|
Being able to detect and identify offshore oil slicks and in particular oil pollutions is important to reduce the detrimental effect of pollution on ecosystems. For decades, SAR images have been used to detect pollutions. In some cases, SAR polarimetric derived parameters are used to improve detection or discrimination. However, the acquisition of multi-polarization channels is at the expense of the extent of the monitored area. Also, some papers question the effectiveness of polarization derived parameters for detection or identification. SAR single polarization is quite efficient to detect hydrocarbons but there are many look-alikes and in general SAR is not much sensitive to oil thickness which is a very important element for pollution remediation. Optical data on the other side have been shown to be sensitive to oil thickness and to some extent to the type of oil and the physical state of the oil: emulsion vs plain film. The main drawback of optical data is that they will not show the ocean surface when clouds are presents or during the night, which is a big constrain in tropical areas, Northern areas and winter high latitude spots. But when clouds are not hiding the area of interest, optical data offer good insight into the nature of the sea surface. Lab measurements helps us to find means to detect, identify, and qualitatively quantify oil on water. But in real life other features will impact the detection and identification: surface ripples, clouds, algae, smoke, shadows, glint. In this paper we will use airborne hyperspectral proprietary data and satellite multispectral data, to understand the spectral signature of oil and non-oil slicks, the impact of oil thickness and a combination of spectral indices to discriminate oil from look-alike on different case studies. When a big pollution occurs, planes are mobilized carrying optical, radar, thermal as well as UV systems which together will discriminate oil slicks from look-alikes. But in most cases only open-source satellite data will be available and it will be necessary to use them at best to identify and characterize the floating slicks. VV single pol SAR data remain the best images to detect hydrocarbons but with some potential false alarms. On the optical side, the hydrocarbon index and Area1700 index are good at detecting thick hydrocarbons but cannot be used with multispectral data. The Fluorescence index can discriminate hydrocarbons but rejection of clouds, and infrastructure require additional indices or processing. Some Algae are potentially the most difficult features to discriminate from hydrocarbon. A new index is proposed to discard them. The objective is to detect hydrocarbons with the minimum of in-situ knowledge.
|
