...
Configurations for INT2LM and COSMO for NWP as well as climate applications are available together with a short description of the configuration (namelist) parameters. The configurations are made available for the current operational applications setups of COSMO and the COSMO-CLM setup which is recommended by the CLM-Community. Please visit the namelist-tool portal for detailed information 🔑 for detailed information.
WG SUPTECH strongly recommends to use ONLY the recommended COSMO-CLM configuration for simulations of the EURO-CORDEX domain (visit the namelist-tool 🔑 for detailed information about the setup). For every other configuration, even if it is only slightly different compared to the recommended setup (e.g., domain size, resolution, atmosphere height, time step, tuning parameters, etc.), tests are necessary to verify the reliability of the new configuration. These tests should include tests with at least two different time steps and the results for PMSL should be analyzied (e.g. RMSE). If the values differ significantly, the results should be compared to observations or a simulation with a third time step should be performed to figure out which of the previous simulations is more reliable. Please be aware, that the time steps have to fulfill the CFL criterion in any case.
If you want to run the model on another domain than Europe, you should follow the following steps to get a reasonable model setup:
Try to get a namelist as a starting point:
Check if somebody has already performed a simulation on this domain or a similar domain. If this is the case try to get namelist from the namelist tool or the person directly and use this namelist as a starting point.
If there is no simulation for your target domain, or if you can't get a namelist, check if a simulation for a domain in a similar climate zone (mid-latitudes, tropics, maritime, continental, flat, mountainous) has been done and use this namelist as starting point.
Take the recommended COSMO-CLM version and incorporate the domain specific changes from the namelist of step 1.
Perform an ERA-Interim simulation (ideally for 30 years but at least for 10 years) and analyze if the results are reasonable.
Test as many different
setupsconfigurations as possible to find the best possible setup for your domain. For this step, the method proposed by Bellprat et at. can be used (Bellprat. O, S. Kotlarski, D. Lüthi, C. Schär (2012): Objective calibration of regional climate models. Journal of Geophysical Research, 117, D23115, DOI: 10.1029/2012JD018262).
Perform tests with a similar setup but different time steps to avoid wrong results as a consequence of the time step problem
(Qne. One time step can give very different results for e.g. pressure, precipitation compared to all other timesteps. This time step should not be used in climate simulations. For the EURO-CORDEX domain, the recommended configuration gives
wrongunrealistic results with a time step of 120 seconds. Since the problem is very likely resolution and domain dependent, no general recommendation can be given.)
Choose the best setup for your domain according to your tests.
Send your namelist to Andreas Will (andreas.will[at]b-tu.de) for inclusion in the namelist tool.
Write a short report on your tests and the outcome and send it to the coordinator of the WG EVAL.
WG SUPTECH strongly recommends to use ONLY the recommended COSMO-CLM configuration for simulations in the EURO-CORDEX domain (visit the namelist-tool for detailed information about the setup).
For every other configuration, even if it is only slightly different to the recommended setup (e.g., domain size, resolution, atmosphere height, time step, tuning parameters, etc.), tests are necessary to verify the reliability of the new configuration.)
In these tests the tested configuration should be run with at least 2 different time steps and the results should be compared w.r.t. the RMSE of the PSML. If they differ significantly, the results should be analysed against observations or, if a simulation optained by using a third time step gives comparable results to one of two other simulations, then as far as it is known up-to-now, the time steps with comparable results are assumed to be more realistic. Please consider, that the time steps of-course need to fulfil the CFL criterion.
For a completely different domain extensive tests are necessary to find a suitable configuration for that domain.
For more details on the proceeding, please use the forum!Please visit the forum for more details on the procedure and in case of questions!
🔑 = CLM-Community login required