COSMO–SkyMed Synthetic Aperture Radar data to observe the Deepwater Horizon oil spill

Oil spills are adverse events that may be very harmful to ecosystems and food chain. In particular, large sea oil spills are very dramatic occurrence often affecting sea and coastal areas. Therefore the sustainability of oil rig infrastructures and oil transportation via oil tankers are linked to law enforcement based on proper monitoring techniques which are also fundamental to mitigate the impact of such pollution. Within this context, in this study a meaningful showcase is analyzed using remotely sensed measurements collected by by Synthetic Aperture Radar (SAR) satellites. The Deepwater Horizon (DWH) oil accident that occurred in the Gulf of Mexico in 2010 is here analyzed. It is one of the world’s largest accidental oil pollution event that affected a sea area larger than 10,000 km2. In this study we exploit SAR data collected by the Italian COSMO–SkyMed (CSK) X–band SAR constellation showing the key benefits of multi–polarization HH–VV SAR measurements in observing such a huge oil pollution event.

known as speckle, which hampers the interpretability of such images. Furthermore, there are other 85 physical phenomena, known as look-alikes, which can generate dark areas in SAR images not related to 86 oil spills, such as biogenic films, low-wind areas, rain cells, internal waves and oceanic or atmospheric 87 fronts [15]. Accordingly, tailored filtering techniques must be developed in order to minimize the 88 number of false alarms. They are generally based on the use of single-polarization SAR data together 89 with ancillary data [5,14,16]. In some cases, the distinction between oil slicks and biogenic films is based 90 on optical data [5]. The importance of dual-polarization SAR measurements has been demonstrated in 91 literature for oil slicks observation purposes [17,18,19]. Nevertheless, although it has been physically 92 demonstrated by theoretical modelling and experiments that polarimetric SAR measurements are the 93 most adequate source to monitor oil slicks at sea [10,20], it is important to analyze, especially in the 94 occurrence of large oil spill accidents, how all the available SAR measurements can be exploited at 95 best. 96 In this study a multi-polarization analysis of the capabilities of dual-polarization PP mode X-band 97 CSK SAR data is first undertaken focusing on the DWH oil spill. The latter was extensively monitored 98 by means of L-, C-and X-band SAR systems but, to the best of our knowledge, no study exploited 99 the incoherent CSK PP mode to consider such a huge oil spill event [21,22,23,24]. Oil spill detection 100 and estimation of the polluted area is undertaken using a textural-based image processing approach, 101 while a multi-polarization analysis is undertaken in order to characterize the contrast, i. e., the ratio 102 between the Normalized Radar Cross Section (NRCS) relevant to the slick-free and oil-covered sea 103 surface, both in the HH and VV channels.  • the spill originated from a water-depth of 1500 m. This has confounded many problems on 126 understanding the behaviour of the oil [28,29]. In general, oil at sea is influenced by a number of 127 advective processes, e.g. wind and wave advection, spreading, etc., and weathering. The latter is 128 a non-advective process that alters the oil's chemical and physical properties. In addition to the 129 conventional weathering process on the surface, the DWH oil was subjected to weathering as it • fresh oil was continuously released. Unlike "conventional" tankers oil spills, where oil is 134 released at once, the DWH oil spill was far more challenging due to continuous fresh oil release. • a massive use of dispersants was made to mitigate the oil's impact on the environment [26,28].

139
The dispersants help to reduce the oil-water interfacial tension which, when aided by the 140 addition of energy in the form of wind/waves, can help to enhance natural dispersion of the oil.  In addition, such huge oil spillage may have a critical long-term impact over the whole marine 150 and coastal ecosystem and, therefore, still needs to be continuosly monitored [31,32].

151
In summary, this unprecedented oil spill accident triggered the operational use of SAR techniques 152 to provide detailed information on the surfactants related to the DWH accident. Nevertheless, kind of surfactants. This implies that a synergistic use of different SAR operating modes is needed. In 156 fact, large-swath imaging modes, i. e., ScanSAR, allow obtaining information on the extent of the oil 157 spill, while narrower swath polarimetric modes, i. e., PP, allow extracting deeper information on the 158 oil's backscattering. provided by ancillary remotely sensed data, e. g., scatterometer/radiometer or buoy measurements.

177
Unfortunately, very often the information coming from other remotely sensed sources is not co-located  information once a priori wind direction information is available [34,35,36]. In this study, a spectral 184 approach is considered that does not require any a priori wind direction information to provide the 185 wind speed map. This approach is based on the inherent SAR peculiarities, i. e., the low-pass filtering 186 in the azimuth direction due to the orbital motion of the sea surface waves that distorts the Doppler 187 history of the backscattered waves [37,38]. The wind map, generated using the azimuth cut-off method, 188 is shown in Figure 3 where the oil-covered area is masked out. It can be noted that low-to-moderate 189 wind regime applies that is characterized by a mean wind speed of 7 m/s at the SAR acquisition time.

190
Hence, SAR sea oil slick detection can be effectively undertaken. In this subsection a texture-based oil spill detection procedure is undertaken to assess the potential 193 of CSK SAR data to detect the DWH oil spill and to estimate its surface extent.

194
In order to extract suitable intensity-based features that allow obtaining the oil spill detection binary 195 mask, a textural-based feature extraction algorithm is adopted using the Gray-Level Co-occurrence   the VV NRCS clearly separates the polluted area, that calls for ASM values larger than 1 due to its 215 homogeneity, from the surrounding sea, that represents a more heterogeneous scenarios resulting in 216 a lower ASM values (see Figure 4 (b)). Please note also that the few isolated black spots related to 217 metallic targets at sea involved in cleaning-up operations (see bright spots in Figure 2) are visible in the 218 oil spill detection map. This is likely due to the fact that they behave as very homogeneous scatterers.

219
The oil spill can be detected even from the HH NRCS, although a very slightly larger number of false 220 alarms and missed oil pixels within the slick are observed, see Figure 4 (a).

221
Hence, according to the detection map of Figure 4 (b), the extent of the DWH oil spill can be estimated 222 to be approximately 100 km 2 at the SAR acquisition time, i. e., 3 days after the accident. In this subsection a multi-polarization analysis is undertaken to discuss the sensitivity of HH-225 and VV-polarized NRCS, σ 0 HH and σ 0 VV , respectively, to slick-free and oil-covered backscattering.

226
The two intensity channels are jointly used to generate the Pauli false-color RGB images of Figure 5 227 where the following color-coding is adopted: R (σ 0 VV ); G (σ 0 HH ) and B (σ 0 HHσ 0 VV ). It can be noted 228 that the joint use of VV and HH channels provides further information that can be exploited to gain    Table 1 245 where the contrast ∆, i. e., the slick-free to oil-covered σ 0 ratio, is also listed for both the channels. As

263
In this study, the capability of multi-polarization CSK SAR data, gathered in dual-polarization PP • The mean σ 0 VV sea-oil contrast is always larger than the σ 0 HH one;