Sentinel-2 Images Research Articles

Identification of Pine Wilt Disease-Infested Stands Based on Single- and Multi-Temporal Medium-Resolution Satellite Data

Pine wilt disease (PWD) is known for its high lethality and rapid transmission, earning it the name “cancer of the pine tree”. The prompt removal of infested pine trees is an effective measure for preventing and controlling pine wilt disease. Accurate and efficient monitoring technologies are crucial for the scientific prevention and control of this plant disease. Currently, numerous remote sensing monitoring studies have been conducted on pine wilt disease. However, there is limited research on the temporal identification of PWD-infested forest stands over large areas. To build classification models, this study utilized three machine learning algorithms: artificial neural network (ANN), random forest (RF), and support vector machine (SVM). We aimed to investigate the effectiveness of single-temporal and multi-temporal Landsat and Sentinel-2 satellite images PWD-infested forest stands detection. The results indicated that, at a spatial resolution of 30 m, Landsat-9 and Sentinel-2 remote sensing images effectively identified PWD-infested forest stands, with classification accuracies of 77.87% and 78.91%, respectively. Higher spatial resolutions in Sentinel-2 remote sensing images were associated with improved identification capabilities. Furthermore, multi-temporal Landsat satellite data (with a classification accuracy of 85.95%) significantly enhanced the performance of the monitoring model compared to single-temporal Landsat satellite data (with a classification accuracy of 77.87%). The RGI difference was found to be the optimal vegetation index. In conclusion, by combining multi-temporal and single-time-phase Landsat remote sensing data, a monitoring model for PWD-infested forest stands was constructed. It achieved a classification accuracy of 88.26%. In this study, a higher accuracy in identifying pine wilt disease and a lower economic cost were achieved by Landsat and Sentinel images, offering valuable insights for the management of pine wilt disease.

Vegetation Classification and Evaluation of Yancheng Coastal Wetlands Based on Random Forest Algorithm from Sentinel-2 Images

The identification of wetland vegetation is essential for environmental protection and management as well as for monitoring wetlands’ health and assessing ecosystem services. However, some limitations on vegetation classification may be related to remote sensing technology, confusion between plant species, and challenges related to inadequate data accuracy. In this paper, vegetation classification in the Yancheng Coastal Wetlands is studied and evaluated from Sentinel-2 images based on a random forest algorithm. Based on consistent time series from remote sensing observations, the characteristic patterns of the Yancheng Coastal Wetlands were better captured. Firstly, the spectral features, vegetation indices, and phenological characteristics were extracted from remote sensing images, and classification products were obtained by constructing a dense time series using a dataset based on Sentinel-2 images in Google Earth Engine (GEE). Then, remote sensing classification products based on the random forest machine learning algorithm were obtained, with an overall accuracy of 95.64% and kappa coefficient of 0.94. Four indicators (POP, SOS, NDVIre, and B12) were the main contributors to the importance of the weight analysis for all features. Comparative experiments were conducted with different classification features. The results show that the method proposed in this paper has better classification.

Joint multitemporal SAR and optical mapping of urban changes

AbstractThis paper presents a methodology designed to leverage multitemporal sequences of synthetic aperture radar (SAR) and multispectral data and automatically extract urban changes. The approach compares results using different radar and optical sensors, describing the advantages and drawbacks of using SAR data from the COnstellation of small Satellites for the Mediterranean basin Observation (COSMO)/SkyMed, SAtélite Argentino de Observación COn Microondas (SAOCOM), and Sentinel-1 constellations, as well as nighttime light data or Sentinel-2 images. Multiple indexes obtained from multispectral data are compared, too, and results obtained by an unsupervised clustering procedure are analyzed. The results show that using different datasets it is possible to obtain consistent results about different types of changes in urban areas (e.g., demolition, development, and densification) with different levels of spatial details.

Image Quality Assessment of Multi-Satellite Pan-Sharpening Approach: A Case Study using Sentinel-2 Synthetic Panchromatic Image and Landsat-8

INTRODUCTION: The satellite's physical and technical capabilities limit high spectral and spatial resolution image acquisition. In Remote Sensing (RS), when high spatial and spectral resolution data is essential for specific Geographic Information System (GIS) applications, Pan Sharpening (PanS) becomes imperative in obtaining such data. 
 OBJECTIVES: Study aims to enhance the spatial resolution of the multispectral Landsat-8 (L8) images using a synthetic panchromatic band generated by averaging four fine-resolution bands in the Sentinel-2 (S2) images. 
 METHODS: Evaluation of the proposed multi-satellite PanS approach, three different PanS techniques, Smoothed Filter Intensity Modulation (SFIM), Gram-Schmidt (GS), and High Pass Filter Additive (HPFA) are used for two different study areas. The techniques' effectiveness was evaluated using well-known Image Quality Assessment Metrics (IQAM) such as Root Mean Square Error (RMSE), Correlation Coefficient (CC), Erreur Relative Globale Adimensionnelle de Synthèse (ERGAS), and Relative Average Spectral Error (RASE). This study leveraged the GEE platform for datasets and implementation.
 RESULTS: The promising values were provided by the GS technique, followed by the SFIM technique, whereas the HPFA technique produced the lowest quantitative result. 
 CONCLUSION: In this study, the spectral bands of the MS image’s performance show apparent variation with respect to that of the different PanS techniques used.

Optimal Configuration of Omega-Kappa FF-SAR Processing for Specular and Non-Specular Targets in Altimetric Data: The Sentinel-6 Michael Freilich Study Case

In this study, the full-focusing (FF) algorithm is reviewed with the objective of optimizing it for processing data from different types of surfaces probed in altimetry. In particular, this work aims to provide a set of optimal FF processing parameters for the Sentinel-6 Michael Freilich (S6-MF) mission. The S6-MF satellite carries an advanced radar altimeter offering a wide range of potential FF-based applications which are just beginning to be explored and require prior optimization of this processing. In S6-MF, the Synthetic Aperture Radar (SAR) altimeter acquisitions are known to be aliased in the along-track direction. Depending on the target, aliasing can be tolerated or may be a severe impairment to provide the level of performance expected from FF processing. Another key aspect to consider in this optimization study is the unprecedented resolution of the FF processing, which results in a higher posting rate than the standard SAR processing. This work investigates the relationship between posting rate and noise levels and provides recommendations for optimal algorithm configurations in various scenarios, including transponder, open ocean, and specular targets like sea-ice and inland water scenes. The Omega–Kappa (WK) algorithm, which has demonstrated superior CPU efficiency compared to the back-projection (BP) algorithm, is considered for this study. But, unlike BP, it operates in the Doppler frequency domain, necessitating further precise spectral and time domain settings. Based on the results of this work, real case studies using S6-MF acquisitions are presented. We first compare S6-MF FF radargrams with Sentinel-1 (S1) images to showcase the potential of optimally configured FF processing. For highly specular surfaces such as sea-ice, distinct techniques are employed for lead signature identification. S1 relies on image-based lineic reconstruction, while S6-MF utilizes phase coherency of focalized pulses for lead detection. The study also delves into two-dimensional wave spectra derived from the amplitude modulation of image/radargrams, with a focus on a coastal example. This case is especially intriguing, as it vividly illustrates different sea states characterized by varying spectral peak positions over time.

Analyzing Flood Damage and Mapping Flood Hazard Zones Using AHP Model: A Case Study of Pol-e-Alam, Logar Province, Afghanistan

This research aims to evaluate the impact of the most recent floods that occurred on August 20, 2022, in Logar province in southern Afghanistan. For this purpose, changes in land use and land cover (LULC) of the study area were created from the Sentinel-2 image with a spatial resolution of 10 meters. To achieve this, the study utilized Sentinel-2 images to analyze LULC changes before and after the flood event and employed a support vector machine for supervised classification. The study also applied the analytical hierarchy process (AHP) to evaluate the future risks of flooding in the study area, focusing on factors related to hydrological phenomena. Overall, the study demonstrates the effectiveness of geospatial technologies and remote sensing in assessing the impacts of floods and creating flood risk maps. This can significantly reduce the consequences of flooding and inform decision-making for disaster management and mitigation.

Feature Optimization-Based Machine Learning Approach for Czech Land Cover Classification Using Sentinel-2 Images

The Czech Republic is one of the countries along the Belt and Road Initiative, and classifying land cover in the Czech Republic helps to understand the distribution of its forest resources, laying the foundation for forestry cooperation between China and the Czech Republic. This study aims to develop a practical approach for land cover classification in the Czech Republic, with the goal of efficiently acquiring spatial distribution information regarding its forest resources. This approach is based on multi-level feature extraction and selection, integrated with advanced machine learning or deep learning models. To accomplish this goal, the study concentrated on two typical experimental regions in the Czech Republic and conducted a series of classification experiments, using Sentinel-2 and DEM data in 2018 as the main data sources. Initially, this study extracted various features, including spectral, vegetation, and terrain features, from the study area, then assessed and selected key features based on their importance. Additionally, this study also explored multi-level spatial contextual features to improve classification performance. The extracted features include texture and morphological features, as well as deep semantic information learned by utilizing a deep learning model, 3D CNN. Finally, an AdaBoost ensemble learning model with the random forest as the base classifier is designed to produce land cover classification maps, thus obtaining the spatial distribution of forest resources. The experimental results demonstrate that feature optimization significantly enhances the extraction of high-quality features of surface objects, thereby improving classification performance. Specifically, morphological and texture features can effectively enhance the discriminability between different features of surface objects, thereby improving classification accuracy. Utilizing deep learning networks enables more efficient extraction of deep feature information, further enhancing classification accuracy. Moreover, employing an ensemble learning model effectively boosts the accuracy of the original classification results from different individual classifiers. Ultimately, the classification accuracy of the two experimental areas reaches 92.84% and 93.83%, respectively. The user accuracies for forests are 92.24% and 93.14%, while the producer accuracies are 97.71% and 97.02%. This study applies the proposed approach for nationwide classification in the Czech Republic, resulting in an overall classification accuracy of 90.98%, with forest user accuracy at 91.97% and producer accuracy at 96.2%. The results in this study demonstrate the feasibility of combining feature optimization with the 3D Convolutional Neural Network (3DCNN) model for land cover classification. This study can serve as a reference for research methods in deep learning for land cover classification, utilizing optimized features.

LinkNet-Spectral-Spatial-Temporal Transformer Based on Few-Shot Learning for Mangrove Loss Detection with Small Dataset

Mangroves grow in intertidal zones in tropical and subtropical regions, offering numerous advantages to humans and ecosystems. Mangrove monitoring is one of the important tasks to understand the current status of mangrove forests regarding their loss issues, including deforestation and degradation. Currently, satellite imagery is widely employed to monitor mangrove ecosystems. Sentinel-2 is an optical satellite imagery whose data are available for free, and which provides satellite imagery at a 5-day temporal resolution. Analyzing satellite images before and after loss can enhance our ability to detect mangrove loss. This paper introduces a LSST-Former model that considers the situation before and after mangrove loss to categorize non-mangrove areas, intact mangroves, and mangrove loss categories using Sentinel-2 images for a limited number of labels. The LSST-Former model was developed by integrating a fully convolutional network (FCN) and a transformer base with few-shot learning algorithms to extract information from spectral-spatial-temporal Sentinel-2 images. The attention mechanism in the transformer algorithm may effectively mitigate the issue of limited labeled samples and enhance the accuracy of learning correlations between samples, resulting in more successful classification. The experimental findings demonstrate that the LSST-Former model achieves an overall accuracy of 99.59% and an Intersection-over-Union (IoU) score of 98.84% for detecting mangrove loss, and the validation of universal applicability achieves an overall accuracy of more than 92% and a kappa accuracy of more than 89%. LSST-Former demonstrates superior performance compared to state-of-the-art deep-learning models such as random forest, Support Vector Machine, U-Net, LinkNet, Vision Transformer, SpectralFormer, MDPrePost-Net, and SST-Former, as evidenced by the experimental results and accuracy metrics.

Using satellite imagery to estimate CO2 partial pressure and exchange with the atmosphere in the Songhua River

Rivers play a vital role in the global carbon cycle, even though they account for only 0.58 % of the Earth's non-glaciated land surface area. Owing to the significant spatial and temporal variability of the partial pressure of CO2 (pCO2) in river water, there is a large degree of uncertainty in the estimation of CO2 flux from the river water-atmosphere interface. This study assembled more than 200 in-situ measured pCO2 in the mainstream water of Songhua River and matched with Sentinel-2 overpasses within ± 7 days from the GEE platform. Multiple linear regression (MLR) and two machine learning models (Random Forest and XGBoost) algorithms were used to estimate pCO2 in Songhua River. MLR, Random Forest, and XGBoost all demonstrated good performance and have the potential to map pCO2 for different years using high-quality Sentinel-2 images. In all the tested modeling approaches, the XGBoost model exhibited stable performance in the validation set (R2 > 0.79, RMSE < 228 µatm, MAPE < 42.93 %, SMAPE < 30.64 %). The distribution of mapped pCO2 by XGBoost model significantly varied with seasonal change, and the pCO2 in summer (445 ± 125 µatm) was higher than in autumn (373 ± 58 µatm) and spring (351 ± 57 µatm). CO2 emission flux from the mainstream water of Songhua River was calculated based on the inverse pCO2 value with a net annual uptake of 15.6 Gg C during the ice-free period. The findings suggest that the Songhua River functions as a carbon sink during both spring and autumn but as a carbon source during the summer. In addition, the potential control of annual CO2 flux spatial variability was attributed to the characteristics of the watershed landscapes. The study is critical to clarify the role of river systems in the global carbon cycle.

Lead fractions from SAR-derived sea ice divergence during MOSAiC

Abstract. Leads and fractures in sea ice play a crucial role in the heat and gas exchange between the ocean and atmosphere, impacting atmospheric, ecological, and oceanic processes. We estimated lead fractions from high-resolution divergence obtained from satellite synthetic aperture radar (SAR) data and evaluated them against existing lead products. We derived two new lead fraction products from divergence with a spatial resolution of 700 m calculated from daily Sentinel-1 images. For the first lead product, we advected and accumulated the lead fractions of individual time instances. With those accumulated divergence-derived lead fractions, we comprehensively described the presence of up to 10 d old leads and analyzed their deformation history. For the second lead product, we used only divergence pixels that were identified as part of linear kinematic features (LKFs). Both new lead products accurately captured the formation of new leads with widths of up to a few hundred meters. We presented a Lagrangian time series of the divergence-based lead fractions along the drift of the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition in the central Arctic Ocean during winter 2019–2020. Lead activity was high in fall and spring, consistent with wind forcing and ice pack consolidation. At larger scales of 50–150 km around the MOSAiC expedition, lead activity on all scales was similar, but differences emerged at smaller scales (10 km). We compared our lead products with six others from satellite and airborne sources, including classified SAR, thermal infrared, microwave radiometer, and altimeter data. We found that the mean lead fractions varied by 1 order of magnitude across different lead products due to different physical lead and sea ice properties observed by the sensors and methodological factors such as spatial resolution. Thus, the choice of lead product should align with the specific application.

Deep learning models integrating multi-sensor and -temporal remote sensing to monitor landslide traces in Vietnam

Landslides pose significant threats to lives and public infrastructure in mountainous regions. Real-time landslide monitoring presents challenges for scientists, often involving substantial costs and risks due to challenging terrain and instability. Recent technological advancements offer the potential to identify landslide-prone areas and provide timely warnings to local populations when adverse weather conditions arise. This study aims to achieve three key objectives: (1) propose indicators for detecting landslides in both field and remote sensing images; (2) develop deep learning (DL) models capable of automatically identifying landslides from fusion data of Sentinel-1 (SAR) and Sentinel-2 (optical) images; and (3) employ DL-trained models to detect this natural hazard in specific regions of Vietnam. Twenty DL models were trained, utilizing three U-shaped architectures, which include U-Net and U-Net3+, combined with different data-fusion choices. The training data consisted of multi-temporal Sentinel images and increased the accuracy of DL models using Adam optimizer to 99% in landslide detection with low loss function values. Using two bands of the Sentinel-1 could not define the characteristics of landslide traces. However, the integration between Sentinel-2 data and these bands makes the landslide detection process more effective. Therefore, the authors proposed a consolidated strategy based on three models: (1) UNet using four S2-bands, (2) UNet3+ using four S2-bands, (3) UNet using four S2-bands and VV S1-band, and (4) UNet using four S2-bands and VH S1-band for fully detect landslides. This integrated strategy uses the capabilities of each model and overcomes model result constraints to better describe landslide traces in varied geographical locations.

Application of potential machine learning models in landslide susceptibility assessment: A case study of Van Yen district, Yen Bai province, Vietnam

Landslides are natural hazards that cause significant damage to both property and human lives. This study employs potential machine learning models such as Random Forest (RF), Gradient Boosting (GB), and Support Vector Machine (SVM) to assess landslide susceptibility in Van Yen District, Yen Bai Province, Vietnam, that experiences a higher frequency of landslides compared to other localities in the region. The study incorporates thirteen input variables, including elevation, slope angle, aspect, plan curvature, profile curvature, Topographic Wetness Index (TWI), distance to faults, lithology, distance to roads, distance to rivers, land cover, rainfall, and Normalized Difference Vegetation Index (NDVI). To construct the models, landslide statistics reports were utilized, consisting of 302 landslide points collected through field surveys and 52 landslide points determined using Radar Sentinel-1 images. The Google Earth Engine cloud computing platform is utilized for constructing the landslide susceptibility models. The outcome of the research is a landslide susceptibility map with five levels: very low, low, moderate, high, and very high. The Area Under the Curve (AUC) is used as a metric to evaluate the performance of all three models. The findings indicate that, besides similarities observed in landslide susceptibility maps for previously occurred landslides, the Random Forest model demonstrates a favorable performance compared to the other models, with an AUC of 0.883.

A Deep Learning Approach for Ship Detection Using Satellite Imagery

INTRODUCTION: This paper addresses ship detection in satellite imagery through a deep learning approach, vital for maritime applications. Traditional methods face challenges with large datasets, motivating the adoption of deep learning techniques.&#x0D; OBJECTIVES: The primary objective is to present an algorithmic methodology for U-Net model training, focusing on achieving accuracy, efficiency, and robust ship detection. Overcoming manual limitations and enhancing real-time monitoring capabilities are key objectives.&#x0D; METHOD: The methodology involves dataset collection from Copernicus Open Hub, employing run-length encoding for efficient preprocessing, and utilizing a U-Net model trained on Sentinel-2 images. Data manipulation includes run-length encoding, masking, and balanced dataset preprocessing.&#x0D; RESULT: Results demonstrate the proposed deep learning model's effectiveness in handling diverse datasets, ensuring accuracy through U-Net architecture, and addressing imbalances. The algorithmic process showcases proficiency in ship detection.&#x0D; CONCLUSION: In conclusion, this paper contributes a comprehensive methodology for ship detection, significantly advancing accuracy, efficiency, and robustness in maritime applications. The U-Net-based model successfully automates ship detection, promising real-time monitoring enhancements and improved maritime security.

New approach for satellite DEM accuracy enhancement by combing machine learning, fuzzy majority voting, and weighted interpolation techniques

The digital elevation model (DEM) is crucial in many global and regional scientific studies in civilian and military applications. The aim of this research is to develop and test a new DEM approach for correcting the various errors in the Shuttle Radar Topography Mission (SRTM) digital elevation model. Firstly, the DEMs with the feature attributes from Sentinel-2 multispectral imagery are generated. Secondly, SRTM DEM with one band and attributes of a sentinel-2 image with eight bands are used as input data in supervised max-like hood, an artificial neural network (ANN), and support vector machine (SVM) classification models. Thirdly, ANN, supervised max-like hood, and SVM classification models, which have various properties, are fused by fuzzy majority voting (probability fusion). Finally, the fused probability is assigned for each pixel of the image, which has 12 fixed ground control points (GCPs), which is considered new input data for the inverse probability weighted interpolation (IPWI) approach to create the corrected SRTM elevations. The results were contrasted with a reference DEM (RD) created by image matching with Worldview-1 stereo satellite images, which had a 1-m vertical accuracy. The results of this study demonstrated that the RMSE of the original SRTM DEM was 5.95. On the other hand, the RMSE of the estimated elevations by the IPWI approach has been improved to 1.98 compared with that of the MLR method (3.01). The study shows a series of significant improvements in the SRTM when assessed with the reference DEM, with an RMSE reduction of (66.72%) when compared to the widely utilized multiple linear regression (MLR) method. It can be concluded that the elevation error of the original SRTM DEM is clearly reduced by the suggested approach.

Radiometric landscape: a new conceptual framework and operational approach for landscape characterisation and mapping

ABSTRACT Landscape mapping has the potential to address some of the most pressing research issues of our time, including climate change, sustainable development, and human well-being. In this paper, we propose an original method that lays the foundations for landscape mapping and overcomes some of the major limitations of existing biophysical methods. Based on the assumption that the primary components of the landscape can be extracted directly from the radiometric information of satellite image time series, this paper presents a new approach to landscape characterization and mapping based solely on remote sensing data. The approach relies on a conceptual model, which links the description, characteristics, structure and functions of the landscape to a set of Remote Sensing-based Essential Landscape Variables (RS-ELVs). The RS-ELVs are then processed according to geographic object-based image analysis (GEOBIA) approach to produce a radiometric landscape map. The model and the remote sensing data processing chain are tested on a case study in central Madagascar (about 13 000 km2) composed of contrasting landscapes resulting from different climatic conditions and agricultural practices. The RS-ELVs are extracted from MODIS image time series for the temporal and spectral variables, and from MODIS and Sentinel-2 images for the texture variables. The parameterization of the segmentation and clustering algorithms is determined by statistical optimization. The final result is a radiometric landscape map in six classes. The landscape classes are then characterized using an independent set of remote sensing variables, a global land cover map and ground observations. The approach successfully identifies and delineates the gradient and major landscape types of the complex region of central Madagascar, confirming our initial hypothesis. The production of such radiometric landscape maps opens the way for integrated territorial development, including the planning and protection of the living environment and human well-being, and the implementation of sectoral policies.

A temporary soil dump settlement and landslide risk analysis using the improved small baseline subset-InSAR and continuous medium model

Abandoned soil disposal in China has become increasingly challenging, which may pose significant safety hazards such as soil settlement and landslides. Interferometry Synthetic Aperture Radar (InSAR) is effective in surface deformation monitoring and Small Baseline Subset-InSAR (SBAS-InSAR) can obtain sufficient coherent point density from the bare soil surface. This study investigated Chishi soil dump in the Shenzhen-Shanwei Special Cooperation Zone (SSSCZ), and a total of 91 Sentinel-1 images from 2019 to 2022 was processed. Accordingly, an improved SBAS-InSAR method was utilized to analyze the soil dump’s time-series deformation, and multi-source remote sensing data were used for auxiliary interpretation. Our results indicate that precipitation, high temperature, and construction vibrations may cause instability of the soil dump. Therefore, a depth-integrated continuous medium model was introduced to analyze the potential landslide risk of the soil dump, which demonstrate that a landslide will likely occur, harm personnel, and damage the buildings surrounding the Chishi soil dump when it was saturated (λ ≥ 0.50). This research can provide a vital case reference for the analysis, interpretation, disaster prevention, and control evaluation of similar soil dumps.

Satellite-Derived Indicators of Drought Severity and Water Storage in Estuarine Reservoirs: A Case Study of Qingcaosha Reservoir, China

Estuarine reservoirs are critical for freshwater supply and security, especially for regions facing water scarcity challenges due to climate change and population growth. Conventional methods for assessing drought severity or monitoring reservoir water level and storage are often limited by data availability, accessibility and quality. We present an approach for monitoring estuarine reservoir water levels, storage and extreme drought via satellite remote sensing and waterline detection. Based on the CoastSat algorithm, Landsat-8 and Sentinel-2 images from 2013 to 2022 were adopted to extract the waterline of Qingcaosha Reservoir, the largest estuarine reservoir in the world and a key source of freshwater for Shanghai, China. This study confirmed the accuracy of the satellite-extracted results through two main methods: (1) calculating the angle of the central shoal slope in the reservoir using the extracted waterline data and measured water levels and (2) inverting the time series of water levels for comparison with measured data. The correlation coefficient of the estimated water level reached ~0.86, and the root mean square error (RMSE) of the estimated shoal slope was ~0.2°, indicating that the approach had high accuracy and reliability. We analyzed the temporal and spatial patterns of waterline changes and identified two dates (21 February 2014 and 15 October 2022) when the reservoir reached the lowest water levels, coinciding with periods of severe saltwater intrusions in the estuary. The extreme drought occurrences in the Qingcaosha Reservoir were firstly documented through the utilization of remote sensing data. The results also indicate a strong resilience of the Qingcaosha Reservoir and demonstrate that the feasibility and utility of using satellite remote sensing and waterline detection for estuarine reservoir storage can provide timely and accurate information for water resource assessment, management and planning.

Machine learning models for gully erosion susceptibility assessment in the Tensift catchment, Haouz Plain, Morocco for sustainable development

Gully erosion is a widespread environmental danger, threatening global socio-economic stability and sustainable development. This study comprehensively applied seven machine learning (ML) models including SVM, KNN, RF, XGBoost, ANN, DT, and LR, and evaluated gully erosion susceptibility in the Tensift catchment and predict it within the Haouz plain, Morocco. To ensure the reliability of the findings, the study employed a robust combination of gully erosion inventory, sentinel images, and Digital Surface Model. Eighteen predictors, encompassing topographical, geomorphological, environmental, and hydrological factors, were selected after multicollinearity analyses. The gully erosion susceptibility of the study revealed that approximately 28.18% of the Tensift catchment is at a very high risk of erosion. Furthermore, 15.13% and 31.28% of the catchment are categorized as low and very low respectively. These findings extend to the Haouz plain, where 7.84% of the surface area are very highly risking erosion, while 18.25% and 55.18% are characterized as low and very low risk areas. To gauge the performance of the ML models, an array of metrics including specificity, precision, sensitivity, and accuracy were employed. The study highlights XGBoost and KNN as the most promising models, achieving AUC ROC values of 0.96 and 0.93 in the test phase. The remaining models namely RF (AUC ROC = 0.89), LR (AUC ROC = 0.80), SVM (AUC ROC = 0.81), DT (AUC ROC = 0.86), and ANN (AUC ROC = 0.78), also displayed commendable performance. The novelty of this research is its innovative approach to combat gully erosion through cutting edge ML models, offering practical solutions for watershed conservation, sustainable management, and the prevention of land degradation. These insights are invaluable for addressing the challenges posed by gully erosion within the region, and beyond its geographical boundaries and can be used for defining appropriate mitigation strategies at local to national scale.

Understanding urban expansion and shrinkage via green plastic cover mapping based on GEE cloud platform: A case study of Shandong, China

Green Plastic Cover (GPC) has been widely used in construction sites of China as the main measure for dust suppression. As a unique land cover in China, GPC could be viewed as a sign for evaluating the intensity of urban construction. Therefore, time series GPC maps could provide a new perspective to understand the pattern of urban development, including both urban expansion and urban shrinkage. The objective of this study is to present a novel framework for detection of urban development patterns based on high resolution GPC maps generated from Google Earth Engine (GEE) cloud platform. Specifically, we first generated the GPC maps of 137 districts and counties in Shandong Province of China from 2018 to 2022 by combining time series spectral features and textural features derived from Sentinel-1 and Sentinel-2 images provided by GEE cloud platform. Afterwards, we designed a GPC-based urban development vigor index (GPC-UDVI) to evaluate the vitality of each city, which offers a novel perspective to identify shrinking cities. Experimental results demonstrated that the average overall accuracy of GPC maps is 89.696 %, which is accurate enough for the following analysis. Besides, the proposed GPC-UDVI exhibits a consistent trend with statistical data and could reveal urban development patterns with a high performance. Above all, dynamic monitoring of urban GPC through remote sensing is significant for understanding urban expansion modes and aiding in urban reconstruction management.

USING OPTICAL SATELLITE IMAGES AND SATELLITE ALTIMETRY DATA TO ESTIMATE VOLUME VARIATIONS IN DAMS

Abstract. This study focused on monitoring the water volume variations of the Doroudzan dam reservoir in Shiraz, Iran, using satellite observations. In particular, Sentinel-3 altimetry mission (SRAL) Level-1B and Level-2 data were employed to calculate water level changes, addressing the limitations in accuracy for inland and shallow waters. Re-tracking of returned waveforms was applied to improve the accuracy of Level-2 altimetry results. Additionally, Sentinel-2 optical images were utilized to monitor the water surface area of the dam reservoir. The results demonstrated that re-tracking the returned waveforms significantly improved the water level observations compared to Level-2 data. The analysis extended to comparing the time series of water surface area estimated from Sentinel-2 images with in-situ data, revealing a high accuracy of 5.39%. Combining optimum water level and surface area data in Heron's equation facilitated the calculation of water volume variations. A remarkable correlation of 95.27% was found when comparing the time series of estimated water volume variations and in-situ data. This study underscores the effectiveness of Copernicus satellites, particularly Sentinel-3 and Sentinel-2 missions, in monitoring inland water bodies and demonstrates the reliability of the techniques employed for tracking dam reservoir volume variations.