Determination of Secchi Disc depths in Lake Eymir using remotely sensed data

In this study Secchi disk depths (SDD) are determined in an eutrophic Eymir Lake in Ankara using the multi-spectral image obtained from the Quickbird satellite. For this purpose, empirical models given in literature and artificial neural networks (ANN) are used. SDDs at 17 sampling points in Eymir Lake are measured via field studies. In the satellite image, pixel values at the sampling points are determined using ERDAS Imagine. Results indicate very low correlations between the SDD values calculated using the empirical models and the ones measured in-situ. Correlation of determination values (R2) up to 0.92 are achieved when ANN modeling is applied. In ANN models developed, Levenberg-Marquardt (LM) and gradient decent algorithm (GDA) are the training algorithms that provided the best results. This study indicates that ANN is an important tool in obtaining information from the remotely sensed data.