Optical interpretation of oil emulsions in the ocean - Part III: A three-dimensional unmixing model to quantify oil concentration

Abstract

Knowledge of oil distribution and characteristics (e.g., oil type, oil thickness, concentration, volume, etc.) is valuable for oil spill response and post-spill assessment. Optical remote sensing has the advantage of collecting information in multiple spectral bands, thus being able to classify and quantify different oil emulsions. However, most spectral characteristics of oil emulsions have been obtained from laboratory measurements, and it is unclear how to apply these lab-based findings to airborne or satellite observations of highly heterogeneous oil patches. This study proposes a three-dimensional (3D) unmixing model to explain AVIRIS image spectra of oil-containing pixels with both horizontal and vertical mixing effects. Based on the 3D unmixing model, an oil concentration lookup table (LUT) and LUT-based oil concentration mapping method are developed to quantify oil concentrations from AVIRIS observations, with the results validated through the use of a spectral similarity index. Using spatial and spectral resampling simulations, we also propose an approach to extend the LUT to different optical sensors (e.g., Landsat) in order to map oil concentrations, yet such an approach requires further research to validate and improve.