Streamflow Reconstruction Dataset for Central Asian River Basin Using ERA5-Land and Multi-Timescale LSTM

This dataset comprises monthly streamflow records from selected Central Asian river gauging stations, both actual and rebuilt. A Multi-Timescale Long Short-Term Memory (MTS-LSTM) deep learning model trained on ERA5-Land climate reanalysis data was used to build the reconstructed streamflow series. The dataset was created as part of the WE-ACT initiative to help with hydrological modelling, water resource assessment, and climate change research in data-poor areas. The dataset comprises streamflow records from numerous sites, as well as documentation outlining the process, variables, and intended application.