From mesoscale to microscale: Estimating ocean mixing and stirring from observations

Abstract

Ocean mixing consists out of many processes acting on various scales, from millimetres to over 100km and from seconds to months. This way, ocean mixing affects the uptake, transport and storage of tracers, such as heat and carbon, thereby influencing the climate, ocean circulation and other phenomena such as Oxygen Minimum Zones, hydrological cycles and sea-level rise. To increase our understanding of the climate and future projections thereof, a better understanding of the magnitude and drivers of these mixing processes is needed. However, both in modelling and observational studies, the mixing occurs at scales that are not fully resolved. Either because the model grid is too coarse or the spatio-temporal scales of observations are insufficient. Therefore we rely on parameterizations: approximations based on larger scale variables. These parameterizations are sensitive to unconstrained choices, but can be improved by observationally-based constraints. This thesis uses various methods and datasets to obtain observationally-based estimates of the mixing strength. This type of estimates helps to improve our understanding of mixing processes and to improve parameterizations of ocean mixing used in modelling studies. First, a new inverse method is introduced: the Spiralling Inverse Method (SIM). The SIM simultaneously estimates the isoneutral and dianeutral diffusivities from observational temperature and salinity data. The SIM uses a balance between a watermass transformation equation and the thermal wind balance. Novel about the SIM is that it does not require a reference velocity or other additional input data, thus solely requiring temperature and salinity data as input. The SIM was tested on a region in the North Atlantic. The estimated diffusivities were well within the range set by observations and other (inverse) methods. Mesoscale eddies have, beside the diffusive component, also an advective component known as the quasi-Stokes velocity. The widely used passive tracer diffusivity is not directly informative of the quasi-Stokes velocity, as this depends on the buoyancy diffusivity (a.k.a the Gent-McWilliams diffusivity). The first global observational estimates of the quasi-Stokes transport are obtained, by assuming that eddies diffuse PV rather than buoyancy. This quasi-Stokes transport is also used to estimate the corresponding buoyancy diffusivity. Subsequently, the new estimates of the quasi-Stokes transport are used to address an assumption underlying the SIM. With the quasi-Stokes estimates included, the SIM is used to obtain global estimates of the isoneutral and dianeutral diffusivities. Despite their noise and sensitivity to choices in the application, the results identify areas for improvement in the theoretical framework, data quality and handling, and technical implementation of the SIM. Lastly, various observations and parameterized estimates of small-scale mixing are compared. High-resolution microstructure measurements of velocity shear and temperature are used to estimate the turbulent kinetic energy dissipation rate. In weak turbulence, the thermistor-based estimates are preferred over the shear-based estimates due to the absence of a (instrumental) noise floor. In addition to the practical implementation improvements, the parameterizations, based on coarser sampled CTD-data, can estimate the dissipation rate to within a factor 5 of the microstructure estimates, but achieving a higher accuracy becomes difficult.