Steps to setup new nested domains

From Geos-chem
Revision as of 18:10, 1 November 2016 by Linzhang (Talk | contribs) (First version of setup new nested domains (2016.10.31))

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

It’s not difficult to set up new nested-grid domains in the model, but the process can be tedious.

  • First, prepare for the meteorological data for the new nested domain. Obtain the native-resolution global met fields (not sure if Harvard or Dalhousie still keep those files on their disks; if not, they can be downloaded from NASA) and trim them to save out those over the domain of interest. The GC support team may have scripts to do this step.
  • Second, change all the domain-specific parameters in the code to match with the new nested domain. I don’t have a complete list of those parameters. The easiest thing is to find all the places in the code that have the ‘if nested-grid’ statement. A couple of parameters are defined in CMN_SIZE and make sure to change these. If you use one of the existing nested-grid resolutions (0.5x0.667 or 0.25x0.3125), no change is necessary for the advection code. Only need to change the domain-specific parameters in input.geos.
  • Third, make sure the input data for offline emissions and emission-related files are ready for HEMCO. My understanding is that HEMCO should be ready to take care of emissions in the nested-grid runs on the fly, but you should test if this is the case.
  • Fourth, test-run and debug the model. You will need to archive boundary conditions from the global runs in this step. This step can be long, and we often find missing data or wrong parameters in this step, but nothing major since the code’s already set up to deal with nested-grid simulations. My experience is to do the testing in two steps. First, test the transport processes only by turning off emissions, chemistry, and deposition. After the first step passes the quality control (more on this below), turn on emissions/chemistry/deposition to test the full chemistry runs.
  • Fifth, test if tuning factors for dust emissions are needed for your nested domain because these emissions are resolution-dependent due to how they’re parameterized in the code. To get the tuning factors, run the nested-grid model for a full year, archive those emissions (no tuning applied), and compare the domain-wide total emissions with those from the coarser-resolution global run (i.e., 2x2.5) to derive the tuning factors. We did the same for lightning NOx and soil NOx a few years ago, but now that the model treats these emissions in a different way.

I think that’s all. For quality control of the new nested model, I use two criteria:

  • (1) the spatial distribution of all species needs to be consistent with the 2x2.5 runs – I always regrid the finer-resolution outputs onto the coarser-resolution for the purpose of direct comparison. Make sure to check all model levels.
  • (2) do a tracer mass calculation for the nested domain and compare it with the mass from the global run. The difference in mass should be less than a few percent and not enlarge as the simulation period gets longer. There won’t be a 100% match in tracer masses because the finer grids do not completely align with the coarser grids.

Well, we hope you won’t be discouraged by this long to-do-list. This has been done many times before and it’s definitely doable. If you have questions, feel free to contact us.

-- Yuxuan Wang, Lin Zhang, Jun Wang; 10/21/16