Spatiotemporal variability drives dissolved organic matter composition in a tropical river: Insights from PARAFAC and molecular analyses.

Multimodal fusion of optical imagery and time-series Synthetic Aperature Radar (SAR) data is essential for Land Use and Land Cover (LULC) classification in persistently cloudy regions. However, inherent spatiotemporal heterogeneity between modalities frequently impedes effective feature alignment, rendering conventional fusion techniques suboptimal in dynamic scenarios. To circumvent these constraints, we propose the Dynamic-Weighted Network (D-WNet), a dual-branch framework designed to explicitly decouple and adaptively synergize heterogeneous multimodal inputs. The architecture utilizes depthwise separable convolutions for the multiscale spatial–spectral analysis of optical imagery and ConvLSTM units to capture temporal scattering dynamics from SAR sequences. Furthermore, the proposed enhanced adaptive fusion module leverages a local-global dual-attention mechanism to dynamically integrate features, while a cross-layer residual guidance strategy refines boundary delineation and mitigates SAR speckle noise. Evaluations across diverse landscapes—including the Qinghai-Tibet Plateau, northwestern China, and California—demonstrate that the proposed framework outperforms existing methods, achieving overall accuracy improvements of 6.8-12.4%. By effectively fusing optical spectral features with time-series SAR responses, the D-WNet provides a robust solution for precise land-cover mapping in persistently cloudy environments.

Land use and land cover (LULC) change significantly affects environmental processes and sustainable land management in river basins. The Upper Blue Nile River Basin, Ethiopia, has experienced significant LULC changes due to population growth, agricultural expansion, and deforestation. This study examines past and future LULC dynamic patterns using an integrated cloud-based framework implemented in Google Earth Engine. Classification and regression tree, random forest (RF), and support vector machine classifiers were used to process multi-temporal Landsat imagery. RF achieved the highest overall accuracy (94.4%) and kappa (0.879) in 2024. The results reveal pronounced agricultural expansion and substantial forest loss over the past two decades. Using the RF-derived LULC maps, a Cellular Automata-Markov model was calibrated and validated with strong agreement (Kappa = 0.886) to project future changes. Projections to 2034 and then 2044 indicate continued expansion of agricultural and built-up areas at the expense of forests and shrub/grasslands. The results support improved land use planning and environmental sustainability in the basin.

ABSTRACT Protected areas (PAs) provide critical ecosystem services, including global climate regulation through the retention of carbon in biomass and soil. Evaluating the impact of terrestrial PAs on carbon storage is essential for strengthening their role in mitigating climate change, especially in the face of significant Land Use and Land Cover (LULC) changes, which can influence carbon dynamics. This study focuses on Ifrane National Park (INP) in Morocco, where the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model was used to simulate carbon stock changes between 2018 and 2033, with a particular emphasis on Strict PAs. The results reveal a total carbon stock decline of 14.28% within INP over the study period. Forested areas, especially Holm oak forests, showed notable resilience, with their carbon stocks increasing by 20.79% inside Strict PAs and by 46.63% outside these zones. In contrast, cedar forests experienced a decline of 5.24% within Strict PAs and 17.56% outside. This highlights the significant role of Strict PAs in mitigating carbon losses, especially in vulnerable ecosystems like cedar forests. Our study underscores the need to expand Strict PAs and adopt sustainable forest management to enhance carbon sequestration and support climate change mitigation.

Remote sensing has rapidly advanced with the integration of deep learning, enabling more accurate and scalable detection of land use and land cover (LULC) changes, particularly with the increasing availability of Sentinel-1 and Sentinel-2 multispectral imagery. This review traces the evolution of classical machine learning approaches toward modern deep learning architectures, including Convolutional Neural Networks (CNNs), encoder-decoder models, Siamese and dual-stream networks, attention-based frameworks, and, more recently, Transformer-based models. Recent developments in Earth observation foundation models, trained on large-scale, multimodal datasets, have introduced new capabilities, including zero-shot inference, cross-sensor transferability, and improved generalization across diverse geographic regions. Despite these advances, significant challenges remain. The fusion of multimodal data, including optical, SAR, and ancillary sources, is complicated by differences in spatial, spectral, and temporal characteristics. Furthermore, domain adaptation, label noise, and limited geographic transferability continue to constrain the robustness of change detection pipelines. The quantification of uncertainty and model interpretability has also become increasingly important for operational applications in urban planning, agriculture, ecosystem monitoring, and disaster response. In addition, the growing computational and environmental costs of large-scale model pretraining underscore the need for more sustainable AI practices. Future research should therefore focus on advancing the Earth observation foundation and generative models, developing temporal AI methods for long-term sequence analysis, and promoting responsible, energy-efficient geospatial artificial intelligence. Integrating advances in remote sensing, machine learning, and environmental science will be essential for building practical, scalable, and reliable planetary monitoring systems.

ABSTRACT The Brahmaputra River (BR) is one of the world’s most dynamic rivers, and its channel migration significantly impacts the land use and land cover (LULC), and socio-economic stability of Assam, India. This study evaluates LULC changes within the river corridor and their impact on socio-economy of the region. Multi-temporal satellite imagery from 1976 to 2020 was analysed using hybrid classification techniques and manual recoding to map seven LULC classes. River dynamics was quantified through bankline shift and overlay analysis. Results indicate a pronounced south-westward shift of the river, with average bankline shifts of 1.71 km on the left bank and 1.74 km on the right bank. Over 44-years, 1552.2 km2 of land was lost to erosion, while 286.5 km2 was formed through deposition. Erosion severely affected agriculture (705.9 km2), threatening food security and livelihoods, and vegetation (446.3 km2), reducing ecological resources. Critically, Built-up losses (78.7 km2) exposed settlements to displacement and infrastructure damage, representing the most immediate socio-economic impact. Conversely, deposition resulted in constructive change supporting agriculture expansion (131.1 km2), vegetation growth (63.6 km2), and settlement development (17.1 km2). These findings underscore that agriculture, vegetation and settlements, the core socio-economic pillars of Assam’s floodplain communities are directly shaped by river dynamics.

This study presents a comprehensive spatiotemporal assessment of key ambient air pollutants (PM2.5, PM10, SO2, NO2, NH3, O3, and CO) across seven urban and semi-urban districts of the Uttar Pradesh National Capital Region (UP-NCR), India, during 2019-2023. Despite an overall decline of approximately 23% in particulate matter concentrations over the study period, annual mean PM2.5 (88 ± 15µg/m3) and PM10 (185 ± 28µg/m3) levels consistently exceeded the National Ambient Air Quality Standards. Land use and land cover (LULC) analysis revealed a 7.06% expansion in built-up areas, reflecting rapid urbanization and its influence on local emission patterns. This urban growth was associated with persistent NO2 enrichment, particularly in Noida, where concentrations increased by 22%. Pronounced seasonal variability was observed, with PM2.5 concentrations peaking during the post-monsoon season particularly in Noida (200 ± 102µg/m3) and Ghaziabad (178 ± 99µg/m3), identified as the regional pollution hotspot. Meteorological analysis revealed strong seasonal influences on pollutant concentrations. Relative humidity exhibited positive correlations with particulate matter during winter (r ≈ 0.44-0.59), reflecting hygroscopic growth and stagnant atmospheric conditions, but strong negative correlations during the monsoon (r ≈ -0.72 to -0.95) due to efficient wet scavenging. Bivariate polar plot analysis identified stagnation-driven pollutant accumulation in densely urbanized districts and wind-induced resuspension of agricultural and crustal dust in peripheral regions. HYSPLIT backward-trajectory clustering further demonstrated substantial contributions from long-range transport originating from western and northwestern source regions during pollution episodes. These findings highlight pronounced spatial heterogeneity and seasonal dynamics in air quality, emphasizing the need for region-specific, airshed-based mitigation strategies across rapidly urbanizing peri-urban corridors.

Glacial lake outburst floods (GLOFs) represent one of the most severe climate-induced hazards in high-mountain regions. The Shisper Glacier in Hassanabad Village, Hunza, Pakistan, has experienced repeated surge events and rapid ice-dammed lake expansion, posing escalating risks to downstream communities and critical infrastructure. This study presents a multi-temporal, multi-index remote sensing and GIS-based assessment of Shisper Glacier dynamics and associated GLOF hazards for the period 2020-2024. Landsat-8 OLI imagery was used to compute the normalized difference snow index (NDSI), normalized difference water index (NDWI), modified normalized difference water index (MNDWI), normalized difference vegetation index (NDVI), land surface temperature (LST), land use and land cover (LULC), and green-red-NIR-shortwave infrared (GRZI) composite indices. Elevation change analysis was performed using the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM), while NASA POWER data provided climatological context for temperature, precipitation, and humidity patterns. The ice-dammed lake area at the glacier's snout was delineated using MNDWI-based water body extraction from Landsat-8 OLI imagery, and lake volume was estimated using a depth-area empirical scaling relationship, consistent with established methodologies for ice-dammed lakes in the Karakoram. Results indicate that the debris-covered glacier expanded from 161 km2 in 2015 to 197 km2 in 2018, while the ice-dammed lake exceeded 0.1 km2 with an estimated volume between 1 × 106 m3 and 10 × 106 m3, representing a substantial and growing GLOF hazard. Rising land surface temperatures and changing precipitation patterns are identified as key drivers of accelerated glacial melt. Approximately 80% of Hassanabad Village's infrastructure, including the Karakoram Highway Bridge and a nearby hydropower facility, lies within the projected GLOF inundation zone. The novelty of this study lies in the first integrated application of seven spectral indices combined with DEM-based surge mapping and infrastructure exposure assessment for Shisper Glacier within a unified hazard framework. These findings underscore the urgent need for continuous satellite-based monitoring, early warning systems, and risk-informed land-use planning in the Hunza region.

Land use and land cover (LULC) change is a major anthropogenic factor influencing flood behavior and hydrological processes. This systematic review synthesizes two decades (2005-2025) of research on hydrological modeling approaches used to assess flood responses under LULC transitions. A total of 114 publications were retrieved from the Scopus database, and after applying PRISMA-based screening, 78 peer-reviewed studies were analyzed using bibliometric and content mapping. The review categorizes hydrological models by spatial scale, process representation, and sensitivity to LULC dynamics. Findings consistently indicate that urban expansion, deforestation, and vegetation loss intensify surface runoff, peak flow, and flood frequency. Despite advancements, significant challenges remain particularly related to data scarcity, model calibration, and the limited integration of socio-economic variables. Emerging tools such as Remote Sensing (RS), Geographic Information Systems (GIS), and machine learning especially within platforms like Google Earth Engine (GEE) enhance LULC detection accuracy and flood prediction capability. The study proposes an integrated decision framework linking bibliometric trends with model selection strategies, enabling researchers to align model choice with data availability and landscape characteristics. Overall, this review emphasizes the importance of interdisciplinary, data-driven modeling to strengthen flood resilience in rapidly transforming land systems.

Spatial land transformation and the psychosocial exposure among climate migrants in southwestern Bangladesh

ABSTRACT Using the Google Earth Engine platform, this study compares three datasets from the Planet, Sentinel-2A, and Landsat 8 satellites for the study of Najaf Province, Iraq, at spatial resolutions of 3.7 m, 10 m, and 30 m, respectively. Seven algorithms (smileCART, Random Forest, Gradient Tree Boost, Support Vector Machine, Minimum Distance, Naive Bayes, and k-Nearest Neighbour) are tested. The primary objectives were to evaluate Land Use and Land Cover (LULC) mapping, and to select a suitable algorithm and dataset. Achieving accurate mapping of LULC categories remains a major challenge, particularly within semi-arid regions characterized by complex farming systems. Supervised classification techniques, specifically the seven algorithms, were applied, and ten classes were successfully identified, namely Irrigated Cropland, Watercourses, Water Bodies, Sandy Areas and Dunes, Arable Land, Bare Areas, Wetland, Urban and Industrial Areas, Palm, and Shrubland. The overall accuracy was computed for each classifier algorithm after the collection of 28–55 samples representing the ground truth for each identified class. Among the classifiers, Planet reached the highest overall accuracy of 92% with the smileCART algorithm, followed by 88% with Gradient Tree Boost and 86% with Random Forest. On the other side, Sentinel-2A achieved a maximum accuracy of 81% (Gradient Tree Boost and Random Forest), while Landsat 8 presented an overall accuracy of 78% using Random Forest. A test using only common bands across all sensors showed that spatial resolution contributes approximately three times more to classification accuracy than spectral richness in this semi-arid environment. It is noteworthy that the Planet satellite demonstrated an ability to distinguish Shrubland from other land cover classes compared to Sentinel-2A and Landsat 8. Since the grassland/shrubland transition is critical in the ecology and degradation of arid and semi-arid regions, the study suggests a role for recently available high-resolution satellite imagery to improve monitoring of such regions.

To assess the survival vulnerability of Oncomelania hupensis in Jiangxi Province under future climate scenarios, and to identify low-vulnerability areas for its survival in this province. Village-level O. hupensis snail survey and O. hupensis snail control with chemical treatments in Jiangxi Province from 2016 to 2024 were captured from the Parasitic Disease Prevention and Control Information Management System of China Disease Prevention and Control Information System. Climatic data were primarily sourced from the Resource and Environmental Science Data Platform, Chinese Academy of Sciences (http://www.resdc.cn/), including annual average temperature, annual average precipitation, annual accumulated temperature above 10 °C, annual accumulated temperature above 0 °C, annual maximum temperature, annual minimum temperature, and annual average relative humidity, and nineteen bioclimatic variables were downloaded from the WorldClim website (https://www.worldclim.org/), including mean diurnal range, isothermality, temperature seasonality, and so on. Elevation and normalized difference vegetation index were catprued from the Resource and Environmental Science Data Platform, Chinese Academy of Sciences (http://www.resdc.cn/), and distance to rivers was downloaded from the WorldPop website (http://www.worldpop.org), and land use and land cover (LULC) data were downloaded from the Big Earth Data Center, Chinese Academy of Sciences (https://data.casearth.cn/), and nature reserve data were obtained from the China Nature Reserve Specimen Resource Sharing Platform (http://www.papc.cn/). Three Shared Socioeconomic Pathways (SSPs) from the Beijing Climate Center-Climate System Model version 2-Medium Resolution (BCC-CSM2-MR) global climate model were employed as future climate scenarios, including SSP126, SSP245, SSP585, and the biomod2 ensemble model in R package was used to simulate suitable habitats for O. hupensis snails in Jiangxi Province in 2050 and 2070 under these scenarios. A snail survival vulnerability index was constructed based on the area of suitable snail habitats, area covered by snail control through chemical treatment, area covered by nature reserves, and changes in snail habitat fragmentation, and a map of snail survival vulnerability distribution was plotted. The real area of snail habitats ranged from 78 486.76 to 85 309.47 hm2, and the area of snail control with chemical treatment ranged from 10 138.98 to 13 240.16 hm2 in Jiangxi Province from 2016 to 2024. There were 429 to 531 villages detected with snails during the nine-year period, and the number of actually snail-infested villages ranged from 645 to 686. A total of 818 snail-present points and 1 996 snail-absent points were obtained from snail survey records. The best performance of the biomod2 ensemble model was achieved if a weighted mean approach was used as the ensemble strategy, with a true skill statistic value of 0.799 and an area under the receiver operating characteristic curve of 0.957, and modeling identified annual average relative humidity and annual average precipitation as two most influencing climatic variables for snail distribution. Relative to the current areas of suitable snail habitats under present climate conditions, the area of suitable snail habitats was projected to expand by 24.49% to 46.28% in Jiangxi Province under future climate scenarios, and the proportion of nature reserves areas in the areas of suitable snail habitats was projected to decrease slightly from the current 2.77% to approximately 2.52%, while the proportion of areas of snail control through chemical treatment in areas of suitable snail habitats varied from 0.64% to 19.57%, and the percentage of changes in snail habitat fragmentation ranged from 3.86% to 12.23%. Based on these four indicators, the snail survival vulnerability index was estimated to range from -1.96 to 0.62 in Jiangxi Province. The arithmetic mean of the snail survival vulnerability index differed under three SSP scenarios (SSP126, SSP245 and SSP585), with the highest mean value (-0.69) in 2070 under SSP126, and the lowest mean value (-0.78) in 2070 under SSP585. The snail survival vulnerability index ranges from -1.96 to 0.62 in Jiangxi Province under future climate scenarios, and the suitable habitats for O. hupensis snails appear an overall tendency towards expansion. Low-vulnerability snail habitats are mainly distributed along the shores of Poyang Lake and the Yangtze River in Jiangxi Province, partially overlapping with nature reserves. Intensified surveillance of O. hupensis snails is recommended in these areas in the future.

ABSTRACT Accurate land use and land cover (LULC) classification is essential for environmental monitoring, resource management, and policy planning. This study evaluates the seasonal performance of six machine learning (ML) algorithms, including random forest (RF), support vector machine (SVM), k-nearest neighbor (k-NN), gradient tree boost (GTB), classification and regression tree (CART), and naive Bayes (NB), using the Google Earth Engine (GEE) platform. The analysis was conducted in Butler County, Ohio, a temperate region with distinct seasonal variability, using optical imagery across spring, summer, fall, and winter. Classifier performance was assessed using overall accuracy (OA), kappa coefficient, and F1-score. Results indicate that classification performance exhibits seasonal variation associated with phenological dynamics, with spring showing the highest average OA (93.84%), kappa (92.30%), and F1-score (0.94), highlighting how seasonal surface conditions influence relative algorithm performance. Among the ML classifiers, k-NN achieved the highest average OA (95.76%), kappa (94.7%), and F1-score (0.96), closely followed by RF with OA (95.45%), kappa (94.32%), and F1-score (0.95), while NB exhibited substantially lower performance across all seasons (average OA: 78.94%, kappa: 73.68%, F1-score: 0.78). These findings underscore how seasonal surface conditions influence relative algorithm performance and highlight the importance of season-aware model selection for reliable year-round LULC classification. The results further demonstrate how seasonal analysis can inform data acquisition timing and support the design of land-monitoring programs across diverse environmental contexts.

ABSTRACT Cloud cover presents a significant challenge for optical image classification, particularly in the analysis of multi-temporal satellite imagery, as it obscures surface features and affects data continuity. This study investigated cloud-filling methods and U-Net-based architectures for time-series Sentinel-2 satellite images in land use and land cover (LULC) classification. Four cloud-filling techniques were evaluated – image mosaicking, null value retention, multiple imputation by chained equations, and principal component analysis (PCA)-enhanced imputation. Among these, PCA-enhanced imputation exhibited the best performance, achieving an overall accuracy of 84% and a macro F1-score of 74.87% by effectively preserving the temporal features and mitigating cloud impact. Four U-Net-based architectures with temporal features—3D U-Net, U-Net with a temporal attention encoder, U-Net convolutional long short-term memory, and U-Net bidirectional convolutional long short-term memory – were compared using the PCA-enhanced imputed dataset. The 3D U-Net outperformed the others due to its ability to capture spatial and temporal continuity. Other models showed limitations in effectively leveraging temporal dependencies. The results demonstrate the effectiveness of multi-temporal imagery and the benefits of cloud-filling methods in enhancing classification accuracy, especially for complex LULC categories such as changed areas. These findings highlight the value of robust preprocessing techniques and temporal modelling for detailed and accurate LULC mapping, offering practical solutions for remote sensing applications.

Changes in land use and land cover (LULC) have a substantial impact on environmental sustainability, urban planning, and agricultural management, necessitating accurate classification using high-resolution satellite imagery. Conventional methods struggle with challenges, such as noise, edge preservation, and misclassification, due to the complexity of multispectral and hyperspectral data. To overcome these issues, this paper presents a Finite-Element Diffusion Banyan Tree Growth Kernel-Driven Attention Network (FED-BTG-KDAN) for LULC mapping. High-resolution satellite images from Sentinel-2, Google Earth Engine (GEE), and Landsat-8 are very useful for multispectral and hyperspectral image classification. Image pre-processing is done by Robust Double-Weighted Guided Image Filtering (RDWGIF) to improve the image quality and spatial resolution. Feature extraction is done by a modified ResNet-152 with Multi-Axis Vision Transformer (MAViT) to extract spatial, spectral, and contextual information. The classification process uses a Finite-Element-Integrated Neural Network (FEINN) with a Diffusion Kernel Attention Network (DKAN) to reduce misclassification and enhance spatial representation. Optimisation is done by Banyan Tree Growth Optimisation (BTGO) to ensure fast convergence and minimise classification errors. The proposed method achieves 99.9% accuracy, demonstrating superior performance in precision, robustness, and computational efficiency compared to conventional approaches. Advantages include enhanced feature extraction through multi-scale attention and optimised classification using bio-inspired learning.

Abstract. Soil moisture is an important part of the land water cycle. It affects ecosystem health, agricultural output, and climate regulation. Several factors influence its variability, especially land use and land cover (LULC). This study looked at how LULC relates to changes in soil moisture in Zaria, Nigeria. The study area is experiencing urban growth and agricultural expansion. Soil moisture data was retrieved from the European Centre for Medium-Range Weather Forecasts (ECMWF). LULC maps were created from satellite images taken from 2000 to 2020. Partial correlation analysis was employed to determine the relationship between various land cover types and soil moisture. The results show strong positive correlations between soil moisture and vegetated areas (r = 0.9801) and also with water bodies (r = 0.9232). These results indicate that thick vegetation and nearby water bodies help retain soil moisture by reducing evaporation, improving infiltration, and increasing water-holding capacity. On the other hand, soil moisture has a strong negative correlation with built-up areas (r = −0.8723) and bare soil (r= −0.997). The study shows that LULC characteristics greatly impact soil moisture changes. The study finally demonstrated the vital role of vegetation and water features in maintaining soil moisture, while also highlighting the negative effects of urban expansion and land degradation in semi-arid areas like Zaria.

Land use and land cover (LULC) classification is essential for environmental monitoring, urban planning, and resource management. This study explores the performance of three state-of-the-art deep learning architectures, MobileNetV3, ResNet34, and GoogleNet, which were enhanced with transfer learning, data augmentation, and adaptive learning rate scheduling. We evaluate these models on two benchmark datasets: EuroSAT, consisting of Sentinel-2 satellite imagery across 10 land cover classes, and PatternNet, a high-resolution aerial dataset with 38 diverse classes. The results demonstrate that MobileNetV3 achieved the highest overall accuracy (97.83% on EuroSAT and 99.23% on PatternNet) with minimal inference time, making it ideal for real-time applications. ResNet34 achieved 97.56% and 99.06% accuracy, respectively, excelling in classifying complex, visually similar classes due to its residual learning blocks. GoogleNet's balanced performance and efficiency achieved 97.36% and 99.58% accuracy across both datasets. An ablation study confirmed that data augmentation, transfer learning, and learning rate scheduling contributed to improvements in accuracy of 5-13%. This research highlights the effectiveness of modern deep learning architectures and optimized training pipelines for LULC classification across diverse datasets, providing a foundation for future advancements in cross-domain remote sensing applications.

Accurate detection of spatiotemporal changes in land use and land cover (LULC) is crucial for urban planning and smart city development, but it faces challenges due to urban environments and satellite imagery complexities. To address these, a novel “Spectral Fusion Autoencoder with Capsule and Convolutional Long Short-Term Memory Network” is proposed for enhanced land cover classification. Additionally, sub-pixel change detection faces challenges due to spectral mixing, where small urban features share similar spectral properties with the surrounding environments, complicating accurate differentiation. To address this, a “Spectral Unmixing Residual Network with Variational Autoencoder” is proposed, separating mixed signals, capturing spatial features, and modeling nonlinear relationships for enhanced sub-pixel detection. Furthermore, geometric distortions in multi-temporal or multi-source imagery, especially in steep terrains, disrupt spatial alignment. Thus, a novel “Dynamic Attention-Driven Capsule Policy Optimization Network” addresses these issues by refining geometric corrections and enhancing spatial consistency through reinforcement learning. Moreover, gradual spectral transitions and intra-class variability complicate inter-class land cover changes. So, a novel, “Warped Spectral Fusion Network with ConvLongShort Net” integrates spectral, spatial, and temporal data to correct illumination and viewing angle variations, capturing subtle transitions effectively. The experimental results demonstrate the proposed model’s efficiency in accurately detecting the spatiotemporal changes with an accuracy of 0.99, loss of 0.003, fractal error of 0.32, and misclassification rate of 0.06.

The Swin Transformer exhibits limitations in fine-grained land use and land cover (LULC) classification, particularly in capturing high-frequency texture details and representing low-contrast regions. To address these issues, we propose a novel network model, termed LACE-Net, which integrates local frequency-domain energy and adaptive contrast enhancement. Built upon the Swin Transformer backbone, the model introduces an innovative Local Frequency-Domain Energy-Adaptive Contrast Enhancement Multi-Scale Attention (LACE). This block consists of parallel branches for frequency-domain perception and contrast enhancement, which effectively combine texture and illumination physical priors. In addition, a texture-adaptive momentum adjustment mechanism is incorporated to refine the spatial enhancement attention weights dynamically. Consequently, LACE-Net greatly strengthens the modeling and representation of high-frequency details and complex spatial structural features. Experiments are performed on a self-constructed Guangxi regional dataset (denoted as GLC-30) and the publicly available remote sensing scene classification benchmark dataset NWPU-RESISC45. The results show that LACE-Net achieves a Top-1 accuracy (Top-1 Acc) of 96.48% and a macro-averaged F1 score (mF1) of 93.13%. These results outperform current mainstream vision models, particularly in mitigating the spectral confusion issue of “same spectrum, different objects.” The model exhibits superior fine-grained classification performance and robust generalization across datasets.

Nature protection practices and agricultural systems face sustainability challenges in mountainous countries regions with limited resources. The dynamics between agriculture and nature involves complex interactions that reflects the responses of farming systems to shifts in nature conservation policy and biophysical factors. Moreover, climate change impacts biodiversity through control of water resources and soil fertility. In Italy, mountain areas have experienced socio-economic changes in recent decades, which has affected traditional agro-forestry activities and resulted in forest expansion. This study examines the sustainability performance of farming systems and forest protection areas in Italy. Existing FAO data on 1990-2025 were used to analyse the dynamics in land use and land cover (LULC) changes in context of social-economic, climate and environmental aspects. The data analysis was performed using R language by its statistical and computing libraries such as readr, ggplot2, reshape2, tidyverse, gridExtra, stats, plotly, latticeExtra, ggpubr as the main ones. The results demonstrated trends in reforestation (increase of forest areas on 9% and shrubland on 21%), climate warming (glacier and snow retreat on 34%), urbanization (increease of artificial surfaces on 5.8 %) and intensification of agriculture activities (stable increase in cropland on 2 %), which indicates sustainable development in nature protection and social-economic activities of Italy.