Multi layer perceptron neural network algorithms for ocean colour applications in coastal waters

D'Alimonte, Davide (2004) Multi layer perceptron neural network algorithms for ocean colour applications in coastal waters. University of Southampton, Doctoral Thesis.

Record type: Thesis (Doctoral)

Abstract

This work focuses on the development, performance assessment and identification of the range of applicability of ocean color algorithms for the retrieval of seawater constituents in optically complex coastal regions. Multi Layer Perception neural network algorithms were implemented using a set of experimental data collected in the northern Adriatic Sea (CoASTS data set). The following quantities were modelled as a function of the remote sensing reflectance: i) Chlorophyll-a concentration (Chl-a); ii) Absorption of the pigmented particulate matter at 443 nm (α_ph (443); iii) Absorption spectra of the coloured dissolved organic matter (α_ys) and of the non pigmented particulate matter (a_dp). The range of applicability of two MLP algorithms developed for the retrieval of the Chl-a in coastal and open sea regions, were compared using SeaWiFS images of European Seas. A method for algorithms blending was also formulated and applied to case studies.