The retrieval of people trapped in buildings in fire or partially destroyed is a priority for rescue teams. Given the widespread use of mobile phones by the general public, an innovative idea to retrieve victims or to track rescue teams [1] consists in capturing, via an antenna array acting as subrogated base station, the signals transmitted by cell phones (even if these are in idle mode). Real-time postprocessing then allows localizing the victim using a robust localization algorithm. The receive antenna array may be mounted under a helicopter for a first rough localization, while a second step involves a pedestrian recue team equipped with portable versions of the same antenna array, providing in real-time localization data of the victim. In order to evaluate the performance of localization algorithms, a precise channel characterization is however required. Collapsed building structures constitute a very complex environment, for which no model exists so far. Also, there is a huge variety of possible scenarios because many different material can be involved; the electromagnetic properties of any given material strongly depends on the structure, humidity, compression, etc., so that the values of permittivity and conductivity are hardly known with precision; and the shape and dimensions of each rubble is highly variable. The goal of the present communication is to propose a channel model to predict the propagation characteristics between an outdoor unit and a buried terminal.