Objectives

The objectives of the trainings course are to provide participants with an overview of inversion methods and models, to prepare bio-optical models and training data sets for inversion methods, and to teach participants how to use various inversion techniques.

The course is intended for scientists or advanced graduate students working with ocean-colour data in coastal waters. Participants should have a strong mathematical background as well as experience in programming.

Background

One of the methodological challenges in remote sensing, here in particular in ocean colour remote sensing, is the determination of geophysical or bio-geochemical variables from the reflected electromagnetic spectrum. Since these geophysical variables are the independent ones it is rather easy to simulate the reflected spectrum, since the radiative transfer equations are well understood, and there are various numerical tools to perform the computation with high accuracy. However, the inverse way is more difficult and requires a number of considerations concerning ambiguities, uncertainties and the scope of an algorithm. This is particularly important for case 2 waters, which, by definition, contain more than 1 water constituent, which determines the reflected spectrum so that a non-linear relationship exists between a number of water constituents and their inherent optical properties and a number of spectral bands of the reflected spectrum.

For the inversion different techniques have been developed including multiple band ratios, linear matrix inversion, iterative adaptation using a radiative transfer model and a non-linear optimization techniques and neural networks, which are used as a non-linear multiple regression procedure.

All of these techniques will be presented and applied in form of exercises. Of particular interest is the issue of the scope of an algorithm and the uncertainties in the retrieval process.