RADARSAT-2 Imagery Research Articles

Monitoring Crop Condition at Gield Scales and at a Daily Time Step Using Synthetic Aperture Radar (SAR)

Operations that rely on optically-based Normalized Difference Vegetation Index (NDVI) to track crop productivity, are negatively impacted by cloud cover. In this research, parameters from Synthetic Aperture Radar (SAR) imagery were calibrated to NDVI with the targeted outcome to integrate SAR into Canadian crop monitoring services. Data from fully polarimetric RADARSAT-2 and dual-polarization Sentinel-1B imagery were calibrated to Sentinel-2 NDVI. Random Forest Regression (RFR) and Least Squares Boosting (LSBoost) were tested to calibrate SAR to NDVI (SARcal-NDVI) for six crops (corn, canola, soybeans, wheat, oats and barley). RADARSAT-2 provided the best fit with R2 exceeding 0.84 for canola, 0.87 for wheat and 0.90 for corn and soybeans. Correlations were lower for oats (R2 of 0.72–0.77) and barley (R2 of 0.43–0.64) due to a limited number of fields. Although a global model yielded lower correlations (R2 of 0.80), this model would be easier to implement into operations. Total power, the first and second eigenvalues and VH backscatter were important in reducing model error. The SARcal-NDVI and Growing Degree Days were then integrated into a Crop Structure Dynamic Model to produce daily estimates of crop condition, at field scales. The next step is to test if these daily estimates of crop condition can improve crop yield forecasting for the Canadian agriculture sector.

IceGCN: An Interactive Sea Ice Classification Pipeline for SAR Imagery Based on Graph Convolutional Network

Monitoring sea ice in the Arctic region is crucial for polar maritime activities. The Canadian Ice Service (CIS) wants to augment its manual interpretation with machine learning-based approaches due to the increasing data volume received from newly launched synthetic aperture radar (SAR) satellites. However, fully supervised machine learning models require large training datasets, which are usually limited in the sea ice classification field. To address this issue, we propose a semi-supervised interactive system to classify sea ice in dual-pol RADARSAT-2 imagery using limited training samples. First, the SAR image is oversegmented into homogeneous regions. Then, a graph is constructed based on the segmentation results, and the feature set of each node is characterized by a convolutional neural network. Finally, a graph convolutional network (GCN) is employed to classify the whole graph using limited labeled nodes automatically. The proposed method is evaluated on a published dataset. Compared with referenced algorithms, this new method outperforms in both qualitative and quantitative aspects.

High-Resolution Detection of Periglacial Landforms Deformation Using Radarsat-2 and GF-7 Stereo Optical Imagery

High-Resolution Detection of Periglacial Landforms Deformation Using Radarsat-2 and GF-7 Stereo Optical Imagery

Surface roughness from in-situ measurements around Indian Antarctic stations

Surface roughness is an important parameter in deriving energy balance over the polar ice-sheets and glaciers. In Antarctica, roughness appears as snow patches and is affected by wind transport. It is also influenced by snowfall, snowdrifts, snowmelt, and snow grain size. In this paper, we derive surface roughness using a laser distometer. The roughness measurements were collected during the 32nd Indian Scientific Expedition to Antarctica 2013 (ISEA-32). After removal of all the possible error sources, it is found that around the Indian Antarctic station Bharati (69.41°S, 76.18°E), roughness is in the range of 0.7–1.2 cm. Snowfall and wind introduce fluctuations in surface roughness measurements. Measurements around Maitri (70.76°S, 11.73°E) showed that different ice surfaces have different values of roughness (0.6–1.4 cm), with the minimum roughness in the interior ice sheet surrounded by hummocks. Sastrugi patterns were also captured in the analysis. Backscattering signatures from RADARSAT-2 imagery were examined in conjunction with the roughness measurements.

Sea Ice–Water Classification of RADARSAT-2 Imagery Based on Residual Neural Networks (ResNet) with Regional Pooling

Sea ice mapping plays an integral role in ship navigation and meteorological modeling in the polar regions. Numerous published studies in sea ice classification using synthetic aperture radar (SAR) have reported high classification rates. However, many of these focus on numerical results based on sample points and ignore the quality of the inferred sea ice maps. We have designed and implemented a novel SAR sea ice classification algorithm where the spatial context, obtained by the unsupervised IRGS segmentation algorithm, is integrated with texture features extracted by a residual neural network (ResNet) and, using regional pooling, classifies ice and water. This algorithm is trained and tested on a published dataset and cross-validated using leave-one-out (LOO) strategy, obtaining an overall accuracy of 99.67% and outperforming several existing algorithms. In addition, visual results show that this new method produces sea ice maps with natural ice–water boundaries and fewer ice and water errors.

Biophysical parameter estimation of crops from polarimetric synthetic aperture radar imagery with data-driven polynomial chaos expansion and global sensitivity analysis

• Polynomial Chaos Expansion regression is suitable for estimation of crop biophysical parameters. • Correlated input vectors prevent reliable Global Sensitivity Analysis with Sobol’s method. • Relationship between polynomial coefficients and Sobol sensitivity indices is an advantage. • Interactions between features affect regression of biophysical parameters in different level. Data-driven machine learning regression methods are easy to implement and applicable to a wide-range of data in biophysical parameter estimation and have become a common approach in the remote sensing field. Among the regression methods, polynomial chaos expansion (PCE) is one of the reliable and interesting ones due to its tight relationship with uncertainty quantification. One of the advantages of PCE is that global sensitivity analysis (GSA) with Sobol’s method can be analytically computed from polynomial coefficients if the input space is statistically independent. However, most of the phenomena include dependent features either statistically or physically. Though the physical independence is provided between inputs, they must be statistically uncorrelated. Therefore, an independent and uncorrelated input space must be created before the regression analysis. In this paper, we performed PCE-based regression analysis for the estimation of biophysical parameters of crops. The study was conducted in the experimental fields of field pea, barley, canola, and oat of the AgriSAR2009 campaign. The input parameters of the regression model were formed by creating polarimetric features derived from RADARSAT-2 imagery. The estimated biophysical parameters were based on the discrete in situ measurements of leaf area index (LAI) and normalized difference vegetation index (NDVI), scattered semi-randomly in each crop field. We implemented neighbourhood component analysis (NCA) to create an independent and uncorrelated input space by eliminating correlations. Finally, we investigated the importance of features, which drive the PCE-based regression models applying GSA with Sobol’s method. Besides the individual effects of each feature, their interactions were found significant.

Sea-Ice Mapping of RADARSAT-2 Imagery by Integrating Spatial Contexture With Textural Features

Mapping different types of sea ice that form, grow, and melt in polar oceans is essential for shipping navigation, climate change modeling, and local community safety. Currently, ice charts are manually generated by analysts at the Canadian Ice Service (CIS) based on dual-polarized RADARSAT-2/RADARSAT Constellation Mission (RCM) imagery on a daily basis. Inspired by the demand for a computer-based mapping system, we have developed an automatic sea ice classification method that integrates spatial contexture (unsupervised segmentation) with textural features (supervised pixel-level labeling). First, the full-scene image is oversegmented, and the segments are merged into homogeneous regions across the entire scene. Second, pixel-based classifiers (support vector machine, random forest) are compared for their ability to label the generated homogeneous regions. Finally, the segmentation and labeling are combined using a proposed energy function. The proposed method was tested on 18 dual-polarization RADARSAT-2 scenes acquired over the Beaufort Sea. This dataset contains water, young ice, first-year ice, and multi-year ice covering melt, summer, and freeze-up seasons. The proposed method obtains an average classification accuracy of 86.33% based on the leave-one-out validation. The experimental results show that the proposed method achieves promising classification results in both quantity and quality measurements compared to benchmark methods. The robustness against incidence angle variance indicates that the proposed method is well-qualified for operational sea ice mapping.

Prediction of crop biophysical variables with panel data techniques and radar remote sensing imagery

Since the late 1970s, remote sensing techniques have been proven to be suitable for characterizing and monitoring plants and crops. In particular, synthetic aperture radar (SAR) missions contribute considerably to this prediction effort. However, the main issue when using SAR image series together with field observations is the scarcity of data due to the difficulty of acquiring field measurements. This research aimed to contribute to solving this problem with an alternative statistical model that can overcome the lack of a long, robust series of field-based ground truth observations. The main novelty of this research is the evaluation of the potential of a panel data approach to radar remote sensing imagery for predicting crop biophysical variables. For this purpose, RADARSAT-2 imagery was acquired over the study area in central Spain. Simultaneously, a field campaign was deployed to estimate crop parameters in the same area and to validate the results of the modelling. The analysis of the influence of the crop type on the incidence angle and the polarimetric parameters showed a strong influence of the co-polar channels (HH, VV), the entropy (H) and the coherence between the co-polar channels (γHHVV), with the differences being higher at 25°. The panel data analysis method demonstrated that good predictions, with R2 greater than 0.78, were achieved for all biophysical variables analysed in this study. Overall, this novel statistical approach with remote sensing data showed great applicability for the prediction of crop variables, even with a short series of observations.

Spring melt pond fraction in the Canadian Arctic Archipelago predicted from RADARSAT-2

Abstract. Melt ponds form on the surface of Arctic sea ice during spring, influencing how much solar radiation is absorbed into the sea ice–ocean system, which in turn impacts the ablation of sea ice during the melt season. Accordingly, melt pond fraction (fp) has been shown to be a useful predictor of sea ice area during the summer months. Sea ice dynamic and thermodynamic processes operating within the narrow channels and inlets of the Canadian Arctic Archipelago (CAA) during the summer months are difficult for model simulations to accurately resolve. Additional information on fp variability in advance of the melt season within the CAA could help constrain model simulations and/or provide useful information in advance of the shipping season. Here, we use RADARSAT-2 imagery to predict and analyze peak melt pond fraction (fpk) and evaluate its utility to provide predictive information with respect to sea ice area during the melt season within the CAA from 2009–2018. The temporal variability of RADARSAT-2 fpk over the 10-year record was found to be strongly linked to the variability of mean April multi-year ice area with a statistically significant detrended correlation (R) of R=-0.89. The spatial distribution of RADARSAT-2 fpk was found to be in excellent agreement with the sea ice stage of development prior to the melt season. RADARSAT-2 fpk values were in good agreement with fpk observed from in situ observations but were found to be ∼ 0.05 larger compared to MODIS fpk observations. Dynamically stable sea ice regions within the CAA exhibited higher detrended correlations between RADARSAT-2 fpk and summer sea ice area. Our results show that RADARSAT-2 fpk can be used to provide predictive information about summer sea ice area for a key shipping region of the Northwest Passage.

RADARSAT-2 Derived Glacier Velocities and Dynamic Discharge Estimates for the Canadian High Arctic: 2015–2020

RADARSAT-2 imagery collected each winter from 2015/2016 to 2019/2020 is used to quantify and characterize the variability in the motion of, and the discharge from, the major marine-terminating ice masses of the Queen Elizabeth Islands (QEI: Devon, Ellesmere and Axel Heiberg Islands) in the Canadian High Arctic. The majority of the glaciers did not experience significant variations in flow speeds over the observation period, and for most that did the variations are attributed to pulse and surge processes. However, there are exceptions where the velocity record indicates continued acceleration of the glaciers by processes that appear distinct from surging or pulsing, such as dynamic thinning. These include Trinity and Wykeham glaciers (Prince of Wales Icefield) and Belcher Glacier (Devon Ice Cap). The combination of surface velocities with ice thicknesses indicates that average ice discharge to the ocean for the QEI over the observation period was 2.78 ± 0.52 Gt a−1 (ranging between ∼2.37 ± 0.48 Gt a−1 and ∼3.20 ± 0.55 Gt a−1), ∼50% of which was channeled through the Trinity-Wykeham glacier basin alone. The results presented here, combined with those of previous studies, provide a comprehensive record of ice motion and discharge from the QEI between 2008 and 2020.

Automated surface water detection from space: a Canada-wide, open-source, automated, near-real time solution

The goal of this research was to develop a fully automated method to map open water extent that is operationally practical on a national scale. Such a system needs to produce acceptable results in all regions of the country and particularly in the Prairie Potholes Region where understanding water surface dynamics is important for predicting flooding, agriculture/water availability and for evaporation calculations in weather models. A system was developed to automate, ingest, and process Radarsat-2 (RS2) imagery, from which mapping open water body extents in near-real time was carried out using a machine learning classification technique. A Random Forest classification algorithm was trained using the data extracted from the Global Surface Water (GSW) occurrence dataset. The GSW occurrence thresholds used to extract the training data were examined and there was little influence of uncertainty on the classification. The quality of classifications generated from RS2 Fine Wide mode imagery improved with increasing incidence angle. All Fine Quad incident angles produced acceptable results, but Standard and Wide mode imagery produced results below the accuracy thresholds deemed acceptable for this operational solution. Validation was carried out by comparing mapped water extents to temporally coincident high resolution multi-spectral imagery and to the USGS Global Land Cover Characteristics dataset, that is currently used as a land-water mask by ECCC in weather numerical weather modelling. The system that has been developed will allow new image datasets (e.g. Radarsat Constellation Mission) or training data that becomes available to be used to improve models. The open source code will be made available on Github.

Investigating the Potential Use of RADARSAT-2 and UAS imagery for Monitoring the Restoration of Peatlands

The restoration of peatlands is critical to help reduce the effects of climate change and further prevent the loss of habitat for many species of flora and fauna. The objective of this research was to evaluate RADARSAT-2 satellite imagery and high-resolution Unmanned Aerial Systems (UASs) to determine if they could be used as surrogates for monitoring the success of peatland restoration. Areas of peatland that were being actively harvested, had been restored from past years (1994–2003), and natural shrub bog in Lac St. Jean, Quebec were used as a test case. We compared the Freeman–Durden and Touzi decompositions by applying the Bhattacharyya Distance (BD) statistic to see if the spectral signatures of restored peatland could be separated from harvested peat and natural shrub bog. We flew Unmanned Aerial Surveys (UASs) over the study site to identify Sphagnum and Polytrichum strictum, two indicator species of early peatland restoration success. Results showed that the Touzi decomposition was better able to separate the spectral signatures of harvested, restored, and natural shrub bog (BD values closer to 9). Symmetric scattering type αs1, Helicity |τ1,2,3|, a steep incidence angle, and peak growing season appear to be important for separating the spectral signatures. We had moderate success in detecting Sphagnum and Polytrichum strictum visually by using texture and pattern but were unable to use colour due to differences in sun angle and clouds during the UAS flights. Results suggest that RADARSAT-2 data using the Touzi decomposition and UAS imagery show potential for monitoring peatland restoration success over time.

Mineral Oil Slicks Identification Using Dual Co-polarized Radarsat-2 and TerraSAR-X SAR Imagery

This study is devoted to a generalization of C-band Radarsat-2 and X-band TerraSAR-X synthetic aperture radar (SAR) data in the form of a diagram serving to easily identify mineral oil slicks (crude oil and emulsions) and separate them from the other oil slicks. The diagram is based on the multi-polarization parameter called Resonant to Non-resonant signal Damping (RND) introduced by Ivonin et al. in 2016, which is related to the ratio between damping within the slick of the short waves and wave breakings. SAR images acquired in the North Sea during oil-on-water exercises in 2011–2012 containing three types of oil spills (crude oil, emulsion, and plant oil) were used. The analysis was performed under moderate sea conditions (wind speeds of 2–6 m/s and sea wave heights of less than 2 m), the incidence angles of 27°–49°, and the signal-to-noise ratio (SNR) of −3 to 11 dB within slicks. On the diagram plane, created by the RND parameter and the Bragg wave number, the mineral oil samples form a well-outlined zone, called a mineral oil zone. For C-band data, the plant oil samples were clearly distinguished from the mineral oils in the diagram. Determination of the confidence level for the detection of mineral oils versus plant oil was proposed using the mineral oil zone boundaries. The mineral oil data with SNR within slicks better than 2 dB lay within this zone with a confidence level better than 65%. The plant oil data with the same SNR lay outside this zone with a confidence level of better than 80%. For mineral oil with SNR of −3 dB, the confidence level is 55%.

Detection of sea-surface wind patterns associated with convective systems and their evolution using a combination of Sentinel-1 and Radarsat-2 imagery

ABSTRACTSynthetic Aperture Radar (SAR) can detect wind patterns induced by the Convective Systems (CS) downdraughts hitting the sea surface due to their high spatial resolution and large swath. Additionally, the combination of two successive SAR images (Sentinel-1 and Radarsat-2) with appropriate observation time delays may enable the observation of wind pattern evolution. The corresponding METEOSAT-10/11 infrared images to the SAR ones are used for the detection of deep convective clouds that are potentially associated with wind patterns. The current paper investigates two cases of time delays (1-minute and 5-minute) between two satellites. The estimated wind speeds by CMOD5.N on Sentinel-1 and Radarsat-2 images are almost the same for the background wind areas. However, the intensity of wind patterns on Radarsat-2 images (acquired 1 and 5 minutes later) is decreased, possibly due to a change of downdraught intensity according to observation time. In the 1-minute delay case, the drop of wind intensity is mainly observed for the intense spots (20–25 m s−1), while in the 5-minute delay case, it is noticed for both intense and moderate (10–15 m s−1) convection spots. The intensity of the high-wind convection spots decreases more quickly since the associated convective-scale downdraughts reach the sea surface more rapidly.

The 2014 Mw 6.1 Ludian Earthquake: The Application of RADARSAT-2 SAR Interferometry and GPS for this Conjugated Ruptured Event

Although the Zhaotong–Ludian fault is a seismically active zone located in the boundary between the Sichuan–Yunnan block and the South China block, it has not experienced a large earthquake greater than Mw 7 since at least 1700. On 3 August, 2014, an Mw 6.1 earthquake (the Ludian earthquake) ruptured the Zhaotong active belt in Ludian County, Yunnan province, China. This earthquake was the largest earthquake recorded in the region since 2000, and it provides us with a unique opportunity to study the active tectonics in the region. The analysis of the aftershocks showed that two conjugate faults could have been involved in the event. We first used Global Positioning System (GPS) data and C-band RADARSAT-2 imagery to map the coseismic surface deformation. We then inverted the derived coseismic deformation for the slip distribution based on the constructed conjugate fault model. Finally, the coulomb failure stress due to the Ludian earthquake was estimated to investigate the potential seismic hazards in this region. Our investigations showed that the Ludian earthquake was mainly a bilateral rupture event. The major slip of the main shock was located at depths of 0–5 km, which is close but does not superpose with the aftershocks that are mostly located at depths of 5–20 km. Interestingly, the seismic moment released by the aftershocks (6.9 × 1018 N∙m) was greater than that of the main shock (2.6 × 1018 N∙m). This evidence suggests that the accumulated elastic strain at depths of 0–20 km could have been fully released by the Ludian earthquake and its subsequent aftershocks. Furthermore, our analysis of the coulomb failure stress changes due to the main shock showed that the aftershocks could be the result of dynamic triggering rather than static triggering.

Classification of oil spill by thicknesses using multiple remote sensors

Satellite Synthetic Aperture Radar (SAR) is an operational tool for monitoring and assessment of oil spills. Satellite SAR has primarily been used to detect the presence/absence of oil, yet its ability to discriminate oil emulsions within a detected oil slick has not been fully exploited. Additionally, one of the challenges in the past has been the ability to deliver strategic information derived from satellite remote sensing in a timely fashion to responders in the field. This study presents methods for the rapid classification of oil types and estimated thicknesses, from which information about thick oil and oil emulsions (i.e., “actionable” oil) can be delivered in an operational timeframe to responders in the field. Experiments carried out at the OHMSETT test facility in New Jersey demonstrate that under specific viewing conditions, a single polarization satellite SAR image can record a signal variance between thick stable emulsions and non-emulsified oil. During a series of field campaigns in the Gulf of Mexico with in situ measurements of oil thickness, multiple satellite data were obtained including fully polarimetric C-band SAR imagery from RADARSAT-2 and multispectral imagery from ASTER and WorldView-2. One campaign included the airborne polarimetric UAVSAR L-band sensor. An oil/emulsion thickness classification product was generated based on RADARSAT-2 polarimetric imagery using entropy and the damping ratio derivations. Herein, we present the classification methods to generate oil thickness products from SAR, validated by sea-truth observations, the multispectral imagery, and the UAVSAR data. We tested the ability to deliver these products with minimum latency to responding vessels via NOAA. During field operations in the Gulf of Mexico, a satellite SAR-based product of oil delineation by relative thickness was delivered to a responding vessel 42 min after the RADARSAT-2 data acquisition. This proof-of-concept test using satellite SAR and multispectral imagery to detect emulsions and deliver a derived information product to a vessel in near-real-time points directly to methods for satellite-based assets to be used in the near future for oil spill tactical response operations.

Estimation of Wind Direction in Tropical Cyclones Using C-Band Dual-Polarization Synthetic Aperture Radar

Under extreme weather conditions, the imprints of kilometer-scale marine atmospheric boundary layer roll vortices on the ocean surface are clearly visible in synthetic aperture radar (SAR) images of storms. Therefore, information about wind direction in storms can be obtained by analyzing SAR image features caused by boundary layer rolls. VH-polarized SAR imagery captures the structural features of storms well and shows prominent image gradients along the radial directions of the storm. The signal-to-noise ratios of VH-polarized images are small in low wind speed areas, but they are large in the same regions of VV-polarized images. Also, the capability of retrieving the atmospheric rolls orientation in VV-polarization is found to be sensitive to incidence angle, with better performances for larger incidence angles. Thus, there is the potential to retrieve the storm's wind directions using a combination of the VH- and VV-polarized SAR observations. In this article, we use the local gradient method to estimate tropical cyclone (TC) wind directions from C-band RADARSAT-2 and Sentinel-1A dual-polarization (VV + VH) SAR imagery. As a case study, wind directions with a spatial resolution of 25 km are derived by using both wide-swath VV- and VH-polarized SAR imagery over two hurricanes (Earl and Bertha) and one Typhoon (Meranti). We compare wind directions derived from ten dual-polarization SAR images with collocated wind directions from buoys, Global Positioning System (GPS) dropsondes, scatterometer, and radiometer. Statistical comparisons show that the wind direction bias and root-mean-square error are, respectively, -0.54° and 14.78° for VV-polarization, 0.38° and 14.25° for VH-polarization, 0.20° and 13.30° for VV- and VH-polarization, suggesting dual-polarization SAR is more suitable for the estimation of TC wind directions than VV- or VH-polarization SAR.

A review: monitoring of rice production by using applications of remote sensing

Rice is the most significant main food for above three billion people all over the world. Rice paddy lands numerated almost 11.5% of the World’s cultivable field region. Thus, using rice region mapping and forecasting is so essential for food security, where requests mostly invade production. One of the monitoring applications is remote sensing, which is the science of hiving information regarding the earth’s surface without physical contact, is becoming progressively significant in majority of strands these days. Newly, rice monitoring applications are detecting by various methods via remote sensing. Surveys about satellite images have displayed that rice lands crop several brightness forms at various plant development phases, permitting lands to be categorized. Howsoever, there is a demand for major survey into realization the interplays amongst electromagnetic waves and rice lands via the expansion of a methodical sample. This study explains extension theoretical patterns that able to inspect with field measurements to provide accurate explanation of remote sensing data and rice development monitoring by using the revolve detection manner. Not a lot is known concerning how electromagnetic waves interact rice lands and how rice crops distribute waves back to the satellite. Although rice lands able to be divided using the backscattering extent and also the rice development can be monitored, a plenary research of the theoretical scattering method is desperate to avouch accurate application of remote sensing data. This paper begins to peruse the communication amongst microwave backscatter signatures and rice development and to monitor rice development by using satellite images. At the end, it is used to results by using multi-temporal RADARSAT imagery have authenticated that the backscatter can make a good segregation for rice planting phases.

Synergistic RADARSAT-2 and Sentinel-1 SAR Images for Ocean Feature Analysis

Using a case study approach, the utility of synergistic RADARSAT-2 (R2) and Sentinel-1 (S1) synthetic aperture radar (SAR) imagery is demonstrated for ocean feature signature analysis in the vicinity of the Gulf Stream. The R2 and S1 images considered are either spatially adjacent or spatially overlapping, and were quasi-simultaneously collected (i.e., within minutes of each other). Spatially adjacent R2 and S1 imagery allows ocean feature signatures to be delineated over large spatial areas, while spatially overlapping R2 and S1 imagery collected within short time intervals provides independent ‘looks’ at the same ocean features. This permits determination of the surface displacement of features, potentially leading to improved classification of the origin of ocean feature signatures (quasi-stationary features are likely related to sea surface temperature fronts, while mobile features are likely related to atmospheric conditions). Further, we demonstrate how the use of S1 Level-2 products (i.e. radial velocity datasets) can be leveraged as contextual data to improve the interpretation and classification of ocean feature signatures extracted from R2 imagery. Despite the straight-forward approach taken here, this work demonstrates that there are practical, real-world applications that would benefit from exploiting these on-going imaging opportunities in operational environments.

Separability analysis of wetlands in Canada using multi-source SAR data

Accurately classifying and monitoring wetlands using new technologies is important because of many services that wetlands provide to the environment. In this regard, Synthetic Aperture Radar (SAR) systems provide valuable data to separate different wetland classes. Using large amount of field samples collected over three years, 78 SAR features extracted from multi-source satellites were investigated to select the most important features and decomposition methods for discriminating five wetland classes: Bog, Fen, Marsh, Swamp, and Shallow Water. The results indicated that the ratio features obtained from the diagonal elements of the covariance matrix (extracted from full polarimetric data RADARSAT-2 imagery) and the intensity layers of the dual polarimetric data (i.e., the data acquired by Sentinel-1 and ALOS-2) were most useful for distinguishing wetland class pairs as well as all wetland classes. In this regard, the ratio of HH and HV channels had the highest potential especially for discriminating herbaceous (Bog, Fen, Marsh) and woody (e.g., Swamp) wetlands. Moreover, the features derived from eigenvalues of the coherency matrix (e.g., Anisotropy, serd, normalized serd, and normalized derd) were among the most optimum features for wetland classification. Regarding the decomposition techniques, the H/A/Alpha and Freeman-Durden methods were selected as the best to discriminate wetlands. In terms of scattering mechanisms, it was observed that the volume component was generally the most useful element to discriminate wetland classes compared to the two other components (i.e., single- and double-bounce). This study comprehensively discusses the efficiency of various SAR features/decomposition methods for wetland studies and the results are expected to help with creating sustainable policies and management for wetland protection and monitoring using remote sensing methods.