STUDY OF ESTIMATE CONCENTRATION OF WATER CONSTITUENTS AT BADUNG STRAIT BALI USING INVERSE MODEL

Abstract

An algorithm was employed to retrieve the concentrations of three water constituents, chlorophyll-a,suspended matter and colored dissolved organic matter (CDOM) from MODIS (Moderate-ResolutionImaging Spectrometer) in wide range covering from oligotrophic case-1 to turbid case-2 waters at theBadung Strait Bali. The algorithm is a neural network (NN) which is used to parameterize the inverse of aradiative transfer model. It’s used in this study as a multiple nonlinear regression technique. The NN is a feedforward back propagation model with two hidden layers. The NN was trained with computed radiancecovering the range of chlorophyll-a from 0.001 to 64.0 ?g/l, inorganic suspended matter from 0.01 to 50.0mg/l, and CDOM absorption at 440nm from 0.001 to 5.0 m-1. Inputs to the NN are the radiance of the fivespectral channels which were under discussion for MODIS. The outputs are the three water constituentconcentrations. The NN algorithm was tested using in-situ data set on May, September, November 2005 atthe Badung Strait Bali and the north sea of Sumbawa Island and applied to MODIS. The coefficient ofdetermination (R2) between chlorophyll-a concentrations derived from simulation and in-situ data is 0.327,for suspended matter R2 is 0.408. No in-situ measurements of CDOM available for validation. Also, in-situdata were compared with the corresponding distribution obtained by the NASA standard OC4 (OC3M) forMODIS chlorophyll-a algorithm and giving R2 0.188. This study gives better accuracy compare withstandard algorithm. How ever both studies are giving over estimate chlorophyll-a concentration. Since thereare no standard MODIS products available for suspended matter and CDOM, the result of the retrieval by theNN for these two variables could only be assessed by a general knowledge of their concentrations anddistribution patterns