Application of MERIS Data for the Estimation of Clarity Depth in the Caspian Sea using Univariate and Multivariate Linear Regression Techniques

Secchi disc is an indicator of the clarity of the water bodies. In this study, new univariate and multivariate linear regression models are developed for monitoring of Secchi disc depth (SD) in the Caspian Sea using MERIS images and unlike of previous studies, the developed models are tested to determine the real accuracy of developed models for the Secchi depth monitoring. In situ measurements of Secchi disc depth was performed in the southern part of Caspian Sea between July and October 2005 and consequently, 25 training and 12 testing data were acquired. In this study, 25 Level 1B MERIS images of the Caspian Sea, acquired concurrent with in-situ measurements, were employed. In univariate regression, the correlation between Secchi depth and Spectral reflectance data (Rrs) and the ratio of Rrs data were investigated and then the Secchi depth and the Rrs parameters with high correlation coefficient were selected and some univariate models were fitted using the training data on them. In addition, the appropriate multivariate regression models were developed using Cp mallow’s statistics and the best one was selected by the test data. The results showed that the developed multivariate model presents better results than univariate models and it has higher correlation coefficient than the previous studies. The variables of the best multivariate model were the reflectance in 412, 510, 560, 681, 779 nm and the correlation coefficient and percentage error of the best model were about 0.7 and 37.7 %, respectively. Finally the maps of Secchi depth in the Caspian Sea were retrieved using the developed multivariate model.