Opened 2 months ago

Last modified 2 months ago

#2084 new help

Snow ancil problem

Reported by: rm024650 Owned by: um_support
Priority: normal Component: UM Model
Keywords: Cc:
Platform: UM Version: 8.5

Description

Dear CMS helpdesk

I am trying to setup a N96 GA6 run (job id XMKXK) with surface snow amount updated periodically from a modified snow ancillary file. The run setup is mostly a copy of the N96 GA6 template job.

I have never used a snow ancillary before. To make sure the UM is really reading and updating the surface snow amount from the file, I made a modified version of the standard qrclim.smow file and replaced the snow amount over Asia throughout the year with values similar to that over Antarctica. The modified snow ancil file is located at /work/n02/n02/rm024650/snow_ancil/smowv1_mod2.

The NRun finished without error. However, when I look at the STASH output (/work/n02/n02/rm024650/xmkxk/xmkxka.pf19810901) and the restart dumps, it seems that the UM is not reading or updating the surface snow amount from the modified ancil. A quick look at the snow amount over land in the reconfigured dump (xmkxk.astart) with Xconv shows a blank plot with all grids containing the missing value (-1073741824.000000) instead of what is in the modified snow ancillary.

I am aware of a very old ticket from Andy Turner (#138) and the solution seems to be a user prestashmaster file. I copied the file from Jeff’s directory on PUMA, modified it to fit the current format (/home/rm024650/snowdepth_update on PUMA) and included it in the UMUI, but it is still giving me missing values for snow amount over land in the reconfigured dump.

I am wondering if I am doing anything wrong (especially with the prestashmaster file) such that something went wrong when creating the reconfigured dump? Any advice would be great. Thank you.

Mike

Attachments (1)

snow_ancil.pdf (143.1 KB) - added by rm024650 2 months ago.

Download all attachments as: .zip

Change History (4)

comment:1 Changed 2 months ago by jeff

Hi Mike

I don't think the prestashmaster file is needed any more. If you click on the Help button in the umui Soil moisture and snow depth panel you will see they have made changes to how this works from version 8.4.

Also the snow amount being all missing data in the reconfigured dump is normal, all fields set to updated in the umui are set to missing data in the reconfigured dump. This is because data for these fields is read from the ancillary file and not from the start dump.

The model is definitely reading the snow amount ancillary file when running so I suspect it is working but somehow resetting the snow depth to more normal values over your square region. I suggest you run the model with and without your modification and difference the results to see if there are any changes.

Jeff.

comment:2 Changed 2 months ago by rm024650

Hi Jeff

Thank you for the quick reply.

So I think what I should do now is remove the prestashmaster file, then re-run the experiment. First I will run with the standard unchanged ancil, then with a modified ancil with zero snow over Tibet (which is actually what I intended to do anyway).

Since I am updating from the ancil every 1 hour, plotting the snow amount from the 6 hourly diagnostic output (without averaging / min / max) should give a clear view on the difference between both runs.

comment:3 Changed 2 months ago by rm024650

As a follow up, I ran two extra 1 year experiments.

Job XMKXI is a ‘control’ run with the standard unmodified snow ancil, copied from /work/n02/n02/hum/ancil/atmos/n96e/orca025/smc_snow/gswp2_hwsd_vg/v1/qrclim.smow.

Job XMKXJ is a sensitivity test, using a modified version of the standard snow ancil with surface snow amount inside the box 70-110E, 25-55N (x=38:60, y=93:116) set to zero throughout the year (/work/n02/n02/rm024650/snow_ancil/smowv1_mod1).
Apart from the snow ancil, both jobs are identical.

The modified snow ancil was created by reading and changing data from the standard file with matlab, then saving as a new netcdf file. After adjusting the missing data value with xconv, the modified ancil was then created with xancil. Only the surface snow amount was changed and I used the ‘calculate snow edge’ option in xancil for snow edge. See attached pdf (fig 1) for settings of the xancil.

Results from both runs are quite confusing. Fig 2a shows the time series of surface snow amount (6 hourly spontaneous output, no min / max / averaging) averaged over the box for both runs. Clearly the snow amount is far from zero in the sensitivity run. A similar plot fig 2b created with data from the standard snow ancil file is also included for comparison. The model output are located at /work/n02/n02/rm024650, output stream F.

I have set the updating frequency to 1 hour in both runs, so I suppose any surface snow should either melt within an hour or reset to zero in the sensitivity run? Also, why is the time series of the control different from the ancil? Did I do something foolish when setting up the ancil / job? Thanks for any advice.

Changed 2 months ago by rm024650

Note: See TracTickets for help on using tickets.