Q: Why does the bottom of the Kraus-Turner mixed layer look bumpy on occasion?  
A: This is due to a truncation error resulting from the mismatch between the isopycnic layer structure and the horizontally continuous mixed-layer density field.  

Q: Can anything be done about this?  
A: Truncation errors can, by definition, be suppressed by increasing model resolution -- in this case, the number of isopycnic layers. Another possible remedy is adoption of a hybrid (mixed depth-isopycnic) coordinate allowing the use of more sophisticated mixed-layer closures. Such a scheme is currently being developed and tested for hidden pitfalls.  

Q: Under what conditions do surfaces rhor = constant differ from neutral surfaces? (rhor = potential density referenced to r × 100 bar.)  
A: This happens if salinity varies along the rhor surface, and if the local pressure differs markedly from the reference pressure.  

Q: How should one select a reference pressure?  
A: The No.1 concern is prevention of coordinate folding (sign changes in d rhor/dz) in the mass field used for model initialization. One should also try to minimize the reference pressure bias.  

Q: Is there a reference level r that will lead to a set of nonfolding rhor surfaces in today's world ocean?  
A: No -- at least not for arbitrarily fine vertical resolution.  

Q: What does vertical resolution have to do with this?  
A: The smoothing of the density profile needed to prevent folding will be overshadowed by the density discretization process if the density steps are chosen sufficiently large.  

Q: So what should one do if high vertical resolution is essential in a particular application? 
A: The best strategy, again, might be to switch to a hybrid (mixed depth-isopycnic) coordinate in the folding region. 

Q: What happens when rho_r surfaces are non-neutral? 
A: The simplicity of the 2-dimensional transport algorithm will not be affected, regardless of how non-neutral the rho_r surfaces are, because they are still material surfaces under adiabatic conditions. However, the clean separation of isoneutral and dianeutral mixing will be lost to some extent. A partial cure for this would be mixing tensor rotation. 

Q: Given that both neutral and rho_r surfaces are material (under adiabatic conditions) and thus allow transport to be a 2-D problem, why not use neutral surfaces as layer interfaces? 
A: Neutral surfaces are only locally defined. Extending them over large horizontal distances leads to ambiguities, such as helicity. 

Q: Is thermobaricity a problem in simulating the thermohaline circulation in an isopycnic coordinate model? 
A: Not if model dynamics are expressed in terms of virtual potential density (defined in the 1999 paper by Sun et al.). 

Q: Is steep topography a problem in isopycnic models? 
A: No. Isopycnic models are free of the pressure gradient error found in sigma coordinate models.