This repository contains new global bathymetry datasets for ocean models using the ORCA family of grids (Madec and Imbard (1996)). The source data set is GEBCO 2021. Horizonally regridded bathymetries are provided on the eORCA1, eORCA025, and eORCA12 grids (with nominal 1°, 1/4° and 1/12° resolution respectively).
The 2D bathymetries on the model grid can be used to generate the model 3D grid and masks. This process depends on the choice of vertical coordinate. In this release we also provide 3D model grids and masks for a z-level vertical grid with partial cells at the ocean floor.
The bathmetries and 3D grids and masks in this repository were created for use in the NEMO model (Madec et al (2019)) and much of the processing uses grid definition files in NEMO format. In principle there is no reason why these grid and mask definitions should not be used in other ocean models subject to the restriction that they are based on the Arakawa C-grid discretisation, with a land sea mask centred on tracer points.
All code and scripts can be found in the src directory and associated namelists, data files in the etc directory of this repository unless otherwise stated.
- Background and motivation
- Creation of model bathymetries
- Projection onto model vertical grid
- Test results
- References
The nominal resolution of the GEBCO 2021 dataset is 15 arc seconds (although the effective resolution over most of the domain is much less than this). The approach to horizontal regridding is to take the median value of all source data points inside a model grid cell. The python code used is bathy_regrid_horiz.py which is based on the algorithm from NEMOBAT by J-M. Molines whereby the model grid cells are approximated as quadrilaterals in latitude-longitude space.
General options: A minimum depth at sea points of 3.0m was set using the min_depth keyword. Note that as detailed below, the smoothing step effectively resets this to 5.0m.
Coastline: A decision was made to maintain the same coastline as the previous ORCA bathymetries to faciliate the use of the new bathymetries in coupled models, where a change in the surface land-sea mask can necessitate a lot of work [what?]. This was achieved by setting the mask_in keyword to True, which tells the routine to impose the land-sea mask from the mesh file on the new bathymetry. The choice to impose the old coastline raises the issue of how to set the ocean depth at points which would be land points based on GEBCO 2021 but were sea points with the old coastline. These were dealt with as follows.
- Recalculating the depth based on the median of only those source data points in the model grid cell that are below sea level.
- Remaining problematic points after step 1 are filled by iterative extrapolation from neighbouring sea points (set fill_bad_cells keyword to True).
- Any remaining problematic points after steps 1 and 2 are set to the minimum depth (so long as the min_depth keyword is set.)
Inland Seas: Large inland seas such as the Caspian Sea, the North American Great Lakes and Lake Victoria in Africa are included in the ocean model primarily to provide a good lower boundary condition for the atmosphere in coupled models. The bathymetry of the inland seas is included in GEBCO, which provides topography over land, but the elevation of the lake surfaces, which is required as a reference point for the lake bathymetries, is not provided. A very accurate simulation of the inland seas is not required in the context of these global models, so the bathymetries of the inland seas have simply been copied from the previous bathymetries. NEMO provides masks to define the location of inland seas, and these masks have been used to facilitate the masking out the inland seas in the input land-sea mask, and the subsequent copying of the old bathymetry using closea_fill.py and closea_copy_bathy.py. (Note that bathy_regrid_horiz.py currently cannot deal with inland seas: it assumes ocean points are those with elevation below mean sea level).
In order to avoid forcing the model at the gridscale, a single pass of a 2nd order Shapiro filter was applied, following the algorithm of Francis (1975). This filter was chosen since it preferentially damps variations near the gridscale.
The smoothing was done using a modified version of the cdfsmooth.f90 module from the CDFTools package. The modified version is available at the JMMP fork of the main repository.
Where narrow strait and channels are only marginally (or not) resolved by the model grid, the horizonal regridding and smoothing process will partially or fully block these channels. There are many examples of channels in the global ocean which have throughflows, exchange flows or overflows which are very important for interbasin exchange of water masses. For these channels we apply hand editing of the smoothed bathymetry to reset the ocean depths in the channels with the aim of modelling approximately the correct volume throughflow or exchange.
Since the channels in question are not fully resolved, this process can be viewed as a crude subgrid parametrisation. The channels are reset to approximately the observed channel depth, and the model integrated for a few decades. If the volume flow through the channel is judged to be still too far from the observed estimates, the depth in the channel can be adjusted, or alternatively the side or bottom friction in the channel changed. Typically for a low resolution model a single-point channel in the model can be much wider than the real channel and so prone to allowing too much throughflow, and this can be mitigated by increasing the bottom or side friction in the model. The bathymetries in this repository were created for use in the NEMO model. Global NEMO configurations are typically run with a free-slip lateral boundary condition; in order to reduce flow in some of the channels we change the slip-condition in the channels to a partial slip or no slip condition.
Note that it is also possible in NEMO configurations to adjust the width of channels by adjusting the horizontal scale factors e1t and e2t locally but we have not done that for these configurations.
The channels and flows that were checked were the North Atlantic overflows: Denmark Strait and Faroe Bank Channel; exchanges with marginal seas: Gibraltar Strait, Bab el Mandeb and Strait of Hormuz; and the exit channels for the Indonesian Throughflow: Lombok Strait, Ombai Strait and Timor Channel. For ORCA1, it was found necessary to edit all of these apart from the Strait of Hormuz. For ORCA025, it was found necessary to edit the Denmark Strait, Faroe Bank Channel, Strait of Gibraltar, Bab el Mandeb and the Lombok Strait. For ORCA12, no editing was necessary. The code used to reset the bathymetry at chosen points was edit_field.py with associated data files eORCA1_bathy_edits.dat and eORCA025_bathy_edits.dat.
These are the specific linux commands used to create the 2D bathymetries.
!!CHECK THAT THESE COMMANDS WORK!!
closea_fill.py -C domcfg_eORCA025_v2.nc -M mesh_mask_eORCA025-GO6.nc \
-o tmask_eORCA025-GO6-CloseaFill.nc
bathy_regrid_horiz.py -B GEBCO_2021_sub_ice_topo.nc -S gebco \
-M tmask_eORCA025-GO6-CloseaFill.nc -m \
-d3.0 -F -o bathy_eORCA025_noclosea_from_GEBCO2021.nc
fill_zero...
cdfsmooth -f bathy_eORCA025_noclosea_from_GEBCO2021_FillZero.nc -c 2 -t S \
-n namelist_shapiro.txt
closea_copy_bathy.py ...
edit_field.py -i bathy_eORCA025_noclosea_from_GEBCO2021_FillZero_S1TT.nc \
-o bathy_eORCA025_noclosea_from_GEBCO2021_FillZero_S1TT_edit.nc \
-v Bathymetry -e eORCA025_bathy_edits.dat
where the files namelist_shapiro.txt, eORCA1_bathy_edits.dat, and eORCA025_bathy_edits.dat can be found in the etc directory of this repository.
The bathymetries on the model horizontal grid can be used to generate model 3D grids and masks, a process that depends on the choice of vertical coordinate system. Here we provide 3D grids and masks (NEMO domain_cfg.nc files), for a z-coordinate vertical grid with 75 levels and partial grid cells allowed at the base of the water column (REFS). The distribution of the 75 levels is based on a double tanh function with a resolution of 1m at the surface (REF).
The 3D grids and masks were created using the DOMAINcfg tool available as part of the NEMO release, where the link points to the revision used. The namelists used for each configuration are available in the etc directory of this repository.
https://github.com/meom-group/CDFTOOLS
https://github.com/JMMP-group/CDFTOOLS, hash code: 136b95b6ac4fefad6973a4c15ca953dcae65dfc6
GEBCO Compilation Group (2021): GEBCO 2021 Grid, doi:10.5285/c6612cbe-50b3-0cff-e053-6c86abc09f8f
Francis, M. (1975): The use of a multipoint filter as a dissipative mechanism in a numerical model of the general circulation of the atmosphere, Quart. J. Roy. Met. Soc. (1975), 101, pp. 567-582, https://doi.org/10.1002/qj.49710142913
Madec, G. and Imbard, M. (1996): A global ocean mesh to overcome the North Pole singularity, Climate Dynamics, 12, 381–388, https://doi.org/10.1007/BF00211684
Madec, G. and system team, N.: Nemo Ocean Engine - version 4.0.1, Notes du Pôle de modélisation de l’Institut Pierre-Simon Laplace (IPSL): (27)., https://doi.org/10.5281/zenodo.3878122, 2019
Molines, J.-M., NEMOBAT, https://github.com/molines/NEMOBAT