We’re running the IUM13.5 on ARCHER2 suite (dh755), everything has been working smoothly over all sea points but we wanted to investigate a continental setting. I noticed that setting the landmask to a constant value resulted in 1 unresolved point in the reconfiguration step so I created a land mask ancillary for all land and encountered the same error. However if I added in a strip of sea at the boundary then the reconfiguration is successful and the model runs.
Does anyone know if this is expected? Is there a minimum number of sea points? I think I wouldn’t expect a minimum number of sea points as I’m sure I could run a continental only LAM using the standard UM.
The error I received was this
????????????????????????????????????????????????????????????????????????????????
???!!!???!!!???!!!???!!!???!!! ERROR ???!!!???!!!???!!!???!!!???!!!
? Error code: 47
? Error from routine: RCF_SPIRAL_CIRCLE_S
? Error message: A point has been left as still unresolved
? Error from processor: 0
? Error number: 5
????????????????????????????????????????????????????????????????????????????????
Our work out of having a few sea points is working for now but not ideal so any help in understanding this would be much appreciated.
yes that’s the work around mask with sea points included (which reconfigures and runs fine), the original mask looked more like the /work/n02/n02/hburns/H2X/ANCILS/mask ancillary file (but the correct 200x200 grid). I’ll try checking 13.3 version with the constant value set again just incase.
You are trying to reconfigure an aquaplanet start file onto a land-only domain – I’ve never seen that attempted before - this may need help from the MO.
Actually it doesn’t seem unreasonable that the spiral search fails in this case.
Yes I have some island idealised domain examples from Rachel Stratton, but not seen any all land examples so I’ll email some metoffice contacts and see if they know you must have a sea point in an idealised domain and if so why! Thanks for your thoughts and I’ll give an earlier UM version a go when i get a chance just in case.
Just an update on this. Some helpful metoffice contacts have been helping me get to the bottom of this.
You can not run on a Cartesian grid with all land points without code modifications to the coastal adjustment scheme if using spiral circle search (only option in 13.5). I’ve been sent a test branch to pull in the required modifications and some information on generating Cartesian ancillaries.
I’ll update in more detail once its all definitely working just stopping the thread from closing before I can post the fix just in case anyone in the future is wondering!
Huzzah! It’s all running fine and I have my initial IUM sims over land completed.
The issue: the spiral circle search method used for coastal adjustment does not work for a Cartesian grid used by Idealised bi-cyclic LAMS
The current solution: Simply turn off the spiral circle search method that is the only coastal adjustment method available in recent UM versions (>vn12.0) as sally has done here in the recon routines (utility/qxreconf)
Currently this is coded up as just commenting out the offending code. If I have time I might look at adding in this as a switch! I’ll make my own branch either way as this branch is not just for addressing this issue
A note on Ancillaries for Cartesian grids This is not supported by Xancil and ANTS, I’m currently just using constant fields which now work. You can try using the coordinates from the output and setting ignore ancil grid check set to true but it might not work, you may need to perform some mule wizardry to generate the ancil