Wireless mesh networks offer flexibility for industrial automation, but, in these environments with changing propagation conditions, it is challenging to guarantee the radio coverage and connectivity. This paper contributes a new localization-based method for the calibration of radio propagation models. The idea is to find the locations of the mobile stations via localization and to use radio signal strength measurements from them for adjusting the radio model parameters until the model better fits to the real environment. This calibration method is integrated in our previously published fault-tolerance framework for guaranteeing the availability of radio coverage and connectivity of wireless mesh networks. It is used to automatically detect environmental dynamics (errors) at run-time and to propose a network reconfiguration before they lead to service failures. An evaluation in a real industrial scenario shows the practicability of our approach.