Abstract

Recent evidence shows that observational and theoretical considerations support the shear-instability paradigm for ocean interior mixing. It also remains the prominent form of parameterisation of mixing in climate models. On the other hand, the application of machine learning to global observational mixing data sets seems to suggest that the shear-instability paradigm is far from dominant in the ocean interior and, in fact, might be a small contributor. While I won’t be able to resolve this conundrum in this talk, I will share my thoughts on its potential reasons. The reasons fall into three main categories: a) our subjective theoretically-informed view of mixing processes might have restricted how we `look for’ mixing in observations and how we interpret the output of numerical simulations, b) we have paid less attention to weak background mixing by focusing on the strongest events which form the basis of our parameterisations, c) the intermittency of turbulence and its statistics are significantly under-studied. I will argue that a new approach might be required to represent mixing in climate models, where larger-scale resolved dynamics constrain statistics and the energetics of small-scale turbulence.