Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

convert_mpas - questions about missing data in ncview plots

RSAPIAIN_DMC

New member
Hello.

Managed to run MPAS finally, and now converting to lat/lon (using convert_mpas ) the history and diagnostic file output to visualize with ncview, but I'm getting some zones without any data.

Centro-RM was run with the 3km static mesh as a base; Sinoptico-15km used the 15km static mesh as a base, 120 forecast hours, file output every 1-hr-forecast, using GFS 0p25 data ungribbed with WPS 4.1

I'm seeing that weird white area in the space... I guess it means there is no data there. I tested with various forecast hours and the white-area doesn't change at all; and it happens with all the variables.

The nlat and nlon were made in a best-guess given the meshes used in each case:
  • 3km means about 32 or 33 points per lat.degree; Centro-3k has 3-deg lat and 6 deg. lon in the pts file definition
  • 15km... I guess I interpolated at 10km by mistake (Sinoptico15 has only a small white patch on top, but the rest of the data has values)

Code:
cat Centro-RM-3k/target_domain
nlat = 99
nlon = 198
startlat=-32.0
startlon=-68.0
endlat=-35.0
endlon=-74.0

Code:
cat Sinoptico15/target_domain
startlat = -15.0
startlon = -115.0
endlat = -55.0
endlon = -65.0
nlat = 400
nlon = 500

I'm attaching what I get, and the simulation run logs:
only warning on initial conditions, no warning on border condition computations; some warnings on the atmosphere-run; no errors in log.

I haven´t used the files: include_fields , nor the exclude_fields (but definitely I'm thinking that I should).

My questions:

1.- For the MPAS simulation, should the area of simulation be larger by at least 1 deg on each side?
(for example, for Centro: rectangle( (-31.5, -67.5), (-35.5, -73.5 )) for the MPAS-run, and then perform the convert_mpas with the target_domain config specified previously?) )

2.- Or are those 'white areas', are something that is expected to happen ? Or maybe something with the specific GFS data that day?

Thank you in advance.
 

Attachments

  • Centro3km-cap04_diagnostic-file.png
    Centro3km-cap04_diagnostic-file.png
    133.8 KB · Views: 4
  • Sinoptico-15km_cap01.png
    Sinoptico-15km_cap01.png
    110.3 KB · Views: 3
  • Centro3km-cap01.png
    Centro3km-cap01.png
    162.3 KB · Views: 4
  • ncdumps-log-atmosphere-run.zip
    1.1 MB · Views: 0
The latlon plot looks weird. I am not sure whether it is because your original data is wrong or there are issues during the interpolation. Would you please upload a single mpas output file from your 3km run for me to take a look? Thanks.
 
Ming Chen, good morning.


Centro-RM-3km, using static-mesh 3km, is defined from lat,lon: (-32.0, -68.0) to (-35.0, -74.0)
Sinoptico15, using static-mesh 15km, is defined from lat,lon in clockwise order 4-points: (-15.0, -115.0), (-15.0, -65.0), (-55.0, -45.0), (-55.0, -135.0)

Please find them here: output nc files for Centro-RM-3km and Sinoptico15

In the meantime I'll try creating the domains with an additional padding and run sims. again.

EDIT:
The limited area mesh (.pts file) was created using: GitHub - MPAS-Dev/MPAS-Limited-Area: Python tool to create a regional subset of a global MPAS Mesh
 
Last edited:
Hello.

Tried a larger sim. with "padding": plot a smaller section of the full domain.
Now I have no white lines... so maybe that would be required for my smaller ones.

Also, I have a question: seems that convert_mpas has an issue with the point amount (maybe not): I would have more density of points in the longitudes when closer to the pole in latitude, than in the northern part of chile. Just curious about this.
  • would there be some way around this with convert_mpas, or would I have to resort to converting with cdo ? (but in CDO I need to generate a weights file and some other things)

Also, added in output nc files for Centro-RM-3km and Sinoptico15 (v3) a netcdf output file for this sim.
 

Attachments

  • ncview-proyeccion-con-padding.png
    ncview-proyeccion-con-padding.png
    311.9 KB · Views: 1
Sorry for the later answer. This is because I was out of office for a sick leave.

I looked at your data "OUTPUT_history.Centro-RM-3k.2025-01-17_00.00.00.nc", it seems that precip (rainc + rainnc) is zero everywhere. Further check of other variables like u10 gives me the value of "-1e+30" everywhere.

I am not sure whether this is because the file "OUTPUT_history.Centro-RM-3k.2025-01-17_00.00.00.nc" is damaged during transfer?

Looking at the plots you posted, I am suspicious that the white area is attributed to NaN values, which indicates that something went wrong in your case. I don't think 'convert_mpas' is the issue here.

Please double check your data and case, and keep me updated ...
 
Top