Inversion of satellite ocean colour imagery and geoacoustic characterization of seabed properties: Variational data inversion using a semi-automatic adjoint approach

In this paper a semi-automatic adjoint approach for variational data inversion is proposed. To demonstrate the effectiveness of the approach two illustrative examples are presented: the geoacoustic characterization of a Mediterranean shallow water environment using realistic experimental conditions and the estimation of oceanic and atmospheric constituents from satellite ocean colour imagery. In the first case geoacoustic parameters of the seabed (density, sound speed and attenuation) are determined from long/medium range underwater acoustic propagation data in the water column. In the second case the aerosol optical thickness in the atmosphere and the phytoplankton concentration in the ocean (chlorophyll-a) are estimated from solar reflectance measurements obtained with ocean colour sensors on board satellites. The general methodology for both applications is based on a modular graph concept that allows a straightforward adjoint computation by means of gradient backpropagation. Generation and coding of the adjoint models in both cases are accomplished with an algorithmic tool.