Extremely large aperture arrays (ELAAs) represent a promising architecture in the sixth-generation mobile communication systems with improved spatial division multiplexing gain and spectral efficiency (SE) potential. However, spatial non-stationary channel characteristics kick in because of signal blockage and near-field propagation. To deal with them, this letter models the non-stationary channel by considering an environment-based visibility region (VR) and spherical wavefronts, respectively. Then, we propose a low-overhead channel reconstruction scheme, which can simultaneously locate the user and identify its VR. Our method is designed for subarray-based hybrid structures while the training overhead caused by analog beamforming is limited. Numerical results demonstrate that the proposed method achieves accurate UE localization and VR identification, while the achievable SE can be close to that of using a perfectly known channel in a fully digital system.