Earth and Environmental Sciences

Abstract Floods are a recurrent phenomenon in the Indian state of Assam, with nearly 39% land classified as flood-prone. This increases the vulnerability of the agriculture sector that supports around 70% of the state’s population either directly or indirectly. While extensive research exists on the effects of floods and climate change on the agriculture sector, only a few explain their direct long-run consequences and the micro-level adaptive responses. We fill this gap by analysing the impact of floods and climate variability on the primary sector by considering Assam’s agricultural gross value added (GVA), that embodies the sector’s real economic performance in comparison to the aggregate gross domestic product, for the period 1980–2023. We found statistically significant bidirectional relationship between floods and GVA, along with interactions between floods and local climatic conditions. The findings demonstrate that environmental shocks directly affect agricultural performance. We also observe that adaptive capacity and resilience of the agriculture system are impacted indirectly. Therefore, there is a need to understand how adaptation policies need to be designed so that the resilience of the sector improves. Qualitative analysis of survey data collected from a sample of 436 farm households across 16 villages in the Brahmaputra valley, capturing local perceptions, and coping strategies help us in this direction. The study underscore mediating role of adaptive capacity, a key contribution of the study. Results emphasizes the need for a pro-active disaster risk reduction programme. Overall, the study informs policymakers that they should develop targeted strategies to further lower the adverse effects of floods and climate variability on the agricultural sector.

Abstract. Atmospheric total peroxy radicals RO2* (RO2∗=HO2+∑RO2) play central roles in tropospheric chemistry, governing the formation of ozone and secondary aerosols. However, due to their extremely low concentrations and high reactivity, direct observation of RO2* remains challenging. In this study, a compact instrument for in-situ measurement of RO2* was developed by combining the Peroxy Radical Chemical Amplification (PERCA) technique with Cavity-Enhanced Absorption Spectroscopy (CEAS). Using a pure HO2 standard for calibration, the system can achieve a chemical chain length of 30 and an optimal detection limit of 0.38 pptv (1σ, 3 min), enabling highly sensitive measurements of ambient peroxy radicals. The self-constructed PERCA–CEAS system was successfully deployed in a field campaign during autumn in Zhuhai to observe ambient RO2*. During the observation period, the mean daytime RO2* was 31.11 ± 18.87 pptv, which resulted in an average P(O3) of 14.41 ± 17.04 ppbv h−1. The comparison of O3 variation and derived P(O3) indicates that the daytime ozone enhancement in Zhuhai was primarily driven by local photochemical production, while regional transport acted mainly as an export effect. Our results demonstrate that a compact PERCA–CEAS system is capable of ambient RO2* measurements and suggest the need of diagnosing O3 formation pattern with the constraint of high time-resolution RO2* concentration.

Biomass burning organic aerosol (BBOA) is a major source of atmospheric brown carbon (BrC), which contributes to climate through solar radiation absorption. Chemical aging (bleaching) of BrC diminishes light absorption over time. Although recent modeling studies have emphasized the need to account for bleaching when assessing BrC's climate impacts, a parameterization representing the entire BBOA matrix has been lacking. Here, we develop a bleaching parameterization for the whole BrC in BBOA based on laboratory experiments that systematically varied temperature and relative humidity. The bleaching timescale increases under low-humidity and low-temperature conditions, likely due to enhanced aerosol viscosity. Implementing this parameterization in a global model increases the simulated direct radiative effect (DRE) of BrC by 1.5 to 2 times relative to previous estimates, with associated uncertainties exceeding 10% of the total organic aerosol DRE. The present scheme particularly shows elevated fresh BrC concentrations in boreal regions, suggesting that bleaching dynamics may influence not only radiative forcing but also snow darkening effects.

Abstract Barrier layers in the upper ocean suppress the upward entrainment of cold thermocline water, trapping heat and momentum near the surface and thereby influencing tropical air-sea interaction. However, the mechanisms governing their spatial and temporal variability are not fully understood. This study highlights the importance of salinity-induced vertical stratification in shaping the climatology and variability of the barrier layer in the Pacific Ocean on subannual and interannual timescales. Compared to observational and reanalysis data, coupled ocean-atmosphere models simulate a less eastward-extending warm pool, along with a thin barrier layer bias. This bias is linked to a saltier upper western Pacific and can be attributed to weaker precipitation and stronger easterly winds along the equator. Consistently, models with a more eastward-extending warm pool tend to exhibit a thicker barrier layer and lower salinity over the western Pacific. On interannual timescales, models agree with observational and reanalysis data that anomalous westerly winds and increased precipitation develop 10–13 months before the peak eastward shift of the warm pool eastern edge (WPEE), accompanied by an upper-ocean freshening. These anomalies peak with the WPEE shift and persist for another 9–10 months. Subannual variations exhibit more complex temporal patterns. Anomalous westerly winds and increased precipitation emerge 3–4 months prior to the WPEE peak extension and reverse rapidly one month after the peak. While models capture the timing and magnitude of wind and precipitation changes, they fail to reproduce subannual salinity variations. Improving salinity climatology and subannual variability in models remains essential for simulating barrier layers.

Abstract. Satellite retrievals of atmospheric greenhouse gas columns are used to obtain information about greenhouse gas sources and sinks by inverse modeling. Such an application requires high accuracy, as even small biases of the retrieved concentrations may result in large errors of the inferred rates of surface emissions (source) and deposition, surface uptake or removal in the atmosphere (sinks). For example, for the upcoming satellite mission dedicated to carbon dioxide monitoring (CO2M), co-funded by ESA and the European Commission for the Copernicus Programme, the accuracy of the dry-air column-averaged CO2 mole fraction (XCO2) is required to be better than 0.5 ppm. Here we investigate a potentially important systematic error source, namely XCO2 biases due to correlated sub-pixel variability of surface reflectance (albedo) and altitude. To minimize this error source we propose the use of an albedo-weighted surface altitude which better represents the satellite’s spatial sample than the unweighted average by using a linearized theoretical analysis. We use Copernicus Sentinel-2 data combined with Copernicus Digital Elevation Model (DEM) data and the Fast atmOspheric traCe gAs retrievaL (FOCAL) algorithm and create a variety of self-consistent experiments to test this theory. First, we conduct experiments with defined conditions and second, we apply the methodology to some real-world examples: the Bełchatów power plant in Poland, the Black Forest in Germany, the region around Mont Blanc in the European Alps and the whole country of Germany. In all these examples, we find that using the albedo-weighted average of the surface altitude reduces biases at locations with heterogeneous surface structure to values below the requirements for future satellite missions. In addition, we developed a possible post-processing equation to account for this process, because high-resolution albedo currently is not measured simultaneously with satellite instruments. We also find that filter parameters connected to the surface roughness might be relaxed when using the albedo-weighted surface altitude in the retrievals. Furthermore, we examine the dependence of the XCO2 errors on the size of the spatial samples and find that the errors become larger with larger spatial samples, but are generally smaller by more than a factor of four when using the albedo-weighted instead of the unweighted average of the surface altitude. In conclusion, we show that the use of the albedo-weighted surface altitude in the retrieval process results in significant reduction of the XCO2 bias compared to the use of the unweighted mean altitude, as currently used in most retrieval schemes.

Sun-induced chlorophyll fluorescence (SIF) is an effective proxy for vegetation photosynthesis, but tower-based retrieval suffers from atmospheric path interference under humid and variable conditions. We present a DOAS-based SIF retrieval algorithm that operates in Fraunhofer lines (680–686 nm, 745–758 nm) and water vapour-sensitive bands (717–727 nm). It constructs an adaptive reference spectrum from SCOPE simulations and PCA and incorporates H2O absorption cross-sections into the fitting process for active atmospheric correction. The algorithm is implemented in a dedicated tower-based system integrating a 1° scanning gimbal with a high-resolution spectrometer. Validation with simulated and field data demonstrates the following: (1) the algorithm retrieves SIF with high fidelity (correlation coefficients >0.9 across all windows); (2) it exhibits lower water-vapour sensitivity and greater cloudy-sky stability than FLD, 3FLD, and SFM, achieving the lowest coefficient of variation (CV = 0.356); (3) over a complete wheat–rice rotation, the retrieved SIF tracks crop growth and phenological stages. This work provides a reliable solution for automated, high-precision tower-based SIF observation under complex atmospheric conditions.

Abstract. Nitrogen (N) is a critical nutrient, that controls photosynthesis and decomposition processes. It is important to include the N cycle in the land component of climate models to improve the exchange fluxes of CO2 between land and atmosphere. We present here the implementation of the N cycle in the CNRM land surface model, namely ISBA. We evaluate the model on two Free-Air Enrichment (FACE), experiments sites: Duke and Oak Ridge. In particular, the response to elevated CO2 is studied. We compare the reference version without the N cycle (C) and the new version in which it is included (CN). A comparison to a multi model analysis shows encouraging results, since the computed NPP and N assimilation flux fall in the inter model range. The CN version performs better than the C version for NPP. Next, we focus on the carbon cycle by comparing simulation results to observations. The CN version improves the carbon stocks, largely overestimated by the C version. In particular, at elevated CO2, in the CN version, photosynthesis is downregulated by the N limitation. This yields a reduction of C accumulation in soil and biomass in comparison to the C version. In the literature, diverging strategies are observed to overcome N limitation. The model reproduces well the main features but fails to represent some sites characteristics. Finally, a detailed analysis of the simulated N dynamics is presented.

Abstract. Using 14 wind profiler radars and hourly records at 3000 weather stations during 2016–2020, 226 warm-sector heavy rainfall (WSHR) events were identified in Guangdong, South China. Five indices reflecting the vertical structure of the atmosphere were analyzed as precursor dynamical signals of WSHR occurrence – the Low-Level Jet Index (LLJI), Vertical Wind Shear (VWS), Atmospheric Lifting Intensity (ALI), and Boundary Layer Height (BLH). Notable fluctuations in signals were found 1–4 h before precipitation onset, and the sensitivity and occurrence mechanisms of WSHR in the three regions showed marked regional differences. In western Guangdong, LLJI increased sharply 1–2 h before onset and, although not linearly correlated with rainfall intensity, was systematically higher in strong-rainfall events. LLJI was also significantly correlated with upper-level (3–5 km) ascent, indicating that the strengthening of the low-level jet can directly enhance upward motion and thereby increase precipitation intensity. In central Guangdong, VWS at 0.5–1.5 km was significantly positively correlated with precipitation intensity across three consecutive pre-onset periods, which may reflect frictional deceleration of the low-level jet and the resulting enhancement of low-level convergence. In eastern Guangdong, anomalously weak upper-level ascent and higher convective available potential energy appeared before strong-rainfall events, indicating that suppressed mid-to-upper-level ascent allowed unstable energy to accumulate. Strong events were significantly warmer at 925 hPa, but showed no significant differences in large-scale vertical velocity or low-level relative humidity, suggesting that the instability accumulation was driven mainly by boundary layer warming; the anomalously high BLH is a manifestation of this thermal instability build-up.

Abstract Outgoing longwave radiation (OLR) from satellite observations is a key indicator of tropical deep convection and a valuable proxy for monitoring El Niño-Southern Oscillation (ENSO) evolution. This study evaluates the performance of OLR retrievals from Fengyun-3C (FY-3C), a Chinese polar-orbiting meteorological satellite, against the interpolated OLR product from the National Oceanic and Atmospheric Administration (NOAA) over the tropical Pacific Ocean for 2014–2019. The analysis focuses on distinguishing coupled and uncoupled El Niño events. To quantify the convection–sea surface temperature (SST) relationship, we developed a Coupling Strength Index (CSI) based on FY-3C OLR and Niño-3.4 SST anomalies. Temporal correlations between FY-3C and NOAA OLR exceed 0.98, confirming high overall consistency; FY-3C OLR faithfully captures convective responses to Niño-3.4 SST anomalies, especially during the strong 2015–2016 El Niño and 2017–2018 La Niña. Spatial correlations range from 0.76 to 0.97, with root-mean-square errors of 4.04–8.36 W m −2 , further demonstrating strong agreement with NOAA OLR, albeit with localized biases emerging during rapid convective transitions. CSI analysis reveals robust ocean–atmosphere coupling during events such as the 2015–2016 El Niño (CSI < −5 W m −2 °C −1 ) and identifies uncoupled El Niños characterized by SST warming without matching tropical convection (e.g., late 2014 and late 2018). Notably, CSI enables early detection of convection–SST decoupling, signaling OLR reversals roughly one month in advance. These results demonstrate that FY-3C OLR combined with the CSI provides a concise quantitative metric for ENSO air–sea coupling strength and improves operational monitoring of ENSO phase transitions.

High-resolution mapping of urban surface CO2 is essential for refined carbon monitoring, emission management, and low-carbon urban planning. Mobile monitoring provides dense street-level observations, but raw CO2 measurements are often affected by transient traffic disturbances, vehicle idling, and localized plume events, which limits their direct use as stable spatial mapping targets. This study developed an integrated framework for predicting, mapping, and interpreting stable surface CO2 patterns in Shenzhen by combining vehicle mobile observations, CSF processing, multiscale remote sensing predictors, machine learning. A CSF-based lower-envelope filter was used to suppress short-duration positive peaks and extract a more stable CO2 accumulation signal from mobile observations. Multiscale predictors representing transportation, urban activity, surface environment, and built form were constructed to characterize both local and surrounding urban contexts. Compared with raw CO2, the CSF-processed target substantially improved prediction performance. The best validation R2 across the candidate models increased from 0.59 to 0.90 in April and from 0.62 to 0.93 in November. The predicted maps identified persistent high-CO2 areas in central and southwestern Shenzhen. SHAP results showed that transport networks and urban activity reinforced surface CO2 accumulation, whereas vegetation and open-surface contexts weakened accumulation at broader spatial ranges. These findings provide an interpretable framework for high-resolution urban CO2 mapping and refined low-carbon governance.

Abstract Urban areas are a significant source of methane emissions, a potent greenhouse gas with a global warming potential 28-34 times stronger than carbon dioxide over a 100-year timeframe. Urban areas are complex environments where multiple types of collocated methane sources exist, many of which have not been extensively measured. In this work, we address a lack of measurement data from two biogenic urban methane sources: sewers and urban water bodies, using five years of mobile surveying data to guide on-site direct measurements. We analyze data from 56 vehicle-based measurement surveys conducted over five years in the Greater Toronto Area (GTA - Canada) from a cluster of high-emitting sewer covers and an urban engineered water canal (i.e., the Keating Channel), in addition to direct measurements from other urban water bodies (i..e, ponds and rivers). We use a static chamber methodology to directly measure methane emissions from 20 sewer covers and perform 150 methane flux measurements from rivers, ponds, and the Keating Channel. We recorded the highest methane emission rate ever directly measured from a sewer at 89 g h-1 of methane, which was located amongst a cluster of 13 sewer covers that collectively emitted 320 g h-1 of methane. We found that average methane emission fluxes from urban water bodies were highest from the Keating Channel at 14,000 (95% c.i.: 8,300 to 19,000) mg d-1 m-2 and urban ponds at 1,780 (95% c.i.: 40 to 3,500) mg d-1 m-2, both of which were characterized by high fluxes from ebullition rather than diffusion. The combined emissions from urban water bodies and sewers account for 6% (95% c.i.: 0 to 13%) of total methane emissions from the GTA. This work provides valuable empirical measurement data to better characterize the upper tail of emission rates from sewers and urban water bodies

Abstract PM 2.5 , as a key atmospheric pollutant severely threatening human health and sustainable development, has drawn increasing attention regarding the influence of urban structure on its concentration. This study focuses on 260 cities in China from 2011 to 2021. By integrating multi-source remote sensing data, reanalysis data, and statistical records, a high-resolution PM 2.5 concentration dataset based on AOD was constructed. By coupling the LightGBM machine learning model with the Geographically and Temporally Weighted Regression (GTWR) model, the study systematically reveals the influence mechanisms and spatiotemporal heterogeneity of urban spatial, economic, and social structures on PM 2.5 concentrations. The results indicate that population density (POP) in the social structure is the most critical factor influencing PM₂.₅, contributing 23.2%, with a notable threshold effect (586 persons/km²). Air pollutant emission (APE) and the night light index (NL) exhibit significant positive and negative influences, contributing 17.65% and 14.18%, respectively. An increase in the separation degree of construction land (IJI) contributes to reducing PM 2.5 concentrations. The GTWR model further reveals significant spatiotemporal heterogeneity among the influencing factors. The impact intensities of POP and IJI show a weakening trend over time, while those of APE and NL display fluctuating characteristics of first decreasing and then increasing. Spatially, the influences of POP and APE exhibit a pattern of increasing intensity from south to north and from central regions toward coastal and northeastern areas.

Inferring fault locking depth from interseismic geodetic observations is important to assessing seismic hazard. Interseismic deformation changes with time due to Earth’s viscoelastic rheology, but elastic models commonly employed for estimating locking depths assume time-invariant deformation. Consequently, the apparent locking depth depends on the time when the deformation is observed. In this study, with a focus on large strike-slip faults, we propose replacing the elastic models with an existing analytical viscoelastic model that incorporates time-dependent deformation while maintaining operational simplicity. Using InSAR observations around the Altyn Tagh Fault as illustrative examples, we explain the application of this approach and discuss how to address practical complications of real faults: (1) The use of this model requires knowledge of the earthquake recurrence interval and the time since the last earthquake in terms of the viscoelastic relaxation time. We show a trade-off relation between these two parameters that can be used to assess the ambiguity of inferred locking depth due to their uncertainties. (2) A stiffness contrast across the fault causes asymmetric deformation. We provide an empirical scaling relationship to allow the application of the laterally homogeneous viscoelastic model. (3) Shallow creep of the fault causes discontinuity of surface deformation. We explain that the rate of steady shallow creep without afterslip and slow slip events is controlled by the ratio of the widths (i.e., depth ranges) of the creep zone and locked zone, and we accordingly design a procedure to incorporate shallow creep. (4) Postseismic transients from nearby earthquakes can cause underestimation of locking depth. We demonstrate that modelling the postseismic effect using the same viscoelastic model can enable an effective correction.

Abstract. Prediction of shallow landslides at the regional scale generally relies on statistical analyses of landslide inventories. Rainfall-duration thresholds and susceptibility maps are among the most common approaches to anticipate future landslide occurrences. However, the outputs and reliability of these approaches can be strongly affected by the representativeness of the landslides included in the inventory. This study specifically investigates the impact of landslides triggered by an extreme rainfall event on the determination of rainfall-duration thresholds and susceptibility maps. We consider the case of Storm Alex, a millennial return period rainfall event, which hit the Alpes-Maritimes region (France) on 2 October 2020. The analysis is based on an inventory of 5383 shallow landslides, including 1656 landslides triggered by Storm Alex. Cumulative rainfall and rainfall duration associated with each landslide were computed following the process of the CTRL-T algorithm. Landslides sharing identical cumulative rainfall and rainfall durations were aggregated into a single point to avoid over-representing rainfall events that triggered many spatially clustered landslides. Then, a 5 % quantile regression was used to compute statistical rainfall-duration thresholds with and without the inclusion of Storm Alex landslides. A Random Forest approach was used to produce susceptibility maps under the same two configurations, which were subsequently compared. Results show that: (a) Including Storm Alex landslides increased the rainfall–duration thresholds by a factor of 1.1 to 1.5, depending on the assumed landslide occurrence time; (b) the exceptional rainfall intensity triggered landslides in areas having an initial lower susceptibility; and (c) including these events in susceptibility modelling alters the spatial distribution of susceptibility values. This study provides a quantitative analysis of the impact of landslides triggered by extreme rainfall events on statistical prediction models.

Marine Synechococcus is among the most widespread and productive autotrophs in the ocean, yet the quantitative role of mixotrophy in different lineages remains poorly constrained. Here, we compared organic nitrogen (urea and leucine) and carbon (glucose) utilization in nutrient-depleted versus nutrient-rich Synechococcus lineages by combining NanoSIMS-based single-cell measurements from field and laboratory incubations with omics analyses. Our findings revealed distinct mixotrophic strategies in different lineages. In nutrient-depleted lineages, elevated urea uptake supplied approximately 40–63% of the estimated total nitrogen demand. This pattern aligned with genomic evidence of enhanced urea transport, particularly the up-regulation of the high-affinity urea transporter DUR3 in low-nitrogen environments. In contrast, nutrient-rich lineages exhibited greater glucose uptake, although the amended organic substrates contributed only 2 to 4% to the estimated cellular carbon demand. These lineage-specific mixotrophic strategies underpinned niche partitioning in marine Synechococcus , refining our understanding of their trophic differentiation and its implications for marine biogeochemical cycling.

Abstract. In complex hydrological systems, flow path dynamics, water storage and mixing, and biogeochemical processing vary in space and may change rapidly during events. Understanding source areas, connectivity and short-term dynamics in stream water quality therefore requires high-temporal-frequency, multi-source observations both within and across catchments. Revolutions in field-deployable analysers and sensors, together with advancement in automation techniques, now make such observations feasible via true “labs-in-the-field”. This paper details the technical realisation and proof-of-concept for the Water Analysis Trailer for Environmental Research (WATER). The WATER is a mobile, trailer-based platform for environmental sensing and automated, high-temporal-frequency sampling and analysis of water from multiple (currently up to 11) sources. It is currently equipped to measure stable water isotopes, nitrate, electrical conductivity, pH and temperature, though its modular design supports the integration of additional measurement devices in the future. A field test in the 1.03 km2 Schwingbach Environmental Observatory, Germany, demonstrated the ability of the WATER to successfully and autonomously collect and analyse samples from six water sources (2 × stream water, 3 × groundwater, 1 × precipitation) over a period of six months, with collected data offering potential for new understanding of catchment functioning. Insights were also gained into the practical considerations necessary when deploying the WATER for an extended period of time, such as ensuring an adequate self-sufficient power supply and scheduling routine maintenance visits. Simulation of the reduced sampling frequency that would result from extending the WATER to sample at its full capacity of 11 sources also indicated that, over multi-month periods, key distributional characteristics of the collected data would likely be maintained. Overall, the WATER provides a mobile and scalable solution for high-temporal-frequency, multi-source hydrological and hydrochemical monitoring that can be (re-)deployed in different locations or targeted to specific events.

With the rapid advancement of high-resolution remote sensing technology, Remote Sensing Semantic Change Detection (RS-SCD) has emerged as a core technology for monitoring dynamic Earth surfaces. RS-SCD enables simultaneous localization of changed regions and inference of their ‘From-To’ class transitions, overcoming the semantic ambiguity limitations of traditional binary change detection. In this systematic review, we first formalize the RS-SCD task and present a conceptual probabilistic perspective to elucidate the structural coupling between change detection and semantic segmentation. Moving beyond conventional architecture-centric surveys, we propose a task-oriented taxonomy based on five semantic-change coupling strategies: sequential decoupling (primarily CNN-based), parallel coupling (Transformer and Mamba architectures), relational coupling (graph neural networks), cross-modal fusion (optical, SAR, and vision-language modalities), and partial-evidence inference (weakly supervised, self-supervised, and foundation-model fine-tuning). For each strategy, we analyse representative models, highlighting trade-offs among accuracy, efficiency, and robustness. We further discuss practical bottlenecks, including pseudo-change suppression, annotation scarcity, and limited generalization, and outline future trends towards large-model-driven RS-SCD that is generalized, all-weather, and trustworthy. This review provides a systematic, conceptually grounded reference for both research and operational deployment of RS-SCD.

Mangrove canopy height (MCH) is a fundamental structural variable for monitoring ecosystem health and quantifying carbon stocks. However, MCH retrieval from optical satellite imagery is often constrained by spectral saturation in dense stands and environmental noise in intertidal zones. This study investigates the utility of kernel-based spectral features (KVIs) as non-linear topological enhancements for MCH estimation by integrating GEDI spaceborne LiDAR with Sentinel-2 and Sentinel-1 data across three mangrove ecosystems along the South China coast. Utilizing four regression models under spatial cross-validation, we evaluated the performance of traditional indices, KVIs, and integrated feature sets against GEDI reference measurements. Results indicate that traditional optical indices exhibit limited linear sensitivity to MCH. Rather than serving as universal accuracy boosters, KVIs function as non-linear stabilizers by redistributing spectral values in Hilbert space, which effectively enhances feature representation in high biomass stands and mitigates background noise. Furthermore, model comparisons reveal that while traditional indices yield competitive baseline accuracy in specific architectures (e.g., 1D-CNN), kernel-based features provide nuanced advantages in spatial stability and ensemble dispersion reduction. Ultimately, this study demonstrates that kernel-based features enhance model robustness under rigorous cross-validation, providing a reliable structural foundation for large-scale ecological monitoring and regional carbon dynamics assessments in heterogeneous coastal environments.

Pixel-level annotation of remote sensing imagery is costly, motivating weakly supervised semantic segmentation (WSSS) using only image-level labels. However, class activation maps (CAMs) often highlight only discriminative sub-regions and fail to separate adjacent land-cover regions, particularly in remote sensing scenes characterized by densely co-occurring land-cover classes and substantial variations in object scale. To address these limitations, we propose a three-stage framework that integrates complementary priors from Contrastive Language–Image Pre-training (CLIP), Self-Distillation with No Labels version 2 (DINOv2), and the Segment Anything Model (SAM). First, a lightweight CLIP adapter aligns vision–language priors with remote sensing imagery, while sigmoid-based multi-label decoupled distillation replaces class-competitive distillation with independent class-wise supervision, producing more complete CAMs. Second, DINOv2-guided feature clustering decomposes large merged regions before SAM prompt generation, while Spatial–Semantic Constraints are used to construct confidence-guided point-and-box prompts and reject excessively expanded or semantically inconsistent masks, thereby generating reliable pseudo-labels. Finally, a compact segmentation network is initialized with the weights learned in Stage 1 and retrained using the refined pseudo-labels generated in Stage 2, eliminating the need for foundation models during inference. Experiments on the Potsdam, LoveDA, and DeepGlobe datasets show that the proposed method achieves mean intersection over union (mIoU) scores of 53.16%, 52.66%, and 62.98%, respectively, outperforming state-of-the-art WSSS baselines by 6.55, 1.16, and 1.27 percentage points, respectively. These results demonstrate the effectiveness and generalizability of the proposed framework across diverse remote sensing scenarios under image-level supervision.

In Deep Learning Remote Sensing, data quantity is rarely the limiting factor. A single high-resolution satellite image can yield thousands of training patches. What determines model performance, yet remains largely overlooked, is the quality of those patches. To date, the choice of sampling method has rarely been treated as a methodological decision. Conventional approaches, namely sliding-window and random sampling, introduce two compounding data-quality problems: severe class imbalance caused by the overproduction of background-only patches and negative learning arising from incomplete annotations, where unlabeled objects are implicitly treated as negative examples during training. To address these limitations at the data construction stage, we propose object-centric patch sampling, a model-independent strategy that anchors each training patch to the geometric centroid of an annotated object. This design ensures that every object-anchored patch contains at least one target instance and substantially reduces exposure to unlabeled regions that generate false-negative supervision signals; only a small, deliberately controlled proportion of background-only patches is retained to preserve contextual variety without reinstating background dominance. The method is evaluated on three heterogeneous remote sensing datasets spanning satellite (Sentinel-2, 10 m), aerial (NAIP, 1 m), and UAV (0.25 m) imagery, covering cotton field segmentation, rural building extraction, and water body delineation, respectively. Using a U-Net architecture under identical training conditions, the proposed approach achieves IoU scores of 0.929, 0.896, and 0.912 on the three datasets, respectively, outperforming sliding-window sampling by up to 19.6 percentage points in IoU and consistently delivering higher F1-scores across all experimental configurations. Evaluation under DeepLabV3+ gives a mean IoU of 0.9504 and a mean F1-score of 0.9772 in a multi-class segmentation task, indicating that the gains are not specific to a single architecture. Unlike model-level solutions such as focal loss or class reweighting, the proposed method improves training data quality at its source and integrates seamlessly into any deep learning pipeline without architectural modifications.

Abstract Dissolved oxygen (DO) is crucial for maintaining ecological integrity in tidal river tributaries. However, the relative roles of physical, chemical, and biological factors in urban tidal rivers remain unclear. In this study, controlled laboratory experiments were conducted to investigate the mechanisms by which these factors and their interactions influence DO dynamics in an urban tidal river. Temporal variations in DO and three nitrogen species were analyzed, and the implications for low-oxygen control were evaluated. The results indicate that physical factors mainly determine the background trajectory and overall range of DO variation. Higher flow velocities increase DO concentrations and accelerate recovery. Salinity showed no clear monotonic effect across the low-to-moderate range tested, whereas the highest salinity treatment produced lower DO levels. Among the chemical treatments, NH₃-N removal maintained substantially higher DO than COD removal or raw-water conditions. Sediment addition caused rapid DO depletion under both chemical and flow treatments, followed by only partial recovery. The influence of biota on DO was condition-dependent and became clearer mainly under high-temperature and still-water conditions. These findings suggest that prioritizing NH₃-N reduction, controlling sediment-related internal loading, and moderately enhancing mixing and reaeration are effective strategies for increasing and stabilizing DO levels.

Study region The study investigates the Corabia – Turnu Măgurele sector (rKm 601 – rKm626) of the Lower Danube River in Romania. This is a hydromorphologically dynamic area of regional importance for navigation. Spatial and temporal dynamics modify its hydrodynamic behaviour, sediment transport, and channel morphology, changing it into a critical navigation sector. Study focus Modern hydrographic and acoustic survey equipment was used to describe the study area's hydromorphological characteristics. The synergistic use of MBES, ADCP, and SSP delineated five morphological areas grouped into three broad classes: shallow-water zones associated with sandbar formation; moderate-depth sections characterized by relatively stable hydrodynamic conditions and requiring long-term monitoring; and deep-water zones within the main navigable channel, characterized by enhanced erosion and sediment transport. Principal component analysis, K-means clustering, and spatial principal component analysis were used to classify areas with different hydromorphological characteristics. The analysis offers insights into water flow patterns and areas for this regional critical area where the sedimentation rate is excessive. New hydrological insights for the region The integrated analysis identified water-flow patterns and zones of excessive sedimentation. The bathymetric surveys and ADCP measurements were complemented by sediment suspended profiling. The result reveals that the majority of sediment particles fall within the 100–1000 µm range (fine to medium sands). A dominant fraction (100–300 µm) underscores the prevalence of fine sand, indicative of moderate-energy depositional behavior. The low content of fine particles (<10 µm) reflects limited silt and clay retention. The statistical analysis (PCA, K-means, spatial PCA) grouped datasets into three types of areas with different hydro-morphological characteristics. Under extreme low-flow conditions, water depths within the navigable channel often locally drop below 2 m, imposing operational limitations on vessel draught and increasing the risk of navigation disruption. The proposed approach provides a valuable method to understand the importance of the synergistic application of bathymetric, hydrodynamic, and sedimentological analysis, contributes to increasing the level of knowledge on sediment dynamics in regional river systems, such as the Lower Danube, and provides support in decision-making regarding the prioritization of dredging interventions.