitude larger than previously suggested and suggests that the volumes of many other ignimbrites from the RMP may be severely underestimated, potentially qualifying the RMP as an ignimbrite flare-up system. Geochemical characterization of the TRSNs residual glass allows us to test the efficiency of rhyolite-MELTS geobarometry for phonolite compositions and discuss the geometry and structure of the plumbing system leading to the TRSNs caldera-forming event. These data highlight the occurrence of a zoned magma reservoir, which fed first the initial Plinian phase, followed by the main ignimbrite and caldera-forming phase.

Study region The Qilian Mountains, northwestern China Study focus Using the PML-V2.8 evapotranspiration (ET) dataset together with hydroclimatic, vegetation, and topographic variables, this study examined the spatiotemporal variations of ET across forests, shrublands, and grasslands in the Qilian Mountains during 2000–2022. Ensemble machine-learning models combined with SHapley Additive exPlanations (SHAP) were further used to identify dominant controls, monthly shifts in driving factors, and response patterns among different vegetation types. New hydrological insights for the region From 2000–2022, ET increased significantly in the Qilian Mountains, mainly in spring and summer. Forest ET increased fastest, whereas grassland ET showed the strongest interannual variability. The dominant controls differed among vegetation types: forest and shrubland ET were mainly regulated by hydroclimatic factors, particularly solar radiation, whereas grassland ET was jointly constrained by energy conditions and elevation-related topographic gradients represented by DEM. Monthly SHAP analysis revealed clear seasonal shifts in dominant controls, with LAI dominating forest and shrubland ET during June-August and DEM dominating grassland ET in most months. Response curves indicated no strong nonlinear responses or clear threshold effects, highlighting the combined roles of vegetation type, seasonal driver shifts, and topographic constraints in regulating ET across mountainous ecosystems.