Modeling complex urban systems for effective policy analysis remains a significant challenge, often caught between the subjective structures of traditional simulation models and the predictive yet opaque nature of modern machine learning. Our team at the Community Co-Financed Flood and Energy Resilience (CCOFFER) initiative introduces the Covariation Mining-Supported System Dynamics (CM-SSD) framework, a novel methodological synthesis designed to bridge this gap. CM-SSD leverages data-driven covariation mining from multivariate time-series data to empirically derive the dependency structure of a system, which then informs and parameterizes a formal System Dynamics model. This integration provides an objective, evidence-based foundation for simulating the nonlinear feedback and time-delayed effects inherent in urban dynamics. The framework’s utility is demonstrated through an application to the urban-economic system of Lake Charles, Louisiana, using annual data from 2007 to 2024. The analysis focuses on uncovering the pathways of influence through which Green Stormwater Infrastructure (GSI) and Federal Spending affect city-level house prices. The discovered network of influence reveals a critical, policy-relevant reinforcing feedback loop: GSI investments not only directly support property values but also catalyze increased federal per capita spending, which in turn provides a secondary stimulus to the housing market. By moving from observational data to a validated simulation model, the CM-SSD framework offers a robust, transparent, and powerful tool for ex ante policy evaluation in complex systems, representing a significant contribution to the fields of data mining, knowledge discovery, and computational social science.