Keyword: Oil slicks, detection and mapping, occurrence intensity, ERS SAR

Abstract
With a large coverage, short repeativity, day and night, and all weather capability, ERS SAR imagery has been proven to be a very effective tool for ocean oil spill monitoring. This paper discusses ocean oil pollution detection and mapping in the Southeast Asian waters by using ERS SAR data. Possible oil slicks are detected first from each image and a normalized average oil slick occurrence intensity is then computed after all the available multi-temporal scenes in an ERS frame have been studied. The scene in which most slicks have been detected in a frame is then selected for re-examination. All oil slicks detected on this image are then further classified into either high or low probability categories. A new normalised maximum oil slick occurrence intensity is computed for each of the total and high probability oil slicks. Spatial distribution maps of the average and worst oil pollution situation for the Southeast Asian waters are then derived.

Introduction
Oil pollution surveillance and mapping over a large ocean area has become easy with the launch of remote sensing satellites. Space-borne synthetic aperture radar (SAR) has, in particular, become a very popular tool for ocean oil slick monitoring due to its wide area coverage under all weather conditions, day and night. Some very successful examples of ERS SAR applications in marine oil spill detection have been documented, such as the near-real-time oil spill monitoring in Norway (Pedersen et al, 1995) and a successful anti-oil spill law prosecution case in Singapore (The Straits Times, 1997). In comparison with the already reported oil spill monitoring case studies by space-borne SAR data, spatial oil spill distribution mapping is still under exploited, especially over a large ocean area. A study on ocean oil spill statistics using the ERS SAR imagery in the Southeast Asian waters is reported (Lu et al, 1999). It is probably the first attempt in marine oil pollution mapping with satellite data over a large area. The present study first compiles the oil spill statistics extended over a much larger area in the Southeast Asia region, and then maps the worst oil pollution scenario in the region.

Oceans in the Southeast Asia are among the busiest in the world. Several major international shipping routes pass through this region connecting the Pacific Ocean to the Indian Ocean. Oil is transported mainly from the Middle East to Japan, United States and the Southeast Asian countries. This region is also rich itself in ocean oil deposits with intensive oil exploitation and inter-regional transport activities already in place. Large oil/gas fields can be found in the Gulf of Thailand, the South China Sea off the Brunei coast, and so on. Mapping the spatial distribution of oil pollution is the first critical step towards the protection of the region’s marine environment.

The study area lies between longitudes 90° E and 125° E, latitudes 10° S and 20° N. Three years of ERS-1/2 data from September 1995 to September 1998 archived in CRISP’s database was used. This includes a total of 657 frames with 5029 scenes. Since thousands of scenes would be required, use of high-resolution PRI products would impose a heavy load on the processing facility and also slow down the slick detection process. Thus, our in-house quick-look products were used instead. With 1000x1013 pixels per scene of 100km x 100km, this version of quick-look images is adequate for fast oil slick detection (ESA, 1998; Lu & Bao, 1999).

Oil Slick Detection and Mapping with ERS SAR
Oil films are very effective in damping wind-generated gravity-capillary short waves on the sea surface and hence appear dark against a brighter background on SAR images. The detectability depends on ocean surface wind speed. If wind speed is too low (typically below 2 to 3 m/s), the sea surface background does not have sufficient roughness to contrast with that of oil films. On the other hand, if wind speed is too high (typically above 15 m/s), oil slicks can be dispersed by the surface waves and disappear below the sea surface.

Not all-dark sea surface areas in SAR images are real oil slicks. Sea surface may also appear dark due to natural slicks, low wind, certain atmospheric and oceanic phenomena and other reasons. Contextual information such as the shapes and locations of the suspected oil slicks can also help to avoid false detection. The main challenge for successful oil slick detection depends on the exclusion of all those possible look-alike slicks. Examples of some typical signatures of oil slicks on ERS SAR images are shown in Figure 1.


Figure 1. Examples of various oil slicks on ERS SAR images:
a). a long linear slick from a moving ship,
b). slicks from offshore oil rigs,
c). slicks probably from natural seepage, and
d). slicks along a major shipping route
Mapping of the average oil pollution distribution is performed in three steps. First, possible oil slicks in each scene are visually identified and discriminated against other look-alike features. Next, these identified oil slicks are further classified into one of the three categories based on their shapes, namely linear, curved and patchy. In addition, the size of each slick is also estimated and grouped into one of four categories, namely “<1km”, “1-5km”, “5-10km”, and "³10km".Lastly, a normalized average oil slick occurrence intensity is then computed for each frame after all the available scenes in this frame have been examined. This pollution index is defined to be the average slick number detected each time per 10,000 km² sea surface. For a given frame, it is expressed by the equation:

where N_i is the total number of slicks detected in the i-th scene, f_sea is the sea surface fraction of the frame, and f_i is the scene fraction of the i-th scene.