Remote sensing provides an effective tool for timely oil pollution response. In this paper, the spectral signature in the optical and infrared domains of oil slicks observed in shallow coastal waters of the Arabian Gulf was investigated with MODIS, MERIS, and Landsat data. Images of the Floating Algae Index (FAI) and estimates of sea currents from hydrodynamic models supported the multi-sensor oil tracking technique. Scenes with and without sunglint were studied as the spectral signature of oil slicks in the optical domain depends upon the viewing geometry and the solar angle in addition to the type of oil and its thickness. Depending on the combination of those factors, oil slicks may exhibit dark or bright contrasts with respect to oil-free waters. Three oil spills events were thoroughly analyzed, namely, those detected on May 26 2000 by Landsat 7 ETM + and MODIS/Terra, on October 21 2007 by MERIS and MODIS, and on August 17 2013 by Landsat 8 and MODIS/Aqua. The oil slick with bright contrast observed by Landsat 7 ETM + on May 26 2000 showed lower temperature than oil-free areas. The spectral Rayleigh-corrected reflectance (Rrc) signature of oil-covered areas indicated higher variability due to differences in oil fractions while the Rrc spectra of the oil-free area were persistent. Combined with RGB composites, FAI images showed potentials in differentiating oil slicks from algal blooms. Ocean circulation and wind data were used to track oil slicks and forecast their potential landfall. The developed oil spill maps were in agreement with official records. The synergistic use of satellite observations and hydrodynamic modeling is recommended for establishing an early warning and decision support system for oil pollution response.
© 2014 Optical Society of America
Oil pollution threatens coastal environment and infrastructure particularly in oil-rich regions, like the Arabian Gulf where desalination plants, the main source of potable water in the region, are constantly at risk. The adverse effects of oil spills generated concerns related to marine resource exploration, recovery, transportation, consequent oil pollution contingency planning, mitigation, and remediation . Oil pollution usually occurs around major shipping routes [2,3] and indicates close connection with offshore installations . According to , 48% of the oil pollution in the marine environment was caused by fuels and 29% was attributed to crude oil. Reports from  showed that 52% of the marine oil pollution was engendered by urban runoff and industrial discharges, 21% was arisen by oil production, 13% was caused by particle settlement in the atmosphere, 9% resulted from natural oil seepage, and tanker accidents contributed only 5% of the marine oil pollution. We expect that this figure changes dramatically in arid and oil-rich regions, like the Arabian Gulf, where urban runoff can be negligible and the traffic rate of oil tankers is exceptionally high. Pavlakis et al.  reported that deliberate oil pollutions occurred more often than those due to ship accidents. Statistics from  indicated that 45% of the global marine oil pollution was caused by operative ship discharges. With respect to the Arabian Gulf, the major contributors to the oil pollution include oil exploration, production and transportation .
Remote Sensing provides frequent and synoptic measurements of the vast ocean surface. It has been widely utilized for aquatic oil pollution monitoring. Published literature using remote sensing approaches provides insights to the impacts of oil pollution on the coastal marine environment [5,10–14].
Active microwave sensors, like Synthetic Aperture Radar (SAR), are commonly used for remote sensing of oil pollution . However, the common use is limited by their high costs, relatively small swath width (e.g. 100 km for Radarsat-1 at an altitude of 798 km), and low revisit frequency (e.g. 24 days for Radarsat-1). Another challenge in oil detection with this technique is the inaccurate discrimination between oil slicks and look-alikes , like algal blooms and low wind areas as they may both appear as dark features in SAR images.
In addition to microwave sensors, several optical satellite sensors have been used to aid oil pollution response. The capability of Advanced Very High Resolution Radiometer (AVHRR, 1978-present) for early monitoring and detection of oil spills has been examined by . They studied the oil spills in 1991 in the Arabian Gulf and showed that a) the use of the infrared (IR) channel to detect oil spills was possible, b) oil spills may not show significantly discernible temperature signatures from the surrounding waters at night, and c) oil spills can be detected in the visible images only under favorable lighting and sea conditions. Casciello et al.  and Grimaldi et al.  also used thermal IR data of AVHRR for oil detection based on Robust Satellite Technique (RST) approaches. But the 1-km coarse resolution of AVHRR, with respect to the extent of oil slicks, can lead to slick-estimation biases inasmuch as that oil slicks exhibit small-scale heterogeneity. The Sea-viewing Wide Field-of-view Sensor (SeaWiFS, 1997-2010) on board NASA/GeoEye’s OrbView-2 was also used for oil detection in spite of its coarse spatial resolution . Moderate Resolution Imaging Spectroradiometer (MODIS) aboard Terra (2000-present) and Aqua (2002-present) satellites provides global coverage twice daily in near real time and is equipped with two 250-m and five 500-m resolution bands. These medium resolution data are of great benefits for oil detection [13,14,20–22]. Hu et al.  utilized MODIS 250-m resolution data to monitor oil spills in a turbid lake in Venezuela, which indicated dark contrast against the bright background. Hu et al.  showed that MODIS imagery containing sun glint was capable of detecting oil slicks in oligotrophic open ocean waters. MEdium Resolution Imaging Spectrometer (MERIS, 2002-2012) on board Envisat also showed great potentials for oil spill monitoring [13,25]. Oil identifications using MODIS and MERIS visible images with and without sun glint contamination have been reported [21,22]. In the presence of sun glint, which is strongly dependent upon the solar geometry and the wave fields, oil impacted areas can be well visualized with high contrast compared to oil-free areas. In the absence of sun glint, the contrast between oil slick and background depends on their different spectral characteristics.
With respect to the use of high resolution imagery, Landsat missions have proven their positive values in mapping oil slicks despite narrower swath width (185 km) and lower revisit frequency (16 days). For example, Landsat 5 TM (1984-2013) captured the image of an oil spill emergency in the Guanabara Bay, Brazil , and Essa et al.  successfully used Landsat 7 ETM + (1999-present) to detect oil slicks in the Arabian Gulf. Landsat 8 was launched in February 2013 and will provide continuity of 40-year imaging data set. It should offer the same potential for oil detection like the previous generations of Landsat missions. All of these optical satellite sensors have performed and will continue to perform important roles for detecting and monitoring oil pollutions. On the other hand, relying on a single sensor is insufficient to meet the requirements of the sequential and operational monitoring of oil pollution due to improper weather conditions (such as cloud cover), different geometry of acquisition, and the lifetime of oil spill. Due to differences in their specifications and configurations, the sensitivity to the presence of oil varies from one sensor to another. There is no single sensor that could provide all necessary information on oil pollution management. Furthermore, relatively long revisit schedule prevents some sensors from being utilized for operational oil pollution monitoring. Therefore, integrating multi-sensor data should be beneficial for routine monitoring of oil pollutions and overall surveillance operations. In addition, multi-sensor satellite imagery can provide comprehensive spatial coverage over the affected areas and complementary information on oil slicks that could cover their extent, and their thicknesses and temperature, which may substantially improve the modeling of the diffusion of oil pollution.
The Arabian Gulf is under considerable threat of oil pollution [9,28]. Timely and effective response is required. Essa et al.  investigated the feasibility of using multiple sources of satellite imagery (including Landsat 7 ETM + , SAR, ASAR, etc.) for oil pollution detection, and their findings demonstrated the frequent occurrence of oil pollutions in the Arabian Gulf and the Gulf of Oman, mainly caused by deliberate oil sludge dumping from giant oil tankers. Alawadi et al.  examined the capability of MODIS 250-m bands in differentiating oil slicks from algal blooms in the Regional Organization for the Protection of the Marine Environment (ROPME) region and developed a classification algorithm to identify different phenomena on the sea surface. However, spectral characteristics of oil pollution in the area have scarcely been reported. Furthermore, to the best of our knowledge, the potential of numerical ocean circulation models have not been sufficiently explored for oil pollution tracking in the shallow waters of the Arabian Gulf . Such models have proven to be of great benefit for oil tracking in the Deepwater Horizon oil disaster in 2010 [30–33].
In this study, we aim to detect oil slicks and depict their potential movements in the Arabian Gulf with multi-source optical and thermal remote sensing imagery. On satellite imagery other features may appear similar to oil slicks. These features include low wind areas, organic film, wind front areas, wind shadow and wave shadow behind land, rain cells, current shear zone, internal waves, and eddies . Hence, another aim of this study is to attempt to differentiate oil slicks from look-alikes. Following these purposes, Landsat 7 Enhanced Thematic Mapper Plus (ETM + ), MODIS, and MERIS medium resolution images are used to observe oil slicks in the Arabian Gulf with several examples illustrated. The capability of the recently launched Landsat 8 is also assessed. Multi-spectral characteristics of oil slicks are extracted and elucidated. To meet the needs of differentiating oil spills from look-alikes, e.g. algal blooms, Floating Algae Index (FAI)  images are examined. Finally, ocean circulation models that are driven by appropriate meteorological data are also integrated in the analysis along with satellite imagery for tracking oil slicks and predicting their landfall in the gulf region.
2. Study area
The Arabian Gulf (Fig. 1) is a shallow semi-enclosed marginal sea, surrounded by the United Arab Emirates (UAE), Oman, Qatar, Bahrain, Saudi Arabia, Kuwait, Iraq, and Iran. Its average depth is ~35 m. It has asymmetric bathymetrical features along the main axis with a deeper zone off the Iranian coast and broad shallow shelf along the southern and western coasts from Kuwait to the UAE. It is connected to the Gulf of Oman through the Strait of Hormuz. With arid land surrounding the gulf region, salinity of the gulf waters can reach up to 44.3‰ , and the water body becomes one of the warmest on earth with temperature reaching 32 °C during summer . In spite of the extreme conditions in the marine environment, fringing and patchy coral reefs and productive sea-grass beds were found in the Arabian Gulf. Some mangrove vegetation also resides in some coastal areas. The gulf region is also rich in marine life, such as sea turtle, sea bird, dugong, whale, dolphin, and fish, most of which are endemic.
The Arabian Gulf has the largest hydrocarbon reserve in the world and is also well known as the most active oil production area. 25% of the world’s oil is produced by countries surrounding the gulf. Most of the oil production is transported worldwide by means of tankers with an annual estimate of 35,000 tankers crossing the Strait of Hormuz . These exceptionally high maritime traffic in the region in addition to other potential sources (e.g. oil seepage, ship accident, military actions, ballast water dumping) constantly exposed the gulf to a high risk of oil pollution [39,40]. And this is specifically critical in the region as oil pollution affects the potable water supply which relies mainly on desalination. Major on-going and future activities to combat oil pollution in the gulf region have been conducted by ROPME member countries and coordinated by Marine Emergency Mutual Aid Center (MEMAC). Other organizations or companies, such as the International Tanker Owners Pollution Federation Limited (ITOPF), also offer services for oil pollution in the Arabian Gulf.
3. Data and method
3.1 Data acquisition and processing
This study utilized MODIS 250-m and 500-m resolution data, Landsat 7 ETM + and Landsat 8 30-m resolution data, and MERIS 300-m resolution data. All satellite data used in this study and their corresponding information are summarized in Table 1. Tables 2–5 list the spatial resolution and center wavelengths for MODIS, MERIS, Landsat 7 ETM + , and Landsat 8, respectively. MODIS Terra and Aqua data (level 0) were acquired from NASA ocean color data archive. These level 0 data were converted to calibrated radiance data using SeaDAS V7.0  with the most recent calibrations. Then correction for gaseous absorption and Rayleigh scattering was carried out. The resulting Rayleigh-corrected reflectance (Rrc) data were derived for MODIS bands 1-16 (Table 2). These Rrc data were mapped to a rectangular projection. All Rrc data were then resampled to 250 m. Red-green-blue (RGB) imagery was generated using the georeferenced Rrc at 645 (R), 555 (G), and 469 (B) nm. FAI imagery was produced using Rrc at 645, 859, and 1240 nm according to Hu . FAI used wavelengths in the red and near-infrared (NIR) regions, and is less sensitive to atmospheric effect and colored dissolved organic matter (CDOM) . MERIS L1B data were also obtained from NASA ocean color data archive and processed similarly to MODIS data in SeaDAS V7.0. Quasi true color images were generated with Rrc at 665, 560, and 443 nm.
Landsat 7 ETM + and Landsat 8 data at a resolution of 30 m, which are level-1 georeferenced total radiance, were obtained from U.S. Geological Survey (USGS). Landsat data were processed similarly to MODIS data. Rrc data were derived at bands 1-5 (Table 4) for Landsat 7 ETM + and at bands 1-6 (Table 5) for Landsat 8. Quasi true color images were composited with Rrc at 660, 565, and 483 nm for Landsat 7 ETM + and at 655, 562, and 443 nm for Landsat 8. FAI imagery was produced using Rrc at 660, 825, and 1650 nm for Landsat 7 ETM + and at 655, 865, and 1610 nm for Landsat 8. Although the band settings for calculations of FAI in Landsat imagery is different from those in MODIS imagery, comparisons between MODIS and Landsat FAI imagery by  suggested that surface slicks can be detected on both imagery, and Landsat FAI imagery was more functional in identifying small-sized features. In order to strengthen the contrast between areas with and without oil slicks, all RGB images were enhanced with a Gaussian enhancement . Landsat thematic images (level 1) were also obtained from USGS with no further processing.
3.2 Identification of oil slicks in satellite imagery
3.2.1 Optical imagery
Identification of oil slicks was carried out through a comprehensive analysis of their spectral signature, which was corroborated with visual examination of a variety of satellite ancillary products. First, images with minimal cloud cover, which were acquired on or close to dates of reported oil spills in the regions, were obtained. We used published literature and reports issued by regional organizations (like ROPME) to determine the dates of oil spill occurrence. Then, the selected images were processed according to steps described in Section 3.1. Areas that exhibit clear contrast with respect to neighboring pixels in the quasi true color RGB composites were delineated. Then, the change in reflectance through transects that involved the impacted areas were analyzed to characterize the spectral signature of oil spills in the gulf region.
Other factors such as ship wake and high concentration of phytoplankton as well as internal wave and wave shadow as stated by  can also indicate distinctive features in satellite imagery and might be misidentified as oil slicks. In this study, we investigate ways to identify their signatures and differentiate them from oil slicks which may lead to a significant reduction in false alarms.
For the sake of validating oil detection, airborne sensors and in situ surveys are the most commonly used methods. However, these methods are not straightforward or cost-effective and the short-life nature of oil slicks may hinder the implementation of these methods. In this study, validation of oil detection mainly relies on documented events of oil pollution in the gulf region, records from local agencies (such as Environmental Agency of Abu Dhabi, MEMAC, and ITOPF), and reports in published literatures. For example, the oil slicks found in Landsat 7 ETM + image collected on May 26 2000 have been verified by in situ survey conducted by . Another oil pollution event reported in  has also been confirmed by their in situ campaigns. According to a report issued by MEMAC, a major oil spill was recorded in the Arabian Gulf on August 6 2013, which caused the oil slicks detected by Landsat 8 and Aqua on August 17 2013.
3.2.2 Thermal observations
Thick oil slicks tend to appear warmer than intermediate-thickness ones when observed with IR sensors. As stated in [13,16], oil slicks with thickness of > 150 µm appear warmer than adjacent areas and thinner slicks appear cooler during daytime while at night a thick spill can appear cooler than surrounding waters due to the fact that the spill can release heat quicker. Salisbury et al.  found that an emissivity (ԑ) difference of 0.01 between seawater and crude oil can produce an apparent difference of 0.6 °C in brightness temperature (TB) at the thermal IR band of Landsat TM at room temperature. Brightness temperature is linked to surface temperature (TS) via emissivity (ԑ) with TB = TS *ԑ0.25. In this study, Landsat 7 ETM + data in the IR band (band 6) were also examined to infer the temperature difference between oil-spilled and oil-free areas.
3.3 Hydrodynamic modeling
The consistency between detected oil slicks observed from consecutive satellite images and from hydrodynamic models is an indicator of the reliability of the oil detection technique. Hydrodynamic models can augment our capability to track oil pollution by simulating their dynamics from consecutive satellite images. The positions of oil slicks can be reinitialized as soon as a new satellite update is available. In cases of persistent cloud coverage and heavy dust, hydrodynamic models can complement satellite imagery and simulate the propagation of oil slicks and their possible landfall. In addition, such models in their three-dimensional configurations can simulate dynamics of submerged oil pollution which cannot be detected by satellite imagery.
In this paper, ocean circulation and wind data were used to complement the information on oil presence that is inferred from satellite imagery, track potential movements of oil slicks, and help in predicting oil pollution impact. The HYbrid Coordinate Ocean Model (HYCOM), a primitive equation ocean general circulation model [45,46], has been used in numerous biogeochemical studies [46, and references therein]. Surface ocean current data were downloaded from HYCOM global data archive. Details can be found at www.hycom.org/dataserver. These data have a spatial resolution of ~9 km and span the period from 2003 to present. Surface ocean current data corresponding to the time of satellite passage when oil slicks were detected (i.e., 10:46:57 local time (LT) for MERIS on October 21 2007 and 10:48:09 LT for Landsat 8 on August 26 2013) were acquired. Blended Sea Winds from NOAA National Climatic Data Center provide globally gridded ocean surface wind data with a spatial resolution of 0.25 degree and with a temporal resolution of 6 hours. They were also used for the purpose of oil tracking. Wind data determined around the overpass times of MERIS on October 21 2007 and Landsat 8 on August 26 2013, respectively, were extracted for the gulf region.
4.1 Satellite observation of oil slicks
The sample of images analyzed in this study that were obtained from different sensors under different solar/viewing geometry included cases with and without sunglint. Both cases were addressed in this study and the difference in the sensitivity of different sensors to oil spills under both conditions was investigated.
Figure 2 shows an example of features observed by consecutive satellite images over the same region on the same day (i.e. May 26 2000). The recurrence of the observed features in the same areas in different satellite images may imply that they are likely to be caused by an oil spill. Figure 2(a) shows the scene observed by Landsat 7 ETM + on May 26 2000. This case was under sunglint contamination. Two areas, namely area 1 and area 2 shown in Fig. 2(a), with dark features can be noticed. The occurrence of oil slick in area 1 has been proven by in situ survey conducted by . The spatial texture of area 2 indicated that area 2 could be attributed to an oil slick although there was no in situ validation to confirm it. However, it is important to add that areas 1 and 2 had high marine traffic. Area 3, as outlined with a red ellipse in Fig. 2(a), showed bright contrast against adjacent regions and was also attributed to oil pollution . The inset figure in Fig. 2(a) shows the thermal IR band image for the oil polluted area and indicates that the oil slick had lower temperature than surrounding waters. The slick of lower temperature found in area 3 seems to be related to decreased emissivity of the oil layer relative to water. Areas 1 and 2 were undetectable at the corresponding IR channels and this is in good agreement with previous findings . Figure 2(b) shows MODIS/Terra image on the same day over the same region as in Fig. 2(a) 41 minutes after Landsat 7 ETM + . Areas 1 and 3 are difficult to identify without prior knowledge of the oil polluted areas observed by Landsat 7 ETM + , and area 2 cannot be detected probably due to the effects of cloud cover and the coarser resolution of MODIS.
The Landsat 7 ETM + Rrc spectra for two transects through the representative oil slicks withdark and bright contrasts (green lines in Fig. 2(a)) are plotted in Figs. 2(c) and 2(d), and Rrc at the five bands against the pixel number along the transect is plotted in Figs. 2(e) and 2(f). The oil slicked area 1 (Figs. 2(c) and 2(e)) generally showed lower Rrc than surrounding waters while the oil covered area 3 (Figs. 2(d) and 2(f)) demonstrated higher Rrc than surrounding waters at all bands. This suggests that the observed feature with bright contrast may be weathered oil suspended in the water column while the dark contrast was due to oil film floating on the sea surface . Simulation results from [47–49] showed that oil films presented dark contrast while oil in water emulsions displayed bright contrast. With respect to area 1, the Rrc spectral shape for the oiled area did not demonstrate any special patterns. As for area 3, some of the Rrc spectra showed peaks between 483 and 660 nm and others showed monotonically ascending trends. For both oil slicked areas 1 and 3, Rrc spectra indicated significant fluctuations. In contrast, Rrc spectra for the oil-free areas showed relatively consistent behavior.
Another example of oil pollution in the absence of sun glint is shown in Fig. 3. Both MERIS (Fig. 3(a)) and MODIS/Terra (Fig. 3(b)) images for October 21 2007 captured features in the gulf region. They exhibited dark contrasts relative to the surrounding waters. However, MODIS/Aqua data (not shown) on the same day did not show the features over the region and this may suggest that the features were not due to variations in optical properties in the water column. Therefore, the features could not be caused by an algal bloom. Furthermore, examining the wind data, and the location and the spatial texture of the affected area implied that these features were probably attributed to oil slicks, which has been validated by field surveys reported by . The medium resolution MODIS/Terra multispectral Rrc is shown in Fig. 3(c) and Rrc for bands 1-7 against the pixel number along the transect is representatively shown in Figs. 3(d)-3(f). Rrc spectra for both oiled and unoiled areas presented peaks at 469 or 488 nm. The oil covered areas had lower Rrc than surrounding waters at all bands except 412 nm. Rrc(412) for the oiled area cannot be distinguished from the unoiled area and this may be related to the strong effects of dust at 412 nm since the atmospheric dust level over the gulf region is high with a total average dust flux of 6.385 g m−2 day−1 for an average of 33 days per year . The similar spectral shape near the chlorophyll-a fluorescence peak for oiled and unoiled areas revealed that chlorophyll-a concentration was low during the period, which corroborated the finding by  that chlorophyll-a showed a subtropical seasonal trend with minimum in spring-summer and maximum in winter. For the IR bands, Rrc values were smaller in the oiled area than in the unoiled area. Especially at 2130 nm Rrc of the oil slicked area was even one fold smaller than that of adjacent oil-free waters. Similar to the Rrc behavior found in Fig. 2, the Rrc spectra for the oil contaminated areas observed in Fig. 3 also displayed high variability (Fig. 3(c)) when compared to oil-free areas. And this is likely due to the variable fractions of oil in the affected areas. Compared with MODIS/Terra derived Rrc spectra, the Landsat 7 ETM + collected Rrc spectra showed less modulation, which is probably correlated with sub-pixel variations . A pixel from the MODIS measurement (250 m resolution) encompasses ~8 x 8 Landsat (30 m resolution) pixels. As a result of the heterogeneous oil slicks and coarser resolution, MODIS Rrc spectra presented relatively large modulations.
A 10-mile long oil spill was reported on August 6 2013. Figure 4(a) presents the Landsat 8 RGB image collected on August 17 2013 with cloud-free conditions. A long oil slick characterized by both dark and bright contrasts against surrounding oil-free waters can be seen. Some small patches of oil spills can also be noticed, which also showed bright and dark features in the image. The fact that oil slicks showed dark and bright contrast in the same RGB image can be related to the variable sunglint level and spatial heterogeneity of oil. These recognized features indicated similar Rrc spectral behaviors (data not shown) to those found in Fig. 2. Some dark features, rather than oil slicks, can also be seen in the RGB image, and they can be probably caused by internal wave, ship wake, natural slick, etc. MODIS/Aqua passed over the same region after 2.5 hours. The oil slicks indicated only bright contrast relative to adjacent unoiled areas, and this was probably caused by higher sunglint conditions for MODIS/Aqua than that for Landsat 8 during their different overpass time. The faint appearance of the slicks in MODIS/Aqua imagery could be related to the coarser resolution of MODIS/Aqua than Landsat 8. A visual inspection of the full pass of MODIS/Aqua imagery over the region for August 17 2013 indicated that the oil slick dispersed southeastward to the shallow areas of the middle gulf region and extended to >100 miles long.
For all of these three oil pollution events delineated above, oil slicks showed dark or bright features in satellite derived quasi true color (RGB) images as a consequence of the variable solar and viewing geometry, and oil types and thickness. With respect to the first two factors, viewing geometry seemed to have more effects on the appearance of oil slicks in satellite images than solar geometry.
4.2 Differentiating oil slicks from algal blooms
Algal blooms indicate dark features in satellite derived true color images due to the strong absorption of phytoplankton and/or CDOM [53,54]. Therefore, they can be easily confused with oil slicks. An effective approach to distinguish oil slicks from algal blooms is of significant importance.
Efforts have been made for detecting algal blooms using various satellite products [54–56]. However, oil contamination stalls the applicability of the standard atmospheric correction scheme, based on which over 90% of the satellite received signal from aerosol was removed. Thus, ocean color products over oil contaminated areas can be misinterpreted. There are two approaches to address the issue. The first one is to develop region-specific atmospheric correction algorithms. The second one is to avoid correcting the contribution from aerosols. Considering the requirements for ancillary information about oil with respect to the first approach, the second one could be a better alternative for regular operation.
FAI is based on Rayleigh corrected reflectance, and the band-subtraction designs make it less sensitive than band-ratio algorithms to changes in environmental and observation conditions (i.e. aerosol type and thickness, solar and viewing geometry, and sun glint). It has been successfully utilized for recognizing and delineating algal blooms in various waters . MODIS FAI image for October 21 2007 is shown in Fig. 5. Oil slicks exhibited lower FAI values than neighboring waters. In contrast, bloom waters indicated higher FAI values than non-bloom waters, as suggested by previous studies . In order to further validate the feasibility of FAI for identifying oil slicks, Landsat 8 data were examined. As shown in Fig. 6(a), dark features observed on August 26 2013 can be apparently discernible against the adjacent areas and a small patch close to the oil platform with bright features can also be seen clearly. Considering the spatial texture and their proximity to an oil platform, the features were very likely caused by oil slicks leaking from the neighboring oil platform. The corresponding FAI image (Fig. 6(b)) showed lower values in the oil slicked areas relative to the surrounding waters. Therefore, FAI can be exploited to separate oil slicks from algal blooms.
4.3 Oil tracking with ocean circulation and wind data
The measurements by MERIS and MODIS/Terra shown in Figs. 3(a) and 3(b) were ~20 minutes apart and the oil slicks moved ~500 m. Although the movement of the oil slicks was slow, comparisons of the two images indicate that the oil slicks were moving towards the west coast of the UAE. This movement pattern of the oil slicks is consistent with results from the surface current (Fig. 7(a)) and wind data (Fig. 7(c)), both of which moved southeast. Driven by the surface current and the wind with an estimated total daily average velocity of 0.42 m s−1, it would take ~23 hours for the rightmost patch in Figs. 3(a) and 3(b) to reach the west coast of the UAE. Similarly, the movement of the oil slicks observed by Landsat 8 on August 17 2013 in Fig. 6 can be predicted and forecasted. Figures 7(b) and 7(d) show the surface current and wind during the overpass of Landsat 8 on August 17 2013. Inspection of surface current data showed that after August 6 2013 the oil slicks were entrained into an eddy with a cyclonic rotation. A clockwise spinning eddy can be identified from the current vectors for August 17 2013 and this can cause the movement of the oil slicks to the shallow area of the middle gulf region. The interpretation of the flow patterns of the oil slicks found on August 17 2013 is not straightforward since the surface current moved southeast while the surface wind blew westward. For example, as shown in in Fig. 4, the oil slick outlined with red ellipses moved southeast while another oil slick outlined with green ellipse moved southwest. Thiswas probably related to the combined effects of ocean current and wind. Following this procedure, potential oil pollution alarm can be issued in a forecast mode. Then timely actions can be taken to minimize the potential damages. This is of great value for management policy-making.
The obtained results in this study demonstrated that satellite sensors play significant roles in oil pollution response since they can provide preliminary spill assessment for remote locations and synoptic scale data. However, limitations exist from these synoptic techniques, including revisit frequency and clear sky requirements. The trade-off between coarser (finer) spatial resolution and larger (smaller) swath width, such as MODIS and Landsat, is well known. Since oil slicks exhibit sub-pixel heterogeneity and usually stretch over few km2, coarser resolution sensors would lead to biases. On the other hand, much of oil pollution could be missed by sensors with fine-scale resolution but infrequent revisit time. Hence, ensemble of multi-source satellite data is a fruitful strategy.
Because oil slicks evolve on hourly to daily time scales, frequent false positives may happen from multi-day satellite overpasses, especially given the paucity of trained observers, and limit utility of satellite observations . Therefore, there are needs to take ancillary data, such as wind and wave height, into consideration for oil slick identification and classification in automatic algorithms [57–59] which should continuously monitored by experienced analysts because of the complexity of the processes and the large number of factors influencing it.
Atmospheric correction has been a preeminent topic, especially for turbid coastal waters, and a lot of efforts have been carried out [60–63]. The operational atmospheric correction algorithm assumes that water-leaving radiance in the NIR region is negligible. However, this assumption progressively degrades with increasing turbidity. The atmospheric dust level over the gulf region is high, which has posed major challenges for atmospheric correction. The default thresholds of cloud albedo and aerosol optical depth at 865 nm (AOD_865) in SeaDAS are 0.027 and 0.3, respectively, and thus all pixels with cloud albedo > 0.027 or AOD_865 > 0.3 will be masked during processing from level 1 to level 2. Ad hoc atmospheric correction scheme is required if level 2 products are needed for application in detecting oil spills. Using level 2 products would also suffer from algorithms for retrieving water properties, especially for shallow coastal waters. For example, the chlorophyll-a concentration derived from the operational algorithm is actually combined effects of all optically active water constituents. Alawadi et al.  used chlorophyll-a product to differentiate between oil spills and look-alikes. Mixed results were obtained in the case of chlorophyll-a concentration, and its usefulness to separate oil slicks from surrounding waters was uncertain. However, in this study, FAI, which is insensitive to aerosol type and can successfully remove cloud contamination, is exploited. Therefore atmospheric correction scheme has negligible effect on the results. It can be concluded that discrimination of oil slicks from algal blooms can be effectively achieved by means of the FAI technique.
In forthcoming studies, algorithms will be developed to automatically detect oil slicks. Thus, oil slicks can be routinely mapped. Along with coastal hydrodynamic models, their motion and potential impacted areas can be forecasted and predicted. It is worth noting that oil detection techniques in this study can also be expanded to other satellite sensors such as Hyperspectral Imager for the Coastal Ocean (HICO), Visible Infrared Imaging Radiometer Suite (VIIRS), Geostationary Ocean Color Imager (GOCI), Spinning Enhanced Visible and Infrared Imager (SEVIRI), and Sentinel 3 that has similar specifications to MERIS and is planned for launch in mid-2015. By merging multi-sensor observations, the temporal coverage for oil detection can be larger and thus reliability of oil slick identification from satellite imagery can be increased.
In this study, oil pollution was detected with MODIS, Landsat, and MERIS images in the Arabian Gulf. Oil slicks in the presence and absence of sun glint were examined. The appearance of oil slicks, either darker or brighter than the surrounding oil-free water, was dependent upon viewing geometry, and oil types and thickness. The spectral signature of oil-free pixels showed higher consistency when compared to oil-covered pixels as the satellite derived reflectance was significantly variable when oil was present and this may be attributed to the variability of oil fractions among the pixels. FAI images showed potentials for differentiating oil slicks from algal bloom mats with the former showing lower FAI while the latter showing higher FAI than surrounding waters. Ocean circulation and wind data were utilized for oil tracking.
Oil detection and monitoring are of great importance for managing marine resources and minimizing its potential adverse effects. Remote sensing represents a significant element to capture oil occurrence and aid in surveillance of its extent and development. With the advent of future satellite missions like sentinel-3 and GOCI-2, remote sensing will augment our ability to promptly detect oil pollution. Integration of multi-source satellite data sets can provide guidance on potential oil identification for further investigation. By marshaling an ensemble of satellite observation and ocean circulation model, an effective warning and forecasting system can be established for oil pollution response.
This study was funded by Masdar Institute of Science and Technology, Abu Dhabi (UAE), and by the U.S. NASA through its Ocean Biology and Biogeochemistry program, and by the BP/Gulf of Mexico Research Initiative through C-IMAGE. We would like to thank the NASA OBPG group for providing MODIS Aqua and Terra images and thank the USGS for providing Landsat 7 ETM + and Landsat 8 data. We also want to give our appreciations to two anonymous reviewers for their suggestions to improve the manuscript.
References and links
1. J. R. Jensen, E. W. Ramsey III, J. M. Holmes, J. E. Michel, B. Savitsky, and B. A. Davis, “Environmental sensitivity Index (ESI) mapping for oil spills using remote sensing and geographic information system technology,” Int. J. Geogr. Info. Syst. 4(2), 181–201 (1990). [CrossRef]
2. J. Lu, “Marine oil spill detection, statistics, and mapping with ERS SAR imagery in south-east Asia,” Int. J. Remote Sens. 24(15), 3013–3032 (2003). [CrossRef]
3. J. Lu, H. Lim, S. C. Liew, M. Bao, and L. K. Kwoh, “Ocean oil pollution mapping with ERS synthetic aperture radar imagery,” in Proceedings of the International Geoscience and Remote Sensing Symposium (Hamburg, 1999), pp. 212–214.
4. H. A. Espedal and O. M. Johannessen, “Detection of oil spills near offshore installations using synthetic aperture radar (SAR),” Int. J. Remote Sens. 21(11), 2141–2144 (2000). [CrossRef]
5. C. Brekke and A. H. Solberg, “Oil spill detection by satellite remote sensing,” Remote Sens. Environ. 95(1), 1–13 (2005). [CrossRef]
6. US Coast Guard, Update of Inputs of Petroleum Hydrocarbons into the Oceans due to Marine Transportation Activities, National Research Council (National Academy Press, Washington, D. C, 1990).
7. P. Pavlakis, A. Sieber, and S. Alexandry, “Monitoring oil-spill pollution in the Mediterranean with ERS SAR,” ESA Earth Observation Quarterly 52 (1996).
8. European Space Agency, “Oil pollution monitoring. ESA Brochure,” ERS and Its Applications-Marine BR-128, 1 (1998).
9. MEMAC (Marine Emergency Mutual Aid Center), “Oil Spill Incidents in ROPME Sea Area (1965-2002),” MEMAC, Bahrain (2003).
10. M. Fingas and C. E. Brown, “Review of oil spill remote sensing,” Spill Sci. Technol. Bull. 4(4), 199–208 (1997). [CrossRef]
11. M. Fingas and C. E. Brown, Oil spill Remote Sensing: A Review (Gulf Professional Publishing, USA, 2011), Chapter 6.
12. M. Fingas and C. E. Brown, “Review of oil spill remote sensors,” in Proceedings of the 7th international conference on Remote Sensing for Marine and Coastal Environments, (ERIM International, Miami, Florida, 2002).
13. I. Leifer, W. J. Lehr, D. Simecke-Beatty, E. Bradley, R. Clark, P. Dennison, Y. Hu, S. Matheson, C. E. Jones, B. Holt, M. Reif, D. A. Roberts, J. Svejkovsky, G. Swayze, and J. Wozencraft, “State of the art satellite and airborne marine oil spill remote sensing: Application to the BP Deepwater Horizon oil spill,” Remote Sens. Environ. 124, 185–209 (2012). [CrossRef]
14. A. H. S. Solberg, “Remote Sensing of Ocean Oil-Spill Pollution,” Proc. IEEE 100(10), 2931–2945 (2012). [CrossRef]
15. C. Hu, X. Li, and W. G. Pichel, “Detection of oil slicks using MODIS and SAR imagery,” in Handbook of Satellite Remote Sensing Image Interpretation: Applications for Marine Living Resources Conversation and Management, J.V. Morales, T.P. Stuart, and S. Sathyendranath, eds. (EU PRESPO and IOCCG, Dartmouth, Canada, 2011).
16. W. Y. Tseng and L. S. Chiu, “AVHRR observations of Persian Gulf oil spills,” in Proceedings of the International Geoscience and Remote Sensing Symposium (Pasadena, California, 1994), pp. 779–782. [CrossRef]
17. D. Casciello, T. Lacava, N. Pergola, and V. Tramutoli, “Robust satellite techniques for oil spill detection and monitoring using AVHRR thermal infrared bands,” Int. J. Remote Sens. 32(14), 4107–4129 (2011). [CrossRef]
18. C. S. L. Grimaldi, D. Casciello, I. Coviello, T. Lacava, N. Pergola, and V. Tramutoli, “An improved RST approach for timely alert and Near Real Time monitoring of oil spill disasters by using AVHRR data,” Nat. Hazards Earth Syst. Sci. 11(5), 1281–1291 (2011). [CrossRef]
19. C. M. Bentz and J. O. Barros, “A multi-sensor approach for oil spill and sea surface monitoring, in southeastern Brazil,” Int. Oil Spill Conf. Proc. 2005, 703–706 (2005). [CrossRef]
20. M. Amado, G. De Carolis, V. De Pasquale, and G. Pasquariello, “Detection and tracking oil slicks on sun-glittered visible and near infrared satellite imagery,” Int. J. Remote Sens. 30(24), 6403–6427 (2009). [CrossRef]
21. F. Alawadi, C. Amos, V. Byfield, and P. Petrov, “The application of hyperspectral image techniques on MODIS data for the detection of oil spills in the RSA,” Proc. SPIE 7110, 71100Q (2008), doi:. [CrossRef]
22. B. Bulgarelli and S. Djavidnia, “On MODIS retrieval of oil spill spectral properties in the marine environment,” IEEE Geosci. Remote Sens. Lett. 9(3), 398–402 (2012). [CrossRef]
23. C. Hu, F. E. Müller-Karger, C. J. Taylor, D. Myhre, B. Murch, A. L. Odriozola, and G. Godoy, “MODIS detects oil spills in lake Maracaibo, Venezuela,” Eos Trans. AGU 84(33), 313–319 (2003). [CrossRef]
24. C. Hu, X. Li, W. G. Pichel, and F. E. Muller-Karger, “Detection of natural oil slicks in the NW Gulf of Mexico using MODIS imagery,” Geophys. Res. Lett. 36(1), L01604 (2009), doi:. [CrossRef]
25. G. De Carolis, M. Adamo, and G. Pasquariello, “Thickness estimation of marine oil slicks with near-infrared MERIS and MODIS imagery: The Lebanon oil spill case study,” in Proceedings of the International Geoscience and Remote Sensing Symposium, (Munich, 2012), pp. 3002–3005. [CrossRef]
26. C. Bentz and F. P. de Miranda, “Application of remote sensing data for oil spill monitoring in the Guanabara Bay, Rio de Janeiro, Brazil,” in Proceedings of the International Geoscience and Remote Sensing Symposium, (Sydney, NSW, 2001), pp. 333–335. [CrossRef]
27. S. Essa, H. Harahsheh, M. Shiobara, and T. Nishidai, Operational Remote Sensing for Detection and Monitoring of Oil Pollution in the Arabian Gulf: Case Studies from the United Arab Emirates (Elsevier B. V., Amsterdam, the Netherlands, 2005), Chap. 3.
28. M. S. Massoud, F. Al-Abdali, and A. N. Al-Ghadban, “The status of oil pollution in the Arabian Gulf by the end of 1993,” Environ. Int. 24(1-2), 11–22 (1998). [CrossRef]
29. A. A. Elhakeem, W. Elshorbagy, and R. Chebbi, “Oil spill simulation and validation in the Arabian (Persian) Gulf with special reference to the UAE coast,” Water Air Soil Pollut. 184(1-4), 243–254 (2007). [CrossRef]
30. Y. Liu, R. H. Weisberg, C. Hu, and L. Zheng, “Tracking the Deepwater Horizon oil spill: A modeling perspective,” Eos Trans. AGU 92(6), 45–46 (2011). [CrossRef]
31. Y. Liu, R. H. Weisberg, C. Hu, and L. Zheng, “Combing numerical ocean circulation models with satellite observations in a trajectory forecast systems: A rapid response to the Deepwater Horizon oil spill,” Proc. SPIE 8030, 80300K (2011), doi:. [CrossRef]
32. Y. Liu, R. H. Weisberg, C. Hu, and L. Zheng, “Satellites, model combine to track Deepwater Horizon oil spill,” SPIE Newsroom, (2011).
33. Y. Liu, R. H. Weisberg, C. Hu, C. Kovach, and R. Riethmüller, Evolution of the Loop Current System during the Deepwater Horizon Oil Spill Event as Observed with Drifters and Satellites, (AGU, 2011).
34. W. Alpers, V. Wismann, R. Theis, H. Huhnerfuss, N. Bartsch, J. Moreira, and J. Lyden, “The damping of ocean surface waves by monomolecular sea slicks measured by airborne multi-frequency radars during the saxon-fpn experiment,” in Geoscience and Remote Sensing Syposium, 1991. IGARSS ’91. Remote Sensing: Global Monitoring for Earth Management, International, 4, pp. 1987, 1990, 3–6 Jun 1991.
35. C. Hu, “A novel ocean color index to detect floating algae in the global oceans,” Remote Sens. Environ. 113(10), 2118–2129 (2009). [CrossRef]
36. P. G. Jacob and A. Al-Muzaini, “Marine plants of the Arabian Gulf: a literature review,” Kuwait Institute for Scientific Research, Safat, Kuwait (KISR Tech Report No. 3426) (1990).
37. ROPME (Regional Organization for the Protection of the Marine Environment). 1999. Regional report of the state of the marine environment, Regional Organization for the Protection of the Marine Environment, State of Kuwait, March 1999.
38. A. R. G. Price, C. R. C. Sheppard, and C. M. Roberts, “The Persian Gulf: its biological setting,” Mar. Pollut. Bull. 27, 9–15 (1993). [CrossRef]
39. P. F. Sale, D. A. Feary, J. A. Burt, A. G. Bauman, G. H. Cavalcante, K. G. Drouillard, B. Kjerfve, E. Marquis, C. G. Trick, P. Usseglio, and H. Lavieren, “The growing need for sustainable ecological management of marine communities of the Persian Gulf,” Ambio 40(1), 4–17 (2011). [CrossRef] [PubMed]
41. K. Baith, R. Lindsay, G. Fu, and C. R. McClain, “Data analysis system developed for ocean color satellite sensors,” EOS 82(18), 202 (2001). [CrossRef]
42. C. Hu, Z. Lee, R. Ma, K. Yu, D. Li, and S. Shang, “Moderate Resolution Imaging Spectroradiometer (MODIS) observations of cyanobacteria blooms in Taihu Lake, China,” J. Geophys. Res. 115(C4), C04002 (2010). [CrossRef]
43. C. Jackson, “Internal wave detection using the Moderate Resolution Imaging Spectroradiometer (MODIS),” J. Geophys. Res. 112(C11), C11012 (2007), doi:. [CrossRef]
44. J. W. Salisbury, D. M. D’Aria, and F. F. SabinsJr., “Thermal infrared remote sensing of crude oil slicks,” Remote Sens. Environ. 45(2), 225–231 (1993). [CrossRef]
45. R. Bleck, “An oceanic general circulation model framed in hybrid isopycnic-Cartesian coordinates,” Ocean Model. 4(1), 55–88 (2002). [CrossRef]
46. E. P. Chassignet, H. E. Hurlburt, E. J. Metzger, O. M. Smedstad, J. A. Cummings, A. J. Wallcraft, C. Lozano, H. L. Tolman, A. Srinivasan, S. Hankin, P. Cornillon, R. Weisburg, A. Barth, R. He, F. Werner, and J. Wilkin, “U.S. GODAE: Global Ocean Prediction With the Hybrid Coordinate Ocean Model (HYCOM),” Oceanogr. 22(2), 48–59 (2009).
47. F. Carnesecchi, V. Byfield, P. Cipollini, G. Corsini, and M. Diani, “An optical model for the interpretation of remotely sensed multispectral images of oil spill,” Proc. SPIE 7105, 710504 (2008), doi:. [CrossRef]
50. W. Hamza, M. R. Enan, H. Al-Hassini, J.-B. Stuut, and D. de-Beer, “Dust storms over the Arabian Gulf: a possible indicators of climate changes consequences,” Aquat. Ecosyst. Health Manage. 14(3), 260–268 (2011). [CrossRef]
51. N. P. Nezlin, I. G. Polikarpov, F. Y. Al-Yamani, D. V. Subba Rao, and A. M. Ignatov, “Satellite monitoring of climatic factors regulating phytoplankton variability in the Arabian (Persian) Gulf,” J. Mar. Syst. 82(1-2), 47–60 (2010). [CrossRef]
53. J. Zhao, C. Hu, B. Lapointe, N. Melo, E. M. Johns, and R. H. Smith, “Satellite-observed black water events off Southwest Florida: Implications for coral reef health in the Florida Keys National Marine Sanctuary,” Remote Sens. 5(1), 415–431 (2013). [CrossRef]
55. J. Gower, S. King, G. Borstad, and L. Brown, “Detection of intense plankton blooms using the 709 nm band of the MERIS imaging spectrometer,” Int. J. Remote Sens. 26(9), 2005–2012 (2005). [CrossRef]
56. R. P. Stumpf, M. E. Culver, P. A. Tester, M. Tomlinson, G. J. Kirkpatrick, B. A. Pederson, E. Truby, V. Ransibrahmanakul, and M. Soracco, “Monitoring Karenia brevis blooms in the Gulf of Mexico using satellite ocean color imagery and other data,” Harmful Algae 35, 1–14 (2003).
57. A. H. S. Solberg, G. Storvik, R. Solberg, and E. Volden, “Automatic detection of oil spills in ERS SAR images,” IEEE Trans. Geosci. Rem. Sens. 37(4), 1916–1924 (1999). [CrossRef]
58. M. Marghany, “Genetic algorithm for oil spill automatic detection from Envisat satellite data. Computational Science and Its applications-ICCSA,” Lect. Notes Comput. Sci. 7972, 587–598 (2013). [CrossRef]
59. M. Marghany and M. Hashim, “Comparative algorithms for automatic detection of oil spill in multisar of RADARSAT-1 SAR and ENVISAT data,” in Proceedings of IEEE Conference on Signal and Image Processing Applications (Kuala Lumpur, 2011), pp. 559–562. [CrossRef]
60. C. Goyens, C. Jamet, and T. Schroeder, “Evaluation of four atmospheric correction algorithms for MODIS-Aqua images over contrasted coastal waters,” Remote Sens. Environ. 131, 63–75 (2013). [CrossRef]
62. M. Wang, S. Son, and W. Shi, “Evaluation of MODIS SWIR and NIR-SWIR atmospheric correction algorithms using SeaBASS data,” Remote Sens. Environ. 113(3), 635–644 (2009). [CrossRef]