Red tide causes significant damage to marine resources such as aquaculture and fisheries in coastal regions. Such red tide events occur globally, across latitudes and ocean ecoregions. Satellite observations can be an effective tool for tracking and investigating red tides and have great potential for informing strategies to minimize their impacts on coastal fisheries. However, previous satellite-based red tide detection algorithms have been mostly conducted over short time scales and within relatively small areas, and have shown significant differences from actual field data, highlighting a need for new, more accurate algorithms to be developed. In this study, we present the newly developed normalized red tide index (NRTI). The NRTI uses Geostationary Ocean Color Imager (GOCI) data to detect red tides by observing in situ spectral characteristics of red tides and sea water using spectroradiometer in the coastal region of Korean Peninsula during severe red tide events. The bimodality of peaks in spectral reflectance with respect to wavelengths has become the basis for developing NRTI, by multiplying the heights of both spectral peaks. Based on the high correlation between the NRTI and the red tide density, we propose an estimation formulation to calculate the red tide density using GOCI data. The formulation and methodology of NRTI and density estimation in this study is anticipated to be applicable to other ocean color satellite data and other regions around the world, thereby increasing capacity to quantify and track red tides at large spatial scales and in real time. Full article

Mangroves are important ecosystems and their distribution and dynamics can provide an understanding of the processes of ecological change. Meanwhile, mangroves protection is also an important element of the Maritime Silk Road (MSR) Cooperation Project. Large amounts of accessible satellite remote sensing data can provide timely and accurate information on the dynamics of mangroves, offering significant advantages in space, time, and characterization. In view of the capability of deep learning in processing massive data in recent years, we developed a new deep learning model—Capsules-Unet, which introduces the capsule concept into U-net to extract mangroves with high accuracy by learning the spatial relationship between objects in images. This model can significantly reduce the number of network parameters to improve the efficiency of data processing. This study uses Landsat data combined with Capsules-Unet to map the dynamics of mangrove changes over the 25 years (1990–2015) along the MSR. The results show that there was a loss in the mangrove area of 1,356,686 ha (about 21.5%) between 1990 and 2015, with anthropic activities such as agriculture, aquaculture, tourism, urban development, and over-development appearing to be the likely drivers of this decline. This information contributes to the understanding of ecological conditions, variability characteristics, and influencing factors along the MSR. Full article

Cracks in concrete can cause the degradation of stiffness, bearing capacity and durability of civil infrastructure. Hence, crack diagnosis is of great importance in concrete research. On the basis of multiple image features, this work presents a novel approach for crack identification of concrete structures. Firstly, the non-local means method is adopted to process the original image, which can effectively diminish the noise influence. Then, to extract the effective features sensitive to the crack, different filters are employed for crack edge detection, which are subsequently tackled by integral projection and principal component analysis (PCA) for optimal feature selection. Moreover, support vector machine (SVM) is used to design the classifiers for initial diagnosis of concrete surface based on extracted features. To raise the classification accuracy, enhanced salp swarm algorithm (ESSA) is applied to the SVM for meta-parameter optimization. The Dempster–Shafer (D–S) fusion algorithm is utilized to fuse the diagnostic results corresponding to different filters for decision making. Finally, to demonstrate the effectiveness of the proposed framework, a total of 1200 images are collected from a real concrete bridge including intact (without crack), longitudinal crack, transverse crack and oblique crack cases. The results validate the performance of proposed method with promising results of diagnosis accuracy as high as 96.25%. Full article

The rapid advance of remote sensing technology during the last few decades provides a new opportunity for measuring detectable estuarine spatial change. Although estuarine surface area and convergence are important hydraulic parameters often used to predict long-term estuarine evolution, the majority of automated analyses of channel plan view dynamics have been specifically written for riverine systems and have limited applicability to most of the estuaries in the world. This study presents MorphEst, a MATLAB-based collection of analysis tools that automatically measure estuarine planform geometry. MorphEst uses channel masks to extract estuarine length, convergence length, estuarine shape, and areal gain and loss of estuarine surface area due to natural or human factors. Comparisons indicated that MorphEst estimates closely matched with independent measurements of estuarine surface area (r = 0.99) and channel width (r = 0.92) of 39 estuaries along the South Korean coast. Overall, this toolbox will help to improve the ability to solve research questions commonly associated with estuarine evolution as it introduces a tool to automatically measure planform geometric features from remotely sensed imagery. Full article

Precise vegetation maps of mountainous areas are of great significance to grasp the situation of an ecological environment and forest resources. In this paper, while multi-source geospatial data can generally be quickly obtained at present, to realize effective vegetation mapping in mountainous areas when samples are difficult to collect due to their perilous terrain and inaccessible deep forest, we propose a novel and intelligent method of sample collection for machine-learning (ML)-based vegetation mapping. First, we employ geo-objects (i.e., polygons) from topographic partitioning and constrained segmentation as basic mapping units and formalize the problem as a supervised classification process using ML algorithms. Second, a previously available vegetation map with rough-scale label information is overlaid on the geo-object-level polygons, and candidate geo-object-based samples can be identified when all the grids’ labels of vegetation types within the geo-objects are the same. Third, various kinds of geo-object-level features are extracted according to high-spatial-resolution remote sensing (HSR-RS) images and multi-source geospatial data. Some unreliable geo-object-based samples are rejected in the candidate set by comparing their features and the rules based on local expert knowledge. Finally, based on these automatically collected samples, we train the model using a random forest (RF)-based algorithm and classify all the geo-objects with labels of vegetation types. A case experiment of Taibai Mountain in China shows that the methodology has the ability to achieve good vegetation mapping results with the rapid and convenient sample collection scheme. The map with a finer geographic distribution pattern of vegetation could clearly promote the vegetation resources investigation and monitoring of the study area; thus, the methodological framework is worth popularizing in the mapping areas such as mountainous regions where the field survey sampling is difficult to implement. Full article

Insect outbreaks are a recurrent natural phenomenon in forest ecosystems expected to increase due to climate change. Recent advances in Unmanned Aerial Vehicles (UAV) and Deep Learning (DL) Networks provide us with tools to monitor them. In this study we used nine orthomosaics and normalized Digital Surface Models (nDSM) to detect and classify healthy and sick Maries fir trees as well as deciduous trees. This study aims at automatically classifying treetops by means of a novel computer vision treetops detection algorithm and the adaptation of existing DL architectures. Considering detection alone, the accuracy results showed 85.70% success. In terms of detection and classification, we were able to detect/classify correctly 78.59% of all tree classes (39.64% for sick fir). However, with data augmentation, detection/classification percentage of the sick fir class rose to 73.01% at the cost of the result accuracy of all tree classes that dropped 63.57%. The implementation of UAV, computer vision and DL techniques contribute to the development of a new approach to evaluate the impact of insect outbreaks in forest. Full article

The province of León preserves a unique hydraulic infrastructure 1200 km-long, used for the exploitation of auriferous deposits in Roman times. It represents the most extensive waterworks in Europe and is one of the best-preserved examples of mining heritage in Antiquity. In this work, three mining exploitation sectors (upper, middle, and lower) characterized by channels and leats developed in different geological materials were examined, using Unmanned Aerial Vehicles (UAVs). A multi-approach based on a comparison of photogrammetric and multispectral data improved the identification and description of the hydraulic network. Comparison with traditional orthoimages and LiDAR data suggests that UAV-derived multispectral images are of great interest in areas where these sets of data have low resolution or areas that are densely covered by vegetation. The results showed that the size of the channel box and its width were factors that do not depend exclusively on the available water resources, as previously suggested, but also on the geological and hydraulic conditioning factors that intervene in each sector. Additionally, the detailed study allowed the establishment of a water sheet maximum height that was much lower than previously thought. All in all, these inferences might help researchers develop new strategies for mapping the Roman mining infrastructure and establishing the importance of geological inheritance on the construction of the hydraulic system that led the Romans to the accomplishment of the largest mining infrastructure ever known in Europe. Full article

Phytoplankton phenology and size structure are key ecological indicators that influence the survival and recruitment of higher trophic levels, marine food web structure, and biogeochemical cycling. For example, the presence of larger phytoplankton cells supports food chains that ultimately contribute to fisheries resources. Monitoring these indicators can thus provide important information to help understand the response of marine ecosystems to environmental change. In this study, we apply the phytoplankton size model of Gittings et al. (2019b) to 20-years of satellite-derived ocean colour observations in the northern and central Red Sea, and investigate interannual variability in phenology metrics for large phytoplankton (>2 µm in cell diameter). Large phytoplankton consistently bloom in the winter. However, the timing of bloom initiation and termination (in autumn and spring, respectively) varies between years. In the autumn/winter of 2002/2003, we detected a phytoplankton bloom, which initiated ~8 weeks earlier and lasted ~11 weeks longer than average. The event was linked with an eddy dipole in the central Red Sea, which increased nutrient availability and enhanced the growth of large phytoplankton. The earlier timing of food availability directly impacted the recruitment success of higher trophic levels, as represented by the maximum catch of two commercially important fisheries (Sardinella spp. and Teuthida) in the following year. The results of our analysis are essential for understanding trophic linkages between phytoplankton and fisheries and for marine management strategies in the Red Sea. Full article

Leptospirosis, an environmental infectious disease of bacterial origin, is the infectious disease with the highest associated mortality in Seychelles. In small island territories, the occurrence of the disease is spatially heterogeneous and a better understanding of the environmental factors that contribute to the presence of the bacteria would help implement targeted control. The present study aimed at identifying the main environmental parameters correlated with animal reservoirs distribution and Leptospira infection in order to delineate habitats with highest prevalence. We used a previously published dataset produced from a large collection of rodents trapped during the dry and wet seasons in most habitats of Mahé, the main island of Seychelles. A land use/land cover analysis was realized in order to describe the various environments using SPOT-5 images by remote sensing (object-based image analysis). At each sampling site, landscape indices were calculated and combined with other geographical parameters together with rainfall records to be used in a multivariate statistical analysis. Several environmental factors were found to be associated with the carriage of leptospires in Rattus rattus and Rattus norvegicus, namely low elevations, fragmented landscapes, the proximity of urbanized areas, an increased distance from forests and, above all, increased precipitation in the three months preceding trapping. The analysis indicated that Leptospira renal carriage could be predicted using the species identification and a description of landscape fragmentation and rainfall, with infection prevalence being positively correlated with these two environmental variables. This model may help decision makers in implementing policies affecting urban landscapes and/or in balancing conservation efforts when designing pest control strategies that should also aim at reducing human contact with Leptospira-laden rats while limiting their impact on the autochthonous fauna. Full article

The classification of high-resolution (HR) synthetic aperture radar (SAR) images is of great importance for SAR scene interpretation and application. However, the presence of intricate spatial structural patterns and complex statistical nature makes SAR image classification a challenging task, especially in the case of limited labeled SAR data. This paper proposes a novel HR SAR image classification method, using a multi-scale deep feature fusion network and covariance pooling manifold network (MFFN-CPMN). MFFN-CPMN combines the advantages of local spatial features and global statistical properties and considers the multi-feature information fusion of SAR images in representation learning. First, we propose a Gabor-filtering-based multi-scale feature fusion network (MFFN) to capture the spatial pattern and get the discriminative features of SAR images. The MFFN belongs to a deep convolutional neural network (CNN). To make full use of a large amount of unlabeled data, the weights of each layer of MFFN are optimized by unsupervised denoising dual-sparse encoder. Moreover, the feature fusion strategy in MFFN can effectively exploit the complementary information between different levels and different scales. Second, we utilize a covariance pooling manifold network to extract further the global second-order statistics of SAR images over the fusional feature maps. Finally, the obtained covariance descriptor is more distinct for various land covers. Experimental results on four HR SAR images demonstrate the effectiveness of the proposed method and achieve promising results over other related algorithms. Full article

Biomass is a principal variable in crop monitoring and management and in assessing carbon cycling. Remote sensing combined with field measurements can be used to estimate biomass over large areas. This study assessed leaf biomass of Agave sisalana (sisal), a perennial crop whose leaves are grown for fibre production in tropical and subtropical regions. Furthermore, the residue from fibre production can be used to produce bioenergy through anaerobic digestion. First, biomass was estimated for 58 field plots using an allometric approach. Then, Sentinel-2 multispectral satellite imagery was used to model biomass in an 8851-ha plantation in semi-arid south-eastern Kenya. Generalised Additive Models were employed to explore how well biomass was explained by various spectral vegetation indices (VIs). The highest performance (explained deviance = 76%, RMSE = 5.15 Mg ha⁻¹) was achieved with ratio and normalised difference VIs based on the green (R560), red-edge (R740 and R783), and near-infrared (R865) spectral bands. Heterogeneity of ground vegetation and resulting background effects seemed to limit model performance. The best performing VI (R740/R783) was used to predict plantation biomass that ranged from 0 to 46.7 Mg ha⁻¹ (mean biomass 10.6 Mg ha⁻¹). The modelling showed that multispectral data are suitable for assessing sisal leaf biomass at the plantation level and in individual blocks. Although these results demonstrate the value of Sentinel-2 red-edge bands at 20-m resolution, the difference from the best model based on green and near-infrared bands at 10-m resolution was rather small. Full article

Landslides and rock falls are among the many phenomena that have an impact on sustainable construction and infrastructure safety. The main causes of landslides are natural meteorological and hydrological phenomena. In building design and construction, environmental monitoring by identifying geotechnical hazards must be taken into account, as appropriate hazard assessment contributes to ensuring future construction safety. The Carpathian region in southern Poland is particularly predisposed to landslide formation. This may be favored by the nature of the shapes associated with the high and steep slopes of the region’s valleys. Another reason for concern is the flysch geological structure, which is characterized by alternating layers of water-permeable sandstones and poorly permeable shales, clays, and marls. Furthermore, the presence of a quaternary weathering cover makes the geological structure more susceptible to landslide processes and tectonic formations. The paper presents the results of a study whose aim was to elaborate a detailed landslide hazard map for a selected area in the Polish Carpathians, using statistical methods. The approach is based on the Hellwig method, which seems particularly useful in the assessment of susceptibility and landslide hazards on a local scale for a relatively small area. A two-stage study was conducted. The first stage of the research involved the creation of a database associated with environmental parameters and triggering factors, whereas the second stage consisted of the adoption of weights for seven thematic sections and their special features on the basis of expert knowledge. The hazard map developed as a result was compared to the mapping made using the weight-of-evidence method. The proposed data normalization method allows the use and analysis of both qualitative and quantitative data collected from various sources. The advantage of this method is the simple calculation procedure. A large-scale (1:2000) map might be used to assess the landslide hazard for specific cadastral units. Such a map becomes the basis for municipal spatial planning and may be able to influence investment decisions. Detailed landslide hazard maps are crucial for more precise risk evaluation for specific cadastral units. This, in turn, allows one to reduce serious economic and social losses, which might be the future results of landslides. Full article

Ice-rich permafrost landscapes are sensitive to ongoing changes in climate. Permafrost retrogressive thaw slumps (RTSs) represent one of the more abrupt and prolonged disturbances, which occur along Arctic river and lake shorelines. These features impact local travel and infrastructure, and there are many questions regarding associated impacts on biogeochemical cycling. Predicting the duration and magnitude of impacts requires that we enhance our knowledge of RTS geomorphological drivers and rates of change. Here we demonstrate the utility of remotely piloted aircraft systems (RPAS) for documenting the volumetric change, associated drivers and potential impacts of the largest active RTS along the Old Crow River in Old Crow Flats, Yukon, Canada. RPAS surveys revealed that 29,174 m³ of sediment was exported during the initial evacuation in June 2016 and an additional 18,845 m³ continued to be exported until June 2019. More sediment export occurred during the warmer 2017 summer that experienced less cumulative rainfall than summer 2018. However, several rain events during 2017 were of higher intensity than during 2018. Overall mean soil organic carbon (SOC) and total nitrogen (TN) within sampled thaw slump sediment was 1.36% and 0.11%, respectively. A combination of multispectral, thermal and irradiance (derived from the RPAS digital surface model) data provided detailed classification of thaw slump floor terrain types including raised dry clay lobes, shaded and relatively stable, and low-lying evacuation-prone sediments. Notably, the path of evacuation-prone sediments extended to a series of ice wedges in the northern headwall, where total irradiance was highest. Using thaw slump floor mean SOC and TN values in conjunction with sediment bulk density and thaw slump fill volume, we estimated that 713 t SOC and 58 t TN were exported to the Old Crow River during the three-year study. Findings showcase the utility of high-resolution RPAS datasets for refining our knowledge of thaw slump geomorphology and associated impacts. Full article

Generative adversarial networks (GANs) are a type of neural network that are characterized by their unique construction and training process. Utilizing the concept of the latent space and exploiting the results of a duel between different GAN components opens up interesting opportunities for computer vision (CV) activities, such as image inpainting, style transfer, or even generative art. GANs have great potential to support aerial and satellite image interpretation activities. Carefully crafting a GAN and applying it to a high-quality dataset can result in nontrivial feature enrichment. In this study, we have designed and tested an unsupervised procedure capable of engineering new features by shifting real orthophotos into the GAN’s underlying latent space. Latent vectors are a low-dimensional representation of the orthophoto patches that hold information about the strength, occurrence, and interaction between spatial features discovered during the network training. Latent vectors were combined with geographical coordinates to bind them to their original location in the orthophoto. In consequence, it was possible to describe the whole research area as a set of latent vectors and perform further spatial analysis not on RGB images but on their lower-dimensional representation. To accomplish this goal, a modified version of the big bidirectional generative adversarial network (BigBiGAN) has been trained on a fine-tailored orthophoto imagery dataset covering the area of the Pilica River region in Poland. Trained models, precisely the generator and encoder, have been utilized during the processes of model quality assurance and feature engineering, respectively. Quality assurance was performed by measuring model reconstruction capabilities and by manually verifying artificial images produced by the generator. The feature engineering use case, on the other hand, has been presented in a real research scenario that involved splitting the orthophoto into a set of patches, encoding the patch set into the GAN latent space, grouping similar patches latent codes by utilizing hierarchical clustering, and producing a segmentation map of the orthophoto. Full article

Surface oceanic fronts are regions characterized by high biological activity. Here, Sea Surface Temperature (SST) fronts are analyzed for the period 2003–2019 using the Multi-scale Ultra-high Resolution (MUR) SST product in northern Patagonia, a coastal region with high environmental variability through river discharges and coastal upwelling events. SST gradient magnitudes were maximum off Chiloé Island in summer and fall, coherent with the highest frontal probability in the coastal oceanic area, which would correspond to the formation of a coastal upwelling front in the meridional direction. Increased gradient magnitudes in the Inner Sea of Chiloé (ISC) were found primarily in spring and summer. The frontal probability analysis revealed the highest occurrences were confined to the northern area (north of Desertores Islands) and around the southern border of Boca del Guafo. An Empirical Orthogonal Function analysis was performed to clarify the dominant modes of variability in SST gradient magnitudes. The meridional coastal fronts explained the dominant mode (78% of the variance) off Chiloé Island, which dominates in summer, whereas the SST fronts inside the ISC (second mode; 15.8%) were found to dominate in spring and early summer (October–January). Future efforts are suggested focusing on high frontal probability areas to study the vertical structure and variability of the coastal fronts in the ISC and its adjacent coastal ocean. Full article