Machine Learning Based Identification of River Plastic Litter in Coastal Region Using Sentinel-2 Data: An Automated Approach for Marine Pollution Monitoring and Management

The assessment of ecological functions, such as those of forest structure zoning and carbon sinks, heavily relies on forest age classification. Therefore, monitoring forest age is a crucial element of forest resource surveys. With the increased availability of high-quality open-access satellite data and advancements in Unmanned Aerial Vehicle Light Detection and Ranging (UAV-LiDAR) technology, remote sensing has emerged as an essential method for acquiring accurate forest age information. In this study, Sentinel-2 remote sensing data, UAV-LiDAR data, and combined Sentinel-2 and LiDAR data are used as data sources. Three machine learning algorithms, Adaptive Boosting (AdaBoost), Random Forest (RF), and Extreme Random Tree (ERT), are used to predict forest age in a Masson pine (Pinus massoniana Lamb.) forest. The optimal model is used to predict the forest age and simulate the spatial age distribution. The machine learning models based on separate Sentinel-2 and LiDAR data accurately predict the age of the Masson pine forest. Nevertheless, the accuracy of the RF model with combined data was higher than that in other cases, with an accuracy R value of 0.81. The model displayed good stability, and the spatial uncertainty of age estimation was low. Compared with the RF model using only Sentinel-2 data (R = 0.43), the RF model with combined LiDAR and Sentinel-2 data achieved the highest accuracy, with R values 88.37% higher. In addition, the forest canopy structure parameters, such as the average height of the forest stand extracted from UAV-LiDAR data, had a significant impact on the estimation of forest age. Thus, when the combined Sentinel-2 and LiDAR data were used to establish these parameters, the highest accuracy in the estimation of Masson pine was obtained. The findings of this study provide new insights for forest age estimation based on multi-source remote sensing data.

Rural communities rely on smallholder maize farms for subsistence agriculture, the main driver of local economic activity and food security. However, their planted area estimates are unknown in most developing countries. This study explores the use of Sentinel-1 and Sentinel-2 data to map smallholder maize farms. The random forest (RF), support vector (SVM) machine learning algorithms and model stacking (ST) were applied. Results show that the classification of combined Sentinel-1 and Sentinel-2 data improved the RF, SVM and ST algorithms by 24.2%, 8.7%, and 9.1%, respectively, compared to the classification of Sentinel-1 data individually. Similarities in the estimated areas (7001.35 ± 1.2 ha for RF, 7926.03 ± 0.7 ha for SVM and 7099.59 ± 0.8 ha for ST) show that machine learning can estimate smallholder maize areas with high accuracies. The study concludes that the single-date Sentinel-1 data were insufficient to map smallholder maize farms. However, single-date Sentinel-1 combined with Sentinel-2 data were sufficient in mapping smallholder farms. These results can be used to support the generation and validation of national crop statistics, thus contributing to food security.

We train and compare the performance of two different machine learning algorithms to learn changes in winter wheat production for fields from the southwest of Sweden. As input to these algorithms, we use cloud-penetrating Sentinel-1 polarimetry radar data together with respective field topography and local weather over four different years. We note that all of the input data were freely available. During training, we used information on winter wheat production over the fields of interest which was available from participating farmers. The two machine learning models we trained are the Light Gradient-Boosting Machine and a Feed-forward Neural Network. Our results show that Sentinel-1 data contain valuable information which can be used for training to predict winter wheat yield once two important steps are taken: performing a critical transformation of each pixel in the images to align it to the training data and then following up with despeckling treatment. Using this approach, we were able to achieve a top root mean square error of 0.75 tons per hectare and a top accuracy of 86% using a k-fold method with k=5. More importantly, however, we established that Sentinel-1 data alone are sufficient to predict yield with an average root mean square error of 0.89 tons per hectare, making this method feasible to employ worldwide.

<p>The climate change induced increased warming of the Arctic is leading to an accelerated thawing of permafrost, which can cause ground subsidence. In consequence, buildings and other infrastructure of local settlements are endangered from destabilization and collapsing in many Arctic regions. The increase of the exploitation of Arctic natural resources has led to the establishment of large industrial infrastructures that are at risk likewise. Most of the human activity in the Arctic is located near permafrost coasts. The thawing of coastal permafrost additionally leads to coastal erosion, which makes Arctic coastal settlements even more vulnerable.</p><p>The European Union (EU) Horizon2020 project “Nunataryuk” aims to assess the impacts of thawing land, coast and subsea permafrost on the climate and on local communities in the Arctic. One task of the project is to determine the impacts of permafrost thaw on coastal Arctic infrastructures and to provide appropriate adaptation and mitigation strategies. For that purpose, a circumpolar account of infrastructure is needed.</p><p>During recent years, the two polar-orbiting Sentinel-2 satellites of the Copernicus program of the EU have been acquiring multi-spectral imagery at high spatial and temporal resolution. Sentinel-2 data is a common choice for land cover mapping. Most land cover products only include one class for built-up areas, however. The fusion of optical and Synthetic Aperture Radar (SAR) data for land cover mapping has gained more and more attention over the last years. By combining Sentinel-2 and Sentinel-1 SAR data, the classification of multiple types of infrastructure can be anticipated. Another emerging trend is the application machine learning and deep learning methods for land cover mapping.</p><p>We present an automated workflow for downloading, processing and classifying Sentinel-2 and Sentinel-1 data in order to map coastal infrastructure with circum-Arctic extent, developed on a highly performant virtual machine (VM) provided by the Copernicus Research and User Support (RUS). We further assess the first classification results mapped with two different methods, one being a pixel-based classification using a Gradient Boosting Machine and the other being a windowed semantic segmentation approach using the deep-learning framework keras.</p>

Climate change is significantly affecting mountain plant communities, causing dynamic alterations in species composition as well as spatial distribution. This raises the need for constant monitoring. The Tatra Mountains are the highest range of the Carpathians which are considered biodiversity hotspots in Central Europe. For this purpose, microwave Sentinel-1 and optical multi-temporal Sentinel-2 data, topographic derivatives, and iterative machine learning methods incorporating classifiers random forest (RF), support vector machines (SVMs), and XGBoost (XGB) were used for the identification of thirteen non-forest plant communities (various types of alpine grasslands, shrublands, herbaceous heaths, mountain hay meadows, rocks, and scree communities). Different scenarios were tested to identify the most important variables, retrieval periods, and spectral bands. The overall accuracy results for the individual algorithms reached RF (0.83–0.96), SVM (0.87–0.93), and lower results for XGBoost (0.69–0.82). The best combination, which included a fusion of Sentinel-1, Sentinel-2, and topographic data, achieved F1-scores for classes in the range of 0.73–0.97 (RF) and 0.66–0.95 (SVM). The inclusion of topographic variables resulted in an improvement in F1-scores for Sentinel-2 data by one–four percent points and Sentinel-1 data by 1%–9%. For spectral bands, the Sentinel-2 10 m resolution bands B4, B3, and B2 showed the highest mean decrease accuracy. The final result is the first comprehensive map of non-forest vegetation for the Tatra Mountains area.

Abstract. Coastal erosion poses a continuous threat to ecosystems, infrastructure, and property. To address these challenges and mitigate the effects of coastal changes, effective and current monitoring is essential. It is particularly important to monitor coastlines and coastal changes in Africa, where a significant portion of the population resides in coastal regions. While optical satellite imagery has been used for large-scale annual coastlines and change monitoring for Africa, its availability and quality are largely limited by the presence of cloud and cloud shadow. In comparison, using radar satellite observations such as Sentinel-1 data can provide consistent coastal mapping and change detection regardless of cloud presence. This paper outlines a fully automated supervised machine learning workflow using Sentinel-1 data and training samples extracted from Sentinel-2 data. It also explores the performance of the workflow for different coastal morphology types across the African coast. The workflow has proved to perform better and produced results that were visually more consistent with Sentinel-2 data compared to thresholding methods. While challenges exist to distinguish between land and water over smooth sandy beaches and rough near-shore water surfaces, our workflow provides an alternative method for coastal change mapping where optical satellites provide insufficient observations free from clouds. Python code of the proposed methodology has been made publicly available.

Coastal regions are increasingly vulnerable to ground deformation hazards, which can cause structural damage and amplify flood risks. Accurately monitoring the evolution of such deformation is crucial for hazard assessment. However, the precision of multi-temporal interferometric synthetic aperture radar (MT-InSAR), a powerful geodetic technique for mapping ground deformation, is often compromised by the atmospheric phase screen (APS) effect. This is more serious in coastal zones, where it can bias the detection of deformation turning points and mislead the interpretation. In this paper, we introduce a novel approach that designs a non-stationarity test for independent component analysis (ICA) to effectively separate the APS and deformation phase components in MT-InSAR applications. Through simulated experiments, the proposed method demonstrated a 50% improvement in deformation accuracy and can effectively track deformation progression. We validated the new method with a case study in Nantong, a coastal region along the northern Yangtze River estuary in China, using Sentinel-1 data from 2015 to 2023. The proposed method retrieved the ground deformation over Nantong with a root-mean-square-error (RMSE) of less than 5.6 mm when compared to ground leveling measurements, which surpasses the traditional MT-InSAR methods. The study results identified diverse ground deformation patterns in Nantong, with deformation rates ranging from −56.3 to 45.9 mm/year, attributed to groundwater extraction, urbanization activities, and land reclamation efforts. The study also highlights the significant coastal accretion and land reclamation processes in the study area, demonstrating the potential capability of the MT-InSAR technique in detecting coastal erosion detection and informing land reservation.

Semantic segmentation of slums in satellite images using transfer learning on fully convolutional neural networks

Prediction of topsoil organic carbon content with Sentinel-2 imagery and spectroscopic measurements under different conditions using an ensemble model approach with multiple pre-treatment combinations

The sustainability index (SI) is a relatively new concept for measuring the performance of water resource systems over long time periods. The purpose of its definition is to provide an indication of the integral behavior of the system with regard to possible undesired consequences if a misbalance in available and required waters occurs. Therefore, the tidal river management (TRM) approach has been implemented for the past three decades (from 1990 to 2020) within the polder system in Southwest Bangladesh to achieve water sustainability. TRM plan and watershed management plan (WMP) have commonalities as both are aimed at ensuring the sustainable use of watershed resources with the management of land, water, and the wider ecosystem of the watershed in an integrated way. The TRM plan focuses mostly on coastal regions, whereas the WMP focuses on both coastal and non-coastal regions. According to this, the aim of this study was to explore the application of the sustainability index of tidal river management (SITRM) in measuring the sustainability of tidal river management in the coastal area of the Lower Ganges–Brahmaputra–Meghna (GBM) delta. In order to quantify the sustainability of tidal river management, this research first provided the components and indicators of SITRM for the coastal region. The study follows a 5-point Likert scale for opinion survey of key informants and comprises households’ survey of farmers. In addition, it includes Landsat satellite images from Earth Explorer of the United States Geological Survey (USGS) and direct field observation to collect information regarding the indicators of SITRM. The study measures the index value of SITRM for identifying the water sustainability of Beel East Khukshia-TRM. The index value was 71.8 out of 100, showing good tidal river management for the Hari–Teka–Bhadra catchment. To achieve water sustainability and aid stakeholders and water managers in decision making, it may be possible to include the SITRM framework in tidal river management projects. In addition, the SITRM is more capable of facing drainage congestion, waterlogging, and climate change issues than watershed sustainability index (WSI), Canadian water sustainability index (CWSI), West Java water sustainability index (WJWSI), and water poverty index (WPI). Therefore, water professionals and policymakers can apply SITRM to assess the resilience of specific TRM schemes for greater sustainability in different coastal regions of the world.

Paddy rice mapping is crucial for cultivation management, yield estimation, and food security. Guangdong, straddling tropics and subtropics, is a major rice-producing region in China. Mapping paddy rice in Guangdong is essential. However, there are 2 main difficulties in tropical and subtropical paddy rice mapping, including the lack of high-quality optical images and differences in paddy rice planting times. This study proposed a paddy rice mapping framework using phenology matching, integrating Sentinel-1 and Sentinel-2 data to incorporate prior knowledge into the classifiers. The transplanting periods of paddy rice were identified with Sentinel-1 data, and the subsequent 3 months were defined as the growth periods. Features during growth periods obtained by Sentinel-1 and Sentinel-2 were inputted into machine learning classifiers. The classifiers using matched features substantially improved mapping accuracy compared with those using unmatched features, both for early and late rice mapping. The proposed method also improved the accuracy by 6.44% to 16.10% compared with 3 other comparison methods. The model, utilizing matched features, was applied to early and late rice mapping in Guangdong in 2020. Regression results between mapping area and statistical data validate paddy rice mapping credibility. Our analysis revealed that thermal conditions, especially cold severity during growing stages, are the primary determinant of paddy rice phenology. Spatial patterns of paddy rice in Guangdong result from a blend of human and physical factors, with slope and minimum temperature emerging as the most important limitations. These findings enhance our understanding of rice ecosystems’ dynamics, offering insights for formulating relevant agricultural policies.

This research compares the use of the SAR (Sentinel-1) and Optical (Sentinel-2) sensors in identifying and mapping burnt and unburnt scars are rising during a bushfire in southeastern Australia and Margalla Hills, Islamabad, Pakistan, in 2019 and 2020. In order to evaluate the backscatter strength along with the Polarimetric decomposition portion, the C-band dual-polarized Sentinel-1 data was investigated to determine the magnitude of the burnt areas of forest cover in the study area. We could derive texture measurements from locally-based statistics using the Grey Level Co-occurrence Matrix (GLCM) and the backscatter coefficient. This was because of how well it picked up on differences in texture between burned and unburned scars. In contrast, Sentinel-2 optical remote sensing was employed to evaluate the extent of the burnt intensity levels for both regions utilizing the differential Normalized Burnt Ratio (dNBR). A Support Vector Machine (SVM) and Markov Random Field (MRF) classifier were utilized to investigate the study’s context. The ideal smoothing parameter is the result of incorporating the image’s spectral characteristics and spatial meaning. Sentinel-2 images were used as a foundation for both the test and training datasets, which were built from images of both unburned and burned areas broken down pixel by pixel. In both types, including spectral sensitivity and sensitivity of Polarimetric for the two groups identified after classification, the experimental findings showed a clear association between them. The algorithm’s efficiency was evaluated using the kappa coefficient and F-score calculation. Except for Sentinel-1 data in Pakistan, all fire areas have more than 0.80 accuracies. The highest precision of both Sentinel-1 and Sentinel-2 was also provided by the performance of users’ and producers’ accuracy. The entropy alpha decomposition helped define the target given by the H-a plane based on its physical properties. After the burn, the entropy and alpha values diminished and formed a pattern. However, the findings in this field validate the effectiveness of SAR sensors data and optical satellite in forest applications. The related sensitivity is highly dependent on the composition of the landscape, the geographical nature of the study area, and the severity of the burn.

This paper aims to develop a supervised classification integrating synthetic aperture radar (SAR) Sentinel-1 (S1) and optical Sentinel-2 (S2) data for land use/land cover (LULC) mapping in a heterogeneous Mediterranean forest area. The time-series of each SAR and optical bands, three optical indices (normalized difference vegetation index, NDVI; normalized burn ratio, NBR; normalized difference red-edge index, NDRE), and two SAR indices (radar vegetation index, RVI; radar forest degradation index, RFDI), constituted the dataset. The coherence information from SAR interferometry (InSAR) analysis and three optical biophysical variables (leaf area index, LAI; fraction of green vegetation cover, fCOVER; fraction of absorbed photosynthetically active radiation, fAPAR) of the single final month of the time-series were added to exploit their correlation with the canopy structure and improve the classification. The random forests (RF) algorithm was used to train and classify the final dataset, and an exhaustive grid search analysis was applied to set the optimal hyperparameters. The overall accuracy reached an F-scoreM of 90.33% and the integration of SAR improved it by 2.53% compared to that obtained using only optical data. The whole process was performed using freely available data and open-source software and libraries (SNAP, Google Earth Engine, Scikit-Learn) executed in Python-script language.

Land use and land cover (LULC) mapping can be of great help in changing land use decisions, but accurate mapping of LULC categories is challenging, especially in semi-arid areas with extensive farming systems and seasonal vegetation phenology. Machine learning algorithms are now widely used for LULC mapping because they provide analytical capabilities for LULC classification. However, the use of machine learning algorithms to improve classification performance is still being explored. The objective of this study is to investigate how to improve the performance of LULC models to reduce prediction errors. To address this question, the study applied a Random Forest (RF) based feature selection approach using Sentinel-1, -2, and Shuttle Radar Topographic Mission (SRTM) data. Results from RF show that the Sentinel-2 data only achieved an out-of-bag overall accuracy of 84.2%, while the Sentinel-1 and SRTM data achieved 83% and 76.44%, respectively. Classification accuracy improved to 89.1% when Sentinel-2, Sentinel-1 backscatter, and SRTM data were combined. This represents a 4.9% improvement in overall accuracy compared to Sentinel-2 alone and a 6.1% and 12.66% improvement compared to Sentinel-1 and SRTM data, respectively. Further independent validation, based on equally sized stratified random samples, consistently found a 5.3% difference between the Sentinel-2 and the combined datasets. This study demonstrates the importance of the synergy between optical, radar, and elevation data in improving the accuracy of LULC maps. In principle, the LULC maps produced in this study could help decision-makers in a wide range of spatial planning applications.

Invasive annual grasses, such as cheatgrass (Bromus tectorum L.), have proliferated in dryland ecosystems of the western United States, promoting increased fire activity and reduced biodiversity that can be detrimental to socio-environmental systems. Monitoring exotic annual grass cover and dynamics over large areas requires the use of remote sensing that can support early detection and rapid response initiatives. However, few studies have leveraged remote sensing technologies and computing frameworks capable of providing rangeland managers with maps of exotic annual grass cover at relatively high spatiotemporal resolutions and near real-time latencies. Here, we developed a system for automated mapping of invasive annual grass (%) cover using in situ observations, harmonized Landsat and Sentinel-2 (HLS) data, maps of biophysical variables, and machine learning techniques. A robust and automated cloud, cloud shadow, water, and snow/ice masking procedure (mean overall accuracy >81%) was implemented using time-series outlier detection and data mining techniques prior to spatiotemporal interpolation of HLS data via regression tree models (r = 0.94; mean absolute error (MAE) = 0.02). Weekly, cloud-free normalized difference vegetation index (NDVI) image composites (2016–2018) were used to construct a suite of spectral and phenological metrics (e.g., start and end of season dates), consistent with information derived from Moderate Resolution Image Spectroradiometer (MODIS) data. These metrics were incorporated into a data mining framework that accurately (r = 0.83; MAE = 11) modeled and mapped exotic annual grass (%) cover throughout dryland ecosystems in the western United States at a native, 30-m spatial resolution. Our results show that inclusion of weekly HLS time-series data and derived indicators improves our ability to map exotic annual grass cover, as compared to distribution models that use MODIS products or monthly, seasonal, or annual HLS composites as primary inputs. This research fills a critical gap in our ability to effectively assess, manage, and monitor drylands by providing a framework that allows for an accurate and timely depiction of land surface phenology and exotic annual grass cover at spatial and temporal resolutions that are meaningful to local resource managers.