The sensitivity of ocean heat uptake to numerical mixing in forced and coupled ocean models

The oceans are believed to be responsible for taking up over 90% of the heat retained by anthropogenic greenhouse gases in the atmosphere. For this reason it is important for the ocean component in climate models to have an acceptably realistic representation of the processes that transport heat from the surface into the ocean interior. It is known that ocean models tend to have significant levels of spurious numerical diapycnal mixing, arising chiefly from truncation errors in the tracer advection scheme. It is therefore important to quantify the sensitivity of the simulated climate to numerical mixing, and then to evaluate remedies for the numerical mixing.     

A large ensemble of forced and coupled simulations with a 1/4° NEMO ocean is shown to have a global mean surface heat flux ranging from -1.5 to +1.5 W/m2. We demonstrate a strong negative correlation between global ocean surface heat flux and the global mean effective diapycnal diffusivity, a metric of total ocean mixing that includes both explicit and numerical contributions. Several potential mechanisms for this apparently paradoxical result are discussed, including changes in upwelling in the Southern Ocean, mixing of bottom waters in the Pacific and Atlantic, and the Atlantic meridional overturning circulation. Approaches to reducing numerical mixing will be described.