Land-sea mask problems

The problem I asked in the latter half of ticket #3561 in the previous system has not been resolved.

In recent nesting suite I have 3 nests. The land sea mask is created fine when I have the smallest nest off Porto (u-cf930, u-cg028). However, the model suffers from the LSM problem when I set the smallest nest off SW England/Wales (u-cf597).

As shown here http://cms.ncas.ac.uk/ticket/3561#comment:13 the LSM has value 1 everywhere even though the nest includes areas over sea. Sometimes the nest is completely corrupted (also shown in the same ticket).

I initially had a problem in the large nest covering the North Atlantic that land-sea mask was wrong in the east of the prime meridian.
This problem has not been resolved but avoided by finding a domain that does not cause the problem. It looked like the problem happened when a region is set so that it extends to the Eastern Hemisphere only a little bit.

I feel like these problems may be related.

People suggest that I should do ANTS but it isn’t currently working. Also when I turn on ANTS LSM seems to be created before ANTS fails. So I think ANTS might not provide a solution.

Any idea?

I have another problem, which is possibly related to this.

This block can be skipped:
I have got advice and now I may be able to create land-sea mask and veg fraction ancils with my copy of u-bu503 (u-cg973) using ANTS. (the land region has value not 1 but a strange value when viewed in xconv so I’m still not completely sure if this is working. I have to test. If this is proved to be working,) I think I can copy these ancils to my suite.

Assuming the LSM problem is solved, my next problem is that ancil_sulpdms for the smallest nest (swBrit) in my latest suite u-ch092 (as well as older suite) fails. It says this;

 Writing out dataset
Number of fields=   12. POINT        163 LAT    52.0625 LONG   355.9393
 POINT        164 LAT    52.0623 LONG   355.9429
 POINT        411 LAT    52.0165 LONG   356.8369
 POINT       1039 LAT    52.0147 LONG   356.8293
.............................
 POINT     293733 LAT    51.0234 LONG   355.7759
 POINT     362250 LAT    50.6910 LONG   357.4272
UNRESOLVED GRID POINTS IN DATASET
 Number of points unresolved is  22
 Search radius  1
 NO DATA FROM WHICH TO SET UNRESOLVED POINTS
 ***ERROR: No source data available in target domain

I tried increasing search_radius to 1000 but the number of points unresolved does not change. What else can I do?

Masaru

Hi Masaru

The mask used here still appears to be incorrect (comprising all land) - the CAP uses the mask to determine unresolved points, so this may be the cause of the problem – can you try running with the correct mask from ANTS?

Grenville

Hi Grenville,

I turned on ANTS and ancil_ants_vegfrac failed for both of two nests. But ancil_sulpdms succeeded for the larger nest (Brit2Port) so I guess ANTS may not be the cause of the failure of ancil_ants_vegfrac? What do you think?

I heard that ANTS environment has been removed from Monsoon. However, for some reason it may be working fine for u-cg973 which is my copy of Claudio Sanchez’s u-bu503. So I was thinking maybe I can copy mask and vegfrac ancils from this job. The only problem is the failure of sulpdms. What do you think about this?

Thanks,
Masaru

I thought maybe I could

  1. let the run fail just as above,
  2. copy qrparm.mask_cci and qrparm.veg.frac_cci created in u-cg973 and link them to qrparm.mask and qrparm.veg.frac respectively,
  3. reset status of ancil_ants_vegfrac to succeeded, and then
  4. trigger run on ancil_sulpdms.

I thought it might either run fine or get the same error as before, but it only says “file %r not found…”. I have no clue what file it is looking for. Any idea?

Thanks,
Masaru

But the swBrit mask in u-cg973 is all land? I am at a loss to know why the land fraction appears to be correct but not the mask.
Grenville

Ahh - I’m looking at the wrong file – how was swBrit/qrparm.mask_cci created?

And yes, I would try using this mask for calculating ancil_sulpdms

Grenville

Has the cci file been created from a different source dataset?

I think so. The resolution is way higher in cci (if this is what you meant).

Yes - thanks, I finally begin to follow what is happening in these suites.

“file %r not found…” was caused by my mistake. I had accidentally uncommented a comment line in my .bashrc.

So I did

  1. let the run fail just as above,
  2. copy qrparm.mask_cci and qrparm.veg.frac_cci created in u-cg973 and link them to qrparm.mask and qrparm.veg.frac respectively,
  3. reset status of ancil_ants_vegfrac to succeeded, and then
  4. trigger run on ancil_sulpdms.

and got exactly the same error as above;

UNRESOLVED GRID POINTS IN DATASET
Number of points unresolved is 22
Search radius 1
NO DATA FROM WHICH TO SET UNRESOLVED POINTS
***ERROR: No source data available in target domain

I’ve increased search radius to 1000 but it says ‘Search radius 1’ above.

app/ancil_sulpdms/rose-app.conf:13:SEARCH_RADIUS=1000
app/ancil_sulpdms/rose-app.conf:14:SEARCH_RADIUS_COL=-1
app/ancil_sulpdms/rose-app.conf:15:SEARCH_RADIUS_ROW=-1

Did I need to change SEARCH_RADIUS_COL and SEARCH_RADIUS_ROW in some way?

What is the full path of the file these error messages come from?

This is from
/home/d03/myosh/cylc-run/u-ch092/log/job/20190710T0000Z/Regn1_swBrit_ancil_sulpdms/NN/job.out

Thanks.

It looks like it is using the mask file Regn1/swBrit/qrparm.mask_sea_nolakes to create this ancil and this mask is all land points, hence why you still can’t create the sulpdms ancil.

HI Jeff,

Thanks for looking into the problem.
How do I find out that kind of things? Did you check the built codes? Where are they?

But qrparm.mask_sea_nolakes is not created correctly even in u-cg973. So there is nothing I can do.

Why do you think the mask is not created correctly for this nest only? Is there any remedy for it?

If you search the output file it says which files are opened, search for Attempting to open file.

I don’t know why the mask isn’t created correctly, I’ll see if I can find any reason and let you know if I do.

Thank you, Jeff.
I created symbolic links to the files created by ANTS in u-cg973 like below (this is in /home/d03/myosh/cylc-run/u-ch352/share/data/ancils/Regn1/swBrit/ ).

qrparm.mask -> qrparm.mask_cci
qrparm.mask_cci -> /home/d03/myosh/cylc-run/u-cg973/share/data/ancils/Regn1/swBrit/qrparm.mask_cci
qrparm.mask_sea_nolakes -> /home/d03/myosh/cylc-run/u-cg973/share/data/ancils/Regn1/swBrit/qrparm.mask_cci
qrparm.mask_sea_nolakes_bu (**original file renamed)
qrparm.veg.frac -> qrparm.veg.frac_cci
qrparm.veg.frac_cci -> /home/d03/myosh/cylc-run/u-cg973/share/data/ancils/Regn1/swBrit/qrparm.veg.frac_cci

Now sulpdms fails for wallclock time limit. I extended it to 60 minutes but it fails after running for an hour.
This may mean qrparm.mask_cci cannot be used in place of qrparm.mask_sea_nolakes in sulpdms…

Masaru

You can tell sulpdms where to get its mask by setting MASKIN in ancil_sulpdms->env->General interpolation ancillary settings

qrparm.mask_cc has a grid of 630x630 points but the swBrit domain in u-ch092 is 315x315 – the mask needs to be on the same grid. Can you set up u-cg973 to create a mask on the correct grid?

Grenville

Thank you, Grenville.
But the latest suite is u-ch352 and they do seem to me to have the same nest settings to u-cg973. Maybe I forgot committing the suite…just did it.

Masaru

The sulpdms is getting stuck doing a spiral search for unresolved points. A qrclim.sulpdms does have data for the first 4 months - the field value is constant in the swBrit region – not surprising given the very low resolution of the source data.

It might be simpler to create the remaining sulpdms data by hand (with a bit of python.)

Grenville