A new approach for earthquake early warning systems (EEWS) is presented that uses wireless, self-organizing mesh sensor networks. To develop the prototype of such IT infrastructures, we follow a model-driven system development paradigm. Structure and behaviour models of network topologies in specific ge-ographic regions are coupled with wave signal analysing algorithms, alarming protocols, convenient visualisations and earthquake data bases to form the basis for various simulation experiments ahead of system im-plementation and installation. The general objective of these studies is to test the functionality of an EEWS and to optimize it under the real-time, reliability and cost-depended requirements of potential end-users. For modelling a technology mix of SDL/ASN.1/UML/C++ is used to generate the code for different kind of simulators, and for the target platform (several node types). This approach is used for realizing a prototype EEWS developed within the EU project SAFER (Seismic Early Warning for Europe) in cooperation with the Geoforschungszentrum Potsdam (GFZ). The first operational area of that EEWS is preparation for Istanbul in a region threatened by strong earthquakes. The presented paper focuses on our adopted and developed tool-based modelling and data base techniques used in that project, that are general and flexible enough for addressing similar prototyping use cases of self-organizing sensor-based IT infrastructures.