SPARC is a physics-constrained spatial machine-learning pipeline that trains an ensemble of geographically-weighted models, validates causal relationships via directed acyclic graphs (DAGs), and simulates "what-if" intervention scenarios with built-in uncertainty quantification. It is designed to be domain-agnostic.