To meet strict dependability requirements in hostile and highly-varying environments, IoT communication protocols need to be carefully tuned in relation to the expected environmental changes. However, this is difficult to attain, as every application has unique properties and requirements. Tuning communication protocols correctly requires indeed significant expertise as well as a clear understanding on how hardware and software components are affected by environmental changes. In this paper, we propose a novel framework to automate the parametrization of IoT communication protocols. The framework uses models of the environment as well as the employed hardware and protocols to predict the effects of environmental changes on network performance and to automatically select a configuration that meets user-specified dependability requirements. We demonstrate how to use this framework to configure a state-of-the-art MAC protocol for an IoT application deployed in a challenging outdoor environment and evaluate its accuracy in predicting how environmental changes affect network performance. We further evaluate the performance with different optimization strategies and show that the average run-time necessary to find a solution is sufficiently low to enable the use of our system in a typical IoT design process.