Atmospheric and Oceanic Sciences

Deep learning methods that integrate convolutional neural networks (CNNs) and Transformers have achieved remarkable progress in hyperspectral image (HSI) classification. However, existing methods still suffer from insufficient multi-scale spatial–spectral feature modeling, a lack of efficient interaction mechanisms between local and global features, and the inherent high computational complexity and redundant information of Transformers, which limit model performance. To address these issues, a Multi-Scale Spatial–Spectral Convolutional Hybrid Transformer model (MS3CHFormer) is proposed in this article. Specifically, a Multi-Scale Spatial–Spectral Convolution Module (MS3ConvM) is first constructed. Through a multi-branch and multi-receptive-field design, it jointly models spatial and spectral features at different scales, thereby enhancing the representation capability of complex ground objects. Then, a Token-Selective Sparse Transformer Encoder (TSSTE) is designed, which adaptively selects tokens and performs sparse modeling via a Dynamic Correlation-Aware Attention (DCAA) mechanism, effectively reducing computational complexity while suppressing redundant information and further reinforcing key feature representations. Furthermore, a Local–Global Feature Fusion Module (LGFFM) is designed to achieve deep complementary fusion of CNN and Transformer features by mapping them into different representation spaces. Finally, a Detail-Preserving Enhancement Module (DPEM) introduces original detail information through residual connections to compensate for detail loss in high-level semantic representations, thereby enhancing the representation capability of boundaries and fine-grained structures. Experiments and comparative analyses on four public HSI datasets demonstrate that the proposed MS3CHFormer outperforms state-of-the-art methods and achieves superior classification accuracy under limited training samples, exhibiting excellent robustness and generalization ability.

Solute transport in rock fractures is strongly influenced by hydrodynamic conditions, and clarifying the Péclet-number-controlled transition of transport regimes is important for understanding contaminant migration in fractured aquifers. Based on three-dimensional numerical simulations, this study investigates conservative solute transport in idealized rough fractures with perfectly mated walls and uniform aperture under a wide range of Péclet numbers (Pe). The evolution of concentration fields, breakthrough curves (BTCs), and diffusive and advective fluxes was analyzed to identify the dominant transport regimes. The results show that, as Pe increases, solute transport changes from a diffusion-dominated regime (Pe < 0.1), to a mixed macro-dispersion-dominated regime (0.1 < Pe < 1000), and finally to a high-Pe advection-controlled regime with Taylor-dispersion-like characteristics (Pe > 1000). Correspondingly, the concentration field evolves from rapid diffusion-driven spreading to a sharper advective front, while the BTCs change from early diffusion-breakthrough curves to step-like breakthrough behavior. Fracture aperture promotes solute spreading and broadens the mixing zone, especially under low-to-intermediate Pe conditions. In contrast, under the perfectly mated and uniform-aperture fracture conditions considered here, increasing roughness mainly induces local tortuosity of the concentration front and has limited influence on the overall BTCs. Flux decomposition further confirms that diffusive flux dominates at low Pe, whereas advective flux becomes increasingly dominant as Pe increases. These findings provide a mechanistic basis for interpreting Pe-controlled solute transport in idealized fracture channels and offer fracture-scale insights for classified groundwater contamination risk assessment. The implications should be interpreted within the assumptions of conservative transport without matrix diffusion, adsorption, or reactive processes.

In Part III of the series, we evaluate the accuracy and applicability of the tomographic algorithm introduced in Part I and applied to real measurements by the research scanning polarimeter in Part II. We focus on the core part of the algorithm, producing a nested family of cloud shapes corresponding to a range of brightness thresholds. This family is then used to derive a 2D field of cloud extinction coefficient. We relate the resolution of the multi-angle measurements to the spatial accuracy of the cloud shape retrievals and determine constraints on the cloud aspect ratios required for the applicability of the algorithm. The expressions for overpass length and time derived in this study allow for estimating how much the cloud can move or change during the measurement process. We estimate biases in cloud size and position retrievals caused by the cloud’s advection during the measurements. Our accuracy estimation techniques are applied to previously published examples of clouds, both simulated and real.

Remote sensing change detection (CD) aims to localize land-surface changes from bi-temporal imagery and plays an important role in applications such as urban monitoring, disaster assessment, and environmental analysis. In high-resolution scenarios, CD performance is often degraded by cross-temporal appearance inconsistency, large variations in target scale, and boundary ambiguity introduced during multi-level decoding. To address these challenges, we propose MIGA-Net, an end-to-end framework that jointly models spatio-temporal interaction, adaptive multi-scale context aggregation, and hierarchical boundary refinement. Specifically, the Spatio-Temporal Graph Interaction Module (ST-GIM) combines interactive attention and graph reasoning to suppress pseudo-changes caused by illumination or seasonal shifts; the Adaptive Gated Context Pyramid Module (AGCP) performs content-driven scale selection and regulates context injection through a gated residual mechanism to reduce noise amplification; and the Hierarchical Boundary-Aware Refinement Module (HBAR) integrates semantic channel filtering and explicit boundary attention for progressive contour recovery. Experiments on LEVIR-CD, WHU-CD, and SYSU-CD demonstrate that MIGA-Net achieves F1 scores of 91.84%, 92.52%, and 82.92%, and IoU scores of 84.91%, 86.08%, and 70.83%, respectively. The proposed method yields consistent improvements in both quantitative metrics and structural boundary quality, indicating its effectiveness for robust pseudo-change suppression and structurally faithful prediction in high-resolution remote sensing CD.

With the elimination of absolute poverty in China, promoting coordinated regional development has become a key agenda. However, the county-level resolution of the socioeconomic development dataset released by the statistics yearbook has limited its use at the local level. Thus, benefiting from fine-grained observation of remote sensing data, we propose a novel socioeconomic downscaling framework accounting for integrated spatial-attribute proximity, namely the Multi-Distance Geographically Neural Network Weighted Regression (MD-GNNWR). Taking the Yangtze River Delta (YRD) as a case study, we first construct a Multi-dimensional Relative Development Index (MRDI) to reflect comprehensive development, by integrating living standards, education, and health dimensions. Subsequently, based on multi-source remote sensing data, i.e. night-time light (NTL), land cover, road networks, points of interest (POI), and terrain data, we develop the first estimates of MRDI for the township-level and for a 1-km grid. Results show that the MD-GNNWR model with spatial proximity achieves R2 = 0.851, over 0.1 higher than the classical Random Forest (RF) model. The township-scale MRDI is significantly correlated with a survey-derived wealth index (Pearson’s r=0.60), while comparison with an external Human Development Index (HDI) product further supports the consistency of the grid-scale MRDI, with Pearson’s r>0.70. We also illustrate how these data can improve decision-making. The grid-scale MRDI Gini coefficient of Anhui Province reaches 0.429, highlighting pronounced inequality and the urgent action. Geographical detector analysis shows bivariate enhancement among variables, with average NTL intensity (ANTL) contributing most (q-value = 0.630). This framework enables fine-scale monitoring of regional development, supporting the identification of spatial disparities and informing coordinated, sustainable strategies.

The calibration accuracy of key weather radar antenna parameters, including beam pointing, beamwidth, and antenna gain, directly affects quantitative precipitation estimation (QPE) and multi-radar network products. Conventional calibration approaches such as external field beacons and far-field tests are often constrained by site conditions and high implementation costs, making them difficult to apply routinely in operational radar networks. To address this limitation, this study proposes a robust solar calibration method for key antenna parameters of weather radars based on a dedicated Volume Coverage Pattern for Sun calibration, hereafter referred to as VCPSun. The proposed method uses a high-density solar scanning strategy with midpoint time alignment and feed-forward control of solar apparent motion. Combined with solar sample identification, propagation path correction, two-dimensional Gaussian surface fitting, and deconvolution of solar-source broadening and scan-smearing effects, the method enables reliability retrieval of beam pointing, beamwidth, and antenna gain. A high-frequency intensive observing experiment was conducted using a China New Generation Weather Radar, model SA-D (CINRAD/SA-D), deployed at the Changsha Meteorological Radar Calibration Center, with independent far-field test results used for validation. The results show that the retention rate of quality-controlled solar samples reached 85.7%, supporting stable reconstruction of the main-lobe power pattern. The retrieved mean beam pointing biases for both polarizations were within ±0.05°. After correction, the relative differences in beamwidth with respect to far-field measurements were respectively 3.26% and 1.52% for the H-polarization azimuth and elevation directions and 2.09% and 1.84% for the V-polarization azimuth and elevation directions, with the overall mean relative difference being less than 3.5%. The antenna gain differences relative to the independent far-field reference values were within 0.2 dB, at −0.062 dB for H-polarization and −0.144 dB for V-polarization. Comparative analysis with historical one-dimensional SunCheck records and an ablation test of the beamwidth correction chain further demonstrate that high-density two-dimensional sampling and physical deconvolution corrections improve the robustness and quantitative accuracy of the solar-based retrieval. These results demonstrate the feasibility of reliable in situ calibration of key antenna parameters for operational weather radars. The proposed method provides a potential technical pathway for in situ quantitative assessment of antenna performance in S-band CINRAD/SA-D radars, although further validation using additional radars and longer observation periods is required prior to network-wide application.

Lithological mapping with multispectral remote sensing remains challenging when diagnostic spectral information is limited and reliable labeled samples are scarce. This problem is particularly relevant when convolutional neural networks (CNNs) are applied to lithological classification, because limited spectral dimensionality and scarce training samples may hinder the learning of discriminative spatial–spectral features. In this study, we developed a limited-sample lithological mapping framework for the Shibaocheng area of Subei County, Gansu Province, China, using band-integrated ASTER and Sentinel-2A multispectral imagery. ASTER shortwave infrared (SWIR) bands were co-registered and resampled to Sentinel-2A imagery, and then integrated with Sentinel-2A visible and near-infrared (VNIR) and red-edge bands to construct a complementary multispectral dataset. A compact spectrally enhanced multi-scale CNN was designed, incorporating a residual spectral feature enhancement module for inter-band representation learning and a parallel multi-scale hybrid convolution module for capturing spatial–spectral features. Eight lithological units were classified under limited-label conditions using 8158 training samples and 3497 spatially independent validation samples. Experimental results show that the band-integrated ASTER–Sentinel-2A dataset improved classification performance compared with single-sensor inputs. Using the proposed model, the band-integrated dataset achieved an overall accuracy (OA) of 94.12%, average accuracy (AA) of 94.04%, and Kappa coefficient of 0.932, compared with OA values of 93.14% and 92.40% obtained using ASTER and Sentinel-2A alone, respectively. The positive effect of band-level integration was also observed for spectral angle mapper (SAM), support vector machine (SVM), and 3D-CNN, whose OA values increased to 54.33%, 86.12%, and 92.29%, respectively. The proposed CNN achieved the highest OA among the evaluated methods, outperforming SAM, SVM, and the conventional 3D-CNN. In addition, t-SNE visualization indicated that incorporating spatial texture features produced more compact and better-separated lithological clusters than using spectral features alone. Ablation experiments further demonstrated that the proposed spectral feature enhancement and multi-scale hybrid convolution modules each contributed to improving lithological classification performance. These results demonstrate that integrating freely available multispectral data with a lightweight spectral–spatial CNN provides a practical and cost-effective solution for lithological mapping in bedrock-exposed arid to semi-arid regions, especially where hyperspectral imagery and dense field samples are unavailable.

Mangroves are widely promoted as coastal bio-shields, yet their capacity to stabilize shorelines in sediment-deprived, high-energy deltas remains poorly quantified at fine spatial scales. This study addresses whether sustained mangrove cover actively mitigates long-term shoreline erosion, or whether these ecosystems instead function as passive indicators of underlying geomorphic change, in the Muthupet wetland (1995–2025). Using Landsat imagery, the Digital Shoreline Analysis System (DSAS), and the Combined Mangrove Recognition Index (CMRI), shoreline movement was coupled with localized biological cover change through Thiessen polygon interpolation. Over three decades, the coastline was highly unstable and retreating, with over 76% of transects actively eroding under sediment deprivation. Contrary to the bio-shield assumption, segments supporting denser mangrove cover in 1995 underwent the greatest landward retreat, whereas sparsely vegetated transects remained comparatively stable, and sustained cover did not significantly reduce shoreline fluctuation. The strong synchronization between physical shoreline retreat and habitat contraction (rs = 0.611, p < 0.0001) is consistent with the Muthupet mangroves functioning as passive indicators of coastal instability rather than active stabilizers. Consistent with observations from other sediment-deprived deltas such as Mekong, these findings necessitate a shift in regional coastal management toward a geomorphology-first framework that prioritizes sediment–budget restoration over vegetation planting alone.

The rapid cryospheric transformation of the Arctic presents complex challenges for global supply chains, necessitating data-driven decision-making frameworks to optimize route planning and resource allocation. Standard aggregate statistics fail to capture the structural nuances required for operational research models in this evolving landscape. This study analyzes the spatiotemporal dynamics of Arctic navigation intensity to determine the extent of the operational window expansion and the spatial displacement of shipping routes. We deployed the Arctic Traffic Intensity Framework (ATIF) to generate a high-resolution spatiotemporal dataset spanning 2012–2024. To address variance instability and inform strategic forecasting, a log-linear regression model was applied to calculate the Annual Percentage Change, allowing for a dual analysis of absolute linear trends (volume) versus relative growth (proportional intensity) across heterogeneous baselines. A bilateral expansion of the high-intensity window to nearly five months (June–October) was observed, driven by earlier spring break-up and delayed autumn freeze-up. Spatially, the geometric navigational centroid has shifted southwestward, highlighting a concentration of activity in the Barents and Kara Seas rather than a uniform Transpolar dispersion. It can be concluded that the Arctic shipping system has transitioned from a seasonally restricted frontier to a standardized resource extraction corridor. However, the traffic is heavily clustered in the western sector, creating high-density logistic bottlenecks rather than a homogenized international transit route.

Abstract. The molecules NO2, O3, OClO and BrO play a major role in the photochemistry of stratospheric ozone, notably in the formation of the springtime Antarctic ozone hole. For this reason, these species have been monitored by Differential Optical Absorption Spectroscopy (DOAS) instrumentation for many decades. To transform DOAS Slant Column Densities (SCDs) into Vertical Column Densities (VCDs), independent of the viewing geometry, the Air Mass Factors (AMFs) relating these quantities are needed. Ground-based stratospheric trace gas measurements are performed in zenith-viewing geometry at twilight, around and beyond 90° solar zenith angle (SZA). At those solar angles, the Earth's sphericity and the rapid changes in photochemical parameters (e.g., photolysis rate coefficients) affect the calculation of the AMFs, particularly for photochemically active species such as NO2, OClO and BrO. This study presents a methodology to infer AMFs that account for sphericity and photochemical effects. We estimate stratospheric AMFs of NO2, O3, OClO and BrO for Belgrano and Marambio Antarctic stations using the MYSTIC (Mayer, 2009; Emde et al., 2010) Radiative Transfer Model (RTM). The photochemical changes taking place during twilight are considered using a photochemical box-model based on the SLIMCAT chemistry transport model (Chipperfield, 1999, 2006). Vertical profile concentrations obtained by this model are “averaged” over the optical paths. That is, for each SZA observed at the station, a vertical concentration equivalent to all the concentrations encountered by the solar beams in different parts of the atmosphere is calculated, considering the different “local” SZAs and the partial optical paths in each layer. These concentration profiles, representative of a complete two-dimensional atmosphere, are then used as input for the one-dimensional fully-spherical version of MYSTIC RTM. The robustness of the proposed methodology is tested against measurements of NO2, O3, OClO and BrO SCDs obtained at Marambio Antarctic station. A good agreement is observed between modelled and measured values of NO2, O3 and OClO SCDs. For BrO, larger differences are obtained but they have been attributed to the tropospheric BrO contribution that has not been included in the model. Our results for Marambio 2018 show that monthly averaged AMFs can be considered as a good approximation for O3 and BrO, but more temporally resolved sampling is recommended for NO2 and especially OClO during July, probably due to vortex dynamics above that site. This work shows the large impact that photochemistry and Earth's sphericity can have on both the magnitude and the SZA dependence of the AMFs during twilight.

ICESat-2 ATL08 is an important data source for global land surface elevation monitoring, while its accuracy has not been systematically evaluated in the complex terrain areas of central and southern China. Taking high-resolution digital surface models as reference data, this study carries out systematic verification with a total of 949 valid verification points covering 18 typical geomorphological areas in central and southern China. The verification sites cover various terrain types including plains, hills, mountains and alpine canyons. The results show that the average root mean square error of all sites is 3.318 m, ranging from 1.044 m to 5.120 m. Among them, plain areas have the highest accuracy (HS, RMSE = 1.044 m), followed by hilly areas with RMSE of approximately 1.610–3.871 m, and mountainous and alpine canyon areas show relatively poorer accuracy with RMSE of approximately 2.374–5.120 m. The overall mean error (ME) is −1.032 m, with ME values ranging from −4.575 m to +2.548 m across sites. The accuracy of ICESat-2 ATL08 in central-southern China is highly terrain-dependent: RMSE is 1.044 m in the plain site and ranges from 1.610 to 3.871 m in hilly areas and from 2.374 to 5.120 m in mountainous and alpine canyon areas. Therefore, users should consider terrain complexity when applying this product, and post-processing correction incorporating topographic information is recommended for alpine canyon areas where RMSE exceeds 5 m.

Abstract Urbanization can negatively impact water quality, threatening key community resilience benefits including ecological health, recreation, and cultural significance. In Austin, Texas, urbanized central and eastern watersheds have poorer water quality and fewer historical protections than rural western watersheds. Boggy Creek, in East Austin, is amongst the most environmentally degraded watersheds in Austin, yet community members report that fifty years ago it was a valuable high-quality amenity. Our objective is to investigate the processes and sources of water quality degradation in Boggy Creek and identify potential remediation strategies to improve environmental health. Monthly stream water monitoring shows that elevated levels of anthropogenic indicators (e.g., Na, Cl, Escherichia coli [E. coli], and 87Sr/86Sr isotope ratio) in stream water in the upstream sections of Boggy Creek are correlated with the spatial distribution of older infrastructure, revealing more pronounced effects of urbanization and municipal water input than in the downstream areas. Unexpected high discharge occurred during drought conditions at a previously dry upstream site. Water quality parameters for the high discharge (pH=9.2, total dissolved solids [TDS]=264 ppm, E. coli=1.0 Most Probable Number [MPN]/100 mL, Ca=12 ppm, and SO4=31 ppm) align closely with that of the municipal supply water within the watershed (n=3, pH=9.3-9.5, TDS=245-266 ppm, E. coli=<1.0 MPN/100 mL, Ca = 11-14 ppm, and SO4 = 29-31 ppm). This indicates that essentially 100% of streamflow was from municipal supply leaks. At a second upstream site, E. coli concentrations markedly increased from 5.2x102 MPN/100 mL in April 2024 to >2.4x104 MPN/100 mL in May and June 2024, indicating that streamflow was predominantly comprised of municipal wastewater. Through collaboration with the city’s Watershed Protection Department, a leak in a private wastewater line near the sampling site was identified and repaired, which led to a significant decrease in E. coli concentrations in September 2024 to 29 MPN/100 mL, which is below Texas’s health guideline levels for water recreation. Our collaborative research approach has value for advancing the resilience of this water resource and can be implemented in other urban hydrologic systems with aging infrastructure.

Estuaries are the transition zones between rivers and the ocean, and act as final buffer zones for plastic pollution before entering the sea. Plastic transport and retention in estuaries are the result of a complex interplay between tidal dynamics, freshwater discharge, and estuary characteristics. Despite its importance, net plastic transport between rivers and the sea is poorly understood and quantified. Here, we show that the net plastic transport in estuaries is highly variable over time and between rivers. We combined plastic concentration data and simulated discharge data to estimate plastic transport for a one-year period in the Saigon, Vietnam, and a five-month period in the Rhine, Netherlands. We estimated the tidal-cycle averaged net plastic transport. We defined the delivery ratio as the relative net transport over a full tidal cycle (ebb and flood phases), ranging from 1 (all plastic transport to sea) to -1 (all plastic transported upstream). We found that the delivery ratio varied between -0.87 and 1.0 for the Saigon and -0.05 and 0.84 for the Rhine. Negative delivery ratios indicate that estuaries can have a net upstream transport of plastics during certain periods. We quantified the uncertainty stemming from incomplete monitoring of full tidal cycles. We demonstrate that the uncertainty for individual tidal cycles remains large, even at 90% coverage of full tidal cycles. Our results emphasize the importance of reliable observations for full tidal cycles. We anticipate our study to provide guidance on improved monitoring, understanding, and accounting of plastic transport and retention in estuaries.

Trophic transfer efficiency (TTE) describes the proportion of energy or nutrients transferred from one trophic level to the next. A common assumption holds that ~10% of energy is transferred upward. This rule of thumb has shaped ecological models and discussions of sustainability, but broad empirical tests are lacking. We compiled 2052 TTE estimates from 122 studies across ecosystems. Average energy transfer efficiency (TTE e ) was 5.92%, well below 10%, while nutrient transfer efficiency (TTE n ) averaged 11.13%. Marine ecosystems had the highest TTE e (8.13%), followed by freshwater (5.53%) and terrestrial (1.52%) systems. TTE e declined with temperature in freshwater and was lower for consumers feeding on autotrophs, endotherms, and higher trophic levels. Our findings challenge long-standing assumptions and highlight the need for better understanding of TTE variation.

Abstract Inflows of phosphorus (P) to freshwaters from the wastewater and agriculture sectors of our food systems continue to reduce aquatic biodiversity and threaten human health. A range of analytical tools were applied to investigate the contributions of these two sectors to riverine P pollution in the catchments of the Somerset Levels and Moors (SLM) region of England. A substance flow analysis of current sector P use in 2021 identified a low food system P use efficiency (48–75%) and variable P input pressures of unused P of 2.1–8.5 kg ha −1 year −1 driven by livestock feed imports and a high population density. Despite recent reductions in wastewater P discharges, concentration (C) and flow (Q) analysis found both point source and diffuse source signals in the river P record. River soluble reactive P (SRP) and total P (TP) flux in a range of SLM sub-catchments ranged from 0.5–1.5 and 0.75–2.4 kg ha −1 year −1 , respectively. The flux of SRP and TP associated with diffuse sources (calculated by CQ analysis) averaged 0.3–0.5 and 0.5–0.95 kg ha −1 year −1 , respectively, and was significantly positively correlated ( r 2 0.6, p < 0.01) to the agricultural P surplus in these sub-catchments. The large intercept (0.28 kg SRP ha −1 and 0.56 kg TP ha −1 ) of this relationship suggested that the historical legacy P store was the main pollution threat from agriculture. An inventory of agricultural P inputs and P outputs over the last 150 years together with a catchment soil analysis programme confirmed a legacy of soil P accumulation amounting to an average 2 t ha −1 . An analysis of the potential impact of reductions in the sector P input pressure on river P concentrations suggested that on-going and planned increases in wastewater P removal efficiency must be supplemented with a major system change towards drawdown of legacy soil P reserves to lower the P pollution threat in this nationally important region.

The Aura Microwave Limb Sounder (Aura/MLS) measures temperature profiles with a horizontal spacing of about 170 km along its near polar orbit. We highpass-filtered the horizontal temperature fluctuations along the suborbital track in the middle atmosphere. The characteristics of inertia gravity waves with horizontal wavelengths between 200 and 825 km are evaluated for the equatorial region (10∘ S to 10∘ N), northern polar region (70∘ N to 82∘ N), and southern polar region (70∘ S to 82∘ S) over the time interval from August 2004 to December 2021. A modulation of gravity wave activity by quasi-biennial oscillations is present in the equatorial stratosphere but not in the equatorial mesosphere. The gravity wave activity in the southern polar mesosphere is stronger by a factor of up to 2 than in the northern polar mesosphere. The seasonal variation in the vertical structure of gravity wave activity shows strong interhemispheric differences. There are double layers of enhanced gravity wave activity in the upper mesosphere over Antarctica in the summer and the winter, while the northern polar region does not show a double layer structure of gravity wave activity. In the northern polar region, upper mesospheric gravity wave activity is decreased after the onset of major sudden stratospheric warmings.

Improving the spatial resolution of thermal imagery is essential for agricultural field management, especially in developing countries where fields are fragmented and heterogeneous. The existing downscale land surface temperature (LST) products are a promising solution while awaiting for further advances in satellite-based thermal sensors. In this study, a linear regression-based approach was proposed to downscale LST from 1 km to 10 m spatial resolution. The proposed method is hereafter referred to as Stratified and Adaptive Regression for Land Surface Temperature Downscaling (STAR-LST). Taking advantage of the linearity between LST and NDVI under homogeneous conditions, multiple linear regression models were first derived for different regions in the (LST, NDVI) feature space at coarse resolution. These regions were defined by partitioning the triangular form of the (LST, NDVI) feature space into sub-triangles, sharing a common top vertex. Each model is then applied to the corresponding NDVI range, allowing thus to derive LST at finer resolution. LST estimates were validated against in-situ measurements collected over olive trees, Kernza, wheat, and barley. Good results were obtained with RMSE values ranging from 2.79 to 4.49 °C for STAR-LST. This presents an improvement of an average of 33 % compared to the two classical methods DisTrad and TsHARP (RMSE in the range 4.18–6.83 °C), which showed similar accuracy. In addition, STAR-LST enabled daily LST estimation whenever coarse-resolution data were available, using interpolated NDVI to generate a high spatio-temporal LST product suitable for managing agricultural fields, particularly the small size fields.

Urban Heat Islands (UHIs) and intensifying heat stress in cities pose growing risks to public health, outdoor thermal comfort, and climate resilience. Shade is a low-energy and cost-effective strategy for reducing outdoor heat exposure. While a variety of shade infrastructures have been developed to protect people from direct sunlight during hot weather, shelter shade remains largely overlooked in urban studies compared with building and vegetation shade. This study examined the spatial-temporal pattern, demographic equity, and supply-demand relationship of shelter shade in Singapore, a compact tropical city with high outdoor activity levels. A Geographic Information System (GIS)-based simulation model was used to estimate hourly shelter shade coverage ratio on the typical hot day. Shelter shade exposure and equity across socio-economic groups were assessed using a population-weighted exposure model and a Gini index. Shelter shade supply-demand analysis was conducted using a four-quadrant analysis and a priority index. Key findings include: (1) considerable spatial-temporal variation in shelter shade, with coverage peaking at midday and concentrated in the Central and West Regions (average: 0.052 and 0.051, respectively); (2) the elderly and Malays exhibited lower average shelter shade exposure compared to the total population (0.328 and 0.333, respectively); and (3) remarkable supply-demand mismatch (25% of the study area), mainly in central, eastern, and northeastern areas with intensive population activities but limited shelter distribution. The study proposes a shelter shade planning framework to support more equitable and effective shading, informing targeted design guidelines and heat-resilient transport and public space planning.

In the daily operation of hydropower stations, the tailwater level is a fundamental parameter for calculating hydropower output, and it is essential for reservoir operation and management. Therefore, this study explores the influencing factors of tailwater level prediction, including the reservoir’s own downstream water level and the headwater level of the downstream reservoir as the characteristic factors. A Stacking ensemble model, using Ridge, Random Forest, Light Gradient Boosting Machine, and Support Vector Regression models as base learners and Ridge as the meta-learner, is established to predict the tailwater level of the Xiluodu Reservoir. We analyze results from the stacked ensemble model and the single model among different stacked combination models and across different quarters for each model. The tailwater level derived from the Stacking ensemble model is found to be in closer agreement with the measured water level than that obtained via interpolation. The model proposed in this study delivers superior predictive performance compared to the four baseline models, with the average absolute error achieving a maximum reduction of 38%. This confirms the effectiveness of the stacking ensemble strategy in predicting tailwater levels, thereby providing accurate boundary conditions for reservoir scheduling calculations.

Soil salinization severely constrains sustainable agricultural development. Accurate monitoring of soil salinity in farmland is therefore of great significance for saline – alkali land management and food security. In this study, typical saline farmland in Caofeidian District, Tangshan, Hebei, China, was selected as the study area, and three experimental plots with different salinization levels were investigated. Considering that the spectral response of soil salinity is relatively weak and that traditional multispectral satellite remote sensing data have limited capability to finely characterize the spatial heterogeneity of salinity, a UAV hyperspectral soil salinity inversion model based on a stacking ensemble framework was developed. Spectral bands were first preprocessed using a combination of logarithmic transformation and standard normal variate transformation. Subsequently, a two-stage feature selection strategy, CARS – SHAP, was proposed to identify informative spectral bands. Specifically, the Competitive Adaptive Reweighted Sampling (CARS) method was first used to select 50 major spectral bands based on regression coefficient weights to reduce redundancy, and multi-model SHAP analysis was then applied to identify 10 key salinity-sensitive spectral bands with stable contributions across different models, thereby improving feature selection stability and prediction consistency under different salinization conditions. Finally, a soil salinity inversion model was constructed using CatBoost, XGBoost, and a multilayer perceptron (MLP) as base learners and partial least squares regression (PLSR) as the meta-learner within the stacking framework. Experimental results demonstrate that the proposed model outperforms conventional single models in both prediction accuracy and stability, achieving an R2 of 0.925 and an RMSE of 0.560 g kg−1 on the validation set. The results provide technical support for high-precision remote sensing monitoring of farmland soil salinity and precision agricultural management.

Predicting canopy traits non-destructively is important for understanding crop growth and improving phenotyping efficiency. Hyperspectral reflectance provides detailed spectral information, but the role of band selection in regression-based trait prediction at the canopy scale remains unclear. In this study, we evaluated the effects of different band-selection algorithms on the prediction accuracy of aboveground biomass (AGB), leaf area index (LAI), and canopy cover (CC) in soybeans using canopy hyperspectral reflectance in the visible to near-infrared (VNIR) range from 501 to 801 nm. The dataset included multiple sites, years, cultivars, and irrigation treatments. We compared a full-band partial least squares regression (PLS) model with three band-selection methods (PLS-Variable Importance in Projection (VIP), Bootstrapped least absolute shrinkage and selection operator (LASSO) (BoLASSO), and an ensemble approach). Model performance was assessed using Kennard–Stone validation and leave-one-year-out cross-validation. The results showed that the effectiveness of band selection depended on the target trait. Full-band PLS performed well for AGB under Kennard–Stone validation, whereas BoLASSO achieved comparable accuracy to PLS for LAI and CC using a reduced number of selected bands. Leave-one-year-out cross-validation showed that year-to-year transferability was more difficult for AGB than for LAI and CC. The selected wavelengths were located mainly in the visible, red-edge, and near-infrared regions. These results indicate that band-selection strategies should be tailored to the target trait and that selected VNIR bands can provide candidate spectral regions for simplified sensing of soybean canopy traits.

Remote Sensing Image-Text Retrieval (RSITR) is the task of learning a shared representation to measure the semantic similarity between remote sensing (RS) images and their textual descriptions. This technology is critical for applications such as disaster assessment and urban management. However, this task is highly challenging due to the varying degrees of alignment reliability between information-dense RS images and their sparse textual descriptions, ranging from explicit mismatches to weak correspondences, thereby fundamentally limiting model retrieval accuracy and robustness. Mainstream embedding-based methods typically rely on discrete supervision paradigms that fail to handle this continuous spectrum of reliability alignment. To resolve this, we propose the continuous reliability-calibrated alignment (CRCA) framework, which pioneers a reliability-aware learning paradigm. The core of our approach is a new supervision framework featuring two key innovations: reliability-calibration weighting (RCW) module, which assigns continuous weights to each pair, with a confidence-gated triplet fusion (CGTF) loss for stable and discriminative learning. To provide a robust foundation for RCW’s assessment, we first employ attentive token condensation (ATC) to purify features by filtering background noise. Furthermore, to deepen the model’s fine-grained semantic understanding, we introduce a Text-guided visual reconstruction (TVR) auxiliary task that compels the model to learn robust local region–word correspondences. Extensive experiments on the RSICD and RSITMD benchmarks demonstrate that CRCA achieves highly competitive performance, with remarkable mR scores of 38.62% and 50.80%, respectively.

Abstract Field‐scale runoff prediction is critical for managing nutrient losses. Ford et al. (2022, https://doi.org/10.1029/2022gl100667 ) present an innovative hybrid modeling and regionalization framework that integrates cluster analysis, National Water Model (NWM) outputs, and machine learning to extend edge‐of‐field (EOF) runoff prediction across the Great Lakes region. In this commentary, we highlight a methodological challenge common in EOF event prediction: when runoff events are rare relative to non‐events, accuracy‐based evaluation can obscure poor event detection. We show that the primary gains in runoff event detection stem from training strategies tailored to datasets dominated by non‐runoff days, with the inclusion of additional soil and meteorological information providing further, complementary improvements while maintaining reasonable overall accuracy. Our results reinforce the promise of the Ford et al. framework, while emphasizing the need for event‐centered evaluation when developing EOF prediction models.

ABSTRACT This study investigates the relationship between the Azores High and Indian summer monsoon during June and September. An opposite correlation pattern is observed during June and September, with significant positive correlation over the Gangetic Plain and north Peninsular India during June, and significant negative correlation over Central India during September. A diagnosis based on monthly ERA5 reanalyzed circulation products archived on finer grids reveals that the sustenance of positive rainfall anomalies over the Gangetic Plain and north Peninsular India during June is supported by the strengthened Azores High near its mean position, zonal extension and northward shift of the Tibetan Anticyclone from its mean position, and strengthening of the Asian jet over the Eurasian region, whereas the sustenance of negative rainfall anomalies over Central India during September is supported by the Azores High, which is shifted to northeast of the North Atlantic, that is, away from its mean position, mid‐latitude waves propagating at higher latitude, hence not affecting Indian summer monsoon rainfall due to their absence towards north India, weakened Tibetan Anticyclone, and weakening of the Asian jet over the Eurasian region.