Remote Sensing Applications-Society and Environment最新文献

{"title":"Advancing climate change Research: Robust methodology for precise mapping of sea level rise using satellite-derived bathymetry and the google Earth Engine API","authors":"Mohammad Ashphaq ,&nbsp;Pankaj K. Srivastava ,&nbsp;D. Mitra","doi":"10.1016/j.rsase.2025.101557","DOIUrl":"10.1016/j.rsase.2025.101557","url":null,"abstract":"<div><div>Sea level rise (SLR), linked to climate change, poses risks to coastal areas and requires urgent action. Traditional methods to measure SLR, such as tide gauges, satellite altimetry, and GNSS-based techniques, have limitations in coverage, accuracy, and data continuity. This study applies Random Forest regression in Google Earth Engine (GEE) to automate satellite-derived bathymetry (SDB) prediction for accurate SLR mapping and time-series analysis. The SDB has been predicted using Landsat series satellite data and derived products, including Chlorophyll, Total Suspended Material, and Turbidity, for the years 1993, 2003, 2013, and 2023. The results demonstrated high accuracy, strong correlation coefficients between in-situ bathymetry and SDB, and low error measures. The correlation coefficients with in-situ bathymetry were 0.8924 in 1993, 0.9386 in 2003, 0.9638 in 2013, and 0.9444 in 2023. Tidal correction was applied to the SDB maps to calculate SLR changes between 1993 and 2023. The analysis delineated a consistent rise in mean SDB values, suggesting a potential increase in sea level over the past four decades. A robust methodology for SLR time-series analysis has been proposed, with all codes accessible for deployment through Landsat collections and temporal parameters.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101557"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143850538","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Ensemble machine learning models for monitoring riparian vegetation dynamics using historical aerial orthophotos","authors":"Afzali Hamid ,&nbsp;Rusnák Miloš","doi":"10.1016/j.rsase.2025.101545","DOIUrl":"10.1016/j.rsase.2025.101545","url":null,"abstract":"<div><div>Historical aerial photographs are well-known as a reliable source of information on historical land cover and land use. However, extracting this information can be challenging due to the limited spectral characteristics in black-and-white images. In this study, we evaluate a textural-based approach using Machine Learning (ML) models to detect the spatial pattern of the braided-wandering multichannel system from historical aerial images with an emphasis on riparian vegetation.</div><div>Five aerial datasets (1949–1992) were used to extract textural information through Gray level Co-occurrence Matrix (GLCM) and geomorphological operations on High-resolution, preprocessed, and normalized orthophotos. We used Random Forest (RF), Light gradient boosting machines (LightGBM), and Extreme Gradient Boosting (XGBoost) ML methods through two classification schemes to classify images into five main classes. GridSearchCV hyperparameter optimization tool were utilized to optimize models and Sequential Feature Selection (SFS) algorithm to reduce the dimensionality of the data cube. The results indicated the efficacy of Morphological operations (Gradient, Eroded, and Dilated) and GLCM features (contrast, entropy) in the final classified map. The RF model demonstrated greater stability and higher median accuracy across datasets. While there was no significant difference between LightGBM and XGBoost in terms of accuracy metrics, XGBoost's performance was notably more variable but significantly faster. In our study, the shadow effects, distortion, and radiometric differences within the orthophotos remain challenging. Despite limitations, the proposed approach addresses key challenges in extracting information from historical orthophotos and can be extended to broader ecological and environmental applications.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101545"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143835132","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Satellite-based remote sensing analysis for the exploration of MVT Pb-Zn mineralization using an integrated approach of minimum distance classification, deep autoencoder and fuzzy logic modeling","authors":"Soran Qaderi ,&nbsp;Abbas Maghsoudi ,&nbsp;Amin Beiranvand Pour ,&nbsp;Mahyar Yousefi","doi":"10.1016/j.rsase.2025.101561","DOIUrl":"10.1016/j.rsase.2025.101561","url":null,"abstract":"<div><div>Mississippi Valley-type (MVT) Pb-Zn mineralization is a key economic resource, yet its exploration is challenging due to complex alteration patterns and high costs. This study integrates ASTER satellite imagery with deep learning to enhance prospectivity mapping. We applied image processing techniques, including Principal Component Analysis (PCA), Band Ratios (BR), Band Math (BM), and Spectral Angle Mapper (SAM), to identify alteration zones. The Minimum Distance Classification (MDC) method classified these zones, extracting key evidence layers. These layers—dolomitization (MDC-PCA, SAM) and carbonate-iron oxide (MDC-BR, MDC-BM)—were integrated using Deep Autoencoder (DAE) and Fuzzy Logic Modeling (GFO) to generate prospectivity maps. Prediction-area (P-A) plots showed the DAE model outperformed GFO, achieving a normalized density (N_d) of 4.1 compared to 3.61 for GFO, indicating a more precise delineation of high-potential mineralization zones. Field validation confirmed strong alignment with known Pb-Zn occurrences. This study highlights the effectiveness of remote sensing and deep learning in cost-effective mineral exploration and provides a scalable framework for similar metallogenic provinces.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101561"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development of topo-bathymetric continuum profiles for coastal barriers with global open-access data","authors":"Valeria Fanti ,&nbsp;Carlos Loureiro ,&nbsp;Óscar Ferreira","doi":"10.1016/j.rsase.2025.101528","DOIUrl":"10.1016/j.rsase.2025.101528","url":null,"abstract":"<div><div>Coastal barriers are dynamic and vulnerable coastal environments exposed to storms and rising sea levels, requiring a thorough understanding of their physical and geomorphological characteristics. Despite this, high-resolution topo-bathymetric data are not openly available for most of the world’s coastal areas, preventing accurate estimation of the exposure to storms and associated risks. Global models of topography and bathymetry, derived from remote sensing techniques, are available worldwide as an open-source solution to characterise coastal morphology. However, their coarse resolution, limited vertical and horizontal accuracy, alongside inconsistencies in the transition from land to the shallow nearshore zone, make their use in coastal areas challenging, requiring careful evaluation. This study investigates the potential and limitations of four recent open-access satellite-derived topographic models (Copernicus GLO-30 DEM, AW3D30, TanDEM-X, Euro-Maps 3D) and three bathymetric models (GEBCO_2023, SRTM15+, ETOPO 2022) in five coastal barriers. It proposes a new approach to integrate global models to provide a consistent representation of the topo-bathymetric continuum profile in coastal areas characterised by a barrier morphology. Coastal barrier profiles, representative of natural sectors and characterized by morphological homogeneity, were derived by merging global topographic and bathymetric digital elevation models and implementing an equilibrium profile in the transition zone. The profiles obtained from the global models were compared with higher resolution local or regional topo-bathymetry. The global topographies tend to underestimate the dune top, with TanDEM-X giving the best results in terms of dune crest height and beach slope. The barrier continuum profiles that merged TanDEM-x and ETOPO 2022 global models were found to have the lowest error, with a vertical RMSE of 0.76 m. Based on integration of these remotely sensed models, it is possible to determine average representative coastal barrier profiles suitable for use in global to regional coastal studies or in data-poor areas, potentially serving as a cost-effective tool for preliminary coastal hazard assessments and early warning systems at wide spatial scales.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101528"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143839627","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multitemporal satellite images for monitoring the volcanic activities and geothermal potential of Ternate Island's Gamalama Volcano, Indonesia's densest active volcanic island","authors":"Bondan Galih Dewanto ,&nbsp;Danang Sri Hadmoko ,&nbsp;Nurul Fitrah Ramadhani ,&nbsp;Admiral Musa Julius","doi":"10.1016/j.rsase.2025.101555","DOIUrl":"10.1016/j.rsase.2025.101555","url":null,"abstract":"<div><div>Throughout history, Ternate, a diminutive volcanic island located in the North Maluku Province of Indonesia, has functioned as a significant center of the nation's social and economic activity. The Gamalama Volcano constituted a significant element of Ternate Island's topography, and its eruption resulted in substantial disruption. The aims of this current contribution are: to monitor the historical activities of the Gamalama volcano and understanding the geothermal potential to support the energy needs in Ternate Island. The multi-temporal analysis was conducted to monitor the activity of Gamalama Volcano, utilizing satellite imagery spanning a period of 50 years. The imagery sources included Landsat 1, Landsat 4, Landsat 5, Landsat 7, ASTER, and Landsat 8. The present study employed the single-channel algorithm to derive the land surface temperature (LST). The band combination and ratio were utilized to infer the geological context and geothermal capacity of the Gamalama Volcano. The analysis of normalized differential vegetation index (NDVI) utilized in the calculation of LST has revealed that vegetation growth has occurred subsequent to certain volcanic eruptions. As per the LST data, the average temperature of the surface within the crater escalated to 38.472 °C during the eruption of 1997, thereby establishing it as the maximum temperature recorded in the past half-century. The volcanic activity of Gamalama Volcano was elucidated through the utilization of the LST technique, which has the capacity to cover various temporal intervals. The congruence between the LST data derived from Landsat and ASTER data substantiates the dependability of the LST methodology. The geothermal potential of approximately 16 °C has been observed in the crater and sand region of the volcano, along with the identification of supplementary hot spots in the north-eastern and western regions of the volcano's primary structure. The utilization of Landsat 8 band combinations and band ratios has substantiated the presence of an area exhibiting elevated geothermal potential within the andesite and basaltic andesite geological formations. The practicality of utilizing multi-source optical satellites for monitoring volcanic activity has been exemplified by the multiple eruptions that have occurred at Gamalama Volcano. Furthermore, this technology could potentially be employed for conducting exploratory research into the geothermal potential of the region.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101555"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143844315","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Monitoring sustainable forest management plans in the Amazon: Integrating LiDAR data and PlanetScope imagery","authors":"Afonso Henrique Moraes Oliveira ,&nbsp;José Humberto Chaves ,&nbsp;Eraldo Aparecido T. Matricardi ,&nbsp;Iara Musse Felix ,&nbsp;Mauro Mendonça Magliano ,&nbsp;Lucietta Guerreiro Martorano","doi":"10.1016/j.rsase.2025.101535","DOIUrl":"10.1016/j.rsase.2025.101535","url":null,"abstract":"<div><div>Selective logging monitoring has traditionally relied on either medium-resolution optical imagery or LiDAR data alone, limiting the detection of both spectral and structural changes in forest cover. This study proposes a integrated analytical approach in parallel of LiDAR data and PlanetScope imagery to enhance monitoring of forest disturbances caused by selective logging in the Amazon. Notably, the correlation between the volume of wood extracted and LiDAR-detected areas is high (r² = 0.9), demonstrating the accuracy of this method in detecting logging-impacted areas. In contrast, the correlation between wood volume and PlanetScope-based mapping is moderate (r² = 0.7), indicating that while this approach effectively detects logging-related disturbances, its accuracy is influenced by factors such as canopy structure and image resolution. LiDAR mapping detected 15.5 % of the total impacted area, compared to 13.7 % detected by PlanetScope. LiDAR achieved higher accuracy in detecting subtle structural changes, such as small clearings (&lt;0.2 ha). Globally, PlanetScope mapping underestimated the total area of clearings, identifying 63.3 ha, whereas LiDAR detected 113.8 ha. The global accuracy of PlanetScope mapping was moderate (P = 0.62) with low recall (R = 0.41), indicating significant underestimation of disturbed forest areas. Metrics such as the global F1-Score (0.50), IoU (0.33), and relatively high RMSE (50.51) further highlight the differences between the two methods. Despite these limitations, PlanetScope mapping was more effective than systems like DETER and SAD in detecting clearings smaller than 1 ha. The integration of these technologies provides more precise and reliable data, strengthening sustainable forest management monitoring and offering critical insights to inform public policies for the Amazon forest sector.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101535"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143767921","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Generating national very high-resolution land cover product of France without any labels: A comparative study","authors":"Junshi Xia ,&nbsp;Clifford Broni-Bediako ,&nbsp;Naoto Yokoya","doi":"10.1016/j.rsase.2025.101542","DOIUrl":"10.1016/j.rsase.2025.101542","url":null,"abstract":"<div><div>Generating a national very high-resolution (VHR) land cover product is crucial for various applications, including environmental monitoring and urban planning. However, creating such a product often requires a large amount of labeled data over a target area, which can be expensive and challenging. In tackling these challenges, this work introduces a comparative analysis of three label-free techniques, including source-domain pretraining, pseudo-labels, and unsupervised domain adaptation (UDA), for developing the French national VHR land cover product. Three label-free techniques leverage the recent OpenEarthMap datasets and employ an advanced segmentation model, a fully Transformer-based network (FT-UNetFormer). The evaluation of these methods utilized the reference offered by the French datasets: FLAIR. Results indicated an overall product accuracy ranging from 82.1% to 85.5%, with a mean intersection over union (mIoU) fluctuating between 57% and 59%. Notably, the highest accuracy was achieved for buildings, while the lowest accuracy was obtained for bareland. Among the three methods, source-domain pretraining demonstrated adequacy but yielded lower accuracy. UDA exhibited very high accuracy; however, it came with considerable computational complexity. The pseudo-labels methods were identified as a viable trade-off between accuracy and computational efficiency. Ultimately, we will release the products derived from the three label-free techniques. The open availability of these products can contribute significantly to informed decision-making and sustainable development across various sectors.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101542"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143873458","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Examining global trends of satellite-derived water quality variables in shallow lakes","authors":"Rossana Caroni ,&nbsp;Anna Joelle Greife ,&nbsp;Mariano Bresciani ,&nbsp;Claudia Giardino ,&nbsp;Giulio Tellina ,&nbsp;Laura Carrea ,&nbsp;Xiaohan Liu ,&nbsp;Stefan Simis ,&nbsp;Clément Albergel ,&nbsp;Monica Pinardi","doi":"10.1016/j.rsase.2025.101565","DOIUrl":"10.1016/j.rsase.2025.101565","url":null,"abstract":"<div><div>Lakes are a vital resource for freshwater supply and key sentinels of climate change, and it is projected that global warming will more persistently affect hydrology, nutrient cycling and biodiversity. In this context, shallow lakes are considered particularly sensitive to a changing environment and it is essential to acknowledge their water quality conditions and recent trends to guide effective water resource management and mitigation strategies. The European Space Agency Climate Change Initiative (ESA-CCI) offers globally consistent satellite observations of the Lakes Essential Climate Variable (ECV) including satellite products such as chlorophyll-a (Chl-a), turbidity and surface water temperature (LSWT) for over 2000 lakes during 1992–2020. From this dataset, we extracted a subset of 347 lakes with mean depth ≤ 3 m distributed globally to investigate a long-term timeseries (2002–2020) for Chl-a and turbidity. Theil-Sen trend analysis showed that Chl-a did not change significantly in 33 % of lakes, significantly increased in 45 % and decreased in 22 % of the lakes, while turbidity significantly increased in 60 % and decreased in 17 % of lakes. Most lakes with increasing Chl-a and turbidity trends were located in lowland areas, and had relatively large areas (surface area &gt;50 km²). Further analysis revealed that the majority of lakes showed a concurrent increase in both Chl-a (48 %) and turbidity (50 %) with LSWT, indicating the potential influence of climate warming on lake water quality. A structural equations model-based analysis used for modelling the interactions between climatic, socioeconomic features and water conditions overall showed that Chl-a and turbidity had a concurrent positive increase with population and gross regional product in most lakes. This finding suggests that the impact of human population growth in lake catchments represents an important factor driving pressures on the water quality of shallow lakes.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101565"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143873460","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Weeds detection in a Citrus orchard using multispectral UAV data and machine learning algorithms: A case study from Souss-Massa basin, Morocco","authors":"Mohammed El Hafyani ,&nbsp;Amine Saddik ,&nbsp;Mohammed Hssaisoune ,&nbsp;Adnane Labbaci ,&nbsp;Abdellaali Tairi ,&nbsp;Fatima Abdelfadel ,&nbsp;Lhoussaine Bouchaou","doi":"10.1016/j.rsase.2025.101553","DOIUrl":"10.1016/j.rsase.2025.101553","url":null,"abstract":"<div><div>Recently, Unmanned Aerial Vehicles (UAVs) have been used extensively in agriculture, especially at the farm scale. In this work, the images acquired using the DJI Phantom 4 Pro Multispectral (P4M) were used to detect weeds taking two sites in a Citrus orchard farm located in the Souss-Massa region as a case study. A variety of processing steps were employed to prepare the data. Starting by the image's alignment, followed by the georeferencing using ground control points (GCPs), the creation of dense clouds, generation of the digital elevation model (DEM), digital surface model (DSM) and finishing by the extraction of the orthomosaic and multispectral image. Then, the spectral indices including the Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI) were calculated. And the machine learning (ML) algorithms such as Maximum Likelihood Classification (MLC), Mahalanobis Distance Classification (MHDC), Minimum Distance Classification (MDC), Spectral Angle Mapper (SAM), and Support Vector Machine (SVM) were applied. In the site 1, the results showed an overall accuracy of 90.00 %, 70.50 %, 76.25 %, 57.50 %, and 88.48 %, and the Cohen's kappa coefficient of 0.81, 0.79, 0.63, 0.44, and 0.77 for the MLC, MHLC, MDC, SAM, and SVM respectively. In the site 2, the results showed an overall accuracy of 97.05 %, 94.12 %, 89.70 %, 52.94 %, 66.33 %, and the Cohen's kappa coefficient of 0.95, 0.90, 0.83, 0.39, and 0.43 for the MLC, MHLC, MDC, SAM, and SVM respectively. This study has therefore shown the potential of UAVs data, and the opportunity that presents this new technology for farmers to develop their production and optimize the water and fertilizers consumption.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101553"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143859511","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"BSANet: A Bilateral Segregation and Aggregation Network for Real-time Cloud Segmentation","authors":"Yijie Li ,&nbsp;Hewei Wang ,&nbsp;Shaofan Wang ,&nbsp;Jinfeng Xu ,&nbsp;Yee Hui Lee ,&nbsp;Soumyabrata Dev","doi":"10.1016/j.rsase.2025.101536","DOIUrl":"10.1016/j.rsase.2025.101536","url":null,"abstract":"<div><div>Segmenting clouds from intensity images is an essential research topic at the intersection of atmospheric science and computer vision, which plays a vital role in weather forecasts and climate evolution analysis. The ground-based sky/cloud image segmentation can help extract the cloud from the original image and analyze the shape or additional features. With the development of deep learning, neural network-based cloud segmentation models can have better performance. In this paper, we introduced a novel sky/cloud segmentation network named Bilateral Segregation and Aggregation Network (BSANet) large version with 4.29 million parameters, which benefits from our designed BSAM, achieving almost the same performance compared with the state-of-the-art method. BSAM uses the rough segmentation map from the previous stage to produce two new weighted feature maps representing the sky and cloud features, and two network branches are utilized to process the features separately. After the deployment via TensorRT, the BSANet-large configuration can achieve 392 fps in FP16 while BSANet-lite with only 90K parameters can achieve 1390 fps, which all exceed real-time standards. Additionally, we proposed a novel and efficient pre-training strategy for sky/cloud segmentation, which can improve segmentation performance when ImageNet pre-training is not available. In the spirit of reproducible research, the model code, dataset, and results of the experiments in this paper are available at: <span><span>https://github.com/Att100/BSANet-cloudseg</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"38 ","pages":"Article 101536"},"PeriodicalIF":3.8,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143873459","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}