Open‐ocean boundary conditions from interior data: Local and remote forcing of Massachusetts Bay

Authors:
  • PBPhilip S. Bogden
    Philip S. Bogden
  • PMPaola Malanotte‐Rizzoli
    Paola Malanotte‐Rizzoli
  • RSRich Signell
    Rich Signell
Publisher website
Google Scholar
Abstract
Massachusetts and Cape Cod Bays form a semienclosed coastal basin that opens onto the much larger Gulf of Maine. Subtidal circulation in the bay is driven by local winds and remotely driven flows from the gulf. The local‐wind forced flow is estimated with a regional shallow water model driven by wind measurements. The model uses a gravity wave radiation condition along the open‐ocean boundary. Results compare reasonably well with observed currents near the coast. In some offshore regions, however, modeled flows are an order of magnitude less energetic than the data. Strong flows are observed even during periods of weak local wind forcing. Poor model‐data comparisons are attributable, at least in part, to open‐ocean boundary conditions that neglect the effects of remote forcing. Velocity measurements from within Massachusetts Bay are used to estimate the remotely forced component of the flow. The data are combined with shallow water dynamics in an inverse‐model formulation that follows the theory of Bennett and McIntosh [1982], who considered tides. We extend their analysis to consider the subtidal response to transient forcing. The inverse model adjusts the a priori open‐ocean boundary condition, thereby minimizing a combined measure of model‐data misfit and boundary condition adjustment. A “consistency criterion” determines the optimal trade‐off between the two. The criterion is based on a measure of plausibility for the inverse solution. The “consistent” inverse solution reproduces 56% of the average squared variation in the data. The local‐wind‐driven flow alone accounts for half of the model skill. The other half is attributable to remotely forced flows from the Gulf of Maine. The unexplained 44% comes from measurement errors and model errors that are not accounted for in the analysis.