EMPRISE® is a software package under developmentused by sensor and radar system companies, as well as the French government for its equipment programs. The software simulates the IQ signals received by a radar system from the digital twin of a radar scene. The goal of any simulation software, such as EMPRISE®, is to quantify the performance of a given radar system in a given environment. To improve the software representativeness of these simulators, it is necessary to calibrate the radar material cross-sections used by the scene model according to measurement data. In this paper, we propose a calibration method, applied to EMPRISE®, based on an adaptive surrogate model and a minimizing search algorithm.