1. Introduction
Due to global warming and climate change, sea ice thickness and its extent have drastically decreased in recent years [
1,
2]. Sea ice affects weather patterns, ocean circulation, coastal communities, and ecosystems, so the changes bring new challenges for transportation and natural resource development [
1]. The reduction of Arctic sea ice has increased the number of ships operating in this region to access natural resources (e.g., fuel resupply ships, and mining) [
3]. These activities are associated with a high risk of oil spills and other chemical releases, which negatively impact local populations of animals, plants, and humans [
4,
5,
6,
7]. These factors can also threaten on an international scale due to the crucial biological, cultural, and economic influence of the polar region [
8]. Owing to the tragic consequences of the Deepwater Horizon incident in the Gulf of Mexico, many countries, including Canada, advised extra attempts to address immediate solutions for future oil spill events, particularly in the Arctic [
9]. The ability to respond effectively and efficiently has been identified by the Arctic Council as a critical priority [
10]. Among major challenges are remoteness, short windows of open water and daylight, inadequate infrastructure (i.e., not enough equipment to sustain large groups), and communication. Therefore, remote detection and surveillance techniques are critical to confront effectively and respond to potential oil spills in the Arctic [
11].
It is expected that shipping traffic in the late summer and early fall will continue to increase in response to the later freeze-up [
12]. Fall freeze-up is a complex time of year in the marine environment, as ice can develop rapidly in various forms depending upon the prevalent meteorological conditions [
13,
14]. When sea ice forms during calm conditions, it begins as frazil ice that accumulates at the water’s surface. If an oil spill were to occur during this time, the oil would readily reach the surface of the water and would form a slurry of seawater, ice, and oil. As a separate scenario, if the sea ice grows under typical congelation growth, the ice provides a semipermeable physical barrier between the ocean and the atmosphere. This type of ice is called nilas; it is highly saline, flexible with wave motion, and can easily be mobilized by dynamic ocean conditions that cause overlapping (rafting), leads (openings in the sea ice), and ridging. If an oil spill occurs in this situation, an oil horizon will form beneath the sea ice and seek the surface, as the oil is less dense than seawater. In time, the oil would likely find its way to the surface through brine channels, pockets, and cracks in the ice. As a third scenario, we consider oil spills on the surface of the seawater. With subzero air temperatures, sea ice would begin to form beneath the oil spill. However, with wind and wave action, the oil could readily distribute laterally in the region, and some combination of dynamic forms of sea ice (e.g., pancake ice) would be seen. Detecting the location, extent, and trajectory of an oil spill is crucial for enacting an effective response plan. Current methods for detecting oil spills in the marine environment have detection times that are highly dependent upon the conditions. For example, in open water, an oil spill can be readily detected by a nearby observer or, if in a remote location, with an optical remote sensing satellite. When sea ice is present, the situation is significantly more complex. Various airborne and spaceborne remote sensing techniques can be used to increase the possibility of detecting the location and extent of an oil spill [
15,
16]. Spaceborne sensing operates over a large-scale area but is limited by long revisit cycles (e.g., an 8-day revisit cycle for both Landsat-5 and -7 and 2 days for MODIS), which fails to record significant changes between image acquisition intervals [
15]. Conversely, airborne sensing methods can obtain high temporal resolution, although in a much smaller region. In addition, other remote sensing techniques (e.g., optical and radar sensors) have been evaluated likewise in spotting oil slicks on both airborne and spaceborne platforms [
17]. Microwave radars are especially advantageous for transmitting and receiving signals in a range of frequencies independent of weather conditions and solar illumination which makes them a key component for polar science studies.
Nowadays, monitoring oil pollution in open water utilizes a combination of satellite- and aircraft-based single-polarization, active, microwave remote sensing systems (e.g., synthetic aperture radar (SAR) and scatterometers). In a recent study [
18], single-polarization SAR was used to determine the impacts of oil on smoothing ocean capillary waves, which resulted in an area of decreased backscatter. The presence of sea ice complicates the behavior of spilled oil in which oil can migrate, partition, and accumulate below, within, and above the ice [
19]. The freeze-up period can be a potential challenge for oil investigation where the formation of newly formed sea ice (NI), such as frazil, grease, and nilas, is prevalent. The challenge is that the oil-polluted NI and its surrounding clean ice can appear identical in SAR imagery [
20]. NI is a solid heterogeneous mixture, which contains high volumes of liquid brine as well as air pockets (with a lesser fraction) in a pure ice background. During the formation of NI, the ice surface can be accompanied by the presence of moisture or frost flowers (depending on atmospheric conditions) due to its high salinity at its topmost layer.
To date, only a few studies have attempted to improve the contrast between NI and oil-contaminated sea ice using polarimetric SAR [
20,
21,
22,
23]. To improve the contrast ability in a polarimetric radar system, we can mathematically combine multipolarization signals to synthesize polarimetric parameters (e.g., copolarization ratio (
Rco), cross-polarization ratio (
Rxo), entropy (
H), mean-alpha (α), conformity coefficient (μ), copolarization correlation coefficient (ρ
co), etc.). Previous studies were conducted using
Rco,
H, and
to distinguish between oil-contaminated and oil-free ice scenes on SAR images [
20,
21,
23,
24]. Their results revealed that the
Rco parameter had good separation potential, alongside
H and
, notably at lower incidence angles (<35°). A recent study [
25] found that oil-contaminated ice can be discriminated from oil-free ice using 0.3-
H and 18-degree mean-alpha. This finding was limited to crude oil, and our study is motivated to investigate this threshold in delineating diesel-contaminated ice. There are complementary studies that used alternative parameters to replace the
H/α classification for polarimetric synthetic aperture radar (PolSAR) data. For instance, [
26] demonstrated random similarity (with the similarity-based angle α
s and entropy
Hs, which measure the scattering mechanism and randomness of polarimetric scatterers, respectively), while [
27] presented scattering diversity (or specific scattering predominance) and surface scattering fraction as
H/α merited alternatives. Furthermore, several studies pointed out that ρ
co and μ parameters can be informative in distinguishing oil-contaminated and uncontaminated open water. In a recent study [
25], conformity coefficient (μ) was also demonstrated as a complementary means to differentiate between crude oil-contaminated and uncontaminated ice.
The polarimetric microwave scattering response of sea ice is dependent upon the physical and thermodynamic state of the ice [
28], thereby giving the potential to provide an additional metric for detecting or monitoring oil spills. The behaviors are mainly dependent on physical parameters such as sea-ice surface roughness and the heterogeneity level within the sea-ice medium. The
Rco parameter is associated with changes in surface scattering, which increases with an increase in incidence angle [
20,
29]. Additionally, changing the volume scattering can potentially enhance the
Rxo values through depolarization mechanisms (e.g., brine pocket rearrangements or oil movements within a homogenous ice volume) [
24,
30,
31]. According to previous research (e.g., [
25,
32]), target decomposition (TD) theorems have been used to differentiate between surface and volume scattering behaviors.
However, true validations of theoretical studies are limited as the ability to conduct oil spill experiments remains heavily regulated in the Arctic regions [
33]. Another limitation of these studies is the calibration accuracy of spaceborne and airborne systems, which can cause misinterpretation of oil slicks [
22]. To address these issues, one useful approach is to simulate the Arctic environment in a controlled mesocosm with a well-calibrated, surface-based scatterometer. Importantly, oil products (e.g., diesel, light oil, crude oil) have physically different characteristics, and therefore it is crucial to generate the scientific basis for the detection of a variety of substances that may be spilled in the marine environment.
In our previous experiments at the Sea-ice Environmental Research Facility (SERF), we performed experiments to detect oil spills in sea ice using a sweet crude product [
25,
34]. In this study, we investigated remote sensing techniques for diesel spill detection and monitoring. The focus of this study was to answer two main questions: (1) how does the presence of diesel fuel and the effect of wind upon seawater alter the growth, thermophysical, and microwave C-band scattering response of newly forming sea ice? and (2) which polarimetric parameters are most sensitive to the presence of diesel fuel on newly formed sea ice? In this study, we opted to use the C-band frequency as it is known for its ability to detect and monitor sea ice in all seasons [
35]. The C-band multipolarization measurement that has been applied in this research provides future comparison and validation with existing SAR satellites (e.g., the Canadian Space Agency RADARSAT 2, RCM, and the European Space Agency’s Sentinel 1). To achieve this goal, an oil-in-ice mesocosm experiment was conducted at the SERF, located at the University of Manitoba, Winnipeg, MB, Canada.
The remainder of this manuscript is structured as follows.
Section 2 provides an overview of our experimental setup and data collection.
Section 3 presents the results of the NRCS and polarimetric parameters.
Section 4 discusses the results presented in
Section 3. Finally,
Section 5 concludes with a summary and recommendations.
5. Conclusions
This study investigated the ability of C-band polarimetric radar to detect diesel spills on seawater under freezing and windy conditions. We performed an experiment at the SERF using a C-band polarimetric scatterometer and examined the microwave scattering response of diesel-contaminated seawater as it froze under prevalent environmental conditions. We addressed two key questions to improve scientific understanding of the behavior of diesel in freezing seawater so that the remote sensing information can be used to guide and inform a response in the event of an Arctic oil spill. Firstly, (1) how does the presence of diesel fuel and the effect of wind upon seawater alter the growth, thermophysical, and microwave C-band scattering response of newly forming sea ice? And secondly, (2) which polarimetric parameters are most sensitive to the presence of diesel fuel on newly formed sea ice?
We discovered three distinguishable stages using our radar system, which corresponded to the physical and thermodynamic conditions of the diesel-contaminated seawater as it froze. The strong, windy conditions during the study period played a leading role in our analysis and had notably altered radar responses. We measured the normalized radar cross section (NRCS) alongside six polarimetric parameters (i.e., Rco, Rxo, μ, ρco, H, and α), which were derived from the NRCS, covariance matrix, and coherency matrix of diesel-free and diesel-contaminated NI. Stage 1 was described as a calm condition, where the NRCS results were relatively low and had low variability. Stage 2 results revealed that the NRCS was sensitive to the presence of wind, which mixed and distributed the diesel in an emulsion. Stage 3 occurred when sea ice formed beneath the diesel, and we observed a drastic increase in backscattering responses. These behaviors were comparatively more considerable at the 25° incidence angle. Regarding these observations altogether, we can state that the NRCS at 25° incidence angle is more sensitive to the presence of diesel fuel, seawater, and NI. Therefore, the NRCS can be informative in detecting the presence of wind and sea ice in diesel-contaminated seawater.
We examined polarimetric parameters following the same three stages previously described. The Rco and Rxo parameters were relatively stable, with the presence of diesel upon seawater under calm atmospheric conditions. However, in Stage 2, they showed sensitivity to the presence of wind by presenting oscillatory behaviors. This sensitivity was greater at 25° than for the 20° incidence angle. Moreover, both Rco and Rxo parameters showed no behavior in detecting NI beneath the diesel layer (Stage 3). Therefore, the Rco, Rxo parameters can effectively detect the presence of wind on oil-contaminated seawater, especially at 25°.
The polarimetric parameter, μ at the 20° incidence angle, maintained positive values throughout the experiment, even when calm atmospheric conditions were prevalent. This behavior demonstrates that μ is more sensitive to the surface roughness of diesel, seawater, and NI at lower incidence angles. In response to the presence of wind, the μ parameter at the 25° incidence angle exhibited a transitory behavior (from negative to positive values). The ρco parameter also showed a significant response with a drastic increase (within positive values). This increase was greater at 25° and clearly delineated the transition to Stage 2. Thus, the μ and ρco parameters are both sensitive to the presence of wind. Stage 3 demonstrated that μ and ρco parameters are informative to the ice formation by illustrating transient plunges in magnitude. These plunges at 20° were stronger, which reveals that the μ and ρco parameters at lower incidence angles are more sensitive to the NI formation. In conclusion, the μ and ρco parameters show potential for discriminating between the presence or absence of wind and NI on diesel-contaminated seawater.
Examination of polarimetric parameters H and α showed that both seawater and NI measurement points (clusters) with and without the presence of wind fell within the SSLE zone. During the study period, the clusters of each incidence angle were mostly concentrated in distinct segments, particularly when the wind conditions were calm. This concentration at the 20° incidence angle was H = 0.28 and α = 21°, and for 25° it was H = 0.39 and α = 15°. The distribution of clusters at 25° was expanded, which discloses that the addition of the wind factor can conceivably increase the difference between incidence angles. Furthermore, both 20° and 25° incidence angles demonstrated linear behavior in distributing the clusters. However, the orientation of clusters in a diesel-spill scenario is significantly distinct and unique from the orientation of clusters in the crude oil scenario.
The findings of this article can greatly improve the capability of detecting diesel spills on seawater under freezing and windy conditions. The results of NRCS alongside the combination of polarimetric parameters as the complimentary criteria can play a major role in diagnosing diesel pollution in the Arctic marine environment. However, our study was limited to the experimental area, low incidence angles, absence of snow, and freezing temperatures. Future work will consider higher incidence angles (>25°) with longer wavelengths such as L-bands. Moreover, in order to improve the accuracy of our analysis, a larger experimental site in the Arctic is recommended, where a well-representative set of radar responses and physical samples can be obtained. To reach these requirements, Canada’s Churchill Marine Observatory facility has been proposed by our team to continue the related studies. Given the rapid changes in the NRCS and polarimetric parameters, it is evident that near real-time monitoring is important for seeing changes, and therefore surface-based radar systems or drone-based radars can play a key role in an oil spill response plan in Arctic waters.