Digital twins (DT) increasingly rely on large volumes of simulation data to represent, predict, and analyze behavior of complex physical systems and assets. Data management techniques in the context of digital twins, however, are still in their early stages [1]. In practice, simulation data generated during scientific research is often stored in heterogeneous file and data formats on local machines, following ad hoc naming conventions and directory structures. The resulting fragmentation significantly hampers data discoverability, reproducibility, and efficient reuse, particularly for simulations defined over multiple spatial, temporal, and parametric dimensions. Retrieving specific quantities of interest – such as the value at a given mesh node, timestep, and for a particular simulation run, as defined by its set of input parameters – can become prohibitively time-consuming. We demonstrate the usage of a dedicated database approach for storing and accessing simulation data within the context of an urban digital twin. As a representative use case, we consider data from a meshbased simulation with a reduced-order model (ROM), describing the dispersion of an airborne contaminant in an urban setting [2]. Our example highlights how structured storage of spatial, temporal and parametric dimensions enables efficient querying, consistency across simulation runs, and clear semantic separation between model metadata and results. The proposed database-centric workflow supports systematic data ingestion, standardized metadata annotation, and scalable access patterns. By replacing file-based data storage with a unified and expandable data model, this approach facilitates rapid and flexible querying of specific simulation results, integration with other digital twin services, and improved communication and collaboration between stakeholders, thus enabling enhanced decision-making capabilities. We aim to illustrate how such a data management strategy can serve as a foundation for robust, maintainable, and extensible digital twin implementations.