A Model for Ocean Vertical Mixing Including Convection

Columbia University, New York NY

Investigators

Abstract

ABSTRACT PI/Institution: Canuto / LDEO Proposal No: OCE-0241668 The main goal of this proposal is to build and test a unified, non-local model to describe non-convective and convective vertical mixing processes in the ocean. The model is intended for use in coarse resolution ocean models. The first part of this work has already been carried out (Canuto et al., 2001, 2002, C1-2). A turbulence model based on the Reynolds Stress Model was constructed which included the velocity field and two active scalars fields, temperature and salinity. Correspondingly, the model provided the eddy diffusivities of momentum, heat and salinity. The model also provided the eddy diffusivities for a passive scalar. All the diffusivities were expressed analytically in terms of the large-scale fields calculable form the OGCMs. The model was tested directly against the data from NATRE (North Atlantic Tracer Release Experiment) specifically, the heat, salt and concentration diffusivities vs. depth. The data were reproduced quite satisfactory including the ratio of salt to heat diffusivity. Recent but yet unpublished data from measurements at Barbados (reported by R.W. Schmitt of WHOI at the Hawaii Ocean Sciences Meeting, Feb. 2002) of a much larger mixing at Barbados than at NATRE, are also in agreement with the predictions of the model. When used in a coarse resolution OGCM, the model reproduced data closer to Levitus data than previous models (e.g., a better agreement of the T vs.z profile in the Arctic Ocean). The model is however local and cannot account for key dynamical feature such as the entrainment process, a precursor of convective overturning or Deep Convection, DC. The latter process is one of the ocean's major features since it represents the initial stage of the global-scale ventilation loops of the world ocean and is also the predominant mechanism involved in the production of NADW (North Atlantic Deep Water). Many studies in the last ten years have highlighted major features of DC among which: spatial localization (e.g., plumes-like behavior), non-locality (a tracer injected in a plume will not get mixed into the ambient fluid until the plume "creases to entrain"; a property is thus transported directly from the surface to the bottom, a non-local process); dependence on rotation. In contrast to this wealth of data, the representation in OGCMs of the rich phenomenon of DC is still rather rudimentary. The most commonly used is the Convective Adjustment proposed in 1969. Many shortcomings have, however, been highlighted by recent studies the most prominent being perhaps the excessive level of convection that ultimately yields a poor representation of the ice data (too little ice). A more promising model is based on an oceanic adaptation of a Plume Model originally devised by G.I. Taylor. The model has produced considerably more acceptable results. Thus, at present, vertical mixing processes can only be dealt with by using one model for convective process and another model for non-convective (diapycnal) mixing. Here, we propose a unified treatment of both processes. As stated above, the non-convective mixing part has already been studied and we now plan to extent the previous local model to become non-local so as to encompass the convective processes. Not only have we used such a model to study other types of convection but we have recently shown that the non-local turbulence model reproduces the Plume Model which assumes that Plumes occupy a small fraction of the convective patch which holds true only in the initial stages. Our mixing model will be first tested against LES data and then implemented in HYCOM (an OGCM in isopycnal coordinates). If our project is successful, not only will the convective and non-convective processes be part of the same mixing model but we shall avoid the limitation of the Plume Model just discussed.