This chapter covers oil spill detection and classification with fully and compact synthetic aperture radar. First, it introduces fundamental theory and techniques of oil spills detection by SAR, as well as the paradigm for fully and compact polarimetric SAR signal processing and features extraction. Next, statistical distances are proposed to evaluate the features that derived from the simulated compact polarimetric SAR data by either pseudo quad-pol reconstruction algorithm or directly from the universal feature extraction methods. Then several widely used supervised classifiers are applied on fully and compact polarimetric SAR features of different modes to evaluate their performance in distinguishing mineral oil between a clean sea surface and biogenic look-alikes. Experiments are conducted on airborne UAVSAR and spaceborne RADARSAT-2 data, which was obtained during the Deepwater Horizon oil spill 2010 in the Gulf of Mexico and Norwegian oil-on-water exercise 2011 in the North Sea, respectively. Some key issues relevant to the best polarimetric mode, feature extraction method, and classification algorithm are investigated and analyzed. In this chapter the possibility of using compact polarimetric SAR modes to obtain better oil spill classification performances while maintaining the large swath-width of single polarimetric SAR system is well proved.