In many land seismic situations, the complex seismic wave propagation effects in the near‐surface area, due to its unconsolidated character, deteriorate the image quality. Although several methods have been proposed to address this problem, the negative impact of 3D complex near‐surface structures is still unsolved to a large extent. This paper presents a complete 3D data‐driven solution for the near‐surface problem based on 3D one‐way traveltime operators, which extends our previous attempts that were limited to a 2D situation. Our solution is composed of four steps: 1) seismic wave propagation from the surface to a suitable datum reflector is described by parametrized one‐way propagation operators, with all the parameters estimated by a new genetic algorithm, the self‐adjustable input genetic algorithm, in an automatic and purely data‐driven way; 2) surface‐consistent residual static corrections are estimated to accommodate the fast variations in the near‐surface area; 3) a replacement velocity model based on the traveltime operators in the good data area (without the near‐surface problem) is estimated; 4) data interpolation and surface layer replacement based on the estimated traveltime operators and the replacement velocity model are carried out in an interweaved manner in order to both remove the near‐surface imprints in the original data and keep the valuable geological information above the datum. Our method is demonstrated on a subset of a 3D field data set from the Middle East yielding encouraging results.