A novel spectral index for ocean oil spill detection

Abstract

In this work, the authors propose a new spectral index for oil spill detection, the Normalized Difference Oil Index (NDOI). Its main characteristic is that it obviates suspended sand in the sea, making it ideal for the detection of spills in coastal areas. A comparison is made of the 10 most important spectral indeces used to detect spills in the specific case of the Deep Water Horizon (DWH). An assessment is made based on confusion matrices, accuracy and F1 -score metrics resulting from the kNN classification. The results obtained present NDOI as a good candidate for future coastal spill monitoring due to its balanced tradeoff among the different error metrics and its performance.