As smart cities increasingly integrate advanced technologies like the Internet of Things (IoT) and Artificial Intelligence (AI), the need for robust data ecosystems to support Digital Twins (DTs) becomes essential. DTs offer real-time digital representations of physical urban systems, enabling predictive analytics, intelligent decision-making, and operational optimization. This paper presents the development of a DT data ecosystem based on stakeholder workshops and use case elicitation. The ecosystem was modelled at micro, meso, and macro levels to reflect different layers of abstraction and ensure scalability and adaptability. Through this modelling process, recurring patterns were identified and synthesized into a multi-level data ecosystem framework. The Capability-Driven Development (CDD) metamodel was extended to incorporate DT-specific elements. The resulting framework also addresses gaps in available datasets and proposes integration of real-time, high-resolution data to improve DT functionality. This approach fosters modularity and reusability, supporting efficient development, validation, and long-term resilience of DT solutions for smart cities.