Semantic segmentation of cropland is critical for accurately extracting crop distribution from satellite remote sensing (RS) images. However, the dynamic temporal patterns caused by crop rotations and the heterogeneous spatial characteristics of cropland pose significant challenges for achieving high-precision segmentation. To tackle these issues, we propose a novel spatiotemporal feature-enhanced network (STFE) designed specifically for cropland segmentation in remote sensing time-series images (RSTI). The STFE network effectively integrates temporal and spatial features by introducing key innovations. First, we design an edge-guided spatial attention (EGSA) module to enhance spatial detail extraction, particularly for delineating ambiguous boundaries. Second, a progressive feature enhancement (PFE) strategy is developed to capture and fuse multi-scale features progressively across network layers. Third, for temporal feature extraction, we incorporate a differential awareness attention (DAA) module, built on ConvLSTM, to dynamically aggregate temporal information, enabling the model to better capture crop rotation patterns and temporal variations. Experimental results on three benchmark datasets—PASTIS, ZueriCrop, and DNETHOR—demonstrate the superior performance of STFE compared to state-of-the-art methods, achieving mean IoU improvements of 3.2% over the best-performing baseline. The model excels particularly in handling challenging scenarios such as irregular crop shapes and mixed cropping patterns. Its adaptability to complex and evolving agricultural landscapes provides a scalable and reliable solution for supporting sustainable farming practices and informed decision-making.

How to cite: chang, M. and li, S.: Cropland segmentation leveraging a synergistic edge enhancement and temporal difference-aware network with Sentinel-2 time-series imagery, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20365, https://doi.org/10.5194/egusphere-egu25-20365, 2025.

KARI (Korea Aerospace Research Institute) is responsible for operating three geostationary satellites, including the GK-2A (Geo-KOMPSAT-2A), and plays a key role in ensuring the stable operation of these satellites. The institute contributes to the continuous acquisition of satellite imagery data, providing around-the-clock support for satellite operations and monitoring. GK-2A was launched on July 5, 2018, from the Guiana Space Centre in French Guiana. It is part of Korea's next-generation geostationary meteorological satellite program, designed to enhance weather forecasting capabilities. This satellite is equipped with advanced payloads, including the Advanced Meteorological Imager (AMI) and the Korea Space Environment Monitor (KSEM). The AMI is dedicated to observing atmospheric conditions in real time, providing high-resolution imagery for weather analysis and forecasting, while the KSEM is tasked with monitoring space weather phenomena, such as solar radiation and geomagnetic storms, which can impact satellite operations and communication systems. The AMI operates across multiple spectral bands, enabling detailed observations of clouds, precipitation, and other atmospheric phenomena. It covers a wide area, including the East Asia region, with a temporal resolution that allows for frequent imaging of the Earth’s atmosphere. One of the AMI observation modes, Local Area (LA), typically covers the Korean Peninsula. However, in the case of special observations, the AMI can perform LA observations, where it is capable of imaging any region within its Field of View (FOV), beyond the standard observation area. This flexibility enhances its capacity for targeted monitoring, making it particularly useful for high-priority events, localized weather phenomena and global disasters requiring rapid observations.
This paper presents an overview of the operational results since the launch of the GK-2A, with a particular focus on special observations conducted using the AMI. The results from the special observation operations during the normal operational period of GK-2A are expected to provide insights into the future direction for the development of special observation operations using the AMI. 

How to cite: Kim, H.-W. and Lee, S. C.: Operational Results of GK-2A AMI Special Observations, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17511, https://doi.org/10.5194/egusphere-egu25-17511, 2025.

The dense Satellite Images Time Series (SITS) plays an important role in the agriculture semantic segmentation task. However, in real-world scenarios, cloud contamination and temporary sensor outages can lead to significant data missing in SITS, which declines the performance of models trained on ideal scenarios. A common approach is to reconstruct the complete SITS before the model’s prediction, where the reconstruction is independent of the prediction. This approach not only leads to the error accumulation from reconstruction to prediction, but also the detailed rebuilding of complete SITS may be redundant for the prediction. In this paper, we proposed a features reconstruction and prediction joint learning framework. The collaborative optimization of the two tasks aims to encourage the model to efficiently reconstruct complete features beneficial for prediction from incomplete SITS. Specifically, we simulate the data-missing scenarios with masks. The prediction task of masked data is supervised by labels. Meanwhile, by using the model that is well-trained on ideal scenarios as a teacher, we leverage its extracted temporal features from the data before masking as the target of the feature reconstruction task. The gradient flow of two tasks will be shared, which enables mutual supervision between them. Feature reconstruction prevents the model from acquiring incorrect reasoning ability caused by the shortest path problem during prediction, whereas prediction keeps reliability and reduces redundancy of reconstructed information. Furthermore, after training with the proposed framework, the model architecture remains unchanged and still maintains its robustness of complete SITS, which enhances the model's feasibility in practical applications. The experiments were conducted across multiple agricultural semantic segmentation datasets with incomplete SITS, sourced from Sentinel-2 and Planet satellites. We also validate its robustness for the common model architectures, and visualize the intermediate features to explore the mutual influence between the two tasks.

How to cite: Wang, Y., Belgiu, M., Hu, A., Xiao, R., and Tao, C.: A Features Reconstruction and Prediction Joint Learning Framework with Incomplete SITS for Agriculture Semantic Segmentation , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19832, https://doi.org/10.5194/egusphere-egu25-19832, 2025.

Global water occurrence data derived from satellite imagery provide critical insights into surface water dynamics, informing science and management of key issues like climate change, water scarcity, and biodiversity loss. The Landsat-based Global Surface Water (GSW) dataset (Pekel et al., 2016) has notably provided an important archive of the global surface water areas and its changes over time. However, its 30-m resolution limits its applicability for smaller river systems. Since the launch of the Copernicus Sentinel-2 (S2) program, higher-resolution imagery (10 m) at recurrence times of 5 days is available, but has not yet been fully exploited. The only large-scale, temporally explicit layer of water occurrence based on S2 was provided by Yang et al. (2020) for the French Metropolitan region, but is limited by noise from clouds, terrain shadows, and seasonal snow. 

The recently developed Dynamic World database (Brown et al., 2022) provides a probabilistic, pixel-scale land cover classification of S2 images updated globally in near-real time, potentially enabling computationally efficient, temporally continuous water mapping at high resolution. Here we evaluate DW’s water detection capabilities and propose a workflow for large-scale, monthly surface water occurrence mapping. Our approach integrates probabilistic and physical-based water classification, topographic filtering, and cloud masking to overcome limitations of GSW and existing Sentinel-2 applications. DW’s water probabilities were compared to spectral indices (NDWI, MNDWI) and combinations of these metrics were explored. We also assessed the potential for topographic data (FABDEM) and pixel-quality measures (CloudScore+) to reduce misclassification and allow the inclusion of more observations. The analysis is applied to the French Rhône-Mediterranean basin, a region chosen due to its diverse hydrological, climatic and geomorphological conditions. Verification is performed using a recently developed high-resolution annual land use product for mainland France (Manière, 2023) and results are compared to the GSW layer.

Preliminary results demonstrate that DW natively detects water in most areas well, but noise from shadow remains a challenge. Through combination with NDWI and further filtering with topographical data, significant classification improvements can be achieved. In addition, the pixel-based cloud-filtering with CloudScore+ enables the inclusion of more observations compared to previous methods. We implemented this approach on Google Earth Engine with a simple and efficient algorithm providing monthly water occurrence observations for a whole year. This scalable workflow holds the potential to address significant limitations of prior methods and facilitate large-scale surface water mapping at high resolution. The results are especially significant in areas where in-situ hydrological monitoring is scarce.

References

Brown, C. F., Brumby, S. P., Guzder-Williams, B., et al. (2022). Dynamic World, Near real-time global 10 m land use land cover mapping. Scientific Data, 9(1), Article 1. https://doi.org/10.1038/s41597-022-01307-4

Manière, L. (2023). Projet MAPD’O. https://bassinversant.org/wp-content/uploads/2023/03/presentation_mapdo.pdf

Pekel, J.-F., Cottam, A., Gorelick, N., & Belward, A. S. (2016). High-resolution mapping of global surface water and its long-term changes. Nature, 540(7633), Article 7633. https://doi.org/10.1038/nature20584

Yang, X., Qin, Q., Yésou, H., et al. (2020). Monthly estimation of the surface water extent in France at a 10-m resolution using Sentinel-2 data. Remote Sensing of Environment, 244, 111803. https://doi.org/10.1016/j.rse.2020.111803

How to cite: Helling, L., Belletti, B., Messager, M., Rey, L., Parmentier, H., and Piégay, H.: Towards Sentinel-2-based monthly water occurrence mapping with the Dynamic World data suite, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18783, https://doi.org/10.5194/egusphere-egu25-18783, 2025.

Accurate and efficient mapping of crop spatial distribution is crucial for agricultural monitoring, yield prediction, and environmental sustainability.  In this study, we developed a novel workflow, GEDI-Guided Crop Mapping Framework (GGCMF), for high-resolution mapping of corn and sorghum by integrating GEDI data, Sentinel-2 imagery, and machine learning classifiers within the Google Earth Engine (GEE) platform. The GGCMF workflow begins by utilizing historical CDL crop type maps to extract canopy height and vertical structural differences from GEDI L2A Vector data, which are processed within a newly developed GEE-compatible framework.  This ensures minimal geolocation errors and allows the accurate differentiation of high- and low-vegetation classes (e.g., corn + sorghum vs. other crops).  Subsequently, Sentinel-2 imagery is employed to capture unique phenological and spectral features, enabling the generation of high-quality training samples for the fine-scale differentiation of corn and sorghum.

This automated approach was applied to multiple years (2019–2022) and regions (China and the U.S.), assessing its transferability and robustness.  Validation of corn classification achieved an average overall accuracy (OA) of 0.91, with strong correlations to independent labels, published mapping products (R² = 0.98), and official statistics (R² = 0.96).  The current results for corn show that the GGCMF method is not only highly accurate but also robust across different temporal and spatial scales. The integration of GEDI and Sentinel-2 data within GEE offers a cost-effective and scalable solution for mapping structurally distinct crops.  By leveraging GEDI's canopy height data for automatic labeling and combining it with Sentinel-2's high-resolution imagery, GGCMF presents a novel, automated workflow for crop mapping.  This approach has significant potential for large-scale agricultural monitoring, providing timely and reliable data to support sustainable agricultural management.

How to cite: Li, Z.: Automated and Scalable Corn and Sorghum Mapping Across Diverse Regions Using GEDI and Time-Series Sentinel-2 Imagery, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16846, https://doi.org/10.5194/egusphere-egu25-16846, 2025.

High Altitude Platform Stations (HAPS) are an emerging technology solution for Earth Observation that are beginning to reach fruition. Operating for weeks at a time in the lower stratosphere (~ 20 km, 65,000 ft), the solar-powered long endurance aircraft provide a transformative ability to monitor areas of interest at unprecedented temporal and spatial resolution. At these improved resolutions, HAPS provide a significant advantage over satellite-based sensors and have a broad range of scientific and operational applications. Large-scale deployment of HAPS technology will revolutionise Earth Sciences with direct benefit to any science question attempting to improve understanding of Earth-surface system processes. Key industry applications include environmental monitoring, precision agriculture, forestry, smart cities and atmospheric sounding. The technology could play a critical operational role in advancing maritime domain awareness and disaster response.

At Kea Aerospace we are currently conducting flight operations with our Mk1 Kea Atmos aircraft capable of stratospheric flight to an optimal altitude of ~50,000 ft. The Mk1 has a 12.5 m wingspan and can deploy a 2.5 kg payload, with a 200L x 200W x 300H mm volume. The average payload power consumption will influence the mission profile and thermal control requirements with power availability mission and payload specific. We aim to deploy optical hyperspectral, synthetic aperture radar and atmospheric sampling instrumentation with key scientific and industry partners including the National Aeronautics and Space Administration (NASA) and German Aerospace Center (DLR).

We present preliminary findings from our Mk1 flight test programme in New Zealand and an overview of our future aspirations and upcoming Mk2 stratospheric long endurance aircraft programme.

How to cite: Price, D., Sueltrop, P., Rocket, M., Fladeland, M., and Baumgartner, S.: High Altitude Platform Stations: A Novel Earth Observation Technique from the Stratosphere, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13947, https://doi.org/10.5194/egusphere-egu25-13947, 2025.

As the Copernicus program matures, ever more gridded data becomes available to researchers to incorporate into their studies. This data is partially raw satellite data, but an increasing amount of derived products are becoming available. In addition, data from various terrestrial sources is being aggregated to gridded formats, enabling integration with products derived from satellite data.

The technologies available for provision, sharing and processing of gridded data have traditionally been developed by the EO community, with functionality tailored towards the requirements of satellite data. Where these technologies have been applied to the more terrestrial products, both derived from satellite data as well as that generated from terrestrial sources, gaps become apparent in the metadata provided. 

These gaps pertain to concepts not required for satellite data, as they are either not relevant, or have clear default values. For example in ISO 19123-1:2023, while one can define if the value being provided pertains to the center of the grid cell or one of the corners (Pixel-in-center, pixel-in-corner), it is not possible to indicate that the value pertains to the entire area of the cell, as required for land cover or population grids.

A further gap becomes apparent regarding the Observable Property that is being conveyed by the provided data. When dealing with satellite data, the only Observable Property being provided tends to be radiance, the only additional metadata to be provided details the individual frequency bands. When dealing with terrestrial products there are almost infinite lists of Observable Properties for which data is collected or generated, clean indication of what exactly the data represents is essential.

In some cases such information is provided through the use of relevant extensions, e.g. the STAC raster extension, that foresees a link to a semantic resource defining what the data actually represents. However, often, this information is not provided in a structured form. The user must extract this information from textual documentation to understand what the data actually represents. 

Proper provision of Observable Property concepts with gridded data would greatly enhance both data discoverability and reuse, as essential concepts describing the data are cleanly exposed, not requiring the user to guess from titles or poorly defined keywords. Proper integration of Observable Property concepts in core metadata structures would greatly increase the FAIRness of provided data.

Further issues encountered in sharing gridded data from diverse sources have to do with the currently available standardized web services and APIs. OGC WCS has been shown to have integral errors providing data over time, while work on OGC API - Coverage has yet to be completed. The openEO API is an interesting alternative, but as this is a processing API, deployment of this API purely for data accessibility entails a great deal of unnecessary overhead.

In conclusion, in order to reap the potential that can be gained from the diverse gridded data products emerging from both terrestrial and satellite sources, there are still a number of issues to be resolved, both in their description and accessibility.

This work was enabled by the FAIRiCUBE EU Horizon Project.

How to cite: Schleidt, K. and Jetschny, S.: Between Heaven and Earth, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12032, https://doi.org/10.5194/egusphere-egu25-12032, 2025.

Satellite imagery is essential for continuously monitoring Earth phenomena, detecting disasters and hazards, and effectively identifying large and small-scale changes across wide areas. Over the past few decades, advancements in satellite technology have significantly increased the use of satellite imagery. In particular, in change detection studies or disaster monitoring research utilizing multi-temporal and multi-satellite imagery, the fusion of images from two or more time periods for the same region is indispensable. However, due to the inherent characteristics of satellite imagery being captured from a distance, geometric distortions are likely to occur, potentially resulting in misalignment between the images and the actual ground surface. The accuracy of high-resolution satellite imagery is determined by the precision of geometric corrections, which becomes an even more critical factor when using multi-satellite and multi-temporal imagery. Consequently, image registration is an essential process in studies that utilize the fusion of high-resolution satellite imagery. In this study, we propose a highly accurate image registration method using high-resolution satellite imagery from CAS500-1, KOMPSAT-3A, and KOMPSAT-3. To overcome the limitations of feature point detection, a ResShift-based super-resolution technique was applied to generate a dataset with higher resolution than the original data, maximizing the performance of the feature matching models. For deep learning-based feature point detection and matching models, SuperPoint, SuperGlue, LightGlue, and RoMa were utilized. Notably, the RoMa model demonstrated exceptional performance by recording over 2,300 correct matches on the super-resolved dataset. The results of this study are expected to contribute to effective image registration in various fields that utilize multi-temporal and multi-satellite imagery.

This work is supported by the Korea Agency for Infrastructure Technology Advancement(KAIA) grant funded by the Ministry of Land, Infrastructure and Transport (Grant RS-2022-00155763).

How to cite: Im, Y. and Lee, Y.: Image Registration of CAS500-1 and KOMPSAT-3/3A Satellite Images Using Deep Learning-Based Feature Matching and Super-Resolution Techniques, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6359, https://doi.org/10.5194/egusphere-egu25-6359, 2025.