Phytoplankton determination in an optically complex coastal region using a multilayer perceptron neural network

The determination of phytoplankton in seawater, quantified as chlorophyll-a concentration (Chl-a) or absorption of pigmented matter (a/sub ph/), is a major objective of optical remote sensing. The accuracy of multilayer perceptron (MLP) neural network algorithms in determining Chl-a and a/sub ph/ at 443 nm as a function of the multispectral remote sensing reflectance (R/sub rs/) was investigated for optically complex waters. The implementation of the MLP algorithms was carried out relying on an experimental dataset collected in a coastal region of the northern Adriatic Sea. The performance of the algorithms was assessed on both separate and combined Case 1 and Case 2 water types. The proposed MLP algorithms showed a better accuracy both with respect to other algorithms developed on the basis of the same dataset as well as with respect to independent algorithms operationally used for the processing of Sea-viewing Wide Field-of-view Sensor (SeaWiFS) data. The study also showed a high accuracy in determining a/sub ph/(443) and, thus, further confirmed the possibility of computing the inherent optical properties of seawater significant components from the R/sub rs/ spectra.