GIS and Earth Observation University

SAR technology

Synthetic Aperture Radar (SAR) has entered into a successful period, while more than 10 spaceborne SAR sensors are being operated today such as RADARSAT-2, RISAT-2, HJ-1C [8], which are the main providers of satellite SAR images for oil spill monitoring, and 10 new SAR systems (NovaSAR-X) will be launched within next 5 years[9]. Spaceborne SAR sensors are extensively operated for the detection of oil spills in the marine environment, as they are independent from sun light, they are independent by cloudiness, they can cover quite large areas and are more cost-effective comparing to airborne monitoring systems. SAR is an active microwave sensor that has a unique imaging capability.  It captures high spatial resolution two-dimensional images [5], with a side-looking imaging geometry and is based on a pulsed radar, which is installed on a platform with a forward movement [9]. Thus, it can be used as a tool to detect the possible location of an oil spill. SLAR (Side- Looking Airborne Radar) is another type of Radar, which can be used for oil spill detection and identification. SAR has superior spatial resolution and range than SLAR due to the fact that SLAR needs a quite big antenna to be installed in order to get high spatial resolution [2]. On the other hand, SLAR is less expensive and mainly used for airborne remote sensing while SAR sensors can be installed on both aircrafts and satellites [10].

Several spaceborne SAR systems have been used over the years for oil spill monitoring and they are described by their frequency band, polarization that refers to the geometry of the electromagnetic signal, the incidence angle, the swath width and the image resolution [5]. Incidence angle is the angular relationship between the radar beam and the ground target, while swath width is actually the distance, centered on the sensor path [11]. The swath width also varies in airborne or spaceborne cases from 20km to 500km. Frequency bands that are commonly used for SAR systems and more specifically for agriculture, ocean and ice monitoring are L, S, X and C-Band [9]. Furthermore there are four typical polarization combinations (HH, VV, HV and VH) where the first letter indicates the transmitted polarization and second one the received polarization [12]. Multi-polarization SAR sensors are quite new in SAR domain and have been used in order to improve the oil spill detection[11]. Dual-polarized (VV and HH) C-Band combination seems to be one of the most appropriate choices for oceans monitoring [11, 12]. Table 1 presents few satellites models with their basic characteristics, of the last two decades, that carrying SAR sensor and is commonly used for ocean monitoring.

Table 1 Satellites models [5]

Oil spill detection by SAR is a complex procedure that is described by several steps, which will be briefly explained below. Once the area of interest is chosen such as open sea, coastal zones etc. a plan is designed by an operator to get SAR data.  Satellite acquires images from the selected area of concern and then these satellite’s data are received from the ground station. Pre-processing of data includes calibration, image cutting and georeferencing is the most time critical step [13]. The most crucial step is the oil spill detection in SAR image that can be described by three different approaches. The manual approach, which is carried out by experienced operators that are trained to analyze SAR images to detect oil spills [5]. Semi automatic and automatic are the other two approaches that are not very popular for oil spill detection because they are complex. In fully automatic method various algorithms are applied to detect oil spills and are usually divided into three sub-steps: dark spot detection i.e. identification of region of interest, features extraction and classification[3]. Dark formations and their detection are a long and analytical section, so wisely are briefly described in the next paragraph. Features are very important for the last sub-step (classification) because they are used as inputs to the classifier. They are also based on the geometry and shape of the selected area, on the backscatter values of the dark spot and there textural and contextual features. Several experiments have shown that textural features that are contained in SAR images are very useful of the analysis and classification of the image[14]. The purpose of the classifier is to identify the oil spill, which means that classifier follows a process in order to distinguish oil spills from other dark formations that are presented on the image. There are many different techniques that are commonly used for sake of the automatic approach and especially for the classification procedure, such as the Artificial Neural Network (ANN) [10], Wavelet Neural Network (WNN), Hybrid Polarimetric,  [15] etc.  The need for automatic algorithms is extremely high and depends on the number of images to be analyzed, but for monitoring large sea areas it is a cost-effective method comparing to the manual approach [3]. The last step of the cycle is to make the end user aware of current status for the decision making. Experienced technologists and experts carry out all stages, steps and approaches of the above described cycle [13]. 

As it mentioned above any dark formation that is detected on the image and is darker than the surrounding area, it will probably be an oil spill and needs classification methods to be applied to characterize them as oil spills or another object i.e. look-alike object [4]. Apart from oil spills dark formations can be oil spills, low wind areas, organic film, fronts, areas sheltered by land, rain cells, current shear zones, grease ice, internal waves, upwelling zones, down welling zones and eddies[16, 17]. If dark formations are not detected in this step they will never be identified. The reason why oil spills appear as dark areas in SAR images is that the oil dampens the capillary waves of the sea surface [7]. 

Although this procedure seems to be quite simple for a human specialist-operator (manual method) it includes few important difficulties for semi-automatic or fully automatic approach. At first fresh oil seems brighter on the image than the older oil spills. They have weak backscattering contrast comparing with the surrounding area and hence discrimination still being a challenging task. Furthermore dark formations can have various contrast values, depending on local sea state, oil type, image resolution and incidence angle. Last but not least the famous look-alike phenomenon which is also presented on the SAR image as a dark formation and it will briefly defined in later sections while it constitutes on of the most crucial limitations-difficulties in oil spills detection via SAR[14, 18]. Taking into account all the above, Synthetic Aperture Radar (SAR) can effectively be used for oil spills detection and identification but several meteorological phenomena and similar dark formations that may arise from a variety of reasons create many misidentifications. A complementation with other sensors it may give important and encouraging solutions. 

Other sensors complementation

The combined use of satellite-based synthetic aperture radar (SAR) images and aircraft surveillance flights is the most cost-effective way to monitor big areas and detect the source of the pollution (oil platforms, tankers, etc.). In overall comparison SAR is dominant because its capabilities for oil spill detection over the oceans are well known and proven even if it cannot combine all of its advantages[5, 6]. On the other hand there are several other sensors based on a different technologies, which can complement with a remarkable extent.

Optical sensors are based in a totally different technology and they can provide a quite good coverage like Terra/Aqua MODIS and ENVISAT MERIS, by imaging oil spills with high resolution but under certain weather conditions. UV technology can detect oil spills because spill displays highly reflectivity of UV radiation even at thin layers. Cloudiness, extreme weather conditions, and daytime operation, are the main disadvantages that can really affect detection and as a result the identification of oil spill [3, 6]. Nevertheless, sensors like UV and IR (infrared) can be easily mounted on aircraft and that makes operational use of these sensors convenient. Moreover, the cost and size of sensors can play a significant role in using sensors for oil spill surveillance. IR sensors are cheap and this has led to their widespread use for oil spill surveillance. Oil absorbs solar radiation and re-emits a portion of this energy as thermal energy. Also IR sensors observe thick oil slicks as hot, intermediate thickness of oil as cool, while thin oil is not possible to detect [2]. A combination of IR and UV can provide a more reliable indication of oil and can be used for oil thickness estimation. Apart from that, another way for oil thickness quantification is the microwave radiometer (MWR) [19]. Thickness is an important limitation-challenge for oil spill detection by SAR. 

Advanced sensors such as laser fluorosensors are very costly but they can complement to SAR sensor because of their capability to classify oil, which is a difficult task for SAR. Fluorosensors can also provide further information about the type and amount of oil on various backgrounds including ice and shoreline[2, 19]. Oil can be classified into heavy, medium and light crude or refined oil. Once classified, it is easier to respond to the oil spill and to model the oil spill drift and spreading. Hyperspectral sensor is an effective solution for oil spill distinction due to the fact that can provide more than 200 wavelengths [3]. 

As it mentioned above, different remote sensing sensors with different technology and technique can highly complement to the oil spills detection with Synthetic Aperture Radar (SAR). Real sampling is something that can provide the most accurate information about oil spill type and thickness. Only vessels that are equipped with specialized radars can carry out this procedure and then they can provide these data (oil type, thickness) to several sensors for a perfect complementation. However they cannot cover large areas and are costly enough [20]