Natural seepage is a significant contributor to marine hydrocarbon inputs. Remote and intermittent seeps are difficult to monitor in the field, yet surface oil slicks can be observed by spaceborne synthetic aperture radar (SAR) because they reduce backscatter, creating potential for automatic mapping. In mapping tasks like segmentation, deep learning models excel, albeit needing large amounts of labeled images. To deal with the scarcity of labeled images, transfer learning is an approach which makes use of knowledge from related domains. In the case of oil slicks, differences between Sentinel-1 acquisition modes, such as the interferometric wide (IW) in the North Sea and extra wide (EW) in the Arctic, complicate direct model transfer. Here, we present a use-case where transfer learning enhances the few-shot segmentation of natural oil slicks. We used labeled slicks in IW images in the North Sea to pretrain a series of DeepLabv3 and SAM models. These models were then fine-tuned on EW-labeled slicks from two documented Arctic seeps on which we have only limited observations. Our results show clear evidence that transfer learning improves segmentation, notably in challenging and noisy images. Overall, few studies have addressed transfer learning between SAR acquisition modes. This work contributes to improved monitoring of poorly understood or yet undiscovered hydrocarbon seeps.

We compare in-situ water volume measurements of mineral oil emulsion sampled from an oil slick in Santa Barbara, California, to acquisitions of airborne UAVSAR data acquired in June 2022. Estimating the water-to-oil fraction using the UAVSAR imagery, we find that low SNR in the co- and cross-polarimetric channels limits this capability above a certain oil-to-water volumetric threshold. Higher SNR regions of the slick had water volume fractions below 20%, while lower SNR regions had water volume fractions above 20%. Calculated damping ratio values align with the noise analysis, indicating that a lower SNR corresponds to higher damping values, while a higher SNR corresponds to lower damping ratio values. For the high SNR case, water fractions calculated using the co-polarimetric ratio (VV/HH) and a theoretical backscattering model were slightly underestimated when compared with in-situ measurements. This observation could be due to potential sampling bias during the collection of in-situ samples, favoring thicker oil with a higher water cut.

The damping ratio is a calculated feature that measures the contrast between oil-slicked water and the open ocean in SAR data. To implement the damping ratio, the current literature suggests estimating the open water backscatter by taking strips of undefined width across the range direction, obtaining the damping ratio as a function of incidence angle. We show in this paper that the method proposed in the literature can be improved by instead sampling open water pixels randomly. The method is tested on RADARSAT-2 quad-polarimetric SAR imagery of a verified oil slick acquired during the 2013 NOFO oil-on-water exercise conducted in the North Sea. The results suggest that deviations in the derived damping ratio encountered by implementing the method proposed in the literature can be reduced from of order 100– 10−1to 10−3.

In September 2019, the German research icebreaker Polarstern started the largest multidisciplinary Arctic expedition, the MOSAiC (Multidisciplinary drifting Observatory for the Study of Arctic Climate) drift experiment. Being moored to ice floes at high Arctic for a whole year, thus including the winter season, the main goal of the expedition is to better understand and quantify relevant processes within the atmosphere-ice-ocean system that impact the sea ice, ultimately leading to improved climate models. Satellite remote sensing, specially multi-frequency synthetic aperture radar (SAR) plays a major role to achieve this goal. Two major objectives in SAR based remote sensing of sea ice is on the one hand to have a large coverage, and on the other hand to obtain a radar response that carries as much information as possible. A comprehensive set of C- and L- band SAR images were acquired during the course of MOSAiC. In this initial study we evaluate the effects of seasonal changes on C- and L-band backscatter in respect to three different sea ice types, i.e., Young Ice, Smooth Ice and Rough/Deformed Ice along with the performance of sea ice type retrieval of a established algorithm. Areas of deformed, smooth and young sea ice were observed in the vicinity of R/V Polarstern and these areas are included whenever possible in the yearlong time series. For both frequencies a change in all backscatter channels values can be observed during the early melt season. This is first noticeable in the C-band images and later followed by a change in the L-band images, probably caused by their different penetration depth and volume scattering sensitivities.

In recent years, space-borne synthetic aperture radar (SAR) polarimetry has become a valuable tool for sea ice type retrieval. L-band SAR has proven to be sensitive toward deformed sea ice and is complementary compared with operationally used C-band SAR for sea ice type classification during the early and advanced melt seasons. Here, we employ an artificial neural network (ANN)-based sea ice type classification algorithm on a comprehensive data set of ALOS-2 PALSAR-2 fully polarimetric images acquired with a range of incidence angles and during different environmental conditions. The variability within the data set means that it is ideal for making a novel assessment of the robustness of the sea ice classification, investigating the intraclass variability, the seasonal variations, and the incidence angle effect on the sea ice classification results. The images coincide with two different Arctic campaigns in 2015: the Norwegian Young Sea Ice Cruise 2015 (N-ICE2015) and the Polarstern’s (PS92) Transitions in the Arctic Seasonal Sea Ice Zone (TRANSSIZ). We find that it is essential to take into account seasonality and intraclass variability when establishing training data for machine learning-based algorithms though moderate differences in incidence angle are possible to accommodate by the classifier during the dry and cold winter season. We also conclude that the incidence angle dependence of backscatter for a given ice type is consistent for different Arctic regions.

Increased human presence and commercial activities in the Barents Sea (fishing, offshore oil and gas exploration) are amplifying the need for large-scale operational ocean monitoring of the eventual oil spills in the region. The geographical location and climate impose additional constraints on satellite-based monitoring, making it necessary to use Synthetic Aperture Radar (SAR). Dark features or low backscatter areas are frequent within the SAR images and their occurrence may indicate oil spills or so-called lookalikes. Automatic oil spill detection hinges on accurate separation of the lookalikes from actual oil spills. Two main types exist in the Barents Sea: newly formed sea ice and low wind regions, where the former occur during the freezing part of the year (approx. November - April) and the other year around. Mapping the occurrence of oil spills and lookalikes in the Barents Sea on a seasonal basis would add to our understanding and knowledge of the low backscatter phenomena. Awareness of the major locations of oil spills, natural oil seeps, or lookalikes, are important for operational services and their effort to reduce false alarms. Here, we explore the use of a segmentation-based dark feature detection method with Sentinel-l Extra Wide-Swath SAR images. We test the method on images acquired over the Barents Sea during the freezing season, and cross-validate the results with two sets of dark features segmented by operational expert oil spill and sea ice monitoring services. The results are discussed, together with currently developing method improvements, all while working towards a fully-automated method for monitoring dark features in the Barents Sea.

We attempt to understand how a set of well known polarimetric Synthetic Aperture Radar (SAR) features are impacted by the additive system noise for mineral oil and produced water slicks. For this, we use quad-polarimetric SAR scenes from Radarsat-2. Oil slicks at sea can be detected using SAR instruments, and the dual- (HH-VV) and quad-polarimetric modes can provide additional information about the characteristics of the oil. Therefore the increase in polarization dimensionality may be beneficial in a potential clean-up situation. For example, characterization could aid in separating different types of oil slicks, like mineral oil and produced water as studied here. Oil slick characterization using scattering properties can only be performed if the returned signal is well above the noise floor. To avoid misinterpretation it is important to understand how the noise impacts the measured radar signal. Most of the features investigated in this study were to a larger degree influenced by the additive noise. Further, a backscatter signal level of 10 dB above the noise floor is identified as necessary to support analysis of the scattering properties within the oil slicks without too much noise contamination of the signal. The mineral oils and produced water slicks showed similar polarimetric behavior, despite their chemical and physical differences at release.

Maritime activities in the Arctic Ocean is increasing and consequently the risk for an oil spill there is rising. Synthetic Aperture Radar (SAR) is used operationally to detect and monitor oil slicks and for sea ice monitoring and observations. Leads are often used for ship routing and within the leads newly formed sea ice is often present. Separation between the low backscatter areas that constitutes oil slicks and newly formed sea ice is therefore important. Here we compare fully polarimetric L-band SAR images overlapping both oil slicks and newly formed sea ice. For the oil slicks airborne Uninhabited Aerial Vehicle SAR (UAVSAR) images are used and for the newly formed sea ice an ALOS-2 PALSAR-2 image is used. Using a set of multi-polarization features we observe that the coefficient of variation of the polarization difference can be used to separate the two.

Operational ice services around the world have recognized the economic and environmental benefits that come from the increased capabilities and uses of space-borne Synthetic Aperture Radar (SAR) observation system. The two major objectives in SAR based remote sensing of sea ice is on the one hand to have a large areal coverage, and on the other hand to obtain a radar response that carries as much information as possible. Although until now, L-Band SAR sensors are rarely used in an operational context, it offers greater capabilities for sea ice type retrieval and is more robust during the melt season compared to higher frequency bands. With the help of JAXA's ALOS-2 PALSAR-2 sensor, we are able to explore the potential of polarimetric L-band acquisitions for sea ice analysis and classification in an operational environment. In this study we investigated the incidence angle related variation on the L-band backscatter and recommended optimal scenarios for Artificial Neural Network based sea ice type retrieval schemes.

In recent years, spaceborne synthetic aperture radar (SAR) polarimetry has become a valuable tool for sea ice analysis. Here, we employ an automatic sea ice classification algorithm on two sets of spatially and temporally near coincident fully polarimetric acquisitions from the ALOS-2, Radarsat-2, and TerraSAR-X/TanDEM-X satellites. Overlapping coincident sea ice freeboard measurements from airborne laser scanner data are used to validate the classification results. The automated sea ice classification algorithm consists of two steps. In the first step, we perform a polarimetric feature extraction procedure. Next, the resulting feature vectors are ingested into a trained neural network classifier to arrive at a pixelwise supervised classification. Coherency matrix-based features that require an eigendecomposition are found to be either of low relevance or redundant to other covariance matrix-based features, which makes coherency matrix-based features dispensable for the purpose of sea ice classification. Among the most useful features for classification are matrix invariant-based features (geometric intensity, scattering diversity, and surface scattering fraction). Classification results show that 100% of the open water is separated from the surrounding sea ice and that the sea ice classes have at least 96.9% accuracy. This analysis reveals analogous results for both X-band and C-band frequencies and slightly different for the L-band. The subsequent classification produces similarly promising results for all four acquisitions. In particular, the overlapping image portions exhibit a reasonable congruence of detected sea ice when compared with high-resolution airborne measurements.

Synthetic aperture radar is used to identify and monitor oil spills. Separation from oil spill look-alikes is an important part of a fully automatic oil spill detection scheme. Here we investigate the polarimetric signatures for oil spills and newly formed sea ice (a well-known look-alike) in fully polarimetric Radarsat-2 satellite scenes. Using the fully polarimetric scenes we calculate four different parameters, co-polarization ratio, polarization difference, scattering entropy, and mean alpha angle. Three pairs of satellite scenes with comparable incidence angles are used. We observe that a combination of the co-polarization ratio and the polarization difference enables us to delineate the spills from their surrounding and also to discriminate the oil spills from the newly formed sea ice. The scattering entropy and the alpha values provide additional information about the scattering mechanisms of sea ice and oil spills.