Bo-Tsen WangYu-Li WangChia-Hao ChangChin-Tsai HsiaoJordi Mahardika PuntuPing-Yu ChangJui-Pin Tsai2025-03-182025-03-182025-0600221694https://www.scopus.com/record/display.uri?eid=2-s2.0-85218233568&origin=recordpagehttps://scholars.lib.ntu.edu.tw/handle/123456789/725769Article number: 132828River stage tomography (RST) is a potential method for delineating spatial heterogeneity in regional aquifers by analyzing groundwater head variations in response to river stage fluctuations. However, groundwater head data often reflect mixed signals from external stimuli, such as rainfall and river stage, which can compromise parameter estimation. To address this issue, we propose an integrated method combining empirical mode decomposition method (EMD), dynamically dimensional search algorithm (DDS), and RST. Synthetic case results demonstrate that the proposed method reconstructs river-induced head variations with high accuracy (R2 = 0.9976, RMSE = 0.0099 m) and yields estimated hydraulic diffusivity (D) field closely matches the true D field. In the real case, the estimated D values align well with the sampled values (differences below 5 % for most wells), and the estimated D field is consistent with the expected aquifer structure within the study area. These findings demonstrate that the proposed method successfully extracts and reconstructs river-induced head variations from original head observations and accurately delineates regional aquifer features. This method shows the significant potential for enhancing RST studies by offering a robust approach for mixed-feature signal decomposition and reconstruction.falseHydraulic diffusivity (D)Parameters estimationRiver stage tomography (RST)Signal decomposition[SDGs]SDG6journal article10.1016/j.jhydrol.2025.1328282-s2.0-85218233568