The aim of this work is to develop an automated approach for sea ice mapping and ice concentration determination using visible/infrared measurements provided by a geostationary satellite. This study is a part of the algorithm development activities of the future GOES-R ABI sensor. The data used in this study as prototype of the future GOES-R ABI sensor are provided by the SEVIRI sensor onboard of the METEOSAT Second Generation (MSG) satellite. The algorithm developed in this study is completely autonomous. Images of the prospective GOES-R will be the sole input of the algorithm which returns as an output ice charts and ice concentration maps. To achieve the ultimate objective of this study, two issues have been addressed. Firstly, it was necessary to accurately detect and map clouds over the study area in order to estimate ice fraction exclusively over cloud-free pixels. This primary step has been performed using the same proxy data provided by the SEVIRI instrument. Secondly, reflectances of all the Caspian Sea pixels were simulated for all the possible sun-satellite geometries and for pure ice and ice free pixels. The neural network was used in order to obtain these reflectances. An exhaustive sample has been selected using MODIS images to train the network. The obtained water and ice reflectances have been used to estimate the ice fraction as a ratio of the difference of the observed reflectance at channel R01 (0.6 μm) and the water reflectance over the difference of the ice reflectance and water reflectance. The obtained results were validated using MODIS images and IMS charts. A good agreement has been observed between observed and simulated values. This implies that visible and IR channels of the upcoming GEOS-R ABI have an interesting potential for sea ice mapping.