Surface water occurrence in river deltas is governed by precipitation, evaporation, and the influx and outflux of water to and from the delta. Although studies of changes in water occurrence have been conducted at large scales, precise detection of changes in water occurrence is missing for most important river deltas. We take the case of the endorheic Selenga River Delta in Russia and train an accurate classification and quantification of water occurrence in its domain. We utilize remotely sensed observations of the Landsat satellite imagery during the last 33 years and implement supervised classification to map the surface water extent and its changes between periods of 1987-2002 and 2003-2019. We find that water occurrence has decreased in the Delta, with seasonally inundated areas presenting more pronounced decreases in water occurrence than permanent water bodies. We show that the change in the surface runoff is the main driver of changes in the spatial patterns of surface water with R2 = 0.58, while changes in water level in the recipient Lake Baikal do not influence water occurrence in the Delta. Our results show that the shrinkage and expansion of the water surface reflect the change in the freshwater supply of the Delta, and the management of the Selenga River needs to consider the impact of changes on the water occurrence.