The applicability of conflict rate models to controlled, structured, terminal airspace is investigated. A collision rate model is adapted to conflicts between flows of aircraft, and its validity is empirically confirmed. Conflicts are divided into between-flow conflicts and within-flow conflicts. Applicability to a terminal airspace without strictly adhered-to procedural routes is investigated by clustering logged mode S-surveillance data into flows using a self-tuning spectral clustering method. This clustering method can effectively partition flight tracks into clusters, but is sensitive to the stop conditions, and unclusterable flight tracks make up a small but significant proportion of conflicts. The predictions of the conflict rate model correlate with observed conflict rates in a replay simulation, but are initially biased towards underpredicting the conflict rates. Altering the timing, intensity and separation requirements for the replay greatly reduces this bias for converging flows, but leads to an overprediction of other conflict types, and altered replay results should be interpreted with care. Analytical predictions for between-flow conflicts can be made, but applicability to within-flow conflicts is limited due to the different types of interactions. The resulting conflict rate model explicitly defines the relation between separation requirements, relative velocities, and flow intensities.