Canonical correlation analysis is used to formulate linear estimates of independent or orthogonal (incremental) information on the performance of water resource systems. Simple economic and resilience indicators are compared, and for the particular configurations chosen in this paper the replicated and independent information sets are explained in terms of the connectivity of the various reservoirs. The canonical correlations are relatively high, indicating that there is significant replication of information; a taxonomy is developed which suggests those basin and structural characteristics which would indicate less replication and consequently improved description of system performance. The procedure is based on simulation studies reported in an earlier paper in this series.