The accuracy and universality of the application of remote sensing images in water surface extraction have been significantly improved, while the complexity and difficulty of remote sensing data processing have also been brought about. Based on the computing platform Google Earth Engine, this study focused on the key rivers in the arid and semi-arid regions of western Northeast of China, and carried out rapid quantitative calculation and analysis of water surface dynamics. By comparing various indexes, this study evaluated the differences in river water surface recognition accuracy, selected the most suitable index for this type of water body, extracted river water surface information combined with field investigation data, and further calculated the length and duration of the river with water to determine the annual cut-off situation of the river. This study aims to provide scientific basis and new technical ideas for river water restoration.