Abstract: To facilitate dynamic monitoring of the spatial distribution of water bodies in the Three Gorges reservoir area, we have developed a framework of water body detection, reconstruction and extraction using China’s GF-1 satellite images of high spatiotemporal resolution. Results show the method of object-oriented change detection has an overall accuracy of cloud detection up to 97%, with the producer's accuracy and user's accuracy of cloud detection both higher than 76%. Large cloud-contaminated areas can be effectively recovered using a spatiotemporal reconstruction approach, which can reflect the complete spatial variations. In addition, we find the water surface area of this reservoir, with a mean of 540.6 km2, did not significantly increase or decrease during 2013-2017. This study is useful for comprehensive, integrated management of the reservoir area, with important implications for applying GF-1 satellite images in this region.

Key words: Three Gorges Reservoir area, GF-1 satellite images, cloud detection, spatiotemporal reconstruction, variation in water surface area