The presence of turbidity and suspended solids in freshwater environments is a natural phenomenon. However, due to human activities, the levels of suspended solids have escalated in numerous habitats. The implications of suspended matter vary depending on its composition, particularly with organic matter's ability to deplete dissolved oxygen within the water column. Excessive sedimentation can have adverse effects on primary productivity and aquatic life by diminishing light penetration and restricting habitat availability. The monitoring of water quality holds immense significance, encompassing vital aspects such as human consumption, agriculture, industry, and overall ecosystem health. Remote sensing techniques, with the utilization of the Sentinel-2 mission, have emerged as valuable tools for monitoring and assessing these critical variables. In the scope of this study, specific algorithms were meticulously calibrated and validated to accurately estimate turbidity and suspended solids in the Albufera de Valencia, leveraging the rich dataset provided by Sentinel-2 satellite imagery. The performance of these algorithms exhibited variability, with the R705 model emerging as the most reliable for estimating both turbidity and suspended solids. However, the validity of its turbidity estimates faced scrutiny. Alternative models, including R705 × R705/R490 and R783 × R705/R490, exhibited less satisfactory outcomes for turbidity estimation but showcased enhanced performance in estimating suspended solids. Further research endeavors are warranted to enhance the precision and robustness of these algorithms, while considering the unique characteristics inherent to the study area. Collaboration between disciplines such as limnology, optics and water chemistry is crucial to advance in water quality estimation models in lakes and lagoons such as Albufera. By integrating expertise and approaches from these diverse fields, new knowledge can be gained and the basis for more effective management and conservation strategies can be laid.

eXplainable Artificial Intelligence (XAI) has emerged as an essential requirement when dealing with mission-critical applications, ensuring transparency and interpretability of the employed black box AI models. XAI significance spans various domains, from healthcare to finance, where understanding the decision-making process of deep learning algorithms is essential. Most computer vision AI-based models are often black boxes; hence, providing explainability of image processing deep neural networks is crucial for their wide adoption and deployment in medical image analysis, autonomous driving, and remote sensing applications.

Existing XAI methods aim to provide insights about the methodology used by the black box model in making decisions by highlighting the most relevant regions within the input image that contribute to the model's prediction. Recently, several XAI methods for image classification tasks have been introduced. On the contrary, image segmentation has received comparatively less attention in the context of explainability, even though it is a fundamental task in computer vision applications, especially in remote sensing. Only some research proposes gradient-based XAI algorithms for image segmentation.

This paper adapts the recent gradient-free Sobol XAI method for semantic segmentation. To measure the performance of the Sobol method for segmentation, we propose a quantitative XAI evaluation method based on a learnable noise model. The main objective of this model is to induce noise on the explanation maps, where higher induced noise signifies low accuracy and vice-versa. A thorough benchmark is conducted using high-resolution satellite images focusing on buildings' segmentation tasks. Our results reveal that the proposed noise-based evaluation technique can effectively compare the fidelity of different XAI methods.

Coastal areas play a key role in the adaptation of ocean-climate change due to their land-sea interface. The mapping and monitoring of their use and cover are crucial to understand where are the most exposed and vulnerable zones and how to manage them in a sustainable way. The finest spatial resolution possible is required to empower the diagnosis and prognosis of coastal objects subject to current and future erosion and/or submersion risks. To date, unmanned aerial vehicles (UAVs) consist of the best platforms to bear sensors capable to provide centimeter-scale 2D and 3D coastal information. The active lidar instrument scans coastal landscapes with a rate of hundreds of thousands points per second propagating at the speed of light. UAV-based lidar products enable to reach the best accuracy and precision in xyz data among the airborne/spaceborne tools. However lidar intensity remains poorly harnessed in Earth Observation from satellite to drone.

Along the bay of Mont-Saint-Michel (France), classifications of nine representative coastal habitats (sediment, soil, salt marsh, dry grass, grass, shrub, tree, car, road) at 1 cm spatial resolution were run based on 2300 pixels of calibration and 2300 pixels of validation for every class, using the DJI Zenmuse L1 data, mounted on a DJI M300-RTK quadcopter. The L1 sensor gathers an active lidar Livox Avia, a passive one-inch blue-green-red (BGR) 20 MP camera, and an inertial measurement unit. The 450m-range Avia instrument emits a 905nm laser at 240 kHz while receiving up to 2 returns.

Landscape-scale classification results were satisfactory based on BGR data (Overall Accuracy, OA: 84,57%), and were substantially improved by 4,14% when adding the mean lidar intensity (OA: 88,71%). At the class-level, road, grass and soil showed better producer’s accuracies (12,83%, 11,3% and 8,95%, respectively), while soil, tree, salt marsh and dry grass were better classified (9,48%, 9,28%, 4,56% and 2,35% of user’s accuracies, respectively) when mean lidar intensity was integrated.

Identifying changes in Earth's phenomena is vital for understanding and mitigating the impacts of environmental issues. Monitoring Earth's surface phenomena can be done effectively using satellite images acquired at different times. In addition to spectral features, spatial features play a significant role in detecting precise changes. However, classical change detection (CD) methods rarely consider spatial information and fail to account for scale variations within images. The present introduces a novel deep learning-based CD method that hierarchically extracts spatial-spectral features in various scales to address these issues. The proposed deep neural network generates a binary change map by employing a multi-scale approach that integrates the information of patches of varied sizes at the decision level. We conducted experiments using Landsat-8 images from Sahand City, East Azarbaijan, Iran, because of their remarkable capacity to represent Earth's surface details. Tabriz's population growth has led to rapid development in Sahand city to accommodate citizens. Studying these changes can offer valuable insights into urban planning. The performance of the proposed deep model is evaluated in comparison to two classical methods, including the Change Vector Analysis (CVA) method and a random forest (RF) algorithm. Based on the change detection results, the proposed deep learning network demonstrates a significant improvement in the kappa coefficient (K.C.) compared to the RF and CVA methods, with an increase of approximately 11.86% and 29.36%, respectively. Furthermore, in terms of overall accuracy (O.A.), the proposed network outperforms both the RF and CVA methods by approximately 17.08% and 29.16%, respectively. The proposed multi-scale deep network performs better in detecting changes across all metrics. As a result, CVA fails to identify changes with sufficient accuracy.

Buildings are objects of great importance that need to be observed continuously. Satellite and aerial images provide valuable resources nowadays for building footprint extraction. Since these images cover large areas, manually detecting buildings will be a time-consuming task. Recent studies have proven the capability of deep learning algorithms in building footprint extraction automatically. But these algorithms need vast amounts of data for training and they may not perform well under the low-data conditions. Digital surface models provide height information which helps discriminate buildings from their surrounding objects. However, they may suffer from noises, especially on the edges of buildings which may result in low boundary resolution. In this research, we aim to address this problem by using edge bands detected by a deep learning model alongside the digital surface models to improve the building footprint extraction when training data is low. Since satellite images have complex backgrounds, using conventional edge detection methods like canny or Sobel filter will produce a lot of noisy edges which can deteriorate the model performance. For this purpose, first, we train a U-Net model for building edge detection with the WHU dataset and fine-tune the model with our target training dataset which contains a low quantity of satellite images. Then, the building edges of the target test images are predicted using this fine-tuned U-Net and concatenated with our RGB-DSM test images to form 5-band RGB-DSM-Edge images. Finally, we train a U-Net with 5-band training images of our target dataset which contains precise building edges in their fifth band. Then we use this model for building footprint extraction from 5-band test images which contain building edges in their fifth band that are predicted by a deep learning model in the first stage. We compared the results of our proposed method with 4-band RGB-DSM and 3-band RGB images. Our method obtained 82.88% in IoU and 90.45% in f1-score metrics which indicates that by using edge bands alongside the digital surface models, the performance of the model improved 2.57% and 1.59% in IoU and F1-score metrics, respectively. Also, the predictions made by 5-band images have sharper building boundaries than RGB-DSM images.

Volcanic eruptions are natural disasters that can have severe impacts on human life both locally and globally. Remote sensing using satellites enables the monitoring and tracking of volcanic clouds, even in locations with difficult or no access, and helps determine eruption source parameters (e.g., erupted volume, plume height, and mass eruption rate), which are essential for describing eruption dynamics and evaluating the associated natural risks.

A systematic literature review was conducted to understand how satellite remote sensing can be used to monitor and detect ash and SO₂. This review ranges from optical (multispectral, hyperspectral and LiDAR) to radar and thermal sensors. This review aims to identify the accuracy, advantages, and limitations of different sensors and algorithms. The Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) was used to conduct the review, and the search parameters included keywords related to the review topic and articles published in peer-reviewed journals between January 1^st, 2010, to September 30^th, 2022.

From this search, 53 papers were chosen based on the use of satellites to detect and monitor volcanic clouds and estimate eruptive source parameters. This review revealed that SEVIRI, AHI, AVHRR, and MODIS are commonly used sensors because of their availability and near real-time capabilities. MISR and CALIOP are also frequently used because of their high spatial and vertical resolutions. The traditional brightness temperature difference (BTD) with radiative transfer models (RTM) is the most used approach to detect and retrieve volcanic clouds parameters despite its limitations. Hyperspectral sensors, such as IASI and TROPOMI, are commonly utilized for SO₂ detection and estimation. Limitations related to scattering, which occur during cloudy conditions, make accurate measurements difficult or impossible under extreme cloudy conditions. When used with the COBRA algorithm, the TROPOMI sensor shows improved resolution and accuracy, reducing noise and scattering.

Although individual approaches and their integration have contributed to the study of volcanic ash and SO₂ emissions, there are limitations both to remote sensing sensors and algorithms used. To overcome the limitations of current sensors and retrieval methods, future research should prioritize the integration of multi-sensor data, address limitations in spatial and temporal resolutions, and utilize computational methods, such as statistical methods, neural networks, and deep learning algorithms, to eliminate the need for fixed threshold constraints present in traditional methods.

Soil moisture (SM) is a crucial hydrological variable that connects land surface and atmospheric processes. Accurate soil moisture monitoring is necessary for understanding energy, water cycles, and ecological system processes. Satellite-based microwave remote sensing is an effective method for obtaining information on land surface hydrology around the globe. Moreover, satellite soil moisture products such as SMAP have been reported with temperature dependence error which is caused by the so-called "temperature effect (TE)," as same as in-situ soil water content (SWC) data, which was measured using the dielectrically measured method. This study aims to remove TE in AMSR2 Level 2 soil moisture products.
In this study, a data-driven method to remove the temperature effect in in-situ SWC data was improved to allow the direct use of satellite products of land surface temperature or brightness temperature measurements. The Mongolia site was selected, and the SWC and soil temperature data at 3 cm depth from 5th September 2016 to 31st October 2019 were used as the reference for developing and evaluating the new TE removal algorithm. To fit the large satellite footprint, areal mean SWC and soil temperature were prepared using the Thiessen polygon method.
According to the results of the newly developed removal algorithm of the temperature effect error, the difference between average ascending and descending crossing time data was eliminated by directly using brightness temperature. The correlation coefficient (R) was used to evaluate the agreement between ascending and descending after using the newly developed TE removal method. According to the results of R, the corrected AMSR2 products exhibited the best performance over the original AMSR2 products in the Mongolia network region, with a relatively higher R-value. This finding reveals that the correction improved the correlation between in-situ and satellite soil water content values at each of the chosen sites and shows that the corrected AMSR2 SWC values better agree with the corrected in-situ measurements.

Soil moisture is one of the popular variables for various processes and models in hydrological research and the Earth sciences due to its high interaction among the land surface and atmosphere, land surface, and underground. Many satellite missions, such as SMAP, SMOS, or AMSR-E/2, have been launched to observe precise soil moisture in extensive or global spatial coverage, which is the limitation of in situ ones. In recent years, the satellite soil moisture products such as SMAP have also been reported to comprise errors caused by the so-called "temperature effects" found in the in situ soil moisture observed by dielectric sensors several decades ago. Though the temperature effect removal methods proposed for in situ data can apply to satellite soil moisture, there are limitations, such as global application. In this work, we also confirmed the existence of these errors in AMSR2 soil moisture products. We developed a new algorithm to remove them using satellite data at ascending and descending times. Three-centimeter depth soil water content (SWC) and soil temperature from 5^th September 2016 to 31^st October 2019 at the Mongolia site were used as the reference data. The correlation coefficient values between the corrected AMSR2 and corrected in situ SWC are between 0.1595 and 0.3542, while the values of the original ones range from 0.1846 to 0.3153. As well as, the values of MD, MAE, RMSE, and ubRMSE became lower in corrected data when compared to the original data. These indicate that the corrected AMSR2 products are better than the original AMSR2 products in the Mongolia network region. These results reveal that the correction method can successfully remove the temperature effects from AMSR2 soil moisture products.

We present new tools designed for a wider application of the spheroidal model of real particles in light scattering and polarised radiative transfer simulations. We briefly describe the main theoretical novations in our solution to the light scattering problems for single homogeneous spheroids that was obtained by using the field expansions in an optimised non-orthogonal spheroidal basis. Our exact solution for spheroids with non-confocal layer boundaries is a theoretical break through. We note our efforts to transform the naturally arising spheroidal T-matrix into the standard spherical one widely used in applications to ensembles of spheroids as well as our modification of the
package CosTuuM used to prepare the optical data on spheroid ensembles with given distributions for polarised radiative transfer simulations. We discuss the applicability range of our tools as well as various ways used by us to reduce the computational time. As the tools can treat scatterers of as large diffraction parameter xa = 2πa/λ as above 300, where a and λ are the major semiaxis and the wavelength, respectively, as an illustration we compare our results with those derived by the Ray Tracing method.

Precision agriculture has seen significant advancements with the integration of remote sensing technologies. However, challenges such as real-time data availability, standardization, and computing limitations in rural settings persist. This study aimed to develop a standardized method for generating spatial variability maps for crop management in vineyards using UAV imagery. Using IDW (Inverse Distance Weight),a geostatistical interpolation method, nadir images with geotagged locations were processed to extract spectral information and EXIF metadata. The results demonstrated that interpolation methods are effective compared to traditional photogrammetry-based methods, with the approach being more than 90% faster, highlighting their potential in real-time applications. Notably, IDW's correlation with Sentinel 2 imagery reached values as high as r = 0.8, comparable to orthomosaics. This method offers a faster, less resource-intensive alternative to existing techniques for crop mapping, addressing the current challenges in precision agriculture. Its practical implications suggest that farmers and agricultural professionals can achieve accurate spatial variability assessments without the need for high-end equipment or extensive computing power, making it a cost-effective and efficient solution for modern agriculture.