A systematic review was conducted on the downscaling methods of GRACE satellite data, which were classified into three categories:statistical downscaling, dynamic downscaling, and multi-source data fusion. The principles and evolution of research methods ranging from classical regression to emerging physics-constrained models were thoroughly analyzed. By evaluating application cases worldwide, the advantages and limitations of each method were revealed, and core issues, including data defects, insufficient physical authenticity of models, difficulty in coupling human activities, and incomplete verification system were identified. The future directions of physics-guided artificial intelligence, spatiotemporally consistent data fusion, and the development of operational high-resolution datasets were discussed. Based on a systematic assessment of existing methods, a new physics-data dual driven paradigm for intelligent fusion was proposed. Keywords: GRACE satellite data; downscaling; terrestrial water storage; groundwater; hydrological model; remote sensing hydrology 〖FL