The increasing spatial resolution of satellite sensors - now reaching sub-meter scale - enables advanced applications in aquatic remote sensing, including highly detailed bathymetry or habitat mapping. However, this fine resolution also increases pixel-level susceptibility to wave glint and whitecap contamination. These wave-induced reflectance anomalies (WIA) can severely degrade image quality and reduce the accuracy of derived products. Although strategic acquisition planning can help to minimize WIA, users often have limited control over satellite tasking. As a result, WIA contamination is difficult to prevent in practice. In this study, we present a novel WIA correction method for pan-sharpened Pl & eacute;iades-1 imagery that detects and reconstructs affected pixel values. WIA features are identified with an ensemble of similarity-weighted support vector machines, which exploit spatial curvature and near-infrared reflectance. Detected anomalies are then corrected with a spatially adaptive gap-filling technique. Our method operates on top-of-atmosphere reflectances, allowing users flexibility in further data preprocessing steps like atmospheric correction. Our approach demonstrates very high accuracy in WIA detection and shows comparable variance reductions in optically deep water compared to the popular Hedley glint correction and the water anomaly filter. In optically shallow environments, our method outperforms these benchmarks by effectively removing WIA while preserving benthic spatial detail. Requiring minimal parameter tuning, the approach is well-suited for integration into automated preprocessing pipelines. It provides a practical tool for improving data quality and enhancing the reliability of downstream applications. Our approach is also easily adaptable to other very high-resolution sensors with similar spatial resolutions.

Accurate shallow-water bathymetric mapping is essential for understanding coastal dynamics, supporting marine infrastructure design, and managing shallow-water ecosystems. Bathymetric LiDAR surveys are expensive, spatially limited, and not routinely performed everywhere. NASA's ICESat-2 mission, equipped with a photon-counting Advanced Topographic Laser Altimeter System (ATLAS), enables cost-effective depth retrieval in optically shallow waters by detecting photon returns from subaqueous surface features. To address the sparse coverage and variable signal quality of ICESat-2 bathymetry, we introduce a photon confidence-aware multimodal machine-learning framework that integrates ICESat-2 photon returns with airborne multispectral data. Using ICESat-2 depth estimates, MagicBathyNet imagery, and co-registered reference bathymetry from our primary site in Puck Lagoon, Poland, we trained and validated random forest (RF) and extreme gradient boosting (XGB) models. RF was more robust when training labels were sparse or noisy, whereas XGB achieved slightly lower errors when high-confidence labels were abundant. We further propose a scene-level bias correction step using sparse, high-confidence ICESat-2 photons. When correction pings were selected from the higher-confidence YAPC regime (150–255) rather than the lower-confidence regime (0–150), MAE and RMSE reductions were generally larger. This indicated that ICESat-2 photon-ping confidence influences the effectiveness of bias correction applied to predicted depths. RF achieved the lowest corrected errors when both training samples and correction pings were drawn from the high-YAPC category. A cross-site validation over Agia Napa, Cyprus, showed the same confidence-regime ranking. Our framework provides a scalable, cost-effective alternative to LiDAR surveys and establishes practical guidelines for confidence-aware photon selection, bias correction, and cross-site validation.

Accurate shallow-water bathymetric mapping is essential for understanding coastal dynamics, supporting marine infrastructure design, and managing shallow-water ecosystems. Bathymetric LiDAR surveys are expensive, spatially limited, and not routinely performed everywhere. NASA's ICESat-2 mission, equipped with a photon-counting Advanced Topographic Laser Altimeter System (ATLAS), enables cost-effective depth retrieval in optically shallow waters by detecting photon returns from subaqueous surface features. To address the sparse coverage and variable signal quality of ICESat-2 bathymetry, we introduce a photon confidence-aware multimodal machine-learning framework that integrates ICESat-2 photon returns with airborne multispectral data. Using ICESat-2 depth estimates, MagicBathyNet imagery, and co-registered reference bathymetry from our primary site in Puck Lagoon, Poland, we trained and validated random forest (RF) and extreme gradient boosting (XGB) models. RF was more robust when training labels were sparse or noisy, whereas XGB achieved slightly lower errors when high-confidence labels were abundant. We further propose a scene-level bias correction step using sparse, high-confidence ICESat-2 photons. When correction pings were selected from the higher-confidence YAPC regime (150-255) rather than the lower-confidence regime (0-150), MAE and RMSE reductions were generally larger. This indicated that ICESat-2 photon-ping confidence influences the effectiveness of bias correction applied to predicted depths. RF achieved the lowest corrected errors when both training samples and correction pings were drawn from the high-YAPC category. A cross-site validation over Agia Napa, Cyprus, showed the same confidence-regime ranking. Our framework provides a scalable, cost-effective alternative to LiDAR surveys and establishes practical guidelines for confidence-aware photon selection, bias correction, and cross-site validation.

Grasslands deliver a variety of ecosystem services, such as the provision of biomass, carbon sequestration or water retention and with that play a key role for climate change mitigation and preservation of biological diversity. The intensity of grassland management directly impacts these ecosystem services and functions. However, spatial information on grassland use intensity is scarce. Remote sensing time series from optical and/or SAR sensors help to overcome this data scarcity, as they enable to derive dates and frequency of mowing events as a proxy of grassland use intensity. A growing number of published algorithms relate abrupt changes in remote sensing time series to grassland management activities either by defining threshold-based rules or by making use of machine/deep learning techniques. So far, the different algorithms have not been compared and, due to a lack of suitable reference data, have usually not been tested for spatial and temporal transferability. We present the results of a comparison exercise based on an unprecedented set of independent reference data, containing information on more than 5000 grassland mowing dates that were compiled from eight European countries over a five-year period. We analyzed the performance of ten mowing detection algorithms across different geographic regions, years, mowing intensity, levels of reference data quality, and method and satellite sensor domains. The overall results show that when using all available reference data, F1 scores ranged from 0.55 to 0.74, and from 0.56 to 0.71 when only the highest quality reference data were considered. This decision, however, reduced the number of reference events to around 1500 with a regional bias to Austria, Germany, and Switzerland. We found that algorithm performance varies across space and time and that overall, the highest accuracies were achieved by machine learning based algorithms, although not substantially outperforming rule-based algorithms. The results did not confirm a consistently positive influence of the combined use of optical and SAR data in the prediction of mowing events, but we observed variations in algorithm performance, towards the lower and higher ends of grassland use intensity. Despite testing a variety of algorithms from different method and sensor domains, we observed a general upper limit of model performance and could not identify one single algorithm that performed best in all cases. The results of this comparison exercise can guide practitioners to choose approaches and input data that are most suitable for their specific use-case. The comparison exercise also highlights the importance of consistent, representative, and reliable reference data. We therefore recommend maintaining and extending this baseline dataset for the evaluation of upcoming algorithms, Earth Observation missions, and derived products for comprehensive monitoring of grassland use intensity.

Light transmission through a sea ice cover has strong implications for the heat content of the upper ocean, the magnitude of bottom and lateral ice melt, and primary productivity in the ocean. Light transmittance in the vicinity of the Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC) Central Observatory was estimated by driving a two-stream radiative transfer model with physical property observations. Data include point and transect observations of snow depth, surface scattering layer thickness, ice thickness, and pond depth. The temporal evolution of light transmittance at specific sites and the spatial variability along transect lines were computed. Ponds transmitted 4–6 times as much solar energy per unit area as bare ice. On July 25, ponds covered about 18% of the area and contributed roughly 50% of the sunlight transmitted through the ice cover. Approximating the transmittance along a transect line using average values for the physical properties will always result in lower light transmittance than finding the average light transmittance using the full distribution of points. Transmitted solar energy calculated using the standard five ice thickness categories and three surface types used in the Los Alamos sea ice model CICE, the sea ice component of many weather and climate models, was only about 1 W m−2 less than using all the points along the transect. This minor difference suggests that the important processes and resulting feedbacks relating to solar transmittance can be represented in models that use five or more categories of ice thickness distributions.

Conservation and restoration of seagrass ecosystems critically depend on accurate habitat distribution data. Traditional monitoring methods for subtidal seagrass meadows are costly and labour-intensive for large-scale, repetitive mapping efforts. Satellite Earth Observation (EO) offer a more cost-effective alternative, but best practices for EO-based seagrass monitoring are lacking. Selecting the most suitable satellite sensor is particularly challenging and requires balancing budget and sensor specifications. The growing number of commercial, very high-resolution (<10 m) sensors further complicates this decision. To address this, we evaluate the suitability of four commonly used satellite sensors (Pléiades-1, WorldView-3, Planet SuperDove, and Sentinel-2) across seven configurations (single scenes, compositing, de-striping) for key seagrass monitoring tasks, including detecting submerged aquatic vegetation, differentiating seagrass from algae, and estimating seagrass cover. Using a sensor-agnostic pipeline for preprocessing, feature selection, and XGBoost model training, we show that Sentinel-2 and Planet SuperDove face limitations in capturing fine-scale meadow fragmentation due to their large Spatial Resolution Distance. In contrast, WorldView-3 and especially Pléiades-1 resolve meter-scale habitat fragmentation. However, Sentinel-2 and Planet SuperDove remain valuable for large-scale assessments due to significant price-per-area advantages. For spatially aggregated metrics like total vegetation cover or biomass, these sensors can provide reliable estimates when mixed-pixel effects are considered. Our results highlight the need for different preprocessing steps, such as water column correction, segmentation, compositing and de-striping, to maximize model performance based on sensor characteristics. Nevertheless, none of our tested sensors could distinguish seagrass from algae, highlighting a significant challenge for advancing EO-based seagrass monitoring in the future.

The adjacency effect (AE) in optical remote sensing emerges from atmospheric scattering of photons reflected from surrounding areas into the sensor's instantaneous field of view (IFOV). This leads to spectral blending over heterogeneous surfaces with high contrast, expanding over horizontal distances of several kilometers and distorting the target signal. Therefore, AE can cause significant errors in quantitative retrieval of surface reflectance in high-contrast environments and further derived information, such as pixel-based chlorophyll-a concentration. This review aims to make the niche topic of AE in optical remote sensing more accessible to a broader audience. It provides non-specialists with an understanding of AE emergence, implications, and correction methods suitable for aquatic high-contrast environments. We summarized key advancements in the field, highlighting studies that introduced correction methods. We guide readers on how these studies theoretically and practically work, emphasizing their strengths and weaknesses. In the highly dynamic research field, some promising AE correction algorithms have been published recently. RAdCor and T-Mart stand out with their promising results and freely accessible nature. We applied both methods to a case study of the Osterseen Lake District south of Munich (Germany), demonstrating their application and effectiveness in correcting AE in the near-infrared (NIR) while identifying room for improvement in the visible light (VIS). These findings also reflect the broader conclusions of this review. While recent algorithms mainly concentrate on operational applicability, future research could focus on expanding AE correction to non-vegetated surfaces and regions beyond mid-latitudes.

Marine plastic pollution poses significant risks to ecosystems and human health, particularly in vulnerable Arctic regions. Therefore, assessing the extent and impacts of plastic contamination in these regions through a One Health framework is essential for developing strategies to build resilience in Arctic populations and ecosystems. The Horizon Europe ICEBERG Project addresses these challenges at the ocean-coast-land continuum, working in partnership with Indigenous peoples and local communities in south Greenland, northwest Iceland, and in Svalbard. ICEBERG integrates technology-enhanced community monitoring, such as drones and time-lapse cameras, with manual beach clean-ups for marine litter, citizen science and interactive data-sharing platforms. Researchers, local communities and civil society map pollution hotspots, trace sources, and assess ecological and health impacts of macro, micro-, nanoplastics. This community-driven data collection, supported by both low-tech and advanced technologies, offers a replicable model for tracking marine plastic pollution, studying its impacts and informing local policies to mitigate its effects. ICEBERG's multidisciplinary, multi-stakeholder approach offers valuable insights into the impacts of plastic pollution in Arctic socio-ecological systems and emphasizes the need for sustainable, inclusive, and data-driven environmental management in polar regions.

Europe’s high-nature value (HNV) grasslands have significantly declined in recent decades. European conservation strategies are mainly confined to protected areas, and many national initiatives aiming for comprehensive coverage suffer from long and irregular monitoring intervals. Addressing this, we propose a data-driven approach to derive information about the location, extent and HNV status of grasslands to improve the efficiency of large-scale field mappings. Serving as a representative example of European and national grassland monitoring, we utilize the regional habitat map of Schleswig-Holstein, Germany, in conjunction with Harmonized Landsat Sentinel-2 time series data to train XGBoost models for the period of 2017-2022. Our models achieved high classification performance, distinguishing eight grassland classes with average F1-scores of 0.89 before and 0.86 after feature selection. We examined model decision-making patterns using an adapted version of SHapley Additive exPlanation values, finding that start-of-season, end-of-season, Red-Edge, and spectral change features significantly impacted predictions. We produced annual HNV grassland maps and, by aggregating yearly results, derived a robust estimate of the HNV status in our study area. Applying our HNV estimate to an independent dataset comprising 2363 km2 of grassland plots with unknown HNV status, we identified 84 km2 as HNV, highlighting the significance of our result. Overall, our study demonstrates how integrating remote sensing data enhances the efficiency and comprehensiveness of large-scale mapping initiatives.

The Arctic is one of the most vulnerable regions on Earth concerning climate change and is increasingly affected by pollution from human activities. The ICEBERG project (Innovative Community Engagement for Building Effective Resilience and Arctic Ocean Pollution-Control Governance in the Context of Climate Change) is a multidisciplinary initiative funded by the European Union. It focuses on assessing types, sources, distributions, and impacts of pollution on ecosystems and coastal communities across the European Arctic. Case studies in West Svalbard, South Greenland, and North Iceland are being used to develop community-driven strategies to enhance resilience and reduce pollution. The project addresses a range of pollutants, including macro-, micro-, and nanoplastics, ship emissions, sewage, persistent organic pollutants, and heavy metals. As part of ICEBERG, our team from the Earth Observation and Modelling (EOM) group at Kiel University deployed time-lapse cameras to monitor the accumulation of marine litter along Arctic beaches. Using machine learning, we aim to automate the detection and classification of marine litter, offering new insights into its types, sizes, and seasonal variations. The results will be combined with drone-based data and coastal marine observatory artificial intelligence processing, which aims to map and monitor the spatiotemporal trends of marine litter in specified areas. By leveraging the high temporal mapping capabilities of small drones with machine learning algorithms, combining both will offer a comprehensive and advanced method for mapping marine litter across various spatial and temporal scales.In the initial phase of ICEBERG, we deployed an autonomous camera system in West Svalbard to collect year-round data from an uninhabited site, while we held community consultation meetings in Iceland and Greenland to introduce the project and jointly explore opportunities for citizen science collaborations. By adopting a citizen science approach, we are actively partnering with academic & non-academic actors, including local and Indigenous stakeholders and non-governmental organizations in Iceland and Greenland who are supporting the installation and maintenance of the cameras. Additionally, through partnerships with high school teachers and students, we are also engaging young people to raise awareness of ongoing pollution challenges and explore actionable measures for mitigation and adaptation. By developing an interactive data-sharing platform, citizen scientists have the opportunity to upload their observations of any kind of pollution, serving as data crowdsourcing along with the data from the time-lapse cameras and drones. ICEBERG empowers communities to actively contribute to the process of identifying pollution sources, monitoring coastal litter, and developing meaningful interventions. We will present our innovative approach for monitoring pollution on Arctic beaches, emphasizing the role of community engagement and potential future co-created solutions. By integrating artificial intelligence tools and fostering local collaborations, ICEBERG offers a sustainable and inclusive approach for addressing environmental challenges in vulnerable Arctic regions. Our presentation will highlight the use of citizen science to enhance Arctic resilience and governance, share preliminary time-lapse data from Svalbard and Iceland, and explore the opportunities and challenges of community engagement in Arctic environmental monitoring.

Marine carbon dioxide removal (mCDR) and geological carbon storage in the marine environment (mCS) promise to help mitigate global climate change alongside drastic emission reductions. However, the implementable potential of mCDR and mCS depends, apart from technology readiness, also on site‐specific conditions. In this work, we explore different options for mCDR and mCS, using the German context as a case study. We challenge each option to remove 10 Mt CO2 yr−1, accounting for 8%–22% of projected hard‐to‐abate and residual emissions of Germany in 2045. We focus on the environmental, resource, and infrastructure requirements of individual mCDR and mCS options at specific sites, within the German jurisdiction when possible. This serves as an entry point to discuss main uncertainty factors and research needs beyond technology readiness, and, where possible, cost estimates, expected environmental effects, and monitoring approaches. In total, we describe 10 mCDR and mCS options; four aim at enhancing the chemical carbon uptake of the ocean through alkalinity enhancement, four aim at enhancing blue carbon ecosystems' sink capacity, and two employ geological off‐shore storage. Our results indicate that five out of 10 options would potentially be implementable within German jurisdiction, and three of them could potentially meet the challenge. Our exercise serves as an example on how the creation of more tangible and site‐specific CDR options can provide a basis for the assessment of socio‐economic, ethical, political, and legal aspects for such implementations. The approach presented here can easily be applied to other regional or national CDR capacity considerations.

Achieving global climate goals while ensuring food security in a changing climate presents significant challenges, particularly when relying solely on land-based solutions. Covering over 70% of the Earth's surface, the ocean remains an underutilized resource for climate mitigation. Ocean alkalinity enhancement (OAE) is one such strategy, designed to strengthen the ocean's natural carbon sink, reduce atmospheric CO2, and mitigate ocean acidification. However, its implications for fisheries, critical for food security and livelihoods, remain uncertain. This study examines the interplay between global fisheries, OAE, and different future socioeconomic and climatic conditions, using the Shared Socioeconomic Pathways (SSPs) and Representative Concentration Pathways framework. We explore how global fisheries and OAE could evolve under three combined scenarios: SSP1-2.6 (sustainability-focused), SSP3-7.0 (regional rivalry), and SSP5-8.5 (high fossil fuel dependency). By integrating ecological, economic, societal, and technological perspectives, we develop scenario narratives and quantify key bio-economic parameters, including technological progress, fishing costs, fisheries management, marine aquaculture, and ecosystem carrying capacity. High-emission (SSP5-8.5) and fragmented development (SSP3-7.0) scenarios present significant barriers to the coexistence of OAE and fisheries, whereas sustainability-focused pathways (SSP1-2.6) offer the most favorable conditions for their alignment. Successfully integrating OAE with fisheries management will likely depend on technological advancements, international cooperation, and socio-economic developments. These scenarios are aligned with those used in model-based scenario studies conducted under the frameworks of the Intergovernmental Panel on Climate Change and the Intergovernmental Platform on Biodiversity and Ecosystem Services (IPBES), providing a shared foundation for future work.

Macrophytes are key species in freshwater ecosystems and ecological indicators of water quality. They produce biomass, build three-dimensional habitats and provide detritus and shelter for associated species. Since the 1990s researchers tried to model the spatial distribution of macrophytes using Species Distribution Models (SDM). Early studies relied on GIS-applications and correlative regression models, but with technological advancements, machine learning-based approaches have become dominant in lake macrophyte SDMs. Still, available modelling techniques remain diverse, each with its own strengths and limitations depending on data, scope, scale, and target species. To better understand different approaches in the context of lake macrophyte research we conducted an extensive systematic literature search on macrophyte SDMs and quantified the use of different algorithms, data, studied species and more for 80 publications. A majority of studies employed correlative approaches at local to regional scales. Furthermore, many studies focussed on widespread macrophyte species, particularly in the temperate regions of the Northern Hemisphere. Commonly used predictors include bioclimatic variables like air temperature and precipitation as well as lake-specific factors such as bathymetry or light availability. We identified a special interest of researchers in forecasts of the potential distribution of invasive macrophytes. In this context, evaluating model transferability for different macrophyte species remains an important yet partially unresolved challenge. For future research, we recommend researchers to take a critical look at chosen species, predictors, source data and study scope and report model uncertainty accordingly.

Seagrass meadows are blue carbon hotpots. Mapping is an integral part to manage and understand meadows and their dynamics and thereby assist conservation and restoration efforts. Since traditional monitoring methods often become unviable at large spatial scales, satellite remote sensing has emerged as a supplementary tool. However, this approach is often constraint by the sensors' spatial resolution and required training data. Here, we test a transformer-based approach to segment seagrass and compare it against a ResNet50 and MobileNetV3 on very high-resolution (VHR) Pléiades data (0.5 m spatial resolution). Our cross-validation approach demonstrates high performances of all methods, with transformers and CNN approaches being almost equal in performance (~ 95 %), whereby our small dataset might have have promoted overfitting. Large seagrass areas are well recognised, while very small patches of just a few pixels size are detected less accurately. However, the smallest patches detected by our models are only few meters in size, demonstrating that VHR data allows to resolve significantly more spatial details compared to data of publicly available medium-resolution sensors like Sentinel-2.

Melt ponds are a core component of the summer sea ice system in the Arctic, increasing the uptake of solar energy and impacting the ice-associated ecosystem. They were thus one of the key topics during the one-year drift campaign MOSAiC in the Transpolar Drift 2019/2020. Pond depth is a dominating factor in the description of the surface meltwater volume, necessary to estimate budgets, and used in model parametrization to simulate pond coverage evolution. However, observational data on pond depth is spatially and temporally strongly limited to a few in situ measurements. Pond bathymetry, which is pond depth spatially fully resolved, remains entirely unexplored. Here, we present a newly developed method to derive pond bathymetry from aerial images. We determine it from a photogrammetric multi-view reconstruction of the summer ice surface topography. Based on images recorded on dedicated grid flights and facilitated assumptions, we were able to obtain pond depth with a mean deviation of 3.5 cm compared to manual in situ observations. The method is independent of pond color and sky conditions, which is an advantage over recently developed radiometric airborne retrieval methods. It can furthermore be implemented in any typical photogrammetry workflow. We present the retrieval algorithm, including requirements for the data recording and survey planning, and a correction method for refraction at the air—pond interface. In addition, we show how the retrieved surface topography model synergizes with the initial image data to retrieve the water level of individual ponds from the visually determined pond margins. We use the method to give a profound overview of the pond coverage on the MOSAiC floe, on which we found unexpected steady pond coverage and volume. We were able to derive individual pond properties of more than 1600 ponds on the floe, including their size, bathymetry, volume, surface elevation above sea level, and temporal evolution. We present a scaling factor for single in situ depth measurements, discuss the representativeness of in situ pond measurements, and show indications for non-rigid pond bottoms. The study points out the great potential to derive geometric properties of the summer sea-ice surface emerging from the increasingly available visual image data recorded from UAVs or aircraft, allowing for an integrated understanding and improved formulation of the thermodynamic and hydrological pond system in models.

Melt ponds play a crucial role in the melting of Arctic sea ice. Studying the evolution of melt ponds is essential for understanding changes in Arctic sea ice. In this study, we used a revised sea ice model to simulate the evolution of melt ponds along the MOSAiC drift at a resolution of 10 m. A novel melt pond parameterization scheme simulates the movement of meltwater under the influence of gravity over a realistic sea ice topography. We evaluated different melt pond parameterization schemes based on remote sensing observations. The absolute deviation of the maximum pond coverage simulated by the new scheme is within 3%, while differences among parameterization schemes exceed 50%. Errors were found to be primarily due to the calculation of macroscopic meltwater loss, which is related to sea ice surface topography. Previous studies have indicated that sea ice with a lower surface roughness has a larger catchment area, resulting in larger pond coverage during the melt season. This study has identified an opposing mechanism: sea ice with lower surface roughness has a larger catchment area connected to the macroscopic flaws of the sea ice surface, which leads to more macroscopic drainage into the ocean and thereby a decrease in melt pond coverage. Experimental simulations showed that sea ice with 46% higher surface roughness, resulting in 12% less macroscopic drainage, exhibited a 38% higher maximum pond fraction. The presence of macroscopic flaws is related to the fragmentation of sea ice cover. As Arctic sea ice cover becomes increasingly fragmented and mobile, this mechanism will become more significant.

Melt ponds are a core component of the summer sea ice system in the Arctic, increasing the uptake of solar energy and impacting the ice-associated ecosystem. They were thus one of the key topics during the 1-year drift campaign Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) in the Transpolar Drift 2019/2020. Pond depth is a dominating factor in describing the surface meltwater volume; it is necessary to estimate budgets and used in model parameterization to simulate pond coverage evolution. However, observational data on pond depth are spatially and temporally strongly limited to a few in situ measurements. Pond bathymetry, which is pond depth spatially fully resolved, remains unexplored. Here, we present a newly developed method to derive pond bathymetry from aerial images. We determine it from a photogrammetric multi-view reconstruction of the summer ice surface topography. Based on images recorded on dedicated grid flights and facilitated assumptions, we were able to obtain pond depth with a mean deviation of 3.5 cm compared to manual in situ observations. The method is independent of pond color and sky conditions, which is an advantage over recently developed radiometric airborne retrieval methods. It can furthermore be implemented in any typical photogrammetry workflow. We present the retrieval algorithm, including requirements for the data recording and survey planning, and a correction method for refraction at the air–pond interface. In addition, we show how the retrieved surface topography model synergizes with the initial image data to retrieve the water level of individual ponds from the visually determined pond margins. We use the method to give a profound overview of the pond coverage on the MOSAiC floe, on which we found unexpected steady pond coverage and volume. We were able to derive individual pond properties of more than 1600 ponds on the floe, including their size, bathymetry, volume, surface elevation above sea level, and temporal evolution. We present a scaling factor for single in situ depth measurements, discuss the representativeness of in situ pond measurements and the importance of such high-resolution data for new satellite retrievals, and show indications for non-rigid pond bottoms. The study points out the great potential to derive geometric properties of the summer sea ice surface emerging from the increasingly available visual image data recorded from uncrewed aerial vehicles (UAVs) or aircraft, allowing for an integrated understanding and improved formulation of the thermodynamic and hydrological pond system in models.

Changes and disturbances to water diversity and quality are complex and multi-scale in space and time. Although in situ methods provide detailed point information on the condition of water bodies, they are of limited use for making area-based monitoring over time, as aquatic ecosystems are extremely dynamic. Remote sensing (RS) provides methods and data for the cost-effective, comprehensive, continuous and standardised monitoring of characteristics and changes in characteristics of water diversity and water quality from local and regional scales to the scale of entire continents. In order to apply and better understand RS techniques and their derived spectral indicators in monitoring water diversity and quality, this study defines five characteristics of water diversity and quality that can be monitored using RS. These are the diversity of water traits, the diversity of water genesis, the structural diversity of water, the taxonomic diversity of water and the functional diversity of water. It is essential to record the diversity of water traits to derive the other four characteristics of water diversity from RS. Furthermore, traits are the only and most important interface between in situ and RS monitoring approaches. The monitoring of these five characteristics of water diversity and water quality using RS technologies is presented in detail and discussed using numerous examples. Finally, current and future developments are presented to advance monitoring using RS and the trait approach in modelling, prediction and assessment as a basis for successful monitoring and management strategies.