Hi there, I’ve been running a 1.5km suite (u-ed824, up-to-date version is committed) for the northern Antarctic Peninsula. There is a strange spatial pattern emerging, even when I run the suite with the default topography, and I’m having trouble identifying where it’s coming from, even when looking into the ancils. It looks like a surface-only issue, as the feature is most visible in the surface net radiative and heat fluxes. Do you have any suggestions? I’m attaching screenshots of some of the output here, showing the pattern and a zoom-in.
These are the forcing ancil files rather the model output but I think I’ve found an example file with the issue that you reported. In the following I’ve looked /work/n02/n02/miclen/cylc-run/u-ed824/share/./cycle/20240914T1200Z/AP/km1p5/RAL3/um/umnsaa_diag000.nc.
Firstly when looking at the file in xconv2 the spatial patterns look a lot more continuous than those shown above. E.g. this field of surface net downward longwave…
Ah apologies, I read your earlier message too quickly and yes, as you say, sent you the path to the ancils. But based on the fact that this boundary is only present in the fields directly at the surface (heat fluxes, net SW and LW, etc) I agree that I think it is one of the ancils producing this issue. However, I’ve plotted up almost all of them and haven’t yet found the source of the issue, which is why I posted and am hoping to get a second opinion / some help
It’s possible that this is a post-processing issue, i.e. a red herring. That said if this is raw data then this is unlikely.
Due to the polar nature of the data, looking at the sea ice edge would be where I’d look first, although it sounds like you have possibly already done this. Plotting this boundary and/or the land sea mask on top of the dodgy-looking data may shed some light.
This might be a plotting artefact; NEMO data for example is notoriously tricky to get right due to its tripolar nature. Xconv (the original one not the new one, xconv2) is good for ‘quick looks’ at atmosphere on regular lat-lon grids but it looks like this one isn’t?
Look at equivalent data from model/confgs that have already been published. Relevant data will presumably be on MASS or perhaps on your local group’s local disk.
Okay, thanks for the recommendations! I appreciate it. I’ll take a look into the steps you’ve laid out. I think my first task is combing through the ancils more closely to see what I might have missed the first time around, especially regarding the sea ice data. I’ll get back to you on what I find!
Also, until you mentioned it earlier, I had no idea there was an xconv2 I’ll definitely double check whether it’s a plotting artifact. In the past, any sort of sharp boundary in my output data was related to the ancils or a mistake I made in the configuration for the run (such as the global model grid).
Just on the topic of sea ice, I’ll quickly add these plots in. So the sea ice ancils look okay to me (three datasets within qrclim.seaice) - at least, there’s no indication of that sharp vertical boundary. However, this suite is meant to take sea ice data from ERA5 and use that as forcing. I double checked the ERA5 sea ice data and it also looks fine (global dataset).
Yes xconv2 is brand new and still under active development. It’s already available to install by users and will be centrally installed on Cirrus, the ‘new archer’.
)…
) from the code owners list here 
I’ll definitely double check whether it’s a plotting artifact. In the past, any sort of sharp boundary in my output data was related to the ancils or a mistake I made in the configuration for the run (such as the global model grid).
