Abstract
Constant-depth (or “z-coordinate”) ocean models such as MOM4 and NEMO have become the de facto workhorse in climate applications, having attained a mature stage in their development and are well understood. A generic shortcoming of this model type, however, is a tendency for the advection scheme to produce unphysical numerical diapycnal mixing, which in some cases may exceed the explicitly parameterised mixing based on observed physical processes, and this is likely to have effects on the long-timescale evolution of the simulated climate system. Despite this, few quantitative estimates have been made of the typical magnitude of the effective diapycnal diffusivity due to numerical mixing in these models. GO5.0 is a recent ocean model configuration developed jointly by the UK Met Office and the National Oceanography Centre. It forms the ocean component of the GC2 climate model, and is closely related to the ocean component of the UKESM1 Earth System Model, the UK's contribution to the CMIP6 model intercomparison. GO5.0 uses version 3.4 of the NEMO model, on the ORCA025 global tripolar grid. An approach to quantifying the numerical diapycnal mixing in this model, based on the isopycnal watermass analysis of Lee et al. (2002), is described, and the estimates thereby obtained of the effective diapycnal diffusivity in GO5.0 are compared with the values of the explicit diffusivity used by the model. It is shown that the effective mixing in this model configuration is up to an order of magnitude higher than the explicit mixing in much of the ocean interior, implying that mixing in the model below the mixed layer is largely dominated by numerical mixing. This is likely to have adverse consequences for the representation of heat uptake in climate models intended for decadal climate projections, and in particular is highly relevant to the interpretation of the CMIP6 class of climate models, many of which use constant-depth ocean models at ¼° resolution
Highlights
The importance of using a correct distribution of the diapycnal mixing, and of the watermass transformation rate, to the largescale ocean circulation in climate models is evident: the upwelling regions of the global overturning streamfunction are associated with mixing processes (Munk and Wunsch, 1998), while the formation of a realistic thermocline relies on appropriate rates of mixing above and below the thermocline (Luyten et al, 1983)
We note that of the thirty-nine climate models contributing to the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (IPCC, 2013), thirty-two used ocean components formulated with depth coordinates, three used terrainfollowing coordinates, and four used an isopycnic ocean model
We have applied the analysis of Lee et al (2002), itself based on the isopycnal watermass transformation framework of Walin (1982) to diagnose the effective diapycnal diffusivity κeff in GO5.0, a 0.25° global ocean-only configuration of the NEMO model
Summary
The importance of using a correct distribution of the diapycnal mixing, and of the watermass transformation rate, to the largescale ocean circulation in climate models is evident: the upwelling regions of the global overturning streamfunction are associated with mixing processes (Munk and Wunsch, 1998), while the formation of a realistic thermocline relies on appropriate rates of mixing above and below the thermocline (Luyten et al, 1983). Small-scale turbulent mixing in ocean models is represented by a variety of parameterisations, including bulk schemes Numerical diapycnal mixing is an intrinsic property of the advection scheme in this class of models, and occurs whenever an advective flux crosses density surfaces, which in general do not follow the horizontal coordinate surfaces (Griffies et al, 2000). It may be reduced by the use of higher-order advection schemes (Hofmann and Morales Maqueda, 2006), and is absent, by construction, in the ocean interior in pure isopycnic models like MICOM
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
