...
- performance (Philip Jones)
- Change number of layers to 64 or 128 for vector performance
- Always loop vertically to nVertLevels, or multiples of 8, with a mask ("Strip-mining")
Remove all allocatable arrays for GPUs: change scratch arrays to permanent arrays.
- reduce excessive diagnostics computations, other items, see: /wiki/spaces/OCNICE/pages/82673686
- analysis
- Continued work on analysis members and MPAS Analysis
- smooth automated implementation in ACME, working at high resolution.
- Sea-Level Rise Topic
- non-Boussinseq / TEOS-10 (Mark Petersen)
- needed for steric SLR
- expected for next generation ocean models
- wetting-drying / massless layers (Xylar Asay-Davis)
- needed for ocean / land-ice coupling, moving grounding line
- needed for coastal inundation
- high-order pressure gradient (Mark Petersen)
- mitigates errors associated with strongly sloped layer interfaces
- quasi-Lagrangian vertical coordinate (regridding/remapping, hybrid vertical coordinates) (Luke Van Roekel (Unlicensed))
- exploits a capability built into MPAS-Ocean
- should improve representation / preservation of water masses
- super-cycling of tracer transport
- needed to make BGC more computationally tractable
- Mesoscale eddy closure
- funded by SciDAC3, in code base, needed to be validated in global configurations
- thin-wall approach for bottom bathymetry
- a great idea that we never have time to implement
- improved vertical turbulence scheme (Luke Van Roekel (Unlicensed))
- needed to get beyond KPP
- prognostic SW penetration model (LV notes: mostly done already, just needs hooks to BGC)
- needed to improve ocean boundary layer representation
...