Abstarct: Monitoring of water quality parameters (e.g. chlorophyll-a (Chl-a), suspended particulate matters (SPM), colored dissolved organic matters (CDOM)) in the water bodies is vital. But the monitoring of spatio-temporal distribution of water quality parameters using traditional in-situ measurements is not cost effective and it is very complicated and time consuming. In the last decades, remote sensing has been appeared as a useful tool for water quality studies. Now, some the water quality parameters are estimated using satellite images reasonably. But one of the most important challenges in this subject is the finding or development of the appropriate methods for estimation. In this study, the performance of five different inverse models of MERIS sensor (MEGS, C2R, FUB/Wew, ALM and Hybrid ALM-ANN) for the extraction of Chl-a, SPM and CDOM in the Caspian Sea using MERIS images were evaluated and validated. In addition the adjacency effects on the results of different inverse models are evaluated. The different empirical approaches (linear regression, multi-variate regression, ALM) were utilized for the extraction of Secchi disk depth in the Caspian sea using MERIS images. For the coincidence of in-situ measurements and Satellite data, three different data extraction methods (central pixel, mean of 3*3 pixels and median of 3*3 pixels) of satellite images were performed. The results demonstrated that mean of 3*3 pixels is the best method for data extraction of satellite images in the modeling of Chl-a, SPM and CDOM. Comparison between the inverse models for the Chl-a and pigment in the Caspian Sea showed that hybrid of ALM-ANN method is the best one and MEGS is the worst one. For the retrieval of the average CDOM in the water column, the models can be ranked as ALM, C2R, FUB/Wew and MEGS. For the SPM retrieval the ALM was the best model and MEGS was the worst one. Utilizing of ICOL processor for the correction of adjacency effects, improved the results of ALM for Chl-a and pigment retrieval and the other models did not present significant improvements. For the CDOM retrieval, ICOL improved the results of C2R model and deteriorate the FUB/Wew results. ALM results did not show any significant improvement. In the SPM modeling, ICOL improved the results of FUB/Wew and C2R models. In addition, the performance of different inverse models in a cross section in the Caspian sea was evaluated and using this cross section the boundary between coastal zone and open sea was determined (8 km). The MEGS and C2R inverse models were validated using the development of new bio-optical models for them for Chl-a retrieval in the Caspian Sea. The validation highly improved the results of MEGS method but it has no significant effect on the C2R results. The validated models were compared with hybrid ALM-ANN model and the results showed that hybrid ALM-ANN model is the best one for Chl-a retrieval. Validation using the shifting method was applied on the inverse models for CDOM and SPM retrieval according to their biasness. The results highly improved and the C2R and ALM were introduced as the best models for CDOM and SPM retrieval, respectively. Results of Secchi disk depth modeling using the reflectance data in the TOA (Top of Atmosphere) and BOA (Bottom of Atmosphere) showed that the linear regression can present an easy and reasonable model , but the results of ALM is better than the other methods and the results of models using BOA data is better than the using the TOA data. Finally, using different inverse models the time series of spatial distribution maps of different water quality parameters in the Caspian sea were generated using the time series of MERIS images.