Replicating the behavior of the prior timeseries diagnostics

From Geos-chem
Revision as of 16:50, 19 July 2019 by Bmy (Talk | contribs) (Created page with "On this page we discuss our plans to replicate specialty diagnostic outputs (timeseries, local-time diagnostics, etc.) in with the GEOS-Chem History diagnostics.. == ND40 pla...")

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

On this page we discuss our plans to replicate specialty diagnostic outputs (timeseries, local-time diagnostics, etc.) in with the GEOS-Chem History diagnostics..

ND40 planeflight diagnostic

Bob Yantosca wrote:

For the “planeflight” diagnostic, we will investigate using a separate package (such as NOAA ObsPack, as recommended by Andy Jacobsen at IGC8) that can more efficiently store point data. Then new diagnostic package we are developing works well for array data but not as well for point data.

Mat Evans replied:

I’ve had a quick look at obspack. I’m not sure its well suited to our needs.
I personally don’t see much need to update the plane flight files. They are easy to understand at the moment and easy to manipulate being in ascii. The biggest problem at the moment is that the rate constants have fallen out of them because of the move to FlexChem. If we can get that sorted I don’t think there is a pressing need to update the format.

--Bob Yantosca (talk) 16:20, 15 November 2017 (UTC)

ND48 station diagnostics

We will keep the current ND48 station diagnostics (bpch format) in GEOS-Chem v11-02 until we can find a suitable replacement.

Jenny Fisher wrote:

It seems we are losing compatibility for both ND48 (station) and ND51 (satellite) diagnostics. These are probably the two my group uses most frequently (and replacing with e.g. hourly output everywhere will require a lot of extra disk space). If we are going to have regional subsetting of diagnostics available, then ND48 can probably be dealt with by subsetting down to the level of a single box – but ND51 is harder.

ND49 instantaneous timeseries diagnostic

You can replicate the ND49 bpch diagnostic (instantaneous timeseries) as follows:

  1. Define a diagnostic collection in HISTORY.rc:
    • Specify instantaneous output:
      • Set mode: instantaneous
    • Specify the desired output frequency:
      • Set frequency: 010000 for hourly output
      • Set frequency: 030000 for 3-hourly output
      • Set frequency: 120000 for 12-hourly output, etc.
        • You can also use the full date/time descriptor (00000000 010000, etc.)
    • Specify that a new netCDF file should be created every 24 hours:
      • Set duration: 240000
        • You can also use the full date/time descriptor (00000001 000000)
  2. Within this diagnostic collection, you can add archive any available quantity from the State_Met, State_Chm, or State_Diag objects.

Example:

#===================================================================
# Add an instantaneous timeseries collection to the master list
#===================================================================
COLLECTIONS: 'inst3hr',
::
#====================================================================
# Define the properties of the instantaneous timeseries collection 
#====================================================================
  inst3hr.template:      '%y4%m2%d2_%h2%n2z.nc4',
  inst3hr.frequency:      030000,
  inst3hr.duration:       240000,
  inst3hr.mode:          'instantaneous',
  inst3hr.fields:        'SpeciesConc_?ADV?', 'GIGCchem',
                         'Met_T',             'GIGCchem',
                         'DryDepVel_?DRY?',   'GIGCchem',
::

--Bob Yantosca (talk) 21:55, 7 June 2018 (UTC)

ND50 24-hr average timeseries diagnostic

You can replicate the ND50 bpch diagnostic (24-hr average timeseries) as follows:

  1. Define a diagnostic collection in HISTORY.rc:
    • Specify time-averaged output:
      • Set mode: time-averaged
    • Specify daily output frequency. This will also set the averaging interval to be daily.
      • Set frequency: 240000
        • You can also use the full date/time descriptor: 00000001 000000
    • Specify that a new netCDF file should be created every 24 hours:
      • Set duration: 240000
        • You can also use the full date/time descriptor: 00000001 000000
  2. Within this diagnostic collection, you can add archive any available quantity from the State_Met, State_Chm, or State_Diag objects.

Example:

#===================================================================
# Add an 24-hr average timeseries collection to the master list
#===================================================================
COLLECTIONS: avg24hr',
::
#===================================================================
# Define the properties of the 24-hr average timeseries collection
#===================================================================
  avg24hr.template:      '%y4%m2%d2_%h2%n2z.nc4',
  avg24hr.frequency:      240000,
  avg24hr.duration:       240000,
  avg24hr.mode:          'time-averaged',
  avg24hr.fields:        'SpeciesConc_?ADV?', 'GIGCchem',
                         'Met_T',             'GIGCchem',
                         'DryDepVel_?DRY?',   'GIGCchem',
::

--Bob Yantosca (talk) 21:58, 7 June 2018 (UTC)

ND51 satellite diagnostic

Aaron van Donkelaar wrote:

I wanted to double check about how the satellite-overpass diagnostic (ND51) was being included. There was some mention of it within the (netCDF diagnostic) presentation, but I wasn’t clear to me how the time-averaged netcdf collection could be made to output something similar to the current ND51.

My main concern was that ND51 samples between two defined local times once each day. From what I understood of the time-averaged collection via netcdf, it could be set to regularly sample between a defined interval of time (e.g. every six hours), but I didn’t understand how this could be set to sample between, say, 10am-12pm local time each day. Neither local time, nor once per day seemed an option.

Apologies if I’ve just misunderstood, but I thought I should check that an ND51-type output would still be possible.

Bob Yantosca replied:

At this time we don't yet have a good substitute for the ND51 satellite timeseries diagnostics. We may be able to replicate this with some type of satellite simulator package, but that remains to be seen as of yet. Right now we are focusing our efforts on implementing the netCDF diagnostics that are required for benchmarking. (Specialty diagnostics will be added later.)

We have designed the netCDF diagnostics for GC-Classic to replicate the behavior of the diagnostics in GCHP. Having the diagnostics work the same way in both configurations of GC facilitates benchmarking and comparing GC-Classic to GCHP. It also makes it easier for users to switch between GC-Classic to GCHP. But GCHP does not allow for local-time diagnostics. GCHP relies on the MAPL library, and its diagnostic package (aka "History") only allows for instantaneous or time-averaged output.

One could save e.g hourly timeseries, along with a local time field, to netCDF, and apply the local time comparison in post-processing. This would be doable but it would also end up generating a ton of output.

The good news is that we will preserve the bpch diagnostics in v11-02 (I need to update the wiki better to reflect that). We will not remove the bpch diagnostics until we have found a suitable replacement for the local-time diagnostics. So for the time being you can still rely on ND51.

--Bob Yantosca (talk) 15:33, 15 November 2017 (UTC)

netCDF diagnostics for GEOS-Chem specialty simulations

In order to prioritize the GEOS-Chem v11-02 (aka 12.0.0) release, we focused on creating netCDF diagnostics that correspond to the bpch diagnostics used for benchmarking GEOS-Chem. Following v11-02 we will start to add the netCDF diagnostics for the following specialty simulations.

NOTE: Emissions for any specialty simulation can be archived as diagnostic output directly by HEMCO. See the following section for more information.

--Bob Yantosca (talk) 13:50, 8 June 2018 (UTC)