Atmospheric dispersion of hazardous materials poses a significant threat to critical infrastructures. In emergency situations, real-time predictions of contaminant transport are urgently needed for informed decision-making. To illustrate the essential computational steps in the early response phase, this contribution selects a simple description of contaminant dispersion. Using the incompressible Navier-Stokes equations, the wind field in an urban environment is computed, and contaminant transport is modeled using an advection-diffusion equation. Numerical methods for sensor-integrated predictions and source detection are explored and tested for their applicability within a digital twin framework. In the digital twinning process, our goal is to establish a highly automated workflow that acquires current urban geometry directly from a database, such as Open Street Map. This involves generating an initial simulation mesh and adjusting its resolution based on the required precision. During an offline preparation phase, computationally intensive flow simulations are conducted for various wind conditions to provide the snapshots for a reduced order model (ROM) of the flow field enabling rapid assessments. Additionally, we explore scientific machine learning techniques to integrate physics and data, striving to create a dependable predictive model for a hybrid digital twin [1]. The resulting computational model is integrated into a digital twin framework to gather real-time sensor data from the physical environment and present the outcomes in an informative manner for human interaction. Additionally, the simulation model aims to aid in developing optimal sensor placement strategies and conducting analyses for hypothetical scenarios. REFERENCES [1] von Danwitz, M., Kochmann, T.T., Sahin, T., Wimmer, J., Braml, T. and Popp, A. (2023), Hybrid Digital Twins: A Proof of Concept for Reinforced Concrete Beams. Proc. Appl. Math. Mech., 22: e202200146. doi.org/10.1002/pamm.202200146