Abstract: This article takes Liaoning province as the research area, based on the GEE (Google Earth Engine) remote sensing cloud platform, and using Sentinel-2 remote sensing images, proposes a multi-feature and multi-level algorithm for extracting lake and reservoir water bodies. This algorithm selects the automatic water index (AWEI_sh) and the improved normalized water index (MNDWI) to extract water bodies, and uses the normalized vegetation index (NDVI), normalized building index (NDBI), normalized difference red edge index (NDREI), Sentinel-2's B8 and B9 bands, as well as DEM data to multi-level eliminate dark and bright ground noise, and to repair partially missing water bodies in the extraction results that are obscured by clouds and mist. Finally, remove the river and small pixels. This algorithm is used to extract lake and reservoir water bodies in Liaoning province from April, July, and October of each year from 2017 to 2021. Different water body extraction algorithms and water body data products were compared. The experimental results showed that the proposed algorithm had good performance in extracting water bodies under large-scale conditions, with an overall accuracy of over 96%. It can effectively remove dark pixel surfaces such as vegetation and shadows, and ensure the integrity of water body information, It has certain applicability and stability in large-scale water extraction.

Key words: GEE, Sentinel-2, lake and reservoir water bodies, cloud cover repair, denoising