Google Earth Engine, Open-Access Satellite Data, and Machine Learning in Support of Large-Area Probabilistic Wetland Mapping

Carbon footprints of tailings dams' disasters: A study in the Brumadinho region (Brazil)

Tailings dams' breaks are environmental disasters with direct and intense degradation of soil. This study analyzed the impacts of B1 tailings dam rupture occurred in the Ribeirão Ferro-Carvão watershed (Brumadinho, Brazil) in January 25, 2019. Soil organic carbon (SOC) approached environmental degradation. The analysis encompassed wetlands (high-SOC pools) located in the so-called Zones of Decreasing Destructive Capacity (DCZ5 to DCZ1) defined along the Ferro-Carvão's stream bed and banks after the disaster. Remote sensed water indices were extracted from Landsat 8 and Sentinel-2 satellite images spanning the 2017-2021 period and used to distinguish the wetlands from other land covers. The annual SOC was extracted from the MapBiomas repository inside and outside the DCZs in the same period, and assessed in the field in 2023. Before the dam collapse, the DCZs maintained stable levels of SOC, while afterwards they decreased substantially reaching minimum values in 2023. The reductions were abrupt: for example, in the DCZ3 the decrease was from 51.28 ton/ha in 2017 to 4.19 ton/ha in 2023. Besides, the SOC increased from DCZs located near to DCZs located farther from the dam site, a result attributed to differences in the percentages of clay and silt in the tailings, which also increased in the same direction. The Ferro-Carvão stream watershed as whole also experienced a slight reduction in the average SOC levels after the dam collapse, from nearly 43 ton/ha in 2017 to 38 ton/ha in 2021. This result was attributed to land use changes related with the management of tailings, namely opening of accesses to remove them from the stream valley, creation of spaces for temporary deposits, among others. Overall, the study highlighted the footprints of tailings dams' accidents on SOC, which affect not only the areas impacted with the mudflow but systemically the surrounding watersheds. This is noteworthy.

Disturbance-mediated changes to boreal mammal spatial networks in industrializing landscapes

Compound effects of anthropogenic disturbances on wildlife emerge through a complex network of direct responses and species interactions. Land-use changes driven by energy and forestry industries are known to disrupt predator-prey dynamics in boreal ecosystems, yet how these disturbance effects propagate across mammal communities remains uncertain. Using structural equation modeling, we tested disturbance-mediated pathways governing the spatial structure of multipredator multiprey boreal mammal networks across a landscape-scale disturbance gradient within Canada's Athabasca oil sands region. Linear disturbances had pervasive direct effects, increasing site use for all focal species, except black bears and threatened caribou, in at least one landscape. Conversely, block (polygonal) disturbance effects were negative but less common. Indirect disturbance effects were widespread and mediated by caribou avoidance of wolves, tracking of primary prey by subordinate predators, and intraguild dependencies among predators and large prey. Context-dependent responses to linear disturbances were most common among prey and within the landscape with intermediate disturbance. Our research suggests that industrial disturbances directly affect a suite of boreal mammals by altering forage availability and movement, leading to indirect effects across a range of interacting predators and prey, including the keystone snowshoe hare. The complexity of network-level direct and indirect disturbance effects reinforces calls for increased investment in addressing habitat degradation as the root cause of threatened species declines and broader ecosystem change.

Enhancing Wetland Mapping: Integrating Sentinel-1/2, GEDI Data, and Google Earth Engine

Wetlands are amongst Earth's most dynamic and complex ecological resources, serving productive and biodiverse ecosystems. Enhancing the quality of wetland mapping through Earth observation (EO) data is essential for improving effective management and conservation practices. However, the achievement of reliable and accurate wetland mapping faces challenges due to the heterogeneous and fragmented landscape of wetlands, along with spectral similarities among different wetland classes. The present study aims to produce advanced 10 m spatial resolution wetland classification maps for four pilot sites on the Island of Newfoundland in Canada. Employing a comprehensive and multidisciplinary approach, this research leverages the synergistic use of optical, synthetic aperture radar (SAR), and light detection and ranging (LiDAR) data. It focuses on ecological and hydrological interpretation using multi-source and multi-sensor EO data to evaluate their effectiveness in identifying wetland classes. The diverse data sources include Sentinel-1 and -2 satellite imagery, Global Ecosystem Dynamics Investigation (GEDI) LiDAR footprints, the Multi-Error-Removed Improved-Terrain (MERIT) Hydro dataset, and the European ReAnalysis (ERA5) dataset. Elevation data and topographical derivatives, such as slope and aspect, were also included in the analysis. The study evaluates the added value of incorporating these new data sources into wetland mapping. Using the Google Earth Engine (GEE) platform and the Random Forest (RF) model, two main objectives are pursued: (1) integrating the GEDI LiDAR footprint heights with multi-source datasets to generate a 10 m vegetation canopy height (VCH) map and (2) seeking to enhance wetland mapping by utilizing the VCH map as an input predictor. Results highlight the significant role of the VCH variable derived from GEDI samples in enhancing wetland classification accuracy, as it provides a vertical profile of vegetation. Accordingly, VCH reached the highest accuracy with a coefficient of determination (R2) of 0.69, a root-mean-square error (RMSE) of 1.51 m, and a mean absolute error (MAE) of 1.26 m. Leveraging VCH in the classification procedure improved the accuracy, with a maximum overall accuracy of 93.45%, a kappa coefficient of 0.92, and an F1 score of 0.88. This study underscores the importance of multi-source and multi-sensor approaches incorporating diverse EO data to address various factors for effective wetland mapping. The results are expected to benefit future wetland mapping studies.

JAXA's new high-resolution land use land cover map for Vietnam using a time-feature convolutional neural network

Land use land cover (LULC) maps are crucial for various applications, such as disaster management, natural resource conservation, biodiversity evaluation, climate modeling, etc. The Japan Aerospace Exploration Agency (JAXA) has released several high-resolution LULC maps for national and regional scales. Vietnam, due to its rich biodiversity and cultural diversity, is a target country for the production of high-resolution LULC maps. This study introduces a high-resolution and high-accuracy LULC map for Vietnam, utilizing a CNN approach that performs convolution over a time-feature domain instead of the typical geospatial domain employed by conventional CNNs. By using multi-temporal data spanning 6 seasons, the produced LULC map achieved a high overall accuracy of 90.5% ± 1.2%, surpassing other 10-meter LULC maps for Vietnam in terms of accuracy and/or the ability to capture detailed features. In addition, a straightforward and practical approach was proposed for generating cloud-free multi-temporal Sentinel-2 images, particularly suitable for cloudy regions. This study marks the first implementation of the time-feature CNN approach for the creation of a high-accuracy LULC map in a tropical cloudy country.

Mapping small inland wetlands in the South-Kivu province by integrating optical and SAR data with statistical models for accurate distribution assessment

There are several techniques for mapping wetlands. In this study, we examined four statistical models to assess the potential distribution of wetlands in the South-Kivu province by combining optical and SAR images. The approach involved integrating topographic, hydrological, and vegetation indices into the four most used classifiers, namely Artificial Neural Network (ANN), Random Forest (RF), Boosted Regression Tree (BRT), and Maximum Entropy (MaxEnt). A wetland distribution map was generated and classified into 'wetland' and 'non-wetland.' The results showed variations in predictions among the different models. RF exhibited the most accurate predictions, achieving an overall classification accuracy of 95.67% and AUC and TSS values of 82.4%. Integrating SAR data improved accuracy and precision, particularly for mapping small inland wetlands. Our estimations indicate that wetlands cover approximately 13.5% (898,690 ha) of the entire province. BRT estimated wetland areas to be ~ 16% (1,106,080 ha), while ANN estimated ~ 14% (967,820 ha), MaxEnt ~ 15% (1,036,950 ha), and RF approximately ~ 10% (691,300 ha). The distribution of these areas varied across different territories, with higher values observed in Mwenga, Shabunda, and Fizi. Many of these areas are permanently flooded, while others experience seasonal inundation. Through digitization, the delineation process revealed variations in wetland areas, ranging from tens to thousands of hectares. The geographical distribution of wetlands generated in this study will serve as an essential reference for future investigations and pave the way for further research on characterizing and categorizing these areas.

Niche suitability and spatial distribution patterns of anurans in a unique Ecoregion mosaic of Northern Pakistan

The lack of information regarding biodiversity status hampers designing and implementing conservation strategies and achieving future targets. Northern Pakistan consists of a unique ecoregion mosaic which supports a myriad of environmental niches for anuran diversity in comparison to the deserts and xeric shrublands throughout the rest of the country. In order to study the niche suitability, species overlap and distribution patterns in Pakistan, we collected observational data for nine anuran species across several distinct ecoregions by surveying 87 randomly selected locations from 2016 to 2018 in Rawalpindi District and Islamabad Capital Territory. Our model showed that the precipitation of the warmest and coldest quarter, distance to rivers and vegetation were the greatest drivers of anuran distribution, expectedly indicating that the presence of humid forests and proximity to waterways greatly influences the habitable range of anurans in Pakistan. Sympatric overlap between species occurred at significantly higher density in tropical and subtropical coniferous forests than in other ecoregion types. We found species such as Minervarya spp., Hoplobatrachus tigerinus and Euphlyctis spp. preferred the lowlands in proximal, central and southern parts of the study area proximal to urban settlements, with little vegetation and higher average temperatures. Duttaphrynus bengalensis and D. stomaticus had scattered distributions throughout the study area with no clear preference for elevation. Sphaerotheca pashchima was patchily distributed in the midwestern extent of the study area as well as the foothills to the north. Microhyla nilphamariensis was widely distributed throughout the study area with a preference for both lowlands and montane terrain. Endemic frogs (Nanorana vicina and Allopaa hazarensis) were observed only in locations with higher elevations, higher density of streams and lower average temperatures as compared to the other seven species sampled. It is recommended to provide legal protection to amphibians of Pakistan, especially endemic species, through revision in the existing wildlife laws. We suggest studying the effectiveness of existing amphibian tunnels and corridors or designing new ones tailored to the needs of our species to prevent their local extinction due to ongoing or proposed urban development which might affect their dispersal and colonization.

Monitoring spatio-temporal changes in wetlands with harmonized image series in Google Earth Engine

Study of rapidly changing lakes and wetlands with remote sensing methods is critical for understanding the climatic and anthropogenic effects. However, most of the studies search for the change of water body in specific time periods. Although this approach reduces the workload related to downloading and processing a large number of satellite images in computer environment, it actually causes ignoring some critical changes that occurred out of specified time periods. On the other hand, this situation reduces the data volume and the limited data causes problems for the management of water resources.  The Google Earth Engine (GEE) platform allows the opportunity to rapidly and practically process large-scale temporal data without downloading. In this study, areal changes in Lake Akşehir in Türkiye, from 1985 to 2020, were calculated and mapped by the GEE as a case. In order to calculate the changes, the Landsat 5 TM, 7 ETM + and 8 OLI&TIRS images were harmonized and created annual mosaics. The Normalized difference water index (NDWI) and the automated water extraction index (AWEI) were applied to these annual mosaics. By this approach, the change in the water area representing a shrank by 87% on average (according to the calculations 91% for the NDWI and 83% for the AWEI) from 1985 to 2020 was assessed practically and rapidly on annual mosaics created from all images between the studied period, instead of assessment based on images taken on only one date in the chosen years as in previous studies. Such an approach will provide time and labour savings and provide more meaningful and uninterrupted data for studies about changes in other wetland areas.

High-frequency time series comparison of Sentinel-1 and Sentinel-2 satellites for mapping open and vegetated water across the United States (2017-2021)

Frequent observations of surface water at fine spatial scales will provide critical data to support the management of aquatic habitat, flood risk and water quality. Sentinel-1 and Sentinel-2 satellites can provide such observations, but algorithms are still needed that perform well across diverse climate and vegetation conditions. We developed surface inundation algorithms for Sentinel-1 and Sentinel-2, respectively, at 12 sites across the conterminous United States (CONUS), covering a total of >536,000 km2 and representing diverse hydrologic and vegetation landscapes. Each scene in the 5-year (2017-2021) time series was classified into open water, vegetated water, and non-water at 20 m resolution using variables from Sentinel-1 and Sentinel-2, as well as variables derived from topographic and weather datasets. The Sentinel-1 algorithm was developed distinct from the Sentinel-2 model to explore if and where the two time series could potentially be integrated into a single high-frequency time series. Within each model, open water and vegetated water (vegetated palustrine, lacustrine, and riverine wetlands) classes were mapped. The models were validated using imagery from WorldView and PlanetScope. Classification accuracy for open water was high across the 5-year period, with an omission and commission error of only 3.1% and 0.9% for the Sentinel-1 algorithm and 3.1% and 0.5% for the Sentinel-2 algorithm, respectively. Vegetated water accuracy was lower, as expected given that the class represents mixed pixels. The Sentinel-2 algorithm showed higher accuracy (10.7% omission and 7.9% commission error) relative to the Sentinel-1 algorithm (28.4% omission and 16.0% commission error). Patterns over time in the proportion of area mapped as open or vegetated water by the Sentinel-1 and Sentinel-2 algorithms were charted and correlated for a subset of all 12 sites. Our results showed that the Sentinel-1 and Sentinel-2 algorithm open water time series can be integrated at all 12 sites to improve the temporal resolution, but sensor-specific differences, such as sensitivity to vegetation structure versus pixel color, complicate the data integration for mixed-pixel, vegetated water. The methods developed here provide inundation at 5-day (Sentinel-2 algorithm) and 12-day (Sentinel-1 algorithm) time steps to improve our understanding of the short- and long-term response of surface water to climate and land use drivers in different ecoregions.

Headwater streams and inland wetlands: Status and advancements of geospatial datasets and maps across the United States

Headwater streams and inland wetlands provide essential functions that support healthy watersheds and downstream waters. However, scientists and aquatic resource managers lack a comprehensive synthesis of national and state stream and wetland geospatial datasets and emerging technologies that can further improve these data. We conducted a review of existing United States (US) federal and state stream and wetland geospatial datasets, focusing on their spatial extent, permanence classifications, and current limitations. We also examined recent peer-reviewed literature for emerging methods that can potentially improve the estimation, representation, and integration of stream and wetland datasets. We found that federal and state datasets rely heavily on the US Geological Survey's National Hydrography Dataset for stream extent and duration information. Only eleven states (22%) had additional stream extent information and seven states (14%) provided additional duration information. Likewise, federal and state wetland datasets primarily use the US Fish and Wildlife Service's National Wetlands Inventory (NWI) Geospatial Dataset, with only two states using non-NWI datasets. Our synthesis revealed that LiDAR-based technologies hold promise for advancing stream and wetland mapping at limited spatial extents. While machine learning techniques may help to scale-up these LiDAR-derived estimates, challenges related to preprocessing and data workflows remain. High-resolution commercial imagery, supported by public imagery and cloud computing, may further aid characterization of the spatial and temporal dynamics of streams and wetlands, especially using multi-platform and multi-temporal machine learning approaches. Models integrating both stream and wetland dynamics are limited, and field-based efforts must remain a key component in developing improved headwater stream and wetland datasets. Continued financial and partnership support of existing databases is also needed to enhance mapping and inform water resources research and policy decisions.

Long-term detection and spatiotemporal variation analysis of open-surface water bodies in the Yellow River Basin from 1986 to 2020

Accurately investigating long-term information about open-surface water bodies can contribute to water resource protection and management. However, due to the limits of big-data calculations for remote sensing, there has been no specific study on the long-term changes in the water bodies in the Yellow River Basin. Thus, in this study, we developed a new combined extraction rule to build an entire annual-scale open-surface water body dataset for 1986-2020 with excellent effectiveness in eliminating the interference of shadows in the Yellow River Basin using all of the available Landsat images. For the first time, the spatial distribution, change trends, conversion processes, and the heterogeneity of the surface water bodies in the Yellow River Basin were analyzed comprehensively to the best of our knowledge. The extraction results had an overall accuracy of 99.70 % and a kappa coefficient of 0.90, which were validated using 34,073 verification points selected on high-resolution Google Earth images and random Landsat images. The total area of water bodies initially decreased (1986-2000) and then increased (2001-2020); however, only the size of the permanent water bodies increased in most areas, while the size of most of the seasonal water bodies decreased. In regions with human-made water bodies, the non-water areas were substantially converted to seasonal and permanent water bodies; however, in areas with natural water bodies, many permanent and seasonal water bodies were gradually converted to non-water areas. Thus, most of the increases in the water bodies occurred in the form of artificial lakes and reservoirs, while most of the decreases in the water body area occurred in natural wetlands and lakes. The areas of both the permanent and seasonal water bodies were positively correlated with precipitation, but only the area of the seasonal water bodies was negatively correlated with temperature.

Predicting wetland area and water depth in Barind plain of India

The present study attempts to delineate wetlands in the lower Tangon river basin in the Barind flood plain region using spectral water body extraction indices. The main objectives of this present study are simulating and predicting wetland areas using the advanced artificial neural network-based cellular automata (ANN-CA) model and water depth using statistical (adaptive exponential smoothing) as well as advanced machine learning algorithms such as Bagging, Random Subspace, Random Forest, Support vector machine, etc. The result shows that RmNDWI and NDWI are the representative wetland delineating indices. NDWI map was used for water depth prediction. Regarding the prediction of wetland areas, a remarkable decline is likely to be identified in the upcoming two decades. The small wetland patches away from the master stream are expected to dry out during the predicted period, where the major wetland patches nearer to the master stream with greater water depth are rather sustainable, but their depth of water is predicted to be reduced in the next decades. All models show satisfactory performance for wetland depth mapping, but the random subspace model was identified as the best-suited water depth predicting method with an acceptable prediction accuracy (root mean square error <0.34 in all the years) and the machine learning models explored better result than adaptive exponential smoothing. This recent study will be very helpful for the policymakers for managing wetland landscape as well as the natural environment.

Creating a Detailed Wetland Inventory with Sentinel-2 Time-Series Data and Google Earth Engine in the Prairie Pothole Region of Canada

Wetlands in the Prairie Pothole Region (PPR) of Canada and the United States represent a unique mapping challenge. They are dynamic both seasonally and year-to-year, are very small, and frequently altered by human activity. Many efforts have been made to estimate the loss of these important habitats but a high-quality inventory of pothole wetlands is needed for data-driven conservation and management of these resources. Typical landcover classifications using one or two image dates from optical or Synthetic Aperture Radar (SAR) Earth Observation (EO) systems often produce reasonable wetland inventories for less dynamic, forested landscapes, but will miss many of the temporary and seasonal wetlands in the PPR. Past studies have attempted to capture PPR wetland dynamics by using dense image stacks of optical or SAR data. We build upon previous work, using 2017–2020 Sentinel-2 imagery processed through the Google Earth Engine (GEE) cloud computing platform to capture seasonal flooding dynamics of wetlands in a prairie pothole wetland landscape in Alberta, Canada. Using 36 different image dates, wetland flood frequency (hydroperiod) was calculated by classifying water/flooding in each image date. This product along with the Global Ecosystem Dynamics Investigation (GEDI) Canopy Height Model (CHM) was then used to generate a seven-class wetland inventory with wetlands classified as areas with seasonal but not permanent water/flooding. Overall accuracies of the resulting inventory were between 95% and 96% based on comparisons with local photo-interpreted inventories at the Canadian Wetland Classification System class level, while wetlands themselves were classified with approximately 70% accuracy. The high overall accuracy is due, in part, to a dominance of uplands in the PPR. This relatively simple method of classifying water through time generates reliable wetland maps but is only applicable to ecosystems with open/non-complex wetland types and may be highly sensitive to the timing of cloud-free optical imagery that captures peak wetland flooding (usually post snow melt). Based on this work, we suggest that expensive field or photo-interpretation training data may not be needed to map wetlands in the PPR as self-labeling of flooded and non-flooded areas in a few Sentinel-2 images is sufficient to classify water through time. Our approach demonstrates a framework for the operational mapping of small, dynamic PPR wetlands that relies on open-access EO data and does not require costly, independent training data. It is an important step towards the effective conservation and management of PPR wetlands, providing an efficient method for baseline and ongoing mapping in these dynamic environments.

Countrywide Mapping of Plant Ecological Communities with 101 Legends including Land Cover Types for the First Time at 10 m Resolution through Convolutional Learning of Satellite Images

This paper presents next-generation mapping of plant ecological communities including land cover and agricultural types at 10 m spatial resolution countrywide. This research introduces modelling and mapping of land cover and ecological communities separately in small regions-of-interest (prefecture level), and later integrating the outputs into a large scale (country level) for dealing with regional distribution characteristics of plant ecological communities effectively. The Sentinel-2 satellite images were processed for cloud masking and half-monthly median composite images consisting of ten multi-spectral bands and seven spectral indexes were generated. The reliable ground truth data were prepared from extant multi-source survey databases through the procedure of stratified sampling, cross-checking, and noisy-labels pruning. Deep convolutional learning of the time-series of the satellite data was employed for prefecture-wise classification and mapping of 29–62 classes. The classification accuracy computed with the 10-fold cross-validation method varied from 71.1–87.5% in terms of F1-score and 70.9–87.4% in terms of Kappa coefficient across 48 prefectural regions. This research produced seamless maps of 101 ecological communities including land cover and agricultural types for the first time at a country scale with an average accuracy of 80.5% F1-score.

Google Earth Engine and Artificial Intelligence (AI): A Comprehensive Review

Remote sensing (RS) plays an important role gathering data in many critical domains (e.g., global climate change, risk assessment and vulnerability reduction of natural hazards, resilience of ecosystems, and urban planning). Retrieving, managing, and analyzing large amounts of RS imagery poses substantial challenges. Google Earth Engine (GEE) provides a scalable, cloud-based, geospatial retrieval and processing platform. GEE also provides access to the vast majority of freely available, public, multi-temporal RS data and offers free cloud-based computational power for geospatial data analysis. Artificial intelligence (AI) methods are a critical enabling technology to automating the interpretation of RS imagery, particularly on object-based domains, so the integration of AI methods into GEE represents a promising path towards operationalizing automated RS-based monitoring programs. In this article, we provide a systematic review of relevant literature to identify recent research that incorporates AI methods in GEE. We then discuss some of the major challenges of integrating GEE and AI and identify several priorities for future research. We developed an interactive web application designed to allow readers to intuitively and dynamically review the publications included in this literature review.

Improved Object-Based Estimation of Forest Aboveground Biomass by Integrating LiDAR Data from GEDI and ICESat-2 with Multi-Sensor Images in a Heterogeneous Mountainous Region

Accurate and effective mapping of forest aboveground biomass (AGB) in heterogeneous mountainous regions is a huge challenge but an urgent demand for resource managements and carbon storage monitoring. Conventional studies have related the plot-measured or LiDAR-based biomass to remote sensing data using pixel-based approaches. The object-based relationship between AGB and multi-source data from LiDAR, multi-frequency radar, and optical sensors were insufficiently studied. It deserves the further exploration that maps forest AGB using the object-based approach and combines LiDAR data with multi-sensor images, which has the smaller uncertainty of positional discrepancy and local heterogeneity, in heterogeneous mountainous regions. To address the improvement of mapping accuracy, satellite LiDAR data from GEDI and ICEsat-2, and images of ALOS-2 yearly mosaic L band SAR (Synthetic Aperture Radar), Sentinel-1 C band SAR, Sentinel-2 MSI, and ALOS-1 DSM were combined for pixel- and object-based forest AGB mapping in a vital heterogeneous mountainous forest. For the object-based approach, optimized objects during a multiresolution segmentation were acquired by the ESP (Estimation of the Scale Parameter) tool, and suitable predictors were selected using an algorithm named VSURF (Variable Selection Using Random Forests). The LiDAR variables at the footprint-level were extracted to connect field plots to the multi-sensor objects as a linear bridge. It was shown that forests’ AGB values varied by elevations with a mean value of 142.58 Mg/ha, ranging from 12.61 to 514.28 Mg/ha. The north slope with the lowest elevation (<1100 m) had the largest mean AGB, while the smallest mean AGB was located in the south slope with the altitude above 2000 m. Using independent validation samples, it was indicated by the accuracy comparison that the object-based approach performed better on the precision with relative improvement based on root-mean-square errors (RIRMSE) of 4.46%. The object-based approach also selected more optimized predictors and markedly decreased the prediction time than the pixel-based analysis. Canopy cover and height explained forest AGB with their effects on biomass varying according to the elevation. The elevation from DSM and variables involved in red-edge bands from MSI were the most contributive predictors in heterogeneous temperate forests. This study is a pioneering exploration of object-based AGB mapping by combining satellite data from LiDAR, MSI, and SAR, which offers an improved methodology for regional carbon mapping in the heterogeneous mountainous forests.

Monitoring Sand Spit Variability Using Sentinel-2 and Google Earth Engine in a Mediterranean Estuary

Estuarine degradation is a major concern worldwide, and is rapidly increasing due to anthropogenic pressures. The Mediterranean Guadiaro estuary, located in San Roque (Cadiz, Spain), is an example of a highly modified estuary, showing severe negative effects of eutrophication episodes and beach erosion. The migration of its river mouth sand spit causes the closure of the estuary, resulting in serious water quality issues and flora and fauna mortality due to the lack of water renewal. With the aim of studying the Guadiaro estuary throughout a 4-year period (2017–2020), the Sentinel-2 A/B twin satellites of the Copernicus programme were used thanks to their 5-day and 10 m temporal and spatial resolution, respectively. Sea–land mapping was performed using the Normalized Difference Water Index (NDWI) in the Google Earth Engine (GEE) platform, selecting cloud-free Sentinel-2 Level 2A images and computing statistics. Results show a closure trend of the Guadiaro river mouth and no clear sand spit seasonal patterns. The study also reveals the potential of both Sentinel-2 and GEE for estuarine monitoring by means of an optimized processing workflow. This improvement will be useful for coastal management to ensure a continuous and detailed monitoring in the area, contributing to the development of early-warning tools, which can be helpful for supporting an ecosystem-based approach to coastal areas.

Seasonality of inundation in geographically isolated wetlands across the United States

Inundation area is a major control on the ecosystem services provisioned by geographically isolated wetlands. Despite its importance, there has not been any comprehensive study to map out the seasonal inundation characteristics of geographically isolated wetlands over the continental United States (CONUS). This study fills the aforementioned gap by evaluating the seasonality or the long-term intra-annual variations of wetland inundation in ten wetlandscapes across the CONUS. We also assess the consistency of these intra-annual variations. Finally, we evaluate the extent to which the seasonality can be explained based on widely available hydrologic fluxes. Our findings highlight significant intra-annual variations of inundation within most wetlandscapes, with a standard deviation of the long-term averaged monthly inundation area ranging from 15% to 151% of its mean across the wetlandscapes. Stark differences in inundation seasonality are observed between snow-affected vs. rain-fed wetlandscapes. The former usually shows the maximum monthly inundation in April following spring snowmelt (SM), while the latter experiences the maximum in February. Although the magnitude of inundation fraction has changed over time in several wetlandscapes, the seasonality of these wetlands shows remarkable constancy. Overall, commonly available regional hydrologic fluxes (e.g. rainfall, SM, and evapotranspiration) are found to be able to explain the inundation seasonality at wetlandscape scale with determination coefficients greater than 0.57 in 7 out of 10 wetlandscapes. Our methodology and presented results may be used to map inundation seasonality and consequently account for its impact on wetland functions.

A deep learning method for cyanobacterial harmful algae blooms prediction in Taihu Lake, China

Cyanobacterial Harmful Algae Blooms (CyanoHABs) in the eutrophic lakes have become a global environmental and ecological problem. In this study, a CNN-LSTM integrated model for predicting the CyanoHABs area was proposed and applied to the prediction of the CyanoHABs area in Taihu Lake. Firstly, the time-series data of the CyanoHABs area in Taihu Lake for 20 years were accurately obtained using MODIS images from 2000 to 2019 based on the FAI method. Then, a principal component analysis was performed on the daily meteorological data for the month before the outbreak of CyanoHABs in Taihu Lake from 2000 to 2019 to determine the meteorological factors closely related to the outbreak of CyanoHABs. Finally, the features of CyanoHABs area and meteorological data were extracted by Convolutional Neural Networks (CNN) model and used as the input of Long Short Term Memory Network (LSTM). An integrated CNN-LSTM model approach was constructed for predicting the CyanoHABs area. The results show that high R² (0.91) and low mean relative error (17.42%) verified the validity of the FAI index to extract the CyanoHABs area in Taihu Lake; the meteorological factors closely related to the CyanoHABs outbreak in Taihu Lake are mainly temperature, relative humidity, wind speed, and precipitation; the CNN-LSTM integrated model has better prediction effect for both training and test sets compared with the CNN and LSTM models. This study provides an effective method for predicting temporal changes in the CyanoHABs area and offers new ideas for scientific and effective regulation of inland water safety.

Impacts of meander migration on the Amazon riverine communities using Landsat time series and cloud computing

River meander migration is a process that maintains biodiverse riparian ecosystems by producing highly sinuous rivers, and oxbow lakes. However, although the floodplains support communities with fish and other practices in the region, meandering rivers can directly affect the life of local communities. For example, erosion of river banks promotes the loss of land on community shores, while sedimentation increases the distance from house to the river. Therefore, communities living along the Juruá River, one of the most sinuous rivers on Earth, are vulnerable to long-term meander migration. In this study, the river meander migration was detected by using Landsat 5-8 data from 1984 to 2020. A per-pixel Water Surface Change Detection Algorithm (WSCDA) was developed to classify regions subject to erosion and sedimentation processes by applying temporal regressions on the water index, called Modified Normalized Difference Water Index (mNDWI). The WSCDA classified the meander migration with omission and commission errors lower than 13.44% and 7.08%, respectively. Then, the number of riparian communities was mapped using high spatial resolution SPOT images. A total of 369 communities with no road access were identified, the majority of which living in stable regions (58.8%), followed by sedimentation (26.02%) and erosion (15.18%) areas. Furthermore, we identified that larger communities (>20 houses) tend to live in more stable locations (70%) compared to smaller communities (1-10 houses) with 55.6%. A theoretical model was proposed to illustrate the main impacts of meander migration on the communities, related to Inundation, Mobility Change, and Food Security. This is the first study exploring the relationship between meander migration and riverine communities at watershed-level, and the results support the identification of vulnerable communities to improve local planning and floodplain conservation.

Assessing the Ecosystem Health of Coastal Wetland Vegetation (Suaeda salsa) Using the Pressure State Response Model, a Case of the Liao River Estuary in China

Suaeda salsa (S. salsa) is an important ecological barrier and tourism resource in coastal wetland resources, and assessing changes in its health is beneficial for protecting the ecological health of wetlands and increasing finances. The aim was to explore improvements in the degradation of S. salsa communities in the Liao River Estuary National Nature Reserve since a wetland restoration project was carried out in Panjin, Liaoning Province, China, in 2015. In this study, landscape changes in the reserve were assessed based on Sentinel-2 images classification results from 2016 to 2019. A pressure-state-response framework was constructed to assess the annual degradation of S. salsa communities within the wetlands. The assessment results show that the area of S. salsa communities and water bodies decreased annually from 2016 to 2019, and the increased degradation indicators indicate a state of continued degradation. The area of types such as aquaculture ponds and Phragmites australis communities did not change much, while the estuarine mudflats increased year by year. The causes of S. salsa community degradation include anthropogenic impacts from abandoned aquaculture ponds and sluice control systems but also natural impacts from changes in the tidal amplitude and soil properties of the mudflats. The results also indicate that the living conditions of S. salsa in the Liao River estuary wetlands are poor and that anthropogenic disturbance is necessary to restore the original vegetation abundance.

Wetland Change Mapping Using Machine Learning Algorithms, and Their Link with Climate Variation and Economic Growth: A Case Study of Guangling County, China

Wetlands are a distinctive terrestrial ecosystem that benefits living things, including people, in various ways. Sustainable wetland ecosystem resources are needed to protect the global environment. Wetlands in China have undergone positive and negative changes in response to several factors, but studies documenting their long-term dynamicity have been few, particularly in Guangling County. This study examines the change of wetlands area based on remotely sensed data while exploring trends associated with climate variations and economic growth in Guangling County, China. Analysis of remotely sensed imagery, mainly in hilly and nonhomogeneous environments is problematic, largely as a result of interference and their high spectral non-homogeneity. We conducted experiments using five classical machine learning algorithms based on the Google Earth Engine (GEE) and obtained the greatest robustness and accuracy using a Support Vector Machine (SVM)—Radial Basis Function (RBF) kernel approach, with overall accuracy and kappa statistics ranging from 86% to 98.1% and from 0.789 to 0.960, respectively. Based on the SVM-RBF model’s outperformance of four other algorithms, we identified spatial distributions of wetland in the study area and associated change trends. We found that 45.71 km2 of wetland area was lost over the past 3.7 decades (January 1984–December 2020), or 81.82% of wetland area coverage. In this paper, we explore how factors such as county economic growth (GDP), humidity, and temperature variations are tightly linked with wetland change.

Modelling Aboveground Biomass Carbon Stock of the Bohai Rim Coastal Wetlands by Integrating Remote Sensing, Terrain, and Climate Data

Remotely sensed vegetation indices (VIs) have been widely used to estimate the aboveground biomass (AGB) carbon stock of coastal wetlands by establishing Vis-related linear models. However, these models always have high uncertainties due to the large spatial variation and fragmentation of coastal wetlands. In this paper, an efficient coastal wetland AGB model for the Bohami Rim coastal wetlands was presented based on multiple data sets. The model was developed statistically with 7 independent variables from 23 metrics derived from remote sensing, topography, and climate data. Compared to previous models, it had better performance, with a root mean square error and r value of 188.32 g m−2 and 0.74, respectively. Using the model, we firstly generated a regional coastal wetland AGB map with a 10 m spatial resolution. Based on the AGB map, the AGB carbon stock of the Bohai Rim coastal wetland was 2.11 Tg C in 2019. The study demonstrated that integrating emerging high spatial resolution multi-remote sensing data and several auxiliary metrics can effectively improve VIs-based coastal wetland AGB models. Such models with emerging freely available data sets will allow for the rapid monitoring and better understanding of the special role that “blue carbon” plays in global carbon cycle.

A Training Sample Migration Method for Wetland Mapping and Monitoring Using Sentinel Data in Google Earth Engine

Wetlands are one of the most important ecosystems due to their critical services to both humans and the environment. Therefore, wetland mapping and monitoring are essential for their conservation. In this regard, remote sensing offers efficient solutions due to the availability of cost-efficient archived images over different spatial scales. However, a lack of sufficient consistent training samples at different times is a significant limitation of multi-temporal wetland monitoring. In this study, a new training sample migration method was developed to identify unchanged training samples to be used in wetland classification and change analyses over the International Shadegan Wetland (ISW) areas of southwestern Iran. To this end, we first produced the wetland map of a reference year (2020), for which we had training samples, by combining Sentinel-1 and Sentinel-2 images and the Random Forest (RF) classifier in Google Earth Engine (GEE). The Overall Accuracy (OA) and Kappa coefficient (KC) of this reference map were 97.93% and 0.97, respectively. Then, an automatic change detection method was developed to migrate unchanged training samples from the reference year to the target years of 2018, 2019, and 2021. Within the proposed method, three indices of the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and the mean Standard Deviation (SD) of the spectral bands, along with two similarity measures of the Euclidean Distance (ED) and Spectral Angle Distance (SAD), were computed for each pair of reference–target years. The optimum threshold for unchanged samples was also derived using a histogram thresholding approach, which led to selecting the samples that were most likely unchanged based on the highest OA and KC for classifying the test dataset. The proposed migration sample method resulted in high OAs of 95.89%, 96.83%, and 97.06% and KCs of 0.95, 0.96, and 0.96 for the target years of 2018, 2019, and 2021, respectively. Finally, the migrated samples were used to generate the wetland map for the target years. Overall, our proposed method showed high potential for wetland mapping and monitoring when no training samples existed for a target year.

Remote Sensing of Wetlands in the Prairie Pothole Region of North America

The Prairie Pothole Region (PPR) of North America is an extremely important habitat for a diverse range of wetland ecosystems that provide a wealth of socio-economic value. This paper describes the ecological characteristics and importance of PPR wetlands and the use of remote sensing for mapping and monitoring applications. While there are comprehensive reviews for wetland remote sensing in recent publications, there is no comprehensive review about the use of remote sensing in the PPR. First, the PPR is described, including the wetland classification systems that have been used, the water regimes that control the surface water and water levels, and the soil and vegetation characteristics of the region. The tools and techniques that have been used in the PPR for analyses of geospatial data for wetland applications are described. Field observations for ground truth data are critical for good validation and accuracy assessment of the many products that are produced. Wetland classification approaches are reviewed, including Decision Trees, Machine Learning, and object versus pixel-based approaches. A comprehensive description of the remote sensing systems and data that have been employed by various studies in the PPR is provided. A wide range of data can be used for various applications, including passive optical data like aerial photographs or satellite-based, Earth-observation data. Both airborne and spaceborne lidar studies are described. A detailed description of Synthetic Aperture RADAR (SAR) data and research are provided. The state of the art is the use of multi-source data to achieve higher accuracies and hybrid approaches. Digital Surface Models are also being incorporated in geospatial analyses to separate forest and shrub and emergent systems based on vegetation height. Remote sensing provides a cost-effective mechanism for mapping and monitoring PPR wetlands, especially with the logistical difficulties and cost of field-based methods. The wetland characteristics of the PPR dictate the need for high resolution in both time and space, which is increasingly possible with the numerous and increasing remote sensing systems available and the trend to open-source data and tools. The fusion of multi-source remote sensing data via state-of-the-art machine learning is recommended for wetland applications in the PPR. The use of such data promotes flexibility for sensor addition, subtraction, or substitution as a function of application needs and potential cost restrictions. This is important in the PPR because of the challenges related to the highly dynamic nature of this unique region.

Progress and Trends in the Application of Google Earth and Google Earth Engine

Earth system science has changed rapidly due to global environmental changes and the advent of Earth observation technology. Therefore, new tools are required to monitor, measure, analyze, evaluate, and model Earth observation data. Google Earth (GE) was officially launched by Google in 2005 as a ”geobrowser”, and Google Earth Engine (GEE) was released in 2010 as a cloud computing platform with substantial computational capabilities. The use of these two tools or platforms in various applications, particularly as used by the remote sensing community, has developed rapidly. In this paper, we reviewed the applications and trends in the use of GE and GEE by analyzing peer-reviewed articles, dating up to January 2021, in the Web of Science (WoS) core collection using scientometric analysis (i.e., by using CiteSpace) and meta-analysis. We found the following: (1) the number of articles describing the use of GE or GEE increased substantially from two in 2006 to 530 in 2020. The number of GEE articles increased much faster than those concerned with the use of GE. (2) Both GE and GEE were extensively used by the remote sensing community as multidisciplinary tools. GE articles covered a broader range of research areas (e.g., biology, education, disease and health, economic, and information science) and appeared in a broader range of journals than those concerned with the use of GEE. (3) GE and GEE shared similar keywords (e.g., “land cover”, “water”, “model”, “vegetation”, and “forest”), which indicates that their application is of great importance in certain research areas. The main difference was that articles describing the use of GE emphasized its use as a visual display platform, while those concerned with GEE placed more emphasis on big data and time-series analysis. (4) Most applications of GE and GEE were undertaken in countries, such as the United States, China, and the United Kingdom. (5) GEE is an important tool for analysis, whereas GE is used as an auxiliary tool for visualization. Finally, in this paper, the merits and limitations of GE and GEE, and recommendations for further improvements, are summarized from an Earth system science perspective.

InSAR Coherence Analysis for Wetlands in Alberta, Canada Using Time-Series Sentinel-1 Data

Wetlands are valuable natural resources which provide numerous services to the environment. Many studies have demonstrated the potential of various types of remote sensing datasets and techniques for wetland mapping and change analysis. However, there are a relatively low number of studies that have investigated the application of the Interferometric Synthetic Aperture Radar (InSAR) coherence products for wetland studies, especially over large areas. Therefore, in this study, coherence products over the entire province of Alberta, Canada (~661,000 km2) were generated using the Sentinel-1 data acquired from 2017 to 2020. Then, these products along with large amount of wetland reference samples were employed to assess the separability of different wetland types and their trends over time. Overall, our analyses showed that coherence can be considered as an added value feature for wetland classification and monitoring. The Treed Bog and Shallow Open Water classes showed the highest and lowest coherence values, respectively. The Treed Wetland and Open Wetland classes were easily distinguishable. When analyzing the wetland subclasses, it was observed that the Treed Bog and Shallow Open Water classes can be easily discriminated from other subclasses. However, there were overlaps between the signatures of the other wetland subclasses, although there were still some dates where these classes were also distinguishable. The analysis of multi-temporal coherence products also showed that the coherence products generated in spring/fall (e.g., May and October) and summer (e.g., July) seasons had the highest and lowest coherence values, respectively. It was also observed that wetland classes preserved coherence during the leaf-off season (15 August–15 October) while they had relatively lower coherence during the leaf-on season (i.e., 15 May–15 August). Finally, several suggestions for future studies were provided.

Estimation of deracinated trees area in temperate forest with satellite images employing machine learning methods

Climate change can increase the number of uprooted trees. Although there have been an increasing number of machine learning applications for satellite image analysis, the estimation of deracinated tree area by satellite image is not well developed. Therefore, we estimated the deracinated tree area of forests via machine-learning classification using Landsat 8 satellite images. We employed support vector machines (SVMs), random forests (RF), and convolutional neural networks (CNNs) as potential machine learning methods, and tested their performance in estimating the deracinated tree area. We collected satellite images of upright trees, deracinated trees, soil, and others (e.g., waterbodies and cities), and trained them with the training data. We compared the accuracy represented by the correct classification rate of these methods, to determine the deracinated tree area. It was found that the SVM and RF performed better than the CNN for two-class classification (deracinated and upright trees), and the correct classification rates of all methods were up to 93%. We found that the CNN and RF performed significantly higher for the four- and two-class classification compared to the other methods, respectively. We conclude that the CNN is useful for estimating deracinated tree areas using Landsat 8 satellite images.

AlgaeMAp: Algae Bloom Monitoring Application for Inland Waters in Latin America

Due to increasing algae bloom occurrence and water degradation on a global scale, there is a demand for water quality monitoring systems based on remote sensing imagery. This paper describes the scientific, theoretical, and methodological background for creating a cloud-computing interface on Google Earth Engine (GEE) which allows end-users to access algae bloom related products with high spatial (30 m) and temporal (~5 day) resolution. The proposed methodology uses Sentinel-2 images corrected for atmospheric and sun-glint effects to generate an image collection of the Normalized Difference Chlorophyll-a Index (NDCI) for the entire time-series. NDCI is used to estimate both Chl-a concentration, based on a non-linear fitting model, and Trophic State Index (TSI), based on a tree-decision model classification into five classes. Once the Chl-a and TSI algorithms had been calibrated and validated they were implemented in GEE as an Earth Engine App, entitled Algae Bloom Monitoring Application (AlgaeMAp). AlgaeMAp is the first online platform built within the GEE platform that offers high spatial resolution of water quality parameters. The App benefits from the huge processing capability of GEE that allows any user with internet access to easily extract detailed spatial (30 m) and long temporal Chl-a and TSI information (from August 2015 and with images every 5 days) throughout the most important reservoirs in the State of São Paulo/Brazil. The application will be adapted to extend to other relevant areas in Latin America.

A Machine Learning-Based Approach for Surface Soil Moisture Estimations with Google Earth Engine

Due to its relation to the Earth’s climate and weather and phenomena like drought, flooding, or landslides, knowledge of the soil moisture content is valuable to many scientific and professional users. Remote-sensing offers the unique possibility for continuous measurements of this variable. Especially for agriculture, there is a strong demand for high spatial resolution mapping. However, operationally available soil moisture products exist with medium to coarse spatial resolution only (≥1 km). This study introduces a machine learning (ML)—based approach for the high spatial resolution (50 m) mapping of soil moisture based on the integration of Landsat-8 optical and thermal images, Copernicus Sentinel-1 C-Band SAR images, and modelled data, executable in the Google Earth Engine. The novelty of this approach lies in applying an entirely data-driven ML concept for global estimation of the surface soil moisture content. Globally distributed in situ data from the International Soil Moisture Network acted as an input for model training. Based on the independent validation dataset, the resulting overall estimation accuracy, in terms of Root-Mean-Squared-Error and R², was 0.04 m3·m−3 and 0.81, respectively. Beyond the retrieval model itself, this article introduces a framework for collecting training data and a stand-alone Python package for soil moisture mapping. The Google Earth Engine Python API facilitates the execution of data collection and retrieval which is entirely cloud-based. For soil moisture retrieval, it eliminates the requirement to download or preprocess any input datasets.

Sentinel-1 SAR Backscatter Analysis Ready Data Preparation in Google Earth Engine

Sentinel-1 satellites provide temporally dense and high spatial resolution synthetic aperture radar (SAR) imagery. The open data policy and global coverage of Sentinel-1 make it a valuable data source for a wide range of SAR-based applications. In this regard, the Google Earth Engine is a key platform for large area analysis with preprocessed Sentinel-1 backscatter images available within a few days after acquisition. To preserve the information content and user freedom, some preprocessing steps (e.g., speckle filtering) are not applied on the ingested Sentinel-1 imagery as they can vary by application. In this technical note, we present a framework for preparing Sentinel-1 SAR backscatter Analysis-Ready-Data in the Google Earth Engine that combines existing and new Google Earth Engine implementations for additional border noise correction, speckle filtering and radiometric terrain normalization. The proposed framework can be used to generate Sentinel-1 Analysis-Ready-Data suitable for a wide range of land and inland water applications. The Analysis Ready Data preparation framework is implemented in the Google Earth Engine JavaScript and Python APIs.

Automated Global Shallow Water Bathymetry Mapping Using Google Earth Engine

Global shallow water bathymetry maps offer critical information to inform activities such as scientific research, environment protection, and marine transportation. Methods that employ satellite-based bathymetric modeling provide an alternative to conventional shipborne measurements, offering high spatial resolution combined with extensive coverage. We developed an automated bathymetry mapping approach based on the Sentinel-2 surface reflectance dataset in Google Earth Engine. We created a new method for generating a clean-water mosaic and a tailored automatic bathymetric estimation algorithm. We then evaluated the performance of the models at six globally diverse sites (Heron Island, Australia; West Coast of Hawaiʻi Island, Hawaiʻi; Saona Island, Dominican Republic; Punta Cana, Dominican Republic; St. Croix, United States Virgin Islands; and The Grenadines) using 113,520 field bathymetry sampling points. Our approach derived accurate bathymetry maps in shallow waters, with Root Mean Square Error (RMSE) values ranging from 1.2 to 1.9 m. This automatic, efficient, and robust method was applied to map shallow water bathymetry at the global scale, especially in areas which have high biodiversity (i.e., coral reefs).

Classification of Nemoral Forests with Fusion of Multi-Temporal Sentinel-1 and 2 Data

Mapping forest extent and forest cover classification are important for the assessment of forest resources in socio-economic as well as ecological terms. Novel developments in the availability of remotely sensed data, computational resources, and advances in areas of statistical learning have enabled the fusion of multi-sensor data, often yielding superior classification results. Most former studies of nemoral forests fusing multi-sensor and multi-temporal data have been limited in spatial extent and typically to a simple classification of landscapes into major land cover classes. We hypothesize that multi-temporal, multi-sensor data will have a specific strength in the further classification of nemoral forest landscapes owing to the distinct seasonal patterns in the phenology of broadleaves. This study aimed to classify the Danish landscape into forest/non-forest and further into forest types (broadleaved/coniferous) and species groups, using a cloud-based approach based on multi-temporal Sentinel 1 and 2 data and a random forest classifier trained with National Forest Inventory (NFI) data. Mapping of non-forest and forest resulted in producer accuracies of 99% and 90%, respectively. The mapping of forest types (broadleaf and conifer) within the forested area resulted in producer accuracies of 95% for conifer and 96% for broadleaf forest. Tree species groups were classified with producer accuracies ranging 34–74%. Species groups with coniferous species were the least confused, whereas the broadleaf groups, especially Quercus species, had higher error rates. The results are applied in Danish national accounting of greenhouse gas emissions from forests, resource assessment, and assessment of forest biodiversity potentials.

Machine Learning Classification of Mediterranean Forest Habitats in Google Earth Engine Based on Seasonal Sentinel-2 Time-Series and Input Image Composition Optimisation

The sustainable management of natural heritage is presently considered a global strategic issue. Owing to the ever-growing availability of free data and software, remote sensing (RS) techniques have been primarily used to map, analyse, and monitor natural resources for conservation purposes. The need to adopt multi-scale and multi-temporal approaches to detect different phenological aspects of different vegetation types and species has also emerged. The time-series composite image approach allows for capturing much of the spectral variability, but presents some criticalities (e.g., time-consuming research, downloading data, and the required storage space). To overcome these issues, the Google Earth engine (GEE) has been proposed, a free cloud-based computational platform that allows users to access and process remotely sensed data at petabyte scales. The application was tested in a natural protected area in Calabria (South Italy), which is particularly representative of the Mediterranean mountain forest environment. In the research, random forest (RF), support vector machine (SVM), and classification and regression tree (CART) algorithms were used to perform supervised pixel-based classification based on the use of Sentinel-2 images. A process to select the best input image (seasonal composition strategies, statistical operators, band composition, and derived vegetation indices (VIs) information) for classification was implemented. A set of accuracy indicators, including overall accuracy (OA) and multi-class F-score (Fm), were computed to assess the results of the different classifications. GEE proved to be a reliable and powerful tool for the classification process. The best results (OA = 0.88 and Fm = 0.88) were achieved using RF with the summer image composite, adding three VIs (NDVI, EVI, and NBR) to the Sentinel-2 bands. SVM and RF produced OAs of 0.83 and 0.80, respectively.

Mapping Coastal Wetlands of the Bohai Rim at a Spatial Resolution of 10 m Using Multiple Open-Access Satellite Data and Terrain Indices

Coastal wetlands provide essential ecosystem services and are closely related to human welfare. However, they can experience substantial degradation, especially in regions in which there is intense human activity. To control these increasingly severe problems and to develop corresponding management policies in coastal wetlands, it is critical to accurately map coastal wetlands. Although remote sensing is the most efficient way to monitor coastal wetlands at a regional scale, it traditionally involves a large amount of work, high cost, and low spatial resolution when mapping coastal wetlands at a large scale. In this study, we developed a workflow for rapidly mapping coastal wetlands at a 10 m spatial resolution, based on the recently emergent Google Earth Engine platform, using a machine learning algorithm, open-access Synthetic Aperture Radar (SAR) and optical images from the Sentinel satellites, and two terrain indices. We then generated a coastal wetland map of the Bohai Rim (BRCW10) based on the workflow. It has a producer accuracy of 82.7%, according to validation using 150 wetland samples. The BRCW10 data reflected finer information when compared to wetland maps derived from two sets of global high-spatial-resolution land cover data, due to the fusion of multiple data sources. The study highlights the benefits of simultaneously merging SAR and optical remote sensing images when mapping coastal wetlands.

Visual spectrogram scanning paired with an observation–confirmation occupancy model improves the efficiency and accuracy of bioacoustic anuran data

Passive acoustic monitoring using autonomous recording units has improved anuran amphibian call survey data collection. A challenge associated with this approach is the time required for audio data processing. Our objective was to develop a more efficient method of processing and analyzing acoustic data through visual spectrogram scanning and the application of an observation–confirmation occupancy model. We compared detection rates between methods of standard recording listening and visually scanning spectrogram images using different spectrogram parameters. Relative to listening, we found that 1 min spectrograms in two 30 s frames yield the best time efficiency–accuracy trade-off. A standard occupancy model applied to visual scanning data underestimated occupancy estimates relative to listening data for three species and overestimated occupancy for one species. The observation–confirmation model used a subset of listening data to improve the estimates of detection probability from visual scanning and therefore reduced bias in occupancy estimates when compared with using visual scanning data alone. Overall, the combination of the visual scanning method and the observation–confirmation model allowed us to maintain the accuracy of occupancy estimates while greatly increasing the efficiency of anuran data processing. These methods are widely applicable and can increase sample size and precision for acoustic monitoring programs using autonomous recording units.

Probabilistic Mapping and Spatial Pattern Analysis of Grazing Lawns in Southern African Savannahs Using WorldView-3 Imagery and Machine Learning Techniques

Savannah grazing lawns are a key food resource for large herbivores such as blue wildebeest (Connochaetes taurinus), hippopotamus (Hippopotamus amphibius) and white rhino (Ceratotherium simum), and impact herbivore densities, movement and recruitment rates. They also exert a strong influence on fire behaviour including frequency, intensity and spread. Thus, variation in grazing lawn cover can have a profound impact on broader savannah ecosystem dynamics. However, knowledge of their present cover and distribution is limited. Importantly, we lack a robust, broad-scale approach for detecting and monitoring grazing lawns, which is critical to enhancing understanding of the ecology of these vital grassland systems. We selected two sites in the Lower Sabie and Satara regions of Kruger National Park, South Africa with mesic and semiarid conditions, respectively. Using spectral and texture features derived from WorldView-3 imagery, we (i) parameterised and assessed the quality of Random Forest (RF), Support Vector Machines (SVM), Classification and Regression Trees (CART) and Multilayer Perceptron (MLP) models for general discrimination of plant functional types (PFTs) within a sub-area of the Lower Sabie landscape, and (ii) compared model performance for probabilistic mapping of grazing lawns in the broader Lower Sabie and Satara landscapes. Further, we used spatial metrics to analyse spatial patterns in grazing lawn distribution in both landscapes along a gradient of distance from waterbodies. All machine learning models achieved high F-scores (F1) and overall accuracy (OA) scores in general savannah PFTs classification, with RF (F1 = 95.73±0.004%, OA = 94.16±0.004%), SVM (F1 = 95.64±0.002%, OA = 94.02±0.002%) and MLP (F1 = 95.71±0.003%, OA = 94.27±0.003%) forming a cluster of the better performing models and marginally outperforming CART (F1 = 92.74±0.006%, OA = 90.93±0.003%). Grazing lawn detection accuracy followed a similar trend within the Lower Sabie landscape, with RF, SVM, MLP and CART achieving F-scores of 0.89, 0.93, 0.94 and 0.81, respectively. Transferring models to the Satara landscape however resulted in relatively lower but high grazing lawn detection accuracies across models (RF = 0.87, SVM = 0.88, MLP = 0.85 and CART = 0.75). Results from spatial pattern analysis revealed a relatively higher proportion of grazing lawn cover under semiarid savannah conditions (Satara) compared to the mesic savannah landscape (Lower Sabie). Additionally, the results show strong negative correlation between grazing lawn spatial structure (fractional cover, patch size and connectivity) and distance from waterbodies, with larger and contiguous grazing lawn patches occurring in close proximity to waterbodies in both landscapes. The proposed machine learning approach provides a novel and robust workflow for accurate and consistent landscape-scale monitoring of grazing lawns, while our findings and research outputs provide timely information critical for understanding habitat heterogeneity in southern African savannahs.

Multitemporal optical and radar metrics for wetland mapping at national level in Albania

Wetlands are highly dynamic, with many natural and anthropogenic drivers causing seasonal, periodic or permanent changes in their structure and composition. Thus, it is necessary to use time series of images for accurate classifications and monitoring. We used all available Sentinel-1 and Sentinel-2 images to produce a national wetlands map for Albania. We derived different indices and temporal metrics and investigated their impacts and synergies in terms of mapping accuracy. Best results were achieved when combining Sentinel-1 with Sentinel-2 and its derived indices. We reduced systematic errors and increased the thematic resolution using morphometric characteristics and knowledge-based rules, achieving an overall accuracy of 82%. Results were also validated against field inventories. This methodology can be reproducible to other countries and can be made operational for an integrated planning that considers the food, water, and energy nexus.

Application of Image Segmentation in Surface Water Extraction of Freshwater Lakes using Radar Data

Freshwater lakes supply a large amount of inland water resources to sustain local and regional developments. However, some lake systems depend upon great fluctuation in water surface area. Poyang lake, the largest freshwater lake in China, undergoes dramatic seasonal and interannual variations. Timely monitoring of Poyang lake surface provides essential information on variation of water occurrence for its ecosystem conservation. Application of histogram-based image segmentation in radar imagery has been widely used to detect water surface of lakes. Still, it is challenging to select the optimal threshold. Here, we analyze the advantages and disadvantages of a segmentation algorithm, the Otsu Method, from both mathematical and application perspectives. We implement the Otsu Method and provide reusable scripts to automatically select a threshold for surface water extraction using Sentinel-1 synthetic aperture radar (SAR) imagery on Google Earth Engine, a cloud-based platform that accelerates processing of Sentinel-1 data and auto-threshold computation. The optimal thresholds for each January from 2017 to 2020 are − 14.88 , − 16.93 , − 16.96 and − 16.87 respectively, and the overall accuracy achieves 92 % after rectification. Furthermore, our study contributes to the update of temporal and spatial variation of Poyang lake, confirming that its surface water area fluctuated annually and tended to shrink both in the center and boundary of the lake on each January from 2017 to 2020.

Characterizing Spatiotemporal Patterns of Mangrove Forests in Can Gio Biosphere Reserve Using Sentinel-2 Imagery

This study aimed at evaluating the spatiotemporal patterns of mangrove forest variations for three ecological zones of the Can Gio biosphere reserve (i.e., core, buffer, and transition zones) and its relation to land use/land cover changes. Time series Sentinel-2 Imagery—which presents the Normalized Different Vegetation Index (NDVI), obtained through the Google Earth Engine and Overlap Similarity Algorithm—was used to characterize vegetation cover in the study area. Furthermore, the Cohen’s Kappa agreement was applied to examine the accuracy of mangrove classification, and the Mann–Kendal (MK) significance was used to analyze the spatiotemporal trends of mangrove forests. The results showed that an NDVI value greater than 0.3 recorded the reflected signal of mangrove population in the study area with an O-index greater than 0.85. A Cohen’s Kappa statistic of agreement of 0.7 and an overall classification accuracy of 83% was obtained. Regarding the trend in mangrove forest patterns, an increase in area of 669 ha and 579 ha explored at the buffer and core zones, respectively, while the largest declined mangrove area of 350 ha was investigated at the buffer zone, followed by a transition at 314 ha during the study period due to the interconversion of shrimp farming and the expansion of built-up areas. Moreover, the study also described the negative impacts of the sea-encroached urban-tourism zone on mangrove patterns in the foreseeable future. The results from this study will act as a basic fundamental authentic report for local governments in proposing strategies for the shielding of mangrove forests and economic development from negative consequences in foreseeable future.

Meta-Analysis of Wetland Classification Using Remote Sensing: A Systematic Review of a 40-Year Trend in North America

North America is covered in 2.5 million km2 of wetlands, which is the remainder of an estimated 56% of wetlands lost since the 1700s. This loss has resulted in a decrease in important habitat and services of great ecological, economic, and recreational benefits to humankind. To better manage these ecosystems, since the 1970s, wetlands in North America have been classified with increasing regularity using remote sensing technology. Since then, optimal methods for wetland classification by numerous researchers have been examined, assessed, modified, and established. Over the past several decades, a large number of studies have investigated the effects of different remote sensing factors, such as data type, spatial resolution, feature selection, classification methods, and other parameters of interest on wetland classification in North America. However, the results of these studies have not yet been synthesized to determine best practices and to establish avenues for future research. This paper reviews the last 40 years of research and development on North American wetland classification through remote sensing methods. A meta-analysis of 157 relevant articles published since 1980 summarizes trends in 23 parameters, including publication, year, study location, application of specific sensors, and classification methods. This paper also examines is the relationship between several remote sensing parameters (e.g., spatial resolution and type of data) and resulting overall accuracies. Finally, this paper discusses the future of remote sensing of wetlands in North America with regard to upcoming technologies and sensors. Given the increasing importance and vulnerability of wetland ecosystems under the climate change influences, this paper aims to provide a comprehensive review in support of the continued, improved, and novel applications of remote sensing for wetland mapping across North America and to provide a fundamental knowledge base for future studies in this field.

Sentinel-1-Imagery-Based High-Resolution Water Cover Detection on Wetlands, Aided by Google Earth Engine

Saline wetlands experience large temporal fluctuations in water supply during the year and are recharged only or mainly through precipitation, meaning they are vulnerable to climate-change-induced aridification. Most passive satellite sensors are unsuitable for continuous wetland monitoring due to cloud cover and their relatively low temporal resolution. However, active satellite sensors such as the C-band synthetic aperture radar of Sentinel-1 satellites offer free, cloud-independent data. We examined surface water cover changes from October 2014 to November 2018 in the strictly protected area (13,000 ha) of the Upper-Kiskunság Alkaline Lakes region in the Danube–Tisza Interfluve in Hungary, with the aim of helping with nature protection planning. Changes and sensitivity can be defined based on the knowledge of variability. We developed a method for water cover detection based on automatic classification, applying the so-called WEKA K-Means clustering algorithm. For satellite data processing and analysis, we used the Google Earth Engine cloud processing platform. In terms of validation, we compared our results with the multispectral Modified Normalized Difference Water Index (MNDWI) derived from Landsat 8 and Sentinel-2 top-of-atmosphere reflectance images using a threshold-based binary classifier (receiver operator characteristics) for the MNDWI data. Using two completely distinct methods operating in distinct wavelength ranges, we obtained adequately matching results, with Spearman’s correlation coefficients (ρ) ranging from 0.54 to 0.80.

Remote Sensing of Boreal Wetlands 1: Data Use for Policy and Management

Wetlands have and continue to undergo rapid environmental and anthropogenic modification and change to their extent, condition, and therefore, ecosystem services. In this first part of a two-part review, we provide decision-makers with an overview on the use of remote sensing technologies for the ‘wise use of wetlands’, following Ramsar Convention protocols. The objectives of this review are to provide: (1) a synthesis of the history of remote sensing of wetlands, (2) a feasibility study to quantify the accuracy of remotely sensed data products when compared with field data based on 286 comparisons found in the literature from 209 articles, (3) recommendations for best approaches based on case studies, and (4) a decision tree to assist users and policymakers at numerous governmental levels and industrial agencies to identify optimal remote sensing approaches based on needs, feasibility, and cost. We argue that in order for remote sensing approaches to be adopted by wetland scientists, land-use managers, and policymakers, there is a need for greater understanding of the use of remote sensing for wetland inventory, condition, and underlying processes at scales relevant for management and policy decisions. The literature review focuses on boreal wetlands primarily from a Canadian perspective, but the results are broadly applicable to policymakers and wetland scientists globally, providing knowledge on how to best incorporate remotely sensed data into their monitoring and measurement procedures. This is the first review quantifying the accuracy and feasibility of remotely sensed data and data combinations needed for monitoring and assessment. These include, baseline classification for wetland inventory, monitoring through time, and prediction of ecosystem processes from individual wetlands to a national scale.

Salt marsh monitoring along the mid-Atlantic coast by Google Earth Engine enabled time series

Salt marshes provide a bulwark against sea-level rise (SLR), an interface between aquatic and terrestrial habitats, important nursery grounds for many species, a buffer against extreme storm impacts, and vast blue carbon repositories. However, salt marshes are at risk of loss from a variety of stressors such as SLR, nutrient enrichment, sediment deficits, herbivory, and anthropogenic disturbances. Determining the dynamics of salt marsh change with remote sensing requires high temporal resolution due to the spectral variability caused by disturbance, tides, and seasonality. Time series analysis of salt marshes can broaden our understanding of these changing environments. This study analyzed aboveground green biomass (AGB) in seven mid-Atlantic Hydrological Unit Code 8 (HUC-8) watersheds. The study revealed that the Eastern Lower Delmarva watershed had the highest average loss and the largest net reduction in salt marsh AGB from 1999–2018. The study developed a method that used Google Earth Engine (GEE) enabled time series of the Landsat archive for regional analysis of salt marsh change and identified at-risk watersheds and salt marshes providing insight into the resilience and management of these ecosystems. The time series were filtered by cloud cover and the Tidal Marsh Inundation Index (TMII). The combination of GEE enabled Landsat time series, and TMII filtering demonstrated a promising method for historic assessment and continued monitoring of salt marsh dynamics.

Estimating Coarse Woody Debris Volume Using Image Analysis and Multispectral LiDAR

Coarse woody debris (CWD, parts of dead trees) is an important factor in forest management, given its roles in promoting local biodiversity and unique microhabitats, as well as providing carbon storage and fire fuel. However, parties interested in monitoring CWD abundance lack accurate methods to measure CWD accurately and extensively. Here, we demonstrate a novel strategy for mapping CWD volume (m3) across a 4300-hectare study area in the boreal forest of Alberta, Canada using optical imagery and an infra-canopy vegetation-index layer derived from multispectral aerial LiDAR. Our models predicted CWD volume with a coefficient of determination (R2) value of 0.62 compared to field data, and a root-mean square error (RMSE) of 0.224 m3/100 m2. Models using multispectral LiDAR data in addition to image-analysis data performed with up to 12% lower RMSE than models using exclusively image-analysis layers. Site managers and researchers requiring reliable and comprehensive maps of CWD volume may benefit from the presented workflow, which aims to streamline the process of CWD measurement. As multispectral LiDAR radiometric calibration routines are developed and standardized, we expect future studies to benefit increasingly more from such products for CWD detection underneath canopy cover.

Estimating Forest Stock Volume in Hunan Province, China, by Integrating In Situ Plot Data, Sentinel-2 Images, and Linear and Machine Learning Regression Models

The forest stock volume (FSV) is one of the key indicators in forestry resource assessments on local, regional, and national scales. To date, scaling up in situ plot-scale measurements across landscapes is still a great challenge in the estimation of FSVs. In this study, Sentinel-2 imagery, the Google Earth Engine (GEE) cloud computing platform, three base station joint differential positioning technology (TBSJDPT), and three algorithms were used to build an FSV model for forests located in Hunan Province, southern China. The GEE cloud computing platform was used to extract the imagery variables from the Sentinel-2 imagery pixels. The TBSJDPT was put forward and used to provide high-precision positions of the sample plot data. The random forests (RF), support vector regression (SVR), and multiple linear regression (MLR) algorithms were used to estimate the FSV. For each pixel, 24 variables were extracted from the Sentinel-2 images taken in 2017 and 2018. The RF model performed the best in both the training phase (i.e., R2 = 0.91, RMSE = 35.13 m3 ha−1, n = 321) and in the test phase (i.e., R2 = 0.58, RMSE = 65.03 m3 ha−1, and n = 138). This model was followed by the SVR model (R2 = 0.54, RMSE = 65.60 m3 ha−1, n = 321 in training; R2 = 0.54, RMSE = 66.00 m3 ha−1, n = 138 in testing), which was slightly better than the MLR model (R2 = 0.38, RMSE = 75.74 m3 ha−1, and n = 321 in training; R2 = 0.49, RMSE = 70.22 m3 ha−1, and n = 138 in testing) in both the training phase and test phase. The best predictive band was Red-Edge 1 (B5), which performed well both in the machine learning methods and in the MLR method. The Blue band (B2), Green band (B3), Red band (B4), SWIR2 band (B12), and vegetation indices (TCW, NDVI_B5, and TCB) were used in the machine learning models, and only one vegetation index (MSI) was used in the MLR model. We mapped the FSV distribution in Hunan Province (3.50 × 108 m3) based on the RF model; it reached a total accuracy of 63.87% compared with the official forest report in 2017 (5.48 × 108 m3). The results from this study will help develop and improve satellite-based methods to estimate FSVs on local, regional and national scales.

Comparing Deep Learning and Shallow Learning for Large-Scale Wetland Classification in Alberta, Canada

Advances in machine learning have changed many fields of study and it has also drawn attention in a variety of remote sensing applications. In particular, deep convolutional neural networks (CNNs) have proven very useful in fields such as image recognition; however, the use of CNNs in large-scale remote sensing landcover classifications still needs further investigation. We set out to test CNN-based landcover classification against a more conventional XGBoost shallow learning algorithm for mapping a notoriously difficult group of landcover classes, wetland class as defined by the Canadian Wetland Classification System. We developed two wetland inventory style products for a large (397,958 km2) area in the Boreal Forest region of Alberta, Canada, using Sentinel-1, Sentinel-2, and ALOS DEM data acquired in Google Earth Engine. We then tested the accuracy of these two products against three validation data sets (two photo-interpreted and one field). The CNN-generated wetland product proved to be more accurate than the shallow learning XGBoost wetland product by 5%. The overall accuracy of the CNN product was 80.2% with a mean F1-score of 0.58. We believe that CNNs are better able to capture natural complexities within wetland classes, and thus may be very useful for complex landcover classifications. Overall, this CNN framework shows great promise for generating large-scale wetland inventory data and may prove useful for other landcover mapping applications.