Remote Sensing for Agriculture, Ecosystems, and Hydrology XXVI

Remote sensing technologies have substantial potential in detecting marine oil spill incidents effective for emergency response and mitigation efforts. Offshore and onshore oil spill incidents remain underreported yet are frequently occurring and challenges remain around effective targeting of oil-type specific mitigation efforts. Here we demonstrate an approach of classifying imagery acquired from RGB, thermal, multi-spectral and hyperspectral data collected from UAV platforms of simulated oil spill incidents implemented in an outdoor environment. Each dataset was classified using SVM, RF and NN The methods are being compared and assessed for accuracy using Kappa coefficient. This paper demonstrates that combining the use of hyperspectral, multispectral and thermal infrared sensors proved to effectively improve the recognition accuracy of oil types in seawater which provide key insights about the oil spill incident necessary for mitigation efforts.

This paper establishes a matching dataset between the near-infrared band of AHI and the microwave radiometer band of ATMS with in-situ SST. The matching dataset is divided into two subsets based on cloud presence at the central pixel. Three machine learning methods are trained and tested on the two sub-matching datasets to evaluate the SST retrieval accuracy under cloudy and cloud-free conditions. The experimental results show that SST retrieval using the AHI/ATMS bands as input features significantly improves the accuracy of SST data. Compared with the results of SST retrieval using only the AHI thermal infrared band as input, the AHI/ATMS metrics of R^2 increases by 7.7% and RMSE reduces by 0.896°under cloudy conditions. At the same time, the fusion of ATMS band data can effectively retrieve the SST under cloud condition. The method proposed in this paper will provide an important technical reference for SST retrieval under all-weather conditions.

In this demonstration project, DLR will collaborate with the ERATOSTHENES Centre of Excellence scientific team of the Marine Safety and Security Sector, in order to derive different information products based on SAR EO Data available from Copernicus Data Hub or DLR archive and/or to support maritime situation awareness in the eastern Mediterranean Sea in near real-time. This research capacity demonstration aims to provide the end-user(s) with SAR-derived maritime surveillance information to support the maritime domain awareness around Cyprus or for a specific area defined for this trail. For example, the demonstration could include identifying critical maritime areas and characterising typical scenarios related to maritime security aspects in the area.

Automated methods for extracting water features from multispectral and hyperspectral imagery. Automated or AI based techniques can be applied to scientifically detect subsurface bottom types and water column properties using fast computational algorithms and image processing techniques based upon synthetic channels created from multiband sensing systems. In this research a review of image analysis techniques is presented and described with the context of modern real time methods which are called artificial techniques. One basis and description of these techniques relies upon generating synthetic channels using wavelet imaging techniques in combination with multiple wavelength contrast algorithms. In this paper and presentation techniques are demonstrated using multispectral-hyperspectral mages flown over Space Coast Florida waters. The results demonstrate the value of modern image analysis approaches to examine environmental and ecologically relevant diversity indices useful for characterizing the quality of marine habitats.

The SIMAGAZ gas monitoring camera developped at Onera was deployed during the IPOMAC and MANIFEST campaigns in 2021 and 2022. These measurement campaigns were carried out several miles off the west coast of France as part of the fight against pollution at sea. Various chemicals, legally classified as HNS (Hazardous Noxious Substances), were discharged at the surface of the sea, and SIMAGAZ measurements were taken from the ship and from the air. The processing of the camera data is based on a physical radiometric model, a spectroscopic database containing the compounds of interest and includes a morphological analysis of the images. SIMAGAZ has enabled the detection, identification and quantification of several gases evaporating from spilled slicks (acetone, butyl acetate, propyl acetate, MTBE and heptane), sometimes for more than half an hour. For other products with absorption lines outside the SIMAGAZ spectral range, no gas was observed as expected. On the basis of this success, the analysis of the concentration fields obtained and their dynamics can be compared with the results of evaporation and atmospheric dispersion modelling tools for HNS.

The capability of detecting ocean depth has core implications on many marine sciences, ranging from navigation, environmental monitoring, and resource exploration. Oceanographic surveying is conventionally based on static data processing methods, where real-time adaptability and user interaction support can be compromised. In this study, we suggest a new idea of depth detection for oceans via the use of Retrieval-Augmented Generation (RAG) as a method to formalize and improve reliability, flexibility, and user-interactivity of depth mapping systems. Our method consists of processing raw sensor data through the dynamic feature extraction module that identifies essential details of the seabed and water column. One of the new wrinkles in our approach will be the ability to introduce user-driven corrections and augmentations in real-time. This allows our system to iteratively improve the generated maps by adjusting to the users (marine scientists, surveyors, educators) in near real-time, increasing the accuracy and clarity of the produced maps.

Chlorophyll a (Chla) serves as an essential pigment in phytoplankton, playing a critical role in determining phytoplankton biomass and assessing the nutrient status of aquatic environments. his study advocates a one-dimensional convolutional neural network (1D CNN) approach aimed at enhancing the accuracy of Chla concentration estimates in optically complex aquatic systems. The experimental results indicate 1D CNN surpasses state-of-the-art algorithms, including OCI, SVR, and RFR, with metrics values of 0.892 for R², 11.243 for RMSE, 0.052 for RMLSE, 1.056 for Bias, and 1.444 for MAE. 1D CNN exhibits superior inversion performance in mid- to high-latitude regions, particularly displaying enhanced detail and greater tolerance to noise in coastal water. The approach provides a reliable and enhanced technique for estimating Chla concentration in optical complex waters.

Satellite remote sensing has the potential to provide a means for global water monitoring and can offer a wider range of long-term observation of chlorophyll-a concentration products. In this study, a data-interpolating empirical orthogonal function method is explored to composite the U.S. SNPP and NOAA-20 Visible Infrared Imaging Radiometer (VIIRS) and FY-3 Meteorological Satellite Medium Resolution Spectral Imager (MERSI) satellite-derived chlorophyll-a data to produce a daily global chlorophyll-a data sets. Processes of outlier detection and removal enhance the whole performance of this interpolating technique. Satellite-derived chlorophyll-a data at the L2 and L3 levels have been utilized to recover missing information from cloudy images. The present study indicates that this technique can fill the missing chlorophyll-a concentration data from geophysical fields. The resulting chlorophyll-a concentration products reveal large and mesoscale spatial distribution patterns of marine water color, highlighting significant seasonal variation characteristics of chlorophyll-a concentration.

KaRIn (Ka-band Radar Interferometer) is the first spaceborne 2D altimeter, designed to measure water elevations from inland and marine water bodies. By their nature, tide gauges and 1D altimetry have gaps along the coastline where the sea level variability is of great importance and in areas of complex coastal morphology and bathymetry, this can lead to large uncertainties. KaRIn 2D altimetry has the potential to improve flood management assessments through a greater understanding of spatial variability of water levels due to tide, storm surge, wind, waves, and river discharge. These novel 2D data will lead to better informed and validated hydrodynamic models for storm surge and flood prediction. Data from tide gauges and Cryosat 2 have been used to validate measurements from KaRIn and to look at coastal processes and features seen in the SWOT data, including storm surge, intertidal morphology, and nearshore processes.

A passive multiplexed multi-parameter marinized sensory array is described. This was deployed in 2 different array configurations across 3 different marine sites in Orkney. The chosen deployment sites were to test the response to long period oceanic waves, the effects of passing marine traffic for harbour security and highly energetic tidal flows for power generation . The sensor array is made up of 4 measurement stations which are connected in series via an optical fibre umbilical. Across these measurement stations, a total of 16 temperature sensors, 4 attitude sensors (each consisting of 3 individual fiber sensors) and 16 flow sensors were successfully deployed. Tank testing demonstrated measurement capability in the 0.05-2.5 ms-1 range, and seaborne deployment demonstrated the system in a practical application.

Accurate bathymetry estimation is crucial for various marine and coastal applications, including resource management, navigation, and environmental protection. This study presents a comparison, on sandy bottom, of bathymetry estimated using optical satellite imagery versus bathymetric data acquired through a multibeam instrument. The study area was the Gulf of Sciacca (on the southern coast of Sicily island, Italy, Mediterranean basin). High-resolution PlanetScope SuperDove satellite images were preprocessed to obtain reflectance just below the water surface. Then, a simplified version of the original radiative transfer equation from Jain and Miller (1977) was used to estimate water depth. The model was calibrated and then validated using high-res multibeam data through a k-fold cross-validation. Multibeam data were acquired using a Reason 8125 multibeam Eco-sounder while PDS2000 software provided the functionalities for survey planning, data processing and editing. Multibeam data reveal in the western sector the presence of structural highs, shoals, rocky substrate and Posidonia oceanica meadows extending even beyond ⁓30 m, while, the easternmost sector from Verdura delta up to Capo Bianco shows the presence of sedimentary structures, prairies of Posidonia oceanica and Cymodocea nodosa, and rocky substrate of different nature. Bathymetry was estimated from optical data from the shoreline down to ⁓26 m water depth; while approaching this depth, the reflectance assumed an asymptotic value. The results show a strong correlation between the depths estimated from satellite images and those obtained from multibeam data, with a root mean square error lower than 11% with a confidence interval of 95%. However, some discrepancies were observed in areas in mixed pixels (with submerged vegetation, complex seabeds) or reflectance surface effects (glint, waves due for instance to moving boat wakes).

Chlorophyll fluorescence is the emission of light by chlorophyll molecules when they are excited by absorbed light. Chlorophyll is the pigment responsible for photosynthesis, the process by which plants and other photosynthetic organisms convert light energy into chemical energy, which varies in different values for different latitudes. The aim of the research is an attempt to compare the spectral characteristics of mosses and lichens, and note any variations in their fluorescence intensity from the region of Livingston Island, Antarctica and Rila Mountain in Bulgaria during the summer season in the southern and northern hemispheres. Field research was carried out in Antarctica and Bulgaria, in order to verify the data from Sentinel 2MSI, drone photography and photogrammetry, as well as photography by a thermal camera with a measurement accuracy of +/- 2°C and a wavelength of 8 - 14 μm . A spectrometer was used to analyze the visible range from 380 to 780 nm and the spectral range in which Sentinel 2MSI and Sentinel 3 SLSTR images are generated. The main research methods are through chlorophyll fluorescence response and the use of several optical indices for remote sensing.

Sun glint in satellite imagery of the water surface contaminates the upwelling signal received by a detector. Many models exist that attempt to correct for this wave facet effect and phenomena. In this work a model for sun glint correction is creating using the comparison of image transects between two nearly simultaneously collected images of the same area, although with differing sensor geometry. One image utilized in this research is almost entirely glint free while the other is contaminated by water wave facet glint. Although many models for removing sun glint exist based on various techniques, none are completely accurate, and there is always a need to improve our understanding of this phenomena and to decontaminate the sun glint pixels. The model developed in this research is based on the statistical properties of the images related to azimuth angles, fetch distances, wind speed and direction, and other factors in attempt to test a new mathematical model for sun glint removal.

Plant Functional Types (PFTs) categorize plant species with similar functioning, effects on ecosystems, or responses to environmental factors. Remote sensing provides biogeophysical data across temporal and spatial scales, aiding PFT extraction. Mapping methods include utilizing existing land cover datasets, decision tree classifiers, evidential reasoning, Fourier transforms, and machine learning like random forest. PFTs are characterized using variables from phenology (e.g., NDVI, EVI from MODIS), plant physiognomy (e.g., tree height from GEDI), NPP, FPAR, ET from MODIS, topography (e.g., DEM, slope, aspect), and environmental conditions (e.g., bioclimatic variables, precipitation, temperature). Variables were standardized to 500m and analyzed for multicollinearity. A random forest model in R achieved a PFT classification accuracy of 0.6769 and a kappa value of 0.6354.

Automatic and accurate identification and extraction of rich and complex mixed land cover types with weak independence can reduce the interference of human factors on classification results, greatly simplify the classification process, enrich the classification algorithm of land cover in remote sensing images and the technical platform for analysis and processing, and provide references for real-time monitoring of land resources. Took global land cover products MCD12Q1 Version 6 and Landsat-8 as the main remote sensing images, mixed and multiple land cover types with weak independence in the study area were extracted automatically and accurately by combined the SRTM V4.1 auxiliary dataset and thematic indexes to construct a multidimensional classification feature set of the spectrum, texture, and terrain.

Satellite Image Time Series (SITS) provide a unique opportunity to monitor the changes and evolution of floodplains and vegetation after the partial destruction of the Living Lab Hedwige-Prosperpolder (LLHPP, Belgium, Antwerp). In this study, the Time-Weighted Dynamic Time Warping, designed to handle irregular SITS for vegetation mapping, was combined with superpixel generation and unsupervised hierarchical clustering to perform an analysis of vegetation changes on LLHP using Sentinel-1 and Sentinel-2 images.

Potato is the world's most popular non-cereal food in terms of global food security. In the recent years, Remote Sensing and Earth Observation in general advanced the field of yield prediction and estimation using satellite-derived data. Moreover, potato is one of the most exportable agricultural products in Cyprus. The combination of Artificial Intelligence (AI) technologies in yield estimation/prediction tasks can be vital for the agricultural industry and enhance the sustainability of food market worldwide. In this study, invariant learning is compared to traditional Machine Learning in in terms of data drift problems and spatiotemporal robustness on a dataset which encompasses in-situ meteorological data, irrigation data, satellite vegetation indices and potato yield data for the period of 2017-2023.

A comparative analysis of different methods for accurate durum wheat yield assessment using Sentinel-2 satellite data is presented for 157 fields in Thessaly plain (Greece) across five growing periods. Two modeling approaches are presented: vegetation index-based multiple linear regression (VI-MLR) and machine learning algorithms (Random Forest, K-Nearest Neighbors, Boosting Regressions). The results demonstrate the potential of machine learning for accurate yield estimation, even in early growth stages. Specifically, while VI-MLR shows moderate accuracy, machine learning algorithms significantly improve it, particularly when utilizing all 12 bands from Sentinel-2. This study emphasizes the effectiveness of machine learning in capturing complex relationships from remote sensing data, underscoring its advantages over traditional methods.

Urban expansion has led to significant changes in urban green spaces impacting the urban environment and residents’ well-being. Therefore, monitoring changes in urban vegetation using remote sensing techniques is crucial. This study aims to address the limitations of traditional remote sensing techniques by integrating terrestrial laser scanning and UAV photogrammetry for change detection. The study concentrates on change detection within Helsinki's Malminkartano region during the leaf-off and leaf-on seasons for the year 2022. 3D point cloud data are compared using the M3C2-algorithm. The results illustrate their efficacy in detecting changes up to 2.8 meters. Moreover, the accuracy assessment of datasets revealed that 95% confidence threshold corresponded to approximately 4 cm differences in both TLS and UAV photogrammetry datasets. The study emphasizes on data processing uncertainties related to point density, registration, vertical height, and scale differences. Future research should address these uncertainties to ensure an accurate assessment of tree parameters.

The use of unmanned aerial spraying systems is increasing due to their many advantages, but there is a lack of research on how to evaluate their performance. In general, the spraying performance is evaluated by collecting the spray droplets with water-sensitive paper and analyzing the images. However, there is a disadvantage that the performance is affected by humidity. In this study, an image analysis program was developed to measure the spraying performance when using pigments and collectors instead of water-sensitive paper. The program was developed in Python and utilizes OpenCV related functions. To overcome the problem of binarization processing, HSV color system was used. The program is able to generate ROIs regardless of the size or shape of the collector and calculate the percentage of deposited area and droplet size distribution of the sprayed droplets.

Site Specific Weed Management is critical to achieve pesticide savings. However, as not all weed species have a negative impact on crops, a species-specific weed management can contribute even more to pesticide savings and improve biodiversity in agricultural landscapes. This requires robust classifiers for a species specific single-plant detection with the ability to generalize across different environmental conditions as well as across different plant phenotypes of the same species. As an example for this, we propose YOLOv8 as an instance segmentation classifier for the detection of five weed groups in beet crops in high resolution UAV imagery.

Crop-type mapping is crucial for precision agriculture, facilitating informed decision-making, particularly in regions where agriculture is a fundamental component of the economy. However, accurate crop-type classification impedes yield monitoring which is crucial for strategic planning. This paper presents a methodology for crop mapping, which involves integrating Sentinel-2 satellite imagery with high-resolution unmanned aerial vehicles (UAV) based images. By leveraging the multispectral bands of Sentinel-2 imagery and the finer spatial resolution of UAV-based images, our approach aims to advance the precision and efficacy of crop mapping models by enriching the information available for analysis. The integration of Sentinel-2 and UAV-based data substantially improves the identification of crop types. These findings underscore the potential of data fusion leveraging the strengths of individual sensors to help overcome weaknesses in alternative sensors. This research contributes to the advancement of precision agriculture methodologies and underscores the transformative impact of data fusion techniques in agricultural remote sensing applications.

The extensive presence of macroplastic pollutants along coastlines represents one of the most pressing environmental challenges. Accurately identifying their locations is crucial for planning effective cleanup activities aimed at their removal. RPAS offers cost-effective, viable solutions through high-resolution images and customized monitoring programs for data acquisition. Hence, this study aims to explore the potential of RPAS high-resolution imagery for mapping macroplastic litter on beaches and distinguishing among various types of identified contaminants. The Brindisi shoreline was surveyed using a DJI MAVIC MINI drone equipped with an RGB camera. The mission allowed for the acquisition of 66 pictures, which were subsequently processed using Metashape software along with GCPs gathered using the integrated GNSS in the Garmin Forerunner 245. The resulting highly detailed orthophoto was analyzed using different spectral separability algorithms. This study demonstrated that RPAS image resolution was satisfactory in detecting macroplastic items, although the accuracy of the final classification maps was affected by the separability algorithms used.

This study explores using drones and AI to improve the management of small streams, especially in rural areas facing increased flood risks due to climate change. This study used drones to capture detailed images of a stream and then applied AI to analyze the images, successfully identifying different areas like vegetation, soil, and water with 94% accuracy. This approach offers valuable insights for: Better monitoring and managing streams: Identifying areas prone to flooding and informing restoration efforts. Improved stream health assessment: Understanding the role of vegetation and monitoring changes in the health of the stream. Promoting sustainable rural development: Protecting communities and infrastructure from floods while supporting sustainable food production. While the study shows promise, the researchers acknowledge limitations and suggest collecting more diverse data and developing more advanced AI models to handle complex areas and deep water zones. Overall, this research highlights the potential of AI and drones to revolutionize rural stream management, leading to a more sustainable future for rural communities and ecosystems.

This study focuses on the critical task of estimating wildfire-burned areas, utilizing remote sensing via cubesats to obtain high-resolution imagery for analysis. Using PlanetScope imagery and surface reflectance from visible and NIR bands, along with vegetation indices, the research mapped burned areas in two locations. A U-Net model based on CNN architecture was employed for estimations, with the data divided into training, testing, and validation sets. Model performance was assessed using metrics like IoU and F1-score, demonstrating the usefulness of vegetation indices combined with visible imagery for detecting burned areas, despite challenges from false alarms in snow and rocky terrains. The study also examined the model's transferability across different regions, indicating its adaptability despite variations in accuracy due to local conditions. This underscores the model's potential for broad geographic application in wildfire analysis.

In this study, the land cover of a coffee production farm was classified into five labels. Aerial images were acquired using a drone equipped with visible and multispectral cameras. Image processing resulted in an orthomosaic for each camera, a vegetative index map (NDVI) and digital elevation models. From a statistical analysis and data fusion techniques, suitable thresholds were found to run a thresholding algorithm, resulting in five masks corresponding to each class. This study allows to increase the understanding of remote sensing methodologies and land use classification applied to the geographical peculiarities of the Colombian territory and is a support point to apply agricultural technological innovation models for the country.

Hydroponic systems demand precise nutrient monitoring for optimal plant growth. Current methods are labor-intensive and require sample preparation. We propose an optically modified direct spectrograph-coupled Laser-Induced Breakdown Spectroscopy (LIBS) system for in-situ monitoring of key nutrients in hydroponic lettuce crops. Additionally, we assess the feasibility and accuracy of LIBS in capturing dynamic changes in nutrient concentrations, providing valuable insights into nutrient uptake kinetics and potential imbalances. This approach offers real-time nutrient monitoring, providing insights into uptake kinetics and potential imbalances. This research highlights LIBS as a valuable tool for real-time remote nutrient monitoring in hydroponic systems, improving decision-making and nutrient management efficiency in agriculture.

The tree density in several areas of Romania has been greatly affected in the last couple of decades. This has been mostly attributed to excessive and unregulated tree harvesting caused by illegal tree cutting operations throughout the country. In order to monitor the situation, various methods have been used: ground based, aerial and spatial surveys. Of all of the aforementioned, spatial surveys using satellite-imaging Remote Sensing has proven particularly useful. The aim of this study is to explore the feasibility of utilizing a simple image-based analysis (in this example using MATLAB) in order to monitor and compare tree density in various Romanian forests. Open source materials and images from recent literature that were meant to draw attention to deforestation in Romania are utilized. Such images were processed, and their analysis has resulted in a better visualization of deforestation. Data that can be further processed were obtained. The results show a series of organized tree harvesting operations that continue in different parts of the country. Future work includes the expansion of collected data, materials, and image parameters.

The following study presents results of the application of satellite optical data for eutrophication monitoring in glacial lakes located in the highest mountain in Balkans - Rila mountain. The group of the Seven Rial lakes and its second lake called Ribnoto was used as a test area in the presented research. Several differential water and vegetation indices including Normalized Differential Water Index (NDWI), Normalized Differential Vegetation Index (NDVI) and Tasseled Cap Transformation for segmentation of Sentinel-2 data were used in order to be tested, and the better ones to be implemented as a reliable method for monitoring levels of eutrophication in mountain glacial lakes areas. Results about dynamics of the eutrophicated area in the lakes are derived. Territories of the swamped areas in the lake were monitored and calculated for the selected time frame of the study.

Leaf area of agricultural crops is an important indicator for understanding growth conditions and evaluating photosynthetic efficiency. Conventional leaf area measurement methods are often destructive, involving the cutting of crop leaves and stubble and manual measurement. Therefore, the conventional method not only reduces the yield of the destroyed crop, but also requires time and labor for the measurement process. In this study, as a highly versatile estimation method, we investigated a method to estimate leaf area by creating a 3D model of a plant from 3D point cloud data acquired with an RGBD camera or LiDAR data. Our method consists of leaf area detection, leaf surface modeling from 3D point cloud data, and leaf area or plant volume estimation from leaf surface model. As a result of conducting an experiment to estimate leaf surface area using Intel RealSense L515 and a potted faux green, it was found that remote measurement of leaf surface area was possible with an estimation error of approximately 6.6%.

The study explores the correlation between NDVI, Canopy Cover (CC), and Leaf Area Index (LAI) in banana plantations (Musa acuminata Colla AAA cv. Dwarf Cavendish) in Tenerife(Canary Islands, Spain). NDVI proved effective in determining CC with an R² higher than 0.7. However, due to three-dimensional considerations, the correlation with LAI was weaker (R² about 0.3). It is suggested to explore these relationships to establish an NDVI-based crop coefficient. This would enhance the accuracy of the current irrigation recommendations.

This study focuses on arid regions, where agricultural changes attract desert-dwelling birds like the Arabian Babbler, leading to increased risks such as roadkill and predation. Using Species Distribution Models and remotely sensed satellite data, we analyzed the habitat preferences of 572 babblers over six years in Israel's Arava Valley. Vegetation density and soil moisture were used to distinguish between modified and natural habitats. This classification aids in understanding anthropogenic pressures on survival rates.

This study investigates the migration and stability of colloidal suspensions (black carbon dispersed in water and water plus SDS) through quasi-surfacic resonant analyses. Sensors, composed of organic UV210 micro-resonators onto oxidized silicon substrate, are integrated into a test platform with real-time detection. Experimentations explore the influence of black carbon nano-powder size and concentration on colloidal dispersion stability. Findings highlight the impact of black carbon concentration on its migration and emphasize the anionic surfactant’s effect on increasing stability. These conclusions, corroborated by rheological and zeta potential measurements, provide insights into determining colloidal dispersion stability and migration via optical resonant analysis.

A study was carried out to validate the METRIC (Mapping EvapoTranspiration at high Resolution with Internalized Calibration) model to estimate actual evapotranspiration (ETa) of a drip-irrigated hazelnut orchard located in Maule Region, Chile (Lat. 35º25´ S; Long 71º23´ W; 189 m above mean sea level). To estimate ETa using the METRIC model, 30 satellite images (Landsat 7 ETM+ and 8 OLI) acquired during clear sky days were used for the 2019-2021 and 2020-2021 growing seasons. The performance of METRIC was evaluated using measurements of ETa from an eddy covariance system (EC) at the time of satellite overpass (11:30 h). The statistical analysis indicated that the METRIC model overestimated ETa values by about 10 % with a root mean square error (RMSE), mean absolute error (MAE) and index of agreement (d) of 0.98 mm d-1, 0.90 mm d-1 y 0.70, respectively. Results demonstrated that METRIC model provides an irrigation tool for mapping spatial and temporal variability of water requirements in the hazelnut orchard (1,500 ha).

In the Arkansas River basin, Colorado, USA, water up-flux from saline shallow groundwater tables leaves salts in the vadose zone. These salts accumulate over decades threatening agricultural sustainability. Remote sensing is an economical and practical tool that can be used to monitor soil salinity (EC, electrical conductivity) at large spatial scales. In this study, several EC algorithms were developed utilizing different combinations of explanatory variables, including a salinity index and actual crop evapotranspiration estimates for surface irrigated maize fields located within a (sub-surface) drainage district in the Arkansas River basin. Estimates of crop water use were performed using a surface energy balance approach and Landsat (satellite) multispectral imagery. Results indicate that the developed EC algorithms that included crop water use and a salinity index improved the accuracy of soil EC mapping over models that included a vegetation index alone.

Pseudo-steppe with grasses hosts a variety of plants, vertebrates and invertebrates; mainly found in the Mediterranean southern areas, this habitat type is currently under the protection of the EU Habitat Directive, with several hosting sites being included in the Natura 2000 network. In this preliminary study, different multispectral and SAR data have been integrated and linked to habitat extent and structural parameters of pseudo-steppe with grasses. The study was carried out over different sites of the Apulia region, in southern Italy, where field calibration and validation data were collected in spring and summer 2024. The aim was to investigate the value of different remote sensing data to support habitat conservation in the Natura 2000 network.

The Mojana region in Colombia, which receives the flows of the Cauca, San Jorge and Magdalena rivers, is facing challenges from climate change and rising sea levels. This study examines recent La Niña events that have coincided with flood disasters, in par ticular the 2021 flood, the longest flood in the region's history, which has affected agriculture. Earth observation data is crucial in vulnerable regions such as the Mojana, reveals significant agricultural transformations following the flooding of the Ca uca River. Machine learning helps map affected areas and analyse recovery, which is essential in isolated regions like the Mojana region. ML improves accuracy in land classification, especially with dual polarization SAR data, and is indispensable for clou d removal in regions with low optical data availability.

Floods are one of the costliest disasters in the world affecting human life, agriculture and infrastructure. Accurate, reliable and near-real time flood extent generation is crucial for efficient flood management. Earth observation data be it optical or microwave has always played a very prominent role in this regard. Microwave data due to its cloud penetration and all weather working capabilities has gained immense popularity over optical data. The scope of single image flood mapping approaches is limited in urban environments due to double bounce phenomenon. Urban areas generally have high interferometric coherence, but the coherence drastically decreases when there is change in scenario bought by floods. In this study we have used decrease in coherence to detect flooded urban areas and thresholding of intensity part of Synthetic Aperture Radar (SAR) data to detect open water floods. The algorithm is applied to Sentinel-1 SAR images acquired during the major flood event that hit Larissa (Greece) in September 2023 and the results are validated with a flood reference map derived from optical imagery.

With the increase in climate change and extreme weather events, natural disasters such as floods have become more frequent, necessitating timely flood detection to mitigate losses to human settlements and the environment. While both Synthetic Aperture Radar (SAR) and Electro-Optical (EO) remote sensing data have been employed to identify flooded regions, SAR is preferred due to its ability to penetrate clouds, which often accompany floods. However, flood-inundated areas often present challenges in estimation due to volume and double bounce scattering in SAR. We address this problem by fusing the SAR data with Digital Elevation Models (DEMs) and land cover maps for improved flood prediction. By leveraging these multi-modal inputs, we generate novel features based on proximity to water bodies and vulnerable low-lying floodplains to train a fusion model. Extensive experimentation, utilizing Sentinel-1 SAR data with various multi-modal input combinations, is presented along with comparisons against state-of-the-art methods that demonstrate the generalization ability of the proposed method for flood detection.

The phenomenon of unauthorized withdrawals of water from underground aquifers is unfortunately very widespread in all rural areas, with significant implications for the management of water resources, soil and the environment. Many studies have shown that Earth Observation (EO) data are an effective tool for mapping irrigated areas at different spatial scales (global, regional, local) and for quantifying the volumes of water used for irrigation. The study examines an interesting area of 36 municipalities in Southern Italy. Here, integrated technologies based on the use of ESA Sentinel-2 (S2) data, Machine Learning (ML) classification algorithm, and Geographic Information Systems (GIS) for the monitoring of irrigated areas have been applied. By cross-referencing this information with the archive of data regarding the authorized irrigation wells, the methodology for mapping the extra-consortium irrigated areas, potentially without concessions, was identified. The geospatial outcomes obtained in this study provide very useful elements in the management of water resources for irrigation purposes.

In the current survey the status of some of the most representative wetlands in Bulgaria has been analyzed (Srebarna, Shabla Lake - Ezerets, Pomorie, Mandra-Poda, Kalimok-Brashlen, Kalimok Complex, Pozharevo-Garvan, Durankulak Lake, Burgas Lake, Atanasovsko Lake and Aheloy-Ravda-Nesebar) based on remote sensing for period of more than 10 years to establish the availability or lack of changes mainly in wetlands’ water surface. The quantitative assessment and the degree of wetands changes have been made by calculating remotely-sensed indices NDVI, NDGI, NDWI, MNDWI and others.

Since 2017, cyanobacteria blooms mainly caused by Microcystis aeruginosa frequently occur in the Moselle river, which leads to the formation of algal scum in extreme cases. Yet, the temporal and spatial dynamics of these scums are largely unknown. Therefore, we developed combined models based on Sentinel-2 data with 10 m resolution (R, G, B, NIR) and corresponding Planet SuperDove data with 3 m resolution to differentiate algal scum from open water and riparian vegetation. For model development, cloud-free almost simultaneous data of Planet SuperDove and Sentinel-2 was retrieved in August 2022. Areas visually detected as scum, water or vegetation were digitalized and the spectral information for each class was retrieved. Based on this information, decision-tree based models were developed to differentiate algal scum. Here, we present the composition and accuracy of these models, especially with respect to the spatial and spectral resolution of the input data.

In 2013, in the Apulia region (Southern Italy) it was detected the first outbreak of Xylella fastidiosa (Xf), a dangerous plant pathogenic bacterium infecting olive trees. “Leccino” and “FS-17” are two varieties tolerant to the bacterium. For this reason, it could be particularly interesting to use proximal and remote sensing techniques to detect and classify different olive tree cultivars. This work aims to evaluate the capabilities of vis-NIR spectroscopy/hyperspectral data analysis in the classification of three olive cultivars based on leaf spectral measurements by the means of a portable spectroradiometer (FieldSpec4 Pro). The results obtained not only demonstrate the efficacy of spectral signature in cultivar classification but also provide important information to optimize ground measurement campaigns and to set up the hyperspectral data collection, on wide areas, acquired by sensors installed on UAV, airborne or satellite.

Hyperspectral imaging is beneficial for non-destructive agricultural inspections, and three-dimensional reconstruction modeling is a powerful tool for inspecting the phenotype of plants. This study proposes an approach to combine three-dimensional reconstruction modeling and hyperspectral images into four-dimensional data. This data not only contains the three-dimensional structural information of an interesting object but also includes the spectral information of every point on the surface of this object. Firstly, the hyperspectral and visible images of an interesting object are acquired from hyperspectral and visible cameras. Secondly, high-resolution visible images are used to reconstruct a three-dimensional surface model of an interesting object. Thirdly, matching hyperspectral images with visible images establishes the correspondence between hyperspectral images and the three-dimensional model. Furthermore, the biomarker index can be derived from hyperspectral data. The biomarker index can be transformed into surface textures and combined with the three-dimensional model to form a three-dimensional biomarker model.

The ecosystem functional properties (EFPs) are quantities characterizing key ecosystem processes, useful to monitor changes also in relationship to climate impacts. The Italian PRISMA innovative satellite mission, with 240 tiny bands and high spatial resolution is able to revolutionize the opportunities to monitoring the Earth and its resources. Here EFPs data derived from flux tower data in different EU ecosystems are linked to different vegetation indices from the hyperspectral PRISMA mission. This research investigates the PRISMA capability to model inter- and intra- ecosystem differences in EFPs.

Submerged aquatic vegetation (SAV) plays a critical role in and serves as hotspots for biodiversity in healthy aquatic ecosystems. However, SAV faces pressures such as grazing, hydrodynamic disturbance and increased temperatures, altering canopy structure. Mapping stress in SAV is challenging due to light attenuation in water. Our project aims to create a detailed SAV model using Structure from Motion (SfM) photogrammetry, correcting above-water radiance for water column attenuation to detect stress variations. In the first project phase, we collected multispectral imagery of SAV in a lab. Plants underwent simulated stresses to assess detectability in imagery. A field spectroradiometer was used to validate image values. Colour photos were processed into 3D models using Agisoft Metashape SfM software, helping determine submergence depths for canopy parts. This data was used to correct spectral reflectance and assess spatial/temporal variation. Future work will explore detecting stress signals using WorldView satellite data along a section of river Wye.

This study investigates the use of high-resolution multispectral images from Sentinel-2 and PlanetScope satellites, combined with Near-Infrared Spectroscopy (NIRS), to enhance Precision Agriculture practices in assessing silage maize yield, moisture, and quality (Neutral Detergent Fiber, Acid Detergent Fiber, and starch content) on a 13-ha field in North-East Italy. By analyzing over 10,000 georeferenced points collected from a combined harvester equipped with a yield monitor and NIRS sensor, and applying linear regression on vegetation indices (VIs) like GNDVI and NDVI, the research highlights the effectiveness of PlanetScope GNDVI in early-season yield variability prediction and moisture content at harvest, with significant r values. The study demonstrates the potential of these technologies in optimizing crop moisture management at harvest and precision fertilization, contributing to the advancement of Precision Agriculture.

The study was conducted in two provinces (Suphan Buri and Ang Thong province) located in the major rice cultivated areas in the Chao Phraya River delta, Thailand. Normally, crop yield information is gained nearly harvesting; thus, there may be delay information for agricultural planning. Agricultural agencies attempt to develop programs or applications in crop monitoring and yield forecasting based on advanced technology. The Office of Agricultural and Economics has been seeking Geo-Informatics (GI) technology, especially remote sensing, to develop crop yield model. The project is begin in rice crop in the important rice cultivation area in Thailand. The study area is fully irrigation system. The essential information on rice biophysical is collected in five main growth stages (e.g., seeding, tillering, panicle, flowering, and harvesting) and combined with remotely sensed data. In this study, Sentinel-2 multispectral instrument (MSI) data and Sentinel-1 synthetic aperture radar (SAR) were used together. Then, the linear regression model, both simple and multiple, are developed combining with significant rice biophysical variables. The results showed that SAR data was appropriate.

In this study, a Sentinel-2 time series obtained over mainland Portugal is used to classify yearly maps of irrigated spring-summer crops. Identification of these crops is performed through simple expert-knowledge rules, rather than using supervised classification based on training. Normalized Difference Vegetation Index (NDVI) profile analysis is used to identifying similar phenological responses usually verified for irrigated crops. An historical map of irrigated crops is generated, from 2017 to 2022, based on annual irrigation maps. Validation is performed using 700 points labelled using photointerpretation of orthophoto-maps and monthly satellite images. An overall accuracy of 97.4% is obtained for the 2021 irrigated map, with high users and producers accuracy for irrigated crops, 91.7% and 80.9% respectively.

The use of technological tools for monitoring crops and environmental conditions is essential in scenarios with water availability limitations. INIA-Chile, INTA-Argentina, Agrosavia-Colombia, INIA-Uruguay, Irrigation-Mendoza, and the Institute of Regional Development of UCLM-Spain, with co-financing from FONTAGRO, are developing the initiative "New technologies for increasing efficiency in LAC agriculture by 2030," in which tools for efficient water resource management are being modernized. The technological axis revolves around the use of time series of satellite images (NDVI) at the plot and regional scales (entire basins). Results confirm the operability of the proposed conceptual framework, the effectiveness of technologies for crop monitoring, water consumption, and improvement in water use efficiency.

Greenhouse gas (GHG) emissions and removals from the Land Use, Land Use Change and Forestry (LULUCF) sector are subject to an inventory which must be reported annually under the United Nations Framework Convention on Climate Change. This study investigates the use of Copernicus data for detection of LU/LC changes associated to Land Use, Land Use Change and Forestry sector in the Yougoiztochen (Southeastern) region in Bulgaria. Data from different sources have been processed and interpreted and lands conversion analysis has been completed considering the impact of climate change factors on different land use categories. As a result, we used all the data to map Land cover/Land use changes for a 5-year period. Scrutinizing the climate change impact could contribute further in the GHG emissions inventory process.

Potato (Solanum tuberosum L.) is an important staple food, adapted to a wide range of environments. The use of remoting sensing can help to understand the relationship between canopy development and crop performance, and to improve yield in local areas. One of the preliminary results shows that the green area fell the least at the end of the season, obtained the highest yields, among them a variety with a short development cycle.