Changeset 0b00607 in flex_extract.git for for_developers/Sphinx/source/Documentation/vertco.rst
- Timestamp:
- Jul 29, 2019, 8:23:57 AM (5 years ago)
- Branches:
- master, ctbto, dev
- Children:
- bc27d19
- Parents:
- 41c9dbc
- File:
-
- 1 edited
Legend:
- Unmodified
- Added
- Removed
-
for_developers/Sphinx/source/Documentation/vertco.rst
r41c9dbc r0b00607 1 =================== 1 ******************* 2 2 Vertical Coordinate 3 =================== 3 ******************* 4 5 Calculation of vertical velocity and preparation of Output-files 6 ================================================================ 7 8 ``flex_extract`` has two ways to calculate the vertical velocity for ``FLEXTRA``/``FLEXPART``: 9 (i) from the horizontal wind field, 10 (ii) from the MARS parameter 77, which is available for operational forecasts and analyses since September 2008 and for reanalysis datasets **ERA5** and **CERA-20C**. 11 12 Especially for high resolution data, use of the ``MARS`` parameter 77 is recommended, 13 since the computational cost (measured in ECMWF HPC units) is reduced by 90-95% at 14 T799. The extraction time, which depends heavily also on the performance of ``MARS``, is 15 generally reduced by 50% as well. The ``MARS`` parameter 77 is then multiplied by ``dp/deta`` to 16 give a vertical velocity in Pa/s as needed by ``FLEXPART``. 17 18 Calculation from the horizontal wind field is still required for historical case studies using 19 **ERA-40**, **ERA-Interim** or operational data prior to September 2008. 4 20 5 UNDER CONSTRUCTION 21 22 Calculation of vertical velocity from horizontal wind using the continuity equation 23 =================================================================================== 24 25 The vertical velocity is computed by the FORTRAN90 program ``CONVERT2`` in the ECMWF 26 vertical coordinate system by applying the equation of continuity and thereby ensuring mass consistent 3D wind fields. A detailed description of ``CONVERT2`` can be found in the 27 documents v20_update_protocol.pdf, V30_update_protocol.pdf and 28 V40_update_protocol.pdf. The computational demand and accuracy of ``CONVERT2`` is highly 29 dependent on the specification of parameters ``GAUSS``, ``RESOL`` and ``SMOOTH``. The 30 following guidance can be given for choosing the right parameters: 31 32 * For very fine output grids (0.25 degree or finer) the full resolution T799 or even T1279 of the operational model is required (``RESOL=799``, ``SMOOTH=0``). The highest available resolution (and the calculation of vertical velocity on the Gaussian grid (``GAUSS=1``) is, however, rather demanding and feasible only for resolutions up to T799. Higher resolutions are achievable on the HPC. If data retrieval at T1279 needs to be performed on *ecgate*, the computation of the vertical velocity is feasible only on the lat/lon grid (``GAUSS=0``), which also yields very good results. Please read document v20_update_protocol.pdf-v60_update_protocol.pdf to see if the errors incurred are acceptable for the planned application. 33 * For lower resolution (often global) output grids, calculation of vertical velocities with lower than operational spectral resolution is recommended. For global grids the following settings appear optimal: 34 - For 1.0 degree grids: ``GAUSS=1``, ``RESOL=255``, ``SMOOTH=179`` 35 - For 0.5 degree grids: ``GAUSS=1``, ``RESOL=399``, ``SMOOTH=359`` 36 - Calculation on the lat/lon grid is not recommended for less than the operational (T1279) resolution. 37 - If ``GAUSS`` is set to 1, only the following choices are possible for ``RESOL`` on *ecgate*: 159,255,319,399,511,799, (on the HPC also 1279, 2047 in future models). This choice is restricted because a reduced Gaussian grid is defined in then ECMWF EMOSLIB only for these spectral resolutions. For ``GAUSS=0``, ``RESOL`` can be any value below the operational resolution. 38 - For ``SMOOTH`` any resolution lower than ``RESOL`` is possible. If no smoothing is desired, ``SMOOTH=0`` should be chosen. ``SMOOTH`` has no effect if vertical velocity is calculated on lat\/lon grid (``GAUSS=0``). 39 * The on demand scripts send an error message for settings where ``SMOOTH`` (if set) and ``RESOL`` are larger than 360./``GRID``/2, since in this case, the output grid cannot resolve the highest wave numbers. The scripts continue operations, however. 40 * Regional grids are not cyclic in zonal directions, but global grids are. The software assumes a cyclic grid if ``RIGHT``-``LEFT`` is equal to ``GRID`` or is equal to ``GRID``-360. 41 * Finally, model and flux data as well as the vertical velocity computed are written to files ``<prefix>yymmddhh`` for application in ATM modelling. If the parameters ``OMEGA`` or ``OMEGADIFF`` are set, also files ``OMEGAyymmddhh`` are created, containing the pressure vertical velocity (omega) and the difference between omega from ``MARS`` and the surface pressure tendency. ``OMEGADIFF`` should be zero except for debugging, since it triggers expensive calculations on the Gaussian grid. 42 43 44 Calculation of vertical velocity from pre-calculated MARS parameter 77 45 ====================================================================== 46 47 Since November 2008, the parameter 77 (deta/dt) is stored in ``MARS`` on full model levels. ``FLEXTRA``/``FLEXPART`` in its current version requires ``deta/dt`` on model half levels, multiplied by ``dp/deta``. In ``flex_extract``, the program ``CONVERT2`` assumes that this parameter is available if the ``CONTROL`` parameter ``ETA`` is set to 1. 48 49 It is recommended to use the pre-calculated parameter 77 by setting ``ETA`` to 1 whenever possible. 50 51 Setting parameter ``ETA`` to 1 normally disables calculation of vertical velocity from the horizontal wind field, which saves a lot of computational time. 52 53 .. note:: 54 However, the calculation on the Gaussian grid are avoided only if both ``GAUSS`` and ``ETADIFF`` are set to 0. Please set ``ETADIFF`` to 1 only if you are really need it for debugging since this is a very expensive option. In this case ``ETAyymmddhh`` files are produced that contain the vertical velocity from horizontal winds and the difference to the pre-calculated vertical velocity. 55 56 The parameters ``RESOL``, ``GRID``, ``UPPER``, ``LOWER``, ``LEFT``, ``RIGHT`` still apply. As for calculations on the Gaussian grid, the spectral resolution parameter ``RESOL`` should be compatible with the grid resolution (see previous subsection). 57 6 58 7 59 .. toctree::
Note: See TracChangeset
for help on using the changeset viewer.