We hypothesize that the primitive-equation ocean model of Bryan & Cox, with a free surface after Killworth and standard parameterizations, can be reconciled with surface and subsurface data from the Tropical Atmosphere-Ocean (TAO) array in the tropical Pacific Ocean. Our hypothesis is expressed as first and second moments of errors in the dynamics, surface fluxes, initial conditions and data. The second moments include inhomogeneous autocovariances with respect to time and space. Development of the error hypothesis especially taxes our understanding of divergences of subsurface flux parameterizations. We seek a weighted best-fit to all the constraints by variational methods, implemented under the Inverse Ocean Modeling system (modular software exploiting functional programming techniqes). The variational formulation allows for an indiviudally weighted error in each flux component. The IOM diagnostics include not only residuals in all constraints and significance tests for our hypothesis, but also a quantitative assessment of the efficiency of the TAO array on SI scales.