International journal of applied earth observation and geoinformation : ITC journal最新文献

{"title":"Advanced deep learning techniques for automated extraction of non-debris-covered areas of glaciers in High-Mountain Asia using time-series remote sensing data","authors":"Gexia Qin ,&nbsp;Ninglian Wang ,&nbsp;Bo Jiang ,&nbsp;Yuwei Wu ,&nbsp;Yanchao Yin ,&nbsp;Zhijie Li","doi":"10.1016/j.jag.2025.104680","DOIUrl":"10.1016/j.jag.2025.104680","url":null,"abstract":"<div><div>Deep learning approaches have gained prominence for automatic glacier boundary extraction due to their localized nature of convolutional operations, potentially leading to incomplete or fragmented glacier pixel representations. Moreover, the accuracy of extracting glacier boundaries from a single remote sensing image (RSIs) is often influenced by seasonal snow, clouds, shadows, and frozen lakes. To overcome these challenges, we introduce a novel model for extracting the non-debris-covered areas of glaciers (NDCAG) from RSIs, termed GlacierSTR-UNet. This model enhances information flow and overall performance by embedding the Swin Transformer (ST) as an encoder into a U-shaped architecture and reduces training time and improves gradient handling by incorporating the ResNet block in the decoder. We deploy the GlacierSTR-UNet model on the Google Earth Engine (GEE) platform to efficiently generate multiple NDCAG results from RSIs taken at different periods. A pixel-by-pixel synthesis algorithm is then applied to aggregate the multiple NDCAG extraction results, producing the final NDCAG. Accuracy assessments indicate that GlacierSTR-UNet achieves an overall accuracy of 0.8817, and the relative deviation between automatically extracted and manually interpreted NDCAG remains within 2 %. Finally, we obtain the NDCAG datasets for the periods of 2015/2016 and 2022/2023 in High-Mountain Asia, revealing a reduction of 4,185.12 ± 7,870.96 km² in NDCAG from 2015/2016 to 2022/2023. These findings demonstrate the effectiveness of our approach in efficiently and accurately extracting NDCAG, highlighting its potential for monitoring glacier changes and supporting glacier inventory efforts.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"142 ","pages":"Article 104680"},"PeriodicalIF":7.6,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144313383","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":"Harnessing deep learning and CRF for prior-knowledge modeling of crop dynamics","authors":"Laura Elena Cué La Rosa ,&nbsp;Dario Augusto Borges Oliveira ,&nbsp;Raul Queiroz Feitosa","doi":"10.1016/j.jag.2025.104616","DOIUrl":"10.1016/j.jag.2025.104616","url":null,"abstract":"<div><div>Remote sensing has revolutionized crop mapping and monitoring, providing valuable insights for sustainable agricultural practices. However, successfully implementing remote sensing-based crop type identification in tropical regions remains challenging. Unlike temperate regions, tropical areas benefit from favorable weather conditions that support diverse land management practices, resulting in complex crop dynamics that are difficult to model. Additionally, frequent cloud cover in tropical regions limits the use of optical data during extended periods of the year, making SAR (Synthetic Aperture Radar) a more attractive alternative. Traditional models like Conditional Random Fields (CRFs) have been useful in classifying crop types using spatial and temporal contexts. However, these models often use fixed inputs, optimizing them towards a specific task, and fail to consider the CRFs’ feedback for end-to-end learning features and the spatiotemporal dependencies. This work addresses this gap by introducing an end-to-end hybrid model that combines deep learning with CRFs to incorporate prior knowledge-based modeling of crop dynamics. The proposed framework integrates a backbone encoder–decoder to capture spatial and temporal contexts, with a CRF block that models temporal dynamics by accounting for label dependencies between adjacent epochs. This block facilitates the integration of both data-driven and expert-domain temporal constraints, allowing the framework to adapt to local agricultural practices and enhance crop dynamics modeling. We evaluate the framework using multi-temporal SAR image sequences from two municipalities in Brazil, comparing the effectiveness of learned versus prior-knowledge temporal constraints. The results show improvements of up to 30% in per-class F1 score and 12% in average F1 score compared to a baseline model that excludes temporal dependencies. These findings highlight the value of incorporating prior knowledge-driven temporal constraints into crop mapping models.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"142 ","pages":"Article 104616"},"PeriodicalIF":7.6,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144313384","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":"Enhancing drone-based fire detection with flame-specific attention and optimized feature fusion","authors":"Qiang Wang ,&nbsp;Shiyu Guan ,&nbsp;Shuchang Lyu ,&nbsp;Guangliang Cheng","doi":"10.1016/j.jag.2025.104655","DOIUrl":"10.1016/j.jag.2025.104655","url":null,"abstract":"<div><div>Recent advancements in drone-based fire early warning technologies have significantly improved fire detection, particularly in remote and forested areas where drones are widely utilized. However, the constrained battery life and limited computational resources of drones present challenges for real-time fire detection. Existing methods primarily focus on fire target identification without considering the distinct color and thermal characteristics of flames, leading to suboptimal detection accuracy. To address these issues, we propose a Flame-Specific Attention (FSA) mechanism, which integrates heat conduction principles and flame shape features to enhance receptive field expansion while maintaining computational efficiency. Furthermore, the Neck of the model is optimized with a Focal Modulation module to improve feature fusion, and a variable multi-attention detection head is introduced to refine detection precision. Experimental results on our Comprehensive Fire Scene Dataset (containing 3,905 images) demonstrate that our model achieves a mean Average Precision ([email protected]) of 87.7%, surpassing both Vision Transformers (ViTs) and traditional CNN approaches. Compared to the YOLOv10 baseline, our approach improves precision by 5.7% while maintaining an inference speed of 182 FPS, enabling real-time deployment in edge-computing scenarios such as drone-based fire detection. Additionally, the model effectively detects small- and medium-sized flames, reducing false positives in challenging lighting conditions (e.g., sunset and urban illumination). These enhancements make our approach highly suitable for early fire warning applications in forest and urban environments.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"142 ","pages":"Article 104655"},"PeriodicalIF":7.6,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144307088","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":"Speeding-up diffusion models for remote sensing semantic segmentation","authors":"Ruben Pascual ,&nbsp;Christian Ayala ,&nbsp;Ruben Sesma ,&nbsp;Aranzazu Jurio ,&nbsp;Daniel Paternain ,&nbsp;Mikel Galar","doi":"10.1016/j.jag.2025.104636","DOIUrl":"10.1016/j.jag.2025.104636","url":null,"abstract":"<div><div>Denoising Diffusion Probabilistic Models (DDPMs) have demonstrated exceptional potential across various generative modeling tasks. Despite evident promise in semantic segmentation, their adoption for remote sensing remains limited primarily due to computationally demanding inference. While initial approaches using DDPMs in remote sensing achieve competitive accuracy with state-of-the-art models, the multi-step nature of their image generation process poses a major bottleneck. To address this limitation, this paper investigates three key strategies for accelerating inference: (1) optimizing training and inference steps, (2) applying DDPM acceleration techniques adapted to segmentation task (including Denoising Diffusion Implicit Models, Improved Denoising Diffusion Models, and Progressive Distillation), and (3) thoroughly analyzing the trade-off between accuracy improvement and additional inference time when using test-time augmentation. These strategies are extensively tested with two established remote sensing semantic segmentation datasets focused on buildings and roads. Finally, we compare the optimized diffusion-based model with state-of-the-art convolutional-based models in terms of accuracy and inference times, showing the narrowing gap between both approaches and the increasing viability of diffusion-based segmentation for practical applications.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"142 ","pages":"Article 104636"},"PeriodicalIF":7.6,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144313382","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":"Fine-scale Antarctic grounded ice cliff 3D calving monitoring based on multi-temporal UAV photogrammetry without ground control","authors":"Shuhang Zhang ,&nbsp;Lei Zheng ,&nbsp;Huizhou Zhou ,&nbsp;Qiuyang Zhao ,&nbsp;Jie Li ,&nbsp;Yinyue Xia ,&nbsp;Wuming Zhang ,&nbsp;Xiao Cheng","doi":"10.1016/j.jag.2025.104620","DOIUrl":"10.1016/j.jag.2025.104620","url":null,"abstract":"<div><div>Grounded ice cliffs in Antarctica contribute directly to global sea level rise through calving, yet they are less studied compared to floating ice shelves. Traditional satellite remote sensing methods face limitations in temporal and spatial resolution for monitoring these small-scale calving events. This research aims to develop a method for fine-scale monitoring of grounded ice cliff calving using multi-temporal UAV photogrammetry without ground control points, with assistance from in-situ snow pit measurements. The methodology involves co-aligning UAV images to create consistent 3D models, followed by detecting calving events based on significant local volume changes between successive models. Noise filtering strategies were implemented to exclude outliers, and calving volume and mass were calculated using measurements of snow depth and density. Our method was conducted near the China’s Qinling Station on Inexpressible Island in Victoria Land, East Antarctica, and achieved a co-alignment accuracy of 2.69 cm. Over a 26-day observation period with 10 repeated flights, 44 calving events were identified along the coastline of 0.89 km, resulting in a total calving volume of 4506.69 m³ and a mass of 3078.45 tons. The average calving rate was determined to be 262.93 tons per kilometre per day along the coastline. This study demonstrates the effectiveness of fine-scale monitoring of grounding ice-cliff calving using UAV photogrammetry in polar environments. Further research is necessary to identify the spatial distribution of the Antarctic grounded ice cliffs and to quantify the total mass loss.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"142 ","pages":"Article 104620"},"PeriodicalIF":7.6,"publicationDate":"2025-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144307087","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":"Model morphing supported large scale crop type mapping: A case stuy of cotton mapping in Xinjiang, China","authors":"Longcai Zhao ,&nbsp;Taifeng Dong ,&nbsp;Xin Du ,&nbsp;Bing Dong ,&nbsp;Qiangzi Li","doi":"10.1016/j.jag.2025.104667","DOIUrl":"10.1016/j.jag.2025.104667","url":null,"abstract":"<div><div>Long-term, large-scale crop distribution mapping is crucial for agricultural policy and resource management. While high-resolution multispectral remote sensing has been widely used for crop type mapping, three major challenges remain: 1) spatiotemporal heterogeneity in cloud-free and shadow-free observations, 2) the lack of sufficient ground truth samples, and 3) limited generalization of identification models over extended periods. To address these challenges, this paper constructs a time-continuous sequence model that captures the unique feature pattern between the target-crop and non-target crops (referred to as the knowledge model). Specifically, a morphing approach was first employed to interpolate intermediate models between two pre-trained non-adjacent knowledge models. Then, a date-continuous sequence model that estimate the probabilistic of growth patterns of target crop was generated. This date-continuous sequence model mitigates spatiotemporal heterogeneity issues at the pixel level across large regions. Additionally, crop-specific knowledge model addresses sample scarcity and enhances generalization during long-term applications. The method was test using a long-term cotton mapping task (2000, 2005–2023) in Xinjiang, China. The results demonstrate that: 1) The sequence of knowledge model can effectively capture feature differences between cotton and non-cotton throughout the growing period, resulting in knowledge feature has a higher separability compared to original spectral and vegetation index features; 2) Segmenting knowledge features with Unet enables effective mapping cotton and non-cotton without ground samples. The estimated planting area from our mapping results shows excellent consistency with official statistics (R² = 0.97). The correlation between our 2018–2021 results and previously published data reached 0.8, 0.88, 0.88, and 0.89. 3). The stable and excellent mapping accuracy proves that resonation of connectivity and reachability in parameter space between two networks with identical architecture, and model morphing is a feasible way to overcome the spatial–temporal heterogeneity in valid observations in large regions.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104667"},"PeriodicalIF":7.6,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144298126","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":"Attention mechanism augmented random forest model for multiple air pollutants estimation","authors":"Xinyu Yu ,&nbsp;Man Sing Wong ,&nbsp;Kwon-Ho Lee","doi":"10.1016/j.jag.2025.104661","DOIUrl":"10.1016/j.jag.2025.104661","url":null,"abstract":"<div><div>Machine learning techniques based on satellite observations energize the derivation of near-surface air pollutant concentrations. However, most of previous studies mainly focused on estimating single air pollutant concentration, ignoring the interactions and dependencies between different air pollutants. Therefore, we proposed a Multiple Pollutants simultaneous estimation method based on Attention mechanism augmented Random Forest model (MPA-RF), including PM_2.5, PM₁₀, O₃, NO₂, CO and SO₂. Specifically, self-attention mechanism was incorporated with the multi-output random forest first to emphasize pertinent features in inputs during model training. Additionally, the multi-head self-attention was also integrated to derive the interactions and temporal dependencies of different air pollutants from historical data. Satellite observations from Advanced Himawari Imager (AHI) in three major urban agglomerations in China were extracted to demonstrate the model performance using sample- and site-based cross-validation schemes. Results elucidate that the proposed model is capable of deriving simultaneous estimations of six air pollutants with high accuracy, R² ranging from 0.74 to 0.93. Benefiting from the consideration of interactions and dependencies between different air pollutants, the proposed model outperforms other single-task contrast models with an R² improvement ranging from 9% to 26%. Moreover, the derived seamless estimations offer a basis for air pollution spatio-temporal patterns and dynamic evolution analysis with time-saving and efficient manner.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104661"},"PeriodicalIF":7.6,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144291355","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":"Shape distortion analysis of hexagonal discrete global grid systems based on standard deviation of neighbor distance","authors":"Zhang Xin,&nbsp;Cao Yibing,&nbsp;Li Tingting","doi":"10.1016/j.jag.2025.104660","DOIUrl":"10.1016/j.jag.2025.104660","url":null,"abstract":"<div><div>Hexagonal Discrete Global Grid Systems (HDGGS) are spatial reference frameworks based on the spatial discretization of the Earth’s surface, dividing it into a network of uniform hexagonal cells. They have been widely applied in geospatial analysis and environmental science fields. However, the hexagonal cells are not perfectly uniform, exhibiting inevitable shape distortion and area deformation. This study proposes a neighbor distance standard deviation method to characterize the shape deformation of individual hexagonal cells, and introduces a global deformation index to assess overall distortion magnitude, thereby addressing the limitations of deformation analysis based solely on cell area and perimeter differences. Finally, we experimentally analyzed deformation characteristics across three hexagonal discrete global grid systems projection types: Fuller3H, Fuller4H, ISEA3H, ISEA4H, and Uber H3. The experimental results show that, compared with ISEA-projection-based HDGGS grids, the Fuller-projection-based HDGGS grids’ cell neighbor distance standard deviation, grid neighboring distance standard deviation, and global deformation index are 72.09 % to 76.12 %, 74.25 % to 81.92 %, and 72.25 % to 76.29 % of the former’s values, respectively. In comparison, the Gnomonic-projection-based HDGGS’s global deformation index is 55.28 % and 40.35 % of the previous two’s values, respectively. Therefore, Gnomonic-projection-based HDGGS demonstrates the least cell shape deformation and optimal equidistant characteristics. This study pioneers quantifying local distance consistency as standardized neighbor distance standard deviation, transcending conventional evaluation paradigms that solely rely on cell perimeter mean squared error or compactness. Experimental results confirm the method’s effectiveness in quantitatively evaluating HDGGS shape deformation, providing a decision-support tool for HDGGS projection type selection.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104660"},"PeriodicalIF":7.6,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144298127","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":"Delineating flood susceptibility zones using novel ensemble models – An application of evidential belief function, relative frequency ratio, and Shannon entropy","authors":"Samuel Yaw Danso ,&nbsp;Yi Ma ,&nbsp;Isaac Yeboah Addo","doi":"10.1016/j.jag.2025.104669","DOIUrl":"10.1016/j.jag.2025.104669","url":null,"abstract":"<div><div>This paper contributes to developing novel ensemble models for delineating flood-prone areas in a West African context. One critical West African city with a history of flooding in Ghana, the Cape Coast Metropolis (CCM), was chosen with flood inventories, comprising 70% training and 30% validation, prepared as the basis for accurate prediction modeling. Furthermore, 13 conditioning parameters were chosen via multicollinearity evaluation. Three bivariate statistical algorithms, namely evidential belief function (EBF), relative frequency ratio (RFR), and Shannon entropy (SE) were combined through basic arithmetic operations to produce nine ensemble scenarios. Model performances were adjudged using the area under receiver operating characteristic curve (AUC/ROC) ratings and the overall best-performing model with a predictive accuracy of 99.6% was selected. Based on the findings, CCM’s total area was categorized into very low (20.0%), low (22.1%), moderate (20.2%), high (18.8%), and very high (18.9%) susceptibility zones. Moreover, the resultant map revealed middle portions down to the coast are most sensitive to floods compared to the northern part due to flat slope surfaces, decreasing vegetative cover, and low elevated lands. These delineated flood zones have substantial implications for national and local flood management to proactively plan and manage floods within the region and contribute to the global agenda of sustainable cities and communities.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104669"},"PeriodicalIF":7.6,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144279365","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":"Synergistic use of multi-sensor satellite data for mapping crop types and land cover dynamics from 2021 to 2023 in Northeast Thailand","authors":"Savittri Ratanopad Suwanlee ,&nbsp;Surasak Keawsomsee ,&nbsp;Emma Izquierdo-Verdiguier ,&nbsp;Álvaro Moreno-Martínez ,&nbsp;Sarawut Ninsawat ,&nbsp;Jaturong Som-ard","doi":"10.1016/j.jag.2025.104673","DOIUrl":"10.1016/j.jag.2025.104673","url":null,"abstract":"<div><div>Accurate and timely information on the spatiotemporal distribution of crops is essential for sustainable agricultural practices and ensuring food security. The significant challenges persist in accurately classifying crop types in highly fragmented cropland regions characterized by small field sizes, complex landscapes, and highly frequent cloud cover. This study presents a novel classification workflow designed to generate archaic/historic and reliable land cover (LC) maps from integrating time series data from multiple EO sources—Sentinel-1, Sentinel-2, and the Highly Scalable Temporal Adaptive Reflectance Fusion Model (HISTARFM)—with the Random Forest (RF) classifier and cloud computing technology. To the evaluate the effectiveness of this approach, Northeast (NE) Thailand was selected as a case study region, focusing on the classification of 14 crop types between 2021 and 2023. Different combinations of EO datasets and a RF classifier were evaluated using a substantial dataset of 13,453 reference points. The crop type/LC transitions from 2021 to 2023 were then analysed and a temporal transfer model was employed to map historical crop fields. The combined all EO datasets in this work achieved high overall accuracy and F1 scores (&gt;85 %) with the high spatial consistency of crop fields when compared to the use of combined both datasets. Results demonstrated the high potential and excellent efficiency of the RF, utilising an extensive reference dataset and the continuous temporal monthly information of gap-filled data. The most dominant crops were rice, followed by cassava, sugarcane and rubber trees throughout the three study years. The transfer learning RF model proved effective in mapping historical crop types and LC even when ground data was limited. Transitions of 7,287 km² (∼5%) appeared from 2021 to 2022, with major crop decreases in rice and sugarcane. From 2022 to 2023, cropland changes totaled 8,466 km² (∼6%), primarily as reductions in sugarcane and rubber trees. Our findings highlight the effectiveness of integrating multiple EO datasets in this study for mapping crop types across large areas and confirm the benefit of using monthly temporal data to obtain historic LC maps, providing valuable insights for a large range of stakeholders.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104673"},"PeriodicalIF":7.6,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144279229","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}