2.1. Design and Response Capabilities of Airborne SAR
Oil slicks show up in SAR images from all platforms as relatively dark areas, often contrasting with the radar signature of surrounding water [14
], and can be observed with X-band, C-band or L-band instruments operating in either horizontal (H) or vertical (V) polarization. The backscattered signal amplitude is reduced by oil primarily through damping of the small-scale ocean surface roughness and has been shown to agree well, from both satellite and aircraft platforms [16
], with the tilted Bragg scattering kB
]. In the Bragg model, the measured returns are the coherent backscatter from the ocean wave spectral component of the same wavenumber as the radar wave projected on the ocean surface, kB
), where kR
is the radar wavenumber and φi
is the incidence angle. For incidence angles between ~20 and 70 degrees, the highest return power is given by the vertical polarization returns (VV, or transmit V, receive V), followed by HH, then the cross-polarization signal, HV, which is generally substantially lower (~6 dB) than the co-polarized returns [22
]. Satellite-borne radars have been used for decades in operational detection of slicks [12
Compared to satellite SAR, airborne SAR has important advantages for urgent response, which include targeted deployment to a specific area, rapidity of image acquisition, short time repeat imaging capability and the potential of having a much higher signal-to-noise ratio (SNR) radar instrument. Because of the importance of rapid deployment, the most useful instruments are deployed on mobile platforms that can be targeted to specific areas without substantial delay, and aircraft have the advantage that they can travel long distances relatively quickly. However, as drones become more common and capable of carrying more volume and mass, they could provide the same advantages of aircraft, possibly at lower cost.
The primary key capability provided by airborne SAR relative to satellite instruments is rapid repeat imaging for slick tracking. Slick position vs. time can be used to initiate models that then forecast where the oil will be transported. This capability has been demonstrated with UAVSAR in a controlled oil release experiment in the North Sea [24
]. In that experiment, a comparison of an oil slick transport model to a series of SAR images acquired at ~30-min intervals and local current measurements showed that knowledge of the local currents near the slick significantly improved the forecasted slick transport relative to using currents estimated from an operational ocean wave model alone. Local currents were measured through the release of a self-locating datum marker drift buoy (SLDMB) [25
] within the slick, and this was used to estimate the near-surface currents by subtracting the wave-induced Stokes drift component from the SLDMB trajectory.
The advantage of high SNR cannot be overstated. If the SNR is too low, the radar return power from an oil slick will be below the instrument noise floor, in which case, the signal can be used for the detection of slick extent, but not for differentiating zones of more or less oil within the slick [20
]. Most satellite SAR instruments have little or no margin between the oil slick return and their noise floor [20
], which seriously contaminates the returns and makes the instruments insensitive to oil characteristics. Observations indicate that the slick thickness is related to the damping of the capillary and gravity-capillary waves, with thicker layers causing more damping [16
], i.e., less return power and darker images. Using imagery collected during the Deepwater Horizon oil spill in 2010, researchers have made advances in determining the volumetric oil fraction [16
] and estimates of layer thickness [29
] from SAR. Accurate information about thickness or volume is useful even in the absence of low latency product generation because it provides scientifically-supportable release volumes for remediation or restitution, information that is needed after clean-up has ended.
High spatial resolution is needed to identify small slicks, delineate slicks in near-shore areas and locate zones of thicker oil within slicks. Spatial averaging of high resolution single-look pixels is also an effective way to reduce noise and identify areas showing higher damping within slicks. The spatial resolution of a SAR instrument is primarily a function of the instrument bandwidth, which can be the same for an airborne or spaceborne instrument. In practice, spaceborne SARs have been limited by a combination of power requirements, on-board data storage and data downlink capacity, and tradeoffs are made between temporal sampling, swath coverage, multi-vs-single polarization and spatial resolution to fit within the system’s capabilities. On-board data storage is not a limitation for airborne SARs, so data can be nominally acquired at finer resolution with full swath capability and even with full polarization diversity.
Finally, airborne SAR needs the capability of rapid product generation and delivery to be useful for response. On-board processing can produce low resolution images that can be georeferenced, then downlinked as GIS-ready files or simple image files to responders in the field.
2.2. UAVSAR: A Testbed for Oil Spill Response
NASA first began to study oil slick characterization using airborne L-band SAR during the 2010 Deepwater Horizon oil spill in the Gulf of Mexico when images were acquired both over the main slick and along the Gulf coast. Following the encouraging results showing the sensitivity of SAR backscatter to the dielectric properties of the oil [16
], which is related to the volumetric fraction of oil in the surface layer, a controlled release experiment in the North Sea was undertaken [31
] in which weathering and transport of thin films (<1 μm) were studied. In 2016, the work continued in the Gulf of Mexico at the site of a persistent seep off the coast of Louisiana to study transport and weathering of a complex slick with thickness ranging from thin sheen through thick emulsions under natural conditions. All of the work was done using the UAVSAR platform, which is a NASA airborne science instrument.
UAVSAR operates in the L-band (1.2575-GHz center frequency) with an 80-MHz bandwidth, which provides 1.7-m resolution in the slant range direction and has 0.8-m resolution in the along-track direction [32
]. Typical products are generated with averaging to ~7-m spatial resolution. The radar operates in fully-polarimetric mode, meaning that it sequentially transmits vertical and horizontal polarization pulses and receives returns from each in both polarizations. The instrument is deployed on a Gulfstream-3 aircraft within a pod that mounts below the fuselage near the center of the aircraft (Figure 1
). The antenna is flush mounted to the left side of the pod (left-looking), and the instrument images a swath ~22 km wide, with an incidence angle ~22° (near range) to ~67° (far range) in standard SAR side-looking geometry [30
]. Typical flight operation is at a 12.5-km altitude (41,000’). The signal-to-noise ratio varies across the swath and is lowest at the edges and maximum near the swath center. Information on the instrument noise figure is provided in [33
]. UAVSAR radar system characteristics are summarized in Table 1
. Figure 2
shows a typical UAVSAR flight line of a swath width of 22.2-km overlaid on Google Earth. The scene shown was imaged in 10 min. Flight duration ranges from 5–6 h depending on weather conditions and altitude.
UAVSAR flight plans are developed prior to deployment and submitted to the U.S. Federal Aviation Administration (FAA) at least 24 h before radar operation. This is a strict requirement for radiating because UAVSAR operates at the same frequency as some of the commercial air traffic control (ATC) radars, and ATC centers are notified in advance that UAVSAR will be operating in their area. All flight lines (imaged swaths) potentially to be acquired must be in the plan, so re-planning the location of flight lines on-board can only involve selection between a fixed set of pre-defined lines that have been submitted to the FAA. For certain missions, in particular those involving imaging of ocean features, the exact location that should be imaged is not necessarily known in advance. For those studies, it is critical that there be some on-board processing capability so that quick-look views can be generated to determine whether the feature is imaged in a specific flight line.
An on-board processor (OBP) unit has been developed for UAVSAR and is installed inside the cabin on the aircraft when needed for specific experiments. OBP images are generated as Google Earth files (kml) with low resolution and latency of approximately one minute between when the radar images the surface and when the kml file is produced. The images are provided in segments whose width is set by the processing algorithm, and are generated as the radar returns are acquired. Figure 3
shows the Google Earth screen with a series of 11 UAVSAR image segments displayed. Figure 4
shows a close-up OBP image of part of the slick, compared to the fully-processed image to show that some of the darker areas within the slick are visible in the near-real-time product. Higher resolution could be achieved on-board with longer processing time, more computing power and more storage. UAVSAR’s on-board processing capability includes file downlink to the ground, and other image formats could be accommodated with minor modification to the system. Downlink through a satellite phone has been demonstrated in the past, but the capability is not currently available on the aircraft.